Index: stable/10/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_dataset.c =================================================================== --- stable/10/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_dataset.c (revision 290756) +++ stable/10/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_dataset.c (revision 290757) @@ -1,4779 +1,4785 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2013, Joyent, Inc. All rights reserved. * Copyright (c) 2011, 2014 by Delphix. All rights reserved. * Copyright (c) 2012 DEY Storage Systems, Inc. All rights reserved. * Copyright (c) 2011-2012 Pawel Jakub Dawidek . * All rights reserved. * Copyright (c) 2012 Martin Matuska . All rights reserved. * Copyright (c) 2013 Steven Hartland. All rights reserved. * Copyright 2013 Nexenta Systems, Inc. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "zfs_namecheck.h" #include "zfs_prop.h" #include "libzfs_impl.h" #include "zfs_deleg.h" static int userquota_propname_decode(const char *propname, boolean_t zoned, zfs_userquota_prop_t *typep, char *domain, int domainlen, uint64_t *ridp); /* * Given a single type (not a mask of types), return the type in a human * readable form. */ const char * zfs_type_to_name(zfs_type_t type) { switch (type) { case ZFS_TYPE_FILESYSTEM: return (dgettext(TEXT_DOMAIN, "filesystem")); case ZFS_TYPE_SNAPSHOT: return (dgettext(TEXT_DOMAIN, "snapshot")); case ZFS_TYPE_VOLUME: return (dgettext(TEXT_DOMAIN, "volume")); } return (NULL); } /* * Given a path and mask of ZFS types, return a string describing this dataset. * This is used when we fail to open a dataset and we cannot get an exact type. * We guess what the type would have been based on the path and the mask of * acceptable types. */ static const char * path_to_str(const char *path, int types) { /* * When given a single type, always report the exact type. */ if (types == ZFS_TYPE_SNAPSHOT) return (dgettext(TEXT_DOMAIN, "snapshot")); if (types == ZFS_TYPE_FILESYSTEM) return (dgettext(TEXT_DOMAIN, "filesystem")); if (types == ZFS_TYPE_VOLUME) return (dgettext(TEXT_DOMAIN, "volume")); /* * The user is requesting more than one type of dataset. If this is the * case, consult the path itself. If we're looking for a snapshot, and * a '@' is found, then report it as "snapshot". Otherwise, remove the * snapshot attribute and try again. */ if (types & ZFS_TYPE_SNAPSHOT) { if (strchr(path, '@') != NULL) return (dgettext(TEXT_DOMAIN, "snapshot")); return (path_to_str(path, types & ~ZFS_TYPE_SNAPSHOT)); } /* * The user has requested either filesystems or volumes. * We have no way of knowing a priori what type this would be, so always * report it as "filesystem" or "volume", our two primitive types. */ if (types & ZFS_TYPE_FILESYSTEM) return (dgettext(TEXT_DOMAIN, "filesystem")); assert(types & ZFS_TYPE_VOLUME); return (dgettext(TEXT_DOMAIN, "volume")); } /* * Validate a ZFS path. This is used even before trying to open the dataset, to * provide a more meaningful error message. We call zfs_error_aux() to * explain exactly why the name was not valid. */ int zfs_validate_name(libzfs_handle_t *hdl, const char *path, int type, boolean_t modifying) { namecheck_err_t why; char what; (void) zfs_prop_get_table(); if (dataset_namecheck(path, &why, &what) != 0) { if (hdl != NULL) { switch (why) { case NAME_ERR_TOOLONG: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "name is too long")); break; case NAME_ERR_LEADING_SLASH: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "leading slash in name")); break; case NAME_ERR_EMPTY_COMPONENT: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "empty component in name")); break; case NAME_ERR_TRAILING_SLASH: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "trailing slash in name")); break; case NAME_ERR_INVALCHAR: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid character " "'%c' in name"), what); break; case NAME_ERR_MULTIPLE_AT: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "multiple '@' delimiters in name")); break; case NAME_ERR_NOLETTER: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool doesn't begin with a letter")); break; case NAME_ERR_RESERVED: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "name is reserved")); break; case NAME_ERR_DISKLIKE: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "reserved disk name")); break; } } return (0); } if (!(type & ZFS_TYPE_SNAPSHOT) && strchr(path, '@') != NULL) { if (hdl != NULL) zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "snapshot delimiter '@' in filesystem name")); return (0); } if (type == ZFS_TYPE_SNAPSHOT && strchr(path, '@') == NULL) { if (hdl != NULL) zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "missing '@' delimiter in snapshot name")); return (0); } if (modifying && strchr(path, '%') != NULL) { if (hdl != NULL) zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid character %c in name"), '%'); return (0); } return (-1); } int zfs_name_valid(const char *name, zfs_type_t type) { if (type == ZFS_TYPE_POOL) return (zpool_name_valid(NULL, B_FALSE, name)); return (zfs_validate_name(NULL, name, type, B_FALSE)); } /* * This function takes the raw DSL properties, and filters out the user-defined * properties into a separate nvlist. */ static nvlist_t * process_user_props(zfs_handle_t *zhp, nvlist_t *props) { libzfs_handle_t *hdl = zhp->zfs_hdl; nvpair_t *elem; nvlist_t *propval; nvlist_t *nvl; if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) { (void) no_memory(hdl); return (NULL); } elem = NULL; while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { if (!zfs_prop_user(nvpair_name(elem))) continue; verify(nvpair_value_nvlist(elem, &propval) == 0); if (nvlist_add_nvlist(nvl, nvpair_name(elem), propval) != 0) { nvlist_free(nvl); (void) no_memory(hdl); return (NULL); } } return (nvl); } static zpool_handle_t * zpool_add_handle(zfs_handle_t *zhp, const char *pool_name) { libzfs_handle_t *hdl = zhp->zfs_hdl; zpool_handle_t *zph; if ((zph = zpool_open_canfail(hdl, pool_name)) != NULL) { if (hdl->libzfs_pool_handles != NULL) zph->zpool_next = hdl->libzfs_pool_handles; hdl->libzfs_pool_handles = zph; } return (zph); } static zpool_handle_t * zpool_find_handle(zfs_handle_t *zhp, const char *pool_name, int len) { libzfs_handle_t *hdl = zhp->zfs_hdl; zpool_handle_t *zph = hdl->libzfs_pool_handles; while ((zph != NULL) && (strncmp(pool_name, zpool_get_name(zph), len) != 0)) zph = zph->zpool_next; return (zph); } /* * Returns a handle to the pool that contains the provided dataset. * If a handle to that pool already exists then that handle is returned. * Otherwise, a new handle is created and added to the list of handles. */ static zpool_handle_t * zpool_handle(zfs_handle_t *zhp) { char *pool_name; int len; zpool_handle_t *zph; len = strcspn(zhp->zfs_name, "/@#") + 1; pool_name = zfs_alloc(zhp->zfs_hdl, len); (void) strlcpy(pool_name, zhp->zfs_name, len); zph = zpool_find_handle(zhp, pool_name, len); if (zph == NULL) zph = zpool_add_handle(zhp, pool_name); free(pool_name); return (zph); } void zpool_free_handles(libzfs_handle_t *hdl) { zpool_handle_t *next, *zph = hdl->libzfs_pool_handles; while (zph != NULL) { next = zph->zpool_next; zpool_close(zph); zph = next; } hdl->libzfs_pool_handles = NULL; } /* * Utility function to gather stats (objset and zpl) for the given object. */ static int get_stats_ioctl(zfs_handle_t *zhp, zfs_cmd_t *zc) { libzfs_handle_t *hdl = zhp->zfs_hdl; (void) strlcpy(zc->zc_name, zhp->zfs_name, sizeof (zc->zc_name)); while (ioctl(hdl->libzfs_fd, ZFS_IOC_OBJSET_STATS, zc) != 0) { if (errno == ENOMEM) { if (zcmd_expand_dst_nvlist(hdl, zc) != 0) { return (-1); } } else { return (-1); } } return (0); } /* * Utility function to get the received properties of the given object. */ static int get_recvd_props_ioctl(zfs_handle_t *zhp) { libzfs_handle_t *hdl = zhp->zfs_hdl; nvlist_t *recvdprops; zfs_cmd_t zc = { 0 }; int err; if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0) return (-1); (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); while (ioctl(hdl->libzfs_fd, ZFS_IOC_OBJSET_RECVD_PROPS, &zc) != 0) { if (errno == ENOMEM) { if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { return (-1); } } else { zcmd_free_nvlists(&zc); return (-1); } } err = zcmd_read_dst_nvlist(zhp->zfs_hdl, &zc, &recvdprops); zcmd_free_nvlists(&zc); if (err != 0) return (-1); nvlist_free(zhp->zfs_recvd_props); zhp->zfs_recvd_props = recvdprops; return (0); } static int put_stats_zhdl(zfs_handle_t *zhp, zfs_cmd_t *zc) { nvlist_t *allprops, *userprops; zhp->zfs_dmustats = zc->zc_objset_stats; /* structure assignment */ if (zcmd_read_dst_nvlist(zhp->zfs_hdl, zc, &allprops) != 0) { return (-1); } /* * XXX Why do we store the user props separately, in addition to * storing them in zfs_props? */ if ((userprops = process_user_props(zhp, allprops)) == NULL) { nvlist_free(allprops); return (-1); } nvlist_free(zhp->zfs_props); nvlist_free(zhp->zfs_user_props); zhp->zfs_props = allprops; zhp->zfs_user_props = userprops; return (0); } static int get_stats(zfs_handle_t *zhp) { int rc = 0; zfs_cmd_t zc = { 0 }; if (zcmd_alloc_dst_nvlist(zhp->zfs_hdl, &zc, 0) != 0) return (-1); if (get_stats_ioctl(zhp, &zc) != 0) rc = -1; else if (put_stats_zhdl(zhp, &zc) != 0) rc = -1; zcmd_free_nvlists(&zc); return (rc); } /* * Refresh the properties currently stored in the handle. */ void zfs_refresh_properties(zfs_handle_t *zhp) { (void) get_stats(zhp); } /* * Makes a handle from the given dataset name. Used by zfs_open() and * zfs_iter_* to create child handles on the fly. */ static int make_dataset_handle_common(zfs_handle_t *zhp, zfs_cmd_t *zc) { if (put_stats_zhdl(zhp, zc) != 0) return (-1); /* * We've managed to open the dataset and gather statistics. Determine * the high-level type. */ if (zhp->zfs_dmustats.dds_type == DMU_OST_ZVOL) zhp->zfs_head_type = ZFS_TYPE_VOLUME; else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZFS) zhp->zfs_head_type = ZFS_TYPE_FILESYSTEM; else abort(); if (zhp->zfs_dmustats.dds_is_snapshot) zhp->zfs_type = ZFS_TYPE_SNAPSHOT; else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZVOL) zhp->zfs_type = ZFS_TYPE_VOLUME; else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZFS) zhp->zfs_type = ZFS_TYPE_FILESYSTEM; else abort(); /* we should never see any other types */ if ((zhp->zpool_hdl = zpool_handle(zhp)) == NULL) return (-1); return (0); } zfs_handle_t * make_dataset_handle(libzfs_handle_t *hdl, const char *path) { zfs_cmd_t zc = { 0 }; zfs_handle_t *zhp = calloc(sizeof (zfs_handle_t), 1); if (zhp == NULL) return (NULL); zhp->zfs_hdl = hdl; (void) strlcpy(zhp->zfs_name, path, sizeof (zhp->zfs_name)); if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0) { free(zhp); return (NULL); } if (get_stats_ioctl(zhp, &zc) == -1) { zcmd_free_nvlists(&zc); free(zhp); return (NULL); } if (make_dataset_handle_common(zhp, &zc) == -1) { free(zhp); zhp = NULL; } zcmd_free_nvlists(&zc); return (zhp); } zfs_handle_t * make_dataset_handle_zc(libzfs_handle_t *hdl, zfs_cmd_t *zc) { zfs_handle_t *zhp = calloc(sizeof (zfs_handle_t), 1); if (zhp == NULL) return (NULL); zhp->zfs_hdl = hdl; (void) strlcpy(zhp->zfs_name, zc->zc_name, sizeof (zhp->zfs_name)); if (make_dataset_handle_common(zhp, zc) == -1) { free(zhp); return (NULL); } return (zhp); } zfs_handle_t * make_dataset_simple_handle_zc(zfs_handle_t *pzhp, zfs_cmd_t *zc) { zfs_handle_t *zhp = calloc(sizeof (zfs_handle_t), 1); if (zhp == NULL) return (NULL); zhp->zfs_hdl = pzhp->zfs_hdl; (void) strlcpy(zhp->zfs_name, zc->zc_name, sizeof (zhp->zfs_name)); zhp->zfs_head_type = pzhp->zfs_type; zhp->zfs_type = ZFS_TYPE_SNAPSHOT; zhp->zpool_hdl = zpool_handle(zhp); return (zhp); } zfs_handle_t * zfs_handle_dup(zfs_handle_t *zhp_orig) { zfs_handle_t *zhp = calloc(sizeof (zfs_handle_t), 1); if (zhp == NULL) return (NULL); zhp->zfs_hdl = zhp_orig->zfs_hdl; zhp->zpool_hdl = zhp_orig->zpool_hdl; (void) strlcpy(zhp->zfs_name, zhp_orig->zfs_name, sizeof (zhp->zfs_name)); zhp->zfs_type = zhp_orig->zfs_type; zhp->zfs_head_type = zhp_orig->zfs_head_type; zhp->zfs_dmustats = zhp_orig->zfs_dmustats; if (zhp_orig->zfs_props != NULL) { if (nvlist_dup(zhp_orig->zfs_props, &zhp->zfs_props, 0) != 0) { (void) no_memory(zhp->zfs_hdl); zfs_close(zhp); return (NULL); } } if (zhp_orig->zfs_user_props != NULL) { if (nvlist_dup(zhp_orig->zfs_user_props, &zhp->zfs_user_props, 0) != 0) { (void) no_memory(zhp->zfs_hdl); zfs_close(zhp); return (NULL); } } if (zhp_orig->zfs_recvd_props != NULL) { if (nvlist_dup(zhp_orig->zfs_recvd_props, &zhp->zfs_recvd_props, 0)) { (void) no_memory(zhp->zfs_hdl); zfs_close(zhp); return (NULL); } } zhp->zfs_mntcheck = zhp_orig->zfs_mntcheck; if (zhp_orig->zfs_mntopts != NULL) { zhp->zfs_mntopts = zfs_strdup(zhp_orig->zfs_hdl, zhp_orig->zfs_mntopts); } zhp->zfs_props_table = zhp_orig->zfs_props_table; return (zhp); } boolean_t zfs_bookmark_exists(const char *path) { nvlist_t *bmarks; nvlist_t *props; char fsname[ZFS_MAXNAMELEN]; char *bmark_name; char *pound; int err; boolean_t rv; (void) strlcpy(fsname, path, sizeof (fsname)); pound = strchr(fsname, '#'); if (pound == NULL) return (B_FALSE); *pound = '\0'; bmark_name = pound + 1; props = fnvlist_alloc(); err = lzc_get_bookmarks(fsname, props, &bmarks); nvlist_free(props); if (err != 0) { nvlist_free(bmarks); return (B_FALSE); } rv = nvlist_exists(bmarks, bmark_name); nvlist_free(bmarks); return (rv); } zfs_handle_t * make_bookmark_handle(zfs_handle_t *parent, const char *path, nvlist_t *bmark_props) { zfs_handle_t *zhp = calloc(sizeof (zfs_handle_t), 1); if (zhp == NULL) return (NULL); /* Fill in the name. */ zhp->zfs_hdl = parent->zfs_hdl; (void) strlcpy(zhp->zfs_name, path, sizeof (zhp->zfs_name)); /* Set the property lists. */ if (nvlist_dup(bmark_props, &zhp->zfs_props, 0) != 0) { free(zhp); return (NULL); } /* Set the types. */ zhp->zfs_head_type = parent->zfs_head_type; zhp->zfs_type = ZFS_TYPE_BOOKMARK; if ((zhp->zpool_hdl = zpool_handle(zhp)) == NULL) { nvlist_free(zhp->zfs_props); free(zhp); return (NULL); } return (zhp); } /* * Opens the given snapshot, filesystem, or volume. The 'types' * argument is a mask of acceptable types. The function will print an * appropriate error message and return NULL if it can't be opened. */ zfs_handle_t * zfs_open(libzfs_handle_t *hdl, const char *path, int types) { zfs_handle_t *zhp; char errbuf[1024]; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot open '%s'"), path); /* * Validate the name before we even try to open it. */ if (!zfs_validate_name(hdl, path, ZFS_TYPE_DATASET, B_FALSE)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid dataset name")); (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); return (NULL); } /* * Try to get stats for the dataset, which will tell us if it exists. */ errno = 0; if ((zhp = make_dataset_handle(hdl, path)) == NULL) { (void) zfs_standard_error(hdl, errno, errbuf); return (NULL); } if (zhp == NULL) { char *at = strchr(path, '@'); if (at != NULL) *at = '\0'; errno = 0; if ((zhp = make_dataset_handle(hdl, path)) == NULL) { (void) zfs_standard_error(hdl, errno, errbuf); return (NULL); } if (at != NULL) *at = '@'; (void) strlcpy(zhp->zfs_name, path, sizeof (zhp->zfs_name)); zhp->zfs_type = ZFS_TYPE_SNAPSHOT; } if (!(types & zhp->zfs_type)) { (void) zfs_error(hdl, EZFS_BADTYPE, errbuf); zfs_close(zhp); return (NULL); } return (zhp); } /* * Release a ZFS handle. Nothing to do but free the associated memory. */ void zfs_close(zfs_handle_t *zhp) { if (zhp->zfs_mntopts) free(zhp->zfs_mntopts); nvlist_free(zhp->zfs_props); nvlist_free(zhp->zfs_user_props); nvlist_free(zhp->zfs_recvd_props); free(zhp); } typedef struct mnttab_node { struct mnttab mtn_mt; avl_node_t mtn_node; } mnttab_node_t; static int libzfs_mnttab_cache_compare(const void *arg1, const void *arg2) { const mnttab_node_t *mtn1 = arg1; const mnttab_node_t *mtn2 = arg2; int rv; rv = strcmp(mtn1->mtn_mt.mnt_special, mtn2->mtn_mt.mnt_special); if (rv == 0) return (0); return (rv > 0 ? 1 : -1); } void libzfs_mnttab_init(libzfs_handle_t *hdl) { assert(avl_numnodes(&hdl->libzfs_mnttab_cache) == 0); avl_create(&hdl->libzfs_mnttab_cache, libzfs_mnttab_cache_compare, sizeof (mnttab_node_t), offsetof(mnttab_node_t, mtn_node)); } void libzfs_mnttab_update(libzfs_handle_t *hdl) { struct mnttab entry; rewind(hdl->libzfs_mnttab); while (getmntent(hdl->libzfs_mnttab, &entry) == 0) { mnttab_node_t *mtn; if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0) continue; mtn = zfs_alloc(hdl, sizeof (mnttab_node_t)); mtn->mtn_mt.mnt_special = zfs_strdup(hdl, entry.mnt_special); mtn->mtn_mt.mnt_mountp = zfs_strdup(hdl, entry.mnt_mountp); mtn->mtn_mt.mnt_fstype = zfs_strdup(hdl, entry.mnt_fstype); mtn->mtn_mt.mnt_mntopts = zfs_strdup(hdl, entry.mnt_mntopts); avl_add(&hdl->libzfs_mnttab_cache, mtn); } } void libzfs_mnttab_fini(libzfs_handle_t *hdl) { void *cookie = NULL; mnttab_node_t *mtn; while (mtn = avl_destroy_nodes(&hdl->libzfs_mnttab_cache, &cookie)) { free(mtn->mtn_mt.mnt_special); free(mtn->mtn_mt.mnt_mountp); free(mtn->mtn_mt.mnt_fstype); free(mtn->mtn_mt.mnt_mntopts); free(mtn); } avl_destroy(&hdl->libzfs_mnttab_cache); } void libzfs_mnttab_cache(libzfs_handle_t *hdl, boolean_t enable) { hdl->libzfs_mnttab_enable = enable; } int libzfs_mnttab_find(libzfs_handle_t *hdl, const char *fsname, struct mnttab *entry) { mnttab_node_t find; mnttab_node_t *mtn; if (!hdl->libzfs_mnttab_enable) { struct mnttab srch = { 0 }; if (avl_numnodes(&hdl->libzfs_mnttab_cache)) libzfs_mnttab_fini(hdl); rewind(hdl->libzfs_mnttab); srch.mnt_special = (char *)fsname; srch.mnt_fstype = MNTTYPE_ZFS; if (getmntany(hdl->libzfs_mnttab, entry, &srch) == 0) return (0); else return (ENOENT); } if (avl_numnodes(&hdl->libzfs_mnttab_cache) == 0) libzfs_mnttab_update(hdl); find.mtn_mt.mnt_special = (char *)fsname; mtn = avl_find(&hdl->libzfs_mnttab_cache, &find, NULL); if (mtn) { *entry = mtn->mtn_mt; return (0); } return (ENOENT); } void libzfs_mnttab_add(libzfs_handle_t *hdl, const char *special, const char *mountp, const char *mntopts) { mnttab_node_t *mtn; if (avl_numnodes(&hdl->libzfs_mnttab_cache) == 0) return; mtn = zfs_alloc(hdl, sizeof (mnttab_node_t)); mtn->mtn_mt.mnt_special = zfs_strdup(hdl, special); mtn->mtn_mt.mnt_mountp = zfs_strdup(hdl, mountp); mtn->mtn_mt.mnt_fstype = zfs_strdup(hdl, MNTTYPE_ZFS); mtn->mtn_mt.mnt_mntopts = zfs_strdup(hdl, mntopts); avl_add(&hdl->libzfs_mnttab_cache, mtn); } void libzfs_mnttab_remove(libzfs_handle_t *hdl, const char *fsname) { mnttab_node_t find; mnttab_node_t *ret; find.mtn_mt.mnt_special = (char *)fsname; if (ret = avl_find(&hdl->libzfs_mnttab_cache, (void *)&find, NULL)) { avl_remove(&hdl->libzfs_mnttab_cache, ret); free(ret->mtn_mt.mnt_special); free(ret->mtn_mt.mnt_mountp); free(ret->mtn_mt.mnt_fstype); free(ret->mtn_mt.mnt_mntopts); free(ret); } } int zfs_spa_version(zfs_handle_t *zhp, int *spa_version) { zpool_handle_t *zpool_handle = zhp->zpool_hdl; if (zpool_handle == NULL) return (-1); *spa_version = zpool_get_prop_int(zpool_handle, ZPOOL_PROP_VERSION, NULL); return (0); } /* * The choice of reservation property depends on the SPA version. */ static int zfs_which_resv_prop(zfs_handle_t *zhp, zfs_prop_t *resv_prop) { int spa_version; if (zfs_spa_version(zhp, &spa_version) < 0) return (-1); if (spa_version >= SPA_VERSION_REFRESERVATION) *resv_prop = ZFS_PROP_REFRESERVATION; else *resv_prop = ZFS_PROP_RESERVATION; return (0); } /* * Given an nvlist of properties to set, validates that they are correct, and * parses any numeric properties (index, boolean, etc) if they are specified as * strings. */ nvlist_t * zfs_valid_proplist(libzfs_handle_t *hdl, zfs_type_t type, nvlist_t *nvl, uint64_t zoned, zfs_handle_t *zhp, const char *errbuf) { nvpair_t *elem; uint64_t intval; char *strval; zfs_prop_t prop; nvlist_t *ret; int chosen_normal = -1; int chosen_utf = -1; if (nvlist_alloc(&ret, NV_UNIQUE_NAME, 0) != 0) { (void) no_memory(hdl); return (NULL); } /* * Make sure this property is valid and applies to this type. */ elem = NULL; while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) { const char *propname = nvpair_name(elem); prop = zfs_name_to_prop(propname); if (prop == ZPROP_INVAL && zfs_prop_user(propname)) { /* * This is a user property: make sure it's a * string, and that it's less than ZAP_MAXNAMELEN. */ if (nvpair_type(elem) != DATA_TYPE_STRING) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' must be a string"), propname); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } if (strlen(nvpair_name(elem)) >= ZAP_MAXNAMELEN) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "property name '%s' is too long"), propname); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } (void) nvpair_value_string(elem, &strval); if (nvlist_add_string(ret, propname, strval) != 0) { (void) no_memory(hdl); goto error; } continue; } /* * Currently, only user properties can be modified on * snapshots. */ if (type == ZFS_TYPE_SNAPSHOT) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "this property can not be modified for snapshots")); (void) zfs_error(hdl, EZFS_PROPTYPE, errbuf); goto error; } if (prop == ZPROP_INVAL && zfs_prop_userquota(propname)) { zfs_userquota_prop_t uqtype; char newpropname[128]; char domain[128]; uint64_t rid; uint64_t valary[3]; if (userquota_propname_decode(propname, zoned, &uqtype, domain, sizeof (domain), &rid) != 0) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' has an invalid user/group name"), propname); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } if (uqtype != ZFS_PROP_USERQUOTA && uqtype != ZFS_PROP_GROUPQUOTA) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' is readonly"), propname); (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf); goto error; } if (nvpair_type(elem) == DATA_TYPE_STRING) { (void) nvpair_value_string(elem, &strval); if (strcmp(strval, "none") == 0) { intval = 0; } else if (zfs_nicestrtonum(hdl, strval, &intval) != 0) { (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } } else if (nvpair_type(elem) == DATA_TYPE_UINT64) { (void) nvpair_value_uint64(elem, &intval); if (intval == 0) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "use 'none' to disable " "userquota/groupquota")); goto error; } } else { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' must be a number"), propname); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } /* * Encode the prop name as * userquota@-domain, to make it easy * for the kernel to decode. */ (void) snprintf(newpropname, sizeof (newpropname), "%s%llx-%s", zfs_userquota_prop_prefixes[uqtype], (longlong_t)rid, domain); valary[0] = uqtype; valary[1] = rid; valary[2] = intval; if (nvlist_add_uint64_array(ret, newpropname, valary, 3) != 0) { (void) no_memory(hdl); goto error; } continue; } else if (prop == ZPROP_INVAL && zfs_prop_written(propname)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' is readonly"), propname); (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf); goto error; } if (prop == ZPROP_INVAL) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid property '%s'"), propname); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } if (!zfs_prop_valid_for_type(prop, type)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' does not " "apply to datasets of this type"), propname); (void) zfs_error(hdl, EZFS_PROPTYPE, errbuf); goto error; } if (zfs_prop_readonly(prop) && (!zfs_prop_setonce(prop) || zhp != NULL)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' is readonly"), propname); (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf); goto error; } if (zprop_parse_value(hdl, elem, prop, type, ret, &strval, &intval, errbuf) != 0) goto error; /* * Perform some additional checks for specific properties. */ switch (prop) { case ZFS_PROP_VERSION: { int version; if (zhp == NULL) break; version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION); if (intval < version) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Can not downgrade; already at version %u"), version); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } break; } case ZFS_PROP_VOLBLOCKSIZE: case ZFS_PROP_RECORDSIZE: { int maxbs = SPA_MAXBLOCKSIZE; if (zhp != NULL) { maxbs = zpool_get_prop_int(zhp->zpool_hdl, ZPOOL_PROP_MAXBLOCKSIZE, NULL); } /* * Volumes are limited to a volblocksize of 128KB, * because they typically service workloads with * small random writes, which incur a large performance * penalty with large blocks. */ if (prop == ZFS_PROP_VOLBLOCKSIZE) maxbs = SPA_OLD_MAXBLOCKSIZE; /* * The value must be a power of two between * SPA_MINBLOCKSIZE and maxbs. */ if (intval < SPA_MINBLOCKSIZE || intval > maxbs || !ISP2(intval)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' must be power of 2 from 512B " "to %uKB"), propname, maxbs >> 10); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } break; } case ZFS_PROP_MLSLABEL: { #ifdef sun /* * Verify the mlslabel string and convert to * internal hex label string. */ m_label_t *new_sl; char *hex = NULL; /* internal label string */ /* Default value is already OK. */ if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) break; /* Verify the label can be converted to binary form */ if (((new_sl = m_label_alloc(MAC_LABEL)) == NULL) || (str_to_label(strval, &new_sl, MAC_LABEL, L_NO_CORRECTION, NULL) == -1)) { goto badlabel; } /* Now translate to hex internal label string */ if (label_to_str(new_sl, &hex, M_INTERNAL, DEF_NAMES) != 0) { if (hex) free(hex); goto badlabel; } m_label_free(new_sl); /* If string is already in internal form, we're done. */ if (strcmp(strval, hex) == 0) { free(hex); break; } /* Replace the label string with the internal form. */ (void) nvlist_remove(ret, zfs_prop_to_name(prop), DATA_TYPE_STRING); verify(nvlist_add_string(ret, zfs_prop_to_name(prop), hex) == 0); free(hex); break; badlabel: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid mlslabel '%s'"), strval); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); m_label_free(new_sl); /* OK if null */ #else /* !sun */ zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "mlslabel is not supported on FreeBSD")); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); #endif /* !sun */ goto error; } case ZFS_PROP_MOUNTPOINT: { namecheck_err_t why; if (strcmp(strval, ZFS_MOUNTPOINT_NONE) == 0 || strcmp(strval, ZFS_MOUNTPOINT_LEGACY) == 0) break; if (mountpoint_namecheck(strval, &why)) { switch (why) { case NAME_ERR_LEADING_SLASH: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' must be an absolute path, " "'none', or 'legacy'"), propname); break; case NAME_ERR_TOOLONG: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "component of '%s' is too long"), propname); break; } (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } } /*FALLTHRU*/ case ZFS_PROP_SHARESMB: case ZFS_PROP_SHARENFS: /* * For the mountpoint and sharenfs or sharesmb * properties, check if it can be set in a * global/non-global zone based on * the zoned property value: * * global zone non-global zone * -------------------------------------------------- * zoned=on mountpoint (no) mountpoint (yes) * sharenfs (no) sharenfs (no) * sharesmb (no) sharesmb (no) * * zoned=off mountpoint (yes) N/A * sharenfs (yes) * sharesmb (yes) */ if (zoned) { if (getzoneid() == GLOBAL_ZONEID) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' cannot be set on " "dataset in a non-global zone"), propname); (void) zfs_error(hdl, EZFS_ZONED, errbuf); goto error; } else if (prop == ZFS_PROP_SHARENFS || prop == ZFS_PROP_SHARESMB) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' cannot be set in " "a non-global zone"), propname); (void) zfs_error(hdl, EZFS_ZONED, errbuf); goto error; } } else if (getzoneid() != GLOBAL_ZONEID) { /* * If zoned property is 'off', this must be in * a global zone. If not, something is wrong. */ zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' cannot be set while dataset " "'zoned' property is set"), propname); (void) zfs_error(hdl, EZFS_ZONED, errbuf); goto error; } /* * At this point, it is legitimate to set the * property. Now we want to make sure that the * property value is valid if it is sharenfs. */ if ((prop == ZFS_PROP_SHARENFS || prop == ZFS_PROP_SHARESMB) && strcmp(strval, "on") != 0 && strcmp(strval, "off") != 0) { zfs_share_proto_t proto; if (prop == ZFS_PROP_SHARESMB) proto = PROTO_SMB; else proto = PROTO_NFS; /* * Must be an valid sharing protocol * option string so init the libshare * in order to enable the parser and * then parse the options. We use the * control API since we don't care about * the current configuration and don't * want the overhead of loading it * until we actually do something. */ if (zfs_init_libshare(hdl, SA_INIT_CONTROL_API) != SA_OK) { /* * An error occurred so we can't do * anything */ zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' cannot be set: problem " "in share initialization"), propname); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } if (zfs_parse_options(strval, proto) != SA_OK) { /* * There was an error in parsing so * deal with it by issuing an error * message and leaving after * uninitializing the the libshare * interface. */ zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' cannot be set to invalid " "options"), propname); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); zfs_uninit_libshare(hdl); goto error; } zfs_uninit_libshare(hdl); } break; case ZFS_PROP_UTF8ONLY: chosen_utf = (int)intval; break; case ZFS_PROP_NORMALIZE: chosen_normal = (int)intval; break; } /* * For changes to existing volumes, we have some additional * checks to enforce. */ if (type == ZFS_TYPE_VOLUME && zhp != NULL) { uint64_t volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE); uint64_t blocksize = zfs_prop_get_int(zhp, ZFS_PROP_VOLBLOCKSIZE); char buf[64]; switch (prop) { case ZFS_PROP_RESERVATION: case ZFS_PROP_REFRESERVATION: if (intval > volsize) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' is greater than current " "volume size"), propname); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } break; case ZFS_PROP_VOLSIZE: if (intval % blocksize != 0) { zfs_nicenum(blocksize, buf, sizeof (buf)); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' must be a multiple of " "volume block size (%s)"), propname, buf); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } if (intval == 0) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' cannot be zero"), propname); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } break; } } } /* * If normalization was chosen, but no UTF8 choice was made, * enforce rejection of non-UTF8 names. * * If normalization was chosen, but rejecting non-UTF8 names * was explicitly not chosen, it is an error. */ if (chosen_normal > 0 && chosen_utf < 0) { if (nvlist_add_uint64(ret, zfs_prop_to_name(ZFS_PROP_UTF8ONLY), 1) != 0) { (void) no_memory(hdl); goto error; } } else if (chosen_normal > 0 && chosen_utf == 0) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' must be set 'on' if normalization chosen"), zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); (void) zfs_error(hdl, EZFS_BADPROP, errbuf); goto error; } return (ret); error: nvlist_free(ret); return (NULL); } int zfs_add_synthetic_resv(zfs_handle_t *zhp, nvlist_t *nvl) { uint64_t old_volsize; uint64_t new_volsize; uint64_t old_reservation; uint64_t new_reservation; zfs_prop_t resv_prop; /* * If this is an existing volume, and someone is setting the volsize, * make sure that it matches the reservation, or add it if necessary. */ old_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE); if (zfs_which_resv_prop(zhp, &resv_prop) < 0) return (-1); old_reservation = zfs_prop_get_int(zhp, resv_prop); if ((zvol_volsize_to_reservation(old_volsize, zhp->zfs_props) != old_reservation) || nvlist_lookup_uint64(nvl, zfs_prop_to_name(resv_prop), &new_reservation) != ENOENT) { return (0); } if (nvlist_lookup_uint64(nvl, zfs_prop_to_name(ZFS_PROP_VOLSIZE), &new_volsize) != 0) return (-1); new_reservation = zvol_volsize_to_reservation(new_volsize, zhp->zfs_props); if (nvlist_add_uint64(nvl, zfs_prop_to_name(resv_prop), new_reservation) != 0) { (void) no_memory(zhp->zfs_hdl); return (-1); } return (1); } void zfs_setprop_error(libzfs_handle_t *hdl, zfs_prop_t prop, int err, char *errbuf) { switch (err) { case ENOSPC: /* * For quotas and reservations, ENOSPC indicates * something different; setting a quota or reservation * doesn't use any disk space. */ switch (prop) { case ZFS_PROP_QUOTA: case ZFS_PROP_REFQUOTA: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "size is less than current used or " "reserved space")); (void) zfs_error(hdl, EZFS_PROPSPACE, errbuf); break; case ZFS_PROP_RESERVATION: case ZFS_PROP_REFRESERVATION: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "size is greater than available space")); (void) zfs_error(hdl, EZFS_PROPSPACE, errbuf); break; default: (void) zfs_standard_error(hdl, err, errbuf); break; } break; case EBUSY: (void) zfs_standard_error(hdl, EBUSY, errbuf); break; case EROFS: (void) zfs_error(hdl, EZFS_DSREADONLY, errbuf); break; case ENOTSUP: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool and or dataset must be upgraded to set this " "property or value")); (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); break; case ERANGE: case EDOM: if (prop == ZFS_PROP_COMPRESSION || prop == ZFS_PROP_RECORDSIZE) { (void) zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "property setting is not allowed on " "bootable datasets")); (void) zfs_error(hdl, EZFS_NOTSUP, errbuf); + } else if (prop == ZFS_PROP_CHECKSUM || + prop == ZFS_PROP_DEDUP) { + (void) zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, + "property setting is not allowed on " + "root pools")); + (void) zfs_error(hdl, EZFS_NOTSUP, errbuf); } else { (void) zfs_standard_error(hdl, err, errbuf); } break; case EINVAL: if (prop == ZPROP_INVAL) { (void) zfs_error(hdl, EZFS_BADPROP, errbuf); } else { (void) zfs_standard_error(hdl, err, errbuf); } break; case EOVERFLOW: /* * This platform can't address a volume this big. */ #ifdef _ILP32 if (prop == ZFS_PROP_VOLSIZE) { (void) zfs_error(hdl, EZFS_VOLTOOBIG, errbuf); break; } #endif /* FALLTHROUGH */ default: (void) zfs_standard_error(hdl, err, errbuf); } } /* * Given a property name and value, set the property for the given dataset. */ int zfs_prop_set(zfs_handle_t *zhp, const char *propname, const char *propval) { zfs_cmd_t zc = { 0 }; int ret = -1; prop_changelist_t *cl = NULL; char errbuf[1024]; libzfs_handle_t *hdl = zhp->zfs_hdl; nvlist_t *nvl = NULL, *realprops; zfs_prop_t prop; boolean_t do_prefix = B_TRUE; int added_resv; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot set property for '%s'"), zhp->zfs_name); if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0 || nvlist_add_string(nvl, propname, propval) != 0) { (void) no_memory(hdl); goto error; } if ((realprops = zfs_valid_proplist(hdl, zhp->zfs_type, nvl, zfs_prop_get_int(zhp, ZFS_PROP_ZONED), zhp, errbuf)) == NULL) goto error; nvlist_free(nvl); nvl = realprops; prop = zfs_name_to_prop(propname); /* We don't support those properties on FreeBSD. */ switch (prop) { case ZFS_PROP_DEVICES: case ZFS_PROP_ISCSIOPTIONS: case ZFS_PROP_XATTR: case ZFS_PROP_VSCAN: case ZFS_PROP_NBMAND: case ZFS_PROP_MLSLABEL: (void) snprintf(errbuf, sizeof (errbuf), "property '%s' not supported on FreeBSD", propname); ret = zfs_error(hdl, EZFS_PERM, errbuf); goto error; } if (prop == ZFS_PROP_VOLSIZE) { if ((added_resv = zfs_add_synthetic_resv(zhp, nvl)) == -1) goto error; } if ((cl = changelist_gather(zhp, prop, 0, 0)) == NULL) goto error; if (prop == ZFS_PROP_MOUNTPOINT && changelist_haszonedchild(cl)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "child dataset with inherited mountpoint is used " "in a non-global zone")); ret = zfs_error(hdl, EZFS_ZONED, errbuf); goto error; } /* * We don't want to unmount & remount the dataset when changing * its canmount property to 'on' or 'noauto'. We only use * the changelist logic to unmount when setting canmount=off. */ if (prop == ZFS_PROP_CANMOUNT) { uint64_t idx; int err = zprop_string_to_index(prop, propval, &idx, ZFS_TYPE_DATASET); if (err == 0 && idx != ZFS_CANMOUNT_OFF) do_prefix = B_FALSE; } if (do_prefix && (ret = changelist_prefix(cl)) != 0) goto error; /* * Execute the corresponding ioctl() to set this property. */ (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); if (zcmd_write_src_nvlist(hdl, &zc, nvl) != 0) goto error; ret = zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc); if (ret != 0) { zfs_setprop_error(hdl, prop, errno, errbuf); if (added_resv && errno == ENOSPC) { /* clean up the volsize property we tried to set */ uint64_t old_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE); nvlist_free(nvl); zcmd_free_nvlists(&zc); if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) goto error; if (nvlist_add_uint64(nvl, zfs_prop_to_name(ZFS_PROP_VOLSIZE), old_volsize) != 0) goto error; if (zcmd_write_src_nvlist(hdl, &zc, nvl) != 0) goto error; (void) zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc); } } else { if (do_prefix) ret = changelist_postfix(cl); /* * Refresh the statistics so the new property value * is reflected. */ if (ret == 0) (void) get_stats(zhp); } error: nvlist_free(nvl); zcmd_free_nvlists(&zc); if (cl) changelist_free(cl); return (ret); } /* * Given a property, inherit the value from the parent dataset, or if received * is TRUE, revert to the received value, if any. */ int zfs_prop_inherit(zfs_handle_t *zhp, const char *propname, boolean_t received) { zfs_cmd_t zc = { 0 }; int ret; prop_changelist_t *cl; libzfs_handle_t *hdl = zhp->zfs_hdl; char errbuf[1024]; zfs_prop_t prop; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot inherit %s for '%s'"), propname, zhp->zfs_name); zc.zc_cookie = received; if ((prop = zfs_name_to_prop(propname)) == ZPROP_INVAL) { /* * For user properties, the amount of work we have to do is very * small, so just do it here. */ if (!zfs_prop_user(propname)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid property")); return (zfs_error(hdl, EZFS_BADPROP, errbuf)); } (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); (void) strlcpy(zc.zc_value, propname, sizeof (zc.zc_value)); if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_INHERIT_PROP, &zc) != 0) return (zfs_standard_error(hdl, errno, errbuf)); return (0); } /* * Verify that this property is inheritable. */ if (zfs_prop_readonly(prop)) return (zfs_error(hdl, EZFS_PROPREADONLY, errbuf)); if (!zfs_prop_inheritable(prop) && !received) return (zfs_error(hdl, EZFS_PROPNONINHERIT, errbuf)); /* * Check to see if the value applies to this type */ if (!zfs_prop_valid_for_type(prop, zhp->zfs_type)) return (zfs_error(hdl, EZFS_PROPTYPE, errbuf)); /* * Normalize the name, to get rid of shorthand abbreviations. */ propname = zfs_prop_to_name(prop); (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); (void) strlcpy(zc.zc_value, propname, sizeof (zc.zc_value)); if (prop == ZFS_PROP_MOUNTPOINT && getzoneid() == GLOBAL_ZONEID && zfs_prop_get_int(zhp, ZFS_PROP_ZONED)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "dataset is used in a non-global zone")); return (zfs_error(hdl, EZFS_ZONED, errbuf)); } /* * Determine datasets which will be affected by this change, if any. */ if ((cl = changelist_gather(zhp, prop, 0, 0)) == NULL) return (-1); if (prop == ZFS_PROP_MOUNTPOINT && changelist_haszonedchild(cl)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "child dataset with inherited mountpoint is used " "in a non-global zone")); ret = zfs_error(hdl, EZFS_ZONED, errbuf); goto error; } if ((ret = changelist_prefix(cl)) != 0) goto error; if ((ret = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_INHERIT_PROP, &zc)) != 0) { return (zfs_standard_error(hdl, errno, errbuf)); } else { if ((ret = changelist_postfix(cl)) != 0) goto error; /* * Refresh the statistics so the new property is reflected. */ (void) get_stats(zhp); } error: changelist_free(cl); return (ret); } /* * True DSL properties are stored in an nvlist. The following two functions * extract them appropriately. */ static uint64_t getprop_uint64(zfs_handle_t *zhp, zfs_prop_t prop, char **source) { nvlist_t *nv; uint64_t value; *source = NULL; if (nvlist_lookup_nvlist(zhp->zfs_props, zfs_prop_to_name(prop), &nv) == 0) { verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0); (void) nvlist_lookup_string(nv, ZPROP_SOURCE, source); } else { verify(!zhp->zfs_props_table || zhp->zfs_props_table[prop] == B_TRUE); value = zfs_prop_default_numeric(prop); *source = ""; } return (value); } static const char * getprop_string(zfs_handle_t *zhp, zfs_prop_t prop, char **source) { nvlist_t *nv; const char *value; *source = NULL; if (nvlist_lookup_nvlist(zhp->zfs_props, zfs_prop_to_name(prop), &nv) == 0) { value = fnvlist_lookup_string(nv, ZPROP_VALUE); (void) nvlist_lookup_string(nv, ZPROP_SOURCE, source); } else { verify(!zhp->zfs_props_table || zhp->zfs_props_table[prop] == B_TRUE); value = zfs_prop_default_string(prop); *source = ""; } return (value); } static boolean_t zfs_is_recvd_props_mode(zfs_handle_t *zhp) { return (zhp->zfs_props == zhp->zfs_recvd_props); } static void zfs_set_recvd_props_mode(zfs_handle_t *zhp, uint64_t *cookie) { *cookie = (uint64_t)(uintptr_t)zhp->zfs_props; zhp->zfs_props = zhp->zfs_recvd_props; } static void zfs_unset_recvd_props_mode(zfs_handle_t *zhp, uint64_t *cookie) { zhp->zfs_props = (nvlist_t *)(uintptr_t)*cookie; *cookie = 0; } /* * Internal function for getting a numeric property. Both zfs_prop_get() and * zfs_prop_get_int() are built using this interface. * * Certain properties can be overridden using 'mount -o'. In this case, scan * the contents of the /etc/mnttab entry, searching for the appropriate options. * If they differ from the on-disk values, report the current values and mark * the source "temporary". */ static int get_numeric_property(zfs_handle_t *zhp, zfs_prop_t prop, zprop_source_t *src, char **source, uint64_t *val) { zfs_cmd_t zc = { 0 }; nvlist_t *zplprops = NULL; struct mnttab mnt; char *mntopt_on = NULL; char *mntopt_off = NULL; boolean_t received = zfs_is_recvd_props_mode(zhp); *source = NULL; switch (prop) { case ZFS_PROP_ATIME: mntopt_on = MNTOPT_ATIME; mntopt_off = MNTOPT_NOATIME; break; case ZFS_PROP_DEVICES: mntopt_on = MNTOPT_DEVICES; mntopt_off = MNTOPT_NODEVICES; break; case ZFS_PROP_EXEC: mntopt_on = MNTOPT_EXEC; mntopt_off = MNTOPT_NOEXEC; break; case ZFS_PROP_READONLY: mntopt_on = MNTOPT_RO; mntopt_off = MNTOPT_RW; break; case ZFS_PROP_SETUID: mntopt_on = MNTOPT_SETUID; mntopt_off = MNTOPT_NOSETUID; break; case ZFS_PROP_XATTR: mntopt_on = MNTOPT_XATTR; mntopt_off = MNTOPT_NOXATTR; break; case ZFS_PROP_NBMAND: mntopt_on = MNTOPT_NBMAND; mntopt_off = MNTOPT_NONBMAND; break; } /* * Because looking up the mount options is potentially expensive * (iterating over all of /etc/mnttab), we defer its calculation until * we're looking up a property which requires its presence. */ if (!zhp->zfs_mntcheck && (mntopt_on != NULL || prop == ZFS_PROP_MOUNTED)) { libzfs_handle_t *hdl = zhp->zfs_hdl; struct mnttab entry; if (libzfs_mnttab_find(hdl, zhp->zfs_name, &entry) == 0) { zhp->zfs_mntopts = zfs_strdup(hdl, entry.mnt_mntopts); if (zhp->zfs_mntopts == NULL) return (-1); } zhp->zfs_mntcheck = B_TRUE; } if (zhp->zfs_mntopts == NULL) mnt.mnt_mntopts = ""; else mnt.mnt_mntopts = zhp->zfs_mntopts; switch (prop) { case ZFS_PROP_ATIME: case ZFS_PROP_DEVICES: case ZFS_PROP_EXEC: case ZFS_PROP_READONLY: case ZFS_PROP_SETUID: case ZFS_PROP_XATTR: case ZFS_PROP_NBMAND: *val = getprop_uint64(zhp, prop, source); if (received) break; if (hasmntopt(&mnt, mntopt_on) && !*val) { *val = B_TRUE; if (src) *src = ZPROP_SRC_TEMPORARY; } else if (hasmntopt(&mnt, mntopt_off) && *val) { *val = B_FALSE; if (src) *src = ZPROP_SRC_TEMPORARY; } break; case ZFS_PROP_CANMOUNT: case ZFS_PROP_VOLSIZE: case ZFS_PROP_QUOTA: case ZFS_PROP_REFQUOTA: case ZFS_PROP_RESERVATION: case ZFS_PROP_REFRESERVATION: case ZFS_PROP_FILESYSTEM_LIMIT: case ZFS_PROP_SNAPSHOT_LIMIT: case ZFS_PROP_FILESYSTEM_COUNT: case ZFS_PROP_SNAPSHOT_COUNT: *val = getprop_uint64(zhp, prop, source); if (*source == NULL) { /* not default, must be local */ *source = zhp->zfs_name; } break; case ZFS_PROP_MOUNTED: *val = (zhp->zfs_mntopts != NULL); break; case ZFS_PROP_NUMCLONES: *val = zhp->zfs_dmustats.dds_num_clones; break; case ZFS_PROP_VERSION: case ZFS_PROP_NORMALIZE: case ZFS_PROP_UTF8ONLY: case ZFS_PROP_CASE: if (!zfs_prop_valid_for_type(prop, zhp->zfs_head_type) || zcmd_alloc_dst_nvlist(zhp->zfs_hdl, &zc, 0) != 0) return (-1); (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_OBJSET_ZPLPROPS, &zc)) { zcmd_free_nvlists(&zc); return (-1); } if (zcmd_read_dst_nvlist(zhp->zfs_hdl, &zc, &zplprops) != 0 || nvlist_lookup_uint64(zplprops, zfs_prop_to_name(prop), val) != 0) { zcmd_free_nvlists(&zc); return (-1); } if (zplprops) nvlist_free(zplprops); zcmd_free_nvlists(&zc); break; case ZFS_PROP_INCONSISTENT: *val = zhp->zfs_dmustats.dds_inconsistent; break; default: switch (zfs_prop_get_type(prop)) { case PROP_TYPE_NUMBER: case PROP_TYPE_INDEX: *val = getprop_uint64(zhp, prop, source); /* * If we tried to use a default value for a * readonly property, it means that it was not * present. */ if (zfs_prop_readonly(prop) && *source != NULL && (*source)[0] == '\0') { *source = NULL; } break; case PROP_TYPE_STRING: default: zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN, "cannot get non-numeric property")); return (zfs_error(zhp->zfs_hdl, EZFS_BADPROP, dgettext(TEXT_DOMAIN, "internal error"))); } } return (0); } /* * Calculate the source type, given the raw source string. */ static void get_source(zfs_handle_t *zhp, zprop_source_t *srctype, char *source, char *statbuf, size_t statlen) { if (statbuf == NULL || *srctype == ZPROP_SRC_TEMPORARY) return; if (source == NULL) { *srctype = ZPROP_SRC_NONE; } else if (source[0] == '\0') { *srctype = ZPROP_SRC_DEFAULT; } else if (strstr(source, ZPROP_SOURCE_VAL_RECVD) != NULL) { *srctype = ZPROP_SRC_RECEIVED; } else { if (strcmp(source, zhp->zfs_name) == 0) { *srctype = ZPROP_SRC_LOCAL; } else { (void) strlcpy(statbuf, source, statlen); *srctype = ZPROP_SRC_INHERITED; } } } int zfs_prop_get_recvd(zfs_handle_t *zhp, const char *propname, char *propbuf, size_t proplen, boolean_t literal) { zfs_prop_t prop; int err = 0; if (zhp->zfs_recvd_props == NULL) if (get_recvd_props_ioctl(zhp) != 0) return (-1); prop = zfs_name_to_prop(propname); if (prop != ZPROP_INVAL) { uint64_t cookie; if (!nvlist_exists(zhp->zfs_recvd_props, propname)) return (-1); zfs_set_recvd_props_mode(zhp, &cookie); err = zfs_prop_get(zhp, prop, propbuf, proplen, NULL, NULL, 0, literal); zfs_unset_recvd_props_mode(zhp, &cookie); } else { nvlist_t *propval; char *recvdval; if (nvlist_lookup_nvlist(zhp->zfs_recvd_props, propname, &propval) != 0) return (-1); verify(nvlist_lookup_string(propval, ZPROP_VALUE, &recvdval) == 0); (void) strlcpy(propbuf, recvdval, proplen); } return (err == 0 ? 0 : -1); } static int get_clones_string(zfs_handle_t *zhp, char *propbuf, size_t proplen) { nvlist_t *value; nvpair_t *pair; value = zfs_get_clones_nvl(zhp); if (value == NULL) return (-1); propbuf[0] = '\0'; for (pair = nvlist_next_nvpair(value, NULL); pair != NULL; pair = nvlist_next_nvpair(value, pair)) { if (propbuf[0] != '\0') (void) strlcat(propbuf, ",", proplen); (void) strlcat(propbuf, nvpair_name(pair), proplen); } return (0); } struct get_clones_arg { uint64_t numclones; nvlist_t *value; const char *origin; char buf[ZFS_MAXNAMELEN]; }; int get_clones_cb(zfs_handle_t *zhp, void *arg) { struct get_clones_arg *gca = arg; if (gca->numclones == 0) { zfs_close(zhp); return (0); } if (zfs_prop_get(zhp, ZFS_PROP_ORIGIN, gca->buf, sizeof (gca->buf), NULL, NULL, 0, B_TRUE) != 0) goto out; if (strcmp(gca->buf, gca->origin) == 0) { fnvlist_add_boolean(gca->value, zfs_get_name(zhp)); gca->numclones--; } out: (void) zfs_iter_children(zhp, get_clones_cb, gca); zfs_close(zhp); return (0); } nvlist_t * zfs_get_clones_nvl(zfs_handle_t *zhp) { nvlist_t *nv, *value; if (nvlist_lookup_nvlist(zhp->zfs_props, zfs_prop_to_name(ZFS_PROP_CLONES), &nv) != 0) { struct get_clones_arg gca; /* * if this is a snapshot, then the kernel wasn't able * to get the clones. Do it by slowly iterating. */ if (zhp->zfs_type != ZFS_TYPE_SNAPSHOT) return (NULL); if (nvlist_alloc(&nv, NV_UNIQUE_NAME, 0) != 0) return (NULL); if (nvlist_alloc(&value, NV_UNIQUE_NAME, 0) != 0) { nvlist_free(nv); return (NULL); } gca.numclones = zfs_prop_get_int(zhp, ZFS_PROP_NUMCLONES); gca.value = value; gca.origin = zhp->zfs_name; if (gca.numclones != 0) { zfs_handle_t *root; char pool[ZFS_MAXNAMELEN]; char *cp = pool; /* get the pool name */ (void) strlcpy(pool, zhp->zfs_name, sizeof (pool)); (void) strsep(&cp, "/@"); root = zfs_open(zhp->zfs_hdl, pool, ZFS_TYPE_FILESYSTEM); (void) get_clones_cb(root, &gca); } if (gca.numclones != 0 || nvlist_add_nvlist(nv, ZPROP_VALUE, value) != 0 || nvlist_add_nvlist(zhp->zfs_props, zfs_prop_to_name(ZFS_PROP_CLONES), nv) != 0) { nvlist_free(nv); nvlist_free(value); return (NULL); } nvlist_free(nv); nvlist_free(value); verify(0 == nvlist_lookup_nvlist(zhp->zfs_props, zfs_prop_to_name(ZFS_PROP_CLONES), &nv)); } verify(nvlist_lookup_nvlist(nv, ZPROP_VALUE, &value) == 0); return (value); } /* * Retrieve a property from the given object. If 'literal' is specified, then * numbers are left as exact values. Otherwise, numbers are converted to a * human-readable form. * * Returns 0 on success, or -1 on error. */ int zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen, zprop_source_t *src, char *statbuf, size_t statlen, boolean_t literal) { char *source = NULL; uint64_t val; const char *str; const char *strval; boolean_t received = zfs_is_recvd_props_mode(zhp); /* * Check to see if this property applies to our object */ if (!zfs_prop_valid_for_type(prop, zhp->zfs_type)) return (-1); if (received && zfs_prop_readonly(prop)) return (-1); if (src) *src = ZPROP_SRC_NONE; switch (prop) { case ZFS_PROP_CREATION: /* * 'creation' is a time_t stored in the statistics. We convert * this into a string unless 'literal' is specified. */ { val = getprop_uint64(zhp, prop, &source); time_t time = (time_t)val; struct tm t; if (literal || localtime_r(&time, &t) == NULL || strftime(propbuf, proplen, "%a %b %e %k:%M %Y", &t) == 0) (void) snprintf(propbuf, proplen, "%llu", val); } break; case ZFS_PROP_MOUNTPOINT: /* * Getting the precise mountpoint can be tricky. * * - for 'none' or 'legacy', return those values. * - for inherited mountpoints, we want to take everything * after our ancestor and append it to the inherited value. * * If the pool has an alternate root, we want to prepend that * root to any values we return. */ str = getprop_string(zhp, prop, &source); if (str[0] == '/') { char buf[MAXPATHLEN]; char *root = buf; const char *relpath; /* * If we inherit the mountpoint, even from a dataset * with a received value, the source will be the path of * the dataset we inherit from. If source is * ZPROP_SOURCE_VAL_RECVD, the received value is not * inherited. */ if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0) { relpath = ""; } else { relpath = zhp->zfs_name + strlen(source); if (relpath[0] == '/') relpath++; } if ((zpool_get_prop(zhp->zpool_hdl, ZPOOL_PROP_ALTROOT, buf, MAXPATHLEN, NULL, B_FALSE)) || (strcmp(root, "-") == 0)) root[0] = '\0'; /* * Special case an alternate root of '/'. This will * avoid having multiple leading slashes in the * mountpoint path. */ if (strcmp(root, "/") == 0) root++; /* * If the mountpoint is '/' then skip over this * if we are obtaining either an alternate root or * an inherited mountpoint. */ if (str[1] == '\0' && (root[0] != '\0' || relpath[0] != '\0')) str++; if (relpath[0] == '\0') (void) snprintf(propbuf, proplen, "%s%s", root, str); else (void) snprintf(propbuf, proplen, "%s%s%s%s", root, str, relpath[0] == '@' ? "" : "/", relpath); } else { /* 'legacy' or 'none' */ (void) strlcpy(propbuf, str, proplen); } break; case ZFS_PROP_ORIGIN: str = getprop_string(zhp, prop, &source); if (str == NULL) return (-1); (void) strlcpy(propbuf, str, proplen); break; case ZFS_PROP_CLONES: if (get_clones_string(zhp, propbuf, proplen) != 0) return (-1); break; case ZFS_PROP_QUOTA: case ZFS_PROP_REFQUOTA: case ZFS_PROP_RESERVATION: case ZFS_PROP_REFRESERVATION: if (get_numeric_property(zhp, prop, src, &source, &val) != 0) return (-1); /* * If quota or reservation is 0, we translate this into 'none' * (unless literal is set), and indicate that it's the default * value. Otherwise, we print the number nicely and indicate * that its set locally. */ if (val == 0) { if (literal) (void) strlcpy(propbuf, "0", proplen); else (void) strlcpy(propbuf, "none", proplen); } else { if (literal) (void) snprintf(propbuf, proplen, "%llu", (u_longlong_t)val); else zfs_nicenum(val, propbuf, proplen); } break; case ZFS_PROP_FILESYSTEM_LIMIT: case ZFS_PROP_SNAPSHOT_LIMIT: case ZFS_PROP_FILESYSTEM_COUNT: case ZFS_PROP_SNAPSHOT_COUNT: if (get_numeric_property(zhp, prop, src, &source, &val) != 0) return (-1); /* * If limit is UINT64_MAX, we translate this into 'none' (unless * literal is set), and indicate that it's the default value. * Otherwise, we print the number nicely and indicate that it's * set locally. */ if (literal) { (void) snprintf(propbuf, proplen, "%llu", (u_longlong_t)val); } else if (val == UINT64_MAX) { (void) strlcpy(propbuf, "none", proplen); } else { zfs_nicenum(val, propbuf, proplen); } break; case ZFS_PROP_REFRATIO: case ZFS_PROP_COMPRESSRATIO: if (get_numeric_property(zhp, prop, src, &source, &val) != 0) return (-1); (void) snprintf(propbuf, proplen, "%llu.%02llux", (u_longlong_t)(val / 100), (u_longlong_t)(val % 100)); break; case ZFS_PROP_TYPE: switch (zhp->zfs_type) { case ZFS_TYPE_FILESYSTEM: str = "filesystem"; break; case ZFS_TYPE_VOLUME: str = "volume"; break; case ZFS_TYPE_SNAPSHOT: str = "snapshot"; break; case ZFS_TYPE_BOOKMARK: str = "bookmark"; break; default: abort(); } (void) snprintf(propbuf, proplen, "%s", str); break; case ZFS_PROP_MOUNTED: /* * The 'mounted' property is a pseudo-property that described * whether the filesystem is currently mounted. Even though * it's a boolean value, the typical values of "on" and "off" * don't make sense, so we translate to "yes" and "no". */ if (get_numeric_property(zhp, ZFS_PROP_MOUNTED, src, &source, &val) != 0) return (-1); if (val) (void) strlcpy(propbuf, "yes", proplen); else (void) strlcpy(propbuf, "no", proplen); break; case ZFS_PROP_NAME: /* * The 'name' property is a pseudo-property derived from the * dataset name. It is presented as a real property to simplify * consumers. */ (void) strlcpy(propbuf, zhp->zfs_name, proplen); break; case ZFS_PROP_MLSLABEL: { #ifdef sun m_label_t *new_sl = NULL; char *ascii = NULL; /* human readable label */ (void) strlcpy(propbuf, getprop_string(zhp, prop, &source), proplen); if (literal || (strcasecmp(propbuf, ZFS_MLSLABEL_DEFAULT) == 0)) break; /* * Try to translate the internal hex string to * human-readable output. If there are any * problems just use the hex string. */ if (str_to_label(propbuf, &new_sl, MAC_LABEL, L_NO_CORRECTION, NULL) == -1) { m_label_free(new_sl); break; } if (label_to_str(new_sl, &ascii, M_LABEL, DEF_NAMES) != 0) { if (ascii) free(ascii); m_label_free(new_sl); break; } m_label_free(new_sl); (void) strlcpy(propbuf, ascii, proplen); free(ascii); #else /* !sun */ propbuf[0] = '\0'; #endif /* !sun */ } break; case ZFS_PROP_GUID: /* * GUIDs are stored as numbers, but they are identifiers. * We don't want them to be pretty printed, because pretty * printing mangles the ID into a truncated and useless value. */ if (get_numeric_property(zhp, prop, src, &source, &val) != 0) return (-1); (void) snprintf(propbuf, proplen, "%llu", (u_longlong_t)val); break; default: switch (zfs_prop_get_type(prop)) { case PROP_TYPE_NUMBER: if (get_numeric_property(zhp, prop, src, &source, &val) != 0) return (-1); if (literal) (void) snprintf(propbuf, proplen, "%llu", (u_longlong_t)val); else zfs_nicenum(val, propbuf, proplen); break; case PROP_TYPE_STRING: str = getprop_string(zhp, prop, &source); if (str == NULL) return (-1); (void) strlcpy(propbuf, str, proplen); break; case PROP_TYPE_INDEX: if (get_numeric_property(zhp, prop, src, &source, &val) != 0) return (-1); if (zfs_prop_index_to_string(prop, val, &strval) != 0) return (-1); (void) strlcpy(propbuf, strval, proplen); break; default: abort(); } } get_source(zhp, src, source, statbuf, statlen); return (0); } /* * Utility function to get the given numeric property. Does no validation that * the given property is the appropriate type; should only be used with * hard-coded property types. */ uint64_t zfs_prop_get_int(zfs_handle_t *zhp, zfs_prop_t prop) { char *source; uint64_t val; (void) get_numeric_property(zhp, prop, NULL, &source, &val); return (val); } int zfs_prop_set_int(zfs_handle_t *zhp, zfs_prop_t prop, uint64_t val) { char buf[64]; (void) snprintf(buf, sizeof (buf), "%llu", (longlong_t)val); return (zfs_prop_set(zhp, zfs_prop_to_name(prop), buf)); } /* * Similar to zfs_prop_get(), but returns the value as an integer. */ int zfs_prop_get_numeric(zfs_handle_t *zhp, zfs_prop_t prop, uint64_t *value, zprop_source_t *src, char *statbuf, size_t statlen) { char *source; /* * Check to see if this property applies to our object */ if (!zfs_prop_valid_for_type(prop, zhp->zfs_type)) { return (zfs_error_fmt(zhp->zfs_hdl, EZFS_PROPTYPE, dgettext(TEXT_DOMAIN, "cannot get property '%s'"), zfs_prop_to_name(prop))); } if (src) *src = ZPROP_SRC_NONE; if (get_numeric_property(zhp, prop, src, &source, value) != 0) return (-1); get_source(zhp, src, source, statbuf, statlen); return (0); } static int idmap_id_to_numeric_domain_rid(uid_t id, boolean_t isuser, char **domainp, idmap_rid_t *ridp) { #ifdef sun idmap_get_handle_t *get_hdl = NULL; idmap_stat status; int err = EINVAL; if (idmap_get_create(&get_hdl) != IDMAP_SUCCESS) goto out; if (isuser) { err = idmap_get_sidbyuid(get_hdl, id, IDMAP_REQ_FLG_USE_CACHE, domainp, ridp, &status); } else { err = idmap_get_sidbygid(get_hdl, id, IDMAP_REQ_FLG_USE_CACHE, domainp, ridp, &status); } if (err == IDMAP_SUCCESS && idmap_get_mappings(get_hdl) == IDMAP_SUCCESS && status == IDMAP_SUCCESS) err = 0; else err = EINVAL; out: if (get_hdl) idmap_get_destroy(get_hdl); return (err); #else /* !sun */ assert(!"invalid code path"); return (EINVAL); // silence compiler warning #endif /* !sun */ } /* * convert the propname into parameters needed by kernel * Eg: userquota@ahrens -> ZFS_PROP_USERQUOTA, "", 126829 * Eg: userused@matt@domain -> ZFS_PROP_USERUSED, "S-1-123-456", 789 */ static int userquota_propname_decode(const char *propname, boolean_t zoned, zfs_userquota_prop_t *typep, char *domain, int domainlen, uint64_t *ridp) { zfs_userquota_prop_t type; char *cp, *end; char *numericsid = NULL; boolean_t isuser; domain[0] = '\0'; *ridp = 0; /* Figure out the property type ({user|group}{quota|space}) */ for (type = 0; type < ZFS_NUM_USERQUOTA_PROPS; type++) { if (strncmp(propname, zfs_userquota_prop_prefixes[type], strlen(zfs_userquota_prop_prefixes[type])) == 0) break; } if (type == ZFS_NUM_USERQUOTA_PROPS) return (EINVAL); *typep = type; isuser = (type == ZFS_PROP_USERQUOTA || type == ZFS_PROP_USERUSED); cp = strchr(propname, '@') + 1; if (strchr(cp, '@')) { #ifdef sun /* * It's a SID name (eg "user@domain") that needs to be * turned into S-1-domainID-RID. */ int flag = 0; idmap_stat stat, map_stat; uid_t pid; idmap_rid_t rid; idmap_get_handle_t *gh = NULL; stat = idmap_get_create(&gh); if (stat != IDMAP_SUCCESS) { idmap_get_destroy(gh); return (ENOMEM); } if (zoned && getzoneid() == GLOBAL_ZONEID) return (ENOENT); if (isuser) { stat = idmap_getuidbywinname(cp, NULL, flag, &pid); if (stat < 0) return (ENOENT); stat = idmap_get_sidbyuid(gh, pid, flag, &numericsid, &rid, &map_stat); } else { stat = idmap_getgidbywinname(cp, NULL, flag, &pid); if (stat < 0) return (ENOENT); stat = idmap_get_sidbygid(gh, pid, flag, &numericsid, &rid, &map_stat); } if (stat < 0) { idmap_get_destroy(gh); return (ENOENT); } stat = idmap_get_mappings(gh); idmap_get_destroy(gh); if (stat < 0) { return (ENOENT); } if (numericsid == NULL) return (ENOENT); cp = numericsid; *ridp = rid; /* will be further decoded below */ #else /* !sun */ return (ENOENT); #endif /* !sun */ } if (strncmp(cp, "S-1-", 4) == 0) { /* It's a numeric SID (eg "S-1-234-567-89") */ (void) strlcpy(domain, cp, domainlen); errno = 0; if (*ridp == 0) { cp = strrchr(domain, '-'); *cp = '\0'; cp++; *ridp = strtoull(cp, &end, 10); } else { end = ""; } if (numericsid) { free(numericsid); numericsid = NULL; } if (errno != 0 || *end != '\0') return (EINVAL); } else if (!isdigit(*cp)) { /* * It's a user/group name (eg "user") that needs to be * turned into a uid/gid */ if (zoned && getzoneid() == GLOBAL_ZONEID) return (ENOENT); if (isuser) { struct passwd *pw; pw = getpwnam(cp); if (pw == NULL) return (ENOENT); *ridp = pw->pw_uid; } else { struct group *gr; gr = getgrnam(cp); if (gr == NULL) return (ENOENT); *ridp = gr->gr_gid; } } else { /* It's a user/group ID (eg "12345"). */ uid_t id = strtoul(cp, &end, 10); idmap_rid_t rid; char *mapdomain; if (*end != '\0') return (EINVAL); if (id > MAXUID) { /* It's an ephemeral ID. */ if (idmap_id_to_numeric_domain_rid(id, isuser, &mapdomain, &rid) != 0) return (ENOENT); (void) strlcpy(domain, mapdomain, domainlen); *ridp = rid; } else { *ridp = id; } } ASSERT3P(numericsid, ==, NULL); return (0); } static int zfs_prop_get_userquota_common(zfs_handle_t *zhp, const char *propname, uint64_t *propvalue, zfs_userquota_prop_t *typep) { int err; zfs_cmd_t zc = { 0 }; (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); err = userquota_propname_decode(propname, zfs_prop_get_int(zhp, ZFS_PROP_ZONED), typep, zc.zc_value, sizeof (zc.zc_value), &zc.zc_guid); zc.zc_objset_type = *typep; if (err) return (err); err = ioctl(zhp->zfs_hdl->libzfs_fd, ZFS_IOC_USERSPACE_ONE, &zc); if (err) return (err); *propvalue = zc.zc_cookie; return (0); } int zfs_prop_get_userquota_int(zfs_handle_t *zhp, const char *propname, uint64_t *propvalue) { zfs_userquota_prop_t type; return (zfs_prop_get_userquota_common(zhp, propname, propvalue, &type)); } int zfs_prop_get_userquota(zfs_handle_t *zhp, const char *propname, char *propbuf, int proplen, boolean_t literal) { int err; uint64_t propvalue; zfs_userquota_prop_t type; err = zfs_prop_get_userquota_common(zhp, propname, &propvalue, &type); if (err) return (err); if (literal) { (void) snprintf(propbuf, proplen, "%llu", propvalue); } else if (propvalue == 0 && (type == ZFS_PROP_USERQUOTA || type == ZFS_PROP_GROUPQUOTA)) { (void) strlcpy(propbuf, "none", proplen); } else { zfs_nicenum(propvalue, propbuf, proplen); } return (0); } int zfs_prop_get_written_int(zfs_handle_t *zhp, const char *propname, uint64_t *propvalue) { int err; zfs_cmd_t zc = { 0 }; const char *snapname; (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); snapname = strchr(propname, '@') + 1; if (strchr(snapname, '@')) { (void) strlcpy(zc.zc_value, snapname, sizeof (zc.zc_value)); } else { /* snapname is the short name, append it to zhp's fsname */ char *cp; (void) strlcpy(zc.zc_value, zhp->zfs_name, sizeof (zc.zc_value)); cp = strchr(zc.zc_value, '@'); if (cp != NULL) *cp = '\0'; (void) strlcat(zc.zc_value, "@", sizeof (zc.zc_value)); (void) strlcat(zc.zc_value, snapname, sizeof (zc.zc_value)); } err = ioctl(zhp->zfs_hdl->libzfs_fd, ZFS_IOC_SPACE_WRITTEN, &zc); if (err) return (err); *propvalue = zc.zc_cookie; return (0); } int zfs_prop_get_written(zfs_handle_t *zhp, const char *propname, char *propbuf, int proplen, boolean_t literal) { int err; uint64_t propvalue; err = zfs_prop_get_written_int(zhp, propname, &propvalue); if (err) return (err); if (literal) { (void) snprintf(propbuf, proplen, "%llu", propvalue); } else { zfs_nicenum(propvalue, propbuf, proplen); } return (0); } /* * Returns the name of the given zfs handle. */ const char * zfs_get_name(const zfs_handle_t *zhp) { return (zhp->zfs_name); } /* * Returns the type of the given zfs handle. */ zfs_type_t zfs_get_type(const zfs_handle_t *zhp) { return (zhp->zfs_type); } /* * Is one dataset name a child dataset of another? * * Needs to handle these cases: * Dataset 1 "a/foo" "a/foo" "a/foo" "a/foo" * Dataset 2 "a/fo" "a/foobar" "a/bar/baz" "a/foo/bar" * Descendant? No. No. No. Yes. */ static boolean_t is_descendant(const char *ds1, const char *ds2) { size_t d1len = strlen(ds1); /* ds2 can't be a descendant if it's smaller */ if (strlen(ds2) < d1len) return (B_FALSE); /* otherwise, compare strings and verify that there's a '/' char */ return (ds2[d1len] == '/' && (strncmp(ds1, ds2, d1len) == 0)); } /* * Given a complete name, return just the portion that refers to the parent. * Will return -1 if there is no parent (path is just the name of the * pool). */ static int parent_name(const char *path, char *buf, size_t buflen) { char *slashp; (void) strlcpy(buf, path, buflen); if ((slashp = strrchr(buf, '/')) == NULL) return (-1); *slashp = '\0'; return (0); } /* * If accept_ancestor is false, then check to make sure that the given path has * a parent, and that it exists. If accept_ancestor is true, then find the * closest existing ancestor for the given path. In prefixlen return the * length of already existing prefix of the given path. We also fetch the * 'zoned' property, which is used to validate property settings when creating * new datasets. */ static int check_parents(libzfs_handle_t *hdl, const char *path, uint64_t *zoned, boolean_t accept_ancestor, int *prefixlen) { zfs_cmd_t zc = { 0 }; char parent[ZFS_MAXNAMELEN]; char *slash; zfs_handle_t *zhp; char errbuf[1024]; uint64_t is_zoned; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot create '%s'"), path); /* get parent, and check to see if this is just a pool */ if (parent_name(path, parent, sizeof (parent)) != 0) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "missing dataset name")); return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); } /* check to see if the pool exists */ if ((slash = strchr(parent, '/')) == NULL) slash = parent + strlen(parent); (void) strncpy(zc.zc_name, parent, slash - parent); zc.zc_name[slash - parent] = '\0'; if (ioctl(hdl->libzfs_fd, ZFS_IOC_OBJSET_STATS, &zc) != 0 && errno == ENOENT) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool '%s'"), zc.zc_name); return (zfs_error(hdl, EZFS_NOENT, errbuf)); } /* check to see if the parent dataset exists */ while ((zhp = make_dataset_handle(hdl, parent)) == NULL) { if (errno == ENOENT && accept_ancestor) { /* * Go deeper to find an ancestor, give up on top level. */ if (parent_name(parent, parent, sizeof (parent)) != 0) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool '%s'"), zc.zc_name); return (zfs_error(hdl, EZFS_NOENT, errbuf)); } } else if (errno == ENOENT) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "parent does not exist")); return (zfs_error(hdl, EZFS_NOENT, errbuf)); } else return (zfs_standard_error(hdl, errno, errbuf)); } is_zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED); if (zoned != NULL) *zoned = is_zoned; /* we are in a non-global zone, but parent is in the global zone */ if (getzoneid() != GLOBAL_ZONEID && !is_zoned) { (void) zfs_standard_error(hdl, EPERM, errbuf); zfs_close(zhp); return (-1); } /* make sure parent is a filesystem */ if (zfs_get_type(zhp) != ZFS_TYPE_FILESYSTEM) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "parent is not a filesystem")); (void) zfs_error(hdl, EZFS_BADTYPE, errbuf); zfs_close(zhp); return (-1); } zfs_close(zhp); if (prefixlen != NULL) *prefixlen = strlen(parent); return (0); } /* * Finds whether the dataset of the given type(s) exists. */ boolean_t zfs_dataset_exists(libzfs_handle_t *hdl, const char *path, zfs_type_t types) { zfs_handle_t *zhp; if (!zfs_validate_name(hdl, path, types, B_FALSE)) return (B_FALSE); /* * Try to get stats for the dataset, which will tell us if it exists. */ if ((zhp = make_dataset_handle(hdl, path)) != NULL) { int ds_type = zhp->zfs_type; zfs_close(zhp); if (types & ds_type) return (B_TRUE); } return (B_FALSE); } /* * Given a path to 'target', create all the ancestors between * the prefixlen portion of the path, and the target itself. * Fail if the initial prefixlen-ancestor does not already exist. */ int create_parents(libzfs_handle_t *hdl, char *target, int prefixlen) { zfs_handle_t *h; char *cp; const char *opname; /* make sure prefix exists */ cp = target + prefixlen; if (*cp != '/') { assert(strchr(cp, '/') == NULL); h = zfs_open(hdl, target, ZFS_TYPE_FILESYSTEM); } else { *cp = '\0'; h = zfs_open(hdl, target, ZFS_TYPE_FILESYSTEM); *cp = '/'; } if (h == NULL) return (-1); zfs_close(h); /* * Attempt to create, mount, and share any ancestor filesystems, * up to the prefixlen-long one. */ for (cp = target + prefixlen + 1; cp = strchr(cp, '/'); *cp = '/', cp++) { *cp = '\0'; h = make_dataset_handle(hdl, target); if (h) { /* it already exists, nothing to do here */ zfs_close(h); continue; } if (zfs_create(hdl, target, ZFS_TYPE_FILESYSTEM, NULL) != 0) { opname = dgettext(TEXT_DOMAIN, "create"); goto ancestorerr; } h = zfs_open(hdl, target, ZFS_TYPE_FILESYSTEM); if (h == NULL) { opname = dgettext(TEXT_DOMAIN, "open"); goto ancestorerr; } if (zfs_mount(h, NULL, 0) != 0) { opname = dgettext(TEXT_DOMAIN, "mount"); goto ancestorerr; } if (zfs_share(h) != 0) { opname = dgettext(TEXT_DOMAIN, "share"); goto ancestorerr; } zfs_close(h); } return (0); ancestorerr: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "failed to %s ancestor '%s'"), opname, target); return (-1); } /* * Creates non-existing ancestors of the given path. */ int zfs_create_ancestors(libzfs_handle_t *hdl, const char *path) { int prefix; char *path_copy; int rc; if (check_parents(hdl, path, NULL, B_TRUE, &prefix) != 0) return (-1); if ((path_copy = strdup(path)) != NULL) { rc = create_parents(hdl, path_copy, prefix); free(path_copy); } if (path_copy == NULL || rc != 0) return (-1); return (0); } /* * Create a new filesystem or volume. */ int zfs_create(libzfs_handle_t *hdl, const char *path, zfs_type_t type, nvlist_t *props) { int ret; uint64_t size = 0; uint64_t blocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE); char errbuf[1024]; uint64_t zoned; dmu_objset_type_t ost; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot create '%s'"), path); /* validate the path, taking care to note the extended error message */ if (!zfs_validate_name(hdl, path, type, B_TRUE)) return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); /* validate parents exist */ if (check_parents(hdl, path, &zoned, B_FALSE, NULL) != 0) return (-1); /* * The failure modes when creating a dataset of a different type over * one that already exists is a little strange. In particular, if you * try to create a dataset on top of an existing dataset, the ioctl() * will return ENOENT, not EEXIST. To prevent this from happening, we * first try to see if the dataset exists. */ if (zfs_dataset_exists(hdl, path, ZFS_TYPE_DATASET)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "dataset already exists")); return (zfs_error(hdl, EZFS_EXISTS, errbuf)); } if (type == ZFS_TYPE_VOLUME) ost = DMU_OST_ZVOL; else ost = DMU_OST_ZFS; if (props && (props = zfs_valid_proplist(hdl, type, props, zoned, NULL, errbuf)) == 0) return (-1); if (type == ZFS_TYPE_VOLUME) { /* * If we are creating a volume, the size and block size must * satisfy a few restraints. First, the blocksize must be a * valid block size between SPA_{MIN,MAX}BLOCKSIZE. Second, the * volsize must be a multiple of the block size, and cannot be * zero. */ if (props == NULL || nvlist_lookup_uint64(props, zfs_prop_to_name(ZFS_PROP_VOLSIZE), &size) != 0) { nvlist_free(props); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "missing volume size")); return (zfs_error(hdl, EZFS_BADPROP, errbuf)); } if ((ret = nvlist_lookup_uint64(props, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &blocksize)) != 0) { if (ret == ENOENT) { blocksize = zfs_prop_default_numeric( ZFS_PROP_VOLBLOCKSIZE); } else { nvlist_free(props); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "missing volume block size")); return (zfs_error(hdl, EZFS_BADPROP, errbuf)); } } if (size == 0) { nvlist_free(props); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "volume size cannot be zero")); return (zfs_error(hdl, EZFS_BADPROP, errbuf)); } if (size % blocksize != 0) { nvlist_free(props); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "volume size must be a multiple of volume block " "size")); return (zfs_error(hdl, EZFS_BADPROP, errbuf)); } } /* create the dataset */ ret = lzc_create(path, ost, props); nvlist_free(props); /* check for failure */ if (ret != 0) { char parent[ZFS_MAXNAMELEN]; (void) parent_name(path, parent, sizeof (parent)); switch (errno) { case ENOENT: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such parent '%s'"), parent); return (zfs_error(hdl, EZFS_NOENT, errbuf)); case EINVAL: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "parent '%s' is not a filesystem"), parent); return (zfs_error(hdl, EZFS_BADTYPE, errbuf)); case EDOM: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "volume block size must be power of 2 from " "512B to 128KB")); return (zfs_error(hdl, EZFS_BADPROP, errbuf)); case ENOTSUP: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be upgraded to set this " "property or value")); return (zfs_error(hdl, EZFS_BADVERSION, errbuf)); #ifdef _ILP32 case EOVERFLOW: /* * This platform can't address a volume this big. */ if (type == ZFS_TYPE_VOLUME) return (zfs_error(hdl, EZFS_VOLTOOBIG, errbuf)); #endif /* FALLTHROUGH */ default: return (zfs_standard_error(hdl, errno, errbuf)); } } return (0); } /* * Destroys the given dataset. The caller must make sure that the filesystem * isn't mounted, and that there are no active dependents. If the file system * does not exist this function does nothing. */ int zfs_destroy(zfs_handle_t *zhp, boolean_t defer) { zfs_cmd_t zc = { 0 }; if (zhp->zfs_type == ZFS_TYPE_BOOKMARK) { nvlist_t *nv = fnvlist_alloc(); fnvlist_add_boolean(nv, zhp->zfs_name); int error = lzc_destroy_bookmarks(nv, NULL); fnvlist_free(nv); if (error != 0) { return (zfs_standard_error_fmt(zhp->zfs_hdl, errno, dgettext(TEXT_DOMAIN, "cannot destroy '%s'"), zhp->zfs_name)); } return (0); } (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); if (ZFS_IS_VOLUME(zhp)) { zc.zc_objset_type = DMU_OST_ZVOL; } else { zc.zc_objset_type = DMU_OST_ZFS; } zc.zc_defer_destroy = defer; if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_DESTROY, &zc) != 0 && errno != ENOENT) { return (zfs_standard_error_fmt(zhp->zfs_hdl, errno, dgettext(TEXT_DOMAIN, "cannot destroy '%s'"), zhp->zfs_name)); } remove_mountpoint(zhp); return (0); } struct destroydata { nvlist_t *nvl; const char *snapname; }; static int zfs_check_snap_cb(zfs_handle_t *zhp, void *arg) { struct destroydata *dd = arg; char name[ZFS_MAXNAMELEN]; int rv = 0; (void) snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name, dd->snapname); if (lzc_exists(name)) verify(nvlist_add_boolean(dd->nvl, name) == 0); rv = zfs_iter_filesystems(zhp, zfs_check_snap_cb, dd); zfs_close(zhp); return (rv); } /* * Destroys all snapshots with the given name in zhp & descendants. */ int zfs_destroy_snaps(zfs_handle_t *zhp, char *snapname, boolean_t defer) { int ret; struct destroydata dd = { 0 }; dd.snapname = snapname; verify(nvlist_alloc(&dd.nvl, NV_UNIQUE_NAME, 0) == 0); (void) zfs_check_snap_cb(zfs_handle_dup(zhp), &dd); if (nvlist_empty(dd.nvl)) { ret = zfs_standard_error_fmt(zhp->zfs_hdl, ENOENT, dgettext(TEXT_DOMAIN, "cannot destroy '%s@%s'"), zhp->zfs_name, snapname); } else { ret = zfs_destroy_snaps_nvl(zhp->zfs_hdl, dd.nvl, defer); } nvlist_free(dd.nvl); return (ret); } /* * Destroys all the snapshots named in the nvlist. */ int zfs_destroy_snaps_nvl(libzfs_handle_t *hdl, nvlist_t *snaps, boolean_t defer) { int ret; nvlist_t *errlist; ret = lzc_destroy_snaps(snaps, defer, &errlist); if (ret == 0) return (0); if (nvlist_empty(errlist)) { char errbuf[1024]; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot destroy snapshots")); ret = zfs_standard_error(hdl, ret, errbuf); } for (nvpair_t *pair = nvlist_next_nvpair(errlist, NULL); pair != NULL; pair = nvlist_next_nvpair(errlist, pair)) { char errbuf[1024]; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot destroy snapshot %s"), nvpair_name(pair)); switch (fnvpair_value_int32(pair)) { case EEXIST: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "snapshot is cloned")); ret = zfs_error(hdl, EZFS_EXISTS, errbuf); break; default: ret = zfs_standard_error(hdl, errno, errbuf); break; } } return (ret); } /* * Clones the given dataset. The target must be of the same type as the source. */ int zfs_clone(zfs_handle_t *zhp, const char *target, nvlist_t *props) { char parent[ZFS_MAXNAMELEN]; int ret; char errbuf[1024]; libzfs_handle_t *hdl = zhp->zfs_hdl; uint64_t zoned; assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT); (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot create '%s'"), target); /* validate the target/clone name */ if (!zfs_validate_name(hdl, target, ZFS_TYPE_FILESYSTEM, B_TRUE)) return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); /* validate parents exist */ if (check_parents(hdl, target, &zoned, B_FALSE, NULL) != 0) return (-1); (void) parent_name(target, parent, sizeof (parent)); /* do the clone */ if (props) { zfs_type_t type; if (ZFS_IS_VOLUME(zhp)) { type = ZFS_TYPE_VOLUME; } else { type = ZFS_TYPE_FILESYSTEM; } if ((props = zfs_valid_proplist(hdl, type, props, zoned, zhp, errbuf)) == NULL) return (-1); } ret = lzc_clone(target, zhp->zfs_name, props); nvlist_free(props); if (ret != 0) { switch (errno) { case ENOENT: /* * The parent doesn't exist. We should have caught this * above, but there may a race condition that has since * destroyed the parent. * * At this point, we don't know whether it's the source * that doesn't exist anymore, or whether the target * dataset doesn't exist. */ zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN, "no such parent '%s'"), parent); return (zfs_error(zhp->zfs_hdl, EZFS_NOENT, errbuf)); case EXDEV: zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN, "source and target pools differ")); return (zfs_error(zhp->zfs_hdl, EZFS_CROSSTARGET, errbuf)); default: return (zfs_standard_error(zhp->zfs_hdl, errno, errbuf)); } } return (ret); } /* * Promotes the given clone fs to be the clone parent. */ int zfs_promote(zfs_handle_t *zhp) { libzfs_handle_t *hdl = zhp->zfs_hdl; zfs_cmd_t zc = { 0 }; char parent[MAXPATHLEN]; int ret; char errbuf[1024]; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot promote '%s'"), zhp->zfs_name); if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "snapshots can not be promoted")); return (zfs_error(hdl, EZFS_BADTYPE, errbuf)); } (void) strlcpy(parent, zhp->zfs_dmustats.dds_origin, sizeof (parent)); if (parent[0] == '\0') { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "not a cloned filesystem")); return (zfs_error(hdl, EZFS_BADTYPE, errbuf)); } (void) strlcpy(zc.zc_value, zhp->zfs_dmustats.dds_origin, sizeof (zc.zc_value)); (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); ret = zfs_ioctl(hdl, ZFS_IOC_PROMOTE, &zc); if (ret != 0) { int save_errno = errno; switch (save_errno) { case EEXIST: /* There is a conflicting snapshot name. */ zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "conflicting snapshot '%s' from parent '%s'"), zc.zc_string, parent); return (zfs_error(hdl, EZFS_EXISTS, errbuf)); default: return (zfs_standard_error(hdl, save_errno, errbuf)); } } return (ret); } typedef struct snapdata { nvlist_t *sd_nvl; const char *sd_snapname; } snapdata_t; static int zfs_snapshot_cb(zfs_handle_t *zhp, void *arg) { snapdata_t *sd = arg; char name[ZFS_MAXNAMELEN]; int rv = 0; if (zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) == 0) { (void) snprintf(name, sizeof (name), "%s@%s", zfs_get_name(zhp), sd->sd_snapname); fnvlist_add_boolean(sd->sd_nvl, name); rv = zfs_iter_filesystems(zhp, zfs_snapshot_cb, sd); } zfs_close(zhp); return (rv); } /* * Creates snapshots. The keys in the snaps nvlist are the snapshots to be * created. */ int zfs_snapshot_nvl(libzfs_handle_t *hdl, nvlist_t *snaps, nvlist_t *props) { int ret; char errbuf[1024]; nvpair_t *elem; nvlist_t *errors; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot create snapshots ")); elem = NULL; while ((elem = nvlist_next_nvpair(snaps, elem)) != NULL) { const char *snapname = nvpair_name(elem); /* validate the target name */ if (!zfs_validate_name(hdl, snapname, ZFS_TYPE_SNAPSHOT, B_TRUE)) { (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot create snapshot '%s'"), snapname); return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); } } if (props != NULL && (props = zfs_valid_proplist(hdl, ZFS_TYPE_SNAPSHOT, props, B_FALSE, NULL, errbuf)) == NULL) { return (-1); } ret = lzc_snapshot(snaps, props, &errors); if (ret != 0) { boolean_t printed = B_FALSE; for (elem = nvlist_next_nvpair(errors, NULL); elem != NULL; elem = nvlist_next_nvpair(errors, elem)) { (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot create snapshot '%s'"), nvpair_name(elem)); (void) zfs_standard_error(hdl, fnvpair_value_int32(elem), errbuf); printed = B_TRUE; } if (!printed) { switch (ret) { case EXDEV: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "multiple snapshots of same " "fs not allowed")); (void) zfs_error(hdl, EZFS_EXISTS, errbuf); break; default: (void) zfs_standard_error(hdl, ret, errbuf); } } } nvlist_free(props); nvlist_free(errors); return (ret); } int zfs_snapshot(libzfs_handle_t *hdl, const char *path, boolean_t recursive, nvlist_t *props) { int ret; snapdata_t sd = { 0 }; char fsname[ZFS_MAXNAMELEN]; char *cp; zfs_handle_t *zhp; char errbuf[1024]; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot snapshot %s"), path); if (!zfs_validate_name(hdl, path, ZFS_TYPE_SNAPSHOT, B_TRUE)) return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); (void) strlcpy(fsname, path, sizeof (fsname)); cp = strchr(fsname, '@'); *cp = '\0'; sd.sd_snapname = cp + 1; if ((zhp = zfs_open(hdl, fsname, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME)) == NULL) { return (-1); } verify(nvlist_alloc(&sd.sd_nvl, NV_UNIQUE_NAME, 0) == 0); if (recursive) { (void) zfs_snapshot_cb(zfs_handle_dup(zhp), &sd); } else { fnvlist_add_boolean(sd.sd_nvl, path); } ret = zfs_snapshot_nvl(hdl, sd.sd_nvl, props); nvlist_free(sd.sd_nvl); zfs_close(zhp); return (ret); } /* * Destroy any more recent snapshots. We invoke this callback on any dependents * of the snapshot first. If the 'cb_dependent' member is non-zero, then this * is a dependent and we should just destroy it without checking the transaction * group. */ typedef struct rollback_data { const char *cb_target; /* the snapshot */ uint64_t cb_create; /* creation time reference */ boolean_t cb_error; boolean_t cb_force; } rollback_data_t; static int rollback_destroy_dependent(zfs_handle_t *zhp, void *data) { rollback_data_t *cbp = data; prop_changelist_t *clp; /* We must destroy this clone; first unmount it */ clp = changelist_gather(zhp, ZFS_PROP_NAME, 0, cbp->cb_force ? MS_FORCE: 0); if (clp == NULL || changelist_prefix(clp) != 0) { cbp->cb_error = B_TRUE; zfs_close(zhp); return (0); } if (zfs_destroy(zhp, B_FALSE) != 0) cbp->cb_error = B_TRUE; else changelist_remove(clp, zhp->zfs_name); (void) changelist_postfix(clp); changelist_free(clp); zfs_close(zhp); return (0); } static int rollback_destroy(zfs_handle_t *zhp, void *data) { rollback_data_t *cbp = data; if (zfs_prop_get_int(zhp, ZFS_PROP_CREATETXG) > cbp->cb_create) { cbp->cb_error |= zfs_iter_dependents(zhp, B_FALSE, rollback_destroy_dependent, cbp); cbp->cb_error |= zfs_destroy(zhp, B_FALSE); } zfs_close(zhp); return (0); } /* * Given a dataset, rollback to a specific snapshot, discarding any * data changes since then and making it the active dataset. * * Any snapshots and bookmarks more recent than the target are * destroyed, along with their dependents (i.e. clones). */ int zfs_rollback(zfs_handle_t *zhp, zfs_handle_t *snap, boolean_t force) { rollback_data_t cb = { 0 }; int err; boolean_t restore_resv = 0; uint64_t old_volsize, new_volsize; zfs_prop_t resv_prop; assert(zhp->zfs_type == ZFS_TYPE_FILESYSTEM || zhp->zfs_type == ZFS_TYPE_VOLUME); /* * Destroy all recent snapshots and their dependents. */ cb.cb_force = force; cb.cb_target = snap->zfs_name; cb.cb_create = zfs_prop_get_int(snap, ZFS_PROP_CREATETXG); (void) zfs_iter_snapshots(zhp, B_FALSE, rollback_destroy, &cb); (void) zfs_iter_bookmarks(zhp, rollback_destroy, &cb); if (cb.cb_error) return (-1); /* * Now that we have verified that the snapshot is the latest, * rollback to the given snapshot. */ if (zhp->zfs_type == ZFS_TYPE_VOLUME) { if (zfs_which_resv_prop(zhp, &resv_prop) < 0) return (-1); old_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE); restore_resv = (old_volsize == zfs_prop_get_int(zhp, resv_prop)); } /* * We rely on zfs_iter_children() to verify that there are no * newer snapshots for the given dataset. Therefore, we can * simply pass the name on to the ioctl() call. There is still * an unlikely race condition where the user has taken a * snapshot since we verified that this was the most recent. */ err = lzc_rollback(zhp->zfs_name, NULL, 0); if (err != 0) { (void) zfs_standard_error_fmt(zhp->zfs_hdl, errno, dgettext(TEXT_DOMAIN, "cannot rollback '%s'"), zhp->zfs_name); return (err); } /* * For volumes, if the pre-rollback volsize matched the pre- * rollback reservation and the volsize has changed then set * the reservation property to the post-rollback volsize. * Make a new handle since the rollback closed the dataset. */ if ((zhp->zfs_type == ZFS_TYPE_VOLUME) && (zhp = make_dataset_handle(zhp->zfs_hdl, zhp->zfs_name))) { if (restore_resv) { new_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE); if (old_volsize != new_volsize) err = zfs_prop_set_int(zhp, resv_prop, new_volsize); } zfs_close(zhp); } return (err); } /* * Renames the given dataset. */ int zfs_rename(zfs_handle_t *zhp, const char *source, const char *target, renameflags_t flags) { int ret; zfs_cmd_t zc = { 0 }; char *delim; prop_changelist_t *cl = NULL; zfs_handle_t *zhrp = NULL; char *parentname = NULL; char parent[ZFS_MAXNAMELEN]; char property[ZFS_MAXPROPLEN]; libzfs_handle_t *hdl = zhp->zfs_hdl; char errbuf[1024]; /* if we have the same exact name, just return success */ if (strcmp(zhp->zfs_name, target) == 0) return (0); (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot rename to '%s'"), target); if (source != NULL) { /* * This is recursive snapshots rename, put snapshot name * (that might not exist) into zfs_name. */ assert(flags.recurse); (void) strlcat(zhp->zfs_name, "@", sizeof(zhp->zfs_name)); (void) strlcat(zhp->zfs_name, source, sizeof(zhp->zfs_name)); zhp->zfs_type = ZFS_TYPE_SNAPSHOT; } /* * Make sure the target name is valid */ if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) { if ((strchr(target, '@') == NULL) || *target == '@') { /* * Snapshot target name is abbreviated, * reconstruct full dataset name */ (void) strlcpy(parent, zhp->zfs_name, sizeof (parent)); delim = strchr(parent, '@'); if (strchr(target, '@') == NULL) *(++delim) = '\0'; else *delim = '\0'; (void) strlcat(parent, target, sizeof (parent)); target = parent; } else { /* * Make sure we're renaming within the same dataset. */ delim = strchr(target, '@'); if (strncmp(zhp->zfs_name, target, delim - target) != 0 || zhp->zfs_name[delim - target] != '@') { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "snapshots must be part of same " "dataset")); return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf)); } } if (!zfs_validate_name(hdl, target, zhp->zfs_type, B_TRUE)) return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); } else { if (flags.recurse) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "recursive rename must be a snapshot")); return (zfs_error(hdl, EZFS_BADTYPE, errbuf)); } if (!zfs_validate_name(hdl, target, zhp->zfs_type, B_TRUE)) return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); /* validate parents */ if (check_parents(hdl, target, NULL, B_FALSE, NULL) != 0) return (-1); /* make sure we're in the same pool */ verify((delim = strchr(target, '/')) != NULL); if (strncmp(zhp->zfs_name, target, delim - target) != 0 || zhp->zfs_name[delim - target] != '/') { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "datasets must be within same pool")); return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf)); } /* new name cannot be a child of the current dataset name */ if (is_descendant(zhp->zfs_name, target)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "New dataset name cannot be a descendant of " "current dataset name")); return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); } } (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot rename '%s'"), zhp->zfs_name); if (getzoneid() == GLOBAL_ZONEID && zfs_prop_get_int(zhp, ZFS_PROP_ZONED)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "dataset is used in a non-global zone")); return (zfs_error(hdl, EZFS_ZONED, errbuf)); } /* * Avoid unmounting file systems with mountpoint property set to * 'legacy' or 'none' even if -u option is not given. */ if (zhp->zfs_type == ZFS_TYPE_FILESYSTEM && !flags.recurse && !flags.nounmount && zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, property, sizeof (property), NULL, NULL, 0, B_FALSE) == 0 && (strcmp(property, "legacy") == 0 || strcmp(property, "none") == 0)) { flags.nounmount = B_TRUE; } if (flags.recurse) { parentname = zfs_strdup(zhp->zfs_hdl, zhp->zfs_name); if (parentname == NULL) { ret = -1; goto error; } delim = strchr(parentname, '@'); *delim = '\0'; zhrp = zfs_open(zhp->zfs_hdl, parentname, ZFS_TYPE_DATASET); if (zhrp == NULL) { ret = -1; goto error; } } else if (zhp->zfs_type != ZFS_TYPE_SNAPSHOT) { if ((cl = changelist_gather(zhp, ZFS_PROP_NAME, flags.nounmount ? CL_GATHER_DONT_UNMOUNT : 0, flags.forceunmount ? MS_FORCE : 0)) == NULL) { return (-1); } if (changelist_haszonedchild(cl)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "child dataset with inherited mountpoint is used " "in a non-global zone")); (void) zfs_error(hdl, EZFS_ZONED, errbuf); goto error; } if ((ret = changelist_prefix(cl)) != 0) goto error; } if (ZFS_IS_VOLUME(zhp)) zc.zc_objset_type = DMU_OST_ZVOL; else zc.zc_objset_type = DMU_OST_ZFS; (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); (void) strlcpy(zc.zc_value, target, sizeof (zc.zc_value)); zc.zc_cookie = flags.recurse ? 1 : 0; if (flags.nounmount) zc.zc_cookie |= 2; if ((ret = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_RENAME, &zc)) != 0) { /* * if it was recursive, the one that actually failed will * be in zc.zc_name */ (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot rename '%s'"), zc.zc_name); if (flags.recurse && errno == EEXIST) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "a child dataset already has a snapshot " "with the new name")); (void) zfs_error(hdl, EZFS_EXISTS, errbuf); } else { (void) zfs_standard_error(zhp->zfs_hdl, errno, errbuf); } /* * On failure, we still want to remount any filesystems that * were previously mounted, so we don't alter the system state. */ if (cl != NULL) (void) changelist_postfix(cl); } else { if (cl != NULL) { changelist_rename(cl, zfs_get_name(zhp), target); ret = changelist_postfix(cl); } } error: if (parentname != NULL) { free(parentname); } if (zhrp != NULL) { zfs_close(zhrp); } if (cl != NULL) { changelist_free(cl); } return (ret); } nvlist_t * zfs_get_user_props(zfs_handle_t *zhp) { return (zhp->zfs_user_props); } nvlist_t * zfs_get_recvd_props(zfs_handle_t *zhp) { if (zhp->zfs_recvd_props == NULL) if (get_recvd_props_ioctl(zhp) != 0) return (NULL); return (zhp->zfs_recvd_props); } /* * This function is used by 'zfs list' to determine the exact set of columns to * display, and their maximum widths. This does two main things: * * - If this is a list of all properties, then expand the list to include * all native properties, and set a flag so that for each dataset we look * for new unique user properties and add them to the list. * * - For non fixed-width properties, keep track of the maximum width seen * so that we can size the column appropriately. If the user has * requested received property values, we also need to compute the width * of the RECEIVED column. */ int zfs_expand_proplist(zfs_handle_t *zhp, zprop_list_t **plp, boolean_t received, boolean_t literal) { libzfs_handle_t *hdl = zhp->zfs_hdl; zprop_list_t *entry; zprop_list_t **last, **start; nvlist_t *userprops, *propval; nvpair_t *elem; char *strval; char buf[ZFS_MAXPROPLEN]; if (zprop_expand_list(hdl, plp, ZFS_TYPE_DATASET) != 0) return (-1); userprops = zfs_get_user_props(zhp); entry = *plp; if (entry->pl_all && nvlist_next_nvpair(userprops, NULL) != NULL) { /* * Go through and add any user properties as necessary. We * start by incrementing our list pointer to the first * non-native property. */ start = plp; while (*start != NULL) { if ((*start)->pl_prop == ZPROP_INVAL) break; start = &(*start)->pl_next; } elem = NULL; while ((elem = nvlist_next_nvpair(userprops, elem)) != NULL) { /* * See if we've already found this property in our list. */ for (last = start; *last != NULL; last = &(*last)->pl_next) { if (strcmp((*last)->pl_user_prop, nvpair_name(elem)) == 0) break; } if (*last == NULL) { if ((entry = zfs_alloc(hdl, sizeof (zprop_list_t))) == NULL || ((entry->pl_user_prop = zfs_strdup(hdl, nvpair_name(elem)))) == NULL) { free(entry); return (-1); } entry->pl_prop = ZPROP_INVAL; entry->pl_width = strlen(nvpair_name(elem)); entry->pl_all = B_TRUE; *last = entry; } } } /* * Now go through and check the width of any non-fixed columns */ for (entry = *plp; entry != NULL; entry = entry->pl_next) { if (entry->pl_fixed && !literal) continue; if (entry->pl_prop != ZPROP_INVAL) { if (zfs_prop_get(zhp, entry->pl_prop, buf, sizeof (buf), NULL, NULL, 0, literal) == 0) { if (strlen(buf) > entry->pl_width) entry->pl_width = strlen(buf); } if (received && zfs_prop_get_recvd(zhp, zfs_prop_to_name(entry->pl_prop), buf, sizeof (buf), literal) == 0) if (strlen(buf) > entry->pl_recvd_width) entry->pl_recvd_width = strlen(buf); } else { if (nvlist_lookup_nvlist(userprops, entry->pl_user_prop, &propval) == 0) { verify(nvlist_lookup_string(propval, ZPROP_VALUE, &strval) == 0); if (strlen(strval) > entry->pl_width) entry->pl_width = strlen(strval); } if (received && zfs_prop_get_recvd(zhp, entry->pl_user_prop, buf, sizeof (buf), literal) == 0) if (strlen(buf) > entry->pl_recvd_width) entry->pl_recvd_width = strlen(buf); } } return (0); } int zfs_deleg_share_nfs(libzfs_handle_t *hdl, char *dataset, char *path, char *resource, void *export, void *sharetab, int sharemax, zfs_share_op_t operation) { zfs_cmd_t zc = { 0 }; int error; (void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name)); (void) strlcpy(zc.zc_value, path, sizeof (zc.zc_value)); if (resource) (void) strlcpy(zc.zc_string, resource, sizeof (zc.zc_string)); zc.zc_share.z_sharedata = (uint64_t)(uintptr_t)sharetab; zc.zc_share.z_exportdata = (uint64_t)(uintptr_t)export; zc.zc_share.z_sharetype = operation; zc.zc_share.z_sharemax = sharemax; error = ioctl(hdl->libzfs_fd, ZFS_IOC_SHARE, &zc); return (error); } void zfs_prune_proplist(zfs_handle_t *zhp, uint8_t *props) { nvpair_t *curr; /* * Keep a reference to the props-table against which we prune the * properties. */ zhp->zfs_props_table = props; curr = nvlist_next_nvpair(zhp->zfs_props, NULL); while (curr) { zfs_prop_t zfs_prop = zfs_name_to_prop(nvpair_name(curr)); nvpair_t *next = nvlist_next_nvpair(zhp->zfs_props, curr); /* * User properties will result in ZPROP_INVAL, and since we * only know how to prune standard ZFS properties, we always * leave these in the list. This can also happen if we * encounter an unknown DSL property (when running older * software, for example). */ if (zfs_prop != ZPROP_INVAL && props[zfs_prop] == B_FALSE) (void) nvlist_remove(zhp->zfs_props, nvpair_name(curr), nvpair_type(curr)); curr = next; } } #ifdef sun static int zfs_smb_acl_mgmt(libzfs_handle_t *hdl, char *dataset, char *path, zfs_smb_acl_op_t cmd, char *resource1, char *resource2) { zfs_cmd_t zc = { 0 }; nvlist_t *nvlist = NULL; int error; (void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name)); (void) strlcpy(zc.zc_value, path, sizeof (zc.zc_value)); zc.zc_cookie = (uint64_t)cmd; if (cmd == ZFS_SMB_ACL_RENAME) { if (nvlist_alloc(&nvlist, NV_UNIQUE_NAME, 0) != 0) { (void) no_memory(hdl); return (NULL); } } switch (cmd) { case ZFS_SMB_ACL_ADD: case ZFS_SMB_ACL_REMOVE: (void) strlcpy(zc.zc_string, resource1, sizeof (zc.zc_string)); break; case ZFS_SMB_ACL_RENAME: if (nvlist_add_string(nvlist, ZFS_SMB_ACL_SRC, resource1) != 0) { (void) no_memory(hdl); return (-1); } if (nvlist_add_string(nvlist, ZFS_SMB_ACL_TARGET, resource2) != 0) { (void) no_memory(hdl); return (-1); } if (zcmd_write_src_nvlist(hdl, &zc, nvlist) != 0) { nvlist_free(nvlist); return (-1); } break; case ZFS_SMB_ACL_PURGE: break; default: return (-1); } error = ioctl(hdl->libzfs_fd, ZFS_IOC_SMB_ACL, &zc); if (nvlist) nvlist_free(nvlist); return (error); } int zfs_smb_acl_add(libzfs_handle_t *hdl, char *dataset, char *path, char *resource) { return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_ADD, resource, NULL)); } int zfs_smb_acl_remove(libzfs_handle_t *hdl, char *dataset, char *path, char *resource) { return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_REMOVE, resource, NULL)); } int zfs_smb_acl_purge(libzfs_handle_t *hdl, char *dataset, char *path) { return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_PURGE, NULL, NULL)); } int zfs_smb_acl_rename(libzfs_handle_t *hdl, char *dataset, char *path, char *oldname, char *newname) { return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_RENAME, oldname, newname)); } #endif /* sun */ int zfs_userspace(zfs_handle_t *zhp, zfs_userquota_prop_t type, zfs_userspace_cb_t func, void *arg) { zfs_cmd_t zc = { 0 }; zfs_useracct_t buf[100]; libzfs_handle_t *hdl = zhp->zfs_hdl; int ret; (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); zc.zc_objset_type = type; zc.zc_nvlist_dst = (uintptr_t)buf; for (;;) { zfs_useracct_t *zua = buf; zc.zc_nvlist_dst_size = sizeof (buf); if (zfs_ioctl(hdl, ZFS_IOC_USERSPACE_MANY, &zc) != 0) { char errbuf[1024]; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot get used/quota for %s"), zc.zc_name); return (zfs_standard_error_fmt(hdl, errno, errbuf)); } if (zc.zc_nvlist_dst_size == 0) break; while (zc.zc_nvlist_dst_size > 0) { if ((ret = func(arg, zua->zu_domain, zua->zu_rid, zua->zu_space)) != 0) return (ret); zua++; zc.zc_nvlist_dst_size -= sizeof (zfs_useracct_t); } } return (0); } struct holdarg { nvlist_t *nvl; const char *snapname; const char *tag; boolean_t recursive; int error; }; static int zfs_hold_one(zfs_handle_t *zhp, void *arg) { struct holdarg *ha = arg; char name[ZFS_MAXNAMELEN]; int rv = 0; (void) snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name, ha->snapname); if (lzc_exists(name)) fnvlist_add_string(ha->nvl, name, ha->tag); if (ha->recursive) rv = zfs_iter_filesystems(zhp, zfs_hold_one, ha); zfs_close(zhp); return (rv); } int zfs_hold(zfs_handle_t *zhp, const char *snapname, const char *tag, boolean_t recursive, int cleanup_fd) { int ret; struct holdarg ha; ha.nvl = fnvlist_alloc(); ha.snapname = snapname; ha.tag = tag; ha.recursive = recursive; (void) zfs_hold_one(zfs_handle_dup(zhp), &ha); if (nvlist_empty(ha.nvl)) { char errbuf[1024]; fnvlist_free(ha.nvl); ret = ENOENT; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot hold snapshot '%s@%s'"), zhp->zfs_name, snapname); (void) zfs_standard_error(zhp->zfs_hdl, ret, errbuf); return (ret); } ret = zfs_hold_nvl(zhp, cleanup_fd, ha.nvl); fnvlist_free(ha.nvl); return (ret); } int zfs_hold_nvl(zfs_handle_t *zhp, int cleanup_fd, nvlist_t *holds) { int ret; nvlist_t *errors; libzfs_handle_t *hdl = zhp->zfs_hdl; char errbuf[1024]; nvpair_t *elem; errors = NULL; ret = lzc_hold(holds, cleanup_fd, &errors); if (ret == 0) { /* There may be errors even in the success case. */ fnvlist_free(errors); return (0); } if (nvlist_empty(errors)) { /* no hold-specific errors */ (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot hold")); switch (ret) { case ENOTSUP: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be upgraded")); (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); break; case EINVAL: (void) zfs_error(hdl, EZFS_BADTYPE, errbuf); break; default: (void) zfs_standard_error(hdl, ret, errbuf); } } for (elem = nvlist_next_nvpair(errors, NULL); elem != NULL; elem = nvlist_next_nvpair(errors, elem)) { (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot hold snapshot '%s'"), nvpair_name(elem)); switch (fnvpair_value_int32(elem)) { case E2BIG: /* * Temporary tags wind up having the ds object id * prepended. So even if we passed the length check * above, it's still possible for the tag to wind * up being slightly too long. */ (void) zfs_error(hdl, EZFS_TAGTOOLONG, errbuf); break; case EINVAL: (void) zfs_error(hdl, EZFS_BADTYPE, errbuf); break; case EEXIST: (void) zfs_error(hdl, EZFS_REFTAG_HOLD, errbuf); break; default: (void) zfs_standard_error(hdl, fnvpair_value_int32(elem), errbuf); } } fnvlist_free(errors); return (ret); } static int zfs_release_one(zfs_handle_t *zhp, void *arg) { struct holdarg *ha = arg; char name[ZFS_MAXNAMELEN]; int rv = 0; nvlist_t *existing_holds; (void) snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name, ha->snapname); if (lzc_get_holds(name, &existing_holds) != 0) { ha->error = ENOENT; } else if (!nvlist_exists(existing_holds, ha->tag)) { ha->error = ESRCH; } else { nvlist_t *torelease = fnvlist_alloc(); fnvlist_add_boolean(torelease, ha->tag); fnvlist_add_nvlist(ha->nvl, name, torelease); fnvlist_free(torelease); } if (ha->recursive) rv = zfs_iter_filesystems(zhp, zfs_release_one, ha); zfs_close(zhp); return (rv); } int zfs_release(zfs_handle_t *zhp, const char *snapname, const char *tag, boolean_t recursive) { int ret; struct holdarg ha; nvlist_t *errors = NULL; nvpair_t *elem; libzfs_handle_t *hdl = zhp->zfs_hdl; char errbuf[1024]; ha.nvl = fnvlist_alloc(); ha.snapname = snapname; ha.tag = tag; ha.recursive = recursive; ha.error = 0; (void) zfs_release_one(zfs_handle_dup(zhp), &ha); if (nvlist_empty(ha.nvl)) { fnvlist_free(ha.nvl); ret = ha.error; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot release hold from snapshot '%s@%s'"), zhp->zfs_name, snapname); if (ret == ESRCH) { (void) zfs_error(hdl, EZFS_REFTAG_RELE, errbuf); } else { (void) zfs_standard_error(hdl, ret, errbuf); } return (ret); } ret = lzc_release(ha.nvl, &errors); fnvlist_free(ha.nvl); if (ret == 0) { /* There may be errors even in the success case. */ fnvlist_free(errors); return (0); } if (nvlist_empty(errors)) { /* no hold-specific errors */ (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot release")); switch (errno) { case ENOTSUP: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be upgraded")); (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); break; default: (void) zfs_standard_error_fmt(hdl, errno, errbuf); } } for (elem = nvlist_next_nvpair(errors, NULL); elem != NULL; elem = nvlist_next_nvpair(errors, elem)) { (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot release hold from snapshot '%s'"), nvpair_name(elem)); switch (fnvpair_value_int32(elem)) { case ESRCH: (void) zfs_error(hdl, EZFS_REFTAG_RELE, errbuf); break; case EINVAL: (void) zfs_error(hdl, EZFS_BADTYPE, errbuf); break; default: (void) zfs_standard_error_fmt(hdl, fnvpair_value_int32(elem), errbuf); } } fnvlist_free(errors); return (ret); } int zfs_get_fsacl(zfs_handle_t *zhp, nvlist_t **nvl) { zfs_cmd_t zc = { 0 }; libzfs_handle_t *hdl = zhp->zfs_hdl; int nvsz = 2048; void *nvbuf; int err = 0; char errbuf[1024]; assert(zhp->zfs_type == ZFS_TYPE_VOLUME || zhp->zfs_type == ZFS_TYPE_FILESYSTEM); tryagain: nvbuf = malloc(nvsz); if (nvbuf == NULL) { err = (zfs_error(hdl, EZFS_NOMEM, strerror(errno))); goto out; } zc.zc_nvlist_dst_size = nvsz; zc.zc_nvlist_dst = (uintptr_t)nvbuf; (void) strlcpy(zc.zc_name, zhp->zfs_name, ZFS_MAXNAMELEN); if (ioctl(hdl->libzfs_fd, ZFS_IOC_GET_FSACL, &zc) != 0) { (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot get permissions on '%s'"), zc.zc_name); switch (errno) { case ENOMEM: free(nvbuf); nvsz = zc.zc_nvlist_dst_size; goto tryagain; case ENOTSUP: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be upgraded")); err = zfs_error(hdl, EZFS_BADVERSION, errbuf); break; case EINVAL: err = zfs_error(hdl, EZFS_BADTYPE, errbuf); break; case ENOENT: err = zfs_error(hdl, EZFS_NOENT, errbuf); break; default: err = zfs_standard_error_fmt(hdl, errno, errbuf); break; } } else { /* success */ int rc = nvlist_unpack(nvbuf, zc.zc_nvlist_dst_size, nvl, 0); if (rc) { (void) snprintf(errbuf, sizeof (errbuf), dgettext( TEXT_DOMAIN, "cannot get permissions on '%s'"), zc.zc_name); err = zfs_standard_error_fmt(hdl, rc, errbuf); } } free(nvbuf); out: return (err); } int zfs_set_fsacl(zfs_handle_t *zhp, boolean_t un, nvlist_t *nvl) { zfs_cmd_t zc = { 0 }; libzfs_handle_t *hdl = zhp->zfs_hdl; char *nvbuf; char errbuf[1024]; size_t nvsz; int err; assert(zhp->zfs_type == ZFS_TYPE_VOLUME || zhp->zfs_type == ZFS_TYPE_FILESYSTEM); err = nvlist_size(nvl, &nvsz, NV_ENCODE_NATIVE); assert(err == 0); nvbuf = malloc(nvsz); err = nvlist_pack(nvl, &nvbuf, &nvsz, NV_ENCODE_NATIVE, 0); assert(err == 0); zc.zc_nvlist_src_size = nvsz; zc.zc_nvlist_src = (uintptr_t)nvbuf; zc.zc_perm_action = un; (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); if (zfs_ioctl(hdl, ZFS_IOC_SET_FSACL, &zc) != 0) { (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot set permissions on '%s'"), zc.zc_name); switch (errno) { case ENOTSUP: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be upgraded")); err = zfs_error(hdl, EZFS_BADVERSION, errbuf); break; case EINVAL: err = zfs_error(hdl, EZFS_BADTYPE, errbuf); break; case ENOENT: err = zfs_error(hdl, EZFS_NOENT, errbuf); break; default: err = zfs_standard_error_fmt(hdl, errno, errbuf); break; } } free(nvbuf); return (err); } int zfs_get_holds(zfs_handle_t *zhp, nvlist_t **nvl) { int err; char errbuf[1024]; err = lzc_get_holds(zhp->zfs_name, nvl); if (err != 0) { libzfs_handle_t *hdl = zhp->zfs_hdl; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot get holds for '%s'"), zhp->zfs_name); switch (err) { case ENOTSUP: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be upgraded")); err = zfs_error(hdl, EZFS_BADVERSION, errbuf); break; case EINVAL: err = zfs_error(hdl, EZFS_BADTYPE, errbuf); break; case ENOENT: err = zfs_error(hdl, EZFS_NOENT, errbuf); break; default: err = zfs_standard_error_fmt(hdl, errno, errbuf); break; } } return (err); } /* * Convert the zvol's volume size to an appropriate reservation. * Note: If this routine is updated, it is necessary to update the ZFS test * suite's shell version in reservation.kshlib. */ uint64_t zvol_volsize_to_reservation(uint64_t volsize, nvlist_t *props) { uint64_t numdb; uint64_t nblocks, volblocksize; int ncopies; char *strval; if (nvlist_lookup_string(props, zfs_prop_to_name(ZFS_PROP_COPIES), &strval) == 0) ncopies = atoi(strval); else ncopies = 1; if (nvlist_lookup_uint64(props, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &volblocksize) != 0) volblocksize = ZVOL_DEFAULT_BLOCKSIZE; nblocks = volsize/volblocksize; /* start with metadnode L0-L6 */ numdb = 7; /* calculate number of indirects */ while (nblocks > 1) { nblocks += DNODES_PER_LEVEL - 1; nblocks /= DNODES_PER_LEVEL; numdb += nblocks; } numdb *= MIN(SPA_DVAS_PER_BP, ncopies + 1); volsize *= ncopies; /* * this is exactly DN_MAX_INDBLKSHIFT when metadata isn't * compressed, but in practice they compress down to about * 1100 bytes */ numdb *= 1ULL << DN_MAX_INDBLKSHIFT; volsize += numdb; return (volsize); } /* * Attach/detach the given filesystem to/from the given jail. */ int zfs_jail(zfs_handle_t *zhp, int jailid, int attach) { libzfs_handle_t *hdl = zhp->zfs_hdl; zfs_cmd_t zc = { 0 }; char errbuf[1024]; unsigned long cmd; int ret; if (attach) { (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot jail '%s'"), zhp->zfs_name); } else { (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot unjail '%s'"), zhp->zfs_name); } switch (zhp->zfs_type) { case ZFS_TYPE_VOLUME: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "volumes can not be jailed")); return (zfs_error(hdl, EZFS_BADTYPE, errbuf)); case ZFS_TYPE_SNAPSHOT: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "snapshots can not be jailed")); return (zfs_error(hdl, EZFS_BADTYPE, errbuf)); } assert(zhp->zfs_type == ZFS_TYPE_FILESYSTEM); (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); zc.zc_objset_type = DMU_OST_ZFS; zc.zc_jailid = jailid; cmd = attach ? ZFS_IOC_JAIL : ZFS_IOC_UNJAIL; if ((ret = ioctl(hdl->libzfs_fd, cmd, &zc)) != 0) zfs_standard_error(hdl, errno, errbuf); return (ret); } Index: stable/10/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_sendrecv.c =================================================================== --- stable/10/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_sendrecv.c (revision 290756) +++ stable/10/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_sendrecv.c (revision 290757) @@ -1,3661 +1,3661 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2015 by Delphix. All rights reserved. * Copyright (c) 2012, Joyent, Inc. All rights reserved. * Copyright (c) 2012 Pawel Jakub Dawidek . * All rights reserved. * Copyright (c) 2013 Steven Hartland. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "zfs_namecheck.h" #include "zfs_prop.h" #include "zfs_fletcher.h" #include "libzfs_impl.h" #include #include #include #include #ifdef __FreeBSD__ extern int zfs_ioctl_version; #endif /* in libzfs_dataset.c */ extern void zfs_setprop_error(libzfs_handle_t *, zfs_prop_t, int, char *); /* We need to use something for ENODATA. */ #define ENODATA EIDRM static int zfs_receive_impl(libzfs_handle_t *, const char *, const char *, recvflags_t *, int, const char *, nvlist_t *, avl_tree_t *, char **, int, uint64_t *); static int guid_to_name(libzfs_handle_t *, const char *, uint64_t, boolean_t, char *); static const zio_cksum_t zero_cksum = { 0 }; typedef struct dedup_arg { int inputfd; int outputfd; libzfs_handle_t *dedup_hdl; } dedup_arg_t; typedef struct progress_arg { zfs_handle_t *pa_zhp; int pa_fd; boolean_t pa_parsable; } progress_arg_t; typedef struct dataref { uint64_t ref_guid; uint64_t ref_object; uint64_t ref_offset; } dataref_t; typedef struct dedup_entry { struct dedup_entry *dde_next; zio_cksum_t dde_chksum; uint64_t dde_prop; dataref_t dde_ref; } dedup_entry_t; #define MAX_DDT_PHYSMEM_PERCENT 20 #define SMALLEST_POSSIBLE_MAX_DDT_MB 128 typedef struct dedup_table { dedup_entry_t **dedup_hash_array; umem_cache_t *ddecache; uint64_t max_ddt_size; /* max dedup table size in bytes */ uint64_t cur_ddt_size; /* current dedup table size in bytes */ uint64_t ddt_count; int numhashbits; boolean_t ddt_full; } dedup_table_t; static int high_order_bit(uint64_t n) { int count; for (count = 0; n != 0; count++) n >>= 1; return (count); } static size_t ssread(void *buf, size_t len, FILE *stream) { size_t outlen; if ((outlen = fread(buf, len, 1, stream)) == 0) return (0); return (outlen); } static void ddt_hash_append(libzfs_handle_t *hdl, dedup_table_t *ddt, dedup_entry_t **ddepp, zio_cksum_t *cs, uint64_t prop, dataref_t *dr) { dedup_entry_t *dde; if (ddt->cur_ddt_size >= ddt->max_ddt_size) { if (ddt->ddt_full == B_FALSE) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Dedup table full. Deduplication will continue " "with existing table entries")); ddt->ddt_full = B_TRUE; } return; } if ((dde = umem_cache_alloc(ddt->ddecache, UMEM_DEFAULT)) != NULL) { assert(*ddepp == NULL); dde->dde_next = NULL; dde->dde_chksum = *cs; dde->dde_prop = prop; dde->dde_ref = *dr; *ddepp = dde; ddt->cur_ddt_size += sizeof (dedup_entry_t); ddt->ddt_count++; } } /* * Using the specified dedup table, do a lookup for an entry with * the checksum cs. If found, return the block's reference info * in *dr. Otherwise, insert a new entry in the dedup table, using * the reference information specified by *dr. * * return value: true - entry was found * false - entry was not found */ static boolean_t ddt_update(libzfs_handle_t *hdl, dedup_table_t *ddt, zio_cksum_t *cs, uint64_t prop, dataref_t *dr) { uint32_t hashcode; dedup_entry_t **ddepp; hashcode = BF64_GET(cs->zc_word[0], 0, ddt->numhashbits); for (ddepp = &(ddt->dedup_hash_array[hashcode]); *ddepp != NULL; ddepp = &((*ddepp)->dde_next)) { if (ZIO_CHECKSUM_EQUAL(((*ddepp)->dde_chksum), *cs) && (*ddepp)->dde_prop == prop) { *dr = (*ddepp)->dde_ref; return (B_TRUE); } } ddt_hash_append(hdl, ddt, ddepp, cs, prop, dr); return (B_FALSE); } static int dump_record(dmu_replay_record_t *drr, void *payload, int payload_len, zio_cksum_t *zc, int outfd) { ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), ==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t)); fletcher_4_incremental_native(drr, offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), zc); if (drr->drr_type != DRR_BEGIN) { ASSERT(ZIO_CHECKSUM_IS_ZERO(&drr->drr_u. drr_checksum.drr_checksum)); drr->drr_u.drr_checksum.drr_checksum = *zc; } fletcher_4_incremental_native(&drr->drr_u.drr_checksum.drr_checksum, sizeof (zio_cksum_t), zc); if (write(outfd, drr, sizeof (*drr)) == -1) return (errno); if (payload_len != 0) { fletcher_4_incremental_native(payload, payload_len, zc); if (write(outfd, payload, payload_len) == -1) return (errno); } return (0); } /* * This function is started in a separate thread when the dedup option * has been requested. The main send thread determines the list of * snapshots to be included in the send stream and makes the ioctl calls * for each one. But instead of having the ioctl send the output to the * the output fd specified by the caller of zfs_send()), the * ioctl is told to direct the output to a pipe, which is read by the * alternate thread running THIS function. This function does the * dedup'ing by: * 1. building a dedup table (the DDT) * 2. doing checksums on each data block and inserting a record in the DDT * 3. looking for matching checksums, and * 4. sending a DRR_WRITE_BYREF record instead of a write record whenever * a duplicate block is found. * The output of this function then goes to the output fd requested * by the caller of zfs_send(). */ static void * cksummer(void *arg) { dedup_arg_t *dda = arg; char *buf = zfs_alloc(dda->dedup_hdl, SPA_MAXBLOCKSIZE); dmu_replay_record_t thedrr; dmu_replay_record_t *drr = &thedrr; FILE *ofp; int outfd; dedup_table_t ddt; zio_cksum_t stream_cksum; uint64_t physmem = sysconf(_SC_PHYS_PAGES) * sysconf(_SC_PAGESIZE); uint64_t numbuckets; ddt.max_ddt_size = MAX((physmem * MAX_DDT_PHYSMEM_PERCENT) / 100, SMALLEST_POSSIBLE_MAX_DDT_MB << 20); numbuckets = ddt.max_ddt_size / (sizeof (dedup_entry_t)); /* * numbuckets must be a power of 2. Increase number to * a power of 2 if necessary. */ if (!ISP2(numbuckets)) numbuckets = 1 << high_order_bit(numbuckets); ddt.dedup_hash_array = calloc(numbuckets, sizeof (dedup_entry_t *)); ddt.ddecache = umem_cache_create("dde", sizeof (dedup_entry_t), 0, NULL, NULL, NULL, NULL, NULL, 0); ddt.cur_ddt_size = numbuckets * sizeof (dedup_entry_t *); ddt.numhashbits = high_order_bit(numbuckets) - 1; ddt.ddt_full = B_FALSE; outfd = dda->outputfd; ofp = fdopen(dda->inputfd, "r"); while (ssread(drr, sizeof (*drr), ofp) != 0) { switch (drr->drr_type) { case DRR_BEGIN: { struct drr_begin *drrb = &drr->drr_u.drr_begin; int fflags; int sz = 0; ZIO_SET_CHECKSUM(&stream_cksum, 0, 0, 0, 0); ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC); /* set the DEDUP feature flag for this stream */ fflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo); fflags |= (DMU_BACKUP_FEATURE_DEDUP | DMU_BACKUP_FEATURE_DEDUPPROPS); DMU_SET_FEATUREFLAGS(drrb->drr_versioninfo, fflags); if (drr->drr_payloadlen != 0) { sz = drr->drr_payloadlen; if (sz > SPA_MAXBLOCKSIZE) { buf = zfs_realloc(dda->dedup_hdl, buf, SPA_MAXBLOCKSIZE, sz); } (void) ssread(buf, sz, ofp); if (ferror(stdin)) perror("fread"); } if (dump_record(drr, buf, sz, &stream_cksum, outfd) != 0) goto out; break; } case DRR_END: { struct drr_end *drre = &drr->drr_u.drr_end; /* use the recalculated checksum */ drre->drr_checksum = stream_cksum; if (dump_record(drr, NULL, 0, &stream_cksum, outfd) != 0) goto out; break; } case DRR_OBJECT: { struct drr_object *drro = &drr->drr_u.drr_object; if (drro->drr_bonuslen > 0) { (void) ssread(buf, P2ROUNDUP((uint64_t)drro->drr_bonuslen, 8), ofp); } if (dump_record(drr, buf, P2ROUNDUP((uint64_t)drro->drr_bonuslen, 8), &stream_cksum, outfd) != 0) goto out; break; } case DRR_SPILL: { struct drr_spill *drrs = &drr->drr_u.drr_spill; (void) ssread(buf, drrs->drr_length, ofp); if (dump_record(drr, buf, drrs->drr_length, &stream_cksum, outfd) != 0) goto out; break; } case DRR_FREEOBJECTS: { if (dump_record(drr, NULL, 0, &stream_cksum, outfd) != 0) goto out; break; } case DRR_WRITE: { struct drr_write *drrw = &drr->drr_u.drr_write; dataref_t dataref; (void) ssread(buf, drrw->drr_length, ofp); /* * Use the existing checksum if it's dedup-capable, * else calculate a SHA256 checksum for it. */ if (ZIO_CHECKSUM_EQUAL(drrw->drr_key.ddk_cksum, zero_cksum) || !DRR_IS_DEDUP_CAPABLE(drrw->drr_checksumflags)) { SHA256_CTX ctx; zio_cksum_t tmpsha256; SHA256Init(&ctx); SHA256Update(&ctx, buf, drrw->drr_length); SHA256Final(&tmpsha256, &ctx); drrw->drr_key.ddk_cksum.zc_word[0] = BE_64(tmpsha256.zc_word[0]); drrw->drr_key.ddk_cksum.zc_word[1] = BE_64(tmpsha256.zc_word[1]); drrw->drr_key.ddk_cksum.zc_word[2] = BE_64(tmpsha256.zc_word[2]); drrw->drr_key.ddk_cksum.zc_word[3] = BE_64(tmpsha256.zc_word[3]); drrw->drr_checksumtype = ZIO_CHECKSUM_SHA256; drrw->drr_checksumflags = DRR_CHECKSUM_DEDUP; } dataref.ref_guid = drrw->drr_toguid; dataref.ref_object = drrw->drr_object; dataref.ref_offset = drrw->drr_offset; if (ddt_update(dda->dedup_hdl, &ddt, &drrw->drr_key.ddk_cksum, drrw->drr_key.ddk_prop, &dataref)) { dmu_replay_record_t wbr_drr = {0}; struct drr_write_byref *wbr_drrr = &wbr_drr.drr_u.drr_write_byref; /* block already present in stream */ wbr_drr.drr_type = DRR_WRITE_BYREF; wbr_drrr->drr_object = drrw->drr_object; wbr_drrr->drr_offset = drrw->drr_offset; wbr_drrr->drr_length = drrw->drr_length; wbr_drrr->drr_toguid = drrw->drr_toguid; wbr_drrr->drr_refguid = dataref.ref_guid; wbr_drrr->drr_refobject = dataref.ref_object; wbr_drrr->drr_refoffset = dataref.ref_offset; wbr_drrr->drr_checksumtype = drrw->drr_checksumtype; wbr_drrr->drr_checksumflags = drrw->drr_checksumtype; wbr_drrr->drr_key.ddk_cksum = drrw->drr_key.ddk_cksum; wbr_drrr->drr_key.ddk_prop = drrw->drr_key.ddk_prop; if (dump_record(&wbr_drr, NULL, 0, &stream_cksum, outfd) != 0) goto out; } else { /* block not previously seen */ if (dump_record(drr, buf, drrw->drr_length, &stream_cksum, outfd) != 0) goto out; } break; } case DRR_WRITE_EMBEDDED: { struct drr_write_embedded *drrwe = &drr->drr_u.drr_write_embedded; (void) ssread(buf, P2ROUNDUP((uint64_t)drrwe->drr_psize, 8), ofp); if (dump_record(drr, buf, P2ROUNDUP((uint64_t)drrwe->drr_psize, 8), &stream_cksum, outfd) != 0) goto out; break; } case DRR_FREE: { if (dump_record(drr, NULL, 0, &stream_cksum, outfd) != 0) goto out; break; } default: (void) fprintf(stderr, "INVALID record type 0x%x\n", drr->drr_type); /* should never happen, so assert */ assert(B_FALSE); } } out: umem_cache_destroy(ddt.ddecache); free(ddt.dedup_hash_array); free(buf); (void) fclose(ofp); return (NULL); } /* * Routines for dealing with the AVL tree of fs-nvlists */ typedef struct fsavl_node { avl_node_t fn_node; nvlist_t *fn_nvfs; char *fn_snapname; uint64_t fn_guid; } fsavl_node_t; static int fsavl_compare(const void *arg1, const void *arg2) { const fsavl_node_t *fn1 = arg1; const fsavl_node_t *fn2 = arg2; if (fn1->fn_guid > fn2->fn_guid) return (+1); else if (fn1->fn_guid < fn2->fn_guid) return (-1); else return (0); } /* * Given the GUID of a snapshot, find its containing filesystem and * (optionally) name. */ static nvlist_t * fsavl_find(avl_tree_t *avl, uint64_t snapguid, char **snapname) { fsavl_node_t fn_find; fsavl_node_t *fn; fn_find.fn_guid = snapguid; fn = avl_find(avl, &fn_find, NULL); if (fn) { if (snapname) *snapname = fn->fn_snapname; return (fn->fn_nvfs); } return (NULL); } static void fsavl_destroy(avl_tree_t *avl) { fsavl_node_t *fn; void *cookie; if (avl == NULL) return; cookie = NULL; while ((fn = avl_destroy_nodes(avl, &cookie)) != NULL) free(fn); avl_destroy(avl); free(avl); } /* * Given an nvlist, produce an avl tree of snapshots, ordered by guid */ static avl_tree_t * fsavl_create(nvlist_t *fss) { avl_tree_t *fsavl; nvpair_t *fselem = NULL; if ((fsavl = malloc(sizeof (avl_tree_t))) == NULL) return (NULL); avl_create(fsavl, fsavl_compare, sizeof (fsavl_node_t), offsetof(fsavl_node_t, fn_node)); while ((fselem = nvlist_next_nvpair(fss, fselem)) != NULL) { nvlist_t *nvfs, *snaps; nvpair_t *snapelem = NULL; VERIFY(0 == nvpair_value_nvlist(fselem, &nvfs)); VERIFY(0 == nvlist_lookup_nvlist(nvfs, "snaps", &snaps)); while ((snapelem = nvlist_next_nvpair(snaps, snapelem)) != NULL) { fsavl_node_t *fn; uint64_t guid; VERIFY(0 == nvpair_value_uint64(snapelem, &guid)); if ((fn = malloc(sizeof (fsavl_node_t))) == NULL) { fsavl_destroy(fsavl); return (NULL); } fn->fn_nvfs = nvfs; fn->fn_snapname = nvpair_name(snapelem); fn->fn_guid = guid; /* * Note: if there are multiple snaps with the * same GUID, we ignore all but one. */ if (avl_find(fsavl, fn, NULL) == NULL) avl_add(fsavl, fn); else free(fn); } } return (fsavl); } /* * Routines for dealing with the giant nvlist of fs-nvlists, etc. */ typedef struct send_data { uint64_t parent_fromsnap_guid; nvlist_t *parent_snaps; nvlist_t *fss; nvlist_t *snapprops; const char *fromsnap; const char *tosnap; boolean_t recursive; /* * The header nvlist is of the following format: * { * "tosnap" -> string * "fromsnap" -> string (if incremental) * "fss" -> { * id -> { * * "name" -> string (full name; for debugging) * "parentfromsnap" -> number (guid of fromsnap in parent) * * "props" -> { name -> value (only if set here) } * "snaps" -> { name (lastname) -> number (guid) } * "snapprops" -> { name (lastname) -> { name -> value } } * * "origin" -> number (guid) (if clone) * "sent" -> boolean (not on-disk) * } * } * } * */ } send_data_t; static void send_iterate_prop(zfs_handle_t *zhp, nvlist_t *nv); static int send_iterate_snap(zfs_handle_t *zhp, void *arg) { send_data_t *sd = arg; uint64_t guid = zhp->zfs_dmustats.dds_guid; char *snapname; nvlist_t *nv; snapname = strrchr(zhp->zfs_name, '@')+1; VERIFY(0 == nvlist_add_uint64(sd->parent_snaps, snapname, guid)); /* * NB: if there is no fromsnap here (it's a newly created fs in * an incremental replication), we will substitute the tosnap. */ if ((sd->fromsnap && strcmp(snapname, sd->fromsnap) == 0) || (sd->parent_fromsnap_guid == 0 && sd->tosnap && strcmp(snapname, sd->tosnap) == 0)) { sd->parent_fromsnap_guid = guid; } VERIFY(0 == nvlist_alloc(&nv, NV_UNIQUE_NAME, 0)); send_iterate_prop(zhp, nv); VERIFY(0 == nvlist_add_nvlist(sd->snapprops, snapname, nv)); nvlist_free(nv); zfs_close(zhp); return (0); } static void send_iterate_prop(zfs_handle_t *zhp, nvlist_t *nv) { nvpair_t *elem = NULL; while ((elem = nvlist_next_nvpair(zhp->zfs_props, elem)) != NULL) { char *propname = nvpair_name(elem); zfs_prop_t prop = zfs_name_to_prop(propname); nvlist_t *propnv; if (!zfs_prop_user(propname)) { /* * Realistically, this should never happen. However, * we want the ability to add DSL properties without * needing to make incompatible version changes. We * need to ignore unknown properties to allow older * software to still send datasets containing these * properties, with the unknown properties elided. */ if (prop == ZPROP_INVAL) continue; if (zfs_prop_readonly(prop)) continue; } verify(nvpair_value_nvlist(elem, &propnv) == 0); if (prop == ZFS_PROP_QUOTA || prop == ZFS_PROP_RESERVATION || prop == ZFS_PROP_REFQUOTA || prop == ZFS_PROP_REFRESERVATION) { char *source; uint64_t value; verify(nvlist_lookup_uint64(propnv, ZPROP_VALUE, &value) == 0); if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) continue; /* * May have no source before SPA_VERSION_RECVD_PROPS, * but is still modifiable. */ if (nvlist_lookup_string(propnv, ZPROP_SOURCE, &source) == 0) { if ((strcmp(source, zhp->zfs_name) != 0) && (strcmp(source, ZPROP_SOURCE_VAL_RECVD) != 0)) continue; } } else { char *source; if (nvlist_lookup_string(propnv, ZPROP_SOURCE, &source) != 0) continue; if ((strcmp(source, zhp->zfs_name) != 0) && (strcmp(source, ZPROP_SOURCE_VAL_RECVD) != 0)) continue; } if (zfs_prop_user(propname) || zfs_prop_get_type(prop) == PROP_TYPE_STRING) { char *value; verify(nvlist_lookup_string(propnv, ZPROP_VALUE, &value) == 0); VERIFY(0 == nvlist_add_string(nv, propname, value)); } else { uint64_t value; verify(nvlist_lookup_uint64(propnv, ZPROP_VALUE, &value) == 0); VERIFY(0 == nvlist_add_uint64(nv, propname, value)); } } } /* * recursively generate nvlists describing datasets. See comment * for the data structure send_data_t above for description of contents * of the nvlist. */ static int send_iterate_fs(zfs_handle_t *zhp, void *arg) { send_data_t *sd = arg; nvlist_t *nvfs, *nv; int rv = 0; uint64_t parent_fromsnap_guid_save = sd->parent_fromsnap_guid; uint64_t guid = zhp->zfs_dmustats.dds_guid; char guidstring[64]; VERIFY(0 == nvlist_alloc(&nvfs, NV_UNIQUE_NAME, 0)); VERIFY(0 == nvlist_add_string(nvfs, "name", zhp->zfs_name)); VERIFY(0 == nvlist_add_uint64(nvfs, "parentfromsnap", sd->parent_fromsnap_guid)); if (zhp->zfs_dmustats.dds_origin[0]) { zfs_handle_t *origin = zfs_open(zhp->zfs_hdl, zhp->zfs_dmustats.dds_origin, ZFS_TYPE_SNAPSHOT); if (origin == NULL) return (-1); VERIFY(0 == nvlist_add_uint64(nvfs, "origin", origin->zfs_dmustats.dds_guid)); } /* iterate over props */ VERIFY(0 == nvlist_alloc(&nv, NV_UNIQUE_NAME, 0)); send_iterate_prop(zhp, nv); VERIFY(0 == nvlist_add_nvlist(nvfs, "props", nv)); nvlist_free(nv); /* iterate over snaps, and set sd->parent_fromsnap_guid */ sd->parent_fromsnap_guid = 0; VERIFY(0 == nvlist_alloc(&sd->parent_snaps, NV_UNIQUE_NAME, 0)); VERIFY(0 == nvlist_alloc(&sd->snapprops, NV_UNIQUE_NAME, 0)); (void) zfs_iter_snapshots_sorted(zhp, send_iterate_snap, sd); VERIFY(0 == nvlist_add_nvlist(nvfs, "snaps", sd->parent_snaps)); VERIFY(0 == nvlist_add_nvlist(nvfs, "snapprops", sd->snapprops)); nvlist_free(sd->parent_snaps); nvlist_free(sd->snapprops); /* add this fs to nvlist */ (void) snprintf(guidstring, sizeof (guidstring), "0x%llx", (longlong_t)guid); VERIFY(0 == nvlist_add_nvlist(sd->fss, guidstring, nvfs)); nvlist_free(nvfs); /* iterate over children */ if (sd->recursive) rv = zfs_iter_filesystems(zhp, send_iterate_fs, sd); sd->parent_fromsnap_guid = parent_fromsnap_guid_save; zfs_close(zhp); return (rv); } static int gather_nvlist(libzfs_handle_t *hdl, const char *fsname, const char *fromsnap, const char *tosnap, boolean_t recursive, nvlist_t **nvlp, avl_tree_t **avlp) { zfs_handle_t *zhp; send_data_t sd = { 0 }; int error; zhp = zfs_open(hdl, fsname, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); if (zhp == NULL) return (EZFS_BADTYPE); VERIFY(0 == nvlist_alloc(&sd.fss, NV_UNIQUE_NAME, 0)); sd.fromsnap = fromsnap; sd.tosnap = tosnap; sd.recursive = recursive; if ((error = send_iterate_fs(zhp, &sd)) != 0) { nvlist_free(sd.fss); if (avlp != NULL) *avlp = NULL; *nvlp = NULL; return (error); } if (avlp != NULL && (*avlp = fsavl_create(sd.fss)) == NULL) { nvlist_free(sd.fss); *nvlp = NULL; return (EZFS_NOMEM); } *nvlp = sd.fss; return (0); } /* * Routines specific to "zfs send" */ typedef struct send_dump_data { /* these are all just the short snapname (the part after the @) */ const char *fromsnap; const char *tosnap; char prevsnap[ZFS_MAXNAMELEN]; uint64_t prevsnap_obj; boolean_t seenfrom, seento, replicate, doall, fromorigin; boolean_t verbose, dryrun, parsable, progress, embed_data, std_out; boolean_t large_block; int outfd; boolean_t err; nvlist_t *fss; nvlist_t *snapholds; avl_tree_t *fsavl; snapfilter_cb_t *filter_cb; void *filter_cb_arg; nvlist_t *debugnv; char holdtag[ZFS_MAXNAMELEN]; int cleanup_fd; uint64_t size; } send_dump_data_t; static int estimate_ioctl(zfs_handle_t *zhp, uint64_t fromsnap_obj, boolean_t fromorigin, uint64_t *sizep) { zfs_cmd_t zc = { 0 }; libzfs_handle_t *hdl = zhp->zfs_hdl; assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT); assert(fromsnap_obj == 0 || !fromorigin); (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); zc.zc_obj = fromorigin; zc.zc_sendobj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID); zc.zc_fromobj = fromsnap_obj; zc.zc_guid = 1; /* estimate flag */ if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_SEND, &zc) != 0) { char errbuf[1024]; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "warning: cannot estimate space for '%s'"), zhp->zfs_name); switch (errno) { case EXDEV: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "not an earlier snapshot from the same fs")); return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf)); case ENOENT: if (zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_SNAPSHOT)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "incremental source (@%s) does not exist"), zc.zc_value); } return (zfs_error(hdl, EZFS_NOENT, errbuf)); case EDQUOT: case EFBIG: case EIO: case ENOLINK: case ENOSPC: case ENXIO: case EPIPE: case ERANGE: case EFAULT: case EROFS: zfs_error_aux(hdl, strerror(errno)); return (zfs_error(hdl, EZFS_BADBACKUP, errbuf)); default: return (zfs_standard_error(hdl, errno, errbuf)); } } *sizep = zc.zc_objset_type; return (0); } /* * Dumps a backup of the given snapshot (incremental from fromsnap if it's not * NULL) to the file descriptor specified by outfd. */ static int dump_ioctl(zfs_handle_t *zhp, const char *fromsnap, uint64_t fromsnap_obj, boolean_t fromorigin, int outfd, enum lzc_send_flags flags, nvlist_t *debugnv) { zfs_cmd_t zc = { 0 }; libzfs_handle_t *hdl = zhp->zfs_hdl; nvlist_t *thisdbg; assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT); assert(fromsnap_obj == 0 || !fromorigin); (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); zc.zc_cookie = outfd; zc.zc_obj = fromorigin; zc.zc_sendobj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID); zc.zc_fromobj = fromsnap_obj; zc.zc_flags = flags; VERIFY(0 == nvlist_alloc(&thisdbg, NV_UNIQUE_NAME, 0)); if (fromsnap && fromsnap[0] != '\0') { VERIFY(0 == nvlist_add_string(thisdbg, "fromsnap", fromsnap)); } if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_SEND, &zc) != 0) { char errbuf[1024]; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "warning: cannot send '%s'"), zhp->zfs_name); VERIFY(0 == nvlist_add_uint64(thisdbg, "error", errno)); if (debugnv) { VERIFY(0 == nvlist_add_nvlist(debugnv, zhp->zfs_name, thisdbg)); } nvlist_free(thisdbg); switch (errno) { case EXDEV: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "not an earlier snapshot from the same fs")); return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf)); case ENOENT: if (zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_SNAPSHOT)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "incremental source (@%s) does not exist"), zc.zc_value); } return (zfs_error(hdl, EZFS_NOENT, errbuf)); case EDQUOT: case EFBIG: case EIO: case ENOLINK: case ENOSPC: #ifdef sun case ENOSTR: #endif case ENXIO: case EPIPE: case ERANGE: case EFAULT: case EROFS: zfs_error_aux(hdl, strerror(errno)); return (zfs_error(hdl, EZFS_BADBACKUP, errbuf)); default: return (zfs_standard_error(hdl, errno, errbuf)); } } if (debugnv) VERIFY(0 == nvlist_add_nvlist(debugnv, zhp->zfs_name, thisdbg)); nvlist_free(thisdbg); return (0); } static void gather_holds(zfs_handle_t *zhp, send_dump_data_t *sdd) { assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT); /* * zfs_send() only sets snapholds for sends that need them, * e.g. replication and doall. */ if (sdd->snapholds == NULL) return; fnvlist_add_string(sdd->snapholds, zhp->zfs_name, sdd->holdtag); } static void * send_progress_thread(void *arg) { progress_arg_t *pa = arg; zfs_cmd_t zc = { 0 }; zfs_handle_t *zhp = pa->pa_zhp; libzfs_handle_t *hdl = zhp->zfs_hdl; unsigned long long bytes; char buf[16]; time_t t; struct tm *tm; (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); if (!pa->pa_parsable) (void) fprintf(stderr, "TIME SENT SNAPSHOT\n"); /* * Print the progress from ZFS_IOC_SEND_PROGRESS every second. */ for (;;) { (void) sleep(1); zc.zc_cookie = pa->pa_fd; if (zfs_ioctl(hdl, ZFS_IOC_SEND_PROGRESS, &zc) != 0) return ((void *)-1); (void) time(&t); tm = localtime(&t); bytes = zc.zc_cookie; if (pa->pa_parsable) { (void) fprintf(stderr, "%02d:%02d:%02d\t%llu\t%s\n", tm->tm_hour, tm->tm_min, tm->tm_sec, bytes, zhp->zfs_name); } else { zfs_nicenum(bytes, buf, sizeof (buf)); (void) fprintf(stderr, "%02d:%02d:%02d %5s %s\n", tm->tm_hour, tm->tm_min, tm->tm_sec, buf, zhp->zfs_name); } } } static void send_print_verbose(FILE *fout, const char *tosnap, const char *fromsnap, uint64_t size, boolean_t parsable) { if (parsable) { if (fromsnap != NULL) { (void) fprintf(fout, "incremental\t%s\t%s", fromsnap, tosnap); } else { (void) fprintf(fout, "full\t%s", tosnap); } } else { if (fromsnap != NULL) { if (strchr(fromsnap, '@') == NULL && strchr(fromsnap, '#') == NULL) { (void) fprintf(fout, dgettext(TEXT_DOMAIN, "send from @%s to %s"), fromsnap, tosnap); } else { (void) fprintf(fout, dgettext(TEXT_DOMAIN, "send from %s to %s"), fromsnap, tosnap); } } else { (void) fprintf(fout, dgettext(TEXT_DOMAIN, "full send of %s"), tosnap); } } if (size != 0) { if (parsable) { (void) fprintf(fout, "\t%llu", (longlong_t)size); } else { char buf[16]; zfs_nicenum(size, buf, sizeof (buf)); (void) fprintf(fout, dgettext(TEXT_DOMAIN, " estimated size is %s"), buf); } } (void) fprintf(fout, "\n"); } static int dump_snapshot(zfs_handle_t *zhp, void *arg) { send_dump_data_t *sdd = arg; progress_arg_t pa = { 0 }; pthread_t tid; char *thissnap; int err; boolean_t isfromsnap, istosnap, fromorigin; boolean_t exclude = B_FALSE; FILE *fout = sdd->std_out ? stdout : stderr; err = 0; thissnap = strchr(zhp->zfs_name, '@') + 1; isfromsnap = (sdd->fromsnap != NULL && strcmp(sdd->fromsnap, thissnap) == 0); if (!sdd->seenfrom && isfromsnap) { gather_holds(zhp, sdd); sdd->seenfrom = B_TRUE; (void) strcpy(sdd->prevsnap, thissnap); sdd->prevsnap_obj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID); zfs_close(zhp); return (0); } if (sdd->seento || !sdd->seenfrom) { zfs_close(zhp); return (0); } istosnap = (strcmp(sdd->tosnap, thissnap) == 0); if (istosnap) sdd->seento = B_TRUE; if (!sdd->doall && !isfromsnap && !istosnap) { if (sdd->replicate) { char *snapname; nvlist_t *snapprops; /* * Filter out all intermediate snapshots except origin * snapshots needed to replicate clones. */ nvlist_t *nvfs = fsavl_find(sdd->fsavl, zhp->zfs_dmustats.dds_guid, &snapname); VERIFY(0 == nvlist_lookup_nvlist(nvfs, "snapprops", &snapprops)); VERIFY(0 == nvlist_lookup_nvlist(snapprops, thissnap, &snapprops)); exclude = !nvlist_exists(snapprops, "is_clone_origin"); } else { exclude = B_TRUE; } } /* * If a filter function exists, call it to determine whether * this snapshot will be sent. */ if (exclude || (sdd->filter_cb != NULL && sdd->filter_cb(zhp, sdd->filter_cb_arg) == B_FALSE)) { /* * This snapshot is filtered out. Don't send it, and don't * set prevsnap_obj, so it will be as if this snapshot didn't * exist, and the next accepted snapshot will be sent as * an incremental from the last accepted one, or as the * first (and full) snapshot in the case of a replication, * non-incremental send. */ zfs_close(zhp); return (0); } gather_holds(zhp, sdd); fromorigin = sdd->prevsnap[0] == '\0' && (sdd->fromorigin || sdd->replicate); if (sdd->verbose) { uint64_t size = 0; (void) estimate_ioctl(zhp, sdd->prevsnap_obj, fromorigin, &size); send_print_verbose(fout, zhp->zfs_name, sdd->prevsnap[0] ? sdd->prevsnap : NULL, size, sdd->parsable); sdd->size += size; } if (!sdd->dryrun) { /* * If progress reporting is requested, spawn a new thread to * poll ZFS_IOC_SEND_PROGRESS at a regular interval. */ if (sdd->progress) { pa.pa_zhp = zhp; pa.pa_fd = sdd->outfd; pa.pa_parsable = sdd->parsable; if (err = pthread_create(&tid, NULL, send_progress_thread, &pa)) { zfs_close(zhp); return (err); } } enum lzc_send_flags flags = 0; if (sdd->large_block) flags |= LZC_SEND_FLAG_LARGE_BLOCK; if (sdd->embed_data) flags |= LZC_SEND_FLAG_EMBED_DATA; err = dump_ioctl(zhp, sdd->prevsnap, sdd->prevsnap_obj, fromorigin, sdd->outfd, flags, sdd->debugnv); if (sdd->progress) { (void) pthread_cancel(tid); (void) pthread_join(tid, NULL); } } (void) strcpy(sdd->prevsnap, thissnap); sdd->prevsnap_obj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID); zfs_close(zhp); return (err); } static int dump_filesystem(zfs_handle_t *zhp, void *arg) { int rv = 0; send_dump_data_t *sdd = arg; boolean_t missingfrom = B_FALSE; zfs_cmd_t zc = { 0 }; (void) snprintf(zc.zc_name, sizeof (zc.zc_name), "%s@%s", zhp->zfs_name, sdd->tosnap); if (ioctl(zhp->zfs_hdl->libzfs_fd, ZFS_IOC_OBJSET_STATS, &zc) != 0) { (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "WARNING: could not send %s@%s: does not exist\n"), zhp->zfs_name, sdd->tosnap); sdd->err = B_TRUE; return (0); } if (sdd->replicate && sdd->fromsnap) { /* * If this fs does not have fromsnap, and we're doing * recursive, we need to send a full stream from the * beginning (or an incremental from the origin if this * is a clone). If we're doing non-recursive, then let * them get the error. */ (void) snprintf(zc.zc_name, sizeof (zc.zc_name), "%s@%s", zhp->zfs_name, sdd->fromsnap); if (ioctl(zhp->zfs_hdl->libzfs_fd, ZFS_IOC_OBJSET_STATS, &zc) != 0) { missingfrom = B_TRUE; } } sdd->seenfrom = sdd->seento = sdd->prevsnap[0] = 0; sdd->prevsnap_obj = 0; if (sdd->fromsnap == NULL || missingfrom) sdd->seenfrom = B_TRUE; rv = zfs_iter_snapshots_sorted(zhp, dump_snapshot, arg); if (!sdd->seenfrom) { (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "WARNING: could not send %s@%s:\n" "incremental source (%s@%s) does not exist\n"), zhp->zfs_name, sdd->tosnap, zhp->zfs_name, sdd->fromsnap); sdd->err = B_TRUE; } else if (!sdd->seento) { if (sdd->fromsnap) { (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "WARNING: could not send %s@%s:\n" "incremental source (%s@%s) " "is not earlier than it\n"), zhp->zfs_name, sdd->tosnap, zhp->zfs_name, sdd->fromsnap); } else { (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "WARNING: " "could not send %s@%s: does not exist\n"), zhp->zfs_name, sdd->tosnap); } sdd->err = B_TRUE; } return (rv); } static int dump_filesystems(zfs_handle_t *rzhp, void *arg) { send_dump_data_t *sdd = arg; nvpair_t *fspair; boolean_t needagain, progress; if (!sdd->replicate) return (dump_filesystem(rzhp, sdd)); /* Mark the clone origin snapshots. */ for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair; fspair = nvlist_next_nvpair(sdd->fss, fspair)) { nvlist_t *nvfs; uint64_t origin_guid = 0; VERIFY(0 == nvpair_value_nvlist(fspair, &nvfs)); (void) nvlist_lookup_uint64(nvfs, "origin", &origin_guid); if (origin_guid != 0) { char *snapname; nvlist_t *origin_nv = fsavl_find(sdd->fsavl, origin_guid, &snapname); if (origin_nv != NULL) { nvlist_t *snapprops; VERIFY(0 == nvlist_lookup_nvlist(origin_nv, "snapprops", &snapprops)); VERIFY(0 == nvlist_lookup_nvlist(snapprops, snapname, &snapprops)); VERIFY(0 == nvlist_add_boolean( snapprops, "is_clone_origin")); } } } again: needagain = progress = B_FALSE; for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair; fspair = nvlist_next_nvpair(sdd->fss, fspair)) { nvlist_t *fslist, *parent_nv; char *fsname; zfs_handle_t *zhp; int err; uint64_t origin_guid = 0; uint64_t parent_guid = 0; VERIFY(nvpair_value_nvlist(fspair, &fslist) == 0); if (nvlist_lookup_boolean(fslist, "sent") == 0) continue; VERIFY(nvlist_lookup_string(fslist, "name", &fsname) == 0); (void) nvlist_lookup_uint64(fslist, "origin", &origin_guid); (void) nvlist_lookup_uint64(fslist, "parentfromsnap", &parent_guid); if (parent_guid != 0) { parent_nv = fsavl_find(sdd->fsavl, parent_guid, NULL); if (!nvlist_exists(parent_nv, "sent")) { /* parent has not been sent; skip this one */ needagain = B_TRUE; continue; } } if (origin_guid != 0) { nvlist_t *origin_nv = fsavl_find(sdd->fsavl, origin_guid, NULL); if (origin_nv != NULL && !nvlist_exists(origin_nv, "sent")) { /* * origin has not been sent yet; * skip this clone. */ needagain = B_TRUE; continue; } } zhp = zfs_open(rzhp->zfs_hdl, fsname, ZFS_TYPE_DATASET); if (zhp == NULL) return (-1); err = dump_filesystem(zhp, sdd); VERIFY(nvlist_add_boolean(fslist, "sent") == 0); progress = B_TRUE; zfs_close(zhp); if (err) return (err); } if (needagain) { assert(progress); goto again; } /* clean out the sent flags in case we reuse this fss */ for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair; fspair = nvlist_next_nvpair(sdd->fss, fspair)) { nvlist_t *fslist; VERIFY(nvpair_value_nvlist(fspair, &fslist) == 0); (void) nvlist_remove_all(fslist, "sent"); } return (0); } nvlist_t * zfs_send_resume_token_to_nvlist(libzfs_handle_t *hdl, const char *token) { unsigned int version; int nread; unsigned long long checksum, packed_len; /* * Decode token header, which is: * -- * Note that the only supported token version is 1. */ nread = sscanf(token, "%u-%llx-%llx-", &version, &checksum, &packed_len); if (nread != 3) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "resume token is corrupt (invalid format)")); return (NULL); } if (version != ZFS_SEND_RESUME_TOKEN_VERSION) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "resume token is corrupt (invalid version %u)"), version); return (NULL); } /* convert hexadecimal representation to binary */ token = strrchr(token, '-') + 1; int len = strlen(token) / 2; unsigned char *compressed = zfs_alloc(hdl, len); for (int i = 0; i < len; i++) { nread = sscanf(token + i * 2, "%2hhx", compressed + i); if (nread != 1) { free(compressed); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "resume token is corrupt " "(payload is not hex-encoded)")); return (NULL); } } /* verify checksum */ zio_cksum_t cksum; - fletcher_4_native(compressed, len, &cksum); + fletcher_4_native(compressed, len, NULL, &cksum); if (cksum.zc_word[0] != checksum) { free(compressed); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "resume token is corrupt (incorrect checksum)")); return (NULL); } /* uncompress */ void *packed = zfs_alloc(hdl, packed_len); uLongf packed_len_long = packed_len; if (uncompress(packed, &packed_len_long, compressed, len) != Z_OK || packed_len_long != packed_len) { free(packed); free(compressed); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "resume token is corrupt (decompression failed)")); return (NULL); } /* unpack nvlist */ nvlist_t *nv; int error = nvlist_unpack(packed, packed_len, &nv, KM_SLEEP); free(packed); free(compressed); if (error != 0) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "resume token is corrupt (nvlist_unpack failed)")); return (NULL); } return (nv); } int zfs_send_resume(libzfs_handle_t *hdl, sendflags_t *flags, int outfd, const char *resume_token) { char errbuf[1024]; char *toname; char *fromname = NULL; uint64_t resumeobj, resumeoff, toguid, fromguid, bytes; zfs_handle_t *zhp; int error = 0; char name[ZFS_MAXNAMELEN]; enum lzc_send_flags lzc_flags = 0; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot resume send")); nvlist_t *resume_nvl = zfs_send_resume_token_to_nvlist(hdl, resume_token); if (resume_nvl == NULL) { /* * zfs_error_aux has already been set by * zfs_send_resume_token_to_nvlist */ return (zfs_error(hdl, EZFS_FAULT, errbuf)); } if (flags->verbose) { (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "resume token contents:\n")); nvlist_print(stderr, resume_nvl); } if (nvlist_lookup_string(resume_nvl, "toname", &toname) != 0 || nvlist_lookup_uint64(resume_nvl, "object", &resumeobj) != 0 || nvlist_lookup_uint64(resume_nvl, "offset", &resumeoff) != 0 || nvlist_lookup_uint64(resume_nvl, "bytes", &bytes) != 0 || nvlist_lookup_uint64(resume_nvl, "toguid", &toguid) != 0) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "resume token is corrupt")); return (zfs_error(hdl, EZFS_FAULT, errbuf)); } fromguid = 0; (void) nvlist_lookup_uint64(resume_nvl, "fromguid", &fromguid); if (flags->embed_data || nvlist_exists(resume_nvl, "embedok")) lzc_flags |= LZC_SEND_FLAG_EMBED_DATA; if (guid_to_name(hdl, toname, toguid, B_FALSE, name) != 0) { if (zfs_dataset_exists(hdl, toname, ZFS_TYPE_DATASET)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' is no longer the same snapshot used in " "the initial send"), toname); } else { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' used in the initial send no longer exists"), toname); } return (zfs_error(hdl, EZFS_BADPATH, errbuf)); } zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET); if (zhp == NULL) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "unable to access '%s'"), name); return (zfs_error(hdl, EZFS_BADPATH, errbuf)); } if (fromguid != 0) { if (guid_to_name(hdl, toname, fromguid, B_TRUE, name) != 0) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "incremental source %#llx no longer exists"), (longlong_t)fromguid); return (zfs_error(hdl, EZFS_BADPATH, errbuf)); } fromname = name; } if (flags->verbose) { uint64_t size = 0; error = lzc_send_space(zhp->zfs_name, fromname, &size); if (error == 0) size = MAX(0, (int64_t)(size - bytes)); send_print_verbose(stderr, zhp->zfs_name, fromname, size, flags->parsable); } if (!flags->dryrun) { progress_arg_t pa = { 0 }; pthread_t tid; /* * If progress reporting is requested, spawn a new thread to * poll ZFS_IOC_SEND_PROGRESS at a regular interval. */ if (flags->progress) { pa.pa_zhp = zhp; pa.pa_fd = outfd; pa.pa_parsable = flags->parsable; error = pthread_create(&tid, NULL, send_progress_thread, &pa); if (error != 0) { zfs_close(zhp); return (error); } } error = lzc_send_resume(zhp->zfs_name, fromname, outfd, lzc_flags, resumeobj, resumeoff); if (flags->progress) { (void) pthread_cancel(tid); (void) pthread_join(tid, NULL); } char errbuf[1024]; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "warning: cannot send '%s'"), zhp->zfs_name); zfs_close(zhp); switch (error) { case 0: return (0); case EXDEV: case ENOENT: case EDQUOT: case EFBIG: case EIO: case ENOLINK: case ENOSPC: #ifdef illumos case ENOSTR: #endif case ENXIO: case EPIPE: case ERANGE: case EFAULT: case EROFS: zfs_error_aux(hdl, strerror(errno)); return (zfs_error(hdl, EZFS_BADBACKUP, errbuf)); default: return (zfs_standard_error(hdl, errno, errbuf)); } } zfs_close(zhp); return (error); } /* * Generate a send stream for the dataset identified by the argument zhp. * * The content of the send stream is the snapshot identified by * 'tosnap'. Incremental streams are requested in two ways: * - from the snapshot identified by "fromsnap" (if non-null) or * - from the origin of the dataset identified by zhp, which must * be a clone. In this case, "fromsnap" is null and "fromorigin" * is TRUE. * * The send stream is recursive (i.e. dumps a hierarchy of snapshots) and * uses a special header (with a hdrtype field of DMU_COMPOUNDSTREAM) * if "replicate" is set. If "doall" is set, dump all the intermediate * snapshots. The DMU_COMPOUNDSTREAM header is used in the "doall" * case too. If "props" is set, send properties. */ int zfs_send(zfs_handle_t *zhp, const char *fromsnap, const char *tosnap, sendflags_t *flags, int outfd, snapfilter_cb_t filter_func, void *cb_arg, nvlist_t **debugnvp) { char errbuf[1024]; send_dump_data_t sdd = { 0 }; int err = 0; nvlist_t *fss = NULL; avl_tree_t *fsavl = NULL; static uint64_t holdseq; int spa_version; pthread_t tid = 0; int pipefd[2]; dedup_arg_t dda = { 0 }; int featureflags = 0; FILE *fout; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot send '%s'"), zhp->zfs_name); if (fromsnap && fromsnap[0] == '\0') { zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN, "zero-length incremental source")); return (zfs_error(zhp->zfs_hdl, EZFS_NOENT, errbuf)); } if (zhp->zfs_type == ZFS_TYPE_FILESYSTEM) { uint64_t version; version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION); if (version >= ZPL_VERSION_SA) { featureflags |= DMU_BACKUP_FEATURE_SA_SPILL; } } if (flags->dedup && !flags->dryrun) { featureflags |= (DMU_BACKUP_FEATURE_DEDUP | DMU_BACKUP_FEATURE_DEDUPPROPS); if (err = pipe(pipefd)) { zfs_error_aux(zhp->zfs_hdl, strerror(errno)); return (zfs_error(zhp->zfs_hdl, EZFS_PIPEFAILED, errbuf)); } dda.outputfd = outfd; dda.inputfd = pipefd[1]; dda.dedup_hdl = zhp->zfs_hdl; if (err = pthread_create(&tid, NULL, cksummer, &dda)) { (void) close(pipefd[0]); (void) close(pipefd[1]); zfs_error_aux(zhp->zfs_hdl, strerror(errno)); return (zfs_error(zhp->zfs_hdl, EZFS_THREADCREATEFAILED, errbuf)); } } if (flags->replicate || flags->doall || flags->props) { dmu_replay_record_t drr = { 0 }; char *packbuf = NULL; size_t buflen = 0; zio_cksum_t zc = { 0 }; if (flags->replicate || flags->props) { nvlist_t *hdrnv; VERIFY(0 == nvlist_alloc(&hdrnv, NV_UNIQUE_NAME, 0)); if (fromsnap) { VERIFY(0 == nvlist_add_string(hdrnv, "fromsnap", fromsnap)); } VERIFY(0 == nvlist_add_string(hdrnv, "tosnap", tosnap)); if (!flags->replicate) { VERIFY(0 == nvlist_add_boolean(hdrnv, "not_recursive")); } err = gather_nvlist(zhp->zfs_hdl, zhp->zfs_name, fromsnap, tosnap, flags->replicate, &fss, &fsavl); if (err) goto err_out; VERIFY(0 == nvlist_add_nvlist(hdrnv, "fss", fss)); err = nvlist_pack(hdrnv, &packbuf, &buflen, NV_ENCODE_XDR, 0); if (debugnvp) *debugnvp = hdrnv; else nvlist_free(hdrnv); if (err) goto stderr_out; } if (!flags->dryrun) { /* write first begin record */ drr.drr_type = DRR_BEGIN; drr.drr_u.drr_begin.drr_magic = DMU_BACKUP_MAGIC; DMU_SET_STREAM_HDRTYPE(drr.drr_u.drr_begin. drr_versioninfo, DMU_COMPOUNDSTREAM); DMU_SET_FEATUREFLAGS(drr.drr_u.drr_begin. drr_versioninfo, featureflags); (void) snprintf(drr.drr_u.drr_begin.drr_toname, sizeof (drr.drr_u.drr_begin.drr_toname), "%s@%s", zhp->zfs_name, tosnap); drr.drr_payloadlen = buflen; err = dump_record(&drr, packbuf, buflen, &zc, outfd); free(packbuf); if (err != 0) goto stderr_out; /* write end record */ bzero(&drr, sizeof (drr)); drr.drr_type = DRR_END; drr.drr_u.drr_end.drr_checksum = zc; err = write(outfd, &drr, sizeof (drr)); if (err == -1) { err = errno; goto stderr_out; } err = 0; } } /* dump each stream */ sdd.fromsnap = fromsnap; sdd.tosnap = tosnap; if (tid != 0) sdd.outfd = pipefd[0]; else sdd.outfd = outfd; sdd.replicate = flags->replicate; sdd.doall = flags->doall; sdd.fromorigin = flags->fromorigin; sdd.fss = fss; sdd.fsavl = fsavl; sdd.verbose = flags->verbose; sdd.parsable = flags->parsable; sdd.progress = flags->progress; sdd.dryrun = flags->dryrun; sdd.large_block = flags->largeblock; sdd.embed_data = flags->embed_data; sdd.filter_cb = filter_func; sdd.filter_cb_arg = cb_arg; if (debugnvp) sdd.debugnv = *debugnvp; if (sdd.verbose && sdd.dryrun) sdd.std_out = B_TRUE; fout = sdd.std_out ? stdout : stderr; /* * Some flags require that we place user holds on the datasets that are * being sent so they don't get destroyed during the send. We can skip * this step if the pool is imported read-only since the datasets cannot * be destroyed. */ if (!flags->dryrun && !zpool_get_prop_int(zfs_get_pool_handle(zhp), ZPOOL_PROP_READONLY, NULL) && zfs_spa_version(zhp, &spa_version) == 0 && spa_version >= SPA_VERSION_USERREFS && (flags->doall || flags->replicate)) { ++holdseq; (void) snprintf(sdd.holdtag, sizeof (sdd.holdtag), ".send-%d-%llu", getpid(), (u_longlong_t)holdseq); sdd.cleanup_fd = open(ZFS_DEV, O_RDWR|O_EXCL); if (sdd.cleanup_fd < 0) { err = errno; goto stderr_out; } sdd.snapholds = fnvlist_alloc(); } else { sdd.cleanup_fd = -1; sdd.snapholds = NULL; } if (flags->verbose || sdd.snapholds != NULL) { /* * Do a verbose no-op dry run to get all the verbose output * or to gather snapshot hold's before generating any data, * then do a non-verbose real run to generate the streams. */ sdd.dryrun = B_TRUE; err = dump_filesystems(zhp, &sdd); if (err != 0) goto stderr_out; if (flags->verbose) { if (flags->parsable) { (void) fprintf(fout, "size\t%llu\n", (longlong_t)sdd.size); } else { char buf[16]; zfs_nicenum(sdd.size, buf, sizeof (buf)); (void) fprintf(fout, dgettext(TEXT_DOMAIN, "total estimated size is %s\n"), buf); } } /* Ensure no snaps found is treated as an error. */ if (!sdd.seento) { err = ENOENT; goto err_out; } /* Skip the second run if dryrun was requested. */ if (flags->dryrun) goto err_out; if (sdd.snapholds != NULL) { err = zfs_hold_nvl(zhp, sdd.cleanup_fd, sdd.snapholds); if (err != 0) goto stderr_out; fnvlist_free(sdd.snapholds); sdd.snapholds = NULL; } sdd.dryrun = B_FALSE; sdd.verbose = B_FALSE; } err = dump_filesystems(zhp, &sdd); fsavl_destroy(fsavl); nvlist_free(fss); /* Ensure no snaps found is treated as an error. */ if (err == 0 && !sdd.seento) err = ENOENT; if (tid != 0) { if (err != 0) (void) pthread_cancel(tid); (void) close(pipefd[0]); (void) pthread_join(tid, NULL); } if (sdd.cleanup_fd != -1) { VERIFY(0 == close(sdd.cleanup_fd)); sdd.cleanup_fd = -1; } if (!flags->dryrun && (flags->replicate || flags->doall || flags->props)) { /* * write final end record. NB: want to do this even if * there was some error, because it might not be totally * failed. */ dmu_replay_record_t drr = { 0 }; drr.drr_type = DRR_END; if (write(outfd, &drr, sizeof (drr)) == -1) { return (zfs_standard_error(zhp->zfs_hdl, errno, errbuf)); } } return (err || sdd.err); stderr_out: err = zfs_standard_error(zhp->zfs_hdl, err, errbuf); err_out: fsavl_destroy(fsavl); nvlist_free(fss); fnvlist_free(sdd.snapholds); if (sdd.cleanup_fd != -1) VERIFY(0 == close(sdd.cleanup_fd)); if (tid != 0) { (void) pthread_cancel(tid); (void) close(pipefd[0]); (void) pthread_join(tid, NULL); } return (err); } int zfs_send_one(zfs_handle_t *zhp, const char *from, int fd, enum lzc_send_flags flags) { int err; libzfs_handle_t *hdl = zhp->zfs_hdl; char errbuf[1024]; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "warning: cannot send '%s'"), zhp->zfs_name); err = lzc_send(zhp->zfs_name, from, fd, flags); if (err != 0) { switch (errno) { case EXDEV: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "not an earlier snapshot from the same fs")); return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf)); case ENOENT: case ESRCH: if (lzc_exists(zhp->zfs_name)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "incremental source (%s) does not exist"), from); } return (zfs_error(hdl, EZFS_NOENT, errbuf)); case EBUSY: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "target is busy; if a filesystem, " "it must not be mounted")); return (zfs_error(hdl, EZFS_BUSY, errbuf)); case EDQUOT: case EFBIG: case EIO: case ENOLINK: case ENOSPC: #ifdef illumos case ENOSTR: #endif case ENXIO: case EPIPE: case ERANGE: case EFAULT: case EROFS: zfs_error_aux(hdl, strerror(errno)); return (zfs_error(hdl, EZFS_BADBACKUP, errbuf)); default: return (zfs_standard_error(hdl, errno, errbuf)); } } return (err != 0); } /* * Routines specific to "zfs recv" */ static int recv_read(libzfs_handle_t *hdl, int fd, void *buf, int ilen, boolean_t byteswap, zio_cksum_t *zc) { char *cp = buf; int rv; int len = ilen; assert(ilen <= SPA_MAXBLOCKSIZE); do { rv = read(fd, cp, len); cp += rv; len -= rv; } while (rv > 0); if (rv < 0 || len != 0) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "failed to read from stream")); return (zfs_error(hdl, EZFS_BADSTREAM, dgettext(TEXT_DOMAIN, "cannot receive"))); } if (zc) { if (byteswap) fletcher_4_incremental_byteswap(buf, ilen, zc); else fletcher_4_incremental_native(buf, ilen, zc); } return (0); } static int recv_read_nvlist(libzfs_handle_t *hdl, int fd, int len, nvlist_t **nvp, boolean_t byteswap, zio_cksum_t *zc) { char *buf; int err; buf = zfs_alloc(hdl, len); if (buf == NULL) return (ENOMEM); err = recv_read(hdl, fd, buf, len, byteswap, zc); if (err != 0) { free(buf); return (err); } err = nvlist_unpack(buf, len, nvp, 0); free(buf); if (err != 0) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid " "stream (malformed nvlist)")); return (EINVAL); } return (0); } static int recv_rename(libzfs_handle_t *hdl, const char *name, const char *tryname, int baselen, char *newname, recvflags_t *flags) { static int seq; zfs_cmd_t zc = { 0 }; int err; prop_changelist_t *clp; zfs_handle_t *zhp; zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET); if (zhp == NULL) return (-1); clp = changelist_gather(zhp, ZFS_PROP_NAME, 0, flags->force ? MS_FORCE : 0); zfs_close(zhp); if (clp == NULL) return (-1); err = changelist_prefix(clp); if (err) return (err); zc.zc_objset_type = DMU_OST_ZFS; (void) strlcpy(zc.zc_name, name, sizeof (zc.zc_name)); if (tryname) { (void) strcpy(newname, tryname); (void) strlcpy(zc.zc_value, tryname, sizeof (zc.zc_value)); if (flags->verbose) { (void) printf("attempting rename %s to %s\n", zc.zc_name, zc.zc_value); } err = ioctl(hdl->libzfs_fd, ZFS_IOC_RENAME, &zc); if (err == 0) changelist_rename(clp, name, tryname); } else { err = ENOENT; } if (err != 0 && strncmp(name + baselen, "recv-", 5) != 0) { seq++; (void) snprintf(newname, ZFS_MAXNAMELEN, "%.*srecv-%u-%u", baselen, name, getpid(), seq); (void) strlcpy(zc.zc_value, newname, sizeof (zc.zc_value)); if (flags->verbose) { (void) printf("failed - trying rename %s to %s\n", zc.zc_name, zc.zc_value); } err = ioctl(hdl->libzfs_fd, ZFS_IOC_RENAME, &zc); if (err == 0) changelist_rename(clp, name, newname); if (err && flags->verbose) { (void) printf("failed (%u) - " "will try again on next pass\n", errno); } err = EAGAIN; } else if (flags->verbose) { if (err == 0) (void) printf("success\n"); else (void) printf("failed (%u)\n", errno); } (void) changelist_postfix(clp); changelist_free(clp); return (err); } static int recv_destroy(libzfs_handle_t *hdl, const char *name, int baselen, char *newname, recvflags_t *flags) { zfs_cmd_t zc = { 0 }; int err = 0; prop_changelist_t *clp; zfs_handle_t *zhp; boolean_t defer = B_FALSE; int spa_version; zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET); if (zhp == NULL) return (-1); clp = changelist_gather(zhp, ZFS_PROP_NAME, 0, flags->force ? MS_FORCE : 0); if (zfs_get_type(zhp) == ZFS_TYPE_SNAPSHOT && zfs_spa_version(zhp, &spa_version) == 0 && spa_version >= SPA_VERSION_USERREFS) defer = B_TRUE; zfs_close(zhp); if (clp == NULL) return (-1); err = changelist_prefix(clp); if (err) return (err); zc.zc_objset_type = DMU_OST_ZFS; zc.zc_defer_destroy = defer; (void) strlcpy(zc.zc_name, name, sizeof (zc.zc_name)); if (flags->verbose) (void) printf("attempting destroy %s\n", zc.zc_name); err = ioctl(hdl->libzfs_fd, ZFS_IOC_DESTROY, &zc); if (err == 0) { if (flags->verbose) (void) printf("success\n"); changelist_remove(clp, zc.zc_name); } (void) changelist_postfix(clp); changelist_free(clp); /* * Deferred destroy might destroy the snapshot or only mark it to be * destroyed later, and it returns success in either case. */ if (err != 0 || (defer && zfs_dataset_exists(hdl, name, ZFS_TYPE_SNAPSHOT))) { err = recv_rename(hdl, name, NULL, baselen, newname, flags); } return (err); } typedef struct guid_to_name_data { uint64_t guid; boolean_t bookmark_ok; char *name; char *skip; } guid_to_name_data_t; static int guid_to_name_cb(zfs_handle_t *zhp, void *arg) { guid_to_name_data_t *gtnd = arg; const char *slash; int err; if (gtnd->skip != NULL && (slash = strrchr(zhp->zfs_name, '/')) != NULL && strcmp(slash + 1, gtnd->skip) == 0) { zfs_close(zhp); return (0); } if (zfs_prop_get_int(zhp, ZFS_PROP_GUID) == gtnd->guid) { (void) strcpy(gtnd->name, zhp->zfs_name); zfs_close(zhp); return (EEXIST); } err = zfs_iter_children(zhp, guid_to_name_cb, gtnd); if (err != EEXIST && gtnd->bookmark_ok) err = zfs_iter_bookmarks(zhp, guid_to_name_cb, gtnd); zfs_close(zhp); return (err); } /* * Attempt to find the local dataset associated with this guid. In the case of * multiple matches, we attempt to find the "best" match by searching * progressively larger portions of the hierarchy. This allows one to send a * tree of datasets individually and guarantee that we will find the source * guid within that hierarchy, even if there are multiple matches elsewhere. */ static int guid_to_name(libzfs_handle_t *hdl, const char *parent, uint64_t guid, boolean_t bookmark_ok, char *name) { char pname[ZFS_MAXNAMELEN]; guid_to_name_data_t gtnd; gtnd.guid = guid; gtnd.bookmark_ok = bookmark_ok; gtnd.name = name; gtnd.skip = NULL; /* * Search progressively larger portions of the hierarchy, starting * with the filesystem specified by 'parent'. This will * select the "most local" version of the origin snapshot in the case * that there are multiple matching snapshots in the system. */ (void) strlcpy(pname, parent, sizeof (pname)); char *cp = strrchr(pname, '@'); if (cp == NULL) cp = strchr(pname, '\0'); for (; cp != NULL; cp = strrchr(pname, '/')) { /* Chop off the last component and open the parent */ *cp = '\0'; zfs_handle_t *zhp = make_dataset_handle(hdl, pname); if (zhp == NULL) continue; int err = guid_to_name_cb(zfs_handle_dup(zhp), >nd); if (err != EEXIST) err = zfs_iter_children(zhp, guid_to_name_cb, >nd); if (err != EEXIST && bookmark_ok) err = zfs_iter_bookmarks(zhp, guid_to_name_cb, >nd); zfs_close(zhp); if (err == EEXIST) return (0); /* * Remember the last portion of the dataset so we skip it next * time through (as we've already searched that portion of the * hierarchy). */ gtnd.skip = strrchr(pname, '/') + 1; } return (ENOENT); } /* * Return +1 if guid1 is before guid2, 0 if they are the same, and -1 if * guid1 is after guid2. */ static int created_before(libzfs_handle_t *hdl, avl_tree_t *avl, uint64_t guid1, uint64_t guid2) { nvlist_t *nvfs; char *fsname, *snapname; char buf[ZFS_MAXNAMELEN]; int rv; zfs_handle_t *guid1hdl, *guid2hdl; uint64_t create1, create2; if (guid2 == 0) return (0); if (guid1 == 0) return (1); nvfs = fsavl_find(avl, guid1, &snapname); VERIFY(0 == nvlist_lookup_string(nvfs, "name", &fsname)); (void) snprintf(buf, sizeof (buf), "%s@%s", fsname, snapname); guid1hdl = zfs_open(hdl, buf, ZFS_TYPE_SNAPSHOT); if (guid1hdl == NULL) return (-1); nvfs = fsavl_find(avl, guid2, &snapname); VERIFY(0 == nvlist_lookup_string(nvfs, "name", &fsname)); (void) snprintf(buf, sizeof (buf), "%s@%s", fsname, snapname); guid2hdl = zfs_open(hdl, buf, ZFS_TYPE_SNAPSHOT); if (guid2hdl == NULL) { zfs_close(guid1hdl); return (-1); } create1 = zfs_prop_get_int(guid1hdl, ZFS_PROP_CREATETXG); create2 = zfs_prop_get_int(guid2hdl, ZFS_PROP_CREATETXG); if (create1 < create2) rv = -1; else if (create1 > create2) rv = +1; else rv = 0; zfs_close(guid1hdl); zfs_close(guid2hdl); return (rv); } static int recv_incremental_replication(libzfs_handle_t *hdl, const char *tofs, recvflags_t *flags, nvlist_t *stream_nv, avl_tree_t *stream_avl, nvlist_t *renamed) { nvlist_t *local_nv, *deleted = NULL; avl_tree_t *local_avl; nvpair_t *fselem, *nextfselem; char *fromsnap; char newname[ZFS_MAXNAMELEN]; char guidname[32]; int error; boolean_t needagain, progress, recursive; char *s1, *s2; VERIFY(0 == nvlist_lookup_string(stream_nv, "fromsnap", &fromsnap)); recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") == ENOENT); if (flags->dryrun) return (0); again: needagain = progress = B_FALSE; VERIFY(0 == nvlist_alloc(&deleted, NV_UNIQUE_NAME, 0)); if ((error = gather_nvlist(hdl, tofs, fromsnap, NULL, recursive, &local_nv, &local_avl)) != 0) return (error); /* * Process deletes and renames */ for (fselem = nvlist_next_nvpair(local_nv, NULL); fselem; fselem = nextfselem) { nvlist_t *nvfs, *snaps; nvlist_t *stream_nvfs = NULL; nvpair_t *snapelem, *nextsnapelem; uint64_t fromguid = 0; uint64_t originguid = 0; uint64_t stream_originguid = 0; uint64_t parent_fromsnap_guid, stream_parent_fromsnap_guid; char *fsname, *stream_fsname; nextfselem = nvlist_next_nvpair(local_nv, fselem); VERIFY(0 == nvpair_value_nvlist(fselem, &nvfs)); VERIFY(0 == nvlist_lookup_nvlist(nvfs, "snaps", &snaps)); VERIFY(0 == nvlist_lookup_string(nvfs, "name", &fsname)); VERIFY(0 == nvlist_lookup_uint64(nvfs, "parentfromsnap", &parent_fromsnap_guid)); (void) nvlist_lookup_uint64(nvfs, "origin", &originguid); /* * First find the stream's fs, so we can check for * a different origin (due to "zfs promote") */ for (snapelem = nvlist_next_nvpair(snaps, NULL); snapelem; snapelem = nvlist_next_nvpair(snaps, snapelem)) { uint64_t thisguid; VERIFY(0 == nvpair_value_uint64(snapelem, &thisguid)); stream_nvfs = fsavl_find(stream_avl, thisguid, NULL); if (stream_nvfs != NULL) break; } /* check for promote */ (void) nvlist_lookup_uint64(stream_nvfs, "origin", &stream_originguid); if (stream_nvfs && originguid != stream_originguid) { switch (created_before(hdl, local_avl, stream_originguid, originguid)) { case 1: { /* promote it! */ zfs_cmd_t zc = { 0 }; nvlist_t *origin_nvfs; char *origin_fsname; if (flags->verbose) (void) printf("promoting %s\n", fsname); origin_nvfs = fsavl_find(local_avl, originguid, NULL); VERIFY(0 == nvlist_lookup_string(origin_nvfs, "name", &origin_fsname)); (void) strlcpy(zc.zc_value, origin_fsname, sizeof (zc.zc_value)); (void) strlcpy(zc.zc_name, fsname, sizeof (zc.zc_name)); error = zfs_ioctl(hdl, ZFS_IOC_PROMOTE, &zc); if (error == 0) progress = B_TRUE; break; } default: break; case -1: fsavl_destroy(local_avl); nvlist_free(local_nv); return (-1); } /* * We had/have the wrong origin, therefore our * list of snapshots is wrong. Need to handle * them on the next pass. */ needagain = B_TRUE; continue; } for (snapelem = nvlist_next_nvpair(snaps, NULL); snapelem; snapelem = nextsnapelem) { uint64_t thisguid; char *stream_snapname; nvlist_t *found, *props; nextsnapelem = nvlist_next_nvpair(snaps, snapelem); VERIFY(0 == nvpair_value_uint64(snapelem, &thisguid)); found = fsavl_find(stream_avl, thisguid, &stream_snapname); /* check for delete */ if (found == NULL) { char name[ZFS_MAXNAMELEN]; if (!flags->force) continue; (void) snprintf(name, sizeof (name), "%s@%s", fsname, nvpair_name(snapelem)); error = recv_destroy(hdl, name, strlen(fsname)+1, newname, flags); if (error) needagain = B_TRUE; else progress = B_TRUE; sprintf(guidname, "%lu", thisguid); nvlist_add_boolean(deleted, guidname); continue; } stream_nvfs = found; if (0 == nvlist_lookup_nvlist(stream_nvfs, "snapprops", &props) && 0 == nvlist_lookup_nvlist(props, stream_snapname, &props)) { zfs_cmd_t zc = { 0 }; zc.zc_cookie = B_TRUE; /* received */ (void) snprintf(zc.zc_name, sizeof (zc.zc_name), "%s@%s", fsname, nvpair_name(snapelem)); if (zcmd_write_src_nvlist(hdl, &zc, props) == 0) { (void) zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc); zcmd_free_nvlists(&zc); } } /* check for different snapname */ if (strcmp(nvpair_name(snapelem), stream_snapname) != 0) { char name[ZFS_MAXNAMELEN]; char tryname[ZFS_MAXNAMELEN]; (void) snprintf(name, sizeof (name), "%s@%s", fsname, nvpair_name(snapelem)); (void) snprintf(tryname, sizeof (name), "%s@%s", fsname, stream_snapname); error = recv_rename(hdl, name, tryname, strlen(fsname)+1, newname, flags); if (error) needagain = B_TRUE; else progress = B_TRUE; } if (strcmp(stream_snapname, fromsnap) == 0) fromguid = thisguid; } /* check for delete */ if (stream_nvfs == NULL) { if (!flags->force) continue; error = recv_destroy(hdl, fsname, strlen(tofs)+1, newname, flags); if (error) needagain = B_TRUE; else progress = B_TRUE; sprintf(guidname, "%lu", parent_fromsnap_guid); nvlist_add_boolean(deleted, guidname); continue; } if (fromguid == 0) { if (flags->verbose) { (void) printf("local fs %s does not have " "fromsnap (%s in stream); must have " "been deleted locally; ignoring\n", fsname, fromsnap); } continue; } VERIFY(0 == nvlist_lookup_string(stream_nvfs, "name", &stream_fsname)); VERIFY(0 == nvlist_lookup_uint64(stream_nvfs, "parentfromsnap", &stream_parent_fromsnap_guid)); s1 = strrchr(fsname, '/'); s2 = strrchr(stream_fsname, '/'); /* * Check if we're going to rename based on parent guid change * and the current parent guid was also deleted. If it was then * rename will fail and is likely unneeded, so avoid this and * force an early retry to determine the new * parent_fromsnap_guid. */ if (stream_parent_fromsnap_guid != 0 && parent_fromsnap_guid != 0 && stream_parent_fromsnap_guid != parent_fromsnap_guid) { sprintf(guidname, "%lu", parent_fromsnap_guid); if (nvlist_exists(deleted, guidname)) { progress = B_TRUE; needagain = B_TRUE; goto doagain; } } /* * Check for rename. If the exact receive path is specified, it * does not count as a rename, but we still need to check the * datasets beneath it. */ if ((stream_parent_fromsnap_guid != 0 && parent_fromsnap_guid != 0 && stream_parent_fromsnap_guid != parent_fromsnap_guid) || ((flags->isprefix || strcmp(tofs, fsname) != 0) && (s1 != NULL) && (s2 != NULL) && strcmp(s1, s2) != 0)) { nvlist_t *parent; char tryname[ZFS_MAXNAMELEN]; parent = fsavl_find(local_avl, stream_parent_fromsnap_guid, NULL); /* * NB: parent might not be found if we used the * tosnap for stream_parent_fromsnap_guid, * because the parent is a newly-created fs; * we'll be able to rename it after we recv the * new fs. */ if (parent != NULL) { char *pname; VERIFY(0 == nvlist_lookup_string(parent, "name", &pname)); (void) snprintf(tryname, sizeof (tryname), "%s%s", pname, strrchr(stream_fsname, '/')); } else { tryname[0] = '\0'; if (flags->verbose) { (void) printf("local fs %s new parent " "not found\n", fsname); } } newname[0] = '\0'; error = recv_rename(hdl, fsname, tryname, strlen(tofs)+1, newname, flags); if (renamed != NULL && newname[0] != '\0') { VERIFY(0 == nvlist_add_boolean(renamed, newname)); } if (error) needagain = B_TRUE; else progress = B_TRUE; } } doagain: fsavl_destroy(local_avl); nvlist_free(local_nv); nvlist_free(deleted); if (needagain && progress) { /* do another pass to fix up temporary names */ if (flags->verbose) (void) printf("another pass:\n"); goto again; } return (needagain); } static int zfs_receive_package(libzfs_handle_t *hdl, int fd, const char *destname, recvflags_t *flags, dmu_replay_record_t *drr, zio_cksum_t *zc, char **top_zfs, int cleanup_fd, uint64_t *action_handlep) { nvlist_t *stream_nv = NULL; avl_tree_t *stream_avl = NULL; char *fromsnap = NULL; char *cp; char tofs[ZFS_MAXNAMELEN]; char sendfs[ZFS_MAXNAMELEN]; char errbuf[1024]; dmu_replay_record_t drre; int error; boolean_t anyerr = B_FALSE; boolean_t softerr = B_FALSE; boolean_t recursive; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot receive")); assert(drr->drr_type == DRR_BEGIN); assert(drr->drr_u.drr_begin.drr_magic == DMU_BACKUP_MAGIC); assert(DMU_GET_STREAM_HDRTYPE(drr->drr_u.drr_begin.drr_versioninfo) == DMU_COMPOUNDSTREAM); /* * Read in the nvlist from the stream. */ if (drr->drr_payloadlen != 0) { error = recv_read_nvlist(hdl, fd, drr->drr_payloadlen, &stream_nv, flags->byteswap, zc); if (error) { error = zfs_error(hdl, EZFS_BADSTREAM, errbuf); goto out; } } recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") == ENOENT); if (recursive && strchr(destname, '@')) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot specify snapshot name for multi-snapshot stream")); error = zfs_error(hdl, EZFS_BADSTREAM, errbuf); goto out; } /* * Read in the end record and verify checksum. */ if (0 != (error = recv_read(hdl, fd, &drre, sizeof (drre), flags->byteswap, NULL))) goto out; if (flags->byteswap) { drre.drr_type = BSWAP_32(drre.drr_type); drre.drr_u.drr_end.drr_checksum.zc_word[0] = BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[0]); drre.drr_u.drr_end.drr_checksum.zc_word[1] = BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[1]); drre.drr_u.drr_end.drr_checksum.zc_word[2] = BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[2]); drre.drr_u.drr_end.drr_checksum.zc_word[3] = BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[3]); } if (drre.drr_type != DRR_END) { error = zfs_error(hdl, EZFS_BADSTREAM, errbuf); goto out; } if (!ZIO_CHECKSUM_EQUAL(drre.drr_u.drr_end.drr_checksum, *zc)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "incorrect header checksum")); error = zfs_error(hdl, EZFS_BADSTREAM, errbuf); goto out; } (void) nvlist_lookup_string(stream_nv, "fromsnap", &fromsnap); if (drr->drr_payloadlen != 0) { nvlist_t *stream_fss; VERIFY(0 == nvlist_lookup_nvlist(stream_nv, "fss", &stream_fss)); if ((stream_avl = fsavl_create(stream_fss)) == NULL) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "couldn't allocate avl tree")); error = zfs_error(hdl, EZFS_NOMEM, errbuf); goto out; } if (fromsnap != NULL) { nvlist_t *renamed = NULL; nvpair_t *pair = NULL; (void) strlcpy(tofs, destname, ZFS_MAXNAMELEN); if (flags->isprefix) { struct drr_begin *drrb = &drr->drr_u.drr_begin; int i; if (flags->istail) { cp = strrchr(drrb->drr_toname, '/'); if (cp == NULL) { (void) strlcat(tofs, "/", ZFS_MAXNAMELEN); i = 0; } else { i = (cp - drrb->drr_toname); } } else { i = strcspn(drrb->drr_toname, "/@"); } /* zfs_receive_one() will create_parents() */ (void) strlcat(tofs, &drrb->drr_toname[i], ZFS_MAXNAMELEN); *strchr(tofs, '@') = '\0'; } if (recursive && !flags->dryrun && !flags->nomount) { VERIFY(0 == nvlist_alloc(&renamed, NV_UNIQUE_NAME, 0)); } softerr = recv_incremental_replication(hdl, tofs, flags, stream_nv, stream_avl, renamed); /* Unmount renamed filesystems before receiving. */ while ((pair = nvlist_next_nvpair(renamed, pair)) != NULL) { zfs_handle_t *zhp; prop_changelist_t *clp = NULL; zhp = zfs_open(hdl, nvpair_name(pair), ZFS_TYPE_FILESYSTEM); if (zhp != NULL) { clp = changelist_gather(zhp, ZFS_PROP_MOUNTPOINT, 0, 0); zfs_close(zhp); if (clp != NULL) { softerr |= changelist_prefix(clp); changelist_free(clp); } } } nvlist_free(renamed); } } /* * Get the fs specified by the first path in the stream (the top level * specified by 'zfs send') and pass it to each invocation of * zfs_receive_one(). */ (void) strlcpy(sendfs, drr->drr_u.drr_begin.drr_toname, ZFS_MAXNAMELEN); if ((cp = strchr(sendfs, '@')) != NULL) *cp = '\0'; /* Finally, receive each contained stream */ do { /* * we should figure out if it has a recoverable * error, in which case do a recv_skip() and drive on. * Note, if we fail due to already having this guid, * zfs_receive_one() will take care of it (ie, * recv_skip() and return 0). */ error = zfs_receive_impl(hdl, destname, NULL, flags, fd, sendfs, stream_nv, stream_avl, top_zfs, cleanup_fd, action_handlep); if (error == ENODATA) { error = 0; break; } anyerr |= error; } while (error == 0); if (drr->drr_payloadlen != 0 && fromsnap != NULL) { /* * Now that we have the fs's they sent us, try the * renames again. */ softerr = recv_incremental_replication(hdl, tofs, flags, stream_nv, stream_avl, NULL); } out: fsavl_destroy(stream_avl); if (stream_nv) nvlist_free(stream_nv); if (softerr) error = -2; if (anyerr) error = -1; return (error); } static void trunc_prop_errs(int truncated) { ASSERT(truncated != 0); if (truncated == 1) (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "1 more property could not be set\n")); else (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "%d more properties could not be set\n"), truncated); } static int recv_skip(libzfs_handle_t *hdl, int fd, boolean_t byteswap) { dmu_replay_record_t *drr; void *buf = zfs_alloc(hdl, SPA_MAXBLOCKSIZE); char errbuf[1024]; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot receive:")); /* XXX would be great to use lseek if possible... */ drr = buf; while (recv_read(hdl, fd, drr, sizeof (dmu_replay_record_t), byteswap, NULL) == 0) { if (byteswap) drr->drr_type = BSWAP_32(drr->drr_type); switch (drr->drr_type) { case DRR_BEGIN: if (drr->drr_payloadlen != 0) { (void) recv_read(hdl, fd, buf, drr->drr_payloadlen, B_FALSE, NULL); } break; case DRR_END: free(buf); return (0); case DRR_OBJECT: if (byteswap) { drr->drr_u.drr_object.drr_bonuslen = BSWAP_32(drr->drr_u.drr_object. drr_bonuslen); } (void) recv_read(hdl, fd, buf, P2ROUNDUP(drr->drr_u.drr_object.drr_bonuslen, 8), B_FALSE, NULL); break; case DRR_WRITE: if (byteswap) { drr->drr_u.drr_write.drr_length = BSWAP_64(drr->drr_u.drr_write.drr_length); } (void) recv_read(hdl, fd, buf, drr->drr_u.drr_write.drr_length, B_FALSE, NULL); break; case DRR_SPILL: if (byteswap) { drr->drr_u.drr_write.drr_length = BSWAP_64(drr->drr_u.drr_spill.drr_length); } (void) recv_read(hdl, fd, buf, drr->drr_u.drr_spill.drr_length, B_FALSE, NULL); break; case DRR_WRITE_EMBEDDED: if (byteswap) { drr->drr_u.drr_write_embedded.drr_psize = BSWAP_32(drr->drr_u.drr_write_embedded. drr_psize); } (void) recv_read(hdl, fd, buf, P2ROUNDUP(drr->drr_u.drr_write_embedded.drr_psize, 8), B_FALSE, NULL); break; case DRR_WRITE_BYREF: case DRR_FREEOBJECTS: case DRR_FREE: break; default: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid record type")); return (zfs_error(hdl, EZFS_BADSTREAM, errbuf)); } } free(buf); return (-1); } static void recv_ecksum_set_aux(libzfs_handle_t *hdl, const char *target_snap, boolean_t resumable) { char target_fs[ZFS_MAXNAMELEN]; zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "checksum mismatch or incomplete stream")); if (!resumable) return; (void) strlcpy(target_fs, target_snap, sizeof (target_fs)); *strchr(target_fs, '@') = '\0'; zfs_handle_t *zhp = zfs_open(hdl, target_fs, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); if (zhp == NULL) return; char token_buf[ZFS_MAXPROPLEN]; int error = zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN, token_buf, sizeof (token_buf), NULL, NULL, 0, B_TRUE); if (error == 0) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "checksum mismatch or incomplete stream.\n" "Partially received snapshot is saved.\n" "A resuming stream can be generated on the sending " "system by running:\n" " zfs send -t %s"), token_buf); } zfs_close(zhp); } /* * Restores a backup of tosnap from the file descriptor specified by infd. */ static int zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap, const char *originsnap, recvflags_t *flags, dmu_replay_record_t *drr, dmu_replay_record_t *drr_noswap, const char *sendfs, nvlist_t *stream_nv, avl_tree_t *stream_avl, char **top_zfs, int cleanup_fd, uint64_t *action_handlep) { zfs_cmd_t zc = { 0 }; time_t begin_time; int ioctl_err, ioctl_errno, err; char *cp; struct drr_begin *drrb = &drr->drr_u.drr_begin; char errbuf[1024]; char prop_errbuf[1024]; const char *chopprefix; boolean_t newfs = B_FALSE; boolean_t stream_wantsnewfs; uint64_t parent_snapguid = 0; prop_changelist_t *clp = NULL; nvlist_t *snapprops_nvlist = NULL; zprop_errflags_t prop_errflags; boolean_t recursive; begin_time = time(NULL); (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot receive")); recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") == ENOENT); if (stream_avl != NULL) { char *snapname; nvlist_t *fs = fsavl_find(stream_avl, drrb->drr_toguid, &snapname); nvlist_t *props; int ret; (void) nvlist_lookup_uint64(fs, "parentfromsnap", &parent_snapguid); err = nvlist_lookup_nvlist(fs, "props", &props); if (err) VERIFY(0 == nvlist_alloc(&props, NV_UNIQUE_NAME, 0)); if (flags->canmountoff) { VERIFY(0 == nvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_CANMOUNT), 0)); } ret = zcmd_write_src_nvlist(hdl, &zc, props); if (err) nvlist_free(props); if (0 == nvlist_lookup_nvlist(fs, "snapprops", &props)) { VERIFY(0 == nvlist_lookup_nvlist(props, snapname, &snapprops_nvlist)); } if (ret != 0) return (-1); } cp = NULL; /* * Determine how much of the snapshot name stored in the stream * we are going to tack on to the name they specified on the * command line, and how much we are going to chop off. * * If they specified a snapshot, chop the entire name stored in * the stream. */ if (flags->istail) { /* * A filesystem was specified with -e. We want to tack on only * the tail of the sent snapshot path. */ if (strchr(tosnap, '@')) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid " "argument - snapshot not allowed with -e")); return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); } chopprefix = strrchr(sendfs, '/'); if (chopprefix == NULL) { /* * The tail is the poolname, so we need to * prepend a path separator. */ int len = strlen(drrb->drr_toname); cp = malloc(len + 2); cp[0] = '/'; (void) strcpy(&cp[1], drrb->drr_toname); chopprefix = cp; } else { chopprefix = drrb->drr_toname + (chopprefix - sendfs); } } else if (flags->isprefix) { /* * A filesystem was specified with -d. We want to tack on * everything but the first element of the sent snapshot path * (all but the pool name). */ if (strchr(tosnap, '@')) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid " "argument - snapshot not allowed with -d")); return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); } chopprefix = strchr(drrb->drr_toname, '/'); if (chopprefix == NULL) chopprefix = strchr(drrb->drr_toname, '@'); } else if (strchr(tosnap, '@') == NULL) { /* * If a filesystem was specified without -d or -e, we want to * tack on everything after the fs specified by 'zfs send'. */ chopprefix = drrb->drr_toname + strlen(sendfs); } else { /* A snapshot was specified as an exact path (no -d or -e). */ if (recursive) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot specify snapshot name for multi-snapshot " "stream")); return (zfs_error(hdl, EZFS_BADSTREAM, errbuf)); } chopprefix = drrb->drr_toname + strlen(drrb->drr_toname); } ASSERT(strstr(drrb->drr_toname, sendfs) == drrb->drr_toname); ASSERT(chopprefix > drrb->drr_toname); ASSERT(chopprefix <= drrb->drr_toname + strlen(drrb->drr_toname)); ASSERT(chopprefix[0] == '/' || chopprefix[0] == '@' || chopprefix[0] == '\0'); /* * Determine name of destination snapshot, store in zc_value. */ (void) strcpy(zc.zc_value, tosnap); (void) strncat(zc.zc_value, chopprefix, sizeof (zc.zc_value)); #ifdef __FreeBSD__ if (zfs_ioctl_version == ZFS_IOCVER_UNDEF) zfs_ioctl_version = get_zfs_ioctl_version(); /* * For forward compatibility hide tosnap in zc_value */ if (zfs_ioctl_version < ZFS_IOCVER_LZC) (void) strcpy(zc.zc_value + strlen(zc.zc_value) + 1, tosnap); #endif free(cp); if (!zfs_name_valid(zc.zc_value, ZFS_TYPE_SNAPSHOT)) { zcmd_free_nvlists(&zc); return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); } /* * Determine the name of the origin snapshot, store in zc_string. */ if (drrb->drr_flags & DRR_FLAG_CLONE) { if (guid_to_name(hdl, zc.zc_value, drrb->drr_fromguid, B_FALSE, zc.zc_string) != 0) { zcmd_free_nvlists(&zc); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "local origin for clone %s does not exist"), zc.zc_value); return (zfs_error(hdl, EZFS_NOENT, errbuf)); } if (flags->verbose) (void) printf("found clone origin %s\n", zc.zc_string); } else if (originsnap) { (void) strncpy(zc.zc_string, originsnap, ZFS_MAXNAMELEN); if (flags->verbose) (void) printf("using provided clone origin %s\n", zc.zc_string); } boolean_t resuming = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) & DMU_BACKUP_FEATURE_RESUMING; stream_wantsnewfs = (drrb->drr_fromguid == 0 || (drrb->drr_flags & DRR_FLAG_CLONE) || originsnap) && !resuming; if (stream_wantsnewfs) { /* * if the parent fs does not exist, look for it based on * the parent snap GUID */ (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot receive new filesystem stream")); (void) strcpy(zc.zc_name, zc.zc_value); cp = strrchr(zc.zc_name, '/'); if (cp) *cp = '\0'; if (cp && !zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_DATASET)) { char suffix[ZFS_MAXNAMELEN]; (void) strcpy(suffix, strrchr(zc.zc_value, '/')); if (guid_to_name(hdl, zc.zc_name, parent_snapguid, B_FALSE, zc.zc_value) == 0) { *strchr(zc.zc_value, '@') = '\0'; (void) strcat(zc.zc_value, suffix); } } } else { /* * if the fs does not exist, look for it based on the * fromsnap GUID */ (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot receive incremental stream")); (void) strcpy(zc.zc_name, zc.zc_value); *strchr(zc.zc_name, '@') = '\0'; /* * If the exact receive path was specified and this is the * topmost path in the stream, then if the fs does not exist we * should look no further. */ if ((flags->isprefix || (*(chopprefix = drrb->drr_toname + strlen(sendfs)) != '\0' && *chopprefix != '@')) && !zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_DATASET)) { char snap[ZFS_MAXNAMELEN]; (void) strcpy(snap, strchr(zc.zc_value, '@')); if (guid_to_name(hdl, zc.zc_name, drrb->drr_fromguid, B_FALSE, zc.zc_value) == 0) { *strchr(zc.zc_value, '@') = '\0'; (void) strcat(zc.zc_value, snap); } } } (void) strcpy(zc.zc_name, zc.zc_value); *strchr(zc.zc_name, '@') = '\0'; if (zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_DATASET)) { zfs_handle_t *zhp; /* * Destination fs exists. It must be one of these cases: * - an incremental send stream * - the stream specifies a new fs (full stream or clone) * and they want us to blow away the existing fs (and * have therefore specified -F and removed any snapshots) * - we are resuming a failed receive. */ if (stream_wantsnewfs) { if (!flags->force) { zcmd_free_nvlists(&zc); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "destination '%s' exists\n" "must specify -F to overwrite it"), zc.zc_name); return (zfs_error(hdl, EZFS_EXISTS, errbuf)); } if (ioctl(hdl->libzfs_fd, ZFS_IOC_SNAPSHOT_LIST_NEXT, &zc) == 0) { zcmd_free_nvlists(&zc); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "destination has snapshots (eg. %s)\n" "must destroy them to overwrite it"), zc.zc_name); return (zfs_error(hdl, EZFS_EXISTS, errbuf)); } } if ((zhp = zfs_open(hdl, zc.zc_name, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME)) == NULL) { zcmd_free_nvlists(&zc); return (-1); } if (stream_wantsnewfs && zhp->zfs_dmustats.dds_origin[0]) { zcmd_free_nvlists(&zc); zfs_close(zhp); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "destination '%s' is a clone\n" "must destroy it to overwrite it"), zc.zc_name); return (zfs_error(hdl, EZFS_EXISTS, errbuf)); } if (!flags->dryrun && zhp->zfs_type == ZFS_TYPE_FILESYSTEM && stream_wantsnewfs) { /* We can't do online recv in this case */ clp = changelist_gather(zhp, ZFS_PROP_NAME, 0, 0); if (clp == NULL) { zfs_close(zhp); zcmd_free_nvlists(&zc); return (-1); } if (changelist_prefix(clp) != 0) { changelist_free(clp); zfs_close(zhp); zcmd_free_nvlists(&zc); return (-1); } } /* * If we are resuming a newfs, set newfs here so that we will * mount it if the recv succeeds this time. We can tell * that it was a newfs on the first recv because the fs * itself will be inconsistent (if the fs existed when we * did the first recv, we would have received it into * .../%recv). */ if (resuming && zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT)) newfs = B_TRUE; zfs_close(zhp); } else { /* * Destination filesystem does not exist. Therefore we better * be creating a new filesystem (either from a full backup, or * a clone). It would therefore be invalid if the user * specified only the pool name (i.e. if the destination name * contained no slash character). */ if (!stream_wantsnewfs || (cp = strrchr(zc.zc_name, '/')) == NULL) { zcmd_free_nvlists(&zc); zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "destination '%s' does not exist"), zc.zc_name); return (zfs_error(hdl, EZFS_NOENT, errbuf)); } /* * Trim off the final dataset component so we perform the * recvbackup ioctl to the filesystems's parent. */ *cp = '\0'; if (flags->isprefix && !flags->istail && !flags->dryrun && create_parents(hdl, zc.zc_value, strlen(tosnap)) != 0) { zcmd_free_nvlists(&zc); return (zfs_error(hdl, EZFS_BADRESTORE, errbuf)); } newfs = B_TRUE; } zc.zc_begin_record = *drr_noswap; zc.zc_cookie = infd; zc.zc_guid = flags->force; zc.zc_resumable = flags->resumable; if (flags->verbose) { (void) printf("%s %s stream of %s into %s\n", flags->dryrun ? "would receive" : "receiving", drrb->drr_fromguid ? "incremental" : "full", drrb->drr_toname, zc.zc_value); (void) fflush(stdout); } if (flags->dryrun) { zcmd_free_nvlists(&zc); return (recv_skip(hdl, infd, flags->byteswap)); } zc.zc_nvlist_dst = (uint64_t)(uintptr_t)prop_errbuf; zc.zc_nvlist_dst_size = sizeof (prop_errbuf); zc.zc_cleanup_fd = cleanup_fd; zc.zc_action_handle = *action_handlep; err = ioctl_err = zfs_ioctl(hdl, ZFS_IOC_RECV, &zc); ioctl_errno = errno; prop_errflags = (zprop_errflags_t)zc.zc_obj; if (err == 0) { nvlist_t *prop_errors; VERIFY(0 == nvlist_unpack((void *)(uintptr_t)zc.zc_nvlist_dst, zc.zc_nvlist_dst_size, &prop_errors, 0)); nvpair_t *prop_err = NULL; while ((prop_err = nvlist_next_nvpair(prop_errors, prop_err)) != NULL) { char tbuf[1024]; zfs_prop_t prop; int intval; prop = zfs_name_to_prop(nvpair_name(prop_err)); (void) nvpair_value_int32(prop_err, &intval); if (strcmp(nvpair_name(prop_err), ZPROP_N_MORE_ERRORS) == 0) { trunc_prop_errs(intval); break; } else { (void) snprintf(tbuf, sizeof (tbuf), dgettext(TEXT_DOMAIN, "cannot receive %s property on %s"), nvpair_name(prop_err), zc.zc_name); zfs_setprop_error(hdl, prop, intval, tbuf); } } nvlist_free(prop_errors); } zc.zc_nvlist_dst = 0; zc.zc_nvlist_dst_size = 0; zcmd_free_nvlists(&zc); if (err == 0 && snapprops_nvlist) { zfs_cmd_t zc2 = { 0 }; (void) strcpy(zc2.zc_name, zc.zc_value); zc2.zc_cookie = B_TRUE; /* received */ if (zcmd_write_src_nvlist(hdl, &zc2, snapprops_nvlist) == 0) { (void) zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc2); zcmd_free_nvlists(&zc2); } } if (err && (ioctl_errno == ENOENT || ioctl_errno == EEXIST)) { /* * It may be that this snapshot already exists, * in which case we want to consume & ignore it * rather than failing. */ avl_tree_t *local_avl; nvlist_t *local_nv, *fs; cp = strchr(zc.zc_value, '@'); /* * XXX Do this faster by just iterating over snaps in * this fs. Also if zc_value does not exist, we will * get a strange "does not exist" error message. */ *cp = '\0'; if (gather_nvlist(hdl, zc.zc_value, NULL, NULL, B_FALSE, &local_nv, &local_avl) == 0) { *cp = '@'; fs = fsavl_find(local_avl, drrb->drr_toguid, NULL); fsavl_destroy(local_avl); nvlist_free(local_nv); if (fs != NULL) { if (flags->verbose) { (void) printf("snap %s already exists; " "ignoring\n", zc.zc_value); } err = ioctl_err = recv_skip(hdl, infd, flags->byteswap); } } *cp = '@'; } if (ioctl_err != 0) { switch (ioctl_errno) { case ENODEV: cp = strchr(zc.zc_value, '@'); *cp = '\0'; zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "most recent snapshot of %s does not\n" "match incremental source"), zc.zc_value); (void) zfs_error(hdl, EZFS_BADRESTORE, errbuf); *cp = '@'; break; case ETXTBSY: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "destination %s has been modified\n" "since most recent snapshot"), zc.zc_name); (void) zfs_error(hdl, EZFS_BADRESTORE, errbuf); break; case EEXIST: cp = strchr(zc.zc_value, '@'); if (newfs) { /* it's the containing fs that exists */ *cp = '\0'; } zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "destination already exists")); (void) zfs_error_fmt(hdl, EZFS_EXISTS, dgettext(TEXT_DOMAIN, "cannot restore to %s"), zc.zc_value); *cp = '@'; break; case EINVAL: (void) zfs_error(hdl, EZFS_BADSTREAM, errbuf); break; case ECKSUM: recv_ecksum_set_aux(hdl, zc.zc_value, flags->resumable); (void) zfs_error(hdl, EZFS_BADSTREAM, errbuf); break; case ENOTSUP: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be upgraded to receive this stream.")); (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); break; case EDQUOT: zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "destination %s space quota exceeded"), zc.zc_name); (void) zfs_error(hdl, EZFS_NOSPC, errbuf); break; default: (void) zfs_standard_error(hdl, ioctl_errno, errbuf); } } /* * Mount the target filesystem (if created). Also mount any * children of the target filesystem if we did a replication * receive (indicated by stream_avl being non-NULL). */ cp = strchr(zc.zc_value, '@'); if (cp && (ioctl_err == 0 || !newfs)) { zfs_handle_t *h; *cp = '\0'; h = zfs_open(hdl, zc.zc_value, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); if (h != NULL) { if (h->zfs_type == ZFS_TYPE_VOLUME) { *cp = '@'; } else if (newfs || stream_avl) { /* * Track the first/top of hierarchy fs, * for mounting and sharing later. */ if (top_zfs && *top_zfs == NULL) *top_zfs = zfs_strdup(hdl, zc.zc_value); } zfs_close(h); } *cp = '@'; } if (clp) { err |= changelist_postfix(clp); changelist_free(clp); } if (prop_errflags & ZPROP_ERR_NOCLEAR) { (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Warning: " "failed to clear unreceived properties on %s"), zc.zc_name); (void) fprintf(stderr, "\n"); } if (prop_errflags & ZPROP_ERR_NORESTORE) { (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Warning: " "failed to restore original properties on %s"), zc.zc_name); (void) fprintf(stderr, "\n"); } if (err || ioctl_err) return (-1); *action_handlep = zc.zc_action_handle; if (flags->verbose) { char buf1[64]; char buf2[64]; uint64_t bytes = zc.zc_cookie; time_t delta = time(NULL) - begin_time; if (delta == 0) delta = 1; zfs_nicenum(bytes, buf1, sizeof (buf1)); zfs_nicenum(bytes/delta, buf2, sizeof (buf1)); (void) printf("received %sB stream in %lu seconds (%sB/sec)\n", buf1, delta, buf2); } return (0); } static int zfs_receive_impl(libzfs_handle_t *hdl, const char *tosnap, const char *originsnap, recvflags_t *flags, int infd, const char *sendfs, nvlist_t *stream_nv, avl_tree_t *stream_avl, char **top_zfs, int cleanup_fd, uint64_t *action_handlep) { int err; dmu_replay_record_t drr, drr_noswap; struct drr_begin *drrb = &drr.drr_u.drr_begin; char errbuf[1024]; zio_cksum_t zcksum = { 0 }; uint64_t featureflags; int hdrtype; (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, "cannot receive")); if (flags->isprefix && !zfs_dataset_exists(hdl, tosnap, ZFS_TYPE_DATASET)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "specified fs " "(%s) does not exist"), tosnap); return (zfs_error(hdl, EZFS_NOENT, errbuf)); } if (originsnap && !zfs_dataset_exists(hdl, originsnap, ZFS_TYPE_DATASET)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "specified origin fs " "(%s) does not exist"), originsnap); return (zfs_error(hdl, EZFS_NOENT, errbuf)); } /* read in the BEGIN record */ if (0 != (err = recv_read(hdl, infd, &drr, sizeof (drr), B_FALSE, &zcksum))) return (err); if (drr.drr_type == DRR_END || drr.drr_type == BSWAP_32(DRR_END)) { /* It's the double end record at the end of a package */ return (ENODATA); } /* the kernel needs the non-byteswapped begin record */ drr_noswap = drr; flags->byteswap = B_FALSE; if (drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC)) { /* * We computed the checksum in the wrong byteorder in * recv_read() above; do it again correctly. */ bzero(&zcksum, sizeof (zio_cksum_t)); fletcher_4_incremental_byteswap(&drr, sizeof (drr), &zcksum); flags->byteswap = B_TRUE; drr.drr_type = BSWAP_32(drr.drr_type); drr.drr_payloadlen = BSWAP_32(drr.drr_payloadlen); drrb->drr_magic = BSWAP_64(drrb->drr_magic); drrb->drr_versioninfo = BSWAP_64(drrb->drr_versioninfo); drrb->drr_creation_time = BSWAP_64(drrb->drr_creation_time); drrb->drr_type = BSWAP_32(drrb->drr_type); drrb->drr_flags = BSWAP_32(drrb->drr_flags); drrb->drr_toguid = BSWAP_64(drrb->drr_toguid); drrb->drr_fromguid = BSWAP_64(drrb->drr_fromguid); } if (drrb->drr_magic != DMU_BACKUP_MAGIC || drr.drr_type != DRR_BEGIN) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid " "stream (bad magic number)")); return (zfs_error(hdl, EZFS_BADSTREAM, errbuf)); } featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo); hdrtype = DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo); if (!DMU_STREAM_SUPPORTED(featureflags) || (hdrtype != DMU_SUBSTREAM && hdrtype != DMU_COMPOUNDSTREAM)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "stream has unsupported feature, feature flags = %lx"), featureflags); return (zfs_error(hdl, EZFS_BADSTREAM, errbuf)); } if (strchr(drrb->drr_toname, '@') == NULL) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid " "stream (bad snapshot name)")); return (zfs_error(hdl, EZFS_BADSTREAM, errbuf)); } if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) == DMU_SUBSTREAM) { char nonpackage_sendfs[ZFS_MAXNAMELEN]; if (sendfs == NULL) { /* * We were not called from zfs_receive_package(). Get * the fs specified by 'zfs send'. */ char *cp; (void) strlcpy(nonpackage_sendfs, drr.drr_u.drr_begin.drr_toname, ZFS_MAXNAMELEN); if ((cp = strchr(nonpackage_sendfs, '@')) != NULL) *cp = '\0'; sendfs = nonpackage_sendfs; } return (zfs_receive_one(hdl, infd, tosnap, originsnap, flags, &drr, &drr_noswap, sendfs, stream_nv, stream_avl, top_zfs, cleanup_fd, action_handlep)); } else { assert(DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) == DMU_COMPOUNDSTREAM); return (zfs_receive_package(hdl, infd, tosnap, flags, &drr, &zcksum, top_zfs, cleanup_fd, action_handlep)); } } /* * Restores a backup of tosnap from the file descriptor specified by infd. * Return 0 on total success, -2 if some things couldn't be * destroyed/renamed/promoted, -1 if some things couldn't be received. * (-1 will override -2, if -1 and the resumable flag was specified the * transfer can be resumed if the sending side supports it). */ int zfs_receive(libzfs_handle_t *hdl, const char *tosnap, nvlist_t *props, recvflags_t *flags, int infd, avl_tree_t *stream_avl) { char *top_zfs = NULL; int err; int cleanup_fd; uint64_t action_handle = 0; char *originsnap = NULL; if (props) { err = nvlist_lookup_string(props, "origin", &originsnap); if (err && err != ENOENT) return (err); } cleanup_fd = open(ZFS_DEV, O_RDWR|O_EXCL); VERIFY(cleanup_fd >= 0); err = zfs_receive_impl(hdl, tosnap, originsnap, flags, infd, NULL, NULL, stream_avl, &top_zfs, cleanup_fd, &action_handle); VERIFY(0 == close(cleanup_fd)); if (err == 0 && !flags->nomount && top_zfs) { zfs_handle_t *zhp; prop_changelist_t *clp; zhp = zfs_open(hdl, top_zfs, ZFS_TYPE_FILESYSTEM); if (zhp != NULL) { clp = changelist_gather(zhp, ZFS_PROP_MOUNTPOINT, CL_GATHER_MOUNT_ALWAYS, 0); zfs_close(zhp); if (clp != NULL) { /* mount and share received datasets */ err = changelist_postfix(clp); changelist_free(clp); } } if (zhp == NULL || clp == NULL || err) err = -1; } if (top_zfs) free(top_zfs); return (err); } Index: stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfeature_common.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfeature_common.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfeature_common.c (revision 290757) @@ -1,233 +1,248 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2011, 2015 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2013, Joyent, Inc. All rights reserved. * Copyright (c) 2014, Nexenta Systems, Inc. All rights reserved. */ #ifdef _KERNEL #include #else #include #include #endif #include #include #include #include "zfeature_common.h" /* * Set to disable all feature checks while opening pools, allowing pools with * unsupported features to be opened. Set for testing only. */ boolean_t zfeature_checks_disable = B_FALSE; zfeature_info_t spa_feature_table[SPA_FEATURES]; /* * Valid characters for feature guids. This list is mainly for aesthetic * purposes and could be expanded in the future. There are different allowed * characters in the guids reverse dns portion (before the colon) and its * short name (after the colon). */ static int valid_char(char c, boolean_t after_colon) { return ((c >= 'a' && c <= 'z') || (c >= '0' && c <= '9') || (after_colon && c == '_') || (!after_colon && (c == '.' || c == '-'))); } /* * Every feature guid must contain exactly one colon which separates a reverse * dns organization name from the feature's "short" name (e.g. * "com.company:feature_name"). */ boolean_t zfeature_is_valid_guid(const char *name) { int i; boolean_t has_colon = B_FALSE; i = 0; while (name[i] != '\0') { char c = name[i++]; if (c == ':') { if (has_colon) return (B_FALSE); has_colon = B_TRUE; continue; } if (!valid_char(c, has_colon)) return (B_FALSE); } return (has_colon); } boolean_t zfeature_is_supported(const char *guid) { if (zfeature_checks_disable) return (B_TRUE); for (spa_feature_t i = 0; i < SPA_FEATURES; i++) { zfeature_info_t *feature = &spa_feature_table[i]; if (strcmp(guid, feature->fi_guid) == 0) return (B_TRUE); } return (B_FALSE); } int zfeature_lookup_name(const char *name, spa_feature_t *res) { for (spa_feature_t i = 0; i < SPA_FEATURES; i++) { zfeature_info_t *feature = &spa_feature_table[i]; if (strcmp(name, feature->fi_uname) == 0) { if (res != NULL) *res = i; return (0); } } return (ENOENT); } boolean_t zfeature_depends_on(spa_feature_t fid, spa_feature_t check) { zfeature_info_t *feature = &spa_feature_table[fid]; for (int i = 0; feature->fi_depends[i] != SPA_FEATURE_NONE; i++) { if (feature->fi_depends[i] == check) return (B_TRUE); } return (B_FALSE); } static void zfeature_register(spa_feature_t fid, const char *guid, const char *name, const char *desc, zfeature_flags_t flags, const spa_feature_t *deps) { zfeature_info_t *feature = &spa_feature_table[fid]; static spa_feature_t nodeps[] = { SPA_FEATURE_NONE }; ASSERT(name != NULL); ASSERT(desc != NULL); ASSERT((flags & ZFEATURE_FLAG_READONLY_COMPAT) == 0 || (flags & ZFEATURE_FLAG_MOS) == 0); ASSERT3U(fid, <, SPA_FEATURES); ASSERT(zfeature_is_valid_guid(guid)); if (deps == NULL) deps = nodeps; feature->fi_feature = fid; feature->fi_guid = guid; feature->fi_uname = name; feature->fi_desc = desc; feature->fi_flags = flags; feature->fi_depends = deps; } void zpool_feature_init(void) { zfeature_register(SPA_FEATURE_ASYNC_DESTROY, "com.delphix:async_destroy", "async_destroy", "Destroy filesystems asynchronously.", ZFEATURE_FLAG_READONLY_COMPAT, NULL); zfeature_register(SPA_FEATURE_EMPTY_BPOBJ, "com.delphix:empty_bpobj", "empty_bpobj", "Snapshots use less space.", ZFEATURE_FLAG_READONLY_COMPAT, NULL); zfeature_register(SPA_FEATURE_LZ4_COMPRESS, "org.illumos:lz4_compress", "lz4_compress", "LZ4 compression algorithm support.", ZFEATURE_FLAG_ACTIVATE_ON_ENABLE, NULL); zfeature_register(SPA_FEATURE_MULTI_VDEV_CRASH_DUMP, "com.joyent:multi_vdev_crash_dump", "multi_vdev_crash_dump", "Crash dumps to multiple vdev pools.", 0, NULL); zfeature_register(SPA_FEATURE_SPACEMAP_HISTOGRAM, "com.delphix:spacemap_histogram", "spacemap_histogram", "Spacemaps maintain space histograms.", ZFEATURE_FLAG_READONLY_COMPAT, NULL); zfeature_register(SPA_FEATURE_ENABLED_TXG, "com.delphix:enabled_txg", "enabled_txg", "Record txg at which a feature is enabled", ZFEATURE_FLAG_READONLY_COMPAT, NULL); static spa_feature_t hole_birth_deps[] = { SPA_FEATURE_ENABLED_TXG, SPA_FEATURE_NONE }; zfeature_register(SPA_FEATURE_HOLE_BIRTH, "com.delphix:hole_birth", "hole_birth", "Retain hole birth txg for more precise zfs send", ZFEATURE_FLAG_MOS | ZFEATURE_FLAG_ACTIVATE_ON_ENABLE, hole_birth_deps); zfeature_register(SPA_FEATURE_EXTENSIBLE_DATASET, "com.delphix:extensible_dataset", "extensible_dataset", "Enhanced dataset functionality, used by other features.", 0, NULL); static const spa_feature_t bookmarks_deps[] = { SPA_FEATURE_EXTENSIBLE_DATASET, SPA_FEATURE_NONE }; zfeature_register(SPA_FEATURE_BOOKMARKS, "com.delphix:bookmarks", "bookmarks", "\"zfs bookmark\" command", ZFEATURE_FLAG_READONLY_COMPAT, bookmarks_deps); static const spa_feature_t filesystem_limits_deps[] = { SPA_FEATURE_EXTENSIBLE_DATASET, SPA_FEATURE_NONE }; zfeature_register(SPA_FEATURE_FS_SS_LIMIT, "com.joyent:filesystem_limits", "filesystem_limits", "Filesystem and snapshot limits.", ZFEATURE_FLAG_READONLY_COMPAT, filesystem_limits_deps); zfeature_register(SPA_FEATURE_EMBEDDED_DATA, "com.delphix:embedded_data", "embedded_data", "Blocks which compress very well use even less space.", ZFEATURE_FLAG_MOS | ZFEATURE_FLAG_ACTIVATE_ON_ENABLE, NULL); static const spa_feature_t large_blocks_deps[] = { SPA_FEATURE_EXTENSIBLE_DATASET, SPA_FEATURE_NONE }; zfeature_register(SPA_FEATURE_LARGE_BLOCKS, "org.open-zfs:large_blocks", "large_blocks", "Support for blocks larger than 128KB.", ZFEATURE_FLAG_PER_DATASET, large_blocks_deps); + +#ifdef illumos + zfeature_register(SPA_FEATURE_SHA512, + "org.illumos:sha512", "sha512", + "SHA-512/256 hash algorithm.", + ZFEATURE_FLAG_PER_DATASET, NULL); + zfeature_register(SPA_FEATURE_SKEIN, + "org.illumos:skein", "skein", + "Skein hash algorithm.", + ZFEATURE_FLAG_PER_DATASET, NULL); + zfeature_register(SPA_FEATURE_EDONR, + "org.illumos:edonr", "edonr", + "Edon-R hash algorithm.", + ZFEATURE_FLAG_PER_DATASET, NULL); +#endif } Index: stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfeature_common.h =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfeature_common.h (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfeature_common.h (revision 290757) @@ -1,98 +1,103 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2011, 2015 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2013, Joyent, Inc. All rights reserved. */ #ifndef _ZFEATURE_COMMON_H #define _ZFEATURE_COMMON_H #include #include #ifdef __cplusplus extern "C" { #endif struct zfeature_info; typedef enum spa_feature { SPA_FEATURE_NONE = -1, SPA_FEATURE_ASYNC_DESTROY, SPA_FEATURE_EMPTY_BPOBJ, SPA_FEATURE_LZ4_COMPRESS, SPA_FEATURE_MULTI_VDEV_CRASH_DUMP, SPA_FEATURE_SPACEMAP_HISTOGRAM, SPA_FEATURE_ENABLED_TXG, SPA_FEATURE_HOLE_BIRTH, SPA_FEATURE_EXTENSIBLE_DATASET, SPA_FEATURE_EMBEDDED_DATA, SPA_FEATURE_BOOKMARKS, SPA_FEATURE_FS_SS_LIMIT, SPA_FEATURE_LARGE_BLOCKS, +#ifdef illumos + SPA_FEATURE_SHA512, + SPA_FEATURE_SKEIN, + SPA_FEATURE_EDONR, +#endif SPA_FEATURES } spa_feature_t; #define SPA_FEATURE_DISABLED (-1ULL) typedef enum zfeature_flags { /* Can open pool readonly even if this feature is not supported. */ ZFEATURE_FLAG_READONLY_COMPAT = (1 << 0), /* Is this feature necessary to read the MOS? */ ZFEATURE_FLAG_MOS = (1 << 1), /* Activate this feature at the same time it is enabled. */ ZFEATURE_FLAG_ACTIVATE_ON_ENABLE = (1 << 2), /* Each dataset has a field set if it has ever used this feature. */ ZFEATURE_FLAG_PER_DATASET = (1 << 3) } zfeature_flags_t; typedef struct zfeature_info { spa_feature_t fi_feature; const char *fi_uname; /* User-facing feature name */ const char *fi_guid; /* On-disk feature identifier */ const char *fi_desc; /* Feature description */ zfeature_flags_t fi_flags; /* array of dependencies, terminated by SPA_FEATURE_NONE */ const spa_feature_t *fi_depends; } zfeature_info_t; typedef int (zfeature_func_t)(zfeature_info_t *, void *); #define ZFS_FEATURE_DEBUG extern zfeature_info_t spa_feature_table[SPA_FEATURES]; extern boolean_t zfeature_is_valid_guid(const char *); extern boolean_t zfeature_is_supported(const char *); extern int zfeature_lookup_name(const char *, spa_feature_t *); extern boolean_t zfeature_depends_on(spa_feature_t, spa_feature_t); extern void zpool_feature_init(void); #ifdef __cplusplus } #endif #endif /* _ZFEATURE_COMMON_H */ Index: stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfs_fletcher.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfs_fletcher.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfs_fletcher.c (revision 290757) @@ -1,246 +1,257 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright 2009 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. */ +/* + * Copyright 2013 Saso Kiselkov. All rights reserved. + */ /* * Fletcher Checksums * ------------------ * * ZFS's 2nd and 4th order Fletcher checksums are defined by the following * recurrence relations: * * a = a + f * i i-1 i-1 * * b = b + a * i i-1 i * * c = c + b (fletcher-4 only) * i i-1 i * * d = d + c (fletcher-4 only) * i i-1 i * * Where * a_0 = b_0 = c_0 = d_0 = 0 * and * f_0 .. f_(n-1) are the input data. * * Using standard techniques, these translate into the following series: * * __n_ __n_ * \ | \ | * a = > f b = > i * f * n /___| n - i n /___| n - i * i = 1 i = 1 * * * __n_ __n_ * \ | i*(i+1) \ | i*(i+1)*(i+2) * c = > ------- f d = > ------------- f * n /___| 2 n - i n /___| 6 n - i * i = 1 i = 1 * * For fletcher-2, the f_is are 64-bit, and [ab]_i are 64-bit accumulators. * Since the additions are done mod (2^64), errors in the high bits may not * be noticed. For this reason, fletcher-2 is deprecated. * * For fletcher-4, the f_is are 32-bit, and [abcd]_i are 64-bit accumulators. * A conservative estimate of how big the buffer can get before we overflow * can be estimated using f_i = 0xffffffff for all i: * * % bc * f=2^32-1;d=0; for (i = 1; d<2^64; i++) { d += f*i*(i+1)*(i+2)/6 }; (i-1)*4 * 2264 * quit * % * * So blocks of up to 2k will not overflow. Our largest block size is * 128k, which has 32k 4-byte words, so we can compute the largest possible * accumulators, then divide by 2^64 to figure the max amount of overflow: * * % bc * a=b=c=d=0; f=2^32-1; for (i=1; i<=32*1024; i++) { a+=f; b+=a; c+=b; d+=c } * a/2^64;b/2^64;c/2^64;d/2^64 * 0 * 0 * 1365 * 11186858 * quit * % * * So a and b cannot overflow. To make sure each bit of input has some * effect on the contents of c and d, we can look at what the factors of * the coefficients in the equations for c_n and d_n are. The number of 2s * in the factors determines the lowest set bit in the multiplier. Running * through the cases for n*(n+1)/2 reveals that the highest power of 2 is * 2^14, and for n*(n+1)*(n+2)/6 it is 2^15. So while some data may overflow * the 64-bit accumulators, every bit of every f_i effects every accumulator, * even for 128k blocks. * * If we wanted to make a stronger version of fletcher4 (fletcher4c?), * we could do our calculations mod (2^32 - 1) by adding in the carries * periodically, and store the number of carries in the top 32-bits. * * -------------------- * Checksum Performance * -------------------- * * There are two interesting components to checksum performance: cached and * uncached performance. With cached data, fletcher-2 is about four times * faster than fletcher-4. With uncached data, the performance difference is * negligible, since the cost of a cache fill dominates the processing time. * Even though fletcher-4 is slower than fletcher-2, it is still a pretty * efficient pass over the data. * * In normal operation, the data which is being checksummed is in a buffer * which has been filled either by: * * 1. a compression step, which will be mostly cached, or * 2. a bcopy() or copyin(), which will be uncached (because the * copy is cache-bypassing). * * For both cached and uncached data, both fletcher checksums are much faster * than sha-256, and slower than 'off', which doesn't touch the data at all. */ #include #include #include #include #include +/*ARGSUSED*/ void -fletcher_2_native(const void *buf, uint64_t size, zio_cksum_t *zcp) +fletcher_2_native(const void *buf, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp) { const uint64_t *ip = buf; const uint64_t *ipend = ip + (size / sizeof (uint64_t)); uint64_t a0, b0, a1, b1; for (a0 = b0 = a1 = b1 = 0; ip < ipend; ip += 2) { a0 += ip[0]; a1 += ip[1]; b0 += a0; b1 += a1; } ZIO_SET_CHECKSUM(zcp, a0, a1, b0, b1); } +/*ARGSUSED*/ void -fletcher_2_byteswap(const void *buf, uint64_t size, zio_cksum_t *zcp) +fletcher_2_byteswap(const void *buf, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp) { const uint64_t *ip = buf; const uint64_t *ipend = ip + (size / sizeof (uint64_t)); uint64_t a0, b0, a1, b1; for (a0 = b0 = a1 = b1 = 0; ip < ipend; ip += 2) { a0 += BSWAP_64(ip[0]); a1 += BSWAP_64(ip[1]); b0 += a0; b1 += a1; } ZIO_SET_CHECKSUM(zcp, a0, a1, b0, b1); } +/*ARGSUSED*/ void -fletcher_4_native(const void *buf, uint64_t size, zio_cksum_t *zcp) +fletcher_4_native(const void *buf, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp) { const uint32_t *ip = buf; const uint32_t *ipend = ip + (size / sizeof (uint32_t)); uint64_t a, b, c, d; for (a = b = c = d = 0; ip < ipend; ip++) { a += ip[0]; b += a; c += b; d += c; } ZIO_SET_CHECKSUM(zcp, a, b, c, d); } +/*ARGSUSED*/ void -fletcher_4_byteswap(const void *buf, uint64_t size, zio_cksum_t *zcp) +fletcher_4_byteswap(const void *buf, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp) { const uint32_t *ip = buf; const uint32_t *ipend = ip + (size / sizeof (uint32_t)); uint64_t a, b, c, d; for (a = b = c = d = 0; ip < ipend; ip++) { a += BSWAP_32(ip[0]); b += a; c += b; d += c; } ZIO_SET_CHECKSUM(zcp, a, b, c, d); } void fletcher_4_incremental_native(const void *buf, uint64_t size, zio_cksum_t *zcp) { const uint32_t *ip = buf; const uint32_t *ipend = ip + (size / sizeof (uint32_t)); uint64_t a, b, c, d; a = zcp->zc_word[0]; b = zcp->zc_word[1]; c = zcp->zc_word[2]; d = zcp->zc_word[3]; for (; ip < ipend; ip++) { a += ip[0]; b += a; c += b; d += c; } ZIO_SET_CHECKSUM(zcp, a, b, c, d); } void fletcher_4_incremental_byteswap(const void *buf, uint64_t size, zio_cksum_t *zcp) { const uint32_t *ip = buf; const uint32_t *ipend = ip + (size / sizeof (uint32_t)); uint64_t a, b, c, d; a = zcp->zc_word[0]; b = zcp->zc_word[1]; c = zcp->zc_word[2]; d = zcp->zc_word[3]; for (; ip < ipend; ip++) { a += BSWAP_32(ip[0]); b += a; c += b; d += c; } ZIO_SET_CHECKSUM(zcp, a, b, c, d); } Index: stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfs_fletcher.h =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfs_fletcher.h (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfs_fletcher.h (revision 290757) @@ -1,53 +1,54 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright 2009 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. */ +/* + * Copyright 2013 Saso Kiselkov. All rights reserved. + */ #ifndef _ZFS_FLETCHER_H #define _ZFS_FLETCHER_H #include #include #ifdef __cplusplus extern "C" { #endif /* * fletcher checksum functions */ -void fletcher_2_native(const void *, uint64_t, zio_cksum_t *); -void fletcher_2_byteswap(const void *, uint64_t, zio_cksum_t *); -void fletcher_4_native(const void *, uint64_t, zio_cksum_t *); -void fletcher_4_byteswap(const void *, uint64_t, zio_cksum_t *); -void fletcher_4_incremental_native(const void *, uint64_t, - zio_cksum_t *); -void fletcher_4_incremental_byteswap(const void *, uint64_t, - zio_cksum_t *); +void fletcher_2_native(const void *, uint64_t, const void *, zio_cksum_t *); +void fletcher_2_byteswap(const void *, uint64_t, const void *, zio_cksum_t *); +void fletcher_4_native(const void *, uint64_t, const void *, zio_cksum_t *); +void fletcher_4_byteswap(const void *, uint64_t, const void *, zio_cksum_t *); +void fletcher_4_incremental_native(const void *, uint64_t, zio_cksum_t *); +void fletcher_4_incremental_byteswap(const void *, uint64_t, zio_cksum_t *); #ifdef __cplusplus } #endif #endif /* _ZFS_FLETCHER_H */ Index: stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfs_prop.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfs_prop.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/common/zfs/zfs_prop.c (revision 290757) @@ -1,672 +1,687 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2014 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2013, Joyent, Inc. All rights reserved. */ /* Portions Copyright 2010 Robert Milkowski */ #include #include #include #include #include #include #include "zfs_prop.h" #include "zfs_deleg.h" #if defined(_KERNEL) #include #else #include #include #include #endif static zprop_desc_t zfs_prop_table[ZFS_NUM_PROPS]; /* Note this is indexed by zfs_userquota_prop_t, keep the order the same */ const char *zfs_userquota_prop_prefixes[] = { "userused@", "userquota@", "groupused@", "groupquota@" }; zprop_desc_t * zfs_prop_get_table(void) { return (zfs_prop_table); } void zfs_prop_init(void) { static zprop_index_t checksum_table[] = { { "on", ZIO_CHECKSUM_ON }, { "off", ZIO_CHECKSUM_OFF }, { "fletcher2", ZIO_CHECKSUM_FLETCHER_2 }, { "fletcher4", ZIO_CHECKSUM_FLETCHER_4 }, { "sha256", ZIO_CHECKSUM_SHA256 }, { "noparity", ZIO_CHECKSUM_NOPARITY }, +#ifdef illumos + { "sha512", ZIO_CHECKSUM_SHA512 }, + { "skein", ZIO_CHECKSUM_SKEIN }, + { "edonr", ZIO_CHECKSUM_EDONR }, +#endif { NULL } }; static zprop_index_t dedup_table[] = { { "on", ZIO_CHECKSUM_ON }, { "off", ZIO_CHECKSUM_OFF }, { "verify", ZIO_CHECKSUM_ON | ZIO_CHECKSUM_VERIFY }, { "sha256", ZIO_CHECKSUM_SHA256 }, { "sha256,verify", ZIO_CHECKSUM_SHA256 | ZIO_CHECKSUM_VERIFY }, +#ifdef illumos + { "sha512", ZIO_CHECKSUM_SHA512 }, + { "sha512,verify", + ZIO_CHECKSUM_SHA512 | ZIO_CHECKSUM_VERIFY }, + { "skein", ZIO_CHECKSUM_SKEIN }, + { "skein,verify", + ZIO_CHECKSUM_SKEIN | ZIO_CHECKSUM_VERIFY }, + { "edonr,verify", + ZIO_CHECKSUM_EDONR | ZIO_CHECKSUM_VERIFY }, +#endif { NULL } }; static zprop_index_t compress_table[] = { { "on", ZIO_COMPRESS_ON }, { "off", ZIO_COMPRESS_OFF }, { "lzjb", ZIO_COMPRESS_LZJB }, { "gzip", ZIO_COMPRESS_GZIP_6 }, /* gzip default */ { "gzip-1", ZIO_COMPRESS_GZIP_1 }, { "gzip-2", ZIO_COMPRESS_GZIP_2 }, { "gzip-3", ZIO_COMPRESS_GZIP_3 }, { "gzip-4", ZIO_COMPRESS_GZIP_4 }, { "gzip-5", ZIO_COMPRESS_GZIP_5 }, { "gzip-6", ZIO_COMPRESS_GZIP_6 }, { "gzip-7", ZIO_COMPRESS_GZIP_7 }, { "gzip-8", ZIO_COMPRESS_GZIP_8 }, { "gzip-9", ZIO_COMPRESS_GZIP_9 }, { "zle", ZIO_COMPRESS_ZLE }, { "lz4", ZIO_COMPRESS_LZ4 }, { NULL } }; static zprop_index_t snapdir_table[] = { { "hidden", ZFS_SNAPDIR_HIDDEN }, { "visible", ZFS_SNAPDIR_VISIBLE }, { NULL } }; static zprop_index_t acl_mode_table[] = { { "discard", ZFS_ACL_DISCARD }, { "groupmask", ZFS_ACL_GROUPMASK }, { "passthrough", ZFS_ACL_PASSTHROUGH }, { "restricted", ZFS_ACL_RESTRICTED }, { NULL } }; static zprop_index_t acl_inherit_table[] = { { "discard", ZFS_ACL_DISCARD }, { "noallow", ZFS_ACL_NOALLOW }, { "restricted", ZFS_ACL_RESTRICTED }, { "passthrough", ZFS_ACL_PASSTHROUGH }, { "secure", ZFS_ACL_RESTRICTED }, /* bkwrd compatability */ { "passthrough-x", ZFS_ACL_PASSTHROUGH_X }, { NULL } }; static zprop_index_t case_table[] = { { "sensitive", ZFS_CASE_SENSITIVE }, { "insensitive", ZFS_CASE_INSENSITIVE }, { "mixed", ZFS_CASE_MIXED }, { NULL } }; static zprop_index_t copies_table[] = { { "1", 1 }, { "2", 2 }, { "3", 3 }, { NULL } }; /* * Use the unique flags we have to send to u8_strcmp() and/or * u8_textprep() to represent the various normalization property * values. */ static zprop_index_t normalize_table[] = { { "none", 0 }, { "formD", U8_TEXTPREP_NFD }, { "formKC", U8_TEXTPREP_NFKC }, { "formC", U8_TEXTPREP_NFC }, { "formKD", U8_TEXTPREP_NFKD }, { NULL } }; static zprop_index_t version_table[] = { { "1", 1 }, { "2", 2 }, { "3", 3 }, { "4", 4 }, { "5", 5 }, { "current", ZPL_VERSION }, { NULL } }; static zprop_index_t boolean_table[] = { { "off", 0 }, { "on", 1 }, { NULL } }; static zprop_index_t logbias_table[] = { { "latency", ZFS_LOGBIAS_LATENCY }, { "throughput", ZFS_LOGBIAS_THROUGHPUT }, { NULL } }; static zprop_index_t canmount_table[] = { { "off", ZFS_CANMOUNT_OFF }, { "on", ZFS_CANMOUNT_ON }, { "noauto", ZFS_CANMOUNT_NOAUTO }, { NULL } }; static zprop_index_t cache_table[] = { { "none", ZFS_CACHE_NONE }, { "metadata", ZFS_CACHE_METADATA }, { "all", ZFS_CACHE_ALL }, { NULL } }; static zprop_index_t sync_table[] = { { "standard", ZFS_SYNC_STANDARD }, { "always", ZFS_SYNC_ALWAYS }, { "disabled", ZFS_SYNC_DISABLED }, { NULL } }; static zprop_index_t volmode_table[] = { { "default", ZFS_VOLMODE_DEFAULT }, { "geom", ZFS_VOLMODE_GEOM }, { "dev", ZFS_VOLMODE_DEV }, { "none", ZFS_VOLMODE_NONE }, { NULL } }; static zprop_index_t redundant_metadata_table[] = { { "all", ZFS_REDUNDANT_METADATA_ALL }, { "most", ZFS_REDUNDANT_METADATA_MOST }, { NULL } }; /* inherit index properties */ zprop_register_index(ZFS_PROP_REDUNDANT_METADATA, "redundant_metadata", ZFS_REDUNDANT_METADATA_ALL, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "all | most", "REDUND_MD", redundant_metadata_table); zprop_register_index(ZFS_PROP_SYNC, "sync", ZFS_SYNC_STANDARD, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "standard | always | disabled", "SYNC", sync_table); zprop_register_index(ZFS_PROP_CHECKSUM, "checksum", ZIO_CHECKSUM_DEFAULT, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, - "on | off | fletcher2 | fletcher4 | sha256", "CHECKSUM", - checksum_table); + "on | off | fletcher2 | fletcher4 | sha256 | sha512 | " + "skein | edonr", "CHECKSUM", checksum_table); zprop_register_index(ZFS_PROP_DEDUP, "dedup", ZIO_CHECKSUM_OFF, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, - "on | off | verify | sha256[,verify]", "DEDUP", - dedup_table); + "on | off | verify | sha256[,verify], sha512[,verify], " + "skein[,verify], edonr,verify", "DEDUP", dedup_table); zprop_register_index(ZFS_PROP_COMPRESSION, "compression", ZIO_COMPRESS_DEFAULT, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "on | off | lzjb | gzip | gzip-[1-9] | zle | lz4", "COMPRESS", compress_table); zprop_register_index(ZFS_PROP_SNAPDIR, "snapdir", ZFS_SNAPDIR_HIDDEN, PROP_INHERIT, ZFS_TYPE_FILESYSTEM, "hidden | visible", "SNAPDIR", snapdir_table); zprop_register_index(ZFS_PROP_ACLMODE, "aclmode", ZFS_ACL_DISCARD, PROP_INHERIT, ZFS_TYPE_FILESYSTEM, "discard | groupmask | passthrough | restricted", "ACLMODE", acl_mode_table); zprop_register_index(ZFS_PROP_ACLINHERIT, "aclinherit", ZFS_ACL_RESTRICTED, PROP_INHERIT, ZFS_TYPE_FILESYSTEM, "discard | noallow | restricted | passthrough | passthrough-x", "ACLINHERIT", acl_inherit_table); zprop_register_index(ZFS_PROP_COPIES, "copies", 1, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "1 | 2 | 3", "COPIES", copies_table); zprop_register_index(ZFS_PROP_PRIMARYCACHE, "primarycache", ZFS_CACHE_ALL, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT | ZFS_TYPE_VOLUME, "all | none | metadata", "PRIMARYCACHE", cache_table); zprop_register_index(ZFS_PROP_SECONDARYCACHE, "secondarycache", ZFS_CACHE_ALL, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT | ZFS_TYPE_VOLUME, "all | none | metadata", "SECONDARYCACHE", cache_table); zprop_register_index(ZFS_PROP_LOGBIAS, "logbias", ZFS_LOGBIAS_LATENCY, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "latency | throughput", "LOGBIAS", logbias_table); zprop_register_index(ZFS_PROP_VOLMODE, "volmode", ZFS_VOLMODE_DEFAULT, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT | ZFS_TYPE_VOLUME, "default | geom | dev | none", "VOLMODE", volmode_table); /* inherit index (boolean) properties */ zprop_register_index(ZFS_PROP_ATIME, "atime", 1, PROP_INHERIT, ZFS_TYPE_FILESYSTEM, "on | off", "ATIME", boolean_table); zprop_register_index(ZFS_PROP_DEVICES, "devices", 1, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "on | off", "DEVICES", boolean_table); zprop_register_index(ZFS_PROP_EXEC, "exec", 1, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "on | off", "EXEC", boolean_table); zprop_register_index(ZFS_PROP_SETUID, "setuid", 1, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "on | off", "SETUID", boolean_table); zprop_register_index(ZFS_PROP_READONLY, "readonly", 0, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "on | off", "RDONLY", boolean_table); zprop_register_index(ZFS_PROP_ZONED, "jailed", 0, PROP_INHERIT, ZFS_TYPE_FILESYSTEM, "on | off", "JAILED", boolean_table); zprop_register_index(ZFS_PROP_XATTR, "xattr", 1, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "on | off", "XATTR", boolean_table); zprop_register_index(ZFS_PROP_VSCAN, "vscan", 0, PROP_INHERIT, ZFS_TYPE_FILESYSTEM, "on | off", "VSCAN", boolean_table); zprop_register_index(ZFS_PROP_NBMAND, "nbmand", 0, PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "on | off", "NBMAND", boolean_table); /* default index properties */ zprop_register_index(ZFS_PROP_VERSION, "version", 0, PROP_DEFAULT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "1 | 2 | 3 | 4 | 5 | current", "VERSION", version_table); zprop_register_index(ZFS_PROP_CANMOUNT, "canmount", ZFS_CANMOUNT_ON, PROP_DEFAULT, ZFS_TYPE_FILESYSTEM, "on | off | noauto", "CANMOUNT", canmount_table); /* readonly index (boolean) properties */ zprop_register_index(ZFS_PROP_MOUNTED, "mounted", 0, PROP_READONLY, ZFS_TYPE_FILESYSTEM, "yes | no", "MOUNTED", boolean_table); zprop_register_index(ZFS_PROP_DEFER_DESTROY, "defer_destroy", 0, PROP_READONLY, ZFS_TYPE_SNAPSHOT, "yes | no", "DEFER_DESTROY", boolean_table); /* set once index properties */ zprop_register_index(ZFS_PROP_NORMALIZE, "normalization", 0, PROP_ONETIME, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "none | formC | formD | formKC | formKD", "NORMALIZATION", normalize_table); zprop_register_index(ZFS_PROP_CASE, "casesensitivity", ZFS_CASE_SENSITIVE, PROP_ONETIME, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "sensitive | insensitive | mixed", "CASE", case_table); /* set once index (boolean) properties */ zprop_register_index(ZFS_PROP_UTF8ONLY, "utf8only", 0, PROP_ONETIME, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "on | off", "UTF8ONLY", boolean_table); /* string properties */ zprop_register_string(ZFS_PROP_ORIGIN, "origin", NULL, PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "", "ORIGIN"); zprop_register_string(ZFS_PROP_CLONES, "clones", NULL, PROP_READONLY, ZFS_TYPE_SNAPSHOT, "[,...]", "CLONES"); zprop_register_string(ZFS_PROP_MOUNTPOINT, "mountpoint", "/", PROP_INHERIT, ZFS_TYPE_FILESYSTEM, " | legacy | none", "MOUNTPOINT"); zprop_register_string(ZFS_PROP_SHARENFS, "sharenfs", "off", PROP_INHERIT, ZFS_TYPE_FILESYSTEM, "on | off | share(1M) options", "SHARENFS"); zprop_register_string(ZFS_PROP_TYPE, "type", NULL, PROP_READONLY, ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "filesystem | volume | snapshot | bookmark", "TYPE"); zprop_register_string(ZFS_PROP_SHARESMB, "sharesmb", "off", PROP_INHERIT, ZFS_TYPE_FILESYSTEM, "on | off | sharemgr(1M) options", "SHARESMB"); zprop_register_string(ZFS_PROP_MLSLABEL, "mlslabel", ZFS_MLSLABEL_DEFAULT, PROP_INHERIT, ZFS_TYPE_DATASET, "", "MLSLABEL"); zprop_register_string(ZFS_PROP_RECEIVE_RESUME_TOKEN, "receive_resume_token", NULL, PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "", "RESUMETOK"); /* readonly number properties */ zprop_register_number(ZFS_PROP_USED, "used", 0, PROP_READONLY, ZFS_TYPE_DATASET, "", "USED"); zprop_register_number(ZFS_PROP_AVAILABLE, "available", 0, PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "", "AVAIL"); zprop_register_number(ZFS_PROP_REFERENCED, "referenced", 0, PROP_READONLY, ZFS_TYPE_DATASET, "", "REFER"); zprop_register_number(ZFS_PROP_COMPRESSRATIO, "compressratio", 0, PROP_READONLY, ZFS_TYPE_DATASET, "<1.00x or higher if compressed>", "RATIO"); zprop_register_number(ZFS_PROP_REFRATIO, "refcompressratio", 0, PROP_READONLY, ZFS_TYPE_DATASET, "<1.00x or higher if compressed>", "REFRATIO"); zprop_register_number(ZFS_PROP_VOLBLOCKSIZE, "volblocksize", ZVOL_DEFAULT_BLOCKSIZE, PROP_ONETIME, ZFS_TYPE_VOLUME, "512 to 128k, power of 2", "VOLBLOCK"); zprop_register_number(ZFS_PROP_USEDSNAP, "usedbysnapshots", 0, PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "", "USEDSNAP"); zprop_register_number(ZFS_PROP_USEDDS, "usedbydataset", 0, PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "", "USEDDS"); zprop_register_number(ZFS_PROP_USEDCHILD, "usedbychildren", 0, PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "", "USEDCHILD"); zprop_register_number(ZFS_PROP_USEDREFRESERV, "usedbyrefreservation", 0, PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "", "USEDREFRESERV"); zprop_register_number(ZFS_PROP_USERREFS, "userrefs", 0, PROP_READONLY, ZFS_TYPE_SNAPSHOT, "", "USERREFS"); zprop_register_number(ZFS_PROP_WRITTEN, "written", 0, PROP_READONLY, ZFS_TYPE_DATASET, "", "WRITTEN"); zprop_register_number(ZFS_PROP_LOGICALUSED, "logicalused", 0, PROP_READONLY, ZFS_TYPE_DATASET, "", "LUSED"); zprop_register_number(ZFS_PROP_LOGICALREFERENCED, "logicalreferenced", 0, PROP_READONLY, ZFS_TYPE_DATASET, "", "LREFER"); /* default number properties */ zprop_register_number(ZFS_PROP_QUOTA, "quota", 0, PROP_DEFAULT, ZFS_TYPE_FILESYSTEM, " | none", "QUOTA"); zprop_register_number(ZFS_PROP_RESERVATION, "reservation", 0, PROP_DEFAULT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, " | none", "RESERV"); zprop_register_number(ZFS_PROP_VOLSIZE, "volsize", 0, PROP_DEFAULT, ZFS_TYPE_VOLUME, "", "VOLSIZE"); zprop_register_number(ZFS_PROP_REFQUOTA, "refquota", 0, PROP_DEFAULT, ZFS_TYPE_FILESYSTEM, " | none", "REFQUOTA"); zprop_register_number(ZFS_PROP_REFRESERVATION, "refreservation", 0, PROP_DEFAULT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, " | none", "REFRESERV"); zprop_register_number(ZFS_PROP_FILESYSTEM_LIMIT, "filesystem_limit", UINT64_MAX, PROP_DEFAULT, ZFS_TYPE_FILESYSTEM, " | none", "FSLIMIT"); zprop_register_number(ZFS_PROP_SNAPSHOT_LIMIT, "snapshot_limit", UINT64_MAX, PROP_DEFAULT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, " | none", "SSLIMIT"); zprop_register_number(ZFS_PROP_FILESYSTEM_COUNT, "filesystem_count", UINT64_MAX, PROP_DEFAULT, ZFS_TYPE_FILESYSTEM, "", "FSCOUNT"); zprop_register_number(ZFS_PROP_SNAPSHOT_COUNT, "snapshot_count", UINT64_MAX, PROP_DEFAULT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "", "SSCOUNT"); /* inherit number properties */ zprop_register_number(ZFS_PROP_RECORDSIZE, "recordsize", SPA_OLD_MAXBLOCKSIZE, PROP_INHERIT, ZFS_TYPE_FILESYSTEM, "512 to 1M, power of 2", "RECSIZE"); /* hidden properties */ zprop_register_hidden(ZFS_PROP_CREATETXG, "createtxg", PROP_TYPE_NUMBER, PROP_READONLY, ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "CREATETXG"); zprop_register_hidden(ZFS_PROP_NUMCLONES, "numclones", PROP_TYPE_NUMBER, PROP_READONLY, ZFS_TYPE_SNAPSHOT, "NUMCLONES"); zprop_register_hidden(ZFS_PROP_NAME, "name", PROP_TYPE_STRING, PROP_READONLY, ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "NAME"); zprop_register_hidden(ZFS_PROP_ISCSIOPTIONS, "iscsioptions", PROP_TYPE_STRING, PROP_INHERIT, ZFS_TYPE_VOLUME, "ISCSIOPTIONS"); zprop_register_hidden(ZFS_PROP_STMF_SHAREINFO, "stmf_sbd_lu", PROP_TYPE_STRING, PROP_INHERIT, ZFS_TYPE_VOLUME, "STMF_SBD_LU"); zprop_register_hidden(ZFS_PROP_GUID, "guid", PROP_TYPE_NUMBER, PROP_READONLY, ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "GUID"); zprop_register_hidden(ZFS_PROP_USERACCOUNTING, "useraccounting", PROP_TYPE_NUMBER, PROP_READONLY, ZFS_TYPE_DATASET, "USERACCOUNTING"); zprop_register_hidden(ZFS_PROP_UNIQUE, "unique", PROP_TYPE_NUMBER, PROP_READONLY, ZFS_TYPE_DATASET, "UNIQUE"); zprop_register_hidden(ZFS_PROP_OBJSETID, "objsetid", PROP_TYPE_NUMBER, PROP_READONLY, ZFS_TYPE_DATASET, "OBJSETID"); zprop_register_hidden(ZFS_PROP_INCONSISTENT, "inconsistent", PROP_TYPE_NUMBER, PROP_READONLY, ZFS_TYPE_DATASET, "INCONSISTENT"); zprop_register_hidden(ZFS_PROP_PREV_SNAP, "prevsnap", PROP_TYPE_STRING, PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "PREVSNAP"); /* oddball properties */ zprop_register_impl(ZFS_PROP_CREATION, "creation", PROP_TYPE_NUMBER, 0, NULL, PROP_READONLY, ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "", "CREATION", B_FALSE, B_TRUE, NULL); } boolean_t zfs_prop_delegatable(zfs_prop_t prop) { zprop_desc_t *pd = &zfs_prop_table[prop]; /* The mlslabel property is never delegatable. */ if (prop == ZFS_PROP_MLSLABEL) return (B_FALSE); return (pd->pd_attr != PROP_READONLY); } /* * Given a zfs dataset property name, returns the corresponding property ID. */ zfs_prop_t zfs_name_to_prop(const char *propname) { return (zprop_name_to_prop(propname, ZFS_TYPE_DATASET)); } /* * For user property names, we allow all lowercase alphanumeric characters, plus * a few useful punctuation characters. */ static int valid_char(char c) { return ((c >= 'a' && c <= 'z') || (c >= '0' && c <= '9') || c == '-' || c == '_' || c == '.' || c == ':'); } /* * Returns true if this is a valid user-defined property (one with a ':'). */ boolean_t zfs_prop_user(const char *name) { int i; char c; boolean_t foundsep = B_FALSE; for (i = 0; i < strlen(name); i++) { c = name[i]; if (!valid_char(c)) return (B_FALSE); if (c == ':') foundsep = B_TRUE; } if (!foundsep) return (B_FALSE); return (B_TRUE); } /* * Returns true if this is a valid userspace-type property (one with a '@'). * Note that after the @, any character is valid (eg, another @, for SID * user@domain). */ boolean_t zfs_prop_userquota(const char *name) { zfs_userquota_prop_t prop; for (prop = 0; prop < ZFS_NUM_USERQUOTA_PROPS; prop++) { if (strncmp(name, zfs_userquota_prop_prefixes[prop], strlen(zfs_userquota_prop_prefixes[prop])) == 0) { return (B_TRUE); } } return (B_FALSE); } /* * Returns true if this is a valid written@ property. * Note that after the @, any character is valid (eg, another @, for * written@pool/fs@origin). */ boolean_t zfs_prop_written(const char *name) { static const char *prefix = "written@"; return (strncmp(name, prefix, strlen(prefix)) == 0); } /* * Tables of index types, plus functions to convert between the user view * (strings) and internal representation (uint64_t). */ int zfs_prop_string_to_index(zfs_prop_t prop, const char *string, uint64_t *index) { return (zprop_string_to_index(prop, string, index, ZFS_TYPE_DATASET)); } int zfs_prop_index_to_string(zfs_prop_t prop, uint64_t index, const char **string) { return (zprop_index_to_string(prop, index, string, ZFS_TYPE_DATASET)); } uint64_t zfs_prop_random_value(zfs_prop_t prop, uint64_t seed) { return (zprop_random_value(prop, seed, ZFS_TYPE_DATASET)); } /* * Returns TRUE if the property applies to any of the given dataset types. */ boolean_t zfs_prop_valid_for_type(int prop, zfs_type_t types) { return (zprop_valid_for_type(prop, types)); } zprop_type_t zfs_prop_get_type(zfs_prop_t prop) { return (zfs_prop_table[prop].pd_proptype); } /* * Returns TRUE if the property is readonly. */ boolean_t zfs_prop_readonly(zfs_prop_t prop) { return (zfs_prop_table[prop].pd_attr == PROP_READONLY || zfs_prop_table[prop].pd_attr == PROP_ONETIME); } /* * Returns TRUE if the property is only allowed to be set once. */ boolean_t zfs_prop_setonce(zfs_prop_t prop) { return (zfs_prop_table[prop].pd_attr == PROP_ONETIME); } const char * zfs_prop_default_string(zfs_prop_t prop) { return (zfs_prop_table[prop].pd_strdefault); } uint64_t zfs_prop_default_numeric(zfs_prop_t prop) { return (zfs_prop_table[prop].pd_numdefault); } /* * Given a dataset property ID, returns the corresponding name. * Assuming the zfs dataset property ID is valid. */ const char * zfs_prop_to_name(zfs_prop_t prop) { return (zfs_prop_table[prop].pd_name); } /* * Returns TRUE if the property is inheritable. */ boolean_t zfs_prop_inheritable(zfs_prop_t prop) { return (zfs_prop_table[prop].pd_attr == PROP_INHERIT || zfs_prop_table[prop].pd_attr == PROP_ONETIME); } #ifndef _KERNEL /* * Returns a string describing the set of acceptable values for the given * zfs property, or NULL if it cannot be set. */ const char * zfs_prop_values(zfs_prop_t prop) { return (zfs_prop_table[prop].pd_values); } /* * Returns TRUE if this property is a string type. Note that index types * (compression, checksum) are treated as strings in userland, even though they * are stored numerically on disk. */ int zfs_prop_is_string(zfs_prop_t prop) { return (zfs_prop_table[prop].pd_proptype == PROP_TYPE_STRING || zfs_prop_table[prop].pd_proptype == PROP_TYPE_INDEX); } /* * Returns the column header for the given property. Used only in * 'zfs list -o', but centralized here with the other property information. */ const char * zfs_prop_column_name(zfs_prop_t prop) { return (zfs_prop_table[prop].pd_colname); } /* * Returns whether the given property should be displayed right-justified for * 'zfs list'. */ boolean_t zfs_prop_align_right(zfs_prop_t prop) { return (zfs_prop_table[prop].pd_rightalign); } #endif Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/arc.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/arc.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/arc.c (revision 290757) @@ -1,7010 +1,7010 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, Joyent, Inc. All rights reserved. * Copyright (c) 2011, 2015 by Delphix. All rights reserved. * Copyright (c) 2014 by Saso Kiselkov. All rights reserved. * Copyright 2015 Nexenta Systems, Inc. All rights reserved. */ /* * DVA-based Adjustable Replacement Cache * * While much of the theory of operation used here is * based on the self-tuning, low overhead replacement cache * presented by Megiddo and Modha at FAST 2003, there are some * significant differences: * * 1. The Megiddo and Modha model assumes any page is evictable. * Pages in its cache cannot be "locked" into memory. This makes * the eviction algorithm simple: evict the last page in the list. * This also make the performance characteristics easy to reason * about. Our cache is not so simple. At any given moment, some * subset of the blocks in the cache are un-evictable because we * have handed out a reference to them. Blocks are only evictable * when there are no external references active. This makes * eviction far more problematic: we choose to evict the evictable * blocks that are the "lowest" in the list. * * There are times when it is not possible to evict the requested * space. In these circumstances we are unable to adjust the cache * size. To prevent the cache growing unbounded at these times we * implement a "cache throttle" that slows the flow of new data * into the cache until we can make space available. * * 2. The Megiddo and Modha model assumes a fixed cache size. * Pages are evicted when the cache is full and there is a cache * miss. Our model has a variable sized cache. It grows with * high use, but also tries to react to memory pressure from the * operating system: decreasing its size when system memory is * tight. * * 3. The Megiddo and Modha model assumes a fixed page size. All * elements of the cache are therefore exactly the same size. So * when adjusting the cache size following a cache miss, its simply * a matter of choosing a single page to evict. In our model, we * have variable sized cache blocks (rangeing from 512 bytes to * 128K bytes). We therefore choose a set of blocks to evict to make * space for a cache miss that approximates as closely as possible * the space used by the new block. * * See also: "ARC: A Self-Tuning, Low Overhead Replacement Cache" * by N. Megiddo & D. Modha, FAST 2003 */ /* * The locking model: * * A new reference to a cache buffer can be obtained in two * ways: 1) via a hash table lookup using the DVA as a key, * or 2) via one of the ARC lists. The arc_read() interface * uses method 1, while the internal arc algorithms for * adjusting the cache use method 2. We therefore provide two * types of locks: 1) the hash table lock array, and 2) the * arc list locks. * * Buffers do not have their own mutexes, rather they rely on the * hash table mutexes for the bulk of their protection (i.e. most * fields in the arc_buf_hdr_t are protected by these mutexes). * * buf_hash_find() returns the appropriate mutex (held) when it * locates the requested buffer in the hash table. It returns * NULL for the mutex if the buffer was not in the table. * * buf_hash_remove() expects the appropriate hash mutex to be * already held before it is invoked. * * Each arc state also has a mutex which is used to protect the * buffer list associated with the state. When attempting to * obtain a hash table lock while holding an arc list lock you * must use: mutex_tryenter() to avoid deadlock. Also note that * the active state mutex must be held before the ghost state mutex. * * Arc buffers may have an associated eviction callback function. * This function will be invoked prior to removing the buffer (e.g. * in arc_do_user_evicts()). Note however that the data associated * with the buffer may be evicted prior to the callback. The callback * must be made with *no locks held* (to prevent deadlock). Additionally, * the users of callbacks must ensure that their private data is * protected from simultaneous callbacks from arc_clear_callback() * and arc_do_user_evicts(). * * Note that the majority of the performance stats are manipulated * with atomic operations. * * The L2ARC uses the l2ad_mtx on each vdev for the following: * * - L2ARC buflist creation * - L2ARC buflist eviction * - L2ARC write completion, which walks L2ARC buflists * - ARC header destruction, as it removes from L2ARC buflists * - ARC header release, as it removes from L2ARC buflists */ #include #include #include #include #include #include #include #include #include #include #ifdef _KERNEL #include #endif #include #include #include #include #include #include #include #ifdef illumos #ifndef _KERNEL /* set with ZFS_DEBUG=watch, to enable watchpoints on frozen buffers */ boolean_t arc_watch = B_FALSE; int arc_procfd; #endif #endif /* illumos */ static kmutex_t arc_reclaim_lock; static kcondvar_t arc_reclaim_thread_cv; static boolean_t arc_reclaim_thread_exit; static kcondvar_t arc_reclaim_waiters_cv; static kmutex_t arc_user_evicts_lock; static kcondvar_t arc_user_evicts_cv; static boolean_t arc_user_evicts_thread_exit; uint_t arc_reduce_dnlc_percent = 3; /* * The number of headers to evict in arc_evict_state_impl() before * dropping the sublist lock and evicting from another sublist. A lower * value means we're more likely to evict the "correct" header (i.e. the * oldest header in the arc state), but comes with higher overhead * (i.e. more invocations of arc_evict_state_impl()). */ int zfs_arc_evict_batch_limit = 10; /* * The number of sublists used for each of the arc state lists. If this * is not set to a suitable value by the user, it will be configured to * the number of CPUs on the system in arc_init(). */ int zfs_arc_num_sublists_per_state = 0; /* number of seconds before growing cache again */ static int arc_grow_retry = 60; /* shift of arc_c for calculating overflow limit in arc_get_data_buf */ int zfs_arc_overflow_shift = 8; /* shift of arc_c for calculating both min and max arc_p */ static int arc_p_min_shift = 4; /* log2(fraction of arc to reclaim) */ static int arc_shrink_shift = 7; /* * log2(fraction of ARC which must be free to allow growing). * I.e. If there is less than arc_c >> arc_no_grow_shift free memory, * when reading a new block into the ARC, we will evict an equal-sized block * from the ARC. * * This must be less than arc_shrink_shift, so that when we shrink the ARC, * we will still not allow it to grow. */ int arc_no_grow_shift = 5; /* * minimum lifespan of a prefetch block in clock ticks * (initialized in arc_init()) */ static int arc_min_prefetch_lifespan; /* * If this percent of memory is free, don't throttle. */ int arc_lotsfree_percent = 10; static int arc_dead; extern boolean_t zfs_prefetch_disable; /* * The arc has filled available memory and has now warmed up. */ static boolean_t arc_warm; /* * These tunables are for performance analysis. */ uint64_t zfs_arc_max; uint64_t zfs_arc_min; uint64_t zfs_arc_meta_limit = 0; uint64_t zfs_arc_meta_min = 0; int zfs_arc_grow_retry = 0; int zfs_arc_shrink_shift = 0; int zfs_arc_p_min_shift = 0; int zfs_disable_dup_eviction = 0; uint64_t zfs_arc_average_blocksize = 8 * 1024; /* 8KB */ u_int zfs_arc_free_target = 0; static int sysctl_vfs_zfs_arc_free_target(SYSCTL_HANDLER_ARGS); static int sysctl_vfs_zfs_arc_meta_limit(SYSCTL_HANDLER_ARGS); #ifdef _KERNEL static void arc_free_target_init(void *unused __unused) { zfs_arc_free_target = vm_pageout_wakeup_thresh; } SYSINIT(arc_free_target_init, SI_SUB_KTHREAD_PAGE, SI_ORDER_ANY, arc_free_target_init, NULL); TUNABLE_QUAD("vfs.zfs.arc_max", &zfs_arc_max); TUNABLE_QUAD("vfs.zfs.arc_min", &zfs_arc_min); TUNABLE_QUAD("vfs.zfs.arc_meta_limit", &zfs_arc_meta_limit); TUNABLE_QUAD("vfs.zfs.arc_meta_min", &zfs_arc_meta_min); TUNABLE_QUAD("vfs.zfs.arc_average_blocksize", &zfs_arc_average_blocksize); TUNABLE_INT("vfs.zfs.arc_shrink_shift", &zfs_arc_shrink_shift); SYSCTL_DECL(_vfs_zfs); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, arc_max, CTLFLAG_RDTUN, &zfs_arc_max, 0, "Maximum ARC size"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, arc_min, CTLFLAG_RDTUN, &zfs_arc_min, 0, "Minimum ARC size"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, arc_average_blocksize, CTLFLAG_RDTUN, &zfs_arc_average_blocksize, 0, "ARC average blocksize"); SYSCTL_INT(_vfs_zfs, OID_AUTO, arc_shrink_shift, CTLFLAG_RW, &arc_shrink_shift, 0, "log2(fraction of arc to reclaim)"); /* * We don't have a tunable for arc_free_target due to the dependency on * pagedaemon initialisation. */ SYSCTL_PROC(_vfs_zfs, OID_AUTO, arc_free_target, CTLTYPE_UINT | CTLFLAG_MPSAFE | CTLFLAG_RW, 0, sizeof(u_int), sysctl_vfs_zfs_arc_free_target, "IU", "Desired number of free pages below which ARC triggers reclaim"); static int sysctl_vfs_zfs_arc_free_target(SYSCTL_HANDLER_ARGS) { u_int val; int err; val = zfs_arc_free_target; err = sysctl_handle_int(oidp, &val, 0, req); if (err != 0 || req->newptr == NULL) return (err); if (val < minfree) return (EINVAL); if (val > cnt.v_page_count) return (EINVAL); zfs_arc_free_target = val; return (0); } /* * Must be declared here, before the definition of corresponding kstat * macro which uses the same names will confuse the compiler. */ SYSCTL_PROC(_vfs_zfs, OID_AUTO, arc_meta_limit, CTLTYPE_U64 | CTLFLAG_MPSAFE | CTLFLAG_RW, 0, sizeof(uint64_t), sysctl_vfs_zfs_arc_meta_limit, "QU", "ARC metadata limit"); #endif /* * Note that buffers can be in one of 6 states: * ARC_anon - anonymous (discussed below) * ARC_mru - recently used, currently cached * ARC_mru_ghost - recentely used, no longer in cache * ARC_mfu - frequently used, currently cached * ARC_mfu_ghost - frequently used, no longer in cache * ARC_l2c_only - exists in L2ARC but not other states * When there are no active references to the buffer, they are * are linked onto a list in one of these arc states. These are * the only buffers that can be evicted or deleted. Within each * state there are multiple lists, one for meta-data and one for * non-meta-data. Meta-data (indirect blocks, blocks of dnodes, * etc.) is tracked separately so that it can be managed more * explicitly: favored over data, limited explicitly. * * Anonymous buffers are buffers that are not associated with * a DVA. These are buffers that hold dirty block copies * before they are written to stable storage. By definition, * they are "ref'd" and are considered part of arc_mru * that cannot be freed. Generally, they will aquire a DVA * as they are written and migrate onto the arc_mru list. * * The ARC_l2c_only state is for buffers that are in the second * level ARC but no longer in any of the ARC_m* lists. The second * level ARC itself may also contain buffers that are in any of * the ARC_m* states - meaning that a buffer can exist in two * places. The reason for the ARC_l2c_only state is to keep the * buffer header in the hash table, so that reads that hit the * second level ARC benefit from these fast lookups. */ typedef struct arc_state { /* * list of evictable buffers */ multilist_t arcs_list[ARC_BUFC_NUMTYPES]; /* * total amount of evictable data in this state */ uint64_t arcs_lsize[ARC_BUFC_NUMTYPES]; /* * total amount of data in this state; this includes: evictable, * non-evictable, ARC_BUFC_DATA, and ARC_BUFC_METADATA. */ refcount_t arcs_size; } arc_state_t; /* The 6 states: */ static arc_state_t ARC_anon; static arc_state_t ARC_mru; static arc_state_t ARC_mru_ghost; static arc_state_t ARC_mfu; static arc_state_t ARC_mfu_ghost; static arc_state_t ARC_l2c_only; typedef struct arc_stats { kstat_named_t arcstat_hits; kstat_named_t arcstat_misses; kstat_named_t arcstat_demand_data_hits; kstat_named_t arcstat_demand_data_misses; kstat_named_t arcstat_demand_metadata_hits; kstat_named_t arcstat_demand_metadata_misses; kstat_named_t arcstat_prefetch_data_hits; kstat_named_t arcstat_prefetch_data_misses; kstat_named_t arcstat_prefetch_metadata_hits; kstat_named_t arcstat_prefetch_metadata_misses; kstat_named_t arcstat_mru_hits; kstat_named_t arcstat_mru_ghost_hits; kstat_named_t arcstat_mfu_hits; kstat_named_t arcstat_mfu_ghost_hits; kstat_named_t arcstat_allocated; kstat_named_t arcstat_deleted; /* * Number of buffers that could not be evicted because the hash lock * was held by another thread. The lock may not necessarily be held * by something using the same buffer, since hash locks are shared * by multiple buffers. */ kstat_named_t arcstat_mutex_miss; /* * Number of buffers skipped because they have I/O in progress, are * indrect prefetch buffers that have not lived long enough, or are * not from the spa we're trying to evict from. */ kstat_named_t arcstat_evict_skip; /* * Number of times arc_evict_state() was unable to evict enough * buffers to reach it's target amount. */ kstat_named_t arcstat_evict_not_enough; kstat_named_t arcstat_evict_l2_cached; kstat_named_t arcstat_evict_l2_eligible; kstat_named_t arcstat_evict_l2_ineligible; kstat_named_t arcstat_evict_l2_skip; kstat_named_t arcstat_hash_elements; kstat_named_t arcstat_hash_elements_max; kstat_named_t arcstat_hash_collisions; kstat_named_t arcstat_hash_chains; kstat_named_t arcstat_hash_chain_max; kstat_named_t arcstat_p; kstat_named_t arcstat_c; kstat_named_t arcstat_c_min; kstat_named_t arcstat_c_max; kstat_named_t arcstat_size; /* * Number of bytes consumed by internal ARC structures necessary * for tracking purposes; these structures are not actually * backed by ARC buffers. This includes arc_buf_hdr_t structures * (allocated via arc_buf_hdr_t_full and arc_buf_hdr_t_l2only * caches), and arc_buf_t structures (allocated via arc_buf_t * cache). */ kstat_named_t arcstat_hdr_size; /* * Number of bytes consumed by ARC buffers of type equal to * ARC_BUFC_DATA. This is generally consumed by buffers backing * on disk user data (e.g. plain file contents). */ kstat_named_t arcstat_data_size; /* * Number of bytes consumed by ARC buffers of type equal to * ARC_BUFC_METADATA. This is generally consumed by buffers * backing on disk data that is used for internal ZFS * structures (e.g. ZAP, dnode, indirect blocks, etc). */ kstat_named_t arcstat_metadata_size; /* * Number of bytes consumed by various buffers and structures * not actually backed with ARC buffers. This includes bonus * buffers (allocated directly via zio_buf_* functions), * dmu_buf_impl_t structures (allocated via dmu_buf_impl_t * cache), and dnode_t structures (allocated via dnode_t cache). */ kstat_named_t arcstat_other_size; /* * Total number of bytes consumed by ARC buffers residing in the * arc_anon state. This includes *all* buffers in the arc_anon * state; e.g. data, metadata, evictable, and unevictable buffers * are all included in this value. */ kstat_named_t arcstat_anon_size; /* * Number of bytes consumed by ARC buffers that meet the * following criteria: backing buffers of type ARC_BUFC_DATA, * residing in the arc_anon state, and are eligible for eviction * (e.g. have no outstanding holds on the buffer). */ kstat_named_t arcstat_anon_evictable_data; /* * Number of bytes consumed by ARC buffers that meet the * following criteria: backing buffers of type ARC_BUFC_METADATA, * residing in the arc_anon state, and are eligible for eviction * (e.g. have no outstanding holds on the buffer). */ kstat_named_t arcstat_anon_evictable_metadata; /* * Total number of bytes consumed by ARC buffers residing in the * arc_mru state. This includes *all* buffers in the arc_mru * state; e.g. data, metadata, evictable, and unevictable buffers * are all included in this value. */ kstat_named_t arcstat_mru_size; /* * Number of bytes consumed by ARC buffers that meet the * following criteria: backing buffers of type ARC_BUFC_DATA, * residing in the arc_mru state, and are eligible for eviction * (e.g. have no outstanding holds on the buffer). */ kstat_named_t arcstat_mru_evictable_data; /* * Number of bytes consumed by ARC buffers that meet the * following criteria: backing buffers of type ARC_BUFC_METADATA, * residing in the arc_mru state, and are eligible for eviction * (e.g. have no outstanding holds on the buffer). */ kstat_named_t arcstat_mru_evictable_metadata; /* * Total number of bytes that *would have been* consumed by ARC * buffers in the arc_mru_ghost state. The key thing to note * here, is the fact that this size doesn't actually indicate * RAM consumption. The ghost lists only consist of headers and * don't actually have ARC buffers linked off of these headers. * Thus, *if* the headers had associated ARC buffers, these * buffers *would have* consumed this number of bytes. */ kstat_named_t arcstat_mru_ghost_size; /* * Number of bytes that *would have been* consumed by ARC * buffers that are eligible for eviction, of type * ARC_BUFC_DATA, and linked off the arc_mru_ghost state. */ kstat_named_t arcstat_mru_ghost_evictable_data; /* * Number of bytes that *would have been* consumed by ARC * buffers that are eligible for eviction, of type * ARC_BUFC_METADATA, and linked off the arc_mru_ghost state. */ kstat_named_t arcstat_mru_ghost_evictable_metadata; /* * Total number of bytes consumed by ARC buffers residing in the * arc_mfu state. This includes *all* buffers in the arc_mfu * state; e.g. data, metadata, evictable, and unevictable buffers * are all included in this value. */ kstat_named_t arcstat_mfu_size; /* * Number of bytes consumed by ARC buffers that are eligible for * eviction, of type ARC_BUFC_DATA, and reside in the arc_mfu * state. */ kstat_named_t arcstat_mfu_evictable_data; /* * Number of bytes consumed by ARC buffers that are eligible for * eviction, of type ARC_BUFC_METADATA, and reside in the * arc_mfu state. */ kstat_named_t arcstat_mfu_evictable_metadata; /* * Total number of bytes that *would have been* consumed by ARC * buffers in the arc_mfu_ghost state. See the comment above * arcstat_mru_ghost_size for more details. */ kstat_named_t arcstat_mfu_ghost_size; /* * Number of bytes that *would have been* consumed by ARC * buffers that are eligible for eviction, of type * ARC_BUFC_DATA, and linked off the arc_mfu_ghost state. */ kstat_named_t arcstat_mfu_ghost_evictable_data; /* * Number of bytes that *would have been* consumed by ARC * buffers that are eligible for eviction, of type * ARC_BUFC_METADATA, and linked off the arc_mru_ghost state. */ kstat_named_t arcstat_mfu_ghost_evictable_metadata; kstat_named_t arcstat_l2_hits; kstat_named_t arcstat_l2_misses; kstat_named_t arcstat_l2_feeds; kstat_named_t arcstat_l2_rw_clash; kstat_named_t arcstat_l2_read_bytes; kstat_named_t arcstat_l2_write_bytes; kstat_named_t arcstat_l2_writes_sent; kstat_named_t arcstat_l2_writes_done; kstat_named_t arcstat_l2_writes_error; kstat_named_t arcstat_l2_writes_lock_retry; kstat_named_t arcstat_l2_evict_lock_retry; kstat_named_t arcstat_l2_evict_reading; kstat_named_t arcstat_l2_evict_l1cached; kstat_named_t arcstat_l2_free_on_write; kstat_named_t arcstat_l2_cdata_free_on_write; kstat_named_t arcstat_l2_abort_lowmem; kstat_named_t arcstat_l2_cksum_bad; kstat_named_t arcstat_l2_io_error; kstat_named_t arcstat_l2_size; kstat_named_t arcstat_l2_asize; kstat_named_t arcstat_l2_hdr_size; kstat_named_t arcstat_l2_compress_successes; kstat_named_t arcstat_l2_compress_zeros; kstat_named_t arcstat_l2_compress_failures; kstat_named_t arcstat_l2_write_trylock_fail; kstat_named_t arcstat_l2_write_passed_headroom; kstat_named_t arcstat_l2_write_spa_mismatch; kstat_named_t arcstat_l2_write_in_l2; kstat_named_t arcstat_l2_write_hdr_io_in_progress; kstat_named_t arcstat_l2_write_not_cacheable; kstat_named_t arcstat_l2_write_full; kstat_named_t arcstat_l2_write_buffer_iter; kstat_named_t arcstat_l2_write_pios; kstat_named_t arcstat_l2_write_buffer_bytes_scanned; kstat_named_t arcstat_l2_write_buffer_list_iter; kstat_named_t arcstat_l2_write_buffer_list_null_iter; kstat_named_t arcstat_memory_throttle_count; kstat_named_t arcstat_duplicate_buffers; kstat_named_t arcstat_duplicate_buffers_size; kstat_named_t arcstat_duplicate_reads; kstat_named_t arcstat_meta_used; kstat_named_t arcstat_meta_limit; kstat_named_t arcstat_meta_max; kstat_named_t arcstat_meta_min; kstat_named_t arcstat_sync_wait_for_async; kstat_named_t arcstat_demand_hit_predictive_prefetch; } arc_stats_t; static arc_stats_t arc_stats = { { "hits", KSTAT_DATA_UINT64 }, { "misses", KSTAT_DATA_UINT64 }, { "demand_data_hits", KSTAT_DATA_UINT64 }, { "demand_data_misses", KSTAT_DATA_UINT64 }, { "demand_metadata_hits", KSTAT_DATA_UINT64 }, { "demand_metadata_misses", KSTAT_DATA_UINT64 }, { "prefetch_data_hits", KSTAT_DATA_UINT64 }, { "prefetch_data_misses", KSTAT_DATA_UINT64 }, { "prefetch_metadata_hits", KSTAT_DATA_UINT64 }, { "prefetch_metadata_misses", KSTAT_DATA_UINT64 }, { "mru_hits", KSTAT_DATA_UINT64 }, { "mru_ghost_hits", KSTAT_DATA_UINT64 }, { "mfu_hits", KSTAT_DATA_UINT64 }, { "mfu_ghost_hits", KSTAT_DATA_UINT64 }, { "allocated", KSTAT_DATA_UINT64 }, { "deleted", KSTAT_DATA_UINT64 }, { "mutex_miss", KSTAT_DATA_UINT64 }, { "evict_skip", KSTAT_DATA_UINT64 }, { "evict_not_enough", KSTAT_DATA_UINT64 }, { "evict_l2_cached", KSTAT_DATA_UINT64 }, { "evict_l2_eligible", KSTAT_DATA_UINT64 }, { "evict_l2_ineligible", KSTAT_DATA_UINT64 }, { "evict_l2_skip", KSTAT_DATA_UINT64 }, { "hash_elements", KSTAT_DATA_UINT64 }, { "hash_elements_max", KSTAT_DATA_UINT64 }, { "hash_collisions", KSTAT_DATA_UINT64 }, { "hash_chains", KSTAT_DATA_UINT64 }, { "hash_chain_max", KSTAT_DATA_UINT64 }, { "p", KSTAT_DATA_UINT64 }, { "c", KSTAT_DATA_UINT64 }, { "c_min", KSTAT_DATA_UINT64 }, { "c_max", KSTAT_DATA_UINT64 }, { "size", KSTAT_DATA_UINT64 }, { "hdr_size", KSTAT_DATA_UINT64 }, { "data_size", KSTAT_DATA_UINT64 }, { "metadata_size", KSTAT_DATA_UINT64 }, { "other_size", KSTAT_DATA_UINT64 }, { "anon_size", KSTAT_DATA_UINT64 }, { "anon_evictable_data", KSTAT_DATA_UINT64 }, { "anon_evictable_metadata", KSTAT_DATA_UINT64 }, { "mru_size", KSTAT_DATA_UINT64 }, { "mru_evictable_data", KSTAT_DATA_UINT64 }, { "mru_evictable_metadata", KSTAT_DATA_UINT64 }, { "mru_ghost_size", KSTAT_DATA_UINT64 }, { "mru_ghost_evictable_data", KSTAT_DATA_UINT64 }, { "mru_ghost_evictable_metadata", KSTAT_DATA_UINT64 }, { "mfu_size", KSTAT_DATA_UINT64 }, { "mfu_evictable_data", KSTAT_DATA_UINT64 }, { "mfu_evictable_metadata", KSTAT_DATA_UINT64 }, { "mfu_ghost_size", KSTAT_DATA_UINT64 }, { "mfu_ghost_evictable_data", KSTAT_DATA_UINT64 }, { "mfu_ghost_evictable_metadata", KSTAT_DATA_UINT64 }, { "l2_hits", KSTAT_DATA_UINT64 }, { "l2_misses", KSTAT_DATA_UINT64 }, { "l2_feeds", KSTAT_DATA_UINT64 }, { "l2_rw_clash", KSTAT_DATA_UINT64 }, { "l2_read_bytes", KSTAT_DATA_UINT64 }, { "l2_write_bytes", KSTAT_DATA_UINT64 }, { "l2_writes_sent", KSTAT_DATA_UINT64 }, { "l2_writes_done", KSTAT_DATA_UINT64 }, { "l2_writes_error", KSTAT_DATA_UINT64 }, { "l2_writes_lock_retry", KSTAT_DATA_UINT64 }, { "l2_evict_lock_retry", KSTAT_DATA_UINT64 }, { "l2_evict_reading", KSTAT_DATA_UINT64 }, { "l2_evict_l1cached", KSTAT_DATA_UINT64 }, { "l2_free_on_write", KSTAT_DATA_UINT64 }, { "l2_cdata_free_on_write", KSTAT_DATA_UINT64 }, { "l2_abort_lowmem", KSTAT_DATA_UINT64 }, { "l2_cksum_bad", KSTAT_DATA_UINT64 }, { "l2_io_error", KSTAT_DATA_UINT64 }, { "l2_size", KSTAT_DATA_UINT64 }, { "l2_asize", KSTAT_DATA_UINT64 }, { "l2_hdr_size", KSTAT_DATA_UINT64 }, { "l2_compress_successes", KSTAT_DATA_UINT64 }, { "l2_compress_zeros", KSTAT_DATA_UINT64 }, { "l2_compress_failures", KSTAT_DATA_UINT64 }, { "l2_write_trylock_fail", KSTAT_DATA_UINT64 }, { "l2_write_passed_headroom", KSTAT_DATA_UINT64 }, { "l2_write_spa_mismatch", KSTAT_DATA_UINT64 }, { "l2_write_in_l2", KSTAT_DATA_UINT64 }, { "l2_write_io_in_progress", KSTAT_DATA_UINT64 }, { "l2_write_not_cacheable", KSTAT_DATA_UINT64 }, { "l2_write_full", KSTAT_DATA_UINT64 }, { "l2_write_buffer_iter", KSTAT_DATA_UINT64 }, { "l2_write_pios", KSTAT_DATA_UINT64 }, { "l2_write_buffer_bytes_scanned", KSTAT_DATA_UINT64 }, { "l2_write_buffer_list_iter", KSTAT_DATA_UINT64 }, { "l2_write_buffer_list_null_iter", KSTAT_DATA_UINT64 }, { "memory_throttle_count", KSTAT_DATA_UINT64 }, { "duplicate_buffers", KSTAT_DATA_UINT64 }, { "duplicate_buffers_size", KSTAT_DATA_UINT64 }, { "duplicate_reads", KSTAT_DATA_UINT64 }, { "arc_meta_used", KSTAT_DATA_UINT64 }, { "arc_meta_limit", KSTAT_DATA_UINT64 }, { "arc_meta_max", KSTAT_DATA_UINT64 }, { "arc_meta_min", KSTAT_DATA_UINT64 }, { "sync_wait_for_async", KSTAT_DATA_UINT64 }, { "demand_hit_predictive_prefetch", KSTAT_DATA_UINT64 }, }; #define ARCSTAT(stat) (arc_stats.stat.value.ui64) #define ARCSTAT_INCR(stat, val) \ atomic_add_64(&arc_stats.stat.value.ui64, (val)) #define ARCSTAT_BUMP(stat) ARCSTAT_INCR(stat, 1) #define ARCSTAT_BUMPDOWN(stat) ARCSTAT_INCR(stat, -1) #define ARCSTAT_MAX(stat, val) { \ uint64_t m; \ while ((val) > (m = arc_stats.stat.value.ui64) && \ (m != atomic_cas_64(&arc_stats.stat.value.ui64, m, (val)))) \ continue; \ } #define ARCSTAT_MAXSTAT(stat) \ ARCSTAT_MAX(stat##_max, arc_stats.stat.value.ui64) /* * We define a macro to allow ARC hits/misses to be easily broken down by * two separate conditions, giving a total of four different subtypes for * each of hits and misses (so eight statistics total). */ #define ARCSTAT_CONDSTAT(cond1, stat1, notstat1, cond2, stat2, notstat2, stat) \ if (cond1) { \ if (cond2) { \ ARCSTAT_BUMP(arcstat_##stat1##_##stat2##_##stat); \ } else { \ ARCSTAT_BUMP(arcstat_##stat1##_##notstat2##_##stat); \ } \ } else { \ if (cond2) { \ ARCSTAT_BUMP(arcstat_##notstat1##_##stat2##_##stat); \ } else { \ ARCSTAT_BUMP(arcstat_##notstat1##_##notstat2##_##stat);\ } \ } kstat_t *arc_ksp; static arc_state_t *arc_anon; static arc_state_t *arc_mru; static arc_state_t *arc_mru_ghost; static arc_state_t *arc_mfu; static arc_state_t *arc_mfu_ghost; static arc_state_t *arc_l2c_only; /* * There are several ARC variables that are critical to export as kstats -- * but we don't want to have to grovel around in the kstat whenever we wish to * manipulate them. For these variables, we therefore define them to be in * terms of the statistic variable. This assures that we are not introducing * the possibility of inconsistency by having shadow copies of the variables, * while still allowing the code to be readable. */ #define arc_size ARCSTAT(arcstat_size) /* actual total arc size */ #define arc_p ARCSTAT(arcstat_p) /* target size of MRU */ #define arc_c ARCSTAT(arcstat_c) /* target size of cache */ #define arc_c_min ARCSTAT(arcstat_c_min) /* min target cache size */ #define arc_c_max ARCSTAT(arcstat_c_max) /* max target cache size */ #define arc_meta_limit ARCSTAT(arcstat_meta_limit) /* max size for metadata */ #define arc_meta_min ARCSTAT(arcstat_meta_min) /* min size for metadata */ #define arc_meta_used ARCSTAT(arcstat_meta_used) /* size of metadata */ #define arc_meta_max ARCSTAT(arcstat_meta_max) /* max size of metadata */ #define L2ARC_IS_VALID_COMPRESS(_c_) \ ((_c_) == ZIO_COMPRESS_LZ4 || (_c_) == ZIO_COMPRESS_EMPTY) static int arc_no_grow; /* Don't try to grow cache size */ static uint64_t arc_tempreserve; static uint64_t arc_loaned_bytes; typedef struct arc_callback arc_callback_t; struct arc_callback { void *acb_private; arc_done_func_t *acb_done; arc_buf_t *acb_buf; zio_t *acb_zio_dummy; arc_callback_t *acb_next; }; typedef struct arc_write_callback arc_write_callback_t; struct arc_write_callback { void *awcb_private; arc_done_func_t *awcb_ready; arc_done_func_t *awcb_physdone; arc_done_func_t *awcb_done; arc_buf_t *awcb_buf; }; /* * ARC buffers are separated into multiple structs as a memory saving measure: * - Common fields struct, always defined, and embedded within it: * - L2-only fields, always allocated but undefined when not in L2ARC * - L1-only fields, only allocated when in L1ARC * * Buffer in L1 Buffer only in L2 * +------------------------+ +------------------------+ * | arc_buf_hdr_t | | arc_buf_hdr_t | * | | | | * | | | | * | | | | * +------------------------+ +------------------------+ * | l2arc_buf_hdr_t | | l2arc_buf_hdr_t | * | (undefined if L1-only) | | | * +------------------------+ +------------------------+ * | l1arc_buf_hdr_t | * | | * | | * | | * | | * +------------------------+ * * Because it's possible for the L2ARC to become extremely large, we can wind * up eating a lot of memory in L2ARC buffer headers, so the size of a header * is minimized by only allocating the fields necessary for an L1-cached buffer * when a header is actually in the L1 cache. The sub-headers (l1arc_buf_hdr and * l2arc_buf_hdr) are embedded rather than allocated separately to save a couple * words in pointers. arc_hdr_realloc() is used to switch a header between * these two allocation states. */ typedef struct l1arc_buf_hdr { kmutex_t b_freeze_lock; #ifdef ZFS_DEBUG /* * used for debugging wtih kmem_flags - by allocating and freeing * b_thawed when the buffer is thawed, we get a record of the stack * trace that thawed it. */ void *b_thawed; #endif arc_buf_t *b_buf; uint32_t b_datacnt; /* for waiting on writes to complete */ kcondvar_t b_cv; /* protected by arc state mutex */ arc_state_t *b_state; multilist_node_t b_arc_node; /* updated atomically */ clock_t b_arc_access; /* self protecting */ refcount_t b_refcnt; arc_callback_t *b_acb; /* temporary buffer holder for in-flight compressed data */ void *b_tmp_cdata; } l1arc_buf_hdr_t; typedef struct l2arc_dev l2arc_dev_t; typedef struct l2arc_buf_hdr { /* protected by arc_buf_hdr mutex */ l2arc_dev_t *b_dev; /* L2ARC device */ uint64_t b_daddr; /* disk address, offset byte */ /* real alloc'd buffer size depending on b_compress applied */ int32_t b_asize; uint8_t b_compress; list_node_t b_l2node; } l2arc_buf_hdr_t; struct arc_buf_hdr { /* protected by hash lock */ dva_t b_dva; uint64_t b_birth; /* * Even though this checksum is only set/verified when a buffer is in * the L1 cache, it needs to be in the set of common fields because it * must be preserved from the time before a buffer is written out to * L2ARC until after it is read back in. */ zio_cksum_t *b_freeze_cksum; arc_buf_hdr_t *b_hash_next; arc_flags_t b_flags; /* immutable */ int32_t b_size; uint64_t b_spa; /* L2ARC fields. Undefined when not in L2ARC. */ l2arc_buf_hdr_t b_l2hdr; /* L1ARC fields. Undefined when in l2arc_only state */ l1arc_buf_hdr_t b_l1hdr; }; #ifdef _KERNEL static int sysctl_vfs_zfs_arc_meta_limit(SYSCTL_HANDLER_ARGS) { uint64_t val; int err; val = arc_meta_limit; err = sysctl_handle_64(oidp, &val, 0, req); if (err != 0 || req->newptr == NULL) return (err); if (val <= 0 || val > arc_c_max) return (EINVAL); arc_meta_limit = val; return (0); } #endif static arc_buf_t *arc_eviction_list; static arc_buf_hdr_t arc_eviction_hdr; #define GHOST_STATE(state) \ ((state) == arc_mru_ghost || (state) == arc_mfu_ghost || \ (state) == arc_l2c_only) #define HDR_IN_HASH_TABLE(hdr) ((hdr)->b_flags & ARC_FLAG_IN_HASH_TABLE) #define HDR_IO_IN_PROGRESS(hdr) ((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS) #define HDR_IO_ERROR(hdr) ((hdr)->b_flags & ARC_FLAG_IO_ERROR) #define HDR_PREFETCH(hdr) ((hdr)->b_flags & ARC_FLAG_PREFETCH) #define HDR_FREED_IN_READ(hdr) ((hdr)->b_flags & ARC_FLAG_FREED_IN_READ) #define HDR_BUF_AVAILABLE(hdr) ((hdr)->b_flags & ARC_FLAG_BUF_AVAILABLE) #define HDR_L2CACHE(hdr) ((hdr)->b_flags & ARC_FLAG_L2CACHE) #define HDR_L2COMPRESS(hdr) ((hdr)->b_flags & ARC_FLAG_L2COMPRESS) #define HDR_L2_READING(hdr) \ (((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS) && \ ((hdr)->b_flags & ARC_FLAG_HAS_L2HDR)) #define HDR_L2_WRITING(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITING) #define HDR_L2_EVICTED(hdr) ((hdr)->b_flags & ARC_FLAG_L2_EVICTED) #define HDR_L2_WRITE_HEAD(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITE_HEAD) #define HDR_ISTYPE_METADATA(hdr) \ ((hdr)->b_flags & ARC_FLAG_BUFC_METADATA) #define HDR_ISTYPE_DATA(hdr) (!HDR_ISTYPE_METADATA(hdr)) #define HDR_HAS_L1HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L1HDR) #define HDR_HAS_L2HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L2HDR) /* * Other sizes */ #define HDR_FULL_SIZE ((int64_t)sizeof (arc_buf_hdr_t)) #define HDR_L2ONLY_SIZE ((int64_t)offsetof(arc_buf_hdr_t, b_l1hdr)) /* * Hash table routines */ #define HT_LOCK_PAD CACHE_LINE_SIZE struct ht_lock { kmutex_t ht_lock; #ifdef _KERNEL unsigned char pad[(HT_LOCK_PAD - sizeof (kmutex_t))]; #endif }; #define BUF_LOCKS 256 typedef struct buf_hash_table { uint64_t ht_mask; arc_buf_hdr_t **ht_table; struct ht_lock ht_locks[BUF_LOCKS] __aligned(CACHE_LINE_SIZE); } buf_hash_table_t; static buf_hash_table_t buf_hash_table; #define BUF_HASH_INDEX(spa, dva, birth) \ (buf_hash(spa, dva, birth) & buf_hash_table.ht_mask) #define BUF_HASH_LOCK_NTRY(idx) (buf_hash_table.ht_locks[idx & (BUF_LOCKS-1)]) #define BUF_HASH_LOCK(idx) (&(BUF_HASH_LOCK_NTRY(idx).ht_lock)) #define HDR_LOCK(hdr) \ (BUF_HASH_LOCK(BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth))) uint64_t zfs_crc64_table[256]; /* * Level 2 ARC */ #define L2ARC_WRITE_SIZE (8 * 1024 * 1024) /* initial write max */ #define L2ARC_HEADROOM 2 /* num of writes */ /* * If we discover during ARC scan any buffers to be compressed, we boost * our headroom for the next scanning cycle by this percentage multiple. */ #define L2ARC_HEADROOM_BOOST 200 #define L2ARC_FEED_SECS 1 /* caching interval secs */ #define L2ARC_FEED_MIN_MS 200 /* min caching interval ms */ /* * Used to distinguish headers that are being process by * l2arc_write_buffers(), but have yet to be assigned to a l2arc disk * address. This can happen when the header is added to the l2arc's list * of buffers to write in the first stage of l2arc_write_buffers(), but * has not yet been written out which happens in the second stage of * l2arc_write_buffers(). */ #define L2ARC_ADDR_UNSET ((uint64_t)(-1)) #define l2arc_writes_sent ARCSTAT(arcstat_l2_writes_sent) #define l2arc_writes_done ARCSTAT(arcstat_l2_writes_done) /* L2ARC Performance Tunables */ uint64_t l2arc_write_max = L2ARC_WRITE_SIZE; /* default max write size */ uint64_t l2arc_write_boost = L2ARC_WRITE_SIZE; /* extra write during warmup */ uint64_t l2arc_headroom = L2ARC_HEADROOM; /* number of dev writes */ uint64_t l2arc_headroom_boost = L2ARC_HEADROOM_BOOST; uint64_t l2arc_feed_secs = L2ARC_FEED_SECS; /* interval seconds */ uint64_t l2arc_feed_min_ms = L2ARC_FEED_MIN_MS; /* min interval milliseconds */ boolean_t l2arc_noprefetch = B_TRUE; /* don't cache prefetch bufs */ boolean_t l2arc_feed_again = B_TRUE; /* turbo warmup */ boolean_t l2arc_norw = B_TRUE; /* no reads during writes */ SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_write_max, CTLFLAG_RW, &l2arc_write_max, 0, "max write size"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_write_boost, CTLFLAG_RW, &l2arc_write_boost, 0, "extra write during warmup"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_headroom, CTLFLAG_RW, &l2arc_headroom, 0, "number of dev writes"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_feed_secs, CTLFLAG_RW, &l2arc_feed_secs, 0, "interval seconds"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_feed_min_ms, CTLFLAG_RW, &l2arc_feed_min_ms, 0, "min interval milliseconds"); SYSCTL_INT(_vfs_zfs, OID_AUTO, l2arc_noprefetch, CTLFLAG_RW, &l2arc_noprefetch, 0, "don't cache prefetch bufs"); SYSCTL_INT(_vfs_zfs, OID_AUTO, l2arc_feed_again, CTLFLAG_RW, &l2arc_feed_again, 0, "turbo warmup"); SYSCTL_INT(_vfs_zfs, OID_AUTO, l2arc_norw, CTLFLAG_RW, &l2arc_norw, 0, "no reads during writes"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, anon_size, CTLFLAG_RD, &ARC_anon.arcs_size.rc_count, 0, "size of anonymous state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, anon_metadata_lsize, CTLFLAG_RD, &ARC_anon.arcs_lsize[ARC_BUFC_METADATA], 0, "size of anonymous state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, anon_data_lsize, CTLFLAG_RD, &ARC_anon.arcs_lsize[ARC_BUFC_DATA], 0, "size of anonymous state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_size, CTLFLAG_RD, &ARC_mru.arcs_size.rc_count, 0, "size of mru state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_metadata_lsize, CTLFLAG_RD, &ARC_mru.arcs_lsize[ARC_BUFC_METADATA], 0, "size of metadata in mru state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_data_lsize, CTLFLAG_RD, &ARC_mru.arcs_lsize[ARC_BUFC_DATA], 0, "size of data in mru state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_ghost_size, CTLFLAG_RD, &ARC_mru_ghost.arcs_size.rc_count, 0, "size of mru ghost state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_ghost_metadata_lsize, CTLFLAG_RD, &ARC_mru_ghost.arcs_lsize[ARC_BUFC_METADATA], 0, "size of metadata in mru ghost state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_ghost_data_lsize, CTLFLAG_RD, &ARC_mru_ghost.arcs_lsize[ARC_BUFC_DATA], 0, "size of data in mru ghost state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_size, CTLFLAG_RD, &ARC_mfu.arcs_size.rc_count, 0, "size of mfu state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_metadata_lsize, CTLFLAG_RD, &ARC_mfu.arcs_lsize[ARC_BUFC_METADATA], 0, "size of metadata in mfu state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_data_lsize, CTLFLAG_RD, &ARC_mfu.arcs_lsize[ARC_BUFC_DATA], 0, "size of data in mfu state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_ghost_size, CTLFLAG_RD, &ARC_mfu_ghost.arcs_size.rc_count, 0, "size of mfu ghost state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_ghost_metadata_lsize, CTLFLAG_RD, &ARC_mfu_ghost.arcs_lsize[ARC_BUFC_METADATA], 0, "size of metadata in mfu ghost state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_ghost_data_lsize, CTLFLAG_RD, &ARC_mfu_ghost.arcs_lsize[ARC_BUFC_DATA], 0, "size of data in mfu ghost state"); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2c_only_size, CTLFLAG_RD, &ARC_l2c_only.arcs_size.rc_count, 0, "size of mru state"); /* * L2ARC Internals */ struct l2arc_dev { vdev_t *l2ad_vdev; /* vdev */ spa_t *l2ad_spa; /* spa */ uint64_t l2ad_hand; /* next write location */ uint64_t l2ad_start; /* first addr on device */ uint64_t l2ad_end; /* last addr on device */ boolean_t l2ad_first; /* first sweep through */ boolean_t l2ad_writing; /* currently writing */ kmutex_t l2ad_mtx; /* lock for buffer list */ list_t l2ad_buflist; /* buffer list */ list_node_t l2ad_node; /* device list node */ refcount_t l2ad_alloc; /* allocated bytes */ }; static list_t L2ARC_dev_list; /* device list */ static list_t *l2arc_dev_list; /* device list pointer */ static kmutex_t l2arc_dev_mtx; /* device list mutex */ static l2arc_dev_t *l2arc_dev_last; /* last device used */ static list_t L2ARC_free_on_write; /* free after write buf list */ static list_t *l2arc_free_on_write; /* free after write list ptr */ static kmutex_t l2arc_free_on_write_mtx; /* mutex for list */ static uint64_t l2arc_ndev; /* number of devices */ typedef struct l2arc_read_callback { arc_buf_t *l2rcb_buf; /* read buffer */ spa_t *l2rcb_spa; /* spa */ blkptr_t l2rcb_bp; /* original blkptr */ zbookmark_phys_t l2rcb_zb; /* original bookmark */ int l2rcb_flags; /* original flags */ enum zio_compress l2rcb_compress; /* applied compress */ } l2arc_read_callback_t; typedef struct l2arc_write_callback { l2arc_dev_t *l2wcb_dev; /* device info */ arc_buf_hdr_t *l2wcb_head; /* head of write buflist */ } l2arc_write_callback_t; typedef struct l2arc_data_free { /* protected by l2arc_free_on_write_mtx */ void *l2df_data; size_t l2df_size; void (*l2df_func)(void *, size_t); list_node_t l2df_list_node; } l2arc_data_free_t; static kmutex_t l2arc_feed_thr_lock; static kcondvar_t l2arc_feed_thr_cv; static uint8_t l2arc_thread_exit; static void arc_get_data_buf(arc_buf_t *); static void arc_access(arc_buf_hdr_t *, kmutex_t *); static boolean_t arc_is_overflowing(); static void arc_buf_watch(arc_buf_t *); static arc_buf_contents_t arc_buf_type(arc_buf_hdr_t *); static uint32_t arc_bufc_to_flags(arc_buf_contents_t); static boolean_t l2arc_write_eligible(uint64_t, arc_buf_hdr_t *); static void l2arc_read_done(zio_t *); static boolean_t l2arc_compress_buf(arc_buf_hdr_t *); static void l2arc_decompress_zio(zio_t *, arc_buf_hdr_t *, enum zio_compress); static void l2arc_release_cdata_buf(arc_buf_hdr_t *); static uint64_t buf_hash(uint64_t spa, const dva_t *dva, uint64_t birth) { uint8_t *vdva = (uint8_t *)dva; uint64_t crc = -1ULL; int i; ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY); for (i = 0; i < sizeof (dva_t); i++) crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ vdva[i]) & 0xFF]; crc ^= (spa>>8) ^ birth; return (crc); } #define BUF_EMPTY(buf) \ ((buf)->b_dva.dva_word[0] == 0 && \ (buf)->b_dva.dva_word[1] == 0) #define BUF_EQUAL(spa, dva, birth, buf) \ ((buf)->b_dva.dva_word[0] == (dva)->dva_word[0]) && \ ((buf)->b_dva.dva_word[1] == (dva)->dva_word[1]) && \ ((buf)->b_birth == birth) && ((buf)->b_spa == spa) static void buf_discard_identity(arc_buf_hdr_t *hdr) { hdr->b_dva.dva_word[0] = 0; hdr->b_dva.dva_word[1] = 0; hdr->b_birth = 0; } static arc_buf_hdr_t * buf_hash_find(uint64_t spa, const blkptr_t *bp, kmutex_t **lockp) { const dva_t *dva = BP_IDENTITY(bp); uint64_t birth = BP_PHYSICAL_BIRTH(bp); uint64_t idx = BUF_HASH_INDEX(spa, dva, birth); kmutex_t *hash_lock = BUF_HASH_LOCK(idx); arc_buf_hdr_t *hdr; mutex_enter(hash_lock); for (hdr = buf_hash_table.ht_table[idx]; hdr != NULL; hdr = hdr->b_hash_next) { if (BUF_EQUAL(spa, dva, birth, hdr)) { *lockp = hash_lock; return (hdr); } } mutex_exit(hash_lock); *lockp = NULL; return (NULL); } /* * Insert an entry into the hash table. If there is already an element * equal to elem in the hash table, then the already existing element * will be returned and the new element will not be inserted. * Otherwise returns NULL. * If lockp == NULL, the caller is assumed to already hold the hash lock. */ static arc_buf_hdr_t * buf_hash_insert(arc_buf_hdr_t *hdr, kmutex_t **lockp) { uint64_t idx = BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth); kmutex_t *hash_lock = BUF_HASH_LOCK(idx); arc_buf_hdr_t *fhdr; uint32_t i; ASSERT(!DVA_IS_EMPTY(&hdr->b_dva)); ASSERT(hdr->b_birth != 0); ASSERT(!HDR_IN_HASH_TABLE(hdr)); if (lockp != NULL) { *lockp = hash_lock; mutex_enter(hash_lock); } else { ASSERT(MUTEX_HELD(hash_lock)); } for (fhdr = buf_hash_table.ht_table[idx], i = 0; fhdr != NULL; fhdr = fhdr->b_hash_next, i++) { if (BUF_EQUAL(hdr->b_spa, &hdr->b_dva, hdr->b_birth, fhdr)) return (fhdr); } hdr->b_hash_next = buf_hash_table.ht_table[idx]; buf_hash_table.ht_table[idx] = hdr; hdr->b_flags |= ARC_FLAG_IN_HASH_TABLE; /* collect some hash table performance data */ if (i > 0) { ARCSTAT_BUMP(arcstat_hash_collisions); if (i == 1) ARCSTAT_BUMP(arcstat_hash_chains); ARCSTAT_MAX(arcstat_hash_chain_max, i); } ARCSTAT_BUMP(arcstat_hash_elements); ARCSTAT_MAXSTAT(arcstat_hash_elements); return (NULL); } static void buf_hash_remove(arc_buf_hdr_t *hdr) { arc_buf_hdr_t *fhdr, **hdrp; uint64_t idx = BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth); ASSERT(MUTEX_HELD(BUF_HASH_LOCK(idx))); ASSERT(HDR_IN_HASH_TABLE(hdr)); hdrp = &buf_hash_table.ht_table[idx]; while ((fhdr = *hdrp) != hdr) { ASSERT(fhdr != NULL); hdrp = &fhdr->b_hash_next; } *hdrp = hdr->b_hash_next; hdr->b_hash_next = NULL; hdr->b_flags &= ~ARC_FLAG_IN_HASH_TABLE; /* collect some hash table performance data */ ARCSTAT_BUMPDOWN(arcstat_hash_elements); if (buf_hash_table.ht_table[idx] && buf_hash_table.ht_table[idx]->b_hash_next == NULL) ARCSTAT_BUMPDOWN(arcstat_hash_chains); } /* * Global data structures and functions for the buf kmem cache. */ static kmem_cache_t *hdr_full_cache; static kmem_cache_t *hdr_l2only_cache; static kmem_cache_t *buf_cache; static void buf_fini(void) { int i; kmem_free(buf_hash_table.ht_table, (buf_hash_table.ht_mask + 1) * sizeof (void *)); for (i = 0; i < BUF_LOCKS; i++) mutex_destroy(&buf_hash_table.ht_locks[i].ht_lock); kmem_cache_destroy(hdr_full_cache); kmem_cache_destroy(hdr_l2only_cache); kmem_cache_destroy(buf_cache); } /* * Constructor callback - called when the cache is empty * and a new buf is requested. */ /* ARGSUSED */ static int hdr_full_cons(void *vbuf, void *unused, int kmflag) { arc_buf_hdr_t *hdr = vbuf; bzero(hdr, HDR_FULL_SIZE); cv_init(&hdr->b_l1hdr.b_cv, NULL, CV_DEFAULT, NULL); refcount_create(&hdr->b_l1hdr.b_refcnt); mutex_init(&hdr->b_l1hdr.b_freeze_lock, NULL, MUTEX_DEFAULT, NULL); multilist_link_init(&hdr->b_l1hdr.b_arc_node); arc_space_consume(HDR_FULL_SIZE, ARC_SPACE_HDRS); return (0); } /* ARGSUSED */ static int hdr_l2only_cons(void *vbuf, void *unused, int kmflag) { arc_buf_hdr_t *hdr = vbuf; bzero(hdr, HDR_L2ONLY_SIZE); arc_space_consume(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS); return (0); } /* ARGSUSED */ static int buf_cons(void *vbuf, void *unused, int kmflag) { arc_buf_t *buf = vbuf; bzero(buf, sizeof (arc_buf_t)); mutex_init(&buf->b_evict_lock, NULL, MUTEX_DEFAULT, NULL); arc_space_consume(sizeof (arc_buf_t), ARC_SPACE_HDRS); return (0); } /* * Destructor callback - called when a cached buf is * no longer required. */ /* ARGSUSED */ static void hdr_full_dest(void *vbuf, void *unused) { arc_buf_hdr_t *hdr = vbuf; ASSERT(BUF_EMPTY(hdr)); cv_destroy(&hdr->b_l1hdr.b_cv); refcount_destroy(&hdr->b_l1hdr.b_refcnt); mutex_destroy(&hdr->b_l1hdr.b_freeze_lock); ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); arc_space_return(HDR_FULL_SIZE, ARC_SPACE_HDRS); } /* ARGSUSED */ static void hdr_l2only_dest(void *vbuf, void *unused) { arc_buf_hdr_t *hdr = vbuf; ASSERT(BUF_EMPTY(hdr)); arc_space_return(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS); } /* ARGSUSED */ static void buf_dest(void *vbuf, void *unused) { arc_buf_t *buf = vbuf; mutex_destroy(&buf->b_evict_lock); arc_space_return(sizeof (arc_buf_t), ARC_SPACE_HDRS); } /* * Reclaim callback -- invoked when memory is low. */ /* ARGSUSED */ static void hdr_recl(void *unused) { dprintf("hdr_recl called\n"); /* * umem calls the reclaim func when we destroy the buf cache, * which is after we do arc_fini(). */ if (!arc_dead) cv_signal(&arc_reclaim_thread_cv); } static void buf_init(void) { uint64_t *ct; uint64_t hsize = 1ULL << 12; int i, j; /* * The hash table is big enough to fill all of physical memory * with an average block size of zfs_arc_average_blocksize (default 8K). * By default, the table will take up * totalmem * sizeof(void*) / 8K (1MB per GB with 8-byte pointers). */ while (hsize * zfs_arc_average_blocksize < (uint64_t)physmem * PAGESIZE) hsize <<= 1; retry: buf_hash_table.ht_mask = hsize - 1; buf_hash_table.ht_table = kmem_zalloc(hsize * sizeof (void*), KM_NOSLEEP); if (buf_hash_table.ht_table == NULL) { ASSERT(hsize > (1ULL << 8)); hsize >>= 1; goto retry; } hdr_full_cache = kmem_cache_create("arc_buf_hdr_t_full", HDR_FULL_SIZE, 0, hdr_full_cons, hdr_full_dest, hdr_recl, NULL, NULL, 0); hdr_l2only_cache = kmem_cache_create("arc_buf_hdr_t_l2only", HDR_L2ONLY_SIZE, 0, hdr_l2only_cons, hdr_l2only_dest, hdr_recl, NULL, NULL, 0); buf_cache = kmem_cache_create("arc_buf_t", sizeof (arc_buf_t), 0, buf_cons, buf_dest, NULL, NULL, NULL, 0); for (i = 0; i < 256; i++) for (ct = zfs_crc64_table + i, *ct = i, j = 8; j > 0; j--) *ct = (*ct >> 1) ^ (-(*ct & 1) & ZFS_CRC64_POLY); for (i = 0; i < BUF_LOCKS; i++) { mutex_init(&buf_hash_table.ht_locks[i].ht_lock, NULL, MUTEX_DEFAULT, NULL); } } /* * Transition between the two allocation states for the arc_buf_hdr struct. * The arc_buf_hdr struct can be allocated with (hdr_full_cache) or without * (hdr_l2only_cache) the fields necessary for the L1 cache - the smaller * version is used when a cache buffer is only in the L2ARC in order to reduce * memory usage. */ static arc_buf_hdr_t * arc_hdr_realloc(arc_buf_hdr_t *hdr, kmem_cache_t *old, kmem_cache_t *new) { ASSERT(HDR_HAS_L2HDR(hdr)); arc_buf_hdr_t *nhdr; l2arc_dev_t *dev = hdr->b_l2hdr.b_dev; ASSERT((old == hdr_full_cache && new == hdr_l2only_cache) || (old == hdr_l2only_cache && new == hdr_full_cache)); nhdr = kmem_cache_alloc(new, KM_PUSHPAGE); ASSERT(MUTEX_HELD(HDR_LOCK(hdr))); buf_hash_remove(hdr); bcopy(hdr, nhdr, HDR_L2ONLY_SIZE); if (new == hdr_full_cache) { nhdr->b_flags |= ARC_FLAG_HAS_L1HDR; /* * arc_access and arc_change_state need to be aware that a * header has just come out of L2ARC, so we set its state to * l2c_only even though it's about to change. */ nhdr->b_l1hdr.b_state = arc_l2c_only; /* Verify previous threads set to NULL before freeing */ ASSERT3P(nhdr->b_l1hdr.b_tmp_cdata, ==, NULL); } else { ASSERT(hdr->b_l1hdr.b_buf == NULL); ASSERT0(hdr->b_l1hdr.b_datacnt); /* * If we've reached here, We must have been called from * arc_evict_hdr(), as such we should have already been * removed from any ghost list we were previously on * (which protects us from racing with arc_evict_state), * thus no locking is needed during this check. */ ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); /* * A buffer must not be moved into the arc_l2c_only * state if it's not finished being written out to the * l2arc device. Otherwise, the b_l1hdr.b_tmp_cdata field * might try to be accessed, even though it was removed. */ VERIFY(!HDR_L2_WRITING(hdr)); VERIFY3P(hdr->b_l1hdr.b_tmp_cdata, ==, NULL); #ifdef ZFS_DEBUG if (hdr->b_l1hdr.b_thawed != NULL) { kmem_free(hdr->b_l1hdr.b_thawed, 1); hdr->b_l1hdr.b_thawed = NULL; } #endif nhdr->b_flags &= ~ARC_FLAG_HAS_L1HDR; } /* * The header has been reallocated so we need to re-insert it into any * lists it was on. */ (void) buf_hash_insert(nhdr, NULL); ASSERT(list_link_active(&hdr->b_l2hdr.b_l2node)); mutex_enter(&dev->l2ad_mtx); /* * We must place the realloc'ed header back into the list at * the same spot. Otherwise, if it's placed earlier in the list, * l2arc_write_buffers() could find it during the function's * write phase, and try to write it out to the l2arc. */ list_insert_after(&dev->l2ad_buflist, hdr, nhdr); list_remove(&dev->l2ad_buflist, hdr); mutex_exit(&dev->l2ad_mtx); /* * Since we're using the pointer address as the tag when * incrementing and decrementing the l2ad_alloc refcount, we * must remove the old pointer (that we're about to destroy) and * add the new pointer to the refcount. Otherwise we'd remove * the wrong pointer address when calling arc_hdr_destroy() later. */ (void) refcount_remove_many(&dev->l2ad_alloc, hdr->b_l2hdr.b_asize, hdr); (void) refcount_add_many(&dev->l2ad_alloc, nhdr->b_l2hdr.b_asize, nhdr); buf_discard_identity(hdr); hdr->b_freeze_cksum = NULL; kmem_cache_free(old, hdr); return (nhdr); } #define ARC_MINTIME (hz>>4) /* 62 ms */ static void arc_cksum_verify(arc_buf_t *buf) { zio_cksum_t zc; if (!(zfs_flags & ZFS_DEBUG_MODIFY)) return; mutex_enter(&buf->b_hdr->b_l1hdr.b_freeze_lock); if (buf->b_hdr->b_freeze_cksum == NULL || HDR_IO_ERROR(buf->b_hdr)) { mutex_exit(&buf->b_hdr->b_l1hdr.b_freeze_lock); return; } - fletcher_2_native(buf->b_data, buf->b_hdr->b_size, &zc); + fletcher_2_native(buf->b_data, buf->b_hdr->b_size, NULL, &zc); if (!ZIO_CHECKSUM_EQUAL(*buf->b_hdr->b_freeze_cksum, zc)) panic("buffer modified while frozen!"); mutex_exit(&buf->b_hdr->b_l1hdr.b_freeze_lock); } static int arc_cksum_equal(arc_buf_t *buf) { zio_cksum_t zc; int equal; mutex_enter(&buf->b_hdr->b_l1hdr.b_freeze_lock); - fletcher_2_native(buf->b_data, buf->b_hdr->b_size, &zc); + fletcher_2_native(buf->b_data, buf->b_hdr->b_size, NULL, &zc); equal = ZIO_CHECKSUM_EQUAL(*buf->b_hdr->b_freeze_cksum, zc); mutex_exit(&buf->b_hdr->b_l1hdr.b_freeze_lock); return (equal); } static void arc_cksum_compute(arc_buf_t *buf, boolean_t force) { if (!force && !(zfs_flags & ZFS_DEBUG_MODIFY)) return; mutex_enter(&buf->b_hdr->b_l1hdr.b_freeze_lock); if (buf->b_hdr->b_freeze_cksum != NULL) { mutex_exit(&buf->b_hdr->b_l1hdr.b_freeze_lock); return; } buf->b_hdr->b_freeze_cksum = kmem_alloc(sizeof (zio_cksum_t), KM_SLEEP); fletcher_2_native(buf->b_data, buf->b_hdr->b_size, - buf->b_hdr->b_freeze_cksum); + NULL, buf->b_hdr->b_freeze_cksum); mutex_exit(&buf->b_hdr->b_l1hdr.b_freeze_lock); #ifdef illumos arc_buf_watch(buf); #endif /* illumos */ } #ifdef illumos #ifndef _KERNEL typedef struct procctl { long cmd; prwatch_t prwatch; } procctl_t; #endif /* ARGSUSED */ static void arc_buf_unwatch(arc_buf_t *buf) { #ifndef _KERNEL if (arc_watch) { int result; procctl_t ctl; ctl.cmd = PCWATCH; ctl.prwatch.pr_vaddr = (uintptr_t)buf->b_data; ctl.prwatch.pr_size = 0; ctl.prwatch.pr_wflags = 0; result = write(arc_procfd, &ctl, sizeof (ctl)); ASSERT3U(result, ==, sizeof (ctl)); } #endif } /* ARGSUSED */ static void arc_buf_watch(arc_buf_t *buf) { #ifndef _KERNEL if (arc_watch) { int result; procctl_t ctl; ctl.cmd = PCWATCH; ctl.prwatch.pr_vaddr = (uintptr_t)buf->b_data; ctl.prwatch.pr_size = buf->b_hdr->b_size; ctl.prwatch.pr_wflags = WA_WRITE; result = write(arc_procfd, &ctl, sizeof (ctl)); ASSERT3U(result, ==, sizeof (ctl)); } #endif } #endif /* illumos */ static arc_buf_contents_t arc_buf_type(arc_buf_hdr_t *hdr) { if (HDR_ISTYPE_METADATA(hdr)) { return (ARC_BUFC_METADATA); } else { return (ARC_BUFC_DATA); } } static uint32_t arc_bufc_to_flags(arc_buf_contents_t type) { switch (type) { case ARC_BUFC_DATA: /* metadata field is 0 if buffer contains normal data */ return (0); case ARC_BUFC_METADATA: return (ARC_FLAG_BUFC_METADATA); default: break; } panic("undefined ARC buffer type!"); return ((uint32_t)-1); } void arc_buf_thaw(arc_buf_t *buf) { if (zfs_flags & ZFS_DEBUG_MODIFY) { if (buf->b_hdr->b_l1hdr.b_state != arc_anon) panic("modifying non-anon buffer!"); if (HDR_IO_IN_PROGRESS(buf->b_hdr)) panic("modifying buffer while i/o in progress!"); arc_cksum_verify(buf); } mutex_enter(&buf->b_hdr->b_l1hdr.b_freeze_lock); if (buf->b_hdr->b_freeze_cksum != NULL) { kmem_free(buf->b_hdr->b_freeze_cksum, sizeof (zio_cksum_t)); buf->b_hdr->b_freeze_cksum = NULL; } #ifdef ZFS_DEBUG if (zfs_flags & ZFS_DEBUG_MODIFY) { if (buf->b_hdr->b_l1hdr.b_thawed != NULL) kmem_free(buf->b_hdr->b_l1hdr.b_thawed, 1); buf->b_hdr->b_l1hdr.b_thawed = kmem_alloc(1, KM_SLEEP); } #endif mutex_exit(&buf->b_hdr->b_l1hdr.b_freeze_lock); #ifdef illumos arc_buf_unwatch(buf); #endif /* illumos */ } void arc_buf_freeze(arc_buf_t *buf) { kmutex_t *hash_lock; if (!(zfs_flags & ZFS_DEBUG_MODIFY)) return; hash_lock = HDR_LOCK(buf->b_hdr); mutex_enter(hash_lock); ASSERT(buf->b_hdr->b_freeze_cksum != NULL || buf->b_hdr->b_l1hdr.b_state == arc_anon); arc_cksum_compute(buf, B_FALSE); mutex_exit(hash_lock); } static void add_reference(arc_buf_hdr_t *hdr, kmutex_t *hash_lock, void *tag) { ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(MUTEX_HELD(hash_lock)); arc_state_t *state = hdr->b_l1hdr.b_state; if ((refcount_add(&hdr->b_l1hdr.b_refcnt, tag) == 1) && (state != arc_anon)) { /* We don't use the L2-only state list. */ if (state != arc_l2c_only) { arc_buf_contents_t type = arc_buf_type(hdr); uint64_t delta = hdr->b_size * hdr->b_l1hdr.b_datacnt; multilist_t *list = &state->arcs_list[type]; uint64_t *size = &state->arcs_lsize[type]; multilist_remove(list, hdr); if (GHOST_STATE(state)) { ASSERT0(hdr->b_l1hdr.b_datacnt); ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); delta = hdr->b_size; } ASSERT(delta > 0); ASSERT3U(*size, >=, delta); atomic_add_64(size, -delta); } /* remove the prefetch flag if we get a reference */ hdr->b_flags &= ~ARC_FLAG_PREFETCH; } } static int remove_reference(arc_buf_hdr_t *hdr, kmutex_t *hash_lock, void *tag) { int cnt; arc_state_t *state = hdr->b_l1hdr.b_state; ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(state == arc_anon || MUTEX_HELD(hash_lock)); ASSERT(!GHOST_STATE(state)); /* * arc_l2c_only counts as a ghost state so we don't need to explicitly * check to prevent usage of the arc_l2c_only list. */ if (((cnt = refcount_remove(&hdr->b_l1hdr.b_refcnt, tag)) == 0) && (state != arc_anon)) { arc_buf_contents_t type = arc_buf_type(hdr); multilist_t *list = &state->arcs_list[type]; uint64_t *size = &state->arcs_lsize[type]; multilist_insert(list, hdr); ASSERT(hdr->b_l1hdr.b_datacnt > 0); atomic_add_64(size, hdr->b_size * hdr->b_l1hdr.b_datacnt); } return (cnt); } /* * Move the supplied buffer to the indicated state. The hash lock * for the buffer must be held by the caller. */ static void arc_change_state(arc_state_t *new_state, arc_buf_hdr_t *hdr, kmutex_t *hash_lock) { arc_state_t *old_state; int64_t refcnt; uint32_t datacnt; uint64_t from_delta, to_delta; arc_buf_contents_t buftype = arc_buf_type(hdr); /* * We almost always have an L1 hdr here, since we call arc_hdr_realloc() * in arc_read() when bringing a buffer out of the L2ARC. However, the * L1 hdr doesn't always exist when we change state to arc_anon before * destroying a header, in which case reallocating to add the L1 hdr is * pointless. */ if (HDR_HAS_L1HDR(hdr)) { old_state = hdr->b_l1hdr.b_state; refcnt = refcount_count(&hdr->b_l1hdr.b_refcnt); datacnt = hdr->b_l1hdr.b_datacnt; } else { old_state = arc_l2c_only; refcnt = 0; datacnt = 0; } ASSERT(MUTEX_HELD(hash_lock)); ASSERT3P(new_state, !=, old_state); ASSERT(refcnt == 0 || datacnt > 0); ASSERT(!GHOST_STATE(new_state) || datacnt == 0); ASSERT(old_state != arc_anon || datacnt <= 1); from_delta = to_delta = datacnt * hdr->b_size; /* * If this buffer is evictable, transfer it from the * old state list to the new state list. */ if (refcnt == 0) { if (old_state != arc_anon && old_state != arc_l2c_only) { uint64_t *size = &old_state->arcs_lsize[buftype]; ASSERT(HDR_HAS_L1HDR(hdr)); multilist_remove(&old_state->arcs_list[buftype], hdr); /* * If prefetching out of the ghost cache, * we will have a non-zero datacnt. */ if (GHOST_STATE(old_state) && datacnt == 0) { /* ghost elements have a ghost size */ ASSERT(hdr->b_l1hdr.b_buf == NULL); from_delta = hdr->b_size; } ASSERT3U(*size, >=, from_delta); atomic_add_64(size, -from_delta); } if (new_state != arc_anon && new_state != arc_l2c_only) { uint64_t *size = &new_state->arcs_lsize[buftype]; /* * An L1 header always exists here, since if we're * moving to some L1-cached state (i.e. not l2c_only or * anonymous), we realloc the header to add an L1hdr * beforehand. */ ASSERT(HDR_HAS_L1HDR(hdr)); multilist_insert(&new_state->arcs_list[buftype], hdr); /* ghost elements have a ghost size */ if (GHOST_STATE(new_state)) { ASSERT0(datacnt); ASSERT(hdr->b_l1hdr.b_buf == NULL); to_delta = hdr->b_size; } atomic_add_64(size, to_delta); } } ASSERT(!BUF_EMPTY(hdr)); if (new_state == arc_anon && HDR_IN_HASH_TABLE(hdr)) buf_hash_remove(hdr); /* adjust state sizes (ignore arc_l2c_only) */ if (to_delta && new_state != arc_l2c_only) { ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(new_state)) { ASSERT0(datacnt); /* * We moving a header to a ghost state, we first * remove all arc buffers. Thus, we'll have a * datacnt of zero, and no arc buffer to use for * the reference. As a result, we use the arc * header pointer for the reference. */ (void) refcount_add_many(&new_state->arcs_size, hdr->b_size, hdr); } else { ASSERT3U(datacnt, !=, 0); /* * Each individual buffer holds a unique reference, * thus we must remove each of these references one * at a time. */ for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) { (void) refcount_add_many(&new_state->arcs_size, hdr->b_size, buf); } } } if (from_delta && old_state != arc_l2c_only) { ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(old_state)) { /* * When moving a header off of a ghost state, * there's the possibility for datacnt to be * non-zero. This is because we first add the * arc buffer to the header prior to changing * the header's state. Since we used the header * for the reference when putting the header on * the ghost state, we must balance that and use * the header when removing off the ghost state * (even though datacnt is non zero). */ IMPLY(datacnt == 0, new_state == arc_anon || new_state == arc_l2c_only); (void) refcount_remove_many(&old_state->arcs_size, hdr->b_size, hdr); } else { ASSERT3P(datacnt, !=, 0); /* * Each individual buffer holds a unique reference, * thus we must remove each of these references one * at a time. */ for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) { (void) refcount_remove_many( &old_state->arcs_size, hdr->b_size, buf); } } } if (HDR_HAS_L1HDR(hdr)) hdr->b_l1hdr.b_state = new_state; /* * L2 headers should never be on the L2 state list since they don't * have L1 headers allocated. */ ASSERT(multilist_is_empty(&arc_l2c_only->arcs_list[ARC_BUFC_DATA]) && multilist_is_empty(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA])); } void arc_space_consume(uint64_t space, arc_space_type_t type) { ASSERT(type >= 0 && type < ARC_SPACE_NUMTYPES); switch (type) { case ARC_SPACE_DATA: ARCSTAT_INCR(arcstat_data_size, space); break; case ARC_SPACE_META: ARCSTAT_INCR(arcstat_metadata_size, space); break; case ARC_SPACE_OTHER: ARCSTAT_INCR(arcstat_other_size, space); break; case ARC_SPACE_HDRS: ARCSTAT_INCR(arcstat_hdr_size, space); break; case ARC_SPACE_L2HDRS: ARCSTAT_INCR(arcstat_l2_hdr_size, space); break; } if (type != ARC_SPACE_DATA) ARCSTAT_INCR(arcstat_meta_used, space); atomic_add_64(&arc_size, space); } void arc_space_return(uint64_t space, arc_space_type_t type) { ASSERT(type >= 0 && type < ARC_SPACE_NUMTYPES); switch (type) { case ARC_SPACE_DATA: ARCSTAT_INCR(arcstat_data_size, -space); break; case ARC_SPACE_META: ARCSTAT_INCR(arcstat_metadata_size, -space); break; case ARC_SPACE_OTHER: ARCSTAT_INCR(arcstat_other_size, -space); break; case ARC_SPACE_HDRS: ARCSTAT_INCR(arcstat_hdr_size, -space); break; case ARC_SPACE_L2HDRS: ARCSTAT_INCR(arcstat_l2_hdr_size, -space); break; } if (type != ARC_SPACE_DATA) { ASSERT(arc_meta_used >= space); if (arc_meta_max < arc_meta_used) arc_meta_max = arc_meta_used; ARCSTAT_INCR(arcstat_meta_used, -space); } ASSERT(arc_size >= space); atomic_add_64(&arc_size, -space); } arc_buf_t * arc_buf_alloc(spa_t *spa, int32_t size, void *tag, arc_buf_contents_t type) { arc_buf_hdr_t *hdr; arc_buf_t *buf; ASSERT3U(size, >, 0); hdr = kmem_cache_alloc(hdr_full_cache, KM_PUSHPAGE); ASSERT(BUF_EMPTY(hdr)); ASSERT3P(hdr->b_freeze_cksum, ==, NULL); hdr->b_size = size; hdr->b_spa = spa_load_guid(spa); buf = kmem_cache_alloc(buf_cache, KM_PUSHPAGE); buf->b_hdr = hdr; buf->b_data = NULL; buf->b_efunc = NULL; buf->b_private = NULL; buf->b_next = NULL; hdr->b_flags = arc_bufc_to_flags(type); hdr->b_flags |= ARC_FLAG_HAS_L1HDR; hdr->b_l1hdr.b_buf = buf; hdr->b_l1hdr.b_state = arc_anon; hdr->b_l1hdr.b_arc_access = 0; hdr->b_l1hdr.b_datacnt = 1; hdr->b_l1hdr.b_tmp_cdata = NULL; arc_get_data_buf(buf); ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); (void) refcount_add(&hdr->b_l1hdr.b_refcnt, tag); return (buf); } static char *arc_onloan_tag = "onloan"; /* * Loan out an anonymous arc buffer. Loaned buffers are not counted as in * flight data by arc_tempreserve_space() until they are "returned". Loaned * buffers must be returned to the arc before they can be used by the DMU or * freed. */ arc_buf_t * arc_loan_buf(spa_t *spa, int size) { arc_buf_t *buf; buf = arc_buf_alloc(spa, size, arc_onloan_tag, ARC_BUFC_DATA); atomic_add_64(&arc_loaned_bytes, size); return (buf); } /* * Return a loaned arc buffer to the arc. */ void arc_return_buf(arc_buf_t *buf, void *tag) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT(buf->b_data != NULL); ASSERT(HDR_HAS_L1HDR(hdr)); (void) refcount_add(&hdr->b_l1hdr.b_refcnt, tag); (void) refcount_remove(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag); atomic_add_64(&arc_loaned_bytes, -hdr->b_size); } /* Detach an arc_buf from a dbuf (tag) */ void arc_loan_inuse_buf(arc_buf_t *buf, void *tag) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT(buf->b_data != NULL); ASSERT(HDR_HAS_L1HDR(hdr)); (void) refcount_add(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag); (void) refcount_remove(&hdr->b_l1hdr.b_refcnt, tag); buf->b_efunc = NULL; buf->b_private = NULL; atomic_add_64(&arc_loaned_bytes, hdr->b_size); } static arc_buf_t * arc_buf_clone(arc_buf_t *from) { arc_buf_t *buf; arc_buf_hdr_t *hdr = from->b_hdr; uint64_t size = hdr->b_size; ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(hdr->b_l1hdr.b_state != arc_anon); buf = kmem_cache_alloc(buf_cache, KM_PUSHPAGE); buf->b_hdr = hdr; buf->b_data = NULL; buf->b_efunc = NULL; buf->b_private = NULL; buf->b_next = hdr->b_l1hdr.b_buf; hdr->b_l1hdr.b_buf = buf; arc_get_data_buf(buf); bcopy(from->b_data, buf->b_data, size); /* * This buffer already exists in the arc so create a duplicate * copy for the caller. If the buffer is associated with user data * then track the size and number of duplicates. These stats will be * updated as duplicate buffers are created and destroyed. */ if (HDR_ISTYPE_DATA(hdr)) { ARCSTAT_BUMP(arcstat_duplicate_buffers); ARCSTAT_INCR(arcstat_duplicate_buffers_size, size); } hdr->b_l1hdr.b_datacnt += 1; return (buf); } void arc_buf_add_ref(arc_buf_t *buf, void* tag) { arc_buf_hdr_t *hdr; kmutex_t *hash_lock; /* * Check to see if this buffer is evicted. Callers * must verify b_data != NULL to know if the add_ref * was successful. */ mutex_enter(&buf->b_evict_lock); if (buf->b_data == NULL) { mutex_exit(&buf->b_evict_lock); return; } hash_lock = HDR_LOCK(buf->b_hdr); mutex_enter(hash_lock); hdr = buf->b_hdr; ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT3P(hash_lock, ==, HDR_LOCK(hdr)); mutex_exit(&buf->b_evict_lock); ASSERT(hdr->b_l1hdr.b_state == arc_mru || hdr->b_l1hdr.b_state == arc_mfu); add_reference(hdr, hash_lock, tag); DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr); arc_access(hdr, hash_lock); mutex_exit(hash_lock); ARCSTAT_BUMP(arcstat_hits); ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr), demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data, metadata, hits); } static void arc_buf_free_on_write(void *data, size_t size, void (*free_func)(void *, size_t)) { l2arc_data_free_t *df; df = kmem_alloc(sizeof (*df), KM_SLEEP); df->l2df_data = data; df->l2df_size = size; df->l2df_func = free_func; mutex_enter(&l2arc_free_on_write_mtx); list_insert_head(l2arc_free_on_write, df); mutex_exit(&l2arc_free_on_write_mtx); } /* * Free the arc data buffer. If it is an l2arc write in progress, * the buffer is placed on l2arc_free_on_write to be freed later. */ static void arc_buf_data_free(arc_buf_t *buf, void (*free_func)(void *, size_t)) { arc_buf_hdr_t *hdr = buf->b_hdr; if (HDR_L2_WRITING(hdr)) { arc_buf_free_on_write(buf->b_data, hdr->b_size, free_func); ARCSTAT_BUMP(arcstat_l2_free_on_write); } else { free_func(buf->b_data, hdr->b_size); } } static void arc_buf_l2_cdata_free(arc_buf_hdr_t *hdr) { ASSERT(HDR_HAS_L2HDR(hdr)); ASSERT(MUTEX_HELD(&hdr->b_l2hdr.b_dev->l2ad_mtx)); /* * The b_tmp_cdata field is linked off of the b_l1hdr, so if * that doesn't exist, the header is in the arc_l2c_only state, * and there isn't anything to free (it's already been freed). */ if (!HDR_HAS_L1HDR(hdr)) return; /* * The header isn't being written to the l2arc device, thus it * shouldn't have a b_tmp_cdata to free. */ if (!HDR_L2_WRITING(hdr)) { ASSERT3P(hdr->b_l1hdr.b_tmp_cdata, ==, NULL); return; } /* * The header does not have compression enabled. This can be due * to the buffer not being compressible, or because we're * freeing the buffer before the second phase of * l2arc_write_buffer() has started (which does the compression * step). In either case, b_tmp_cdata does not point to a * separately compressed buffer, so there's nothing to free (it * points to the same buffer as the arc_buf_t's b_data field). */ if (hdr->b_l2hdr.b_compress == ZIO_COMPRESS_OFF) { hdr->b_l1hdr.b_tmp_cdata = NULL; return; } /* * There's nothing to free since the buffer was all zero's and * compressed to a zero length buffer. */ if (hdr->b_l2hdr.b_compress == ZIO_COMPRESS_EMPTY) { ASSERT3P(hdr->b_l1hdr.b_tmp_cdata, ==, NULL); return; } ASSERT(L2ARC_IS_VALID_COMPRESS(hdr->b_l2hdr.b_compress)); arc_buf_free_on_write(hdr->b_l1hdr.b_tmp_cdata, hdr->b_size, zio_data_buf_free); ARCSTAT_BUMP(arcstat_l2_cdata_free_on_write); hdr->b_l1hdr.b_tmp_cdata = NULL; } /* * Free up buf->b_data and if 'remove' is set, then pull the * arc_buf_t off of the the arc_buf_hdr_t's list and free it. */ static void arc_buf_destroy(arc_buf_t *buf, boolean_t remove) { arc_buf_t **bufp; /* free up data associated with the buf */ if (buf->b_data != NULL) { arc_state_t *state = buf->b_hdr->b_l1hdr.b_state; uint64_t size = buf->b_hdr->b_size; arc_buf_contents_t type = arc_buf_type(buf->b_hdr); arc_cksum_verify(buf); #ifdef illumos arc_buf_unwatch(buf); #endif /* illumos */ if (type == ARC_BUFC_METADATA) { arc_buf_data_free(buf, zio_buf_free); arc_space_return(size, ARC_SPACE_META); } else { ASSERT(type == ARC_BUFC_DATA); arc_buf_data_free(buf, zio_data_buf_free); arc_space_return(size, ARC_SPACE_DATA); } /* protected by hash lock, if in the hash table */ if (multilist_link_active(&buf->b_hdr->b_l1hdr.b_arc_node)) { uint64_t *cnt = &state->arcs_lsize[type]; ASSERT(refcount_is_zero( &buf->b_hdr->b_l1hdr.b_refcnt)); ASSERT(state != arc_anon && state != arc_l2c_only); ASSERT3U(*cnt, >=, size); atomic_add_64(cnt, -size); } (void) refcount_remove_many(&state->arcs_size, size, buf); buf->b_data = NULL; /* * If we're destroying a duplicate buffer make sure * that the appropriate statistics are updated. */ if (buf->b_hdr->b_l1hdr.b_datacnt > 1 && HDR_ISTYPE_DATA(buf->b_hdr)) { ARCSTAT_BUMPDOWN(arcstat_duplicate_buffers); ARCSTAT_INCR(arcstat_duplicate_buffers_size, -size); } ASSERT(buf->b_hdr->b_l1hdr.b_datacnt > 0); buf->b_hdr->b_l1hdr.b_datacnt -= 1; } /* only remove the buf if requested */ if (!remove) return; /* remove the buf from the hdr list */ for (bufp = &buf->b_hdr->b_l1hdr.b_buf; *bufp != buf; bufp = &(*bufp)->b_next) continue; *bufp = buf->b_next; buf->b_next = NULL; ASSERT(buf->b_efunc == NULL); /* clean up the buf */ buf->b_hdr = NULL; kmem_cache_free(buf_cache, buf); } static void arc_hdr_l2hdr_destroy(arc_buf_hdr_t *hdr) { l2arc_buf_hdr_t *l2hdr = &hdr->b_l2hdr; l2arc_dev_t *dev = l2hdr->b_dev; ASSERT(MUTEX_HELD(&dev->l2ad_mtx)); ASSERT(HDR_HAS_L2HDR(hdr)); list_remove(&dev->l2ad_buflist, hdr); /* * We don't want to leak the b_tmp_cdata buffer that was * allocated in l2arc_write_buffers() */ arc_buf_l2_cdata_free(hdr); /* * If the l2hdr's b_daddr is equal to L2ARC_ADDR_UNSET, then * this header is being processed by l2arc_write_buffers() (i.e. * it's in the first stage of l2arc_write_buffers()). * Re-affirming that truth here, just to serve as a reminder. If * b_daddr does not equal L2ARC_ADDR_UNSET, then the header may or * may not have its HDR_L2_WRITING flag set. (the write may have * completed, in which case HDR_L2_WRITING will be false and the * b_daddr field will point to the address of the buffer on disk). */ IMPLY(l2hdr->b_daddr == L2ARC_ADDR_UNSET, HDR_L2_WRITING(hdr)); /* * If b_daddr is equal to L2ARC_ADDR_UNSET, we're racing with * l2arc_write_buffers(). Since we've just removed this header * from the l2arc buffer list, this header will never reach the * second stage of l2arc_write_buffers(), which increments the * accounting stats for this header. Thus, we must be careful * not to decrement them for this header either. */ if (l2hdr->b_daddr != L2ARC_ADDR_UNSET) { ARCSTAT_INCR(arcstat_l2_asize, -l2hdr->b_asize); ARCSTAT_INCR(arcstat_l2_size, -hdr->b_size); vdev_space_update(dev->l2ad_vdev, -l2hdr->b_asize, 0, 0); (void) refcount_remove_many(&dev->l2ad_alloc, l2hdr->b_asize, hdr); } hdr->b_flags &= ~ARC_FLAG_HAS_L2HDR; } static void arc_hdr_destroy(arc_buf_hdr_t *hdr) { if (HDR_HAS_L1HDR(hdr)) { ASSERT(hdr->b_l1hdr.b_buf == NULL || hdr->b_l1hdr.b_datacnt > 0); ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon); } ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT(!HDR_IN_HASH_TABLE(hdr)); if (HDR_HAS_L2HDR(hdr)) { l2arc_dev_t *dev = hdr->b_l2hdr.b_dev; boolean_t buflist_held = MUTEX_HELD(&dev->l2ad_mtx); if (!buflist_held) mutex_enter(&dev->l2ad_mtx); /* * Even though we checked this conditional above, we * need to check this again now that we have the * l2ad_mtx. This is because we could be racing with * another thread calling l2arc_evict() which might have * destroyed this header's L2 portion as we were waiting * to acquire the l2ad_mtx. If that happens, we don't * want to re-destroy the header's L2 portion. */ if (HDR_HAS_L2HDR(hdr)) { if (hdr->b_l2hdr.b_daddr != L2ARC_ADDR_UNSET) trim_map_free(dev->l2ad_vdev, hdr->b_l2hdr.b_daddr, hdr->b_l2hdr.b_asize, 0); arc_hdr_l2hdr_destroy(hdr); } if (!buflist_held) mutex_exit(&dev->l2ad_mtx); } if (!BUF_EMPTY(hdr)) buf_discard_identity(hdr); if (hdr->b_freeze_cksum != NULL) { kmem_free(hdr->b_freeze_cksum, sizeof (zio_cksum_t)); hdr->b_freeze_cksum = NULL; } if (HDR_HAS_L1HDR(hdr)) { while (hdr->b_l1hdr.b_buf) { arc_buf_t *buf = hdr->b_l1hdr.b_buf; if (buf->b_efunc != NULL) { mutex_enter(&arc_user_evicts_lock); mutex_enter(&buf->b_evict_lock); ASSERT(buf->b_hdr != NULL); arc_buf_destroy(hdr->b_l1hdr.b_buf, FALSE); hdr->b_l1hdr.b_buf = buf->b_next; buf->b_hdr = &arc_eviction_hdr; buf->b_next = arc_eviction_list; arc_eviction_list = buf; mutex_exit(&buf->b_evict_lock); cv_signal(&arc_user_evicts_cv); mutex_exit(&arc_user_evicts_lock); } else { arc_buf_destroy(hdr->b_l1hdr.b_buf, TRUE); } } #ifdef ZFS_DEBUG if (hdr->b_l1hdr.b_thawed != NULL) { kmem_free(hdr->b_l1hdr.b_thawed, 1); hdr->b_l1hdr.b_thawed = NULL; } #endif } ASSERT3P(hdr->b_hash_next, ==, NULL); if (HDR_HAS_L1HDR(hdr)) { ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); kmem_cache_free(hdr_full_cache, hdr); } else { kmem_cache_free(hdr_l2only_cache, hdr); } } void arc_buf_free(arc_buf_t *buf, void *tag) { arc_buf_hdr_t *hdr = buf->b_hdr; int hashed = hdr->b_l1hdr.b_state != arc_anon; ASSERT(buf->b_efunc == NULL); ASSERT(buf->b_data != NULL); if (hashed) { kmutex_t *hash_lock = HDR_LOCK(hdr); mutex_enter(hash_lock); hdr = buf->b_hdr; ASSERT3P(hash_lock, ==, HDR_LOCK(hdr)); (void) remove_reference(hdr, hash_lock, tag); if (hdr->b_l1hdr.b_datacnt > 1) { arc_buf_destroy(buf, TRUE); } else { ASSERT(buf == hdr->b_l1hdr.b_buf); ASSERT(buf->b_efunc == NULL); hdr->b_flags |= ARC_FLAG_BUF_AVAILABLE; } mutex_exit(hash_lock); } else if (HDR_IO_IN_PROGRESS(hdr)) { int destroy_hdr; /* * We are in the middle of an async write. Don't destroy * this buffer unless the write completes before we finish * decrementing the reference count. */ mutex_enter(&arc_user_evicts_lock); (void) remove_reference(hdr, NULL, tag); ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); destroy_hdr = !HDR_IO_IN_PROGRESS(hdr); mutex_exit(&arc_user_evicts_lock); if (destroy_hdr) arc_hdr_destroy(hdr); } else { if (remove_reference(hdr, NULL, tag) > 0) arc_buf_destroy(buf, TRUE); else arc_hdr_destroy(hdr); } } boolean_t arc_buf_remove_ref(arc_buf_t *buf, void* tag) { arc_buf_hdr_t *hdr = buf->b_hdr; kmutex_t *hash_lock = HDR_LOCK(hdr); boolean_t no_callback = (buf->b_efunc == NULL); if (hdr->b_l1hdr.b_state == arc_anon) { ASSERT(hdr->b_l1hdr.b_datacnt == 1); arc_buf_free(buf, tag); return (no_callback); } mutex_enter(hash_lock); hdr = buf->b_hdr; ASSERT(hdr->b_l1hdr.b_datacnt > 0); ASSERT3P(hash_lock, ==, HDR_LOCK(hdr)); ASSERT(hdr->b_l1hdr.b_state != arc_anon); ASSERT(buf->b_data != NULL); (void) remove_reference(hdr, hash_lock, tag); if (hdr->b_l1hdr.b_datacnt > 1) { if (no_callback) arc_buf_destroy(buf, TRUE); } else if (no_callback) { ASSERT(hdr->b_l1hdr.b_buf == buf && buf->b_next == NULL); ASSERT(buf->b_efunc == NULL); hdr->b_flags |= ARC_FLAG_BUF_AVAILABLE; } ASSERT(no_callback || hdr->b_l1hdr.b_datacnt > 1 || refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); mutex_exit(hash_lock); return (no_callback); } int32_t arc_buf_size(arc_buf_t *buf) { return (buf->b_hdr->b_size); } /* * Called from the DMU to determine if the current buffer should be * evicted. In order to ensure proper locking, the eviction must be initiated * from the DMU. Return true if the buffer is associated with user data and * duplicate buffers still exist. */ boolean_t arc_buf_eviction_needed(arc_buf_t *buf) { arc_buf_hdr_t *hdr; boolean_t evict_needed = B_FALSE; if (zfs_disable_dup_eviction) return (B_FALSE); mutex_enter(&buf->b_evict_lock); hdr = buf->b_hdr; if (hdr == NULL) { /* * We are in arc_do_user_evicts(); let that function * perform the eviction. */ ASSERT(buf->b_data == NULL); mutex_exit(&buf->b_evict_lock); return (B_FALSE); } else if (buf->b_data == NULL) { /* * We have already been added to the arc eviction list; * recommend eviction. */ ASSERT3P(hdr, ==, &arc_eviction_hdr); mutex_exit(&buf->b_evict_lock); return (B_TRUE); } if (hdr->b_l1hdr.b_datacnt > 1 && HDR_ISTYPE_DATA(hdr)) evict_needed = B_TRUE; mutex_exit(&buf->b_evict_lock); return (evict_needed); } /* * Evict the arc_buf_hdr that is provided as a parameter. The resultant * state of the header is dependent on it's state prior to entering this * function. The following transitions are possible: * * - arc_mru -> arc_mru_ghost * - arc_mfu -> arc_mfu_ghost * - arc_mru_ghost -> arc_l2c_only * - arc_mru_ghost -> deleted * - arc_mfu_ghost -> arc_l2c_only * - arc_mfu_ghost -> deleted */ static int64_t arc_evict_hdr(arc_buf_hdr_t *hdr, kmutex_t *hash_lock) { arc_state_t *evicted_state, *state; int64_t bytes_evicted = 0; ASSERT(MUTEX_HELD(hash_lock)); ASSERT(HDR_HAS_L1HDR(hdr)); state = hdr->b_l1hdr.b_state; if (GHOST_STATE(state)) { ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT(hdr->b_l1hdr.b_buf == NULL); /* * l2arc_write_buffers() relies on a header's L1 portion * (i.e. it's b_tmp_cdata field) during it's write phase. * Thus, we cannot push a header onto the arc_l2c_only * state (removing it's L1 piece) until the header is * done being written to the l2arc. */ if (HDR_HAS_L2HDR(hdr) && HDR_L2_WRITING(hdr)) { ARCSTAT_BUMP(arcstat_evict_l2_skip); return (bytes_evicted); } ARCSTAT_BUMP(arcstat_deleted); bytes_evicted += hdr->b_size; DTRACE_PROBE1(arc__delete, arc_buf_hdr_t *, hdr); if (HDR_HAS_L2HDR(hdr)) { /* * This buffer is cached on the 2nd Level ARC; * don't destroy the header. */ arc_change_state(arc_l2c_only, hdr, hash_lock); /* * dropping from L1+L2 cached to L2-only, * realloc to remove the L1 header. */ hdr = arc_hdr_realloc(hdr, hdr_full_cache, hdr_l2only_cache); } else { arc_change_state(arc_anon, hdr, hash_lock); arc_hdr_destroy(hdr); } return (bytes_evicted); } ASSERT(state == arc_mru || state == arc_mfu); evicted_state = (state == arc_mru) ? arc_mru_ghost : arc_mfu_ghost; /* prefetch buffers have a minimum lifespan */ if (HDR_IO_IN_PROGRESS(hdr) || ((hdr->b_flags & (ARC_FLAG_PREFETCH | ARC_FLAG_INDIRECT)) && ddi_get_lbolt() - hdr->b_l1hdr.b_arc_access < arc_min_prefetch_lifespan)) { ARCSTAT_BUMP(arcstat_evict_skip); return (bytes_evicted); } ASSERT0(refcount_count(&hdr->b_l1hdr.b_refcnt)); ASSERT3U(hdr->b_l1hdr.b_datacnt, >, 0); while (hdr->b_l1hdr.b_buf) { arc_buf_t *buf = hdr->b_l1hdr.b_buf; if (!mutex_tryenter(&buf->b_evict_lock)) { ARCSTAT_BUMP(arcstat_mutex_miss); break; } if (buf->b_data != NULL) bytes_evicted += hdr->b_size; if (buf->b_efunc != NULL) { mutex_enter(&arc_user_evicts_lock); arc_buf_destroy(buf, FALSE); hdr->b_l1hdr.b_buf = buf->b_next; buf->b_hdr = &arc_eviction_hdr; buf->b_next = arc_eviction_list; arc_eviction_list = buf; cv_signal(&arc_user_evicts_cv); mutex_exit(&arc_user_evicts_lock); mutex_exit(&buf->b_evict_lock); } else { mutex_exit(&buf->b_evict_lock); arc_buf_destroy(buf, TRUE); } } if (HDR_HAS_L2HDR(hdr)) { ARCSTAT_INCR(arcstat_evict_l2_cached, hdr->b_size); } else { if (l2arc_write_eligible(hdr->b_spa, hdr)) ARCSTAT_INCR(arcstat_evict_l2_eligible, hdr->b_size); else ARCSTAT_INCR(arcstat_evict_l2_ineligible, hdr->b_size); } if (hdr->b_l1hdr.b_datacnt == 0) { arc_change_state(evicted_state, hdr, hash_lock); ASSERT(HDR_IN_HASH_TABLE(hdr)); hdr->b_flags |= ARC_FLAG_IN_HASH_TABLE; hdr->b_flags &= ~ARC_FLAG_BUF_AVAILABLE; DTRACE_PROBE1(arc__evict, arc_buf_hdr_t *, hdr); } return (bytes_evicted); } static uint64_t arc_evict_state_impl(multilist_t *ml, int idx, arc_buf_hdr_t *marker, uint64_t spa, int64_t bytes) { multilist_sublist_t *mls; uint64_t bytes_evicted = 0; arc_buf_hdr_t *hdr; kmutex_t *hash_lock; int evict_count = 0; ASSERT3P(marker, !=, NULL); IMPLY(bytes < 0, bytes == ARC_EVICT_ALL); mls = multilist_sublist_lock(ml, idx); for (hdr = multilist_sublist_prev(mls, marker); hdr != NULL; hdr = multilist_sublist_prev(mls, marker)) { if ((bytes != ARC_EVICT_ALL && bytes_evicted >= bytes) || (evict_count >= zfs_arc_evict_batch_limit)) break; /* * To keep our iteration location, move the marker * forward. Since we're not holding hdr's hash lock, we * must be very careful and not remove 'hdr' from the * sublist. Otherwise, other consumers might mistake the * 'hdr' as not being on a sublist when they call the * multilist_link_active() function (they all rely on * the hash lock protecting concurrent insertions and * removals). multilist_sublist_move_forward() was * specifically implemented to ensure this is the case * (only 'marker' will be removed and re-inserted). */ multilist_sublist_move_forward(mls, marker); /* * The only case where the b_spa field should ever be * zero, is the marker headers inserted by * arc_evict_state(). It's possible for multiple threads * to be calling arc_evict_state() concurrently (e.g. * dsl_pool_close() and zio_inject_fault()), so we must * skip any markers we see from these other threads. */ if (hdr->b_spa == 0) continue; /* we're only interested in evicting buffers of a certain spa */ if (spa != 0 && hdr->b_spa != spa) { ARCSTAT_BUMP(arcstat_evict_skip); continue; } hash_lock = HDR_LOCK(hdr); /* * We aren't calling this function from any code path * that would already be holding a hash lock, so we're * asserting on this assumption to be defensive in case * this ever changes. Without this check, it would be * possible to incorrectly increment arcstat_mutex_miss * below (e.g. if the code changed such that we called * this function with a hash lock held). */ ASSERT(!MUTEX_HELD(hash_lock)); if (mutex_tryenter(hash_lock)) { uint64_t evicted = arc_evict_hdr(hdr, hash_lock); mutex_exit(hash_lock); bytes_evicted += evicted; /* * If evicted is zero, arc_evict_hdr() must have * decided to skip this header, don't increment * evict_count in this case. */ if (evicted != 0) evict_count++; /* * If arc_size isn't overflowing, signal any * threads that might happen to be waiting. * * For each header evicted, we wake up a single * thread. If we used cv_broadcast, we could * wake up "too many" threads causing arc_size * to significantly overflow arc_c; since * arc_get_data_buf() doesn't check for overflow * when it's woken up (it doesn't because it's * possible for the ARC to be overflowing while * full of un-evictable buffers, and the * function should proceed in this case). * * If threads are left sleeping, due to not * using cv_broadcast, they will be woken up * just before arc_reclaim_thread() sleeps. */ mutex_enter(&arc_reclaim_lock); if (!arc_is_overflowing()) cv_signal(&arc_reclaim_waiters_cv); mutex_exit(&arc_reclaim_lock); } else { ARCSTAT_BUMP(arcstat_mutex_miss); } } multilist_sublist_unlock(mls); return (bytes_evicted); } /* * Evict buffers from the given arc state, until we've removed the * specified number of bytes. Move the removed buffers to the * appropriate evict state. * * This function makes a "best effort". It skips over any buffers * it can't get a hash_lock on, and so, may not catch all candidates. * It may also return without evicting as much space as requested. * * If bytes is specified using the special value ARC_EVICT_ALL, this * will evict all available (i.e. unlocked and evictable) buffers from * the given arc state; which is used by arc_flush(). */ static uint64_t arc_evict_state(arc_state_t *state, uint64_t spa, int64_t bytes, arc_buf_contents_t type) { uint64_t total_evicted = 0; multilist_t *ml = &state->arcs_list[type]; int num_sublists; arc_buf_hdr_t **markers; IMPLY(bytes < 0, bytes == ARC_EVICT_ALL); num_sublists = multilist_get_num_sublists(ml); /* * If we've tried to evict from each sublist, made some * progress, but still have not hit the target number of bytes * to evict, we want to keep trying. The markers allow us to * pick up where we left off for each individual sublist, rather * than starting from the tail each time. */ markers = kmem_zalloc(sizeof (*markers) * num_sublists, KM_SLEEP); for (int i = 0; i < num_sublists; i++) { markers[i] = kmem_cache_alloc(hdr_full_cache, KM_SLEEP); /* * A b_spa of 0 is used to indicate that this header is * a marker. This fact is used in arc_adjust_type() and * arc_evict_state_impl(). */ markers[i]->b_spa = 0; multilist_sublist_t *mls = multilist_sublist_lock(ml, i); multilist_sublist_insert_tail(mls, markers[i]); multilist_sublist_unlock(mls); } /* * While we haven't hit our target number of bytes to evict, or * we're evicting all available buffers. */ while (total_evicted < bytes || bytes == ARC_EVICT_ALL) { /* * Start eviction using a randomly selected sublist, * this is to try and evenly balance eviction across all * sublists. Always starting at the same sublist * (e.g. index 0) would cause evictions to favor certain * sublists over others. */ int sublist_idx = multilist_get_random_index(ml); uint64_t scan_evicted = 0; for (int i = 0; i < num_sublists; i++) { uint64_t bytes_remaining; uint64_t bytes_evicted; if (bytes == ARC_EVICT_ALL) bytes_remaining = ARC_EVICT_ALL; else if (total_evicted < bytes) bytes_remaining = bytes - total_evicted; else break; bytes_evicted = arc_evict_state_impl(ml, sublist_idx, markers[sublist_idx], spa, bytes_remaining); scan_evicted += bytes_evicted; total_evicted += bytes_evicted; /* we've reached the end, wrap to the beginning */ if (++sublist_idx >= num_sublists) sublist_idx = 0; } /* * If we didn't evict anything during this scan, we have * no reason to believe we'll evict more during another * scan, so break the loop. */ if (scan_evicted == 0) { /* This isn't possible, let's make that obvious */ ASSERT3S(bytes, !=, 0); /* * When bytes is ARC_EVICT_ALL, the only way to * break the loop is when scan_evicted is zero. * In that case, we actually have evicted enough, * so we don't want to increment the kstat. */ if (bytes != ARC_EVICT_ALL) { ASSERT3S(total_evicted, <, bytes); ARCSTAT_BUMP(arcstat_evict_not_enough); } break; } } for (int i = 0; i < num_sublists; i++) { multilist_sublist_t *mls = multilist_sublist_lock(ml, i); multilist_sublist_remove(mls, markers[i]); multilist_sublist_unlock(mls); kmem_cache_free(hdr_full_cache, markers[i]); } kmem_free(markers, sizeof (*markers) * num_sublists); return (total_evicted); } /* * Flush all "evictable" data of the given type from the arc state * specified. This will not evict any "active" buffers (i.e. referenced). * * When 'retry' is set to FALSE, the function will make a single pass * over the state and evict any buffers that it can. Since it doesn't * continually retry the eviction, it might end up leaving some buffers * in the ARC due to lock misses. * * When 'retry' is set to TRUE, the function will continually retry the * eviction until *all* evictable buffers have been removed from the * state. As a result, if concurrent insertions into the state are * allowed (e.g. if the ARC isn't shutting down), this function might * wind up in an infinite loop, continually trying to evict buffers. */ static uint64_t arc_flush_state(arc_state_t *state, uint64_t spa, arc_buf_contents_t type, boolean_t retry) { uint64_t evicted = 0; while (state->arcs_lsize[type] != 0) { evicted += arc_evict_state(state, spa, ARC_EVICT_ALL, type); if (!retry) break; } return (evicted); } /* * Evict the specified number of bytes from the state specified, * restricting eviction to the spa and type given. This function * prevents us from trying to evict more from a state's list than * is "evictable", and to skip evicting altogether when passed a * negative value for "bytes". In contrast, arc_evict_state() will * evict everything it can, when passed a negative value for "bytes". */ static uint64_t arc_adjust_impl(arc_state_t *state, uint64_t spa, int64_t bytes, arc_buf_contents_t type) { int64_t delta; if (bytes > 0 && state->arcs_lsize[type] > 0) { delta = MIN(state->arcs_lsize[type], bytes); return (arc_evict_state(state, spa, delta, type)); } return (0); } /* * Evict metadata buffers from the cache, such that arc_meta_used is * capped by the arc_meta_limit tunable. */ static uint64_t arc_adjust_meta(void) { uint64_t total_evicted = 0; int64_t target; /* * If we're over the meta limit, we want to evict enough * metadata to get back under the meta limit. We don't want to * evict so much that we drop the MRU below arc_p, though. If * we're over the meta limit more than we're over arc_p, we * evict some from the MRU here, and some from the MFU below. */ target = MIN((int64_t)(arc_meta_used - arc_meta_limit), (int64_t)(refcount_count(&arc_anon->arcs_size) + refcount_count(&arc_mru->arcs_size) - arc_p)); total_evicted += arc_adjust_impl(arc_mru, 0, target, ARC_BUFC_METADATA); /* * Similar to the above, we want to evict enough bytes to get us * below the meta limit, but not so much as to drop us below the * space alloted to the MFU (which is defined as arc_c - arc_p). */ target = MIN((int64_t)(arc_meta_used - arc_meta_limit), (int64_t)(refcount_count(&arc_mfu->arcs_size) - (arc_c - arc_p))); total_evicted += arc_adjust_impl(arc_mfu, 0, target, ARC_BUFC_METADATA); return (total_evicted); } /* * Return the type of the oldest buffer in the given arc state * * This function will select a random sublist of type ARC_BUFC_DATA and * a random sublist of type ARC_BUFC_METADATA. The tail of each sublist * is compared, and the type which contains the "older" buffer will be * returned. */ static arc_buf_contents_t arc_adjust_type(arc_state_t *state) { multilist_t *data_ml = &state->arcs_list[ARC_BUFC_DATA]; multilist_t *meta_ml = &state->arcs_list[ARC_BUFC_METADATA]; int data_idx = multilist_get_random_index(data_ml); int meta_idx = multilist_get_random_index(meta_ml); multilist_sublist_t *data_mls; multilist_sublist_t *meta_mls; arc_buf_contents_t type; arc_buf_hdr_t *data_hdr; arc_buf_hdr_t *meta_hdr; /* * We keep the sublist lock until we're finished, to prevent * the headers from being destroyed via arc_evict_state(). */ data_mls = multilist_sublist_lock(data_ml, data_idx); meta_mls = multilist_sublist_lock(meta_ml, meta_idx); /* * These two loops are to ensure we skip any markers that * might be at the tail of the lists due to arc_evict_state(). */ for (data_hdr = multilist_sublist_tail(data_mls); data_hdr != NULL; data_hdr = multilist_sublist_prev(data_mls, data_hdr)) { if (data_hdr->b_spa != 0) break; } for (meta_hdr = multilist_sublist_tail(meta_mls); meta_hdr != NULL; meta_hdr = multilist_sublist_prev(meta_mls, meta_hdr)) { if (meta_hdr->b_spa != 0) break; } if (data_hdr == NULL && meta_hdr == NULL) { type = ARC_BUFC_DATA; } else if (data_hdr == NULL) { ASSERT3P(meta_hdr, !=, NULL); type = ARC_BUFC_METADATA; } else if (meta_hdr == NULL) { ASSERT3P(data_hdr, !=, NULL); type = ARC_BUFC_DATA; } else { ASSERT3P(data_hdr, !=, NULL); ASSERT3P(meta_hdr, !=, NULL); /* The headers can't be on the sublist without an L1 header */ ASSERT(HDR_HAS_L1HDR(data_hdr)); ASSERT(HDR_HAS_L1HDR(meta_hdr)); if (data_hdr->b_l1hdr.b_arc_access < meta_hdr->b_l1hdr.b_arc_access) { type = ARC_BUFC_DATA; } else { type = ARC_BUFC_METADATA; } } multilist_sublist_unlock(meta_mls); multilist_sublist_unlock(data_mls); return (type); } /* * Evict buffers from the cache, such that arc_size is capped by arc_c. */ static uint64_t arc_adjust(void) { uint64_t total_evicted = 0; uint64_t bytes; int64_t target; /* * If we're over arc_meta_limit, we want to correct that before * potentially evicting data buffers below. */ total_evicted += arc_adjust_meta(); /* * Adjust MRU size * * If we're over the target cache size, we want to evict enough * from the list to get back to our target size. We don't want * to evict too much from the MRU, such that it drops below * arc_p. So, if we're over our target cache size more than * the MRU is over arc_p, we'll evict enough to get back to * arc_p here, and then evict more from the MFU below. */ target = MIN((int64_t)(arc_size - arc_c), (int64_t)(refcount_count(&arc_anon->arcs_size) + refcount_count(&arc_mru->arcs_size) + arc_meta_used - arc_p)); /* * If we're below arc_meta_min, always prefer to evict data. * Otherwise, try to satisfy the requested number of bytes to * evict from the type which contains older buffers; in an * effort to keep newer buffers in the cache regardless of their * type. If we cannot satisfy the number of bytes from this * type, spill over into the next type. */ if (arc_adjust_type(arc_mru) == ARC_BUFC_METADATA && arc_meta_used > arc_meta_min) { bytes = arc_adjust_impl(arc_mru, 0, target, ARC_BUFC_METADATA); total_evicted += bytes; /* * If we couldn't evict our target number of bytes from * metadata, we try to get the rest from data. */ target -= bytes; total_evicted += arc_adjust_impl(arc_mru, 0, target, ARC_BUFC_DATA); } else { bytes = arc_adjust_impl(arc_mru, 0, target, ARC_BUFC_DATA); total_evicted += bytes; /* * If we couldn't evict our target number of bytes from * data, we try to get the rest from metadata. */ target -= bytes; total_evicted += arc_adjust_impl(arc_mru, 0, target, ARC_BUFC_METADATA); } /* * Adjust MFU size * * Now that we've tried to evict enough from the MRU to get its * size back to arc_p, if we're still above the target cache * size, we evict the rest from the MFU. */ target = arc_size - arc_c; if (arc_adjust_type(arc_mfu) == ARC_BUFC_METADATA && arc_meta_used > arc_meta_min) { bytes = arc_adjust_impl(arc_mfu, 0, target, ARC_BUFC_METADATA); total_evicted += bytes; /* * If we couldn't evict our target number of bytes from * metadata, we try to get the rest from data. */ target -= bytes; total_evicted += arc_adjust_impl(arc_mfu, 0, target, ARC_BUFC_DATA); } else { bytes = arc_adjust_impl(arc_mfu, 0, target, ARC_BUFC_DATA); total_evicted += bytes; /* * If we couldn't evict our target number of bytes from * data, we try to get the rest from data. */ target -= bytes; total_evicted += arc_adjust_impl(arc_mfu, 0, target, ARC_BUFC_METADATA); } /* * Adjust ghost lists * * In addition to the above, the ARC also defines target values * for the ghost lists. The sum of the mru list and mru ghost * list should never exceed the target size of the cache, and * the sum of the mru list, mfu list, mru ghost list, and mfu * ghost list should never exceed twice the target size of the * cache. The following logic enforces these limits on the ghost * caches, and evicts from them as needed. */ target = refcount_count(&arc_mru->arcs_size) + refcount_count(&arc_mru_ghost->arcs_size) - arc_c; bytes = arc_adjust_impl(arc_mru_ghost, 0, target, ARC_BUFC_DATA); total_evicted += bytes; target -= bytes; total_evicted += arc_adjust_impl(arc_mru_ghost, 0, target, ARC_BUFC_METADATA); /* * We assume the sum of the mru list and mfu list is less than * or equal to arc_c (we enforced this above), which means we * can use the simpler of the two equations below: * * mru + mfu + mru ghost + mfu ghost <= 2 * arc_c * mru ghost + mfu ghost <= arc_c */ target = refcount_count(&arc_mru_ghost->arcs_size) + refcount_count(&arc_mfu_ghost->arcs_size) - arc_c; bytes = arc_adjust_impl(arc_mfu_ghost, 0, target, ARC_BUFC_DATA); total_evicted += bytes; target -= bytes; total_evicted += arc_adjust_impl(arc_mfu_ghost, 0, target, ARC_BUFC_METADATA); return (total_evicted); } static void arc_do_user_evicts(void) { mutex_enter(&arc_user_evicts_lock); while (arc_eviction_list != NULL) { arc_buf_t *buf = arc_eviction_list; arc_eviction_list = buf->b_next; mutex_enter(&buf->b_evict_lock); buf->b_hdr = NULL; mutex_exit(&buf->b_evict_lock); mutex_exit(&arc_user_evicts_lock); if (buf->b_efunc != NULL) VERIFY0(buf->b_efunc(buf->b_private)); buf->b_efunc = NULL; buf->b_private = NULL; kmem_cache_free(buf_cache, buf); mutex_enter(&arc_user_evicts_lock); } mutex_exit(&arc_user_evicts_lock); } void arc_flush(spa_t *spa, boolean_t retry) { uint64_t guid = 0; /* * If retry is TRUE, a spa must not be specified since we have * no good way to determine if all of a spa's buffers have been * evicted from an arc state. */ ASSERT(!retry || spa == 0); if (spa != NULL) guid = spa_load_guid(spa); (void) arc_flush_state(arc_mru, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_mru, guid, ARC_BUFC_METADATA, retry); (void) arc_flush_state(arc_mfu, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_mfu, guid, ARC_BUFC_METADATA, retry); (void) arc_flush_state(arc_mru_ghost, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_mru_ghost, guid, ARC_BUFC_METADATA, retry); (void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_METADATA, retry); arc_do_user_evicts(); ASSERT(spa || arc_eviction_list == NULL); } void arc_shrink(int64_t to_free) { if (arc_c > arc_c_min) { DTRACE_PROBE4(arc__shrink, uint64_t, arc_c, uint64_t, arc_c_min, uint64_t, arc_p, uint64_t, to_free); if (arc_c > arc_c_min + to_free) atomic_add_64(&arc_c, -to_free); else arc_c = arc_c_min; atomic_add_64(&arc_p, -(arc_p >> arc_shrink_shift)); if (arc_c > arc_size) arc_c = MAX(arc_size, arc_c_min); if (arc_p > arc_c) arc_p = (arc_c >> 1); DTRACE_PROBE2(arc__shrunk, uint64_t, arc_c, uint64_t, arc_p); ASSERT(arc_c >= arc_c_min); ASSERT((int64_t)arc_p >= 0); } if (arc_size > arc_c) { DTRACE_PROBE2(arc__shrink_adjust, uint64_t, arc_size, uint64_t, arc_c); (void) arc_adjust(); } } static long needfree = 0; typedef enum free_memory_reason_t { FMR_UNKNOWN, FMR_NEEDFREE, FMR_LOTSFREE, FMR_SWAPFS_MINFREE, FMR_PAGES_PP_MAXIMUM, FMR_HEAP_ARENA, FMR_ZIO_ARENA, FMR_ZIO_FRAG, } free_memory_reason_t; int64_t last_free_memory; free_memory_reason_t last_free_reason; /* * Additional reserve of pages for pp_reserve. */ int64_t arc_pages_pp_reserve = 64; /* * Additional reserve of pages for swapfs. */ int64_t arc_swapfs_reserve = 64; /* * Return the amount of memory that can be consumed before reclaim will be * needed. Positive if there is sufficient free memory, negative indicates * the amount of memory that needs to be freed up. */ static int64_t arc_available_memory(void) { int64_t lowest = INT64_MAX; int64_t n; free_memory_reason_t r = FMR_UNKNOWN; #ifdef _KERNEL if (needfree > 0) { n = PAGESIZE * (-needfree); if (n < lowest) { lowest = n; r = FMR_NEEDFREE; } } /* * Cooperate with pagedaemon when it's time for it to scan * and reclaim some pages. */ n = PAGESIZE * ((int64_t)freemem - zfs_arc_free_target); if (n < lowest) { lowest = n; r = FMR_LOTSFREE; } #ifdef sun /* * check that we're out of range of the pageout scanner. It starts to * schedule paging if freemem is less than lotsfree and needfree. * lotsfree is the high-water mark for pageout, and needfree is the * number of needed free pages. We add extra pages here to make sure * the scanner doesn't start up while we're freeing memory. */ n = PAGESIZE * (freemem - lotsfree - needfree - desfree); if (n < lowest) { lowest = n; r = FMR_LOTSFREE; } /* * check to make sure that swapfs has enough space so that anon * reservations can still succeed. anon_resvmem() checks that the * availrmem is greater than swapfs_minfree, and the number of reserved * swap pages. We also add a bit of extra here just to prevent * circumstances from getting really dire. */ n = PAGESIZE * (availrmem - swapfs_minfree - swapfs_reserve - desfree - arc_swapfs_reserve); if (n < lowest) { lowest = n; r = FMR_SWAPFS_MINFREE; } /* * Check that we have enough availrmem that memory locking (e.g., via * mlock(3C) or memcntl(2)) can still succeed. (pages_pp_maximum * stores the number of pages that cannot be locked; when availrmem * drops below pages_pp_maximum, page locking mechanisms such as * page_pp_lock() will fail.) */ n = PAGESIZE * (availrmem - pages_pp_maximum - arc_pages_pp_reserve); if (n < lowest) { lowest = n; r = FMR_PAGES_PP_MAXIMUM; } #endif /* sun */ #if defined(__i386) || !defined(UMA_MD_SMALL_ALLOC) /* * If we're on an i386 platform, it's possible that we'll exhaust the * kernel heap space before we ever run out of available physical * memory. Most checks of the size of the heap_area compare against * tune.t_minarmem, which is the minimum available real memory that we * can have in the system. However, this is generally fixed at 25 pages * which is so low that it's useless. In this comparison, we seek to * calculate the total heap-size, and reclaim if more than 3/4ths of the * heap is allocated. (Or, in the calculation, if less than 1/4th is * free) */ n = (int64_t)vmem_size(heap_arena, VMEM_FREE) - (vmem_size(heap_arena, VMEM_FREE | VMEM_ALLOC) >> 2); if (n < lowest) { lowest = n; r = FMR_HEAP_ARENA; } #define zio_arena NULL #else #define zio_arena heap_arena #endif /* * If zio data pages are being allocated out of a separate heap segment, * then enforce that the size of available vmem for this arena remains * above about 1/16th free. * * Note: The 1/16th arena free requirement was put in place * to aggressively evict memory from the arc in order to avoid * memory fragmentation issues. */ if (zio_arena != NULL) { n = (int64_t)vmem_size(zio_arena, VMEM_FREE) - (vmem_size(zio_arena, VMEM_ALLOC) >> 4); if (n < lowest) { lowest = n; r = FMR_ZIO_ARENA; } } /* * Above limits know nothing about real level of KVA fragmentation. * Start aggressive reclamation if too little sequential KVA left. */ if (lowest > 0) { n = (vmem_size(heap_arena, VMEM_MAXFREE) < zfs_max_recordsize) ? -((int64_t)vmem_size(heap_arena, VMEM_ALLOC) >> 4) : INT64_MAX; if (n < lowest) { lowest = n; r = FMR_ZIO_FRAG; } } #else /* _KERNEL */ /* Every 100 calls, free a small amount */ if (spa_get_random(100) == 0) lowest = -1024; #endif /* _KERNEL */ last_free_memory = lowest; last_free_reason = r; DTRACE_PROBE2(arc__available_memory, int64_t, lowest, int, r); return (lowest); } /* * Determine if the system is under memory pressure and is asking * to reclaim memory. A return value of TRUE indicates that the system * is under memory pressure and that the arc should adjust accordingly. */ static boolean_t arc_reclaim_needed(void) { return (arc_available_memory() < 0); } extern kmem_cache_t *zio_buf_cache[]; extern kmem_cache_t *zio_data_buf_cache[]; extern kmem_cache_t *range_seg_cache; static __noinline void arc_kmem_reap_now(void) { size_t i; kmem_cache_t *prev_cache = NULL; kmem_cache_t *prev_data_cache = NULL; DTRACE_PROBE(arc__kmem_reap_start); #ifdef _KERNEL if (arc_meta_used >= arc_meta_limit) { /* * We are exceeding our meta-data cache limit. * Purge some DNLC entries to release holds on meta-data. */ dnlc_reduce_cache((void *)(uintptr_t)arc_reduce_dnlc_percent); } #if defined(__i386) /* * Reclaim unused memory from all kmem caches. */ kmem_reap(); #endif #endif for (i = 0; i < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; i++) { if (zio_buf_cache[i] != prev_cache) { prev_cache = zio_buf_cache[i]; kmem_cache_reap_now(zio_buf_cache[i]); } if (zio_data_buf_cache[i] != prev_data_cache) { prev_data_cache = zio_data_buf_cache[i]; kmem_cache_reap_now(zio_data_buf_cache[i]); } } kmem_cache_reap_now(buf_cache); kmem_cache_reap_now(hdr_full_cache); kmem_cache_reap_now(hdr_l2only_cache); kmem_cache_reap_now(range_seg_cache); #ifdef sun if (zio_arena != NULL) { /* * Ask the vmem arena to reclaim unused memory from its * quantum caches. */ vmem_qcache_reap(zio_arena); } #endif DTRACE_PROBE(arc__kmem_reap_end); } /* * Threads can block in arc_get_data_buf() waiting for this thread to evict * enough data and signal them to proceed. When this happens, the threads in * arc_get_data_buf() are sleeping while holding the hash lock for their * particular arc header. Thus, we must be careful to never sleep on a * hash lock in this thread. This is to prevent the following deadlock: * * - Thread A sleeps on CV in arc_get_data_buf() holding hash lock "L", * waiting for the reclaim thread to signal it. * * - arc_reclaim_thread() tries to acquire hash lock "L" using mutex_enter, * fails, and goes to sleep forever. * * This possible deadlock is avoided by always acquiring a hash lock * using mutex_tryenter() from arc_reclaim_thread(). */ static void arc_reclaim_thread(void *dummy __unused) { clock_t growtime = 0; callb_cpr_t cpr; CALLB_CPR_INIT(&cpr, &arc_reclaim_lock, callb_generic_cpr, FTAG); mutex_enter(&arc_reclaim_lock); while (!arc_reclaim_thread_exit) { int64_t free_memory = arc_available_memory(); uint64_t evicted = 0; mutex_exit(&arc_reclaim_lock); if (free_memory < 0) { arc_no_grow = B_TRUE; arc_warm = B_TRUE; /* * Wait at least zfs_grow_retry (default 60) seconds * before considering growing. */ growtime = ddi_get_lbolt() + (arc_grow_retry * hz); arc_kmem_reap_now(); /* * If we are still low on memory, shrink the ARC * so that we have arc_shrink_min free space. */ free_memory = arc_available_memory(); int64_t to_free = (arc_c >> arc_shrink_shift) - free_memory; if (to_free > 0) { #ifdef _KERNEL to_free = MAX(to_free, ptob(needfree)); #endif arc_shrink(to_free); } } else if (free_memory < arc_c >> arc_no_grow_shift) { arc_no_grow = B_TRUE; } else if (ddi_get_lbolt() >= growtime) { arc_no_grow = B_FALSE; } evicted = arc_adjust(); mutex_enter(&arc_reclaim_lock); /* * If evicted is zero, we couldn't evict anything via * arc_adjust(). This could be due to hash lock * collisions, but more likely due to the majority of * arc buffers being unevictable. Therefore, even if * arc_size is above arc_c, another pass is unlikely to * be helpful and could potentially cause us to enter an * infinite loop. */ if (arc_size <= arc_c || evicted == 0) { #ifdef _KERNEL needfree = 0; #endif /* * We're either no longer overflowing, or we * can't evict anything more, so we should wake * up any threads before we go to sleep. */ cv_broadcast(&arc_reclaim_waiters_cv); /* * Block until signaled, or after one second (we * might need to perform arc_kmem_reap_now() * even if we aren't being signalled) */ CALLB_CPR_SAFE_BEGIN(&cpr); (void) cv_timedwait(&arc_reclaim_thread_cv, &arc_reclaim_lock, hz); CALLB_CPR_SAFE_END(&cpr, &arc_reclaim_lock); } } arc_reclaim_thread_exit = FALSE; cv_broadcast(&arc_reclaim_thread_cv); CALLB_CPR_EXIT(&cpr); /* drops arc_reclaim_lock */ thread_exit(); } static void arc_user_evicts_thread(void *dummy __unused) { callb_cpr_t cpr; CALLB_CPR_INIT(&cpr, &arc_user_evicts_lock, callb_generic_cpr, FTAG); mutex_enter(&arc_user_evicts_lock); while (!arc_user_evicts_thread_exit) { mutex_exit(&arc_user_evicts_lock); arc_do_user_evicts(); /* * This is necessary in order for the mdb ::arc dcmd to * show up to date information. Since the ::arc command * does not call the kstat's update function, without * this call, the command may show stale stats for the * anon, mru, mru_ghost, mfu, and mfu_ghost lists. Even * with this change, the data might be up to 1 second * out of date; but that should suffice. The arc_state_t * structures can be queried directly if more accurate * information is needed. */ if (arc_ksp != NULL) arc_ksp->ks_update(arc_ksp, KSTAT_READ); mutex_enter(&arc_user_evicts_lock); /* * Block until signaled, or after one second (we need to * call the arc's kstat update function regularly). */ CALLB_CPR_SAFE_BEGIN(&cpr); (void) cv_timedwait(&arc_user_evicts_cv, &arc_user_evicts_lock, hz); CALLB_CPR_SAFE_END(&cpr, &arc_user_evicts_lock); } arc_user_evicts_thread_exit = FALSE; cv_broadcast(&arc_user_evicts_cv); CALLB_CPR_EXIT(&cpr); /* drops arc_user_evicts_lock */ thread_exit(); } /* * Adapt arc info given the number of bytes we are trying to add and * the state that we are comming from. This function is only called * when we are adding new content to the cache. */ static void arc_adapt(int bytes, arc_state_t *state) { int mult; uint64_t arc_p_min = (arc_c >> arc_p_min_shift); int64_t mrug_size = refcount_count(&arc_mru_ghost->arcs_size); int64_t mfug_size = refcount_count(&arc_mfu_ghost->arcs_size); if (state == arc_l2c_only) return; ASSERT(bytes > 0); /* * Adapt the target size of the MRU list: * - if we just hit in the MRU ghost list, then increase * the target size of the MRU list. * - if we just hit in the MFU ghost list, then increase * the target size of the MFU list by decreasing the * target size of the MRU list. */ if (state == arc_mru_ghost) { mult = (mrug_size >= mfug_size) ? 1 : (mfug_size / mrug_size); mult = MIN(mult, 10); /* avoid wild arc_p adjustment */ arc_p = MIN(arc_c - arc_p_min, arc_p + bytes * mult); } else if (state == arc_mfu_ghost) { uint64_t delta; mult = (mfug_size >= mrug_size) ? 1 : (mrug_size / mfug_size); mult = MIN(mult, 10); delta = MIN(bytes * mult, arc_p); arc_p = MAX(arc_p_min, arc_p - delta); } ASSERT((int64_t)arc_p >= 0); if (arc_reclaim_needed()) { cv_signal(&arc_reclaim_thread_cv); return; } if (arc_no_grow) return; if (arc_c >= arc_c_max) return; /* * If we're within (2 * maxblocksize) bytes of the target * cache size, increment the target cache size */ if (arc_size > arc_c - (2ULL << SPA_MAXBLOCKSHIFT)) { DTRACE_PROBE1(arc__inc_adapt, int, bytes); atomic_add_64(&arc_c, (int64_t)bytes); if (arc_c > arc_c_max) arc_c = arc_c_max; else if (state == arc_anon) atomic_add_64(&arc_p, (int64_t)bytes); if (arc_p > arc_c) arc_p = arc_c; } ASSERT((int64_t)arc_p >= 0); } /* * Check if arc_size has grown past our upper threshold, determined by * zfs_arc_overflow_shift. */ static boolean_t arc_is_overflowing(void) { /* Always allow at least one block of overflow */ uint64_t overflow = MAX(SPA_MAXBLOCKSIZE, arc_c >> zfs_arc_overflow_shift); return (arc_size >= arc_c + overflow); } /* * The buffer, supplied as the first argument, needs a data block. If we * are hitting the hard limit for the cache size, we must sleep, waiting * for the eviction thread to catch up. If we're past the target size * but below the hard limit, we'll only signal the reclaim thread and * continue on. */ static void arc_get_data_buf(arc_buf_t *buf) { arc_state_t *state = buf->b_hdr->b_l1hdr.b_state; uint64_t size = buf->b_hdr->b_size; arc_buf_contents_t type = arc_buf_type(buf->b_hdr); arc_adapt(size, state); /* * If arc_size is currently overflowing, and has grown past our * upper limit, we must be adding data faster than the evict * thread can evict. Thus, to ensure we don't compound the * problem by adding more data and forcing arc_size to grow even * further past it's target size, we halt and wait for the * eviction thread to catch up. * * It's also possible that the reclaim thread is unable to evict * enough buffers to get arc_size below the overflow limit (e.g. * due to buffers being un-evictable, or hash lock collisions). * In this case, we want to proceed regardless if we're * overflowing; thus we don't use a while loop here. */ if (arc_is_overflowing()) { mutex_enter(&arc_reclaim_lock); /* * Now that we've acquired the lock, we may no longer be * over the overflow limit, lets check. * * We're ignoring the case of spurious wake ups. If that * were to happen, it'd let this thread consume an ARC * buffer before it should have (i.e. before we're under * the overflow limit and were signalled by the reclaim * thread). As long as that is a rare occurrence, it * shouldn't cause any harm. */ if (arc_is_overflowing()) { cv_signal(&arc_reclaim_thread_cv); cv_wait(&arc_reclaim_waiters_cv, &arc_reclaim_lock); } mutex_exit(&arc_reclaim_lock); } if (type == ARC_BUFC_METADATA) { buf->b_data = zio_buf_alloc(size); arc_space_consume(size, ARC_SPACE_META); } else { ASSERT(type == ARC_BUFC_DATA); buf->b_data = zio_data_buf_alloc(size); arc_space_consume(size, ARC_SPACE_DATA); } /* * Update the state size. Note that ghost states have a * "ghost size" and so don't need to be updated. */ if (!GHOST_STATE(buf->b_hdr->b_l1hdr.b_state)) { arc_buf_hdr_t *hdr = buf->b_hdr; arc_state_t *state = hdr->b_l1hdr.b_state; (void) refcount_add_many(&state->arcs_size, size, buf); /* * If this is reached via arc_read, the link is * protected by the hash lock. If reached via * arc_buf_alloc, the header should not be accessed by * any other thread. And, if reached via arc_read_done, * the hash lock will protect it if it's found in the * hash table; otherwise no other thread should be * trying to [add|remove]_reference it. */ if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) { ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); atomic_add_64(&hdr->b_l1hdr.b_state->arcs_lsize[type], size); } /* * If we are growing the cache, and we are adding anonymous * data, and we have outgrown arc_p, update arc_p */ if (arc_size < arc_c && hdr->b_l1hdr.b_state == arc_anon && (refcount_count(&arc_anon->arcs_size) + refcount_count(&arc_mru->arcs_size) > arc_p)) arc_p = MIN(arc_c, arc_p + size); } ARCSTAT_BUMP(arcstat_allocated); } /* * This routine is called whenever a buffer is accessed. * NOTE: the hash lock is dropped in this function. */ static void arc_access(arc_buf_hdr_t *hdr, kmutex_t *hash_lock) { clock_t now; ASSERT(MUTEX_HELD(hash_lock)); ASSERT(HDR_HAS_L1HDR(hdr)); if (hdr->b_l1hdr.b_state == arc_anon) { /* * This buffer is not in the cache, and does not * appear in our "ghost" list. Add the new buffer * to the MRU state. */ ASSERT0(hdr->b_l1hdr.b_arc_access); hdr->b_l1hdr.b_arc_access = ddi_get_lbolt(); DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr); arc_change_state(arc_mru, hdr, hash_lock); } else if (hdr->b_l1hdr.b_state == arc_mru) { now = ddi_get_lbolt(); /* * If this buffer is here because of a prefetch, then either: * - clear the flag if this is a "referencing" read * (any subsequent access will bump this into the MFU state). * or * - move the buffer to the head of the list if this is * another prefetch (to make it less likely to be evicted). */ if (HDR_PREFETCH(hdr)) { if (refcount_count(&hdr->b_l1hdr.b_refcnt) == 0) { /* link protected by hash lock */ ASSERT(multilist_link_active( &hdr->b_l1hdr.b_arc_node)); } else { hdr->b_flags &= ~ARC_FLAG_PREFETCH; ARCSTAT_BUMP(arcstat_mru_hits); } hdr->b_l1hdr.b_arc_access = now; return; } /* * This buffer has been "accessed" only once so far, * but it is still in the cache. Move it to the MFU * state. */ if (now > hdr->b_l1hdr.b_arc_access + ARC_MINTIME) { /* * More than 125ms have passed since we * instantiated this buffer. Move it to the * most frequently used state. */ hdr->b_l1hdr.b_arc_access = now; DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr); arc_change_state(arc_mfu, hdr, hash_lock); } ARCSTAT_BUMP(arcstat_mru_hits); } else if (hdr->b_l1hdr.b_state == arc_mru_ghost) { arc_state_t *new_state; /* * This buffer has been "accessed" recently, but * was evicted from the cache. Move it to the * MFU state. */ if (HDR_PREFETCH(hdr)) { new_state = arc_mru; if (refcount_count(&hdr->b_l1hdr.b_refcnt) > 0) hdr->b_flags &= ~ARC_FLAG_PREFETCH; DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr); } else { new_state = arc_mfu; DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr); } hdr->b_l1hdr.b_arc_access = ddi_get_lbolt(); arc_change_state(new_state, hdr, hash_lock); ARCSTAT_BUMP(arcstat_mru_ghost_hits); } else if (hdr->b_l1hdr.b_state == arc_mfu) { /* * This buffer has been accessed more than once and is * still in the cache. Keep it in the MFU state. * * NOTE: an add_reference() that occurred when we did * the arc_read() will have kicked this off the list. * If it was a prefetch, we will explicitly move it to * the head of the list now. */ if ((HDR_PREFETCH(hdr)) != 0) { ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); /* link protected by hash_lock */ ASSERT(multilist_link_active(&hdr->b_l1hdr.b_arc_node)); } ARCSTAT_BUMP(arcstat_mfu_hits); hdr->b_l1hdr.b_arc_access = ddi_get_lbolt(); } else if (hdr->b_l1hdr.b_state == arc_mfu_ghost) { arc_state_t *new_state = arc_mfu; /* * This buffer has been accessed more than once but has * been evicted from the cache. Move it back to the * MFU state. */ if (HDR_PREFETCH(hdr)) { /* * This is a prefetch access... * move this block back to the MRU state. */ ASSERT0(refcount_count(&hdr->b_l1hdr.b_refcnt)); new_state = arc_mru; } hdr->b_l1hdr.b_arc_access = ddi_get_lbolt(); DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr); arc_change_state(new_state, hdr, hash_lock); ARCSTAT_BUMP(arcstat_mfu_ghost_hits); } else if (hdr->b_l1hdr.b_state == arc_l2c_only) { /* * This buffer is on the 2nd Level ARC. */ hdr->b_l1hdr.b_arc_access = ddi_get_lbolt(); DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr); arc_change_state(arc_mfu, hdr, hash_lock); } else { ASSERT(!"invalid arc state"); } } /* a generic arc_done_func_t which you can use */ /* ARGSUSED */ void arc_bcopy_func(zio_t *zio, arc_buf_t *buf, void *arg) { if (zio == NULL || zio->io_error == 0) bcopy(buf->b_data, arg, buf->b_hdr->b_size); VERIFY(arc_buf_remove_ref(buf, arg)); } /* a generic arc_done_func_t */ void arc_getbuf_func(zio_t *zio, arc_buf_t *buf, void *arg) { arc_buf_t **bufp = arg; if (zio && zio->io_error) { VERIFY(arc_buf_remove_ref(buf, arg)); *bufp = NULL; } else { *bufp = buf; ASSERT(buf->b_data); } } static void arc_read_done(zio_t *zio) { arc_buf_hdr_t *hdr; arc_buf_t *buf; arc_buf_t *abuf; /* buffer we're assigning to callback */ kmutex_t *hash_lock = NULL; arc_callback_t *callback_list, *acb; int freeable = FALSE; buf = zio->io_private; hdr = buf->b_hdr; /* * The hdr was inserted into hash-table and removed from lists * prior to starting I/O. We should find this header, since * it's in the hash table, and it should be legit since it's * not possible to evict it during the I/O. The only possible * reason for it not to be found is if we were freed during the * read. */ if (HDR_IN_HASH_TABLE(hdr)) { ASSERT3U(hdr->b_birth, ==, BP_PHYSICAL_BIRTH(zio->io_bp)); ASSERT3U(hdr->b_dva.dva_word[0], ==, BP_IDENTITY(zio->io_bp)->dva_word[0]); ASSERT3U(hdr->b_dva.dva_word[1], ==, BP_IDENTITY(zio->io_bp)->dva_word[1]); arc_buf_hdr_t *found = buf_hash_find(hdr->b_spa, zio->io_bp, &hash_lock); ASSERT((found == NULL && HDR_FREED_IN_READ(hdr) && hash_lock == NULL) || (found == hdr && DVA_EQUAL(&hdr->b_dva, BP_IDENTITY(zio->io_bp))) || (found == hdr && HDR_L2_READING(hdr))); } hdr->b_flags &= ~ARC_FLAG_L2_EVICTED; if (l2arc_noprefetch && HDR_PREFETCH(hdr)) hdr->b_flags &= ~ARC_FLAG_L2CACHE; /* byteswap if necessary */ callback_list = hdr->b_l1hdr.b_acb; ASSERT(callback_list != NULL); if (BP_SHOULD_BYTESWAP(zio->io_bp) && zio->io_error == 0) { dmu_object_byteswap_t bswap = DMU_OT_BYTESWAP(BP_GET_TYPE(zio->io_bp)); arc_byteswap_func_t *func = BP_GET_LEVEL(zio->io_bp) > 0 ? byteswap_uint64_array : dmu_ot_byteswap[bswap].ob_func; func(buf->b_data, hdr->b_size); } arc_cksum_compute(buf, B_FALSE); #ifdef illumos arc_buf_watch(buf); #endif /* illumos */ if (hash_lock && zio->io_error == 0 && hdr->b_l1hdr.b_state == arc_anon) { /* * Only call arc_access on anonymous buffers. This is because * if we've issued an I/O for an evicted buffer, we've already * called arc_access (to prevent any simultaneous readers from * getting confused). */ arc_access(hdr, hash_lock); } /* create copies of the data buffer for the callers */ abuf = buf; for (acb = callback_list; acb; acb = acb->acb_next) { if (acb->acb_done) { if (abuf == NULL) { ARCSTAT_BUMP(arcstat_duplicate_reads); abuf = arc_buf_clone(buf); } acb->acb_buf = abuf; abuf = NULL; } } hdr->b_l1hdr.b_acb = NULL; hdr->b_flags &= ~ARC_FLAG_IO_IN_PROGRESS; ASSERT(!HDR_BUF_AVAILABLE(hdr)); if (abuf == buf) { ASSERT(buf->b_efunc == NULL); ASSERT(hdr->b_l1hdr.b_datacnt == 1); hdr->b_flags |= ARC_FLAG_BUF_AVAILABLE; } ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt) || callback_list != NULL); if (zio->io_error != 0) { hdr->b_flags |= ARC_FLAG_IO_ERROR; if (hdr->b_l1hdr.b_state != arc_anon) arc_change_state(arc_anon, hdr, hash_lock); if (HDR_IN_HASH_TABLE(hdr)) buf_hash_remove(hdr); freeable = refcount_is_zero(&hdr->b_l1hdr.b_refcnt); } /* * Broadcast before we drop the hash_lock to avoid the possibility * that the hdr (and hence the cv) might be freed before we get to * the cv_broadcast(). */ cv_broadcast(&hdr->b_l1hdr.b_cv); if (hash_lock != NULL) { mutex_exit(hash_lock); } else { /* * This block was freed while we waited for the read to * complete. It has been removed from the hash table and * moved to the anonymous state (so that it won't show up * in the cache). */ ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon); freeable = refcount_is_zero(&hdr->b_l1hdr.b_refcnt); } /* execute each callback and free its structure */ while ((acb = callback_list) != NULL) { if (acb->acb_done) acb->acb_done(zio, acb->acb_buf, acb->acb_private); if (acb->acb_zio_dummy != NULL) { acb->acb_zio_dummy->io_error = zio->io_error; zio_nowait(acb->acb_zio_dummy); } callback_list = acb->acb_next; kmem_free(acb, sizeof (arc_callback_t)); } if (freeable) arc_hdr_destroy(hdr); } /* * "Read" the block at the specified DVA (in bp) via the * cache. If the block is found in the cache, invoke the provided * callback immediately and return. Note that the `zio' parameter * in the callback will be NULL in this case, since no IO was * required. If the block is not in the cache pass the read request * on to the spa with a substitute callback function, so that the * requested block will be added to the cache. * * If a read request arrives for a block that has a read in-progress, * either wait for the in-progress read to complete (and return the * results); or, if this is a read with a "done" func, add a record * to the read to invoke the "done" func when the read completes, * and return; or just return. * * arc_read_done() will invoke all the requested "done" functions * for readers of this block. */ int arc_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, arc_done_func_t *done, void *private, zio_priority_t priority, int zio_flags, arc_flags_t *arc_flags, const zbookmark_phys_t *zb) { arc_buf_hdr_t *hdr = NULL; arc_buf_t *buf = NULL; kmutex_t *hash_lock = NULL; zio_t *rzio; uint64_t guid = spa_load_guid(spa); ASSERT(!BP_IS_EMBEDDED(bp) || BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA); top: if (!BP_IS_EMBEDDED(bp)) { /* * Embedded BP's have no DVA and require no I/O to "read". * Create an anonymous arc buf to back it. */ hdr = buf_hash_find(guid, bp, &hash_lock); } if (hdr != NULL && HDR_HAS_L1HDR(hdr) && hdr->b_l1hdr.b_datacnt > 0) { *arc_flags |= ARC_FLAG_CACHED; if (HDR_IO_IN_PROGRESS(hdr)) { if ((hdr->b_flags & ARC_FLAG_PRIO_ASYNC_READ) && priority == ZIO_PRIORITY_SYNC_READ) { /* * This sync read must wait for an * in-progress async read (e.g. a predictive * prefetch). Async reads are queued * separately at the vdev_queue layer, so * this is a form of priority inversion. * Ideally, we would "inherit" the demand * i/o's priority by moving the i/o from * the async queue to the synchronous queue, * but there is currently no mechanism to do * so. Track this so that we can evaluate * the magnitude of this potential performance * problem. * * Note that if the prefetch i/o is already * active (has been issued to the device), * the prefetch improved performance, because * we issued it sooner than we would have * without the prefetch. */ DTRACE_PROBE1(arc__sync__wait__for__async, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP(arcstat_sync_wait_for_async); } if (hdr->b_flags & ARC_FLAG_PREDICTIVE_PREFETCH) { hdr->b_flags &= ~ARC_FLAG_PREDICTIVE_PREFETCH; } if (*arc_flags & ARC_FLAG_WAIT) { cv_wait(&hdr->b_l1hdr.b_cv, hash_lock); mutex_exit(hash_lock); goto top; } ASSERT(*arc_flags & ARC_FLAG_NOWAIT); if (done) { arc_callback_t *acb = NULL; acb = kmem_zalloc(sizeof (arc_callback_t), KM_SLEEP); acb->acb_done = done; acb->acb_private = private; if (pio != NULL) acb->acb_zio_dummy = zio_null(pio, spa, NULL, NULL, NULL, zio_flags); ASSERT(acb->acb_done != NULL); acb->acb_next = hdr->b_l1hdr.b_acb; hdr->b_l1hdr.b_acb = acb; add_reference(hdr, hash_lock, private); mutex_exit(hash_lock); return (0); } mutex_exit(hash_lock); return (0); } ASSERT(hdr->b_l1hdr.b_state == arc_mru || hdr->b_l1hdr.b_state == arc_mfu); if (done) { if (hdr->b_flags & ARC_FLAG_PREDICTIVE_PREFETCH) { /* * This is a demand read which does not have to * wait for i/o because we did a predictive * prefetch i/o for it, which has completed. */ DTRACE_PROBE1( arc__demand__hit__predictive__prefetch, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP( arcstat_demand_hit_predictive_prefetch); hdr->b_flags &= ~ARC_FLAG_PREDICTIVE_PREFETCH; } add_reference(hdr, hash_lock, private); /* * If this block is already in use, create a new * copy of the data so that we will be guaranteed * that arc_release() will always succeed. */ buf = hdr->b_l1hdr.b_buf; ASSERT(buf); ASSERT(buf->b_data); if (HDR_BUF_AVAILABLE(hdr)) { ASSERT(buf->b_efunc == NULL); hdr->b_flags &= ~ARC_FLAG_BUF_AVAILABLE; } else { buf = arc_buf_clone(buf); } } else if (*arc_flags & ARC_FLAG_PREFETCH && refcount_count(&hdr->b_l1hdr.b_refcnt) == 0) { hdr->b_flags |= ARC_FLAG_PREFETCH; } DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr); arc_access(hdr, hash_lock); if (*arc_flags & ARC_FLAG_L2CACHE) hdr->b_flags |= ARC_FLAG_L2CACHE; if (*arc_flags & ARC_FLAG_L2COMPRESS) hdr->b_flags |= ARC_FLAG_L2COMPRESS; mutex_exit(hash_lock); ARCSTAT_BUMP(arcstat_hits); ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr), demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data, metadata, hits); if (done) done(NULL, buf, private); } else { uint64_t size = BP_GET_LSIZE(bp); arc_callback_t *acb; vdev_t *vd = NULL; uint64_t addr = 0; boolean_t devw = B_FALSE; enum zio_compress b_compress = ZIO_COMPRESS_OFF; int32_t b_asize = 0; if (hdr == NULL) { /* this block is not in the cache */ arc_buf_hdr_t *exists = NULL; arc_buf_contents_t type = BP_GET_BUFC_TYPE(bp); buf = arc_buf_alloc(spa, size, private, type); hdr = buf->b_hdr; if (!BP_IS_EMBEDDED(bp)) { hdr->b_dva = *BP_IDENTITY(bp); hdr->b_birth = BP_PHYSICAL_BIRTH(bp); exists = buf_hash_insert(hdr, &hash_lock); } if (exists != NULL) { /* somebody beat us to the hash insert */ mutex_exit(hash_lock); buf_discard_identity(hdr); (void) arc_buf_remove_ref(buf, private); goto top; /* restart the IO request */ } /* * If there is a callback, we pass our reference to * it; otherwise we remove our reference. */ if (done == NULL) { (void) remove_reference(hdr, hash_lock, private); } if (*arc_flags & ARC_FLAG_PREFETCH) hdr->b_flags |= ARC_FLAG_PREFETCH; if (*arc_flags & ARC_FLAG_L2CACHE) hdr->b_flags |= ARC_FLAG_L2CACHE; if (*arc_flags & ARC_FLAG_L2COMPRESS) hdr->b_flags |= ARC_FLAG_L2COMPRESS; if (BP_GET_LEVEL(bp) > 0) hdr->b_flags |= ARC_FLAG_INDIRECT; } else { /* * This block is in the ghost cache. If it was L2-only * (and thus didn't have an L1 hdr), we realloc the * header to add an L1 hdr. */ if (!HDR_HAS_L1HDR(hdr)) { hdr = arc_hdr_realloc(hdr, hdr_l2only_cache, hdr_full_cache); } ASSERT(GHOST_STATE(hdr->b_l1hdr.b_state)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); /* * If there is a callback, we pass a reference to it. */ if (done != NULL) add_reference(hdr, hash_lock, private); if (*arc_flags & ARC_FLAG_PREFETCH) hdr->b_flags |= ARC_FLAG_PREFETCH; if (*arc_flags & ARC_FLAG_L2CACHE) hdr->b_flags |= ARC_FLAG_L2CACHE; if (*arc_flags & ARC_FLAG_L2COMPRESS) hdr->b_flags |= ARC_FLAG_L2COMPRESS; buf = kmem_cache_alloc(buf_cache, KM_PUSHPAGE); buf->b_hdr = hdr; buf->b_data = NULL; buf->b_efunc = NULL; buf->b_private = NULL; buf->b_next = NULL; hdr->b_l1hdr.b_buf = buf; ASSERT0(hdr->b_l1hdr.b_datacnt); hdr->b_l1hdr.b_datacnt = 1; arc_get_data_buf(buf); arc_access(hdr, hash_lock); } if (*arc_flags & ARC_FLAG_PREDICTIVE_PREFETCH) hdr->b_flags |= ARC_FLAG_PREDICTIVE_PREFETCH; ASSERT(!GHOST_STATE(hdr->b_l1hdr.b_state)); acb = kmem_zalloc(sizeof (arc_callback_t), KM_SLEEP); acb->acb_done = done; acb->acb_private = private; ASSERT(hdr->b_l1hdr.b_acb == NULL); hdr->b_l1hdr.b_acb = acb; hdr->b_flags |= ARC_FLAG_IO_IN_PROGRESS; if (HDR_HAS_L2HDR(hdr) && (vd = hdr->b_l2hdr.b_dev->l2ad_vdev) != NULL) { devw = hdr->b_l2hdr.b_dev->l2ad_writing; addr = hdr->b_l2hdr.b_daddr; b_compress = hdr->b_l2hdr.b_compress; b_asize = hdr->b_l2hdr.b_asize; /* * Lock out device removal. */ if (vdev_is_dead(vd) || !spa_config_tryenter(spa, SCL_L2ARC, vd, RW_READER)) vd = NULL; } if (hash_lock != NULL) mutex_exit(hash_lock); /* * At this point, we have a level 1 cache miss. Try again in * L2ARC if possible. */ ASSERT3U(hdr->b_size, ==, size); DTRACE_PROBE4(arc__miss, arc_buf_hdr_t *, hdr, blkptr_t *, bp, uint64_t, size, zbookmark_phys_t *, zb); ARCSTAT_BUMP(arcstat_misses); ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr), demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data, metadata, misses); #ifdef _KERNEL curthread->td_ru.ru_inblock++; #endif if (priority == ZIO_PRIORITY_ASYNC_READ) hdr->b_flags |= ARC_FLAG_PRIO_ASYNC_READ; else hdr->b_flags &= ~ARC_FLAG_PRIO_ASYNC_READ; if (vd != NULL && l2arc_ndev != 0 && !(l2arc_norw && devw)) { /* * Read from the L2ARC if the following are true: * 1. The L2ARC vdev was previously cached. * 2. This buffer still has L2ARC metadata. * 3. This buffer isn't currently writing to the L2ARC. * 4. The L2ARC entry wasn't evicted, which may * also have invalidated the vdev. * 5. This isn't prefetch and l2arc_noprefetch is set. */ if (HDR_HAS_L2HDR(hdr) && !HDR_L2_WRITING(hdr) && !HDR_L2_EVICTED(hdr) && !(l2arc_noprefetch && HDR_PREFETCH(hdr))) { l2arc_read_callback_t *cb; DTRACE_PROBE1(l2arc__hit, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP(arcstat_l2_hits); cb = kmem_zalloc(sizeof (l2arc_read_callback_t), KM_SLEEP); cb->l2rcb_buf = buf; cb->l2rcb_spa = spa; cb->l2rcb_bp = *bp; cb->l2rcb_zb = *zb; cb->l2rcb_flags = zio_flags; cb->l2rcb_compress = b_compress; ASSERT(addr >= VDEV_LABEL_START_SIZE && addr + size < vd->vdev_psize - VDEV_LABEL_END_SIZE); /* * l2arc read. The SCL_L2ARC lock will be * released by l2arc_read_done(). * Issue a null zio if the underlying buffer * was squashed to zero size by compression. */ if (b_compress == ZIO_COMPRESS_EMPTY) { rzio = zio_null(pio, spa, vd, l2arc_read_done, cb, zio_flags | ZIO_FLAG_DONT_CACHE | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY); } else { rzio = zio_read_phys(pio, vd, addr, b_asize, buf->b_data, ZIO_CHECKSUM_OFF, l2arc_read_done, cb, priority, zio_flags | ZIO_FLAG_DONT_CACHE | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY, B_FALSE); } DTRACE_PROBE2(l2arc__read, vdev_t *, vd, zio_t *, rzio); ARCSTAT_INCR(arcstat_l2_read_bytes, b_asize); if (*arc_flags & ARC_FLAG_NOWAIT) { zio_nowait(rzio); return (0); } ASSERT(*arc_flags & ARC_FLAG_WAIT); if (zio_wait(rzio) == 0) return (0); /* l2arc read error; goto zio_read() */ } else { DTRACE_PROBE1(l2arc__miss, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP(arcstat_l2_misses); if (HDR_L2_WRITING(hdr)) ARCSTAT_BUMP(arcstat_l2_rw_clash); spa_config_exit(spa, SCL_L2ARC, vd); } } else { if (vd != NULL) spa_config_exit(spa, SCL_L2ARC, vd); if (l2arc_ndev != 0) { DTRACE_PROBE1(l2arc__miss, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP(arcstat_l2_misses); } } rzio = zio_read(pio, spa, bp, buf->b_data, size, arc_read_done, buf, priority, zio_flags, zb); if (*arc_flags & ARC_FLAG_WAIT) return (zio_wait(rzio)); ASSERT(*arc_flags & ARC_FLAG_NOWAIT); zio_nowait(rzio); } return (0); } void arc_set_callback(arc_buf_t *buf, arc_evict_func_t *func, void *private) { ASSERT(buf->b_hdr != NULL); ASSERT(buf->b_hdr->b_l1hdr.b_state != arc_anon); ASSERT(!refcount_is_zero(&buf->b_hdr->b_l1hdr.b_refcnt) || func == NULL); ASSERT(buf->b_efunc == NULL); ASSERT(!HDR_BUF_AVAILABLE(buf->b_hdr)); buf->b_efunc = func; buf->b_private = private; } /* * Notify the arc that a block was freed, and thus will never be used again. */ void arc_freed(spa_t *spa, const blkptr_t *bp) { arc_buf_hdr_t *hdr; kmutex_t *hash_lock; uint64_t guid = spa_load_guid(spa); ASSERT(!BP_IS_EMBEDDED(bp)); hdr = buf_hash_find(guid, bp, &hash_lock); if (hdr == NULL) return; if (HDR_BUF_AVAILABLE(hdr)) { arc_buf_t *buf = hdr->b_l1hdr.b_buf; add_reference(hdr, hash_lock, FTAG); hdr->b_flags &= ~ARC_FLAG_BUF_AVAILABLE; mutex_exit(hash_lock); arc_release(buf, FTAG); (void) arc_buf_remove_ref(buf, FTAG); } else { mutex_exit(hash_lock); } } /* * Clear the user eviction callback set by arc_set_callback(), first calling * it if it exists. Because the presence of a callback keeps an arc_buf cached * clearing the callback may result in the arc_buf being destroyed. However, * it will not result in the *last* arc_buf being destroyed, hence the data * will remain cached in the ARC. We make a copy of the arc buffer here so * that we can process the callback without holding any locks. * * It's possible that the callback is already in the process of being cleared * by another thread. In this case we can not clear the callback. * * Returns B_TRUE if the callback was successfully called and cleared. */ boolean_t arc_clear_callback(arc_buf_t *buf) { arc_buf_hdr_t *hdr; kmutex_t *hash_lock; arc_evict_func_t *efunc = buf->b_efunc; void *private = buf->b_private; mutex_enter(&buf->b_evict_lock); hdr = buf->b_hdr; if (hdr == NULL) { /* * We are in arc_do_user_evicts(). */ ASSERT(buf->b_data == NULL); mutex_exit(&buf->b_evict_lock); return (B_FALSE); } else if (buf->b_data == NULL) { /* * We are on the eviction list; process this buffer now * but let arc_do_user_evicts() do the reaping. */ buf->b_efunc = NULL; mutex_exit(&buf->b_evict_lock); VERIFY0(efunc(private)); return (B_TRUE); } hash_lock = HDR_LOCK(hdr); mutex_enter(hash_lock); hdr = buf->b_hdr; ASSERT3P(hash_lock, ==, HDR_LOCK(hdr)); ASSERT3U(refcount_count(&hdr->b_l1hdr.b_refcnt), <, hdr->b_l1hdr.b_datacnt); ASSERT(hdr->b_l1hdr.b_state == arc_mru || hdr->b_l1hdr.b_state == arc_mfu); buf->b_efunc = NULL; buf->b_private = NULL; if (hdr->b_l1hdr.b_datacnt > 1) { mutex_exit(&buf->b_evict_lock); arc_buf_destroy(buf, TRUE); } else { ASSERT(buf == hdr->b_l1hdr.b_buf); hdr->b_flags |= ARC_FLAG_BUF_AVAILABLE; mutex_exit(&buf->b_evict_lock); } mutex_exit(hash_lock); VERIFY0(efunc(private)); return (B_TRUE); } /* * Release this buffer from the cache, making it an anonymous buffer. This * must be done after a read and prior to modifying the buffer contents. * If the buffer has more than one reference, we must make * a new hdr for the buffer. */ void arc_release(arc_buf_t *buf, void *tag) { arc_buf_hdr_t *hdr = buf->b_hdr; /* * It would be nice to assert that if it's DMU metadata (level > * 0 || it's the dnode file), then it must be syncing context. * But we don't know that information at this level. */ mutex_enter(&buf->b_evict_lock); ASSERT(HDR_HAS_L1HDR(hdr)); /* * We don't grab the hash lock prior to this check, because if * the buffer's header is in the arc_anon state, it won't be * linked into the hash table. */ if (hdr->b_l1hdr.b_state == arc_anon) { mutex_exit(&buf->b_evict_lock); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT(!HDR_IN_HASH_TABLE(hdr)); ASSERT(!HDR_HAS_L2HDR(hdr)); ASSERT(BUF_EMPTY(hdr)); ASSERT3U(hdr->b_l1hdr.b_datacnt, ==, 1); ASSERT3S(refcount_count(&hdr->b_l1hdr.b_refcnt), ==, 1); ASSERT(!list_link_active(&hdr->b_l1hdr.b_arc_node)); ASSERT3P(buf->b_efunc, ==, NULL); ASSERT3P(buf->b_private, ==, NULL); hdr->b_l1hdr.b_arc_access = 0; arc_buf_thaw(buf); return; } kmutex_t *hash_lock = HDR_LOCK(hdr); mutex_enter(hash_lock); /* * This assignment is only valid as long as the hash_lock is * held, we must be careful not to reference state or the * b_state field after dropping the lock. */ arc_state_t *state = hdr->b_l1hdr.b_state; ASSERT3P(hash_lock, ==, HDR_LOCK(hdr)); ASSERT3P(state, !=, arc_anon); /* this buffer is not on any list */ ASSERT(refcount_count(&hdr->b_l1hdr.b_refcnt) > 0); if (HDR_HAS_L2HDR(hdr)) { mutex_enter(&hdr->b_l2hdr.b_dev->l2ad_mtx); /* * We have to recheck this conditional again now that * we're holding the l2ad_mtx to prevent a race with * another thread which might be concurrently calling * l2arc_evict(). In that case, l2arc_evict() might have * destroyed the header's L2 portion as we were waiting * to acquire the l2ad_mtx. */ if (HDR_HAS_L2HDR(hdr)) { if (hdr->b_l2hdr.b_daddr != L2ARC_ADDR_UNSET) trim_map_free(hdr->b_l2hdr.b_dev->l2ad_vdev, hdr->b_l2hdr.b_daddr, hdr->b_l2hdr.b_asize, 0); arc_hdr_l2hdr_destroy(hdr); } mutex_exit(&hdr->b_l2hdr.b_dev->l2ad_mtx); } /* * Do we have more than one buf? */ if (hdr->b_l1hdr.b_datacnt > 1) { arc_buf_hdr_t *nhdr; arc_buf_t **bufp; uint64_t blksz = hdr->b_size; uint64_t spa = hdr->b_spa; arc_buf_contents_t type = arc_buf_type(hdr); uint32_t flags = hdr->b_flags; ASSERT(hdr->b_l1hdr.b_buf != buf || buf->b_next != NULL); /* * Pull the data off of this hdr and attach it to * a new anonymous hdr. */ (void) remove_reference(hdr, hash_lock, tag); bufp = &hdr->b_l1hdr.b_buf; while (*bufp != buf) bufp = &(*bufp)->b_next; *bufp = buf->b_next; buf->b_next = NULL; ASSERT3P(state, !=, arc_l2c_only); (void) refcount_remove_many( &state->arcs_size, hdr->b_size, buf); if (refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) { ASSERT3P(state, !=, arc_l2c_only); uint64_t *size = &state->arcs_lsize[type]; ASSERT3U(*size, >=, hdr->b_size); atomic_add_64(size, -hdr->b_size); } /* * We're releasing a duplicate user data buffer, update * our statistics accordingly. */ if (HDR_ISTYPE_DATA(hdr)) { ARCSTAT_BUMPDOWN(arcstat_duplicate_buffers); ARCSTAT_INCR(arcstat_duplicate_buffers_size, -hdr->b_size); } hdr->b_l1hdr.b_datacnt -= 1; arc_cksum_verify(buf); #ifdef illumos arc_buf_unwatch(buf); #endif /* illumos */ mutex_exit(hash_lock); nhdr = kmem_cache_alloc(hdr_full_cache, KM_PUSHPAGE); nhdr->b_size = blksz; nhdr->b_spa = spa; nhdr->b_flags = flags & ARC_FLAG_L2_WRITING; nhdr->b_flags |= arc_bufc_to_flags(type); nhdr->b_flags |= ARC_FLAG_HAS_L1HDR; nhdr->b_l1hdr.b_buf = buf; nhdr->b_l1hdr.b_datacnt = 1; nhdr->b_l1hdr.b_state = arc_anon; nhdr->b_l1hdr.b_arc_access = 0; nhdr->b_l1hdr.b_tmp_cdata = NULL; nhdr->b_freeze_cksum = NULL; (void) refcount_add(&nhdr->b_l1hdr.b_refcnt, tag); buf->b_hdr = nhdr; mutex_exit(&buf->b_evict_lock); (void) refcount_add_many(&arc_anon->arcs_size, blksz, buf); } else { mutex_exit(&buf->b_evict_lock); ASSERT(refcount_count(&hdr->b_l1hdr.b_refcnt) == 1); /* protected by hash lock, or hdr is on arc_anon */ ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); arc_change_state(arc_anon, hdr, hash_lock); hdr->b_l1hdr.b_arc_access = 0; mutex_exit(hash_lock); buf_discard_identity(hdr); arc_buf_thaw(buf); } buf->b_efunc = NULL; buf->b_private = NULL; } int arc_released(arc_buf_t *buf) { int released; mutex_enter(&buf->b_evict_lock); released = (buf->b_data != NULL && buf->b_hdr->b_l1hdr.b_state == arc_anon); mutex_exit(&buf->b_evict_lock); return (released); } #ifdef ZFS_DEBUG int arc_referenced(arc_buf_t *buf) { int referenced; mutex_enter(&buf->b_evict_lock); referenced = (refcount_count(&buf->b_hdr->b_l1hdr.b_refcnt)); mutex_exit(&buf->b_evict_lock); return (referenced); } #endif static void arc_write_ready(zio_t *zio) { arc_write_callback_t *callback = zio->io_private; arc_buf_t *buf = callback->awcb_buf; arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(!refcount_is_zero(&buf->b_hdr->b_l1hdr.b_refcnt)); ASSERT(hdr->b_l1hdr.b_datacnt > 0); callback->awcb_ready(zio, buf, callback->awcb_private); /* * If the IO is already in progress, then this is a re-write * attempt, so we need to thaw and re-compute the cksum. * It is the responsibility of the callback to handle the * accounting for any re-write attempt. */ if (HDR_IO_IN_PROGRESS(hdr)) { mutex_enter(&hdr->b_l1hdr.b_freeze_lock); if (hdr->b_freeze_cksum != NULL) { kmem_free(hdr->b_freeze_cksum, sizeof (zio_cksum_t)); hdr->b_freeze_cksum = NULL; } mutex_exit(&hdr->b_l1hdr.b_freeze_lock); } arc_cksum_compute(buf, B_FALSE); hdr->b_flags |= ARC_FLAG_IO_IN_PROGRESS; } /* * The SPA calls this callback for each physical write that happens on behalf * of a logical write. See the comment in dbuf_write_physdone() for details. */ static void arc_write_physdone(zio_t *zio) { arc_write_callback_t *cb = zio->io_private; if (cb->awcb_physdone != NULL) cb->awcb_physdone(zio, cb->awcb_buf, cb->awcb_private); } static void arc_write_done(zio_t *zio) { arc_write_callback_t *callback = zio->io_private; arc_buf_t *buf = callback->awcb_buf; arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT(hdr->b_l1hdr.b_acb == NULL); if (zio->io_error == 0) { if (BP_IS_HOLE(zio->io_bp) || BP_IS_EMBEDDED(zio->io_bp)) { buf_discard_identity(hdr); } else { hdr->b_dva = *BP_IDENTITY(zio->io_bp); hdr->b_birth = BP_PHYSICAL_BIRTH(zio->io_bp); } } else { ASSERT(BUF_EMPTY(hdr)); } /* * If the block to be written was all-zero or compressed enough to be * embedded in the BP, no write was performed so there will be no * dva/birth/checksum. The buffer must therefore remain anonymous * (and uncached). */ if (!BUF_EMPTY(hdr)) { arc_buf_hdr_t *exists; kmutex_t *hash_lock; ASSERT(zio->io_error == 0); arc_cksum_verify(buf); exists = buf_hash_insert(hdr, &hash_lock); if (exists != NULL) { /* * This can only happen if we overwrite for * sync-to-convergence, because we remove * buffers from the hash table when we arc_free(). */ if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { if (!BP_EQUAL(&zio->io_bp_orig, zio->io_bp)) panic("bad overwrite, hdr=%p exists=%p", (void *)hdr, (void *)exists); ASSERT(refcount_is_zero( &exists->b_l1hdr.b_refcnt)); arc_change_state(arc_anon, exists, hash_lock); mutex_exit(hash_lock); arc_hdr_destroy(exists); exists = buf_hash_insert(hdr, &hash_lock); ASSERT3P(exists, ==, NULL); } else if (zio->io_flags & ZIO_FLAG_NOPWRITE) { /* nopwrite */ ASSERT(zio->io_prop.zp_nopwrite); if (!BP_EQUAL(&zio->io_bp_orig, zio->io_bp)) panic("bad nopwrite, hdr=%p exists=%p", (void *)hdr, (void *)exists); } else { /* Dedup */ ASSERT(hdr->b_l1hdr.b_datacnt == 1); ASSERT(hdr->b_l1hdr.b_state == arc_anon); ASSERT(BP_GET_DEDUP(zio->io_bp)); ASSERT(BP_GET_LEVEL(zio->io_bp) == 0); } } hdr->b_flags &= ~ARC_FLAG_IO_IN_PROGRESS; /* if it's not anon, we are doing a scrub */ if (exists == NULL && hdr->b_l1hdr.b_state == arc_anon) arc_access(hdr, hash_lock); mutex_exit(hash_lock); } else { hdr->b_flags &= ~ARC_FLAG_IO_IN_PROGRESS; } ASSERT(!refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); callback->awcb_done(zio, buf, callback->awcb_private); kmem_free(callback, sizeof (arc_write_callback_t)); } zio_t * arc_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, arc_buf_t *buf, boolean_t l2arc, boolean_t l2arc_compress, const zio_prop_t *zp, arc_done_func_t *ready, arc_done_func_t *physdone, arc_done_func_t *done, void *private, zio_priority_t priority, int zio_flags, const zbookmark_phys_t *zb) { arc_buf_hdr_t *hdr = buf->b_hdr; arc_write_callback_t *callback; zio_t *zio; ASSERT(ready != NULL); ASSERT(done != NULL); ASSERT(!HDR_IO_ERROR(hdr)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT(hdr->b_l1hdr.b_acb == NULL); ASSERT(hdr->b_l1hdr.b_datacnt > 0); if (l2arc) hdr->b_flags |= ARC_FLAG_L2CACHE; if (l2arc_compress) hdr->b_flags |= ARC_FLAG_L2COMPRESS; callback = kmem_zalloc(sizeof (arc_write_callback_t), KM_SLEEP); callback->awcb_ready = ready; callback->awcb_physdone = physdone; callback->awcb_done = done; callback->awcb_private = private; callback->awcb_buf = buf; zio = zio_write(pio, spa, txg, bp, buf->b_data, hdr->b_size, zp, arc_write_ready, arc_write_physdone, arc_write_done, callback, priority, zio_flags, zb); return (zio); } static int arc_memory_throttle(uint64_t reserve, uint64_t txg) { #ifdef _KERNEL uint64_t available_memory = ptob(freemem); static uint64_t page_load = 0; static uint64_t last_txg = 0; #if defined(__i386) || !defined(UMA_MD_SMALL_ALLOC) available_memory = MIN(available_memory, ptob(vmem_size(heap_arena, VMEM_FREE))); #endif if (freemem > (uint64_t)physmem * arc_lotsfree_percent / 100) return (0); if (txg > last_txg) { last_txg = txg; page_load = 0; } /* * If we are in pageout, we know that memory is already tight, * the arc is already going to be evicting, so we just want to * continue to let page writes occur as quickly as possible. */ if (curproc == pageproc) { if (page_load > MAX(ptob(minfree), available_memory) / 4) return (SET_ERROR(ERESTART)); /* Note: reserve is inflated, so we deflate */ page_load += reserve / 8; return (0); } else if (page_load > 0 && arc_reclaim_needed()) { /* memory is low, delay before restarting */ ARCSTAT_INCR(arcstat_memory_throttle_count, 1); return (SET_ERROR(EAGAIN)); } page_load = 0; #endif return (0); } void arc_tempreserve_clear(uint64_t reserve) { atomic_add_64(&arc_tempreserve, -reserve); ASSERT((int64_t)arc_tempreserve >= 0); } int arc_tempreserve_space(uint64_t reserve, uint64_t txg) { int error; uint64_t anon_size; if (reserve > arc_c/4 && !arc_no_grow) { arc_c = MIN(arc_c_max, reserve * 4); DTRACE_PROBE1(arc__set_reserve, uint64_t, arc_c); } if (reserve > arc_c) return (SET_ERROR(ENOMEM)); /* * Don't count loaned bufs as in flight dirty data to prevent long * network delays from blocking transactions that are ready to be * assigned to a txg. */ anon_size = MAX((int64_t)(refcount_count(&arc_anon->arcs_size) - arc_loaned_bytes), 0); /* * Writes will, almost always, require additional memory allocations * in order to compress/encrypt/etc the data. We therefore need to * make sure that there is sufficient available memory for this. */ error = arc_memory_throttle(reserve, txg); if (error != 0) return (error); /* * Throttle writes when the amount of dirty data in the cache * gets too large. We try to keep the cache less than half full * of dirty blocks so that our sync times don't grow too large. * Note: if two requests come in concurrently, we might let them * both succeed, when one of them should fail. Not a huge deal. */ if (reserve + arc_tempreserve + anon_size > arc_c / 2 && anon_size > arc_c / 4) { dprintf("failing, arc_tempreserve=%lluK anon_meta=%lluK " "anon_data=%lluK tempreserve=%lluK arc_c=%lluK\n", arc_tempreserve>>10, arc_anon->arcs_lsize[ARC_BUFC_METADATA]>>10, arc_anon->arcs_lsize[ARC_BUFC_DATA]>>10, reserve>>10, arc_c>>10); return (SET_ERROR(ERESTART)); } atomic_add_64(&arc_tempreserve, reserve); return (0); } static void arc_kstat_update_state(arc_state_t *state, kstat_named_t *size, kstat_named_t *evict_data, kstat_named_t *evict_metadata) { size->value.ui64 = refcount_count(&state->arcs_size); evict_data->value.ui64 = state->arcs_lsize[ARC_BUFC_DATA]; evict_metadata->value.ui64 = state->arcs_lsize[ARC_BUFC_METADATA]; } static int arc_kstat_update(kstat_t *ksp, int rw) { arc_stats_t *as = ksp->ks_data; if (rw == KSTAT_WRITE) { return (EACCES); } else { arc_kstat_update_state(arc_anon, &as->arcstat_anon_size, &as->arcstat_anon_evictable_data, &as->arcstat_anon_evictable_metadata); arc_kstat_update_state(arc_mru, &as->arcstat_mru_size, &as->arcstat_mru_evictable_data, &as->arcstat_mru_evictable_metadata); arc_kstat_update_state(arc_mru_ghost, &as->arcstat_mru_ghost_size, &as->arcstat_mru_ghost_evictable_data, &as->arcstat_mru_ghost_evictable_metadata); arc_kstat_update_state(arc_mfu, &as->arcstat_mfu_size, &as->arcstat_mfu_evictable_data, &as->arcstat_mfu_evictable_metadata); arc_kstat_update_state(arc_mfu_ghost, &as->arcstat_mfu_ghost_size, &as->arcstat_mfu_ghost_evictable_data, &as->arcstat_mfu_ghost_evictable_metadata); } return (0); } /* * This function *must* return indices evenly distributed between all * sublists of the multilist. This is needed due to how the ARC eviction * code is laid out; arc_evict_state() assumes ARC buffers are evenly * distributed between all sublists and uses this assumption when * deciding which sublist to evict from and how much to evict from it. */ unsigned int arc_state_multilist_index_func(multilist_t *ml, void *obj) { arc_buf_hdr_t *hdr = obj; /* * We rely on b_dva to generate evenly distributed index * numbers using buf_hash below. So, as an added precaution, * let's make sure we never add empty buffers to the arc lists. */ ASSERT(!BUF_EMPTY(hdr)); /* * The assumption here, is the hash value for a given * arc_buf_hdr_t will remain constant throughout it's lifetime * (i.e. it's b_spa, b_dva, and b_birth fields don't change). * Thus, we don't need to store the header's sublist index * on insertion, as this index can be recalculated on removal. * * Also, the low order bits of the hash value are thought to be * distributed evenly. Otherwise, in the case that the multilist * has a power of two number of sublists, each sublists' usage * would not be evenly distributed. */ return (buf_hash(hdr->b_spa, &hdr->b_dva, hdr->b_birth) % multilist_get_num_sublists(ml)); } #ifdef _KERNEL static eventhandler_tag arc_event_lowmem = NULL; static void arc_lowmem(void *arg __unused, int howto __unused) { mutex_enter(&arc_reclaim_lock); /* XXX: Memory deficit should be passed as argument. */ needfree = btoc(arc_c >> arc_shrink_shift); DTRACE_PROBE(arc__needfree); cv_signal(&arc_reclaim_thread_cv); /* * It is unsafe to block here in arbitrary threads, because we can come * here from ARC itself and may hold ARC locks and thus risk a deadlock * with ARC reclaim thread. */ if (curproc == pageproc) (void) cv_wait(&arc_reclaim_waiters_cv, &arc_reclaim_lock); mutex_exit(&arc_reclaim_lock); } #endif void arc_init(void) { int i, prefetch_tunable_set = 0; mutex_init(&arc_reclaim_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&arc_reclaim_thread_cv, NULL, CV_DEFAULT, NULL); cv_init(&arc_reclaim_waiters_cv, NULL, CV_DEFAULT, NULL); mutex_init(&arc_user_evicts_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&arc_user_evicts_cv, NULL, CV_DEFAULT, NULL); /* Convert seconds to clock ticks */ arc_min_prefetch_lifespan = 1 * hz; /* Start out with 1/8 of all memory */ arc_c = kmem_size() / 8; #ifdef sun #ifdef _KERNEL /* * On architectures where the physical memory can be larger * than the addressable space (intel in 32-bit mode), we may * need to limit the cache to 1/8 of VM size. */ arc_c = MIN(arc_c, vmem_size(heap_arena, VMEM_ALLOC | VMEM_FREE) / 8); #endif #endif /* sun */ /* set min cache to 1/32 of all memory, or 16MB, whichever is more */ arc_c_min = MAX(arc_c / 4, 16 << 20); /* set max to 1/2 of all memory, or all but 1GB, whichever is more */ if (arc_c * 8 >= 1 << 30) arc_c_max = (arc_c * 8) - (1 << 30); else arc_c_max = arc_c_min; arc_c_max = MAX(arc_c * 5, arc_c_max); /* * In userland, there's only the memory pressure that we artificially * create (see arc_available_memory()). Don't let arc_c get too * small, because it can cause transactions to be larger than * arc_c, causing arc_tempreserve_space() to fail. */ #ifndef _KERNEL arc_c_min = arc_c_max / 2; #endif #ifdef _KERNEL /* * Allow the tunables to override our calculations if they are * reasonable (ie. over 16MB) */ if (zfs_arc_max > 16 << 20 && zfs_arc_max < kmem_size()) arc_c_max = zfs_arc_max; if (zfs_arc_min > 16 << 20 && zfs_arc_min <= arc_c_max) arc_c_min = zfs_arc_min; #endif arc_c = arc_c_max; arc_p = (arc_c >> 1); /* limit meta-data to 1/4 of the arc capacity */ arc_meta_limit = arc_c_max / 4; /* Allow the tunable to override if it is reasonable */ if (zfs_arc_meta_limit > 0 && zfs_arc_meta_limit <= arc_c_max) arc_meta_limit = zfs_arc_meta_limit; if (arc_c_min < arc_meta_limit / 2 && zfs_arc_min == 0) arc_c_min = arc_meta_limit / 2; if (zfs_arc_meta_min > 0) { arc_meta_min = zfs_arc_meta_min; } else { arc_meta_min = arc_c_min / 2; } if (zfs_arc_grow_retry > 0) arc_grow_retry = zfs_arc_grow_retry; if (zfs_arc_shrink_shift > 0) arc_shrink_shift = zfs_arc_shrink_shift; /* * Ensure that arc_no_grow_shift is less than arc_shrink_shift. */ if (arc_no_grow_shift >= arc_shrink_shift) arc_no_grow_shift = arc_shrink_shift - 1; if (zfs_arc_p_min_shift > 0) arc_p_min_shift = zfs_arc_p_min_shift; if (zfs_arc_num_sublists_per_state < 1) zfs_arc_num_sublists_per_state = MAX(max_ncpus, 1); /* if kmem_flags are set, lets try to use less memory */ if (kmem_debugging()) arc_c = arc_c / 2; if (arc_c < arc_c_min) arc_c = arc_c_min; zfs_arc_min = arc_c_min; zfs_arc_max = arc_c_max; arc_anon = &ARC_anon; arc_mru = &ARC_mru; arc_mru_ghost = &ARC_mru_ghost; arc_mfu = &ARC_mfu; arc_mfu_ghost = &ARC_mfu_ghost; arc_l2c_only = &ARC_l2c_only; arc_size = 0; multilist_create(&arc_mru->arcs_list[ARC_BUFC_METADATA], sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), zfs_arc_num_sublists_per_state, arc_state_multilist_index_func); multilist_create(&arc_mru->arcs_list[ARC_BUFC_DATA], sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), zfs_arc_num_sublists_per_state, arc_state_multilist_index_func); multilist_create(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA], sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), zfs_arc_num_sublists_per_state, arc_state_multilist_index_func); multilist_create(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA], sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), zfs_arc_num_sublists_per_state, arc_state_multilist_index_func); multilist_create(&arc_mfu->arcs_list[ARC_BUFC_METADATA], sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), zfs_arc_num_sublists_per_state, arc_state_multilist_index_func); multilist_create(&arc_mfu->arcs_list[ARC_BUFC_DATA], sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), zfs_arc_num_sublists_per_state, arc_state_multilist_index_func); multilist_create(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA], sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), zfs_arc_num_sublists_per_state, arc_state_multilist_index_func); multilist_create(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA], sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), zfs_arc_num_sublists_per_state, arc_state_multilist_index_func); multilist_create(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA], sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), zfs_arc_num_sublists_per_state, arc_state_multilist_index_func); multilist_create(&arc_l2c_only->arcs_list[ARC_BUFC_DATA], sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), zfs_arc_num_sublists_per_state, arc_state_multilist_index_func); refcount_create(&arc_anon->arcs_size); refcount_create(&arc_mru->arcs_size); refcount_create(&arc_mru_ghost->arcs_size); refcount_create(&arc_mfu->arcs_size); refcount_create(&arc_mfu_ghost->arcs_size); refcount_create(&arc_l2c_only->arcs_size); buf_init(); arc_reclaim_thread_exit = FALSE; arc_user_evicts_thread_exit = FALSE; arc_eviction_list = NULL; bzero(&arc_eviction_hdr, sizeof (arc_buf_hdr_t)); arc_ksp = kstat_create("zfs", 0, "arcstats", "misc", KSTAT_TYPE_NAMED, sizeof (arc_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL); if (arc_ksp != NULL) { arc_ksp->ks_data = &arc_stats; arc_ksp->ks_update = arc_kstat_update; kstat_install(arc_ksp); } (void) thread_create(NULL, 0, arc_reclaim_thread, NULL, 0, &p0, TS_RUN, minclsyspri); #ifdef _KERNEL arc_event_lowmem = EVENTHANDLER_REGISTER(vm_lowmem, arc_lowmem, NULL, EVENTHANDLER_PRI_FIRST); #endif (void) thread_create(NULL, 0, arc_user_evicts_thread, NULL, 0, &p0, TS_RUN, minclsyspri); arc_dead = FALSE; arc_warm = B_FALSE; /* * Calculate maximum amount of dirty data per pool. * * If it has been set by /etc/system, take that. * Otherwise, use a percentage of physical memory defined by * zfs_dirty_data_max_percent (default 10%) with a cap at * zfs_dirty_data_max_max (default 4GB). */ if (zfs_dirty_data_max == 0) { zfs_dirty_data_max = ptob(physmem) * zfs_dirty_data_max_percent / 100; zfs_dirty_data_max = MIN(zfs_dirty_data_max, zfs_dirty_data_max_max); } #ifdef _KERNEL if (TUNABLE_INT_FETCH("vfs.zfs.prefetch_disable", &zfs_prefetch_disable)) prefetch_tunable_set = 1; #ifdef __i386__ if (prefetch_tunable_set == 0) { printf("ZFS NOTICE: Prefetch is disabled by default on i386 " "-- to enable,\n"); printf(" add \"vfs.zfs.prefetch_disable=0\" " "to /boot/loader.conf.\n"); zfs_prefetch_disable = 1; } #else if ((((uint64_t)physmem * PAGESIZE) < (1ULL << 32)) && prefetch_tunable_set == 0) { printf("ZFS NOTICE: Prefetch is disabled by default if less " "than 4GB of RAM is present;\n" " to enable, add \"vfs.zfs.prefetch_disable=0\" " "to /boot/loader.conf.\n"); zfs_prefetch_disable = 1; } #endif /* Warn about ZFS memory and address space requirements. */ if (((uint64_t)physmem * PAGESIZE) < (256 + 128 + 64) * (1 << 20)) { printf("ZFS WARNING: Recommended minimum RAM size is 512MB; " "expect unstable behavior.\n"); } if (kmem_size() < 512 * (1 << 20)) { printf("ZFS WARNING: Recommended minimum kmem_size is 512MB; " "expect unstable behavior.\n"); printf(" Consider tuning vm.kmem_size and " "vm.kmem_size_max\n"); printf(" in /boot/loader.conf.\n"); } #endif } void arc_fini(void) { mutex_enter(&arc_reclaim_lock); arc_reclaim_thread_exit = TRUE; /* * The reclaim thread will set arc_reclaim_thread_exit back to * FALSE when it is finished exiting; we're waiting for that. */ while (arc_reclaim_thread_exit) { cv_signal(&arc_reclaim_thread_cv); cv_wait(&arc_reclaim_thread_cv, &arc_reclaim_lock); } mutex_exit(&arc_reclaim_lock); mutex_enter(&arc_user_evicts_lock); arc_user_evicts_thread_exit = TRUE; /* * The user evicts thread will set arc_user_evicts_thread_exit * to FALSE when it is finished exiting; we're waiting for that. */ while (arc_user_evicts_thread_exit) { cv_signal(&arc_user_evicts_cv); cv_wait(&arc_user_evicts_cv, &arc_user_evicts_lock); } mutex_exit(&arc_user_evicts_lock); /* Use TRUE to ensure *all* buffers are evicted */ arc_flush(NULL, TRUE); arc_dead = TRUE; if (arc_ksp != NULL) { kstat_delete(arc_ksp); arc_ksp = NULL; } mutex_destroy(&arc_reclaim_lock); cv_destroy(&arc_reclaim_thread_cv); cv_destroy(&arc_reclaim_waiters_cv); mutex_destroy(&arc_user_evicts_lock); cv_destroy(&arc_user_evicts_cv); refcount_destroy(&arc_anon->arcs_size); refcount_destroy(&arc_mru->arcs_size); refcount_destroy(&arc_mru_ghost->arcs_size); refcount_destroy(&arc_mfu->arcs_size); refcount_destroy(&arc_mfu_ghost->arcs_size); refcount_destroy(&arc_l2c_only->arcs_size); multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_DATA]); multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA]); multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_DATA]); multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA]); buf_fini(); ASSERT0(arc_loaned_bytes); #ifdef _KERNEL if (arc_event_lowmem != NULL) EVENTHANDLER_DEREGISTER(vm_lowmem, arc_event_lowmem); #endif } /* * Level 2 ARC * * The level 2 ARC (L2ARC) is a cache layer in-between main memory and disk. * It uses dedicated storage devices to hold cached data, which are populated * using large infrequent writes. The main role of this cache is to boost * the performance of random read workloads. The intended L2ARC devices * include short-stroked disks, solid state disks, and other media with * substantially faster read latency than disk. * * +-----------------------+ * | ARC | * +-----------------------+ * | ^ ^ * | | | * l2arc_feed_thread() arc_read() * | | | * | l2arc read | * V | | * +---------------+ | * | L2ARC | | * +---------------+ | * | ^ | * l2arc_write() | | * | | | * V | | * +-------+ +-------+ * | vdev | | vdev | * | cache | | cache | * +-------+ +-------+ * +=========+ .-----. * : L2ARC : |-_____-| * : devices : | Disks | * +=========+ `-_____-' * * Read requests are satisfied from the following sources, in order: * * 1) ARC * 2) vdev cache of L2ARC devices * 3) L2ARC devices * 4) vdev cache of disks * 5) disks * * Some L2ARC device types exhibit extremely slow write performance. * To accommodate for this there are some significant differences between * the L2ARC and traditional cache design: * * 1. There is no eviction path from the ARC to the L2ARC. Evictions from * the ARC behave as usual, freeing buffers and placing headers on ghost * lists. The ARC does not send buffers to the L2ARC during eviction as * this would add inflated write latencies for all ARC memory pressure. * * 2. The L2ARC attempts to cache data from the ARC before it is evicted. * It does this by periodically scanning buffers from the eviction-end of * the MFU and MRU ARC lists, copying them to the L2ARC devices if they are * not already there. It scans until a headroom of buffers is satisfied, * which itself is a buffer for ARC eviction. If a compressible buffer is * found during scanning and selected for writing to an L2ARC device, we * temporarily boost scanning headroom during the next scan cycle to make * sure we adapt to compression effects (which might significantly reduce * the data volume we write to L2ARC). The thread that does this is * l2arc_feed_thread(), illustrated below; example sizes are included to * provide a better sense of ratio than this diagram: * * head --> tail * +---------------------+----------+ * ARC_mfu |:::::#:::::::::::::::|o#o###o###|-->. # already on L2ARC * +---------------------+----------+ | o L2ARC eligible * ARC_mru |:#:::::::::::::::::::|#o#ooo####|-->| : ARC buffer * +---------------------+----------+ | * 15.9 Gbytes ^ 32 Mbytes | * headroom | * l2arc_feed_thread() * | * l2arc write hand <--[oooo]--' * | 8 Mbyte * | write max * V * +==============================+ * L2ARC dev |####|#|###|###| |####| ... | * +==============================+ * 32 Gbytes * * 3. If an ARC buffer is copied to the L2ARC but then hit instead of * evicted, then the L2ARC has cached a buffer much sooner than it probably * needed to, potentially wasting L2ARC device bandwidth and storage. It is * safe to say that this is an uncommon case, since buffers at the end of * the ARC lists have moved there due to inactivity. * * 4. If the ARC evicts faster than the L2ARC can maintain a headroom, * then the L2ARC simply misses copying some buffers. This serves as a * pressure valve to prevent heavy read workloads from both stalling the ARC * with waits and clogging the L2ARC with writes. This also helps prevent * the potential for the L2ARC to churn if it attempts to cache content too * quickly, such as during backups of the entire pool. * * 5. After system boot and before the ARC has filled main memory, there are * no evictions from the ARC and so the tails of the ARC_mfu and ARC_mru * lists can remain mostly static. Instead of searching from tail of these * lists as pictured, the l2arc_feed_thread() will search from the list heads * for eligible buffers, greatly increasing its chance of finding them. * * The L2ARC device write speed is also boosted during this time so that * the L2ARC warms up faster. Since there have been no ARC evictions yet, * there are no L2ARC reads, and no fear of degrading read performance * through increased writes. * * 6. Writes to the L2ARC devices are grouped and sent in-sequence, so that * the vdev queue can aggregate them into larger and fewer writes. Each * device is written to in a rotor fashion, sweeping writes through * available space then repeating. * * 7. The L2ARC does not store dirty content. It never needs to flush * write buffers back to disk based storage. * * 8. If an ARC buffer is written (and dirtied) which also exists in the * L2ARC, the now stale L2ARC buffer is immediately dropped. * * The performance of the L2ARC can be tweaked by a number of tunables, which * may be necessary for different workloads: * * l2arc_write_max max write bytes per interval * l2arc_write_boost extra write bytes during device warmup * l2arc_noprefetch skip caching prefetched buffers * l2arc_headroom number of max device writes to precache * l2arc_headroom_boost when we find compressed buffers during ARC * scanning, we multiply headroom by this * percentage factor for the next scan cycle, * since more compressed buffers are likely to * be present * l2arc_feed_secs seconds between L2ARC writing * * Tunables may be removed or added as future performance improvements are * integrated, and also may become zpool properties. * * There are three key functions that control how the L2ARC warms up: * * l2arc_write_eligible() check if a buffer is eligible to cache * l2arc_write_size() calculate how much to write * l2arc_write_interval() calculate sleep delay between writes * * These three functions determine what to write, how much, and how quickly * to send writes. */ static boolean_t l2arc_write_eligible(uint64_t spa_guid, arc_buf_hdr_t *hdr) { /* * A buffer is *not* eligible for the L2ARC if it: * 1. belongs to a different spa. * 2. is already cached on the L2ARC. * 3. has an I/O in progress (it may be an incomplete read). * 4. is flagged not eligible (zfs property). */ if (hdr->b_spa != spa_guid) { ARCSTAT_BUMP(arcstat_l2_write_spa_mismatch); return (B_FALSE); } if (HDR_HAS_L2HDR(hdr)) { ARCSTAT_BUMP(arcstat_l2_write_in_l2); return (B_FALSE); } if (HDR_IO_IN_PROGRESS(hdr)) { ARCSTAT_BUMP(arcstat_l2_write_hdr_io_in_progress); return (B_FALSE); } if (!HDR_L2CACHE(hdr)) { ARCSTAT_BUMP(arcstat_l2_write_not_cacheable); return (B_FALSE); } return (B_TRUE); } static uint64_t l2arc_write_size(void) { uint64_t size; /* * Make sure our globals have meaningful values in case the user * altered them. */ size = l2arc_write_max; if (size == 0) { cmn_err(CE_NOTE, "Bad value for l2arc_write_max, value must " "be greater than zero, resetting it to the default (%d)", L2ARC_WRITE_SIZE); size = l2arc_write_max = L2ARC_WRITE_SIZE; } if (arc_warm == B_FALSE) size += l2arc_write_boost; return (size); } static clock_t l2arc_write_interval(clock_t began, uint64_t wanted, uint64_t wrote) { clock_t interval, next, now; /* * If the ARC lists are busy, increase our write rate; if the * lists are stale, idle back. This is achieved by checking * how much we previously wrote - if it was more than half of * what we wanted, schedule the next write much sooner. */ if (l2arc_feed_again && wrote > (wanted / 2)) interval = (hz * l2arc_feed_min_ms) / 1000; else interval = hz * l2arc_feed_secs; now = ddi_get_lbolt(); next = MAX(now, MIN(now + interval, began + interval)); return (next); } /* * Cycle through L2ARC devices. This is how L2ARC load balances. * If a device is returned, this also returns holding the spa config lock. */ static l2arc_dev_t * l2arc_dev_get_next(void) { l2arc_dev_t *first, *next = NULL; /* * Lock out the removal of spas (spa_namespace_lock), then removal * of cache devices (l2arc_dev_mtx). Once a device has been selected, * both locks will be dropped and a spa config lock held instead. */ mutex_enter(&spa_namespace_lock); mutex_enter(&l2arc_dev_mtx); /* if there are no vdevs, there is nothing to do */ if (l2arc_ndev == 0) goto out; first = NULL; next = l2arc_dev_last; do { /* loop around the list looking for a non-faulted vdev */ if (next == NULL) { next = list_head(l2arc_dev_list); } else { next = list_next(l2arc_dev_list, next); if (next == NULL) next = list_head(l2arc_dev_list); } /* if we have come back to the start, bail out */ if (first == NULL) first = next; else if (next == first) break; } while (vdev_is_dead(next->l2ad_vdev)); /* if we were unable to find any usable vdevs, return NULL */ if (vdev_is_dead(next->l2ad_vdev)) next = NULL; l2arc_dev_last = next; out: mutex_exit(&l2arc_dev_mtx); /* * Grab the config lock to prevent the 'next' device from being * removed while we are writing to it. */ if (next != NULL) spa_config_enter(next->l2ad_spa, SCL_L2ARC, next, RW_READER); mutex_exit(&spa_namespace_lock); return (next); } /* * Free buffers that were tagged for destruction. */ static void l2arc_do_free_on_write() { list_t *buflist; l2arc_data_free_t *df, *df_prev; mutex_enter(&l2arc_free_on_write_mtx); buflist = l2arc_free_on_write; for (df = list_tail(buflist); df; df = df_prev) { df_prev = list_prev(buflist, df); ASSERT(df->l2df_data != NULL); ASSERT(df->l2df_func != NULL); df->l2df_func(df->l2df_data, df->l2df_size); list_remove(buflist, df); kmem_free(df, sizeof (l2arc_data_free_t)); } mutex_exit(&l2arc_free_on_write_mtx); } /* * A write to a cache device has completed. Update all headers to allow * reads from these buffers to begin. */ static void l2arc_write_done(zio_t *zio) { l2arc_write_callback_t *cb; l2arc_dev_t *dev; list_t *buflist; arc_buf_hdr_t *head, *hdr, *hdr_prev; kmutex_t *hash_lock; int64_t bytes_dropped = 0; cb = zio->io_private; ASSERT(cb != NULL); dev = cb->l2wcb_dev; ASSERT(dev != NULL); head = cb->l2wcb_head; ASSERT(head != NULL); buflist = &dev->l2ad_buflist; ASSERT(buflist != NULL); DTRACE_PROBE2(l2arc__iodone, zio_t *, zio, l2arc_write_callback_t *, cb); if (zio->io_error != 0) ARCSTAT_BUMP(arcstat_l2_writes_error); /* * All writes completed, or an error was hit. */ top: mutex_enter(&dev->l2ad_mtx); for (hdr = list_prev(buflist, head); hdr; hdr = hdr_prev) { hdr_prev = list_prev(buflist, hdr); hash_lock = HDR_LOCK(hdr); /* * We cannot use mutex_enter or else we can deadlock * with l2arc_write_buffers (due to swapping the order * the hash lock and l2ad_mtx are taken). */ if (!mutex_tryenter(hash_lock)) { /* * Missed the hash lock. We must retry so we * don't leave the ARC_FLAG_L2_WRITING bit set. */ ARCSTAT_BUMP(arcstat_l2_writes_lock_retry); /* * We don't want to rescan the headers we've * already marked as having been written out, so * we reinsert the head node so we can pick up * where we left off. */ list_remove(buflist, head); list_insert_after(buflist, hdr, head); mutex_exit(&dev->l2ad_mtx); /* * We wait for the hash lock to become available * to try and prevent busy waiting, and increase * the chance we'll be able to acquire the lock * the next time around. */ mutex_enter(hash_lock); mutex_exit(hash_lock); goto top; } /* * We could not have been moved into the arc_l2c_only * state while in-flight due to our ARC_FLAG_L2_WRITING * bit being set. Let's just ensure that's being enforced. */ ASSERT(HDR_HAS_L1HDR(hdr)); /* * We may have allocated a buffer for L2ARC compression, * we must release it to avoid leaking this data. */ l2arc_release_cdata_buf(hdr); if (zio->io_error != 0) { /* * Error - drop L2ARC entry. */ list_remove(buflist, hdr); trim_map_free(hdr->b_l2hdr.b_dev->l2ad_vdev, hdr->b_l2hdr.b_daddr, hdr->b_l2hdr.b_asize, 0); hdr->b_flags &= ~ARC_FLAG_HAS_L2HDR; ARCSTAT_INCR(arcstat_l2_asize, -hdr->b_l2hdr.b_asize); ARCSTAT_INCR(arcstat_l2_size, -hdr->b_size); bytes_dropped += hdr->b_l2hdr.b_asize; (void) refcount_remove_many(&dev->l2ad_alloc, hdr->b_l2hdr.b_asize, hdr); } /* * Allow ARC to begin reads and ghost list evictions to * this L2ARC entry. */ hdr->b_flags &= ~ARC_FLAG_L2_WRITING; mutex_exit(hash_lock); } atomic_inc_64(&l2arc_writes_done); list_remove(buflist, head); ASSERT(!HDR_HAS_L1HDR(head)); kmem_cache_free(hdr_l2only_cache, head); mutex_exit(&dev->l2ad_mtx); vdev_space_update(dev->l2ad_vdev, -bytes_dropped, 0, 0); l2arc_do_free_on_write(); kmem_free(cb, sizeof (l2arc_write_callback_t)); } /* * A read to a cache device completed. Validate buffer contents before * handing over to the regular ARC routines. */ static void l2arc_read_done(zio_t *zio) { l2arc_read_callback_t *cb; arc_buf_hdr_t *hdr; arc_buf_t *buf; kmutex_t *hash_lock; int equal; ASSERT(zio->io_vd != NULL); ASSERT(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE); spa_config_exit(zio->io_spa, SCL_L2ARC, zio->io_vd); cb = zio->io_private; ASSERT(cb != NULL); buf = cb->l2rcb_buf; ASSERT(buf != NULL); hash_lock = HDR_LOCK(buf->b_hdr); mutex_enter(hash_lock); hdr = buf->b_hdr; ASSERT3P(hash_lock, ==, HDR_LOCK(hdr)); /* * If the buffer was compressed, decompress it first. */ if (cb->l2rcb_compress != ZIO_COMPRESS_OFF) l2arc_decompress_zio(zio, hdr, cb->l2rcb_compress); ASSERT(zio->io_data != NULL); ASSERT3U(zio->io_size, ==, hdr->b_size); ASSERT3U(BP_GET_LSIZE(&cb->l2rcb_bp), ==, hdr->b_size); /* * Check this survived the L2ARC journey. */ equal = arc_cksum_equal(buf); if (equal && zio->io_error == 0 && !HDR_L2_EVICTED(hdr)) { mutex_exit(hash_lock); zio->io_private = buf; zio->io_bp_copy = cb->l2rcb_bp; /* XXX fix in L2ARC 2.0 */ zio->io_bp = &zio->io_bp_copy; /* XXX fix in L2ARC 2.0 */ arc_read_done(zio); } else { mutex_exit(hash_lock); /* * Buffer didn't survive caching. Increment stats and * reissue to the original storage device. */ if (zio->io_error != 0) { ARCSTAT_BUMP(arcstat_l2_io_error); } else { zio->io_error = SET_ERROR(EIO); } if (!equal) ARCSTAT_BUMP(arcstat_l2_cksum_bad); /* * If there's no waiter, issue an async i/o to the primary * storage now. If there *is* a waiter, the caller must * issue the i/o in a context where it's OK to block. */ if (zio->io_waiter == NULL) { zio_t *pio = zio_unique_parent(zio); ASSERT(!pio || pio->io_child_type == ZIO_CHILD_LOGICAL); zio_nowait(zio_read(pio, cb->l2rcb_spa, &cb->l2rcb_bp, buf->b_data, hdr->b_size, arc_read_done, buf, zio->io_priority, cb->l2rcb_flags, &cb->l2rcb_zb)); } } kmem_free(cb, sizeof (l2arc_read_callback_t)); } /* * This is the list priority from which the L2ARC will search for pages to * cache. This is used within loops (0..3) to cycle through lists in the * desired order. This order can have a significant effect on cache * performance. * * Currently the metadata lists are hit first, MFU then MRU, followed by * the data lists. This function returns a locked list, and also returns * the lock pointer. */ static multilist_sublist_t * l2arc_sublist_lock(int list_num) { multilist_t *ml = NULL; unsigned int idx; ASSERT(list_num >= 0 && list_num <= 3); switch (list_num) { case 0: ml = &arc_mfu->arcs_list[ARC_BUFC_METADATA]; break; case 1: ml = &arc_mru->arcs_list[ARC_BUFC_METADATA]; break; case 2: ml = &arc_mfu->arcs_list[ARC_BUFC_DATA]; break; case 3: ml = &arc_mru->arcs_list[ARC_BUFC_DATA]; break; } /* * Return a randomly-selected sublist. This is acceptable * because the caller feeds only a little bit of data for each * call (8MB). Subsequent calls will result in different * sublists being selected. */ idx = multilist_get_random_index(ml); return (multilist_sublist_lock(ml, idx)); } /* * Evict buffers from the device write hand to the distance specified in * bytes. This distance may span populated buffers, it may span nothing. * This is clearing a region on the L2ARC device ready for writing. * If the 'all' boolean is set, every buffer is evicted. */ static void l2arc_evict(l2arc_dev_t *dev, uint64_t distance, boolean_t all) { list_t *buflist; arc_buf_hdr_t *hdr, *hdr_prev; kmutex_t *hash_lock; uint64_t taddr; buflist = &dev->l2ad_buflist; if (!all && dev->l2ad_first) { /* * This is the first sweep through the device. There is * nothing to evict. */ return; } if (dev->l2ad_hand >= (dev->l2ad_end - (2 * distance))) { /* * When nearing the end of the device, evict to the end * before the device write hand jumps to the start. */ taddr = dev->l2ad_end; } else { taddr = dev->l2ad_hand + distance; } DTRACE_PROBE4(l2arc__evict, l2arc_dev_t *, dev, list_t *, buflist, uint64_t, taddr, boolean_t, all); top: mutex_enter(&dev->l2ad_mtx); for (hdr = list_tail(buflist); hdr; hdr = hdr_prev) { hdr_prev = list_prev(buflist, hdr); hash_lock = HDR_LOCK(hdr); /* * We cannot use mutex_enter or else we can deadlock * with l2arc_write_buffers (due to swapping the order * the hash lock and l2ad_mtx are taken). */ if (!mutex_tryenter(hash_lock)) { /* * Missed the hash lock. Retry. */ ARCSTAT_BUMP(arcstat_l2_evict_lock_retry); mutex_exit(&dev->l2ad_mtx); mutex_enter(hash_lock); mutex_exit(hash_lock); goto top; } if (HDR_L2_WRITE_HEAD(hdr)) { /* * We hit a write head node. Leave it for * l2arc_write_done(). */ list_remove(buflist, hdr); mutex_exit(hash_lock); continue; } if (!all && HDR_HAS_L2HDR(hdr) && (hdr->b_l2hdr.b_daddr > taddr || hdr->b_l2hdr.b_daddr < dev->l2ad_hand)) { /* * We've evicted to the target address, * or the end of the device. */ mutex_exit(hash_lock); break; } ASSERT(HDR_HAS_L2HDR(hdr)); if (!HDR_HAS_L1HDR(hdr)) { ASSERT(!HDR_L2_READING(hdr)); /* * This doesn't exist in the ARC. Destroy. * arc_hdr_destroy() will call list_remove() * and decrement arcstat_l2_size. */ arc_change_state(arc_anon, hdr, hash_lock); arc_hdr_destroy(hdr); } else { ASSERT(hdr->b_l1hdr.b_state != arc_l2c_only); ARCSTAT_BUMP(arcstat_l2_evict_l1cached); /* * Invalidate issued or about to be issued * reads, since we may be about to write * over this location. */ if (HDR_L2_READING(hdr)) { ARCSTAT_BUMP(arcstat_l2_evict_reading); hdr->b_flags |= ARC_FLAG_L2_EVICTED; } /* Ensure this header has finished being written */ ASSERT(!HDR_L2_WRITING(hdr)); ASSERT3P(hdr->b_l1hdr.b_tmp_cdata, ==, NULL); arc_hdr_l2hdr_destroy(hdr); } mutex_exit(hash_lock); } mutex_exit(&dev->l2ad_mtx); } /* * Find and write ARC buffers to the L2ARC device. * * An ARC_FLAG_L2_WRITING flag is set so that the L2ARC buffers are not valid * for reading until they have completed writing. * The headroom_boost is an in-out parameter used to maintain headroom boost * state between calls to this function. * * Returns the number of bytes actually written (which may be smaller than * the delta by which the device hand has changed due to alignment). */ static uint64_t l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz, boolean_t *headroom_boost) { arc_buf_hdr_t *hdr, *hdr_prev, *head; uint64_t write_asize, write_sz, headroom, buf_compress_minsz; void *buf_data; boolean_t full; l2arc_write_callback_t *cb; zio_t *pio, *wzio; uint64_t guid = spa_load_guid(spa); const boolean_t do_headroom_boost = *headroom_boost; int try; ASSERT(dev->l2ad_vdev != NULL); /* Lower the flag now, we might want to raise it again later. */ *headroom_boost = B_FALSE; pio = NULL; write_sz = write_asize = 0; full = B_FALSE; head = kmem_cache_alloc(hdr_l2only_cache, KM_PUSHPAGE); head->b_flags |= ARC_FLAG_L2_WRITE_HEAD; head->b_flags |= ARC_FLAG_HAS_L2HDR; ARCSTAT_BUMP(arcstat_l2_write_buffer_iter); /* * We will want to try to compress buffers that are at least 2x the * device sector size. */ buf_compress_minsz = 2 << dev->l2ad_vdev->vdev_ashift; /* * Copy buffers for L2ARC writing. */ for (try = 0; try <= 3; try++) { multilist_sublist_t *mls = l2arc_sublist_lock(try); uint64_t passed_sz = 0; ARCSTAT_BUMP(arcstat_l2_write_buffer_list_iter); /* * L2ARC fast warmup. * * Until the ARC is warm and starts to evict, read from the * head of the ARC lists rather than the tail. */ if (arc_warm == B_FALSE) hdr = multilist_sublist_head(mls); else hdr = multilist_sublist_tail(mls); if (hdr == NULL) ARCSTAT_BUMP(arcstat_l2_write_buffer_list_null_iter); headroom = target_sz * l2arc_headroom; if (do_headroom_boost) headroom = (headroom * l2arc_headroom_boost) / 100; for (; hdr; hdr = hdr_prev) { kmutex_t *hash_lock; uint64_t buf_sz; uint64_t buf_a_sz; if (arc_warm == B_FALSE) hdr_prev = multilist_sublist_next(mls, hdr); else hdr_prev = multilist_sublist_prev(mls, hdr); ARCSTAT_INCR(arcstat_l2_write_buffer_bytes_scanned, hdr->b_size); hash_lock = HDR_LOCK(hdr); if (!mutex_tryenter(hash_lock)) { ARCSTAT_BUMP(arcstat_l2_write_trylock_fail); /* * Skip this buffer rather than waiting. */ continue; } passed_sz += hdr->b_size; if (passed_sz > headroom) { /* * Searched too far. */ mutex_exit(hash_lock); ARCSTAT_BUMP(arcstat_l2_write_passed_headroom); break; } if (!l2arc_write_eligible(guid, hdr)) { mutex_exit(hash_lock); continue; } /* * Assume that the buffer is not going to be compressed * and could take more space on disk because of a larger * disk block size. */ buf_sz = hdr->b_size; buf_a_sz = vdev_psize_to_asize(dev->l2ad_vdev, buf_sz); if ((write_asize + buf_a_sz) > target_sz) { full = B_TRUE; mutex_exit(hash_lock); ARCSTAT_BUMP(arcstat_l2_write_full); break; } if (pio == NULL) { /* * Insert a dummy header on the buflist so * l2arc_write_done() can find where the * write buffers begin without searching. */ mutex_enter(&dev->l2ad_mtx); list_insert_head(&dev->l2ad_buflist, head); mutex_exit(&dev->l2ad_mtx); cb = kmem_alloc( sizeof (l2arc_write_callback_t), KM_SLEEP); cb->l2wcb_dev = dev; cb->l2wcb_head = head; pio = zio_root(spa, l2arc_write_done, cb, ZIO_FLAG_CANFAIL); ARCSTAT_BUMP(arcstat_l2_write_pios); } /* * Create and add a new L2ARC header. */ hdr->b_l2hdr.b_dev = dev; hdr->b_flags |= ARC_FLAG_L2_WRITING; /* * Temporarily stash the data buffer in b_tmp_cdata. * The subsequent write step will pick it up from * there. This is because can't access b_l1hdr.b_buf * without holding the hash_lock, which we in turn * can't access without holding the ARC list locks * (which we want to avoid during compression/writing). */ hdr->b_l2hdr.b_compress = ZIO_COMPRESS_OFF; hdr->b_l2hdr.b_asize = hdr->b_size; hdr->b_l1hdr.b_tmp_cdata = hdr->b_l1hdr.b_buf->b_data; /* * Explicitly set the b_daddr field to a known * value which means "invalid address". This * enables us to differentiate which stage of * l2arc_write_buffers() the particular header * is in (e.g. this loop, or the one below). * ARC_FLAG_L2_WRITING is not enough to make * this distinction, and we need to know in * order to do proper l2arc vdev accounting in * arc_release() and arc_hdr_destroy(). * * Note, we can't use a new flag to distinguish * the two stages because we don't hold the * header's hash_lock below, in the second stage * of this function. Thus, we can't simply * change the b_flags field to denote that the * IO has been sent. We can change the b_daddr * field of the L2 portion, though, since we'll * be holding the l2ad_mtx; which is why we're * using it to denote the header's state change. */ hdr->b_l2hdr.b_daddr = L2ARC_ADDR_UNSET; hdr->b_flags |= ARC_FLAG_HAS_L2HDR; mutex_enter(&dev->l2ad_mtx); list_insert_head(&dev->l2ad_buflist, hdr); mutex_exit(&dev->l2ad_mtx); /* * Compute and store the buffer cksum before * writing. On debug the cksum is verified first. */ arc_cksum_verify(hdr->b_l1hdr.b_buf); arc_cksum_compute(hdr->b_l1hdr.b_buf, B_TRUE); mutex_exit(hash_lock); write_sz += buf_sz; write_asize += buf_a_sz; } multilist_sublist_unlock(mls); if (full == B_TRUE) break; } /* No buffers selected for writing? */ if (pio == NULL) { ASSERT0(write_sz); ASSERT(!HDR_HAS_L1HDR(head)); kmem_cache_free(hdr_l2only_cache, head); return (0); } mutex_enter(&dev->l2ad_mtx); /* * Note that elsewhere in this file arcstat_l2_asize * and the used space on l2ad_vdev are updated using b_asize, * which is not necessarily rounded up to the device block size. * Too keep accounting consistent we do the same here as well: * stats_size accumulates the sum of b_asize of the written buffers, * while write_asize accumulates the sum of b_asize rounded up * to the device block size. * The latter sum is used only to validate the corectness of the code. */ uint64_t stats_size = 0; write_asize = 0; /* * Now start writing the buffers. We're starting at the write head * and work backwards, retracing the course of the buffer selector * loop above. */ for (hdr = list_prev(&dev->l2ad_buflist, head); hdr; hdr = list_prev(&dev->l2ad_buflist, hdr)) { uint64_t buf_sz; /* * We rely on the L1 portion of the header below, so * it's invalid for this header to have been evicted out * of the ghost cache, prior to being written out. The * ARC_FLAG_L2_WRITING bit ensures this won't happen. */ ASSERT(HDR_HAS_L1HDR(hdr)); /* * We shouldn't need to lock the buffer here, since we flagged * it as ARC_FLAG_L2_WRITING in the previous step, but we must * take care to only access its L2 cache parameters. In * particular, hdr->l1hdr.b_buf may be invalid by now due to * ARC eviction. */ hdr->b_l2hdr.b_daddr = dev->l2ad_hand; if ((HDR_L2COMPRESS(hdr)) && hdr->b_l2hdr.b_asize >= buf_compress_minsz) { if (l2arc_compress_buf(hdr)) { /* * If compression succeeded, enable headroom * boost on the next scan cycle. */ *headroom_boost = B_TRUE; } } /* * Pick up the buffer data we had previously stashed away * (and now potentially also compressed). */ buf_data = hdr->b_l1hdr.b_tmp_cdata; buf_sz = hdr->b_l2hdr.b_asize; /* * We need to do this regardless if buf_sz is zero or * not, otherwise, when this l2hdr is evicted we'll * remove a reference that was never added. */ (void) refcount_add_many(&dev->l2ad_alloc, buf_sz, hdr); /* Compression may have squashed the buffer to zero length. */ if (buf_sz != 0) { uint64_t buf_a_sz; wzio = zio_write_phys(pio, dev->l2ad_vdev, dev->l2ad_hand, buf_sz, buf_data, ZIO_CHECKSUM_OFF, NULL, NULL, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_CANFAIL, B_FALSE); DTRACE_PROBE2(l2arc__write, vdev_t *, dev->l2ad_vdev, zio_t *, wzio); (void) zio_nowait(wzio); stats_size += buf_sz; /* * Keep the clock hand suitably device-aligned. */ buf_a_sz = vdev_psize_to_asize(dev->l2ad_vdev, buf_sz); write_asize += buf_a_sz; dev->l2ad_hand += buf_a_sz; } } mutex_exit(&dev->l2ad_mtx); ASSERT3U(write_asize, <=, target_sz); ARCSTAT_BUMP(arcstat_l2_writes_sent); ARCSTAT_INCR(arcstat_l2_write_bytes, write_asize); ARCSTAT_INCR(arcstat_l2_size, write_sz); ARCSTAT_INCR(arcstat_l2_asize, stats_size); vdev_space_update(dev->l2ad_vdev, stats_size, 0, 0); /* * Bump device hand to the device start if it is approaching the end. * l2arc_evict() will already have evicted ahead for this case. */ if (dev->l2ad_hand >= (dev->l2ad_end - target_sz)) { dev->l2ad_hand = dev->l2ad_start; dev->l2ad_first = B_FALSE; } dev->l2ad_writing = B_TRUE; (void) zio_wait(pio); dev->l2ad_writing = B_FALSE; return (write_asize); } /* * Compresses an L2ARC buffer. * The data to be compressed must be prefilled in l1hdr.b_tmp_cdata and its * size in l2hdr->b_asize. This routine tries to compress the data and * depending on the compression result there are three possible outcomes: * *) The buffer was incompressible. The original l2hdr contents were left * untouched and are ready for writing to an L2 device. * *) The buffer was all-zeros, so there is no need to write it to an L2 * device. To indicate this situation b_tmp_cdata is NULL'ed, b_asize is * set to zero and b_compress is set to ZIO_COMPRESS_EMPTY. * *) Compression succeeded and b_tmp_cdata was replaced with a temporary * data buffer which holds the compressed data to be written, and b_asize * tells us how much data there is. b_compress is set to the appropriate * compression algorithm. Once writing is done, invoke * l2arc_release_cdata_buf on this l2hdr to free this temporary buffer. * * Returns B_TRUE if compression succeeded, or B_FALSE if it didn't (the * buffer was incompressible). */ static boolean_t l2arc_compress_buf(arc_buf_hdr_t *hdr) { void *cdata; size_t csize, len, rounded; ASSERT(HDR_HAS_L2HDR(hdr)); l2arc_buf_hdr_t *l2hdr = &hdr->b_l2hdr; ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT3S(l2hdr->b_compress, ==, ZIO_COMPRESS_OFF); ASSERT(hdr->b_l1hdr.b_tmp_cdata != NULL); len = l2hdr->b_asize; cdata = zio_data_buf_alloc(len); ASSERT3P(cdata, !=, NULL); csize = zio_compress_data(ZIO_COMPRESS_LZ4, hdr->b_l1hdr.b_tmp_cdata, cdata, l2hdr->b_asize); if (csize == 0) { /* zero block, indicate that there's nothing to write */ zio_data_buf_free(cdata, len); l2hdr->b_compress = ZIO_COMPRESS_EMPTY; l2hdr->b_asize = 0; hdr->b_l1hdr.b_tmp_cdata = NULL; ARCSTAT_BUMP(arcstat_l2_compress_zeros); return (B_TRUE); } rounded = P2ROUNDUP(csize, (size_t)1 << l2hdr->b_dev->l2ad_vdev->vdev_ashift); if (rounded < len) { /* * Compression succeeded, we'll keep the cdata around for * writing and release it afterwards. */ if (rounded > csize) { bzero((char *)cdata + csize, rounded - csize); csize = rounded; } l2hdr->b_compress = ZIO_COMPRESS_LZ4; l2hdr->b_asize = csize; hdr->b_l1hdr.b_tmp_cdata = cdata; ARCSTAT_BUMP(arcstat_l2_compress_successes); return (B_TRUE); } else { /* * Compression failed, release the compressed buffer. * l2hdr will be left unmodified. */ zio_data_buf_free(cdata, len); ARCSTAT_BUMP(arcstat_l2_compress_failures); return (B_FALSE); } } /* * Decompresses a zio read back from an l2arc device. On success, the * underlying zio's io_data buffer is overwritten by the uncompressed * version. On decompression error (corrupt compressed stream), the * zio->io_error value is set to signal an I/O error. * * Please note that the compressed data stream is not checksummed, so * if the underlying device is experiencing data corruption, we may feed * corrupt data to the decompressor, so the decompressor needs to be * able to handle this situation (LZ4 does). */ static void l2arc_decompress_zio(zio_t *zio, arc_buf_hdr_t *hdr, enum zio_compress c) { ASSERT(L2ARC_IS_VALID_COMPRESS(c)); if (zio->io_error != 0) { /* * An io error has occured, just restore the original io * size in preparation for a main pool read. */ zio->io_orig_size = zio->io_size = hdr->b_size; return; } if (c == ZIO_COMPRESS_EMPTY) { /* * An empty buffer results in a null zio, which means we * need to fill its io_data after we're done restoring the * buffer's contents. */ ASSERT(hdr->b_l1hdr.b_buf != NULL); bzero(hdr->b_l1hdr.b_buf->b_data, hdr->b_size); zio->io_data = zio->io_orig_data = hdr->b_l1hdr.b_buf->b_data; } else { ASSERT(zio->io_data != NULL); /* * We copy the compressed data from the start of the arc buffer * (the zio_read will have pulled in only what we need, the * rest is garbage which we will overwrite at decompression) * and then decompress back to the ARC data buffer. This way we * can minimize copying by simply decompressing back over the * original compressed data (rather than decompressing to an * aux buffer and then copying back the uncompressed buffer, * which is likely to be much larger). */ uint64_t csize; void *cdata; csize = zio->io_size; cdata = zio_data_buf_alloc(csize); bcopy(zio->io_data, cdata, csize); if (zio_decompress_data(c, cdata, zio->io_data, csize, hdr->b_size) != 0) zio->io_error = EIO; zio_data_buf_free(cdata, csize); } /* Restore the expected uncompressed IO size. */ zio->io_orig_size = zio->io_size = hdr->b_size; } /* * Releases the temporary b_tmp_cdata buffer in an l2arc header structure. * This buffer serves as a temporary holder of compressed data while * the buffer entry is being written to an l2arc device. Once that is * done, we can dispose of it. */ static void l2arc_release_cdata_buf(arc_buf_hdr_t *hdr) { ASSERT(HDR_HAS_L2HDR(hdr)); enum zio_compress comp = hdr->b_l2hdr.b_compress; ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(comp == ZIO_COMPRESS_OFF || L2ARC_IS_VALID_COMPRESS(comp)); if (comp == ZIO_COMPRESS_OFF) { /* * In this case, b_tmp_cdata points to the same buffer * as the arc_buf_t's b_data field. We don't want to * free it, since the arc_buf_t will handle that. */ hdr->b_l1hdr.b_tmp_cdata = NULL; } else if (comp == ZIO_COMPRESS_EMPTY) { /* * In this case, b_tmp_cdata was compressed to an empty * buffer, thus there's nothing to free and b_tmp_cdata * should have been set to NULL in l2arc_write_buffers(). */ ASSERT3P(hdr->b_l1hdr.b_tmp_cdata, ==, NULL); } else { /* * If the data was compressed, then we've allocated a * temporary buffer for it, so now we need to release it. */ ASSERT(hdr->b_l1hdr.b_tmp_cdata != NULL); zio_data_buf_free(hdr->b_l1hdr.b_tmp_cdata, hdr->b_size); hdr->b_l1hdr.b_tmp_cdata = NULL; } } /* * This thread feeds the L2ARC at regular intervals. This is the beating * heart of the L2ARC. */ static void l2arc_feed_thread(void *dummy __unused) { callb_cpr_t cpr; l2arc_dev_t *dev; spa_t *spa; uint64_t size, wrote; clock_t begin, next = ddi_get_lbolt(); boolean_t headroom_boost = B_FALSE; CALLB_CPR_INIT(&cpr, &l2arc_feed_thr_lock, callb_generic_cpr, FTAG); mutex_enter(&l2arc_feed_thr_lock); while (l2arc_thread_exit == 0) { CALLB_CPR_SAFE_BEGIN(&cpr); (void) cv_timedwait(&l2arc_feed_thr_cv, &l2arc_feed_thr_lock, next - ddi_get_lbolt()); CALLB_CPR_SAFE_END(&cpr, &l2arc_feed_thr_lock); next = ddi_get_lbolt() + hz; /* * Quick check for L2ARC devices. */ mutex_enter(&l2arc_dev_mtx); if (l2arc_ndev == 0) { mutex_exit(&l2arc_dev_mtx); continue; } mutex_exit(&l2arc_dev_mtx); begin = ddi_get_lbolt(); /* * This selects the next l2arc device to write to, and in * doing so the next spa to feed from: dev->l2ad_spa. This * will return NULL if there are now no l2arc devices or if * they are all faulted. * * If a device is returned, its spa's config lock is also * held to prevent device removal. l2arc_dev_get_next() * will grab and release l2arc_dev_mtx. */ if ((dev = l2arc_dev_get_next()) == NULL) continue; spa = dev->l2ad_spa; ASSERT(spa != NULL); /* * If the pool is read-only then force the feed thread to * sleep a little longer. */ if (!spa_writeable(spa)) { next = ddi_get_lbolt() + 5 * l2arc_feed_secs * hz; spa_config_exit(spa, SCL_L2ARC, dev); continue; } /* * Avoid contributing to memory pressure. */ if (arc_reclaim_needed()) { ARCSTAT_BUMP(arcstat_l2_abort_lowmem); spa_config_exit(spa, SCL_L2ARC, dev); continue; } ARCSTAT_BUMP(arcstat_l2_feeds); size = l2arc_write_size(); /* * Evict L2ARC buffers that will be overwritten. */ l2arc_evict(dev, size, B_FALSE); /* * Write ARC buffers. */ wrote = l2arc_write_buffers(spa, dev, size, &headroom_boost); /* * Calculate interval between writes. */ next = l2arc_write_interval(begin, size, wrote); spa_config_exit(spa, SCL_L2ARC, dev); } l2arc_thread_exit = 0; cv_broadcast(&l2arc_feed_thr_cv); CALLB_CPR_EXIT(&cpr); /* drops l2arc_feed_thr_lock */ thread_exit(); } boolean_t l2arc_vdev_present(vdev_t *vd) { l2arc_dev_t *dev; mutex_enter(&l2arc_dev_mtx); for (dev = list_head(l2arc_dev_list); dev != NULL; dev = list_next(l2arc_dev_list, dev)) { if (dev->l2ad_vdev == vd) break; } mutex_exit(&l2arc_dev_mtx); return (dev != NULL); } /* * Add a vdev for use by the L2ARC. By this point the spa has already * validated the vdev and opened it. */ void l2arc_add_vdev(spa_t *spa, vdev_t *vd) { l2arc_dev_t *adddev; ASSERT(!l2arc_vdev_present(vd)); vdev_ashift_optimize(vd); /* * Create a new l2arc device entry. */ adddev = kmem_zalloc(sizeof (l2arc_dev_t), KM_SLEEP); adddev->l2ad_spa = spa; adddev->l2ad_vdev = vd; adddev->l2ad_start = VDEV_LABEL_START_SIZE; adddev->l2ad_end = VDEV_LABEL_START_SIZE + vdev_get_min_asize(vd); adddev->l2ad_hand = adddev->l2ad_start; adddev->l2ad_first = B_TRUE; adddev->l2ad_writing = B_FALSE; mutex_init(&adddev->l2ad_mtx, NULL, MUTEX_DEFAULT, NULL); /* * This is a list of all ARC buffers that are still valid on the * device. */ list_create(&adddev->l2ad_buflist, sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l2hdr.b_l2node)); vdev_space_update(vd, 0, 0, adddev->l2ad_end - adddev->l2ad_hand); refcount_create(&adddev->l2ad_alloc); /* * Add device to global list */ mutex_enter(&l2arc_dev_mtx); list_insert_head(l2arc_dev_list, adddev); atomic_inc_64(&l2arc_ndev); mutex_exit(&l2arc_dev_mtx); } /* * Remove a vdev from the L2ARC. */ void l2arc_remove_vdev(vdev_t *vd) { l2arc_dev_t *dev, *nextdev, *remdev = NULL; /* * Find the device by vdev */ mutex_enter(&l2arc_dev_mtx); for (dev = list_head(l2arc_dev_list); dev; dev = nextdev) { nextdev = list_next(l2arc_dev_list, dev); if (vd == dev->l2ad_vdev) { remdev = dev; break; } } ASSERT(remdev != NULL); /* * Remove device from global list */ list_remove(l2arc_dev_list, remdev); l2arc_dev_last = NULL; /* may have been invalidated */ atomic_dec_64(&l2arc_ndev); mutex_exit(&l2arc_dev_mtx); /* * Clear all buflists and ARC references. L2ARC device flush. */ l2arc_evict(remdev, 0, B_TRUE); list_destroy(&remdev->l2ad_buflist); mutex_destroy(&remdev->l2ad_mtx); refcount_destroy(&remdev->l2ad_alloc); kmem_free(remdev, sizeof (l2arc_dev_t)); } void l2arc_init(void) { l2arc_thread_exit = 0; l2arc_ndev = 0; l2arc_writes_sent = 0; l2arc_writes_done = 0; mutex_init(&l2arc_feed_thr_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&l2arc_feed_thr_cv, NULL, CV_DEFAULT, NULL); mutex_init(&l2arc_dev_mtx, NULL, MUTEX_DEFAULT, NULL); mutex_init(&l2arc_free_on_write_mtx, NULL, MUTEX_DEFAULT, NULL); l2arc_dev_list = &L2ARC_dev_list; l2arc_free_on_write = &L2ARC_free_on_write; list_create(l2arc_dev_list, sizeof (l2arc_dev_t), offsetof(l2arc_dev_t, l2ad_node)); list_create(l2arc_free_on_write, sizeof (l2arc_data_free_t), offsetof(l2arc_data_free_t, l2df_list_node)); } void l2arc_fini(void) { /* * This is called from dmu_fini(), which is called from spa_fini(); * Because of this, we can assume that all l2arc devices have * already been removed when the pools themselves were removed. */ l2arc_do_free_on_write(); mutex_destroy(&l2arc_feed_thr_lock); cv_destroy(&l2arc_feed_thr_cv); mutex_destroy(&l2arc_dev_mtx); mutex_destroy(&l2arc_free_on_write_mtx); list_destroy(l2arc_dev_list); list_destroy(l2arc_free_on_write); } void l2arc_start(void) { if (!(spa_mode_global & FWRITE)) return; (void) thread_create(NULL, 0, l2arc_feed_thread, NULL, 0, &p0, TS_RUN, minclsyspri); } void l2arc_stop(void) { if (!(spa_mode_global & FWRITE)) return; mutex_enter(&l2arc_feed_thr_lock); cv_signal(&l2arc_feed_thr_cv); /* kick thread out of startup */ l2arc_thread_exit = 1; while (l2arc_thread_exit != 0) cv_wait(&l2arc_feed_thr_cv, &l2arc_feed_thr_lock); mutex_exit(&l2arc_feed_thr_lock); } Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/ddt.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/ddt.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/ddt.c (revision 290757) @@ -1,1164 +1,1165 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved. - * Copyright (c) 2012, 2014 by Delphix. All rights reserved. + * Copyright (c) 2012, 2015 by Delphix. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include /* * Enable/disable prefetching of dedup-ed blocks which are going to be freed. */ int zfs_dedup_prefetch = 1; SYSCTL_DECL(_vfs_zfs); SYSCTL_NODE(_vfs_zfs, OID_AUTO, dedup, CTLFLAG_RW, 0, "ZFS DEDUP"); TUNABLE_INT("vfs.zfs.dedup.prefetch", &zfs_dedup_prefetch); SYSCTL_INT(_vfs_zfs_dedup, OID_AUTO, prefetch, CTLFLAG_RW, &zfs_dedup_prefetch, 0, "Enable/disable prefetching of dedup-ed blocks which are going to be freed"); static const ddt_ops_t *ddt_ops[DDT_TYPES] = { &ddt_zap_ops, }; static const char *ddt_class_name[DDT_CLASSES] = { "ditto", "duplicate", "unique", }; static void ddt_object_create(ddt_t *ddt, enum ddt_type type, enum ddt_class class, dmu_tx_t *tx) { spa_t *spa = ddt->ddt_spa; objset_t *os = ddt->ddt_os; uint64_t *objectp = &ddt->ddt_object[type][class]; - boolean_t prehash = zio_checksum_table[ddt->ddt_checksum].ci_dedup; + boolean_t prehash = zio_checksum_table[ddt->ddt_checksum].ci_flags & + ZCHECKSUM_FLAG_DEDUP; char name[DDT_NAMELEN]; ddt_object_name(ddt, type, class, name); ASSERT(*objectp == 0); VERIFY(ddt_ops[type]->ddt_op_create(os, objectp, tx, prehash) == 0); ASSERT(*objectp != 0); VERIFY(zap_add(os, DMU_POOL_DIRECTORY_OBJECT, name, sizeof (uint64_t), 1, objectp, tx) == 0); VERIFY(zap_add(os, spa->spa_ddt_stat_object, name, sizeof (uint64_t), sizeof (ddt_histogram_t) / sizeof (uint64_t), &ddt->ddt_histogram[type][class], tx) == 0); } static void ddt_object_destroy(ddt_t *ddt, enum ddt_type type, enum ddt_class class, dmu_tx_t *tx) { spa_t *spa = ddt->ddt_spa; objset_t *os = ddt->ddt_os; uint64_t *objectp = &ddt->ddt_object[type][class]; uint64_t count; char name[DDT_NAMELEN]; ddt_object_name(ddt, type, class, name); ASSERT(*objectp != 0); VERIFY(ddt_object_count(ddt, type, class, &count) == 0 && count == 0); ASSERT(ddt_histogram_empty(&ddt->ddt_histogram[type][class])); VERIFY(zap_remove(os, DMU_POOL_DIRECTORY_OBJECT, name, tx) == 0); VERIFY(zap_remove(os, spa->spa_ddt_stat_object, name, tx) == 0); VERIFY(ddt_ops[type]->ddt_op_destroy(os, *objectp, tx) == 0); bzero(&ddt->ddt_object_stats[type][class], sizeof (ddt_object_t)); *objectp = 0; } static int ddt_object_load(ddt_t *ddt, enum ddt_type type, enum ddt_class class) { ddt_object_t *ddo = &ddt->ddt_object_stats[type][class]; dmu_object_info_t doi; uint64_t count; char name[DDT_NAMELEN]; int error; ddt_object_name(ddt, type, class, name); error = zap_lookup(ddt->ddt_os, DMU_POOL_DIRECTORY_OBJECT, name, sizeof (uint64_t), 1, &ddt->ddt_object[type][class]); if (error != 0) return (error); VERIFY0(zap_lookup(ddt->ddt_os, ddt->ddt_spa->spa_ddt_stat_object, name, sizeof (uint64_t), sizeof (ddt_histogram_t) / sizeof (uint64_t), &ddt->ddt_histogram[type][class])); /* * Seed the cached statistics. */ VERIFY(ddt_object_info(ddt, type, class, &doi) == 0); error = ddt_object_count(ddt, type, class, &count); if (error) return error; ddo->ddo_count = count; ddo->ddo_dspace = doi.doi_physical_blocks_512 << 9; ddo->ddo_mspace = doi.doi_fill_count * doi.doi_data_block_size; return (0); } static void ddt_object_sync(ddt_t *ddt, enum ddt_type type, enum ddt_class class, dmu_tx_t *tx) { ddt_object_t *ddo = &ddt->ddt_object_stats[type][class]; dmu_object_info_t doi; uint64_t count; char name[DDT_NAMELEN]; ddt_object_name(ddt, type, class, name); VERIFY(zap_update(ddt->ddt_os, ddt->ddt_spa->spa_ddt_stat_object, name, sizeof (uint64_t), sizeof (ddt_histogram_t) / sizeof (uint64_t), &ddt->ddt_histogram[type][class], tx) == 0); /* * Cache DDT statistics; this is the only time they'll change. */ VERIFY(ddt_object_info(ddt, type, class, &doi) == 0); VERIFY(ddt_object_count(ddt, type, class, &count) == 0); ddo->ddo_count = count; ddo->ddo_dspace = doi.doi_physical_blocks_512 << 9; ddo->ddo_mspace = doi.doi_fill_count * doi.doi_data_block_size; } static int ddt_object_lookup(ddt_t *ddt, enum ddt_type type, enum ddt_class class, ddt_entry_t *dde) { if (!ddt_object_exists(ddt, type, class)) return (SET_ERROR(ENOENT)); return (ddt_ops[type]->ddt_op_lookup(ddt->ddt_os, ddt->ddt_object[type][class], dde)); } static void ddt_object_prefetch(ddt_t *ddt, enum ddt_type type, enum ddt_class class, ddt_entry_t *dde) { if (!ddt_object_exists(ddt, type, class)) return; ddt_ops[type]->ddt_op_prefetch(ddt->ddt_os, ddt->ddt_object[type][class], dde); } int ddt_object_update(ddt_t *ddt, enum ddt_type type, enum ddt_class class, ddt_entry_t *dde, dmu_tx_t *tx) { ASSERT(ddt_object_exists(ddt, type, class)); return (ddt_ops[type]->ddt_op_update(ddt->ddt_os, ddt->ddt_object[type][class], dde, tx)); } static int ddt_object_remove(ddt_t *ddt, enum ddt_type type, enum ddt_class class, ddt_entry_t *dde, dmu_tx_t *tx) { ASSERT(ddt_object_exists(ddt, type, class)); return (ddt_ops[type]->ddt_op_remove(ddt->ddt_os, ddt->ddt_object[type][class], dde, tx)); } int ddt_object_walk(ddt_t *ddt, enum ddt_type type, enum ddt_class class, uint64_t *walk, ddt_entry_t *dde) { ASSERT(ddt_object_exists(ddt, type, class)); return (ddt_ops[type]->ddt_op_walk(ddt->ddt_os, ddt->ddt_object[type][class], dde, walk)); } int ddt_object_count(ddt_t *ddt, enum ddt_type type, enum ddt_class class, uint64_t *count) { ASSERT(ddt_object_exists(ddt, type, class)); return (ddt_ops[type]->ddt_op_count(ddt->ddt_os, ddt->ddt_object[type][class], count)); } int ddt_object_info(ddt_t *ddt, enum ddt_type type, enum ddt_class class, dmu_object_info_t *doi) { if (!ddt_object_exists(ddt, type, class)) return (SET_ERROR(ENOENT)); return (dmu_object_info(ddt->ddt_os, ddt->ddt_object[type][class], doi)); } boolean_t ddt_object_exists(ddt_t *ddt, enum ddt_type type, enum ddt_class class) { return (!!ddt->ddt_object[type][class]); } void ddt_object_name(ddt_t *ddt, enum ddt_type type, enum ddt_class class, char *name) { (void) sprintf(name, DMU_POOL_DDT, zio_checksum_table[ddt->ddt_checksum].ci_name, ddt_ops[type]->ddt_op_name, ddt_class_name[class]); } void ddt_bp_fill(const ddt_phys_t *ddp, blkptr_t *bp, uint64_t txg) { ASSERT(txg != 0); for (int d = 0; d < SPA_DVAS_PER_BP; d++) bp->blk_dva[d] = ddp->ddp_dva[d]; BP_SET_BIRTH(bp, txg, ddp->ddp_phys_birth); } void ddt_bp_create(enum zio_checksum checksum, const ddt_key_t *ddk, const ddt_phys_t *ddp, blkptr_t *bp) { BP_ZERO(bp); if (ddp != NULL) ddt_bp_fill(ddp, bp, ddp->ddp_phys_birth); bp->blk_cksum = ddk->ddk_cksum; bp->blk_fill = 1; BP_SET_LSIZE(bp, DDK_GET_LSIZE(ddk)); BP_SET_PSIZE(bp, DDK_GET_PSIZE(ddk)); BP_SET_COMPRESS(bp, DDK_GET_COMPRESS(ddk)); BP_SET_CHECKSUM(bp, checksum); BP_SET_TYPE(bp, DMU_OT_DEDUP); BP_SET_LEVEL(bp, 0); BP_SET_DEDUP(bp, 0); BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER); } void ddt_key_fill(ddt_key_t *ddk, const blkptr_t *bp) { ddk->ddk_cksum = bp->blk_cksum; ddk->ddk_prop = 0; DDK_SET_LSIZE(ddk, BP_GET_LSIZE(bp)); DDK_SET_PSIZE(ddk, BP_GET_PSIZE(bp)); DDK_SET_COMPRESS(ddk, BP_GET_COMPRESS(bp)); } void ddt_phys_fill(ddt_phys_t *ddp, const blkptr_t *bp) { ASSERT(ddp->ddp_phys_birth == 0); for (int d = 0; d < SPA_DVAS_PER_BP; d++) ddp->ddp_dva[d] = bp->blk_dva[d]; ddp->ddp_phys_birth = BP_PHYSICAL_BIRTH(bp); } void ddt_phys_clear(ddt_phys_t *ddp) { bzero(ddp, sizeof (*ddp)); } void ddt_phys_addref(ddt_phys_t *ddp) { ddp->ddp_refcnt++; } void ddt_phys_decref(ddt_phys_t *ddp) { ASSERT((int64_t)ddp->ddp_refcnt > 0); ddp->ddp_refcnt--; } void ddt_phys_free(ddt_t *ddt, ddt_key_t *ddk, ddt_phys_t *ddp, uint64_t txg) { blkptr_t blk; ddt_bp_create(ddt->ddt_checksum, ddk, ddp, &blk); ddt_phys_clear(ddp); zio_free(ddt->ddt_spa, txg, &blk); } ddt_phys_t * ddt_phys_select(const ddt_entry_t *dde, const blkptr_t *bp) { ddt_phys_t *ddp = (ddt_phys_t *)dde->dde_phys; for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) { if (DVA_EQUAL(BP_IDENTITY(bp), &ddp->ddp_dva[0]) && BP_PHYSICAL_BIRTH(bp) == ddp->ddp_phys_birth) return (ddp); } return (NULL); } uint64_t ddt_phys_total_refcnt(const ddt_entry_t *dde) { uint64_t refcnt = 0; for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) refcnt += dde->dde_phys[p].ddp_refcnt; return (refcnt); } static void ddt_stat_generate(ddt_t *ddt, ddt_entry_t *dde, ddt_stat_t *dds) { spa_t *spa = ddt->ddt_spa; ddt_phys_t *ddp = dde->dde_phys; ddt_key_t *ddk = &dde->dde_key; uint64_t lsize = DDK_GET_LSIZE(ddk); uint64_t psize = DDK_GET_PSIZE(ddk); bzero(dds, sizeof (*dds)); for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) { uint64_t dsize = 0; uint64_t refcnt = ddp->ddp_refcnt; if (ddp->ddp_phys_birth == 0) continue; for (int d = 0; d < SPA_DVAS_PER_BP; d++) dsize += dva_get_dsize_sync(spa, &ddp->ddp_dva[d]); dds->dds_blocks += 1; dds->dds_lsize += lsize; dds->dds_psize += psize; dds->dds_dsize += dsize; dds->dds_ref_blocks += refcnt; dds->dds_ref_lsize += lsize * refcnt; dds->dds_ref_psize += psize * refcnt; dds->dds_ref_dsize += dsize * refcnt; } } void ddt_stat_add(ddt_stat_t *dst, const ddt_stat_t *src, uint64_t neg) { const uint64_t *s = (const uint64_t *)src; uint64_t *d = (uint64_t *)dst; uint64_t *d_end = (uint64_t *)(dst + 1); ASSERT(neg == 0 || neg == -1ULL); /* add or subtract */ while (d < d_end) *d++ += (*s++ ^ neg) - neg; } static void ddt_stat_update(ddt_t *ddt, ddt_entry_t *dde, uint64_t neg) { ddt_stat_t dds; ddt_histogram_t *ddh; int bucket; ddt_stat_generate(ddt, dde, &dds); bucket = highbit64(dds.dds_ref_blocks) - 1; ASSERT(bucket >= 0); ddh = &ddt->ddt_histogram[dde->dde_type][dde->dde_class]; ddt_stat_add(&ddh->ddh_stat[bucket], &dds, neg); } void ddt_histogram_add(ddt_histogram_t *dst, const ddt_histogram_t *src) { for (int h = 0; h < 64; h++) ddt_stat_add(&dst->ddh_stat[h], &src->ddh_stat[h], 0); } void ddt_histogram_stat(ddt_stat_t *dds, const ddt_histogram_t *ddh) { bzero(dds, sizeof (*dds)); for (int h = 0; h < 64; h++) ddt_stat_add(dds, &ddh->ddh_stat[h], 0); } boolean_t ddt_histogram_empty(const ddt_histogram_t *ddh) { const uint64_t *s = (const uint64_t *)ddh; const uint64_t *s_end = (const uint64_t *)(ddh + 1); while (s < s_end) if (*s++ != 0) return (B_FALSE); return (B_TRUE); } void ddt_get_dedup_object_stats(spa_t *spa, ddt_object_t *ddo_total) { /* Sum the statistics we cached in ddt_object_sync(). */ for (enum zio_checksum c = 0; c < ZIO_CHECKSUM_FUNCTIONS; c++) { ddt_t *ddt = spa->spa_ddt[c]; for (enum ddt_type type = 0; type < DDT_TYPES; type++) { for (enum ddt_class class = 0; class < DDT_CLASSES; class++) { ddt_object_t *ddo = &ddt->ddt_object_stats[type][class]; ddo_total->ddo_count += ddo->ddo_count; ddo_total->ddo_dspace += ddo->ddo_dspace; ddo_total->ddo_mspace += ddo->ddo_mspace; } } } /* ... and compute the averages. */ if (ddo_total->ddo_count != 0) { ddo_total->ddo_dspace /= ddo_total->ddo_count; ddo_total->ddo_mspace /= ddo_total->ddo_count; } } void ddt_get_dedup_histogram(spa_t *spa, ddt_histogram_t *ddh) { for (enum zio_checksum c = 0; c < ZIO_CHECKSUM_FUNCTIONS; c++) { ddt_t *ddt = spa->spa_ddt[c]; for (enum ddt_type type = 0; type < DDT_TYPES; type++) { for (enum ddt_class class = 0; class < DDT_CLASSES; class++) { ddt_histogram_add(ddh, &ddt->ddt_histogram_cache[type][class]); } } } } void ddt_get_dedup_stats(spa_t *spa, ddt_stat_t *dds_total) { ddt_histogram_t *ddh_total; ddh_total = kmem_zalloc(sizeof (ddt_histogram_t), KM_SLEEP); ddt_get_dedup_histogram(spa, ddh_total); ddt_histogram_stat(dds_total, ddh_total); kmem_free(ddh_total, sizeof (ddt_histogram_t)); } uint64_t ddt_get_dedup_dspace(spa_t *spa) { ddt_stat_t dds_total = { 0 }; ddt_get_dedup_stats(spa, &dds_total); return (dds_total.dds_ref_dsize - dds_total.dds_dsize); } uint64_t ddt_get_pool_dedup_ratio(spa_t *spa) { ddt_stat_t dds_total = { 0 }; ddt_get_dedup_stats(spa, &dds_total); if (dds_total.dds_dsize == 0) return (100); return (dds_total.dds_ref_dsize * 100 / dds_total.dds_dsize); } int ddt_ditto_copies_needed(ddt_t *ddt, ddt_entry_t *dde, ddt_phys_t *ddp_willref) { spa_t *spa = ddt->ddt_spa; uint64_t total_refcnt = 0; uint64_t ditto = spa->spa_dedup_ditto; int total_copies = 0; int desired_copies = 0; for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) { ddt_phys_t *ddp = &dde->dde_phys[p]; zio_t *zio = dde->dde_lead_zio[p]; uint64_t refcnt = ddp->ddp_refcnt; /* committed refs */ if (zio != NULL) refcnt += zio->io_parent_count; /* pending refs */ if (ddp == ddp_willref) refcnt++; /* caller's ref */ if (refcnt != 0) { total_refcnt += refcnt; total_copies += p; } } if (ditto == 0 || ditto > UINT32_MAX) ditto = UINT32_MAX; if (total_refcnt >= 1) desired_copies++; if (total_refcnt >= ditto) desired_copies++; if (total_refcnt >= ditto * ditto) desired_copies++; return (MAX(desired_copies, total_copies) - total_copies); } int ddt_ditto_copies_present(ddt_entry_t *dde) { ddt_phys_t *ddp = &dde->dde_phys[DDT_PHYS_DITTO]; dva_t *dva = ddp->ddp_dva; int copies = 0 - DVA_GET_GANG(dva); for (int d = 0; d < SPA_DVAS_PER_BP; d++, dva++) if (DVA_IS_VALID(dva)) copies++; ASSERT(copies >= 0 && copies < SPA_DVAS_PER_BP); return (copies); } size_t ddt_compress(void *src, uchar_t *dst, size_t s_len, size_t d_len) { uchar_t *version = dst++; int cpfunc = ZIO_COMPRESS_ZLE; zio_compress_info_t *ci = &zio_compress_table[cpfunc]; size_t c_len; ASSERT(d_len >= s_len + 1); /* no compression plus version byte */ c_len = ci->ci_compress(src, dst, s_len, d_len - 1, ci->ci_level); if (c_len == s_len) { cpfunc = ZIO_COMPRESS_OFF; bcopy(src, dst, s_len); } *version = cpfunc; /* CONSTCOND */ if (ZFS_HOST_BYTEORDER) *version |= DDT_COMPRESS_BYTEORDER_MASK; return (c_len + 1); } void ddt_decompress(uchar_t *src, void *dst, size_t s_len, size_t d_len) { uchar_t version = *src++; int cpfunc = version & DDT_COMPRESS_FUNCTION_MASK; zio_compress_info_t *ci = &zio_compress_table[cpfunc]; if (ci->ci_decompress != NULL) (void) ci->ci_decompress(src, dst, s_len, d_len, ci->ci_level); else bcopy(src, dst, d_len); if (((version & DDT_COMPRESS_BYTEORDER_MASK) != 0) != (ZFS_HOST_BYTEORDER != 0)) byteswap_uint64_array(dst, d_len); } ddt_t * ddt_select_by_checksum(spa_t *spa, enum zio_checksum c) { return (spa->spa_ddt[c]); } ddt_t * ddt_select(spa_t *spa, const blkptr_t *bp) { return (spa->spa_ddt[BP_GET_CHECKSUM(bp)]); } void ddt_enter(ddt_t *ddt) { mutex_enter(&ddt->ddt_lock); } void ddt_exit(ddt_t *ddt) { mutex_exit(&ddt->ddt_lock); } static ddt_entry_t * ddt_alloc(const ddt_key_t *ddk) { ddt_entry_t *dde; dde = kmem_zalloc(sizeof (ddt_entry_t), KM_SLEEP); cv_init(&dde->dde_cv, NULL, CV_DEFAULT, NULL); dde->dde_key = *ddk; return (dde); } static void ddt_free(ddt_entry_t *dde) { ASSERT(!dde->dde_loading); for (int p = 0; p < DDT_PHYS_TYPES; p++) ASSERT(dde->dde_lead_zio[p] == NULL); if (dde->dde_repair_data != NULL) zio_buf_free(dde->dde_repair_data, DDK_GET_PSIZE(&dde->dde_key)); cv_destroy(&dde->dde_cv); kmem_free(dde, sizeof (*dde)); } void ddt_remove(ddt_t *ddt, ddt_entry_t *dde) { ASSERT(MUTEX_HELD(&ddt->ddt_lock)); avl_remove(&ddt->ddt_tree, dde); ddt_free(dde); } ddt_entry_t * ddt_lookup(ddt_t *ddt, const blkptr_t *bp, boolean_t add) { ddt_entry_t *dde, dde_search; enum ddt_type type; enum ddt_class class; avl_index_t where; int error; ASSERT(MUTEX_HELD(&ddt->ddt_lock)); ddt_key_fill(&dde_search.dde_key, bp); dde = avl_find(&ddt->ddt_tree, &dde_search, &where); if (dde == NULL) { if (!add) return (NULL); dde = ddt_alloc(&dde_search.dde_key); avl_insert(&ddt->ddt_tree, dde, where); } while (dde->dde_loading) cv_wait(&dde->dde_cv, &ddt->ddt_lock); if (dde->dde_loaded) return (dde); dde->dde_loading = B_TRUE; ddt_exit(ddt); error = ENOENT; for (type = 0; type < DDT_TYPES; type++) { for (class = 0; class < DDT_CLASSES; class++) { error = ddt_object_lookup(ddt, type, class, dde); if (error != ENOENT) break; } if (error != ENOENT) break; } ASSERT(error == 0 || error == ENOENT); ddt_enter(ddt); ASSERT(dde->dde_loaded == B_FALSE); ASSERT(dde->dde_loading == B_TRUE); dde->dde_type = type; /* will be DDT_TYPES if no entry found */ dde->dde_class = class; /* will be DDT_CLASSES if no entry found */ dde->dde_loaded = B_TRUE; dde->dde_loading = B_FALSE; if (error == 0) ddt_stat_update(ddt, dde, -1ULL); cv_broadcast(&dde->dde_cv); return (dde); } void ddt_prefetch(spa_t *spa, const blkptr_t *bp) { ddt_t *ddt; ddt_entry_t dde; if (!zfs_dedup_prefetch || bp == NULL || !BP_GET_DEDUP(bp)) return; /* * We only remove the DDT once all tables are empty and only * prefetch dedup blocks when there are entries in the DDT. * Thus no locking is required as the DDT can't disappear on us. */ ddt = ddt_select(spa, bp); ddt_key_fill(&dde.dde_key, bp); for (enum ddt_type type = 0; type < DDT_TYPES; type++) { for (enum ddt_class class = 0; class < DDT_CLASSES; class++) { ddt_object_prefetch(ddt, type, class, &dde); } } } int ddt_entry_compare(const void *x1, const void *x2) { const ddt_entry_t *dde1 = x1; const ddt_entry_t *dde2 = x2; const uint64_t *u1 = (const uint64_t *)&dde1->dde_key; const uint64_t *u2 = (const uint64_t *)&dde2->dde_key; for (int i = 0; i < DDT_KEY_WORDS; i++) { if (u1[i] < u2[i]) return (-1); if (u1[i] > u2[i]) return (1); } return (0); } static ddt_t * ddt_table_alloc(spa_t *spa, enum zio_checksum c) { ddt_t *ddt; ddt = kmem_zalloc(sizeof (*ddt), KM_SLEEP); mutex_init(&ddt->ddt_lock, NULL, MUTEX_DEFAULT, NULL); avl_create(&ddt->ddt_tree, ddt_entry_compare, sizeof (ddt_entry_t), offsetof(ddt_entry_t, dde_node)); avl_create(&ddt->ddt_repair_tree, ddt_entry_compare, sizeof (ddt_entry_t), offsetof(ddt_entry_t, dde_node)); ddt->ddt_checksum = c; ddt->ddt_spa = spa; ddt->ddt_os = spa->spa_meta_objset; return (ddt); } static void ddt_table_free(ddt_t *ddt) { ASSERT(avl_numnodes(&ddt->ddt_tree) == 0); ASSERT(avl_numnodes(&ddt->ddt_repair_tree) == 0); avl_destroy(&ddt->ddt_tree); avl_destroy(&ddt->ddt_repair_tree); mutex_destroy(&ddt->ddt_lock); kmem_free(ddt, sizeof (*ddt)); } void ddt_create(spa_t *spa) { spa->spa_dedup_checksum = ZIO_DEDUPCHECKSUM; for (enum zio_checksum c = 0; c < ZIO_CHECKSUM_FUNCTIONS; c++) spa->spa_ddt[c] = ddt_table_alloc(spa, c); } int ddt_load(spa_t *spa) { int error; ddt_create(spa); error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DDT_STATS, sizeof (uint64_t), 1, &spa->spa_ddt_stat_object); if (error) return (error == ENOENT ? 0 : error); for (enum zio_checksum c = 0; c < ZIO_CHECKSUM_FUNCTIONS; c++) { ddt_t *ddt = spa->spa_ddt[c]; for (enum ddt_type type = 0; type < DDT_TYPES; type++) { for (enum ddt_class class = 0; class < DDT_CLASSES; class++) { error = ddt_object_load(ddt, type, class); if (error != 0 && error != ENOENT) return (error); } } /* * Seed the cached histograms. */ bcopy(ddt->ddt_histogram, &ddt->ddt_histogram_cache, sizeof (ddt->ddt_histogram)); } return (0); } void ddt_unload(spa_t *spa) { for (enum zio_checksum c = 0; c < ZIO_CHECKSUM_FUNCTIONS; c++) { if (spa->spa_ddt[c]) { ddt_table_free(spa->spa_ddt[c]); spa->spa_ddt[c] = NULL; } } } boolean_t ddt_class_contains(spa_t *spa, enum ddt_class max_class, const blkptr_t *bp) { ddt_t *ddt; ddt_entry_t dde; if (!BP_GET_DEDUP(bp)) return (B_FALSE); if (max_class == DDT_CLASS_UNIQUE) return (B_TRUE); ddt = spa->spa_ddt[BP_GET_CHECKSUM(bp)]; ddt_key_fill(&dde.dde_key, bp); for (enum ddt_type type = 0; type < DDT_TYPES; type++) for (enum ddt_class class = 0; class <= max_class; class++) if (ddt_object_lookup(ddt, type, class, &dde) == 0) return (B_TRUE); return (B_FALSE); } ddt_entry_t * ddt_repair_start(ddt_t *ddt, const blkptr_t *bp) { ddt_key_t ddk; ddt_entry_t *dde; ddt_key_fill(&ddk, bp); dde = ddt_alloc(&ddk); for (enum ddt_type type = 0; type < DDT_TYPES; type++) { for (enum ddt_class class = 0; class < DDT_CLASSES; class++) { /* * We can only do repair if there are multiple copies * of the block. For anything in the UNIQUE class, * there's definitely only one copy, so don't even try. */ if (class != DDT_CLASS_UNIQUE && ddt_object_lookup(ddt, type, class, dde) == 0) return (dde); } } bzero(dde->dde_phys, sizeof (dde->dde_phys)); return (dde); } void ddt_repair_done(ddt_t *ddt, ddt_entry_t *dde) { avl_index_t where; ddt_enter(ddt); if (dde->dde_repair_data != NULL && spa_writeable(ddt->ddt_spa) && avl_find(&ddt->ddt_repair_tree, dde, &where) == NULL) avl_insert(&ddt->ddt_repair_tree, dde, where); else ddt_free(dde); ddt_exit(ddt); } static void ddt_repair_entry_done(zio_t *zio) { ddt_entry_t *rdde = zio->io_private; ddt_free(rdde); } static void ddt_repair_entry(ddt_t *ddt, ddt_entry_t *dde, ddt_entry_t *rdde, zio_t *rio) { ddt_phys_t *ddp = dde->dde_phys; ddt_phys_t *rddp = rdde->dde_phys; ddt_key_t *ddk = &dde->dde_key; ddt_key_t *rddk = &rdde->dde_key; zio_t *zio; blkptr_t blk; zio = zio_null(rio, rio->io_spa, NULL, ddt_repair_entry_done, rdde, rio->io_flags); for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++, rddp++) { if (ddp->ddp_phys_birth == 0 || ddp->ddp_phys_birth != rddp->ddp_phys_birth || bcmp(ddp->ddp_dva, rddp->ddp_dva, sizeof (ddp->ddp_dva))) continue; ddt_bp_create(ddt->ddt_checksum, ddk, ddp, &blk); zio_nowait(zio_rewrite(zio, zio->io_spa, 0, &blk, rdde->dde_repair_data, DDK_GET_PSIZE(rddk), NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, ZIO_DDT_CHILD_FLAGS(zio), NULL)); } zio_nowait(zio); } static void ddt_repair_table(ddt_t *ddt, zio_t *rio) { spa_t *spa = ddt->ddt_spa; ddt_entry_t *dde, *rdde_next, *rdde; avl_tree_t *t = &ddt->ddt_repair_tree; blkptr_t blk; if (spa_sync_pass(spa) > 1) return; ddt_enter(ddt); for (rdde = avl_first(t); rdde != NULL; rdde = rdde_next) { rdde_next = AVL_NEXT(t, rdde); avl_remove(&ddt->ddt_repair_tree, rdde); ddt_exit(ddt); ddt_bp_create(ddt->ddt_checksum, &rdde->dde_key, NULL, &blk); dde = ddt_repair_start(ddt, &blk); ddt_repair_entry(ddt, dde, rdde, rio); ddt_repair_done(ddt, dde); ddt_enter(ddt); } ddt_exit(ddt); } static void ddt_sync_entry(ddt_t *ddt, ddt_entry_t *dde, dmu_tx_t *tx, uint64_t txg) { dsl_pool_t *dp = ddt->ddt_spa->spa_dsl_pool; ddt_phys_t *ddp = dde->dde_phys; ddt_key_t *ddk = &dde->dde_key; enum ddt_type otype = dde->dde_type; enum ddt_type ntype = DDT_TYPE_CURRENT; enum ddt_class oclass = dde->dde_class; enum ddt_class nclass; uint64_t total_refcnt = 0; ASSERT(dde->dde_loaded); ASSERT(!dde->dde_loading); for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) { ASSERT(dde->dde_lead_zio[p] == NULL); ASSERT((int64_t)ddp->ddp_refcnt >= 0); if (ddp->ddp_phys_birth == 0) { ASSERT(ddp->ddp_refcnt == 0); continue; } if (p == DDT_PHYS_DITTO) { if (ddt_ditto_copies_needed(ddt, dde, NULL) == 0) ddt_phys_free(ddt, ddk, ddp, txg); continue; } if (ddp->ddp_refcnt == 0) ddt_phys_free(ddt, ddk, ddp, txg); total_refcnt += ddp->ddp_refcnt; } if (dde->dde_phys[DDT_PHYS_DITTO].ddp_phys_birth != 0) nclass = DDT_CLASS_DITTO; else if (total_refcnt > 1) nclass = DDT_CLASS_DUPLICATE; else nclass = DDT_CLASS_UNIQUE; if (otype != DDT_TYPES && (otype != ntype || oclass != nclass || total_refcnt == 0)) { VERIFY(ddt_object_remove(ddt, otype, oclass, dde, tx) == 0); ASSERT(ddt_object_lookup(ddt, otype, oclass, dde) == ENOENT); } if (total_refcnt != 0) { dde->dde_type = ntype; dde->dde_class = nclass; ddt_stat_update(ddt, dde, 0); if (!ddt_object_exists(ddt, ntype, nclass)) ddt_object_create(ddt, ntype, nclass, tx); VERIFY(ddt_object_update(ddt, ntype, nclass, dde, tx) == 0); /* * If the class changes, the order that we scan this bp * changes. If it decreases, we could miss it, so * scan it right now. (This covers both class changing * while we are doing ddt_walk(), and when we are * traversing.) */ if (nclass < oclass) { dsl_scan_ddt_entry(dp->dp_scan, ddt->ddt_checksum, dde, tx); } } } static void ddt_sync_table(ddt_t *ddt, dmu_tx_t *tx, uint64_t txg) { spa_t *spa = ddt->ddt_spa; ddt_entry_t *dde; void *cookie = NULL; if (avl_numnodes(&ddt->ddt_tree) == 0) return; ASSERT(spa->spa_uberblock.ub_version >= SPA_VERSION_DEDUP); if (spa->spa_ddt_stat_object == 0) { spa->spa_ddt_stat_object = zap_create_link(ddt->ddt_os, DMU_OT_DDT_STATS, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DDT_STATS, tx); } while ((dde = avl_destroy_nodes(&ddt->ddt_tree, &cookie)) != NULL) { ddt_sync_entry(ddt, dde, tx, txg); ddt_free(dde); } for (enum ddt_type type = 0; type < DDT_TYPES; type++) { uint64_t add, count = 0; for (enum ddt_class class = 0; class < DDT_CLASSES; class++) { if (ddt_object_exists(ddt, type, class)) { ddt_object_sync(ddt, type, class, tx); VERIFY(ddt_object_count(ddt, type, class, &add) == 0); count += add; } } for (enum ddt_class class = 0; class < DDT_CLASSES; class++) { if (count == 0 && ddt_object_exists(ddt, type, class)) ddt_object_destroy(ddt, type, class, tx); } } bcopy(ddt->ddt_histogram, &ddt->ddt_histogram_cache, sizeof (ddt->ddt_histogram)); } void ddt_sync(spa_t *spa, uint64_t txg) { dmu_tx_t *tx; zio_t *rio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE); ASSERT(spa_syncing_txg(spa) == txg); tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg); for (enum zio_checksum c = 0; c < ZIO_CHECKSUM_FUNCTIONS; c++) { ddt_t *ddt = spa->spa_ddt[c]; if (ddt == NULL) continue; ddt_sync_table(ddt, tx, txg); ddt_repair_table(ddt, rio); } (void) zio_wait(rio); dmu_tx_commit(tx); } int ddt_walk(spa_t *spa, ddt_bookmark_t *ddb, ddt_entry_t *dde) { do { do { do { ddt_t *ddt = spa->spa_ddt[ddb->ddb_checksum]; int error = ENOENT; if (ddt_object_exists(ddt, ddb->ddb_type, ddb->ddb_class)) { error = ddt_object_walk(ddt, ddb->ddb_type, ddb->ddb_class, &ddb->ddb_cursor, dde); } dde->dde_type = ddb->ddb_type; dde->dde_class = ddb->ddb_class; if (error == 0) return (0); if (error != ENOENT) return (error); ddb->ddb_cursor = 0; } while (++ddb->ddb_checksum < ZIO_CHECKSUM_FUNCTIONS); ddb->ddb_checksum = 0; } while (++ddb->ddb_type < DDT_TYPES); ddb->ddb_type = 0; } while (++ddb->ddb_class < DDT_CLASSES); return (SET_ERROR(ENOENT)); } Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu.c (revision 290757) @@ -1,2122 +1,2128 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2015 by Delphix. All rights reserved. */ /* Copyright (c) 2013 by Saso Kiselkov. All rights reserved. */ /* Copyright (c) 2013, Joyent, Inc. All rights reserved. */ /* Copyright (c) 2014, Nexenta Systems, Inc. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef _KERNEL #include #include #endif /* * Enable/disable nopwrite feature. */ int zfs_nopwrite_enabled = 1; SYSCTL_DECL(_vfs_zfs); TUNABLE_INT("vfs.zfs.nopwrite_enabled", &zfs_nopwrite_enabled); SYSCTL_INT(_vfs_zfs, OID_AUTO, nopwrite_enabled, CTLFLAG_RDTUN, &zfs_nopwrite_enabled, 0, "Enable nopwrite feature"); const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = { { DMU_BSWAP_UINT8, TRUE, "unallocated" }, { DMU_BSWAP_ZAP, TRUE, "object directory" }, { DMU_BSWAP_UINT64, TRUE, "object array" }, { DMU_BSWAP_UINT8, TRUE, "packed nvlist" }, { DMU_BSWAP_UINT64, TRUE, "packed nvlist size" }, { DMU_BSWAP_UINT64, TRUE, "bpobj" }, { DMU_BSWAP_UINT64, TRUE, "bpobj header" }, { DMU_BSWAP_UINT64, TRUE, "SPA space map header" }, { DMU_BSWAP_UINT64, TRUE, "SPA space map" }, { DMU_BSWAP_UINT64, TRUE, "ZIL intent log" }, { DMU_BSWAP_DNODE, TRUE, "DMU dnode" }, { DMU_BSWAP_OBJSET, TRUE, "DMU objset" }, { DMU_BSWAP_UINT64, TRUE, "DSL directory" }, { DMU_BSWAP_ZAP, TRUE, "DSL directory child map"}, { DMU_BSWAP_ZAP, TRUE, "DSL dataset snap map" }, { DMU_BSWAP_ZAP, TRUE, "DSL props" }, { DMU_BSWAP_UINT64, TRUE, "DSL dataset" }, { DMU_BSWAP_ZNODE, TRUE, "ZFS znode" }, { DMU_BSWAP_OLDACL, TRUE, "ZFS V0 ACL" }, { DMU_BSWAP_UINT8, FALSE, "ZFS plain file" }, { DMU_BSWAP_ZAP, TRUE, "ZFS directory" }, { DMU_BSWAP_ZAP, TRUE, "ZFS master node" }, { DMU_BSWAP_ZAP, TRUE, "ZFS delete queue" }, { DMU_BSWAP_UINT8, FALSE, "zvol object" }, { DMU_BSWAP_ZAP, TRUE, "zvol prop" }, { DMU_BSWAP_UINT8, FALSE, "other uint8[]" }, { DMU_BSWAP_UINT64, FALSE, "other uint64[]" }, { DMU_BSWAP_ZAP, TRUE, "other ZAP" }, { DMU_BSWAP_ZAP, TRUE, "persistent error log" }, { DMU_BSWAP_UINT8, TRUE, "SPA history" }, { DMU_BSWAP_UINT64, TRUE, "SPA history offsets" }, { DMU_BSWAP_ZAP, TRUE, "Pool properties" }, { DMU_BSWAP_ZAP, TRUE, "DSL permissions" }, { DMU_BSWAP_ACL, TRUE, "ZFS ACL" }, { DMU_BSWAP_UINT8, TRUE, "ZFS SYSACL" }, { DMU_BSWAP_UINT8, TRUE, "FUID table" }, { DMU_BSWAP_UINT64, TRUE, "FUID table size" }, { DMU_BSWAP_ZAP, TRUE, "DSL dataset next clones"}, { DMU_BSWAP_ZAP, TRUE, "scan work queue" }, { DMU_BSWAP_ZAP, TRUE, "ZFS user/group used" }, { DMU_BSWAP_ZAP, TRUE, "ZFS user/group quota" }, { DMU_BSWAP_ZAP, TRUE, "snapshot refcount tags"}, { DMU_BSWAP_ZAP, TRUE, "DDT ZAP algorithm" }, { DMU_BSWAP_ZAP, TRUE, "DDT statistics" }, { DMU_BSWAP_UINT8, TRUE, "System attributes" }, { DMU_BSWAP_ZAP, TRUE, "SA master node" }, { DMU_BSWAP_ZAP, TRUE, "SA attr registration" }, { DMU_BSWAP_ZAP, TRUE, "SA attr layouts" }, { DMU_BSWAP_ZAP, TRUE, "scan translations" }, { DMU_BSWAP_UINT8, FALSE, "deduplicated block" }, { DMU_BSWAP_ZAP, TRUE, "DSL deadlist map" }, { DMU_BSWAP_UINT64, TRUE, "DSL deadlist map hdr" }, { DMU_BSWAP_ZAP, TRUE, "DSL dir clones" }, { DMU_BSWAP_UINT64, TRUE, "bpobj subobj" } }; const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS] = { { byteswap_uint8_array, "uint8" }, { byteswap_uint16_array, "uint16" }, { byteswap_uint32_array, "uint32" }, { byteswap_uint64_array, "uint64" }, { zap_byteswap, "zap" }, { dnode_buf_byteswap, "dnode" }, { dmu_objset_byteswap, "objset" }, { zfs_znode_byteswap, "znode" }, { zfs_oldacl_byteswap, "oldacl" }, { zfs_acl_byteswap, "acl" } }; int dmu_buf_hold_noread(objset_t *os, uint64_t object, uint64_t offset, void *tag, dmu_buf_t **dbp) { dnode_t *dn; uint64_t blkid; dmu_buf_impl_t *db; int err; err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); blkid = dbuf_whichblock(dn, 0, offset); rw_enter(&dn->dn_struct_rwlock, RW_READER); db = dbuf_hold(dn, blkid, tag); rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); if (db == NULL) { *dbp = NULL; return (SET_ERROR(EIO)); } *dbp = &db->db; return (err); } int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, void *tag, dmu_buf_t **dbp, int flags) { int err; int db_flags = DB_RF_CANFAIL; if (flags & DMU_READ_NO_PREFETCH) db_flags |= DB_RF_NOPREFETCH; err = dmu_buf_hold_noread(os, object, offset, tag, dbp); if (err == 0) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)(*dbp); err = dbuf_read(db, NULL, db_flags); if (err != 0) { dbuf_rele(db, tag); *dbp = NULL; } } return (err); } int dmu_bonus_max(void) { return (DN_MAX_BONUSLEN); } int dmu_set_bonus(dmu_buf_t *db_fake, int newsize, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; int error; DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (dn->dn_bonus != db) { error = SET_ERROR(EINVAL); } else if (newsize < 0 || newsize > db_fake->db_size) { error = SET_ERROR(EINVAL); } else { dnode_setbonuslen(dn, newsize, tx); error = 0; } DB_DNODE_EXIT(db); return (error); } int dmu_set_bonustype(dmu_buf_t *db_fake, dmu_object_type_t type, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; int error; DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (!DMU_OT_IS_VALID(type)) { error = SET_ERROR(EINVAL); } else if (dn->dn_bonus != db) { error = SET_ERROR(EINVAL); } else { dnode_setbonus_type(dn, type, tx); error = 0; } DB_DNODE_EXIT(db); return (error); } dmu_object_type_t dmu_get_bonustype(dmu_buf_t *db_fake) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; dmu_object_type_t type; DB_DNODE_ENTER(db); dn = DB_DNODE(db); type = dn->dn_bonustype; DB_DNODE_EXIT(db); return (type); } int dmu_rm_spill(objset_t *os, uint64_t object, dmu_tx_t *tx) { dnode_t *dn; int error; error = dnode_hold(os, object, FTAG, &dn); dbuf_rm_spill(dn, tx); rw_enter(&dn->dn_struct_rwlock, RW_WRITER); dnode_rm_spill(dn, tx); rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); return (error); } /* * returns ENOENT, EIO, or 0. */ int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **dbp) { dnode_t *dn; dmu_buf_impl_t *db; int error; error = dnode_hold(os, object, FTAG, &dn); if (error) return (error); rw_enter(&dn->dn_struct_rwlock, RW_READER); if (dn->dn_bonus == NULL) { rw_exit(&dn->dn_struct_rwlock); rw_enter(&dn->dn_struct_rwlock, RW_WRITER); if (dn->dn_bonus == NULL) dbuf_create_bonus(dn); } db = dn->dn_bonus; /* as long as the bonus buf is held, the dnode will be held */ if (refcount_add(&db->db_holds, tag) == 1) { VERIFY(dnode_add_ref(dn, db)); atomic_inc_32(&dn->dn_dbufs_count); } /* * Wait to drop dn_struct_rwlock until after adding the bonus dbuf's * hold and incrementing the dbuf count to ensure that dnode_move() sees * a dnode hold for every dbuf. */ rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); VERIFY(0 == dbuf_read(db, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH)); *dbp = &db->db; return (0); } /* * returns ENOENT, EIO, or 0. * * This interface will allocate a blank spill dbuf when a spill blk * doesn't already exist on the dnode. * * if you only want to find an already existing spill db, then * dmu_spill_hold_existing() should be used. */ int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags, void *tag, dmu_buf_t **dbp) { dmu_buf_impl_t *db = NULL; int err; if ((flags & DB_RF_HAVESTRUCT) == 0) rw_enter(&dn->dn_struct_rwlock, RW_READER); db = dbuf_hold(dn, DMU_SPILL_BLKID, tag); if ((flags & DB_RF_HAVESTRUCT) == 0) rw_exit(&dn->dn_struct_rwlock); ASSERT(db != NULL); err = dbuf_read(db, NULL, flags); if (err == 0) *dbp = &db->db; else dbuf_rele(db, tag); return (err); } int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)bonus; dnode_t *dn; int err; DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (spa_version(dn->dn_objset->os_spa) < SPA_VERSION_SA) { err = SET_ERROR(EINVAL); } else { rw_enter(&dn->dn_struct_rwlock, RW_READER); if (!dn->dn_have_spill) { err = SET_ERROR(ENOENT); } else { err = dmu_spill_hold_by_dnode(dn, DB_RF_HAVESTRUCT | DB_RF_CANFAIL, tag, dbp); } rw_exit(&dn->dn_struct_rwlock); } DB_DNODE_EXIT(db); return (err); } int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)bonus; dnode_t *dn; int err; DB_DNODE_ENTER(db); dn = DB_DNODE(db); err = dmu_spill_hold_by_dnode(dn, DB_RF_CANFAIL, tag, dbp); DB_DNODE_EXIT(db); return (err); } /* * Note: longer-term, we should modify all of the dmu_buf_*() interfaces * to take a held dnode rather than -- the lookup is wasteful, * and can induce severe lock contention when writing to several files * whose dnodes are in the same block. */ static int dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, uint64_t length, boolean_t read, void *tag, int *numbufsp, dmu_buf_t ***dbpp, uint32_t flags) { dmu_buf_t **dbp; uint64_t blkid, nblks, i; uint32_t dbuf_flags; int err; zio_t *zio; ASSERT(length <= DMU_MAX_ACCESS); /* * Note: We directly notify the prefetch code of this read, so that * we can tell it about the multi-block read. dbuf_read() only knows * about the one block it is accessing. */ dbuf_flags = DB_RF_CANFAIL | DB_RF_NEVERWAIT | DB_RF_HAVESTRUCT | DB_RF_NOPREFETCH; rw_enter(&dn->dn_struct_rwlock, RW_READER); if (dn->dn_datablkshift) { int blkshift = dn->dn_datablkshift; nblks = (P2ROUNDUP(offset + length, 1ULL << blkshift) - P2ALIGN(offset, 1ULL << blkshift)) >> blkshift; } else { if (offset + length > dn->dn_datablksz) { zfs_panic_recover("zfs: accessing past end of object " "%llx/%llx (size=%u access=%llu+%llu)", (longlong_t)dn->dn_objset-> os_dsl_dataset->ds_object, (longlong_t)dn->dn_object, dn->dn_datablksz, (longlong_t)offset, (longlong_t)length); rw_exit(&dn->dn_struct_rwlock); return (SET_ERROR(EIO)); } nblks = 1; } dbp = kmem_zalloc(sizeof (dmu_buf_t *) * nblks, KM_SLEEP); zio = zio_root(dn->dn_objset->os_spa, NULL, NULL, ZIO_FLAG_CANFAIL); blkid = dbuf_whichblock(dn, 0, offset); for (i = 0; i < nblks; i++) { dmu_buf_impl_t *db = dbuf_hold(dn, blkid + i, tag); if (db == NULL) { rw_exit(&dn->dn_struct_rwlock); dmu_buf_rele_array(dbp, nblks, tag); zio_nowait(zio); return (SET_ERROR(EIO)); } /* initiate async i/o */ if (read) (void) dbuf_read(db, zio, dbuf_flags); #ifdef _KERNEL else curthread->td_ru.ru_oublock++; #endif dbp[i] = &db->db; } if ((flags & DMU_READ_NO_PREFETCH) == 0 && read && length <= zfetch_array_rd_sz) { dmu_zfetch(&dn->dn_zfetch, blkid, nblks); } rw_exit(&dn->dn_struct_rwlock); /* wait for async i/o */ err = zio_wait(zio); if (err) { dmu_buf_rele_array(dbp, nblks, tag); return (err); } /* wait for other io to complete */ if (read) { for (i = 0; i < nblks; i++) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbp[i]; mutex_enter(&db->db_mtx); while (db->db_state == DB_READ || db->db_state == DB_FILL) cv_wait(&db->db_changed, &db->db_mtx); if (db->db_state == DB_UNCACHED) err = SET_ERROR(EIO); mutex_exit(&db->db_mtx); if (err) { dmu_buf_rele_array(dbp, nblks, tag); return (err); } } } *numbufsp = nblks; *dbpp = dbp; return (0); } static int dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset, uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp) { dnode_t *dn; int err; err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag, numbufsp, dbpp, DMU_READ_PREFETCH); dnode_rele(dn, FTAG); return (err); } int dmu_buf_hold_array_by_bonus(dmu_buf_t *db_fake, uint64_t offset, uint64_t length, boolean_t read, void *tag, int *numbufsp, dmu_buf_t ***dbpp) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; int err; DB_DNODE_ENTER(db); dn = DB_DNODE(db); err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag, numbufsp, dbpp, DMU_READ_PREFETCH); DB_DNODE_EXIT(db); return (err); } void dmu_buf_rele_array(dmu_buf_t **dbp_fake, int numbufs, void *tag) { int i; dmu_buf_impl_t **dbp = (dmu_buf_impl_t **)dbp_fake; if (numbufs == 0) return; for (i = 0; i < numbufs; i++) { if (dbp[i]) dbuf_rele(dbp[i], tag); } kmem_free(dbp, sizeof (dmu_buf_t *) * numbufs); } /* * Issue prefetch i/os for the given blocks. If level is greater than 0, the * indirect blocks prefeteched will be those that point to the blocks containing * the data starting at offset, and continuing to offset + len. * * Note that if the indirect blocks above the blocks being prefetched are not in * cache, they will be asychronously read in. */ void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset, uint64_t len, zio_priority_t pri) { dnode_t *dn; uint64_t blkid; int nblks, err; if (len == 0) { /* they're interested in the bonus buffer */ dn = DMU_META_DNODE(os); if (object == 0 || object >= DN_MAX_OBJECT) return; rw_enter(&dn->dn_struct_rwlock, RW_READER); blkid = dbuf_whichblock(dn, level, object * sizeof (dnode_phys_t)); dbuf_prefetch(dn, level, blkid, pri, 0); rw_exit(&dn->dn_struct_rwlock); return; } /* * XXX - Note, if the dnode for the requested object is not * already cached, we will do a *synchronous* read in the * dnode_hold() call. The same is true for any indirects. */ err = dnode_hold(os, object, FTAG, &dn); if (err != 0) return; rw_enter(&dn->dn_struct_rwlock, RW_READER); /* * offset + len - 1 is the last byte we want to prefetch for, and offset * is the first. Then dbuf_whichblk(dn, level, off + len - 1) is the * last block we want to prefetch, and dbuf_whichblock(dn, level, * offset) is the first. Then the number we need to prefetch is the * last - first + 1. */ if (level > 0 || dn->dn_datablkshift != 0) { nblks = dbuf_whichblock(dn, level, offset + len - 1) - dbuf_whichblock(dn, level, offset) + 1; } else { nblks = (offset < dn->dn_datablksz); } if (nblks != 0) { blkid = dbuf_whichblock(dn, level, offset); for (int i = 0; i < nblks; i++) dbuf_prefetch(dn, level, blkid + i, pri, 0); } rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); } /* * Get the next "chunk" of file data to free. We traverse the file from * the end so that the file gets shorter over time (if we crashes in the * middle, this will leave us in a better state). We find allocated file * data by simply searching the allocated level 1 indirects. * * On input, *start should be the first offset that does not need to be * freed (e.g. "offset + length"). On return, *start will be the first * offset that should be freed. */ static int get_next_chunk(dnode_t *dn, uint64_t *start, uint64_t minimum) { uint64_t maxblks = DMU_MAX_ACCESS >> (dn->dn_indblkshift + 1); /* bytes of data covered by a level-1 indirect block */ uint64_t iblkrange = dn->dn_datablksz * EPB(dn->dn_indblkshift, SPA_BLKPTRSHIFT); ASSERT3U(minimum, <=, *start); if (*start - minimum <= iblkrange * maxblks) { *start = minimum; return (0); } ASSERT(ISP2(iblkrange)); for (uint64_t blks = 0; *start > minimum && blks < maxblks; blks++) { int err; /* * dnode_next_offset(BACKWARDS) will find an allocated L1 * indirect block at or before the input offset. We must * decrement *start so that it is at the end of the region * to search. */ (*start)--; err = dnode_next_offset(dn, DNODE_FIND_BACKWARDS, start, 2, 1, 0); /* if there are no indirect blocks before start, we are done */ if (err == ESRCH) { *start = minimum; break; } else if (err != 0) { return (err); } /* set start to the beginning of this L1 indirect */ *start = P2ALIGN(*start, iblkrange); } if (*start < minimum) *start = minimum; return (0); } static int dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset, uint64_t length) { uint64_t object_size = (dn->dn_maxblkid + 1) * dn->dn_datablksz; int err; if (offset >= object_size) return (0); if (length == DMU_OBJECT_END || offset + length > object_size) length = object_size - offset; while (length != 0) { uint64_t chunk_end, chunk_begin; chunk_end = chunk_begin = offset + length; /* move chunk_begin backwards to the beginning of this chunk */ err = get_next_chunk(dn, &chunk_begin, offset); if (err) return (err); ASSERT3U(chunk_begin, >=, offset); ASSERT3U(chunk_begin, <=, chunk_end); dmu_tx_t *tx = dmu_tx_create(os); dmu_tx_hold_free(tx, dn->dn_object, chunk_begin, chunk_end - chunk_begin); /* * Mark this transaction as typically resulting in a net * reduction in space used. */ dmu_tx_mark_netfree(tx); err = dmu_tx_assign(tx, TXG_WAIT); if (err) { dmu_tx_abort(tx); return (err); } dnode_free_range(dn, chunk_begin, chunk_end - chunk_begin, tx); dmu_tx_commit(tx); length -= chunk_end - chunk_begin; } return (0); } int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, uint64_t length) { dnode_t *dn; int err; err = dnode_hold(os, object, FTAG, &dn); if (err != 0) return (err); err = dmu_free_long_range_impl(os, dn, offset, length); /* * It is important to zero out the maxblkid when freeing the entire * file, so that (a) subsequent calls to dmu_free_long_range_impl() * will take the fast path, and (b) dnode_reallocate() can verify * that the entire file has been freed. */ if (err == 0 && offset == 0 && length == DMU_OBJECT_END) dn->dn_maxblkid = 0; dnode_rele(dn, FTAG); return (err); } int dmu_free_long_object(objset_t *os, uint64_t object) { dmu_tx_t *tx; int err; err = dmu_free_long_range(os, object, 0, DMU_OBJECT_END); if (err != 0) return (err); tx = dmu_tx_create(os); dmu_tx_hold_bonus(tx, object); dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); dmu_tx_mark_netfree(tx); err = dmu_tx_assign(tx, TXG_WAIT); if (err == 0) { err = dmu_object_free(os, object, tx); dmu_tx_commit(tx); } else { dmu_tx_abort(tx); } return (err); } int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, dmu_tx_t *tx) { dnode_t *dn; int err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); ASSERT(offset < UINT64_MAX); ASSERT(size == -1ULL || size <= UINT64_MAX - offset); dnode_free_range(dn, offset, size, tx); dnode_rele(dn, FTAG); return (0); } int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, void *buf, uint32_t flags) { dnode_t *dn; dmu_buf_t **dbp; int numbufs, err; err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); /* * Deal with odd block sizes, where there can't be data past the first * block. If we ever do the tail block optimization, we will need to * handle that here as well. */ if (dn->dn_maxblkid == 0) { int newsz = offset > dn->dn_datablksz ? 0 : MIN(size, dn->dn_datablksz - offset); bzero((char *)buf + newsz, size - newsz); size = newsz; } while (size > 0) { uint64_t mylen = MIN(size, DMU_MAX_ACCESS / 2); int i; /* * NB: we could do this block-at-a-time, but it's nice * to be reading in parallel. */ err = dmu_buf_hold_array_by_dnode(dn, offset, mylen, TRUE, FTAG, &numbufs, &dbp, flags); if (err) break; for (i = 0; i < numbufs; i++) { int tocpy; int bufoff; dmu_buf_t *db = dbp[i]; ASSERT(size > 0); bufoff = offset - db->db_offset; tocpy = (int)MIN(db->db_size - bufoff, size); bcopy((char *)db->db_data + bufoff, buf, tocpy); offset += tocpy; size -= tocpy; buf = (char *)buf + tocpy; } dmu_buf_rele_array(dbp, numbufs, FTAG); } dnode_rele(dn, FTAG); return (err); } void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, const void *buf, dmu_tx_t *tx) { dmu_buf_t **dbp; int numbufs, i; if (size == 0) return; VERIFY(0 == dmu_buf_hold_array(os, object, offset, size, FALSE, FTAG, &numbufs, &dbp)); for (i = 0; i < numbufs; i++) { int tocpy; int bufoff; dmu_buf_t *db = dbp[i]; ASSERT(size > 0); bufoff = offset - db->db_offset; tocpy = (int)MIN(db->db_size - bufoff, size); ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size); if (tocpy == db->db_size) dmu_buf_will_fill(db, tx); else dmu_buf_will_dirty(db, tx); bcopy(buf, (char *)db->db_data + bufoff, tocpy); if (tocpy == db->db_size) dmu_buf_fill_done(db, tx); offset += tocpy; size -= tocpy; buf = (char *)buf + tocpy; } dmu_buf_rele_array(dbp, numbufs, FTAG); } void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, dmu_tx_t *tx) { dmu_buf_t **dbp; int numbufs, i; if (size == 0) return; VERIFY(0 == dmu_buf_hold_array(os, object, offset, size, FALSE, FTAG, &numbufs, &dbp)); for (i = 0; i < numbufs; i++) { dmu_buf_t *db = dbp[i]; dmu_buf_will_not_fill(db, tx); } dmu_buf_rele_array(dbp, numbufs, FTAG); } void dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset, void *data, uint8_t etype, uint8_t comp, int uncompressed_size, int compressed_size, int byteorder, dmu_tx_t *tx) { dmu_buf_t *db; ASSERT3U(etype, <, NUM_BP_EMBEDDED_TYPES); ASSERT3U(comp, <, ZIO_COMPRESS_FUNCTIONS); VERIFY0(dmu_buf_hold_noread(os, object, offset, FTAG, &db)); dmu_buf_write_embedded(db, data, (bp_embedded_type_t)etype, (enum zio_compress)comp, uncompressed_size, compressed_size, byteorder, tx); dmu_buf_rele(db, FTAG); } /* * DMU support for xuio */ kstat_t *xuio_ksp = NULL; int dmu_xuio_init(xuio_t *xuio, int nblk) { dmu_xuio_t *priv; uio_t *uio = &xuio->xu_uio; uio->uio_iovcnt = nblk; uio->uio_iov = kmem_zalloc(nblk * sizeof (iovec_t), KM_SLEEP); priv = kmem_zalloc(sizeof (dmu_xuio_t), KM_SLEEP); priv->cnt = nblk; priv->bufs = kmem_zalloc(nblk * sizeof (arc_buf_t *), KM_SLEEP); priv->iovp = uio->uio_iov; XUIO_XUZC_PRIV(xuio) = priv; if (XUIO_XUZC_RW(xuio) == UIO_READ) XUIOSTAT_INCR(xuiostat_onloan_rbuf, nblk); else XUIOSTAT_INCR(xuiostat_onloan_wbuf, nblk); return (0); } void dmu_xuio_fini(xuio_t *xuio) { dmu_xuio_t *priv = XUIO_XUZC_PRIV(xuio); int nblk = priv->cnt; kmem_free(priv->iovp, nblk * sizeof (iovec_t)); kmem_free(priv->bufs, nblk * sizeof (arc_buf_t *)); kmem_free(priv, sizeof (dmu_xuio_t)); if (XUIO_XUZC_RW(xuio) == UIO_READ) XUIOSTAT_INCR(xuiostat_onloan_rbuf, -nblk); else XUIOSTAT_INCR(xuiostat_onloan_wbuf, -nblk); } /* * Initialize iov[priv->next] and priv->bufs[priv->next] with { off, n, abuf } * and increase priv->next by 1. */ int dmu_xuio_add(xuio_t *xuio, arc_buf_t *abuf, offset_t off, size_t n) { struct iovec *iov; uio_t *uio = &xuio->xu_uio; dmu_xuio_t *priv = XUIO_XUZC_PRIV(xuio); int i = priv->next++; ASSERT(i < priv->cnt); ASSERT(off + n <= arc_buf_size(abuf)); iov = uio->uio_iov + i; iov->iov_base = (char *)abuf->b_data + off; iov->iov_len = n; priv->bufs[i] = abuf; return (0); } int dmu_xuio_cnt(xuio_t *xuio) { dmu_xuio_t *priv = XUIO_XUZC_PRIV(xuio); return (priv->cnt); } arc_buf_t * dmu_xuio_arcbuf(xuio_t *xuio, int i) { dmu_xuio_t *priv = XUIO_XUZC_PRIV(xuio); ASSERT(i < priv->cnt); return (priv->bufs[i]); } void dmu_xuio_clear(xuio_t *xuio, int i) { dmu_xuio_t *priv = XUIO_XUZC_PRIV(xuio); ASSERT(i < priv->cnt); priv->bufs[i] = NULL; } static void xuio_stat_init(void) { xuio_ksp = kstat_create("zfs", 0, "xuio_stats", "misc", KSTAT_TYPE_NAMED, sizeof (xuio_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL); if (xuio_ksp != NULL) { xuio_ksp->ks_data = &xuio_stats; kstat_install(xuio_ksp); } } static void xuio_stat_fini(void) { if (xuio_ksp != NULL) { kstat_delete(xuio_ksp); xuio_ksp = NULL; } } void xuio_stat_wbuf_copied() { XUIOSTAT_BUMP(xuiostat_wbuf_copied); } void xuio_stat_wbuf_nocopy() { XUIOSTAT_BUMP(xuiostat_wbuf_nocopy); } #ifdef _KERNEL static int dmu_read_uio_dnode(dnode_t *dn, uio_t *uio, uint64_t size) { dmu_buf_t **dbp; int numbufs, i, err; xuio_t *xuio = NULL; /* * NB: we could do this block-at-a-time, but it's nice * to be reading in parallel. */ err = dmu_buf_hold_array_by_dnode(dn, uio->uio_loffset, size, TRUE, FTAG, &numbufs, &dbp, 0); if (err) return (err); #ifdef UIO_XUIO if (uio->uio_extflg == UIO_XUIO) xuio = (xuio_t *)uio; #endif for (i = 0; i < numbufs; i++) { int tocpy; int bufoff; dmu_buf_t *db = dbp[i]; ASSERT(size > 0); bufoff = uio->uio_loffset - db->db_offset; tocpy = (int)MIN(db->db_size - bufoff, size); if (xuio) { dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db; arc_buf_t *dbuf_abuf = dbi->db_buf; arc_buf_t *abuf = dbuf_loan_arcbuf(dbi); err = dmu_xuio_add(xuio, abuf, bufoff, tocpy); if (!err) { uio->uio_resid -= tocpy; uio->uio_loffset += tocpy; } if (abuf == dbuf_abuf) XUIOSTAT_BUMP(xuiostat_rbuf_nocopy); else XUIOSTAT_BUMP(xuiostat_rbuf_copied); } else { err = uiomove((char *)db->db_data + bufoff, tocpy, UIO_READ, uio); } if (err) break; size -= tocpy; } dmu_buf_rele_array(dbp, numbufs, FTAG); return (err); } /* * Read 'size' bytes into the uio buffer. * From object zdb->db_object. * Starting at offset uio->uio_loffset. * * If the caller already has a dbuf in the target object * (e.g. its bonus buffer), this routine is faster than dmu_read_uio(), * because we don't have to find the dnode_t for the object. */ int dmu_read_uio_dbuf(dmu_buf_t *zdb, uio_t *uio, uint64_t size) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)zdb; dnode_t *dn; int err; if (size == 0) return (0); DB_DNODE_ENTER(db); dn = DB_DNODE(db); err = dmu_read_uio_dnode(dn, uio, size); DB_DNODE_EXIT(db); return (err); } /* * Read 'size' bytes into the uio buffer. * From the specified object * Starting at offset uio->uio_loffset. */ int dmu_read_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size) { dnode_t *dn; int err; if (size == 0) return (0); err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); err = dmu_read_uio_dnode(dn, uio, size); dnode_rele(dn, FTAG); return (err); } static int dmu_write_uio_dnode(dnode_t *dn, uio_t *uio, uint64_t size, dmu_tx_t *tx) { dmu_buf_t **dbp; int numbufs; int err = 0; int i; err = dmu_buf_hold_array_by_dnode(dn, uio->uio_loffset, size, FALSE, FTAG, &numbufs, &dbp, DMU_READ_PREFETCH); if (err) return (err); for (i = 0; i < numbufs; i++) { int tocpy; int bufoff; dmu_buf_t *db = dbp[i]; ASSERT(size > 0); bufoff = uio->uio_loffset - db->db_offset; tocpy = (int)MIN(db->db_size - bufoff, size); ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size); if (tocpy == db->db_size) dmu_buf_will_fill(db, tx); else dmu_buf_will_dirty(db, tx); /* * XXX uiomove could block forever (eg. nfs-backed * pages). There needs to be a uiolockdown() function * to lock the pages in memory, so that uiomove won't * block. */ err = uiomove((char *)db->db_data + bufoff, tocpy, UIO_WRITE, uio); if (tocpy == db->db_size) dmu_buf_fill_done(db, tx); if (err) break; size -= tocpy; } dmu_buf_rele_array(dbp, numbufs, FTAG); return (err); } /* * Write 'size' bytes from the uio buffer. * To object zdb->db_object. * Starting at offset uio->uio_loffset. * * If the caller already has a dbuf in the target object * (e.g. its bonus buffer), this routine is faster than dmu_write_uio(), * because we don't have to find the dnode_t for the object. */ int dmu_write_uio_dbuf(dmu_buf_t *zdb, uio_t *uio, uint64_t size, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)zdb; dnode_t *dn; int err; if (size == 0) return (0); DB_DNODE_ENTER(db); dn = DB_DNODE(db); err = dmu_write_uio_dnode(dn, uio, size, tx); DB_DNODE_EXIT(db); return (err); } /* * Write 'size' bytes from the uio buffer. * To the specified object. * Starting at offset uio->uio_loffset. */ int dmu_write_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size, dmu_tx_t *tx) { dnode_t *dn; int err; if (size == 0) return (0); err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); err = dmu_write_uio_dnode(dn, uio, size, tx); dnode_rele(dn, FTAG); return (err); } #ifdef sun int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, page_t *pp, dmu_tx_t *tx) { dmu_buf_t **dbp; int numbufs, i; int err; if (size == 0) return (0); err = dmu_buf_hold_array(os, object, offset, size, FALSE, FTAG, &numbufs, &dbp); if (err) return (err); for (i = 0; i < numbufs; i++) { int tocpy, copied, thiscpy; int bufoff; dmu_buf_t *db = dbp[i]; caddr_t va; ASSERT(size > 0); ASSERT3U(db->db_size, >=, PAGESIZE); bufoff = offset - db->db_offset; tocpy = (int)MIN(db->db_size - bufoff, size); ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size); if (tocpy == db->db_size) dmu_buf_will_fill(db, tx); else dmu_buf_will_dirty(db, tx); for (copied = 0; copied < tocpy; copied += PAGESIZE) { ASSERT3U(pp->p_offset, ==, db->db_offset + bufoff); thiscpy = MIN(PAGESIZE, tocpy - copied); va = zfs_map_page(pp, S_READ); bcopy(va, (char *)db->db_data + bufoff, thiscpy); zfs_unmap_page(pp, va); pp = pp->p_next; bufoff += PAGESIZE; } if (tocpy == db->db_size) dmu_buf_fill_done(db, tx); offset += tocpy; size -= tocpy; } dmu_buf_rele_array(dbp, numbufs, FTAG); return (err); } #else int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, vm_page_t *ma, dmu_tx_t *tx) { dmu_buf_t **dbp; struct sf_buf *sf; int numbufs, i; int err; if (size == 0) return (0); err = dmu_buf_hold_array(os, object, offset, size, FALSE, FTAG, &numbufs, &dbp); if (err) return (err); for (i = 0; i < numbufs; i++) { int tocpy, copied, thiscpy; int bufoff; dmu_buf_t *db = dbp[i]; caddr_t va; ASSERT(size > 0); ASSERT3U(db->db_size, >=, PAGESIZE); bufoff = offset - db->db_offset; tocpy = (int)MIN(db->db_size - bufoff, size); ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size); if (tocpy == db->db_size) dmu_buf_will_fill(db, tx); else dmu_buf_will_dirty(db, tx); for (copied = 0; copied < tocpy; copied += PAGESIZE) { ASSERT3U(ptoa((*ma)->pindex), ==, db->db_offset + bufoff); thiscpy = MIN(PAGESIZE, tocpy - copied); va = zfs_map_page(*ma, &sf); bcopy(va, (char *)db->db_data + bufoff, thiscpy); zfs_unmap_page(sf); ma += 1; bufoff += PAGESIZE; } if (tocpy == db->db_size) dmu_buf_fill_done(db, tx); offset += tocpy; size -= tocpy; } dmu_buf_rele_array(dbp, numbufs, FTAG); return (err); } #endif /* sun */ #endif /* * Allocate a loaned anonymous arc buffer. */ arc_buf_t * dmu_request_arcbuf(dmu_buf_t *handle, int size) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)handle; return (arc_loan_buf(db->db_objset->os_spa, size)); } /* * Free a loaned arc buffer. */ void dmu_return_arcbuf(arc_buf_t *buf) { arc_return_buf(buf, FTAG); VERIFY(arc_buf_remove_ref(buf, FTAG)); } /* * When possible directly assign passed loaned arc buffer to a dbuf. * If this is not possible copy the contents of passed arc buf via * dmu_write(). */ void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, arc_buf_t *buf, dmu_tx_t *tx) { dmu_buf_impl_t *dbuf = (dmu_buf_impl_t *)handle; dnode_t *dn; dmu_buf_impl_t *db; uint32_t blksz = (uint32_t)arc_buf_size(buf); uint64_t blkid; DB_DNODE_ENTER(dbuf); dn = DB_DNODE(dbuf); rw_enter(&dn->dn_struct_rwlock, RW_READER); blkid = dbuf_whichblock(dn, 0, offset); VERIFY((db = dbuf_hold(dn, blkid, FTAG)) != NULL); rw_exit(&dn->dn_struct_rwlock); DB_DNODE_EXIT(dbuf); /* * We can only assign if the offset is aligned, the arc buf is the * same size as the dbuf, and the dbuf is not metadata. It * can't be metadata because the loaned arc buf comes from the * user-data kmem arena. */ if (offset == db->db.db_offset && blksz == db->db.db_size && DBUF_GET_BUFC_TYPE(db) == ARC_BUFC_DATA) { dbuf_assign_arcbuf(db, buf, tx); dbuf_rele(db, FTAG); } else { objset_t *os; uint64_t object; DB_DNODE_ENTER(dbuf); dn = DB_DNODE(dbuf); os = dn->dn_objset; object = dn->dn_object; DB_DNODE_EXIT(dbuf); dbuf_rele(db, FTAG); dmu_write(os, object, offset, blksz, buf->b_data, tx); dmu_return_arcbuf(buf); XUIOSTAT_BUMP(xuiostat_wbuf_copied); } } typedef struct { dbuf_dirty_record_t *dsa_dr; dmu_sync_cb_t *dsa_done; zgd_t *dsa_zgd; dmu_tx_t *dsa_tx; } dmu_sync_arg_t; /* ARGSUSED */ static void dmu_sync_ready(zio_t *zio, arc_buf_t *buf, void *varg) { dmu_sync_arg_t *dsa = varg; dmu_buf_t *db = dsa->dsa_zgd->zgd_db; blkptr_t *bp = zio->io_bp; if (zio->io_error == 0) { if (BP_IS_HOLE(bp)) { /* * A block of zeros may compress to a hole, but the * block size still needs to be known for replay. */ BP_SET_LSIZE(bp, db->db_size); } else if (!BP_IS_EMBEDDED(bp)) { ASSERT(BP_GET_LEVEL(bp) == 0); bp->blk_fill = 1; } } } static void dmu_sync_late_arrival_ready(zio_t *zio) { dmu_sync_ready(zio, NULL, zio->io_private); } /* ARGSUSED */ static void dmu_sync_done(zio_t *zio, arc_buf_t *buf, void *varg) { dmu_sync_arg_t *dsa = varg; dbuf_dirty_record_t *dr = dsa->dsa_dr; dmu_buf_impl_t *db = dr->dr_dbuf; mutex_enter(&db->db_mtx); ASSERT(dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC); if (zio->io_error == 0) { dr->dt.dl.dr_nopwrite = !!(zio->io_flags & ZIO_FLAG_NOPWRITE); if (dr->dt.dl.dr_nopwrite) { blkptr_t *bp = zio->io_bp; blkptr_t *bp_orig = &zio->io_bp_orig; uint8_t chksum = BP_GET_CHECKSUM(bp_orig); ASSERT(BP_EQUAL(bp, bp_orig)); ASSERT(zio->io_prop.zp_compress != ZIO_COMPRESS_OFF); - ASSERT(zio_checksum_table[chksum].ci_dedup); + ASSERT(zio_checksum_table[chksum].ci_flags & + ZCHECKSUM_FLAG_NOPWRITE); } dr->dt.dl.dr_overridden_by = *zio->io_bp; dr->dt.dl.dr_override_state = DR_OVERRIDDEN; dr->dt.dl.dr_copies = zio->io_prop.zp_copies; /* * Old style holes are filled with all zeros, whereas * new-style holes maintain their lsize, type, level, * and birth time (see zio_write_compress). While we * need to reset the BP_SET_LSIZE() call that happened * in dmu_sync_ready for old style holes, we do *not* * want to wipe out the information contained in new * style holes. Thus, only zero out the block pointer if * it's an old style hole. */ if (BP_IS_HOLE(&dr->dt.dl.dr_overridden_by) && dr->dt.dl.dr_overridden_by.blk_birth == 0) BP_ZERO(&dr->dt.dl.dr_overridden_by); } else { dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN; } cv_broadcast(&db->db_changed); mutex_exit(&db->db_mtx); dsa->dsa_done(dsa->dsa_zgd, zio->io_error); kmem_free(dsa, sizeof (*dsa)); } static void dmu_sync_late_arrival_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; dmu_sync_arg_t *dsa = zio->io_private; blkptr_t *bp_orig = &zio->io_bp_orig; if (zio->io_error == 0 && !BP_IS_HOLE(bp)) { /* * If we didn't allocate a new block (i.e. ZIO_FLAG_NOPWRITE) * then there is nothing to do here. Otherwise, free the * newly allocated block in this txg. */ if (zio->io_flags & ZIO_FLAG_NOPWRITE) { ASSERT(BP_EQUAL(bp, bp_orig)); } else { ASSERT(BP_IS_HOLE(bp_orig) || !BP_EQUAL(bp, bp_orig)); ASSERT(zio->io_bp->blk_birth == zio->io_txg); ASSERT(zio->io_txg > spa_syncing_txg(zio->io_spa)); zio_free(zio->io_spa, zio->io_txg, zio->io_bp); } } dmu_tx_commit(dsa->dsa_tx); dsa->dsa_done(dsa->dsa_zgd, zio->io_error); kmem_free(dsa, sizeof (*dsa)); } static int dmu_sync_late_arrival(zio_t *pio, objset_t *os, dmu_sync_cb_t *done, zgd_t *zgd, zio_prop_t *zp, zbookmark_phys_t *zb) { dmu_sync_arg_t *dsa; dmu_tx_t *tx; tx = dmu_tx_create(os); dmu_tx_hold_space(tx, zgd->zgd_db->db_size); if (dmu_tx_assign(tx, TXG_WAIT) != 0) { dmu_tx_abort(tx); /* Make zl_get_data do txg_waited_synced() */ return (SET_ERROR(EIO)); } dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP); dsa->dsa_dr = NULL; dsa->dsa_done = done; dsa->dsa_zgd = zgd; dsa->dsa_tx = tx; zio_nowait(zio_write(pio, os->os_spa, dmu_tx_get_txg(tx), zgd->zgd_bp, zgd->zgd_db->db_data, zgd->zgd_db->db_size, zp, dmu_sync_late_arrival_ready, NULL, dmu_sync_late_arrival_done, dsa, ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, zb)); return (0); } /* * Intent log support: sync the block associated with db to disk. * N.B. and XXX: the caller is responsible for making sure that the * data isn't changing while dmu_sync() is writing it. * * Return values: * * EEXIST: this txg has already been synced, so there's nothing to do. * The caller should not log the write. * * ENOENT: the block was dbuf_free_range()'d, so there's nothing to do. * The caller should not log the write. * * EALREADY: this block is already in the process of being synced. * The caller should track its progress (somehow). * * EIO: could not do the I/O. * The caller should do a txg_wait_synced(). * * 0: the I/O has been initiated. * The caller should log this blkptr in the done callback. * It is possible that the I/O will fail, in which case * the error will be reported to the done callback and * propagated to pio from zio_done(). */ int dmu_sync(zio_t *pio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd) { blkptr_t *bp = zgd->zgd_bp; dmu_buf_impl_t *db = (dmu_buf_impl_t *)zgd->zgd_db; objset_t *os = db->db_objset; dsl_dataset_t *ds = os->os_dsl_dataset; dbuf_dirty_record_t *dr; dmu_sync_arg_t *dsa; zbookmark_phys_t zb; zio_prop_t zp; dnode_t *dn; ASSERT(pio != NULL); ASSERT(txg != 0); SET_BOOKMARK(&zb, ds->ds_object, db->db.db_object, db->db_level, db->db_blkid); DB_DNODE_ENTER(db); dn = DB_DNODE(db); dmu_write_policy(os, dn, db->db_level, WP_DMU_SYNC, &zp); DB_DNODE_EXIT(db); /* * If we're frozen (running ziltest), we always need to generate a bp. */ if (txg > spa_freeze_txg(os->os_spa)) return (dmu_sync_late_arrival(pio, os, done, zgd, &zp, &zb)); /* * Grabbing db_mtx now provides a barrier between dbuf_sync_leaf() * and us. If we determine that this txg is not yet syncing, * but it begins to sync a moment later, that's OK because the * sync thread will block in dbuf_sync_leaf() until we drop db_mtx. */ mutex_enter(&db->db_mtx); if (txg <= spa_last_synced_txg(os->os_spa)) { /* * This txg has already synced. There's nothing to do. */ mutex_exit(&db->db_mtx); return (SET_ERROR(EEXIST)); } if (txg <= spa_syncing_txg(os->os_spa)) { /* * This txg is currently syncing, so we can't mess with * the dirty record anymore; just write a new log block. */ mutex_exit(&db->db_mtx); return (dmu_sync_late_arrival(pio, os, done, zgd, &zp, &zb)); } dr = db->db_last_dirty; while (dr && dr->dr_txg != txg) dr = dr->dr_next; if (dr == NULL) { /* * There's no dr for this dbuf, so it must have been freed. * There's no need to log writes to freed blocks, so we're done. */ mutex_exit(&db->db_mtx); return (SET_ERROR(ENOENT)); } ASSERT(dr->dr_next == NULL || dr->dr_next->dr_txg < txg); /* * Assume the on-disk data is X, the current syncing data (in * txg - 1) is Y, and the current in-memory data is Z (currently * in dmu_sync). * * We usually want to perform a nopwrite if X and Z are the * same. However, if Y is different (i.e. the BP is going to * change before this write takes effect), then a nopwrite will * be incorrect - we would override with X, which could have * been freed when Y was written. * * (Note that this is not a concern when we are nop-writing from * syncing context, because X and Y must be identical, because * all previous txgs have been synced.) * * Therefore, we disable nopwrite if the current BP could change * before this TXG. There are two ways it could change: by * being dirty (dr_next is non-NULL), or by being freed * (dnode_block_freed()). This behavior is verified by * zio_done(), which VERIFYs that the override BP is identical * to the on-disk BP. */ DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (dr->dr_next != NULL || dnode_block_freed(dn, db->db_blkid)) zp.zp_nopwrite = B_FALSE; DB_DNODE_EXIT(db); ASSERT(dr->dr_txg == txg); if (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC || dr->dt.dl.dr_override_state == DR_OVERRIDDEN) { /* * We have already issued a sync write for this buffer, * or this buffer has already been synced. It could not * have been dirtied since, or we would have cleared the state. */ mutex_exit(&db->db_mtx); return (SET_ERROR(EALREADY)); } ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN); dr->dt.dl.dr_override_state = DR_IN_DMU_SYNC; mutex_exit(&db->db_mtx); dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP); dsa->dsa_dr = dr; dsa->dsa_done = done; dsa->dsa_zgd = zgd; dsa->dsa_tx = NULL; zio_nowait(arc_write(pio, os->os_spa, txg, bp, dr->dt.dl.dr_data, DBUF_IS_L2CACHEABLE(db), DBUF_IS_L2COMPRESSIBLE(db), &zp, dmu_sync_ready, NULL, dmu_sync_done, dsa, ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, &zb)); return (0); } int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, int ibs, dmu_tx_t *tx) { dnode_t *dn; int err; err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); err = dnode_set_blksz(dn, size, ibs, tx); dnode_rele(dn, FTAG); return (err); } void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, dmu_tx_t *tx) { dnode_t *dn; /* * Send streams include each object's checksum function. This * check ensures that the receiving system can understand the * checksum function transmitted. */ ASSERT3U(checksum, <, ZIO_CHECKSUM_LEGACY_FUNCTIONS); VERIFY0(dnode_hold(os, object, FTAG, &dn)); ASSERT3U(checksum, <, ZIO_CHECKSUM_FUNCTIONS); dn->dn_checksum = checksum; dnode_setdirty(dn, tx); dnode_rele(dn, FTAG); } void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, dmu_tx_t *tx) { dnode_t *dn; /* * Send streams include each object's compression function. This * check ensures that the receiving system can understand the * compression function transmitted. */ ASSERT3U(compress, <, ZIO_COMPRESS_LEGACY_FUNCTIONS); VERIFY0(dnode_hold(os, object, FTAG, &dn)); dn->dn_compress = compress; dnode_setdirty(dn, tx); dnode_rele(dn, FTAG); } int zfs_mdcomp_disable = 0; TUNABLE_INT("vfs.zfs.mdcomp_disable", &zfs_mdcomp_disable); SYSCTL_INT(_vfs_zfs, OID_AUTO, mdcomp_disable, CTLFLAG_RW, &zfs_mdcomp_disable, 0, "Disable metadata compression"); /* * When the "redundant_metadata" property is set to "most", only indirect * blocks of this level and higher will have an additional ditto block. */ int zfs_redundant_metadata_most_ditto_level = 2; void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, zio_prop_t *zp) { dmu_object_type_t type = dn ? dn->dn_type : DMU_OT_OBJSET; boolean_t ismd = (level > 0 || DMU_OT_IS_METADATA(type) || (wp & WP_SPILL)); enum zio_checksum checksum = os->os_checksum; enum zio_compress compress = os->os_compress; enum zio_checksum dedup_checksum = os->os_dedup_checksum; boolean_t dedup = B_FALSE; boolean_t nopwrite = B_FALSE; boolean_t dedup_verify = os->os_dedup_verify; int copies = os->os_copies; /* * We maintain different write policies for each of the following * types of data: * 1. metadata * 2. preallocated blocks (i.e. level-0 blocks of a dump device) * 3. all other level 0 blocks */ if (ismd) { if (zfs_mdcomp_disable) { compress = ZIO_COMPRESS_EMPTY; } else { /* * XXX -- we should design a compression algorithm * that specializes in arrays of bps. */ compress = zio_compress_select(os->os_spa, ZIO_COMPRESS_ON, ZIO_COMPRESS_ON); } /* * Metadata always gets checksummed. If the data * checksum is multi-bit correctable, and it's not a * ZBT-style checksum, then it's suitable for metadata * as well. Otherwise, the metadata checksum defaults * to fletcher4. */ - if (zio_checksum_table[checksum].ci_correctable < 1 || - zio_checksum_table[checksum].ci_eck) + if (!(zio_checksum_table[checksum].ci_flags & + ZCHECKSUM_FLAG_METADATA) || + (zio_checksum_table[checksum].ci_flags & + ZCHECKSUM_FLAG_EMBEDDED)) checksum = ZIO_CHECKSUM_FLETCHER_4; if (os->os_redundant_metadata == ZFS_REDUNDANT_METADATA_ALL || (os->os_redundant_metadata == ZFS_REDUNDANT_METADATA_MOST && (level >= zfs_redundant_metadata_most_ditto_level || DMU_OT_IS_METADATA(type) || (wp & WP_SPILL)))) copies++; } else if (wp & WP_NOFILL) { ASSERT(level == 0); /* * If we're writing preallocated blocks, we aren't actually * writing them so don't set any policy properties. These * blocks are currently only used by an external subsystem * outside of zfs (i.e. dump) and not written by the zio * pipeline. */ compress = ZIO_COMPRESS_OFF; checksum = ZIO_CHECKSUM_NOPARITY; } else { compress = zio_compress_select(os->os_spa, dn->dn_compress, compress); checksum = (dedup_checksum == ZIO_CHECKSUM_OFF) ? zio_checksum_select(dn->dn_checksum, checksum) : dedup_checksum; /* * Determine dedup setting. If we are in dmu_sync(), * we won't actually dedup now because that's all * done in syncing context; but we do want to use the * dedup checkum. If the checksum is not strong * enough to ensure unique signatures, force * dedup_verify. */ if (dedup_checksum != ZIO_CHECKSUM_OFF) { dedup = (wp & WP_DMU_SYNC) ? B_FALSE : B_TRUE; - if (!zio_checksum_table[checksum].ci_dedup) + if (!(zio_checksum_table[checksum].ci_flags & + ZCHECKSUM_FLAG_DEDUP)) dedup_verify = B_TRUE; } /* - * Enable nopwrite if we have a cryptographically secure - * checksum that has no known collisions (i.e. SHA-256) - * and compression is enabled. We don't enable nopwrite if - * dedup is enabled as the two features are mutually exclusive. + * Enable nopwrite if we have secure enough checksum + * algorithm (see comment in zio_nop_write) and + * compression is enabled. We don't enable nopwrite if + * dedup is enabled as the two features are mutually + * exclusive. */ - nopwrite = (!dedup && zio_checksum_table[checksum].ci_dedup && + nopwrite = (!dedup && (zio_checksum_table[checksum].ci_flags & + ZCHECKSUM_FLAG_NOPWRITE) && compress != ZIO_COMPRESS_OFF && zfs_nopwrite_enabled); } zp->zp_checksum = checksum; zp->zp_compress = compress; zp->zp_type = (wp & WP_SPILL) ? dn->dn_bonustype : type; zp->zp_level = level; zp->zp_copies = MIN(copies, spa_max_replication(os->os_spa)); zp->zp_dedup = dedup; zp->zp_dedup_verify = dedup && dedup_verify; zp->zp_nopwrite = nopwrite; } int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, uint64_t *off) { dnode_t *dn; int err; /* * Sync any current changes before * we go trundling through the block pointers. */ err = dmu_object_wait_synced(os, object); if (err) { return (err); } err = dnode_hold(os, object, FTAG, &dn); if (err) { return (err); } err = dnode_next_offset(dn, (hole ? DNODE_FIND_HOLE : 0), off, 1, 1, 0); dnode_rele(dn, FTAG); return (err); } /* * Given the ZFS object, if it contains any dirty nodes * this function flushes all dirty blocks to disk. This * ensures the DMU object info is updated. A more efficient * future version might just find the TXG with the maximum * ID and wait for that to be synced. */ int dmu_object_wait_synced(objset_t *os, uint64_t object) { dnode_t *dn; int error, i; error = dnode_hold(os, object, FTAG, &dn); if (error) { return (error); } for (i = 0; i < TXG_SIZE; i++) { if (list_link_active(&dn->dn_dirty_link[i])) { break; } } dnode_rele(dn, FTAG); if (i != TXG_SIZE) { txg_wait_synced(dmu_objset_pool(os), 0); } return (0); } void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi) { dnode_phys_t *dnp; rw_enter(&dn->dn_struct_rwlock, RW_READER); mutex_enter(&dn->dn_mtx); dnp = dn->dn_phys; doi->doi_data_block_size = dn->dn_datablksz; doi->doi_metadata_block_size = dn->dn_indblkshift ? 1ULL << dn->dn_indblkshift : 0; doi->doi_type = dn->dn_type; doi->doi_bonus_type = dn->dn_bonustype; doi->doi_bonus_size = dn->dn_bonuslen; doi->doi_indirection = dn->dn_nlevels; doi->doi_checksum = dn->dn_checksum; doi->doi_compress = dn->dn_compress; doi->doi_nblkptr = dn->dn_nblkptr; doi->doi_physical_blocks_512 = (DN_USED_BYTES(dnp) + 256) >> 9; doi->doi_max_offset = (dn->dn_maxblkid + 1) * dn->dn_datablksz; doi->doi_fill_count = 0; for (int i = 0; i < dnp->dn_nblkptr; i++) doi->doi_fill_count += BP_GET_FILL(&dnp->dn_blkptr[i]); mutex_exit(&dn->dn_mtx); rw_exit(&dn->dn_struct_rwlock); } /* * Get information on a DMU object. * If doi is NULL, just indicates whether the object exists. */ int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi) { dnode_t *dn; int err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); if (doi != NULL) dmu_object_info_from_dnode(dn, doi); dnode_rele(dn, FTAG); return (0); } /* * As above, but faster; can be used when you have a held dbuf in hand. */ void dmu_object_info_from_db(dmu_buf_t *db_fake, dmu_object_info_t *doi) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; DB_DNODE_ENTER(db); dmu_object_info_from_dnode(DB_DNODE(db), doi); DB_DNODE_EXIT(db); } /* * Faster still when you only care about the size. * This is specifically optimized for zfs_getattr(). */ void dmu_object_size_from_db(dmu_buf_t *db_fake, uint32_t *blksize, u_longlong_t *nblk512) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; DB_DNODE_ENTER(db); dn = DB_DNODE(db); *blksize = dn->dn_datablksz; /* add 1 for dnode space */ *nblk512 = ((DN_USED_BYTES(dn->dn_phys) + SPA_MINBLOCKSIZE/2) >> SPA_MINBLOCKSHIFT) + 1; DB_DNODE_EXIT(db); } void byteswap_uint64_array(void *vbuf, size_t size) { uint64_t *buf = vbuf; size_t count = size >> 3; int i; ASSERT((size & 7) == 0); for (i = 0; i < count; i++) buf[i] = BSWAP_64(buf[i]); } void byteswap_uint32_array(void *vbuf, size_t size) { uint32_t *buf = vbuf; size_t count = size >> 2; int i; ASSERT((size & 3) == 0); for (i = 0; i < count; i++) buf[i] = BSWAP_32(buf[i]); } void byteswap_uint16_array(void *vbuf, size_t size) { uint16_t *buf = vbuf; size_t count = size >> 1; int i; ASSERT((size & 1) == 0); for (i = 0; i < count; i++) buf[i] = BSWAP_16(buf[i]); } /* ARGSUSED */ void byteswap_uint8_array(void *vbuf, size_t size) { } void dmu_init(void) { zfs_dbgmsg_init(); sa_cache_init(); xuio_stat_init(); dmu_objset_init(); dnode_init(); dbuf_init(); zfetch_init(); zio_compress_init(); l2arc_init(); arc_init(); } void dmu_fini(void) { arc_fini(); /* arc depends on l2arc, so arc must go first */ l2arc_fini(); zfetch_fini(); zio_compress_fini(); dbuf_fini(); dnode_fini(); dmu_objset_fini(); xuio_stat_fini(); sa_cache_fini(); zfs_dbgmsg_fini(); } Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu_send.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu_send.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu_send.c (revision 290757) @@ -1,3166 +1,3167 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2011, 2015 by Delphix. All rights reserved. * Copyright (c) 2014, Joyent, Inc. All rights reserved. * Copyright (c) 2012, Martin Matuska . All rights reserved. * Copyright 2014 HybridCluster. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef __FreeBSD__ #undef dump_write #define dump_write dmu_dump_write #endif /* Set this tunable to TRUE to replace corrupt data with 0x2f5baddb10c */ int zfs_send_corrupt_data = B_FALSE; int zfs_send_queue_length = 16 * 1024 * 1024; int zfs_recv_queue_length = 16 * 1024 * 1024; static char *dmu_recv_tag = "dmu_recv_tag"; const char *recv_clone_name = "%recv"; #define BP_SPAN(datablkszsec, indblkshift, level) \ (((uint64_t)datablkszsec) << (SPA_MINBLOCKSHIFT + \ (level) * (indblkshift - SPA_BLKPTRSHIFT))) static void byteswap_record(dmu_replay_record_t *drr); struct send_thread_arg { bqueue_t q; dsl_dataset_t *ds; /* Dataset to traverse */ uint64_t fromtxg; /* Traverse from this txg */ int flags; /* flags to pass to traverse_dataset */ int error_code; boolean_t cancel; zbookmark_phys_t resume; }; struct send_block_record { boolean_t eos_marker; /* Marks the end of the stream */ blkptr_t bp; zbookmark_phys_t zb; uint8_t indblkshift; uint16_t datablkszsec; bqueue_node_t ln; }; static int dump_bytes(dmu_sendarg_t *dsp, void *buf, int len) { dsl_dataset_t *ds = dmu_objset_ds(dsp->dsa_os); struct uio auio; struct iovec aiov; ASSERT0(len % 8); aiov.iov_base = buf; aiov.iov_len = len; auio.uio_iov = &aiov; auio.uio_iovcnt = 1; auio.uio_resid = len; auio.uio_segflg = UIO_SYSSPACE; auio.uio_rw = UIO_WRITE; auio.uio_offset = (off_t)-1; auio.uio_td = dsp->dsa_td; #ifdef _KERNEL if (dsp->dsa_fp->f_type == DTYPE_VNODE) bwillwrite(); dsp->dsa_err = fo_write(dsp->dsa_fp, &auio, dsp->dsa_td->td_ucred, 0, dsp->dsa_td); #else fprintf(stderr, "%s: returning EOPNOTSUPP\n", __func__); dsp->dsa_err = EOPNOTSUPP; #endif mutex_enter(&ds->ds_sendstream_lock); *dsp->dsa_off += len; mutex_exit(&ds->ds_sendstream_lock); return (dsp->dsa_err); } /* * For all record types except BEGIN, fill in the checksum (overlaid in * drr_u.drr_checksum.drr_checksum). The checksum verifies everything * up to the start of the checksum itself. */ static int dump_record(dmu_sendarg_t *dsp, void *payload, int payload_len) { ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), ==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t)); fletcher_4_incremental_native(dsp->dsa_drr, offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), &dsp->dsa_zc); if (dsp->dsa_drr->drr_type != DRR_BEGIN) { ASSERT(ZIO_CHECKSUM_IS_ZERO(&dsp->dsa_drr->drr_u. drr_checksum.drr_checksum)); dsp->dsa_drr->drr_u.drr_checksum.drr_checksum = dsp->dsa_zc; } fletcher_4_incremental_native(&dsp->dsa_drr-> drr_u.drr_checksum.drr_checksum, sizeof (zio_cksum_t), &dsp->dsa_zc); if (dump_bytes(dsp, dsp->dsa_drr, sizeof (dmu_replay_record_t)) != 0) return (SET_ERROR(EINTR)); if (payload_len != 0) { fletcher_4_incremental_native(payload, payload_len, &dsp->dsa_zc); if (dump_bytes(dsp, payload, payload_len) != 0) return (SET_ERROR(EINTR)); } return (0); } static int dump_free(dmu_sendarg_t *dsp, uint64_t object, uint64_t offset, uint64_t length) { struct drr_free *drrf = &(dsp->dsa_drr->drr_u.drr_free); /* * When we receive a free record, dbuf_free_range() assumes * that the receiving system doesn't have any dbufs in the range * being freed. This is always true because there is a one-record * constraint: we only send one WRITE record for any given * object,offset. We know that the one-record constraint is * true because we always send data in increasing order by * object,offset. * * If the increasing-order constraint ever changes, we should find * another way to assert that the one-record constraint is still * satisfied. */ ASSERT(object > dsp->dsa_last_data_object || (object == dsp->dsa_last_data_object && offset > dsp->dsa_last_data_offset)); /* * If we are doing a non-incremental send, then there can't * be any data in the dataset we're receiving into. Therefore * a free record would simply be a no-op. Save space by not * sending it to begin with. */ if (!dsp->dsa_incremental) return (0); if (length != -1ULL && offset + length < offset) length = -1ULL; /* * If there is a pending op, but it's not PENDING_FREE, push it out, * since free block aggregation can only be done for blocks of the * same type (i.e., DRR_FREE records can only be aggregated with * other DRR_FREE records. DRR_FREEOBJECTS records can only be * aggregated with other DRR_FREEOBJECTS records. */ if (dsp->dsa_pending_op != PENDING_NONE && dsp->dsa_pending_op != PENDING_FREE) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } if (dsp->dsa_pending_op == PENDING_FREE) { /* * There should never be a PENDING_FREE if length is -1 * (because dump_dnode is the only place where this * function is called with a -1, and only after flushing * any pending record). */ ASSERT(length != -1ULL); /* * Check to see whether this free block can be aggregated * with pending one. */ if (drrf->drr_object == object && drrf->drr_offset + drrf->drr_length == offset) { drrf->drr_length += length; return (0); } else { /* not a continuation. Push out pending record */ if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } } /* create a FREE record and make it pending */ bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_FREE; drrf->drr_object = object; drrf->drr_offset = offset; drrf->drr_length = length; drrf->drr_toguid = dsp->dsa_toguid; if (length == -1ULL) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); } else { dsp->dsa_pending_op = PENDING_FREE; } return (0); } static int dump_write(dmu_sendarg_t *dsp, dmu_object_type_t type, uint64_t object, uint64_t offset, int blksz, const blkptr_t *bp, void *data) { struct drr_write *drrw = &(dsp->dsa_drr->drr_u.drr_write); /* * We send data in increasing object, offset order. * See comment in dump_free() for details. */ ASSERT(object > dsp->dsa_last_data_object || (object == dsp->dsa_last_data_object && offset > dsp->dsa_last_data_offset)); dsp->dsa_last_data_object = object; dsp->dsa_last_data_offset = offset + blksz - 1; /* * If there is any kind of pending aggregation (currently either * a grouping of free objects or free blocks), push it out to * the stream, since aggregation can't be done across operations * of different types. */ if (dsp->dsa_pending_op != PENDING_NONE) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } /* write a WRITE record */ bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_WRITE; drrw->drr_object = object; drrw->drr_type = type; drrw->drr_offset = offset; drrw->drr_length = blksz; drrw->drr_toguid = dsp->dsa_toguid; if (bp == NULL || BP_IS_EMBEDDED(bp)) { /* * There's no pre-computed checksum for partial-block * writes or embedded BP's, so (like * fletcher4-checkummed blocks) userland will have to * compute a dedup-capable checksum itself. */ drrw->drr_checksumtype = ZIO_CHECKSUM_OFF; } else { drrw->drr_checksumtype = BP_GET_CHECKSUM(bp); - if (zio_checksum_table[drrw->drr_checksumtype].ci_dedup) + if (zio_checksum_table[drrw->drr_checksumtype].ci_flags & + ZCHECKSUM_FLAG_DEDUP) drrw->drr_checksumflags |= DRR_CHECKSUM_DEDUP; DDK_SET_LSIZE(&drrw->drr_key, BP_GET_LSIZE(bp)); DDK_SET_PSIZE(&drrw->drr_key, BP_GET_PSIZE(bp)); DDK_SET_COMPRESS(&drrw->drr_key, BP_GET_COMPRESS(bp)); drrw->drr_key.ddk_cksum = bp->blk_cksum; } if (dump_record(dsp, data, blksz) != 0) return (SET_ERROR(EINTR)); return (0); } static int dump_write_embedded(dmu_sendarg_t *dsp, uint64_t object, uint64_t offset, int blksz, const blkptr_t *bp) { char buf[BPE_PAYLOAD_SIZE]; struct drr_write_embedded *drrw = &(dsp->dsa_drr->drr_u.drr_write_embedded); if (dsp->dsa_pending_op != PENDING_NONE) { if (dump_record(dsp, NULL, 0) != 0) return (EINTR); dsp->dsa_pending_op = PENDING_NONE; } ASSERT(BP_IS_EMBEDDED(bp)); bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_WRITE_EMBEDDED; drrw->drr_object = object; drrw->drr_offset = offset; drrw->drr_length = blksz; drrw->drr_toguid = dsp->dsa_toguid; drrw->drr_compression = BP_GET_COMPRESS(bp); drrw->drr_etype = BPE_GET_ETYPE(bp); drrw->drr_lsize = BPE_GET_LSIZE(bp); drrw->drr_psize = BPE_GET_PSIZE(bp); decode_embedded_bp_compressed(bp, buf); if (dump_record(dsp, buf, P2ROUNDUP(drrw->drr_psize, 8)) != 0) return (EINTR); return (0); } static int dump_spill(dmu_sendarg_t *dsp, uint64_t object, int blksz, void *data) { struct drr_spill *drrs = &(dsp->dsa_drr->drr_u.drr_spill); if (dsp->dsa_pending_op != PENDING_NONE) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } /* write a SPILL record */ bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_SPILL; drrs->drr_object = object; drrs->drr_length = blksz; drrs->drr_toguid = dsp->dsa_toguid; if (dump_record(dsp, data, blksz) != 0) return (SET_ERROR(EINTR)); return (0); } static int dump_freeobjects(dmu_sendarg_t *dsp, uint64_t firstobj, uint64_t numobjs) { struct drr_freeobjects *drrfo = &(dsp->dsa_drr->drr_u.drr_freeobjects); /* See comment in dump_free(). */ if (!dsp->dsa_incremental) return (0); /* * If there is a pending op, but it's not PENDING_FREEOBJECTS, * push it out, since free block aggregation can only be done for * blocks of the same type (i.e., DRR_FREE records can only be * aggregated with other DRR_FREE records. DRR_FREEOBJECTS records * can only be aggregated with other DRR_FREEOBJECTS records. */ if (dsp->dsa_pending_op != PENDING_NONE && dsp->dsa_pending_op != PENDING_FREEOBJECTS) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } if (dsp->dsa_pending_op == PENDING_FREEOBJECTS) { /* * See whether this free object array can be aggregated * with pending one */ if (drrfo->drr_firstobj + drrfo->drr_numobjs == firstobj) { drrfo->drr_numobjs += numobjs; return (0); } else { /* can't be aggregated. Push out pending record */ if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } } /* write a FREEOBJECTS record */ bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_FREEOBJECTS; drrfo->drr_firstobj = firstobj; drrfo->drr_numobjs = numobjs; drrfo->drr_toguid = dsp->dsa_toguid; dsp->dsa_pending_op = PENDING_FREEOBJECTS; return (0); } static int dump_dnode(dmu_sendarg_t *dsp, uint64_t object, dnode_phys_t *dnp) { struct drr_object *drro = &(dsp->dsa_drr->drr_u.drr_object); if (object < dsp->dsa_resume_object) { /* * Note: when resuming, we will visit all the dnodes in * the block of dnodes that we are resuming from. In * this case it's unnecessary to send the dnodes prior to * the one we are resuming from. We should be at most one * block's worth of dnodes behind the resume point. */ ASSERT3U(dsp->dsa_resume_object - object, <, 1 << (DNODE_BLOCK_SHIFT - DNODE_SHIFT)); return (0); } if (dnp == NULL || dnp->dn_type == DMU_OT_NONE) return (dump_freeobjects(dsp, object, 1)); if (dsp->dsa_pending_op != PENDING_NONE) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } /* write an OBJECT record */ bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_OBJECT; drro->drr_object = object; drro->drr_type = dnp->dn_type; drro->drr_bonustype = dnp->dn_bonustype; drro->drr_blksz = dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT; drro->drr_bonuslen = dnp->dn_bonuslen; drro->drr_checksumtype = dnp->dn_checksum; drro->drr_compress = dnp->dn_compress; drro->drr_toguid = dsp->dsa_toguid; if (!(dsp->dsa_featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) && drro->drr_blksz > SPA_OLD_MAXBLOCKSIZE) drro->drr_blksz = SPA_OLD_MAXBLOCKSIZE; if (dump_record(dsp, DN_BONUS(dnp), P2ROUNDUP(dnp->dn_bonuslen, 8)) != 0) { return (SET_ERROR(EINTR)); } /* Free anything past the end of the file. */ if (dump_free(dsp, object, (dnp->dn_maxblkid + 1) * (dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT), -1ULL) != 0) return (SET_ERROR(EINTR)); if (dsp->dsa_err != 0) return (SET_ERROR(EINTR)); return (0); } static boolean_t backup_do_embed(dmu_sendarg_t *dsp, const blkptr_t *bp) { if (!BP_IS_EMBEDDED(bp)) return (B_FALSE); /* * Compression function must be legacy, or explicitly enabled. */ if ((BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_LEGACY_FUNCTIONS && !(dsp->dsa_featureflags & DMU_BACKUP_FEATURE_EMBED_DATA_LZ4))) return (B_FALSE); /* * Embed type must be explicitly enabled. */ switch (BPE_GET_ETYPE(bp)) { case BP_EMBEDDED_TYPE_DATA: if (dsp->dsa_featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) return (B_TRUE); break; default: return (B_FALSE); } return (B_FALSE); } /* * This is the callback function to traverse_dataset that acts as the worker * thread for dmu_send_impl. */ /*ARGSUSED*/ static int send_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, const zbookmark_phys_t *zb, const struct dnode_phys *dnp, void *arg) { struct send_thread_arg *sta = arg; struct send_block_record *record; uint64_t record_size; int err = 0; ASSERT(zb->zb_object == DMU_META_DNODE_OBJECT || zb->zb_object >= sta->resume.zb_object); if (sta->cancel) return (SET_ERROR(EINTR)); if (bp == NULL) { ASSERT3U(zb->zb_level, ==, ZB_DNODE_LEVEL); return (0); } else if (zb->zb_level < 0) { return (0); } record = kmem_zalloc(sizeof (struct send_block_record), KM_SLEEP); record->eos_marker = B_FALSE; record->bp = *bp; record->zb = *zb; record->indblkshift = dnp->dn_indblkshift; record->datablkszsec = dnp->dn_datablkszsec; record_size = dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT; bqueue_enqueue(&sta->q, record, record_size); return (err); } /* * This function kicks off the traverse_dataset. It also handles setting the * error code of the thread in case something goes wrong, and pushes the End of * Stream record when the traverse_dataset call has finished. If there is no * dataset to traverse, the thread immediately pushes End of Stream marker. */ static void send_traverse_thread(void *arg) { struct send_thread_arg *st_arg = arg; int err; struct send_block_record *data; if (st_arg->ds != NULL) { err = traverse_dataset_resume(st_arg->ds, st_arg->fromtxg, &st_arg->resume, st_arg->flags, send_cb, st_arg); if (err != EINTR) st_arg->error_code = err; } data = kmem_zalloc(sizeof (*data), KM_SLEEP); data->eos_marker = B_TRUE; bqueue_enqueue(&st_arg->q, data, 1); thread_exit(); } /* * This function actually handles figuring out what kind of record needs to be * dumped, reading the data (which has hopefully been prefetched), and calling * the appropriate helper function. */ static int do_dump(dmu_sendarg_t *dsa, struct send_block_record *data) { dsl_dataset_t *ds = dmu_objset_ds(dsa->dsa_os); const blkptr_t *bp = &data->bp; const zbookmark_phys_t *zb = &data->zb; uint8_t indblkshift = data->indblkshift; uint16_t dblkszsec = data->datablkszsec; spa_t *spa = ds->ds_dir->dd_pool->dp_spa; dmu_object_type_t type = bp ? BP_GET_TYPE(bp) : DMU_OT_NONE; int err = 0; ASSERT3U(zb->zb_level, >=, 0); ASSERT(zb->zb_object == DMU_META_DNODE_OBJECT || zb->zb_object >= dsa->dsa_resume_object); if (zb->zb_object != DMU_META_DNODE_OBJECT && DMU_OBJECT_IS_SPECIAL(zb->zb_object)) { return (0); } else if (BP_IS_HOLE(bp) && zb->zb_object == DMU_META_DNODE_OBJECT) { uint64_t span = BP_SPAN(dblkszsec, indblkshift, zb->zb_level); uint64_t dnobj = (zb->zb_blkid * span) >> DNODE_SHIFT; err = dump_freeobjects(dsa, dnobj, span >> DNODE_SHIFT); } else if (BP_IS_HOLE(bp)) { uint64_t span = BP_SPAN(dblkszsec, indblkshift, zb->zb_level); uint64_t offset = zb->zb_blkid * span; err = dump_free(dsa, zb->zb_object, offset, span); } else if (zb->zb_level > 0 || type == DMU_OT_OBJSET) { return (0); } else if (type == DMU_OT_DNODE) { int blksz = BP_GET_LSIZE(bp); arc_flags_t aflags = ARC_FLAG_WAIT; arc_buf_t *abuf; ASSERT0(zb->zb_level); if (arc_read(NULL, spa, bp, arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &aflags, zb) != 0) return (SET_ERROR(EIO)); dnode_phys_t *blk = abuf->b_data; uint64_t dnobj = zb->zb_blkid * (blksz >> DNODE_SHIFT); for (int i = 0; i < blksz >> DNODE_SHIFT; i++) { err = dump_dnode(dsa, dnobj + i, blk + i); if (err != 0) break; } (void) arc_buf_remove_ref(abuf, &abuf); } else if (type == DMU_OT_SA) { arc_flags_t aflags = ARC_FLAG_WAIT; arc_buf_t *abuf; int blksz = BP_GET_LSIZE(bp); if (arc_read(NULL, spa, bp, arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &aflags, zb) != 0) return (SET_ERROR(EIO)); err = dump_spill(dsa, zb->zb_object, blksz, abuf->b_data); (void) arc_buf_remove_ref(abuf, &abuf); } else if (backup_do_embed(dsa, bp)) { /* it's an embedded level-0 block of a regular object */ int blksz = dblkszsec << SPA_MINBLOCKSHIFT; ASSERT0(zb->zb_level); err = dump_write_embedded(dsa, zb->zb_object, zb->zb_blkid * blksz, blksz, bp); } else { /* it's a level-0 block of a regular object */ arc_flags_t aflags = ARC_FLAG_WAIT; arc_buf_t *abuf; int blksz = dblkszsec << SPA_MINBLOCKSHIFT; uint64_t offset; ASSERT0(zb->zb_level); ASSERT(zb->zb_object > dsa->dsa_resume_object || (zb->zb_object == dsa->dsa_resume_object && zb->zb_blkid * blksz >= dsa->dsa_resume_offset)); if (arc_read(NULL, spa, bp, arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &aflags, zb) != 0) { if (zfs_send_corrupt_data) { /* Send a block filled with 0x"zfs badd bloc" */ abuf = arc_buf_alloc(spa, blksz, &abuf, ARC_BUFC_DATA); uint64_t *ptr; for (ptr = abuf->b_data; (char *)ptr < (char *)abuf->b_data + blksz; ptr++) *ptr = 0x2f5baddb10cULL; } else { return (SET_ERROR(EIO)); } } offset = zb->zb_blkid * blksz; if (!(dsa->dsa_featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) && blksz > SPA_OLD_MAXBLOCKSIZE) { char *buf = abuf->b_data; while (blksz > 0 && err == 0) { int n = MIN(blksz, SPA_OLD_MAXBLOCKSIZE); err = dump_write(dsa, type, zb->zb_object, offset, n, NULL, buf); offset += n; buf += n; blksz -= n; } } else { err = dump_write(dsa, type, zb->zb_object, offset, blksz, bp, abuf->b_data); } (void) arc_buf_remove_ref(abuf, &abuf); } ASSERT(err == 0 || err == EINTR); return (err); } /* * Pop the new data off the queue, and free the old data. */ static struct send_block_record * get_next_record(bqueue_t *bq, struct send_block_record *data) { struct send_block_record *tmp = bqueue_dequeue(bq); kmem_free(data, sizeof (*data)); return (tmp); } /* * Actually do the bulk of the work in a zfs send. * * Note: Releases dp using the specified tag. */ static int dmu_send_impl(void *tag, dsl_pool_t *dp, dsl_dataset_t *to_ds, zfs_bookmark_phys_t *ancestor_zb, boolean_t is_clone, boolean_t embedok, boolean_t large_block_ok, int outfd, uint64_t resumeobj, uint64_t resumeoff, #ifdef illumos vnode_t *vp, offset_t *off) #else struct file *fp, offset_t *off) #endif { objset_t *os; dmu_replay_record_t *drr; dmu_sendarg_t *dsp; int err; uint64_t fromtxg = 0; uint64_t featureflags = 0; struct send_thread_arg to_arg = { 0 }; err = dmu_objset_from_ds(to_ds, &os); if (err != 0) { dsl_pool_rele(dp, tag); return (err); } drr = kmem_zalloc(sizeof (dmu_replay_record_t), KM_SLEEP); drr->drr_type = DRR_BEGIN; drr->drr_u.drr_begin.drr_magic = DMU_BACKUP_MAGIC; DMU_SET_STREAM_HDRTYPE(drr->drr_u.drr_begin.drr_versioninfo, DMU_SUBSTREAM); #ifdef _KERNEL if (dmu_objset_type(os) == DMU_OST_ZFS) { uint64_t version; if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &version) != 0) { kmem_free(drr, sizeof (dmu_replay_record_t)); dsl_pool_rele(dp, tag); return (SET_ERROR(EINVAL)); } if (version >= ZPL_VERSION_SA) { featureflags |= DMU_BACKUP_FEATURE_SA_SPILL; } } #endif if (large_block_ok && to_ds->ds_feature_inuse[SPA_FEATURE_LARGE_BLOCKS]) featureflags |= DMU_BACKUP_FEATURE_LARGE_BLOCKS; if (embedok && spa_feature_is_active(dp->dp_spa, SPA_FEATURE_EMBEDDED_DATA)) { featureflags |= DMU_BACKUP_FEATURE_EMBED_DATA; if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_LZ4_COMPRESS)) featureflags |= DMU_BACKUP_FEATURE_EMBED_DATA_LZ4; } if (resumeobj != 0 || resumeoff != 0) { featureflags |= DMU_BACKUP_FEATURE_RESUMING; } DMU_SET_FEATUREFLAGS(drr->drr_u.drr_begin.drr_versioninfo, featureflags); drr->drr_u.drr_begin.drr_creation_time = dsl_dataset_phys(to_ds)->ds_creation_time; drr->drr_u.drr_begin.drr_type = dmu_objset_type(os); if (is_clone) drr->drr_u.drr_begin.drr_flags |= DRR_FLAG_CLONE; drr->drr_u.drr_begin.drr_toguid = dsl_dataset_phys(to_ds)->ds_guid; if (dsl_dataset_phys(to_ds)->ds_flags & DS_FLAG_CI_DATASET) drr->drr_u.drr_begin.drr_flags |= DRR_FLAG_CI_DATA; if (ancestor_zb != NULL) { drr->drr_u.drr_begin.drr_fromguid = ancestor_zb->zbm_guid; fromtxg = ancestor_zb->zbm_creation_txg; } dsl_dataset_name(to_ds, drr->drr_u.drr_begin.drr_toname); if (!to_ds->ds_is_snapshot) { (void) strlcat(drr->drr_u.drr_begin.drr_toname, "@--head--", sizeof (drr->drr_u.drr_begin.drr_toname)); } dsp = kmem_zalloc(sizeof (dmu_sendarg_t), KM_SLEEP); dsp->dsa_drr = drr; dsp->dsa_outfd = outfd; dsp->dsa_proc = curproc; dsp->dsa_td = curthread; dsp->dsa_fp = fp; dsp->dsa_os = os; dsp->dsa_off = off; dsp->dsa_toguid = dsl_dataset_phys(to_ds)->ds_guid; dsp->dsa_pending_op = PENDING_NONE; dsp->dsa_incremental = (ancestor_zb != NULL); dsp->dsa_featureflags = featureflags; dsp->dsa_resume_object = resumeobj; dsp->dsa_resume_offset = resumeoff; mutex_enter(&to_ds->ds_sendstream_lock); list_insert_head(&to_ds->ds_sendstreams, dsp); mutex_exit(&to_ds->ds_sendstream_lock); dsl_dataset_long_hold(to_ds, FTAG); dsl_pool_rele(dp, tag); void *payload = NULL; size_t payload_len = 0; if (resumeobj != 0 || resumeoff != 0) { dmu_object_info_t to_doi; err = dmu_object_info(os, resumeobj, &to_doi); if (err != 0) goto out; SET_BOOKMARK(&to_arg.resume, to_ds->ds_object, resumeobj, 0, resumeoff / to_doi.doi_data_block_size); nvlist_t *nvl = fnvlist_alloc(); fnvlist_add_uint64(nvl, "resume_object", resumeobj); fnvlist_add_uint64(nvl, "resume_offset", resumeoff); payload = fnvlist_pack(nvl, &payload_len); drr->drr_payloadlen = payload_len; fnvlist_free(nvl); } err = dump_record(dsp, payload, payload_len); fnvlist_pack_free(payload, payload_len); if (err != 0) { err = dsp->dsa_err; goto out; } err = bqueue_init(&to_arg.q, zfs_send_queue_length, offsetof(struct send_block_record, ln)); to_arg.error_code = 0; to_arg.cancel = B_FALSE; to_arg.ds = to_ds; to_arg.fromtxg = fromtxg; to_arg.flags = TRAVERSE_PRE | TRAVERSE_PREFETCH; (void) thread_create(NULL, 0, send_traverse_thread, &to_arg, 0, &p0, TS_RUN, minclsyspri); struct send_block_record *to_data; to_data = bqueue_dequeue(&to_arg.q); while (!to_data->eos_marker && err == 0) { err = do_dump(dsp, to_data); to_data = get_next_record(&to_arg.q, to_data); if (issig(JUSTLOOKING) && issig(FORREAL)) err = EINTR; } if (err != 0) { to_arg.cancel = B_TRUE; while (!to_data->eos_marker) { to_data = get_next_record(&to_arg.q, to_data); } } kmem_free(to_data, sizeof (*to_data)); bqueue_destroy(&to_arg.q); if (err == 0 && to_arg.error_code != 0) err = to_arg.error_code; if (err != 0) goto out; if (dsp->dsa_pending_op != PENDING_NONE) if (dump_record(dsp, NULL, 0) != 0) err = SET_ERROR(EINTR); if (err != 0) { if (err == EINTR && dsp->dsa_err != 0) err = dsp->dsa_err; goto out; } bzero(drr, sizeof (dmu_replay_record_t)); drr->drr_type = DRR_END; drr->drr_u.drr_end.drr_checksum = dsp->dsa_zc; drr->drr_u.drr_end.drr_toguid = dsp->dsa_toguid; if (dump_record(dsp, NULL, 0) != 0) err = dsp->dsa_err; out: mutex_enter(&to_ds->ds_sendstream_lock); list_remove(&to_ds->ds_sendstreams, dsp); mutex_exit(&to_ds->ds_sendstream_lock); kmem_free(drr, sizeof (dmu_replay_record_t)); kmem_free(dsp, sizeof (dmu_sendarg_t)); dsl_dataset_long_rele(to_ds, FTAG); return (err); } int dmu_send_obj(const char *pool, uint64_t tosnap, uint64_t fromsnap, boolean_t embedok, boolean_t large_block_ok, #ifdef illumos int outfd, vnode_t *vp, offset_t *off) #else int outfd, struct file *fp, offset_t *off) #endif { dsl_pool_t *dp; dsl_dataset_t *ds; dsl_dataset_t *fromds = NULL; int err; err = dsl_pool_hold(pool, FTAG, &dp); if (err != 0) return (err); err = dsl_dataset_hold_obj(dp, tosnap, FTAG, &ds); if (err != 0) { dsl_pool_rele(dp, FTAG); return (err); } if (fromsnap != 0) { zfs_bookmark_phys_t zb; boolean_t is_clone; err = dsl_dataset_hold_obj(dp, fromsnap, FTAG, &fromds); if (err != 0) { dsl_dataset_rele(ds, FTAG); dsl_pool_rele(dp, FTAG); return (err); } if (!dsl_dataset_is_before(ds, fromds, 0)) err = SET_ERROR(EXDEV); zb.zbm_creation_time = dsl_dataset_phys(fromds)->ds_creation_time; zb.zbm_creation_txg = dsl_dataset_phys(fromds)->ds_creation_txg; zb.zbm_guid = dsl_dataset_phys(fromds)->ds_guid; is_clone = (fromds->ds_dir != ds->ds_dir); dsl_dataset_rele(fromds, FTAG); err = dmu_send_impl(FTAG, dp, ds, &zb, is_clone, embedok, large_block_ok, outfd, 0, 0, fp, off); } else { err = dmu_send_impl(FTAG, dp, ds, NULL, B_FALSE, embedok, large_block_ok, outfd, 0, 0, fp, off); } dsl_dataset_rele(ds, FTAG); return (err); } int dmu_send(const char *tosnap, const char *fromsnap, boolean_t embedok, boolean_t large_block_ok, int outfd, uint64_t resumeobj, uint64_t resumeoff, #ifdef illumos vnode_t *vp, offset_t *off) #else struct file *fp, offset_t *off) #endif { dsl_pool_t *dp; dsl_dataset_t *ds; int err; boolean_t owned = B_FALSE; if (fromsnap != NULL && strpbrk(fromsnap, "@#") == NULL) return (SET_ERROR(EINVAL)); err = dsl_pool_hold(tosnap, FTAG, &dp); if (err != 0) return (err); if (strchr(tosnap, '@') == NULL && spa_writeable(dp->dp_spa)) { /* * We are sending a filesystem or volume. Ensure * that it doesn't change by owning the dataset. */ err = dsl_dataset_own(dp, tosnap, FTAG, &ds); owned = B_TRUE; } else { err = dsl_dataset_hold(dp, tosnap, FTAG, &ds); } if (err != 0) { dsl_pool_rele(dp, FTAG); return (err); } if (fromsnap != NULL) { zfs_bookmark_phys_t zb; boolean_t is_clone = B_FALSE; int fsnamelen = strchr(tosnap, '@') - tosnap; /* * If the fromsnap is in a different filesystem, then * mark the send stream as a clone. */ if (strncmp(tosnap, fromsnap, fsnamelen) != 0 || (fromsnap[fsnamelen] != '@' && fromsnap[fsnamelen] != '#')) { is_clone = B_TRUE; } if (strchr(fromsnap, '@')) { dsl_dataset_t *fromds; err = dsl_dataset_hold(dp, fromsnap, FTAG, &fromds); if (err == 0) { if (!dsl_dataset_is_before(ds, fromds, 0)) err = SET_ERROR(EXDEV); zb.zbm_creation_time = dsl_dataset_phys(fromds)->ds_creation_time; zb.zbm_creation_txg = dsl_dataset_phys(fromds)->ds_creation_txg; zb.zbm_guid = dsl_dataset_phys(fromds)->ds_guid; is_clone = (ds->ds_dir != fromds->ds_dir); dsl_dataset_rele(fromds, FTAG); } } else { err = dsl_bookmark_lookup(dp, fromsnap, ds, &zb); } if (err != 0) { dsl_dataset_rele(ds, FTAG); dsl_pool_rele(dp, FTAG); return (err); } err = dmu_send_impl(FTAG, dp, ds, &zb, is_clone, embedok, large_block_ok, outfd, resumeobj, resumeoff, fp, off); } else { err = dmu_send_impl(FTAG, dp, ds, NULL, B_FALSE, embedok, large_block_ok, outfd, resumeobj, resumeoff, fp, off); } if (owned) dsl_dataset_disown(ds, FTAG); else dsl_dataset_rele(ds, FTAG); return (err); } static int dmu_adjust_send_estimate_for_indirects(dsl_dataset_t *ds, uint64_t size, uint64_t *sizep) { int err; /* * Assume that space (both on-disk and in-stream) is dominated by * data. We will adjust for indirect blocks and the copies property, * but ignore per-object space used (eg, dnodes and DRR_OBJECT records). */ /* * Subtract out approximate space used by indirect blocks. * Assume most space is used by data blocks (non-indirect, non-dnode). * Assume all blocks are recordsize. Assume ditto blocks and * internal fragmentation counter out compression. * * Therefore, space used by indirect blocks is sizeof(blkptr_t) per * block, which we observe in practice. */ uint64_t recordsize; err = dsl_prop_get_int_ds(ds, "recordsize", &recordsize); if (err != 0) return (err); size -= size / recordsize * sizeof (blkptr_t); /* Add in the space for the record associated with each block. */ size += size / recordsize * sizeof (dmu_replay_record_t); *sizep = size; return (0); } int dmu_send_estimate(dsl_dataset_t *ds, dsl_dataset_t *fromds, uint64_t *sizep) { dsl_pool_t *dp = ds->ds_dir->dd_pool; int err; uint64_t size; ASSERT(dsl_pool_config_held(dp)); /* tosnap must be a snapshot */ if (!ds->ds_is_snapshot) return (SET_ERROR(EINVAL)); /* fromsnap, if provided, must be a snapshot */ if (fromds != NULL && !fromds->ds_is_snapshot) return (SET_ERROR(EINVAL)); /* * fromsnap must be an earlier snapshot from the same fs as tosnap, * or the origin's fs. */ if (fromds != NULL && !dsl_dataset_is_before(ds, fromds, 0)) return (SET_ERROR(EXDEV)); /* Get uncompressed size estimate of changed data. */ if (fromds == NULL) { size = dsl_dataset_phys(ds)->ds_uncompressed_bytes; } else { uint64_t used, comp; err = dsl_dataset_space_written(fromds, ds, &used, &comp, &size); if (err != 0) return (err); } err = dmu_adjust_send_estimate_for_indirects(ds, size, sizep); return (err); } /* * Simple callback used to traverse the blocks of a snapshot and sum their * uncompressed size */ /* ARGSUSED */ static int dmu_calculate_send_traversal(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, const zbookmark_phys_t *zb, const dnode_phys_t *dnp, void *arg) { uint64_t *spaceptr = arg; if (bp != NULL && !BP_IS_HOLE(bp)) { *spaceptr += BP_GET_UCSIZE(bp); } return (0); } /* * Given a desination snapshot and a TXG, calculate the approximate size of a * send stream sent from that TXG. from_txg may be zero, indicating that the * whole snapshot will be sent. */ int dmu_send_estimate_from_txg(dsl_dataset_t *ds, uint64_t from_txg, uint64_t *sizep) { dsl_pool_t *dp = ds->ds_dir->dd_pool; int err; uint64_t size = 0; ASSERT(dsl_pool_config_held(dp)); /* tosnap must be a snapshot */ if (!dsl_dataset_is_snapshot(ds)) return (SET_ERROR(EINVAL)); /* verify that from_txg is before the provided snapshot was taken */ if (from_txg >= dsl_dataset_phys(ds)->ds_creation_txg) { return (SET_ERROR(EXDEV)); } /* * traverse the blocks of the snapshot with birth times after * from_txg, summing their uncompressed size */ err = traverse_dataset(ds, from_txg, TRAVERSE_POST, dmu_calculate_send_traversal, &size); if (err) return (err); err = dmu_adjust_send_estimate_for_indirects(ds, size, sizep); return (err); } typedef struct dmu_recv_begin_arg { const char *drba_origin; dmu_recv_cookie_t *drba_cookie; cred_t *drba_cred; uint64_t drba_snapobj; } dmu_recv_begin_arg_t; static int recv_begin_check_existing_impl(dmu_recv_begin_arg_t *drba, dsl_dataset_t *ds, uint64_t fromguid) { uint64_t val; int error; dsl_pool_t *dp = ds->ds_dir->dd_pool; /* temporary clone name must not exist */ error = zap_lookup(dp->dp_meta_objset, dsl_dir_phys(ds->ds_dir)->dd_child_dir_zapobj, recv_clone_name, 8, 1, &val); if (error != ENOENT) return (error == 0 ? EBUSY : error); /* new snapshot name must not exist */ error = zap_lookup(dp->dp_meta_objset, dsl_dataset_phys(ds)->ds_snapnames_zapobj, drba->drba_cookie->drc_tosnap, 8, 1, &val); if (error != ENOENT) return (error == 0 ? EEXIST : error); /* * Check snapshot limit before receiving. We'll recheck again at the * end, but might as well abort before receiving if we're already over * the limit. * * Note that we do not check the file system limit with * dsl_dir_fscount_check because the temporary %clones don't count * against that limit. */ error = dsl_fs_ss_limit_check(ds->ds_dir, 1, ZFS_PROP_SNAPSHOT_LIMIT, NULL, drba->drba_cred); if (error != 0) return (error); if (fromguid != 0) { dsl_dataset_t *snap; uint64_t obj = dsl_dataset_phys(ds)->ds_prev_snap_obj; /* Find snapshot in this dir that matches fromguid. */ while (obj != 0) { error = dsl_dataset_hold_obj(dp, obj, FTAG, &snap); if (error != 0) return (SET_ERROR(ENODEV)); if (snap->ds_dir != ds->ds_dir) { dsl_dataset_rele(snap, FTAG); return (SET_ERROR(ENODEV)); } if (dsl_dataset_phys(snap)->ds_guid == fromguid) break; obj = dsl_dataset_phys(snap)->ds_prev_snap_obj; dsl_dataset_rele(snap, FTAG); } if (obj == 0) return (SET_ERROR(ENODEV)); if (drba->drba_cookie->drc_force) { drba->drba_snapobj = obj; } else { /* * If we are not forcing, there must be no * changes since fromsnap. */ if (dsl_dataset_modified_since_snap(ds, snap)) { dsl_dataset_rele(snap, FTAG); return (SET_ERROR(ETXTBSY)); } drba->drba_snapobj = ds->ds_prev->ds_object; } dsl_dataset_rele(snap, FTAG); } else { /* if full, then must be forced */ if (!drba->drba_cookie->drc_force) return (SET_ERROR(EEXIST)); /* start from $ORIGIN@$ORIGIN, if supported */ drba->drba_snapobj = dp->dp_origin_snap != NULL ? dp->dp_origin_snap->ds_object : 0; } return (0); } static int dmu_recv_begin_check(void *arg, dmu_tx_t *tx) { dmu_recv_begin_arg_t *drba = arg; dsl_pool_t *dp = dmu_tx_pool(tx); struct drr_begin *drrb = drba->drba_cookie->drc_drrb; uint64_t fromguid = drrb->drr_fromguid; int flags = drrb->drr_flags; int error; uint64_t featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo); dsl_dataset_t *ds; const char *tofs = drba->drba_cookie->drc_tofs; /* already checked */ ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC); ASSERT(!(featureflags & DMU_BACKUP_FEATURE_RESUMING)); if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) == DMU_COMPOUNDSTREAM || drrb->drr_type >= DMU_OST_NUMTYPES || ((flags & DRR_FLAG_CLONE) && drba->drba_origin == NULL)) return (SET_ERROR(EINVAL)); /* Verify pool version supports SA if SA_SPILL feature set */ if ((featureflags & DMU_BACKUP_FEATURE_SA_SPILL) && spa_version(dp->dp_spa) < SPA_VERSION_SA) return (SET_ERROR(ENOTSUP)); if (drba->drba_cookie->drc_resumable && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_EXTENSIBLE_DATASET)) return (SET_ERROR(ENOTSUP)); /* * The receiving code doesn't know how to translate a WRITE_EMBEDDED * record to a plan WRITE record, so the pool must have the * EMBEDDED_DATA feature enabled if the stream has WRITE_EMBEDDED * records. Same with WRITE_EMBEDDED records that use LZ4 compression. */ if ((featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_EMBEDDED_DATA)) return (SET_ERROR(ENOTSUP)); if ((featureflags & DMU_BACKUP_FEATURE_EMBED_DATA_LZ4) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LZ4_COMPRESS)) return (SET_ERROR(ENOTSUP)); /* * The receiving code doesn't know how to translate large blocks * to smaller ones, so the pool must have the LARGE_BLOCKS * feature enabled if the stream has LARGE_BLOCKS. */ if ((featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LARGE_BLOCKS)) return (SET_ERROR(ENOTSUP)); error = dsl_dataset_hold(dp, tofs, FTAG, &ds); if (error == 0) { /* target fs already exists; recv into temp clone */ /* Can't recv a clone into an existing fs */ if (flags & DRR_FLAG_CLONE) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EINVAL)); } error = recv_begin_check_existing_impl(drba, ds, fromguid); dsl_dataset_rele(ds, FTAG); } else if (error == ENOENT) { /* target fs does not exist; must be a full backup or clone */ char buf[MAXNAMELEN]; /* * If it's a non-clone incremental, we are missing the * target fs, so fail the recv. */ if (fromguid != 0 && !(flags & DRR_FLAG_CLONE || drba->drba_origin)) return (SET_ERROR(ENOENT)); /* Open the parent of tofs */ ASSERT3U(strlen(tofs), <, MAXNAMELEN); (void) strlcpy(buf, tofs, strrchr(tofs, '/') - tofs + 1); error = dsl_dataset_hold(dp, buf, FTAG, &ds); if (error != 0) return (error); /* * Check filesystem and snapshot limits before receiving. We'll * recheck snapshot limits again at the end (we create the * filesystems and increment those counts during begin_sync). */ error = dsl_fs_ss_limit_check(ds->ds_dir, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, drba->drba_cred); if (error != 0) { dsl_dataset_rele(ds, FTAG); return (error); } error = dsl_fs_ss_limit_check(ds->ds_dir, 1, ZFS_PROP_SNAPSHOT_LIMIT, NULL, drba->drba_cred); if (error != 0) { dsl_dataset_rele(ds, FTAG); return (error); } if (drba->drba_origin != NULL) { dsl_dataset_t *origin; error = dsl_dataset_hold(dp, drba->drba_origin, FTAG, &origin); if (error != 0) { dsl_dataset_rele(ds, FTAG); return (error); } if (!origin->ds_is_snapshot) { dsl_dataset_rele(origin, FTAG); dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EINVAL)); } if (dsl_dataset_phys(origin)->ds_guid != fromguid) { dsl_dataset_rele(origin, FTAG); dsl_dataset_rele(ds, FTAG); return (SET_ERROR(ENODEV)); } dsl_dataset_rele(origin, FTAG); } dsl_dataset_rele(ds, FTAG); error = 0; } return (error); } static void dmu_recv_begin_sync(void *arg, dmu_tx_t *tx) { dmu_recv_begin_arg_t *drba = arg; dsl_pool_t *dp = dmu_tx_pool(tx); objset_t *mos = dp->dp_meta_objset; struct drr_begin *drrb = drba->drba_cookie->drc_drrb; const char *tofs = drba->drba_cookie->drc_tofs; dsl_dataset_t *ds, *newds; uint64_t dsobj; int error; uint64_t crflags = 0; if (drrb->drr_flags & DRR_FLAG_CI_DATA) crflags |= DS_FLAG_CI_DATASET; error = dsl_dataset_hold(dp, tofs, FTAG, &ds); if (error == 0) { /* create temporary clone */ dsl_dataset_t *snap = NULL; if (drba->drba_snapobj != 0) { VERIFY0(dsl_dataset_hold_obj(dp, drba->drba_snapobj, FTAG, &snap)); } dsobj = dsl_dataset_create_sync(ds->ds_dir, recv_clone_name, snap, crflags, drba->drba_cred, tx); if (drba->drba_snapobj != 0) dsl_dataset_rele(snap, FTAG); dsl_dataset_rele(ds, FTAG); } else { dsl_dir_t *dd; const char *tail; dsl_dataset_t *origin = NULL; VERIFY0(dsl_dir_hold(dp, tofs, FTAG, &dd, &tail)); if (drba->drba_origin != NULL) { VERIFY0(dsl_dataset_hold(dp, drba->drba_origin, FTAG, &origin)); } /* Create new dataset. */ dsobj = dsl_dataset_create_sync(dd, strrchr(tofs, '/') + 1, origin, crflags, drba->drba_cred, tx); if (origin != NULL) dsl_dataset_rele(origin, FTAG); dsl_dir_rele(dd, FTAG); drba->drba_cookie->drc_newfs = B_TRUE; } VERIFY0(dsl_dataset_own_obj(dp, dsobj, dmu_recv_tag, &newds)); if (drba->drba_cookie->drc_resumable) { dsl_dataset_zapify(newds, tx); if (drrb->drr_fromguid != 0) { VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_FROMGUID, 8, 1, &drrb->drr_fromguid, tx)); } VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_TOGUID, 8, 1, &drrb->drr_toguid, tx)); VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_TONAME, 1, strlen(drrb->drr_toname) + 1, drrb->drr_toname, tx)); uint64_t one = 1; uint64_t zero = 0; VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_OBJECT, 8, 1, &one, tx)); VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_OFFSET, 8, 1, &zero, tx)); VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_BYTES, 8, 1, &zero, tx)); if (DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) & DMU_BACKUP_FEATURE_EMBED_DATA) { VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_EMBEDOK, 8, 1, &one, tx)); } } dmu_buf_will_dirty(newds->ds_dbuf, tx); dsl_dataset_phys(newds)->ds_flags |= DS_FLAG_INCONSISTENT; /* * If we actually created a non-clone, we need to create the * objset in our new dataset. */ if (BP_IS_HOLE(dsl_dataset_get_blkptr(newds))) { (void) dmu_objset_create_impl(dp->dp_spa, newds, dsl_dataset_get_blkptr(newds), drrb->drr_type, tx); } drba->drba_cookie->drc_ds = newds; spa_history_log_internal_ds(newds, "receive", tx, ""); } static int dmu_recv_resume_begin_check(void *arg, dmu_tx_t *tx) { dmu_recv_begin_arg_t *drba = arg; dsl_pool_t *dp = dmu_tx_pool(tx); struct drr_begin *drrb = drba->drba_cookie->drc_drrb; int error; uint64_t featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo); dsl_dataset_t *ds; const char *tofs = drba->drba_cookie->drc_tofs; /* already checked */ ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC); ASSERT(featureflags & DMU_BACKUP_FEATURE_RESUMING); if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) == DMU_COMPOUNDSTREAM || drrb->drr_type >= DMU_OST_NUMTYPES) return (SET_ERROR(EINVAL)); /* Verify pool version supports SA if SA_SPILL feature set */ if ((featureflags & DMU_BACKUP_FEATURE_SA_SPILL) && spa_version(dp->dp_spa) < SPA_VERSION_SA) return (SET_ERROR(ENOTSUP)); /* * The receiving code doesn't know how to translate a WRITE_EMBEDDED * record to a plain WRITE record, so the pool must have the * EMBEDDED_DATA feature enabled if the stream has WRITE_EMBEDDED * records. Same with WRITE_EMBEDDED records that use LZ4 compression. */ if ((featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_EMBEDDED_DATA)) return (SET_ERROR(ENOTSUP)); if ((featureflags & DMU_BACKUP_FEATURE_EMBED_DATA_LZ4) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LZ4_COMPRESS)) return (SET_ERROR(ENOTSUP)); char recvname[ZFS_MAXNAMELEN]; (void) snprintf(recvname, sizeof (recvname), "%s/%s", tofs, recv_clone_name); if (dsl_dataset_hold(dp, recvname, FTAG, &ds) != 0) { /* %recv does not exist; continue in tofs */ error = dsl_dataset_hold(dp, tofs, FTAG, &ds); if (error != 0) return (error); } /* check that ds is marked inconsistent */ if (!DS_IS_INCONSISTENT(ds)) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EINVAL)); } /* check that there is resuming data, and that the toguid matches */ if (!dsl_dataset_is_zapified(ds)) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EINVAL)); } uint64_t val; error = zap_lookup(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_TOGUID, sizeof (val), 1, &val); if (error != 0 || drrb->drr_toguid != val) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EINVAL)); } /* * Check if the receive is still running. If so, it will be owned. * Note that nothing else can own the dataset (e.g. after the receive * fails) because it will be marked inconsistent. */ if (dsl_dataset_has_owner(ds)) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EBUSY)); } /* There should not be any snapshots of this fs yet. */ if (ds->ds_prev != NULL && ds->ds_prev->ds_dir == ds->ds_dir) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EINVAL)); } /* * Note: resume point will be checked when we process the first WRITE * record. */ /* check that the origin matches */ val = 0; (void) zap_lookup(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_FROMGUID, sizeof (val), 1, &val); if (drrb->drr_fromguid != val) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EINVAL)); } dsl_dataset_rele(ds, FTAG); return (0); } static void dmu_recv_resume_begin_sync(void *arg, dmu_tx_t *tx) { dmu_recv_begin_arg_t *drba = arg; dsl_pool_t *dp = dmu_tx_pool(tx); const char *tofs = drba->drba_cookie->drc_tofs; dsl_dataset_t *ds; uint64_t dsobj; char recvname[ZFS_MAXNAMELEN]; (void) snprintf(recvname, sizeof (recvname), "%s/%s", tofs, recv_clone_name); if (dsl_dataset_hold(dp, recvname, FTAG, &ds) != 0) { /* %recv does not exist; continue in tofs */ VERIFY0(dsl_dataset_hold(dp, tofs, FTAG, &ds)); drba->drba_cookie->drc_newfs = B_TRUE; } /* clear the inconsistent flag so that we can own it */ ASSERT(DS_IS_INCONSISTENT(ds)); dmu_buf_will_dirty(ds->ds_dbuf, tx); dsl_dataset_phys(ds)->ds_flags &= ~DS_FLAG_INCONSISTENT; dsobj = ds->ds_object; dsl_dataset_rele(ds, FTAG); VERIFY0(dsl_dataset_own_obj(dp, dsobj, dmu_recv_tag, &ds)); dmu_buf_will_dirty(ds->ds_dbuf, tx); dsl_dataset_phys(ds)->ds_flags |= DS_FLAG_INCONSISTENT; ASSERT(!BP_IS_HOLE(dsl_dataset_get_blkptr(ds))); drba->drba_cookie->drc_ds = ds; spa_history_log_internal_ds(ds, "resume receive", tx, ""); } /* * NB: callers *MUST* call dmu_recv_stream() if dmu_recv_begin() * succeeds; otherwise we will leak the holds on the datasets. */ int dmu_recv_begin(char *tofs, char *tosnap, dmu_replay_record_t *drr_begin, boolean_t force, boolean_t resumable, char *origin, dmu_recv_cookie_t *drc) { dmu_recv_begin_arg_t drba = { 0 }; bzero(drc, sizeof (dmu_recv_cookie_t)); drc->drc_drr_begin = drr_begin; drc->drc_drrb = &drr_begin->drr_u.drr_begin; drc->drc_tosnap = tosnap; drc->drc_tofs = tofs; drc->drc_force = force; drc->drc_resumable = resumable; drc->drc_cred = CRED(); if (drc->drc_drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC)) { drc->drc_byteswap = B_TRUE; fletcher_4_incremental_byteswap(drr_begin, sizeof (dmu_replay_record_t), &drc->drc_cksum); byteswap_record(drr_begin); } else if (drc->drc_drrb->drr_magic == DMU_BACKUP_MAGIC) { fletcher_4_incremental_native(drr_begin, sizeof (dmu_replay_record_t), &drc->drc_cksum); } else { return (SET_ERROR(EINVAL)); } drba.drba_origin = origin; drba.drba_cookie = drc; drba.drba_cred = CRED(); if (DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo) & DMU_BACKUP_FEATURE_RESUMING) { return (dsl_sync_task(tofs, dmu_recv_resume_begin_check, dmu_recv_resume_begin_sync, &drba, 5, ZFS_SPACE_CHECK_NORMAL)); } else { return (dsl_sync_task(tofs, dmu_recv_begin_check, dmu_recv_begin_sync, &drba, 5, ZFS_SPACE_CHECK_NORMAL)); } } struct receive_record_arg { dmu_replay_record_t header; void *payload; /* Pointer to a buffer containing the payload */ /* * If the record is a write, pointer to the arc_buf_t containing the * payload. */ arc_buf_t *write_buf; int payload_size; uint64_t bytes_read; /* bytes read from stream when record created */ boolean_t eos_marker; /* Marks the end of the stream */ bqueue_node_t node; }; struct receive_writer_arg { objset_t *os; boolean_t byteswap; bqueue_t q; /* * These three args are used to signal to the main thread that we're * done. */ kmutex_t mutex; kcondvar_t cv; boolean_t done; int err; /* A map from guid to dataset to help handle dedup'd streams. */ avl_tree_t *guid_to_ds_map; boolean_t resumable; uint64_t last_object, last_offset; uint64_t bytes_read; /* bytes read when current record created */ }; struct receive_arg { objset_t *os; kthread_t *td; struct file *fp; uint64_t voff; /* The current offset in the stream */ uint64_t bytes_read; /* * A record that has had its payload read in, but hasn't yet been handed * off to the worker thread. */ struct receive_record_arg *rrd; /* A record that has had its header read in, but not its payload. */ struct receive_record_arg *next_rrd; zio_cksum_t cksum; zio_cksum_t prev_cksum; int err; boolean_t byteswap; /* Sorted list of objects not to issue prefetches for. */ list_t ignore_obj_list; }; struct receive_ign_obj_node { list_node_t node; uint64_t object; }; typedef struct guid_map_entry { uint64_t guid; dsl_dataset_t *gme_ds; avl_node_t avlnode; } guid_map_entry_t; static int guid_compare(const void *arg1, const void *arg2) { const guid_map_entry_t *gmep1 = arg1; const guid_map_entry_t *gmep2 = arg2; if (gmep1->guid < gmep2->guid) return (-1); else if (gmep1->guid > gmep2->guid) return (1); return (0); } static void free_guid_map_onexit(void *arg) { avl_tree_t *ca = arg; void *cookie = NULL; guid_map_entry_t *gmep; while ((gmep = avl_destroy_nodes(ca, &cookie)) != NULL) { dsl_dataset_long_rele(gmep->gme_ds, gmep); dsl_dataset_rele(gmep->gme_ds, gmep); kmem_free(gmep, sizeof (guid_map_entry_t)); } avl_destroy(ca); kmem_free(ca, sizeof (avl_tree_t)); } static int restore_bytes(struct receive_arg *ra, void *buf, int len, off_t off, ssize_t *resid) { struct uio auio; struct iovec aiov; int error; aiov.iov_base = buf; aiov.iov_len = len; auio.uio_iov = &aiov; auio.uio_iovcnt = 1; auio.uio_resid = len; auio.uio_segflg = UIO_SYSSPACE; auio.uio_rw = UIO_READ; auio.uio_offset = off; auio.uio_td = ra->td; #ifdef _KERNEL error = fo_read(ra->fp, &auio, ra->td->td_ucred, FOF_OFFSET, ra->td); #else fprintf(stderr, "%s: returning EOPNOTSUPP\n", __func__); error = EOPNOTSUPP; #endif *resid = auio.uio_resid; return (error); } static int receive_read(struct receive_arg *ra, int len, void *buf) { int done = 0; /* some things will require 8-byte alignment, so everything must */ ASSERT0(len % 8); while (done < len) { ssize_t resid; ra->err = restore_bytes(ra, buf + done, len - done, ra->voff, &resid); if (resid == len - done) { /* * Note: ECKSUM indicates that the receive * was interrupted and can potentially be resumed. */ ra->err = SET_ERROR(ECKSUM); } ra->voff += len - done - resid; done = len - resid; if (ra->err != 0) return (ra->err); } ra->bytes_read += len; ASSERT3U(done, ==, len); return (0); } static void byteswap_record(dmu_replay_record_t *drr) { #define DO64(X) (drr->drr_u.X = BSWAP_64(drr->drr_u.X)) #define DO32(X) (drr->drr_u.X = BSWAP_32(drr->drr_u.X)) drr->drr_type = BSWAP_32(drr->drr_type); drr->drr_payloadlen = BSWAP_32(drr->drr_payloadlen); switch (drr->drr_type) { case DRR_BEGIN: DO64(drr_begin.drr_magic); DO64(drr_begin.drr_versioninfo); DO64(drr_begin.drr_creation_time); DO32(drr_begin.drr_type); DO32(drr_begin.drr_flags); DO64(drr_begin.drr_toguid); DO64(drr_begin.drr_fromguid); break; case DRR_OBJECT: DO64(drr_object.drr_object); DO32(drr_object.drr_type); DO32(drr_object.drr_bonustype); DO32(drr_object.drr_blksz); DO32(drr_object.drr_bonuslen); DO64(drr_object.drr_toguid); break; case DRR_FREEOBJECTS: DO64(drr_freeobjects.drr_firstobj); DO64(drr_freeobjects.drr_numobjs); DO64(drr_freeobjects.drr_toguid); break; case DRR_WRITE: DO64(drr_write.drr_object); DO32(drr_write.drr_type); DO64(drr_write.drr_offset); DO64(drr_write.drr_length); DO64(drr_write.drr_toguid); ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_write.drr_key.ddk_cksum); DO64(drr_write.drr_key.ddk_prop); break; case DRR_WRITE_BYREF: DO64(drr_write_byref.drr_object); DO64(drr_write_byref.drr_offset); DO64(drr_write_byref.drr_length); DO64(drr_write_byref.drr_toguid); DO64(drr_write_byref.drr_refguid); DO64(drr_write_byref.drr_refobject); DO64(drr_write_byref.drr_refoffset); ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_write_byref. drr_key.ddk_cksum); DO64(drr_write_byref.drr_key.ddk_prop); break; case DRR_WRITE_EMBEDDED: DO64(drr_write_embedded.drr_object); DO64(drr_write_embedded.drr_offset); DO64(drr_write_embedded.drr_length); DO64(drr_write_embedded.drr_toguid); DO32(drr_write_embedded.drr_lsize); DO32(drr_write_embedded.drr_psize); break; case DRR_FREE: DO64(drr_free.drr_object); DO64(drr_free.drr_offset); DO64(drr_free.drr_length); DO64(drr_free.drr_toguid); break; case DRR_SPILL: DO64(drr_spill.drr_object); DO64(drr_spill.drr_length); DO64(drr_spill.drr_toguid); break; case DRR_END: DO64(drr_end.drr_toguid); ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_end.drr_checksum); break; } if (drr->drr_type != DRR_BEGIN) { ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_checksum.drr_checksum); } #undef DO64 #undef DO32 } static inline uint8_t deduce_nblkptr(dmu_object_type_t bonus_type, uint64_t bonus_size) { if (bonus_type == DMU_OT_SA) { return (1); } else { return (1 + ((DN_MAX_BONUSLEN - bonus_size) >> SPA_BLKPTRSHIFT)); } } static void save_resume_state(struct receive_writer_arg *rwa, uint64_t object, uint64_t offset, dmu_tx_t *tx) { int txgoff = dmu_tx_get_txg(tx) & TXG_MASK; if (!rwa->resumable) return; /* * We use ds_resume_bytes[] != 0 to indicate that we need to * update this on disk, so it must not be 0. */ ASSERT(rwa->bytes_read != 0); /* * We only resume from write records, which have a valid * (non-meta-dnode) object number. */ ASSERT(object != 0); /* * For resuming to work correctly, we must receive records in order, * sorted by object,offset. This is checked by the callers, but * assert it here for good measure. */ ASSERT3U(object, >=, rwa->os->os_dsl_dataset->ds_resume_object[txgoff]); ASSERT(object != rwa->os->os_dsl_dataset->ds_resume_object[txgoff] || offset >= rwa->os->os_dsl_dataset->ds_resume_offset[txgoff]); ASSERT3U(rwa->bytes_read, >=, rwa->os->os_dsl_dataset->ds_resume_bytes[txgoff]); rwa->os->os_dsl_dataset->ds_resume_object[txgoff] = object; rwa->os->os_dsl_dataset->ds_resume_offset[txgoff] = offset; rwa->os->os_dsl_dataset->ds_resume_bytes[txgoff] = rwa->bytes_read; } static int receive_object(struct receive_writer_arg *rwa, struct drr_object *drro, void *data) { dmu_object_info_t doi; dmu_tx_t *tx; uint64_t object; int err; if (drro->drr_type == DMU_OT_NONE || !DMU_OT_IS_VALID(drro->drr_type) || !DMU_OT_IS_VALID(drro->drr_bonustype) || drro->drr_checksumtype >= ZIO_CHECKSUM_FUNCTIONS || drro->drr_compress >= ZIO_COMPRESS_FUNCTIONS || P2PHASE(drro->drr_blksz, SPA_MINBLOCKSIZE) || drro->drr_blksz < SPA_MINBLOCKSIZE || drro->drr_blksz > spa_maxblocksize(dmu_objset_spa(rwa->os)) || drro->drr_bonuslen > DN_MAX_BONUSLEN) { return (SET_ERROR(EINVAL)); } err = dmu_object_info(rwa->os, drro->drr_object, &doi); if (err != 0 && err != ENOENT) return (SET_ERROR(EINVAL)); object = err == 0 ? drro->drr_object : DMU_NEW_OBJECT; /* * If we are losing blkptrs or changing the block size this must * be a new file instance. We must clear out the previous file * contents before we can change this type of metadata in the dnode. */ if (err == 0) { int nblkptr; nblkptr = deduce_nblkptr(drro->drr_bonustype, drro->drr_bonuslen); if (drro->drr_blksz != doi.doi_data_block_size || nblkptr < doi.doi_nblkptr) { err = dmu_free_long_range(rwa->os, drro->drr_object, 0, DMU_OBJECT_END); if (err != 0) return (SET_ERROR(EINVAL)); } } tx = dmu_tx_create(rwa->os); dmu_tx_hold_bonus(tx, object); err = dmu_tx_assign(tx, TXG_WAIT); if (err != 0) { dmu_tx_abort(tx); return (err); } if (object == DMU_NEW_OBJECT) { /* currently free, want to be allocated */ err = dmu_object_claim(rwa->os, drro->drr_object, drro->drr_type, drro->drr_blksz, drro->drr_bonustype, drro->drr_bonuslen, tx); } else if (drro->drr_type != doi.doi_type || drro->drr_blksz != doi.doi_data_block_size || drro->drr_bonustype != doi.doi_bonus_type || drro->drr_bonuslen != doi.doi_bonus_size) { /* currently allocated, but with different properties */ err = dmu_object_reclaim(rwa->os, drro->drr_object, drro->drr_type, drro->drr_blksz, drro->drr_bonustype, drro->drr_bonuslen, tx); } if (err != 0) { dmu_tx_commit(tx); return (SET_ERROR(EINVAL)); } dmu_object_set_checksum(rwa->os, drro->drr_object, drro->drr_checksumtype, tx); dmu_object_set_compress(rwa->os, drro->drr_object, drro->drr_compress, tx); if (data != NULL) { dmu_buf_t *db; VERIFY0(dmu_bonus_hold(rwa->os, drro->drr_object, FTAG, &db)); dmu_buf_will_dirty(db, tx); ASSERT3U(db->db_size, >=, drro->drr_bonuslen); bcopy(data, db->db_data, drro->drr_bonuslen); if (rwa->byteswap) { dmu_object_byteswap_t byteswap = DMU_OT_BYTESWAP(drro->drr_bonustype); dmu_ot_byteswap[byteswap].ob_func(db->db_data, drro->drr_bonuslen); } dmu_buf_rele(db, FTAG); } dmu_tx_commit(tx); return (0); } /* ARGSUSED */ static int receive_freeobjects(struct receive_writer_arg *rwa, struct drr_freeobjects *drrfo) { uint64_t obj; if (drrfo->drr_firstobj + drrfo->drr_numobjs < drrfo->drr_firstobj) return (SET_ERROR(EINVAL)); for (obj = drrfo->drr_firstobj; obj < drrfo->drr_firstobj + drrfo->drr_numobjs; (void) dmu_object_next(rwa->os, &obj, FALSE, 0)) { int err; if (dmu_object_info(rwa->os, obj, NULL) != 0) continue; err = dmu_free_long_object(rwa->os, obj); if (err != 0) return (err); } return (0); } static int receive_write(struct receive_writer_arg *rwa, struct drr_write *drrw, arc_buf_t *abuf) { dmu_tx_t *tx; int err; if (drrw->drr_offset + drrw->drr_length < drrw->drr_offset || !DMU_OT_IS_VALID(drrw->drr_type)) return (SET_ERROR(EINVAL)); /* * For resuming to work, records must be in increasing order * by (object, offset). */ if (drrw->drr_object < rwa->last_object || (drrw->drr_object == rwa->last_object && drrw->drr_offset < rwa->last_offset)) { return (SET_ERROR(EINVAL)); } rwa->last_object = drrw->drr_object; rwa->last_offset = drrw->drr_offset; if (dmu_object_info(rwa->os, drrw->drr_object, NULL) != 0) return (SET_ERROR(EINVAL)); tx = dmu_tx_create(rwa->os); dmu_tx_hold_write(tx, drrw->drr_object, drrw->drr_offset, drrw->drr_length); err = dmu_tx_assign(tx, TXG_WAIT); if (err != 0) { dmu_tx_abort(tx); return (err); } if (rwa->byteswap) { dmu_object_byteswap_t byteswap = DMU_OT_BYTESWAP(drrw->drr_type); dmu_ot_byteswap[byteswap].ob_func(abuf->b_data, drrw->drr_length); } dmu_buf_t *bonus; if (dmu_bonus_hold(rwa->os, drrw->drr_object, FTAG, &bonus) != 0) return (SET_ERROR(EINVAL)); dmu_assign_arcbuf(bonus, drrw->drr_offset, abuf, tx); /* * Note: If the receive fails, we want the resume stream to start * with the same record that we last successfully received (as opposed * to the next record), so that we can verify that we are * resuming from the correct location. */ save_resume_state(rwa, drrw->drr_object, drrw->drr_offset, tx); dmu_tx_commit(tx); dmu_buf_rele(bonus, FTAG); return (0); } /* * Handle a DRR_WRITE_BYREF record. This record is used in dedup'ed * streams to refer to a copy of the data that is already on the * system because it came in earlier in the stream. This function * finds the earlier copy of the data, and uses that copy instead of * data from the stream to fulfill this write. */ static int receive_write_byref(struct receive_writer_arg *rwa, struct drr_write_byref *drrwbr) { dmu_tx_t *tx; int err; guid_map_entry_t gmesrch; guid_map_entry_t *gmep; avl_index_t where; objset_t *ref_os = NULL; dmu_buf_t *dbp; if (drrwbr->drr_offset + drrwbr->drr_length < drrwbr->drr_offset) return (SET_ERROR(EINVAL)); /* * If the GUID of the referenced dataset is different from the * GUID of the target dataset, find the referenced dataset. */ if (drrwbr->drr_toguid != drrwbr->drr_refguid) { gmesrch.guid = drrwbr->drr_refguid; if ((gmep = avl_find(rwa->guid_to_ds_map, &gmesrch, &where)) == NULL) { return (SET_ERROR(EINVAL)); } if (dmu_objset_from_ds(gmep->gme_ds, &ref_os)) return (SET_ERROR(EINVAL)); } else { ref_os = rwa->os; } err = dmu_buf_hold(ref_os, drrwbr->drr_refobject, drrwbr->drr_refoffset, FTAG, &dbp, DMU_READ_PREFETCH); if (err != 0) return (err); tx = dmu_tx_create(rwa->os); dmu_tx_hold_write(tx, drrwbr->drr_object, drrwbr->drr_offset, drrwbr->drr_length); err = dmu_tx_assign(tx, TXG_WAIT); if (err != 0) { dmu_tx_abort(tx); return (err); } dmu_write(rwa->os, drrwbr->drr_object, drrwbr->drr_offset, drrwbr->drr_length, dbp->db_data, tx); dmu_buf_rele(dbp, FTAG); /* See comment in restore_write. */ save_resume_state(rwa, drrwbr->drr_object, drrwbr->drr_offset, tx); dmu_tx_commit(tx); return (0); } static int receive_write_embedded(struct receive_writer_arg *rwa, struct drr_write_embedded *drrwe, void *data) { dmu_tx_t *tx; int err; if (drrwe->drr_offset + drrwe->drr_length < drrwe->drr_offset) return (EINVAL); if (drrwe->drr_psize > BPE_PAYLOAD_SIZE) return (EINVAL); if (drrwe->drr_etype >= NUM_BP_EMBEDDED_TYPES) return (EINVAL); if (drrwe->drr_compression >= ZIO_COMPRESS_FUNCTIONS) return (EINVAL); tx = dmu_tx_create(rwa->os); dmu_tx_hold_write(tx, drrwe->drr_object, drrwe->drr_offset, drrwe->drr_length); err = dmu_tx_assign(tx, TXG_WAIT); if (err != 0) { dmu_tx_abort(tx); return (err); } dmu_write_embedded(rwa->os, drrwe->drr_object, drrwe->drr_offset, data, drrwe->drr_etype, drrwe->drr_compression, drrwe->drr_lsize, drrwe->drr_psize, rwa->byteswap ^ ZFS_HOST_BYTEORDER, tx); /* See comment in restore_write. */ save_resume_state(rwa, drrwe->drr_object, drrwe->drr_offset, tx); dmu_tx_commit(tx); return (0); } static int receive_spill(struct receive_writer_arg *rwa, struct drr_spill *drrs, void *data) { dmu_tx_t *tx; dmu_buf_t *db, *db_spill; int err; if (drrs->drr_length < SPA_MINBLOCKSIZE || drrs->drr_length > spa_maxblocksize(dmu_objset_spa(rwa->os))) return (SET_ERROR(EINVAL)); if (dmu_object_info(rwa->os, drrs->drr_object, NULL) != 0) return (SET_ERROR(EINVAL)); VERIFY0(dmu_bonus_hold(rwa->os, drrs->drr_object, FTAG, &db)); if ((err = dmu_spill_hold_by_bonus(db, FTAG, &db_spill)) != 0) { dmu_buf_rele(db, FTAG); return (err); } tx = dmu_tx_create(rwa->os); dmu_tx_hold_spill(tx, db->db_object); err = dmu_tx_assign(tx, TXG_WAIT); if (err != 0) { dmu_buf_rele(db, FTAG); dmu_buf_rele(db_spill, FTAG); dmu_tx_abort(tx); return (err); } dmu_buf_will_dirty(db_spill, tx); if (db_spill->db_size < drrs->drr_length) VERIFY(0 == dbuf_spill_set_blksz(db_spill, drrs->drr_length, tx)); bcopy(data, db_spill->db_data, drrs->drr_length); dmu_buf_rele(db, FTAG); dmu_buf_rele(db_spill, FTAG); dmu_tx_commit(tx); return (0); } /* ARGSUSED */ static int receive_free(struct receive_writer_arg *rwa, struct drr_free *drrf) { int err; if (drrf->drr_length != -1ULL && drrf->drr_offset + drrf->drr_length < drrf->drr_offset) return (SET_ERROR(EINVAL)); if (dmu_object_info(rwa->os, drrf->drr_object, NULL) != 0) return (SET_ERROR(EINVAL)); err = dmu_free_long_range(rwa->os, drrf->drr_object, drrf->drr_offset, drrf->drr_length); return (err); } /* used to destroy the drc_ds on error */ static void dmu_recv_cleanup_ds(dmu_recv_cookie_t *drc) { if (drc->drc_resumable) { /* wait for our resume state to be written to disk */ txg_wait_synced(drc->drc_ds->ds_dir->dd_pool, 0); dsl_dataset_disown(drc->drc_ds, dmu_recv_tag); } else { char name[MAXNAMELEN]; dsl_dataset_name(drc->drc_ds, name); dsl_dataset_disown(drc->drc_ds, dmu_recv_tag); (void) dsl_destroy_head(name); } } static void receive_cksum(struct receive_arg *ra, int len, void *buf) { if (ra->byteswap) { fletcher_4_incremental_byteswap(buf, len, &ra->cksum); } else { fletcher_4_incremental_native(buf, len, &ra->cksum); } } /* * Read the payload into a buffer of size len, and update the current record's * payload field. * Allocate ra->next_rrd and read the next record's header into * ra->next_rrd->header. * Verify checksum of payload and next record. */ static int receive_read_payload_and_next_header(struct receive_arg *ra, int len, void *buf) { int err; if (len != 0) { ASSERT3U(len, <=, SPA_MAXBLOCKSIZE); err = receive_read(ra, len, buf); if (err != 0) return (err); receive_cksum(ra, len, buf); /* note: rrd is NULL when reading the begin record's payload */ if (ra->rrd != NULL) { ra->rrd->payload = buf; ra->rrd->payload_size = len; ra->rrd->bytes_read = ra->bytes_read; } } ra->prev_cksum = ra->cksum; ra->next_rrd = kmem_zalloc(sizeof (*ra->next_rrd), KM_SLEEP); err = receive_read(ra, sizeof (ra->next_rrd->header), &ra->next_rrd->header); ra->next_rrd->bytes_read = ra->bytes_read; if (err != 0) { kmem_free(ra->next_rrd, sizeof (*ra->next_rrd)); ra->next_rrd = NULL; return (err); } if (ra->next_rrd->header.drr_type == DRR_BEGIN) { kmem_free(ra->next_rrd, sizeof (*ra->next_rrd)); ra->next_rrd = NULL; return (SET_ERROR(EINVAL)); } /* * Note: checksum is of everything up to but not including the * checksum itself. */ ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), ==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t)); receive_cksum(ra, offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), &ra->next_rrd->header); zio_cksum_t cksum_orig = ra->next_rrd->header.drr_u.drr_checksum.drr_checksum; zio_cksum_t *cksump = &ra->next_rrd->header.drr_u.drr_checksum.drr_checksum; if (ra->byteswap) byteswap_record(&ra->next_rrd->header); if ((!ZIO_CHECKSUM_IS_ZERO(cksump)) && !ZIO_CHECKSUM_EQUAL(ra->cksum, *cksump)) { kmem_free(ra->next_rrd, sizeof (*ra->next_rrd)); ra->next_rrd = NULL; return (SET_ERROR(ECKSUM)); } receive_cksum(ra, sizeof (cksum_orig), &cksum_orig); return (0); } /* * Issue the prefetch reads for any necessary indirect blocks. * * We use the object ignore list to tell us whether or not to issue prefetches * for a given object. We do this for both correctness (in case the blocksize * of an object has changed) and performance (if the object doesn't exist, don't * needlessly try to issue prefetches). We also trim the list as we go through * the stream to prevent it from growing to an unbounded size. * * The object numbers within will always be in sorted order, and any write * records we see will also be in sorted order, but they're not sorted with * respect to each other (i.e. we can get several object records before * receiving each object's write records). As a result, once we've reached a * given object number, we can safely remove any reference to lower object * numbers in the ignore list. In practice, we receive up to 32 object records * before receiving write records, so the list can have up to 32 nodes in it. */ /* ARGSUSED */ static void receive_read_prefetch(struct receive_arg *ra, uint64_t object, uint64_t offset, uint64_t length) { struct receive_ign_obj_node *node = list_head(&ra->ignore_obj_list); while (node != NULL && node->object < object) { VERIFY3P(node, ==, list_remove_head(&ra->ignore_obj_list)); kmem_free(node, sizeof (*node)); node = list_head(&ra->ignore_obj_list); } if (node == NULL || node->object > object) { dmu_prefetch(ra->os, object, 1, offset, length, ZIO_PRIORITY_SYNC_READ); } } /* * Read records off the stream, issuing any necessary prefetches. */ static int receive_read_record(struct receive_arg *ra) { int err; switch (ra->rrd->header.drr_type) { case DRR_OBJECT: { struct drr_object *drro = &ra->rrd->header.drr_u.drr_object; uint32_t size = P2ROUNDUP(drro->drr_bonuslen, 8); void *buf = kmem_zalloc(size, KM_SLEEP); dmu_object_info_t doi; err = receive_read_payload_and_next_header(ra, size, buf); if (err != 0) { kmem_free(buf, size); return (err); } err = dmu_object_info(ra->os, drro->drr_object, &doi); /* * See receive_read_prefetch for an explanation why we're * storing this object in the ignore_obj_list. */ if (err == ENOENT || (err == 0 && doi.doi_data_block_size != drro->drr_blksz)) { struct receive_ign_obj_node *node = kmem_zalloc(sizeof (*node), KM_SLEEP); node->object = drro->drr_object; #ifdef ZFS_DEBUG struct receive_ign_obj_node *last_object = list_tail(&ra->ignore_obj_list); uint64_t last_objnum = (last_object != NULL ? last_object->object : 0); ASSERT3U(node->object, >, last_objnum); #endif list_insert_tail(&ra->ignore_obj_list, node); err = 0; } return (err); } case DRR_FREEOBJECTS: { err = receive_read_payload_and_next_header(ra, 0, NULL); return (err); } case DRR_WRITE: { struct drr_write *drrw = &ra->rrd->header.drr_u.drr_write; arc_buf_t *abuf = arc_loan_buf(dmu_objset_spa(ra->os), drrw->drr_length); err = receive_read_payload_and_next_header(ra, drrw->drr_length, abuf->b_data); if (err != 0) { dmu_return_arcbuf(abuf); return (err); } ra->rrd->write_buf = abuf; receive_read_prefetch(ra, drrw->drr_object, drrw->drr_offset, drrw->drr_length); return (err); } case DRR_WRITE_BYREF: { struct drr_write_byref *drrwb = &ra->rrd->header.drr_u.drr_write_byref; err = receive_read_payload_and_next_header(ra, 0, NULL); receive_read_prefetch(ra, drrwb->drr_object, drrwb->drr_offset, drrwb->drr_length); return (err); } case DRR_WRITE_EMBEDDED: { struct drr_write_embedded *drrwe = &ra->rrd->header.drr_u.drr_write_embedded; uint32_t size = P2ROUNDUP(drrwe->drr_psize, 8); void *buf = kmem_zalloc(size, KM_SLEEP); err = receive_read_payload_and_next_header(ra, size, buf); if (err != 0) { kmem_free(buf, size); return (err); } receive_read_prefetch(ra, drrwe->drr_object, drrwe->drr_offset, drrwe->drr_length); return (err); } case DRR_FREE: { /* * It might be beneficial to prefetch indirect blocks here, but * we don't really have the data to decide for sure. */ err = receive_read_payload_and_next_header(ra, 0, NULL); return (err); } case DRR_END: { struct drr_end *drre = &ra->rrd->header.drr_u.drr_end; if (!ZIO_CHECKSUM_EQUAL(ra->prev_cksum, drre->drr_checksum)) return (SET_ERROR(ECKSUM)); return (0); } case DRR_SPILL: { struct drr_spill *drrs = &ra->rrd->header.drr_u.drr_spill; void *buf = kmem_zalloc(drrs->drr_length, KM_SLEEP); err = receive_read_payload_and_next_header(ra, drrs->drr_length, buf); if (err != 0) kmem_free(buf, drrs->drr_length); return (err); } default: return (SET_ERROR(EINVAL)); } } /* * Commit the records to the pool. */ static int receive_process_record(struct receive_writer_arg *rwa, struct receive_record_arg *rrd) { int err; /* Processing in order, therefore bytes_read should be increasing. */ ASSERT3U(rrd->bytes_read, >=, rwa->bytes_read); rwa->bytes_read = rrd->bytes_read; switch (rrd->header.drr_type) { case DRR_OBJECT: { struct drr_object *drro = &rrd->header.drr_u.drr_object; err = receive_object(rwa, drro, rrd->payload); kmem_free(rrd->payload, rrd->payload_size); rrd->payload = NULL; return (err); } case DRR_FREEOBJECTS: { struct drr_freeobjects *drrfo = &rrd->header.drr_u.drr_freeobjects; return (receive_freeobjects(rwa, drrfo)); } case DRR_WRITE: { struct drr_write *drrw = &rrd->header.drr_u.drr_write; err = receive_write(rwa, drrw, rrd->write_buf); /* if receive_write() is successful, it consumes the arc_buf */ if (err != 0) dmu_return_arcbuf(rrd->write_buf); rrd->write_buf = NULL; rrd->payload = NULL; return (err); } case DRR_WRITE_BYREF: { struct drr_write_byref *drrwbr = &rrd->header.drr_u.drr_write_byref; return (receive_write_byref(rwa, drrwbr)); } case DRR_WRITE_EMBEDDED: { struct drr_write_embedded *drrwe = &rrd->header.drr_u.drr_write_embedded; err = receive_write_embedded(rwa, drrwe, rrd->payload); kmem_free(rrd->payload, rrd->payload_size); rrd->payload = NULL; return (err); } case DRR_FREE: { struct drr_free *drrf = &rrd->header.drr_u.drr_free; return (receive_free(rwa, drrf)); } case DRR_SPILL: { struct drr_spill *drrs = &rrd->header.drr_u.drr_spill; err = receive_spill(rwa, drrs, rrd->payload); kmem_free(rrd->payload, rrd->payload_size); rrd->payload = NULL; return (err); } default: return (SET_ERROR(EINVAL)); } } /* * dmu_recv_stream's worker thread; pull records off the queue, and then call * receive_process_record When we're done, signal the main thread and exit. */ static void receive_writer_thread(void *arg) { struct receive_writer_arg *rwa = arg; struct receive_record_arg *rrd; for (rrd = bqueue_dequeue(&rwa->q); !rrd->eos_marker; rrd = bqueue_dequeue(&rwa->q)) { /* * If there's an error, the main thread will stop putting things * on the queue, but we need to clear everything in it before we * can exit. */ if (rwa->err == 0) { rwa->err = receive_process_record(rwa, rrd); } else if (rrd->write_buf != NULL) { dmu_return_arcbuf(rrd->write_buf); rrd->write_buf = NULL; rrd->payload = NULL; } else if (rrd->payload != NULL) { kmem_free(rrd->payload, rrd->payload_size); rrd->payload = NULL; } kmem_free(rrd, sizeof (*rrd)); } kmem_free(rrd, sizeof (*rrd)); mutex_enter(&rwa->mutex); rwa->done = B_TRUE; cv_signal(&rwa->cv); mutex_exit(&rwa->mutex); thread_exit(); } static int resume_check(struct receive_arg *ra, nvlist_t *begin_nvl) { uint64_t val; objset_t *mos = dmu_objset_pool(ra->os)->dp_meta_objset; uint64_t dsobj = dmu_objset_id(ra->os); uint64_t resume_obj, resume_off; if (nvlist_lookup_uint64(begin_nvl, "resume_object", &resume_obj) != 0 || nvlist_lookup_uint64(begin_nvl, "resume_offset", &resume_off) != 0) { return (SET_ERROR(EINVAL)); } VERIFY0(zap_lookup(mos, dsobj, DS_FIELD_RESUME_OBJECT, sizeof (val), 1, &val)); if (resume_obj != val) return (SET_ERROR(EINVAL)); VERIFY0(zap_lookup(mos, dsobj, DS_FIELD_RESUME_OFFSET, sizeof (val), 1, &val)); if (resume_off != val) return (SET_ERROR(EINVAL)); return (0); } /* * Read in the stream's records, one by one, and apply them to the pool. There * are two threads involved; the thread that calls this function will spin up a * worker thread, read the records off the stream one by one, and issue * prefetches for any necessary indirect blocks. It will then push the records * onto an internal blocking queue. The worker thread will pull the records off * the queue, and actually write the data into the DMU. This way, the worker * thread doesn't have to wait for reads to complete, since everything it needs * (the indirect blocks) will be prefetched. * * NB: callers *must* call dmu_recv_end() if this succeeds. */ int dmu_recv_stream(dmu_recv_cookie_t *drc, struct file *fp, offset_t *voffp, int cleanup_fd, uint64_t *action_handlep) { int err = 0; struct receive_arg ra = { 0 }; struct receive_writer_arg rwa = { 0 }; int featureflags; nvlist_t *begin_nvl = NULL; ra.byteswap = drc->drc_byteswap; ra.cksum = drc->drc_cksum; ra.td = curthread; ra.fp = fp; ra.voff = *voffp; if (dsl_dataset_is_zapified(drc->drc_ds)) { (void) zap_lookup(drc->drc_ds->ds_dir->dd_pool->dp_meta_objset, drc->drc_ds->ds_object, DS_FIELD_RESUME_BYTES, sizeof (ra.bytes_read), 1, &ra.bytes_read); } list_create(&ra.ignore_obj_list, sizeof (struct receive_ign_obj_node), offsetof(struct receive_ign_obj_node, node)); /* these were verified in dmu_recv_begin */ ASSERT3U(DMU_GET_STREAM_HDRTYPE(drc->drc_drrb->drr_versioninfo), ==, DMU_SUBSTREAM); ASSERT3U(drc->drc_drrb->drr_type, <, DMU_OST_NUMTYPES); /* * Open the objset we are modifying. */ VERIFY0(dmu_objset_from_ds(drc->drc_ds, &ra.os)); ASSERT(dsl_dataset_phys(drc->drc_ds)->ds_flags & DS_FLAG_INCONSISTENT); featureflags = DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo); /* if this stream is dedup'ed, set up the avl tree for guid mapping */ if (featureflags & DMU_BACKUP_FEATURE_DEDUP) { minor_t minor; if (cleanup_fd == -1) { ra.err = SET_ERROR(EBADF); goto out; } ra.err = zfs_onexit_fd_hold(cleanup_fd, &minor); if (ra.err != 0) { cleanup_fd = -1; goto out; } if (*action_handlep == 0) { rwa.guid_to_ds_map = kmem_alloc(sizeof (avl_tree_t), KM_SLEEP); avl_create(rwa.guid_to_ds_map, guid_compare, sizeof (guid_map_entry_t), offsetof(guid_map_entry_t, avlnode)); err = zfs_onexit_add_cb(minor, free_guid_map_onexit, rwa.guid_to_ds_map, action_handlep); if (ra.err != 0) goto out; } else { err = zfs_onexit_cb_data(minor, *action_handlep, (void **)&rwa.guid_to_ds_map); if (ra.err != 0) goto out; } drc->drc_guid_to_ds_map = rwa.guid_to_ds_map; } uint32_t payloadlen = drc->drc_drr_begin->drr_payloadlen; void *payload = NULL; if (payloadlen != 0) payload = kmem_alloc(payloadlen, KM_SLEEP); err = receive_read_payload_and_next_header(&ra, payloadlen, payload); if (err != 0) { if (payloadlen != 0) kmem_free(payload, payloadlen); goto out; } if (payloadlen != 0) { err = nvlist_unpack(payload, payloadlen, &begin_nvl, KM_SLEEP); kmem_free(payload, payloadlen); if (err != 0) goto out; } if (featureflags & DMU_BACKUP_FEATURE_RESUMING) { err = resume_check(&ra, begin_nvl); if (err != 0) goto out; } (void) bqueue_init(&rwa.q, zfs_recv_queue_length, offsetof(struct receive_record_arg, node)); cv_init(&rwa.cv, NULL, CV_DEFAULT, NULL); mutex_init(&rwa.mutex, NULL, MUTEX_DEFAULT, NULL); rwa.os = ra.os; rwa.byteswap = drc->drc_byteswap; rwa.resumable = drc->drc_resumable; (void) thread_create(NULL, 0, receive_writer_thread, &rwa, 0, &p0, TS_RUN, minclsyspri); /* * We're reading rwa.err without locks, which is safe since we are the * only reader, and the worker thread is the only writer. It's ok if we * miss a write for an iteration or two of the loop, since the writer * thread will keep freeing records we send it until we send it an eos * marker. * * We can leave this loop in 3 ways: First, if rwa.err is * non-zero. In that case, the writer thread will free the rrd we just * pushed. Second, if we're interrupted; in that case, either it's the * first loop and ra.rrd was never allocated, or it's later, and ra.rrd * has been handed off to the writer thread who will free it. Finally, * if receive_read_record fails or we're at the end of the stream, then * we free ra.rrd and exit. */ while (rwa.err == 0) { if (issig(JUSTLOOKING) && issig(FORREAL)) { err = SET_ERROR(EINTR); break; } ASSERT3P(ra.rrd, ==, NULL); ra.rrd = ra.next_rrd; ra.next_rrd = NULL; /* Allocates and loads header into ra.next_rrd */ err = receive_read_record(&ra); if (ra.rrd->header.drr_type == DRR_END || err != 0) { kmem_free(ra.rrd, sizeof (*ra.rrd)); ra.rrd = NULL; break; } bqueue_enqueue(&rwa.q, ra.rrd, sizeof (struct receive_record_arg) + ra.rrd->payload_size); ra.rrd = NULL; } if (ra.next_rrd == NULL) ra.next_rrd = kmem_zalloc(sizeof (*ra.next_rrd), KM_SLEEP); ra.next_rrd->eos_marker = B_TRUE; bqueue_enqueue(&rwa.q, ra.next_rrd, 1); mutex_enter(&rwa.mutex); while (!rwa.done) { cv_wait(&rwa.cv, &rwa.mutex); } mutex_exit(&rwa.mutex); cv_destroy(&rwa.cv); mutex_destroy(&rwa.mutex); bqueue_destroy(&rwa.q); if (err == 0) err = rwa.err; out: nvlist_free(begin_nvl); if ((featureflags & DMU_BACKUP_FEATURE_DEDUP) && (cleanup_fd != -1)) zfs_onexit_fd_rele(cleanup_fd); if (err != 0) { /* * Clean up references. If receive is not resumable, * destroy what we created, so we don't leave it in * the inconsistent state. */ dmu_recv_cleanup_ds(drc); } *voffp = ra.voff; for (struct receive_ign_obj_node *n = list_remove_head(&ra.ignore_obj_list); n != NULL; n = list_remove_head(&ra.ignore_obj_list)) { kmem_free(n, sizeof (*n)); } list_destroy(&ra.ignore_obj_list); return (err); } static int dmu_recv_end_check(void *arg, dmu_tx_t *tx) { dmu_recv_cookie_t *drc = arg; dsl_pool_t *dp = dmu_tx_pool(tx); int error; ASSERT3P(drc->drc_ds->ds_owner, ==, dmu_recv_tag); if (!drc->drc_newfs) { dsl_dataset_t *origin_head; error = dsl_dataset_hold(dp, drc->drc_tofs, FTAG, &origin_head); if (error != 0) return (error); if (drc->drc_force) { /* * We will destroy any snapshots in tofs (i.e. before * origin_head) that are after the origin (which is * the snap before drc_ds, because drc_ds can not * have any snaps of its own). */ uint64_t obj; obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj; while (obj != dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) { dsl_dataset_t *snap; error = dsl_dataset_hold_obj(dp, obj, FTAG, &snap); if (error != 0) break; if (snap->ds_dir != origin_head->ds_dir) error = SET_ERROR(EINVAL); if (error == 0) { error = dsl_destroy_snapshot_check_impl( snap, B_FALSE); } obj = dsl_dataset_phys(snap)->ds_prev_snap_obj; dsl_dataset_rele(snap, FTAG); if (error != 0) break; } if (error != 0) { dsl_dataset_rele(origin_head, FTAG); return (error); } } error = dsl_dataset_clone_swap_check_impl(drc->drc_ds, origin_head, drc->drc_force, drc->drc_owner, tx); if (error != 0) { dsl_dataset_rele(origin_head, FTAG); return (error); } error = dsl_dataset_snapshot_check_impl(origin_head, drc->drc_tosnap, tx, B_TRUE, 1, drc->drc_cred); dsl_dataset_rele(origin_head, FTAG); if (error != 0) return (error); error = dsl_destroy_head_check_impl(drc->drc_ds, 1); } else { error = dsl_dataset_snapshot_check_impl(drc->drc_ds, drc->drc_tosnap, tx, B_TRUE, 1, drc->drc_cred); } return (error); } static void dmu_recv_end_sync(void *arg, dmu_tx_t *tx) { dmu_recv_cookie_t *drc = arg; dsl_pool_t *dp = dmu_tx_pool(tx); spa_history_log_internal_ds(drc->drc_ds, "finish receiving", tx, "snap=%s", drc->drc_tosnap); if (!drc->drc_newfs) { dsl_dataset_t *origin_head; VERIFY0(dsl_dataset_hold(dp, drc->drc_tofs, FTAG, &origin_head)); if (drc->drc_force) { /* * Destroy any snapshots of drc_tofs (origin_head) * after the origin (the snap before drc_ds). */ uint64_t obj; obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj; while (obj != dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) { dsl_dataset_t *snap; VERIFY0(dsl_dataset_hold_obj(dp, obj, FTAG, &snap)); ASSERT3P(snap->ds_dir, ==, origin_head->ds_dir); obj = dsl_dataset_phys(snap)->ds_prev_snap_obj; dsl_destroy_snapshot_sync_impl(snap, B_FALSE, tx); dsl_dataset_rele(snap, FTAG); } } VERIFY3P(drc->drc_ds->ds_prev, ==, origin_head->ds_prev); dsl_dataset_clone_swap_sync_impl(drc->drc_ds, origin_head, tx); dsl_dataset_snapshot_sync_impl(origin_head, drc->drc_tosnap, tx); /* set snapshot's creation time and guid */ dmu_buf_will_dirty(origin_head->ds_prev->ds_dbuf, tx); dsl_dataset_phys(origin_head->ds_prev)->ds_creation_time = drc->drc_drrb->drr_creation_time; dsl_dataset_phys(origin_head->ds_prev)->ds_guid = drc->drc_drrb->drr_toguid; dsl_dataset_phys(origin_head->ds_prev)->ds_flags &= ~DS_FLAG_INCONSISTENT; dmu_buf_will_dirty(origin_head->ds_dbuf, tx); dsl_dataset_phys(origin_head)->ds_flags &= ~DS_FLAG_INCONSISTENT; dsl_dataset_rele(origin_head, FTAG); dsl_destroy_head_sync_impl(drc->drc_ds, tx); if (drc->drc_owner != NULL) VERIFY3P(origin_head->ds_owner, ==, drc->drc_owner); } else { dsl_dataset_t *ds = drc->drc_ds; dsl_dataset_snapshot_sync_impl(ds, drc->drc_tosnap, tx); /* set snapshot's creation time and guid */ dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx); dsl_dataset_phys(ds->ds_prev)->ds_creation_time = drc->drc_drrb->drr_creation_time; dsl_dataset_phys(ds->ds_prev)->ds_guid = drc->drc_drrb->drr_toguid; dsl_dataset_phys(ds->ds_prev)->ds_flags &= ~DS_FLAG_INCONSISTENT; dmu_buf_will_dirty(ds->ds_dbuf, tx); dsl_dataset_phys(ds)->ds_flags &= ~DS_FLAG_INCONSISTENT; if (dsl_dataset_has_resume_receive_state(ds)) { (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_FROMGUID, tx); (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_OBJECT, tx); (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_OFFSET, tx); (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_BYTES, tx); (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_TOGUID, tx); (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_TONAME, tx); } } drc->drc_newsnapobj = dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj; /* * Release the hold from dmu_recv_begin. This must be done before * we return to open context, so that when we free the dataset's dnode, * we can evict its bonus buffer. */ dsl_dataset_disown(drc->drc_ds, dmu_recv_tag); drc->drc_ds = NULL; } static int add_ds_to_guidmap(const char *name, avl_tree_t *guid_map, uint64_t snapobj) { dsl_pool_t *dp; dsl_dataset_t *snapds; guid_map_entry_t *gmep; int err; ASSERT(guid_map != NULL); err = dsl_pool_hold(name, FTAG, &dp); if (err != 0) return (err); gmep = kmem_alloc(sizeof (*gmep), KM_SLEEP); err = dsl_dataset_hold_obj(dp, snapobj, gmep, &snapds); if (err == 0) { gmep->guid = dsl_dataset_phys(snapds)->ds_guid; gmep->gme_ds = snapds; avl_add(guid_map, gmep); dsl_dataset_long_hold(snapds, gmep); } else kmem_free(gmep, sizeof (*gmep)); dsl_pool_rele(dp, FTAG); return (err); } static int dmu_recv_end_modified_blocks = 3; static int dmu_recv_existing_end(dmu_recv_cookie_t *drc) { int error; char name[MAXNAMELEN]; #ifdef _KERNEL /* * We will be destroying the ds; make sure its origin is unmounted if * necessary. */ dsl_dataset_name(drc->drc_ds, name); zfs_destroy_unmount_origin(name); #endif error = dsl_sync_task(drc->drc_tofs, dmu_recv_end_check, dmu_recv_end_sync, drc, dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL); if (error != 0) dmu_recv_cleanup_ds(drc); return (error); } static int dmu_recv_new_end(dmu_recv_cookie_t *drc) { int error; error = dsl_sync_task(drc->drc_tofs, dmu_recv_end_check, dmu_recv_end_sync, drc, dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL); if (error != 0) { dmu_recv_cleanup_ds(drc); } else if (drc->drc_guid_to_ds_map != NULL) { (void) add_ds_to_guidmap(drc->drc_tofs, drc->drc_guid_to_ds_map, drc->drc_newsnapobj); } return (error); } int dmu_recv_end(dmu_recv_cookie_t *drc, void *owner) { drc->drc_owner = owner; if (drc->drc_newfs) return (dmu_recv_new_end(drc)); else return (dmu_recv_existing_end(drc)); } /* * Return TRUE if this objset is currently being received into. */ boolean_t dmu_objset_is_receiving(objset_t *os) { return (os->os_dsl_dataset != NULL && os->os_dsl_dataset->ds_owner == dmu_recv_tag); } Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_dataset.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_dataset.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_dataset.c (revision 290757) @@ -1,3608 +1,3616 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Portions Copyright (c) 2011 Martin Matuska * Copyright (c) 2011, 2015 by Delphix. All rights reserved. * Copyright (c) 2014, Joyent, Inc. All rights reserved. * Copyright (c) 2014 RackTop Systems. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include +#include #include #include SYSCTL_DECL(_vfs_zfs); /* * The SPA supports block sizes up to 16MB. However, very large blocks * can have an impact on i/o latency (e.g. tying up a spinning disk for * ~300ms), and also potentially on the memory allocator. Therefore, * we do not allow the recordsize to be set larger than zfs_max_recordsize * (default 1MB). Larger blocks can be created by changing this tunable, * and pools with larger blocks can always be imported and used, regardless * of this setting. */ int zfs_max_recordsize = 1 * 1024 * 1024; SYSCTL_INT(_vfs_zfs, OID_AUTO, max_recordsize, CTLFLAG_RWTUN, &zfs_max_recordsize, 0, "Maximum block size. Expect dragons when tuning this."); #define SWITCH64(x, y) \ { \ uint64_t __tmp = (x); \ (x) = (y); \ (y) = __tmp; \ } #define DS_REF_MAX (1ULL << 62) extern inline dsl_dataset_phys_t *dsl_dataset_phys(dsl_dataset_t *ds); /* * Figure out how much of this delta should be propogated to the dsl_dir * layer. If there's a refreservation, that space has already been * partially accounted for in our ancestors. */ static int64_t parent_delta(dsl_dataset_t *ds, int64_t delta) { dsl_dataset_phys_t *ds_phys; uint64_t old_bytes, new_bytes; if (ds->ds_reserved == 0) return (delta); ds_phys = dsl_dataset_phys(ds); old_bytes = MAX(ds_phys->ds_unique_bytes, ds->ds_reserved); new_bytes = MAX(ds_phys->ds_unique_bytes + delta, ds->ds_reserved); ASSERT3U(ABS((int64_t)(new_bytes - old_bytes)), <=, ABS(delta)); return (new_bytes - old_bytes); } void dsl_dataset_block_born(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx) { int used = bp_get_dsize_sync(tx->tx_pool->dp_spa, bp); int compressed = BP_GET_PSIZE(bp); int uncompressed = BP_GET_UCSIZE(bp); int64_t delta; dprintf_bp(bp, "ds=%p", ds); ASSERT(dmu_tx_is_syncing(tx)); /* It could have been compressed away to nothing */ if (BP_IS_HOLE(bp)) return; ASSERT(BP_GET_TYPE(bp) != DMU_OT_NONE); ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp))); if (ds == NULL) { dsl_pool_mos_diduse_space(tx->tx_pool, used, compressed, uncompressed); return; } dmu_buf_will_dirty(ds->ds_dbuf, tx); mutex_enter(&ds->ds_lock); delta = parent_delta(ds, used); dsl_dataset_phys(ds)->ds_referenced_bytes += used; dsl_dataset_phys(ds)->ds_compressed_bytes += compressed; dsl_dataset_phys(ds)->ds_uncompressed_bytes += uncompressed; dsl_dataset_phys(ds)->ds_unique_bytes += used; + if (BP_GET_LSIZE(bp) > SPA_OLD_MAXBLOCKSIZE) { ds->ds_feature_activation_needed[SPA_FEATURE_LARGE_BLOCKS] = B_TRUE; } + + spa_feature_t f = zio_checksum_to_feature(BP_GET_CHECKSUM(bp)); + if (f != SPA_FEATURE_NONE) + ds->ds_feature_activation_needed[f] = B_TRUE; + mutex_exit(&ds->ds_lock); dsl_dir_diduse_space(ds->ds_dir, DD_USED_HEAD, delta, compressed, uncompressed, tx); dsl_dir_transfer_space(ds->ds_dir, used - delta, DD_USED_REFRSRV, DD_USED_HEAD, NULL); } int dsl_dataset_block_kill(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx, boolean_t async) { int used = bp_get_dsize_sync(tx->tx_pool->dp_spa, bp); int compressed = BP_GET_PSIZE(bp); int uncompressed = BP_GET_UCSIZE(bp); if (BP_IS_HOLE(bp)) return (0); ASSERT(dmu_tx_is_syncing(tx)); ASSERT(bp->blk_birth <= tx->tx_txg); if (ds == NULL) { dsl_free(tx->tx_pool, tx->tx_txg, bp); dsl_pool_mos_diduse_space(tx->tx_pool, -used, -compressed, -uncompressed); return (used); } ASSERT3P(tx->tx_pool, ==, ds->ds_dir->dd_pool); ASSERT(!ds->ds_is_snapshot); dmu_buf_will_dirty(ds->ds_dbuf, tx); if (bp->blk_birth > dsl_dataset_phys(ds)->ds_prev_snap_txg) { int64_t delta; dprintf_bp(bp, "freeing ds=%llu", ds->ds_object); dsl_free(tx->tx_pool, tx->tx_txg, bp); mutex_enter(&ds->ds_lock); ASSERT(dsl_dataset_phys(ds)->ds_unique_bytes >= used || !DS_UNIQUE_IS_ACCURATE(ds)); delta = parent_delta(ds, -used); dsl_dataset_phys(ds)->ds_unique_bytes -= used; mutex_exit(&ds->ds_lock); dsl_dir_diduse_space(ds->ds_dir, DD_USED_HEAD, delta, -compressed, -uncompressed, tx); dsl_dir_transfer_space(ds->ds_dir, -used - delta, DD_USED_REFRSRV, DD_USED_HEAD, NULL); } else { dprintf_bp(bp, "putting on dead list: %s", ""); if (async) { /* * We are here as part of zio's write done callback, * which means we're a zio interrupt thread. We can't * call dsl_deadlist_insert() now because it may block * waiting for I/O. Instead, put bp on the deferred * queue and let dsl_pool_sync() finish the job. */ bplist_append(&ds->ds_pending_deadlist, bp); } else { dsl_deadlist_insert(&ds->ds_deadlist, bp, tx); } ASSERT3U(ds->ds_prev->ds_object, ==, dsl_dataset_phys(ds)->ds_prev_snap_obj); ASSERT(dsl_dataset_phys(ds->ds_prev)->ds_num_children > 0); /* if (bp->blk_birth > prev prev snap txg) prev unique += bs */ if (dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj == ds->ds_object && bp->blk_birth > dsl_dataset_phys(ds->ds_prev)->ds_prev_snap_txg) { dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx); mutex_enter(&ds->ds_prev->ds_lock); dsl_dataset_phys(ds->ds_prev)->ds_unique_bytes += used; mutex_exit(&ds->ds_prev->ds_lock); } if (bp->blk_birth > ds->ds_dir->dd_origin_txg) { dsl_dir_transfer_space(ds->ds_dir, used, DD_USED_HEAD, DD_USED_SNAP, tx); } } mutex_enter(&ds->ds_lock); ASSERT3U(dsl_dataset_phys(ds)->ds_referenced_bytes, >=, used); dsl_dataset_phys(ds)->ds_referenced_bytes -= used; ASSERT3U(dsl_dataset_phys(ds)->ds_compressed_bytes, >=, compressed); dsl_dataset_phys(ds)->ds_compressed_bytes -= compressed; ASSERT3U(dsl_dataset_phys(ds)->ds_uncompressed_bytes, >=, uncompressed); dsl_dataset_phys(ds)->ds_uncompressed_bytes -= uncompressed; mutex_exit(&ds->ds_lock); return (used); } uint64_t dsl_dataset_prev_snap_txg(dsl_dataset_t *ds) { uint64_t trysnap = 0; if (ds == NULL) return (0); /* * The snapshot creation could fail, but that would cause an * incorrect FALSE return, which would only result in an * overestimation of the amount of space that an operation would * consume, which is OK. * * There's also a small window where we could miss a pending * snapshot, because we could set the sync task in the quiescing * phase. So this should only be used as a guess. */ if (ds->ds_trysnap_txg > spa_last_synced_txg(ds->ds_dir->dd_pool->dp_spa)) trysnap = ds->ds_trysnap_txg; return (MAX(dsl_dataset_phys(ds)->ds_prev_snap_txg, trysnap)); } boolean_t dsl_dataset_block_freeable(dsl_dataset_t *ds, const blkptr_t *bp, uint64_t blk_birth) { if (blk_birth <= dsl_dataset_prev_snap_txg(ds) || (bp != NULL && BP_IS_HOLE(bp))) return (B_FALSE); ddt_prefetch(dsl_dataset_get_spa(ds), bp); return (B_TRUE); } static void dsl_dataset_evict(void *dbu) { dsl_dataset_t *ds = dbu; ASSERT(ds->ds_owner == NULL); ds->ds_dbuf = NULL; unique_remove(ds->ds_fsid_guid); if (ds->ds_objset != NULL) dmu_objset_evict(ds->ds_objset); if (ds->ds_prev) { dsl_dataset_rele(ds->ds_prev, ds); ds->ds_prev = NULL; } bplist_destroy(&ds->ds_pending_deadlist); if (ds->ds_deadlist.dl_os != NULL) dsl_deadlist_close(&ds->ds_deadlist); if (ds->ds_dir) dsl_dir_async_rele(ds->ds_dir, ds); ASSERT(!list_link_active(&ds->ds_synced_link)); list_destroy(&ds->ds_prop_cbs); if (mutex_owned(&ds->ds_lock)) mutex_exit(&ds->ds_lock); mutex_destroy(&ds->ds_lock); if (mutex_owned(&ds->ds_opening_lock)) mutex_exit(&ds->ds_opening_lock); mutex_destroy(&ds->ds_opening_lock); mutex_destroy(&ds->ds_sendstream_lock); refcount_destroy(&ds->ds_longholds); kmem_free(ds, sizeof (dsl_dataset_t)); } int dsl_dataset_get_snapname(dsl_dataset_t *ds) { dsl_dataset_phys_t *headphys; int err; dmu_buf_t *headdbuf; dsl_pool_t *dp = ds->ds_dir->dd_pool; objset_t *mos = dp->dp_meta_objset; if (ds->ds_snapname[0]) return (0); if (dsl_dataset_phys(ds)->ds_next_snap_obj == 0) return (0); err = dmu_bonus_hold(mos, dsl_dir_phys(ds->ds_dir)->dd_head_dataset_obj, FTAG, &headdbuf); if (err != 0) return (err); headphys = headdbuf->db_data; err = zap_value_search(dp->dp_meta_objset, headphys->ds_snapnames_zapobj, ds->ds_object, 0, ds->ds_snapname); dmu_buf_rele(headdbuf, FTAG); return (err); } int dsl_dataset_snap_lookup(dsl_dataset_t *ds, const char *name, uint64_t *value) { objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; uint64_t snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; matchtype_t mt; int err; if (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_CI_DATASET) mt = MT_FIRST; else mt = MT_EXACT; err = zap_lookup_norm(mos, snapobj, name, 8, 1, value, mt, NULL, 0, NULL); if (err == ENOTSUP && mt == MT_FIRST) err = zap_lookup(mos, snapobj, name, 8, 1, value); return (err); } int dsl_dataset_snap_remove(dsl_dataset_t *ds, const char *name, dmu_tx_t *tx, boolean_t adj_cnt) { objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; uint64_t snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; matchtype_t mt; int err; dsl_dir_snap_cmtime_update(ds->ds_dir); if (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_CI_DATASET) mt = MT_FIRST; else mt = MT_EXACT; err = zap_remove_norm(mos, snapobj, name, mt, tx); if (err == ENOTSUP && mt == MT_FIRST) err = zap_remove(mos, snapobj, name, tx); if (err == 0 && adj_cnt) dsl_fs_ss_count_adjust(ds->ds_dir, -1, DD_FIELD_SNAPSHOT_COUNT, tx); return (err); } boolean_t dsl_dataset_try_add_ref(dsl_pool_t *dp, dsl_dataset_t *ds, void *tag) { dmu_buf_t *dbuf = ds->ds_dbuf; boolean_t result = B_FALSE; if (dbuf != NULL && dmu_buf_try_add_ref(dbuf, dp->dp_meta_objset, ds->ds_object, DMU_BONUS_BLKID, tag)) { if (ds == dmu_buf_get_user(dbuf)) result = B_TRUE; else dmu_buf_rele(dbuf, tag); } return (result); } int dsl_dataset_hold_obj(dsl_pool_t *dp, uint64_t dsobj, void *tag, dsl_dataset_t **dsp) { objset_t *mos = dp->dp_meta_objset; dmu_buf_t *dbuf; dsl_dataset_t *ds; int err; dmu_object_info_t doi; ASSERT(dsl_pool_config_held(dp)); err = dmu_bonus_hold(mos, dsobj, tag, &dbuf); if (err != 0) return (err); /* Make sure dsobj has the correct object type. */ dmu_object_info_from_db(dbuf, &doi); if (doi.doi_bonus_type != DMU_OT_DSL_DATASET) { dmu_buf_rele(dbuf, tag); return (SET_ERROR(EINVAL)); } ds = dmu_buf_get_user(dbuf); if (ds == NULL) { dsl_dataset_t *winner = NULL; ds = kmem_zalloc(sizeof (dsl_dataset_t), KM_SLEEP); ds->ds_dbuf = dbuf; ds->ds_object = dsobj; ds->ds_is_snapshot = dsl_dataset_phys(ds)->ds_num_children != 0; mutex_init(&ds->ds_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&ds->ds_opening_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&ds->ds_sendstream_lock, NULL, MUTEX_DEFAULT, NULL); refcount_create(&ds->ds_longholds); bplist_create(&ds->ds_pending_deadlist); dsl_deadlist_open(&ds->ds_deadlist, mos, dsl_dataset_phys(ds)->ds_deadlist_obj); list_create(&ds->ds_sendstreams, sizeof (dmu_sendarg_t), offsetof(dmu_sendarg_t, dsa_link)); list_create(&ds->ds_prop_cbs, sizeof (dsl_prop_cb_record_t), offsetof(dsl_prop_cb_record_t, cbr_ds_node)); if (doi.doi_type == DMU_OTN_ZAP_METADATA) { for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { if (!(spa_feature_table[f].fi_flags & ZFEATURE_FLAG_PER_DATASET)) continue; err = zap_contains(mos, dsobj, spa_feature_table[f].fi_guid); if (err == 0) { ds->ds_feature_inuse[f] = B_TRUE; } else { ASSERT3U(err, ==, ENOENT); err = 0; } } } err = dsl_dir_hold_obj(dp, dsl_dataset_phys(ds)->ds_dir_obj, NULL, ds, &ds->ds_dir); if (err != 0) { mutex_destroy(&ds->ds_lock); mutex_destroy(&ds->ds_opening_lock); mutex_destroy(&ds->ds_sendstream_lock); refcount_destroy(&ds->ds_longholds); bplist_destroy(&ds->ds_pending_deadlist); dsl_deadlist_close(&ds->ds_deadlist); kmem_free(ds, sizeof (dsl_dataset_t)); dmu_buf_rele(dbuf, tag); return (err); } if (!ds->ds_is_snapshot) { ds->ds_snapname[0] = '\0'; if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) { err = dsl_dataset_hold_obj(dp, dsl_dataset_phys(ds)->ds_prev_snap_obj, ds, &ds->ds_prev); } if (doi.doi_type == DMU_OTN_ZAP_METADATA) { int zaperr = zap_lookup(mos, ds->ds_object, DS_FIELD_BOOKMARK_NAMES, sizeof (ds->ds_bookmarks), 1, &ds->ds_bookmarks); if (zaperr != ENOENT) VERIFY0(zaperr); } } else { if (zfs_flags & ZFS_DEBUG_SNAPNAMES) err = dsl_dataset_get_snapname(ds); if (err == 0 && dsl_dataset_phys(ds)->ds_userrefs_obj != 0) { err = zap_count( ds->ds_dir->dd_pool->dp_meta_objset, dsl_dataset_phys(ds)->ds_userrefs_obj, &ds->ds_userrefs); } } if (err == 0 && !ds->ds_is_snapshot) { err = dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_REFRESERVATION), &ds->ds_reserved); if (err == 0) { err = dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_REFQUOTA), &ds->ds_quota); } } else { ds->ds_reserved = ds->ds_quota = 0; } dmu_buf_init_user(&ds->ds_dbu, dsl_dataset_evict, &ds->ds_dbuf); if (err == 0) winner = dmu_buf_set_user_ie(dbuf, &ds->ds_dbu); if (err != 0 || winner != NULL) { bplist_destroy(&ds->ds_pending_deadlist); dsl_deadlist_close(&ds->ds_deadlist); if (ds->ds_prev) dsl_dataset_rele(ds->ds_prev, ds); dsl_dir_rele(ds->ds_dir, ds); mutex_destroy(&ds->ds_lock); mutex_destroy(&ds->ds_opening_lock); mutex_destroy(&ds->ds_sendstream_lock); refcount_destroy(&ds->ds_longholds); kmem_free(ds, sizeof (dsl_dataset_t)); if (err != 0) { dmu_buf_rele(dbuf, tag); return (err); } ds = winner; } else { ds->ds_fsid_guid = unique_insert(dsl_dataset_phys(ds)->ds_fsid_guid); } } ASSERT3P(ds->ds_dbuf, ==, dbuf); ASSERT3P(dsl_dataset_phys(ds), ==, dbuf->db_data); ASSERT(dsl_dataset_phys(ds)->ds_prev_snap_obj != 0 || spa_version(dp->dp_spa) < SPA_VERSION_ORIGIN || dp->dp_origin_snap == NULL || ds == dp->dp_origin_snap); *dsp = ds; return (0); } int dsl_dataset_hold(dsl_pool_t *dp, const char *name, void *tag, dsl_dataset_t **dsp) { dsl_dir_t *dd; const char *snapname; uint64_t obj; int err = 0; dsl_dataset_t *ds; err = dsl_dir_hold(dp, name, FTAG, &dd, &snapname); if (err != 0) return (err); ASSERT(dsl_pool_config_held(dp)); obj = dsl_dir_phys(dd)->dd_head_dataset_obj; if (obj != 0) err = dsl_dataset_hold_obj(dp, obj, tag, &ds); else err = SET_ERROR(ENOENT); /* we may be looking for a snapshot */ if (err == 0 && snapname != NULL) { dsl_dataset_t *snap_ds; if (*snapname++ != '@') { dsl_dataset_rele(ds, tag); dsl_dir_rele(dd, FTAG); return (SET_ERROR(ENOENT)); } dprintf("looking for snapshot '%s'\n", snapname); err = dsl_dataset_snap_lookup(ds, snapname, &obj); if (err == 0) err = dsl_dataset_hold_obj(dp, obj, tag, &snap_ds); dsl_dataset_rele(ds, tag); if (err == 0) { mutex_enter(&snap_ds->ds_lock); if (snap_ds->ds_snapname[0] == 0) (void) strlcpy(snap_ds->ds_snapname, snapname, sizeof (snap_ds->ds_snapname)); mutex_exit(&snap_ds->ds_lock); ds = snap_ds; } } if (err == 0) *dsp = ds; dsl_dir_rele(dd, FTAG); return (err); } int dsl_dataset_own_obj(dsl_pool_t *dp, uint64_t dsobj, void *tag, dsl_dataset_t **dsp) { int err = dsl_dataset_hold_obj(dp, dsobj, tag, dsp); if (err != 0) return (err); if (!dsl_dataset_tryown(*dsp, tag)) { dsl_dataset_rele(*dsp, tag); *dsp = NULL; return (SET_ERROR(EBUSY)); } return (0); } int dsl_dataset_own(dsl_pool_t *dp, const char *name, void *tag, dsl_dataset_t **dsp) { int err = dsl_dataset_hold(dp, name, tag, dsp); if (err != 0) return (err); if (!dsl_dataset_tryown(*dsp, tag)) { dsl_dataset_rele(*dsp, tag); return (SET_ERROR(EBUSY)); } return (0); } /* * See the comment above dsl_pool_hold() for details. In summary, a long * hold is used to prevent destruction of a dataset while the pool hold * is dropped, allowing other concurrent operations (e.g. spa_sync()). * * The dataset and pool must be held when this function is called. After it * is called, the pool hold may be released while the dataset is still held * and accessed. */ void dsl_dataset_long_hold(dsl_dataset_t *ds, void *tag) { ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); (void) refcount_add(&ds->ds_longholds, tag); } void dsl_dataset_long_rele(dsl_dataset_t *ds, void *tag) { (void) refcount_remove(&ds->ds_longholds, tag); } /* Return B_TRUE if there are any long holds on this dataset. */ boolean_t dsl_dataset_long_held(dsl_dataset_t *ds) { return (!refcount_is_zero(&ds->ds_longholds)); } void dsl_dataset_name(dsl_dataset_t *ds, char *name) { if (ds == NULL) { (void) strcpy(name, "mos"); } else { dsl_dir_name(ds->ds_dir, name); VERIFY0(dsl_dataset_get_snapname(ds)); if (ds->ds_snapname[0]) { (void) strcat(name, "@"); /* * We use a "recursive" mutex so that we * can call dprintf_ds() with ds_lock held. */ if (!MUTEX_HELD(&ds->ds_lock)) { mutex_enter(&ds->ds_lock); (void) strcat(name, ds->ds_snapname); mutex_exit(&ds->ds_lock); } else { (void) strcat(name, ds->ds_snapname); } } } } void dsl_dataset_rele(dsl_dataset_t *ds, void *tag) { dmu_buf_rele(ds->ds_dbuf, tag); } void dsl_dataset_disown(dsl_dataset_t *ds, void *tag) { ASSERT3P(ds->ds_owner, ==, tag); ASSERT(ds->ds_dbuf != NULL); mutex_enter(&ds->ds_lock); ds->ds_owner = NULL; mutex_exit(&ds->ds_lock); dsl_dataset_long_rele(ds, tag); dsl_dataset_rele(ds, tag); } boolean_t dsl_dataset_tryown(dsl_dataset_t *ds, void *tag) { boolean_t gotit = FALSE; ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); mutex_enter(&ds->ds_lock); if (ds->ds_owner == NULL && !DS_IS_INCONSISTENT(ds)) { ds->ds_owner = tag; dsl_dataset_long_hold(ds, tag); gotit = TRUE; } mutex_exit(&ds->ds_lock); return (gotit); } boolean_t dsl_dataset_has_owner(dsl_dataset_t *ds) { boolean_t rv; mutex_enter(&ds->ds_lock); rv = (ds->ds_owner != NULL); mutex_exit(&ds->ds_lock); return (rv); } static void dsl_dataset_activate_feature(uint64_t dsobj, spa_feature_t f, dmu_tx_t *tx) { spa_t *spa = dmu_tx_pool(tx)->dp_spa; objset_t *mos = dmu_tx_pool(tx)->dp_meta_objset; uint64_t zero = 0; VERIFY(spa_feature_table[f].fi_flags & ZFEATURE_FLAG_PER_DATASET); spa_feature_incr(spa, f, tx); dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx); VERIFY0(zap_add(mos, dsobj, spa_feature_table[f].fi_guid, sizeof (zero), 1, &zero, tx)); } void dsl_dataset_deactivate_feature(uint64_t dsobj, spa_feature_t f, dmu_tx_t *tx) { spa_t *spa = dmu_tx_pool(tx)->dp_spa; objset_t *mos = dmu_tx_pool(tx)->dp_meta_objset; VERIFY(spa_feature_table[f].fi_flags & ZFEATURE_FLAG_PER_DATASET); VERIFY0(zap_remove(mos, dsobj, spa_feature_table[f].fi_guid, tx)); spa_feature_decr(spa, f, tx); } uint64_t dsl_dataset_create_sync_dd(dsl_dir_t *dd, dsl_dataset_t *origin, uint64_t flags, dmu_tx_t *tx) { dsl_pool_t *dp = dd->dd_pool; dmu_buf_t *dbuf; dsl_dataset_phys_t *dsphys; uint64_t dsobj; objset_t *mos = dp->dp_meta_objset; if (origin == NULL) origin = dp->dp_origin_snap; ASSERT(origin == NULL || origin->ds_dir->dd_pool == dp); ASSERT(origin == NULL || dsl_dataset_phys(origin)->ds_num_children > 0); ASSERT(dmu_tx_is_syncing(tx)); ASSERT(dsl_dir_phys(dd)->dd_head_dataset_obj == 0); dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0, DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx); VERIFY0(dmu_bonus_hold(mos, dsobj, FTAG, &dbuf)); dmu_buf_will_dirty(dbuf, tx); dsphys = dbuf->db_data; bzero(dsphys, sizeof (dsl_dataset_phys_t)); dsphys->ds_dir_obj = dd->dd_object; dsphys->ds_flags = flags; dsphys->ds_fsid_guid = unique_create(); do { (void) random_get_pseudo_bytes((void*)&dsphys->ds_guid, sizeof (dsphys->ds_guid)); } while (dsphys->ds_guid == 0); dsphys->ds_snapnames_zapobj = zap_create_norm(mos, U8_TEXTPREP_TOUPPER, DMU_OT_DSL_DS_SNAP_MAP, DMU_OT_NONE, 0, tx); dsphys->ds_creation_time = gethrestime_sec(); dsphys->ds_creation_txg = tx->tx_txg == TXG_INITIAL ? 1 : tx->tx_txg; if (origin == NULL) { dsphys->ds_deadlist_obj = dsl_deadlist_alloc(mos, tx); } else { dsl_dataset_t *ohds; /* head of the origin snapshot */ dsphys->ds_prev_snap_obj = origin->ds_object; dsphys->ds_prev_snap_txg = dsl_dataset_phys(origin)->ds_creation_txg; dsphys->ds_referenced_bytes = dsl_dataset_phys(origin)->ds_referenced_bytes; dsphys->ds_compressed_bytes = dsl_dataset_phys(origin)->ds_compressed_bytes; dsphys->ds_uncompressed_bytes = dsl_dataset_phys(origin)->ds_uncompressed_bytes; dsphys->ds_bp = dsl_dataset_phys(origin)->ds_bp; /* * Inherit flags that describe the dataset's contents * (INCONSISTENT) or properties (Case Insensitive). */ dsphys->ds_flags |= dsl_dataset_phys(origin)->ds_flags & (DS_FLAG_INCONSISTENT | DS_FLAG_CI_DATASET); for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { if (origin->ds_feature_inuse[f]) dsl_dataset_activate_feature(dsobj, f, tx); } dmu_buf_will_dirty(origin->ds_dbuf, tx); dsl_dataset_phys(origin)->ds_num_children++; VERIFY0(dsl_dataset_hold_obj(dp, dsl_dir_phys(origin->ds_dir)->dd_head_dataset_obj, FTAG, &ohds)); dsphys->ds_deadlist_obj = dsl_deadlist_clone(&ohds->ds_deadlist, dsphys->ds_prev_snap_txg, dsphys->ds_prev_snap_obj, tx); dsl_dataset_rele(ohds, FTAG); if (spa_version(dp->dp_spa) >= SPA_VERSION_NEXT_CLONES) { if (dsl_dataset_phys(origin)->ds_next_clones_obj == 0) { dsl_dataset_phys(origin)->ds_next_clones_obj = zap_create(mos, DMU_OT_NEXT_CLONES, DMU_OT_NONE, 0, tx); } VERIFY0(zap_add_int(mos, dsl_dataset_phys(origin)->ds_next_clones_obj, dsobj, tx)); } dmu_buf_will_dirty(dd->dd_dbuf, tx); dsl_dir_phys(dd)->dd_origin_obj = origin->ds_object; if (spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) { if (dsl_dir_phys(origin->ds_dir)->dd_clones == 0) { dmu_buf_will_dirty(origin->ds_dir->dd_dbuf, tx); dsl_dir_phys(origin->ds_dir)->dd_clones = zap_create(mos, DMU_OT_DSL_CLONES, DMU_OT_NONE, 0, tx); } VERIFY0(zap_add_int(mos, dsl_dir_phys(origin->ds_dir)->dd_clones, dsobj, tx)); } } if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE) dsphys->ds_flags |= DS_FLAG_UNIQUE_ACCURATE; dmu_buf_rele(dbuf, FTAG); dmu_buf_will_dirty(dd->dd_dbuf, tx); dsl_dir_phys(dd)->dd_head_dataset_obj = dsobj; return (dsobj); } static void dsl_dataset_zero_zil(dsl_dataset_t *ds, dmu_tx_t *tx) { objset_t *os; VERIFY0(dmu_objset_from_ds(ds, &os)); bzero(&os->os_zil_header, sizeof (os->os_zil_header)); dsl_dataset_dirty(ds, tx); } uint64_t dsl_dataset_create_sync(dsl_dir_t *pdd, const char *lastname, dsl_dataset_t *origin, uint64_t flags, cred_t *cr, dmu_tx_t *tx) { dsl_pool_t *dp = pdd->dd_pool; uint64_t dsobj, ddobj; dsl_dir_t *dd; ASSERT(dmu_tx_is_syncing(tx)); ASSERT(lastname[0] != '@'); ddobj = dsl_dir_create_sync(dp, pdd, lastname, tx); VERIFY0(dsl_dir_hold_obj(dp, ddobj, lastname, FTAG, &dd)); dsobj = dsl_dataset_create_sync_dd(dd, origin, flags & ~DS_CREATE_FLAG_NODIRTY, tx); dsl_deleg_set_create_perms(dd, tx, cr); /* * Since we're creating a new node we know it's a leaf, so we can * initialize the counts if the limit feature is active. */ if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_FS_SS_LIMIT)) { uint64_t cnt = 0; objset_t *os = dd->dd_pool->dp_meta_objset; dsl_dir_zapify(dd, tx); VERIFY0(zap_add(os, dd->dd_object, DD_FIELD_FILESYSTEM_COUNT, sizeof (cnt), 1, &cnt, tx)); VERIFY0(zap_add(os, dd->dd_object, DD_FIELD_SNAPSHOT_COUNT, sizeof (cnt), 1, &cnt, tx)); } dsl_dir_rele(dd, FTAG); /* * If we are creating a clone, make sure we zero out any stale * data from the origin snapshots zil header. */ if (origin != NULL && !(flags & DS_CREATE_FLAG_NODIRTY)) { dsl_dataset_t *ds; VERIFY0(dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds)); dsl_dataset_zero_zil(ds, tx); dsl_dataset_rele(ds, FTAG); } return (dsobj); } #ifdef __FreeBSD__ /* FreeBSD ioctl compat begin */ struct destroyarg { nvlist_t *nvl; const char *snapname; }; static int dsl_check_snap_cb(const char *name, void *arg) { struct destroyarg *da = arg; dsl_dataset_t *ds; char *dsname; dsname = kmem_asprintf("%s@%s", name, da->snapname); fnvlist_add_boolean(da->nvl, dsname); kmem_free(dsname, strlen(dsname) + 1); return (0); } int dmu_get_recursive_snaps_nvl(char *fsname, const char *snapname, nvlist_t *snaps) { struct destroyarg *da; int err; da = kmem_zalloc(sizeof (struct destroyarg), KM_SLEEP); da->nvl = snaps; da->snapname = snapname; err = dmu_objset_find(fsname, dsl_check_snap_cb, da, DS_FIND_CHILDREN); kmem_free(da, sizeof (struct destroyarg)); return (err); } /* FreeBSD ioctl compat end */ #endif /* __FreeBSD__ */ /* * The unique space in the head dataset can be calculated by subtracting * the space used in the most recent snapshot, that is still being used * in this file system, from the space currently in use. To figure out * the space in the most recent snapshot still in use, we need to take * the total space used in the snapshot and subtract out the space that * has been freed up since the snapshot was taken. */ void dsl_dataset_recalc_head_uniq(dsl_dataset_t *ds) { uint64_t mrs_used; uint64_t dlused, dlcomp, dluncomp; ASSERT(!ds->ds_is_snapshot); if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) mrs_used = dsl_dataset_phys(ds->ds_prev)->ds_referenced_bytes; else mrs_used = 0; dsl_deadlist_space(&ds->ds_deadlist, &dlused, &dlcomp, &dluncomp); ASSERT3U(dlused, <=, mrs_used); dsl_dataset_phys(ds)->ds_unique_bytes = dsl_dataset_phys(ds)->ds_referenced_bytes - (mrs_used - dlused); if (spa_version(ds->ds_dir->dd_pool->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE) dsl_dataset_phys(ds)->ds_flags |= DS_FLAG_UNIQUE_ACCURATE; } void dsl_dataset_remove_from_next_clones(dsl_dataset_t *ds, uint64_t obj, dmu_tx_t *tx) { objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; uint64_t count; int err; ASSERT(dsl_dataset_phys(ds)->ds_num_children >= 2); err = zap_remove_int(mos, dsl_dataset_phys(ds)->ds_next_clones_obj, obj, tx); /* * The err should not be ENOENT, but a bug in a previous version * of the code could cause upgrade_clones_cb() to not set * ds_next_snap_obj when it should, leading to a missing entry. * If we knew that the pool was created after * SPA_VERSION_NEXT_CLONES, we could assert that it isn't * ENOENT. However, at least we can check that we don't have * too many entries in the next_clones_obj even after failing to * remove this one. */ if (err != ENOENT) VERIFY0(err); ASSERT0(zap_count(mos, dsl_dataset_phys(ds)->ds_next_clones_obj, &count)); ASSERT3U(count, <=, dsl_dataset_phys(ds)->ds_num_children - 2); } blkptr_t * dsl_dataset_get_blkptr(dsl_dataset_t *ds) { return (&dsl_dataset_phys(ds)->ds_bp); } void dsl_dataset_set_blkptr(dsl_dataset_t *ds, blkptr_t *bp, dmu_tx_t *tx) { ASSERT(dmu_tx_is_syncing(tx)); /* If it's the meta-objset, set dp_meta_rootbp */ if (ds == NULL) { tx->tx_pool->dp_meta_rootbp = *bp; } else { dmu_buf_will_dirty(ds->ds_dbuf, tx); dsl_dataset_phys(ds)->ds_bp = *bp; } } spa_t * dsl_dataset_get_spa(dsl_dataset_t *ds) { return (ds->ds_dir->dd_pool->dp_spa); } void dsl_dataset_dirty(dsl_dataset_t *ds, dmu_tx_t *tx) { dsl_pool_t *dp; if (ds == NULL) /* this is the meta-objset */ return; ASSERT(ds->ds_objset != NULL); if (dsl_dataset_phys(ds)->ds_next_snap_obj != 0) panic("dirtying snapshot!"); dp = ds->ds_dir->dd_pool; if (txg_list_add(&dp->dp_dirty_datasets, ds, tx->tx_txg)) { /* up the hold count until we can be written out */ dmu_buf_add_ref(ds->ds_dbuf, ds); } } boolean_t dsl_dataset_is_dirty(dsl_dataset_t *ds) { for (int t = 0; t < TXG_SIZE; t++) { if (txg_list_member(&ds->ds_dir->dd_pool->dp_dirty_datasets, ds, t)) return (B_TRUE); } return (B_FALSE); } static int dsl_dataset_snapshot_reserve_space(dsl_dataset_t *ds, dmu_tx_t *tx) { uint64_t asize; if (!dmu_tx_is_syncing(tx)) return (0); /* * If there's an fs-only reservation, any blocks that might become * owned by the snapshot dataset must be accommodated by space * outside of the reservation. */ ASSERT(ds->ds_reserved == 0 || DS_UNIQUE_IS_ACCURATE(ds)); asize = MIN(dsl_dataset_phys(ds)->ds_unique_bytes, ds->ds_reserved); if (asize > dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE)) return (SET_ERROR(ENOSPC)); /* * Propagate any reserved space for this snapshot to other * snapshot checks in this sync group. */ if (asize > 0) dsl_dir_willuse_space(ds->ds_dir, asize, tx); return (0); } typedef struct dsl_dataset_snapshot_arg { nvlist_t *ddsa_snaps; nvlist_t *ddsa_props; nvlist_t *ddsa_errors; cred_t *ddsa_cr; } dsl_dataset_snapshot_arg_t; int dsl_dataset_snapshot_check_impl(dsl_dataset_t *ds, const char *snapname, dmu_tx_t *tx, boolean_t recv, uint64_t cnt, cred_t *cr) { int error; uint64_t value; ds->ds_trysnap_txg = tx->tx_txg; if (!dmu_tx_is_syncing(tx)) return (0); /* * We don't allow multiple snapshots of the same txg. If there * is already one, try again. */ if (dsl_dataset_phys(ds)->ds_prev_snap_txg >= tx->tx_txg) return (SET_ERROR(EAGAIN)); /* * Check for conflicting snapshot name. */ error = dsl_dataset_snap_lookup(ds, snapname, &value); if (error == 0) return (SET_ERROR(EEXIST)); if (error != ENOENT) return (error); /* * We don't allow taking snapshots of inconsistent datasets, such as * those into which we are currently receiving. However, if we are * creating this snapshot as part of a receive, this check will be * executed atomically with respect to the completion of the receive * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this * case we ignore this, knowing it will be fixed up for us shortly in * dmu_recv_end_sync(). */ if (!recv && DS_IS_INCONSISTENT(ds)) return (SET_ERROR(EBUSY)); /* * Skip the check for temporary snapshots or if we have already checked * the counts in dsl_dataset_snapshot_check. This means we really only * check the count here when we're receiving a stream. */ if (cnt != 0 && cr != NULL) { error = dsl_fs_ss_limit_check(ds->ds_dir, cnt, ZFS_PROP_SNAPSHOT_LIMIT, NULL, cr); if (error != 0) return (error); } error = dsl_dataset_snapshot_reserve_space(ds, tx); if (error != 0) return (error); return (0); } static int dsl_dataset_snapshot_check(void *arg, dmu_tx_t *tx) { dsl_dataset_snapshot_arg_t *ddsa = arg; dsl_pool_t *dp = dmu_tx_pool(tx); nvpair_t *pair; int rv = 0; /* * Pre-compute how many total new snapshots will be created for each * level in the tree and below. This is needed for validating the * snapshot limit when either taking a recursive snapshot or when * taking multiple snapshots. * * The problem is that the counts are not actually adjusted when * we are checking, only when we finally sync. For a single snapshot, * this is easy, the count will increase by 1 at each node up the tree, * but its more complicated for the recursive/multiple snapshot case. * * The dsl_fs_ss_limit_check function does recursively check the count * at each level up the tree but since it is validating each snapshot * independently we need to be sure that we are validating the complete * count for the entire set of snapshots. We do this by rolling up the * counts for each component of the name into an nvlist and then * checking each of those cases with the aggregated count. * * This approach properly handles not only the recursive snapshot * case (where we get all of those on the ddsa_snaps list) but also * the sibling case (e.g. snapshot a/b and a/c so that we will also * validate the limit on 'a' using a count of 2). * * We validate the snapshot names in the third loop and only report * name errors once. */ if (dmu_tx_is_syncing(tx)) { nvlist_t *cnt_track = NULL; cnt_track = fnvlist_alloc(); /* Rollup aggregated counts into the cnt_track list */ for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL); pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) { char *pdelim; uint64_t val; char nm[MAXPATHLEN]; (void) strlcpy(nm, nvpair_name(pair), sizeof (nm)); pdelim = strchr(nm, '@'); if (pdelim == NULL) continue; *pdelim = '\0'; do { if (nvlist_lookup_uint64(cnt_track, nm, &val) == 0) { /* update existing entry */ fnvlist_add_uint64(cnt_track, nm, val + 1); } else { /* add to list */ fnvlist_add_uint64(cnt_track, nm, 1); } pdelim = strrchr(nm, '/'); if (pdelim != NULL) *pdelim = '\0'; } while (pdelim != NULL); } /* Check aggregated counts at each level */ for (pair = nvlist_next_nvpair(cnt_track, NULL); pair != NULL; pair = nvlist_next_nvpair(cnt_track, pair)) { int error = 0; char *name; uint64_t cnt = 0; dsl_dataset_t *ds; name = nvpair_name(pair); cnt = fnvpair_value_uint64(pair); ASSERT(cnt > 0); error = dsl_dataset_hold(dp, name, FTAG, &ds); if (error == 0) { error = dsl_fs_ss_limit_check(ds->ds_dir, cnt, ZFS_PROP_SNAPSHOT_LIMIT, NULL, ddsa->ddsa_cr); dsl_dataset_rele(ds, FTAG); } if (error != 0) { if (ddsa->ddsa_errors != NULL) fnvlist_add_int32(ddsa->ddsa_errors, name, error); rv = error; /* only report one error for this check */ break; } } nvlist_free(cnt_track); } for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL); pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) { int error = 0; dsl_dataset_t *ds; char *name, *atp; char dsname[MAXNAMELEN]; name = nvpair_name(pair); if (strlen(name) >= MAXNAMELEN) error = SET_ERROR(ENAMETOOLONG); if (error == 0) { atp = strchr(name, '@'); if (atp == NULL) error = SET_ERROR(EINVAL); if (error == 0) (void) strlcpy(dsname, name, atp - name + 1); } if (error == 0) error = dsl_dataset_hold(dp, dsname, FTAG, &ds); if (error == 0) { /* passing 0/NULL skips dsl_fs_ss_limit_check */ error = dsl_dataset_snapshot_check_impl(ds, atp + 1, tx, B_FALSE, 0, NULL); dsl_dataset_rele(ds, FTAG); } if (error != 0) { if (ddsa->ddsa_errors != NULL) { fnvlist_add_int32(ddsa->ddsa_errors, name, error); } rv = error; } } return (rv); } void dsl_dataset_snapshot_sync_impl(dsl_dataset_t *ds, const char *snapname, dmu_tx_t *tx) { static zil_header_t zero_zil; dsl_pool_t *dp = ds->ds_dir->dd_pool; dmu_buf_t *dbuf; dsl_dataset_phys_t *dsphys; uint64_t dsobj, crtxg; objset_t *mos = dp->dp_meta_objset; objset_t *os; ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock)); /* * If we are on an old pool, the zil must not be active, in which * case it will be zeroed. Usually zil_suspend() accomplishes this. */ ASSERT(spa_version(dmu_tx_pool(tx)->dp_spa) >= SPA_VERSION_FAST_SNAP || dmu_objset_from_ds(ds, &os) != 0 || bcmp(&os->os_phys->os_zil_header, &zero_zil, sizeof (zero_zil)) == 0); dsl_fs_ss_count_adjust(ds->ds_dir, 1, DD_FIELD_SNAPSHOT_COUNT, tx); /* * The origin's ds_creation_txg has to be < TXG_INITIAL */ if (strcmp(snapname, ORIGIN_DIR_NAME) == 0) crtxg = 1; else crtxg = tx->tx_txg; dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0, DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx); VERIFY0(dmu_bonus_hold(mos, dsobj, FTAG, &dbuf)); dmu_buf_will_dirty(dbuf, tx); dsphys = dbuf->db_data; bzero(dsphys, sizeof (dsl_dataset_phys_t)); dsphys->ds_dir_obj = ds->ds_dir->dd_object; dsphys->ds_fsid_guid = unique_create(); do { (void) random_get_pseudo_bytes((void*)&dsphys->ds_guid, sizeof (dsphys->ds_guid)); } while (dsphys->ds_guid == 0); dsphys->ds_prev_snap_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj; dsphys->ds_prev_snap_txg = dsl_dataset_phys(ds)->ds_prev_snap_txg; dsphys->ds_next_snap_obj = ds->ds_object; dsphys->ds_num_children = 1; dsphys->ds_creation_time = gethrestime_sec(); dsphys->ds_creation_txg = crtxg; dsphys->ds_deadlist_obj = dsl_dataset_phys(ds)->ds_deadlist_obj; dsphys->ds_referenced_bytes = dsl_dataset_phys(ds)->ds_referenced_bytes; dsphys->ds_compressed_bytes = dsl_dataset_phys(ds)->ds_compressed_bytes; dsphys->ds_uncompressed_bytes = dsl_dataset_phys(ds)->ds_uncompressed_bytes; dsphys->ds_flags = dsl_dataset_phys(ds)->ds_flags; dsphys->ds_bp = dsl_dataset_phys(ds)->ds_bp; dmu_buf_rele(dbuf, FTAG); for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { if (ds->ds_feature_inuse[f]) dsl_dataset_activate_feature(dsobj, f, tx); } ASSERT3U(ds->ds_prev != 0, ==, dsl_dataset_phys(ds)->ds_prev_snap_obj != 0); if (ds->ds_prev) { uint64_t next_clones_obj = dsl_dataset_phys(ds->ds_prev)->ds_next_clones_obj; ASSERT(dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj == ds->ds_object || dsl_dataset_phys(ds->ds_prev)->ds_num_children > 1); if (dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj == ds->ds_object) { dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx); ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_txg, ==, dsl_dataset_phys(ds->ds_prev)->ds_creation_txg); dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj = dsobj; } else if (next_clones_obj != 0) { dsl_dataset_remove_from_next_clones(ds->ds_prev, dsphys->ds_next_snap_obj, tx); VERIFY0(zap_add_int(mos, next_clones_obj, dsobj, tx)); } } /* * If we have a reference-reservation on this dataset, we will * need to increase the amount of refreservation being charged * since our unique space is going to zero. */ if (ds->ds_reserved) { int64_t delta; ASSERT(DS_UNIQUE_IS_ACCURATE(ds)); delta = MIN(dsl_dataset_phys(ds)->ds_unique_bytes, ds->ds_reserved); dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV, delta, 0, 0, tx); } dmu_buf_will_dirty(ds->ds_dbuf, tx); dsl_dataset_phys(ds)->ds_deadlist_obj = dsl_deadlist_clone(&ds->ds_deadlist, UINT64_MAX, dsl_dataset_phys(ds)->ds_prev_snap_obj, tx); dsl_deadlist_close(&ds->ds_deadlist); dsl_deadlist_open(&ds->ds_deadlist, mos, dsl_dataset_phys(ds)->ds_deadlist_obj); dsl_deadlist_add_key(&ds->ds_deadlist, dsl_dataset_phys(ds)->ds_prev_snap_txg, tx); ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_txg, <, tx->tx_txg); dsl_dataset_phys(ds)->ds_prev_snap_obj = dsobj; dsl_dataset_phys(ds)->ds_prev_snap_txg = crtxg; dsl_dataset_phys(ds)->ds_unique_bytes = 0; if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE) dsl_dataset_phys(ds)->ds_flags |= DS_FLAG_UNIQUE_ACCURATE; VERIFY0(zap_add(mos, dsl_dataset_phys(ds)->ds_snapnames_zapobj, snapname, 8, 1, &dsobj, tx)); if (ds->ds_prev) dsl_dataset_rele(ds->ds_prev, ds); VERIFY0(dsl_dataset_hold_obj(dp, dsl_dataset_phys(ds)->ds_prev_snap_obj, ds, &ds->ds_prev)); dsl_scan_ds_snapshotted(ds, tx); dsl_dir_snap_cmtime_update(ds->ds_dir); spa_history_log_internal_ds(ds->ds_prev, "snapshot", tx, ""); } static void dsl_dataset_snapshot_sync(void *arg, dmu_tx_t *tx) { dsl_dataset_snapshot_arg_t *ddsa = arg; dsl_pool_t *dp = dmu_tx_pool(tx); nvpair_t *pair; for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL); pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) { dsl_dataset_t *ds; char *name, *atp; char dsname[MAXNAMELEN]; name = nvpair_name(pair); atp = strchr(name, '@'); (void) strlcpy(dsname, name, atp - name + 1); VERIFY0(dsl_dataset_hold(dp, dsname, FTAG, &ds)); dsl_dataset_snapshot_sync_impl(ds, atp + 1, tx); if (ddsa->ddsa_props != NULL) { dsl_props_set_sync_impl(ds->ds_prev, ZPROP_SRC_LOCAL, ddsa->ddsa_props, tx); } dsl_dataset_rele(ds, FTAG); } } /* * The snapshots must all be in the same pool. * All-or-nothing: if there are any failures, nothing will be modified. */ int dsl_dataset_snapshot(nvlist_t *snaps, nvlist_t *props, nvlist_t *errors) { dsl_dataset_snapshot_arg_t ddsa; nvpair_t *pair; boolean_t needsuspend; int error; spa_t *spa; char *firstname; nvlist_t *suspended = NULL; pair = nvlist_next_nvpair(snaps, NULL); if (pair == NULL) return (0); firstname = nvpair_name(pair); error = spa_open(firstname, &spa, FTAG); if (error != 0) return (error); needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP); spa_close(spa, FTAG); if (needsuspend) { suspended = fnvlist_alloc(); for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; pair = nvlist_next_nvpair(snaps, pair)) { char fsname[MAXNAMELEN]; char *snapname = nvpair_name(pair); char *atp; void *cookie; atp = strchr(snapname, '@'); if (atp == NULL) { error = SET_ERROR(EINVAL); break; } (void) strlcpy(fsname, snapname, atp - snapname + 1); error = zil_suspend(fsname, &cookie); if (error != 0) break; fnvlist_add_uint64(suspended, fsname, (uintptr_t)cookie); } } ddsa.ddsa_snaps = snaps; ddsa.ddsa_props = props; ddsa.ddsa_errors = errors; ddsa.ddsa_cr = CRED(); if (error == 0) { error = dsl_sync_task(firstname, dsl_dataset_snapshot_check, dsl_dataset_snapshot_sync, &ddsa, fnvlist_num_pairs(snaps) * 3, ZFS_SPACE_CHECK_NORMAL); } if (suspended != NULL) { for (pair = nvlist_next_nvpair(suspended, NULL); pair != NULL; pair = nvlist_next_nvpair(suspended, pair)) { zil_resume((void *)(uintptr_t) fnvpair_value_uint64(pair)); } fnvlist_free(suspended); } #ifdef __FreeBSD__ #ifdef _KERNEL if (error == 0) { for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; pair = nvlist_next_nvpair(snaps, pair)) { char *snapname = nvpair_name(pair); zvol_create_minors(snapname); } } #endif #endif return (error); } typedef struct dsl_dataset_snapshot_tmp_arg { const char *ddsta_fsname; const char *ddsta_snapname; minor_t ddsta_cleanup_minor; const char *ddsta_htag; } dsl_dataset_snapshot_tmp_arg_t; static int dsl_dataset_snapshot_tmp_check(void *arg, dmu_tx_t *tx) { dsl_dataset_snapshot_tmp_arg_t *ddsta = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dataset_t *ds; int error; error = dsl_dataset_hold(dp, ddsta->ddsta_fsname, FTAG, &ds); if (error != 0) return (error); /* NULL cred means no limit check for tmp snapshot */ error = dsl_dataset_snapshot_check_impl(ds, ddsta->ddsta_snapname, tx, B_FALSE, 0, NULL); if (error != 0) { dsl_dataset_rele(ds, FTAG); return (error); } if (spa_version(dp->dp_spa) < SPA_VERSION_USERREFS) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(ENOTSUP)); } error = dsl_dataset_user_hold_check_one(NULL, ddsta->ddsta_htag, B_TRUE, tx); if (error != 0) { dsl_dataset_rele(ds, FTAG); return (error); } dsl_dataset_rele(ds, FTAG); return (0); } static void dsl_dataset_snapshot_tmp_sync(void *arg, dmu_tx_t *tx) { dsl_dataset_snapshot_tmp_arg_t *ddsta = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dataset_t *ds; VERIFY0(dsl_dataset_hold(dp, ddsta->ddsta_fsname, FTAG, &ds)); dsl_dataset_snapshot_sync_impl(ds, ddsta->ddsta_snapname, tx); dsl_dataset_user_hold_sync_one(ds->ds_prev, ddsta->ddsta_htag, ddsta->ddsta_cleanup_minor, gethrestime_sec(), tx); dsl_destroy_snapshot_sync_impl(ds->ds_prev, B_TRUE, tx); dsl_dataset_rele(ds, FTAG); } int dsl_dataset_snapshot_tmp(const char *fsname, const char *snapname, minor_t cleanup_minor, const char *htag) { dsl_dataset_snapshot_tmp_arg_t ddsta; int error; spa_t *spa; boolean_t needsuspend; void *cookie; ddsta.ddsta_fsname = fsname; ddsta.ddsta_snapname = snapname; ddsta.ddsta_cleanup_minor = cleanup_minor; ddsta.ddsta_htag = htag; error = spa_open(fsname, &spa, FTAG); if (error != 0) return (error); needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP); spa_close(spa, FTAG); if (needsuspend) { error = zil_suspend(fsname, &cookie); if (error != 0) return (error); } error = dsl_sync_task(fsname, dsl_dataset_snapshot_tmp_check, dsl_dataset_snapshot_tmp_sync, &ddsta, 3, ZFS_SPACE_CHECK_RESERVED); if (needsuspend) zil_resume(cookie); return (error); } void dsl_dataset_sync(dsl_dataset_t *ds, zio_t *zio, dmu_tx_t *tx) { ASSERT(dmu_tx_is_syncing(tx)); ASSERT(ds->ds_objset != NULL); ASSERT(dsl_dataset_phys(ds)->ds_next_snap_obj == 0); /* * in case we had to change ds_fsid_guid when we opened it, * sync it out now. */ dmu_buf_will_dirty(ds->ds_dbuf, tx); dsl_dataset_phys(ds)->ds_fsid_guid = ds->ds_fsid_guid; if (ds->ds_resume_bytes[tx->tx_txg & TXG_MASK] != 0) { VERIFY0(zap_update(tx->tx_pool->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_OBJECT, 8, 1, &ds->ds_resume_object[tx->tx_txg & TXG_MASK], tx)); VERIFY0(zap_update(tx->tx_pool->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_OFFSET, 8, 1, &ds->ds_resume_offset[tx->tx_txg & TXG_MASK], tx)); VERIFY0(zap_update(tx->tx_pool->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_BYTES, 8, 1, &ds->ds_resume_bytes[tx->tx_txg & TXG_MASK], tx)); ds->ds_resume_object[tx->tx_txg & TXG_MASK] = 0; ds->ds_resume_offset[tx->tx_txg & TXG_MASK] = 0; ds->ds_resume_bytes[tx->tx_txg & TXG_MASK] = 0; } dmu_objset_sync(ds->ds_objset, zio, tx); for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { if (ds->ds_feature_activation_needed[f]) { if (ds->ds_feature_inuse[f]) continue; dsl_dataset_activate_feature(ds->ds_object, f, tx); ds->ds_feature_inuse[f] = B_TRUE; } } } static void get_clones_stat(dsl_dataset_t *ds, nvlist_t *nv) { uint64_t count = 0; objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; zap_cursor_t zc; zap_attribute_t za; nvlist_t *propval = fnvlist_alloc(); nvlist_t *val = fnvlist_alloc(); ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); /* * There may be missing entries in ds_next_clones_obj * due to a bug in a previous version of the code. * Only trust it if it has the right number of entries. */ if (dsl_dataset_phys(ds)->ds_next_clones_obj != 0) { VERIFY0(zap_count(mos, dsl_dataset_phys(ds)->ds_next_clones_obj, &count)); } if (count != dsl_dataset_phys(ds)->ds_num_children - 1) goto fail; for (zap_cursor_init(&zc, mos, dsl_dataset_phys(ds)->ds_next_clones_obj); zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { dsl_dataset_t *clone; char buf[ZFS_MAXNAMELEN]; VERIFY0(dsl_dataset_hold_obj(ds->ds_dir->dd_pool, za.za_first_integer, FTAG, &clone)); dsl_dir_name(clone->ds_dir, buf); fnvlist_add_boolean(val, buf); dsl_dataset_rele(clone, FTAG); } zap_cursor_fini(&zc); fnvlist_add_nvlist(propval, ZPROP_VALUE, val); fnvlist_add_nvlist(nv, zfs_prop_to_name(ZFS_PROP_CLONES), propval); fail: nvlist_free(val); nvlist_free(propval); } static void get_receive_resume_stats(dsl_dataset_t *ds, nvlist_t *nv) { dsl_pool_t *dp = ds->ds_dir->dd_pool; if (dsl_dataset_has_resume_receive_state(ds)) { char *str; void *packed; uint8_t *compressed; uint64_t val; nvlist_t *token_nv = fnvlist_alloc(); size_t packed_size, compressed_size; if (zap_lookup(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_FROMGUID, sizeof (val), 1, &val) == 0) { fnvlist_add_uint64(token_nv, "fromguid", val); } if (zap_lookup(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_OBJECT, sizeof (val), 1, &val) == 0) { fnvlist_add_uint64(token_nv, "object", val); } if (zap_lookup(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_OFFSET, sizeof (val), 1, &val) == 0) { fnvlist_add_uint64(token_nv, "offset", val); } if (zap_lookup(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_BYTES, sizeof (val), 1, &val) == 0) { fnvlist_add_uint64(token_nv, "bytes", val); } if (zap_lookup(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_TOGUID, sizeof (val), 1, &val) == 0) { fnvlist_add_uint64(token_nv, "toguid", val); } char buf[256]; if (zap_lookup(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_TONAME, 1, sizeof (buf), buf) == 0) { fnvlist_add_string(token_nv, "toname", buf); } if (zap_contains(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_EMBEDOK) == 0) { fnvlist_add_boolean(token_nv, "embedok"); } packed = fnvlist_pack(token_nv, &packed_size); fnvlist_free(token_nv); compressed = kmem_alloc(packed_size, KM_SLEEP); compressed_size = gzip_compress(packed, compressed, packed_size, packed_size, 6); zio_cksum_t cksum; - fletcher_4_native(compressed, compressed_size, &cksum); + fletcher_4_native(compressed, compressed_size, NULL, &cksum); str = kmem_alloc(compressed_size * 2 + 1, KM_SLEEP); for (int i = 0; i < compressed_size; i++) { (void) sprintf(str + i * 2, "%02x", compressed[i]); } str[compressed_size * 2] = '\0'; char *propval = kmem_asprintf("%u-%llx-%llx-%s", ZFS_SEND_RESUME_TOKEN_VERSION, (longlong_t)cksum.zc_word[0], (longlong_t)packed_size, str); dsl_prop_nvlist_add_string(nv, ZFS_PROP_RECEIVE_RESUME_TOKEN, propval); kmem_free(packed, packed_size); kmem_free(str, compressed_size * 2 + 1); kmem_free(compressed, packed_size); strfree(propval); } } void dsl_dataset_stats(dsl_dataset_t *ds, nvlist_t *nv) { dsl_pool_t *dp = ds->ds_dir->dd_pool; uint64_t refd, avail, uobjs, aobjs, ratio; ASSERT(dsl_pool_config_held(dp)); ratio = dsl_dataset_phys(ds)->ds_compressed_bytes == 0 ? 100 : (dsl_dataset_phys(ds)->ds_uncompressed_bytes * 100 / dsl_dataset_phys(ds)->ds_compressed_bytes); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRATIO, ratio); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_LOGICALREFERENCED, dsl_dataset_phys(ds)->ds_uncompressed_bytes); if (ds->ds_is_snapshot) { dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_COMPRESSRATIO, ratio); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USED, dsl_dataset_phys(ds)->ds_unique_bytes); get_clones_stat(ds, nv); } else { if (ds->ds_prev != NULL && ds->ds_prev != dp->dp_origin_snap) { char buf[MAXNAMELEN]; dsl_dataset_name(ds->ds_prev, buf); dsl_prop_nvlist_add_string(nv, ZFS_PROP_PREV_SNAP, buf); } dsl_dir_stats(ds->ds_dir, nv); } dsl_dataset_space(ds, &refd, &avail, &uobjs, &aobjs); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_AVAILABLE, avail); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFERENCED, refd); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATION, dsl_dataset_phys(ds)->ds_creation_time); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATETXG, dsl_dataset_phys(ds)->ds_creation_txg); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFQUOTA, ds->ds_quota); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRESERVATION, ds->ds_reserved); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_GUID, dsl_dataset_phys(ds)->ds_guid); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_UNIQUE, dsl_dataset_phys(ds)->ds_unique_bytes); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_OBJSETID, ds->ds_object); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERREFS, ds->ds_userrefs); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_DEFER_DESTROY, DS_IS_DEFER_DESTROY(ds) ? 1 : 0); if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) { uint64_t written, comp, uncomp; dsl_pool_t *dp = ds->ds_dir->dd_pool; dsl_dataset_t *prev; int err = dsl_dataset_hold_obj(dp, dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev); if (err == 0) { err = dsl_dataset_space_written(prev, ds, &written, &comp, &uncomp); dsl_dataset_rele(prev, FTAG); if (err == 0) { dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_WRITTEN, written); } } } if (!dsl_dataset_is_snapshot(ds)) { /* * A failed "newfs" (e.g. full) resumable receive leaves * the stats set on this dataset. Check here for the prop. */ get_receive_resume_stats(ds, nv); /* * A failed incremental resumable receive leaves the * stats set on our child named "%recv". Check the child * for the prop. */ char recvname[ZFS_MAXNAMELEN]; dsl_dataset_t *recv_ds; dsl_dataset_name(ds, recvname); (void) strcat(recvname, "/"); (void) strcat(recvname, recv_clone_name); if (dsl_dataset_hold(dp, recvname, FTAG, &recv_ds) == 0) { get_receive_resume_stats(recv_ds, nv); dsl_dataset_rele(recv_ds, FTAG); } } } void dsl_dataset_fast_stat(dsl_dataset_t *ds, dmu_objset_stats_t *stat) { dsl_pool_t *dp = ds->ds_dir->dd_pool; ASSERT(dsl_pool_config_held(dp)); stat->dds_creation_txg = dsl_dataset_phys(ds)->ds_creation_txg; stat->dds_inconsistent = dsl_dataset_phys(ds)->ds_flags & DS_FLAG_INCONSISTENT; stat->dds_guid = dsl_dataset_phys(ds)->ds_guid; stat->dds_origin[0] = '\0'; if (ds->ds_is_snapshot) { stat->dds_is_snapshot = B_TRUE; stat->dds_num_clones = dsl_dataset_phys(ds)->ds_num_children - 1; } else { stat->dds_is_snapshot = B_FALSE; stat->dds_num_clones = 0; if (dsl_dir_is_clone(ds->ds_dir)) { dsl_dataset_t *ods; VERIFY0(dsl_dataset_hold_obj(dp, dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods)); dsl_dataset_name(ods, stat->dds_origin); dsl_dataset_rele(ods, FTAG); } } } uint64_t dsl_dataset_fsid_guid(dsl_dataset_t *ds) { return (ds->ds_fsid_guid); } void dsl_dataset_space(dsl_dataset_t *ds, uint64_t *refdbytesp, uint64_t *availbytesp, uint64_t *usedobjsp, uint64_t *availobjsp) { *refdbytesp = dsl_dataset_phys(ds)->ds_referenced_bytes; *availbytesp = dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE); if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes) *availbytesp += ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes; if (ds->ds_quota != 0) { /* * Adjust available bytes according to refquota */ if (*refdbytesp < ds->ds_quota) *availbytesp = MIN(*availbytesp, ds->ds_quota - *refdbytesp); else *availbytesp = 0; } *usedobjsp = BP_GET_FILL(&dsl_dataset_phys(ds)->ds_bp); *availobjsp = DN_MAX_OBJECT - *usedobjsp; } boolean_t dsl_dataset_modified_since_snap(dsl_dataset_t *ds, dsl_dataset_t *snap) { dsl_pool_t *dp = ds->ds_dir->dd_pool; ASSERT(dsl_pool_config_held(dp)); if (snap == NULL) return (B_FALSE); if (dsl_dataset_phys(ds)->ds_bp.blk_birth > dsl_dataset_phys(snap)->ds_creation_txg) { objset_t *os, *os_snap; /* * It may be that only the ZIL differs, because it was * reset in the head. Don't count that as being * modified. */ if (dmu_objset_from_ds(ds, &os) != 0) return (B_TRUE); if (dmu_objset_from_ds(snap, &os_snap) != 0) return (B_TRUE); return (bcmp(&os->os_phys->os_meta_dnode, &os_snap->os_phys->os_meta_dnode, sizeof (os->os_phys->os_meta_dnode)) != 0); } return (B_FALSE); } typedef struct dsl_dataset_rename_snapshot_arg { const char *ddrsa_fsname; const char *ddrsa_oldsnapname; const char *ddrsa_newsnapname; boolean_t ddrsa_recursive; dmu_tx_t *ddrsa_tx; } dsl_dataset_rename_snapshot_arg_t; /* ARGSUSED */ static int dsl_dataset_rename_snapshot_check_impl(dsl_pool_t *dp, dsl_dataset_t *hds, void *arg) { dsl_dataset_rename_snapshot_arg_t *ddrsa = arg; int error; uint64_t val; error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val); if (error != 0) { /* ignore nonexistent snapshots */ return (error == ENOENT ? 0 : error); } /* new name should not exist */ error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_newsnapname, &val); if (error == 0) error = SET_ERROR(EEXIST); else if (error == ENOENT) error = 0; /* dataset name + 1 for the "@" + the new snapshot name must fit */ if (dsl_dir_namelen(hds->ds_dir) + 1 + strlen(ddrsa->ddrsa_newsnapname) >= MAXNAMELEN) error = SET_ERROR(ENAMETOOLONG); return (error); } static int dsl_dataset_rename_snapshot_check(void *arg, dmu_tx_t *tx) { dsl_dataset_rename_snapshot_arg_t *ddrsa = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dataset_t *hds; int error; error = dsl_dataset_hold(dp, ddrsa->ddrsa_fsname, FTAG, &hds); if (error != 0) return (error); if (ddrsa->ddrsa_recursive) { error = dmu_objset_find_dp(dp, hds->ds_dir->dd_object, dsl_dataset_rename_snapshot_check_impl, ddrsa, DS_FIND_CHILDREN); } else { error = dsl_dataset_rename_snapshot_check_impl(dp, hds, ddrsa); } dsl_dataset_rele(hds, FTAG); return (error); } static int dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t *dp, dsl_dataset_t *hds, void *arg) { #ifdef __FreeBSD__ #ifdef _KERNEL char *oldname, *newname; #endif #endif dsl_dataset_rename_snapshot_arg_t *ddrsa = arg; dsl_dataset_t *ds; uint64_t val; dmu_tx_t *tx = ddrsa->ddrsa_tx; int error; error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val); ASSERT(error == 0 || error == ENOENT); if (error == ENOENT) { /* ignore nonexistent snapshots */ return (0); } VERIFY0(dsl_dataset_hold_obj(dp, val, FTAG, &ds)); /* log before we change the name */ spa_history_log_internal_ds(ds, "rename", tx, "-> @%s", ddrsa->ddrsa_newsnapname); VERIFY0(dsl_dataset_snap_remove(hds, ddrsa->ddrsa_oldsnapname, tx, B_FALSE)); mutex_enter(&ds->ds_lock); (void) strcpy(ds->ds_snapname, ddrsa->ddrsa_newsnapname); mutex_exit(&ds->ds_lock); VERIFY0(zap_add(dp->dp_meta_objset, dsl_dataset_phys(hds)->ds_snapnames_zapobj, ds->ds_snapname, 8, 1, &ds->ds_object, tx)); #ifdef __FreeBSD__ #ifdef _KERNEL oldname = kmem_alloc(MAXPATHLEN, KM_SLEEP); newname = kmem_alloc(MAXPATHLEN, KM_SLEEP); snprintf(oldname, MAXPATHLEN, "%s@%s", ddrsa->ddrsa_fsname, ddrsa->ddrsa_oldsnapname); snprintf(newname, MAXPATHLEN, "%s@%s", ddrsa->ddrsa_fsname, ddrsa->ddrsa_newsnapname); zfsvfs_update_fromname(oldname, newname); zvol_rename_minors(oldname, newname); kmem_free(newname, MAXPATHLEN); kmem_free(oldname, MAXPATHLEN); #endif #endif dsl_dataset_rele(ds, FTAG); return (0); } static void dsl_dataset_rename_snapshot_sync(void *arg, dmu_tx_t *tx) { dsl_dataset_rename_snapshot_arg_t *ddrsa = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dataset_t *hds; VERIFY0(dsl_dataset_hold(dp, ddrsa->ddrsa_fsname, FTAG, &hds)); ddrsa->ddrsa_tx = tx; if (ddrsa->ddrsa_recursive) { VERIFY0(dmu_objset_find_dp(dp, hds->ds_dir->dd_object, dsl_dataset_rename_snapshot_sync_impl, ddrsa, DS_FIND_CHILDREN)); } else { VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp, hds, ddrsa)); } dsl_dataset_rele(hds, FTAG); } int dsl_dataset_rename_snapshot(const char *fsname, const char *oldsnapname, const char *newsnapname, boolean_t recursive) { dsl_dataset_rename_snapshot_arg_t ddrsa; ddrsa.ddrsa_fsname = fsname; ddrsa.ddrsa_oldsnapname = oldsnapname; ddrsa.ddrsa_newsnapname = newsnapname; ddrsa.ddrsa_recursive = recursive; return (dsl_sync_task(fsname, dsl_dataset_rename_snapshot_check, dsl_dataset_rename_snapshot_sync, &ddrsa, 1, ZFS_SPACE_CHECK_RESERVED)); } /* * If we're doing an ownership handoff, we need to make sure that there is * only one long hold on the dataset. We're not allowed to change anything here * so we don't permanently release the long hold or regular hold here. We want * to do this only when syncing to avoid the dataset unexpectedly going away * when we release the long hold. */ static int dsl_dataset_handoff_check(dsl_dataset_t *ds, void *owner, dmu_tx_t *tx) { boolean_t held; if (!dmu_tx_is_syncing(tx)) return (0); if (owner != NULL) { VERIFY3P(ds->ds_owner, ==, owner); dsl_dataset_long_rele(ds, owner); } held = dsl_dataset_long_held(ds); if (owner != NULL) dsl_dataset_long_hold(ds, owner); if (held) return (SET_ERROR(EBUSY)); return (0); } typedef struct dsl_dataset_rollback_arg { const char *ddra_fsname; void *ddra_owner; nvlist_t *ddra_result; } dsl_dataset_rollback_arg_t; static int dsl_dataset_rollback_check(void *arg, dmu_tx_t *tx) { dsl_dataset_rollback_arg_t *ddra = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dataset_t *ds; int64_t unused_refres_delta; int error; error = dsl_dataset_hold(dp, ddra->ddra_fsname, FTAG, &ds); if (error != 0) return (error); /* must not be a snapshot */ if (ds->ds_is_snapshot) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EINVAL)); } /* must have a most recent snapshot */ if (dsl_dataset_phys(ds)->ds_prev_snap_txg < TXG_INITIAL) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EINVAL)); } /* must not have any bookmarks after the most recent snapshot */ nvlist_t *proprequest = fnvlist_alloc(); fnvlist_add_boolean(proprequest, zfs_prop_to_name(ZFS_PROP_CREATETXG)); nvlist_t *bookmarks = fnvlist_alloc(); error = dsl_get_bookmarks_impl(ds, proprequest, bookmarks); fnvlist_free(proprequest); if (error != 0) return (error); for (nvpair_t *pair = nvlist_next_nvpair(bookmarks, NULL); pair != NULL; pair = nvlist_next_nvpair(bookmarks, pair)) { nvlist_t *valuenv = fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair), zfs_prop_to_name(ZFS_PROP_CREATETXG)); uint64_t createtxg = fnvlist_lookup_uint64(valuenv, "value"); if (createtxg > dsl_dataset_phys(ds)->ds_prev_snap_txg) { fnvlist_free(bookmarks); dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EEXIST)); } } fnvlist_free(bookmarks); error = dsl_dataset_handoff_check(ds, ddra->ddra_owner, tx); if (error != 0) { dsl_dataset_rele(ds, FTAG); return (error); } /* * Check if the snap we are rolling back to uses more than * the refquota. */ if (ds->ds_quota != 0 && dsl_dataset_phys(ds->ds_prev)->ds_referenced_bytes > ds->ds_quota) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EDQUOT)); } /* * When we do the clone swap, we will temporarily use more space * due to the refreservation (the head will no longer have any * unique space, so the entire amount of the refreservation will need * to be free). We will immediately destroy the clone, freeing * this space, but the freeing happens over many txg's. */ unused_refres_delta = (int64_t)MIN(ds->ds_reserved, dsl_dataset_phys(ds)->ds_unique_bytes); if (unused_refres_delta > 0 && unused_refres_delta > dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE)) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(ENOSPC)); } dsl_dataset_rele(ds, FTAG); return (0); } static void dsl_dataset_rollback_sync(void *arg, dmu_tx_t *tx) { dsl_dataset_rollback_arg_t *ddra = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dataset_t *ds, *clone; uint64_t cloneobj; char namebuf[ZFS_MAXNAMELEN]; VERIFY0(dsl_dataset_hold(dp, ddra->ddra_fsname, FTAG, &ds)); dsl_dataset_name(ds->ds_prev, namebuf); fnvlist_add_string(ddra->ddra_result, "target", namebuf); cloneobj = dsl_dataset_create_sync(ds->ds_dir, "%rollback", ds->ds_prev, DS_CREATE_FLAG_NODIRTY, kcred, tx); VERIFY0(dsl_dataset_hold_obj(dp, cloneobj, FTAG, &clone)); dsl_dataset_clone_swap_sync_impl(clone, ds, tx); dsl_dataset_zero_zil(ds, tx); dsl_destroy_head_sync_impl(clone, tx); dsl_dataset_rele(clone, FTAG); dsl_dataset_rele(ds, FTAG); } /* * Rolls back the given filesystem or volume to the most recent snapshot. * The name of the most recent snapshot will be returned under key "target" * in the result nvlist. * * If owner != NULL: * - The existing dataset MUST be owned by the specified owner at entry * - Upon return, dataset will still be held by the same owner, whether we * succeed or not. * * This mode is required any time the existing filesystem is mounted. See * notes above zfs_suspend_fs() for further details. */ int dsl_dataset_rollback(const char *fsname, void *owner, nvlist_t *result) { dsl_dataset_rollback_arg_t ddra; ddra.ddra_fsname = fsname; ddra.ddra_owner = owner; ddra.ddra_result = result; return (dsl_sync_task(fsname, dsl_dataset_rollback_check, dsl_dataset_rollback_sync, &ddra, 1, ZFS_SPACE_CHECK_RESERVED)); } struct promotenode { list_node_t link; dsl_dataset_t *ds; }; typedef struct dsl_dataset_promote_arg { const char *ddpa_clonename; dsl_dataset_t *ddpa_clone; list_t shared_snaps, origin_snaps, clone_snaps; dsl_dataset_t *origin_origin; /* origin of the origin */ uint64_t used, comp, uncomp, unique, cloneusedsnap, originusedsnap; char *err_ds; cred_t *cr; } dsl_dataset_promote_arg_t; static int snaplist_space(list_t *l, uint64_t mintxg, uint64_t *spacep); static int promote_hold(dsl_dataset_promote_arg_t *ddpa, dsl_pool_t *dp, void *tag); static void promote_rele(dsl_dataset_promote_arg_t *ddpa, void *tag); static int dsl_dataset_promote_check(void *arg, dmu_tx_t *tx) { dsl_dataset_promote_arg_t *ddpa = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dataset_t *hds; struct promotenode *snap; dsl_dataset_t *origin_ds; int err; uint64_t unused; uint64_t ss_mv_cnt; size_t max_snap_len; err = promote_hold(ddpa, dp, FTAG); if (err != 0) return (err); hds = ddpa->ddpa_clone; max_snap_len = MAXNAMELEN - strlen(ddpa->ddpa_clonename) - 1; if (dsl_dataset_phys(hds)->ds_flags & DS_FLAG_NOPROMOTE) { promote_rele(ddpa, FTAG); return (SET_ERROR(EXDEV)); } /* * Compute and check the amount of space to transfer. Since this is * so expensive, don't do the preliminary check. */ if (!dmu_tx_is_syncing(tx)) { promote_rele(ddpa, FTAG); return (0); } snap = list_head(&ddpa->shared_snaps); origin_ds = snap->ds; /* compute origin's new unique space */ snap = list_tail(&ddpa->clone_snaps); ASSERT3U(dsl_dataset_phys(snap->ds)->ds_prev_snap_obj, ==, origin_ds->ds_object); dsl_deadlist_space_range(&snap->ds->ds_deadlist, dsl_dataset_phys(origin_ds)->ds_prev_snap_txg, UINT64_MAX, &ddpa->unique, &unused, &unused); /* * Walk the snapshots that we are moving * * Compute space to transfer. Consider the incremental changes * to used by each snapshot: * (my used) = (prev's used) + (blocks born) - (blocks killed) * So each snapshot gave birth to: * (blocks born) = (my used) - (prev's used) + (blocks killed) * So a sequence would look like: * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0) * Which simplifies to: * uN + kN + kN-1 + ... + k1 + k0 * Note however, if we stop before we reach the ORIGIN we get: * uN + kN + kN-1 + ... + kM - uM-1 */ ss_mv_cnt = 0; ddpa->used = dsl_dataset_phys(origin_ds)->ds_referenced_bytes; ddpa->comp = dsl_dataset_phys(origin_ds)->ds_compressed_bytes; ddpa->uncomp = dsl_dataset_phys(origin_ds)->ds_uncompressed_bytes; for (snap = list_head(&ddpa->shared_snaps); snap; snap = list_next(&ddpa->shared_snaps, snap)) { uint64_t val, dlused, dlcomp, dluncomp; dsl_dataset_t *ds = snap->ds; ss_mv_cnt++; /* * If there are long holds, we won't be able to evict * the objset. */ if (dsl_dataset_long_held(ds)) { err = SET_ERROR(EBUSY); goto out; } /* Check that the snapshot name does not conflict */ VERIFY0(dsl_dataset_get_snapname(ds)); if (strlen(ds->ds_snapname) >= max_snap_len) { err = SET_ERROR(ENAMETOOLONG); goto out; } err = dsl_dataset_snap_lookup(hds, ds->ds_snapname, &val); if (err == 0) { (void) strcpy(ddpa->err_ds, snap->ds->ds_snapname); err = SET_ERROR(EEXIST); goto out; } if (err != ENOENT) goto out; /* The very first snapshot does not have a deadlist */ if (dsl_dataset_phys(ds)->ds_prev_snap_obj == 0) continue; dsl_deadlist_space(&ds->ds_deadlist, &dlused, &dlcomp, &dluncomp); ddpa->used += dlused; ddpa->comp += dlcomp; ddpa->uncomp += dluncomp; } /* * If we are a clone of a clone then we never reached ORIGIN, * so we need to subtract out the clone origin's used space. */ if (ddpa->origin_origin) { ddpa->used -= dsl_dataset_phys(ddpa->origin_origin)->ds_referenced_bytes; ddpa->comp -= dsl_dataset_phys(ddpa->origin_origin)->ds_compressed_bytes; ddpa->uncomp -= dsl_dataset_phys(ddpa->origin_origin)-> ds_uncompressed_bytes; } /* Check that there is enough space and limit headroom here */ err = dsl_dir_transfer_possible(origin_ds->ds_dir, hds->ds_dir, 0, ss_mv_cnt, ddpa->used, ddpa->cr); if (err != 0) goto out; /* * Compute the amounts of space that will be used by snapshots * after the promotion (for both origin and clone). For each, * it is the amount of space that will be on all of their * deadlists (that was not born before their new origin). */ if (dsl_dir_phys(hds->ds_dir)->dd_flags & DD_FLAG_USED_BREAKDOWN) { uint64_t space; /* * Note, typically this will not be a clone of a clone, * so dd_origin_txg will be < TXG_INITIAL, so * these snaplist_space() -> dsl_deadlist_space_range() * calls will be fast because they do not have to * iterate over all bps. */ snap = list_head(&ddpa->origin_snaps); err = snaplist_space(&ddpa->shared_snaps, snap->ds->ds_dir->dd_origin_txg, &ddpa->cloneusedsnap); if (err != 0) goto out; err = snaplist_space(&ddpa->clone_snaps, snap->ds->ds_dir->dd_origin_txg, &space); if (err != 0) goto out; ddpa->cloneusedsnap += space; } if (dsl_dir_phys(origin_ds->ds_dir)->dd_flags & DD_FLAG_USED_BREAKDOWN) { err = snaplist_space(&ddpa->origin_snaps, dsl_dataset_phys(origin_ds)->ds_creation_txg, &ddpa->originusedsnap); if (err != 0) goto out; } out: promote_rele(ddpa, FTAG); return (err); } static void dsl_dataset_promote_sync(void *arg, dmu_tx_t *tx) { dsl_dataset_promote_arg_t *ddpa = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dataset_t *hds; struct promotenode *snap; dsl_dataset_t *origin_ds; dsl_dataset_t *origin_head; dsl_dir_t *dd; dsl_dir_t *odd = NULL; uint64_t oldnext_obj; int64_t delta; #if defined(__FreeBSD__) && defined(_KERNEL) char *oldname, *newname; #endif VERIFY0(promote_hold(ddpa, dp, FTAG)); hds = ddpa->ddpa_clone; ASSERT0(dsl_dataset_phys(hds)->ds_flags & DS_FLAG_NOPROMOTE); snap = list_head(&ddpa->shared_snaps); origin_ds = snap->ds; dd = hds->ds_dir; snap = list_head(&ddpa->origin_snaps); origin_head = snap->ds; /* * We need to explicitly open odd, since origin_ds's dd will be * changing. */ VERIFY0(dsl_dir_hold_obj(dp, origin_ds->ds_dir->dd_object, NULL, FTAG, &odd)); /* change origin's next snap */ dmu_buf_will_dirty(origin_ds->ds_dbuf, tx); oldnext_obj = dsl_dataset_phys(origin_ds)->ds_next_snap_obj; snap = list_tail(&ddpa->clone_snaps); ASSERT3U(dsl_dataset_phys(snap->ds)->ds_prev_snap_obj, ==, origin_ds->ds_object); dsl_dataset_phys(origin_ds)->ds_next_snap_obj = snap->ds->ds_object; /* change the origin's next clone */ if (dsl_dataset_phys(origin_ds)->ds_next_clones_obj) { dsl_dataset_remove_from_next_clones(origin_ds, snap->ds->ds_object, tx); VERIFY0(zap_add_int(dp->dp_meta_objset, dsl_dataset_phys(origin_ds)->ds_next_clones_obj, oldnext_obj, tx)); } /* change origin */ dmu_buf_will_dirty(dd->dd_dbuf, tx); ASSERT3U(dsl_dir_phys(dd)->dd_origin_obj, ==, origin_ds->ds_object); dsl_dir_phys(dd)->dd_origin_obj = dsl_dir_phys(odd)->dd_origin_obj; dd->dd_origin_txg = origin_head->ds_dir->dd_origin_txg; dmu_buf_will_dirty(odd->dd_dbuf, tx); dsl_dir_phys(odd)->dd_origin_obj = origin_ds->ds_object; origin_head->ds_dir->dd_origin_txg = dsl_dataset_phys(origin_ds)->ds_creation_txg; /* change dd_clone entries */ if (spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) { VERIFY0(zap_remove_int(dp->dp_meta_objset, dsl_dir_phys(odd)->dd_clones, hds->ds_object, tx)); VERIFY0(zap_add_int(dp->dp_meta_objset, dsl_dir_phys(ddpa->origin_origin->ds_dir)->dd_clones, hds->ds_object, tx)); VERIFY0(zap_remove_int(dp->dp_meta_objset, dsl_dir_phys(ddpa->origin_origin->ds_dir)->dd_clones, origin_head->ds_object, tx)); if (dsl_dir_phys(dd)->dd_clones == 0) { dsl_dir_phys(dd)->dd_clones = zap_create(dp->dp_meta_objset, DMU_OT_DSL_CLONES, DMU_OT_NONE, 0, tx); } VERIFY0(zap_add_int(dp->dp_meta_objset, dsl_dir_phys(dd)->dd_clones, origin_head->ds_object, tx)); } #if defined(__FreeBSD__) && defined(_KERNEL) /* Take the spa_namespace_lock early so zvol renames don't deadlock. */ mutex_enter(&spa_namespace_lock); oldname = kmem_alloc(MAXPATHLEN, KM_SLEEP); newname = kmem_alloc(MAXPATHLEN, KM_SLEEP); #endif /* move snapshots to this dir */ for (snap = list_head(&ddpa->shared_snaps); snap; snap = list_next(&ddpa->shared_snaps, snap)) { dsl_dataset_t *ds = snap->ds; /* * Property callbacks are registered to a particular * dsl_dir. Since ours is changing, evict the objset * so that they will be unregistered from the old dsl_dir. */ if (ds->ds_objset) { dmu_objset_evict(ds->ds_objset); ds->ds_objset = NULL; } /* move snap name entry */ VERIFY0(dsl_dataset_get_snapname(ds)); VERIFY0(dsl_dataset_snap_remove(origin_head, ds->ds_snapname, tx, B_TRUE)); VERIFY0(zap_add(dp->dp_meta_objset, dsl_dataset_phys(hds)->ds_snapnames_zapobj, ds->ds_snapname, 8, 1, &ds->ds_object, tx)); dsl_fs_ss_count_adjust(hds->ds_dir, 1, DD_FIELD_SNAPSHOT_COUNT, tx); /* change containing dsl_dir */ dmu_buf_will_dirty(ds->ds_dbuf, tx); ASSERT3U(dsl_dataset_phys(ds)->ds_dir_obj, ==, odd->dd_object); dsl_dataset_phys(ds)->ds_dir_obj = dd->dd_object; ASSERT3P(ds->ds_dir, ==, odd); dsl_dir_rele(ds->ds_dir, ds); VERIFY0(dsl_dir_hold_obj(dp, dd->dd_object, NULL, ds, &ds->ds_dir)); #if defined(__FreeBSD__) && defined(_KERNEL) dsl_dataset_name(ds, newname); zfsvfs_update_fromname(oldname, newname); zvol_rename_minors(oldname, newname); #endif /* move any clone references */ if (dsl_dataset_phys(ds)->ds_next_clones_obj && spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) { zap_cursor_t zc; zap_attribute_t za; for (zap_cursor_init(&zc, dp->dp_meta_objset, dsl_dataset_phys(ds)->ds_next_clones_obj); zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { dsl_dataset_t *cnds; uint64_t o; if (za.za_first_integer == oldnext_obj) { /* * We've already moved the * origin's reference. */ continue; } VERIFY0(dsl_dataset_hold_obj(dp, za.za_first_integer, FTAG, &cnds)); o = dsl_dir_phys(cnds->ds_dir)-> dd_head_dataset_obj; VERIFY0(zap_remove_int(dp->dp_meta_objset, dsl_dir_phys(odd)->dd_clones, o, tx)); VERIFY0(zap_add_int(dp->dp_meta_objset, dsl_dir_phys(dd)->dd_clones, o, tx)); dsl_dataset_rele(cnds, FTAG); } zap_cursor_fini(&zc); } ASSERT(!dsl_prop_hascb(ds)); } #if defined(__FreeBSD__) && defined(_KERNEL) mutex_exit(&spa_namespace_lock); kmem_free(newname, MAXPATHLEN); kmem_free(oldname, MAXPATHLEN); #endif /* * Change space accounting. * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either * both be valid, or both be 0 (resulting in delta == 0). This * is true for each of {clone,origin} independently. */ delta = ddpa->cloneusedsnap - dsl_dir_phys(dd)->dd_used_breakdown[DD_USED_SNAP]; ASSERT3S(delta, >=, 0); ASSERT3U(ddpa->used, >=, delta); dsl_dir_diduse_space(dd, DD_USED_SNAP, delta, 0, 0, tx); dsl_dir_diduse_space(dd, DD_USED_HEAD, ddpa->used - delta, ddpa->comp, ddpa->uncomp, tx); delta = ddpa->originusedsnap - dsl_dir_phys(odd)->dd_used_breakdown[DD_USED_SNAP]; ASSERT3S(delta, <=, 0); ASSERT3U(ddpa->used, >=, -delta); dsl_dir_diduse_space(odd, DD_USED_SNAP, delta, 0, 0, tx); dsl_dir_diduse_space(odd, DD_USED_HEAD, -ddpa->used - delta, -ddpa->comp, -ddpa->uncomp, tx); dsl_dataset_phys(origin_ds)->ds_unique_bytes = ddpa->unique; /* log history record */ spa_history_log_internal_ds(hds, "promote", tx, ""); dsl_dir_rele(odd, FTAG); promote_rele(ddpa, FTAG); } /* * Make a list of dsl_dataset_t's for the snapshots between first_obj * (exclusive) and last_obj (inclusive). The list will be in reverse * order (last_obj will be the list_head()). If first_obj == 0, do all * snapshots back to this dataset's origin. */ static int snaplist_make(dsl_pool_t *dp, uint64_t first_obj, uint64_t last_obj, list_t *l, void *tag) { uint64_t obj = last_obj; list_create(l, sizeof (struct promotenode), offsetof(struct promotenode, link)); while (obj != first_obj) { dsl_dataset_t *ds; struct promotenode *snap; int err; err = dsl_dataset_hold_obj(dp, obj, tag, &ds); ASSERT(err != ENOENT); if (err != 0) return (err); if (first_obj == 0) first_obj = dsl_dir_phys(ds->ds_dir)->dd_origin_obj; snap = kmem_alloc(sizeof (*snap), KM_SLEEP); snap->ds = ds; list_insert_tail(l, snap); obj = dsl_dataset_phys(ds)->ds_prev_snap_obj; } return (0); } static int snaplist_space(list_t *l, uint64_t mintxg, uint64_t *spacep) { struct promotenode *snap; *spacep = 0; for (snap = list_head(l); snap; snap = list_next(l, snap)) { uint64_t used, comp, uncomp; dsl_deadlist_space_range(&snap->ds->ds_deadlist, mintxg, UINT64_MAX, &used, &comp, &uncomp); *spacep += used; } return (0); } static void snaplist_destroy(list_t *l, void *tag) { struct promotenode *snap; if (l == NULL || !list_link_active(&l->list_head)) return; while ((snap = list_tail(l)) != NULL) { list_remove(l, snap); dsl_dataset_rele(snap->ds, tag); kmem_free(snap, sizeof (*snap)); } list_destroy(l); } static int promote_hold(dsl_dataset_promote_arg_t *ddpa, dsl_pool_t *dp, void *tag) { int error; dsl_dir_t *dd; struct promotenode *snap; error = dsl_dataset_hold(dp, ddpa->ddpa_clonename, tag, &ddpa->ddpa_clone); if (error != 0) return (error); dd = ddpa->ddpa_clone->ds_dir; if (ddpa->ddpa_clone->ds_is_snapshot || !dsl_dir_is_clone(dd)) { dsl_dataset_rele(ddpa->ddpa_clone, tag); return (SET_ERROR(EINVAL)); } error = snaplist_make(dp, 0, dsl_dir_phys(dd)->dd_origin_obj, &ddpa->shared_snaps, tag); if (error != 0) goto out; error = snaplist_make(dp, 0, ddpa->ddpa_clone->ds_object, &ddpa->clone_snaps, tag); if (error != 0) goto out; snap = list_head(&ddpa->shared_snaps); ASSERT3U(snap->ds->ds_object, ==, dsl_dir_phys(dd)->dd_origin_obj); error = snaplist_make(dp, dsl_dir_phys(dd)->dd_origin_obj, dsl_dir_phys(snap->ds->ds_dir)->dd_head_dataset_obj, &ddpa->origin_snaps, tag); if (error != 0) goto out; if (dsl_dir_phys(snap->ds->ds_dir)->dd_origin_obj != 0) { error = dsl_dataset_hold_obj(dp, dsl_dir_phys(snap->ds->ds_dir)->dd_origin_obj, tag, &ddpa->origin_origin); if (error != 0) goto out; } out: if (error != 0) promote_rele(ddpa, tag); return (error); } static void promote_rele(dsl_dataset_promote_arg_t *ddpa, void *tag) { snaplist_destroy(&ddpa->shared_snaps, tag); snaplist_destroy(&ddpa->clone_snaps, tag); snaplist_destroy(&ddpa->origin_snaps, tag); if (ddpa->origin_origin != NULL) dsl_dataset_rele(ddpa->origin_origin, tag); dsl_dataset_rele(ddpa->ddpa_clone, tag); } /* * Promote a clone. * * If it fails due to a conflicting snapshot name, "conflsnap" will be filled * in with the name. (It must be at least MAXNAMELEN bytes long.) */ int dsl_dataset_promote(const char *name, char *conflsnap) { dsl_dataset_promote_arg_t ddpa = { 0 }; uint64_t numsnaps; int error; objset_t *os; /* * We will modify space proportional to the number of * snapshots. Compute numsnaps. */ error = dmu_objset_hold(name, FTAG, &os); if (error != 0) return (error); error = zap_count(dmu_objset_pool(os)->dp_meta_objset, dsl_dataset_phys(dmu_objset_ds(os))->ds_snapnames_zapobj, &numsnaps); dmu_objset_rele(os, FTAG); if (error != 0) return (error); ddpa.ddpa_clonename = name; ddpa.err_ds = conflsnap; ddpa.cr = CRED(); return (dsl_sync_task(name, dsl_dataset_promote_check, dsl_dataset_promote_sync, &ddpa, 2 + numsnaps, ZFS_SPACE_CHECK_RESERVED)); } int dsl_dataset_clone_swap_check_impl(dsl_dataset_t *clone, dsl_dataset_t *origin_head, boolean_t force, void *owner, dmu_tx_t *tx) { int64_t unused_refres_delta; /* they should both be heads */ if (clone->ds_is_snapshot || origin_head->ds_is_snapshot) return (SET_ERROR(EINVAL)); /* if we are not forcing, the branch point should be just before them */ if (!force && clone->ds_prev != origin_head->ds_prev) return (SET_ERROR(EINVAL)); /* clone should be the clone (unless they are unrelated) */ if (clone->ds_prev != NULL && clone->ds_prev != clone->ds_dir->dd_pool->dp_origin_snap && origin_head->ds_dir != clone->ds_prev->ds_dir) return (SET_ERROR(EINVAL)); /* the clone should be a child of the origin */ if (clone->ds_dir->dd_parent != origin_head->ds_dir) return (SET_ERROR(EINVAL)); /* origin_head shouldn't be modified unless 'force' */ if (!force && dsl_dataset_modified_since_snap(origin_head, origin_head->ds_prev)) return (SET_ERROR(ETXTBSY)); /* origin_head should have no long holds (e.g. is not mounted) */ if (dsl_dataset_handoff_check(origin_head, owner, tx)) return (SET_ERROR(EBUSY)); /* check amount of any unconsumed refreservation */ unused_refres_delta = (int64_t)MIN(origin_head->ds_reserved, dsl_dataset_phys(origin_head)->ds_unique_bytes) - (int64_t)MIN(origin_head->ds_reserved, dsl_dataset_phys(clone)->ds_unique_bytes); if (unused_refres_delta > 0 && unused_refres_delta > dsl_dir_space_available(origin_head->ds_dir, NULL, 0, TRUE)) return (SET_ERROR(ENOSPC)); /* clone can't be over the head's refquota */ if (origin_head->ds_quota != 0 && dsl_dataset_phys(clone)->ds_referenced_bytes > origin_head->ds_quota) return (SET_ERROR(EDQUOT)); return (0); } void dsl_dataset_clone_swap_sync_impl(dsl_dataset_t *clone, dsl_dataset_t *origin_head, dmu_tx_t *tx) { dsl_pool_t *dp = dmu_tx_pool(tx); int64_t unused_refres_delta; ASSERT(clone->ds_reserved == 0); ASSERT(origin_head->ds_quota == 0 || dsl_dataset_phys(clone)->ds_unique_bytes <= origin_head->ds_quota); ASSERT3P(clone->ds_prev, ==, origin_head->ds_prev); /* * Swap per-dataset feature flags. */ for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { if (!(spa_feature_table[f].fi_flags & ZFEATURE_FLAG_PER_DATASET)) { ASSERT(!clone->ds_feature_inuse[f]); ASSERT(!origin_head->ds_feature_inuse[f]); continue; } boolean_t clone_inuse = clone->ds_feature_inuse[f]; boolean_t origin_head_inuse = origin_head->ds_feature_inuse[f]; if (clone_inuse) { dsl_dataset_deactivate_feature(clone->ds_object, f, tx); clone->ds_feature_inuse[f] = B_FALSE; } if (origin_head_inuse) { dsl_dataset_deactivate_feature(origin_head->ds_object, f, tx); origin_head->ds_feature_inuse[f] = B_FALSE; } if (clone_inuse) { dsl_dataset_activate_feature(origin_head->ds_object, f, tx); origin_head->ds_feature_inuse[f] = B_TRUE; } if (origin_head_inuse) { dsl_dataset_activate_feature(clone->ds_object, f, tx); clone->ds_feature_inuse[f] = B_TRUE; } } dmu_buf_will_dirty(clone->ds_dbuf, tx); dmu_buf_will_dirty(origin_head->ds_dbuf, tx); if (clone->ds_objset != NULL) { dmu_objset_evict(clone->ds_objset); clone->ds_objset = NULL; } if (origin_head->ds_objset != NULL) { dmu_objset_evict(origin_head->ds_objset); origin_head->ds_objset = NULL; } unused_refres_delta = (int64_t)MIN(origin_head->ds_reserved, dsl_dataset_phys(origin_head)->ds_unique_bytes) - (int64_t)MIN(origin_head->ds_reserved, dsl_dataset_phys(clone)->ds_unique_bytes); /* * Reset origin's unique bytes, if it exists. */ if (clone->ds_prev) { dsl_dataset_t *origin = clone->ds_prev; uint64_t comp, uncomp; dmu_buf_will_dirty(origin->ds_dbuf, tx); dsl_deadlist_space_range(&clone->ds_deadlist, dsl_dataset_phys(origin)->ds_prev_snap_txg, UINT64_MAX, &dsl_dataset_phys(origin)->ds_unique_bytes, &comp, &uncomp); } /* swap blkptrs */ { blkptr_t tmp; tmp = dsl_dataset_phys(origin_head)->ds_bp; dsl_dataset_phys(origin_head)->ds_bp = dsl_dataset_phys(clone)->ds_bp; dsl_dataset_phys(clone)->ds_bp = tmp; } /* set dd_*_bytes */ { int64_t dused, dcomp, duncomp; uint64_t cdl_used, cdl_comp, cdl_uncomp; uint64_t odl_used, odl_comp, odl_uncomp; ASSERT3U(dsl_dir_phys(clone->ds_dir)-> dd_used_breakdown[DD_USED_SNAP], ==, 0); dsl_deadlist_space(&clone->ds_deadlist, &cdl_used, &cdl_comp, &cdl_uncomp); dsl_deadlist_space(&origin_head->ds_deadlist, &odl_used, &odl_comp, &odl_uncomp); dused = dsl_dataset_phys(clone)->ds_referenced_bytes + cdl_used - (dsl_dataset_phys(origin_head)->ds_referenced_bytes + odl_used); dcomp = dsl_dataset_phys(clone)->ds_compressed_bytes + cdl_comp - (dsl_dataset_phys(origin_head)->ds_compressed_bytes + odl_comp); duncomp = dsl_dataset_phys(clone)->ds_uncompressed_bytes + cdl_uncomp - (dsl_dataset_phys(origin_head)->ds_uncompressed_bytes + odl_uncomp); dsl_dir_diduse_space(origin_head->ds_dir, DD_USED_HEAD, dused, dcomp, duncomp, tx); dsl_dir_diduse_space(clone->ds_dir, DD_USED_HEAD, -dused, -dcomp, -duncomp, tx); /* * The difference in the space used by snapshots is the * difference in snapshot space due to the head's * deadlist (since that's the only thing that's * changing that affects the snapused). */ dsl_deadlist_space_range(&clone->ds_deadlist, origin_head->ds_dir->dd_origin_txg, UINT64_MAX, &cdl_used, &cdl_comp, &cdl_uncomp); dsl_deadlist_space_range(&origin_head->ds_deadlist, origin_head->ds_dir->dd_origin_txg, UINT64_MAX, &odl_used, &odl_comp, &odl_uncomp); dsl_dir_transfer_space(origin_head->ds_dir, cdl_used - odl_used, DD_USED_HEAD, DD_USED_SNAP, NULL); } /* swap ds_*_bytes */ SWITCH64(dsl_dataset_phys(origin_head)->ds_referenced_bytes, dsl_dataset_phys(clone)->ds_referenced_bytes); SWITCH64(dsl_dataset_phys(origin_head)->ds_compressed_bytes, dsl_dataset_phys(clone)->ds_compressed_bytes); SWITCH64(dsl_dataset_phys(origin_head)->ds_uncompressed_bytes, dsl_dataset_phys(clone)->ds_uncompressed_bytes); SWITCH64(dsl_dataset_phys(origin_head)->ds_unique_bytes, dsl_dataset_phys(clone)->ds_unique_bytes); /* apply any parent delta for change in unconsumed refreservation */ dsl_dir_diduse_space(origin_head->ds_dir, DD_USED_REFRSRV, unused_refres_delta, 0, 0, tx); /* * Swap deadlists. */ dsl_deadlist_close(&clone->ds_deadlist); dsl_deadlist_close(&origin_head->ds_deadlist); SWITCH64(dsl_dataset_phys(origin_head)->ds_deadlist_obj, dsl_dataset_phys(clone)->ds_deadlist_obj); dsl_deadlist_open(&clone->ds_deadlist, dp->dp_meta_objset, dsl_dataset_phys(clone)->ds_deadlist_obj); dsl_deadlist_open(&origin_head->ds_deadlist, dp->dp_meta_objset, dsl_dataset_phys(origin_head)->ds_deadlist_obj); dsl_scan_ds_clone_swapped(origin_head, clone, tx); spa_history_log_internal_ds(clone, "clone swap", tx, "parent=%s", origin_head->ds_dir->dd_myname); } /* * Given a pool name and a dataset object number in that pool, * return the name of that dataset. */ int dsl_dsobj_to_dsname(char *pname, uint64_t obj, char *buf) { dsl_pool_t *dp; dsl_dataset_t *ds; int error; error = dsl_pool_hold(pname, FTAG, &dp); if (error != 0) return (error); error = dsl_dataset_hold_obj(dp, obj, FTAG, &ds); if (error == 0) { dsl_dataset_name(ds, buf); dsl_dataset_rele(ds, FTAG); } dsl_pool_rele(dp, FTAG); return (error); } int dsl_dataset_check_quota(dsl_dataset_t *ds, boolean_t check_quota, uint64_t asize, uint64_t inflight, uint64_t *used, uint64_t *ref_rsrv) { int error = 0; ASSERT3S(asize, >, 0); /* * *ref_rsrv is the portion of asize that will come from any * unconsumed refreservation space. */ *ref_rsrv = 0; mutex_enter(&ds->ds_lock); /* * Make a space adjustment for reserved bytes. */ if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes) { ASSERT3U(*used, >=, ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes); *used -= (ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes); *ref_rsrv = asize - MIN(asize, parent_delta(ds, asize + inflight)); } if (!check_quota || ds->ds_quota == 0) { mutex_exit(&ds->ds_lock); return (0); } /* * If they are requesting more space, and our current estimate * is over quota, they get to try again unless the actual * on-disk is over quota and there are no pending changes (which * may free up space for us). */ if (dsl_dataset_phys(ds)->ds_referenced_bytes + inflight >= ds->ds_quota) { if (inflight > 0 || dsl_dataset_phys(ds)->ds_referenced_bytes < ds->ds_quota) error = SET_ERROR(ERESTART); else error = SET_ERROR(EDQUOT); } mutex_exit(&ds->ds_lock); return (error); } typedef struct dsl_dataset_set_qr_arg { const char *ddsqra_name; zprop_source_t ddsqra_source; uint64_t ddsqra_value; } dsl_dataset_set_qr_arg_t; /* ARGSUSED */ static int dsl_dataset_set_refquota_check(void *arg, dmu_tx_t *tx) { dsl_dataset_set_qr_arg_t *ddsqra = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dataset_t *ds; int error; uint64_t newval; if (spa_version(dp->dp_spa) < SPA_VERSION_REFQUOTA) return (SET_ERROR(ENOTSUP)); error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds); if (error != 0) return (error); if (ds->ds_is_snapshot) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EINVAL)); } error = dsl_prop_predict(ds->ds_dir, zfs_prop_to_name(ZFS_PROP_REFQUOTA), ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval); if (error != 0) { dsl_dataset_rele(ds, FTAG); return (error); } if (newval == 0) { dsl_dataset_rele(ds, FTAG); return (0); } if (newval < dsl_dataset_phys(ds)->ds_referenced_bytes || newval < ds->ds_reserved) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(ENOSPC)); } dsl_dataset_rele(ds, FTAG); return (0); } static void dsl_dataset_set_refquota_sync(void *arg, dmu_tx_t *tx) { dsl_dataset_set_qr_arg_t *ddsqra = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dataset_t *ds; uint64_t newval; VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds)); dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_REFQUOTA), ddsqra->ddsqra_source, sizeof (ddsqra->ddsqra_value), 1, &ddsqra->ddsqra_value, tx); VERIFY0(dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_REFQUOTA), &newval)); if (ds->ds_quota != newval) { dmu_buf_will_dirty(ds->ds_dbuf, tx); ds->ds_quota = newval; } dsl_dataset_rele(ds, FTAG); } int dsl_dataset_set_refquota(const char *dsname, zprop_source_t source, uint64_t refquota) { dsl_dataset_set_qr_arg_t ddsqra; ddsqra.ddsqra_name = dsname; ddsqra.ddsqra_source = source; ddsqra.ddsqra_value = refquota; return (dsl_sync_task(dsname, dsl_dataset_set_refquota_check, dsl_dataset_set_refquota_sync, &ddsqra, 0, ZFS_SPACE_CHECK_NONE)); } static int dsl_dataset_set_refreservation_check(void *arg, dmu_tx_t *tx) { dsl_dataset_set_qr_arg_t *ddsqra = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dataset_t *ds; int error; uint64_t newval, unique; if (spa_version(dp->dp_spa) < SPA_VERSION_REFRESERVATION) return (SET_ERROR(ENOTSUP)); error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds); if (error != 0) return (error); if (ds->ds_is_snapshot) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(EINVAL)); } error = dsl_prop_predict(ds->ds_dir, zfs_prop_to_name(ZFS_PROP_REFRESERVATION), ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval); if (error != 0) { dsl_dataset_rele(ds, FTAG); return (error); } /* * If we are doing the preliminary check in open context, the * space estimates may be inaccurate. */ if (!dmu_tx_is_syncing(tx)) { dsl_dataset_rele(ds, FTAG); return (0); } mutex_enter(&ds->ds_lock); if (!DS_UNIQUE_IS_ACCURATE(ds)) dsl_dataset_recalc_head_uniq(ds); unique = dsl_dataset_phys(ds)->ds_unique_bytes; mutex_exit(&ds->ds_lock); if (MAX(unique, newval) > MAX(unique, ds->ds_reserved)) { uint64_t delta = MAX(unique, newval) - MAX(unique, ds->ds_reserved); if (delta > dsl_dir_space_available(ds->ds_dir, NULL, 0, B_TRUE) || (ds->ds_quota > 0 && newval > ds->ds_quota)) { dsl_dataset_rele(ds, FTAG); return (SET_ERROR(ENOSPC)); } } dsl_dataset_rele(ds, FTAG); return (0); } void dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t *ds, zprop_source_t source, uint64_t value, dmu_tx_t *tx) { uint64_t newval; uint64_t unique; int64_t delta; dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_REFRESERVATION), source, sizeof (value), 1, &value, tx); VERIFY0(dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_REFRESERVATION), &newval)); dmu_buf_will_dirty(ds->ds_dbuf, tx); mutex_enter(&ds->ds_dir->dd_lock); mutex_enter(&ds->ds_lock); ASSERT(DS_UNIQUE_IS_ACCURATE(ds)); unique = dsl_dataset_phys(ds)->ds_unique_bytes; delta = MAX(0, (int64_t)(newval - unique)) - MAX(0, (int64_t)(ds->ds_reserved - unique)); ds->ds_reserved = newval; mutex_exit(&ds->ds_lock); dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV, delta, 0, 0, tx); mutex_exit(&ds->ds_dir->dd_lock); } static void dsl_dataset_set_refreservation_sync(void *arg, dmu_tx_t *tx) { dsl_dataset_set_qr_arg_t *ddsqra = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dataset_t *ds; VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds)); dsl_dataset_set_refreservation_sync_impl(ds, ddsqra->ddsqra_source, ddsqra->ddsqra_value, tx); dsl_dataset_rele(ds, FTAG); } int dsl_dataset_set_refreservation(const char *dsname, zprop_source_t source, uint64_t refreservation) { dsl_dataset_set_qr_arg_t ddsqra; ddsqra.ddsqra_name = dsname; ddsqra.ddsqra_source = source; ddsqra.ddsqra_value = refreservation; return (dsl_sync_task(dsname, dsl_dataset_set_refreservation_check, dsl_dataset_set_refreservation_sync, &ddsqra, 0, ZFS_SPACE_CHECK_NONE)); } /* * Return (in *usedp) the amount of space written in new that is not * present in oldsnap. New may be a snapshot or the head. Old must be * a snapshot before new, in new's filesystem (or its origin). If not then * fail and return EINVAL. * * The written space is calculated by considering two components: First, we * ignore any freed space, and calculate the written as new's used space * minus old's used space. Next, we add in the amount of space that was freed * between the two snapshots, thus reducing new's used space relative to old's. * Specifically, this is the space that was born before old->ds_creation_txg, * and freed before new (ie. on new's deadlist or a previous deadlist). * * space freed [---------------------] * snapshots ---O-------O--------O-------O------ * oldsnap new */ int dsl_dataset_space_written(dsl_dataset_t *oldsnap, dsl_dataset_t *new, uint64_t *usedp, uint64_t *compp, uint64_t *uncompp) { int err = 0; uint64_t snapobj; dsl_pool_t *dp = new->ds_dir->dd_pool; ASSERT(dsl_pool_config_held(dp)); *usedp = 0; *usedp += dsl_dataset_phys(new)->ds_referenced_bytes; *usedp -= dsl_dataset_phys(oldsnap)->ds_referenced_bytes; *compp = 0; *compp += dsl_dataset_phys(new)->ds_compressed_bytes; *compp -= dsl_dataset_phys(oldsnap)->ds_compressed_bytes; *uncompp = 0; *uncompp += dsl_dataset_phys(new)->ds_uncompressed_bytes; *uncompp -= dsl_dataset_phys(oldsnap)->ds_uncompressed_bytes; snapobj = new->ds_object; while (snapobj != oldsnap->ds_object) { dsl_dataset_t *snap; uint64_t used, comp, uncomp; if (snapobj == new->ds_object) { snap = new; } else { err = dsl_dataset_hold_obj(dp, snapobj, FTAG, &snap); if (err != 0) break; } if (dsl_dataset_phys(snap)->ds_prev_snap_txg == dsl_dataset_phys(oldsnap)->ds_creation_txg) { /* * The blocks in the deadlist can not be born after * ds_prev_snap_txg, so get the whole deadlist space, * which is more efficient (especially for old-format * deadlists). Unfortunately the deadlist code * doesn't have enough information to make this * optimization itself. */ dsl_deadlist_space(&snap->ds_deadlist, &used, &comp, &uncomp); } else { dsl_deadlist_space_range(&snap->ds_deadlist, 0, dsl_dataset_phys(oldsnap)->ds_creation_txg, &used, &comp, &uncomp); } *usedp += used; *compp += comp; *uncompp += uncomp; /* * If we get to the beginning of the chain of snapshots * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap * was not a snapshot of/before new. */ snapobj = dsl_dataset_phys(snap)->ds_prev_snap_obj; if (snap != new) dsl_dataset_rele(snap, FTAG); if (snapobj == 0) { err = SET_ERROR(EINVAL); break; } } return (err); } /* * Return (in *usedp) the amount of space that will be reclaimed if firstsnap, * lastsnap, and all snapshots in between are deleted. * * blocks that would be freed [---------------------------] * snapshots ---O-------O--------O-------O--------O * firstsnap lastsnap * * This is the set of blocks that were born after the snap before firstsnap, * (birth > firstsnap->prev_snap_txg) and died before the snap after the * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist). * We calculate this by iterating over the relevant deadlists (from the snap * after lastsnap, backward to the snap after firstsnap), summing up the * space on the deadlist that was born after the snap before firstsnap. */ int dsl_dataset_space_wouldfree(dsl_dataset_t *firstsnap, dsl_dataset_t *lastsnap, uint64_t *usedp, uint64_t *compp, uint64_t *uncompp) { int err = 0; uint64_t snapobj; dsl_pool_t *dp = firstsnap->ds_dir->dd_pool; ASSERT(firstsnap->ds_is_snapshot); ASSERT(lastsnap->ds_is_snapshot); /* * Check that the snapshots are in the same dsl_dir, and firstsnap * is before lastsnap. */ if (firstsnap->ds_dir != lastsnap->ds_dir || dsl_dataset_phys(firstsnap)->ds_creation_txg > dsl_dataset_phys(lastsnap)->ds_creation_txg) return (SET_ERROR(EINVAL)); *usedp = *compp = *uncompp = 0; snapobj = dsl_dataset_phys(lastsnap)->ds_next_snap_obj; while (snapobj != firstsnap->ds_object) { dsl_dataset_t *ds; uint64_t used, comp, uncomp; err = dsl_dataset_hold_obj(dp, snapobj, FTAG, &ds); if (err != 0) break; dsl_deadlist_space_range(&ds->ds_deadlist, dsl_dataset_phys(firstsnap)->ds_prev_snap_txg, UINT64_MAX, &used, &comp, &uncomp); *usedp += used; *compp += comp; *uncompp += uncomp; snapobj = dsl_dataset_phys(ds)->ds_prev_snap_obj; ASSERT3U(snapobj, !=, 0); dsl_dataset_rele(ds, FTAG); } return (err); } /* * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline. * For example, they could both be snapshots of the same filesystem, and * 'earlier' is before 'later'. Or 'earlier' could be the origin of * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's * filesystem. Or 'earlier' could be the origin's origin. * * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg. */ boolean_t dsl_dataset_is_before(dsl_dataset_t *later, dsl_dataset_t *earlier, uint64_t earlier_txg) { dsl_pool_t *dp = later->ds_dir->dd_pool; int error; boolean_t ret; ASSERT(dsl_pool_config_held(dp)); ASSERT(earlier->ds_is_snapshot || earlier_txg != 0); if (earlier_txg == 0) earlier_txg = dsl_dataset_phys(earlier)->ds_creation_txg; if (later->ds_is_snapshot && earlier_txg >= dsl_dataset_phys(later)->ds_creation_txg) return (B_FALSE); if (later->ds_dir == earlier->ds_dir) return (B_TRUE); if (!dsl_dir_is_clone(later->ds_dir)) return (B_FALSE); if (dsl_dir_phys(later->ds_dir)->dd_origin_obj == earlier->ds_object) return (B_TRUE); dsl_dataset_t *origin; error = dsl_dataset_hold_obj(dp, dsl_dir_phys(later->ds_dir)->dd_origin_obj, FTAG, &origin); if (error != 0) return (B_FALSE); ret = dsl_dataset_is_before(origin, earlier, earlier_txg); dsl_dataset_rele(origin, FTAG); return (ret); } void dsl_dataset_zapify(dsl_dataset_t *ds, dmu_tx_t *tx) { objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; dmu_object_zapify(mos, ds->ds_object, DMU_OT_DSL_DATASET, tx); } boolean_t dsl_dataset_is_zapified(dsl_dataset_t *ds) { dmu_object_info_t doi; dmu_object_info_from_db(ds->ds_dbuf, &doi); return (doi.doi_type == DMU_OTN_ZAP_METADATA); } boolean_t dsl_dataset_has_resume_receive_state(dsl_dataset_t *ds) { return (dsl_dataset_is_zapified(ds) && zap_contains(ds->ds_dir->dd_pool->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_TOGUID) == 0); } Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/edonr_zfs.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/edonr_zfs.c (nonexistent) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/edonr_zfs.c (revision 290757) @@ -0,0 +1,102 @@ +/* + * CDDL HEADER START + * + * The contents of this file are subject to the terms of the + * Common Development and Distribution License (the "License"). + * You may not use this file except in compliance with the License. + * + * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE + * or http://opensource.org/licenses/CDDL-1.0. + * See the License for the specific language governing permissions + * and limitations under the License. + * + * When distributing Covered Code, include this CDDL HEADER in each + * file and include the License file at usr/src/OPENSOLARIS.LICENSE. + * If applicable, add the following below this CDDL HEADER, with the + * fields enclosed by brackets "[]" replaced with your own identifying + * information: Portions Copyright [yyyy] [name of copyright owner] + * + * CDDL HEADER END + */ +/* + * Copyright 2013 Saso Kiselkov. All rights reserved. + * Use is subject to license terms. + */ +#include +#include +#include + +#define EDONR_MODE 512 +#define EDONR_BLOCK_SIZE EdonR512_BLOCK_SIZE + +/* + * Native zio_checksum interface for the Edon-R hash function. + */ +/*ARGSUSED*/ +void +zio_checksum_edonr_native(const void *buf, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp) +{ + uint8_t digest[EDONR_MODE / 8]; + EdonRState ctx; + + ASSERT(ctx_template != NULL); + bcopy(ctx_template, &ctx, sizeof (ctx)); + EdonRUpdate(&ctx, buf, size * 8); + EdonRFinal(&ctx, digest); + bcopy(digest, zcp->zc_word, sizeof (zcp->zc_word)); +} + +/* + * Byteswapped zio_checksum interface for the Edon-R hash function. + */ +void +zio_checksum_edonr_byteswap(const void *buf, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp) +{ + zio_cksum_t tmp; + + zio_checksum_edonr_native(buf, size, ctx_template, &tmp); + zcp->zc_word[0] = BSWAP_64(zcp->zc_word[0]); + zcp->zc_word[1] = BSWAP_64(zcp->zc_word[1]); + zcp->zc_word[2] = BSWAP_64(zcp->zc_word[2]); + zcp->zc_word[3] = BSWAP_64(zcp->zc_word[3]); +} + +void * +zio_checksum_edonr_tmpl_init(const zio_cksum_salt_t *salt) +{ + EdonRState *ctx; + uint8_t salt_block[EDONR_BLOCK_SIZE]; + + /* + * Edon-R needs all but the last hash invocation to be on full-size + * blocks, but the salt is too small. Rather than simply padding it + * with zeros, we expand the salt into a new salt block of proper + * size by double-hashing it (the new salt block will be composed of + * H(salt) || H(H(salt))). + */ + CTASSERT(EDONR_BLOCK_SIZE == 2 * (EDONR_MODE / 8)); + EdonRHash(EDONR_MODE, salt->zcs_bytes, sizeof (salt->zcs_bytes) * 8, + salt_block); + EdonRHash(EDONR_MODE, salt_block, EDONR_MODE, salt_block + + EDONR_MODE / 8); + + /* + * Feed the new salt block into the hash function - this will serve + * as our MAC key. + */ + ctx = kmem_zalloc(sizeof (*ctx), KM_SLEEP); + EdonRInit(ctx, EDONR_MODE); + EdonRUpdate(ctx, salt_block, sizeof (salt_block) * 8); + return (ctx); +} + +void +zio_checksum_edonr_tmpl_free(void *ctx_template) +{ + EdonRState *ctx = ctx_template; + + bzero(ctx, sizeof (*ctx)); + kmem_free(ctx, sizeof (*ctx)); +} Property changes on: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/edonr_zfs.c ___________________________________________________________________ Added: svn:eol-style ## -0,0 +1 ## +native \ No newline at end of property Added: svn:keywords ## -0,0 +1 ## +FreeBSD=%H \ No newline at end of property Added: svn:mime-type ## -0,0 +1 ## +text/plain \ No newline at end of property Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sha256.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sha256.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sha256.c (revision 290757) @@ -1,54 +1,87 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright 2009 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. */ +/* + * Copyright 2013 Saso Kiselkov. All rights reserved. + */ #include #include #ifdef _KERNEL #include #else #include #endif +/*ARGSUSED*/ void -zio_checksum_SHA256(const void *buf, uint64_t size, zio_cksum_t *zcp) +zio_checksum_SHA256(const void *buf, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp) { SHA256_CTX ctx; zio_cksum_t tmp; SHA256_Init(&ctx); SHA256_Update(&ctx, buf, size); SHA256_Final((unsigned char *)&tmp, &ctx); /* * A prior implementation of this function had a * private SHA256 implementation always wrote things out in * Big Endian and there wasn't a byteswap variant of it. * To preseve on disk compatibility we need to force that * behaviour. */ zcp->zc_word[0] = BE_64(tmp.zc_word[0]); zcp->zc_word[1] = BE_64(tmp.zc_word[1]); zcp->zc_word[2] = BE_64(tmp.zc_word[2]); zcp->zc_word[3] = BE_64(tmp.zc_word[3]); } + +#ifdef illumos +/*ARGSUSED*/ +void +zio_checksum_SHA512_native(const void *buf, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp) +{ + SHA2_CTX ctx; + + SHA2Init(SHA512_256, &ctx); + SHA2Update(&ctx, buf, size); + SHA2Final(zcp, &ctx); +} + +/*ARGSUSED*/ +void +zio_checksum_SHA512_byteswap(const void *buf, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp) +{ + zio_cksum_t tmp; + + zio_checksum_SHA512_native(buf, size, ctx_template, &tmp); + zcp->zc_word[0] = BSWAP_64(tmp.zc_word[0]); + zcp->zc_word[1] = BSWAP_64(tmp.zc_word[1]); + zcp->zc_word[2] = BSWAP_64(tmp.zc_word[2]); + zcp->zc_word[3] = BSWAP_64(tmp.zc_word[3]); +} +#endif Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/skein_zfs.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/skein_zfs.c (nonexistent) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/skein_zfs.c (revision 290757) @@ -0,0 +1,91 @@ +/* + * CDDL HEADER START + * + * The contents of this file are subject to the terms of the + * Common Development and Distribution License (the "License"). + * You may not use this file except in compliance with the License. + * + * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE + * or http://opensource.org/licenses/CDDL-1.0. + * See the License for the specific language governing permissions + * and limitations under the License. + * + * When distributing Covered Code, include this CDDL HEADER in each + * file and include the License file at usr/src/OPENSOLARIS.LICENSE. + * If applicable, add the following below this CDDL HEADER, with the + * fields enclosed by brackets "[]" replaced with your own identifying + * information: Portions Copyright [yyyy] [name of copyright owner] + * + * CDDL HEADER END + */ +/* + * Copyright 2013 Saso Kiselkov. All rights reserved. + */ +#include +#include +#include + +/* + * Computes a native 256-bit skein MAC checksum. Please note that this + * function requires the presence of a ctx_template that should be allocated + * using zio_checksum_skein_tmpl_init. + */ +/*ARGSUSED*/ +void +zio_checksum_skein_native(const void *buf, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp) +{ + Skein_512_Ctxt_t ctx; + + ASSERT(ctx_template != NULL); + bcopy(ctx_template, &ctx, sizeof (ctx)); + (void) Skein_512_Update(&ctx, buf, size); + (void) Skein_512_Final(&ctx, (uint8_t *)zcp); + bzero(&ctx, sizeof (ctx)); +} + +/* + * Byteswapped version of zio_checksum_skein_native. This just invokes + * the native checksum function and byteswaps the resulting checksum (since + * skein is internally endian-insensitive). + */ +void +zio_checksum_skein_byteswap(const void *buf, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp) +{ + zio_cksum_t tmp; + + zio_checksum_skein_native(buf, size, ctx_template, &tmp); + zcp->zc_word[0] = BSWAP_64(tmp.zc_word[0]); + zcp->zc_word[1] = BSWAP_64(tmp.zc_word[1]); + zcp->zc_word[2] = BSWAP_64(tmp.zc_word[2]); + zcp->zc_word[3] = BSWAP_64(tmp.zc_word[3]); +} + +/* + * Allocates a skein MAC template suitable for using in skein MAC checksum + * computations and returns a pointer to it. + */ +void * +zio_checksum_skein_tmpl_init(const zio_cksum_salt_t *salt) +{ + Skein_512_Ctxt_t *ctx; + + ctx = kmem_zalloc(sizeof (*ctx), KM_SLEEP); + (void) Skein_512_InitExt(ctx, sizeof (zio_cksum_t) * 8, 0, + salt->zcs_bytes, sizeof (salt->zcs_bytes)); + return (ctx); +} + +/* + * Frees a skein context template previously allocated using + * zio_checksum_skein_tmpl_init. + */ +void +zio_checksum_skein_tmpl_free(void *ctx_template) +{ + Skein_512_Ctxt_t *ctx = ctx_template; + + bzero(ctx, sizeof (*ctx)); + kmem_free(ctx, sizeof (*ctx)); +} Property changes on: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/skein_zfs.c ___________________________________________________________________ Added: svn:eol-style ## -0,0 +1 ## +native \ No newline at end of property Added: svn:keywords ## -0,0 +1 ## +FreeBSD=%H \ No newline at end of property Added: svn:mime-type ## -0,0 +1 ## +text/plain \ No newline at end of property Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c (revision 290757) @@ -1,7026 +1,7060 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2013 by Delphix. All rights reserved. * Copyright (c) 2015, Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2013 Martin Matuska . All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. + * Copyright 2013 Saso Kiselkov. All rights reserved. */ /* * SPA: Storage Pool Allocator * * This file contains all the routines used when modifying on-disk SPA state. * This includes opening, importing, destroying, exporting a pool, and syncing a * pool. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef _KERNEL #include #include #include #endif /* _KERNEL */ #include "zfs_prop.h" #include "zfs_comutil.h" /* Check hostid on import? */ static int check_hostid = 1; SYSCTL_DECL(_vfs_zfs); TUNABLE_INT("vfs.zfs.check_hostid", &check_hostid); SYSCTL_INT(_vfs_zfs, OID_AUTO, check_hostid, CTLFLAG_RW, &check_hostid, 0, "Check hostid on import?"); /* * The interval, in seconds, at which failed configuration cache file writes * should be retried. */ static int zfs_ccw_retry_interval = 300; typedef enum zti_modes { ZTI_MODE_FIXED, /* value is # of threads (min 1) */ ZTI_MODE_BATCH, /* cpu-intensive; value is ignored */ ZTI_MODE_NULL, /* don't create a taskq */ ZTI_NMODES } zti_modes_t; #define ZTI_P(n, q) { ZTI_MODE_FIXED, (n), (q) } #define ZTI_BATCH { ZTI_MODE_BATCH, 0, 1 } #define ZTI_NULL { ZTI_MODE_NULL, 0, 0 } #define ZTI_N(n) ZTI_P(n, 1) #define ZTI_ONE ZTI_N(1) typedef struct zio_taskq_info { zti_modes_t zti_mode; uint_t zti_value; uint_t zti_count; } zio_taskq_info_t; static const char *const zio_taskq_types[ZIO_TASKQ_TYPES] = { "issue", "issue_high", "intr", "intr_high" }; /* * This table defines the taskq settings for each ZFS I/O type. When * initializing a pool, we use this table to create an appropriately sized * taskq. Some operations are low volume and therefore have a small, static * number of threads assigned to their taskqs using the ZTI_N(#) or ZTI_ONE * macros. Other operations process a large amount of data; the ZTI_BATCH * macro causes us to create a taskq oriented for throughput. Some operations * are so high frequency and short-lived that the taskq itself can become a a * point of lock contention. The ZTI_P(#, #) macro indicates that we need an * additional degree of parallelism specified by the number of threads per- * taskq and the number of taskqs; when dispatching an event in this case, the * particular taskq is chosen at random. * * The different taskq priorities are to handle the different contexts (issue * and interrupt) and then to reserve threads for ZIO_PRIORITY_NOW I/Os that * need to be handled with minimum delay. */ const zio_taskq_info_t zio_taskqs[ZIO_TYPES][ZIO_TASKQ_TYPES] = { /* ISSUE ISSUE_HIGH INTR INTR_HIGH */ { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* NULL */ { ZTI_N(8), ZTI_NULL, ZTI_BATCH, ZTI_NULL }, /* READ */ { ZTI_BATCH, ZTI_N(5), ZTI_N(8), ZTI_N(5) }, /* WRITE */ { ZTI_P(12, 8), ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* FREE */ { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* CLAIM */ { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* IOCTL */ }; static void spa_sync_version(void *arg, dmu_tx_t *tx); static void spa_sync_props(void *arg, dmu_tx_t *tx); static boolean_t spa_has_active_shared_spare(spa_t *spa); static int spa_load_impl(spa_t *spa, uint64_t, nvlist_t *config, spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig, char **ereport); static void spa_vdev_resilver_done(spa_t *spa); uint_t zio_taskq_batch_pct = 75; /* 1 thread per cpu in pset */ #ifdef PSRSET_BIND id_t zio_taskq_psrset_bind = PS_NONE; #endif #ifdef SYSDC boolean_t zio_taskq_sysdc = B_TRUE; /* use SDC scheduling class */ #endif uint_t zio_taskq_basedc = 80; /* base duty cycle */ boolean_t spa_create_process = B_TRUE; /* no process ==> no sysdc */ extern int zfs_sync_pass_deferred_free; #ifndef illumos extern void spa_deadman(void *arg); #endif /* * This (illegal) pool name is used when temporarily importing a spa_t in order * to get the vdev stats associated with the imported devices. */ #define TRYIMPORT_NAME "$import" /* * ========================================================================== * SPA properties routines * ========================================================================== */ /* * Add a (source=src, propname=propval) list to an nvlist. */ static void spa_prop_add_list(nvlist_t *nvl, zpool_prop_t prop, char *strval, uint64_t intval, zprop_source_t src) { const char *propname = zpool_prop_to_name(prop); nvlist_t *propval; VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_uint64(propval, ZPROP_SOURCE, src) == 0); if (strval != NULL) VERIFY(nvlist_add_string(propval, ZPROP_VALUE, strval) == 0); else VERIFY(nvlist_add_uint64(propval, ZPROP_VALUE, intval) == 0); VERIFY(nvlist_add_nvlist(nvl, propname, propval) == 0); nvlist_free(propval); } /* * Get property values from the spa configuration. */ static void spa_prop_get_config(spa_t *spa, nvlist_t **nvp) { vdev_t *rvd = spa->spa_root_vdev; dsl_pool_t *pool = spa->spa_dsl_pool; uint64_t size, alloc, cap, version; zprop_source_t src = ZPROP_SRC_NONE; spa_config_dirent_t *dp; metaslab_class_t *mc = spa_normal_class(spa); ASSERT(MUTEX_HELD(&spa->spa_props_lock)); if (rvd != NULL) { alloc = metaslab_class_get_alloc(spa_normal_class(spa)); size = metaslab_class_get_space(spa_normal_class(spa)); spa_prop_add_list(*nvp, ZPOOL_PROP_NAME, spa_name(spa), 0, src); spa_prop_add_list(*nvp, ZPOOL_PROP_SIZE, NULL, size, src); spa_prop_add_list(*nvp, ZPOOL_PROP_ALLOCATED, NULL, alloc, src); spa_prop_add_list(*nvp, ZPOOL_PROP_FREE, NULL, size - alloc, src); spa_prop_add_list(*nvp, ZPOOL_PROP_FRAGMENTATION, NULL, metaslab_class_fragmentation(mc), src); spa_prop_add_list(*nvp, ZPOOL_PROP_EXPANDSZ, NULL, metaslab_class_expandable_space(mc), src); spa_prop_add_list(*nvp, ZPOOL_PROP_READONLY, NULL, (spa_mode(spa) == FREAD), src); cap = (size == 0) ? 0 : (alloc * 100 / size); spa_prop_add_list(*nvp, ZPOOL_PROP_CAPACITY, NULL, cap, src); spa_prop_add_list(*nvp, ZPOOL_PROP_DEDUPRATIO, NULL, ddt_get_pool_dedup_ratio(spa), src); spa_prop_add_list(*nvp, ZPOOL_PROP_HEALTH, NULL, rvd->vdev_state, src); version = spa_version(spa); if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION)) src = ZPROP_SRC_DEFAULT; else src = ZPROP_SRC_LOCAL; spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL, version, src); } if (pool != NULL) { /* * The $FREE directory was introduced in SPA_VERSION_DEADLISTS, * when opening pools before this version freedir will be NULL. */ if (pool->dp_free_dir != NULL) { spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, NULL, dsl_dir_phys(pool->dp_free_dir)->dd_used_bytes, src); } else { spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, NULL, 0, src); } if (pool->dp_leak_dir != NULL) { spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED, NULL, dsl_dir_phys(pool->dp_leak_dir)->dd_used_bytes, src); } else { spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED, NULL, 0, src); } } spa_prop_add_list(*nvp, ZPOOL_PROP_GUID, NULL, spa_guid(spa), src); if (spa->spa_comment != NULL) { spa_prop_add_list(*nvp, ZPOOL_PROP_COMMENT, spa->spa_comment, 0, ZPROP_SRC_LOCAL); } if (spa->spa_root != NULL) spa_prop_add_list(*nvp, ZPOOL_PROP_ALTROOT, spa->spa_root, 0, ZPROP_SRC_LOCAL); if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) { spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL, MIN(zfs_max_recordsize, SPA_MAXBLOCKSIZE), ZPROP_SRC_NONE); } else { spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL, SPA_OLD_MAXBLOCKSIZE, ZPROP_SRC_NONE); } if ((dp = list_head(&spa->spa_config_list)) != NULL) { if (dp->scd_path == NULL) { spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE, "none", 0, ZPROP_SRC_LOCAL); } else if (strcmp(dp->scd_path, spa_config_path) != 0) { spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE, dp->scd_path, 0, ZPROP_SRC_LOCAL); } } } /* * Get zpool property values. */ int spa_prop_get(spa_t *spa, nvlist_t **nvp) { objset_t *mos = spa->spa_meta_objset; zap_cursor_t zc; zap_attribute_t za; int err; VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0); mutex_enter(&spa->spa_props_lock); /* * Get properties from the spa config. */ spa_prop_get_config(spa, nvp); /* If no pool property object, no more prop to get. */ if (mos == NULL || spa->spa_pool_props_object == 0) { mutex_exit(&spa->spa_props_lock); return (0); } /* * Get properties from the MOS pool property object. */ for (zap_cursor_init(&zc, mos, spa->spa_pool_props_object); (err = zap_cursor_retrieve(&zc, &za)) == 0; zap_cursor_advance(&zc)) { uint64_t intval = 0; char *strval = NULL; zprop_source_t src = ZPROP_SRC_DEFAULT; zpool_prop_t prop; if ((prop = zpool_name_to_prop(za.za_name)) == ZPROP_INVAL) continue; switch (za.za_integer_length) { case 8: /* integer property */ if (za.za_first_integer != zpool_prop_default_numeric(prop)) src = ZPROP_SRC_LOCAL; if (prop == ZPOOL_PROP_BOOTFS) { dsl_pool_t *dp; dsl_dataset_t *ds = NULL; dp = spa_get_dsl(spa); dsl_pool_config_enter(dp, FTAG); if (err = dsl_dataset_hold_obj(dp, za.za_first_integer, FTAG, &ds)) { dsl_pool_config_exit(dp, FTAG); break; } strval = kmem_alloc( MAXNAMELEN + strlen(MOS_DIR_NAME) + 1, KM_SLEEP); dsl_dataset_name(ds, strval); dsl_dataset_rele(ds, FTAG); dsl_pool_config_exit(dp, FTAG); } else { strval = NULL; intval = za.za_first_integer; } spa_prop_add_list(*nvp, prop, strval, intval, src); if (strval != NULL) kmem_free(strval, MAXNAMELEN + strlen(MOS_DIR_NAME) + 1); break; case 1: /* string property */ strval = kmem_alloc(za.za_num_integers, KM_SLEEP); err = zap_lookup(mos, spa->spa_pool_props_object, za.za_name, 1, za.za_num_integers, strval); if (err) { kmem_free(strval, za.za_num_integers); break; } spa_prop_add_list(*nvp, prop, strval, 0, src); kmem_free(strval, za.za_num_integers); break; default: break; } } zap_cursor_fini(&zc); mutex_exit(&spa->spa_props_lock); out: if (err && err != ENOENT) { nvlist_free(*nvp); *nvp = NULL; return (err); } return (0); } /* * Validate the given pool properties nvlist and modify the list * for the property values to be set. */ static int spa_prop_validate(spa_t *spa, nvlist_t *props) { nvpair_t *elem; int error = 0, reset_bootfs = 0; uint64_t objnum = 0; boolean_t has_feature = B_FALSE; elem = NULL; while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { uint64_t intval; char *strval, *slash, *check, *fname; const char *propname = nvpair_name(elem); zpool_prop_t prop = zpool_name_to_prop(propname); switch (prop) { case ZPROP_INVAL: if (!zpool_prop_feature(propname)) { error = SET_ERROR(EINVAL); break; } /* * Sanitize the input. */ if (nvpair_type(elem) != DATA_TYPE_UINT64) { error = SET_ERROR(EINVAL); break; } if (nvpair_value_uint64(elem, &intval) != 0) { error = SET_ERROR(EINVAL); break; } if (intval != 0) { error = SET_ERROR(EINVAL); break; } fname = strchr(propname, '@') + 1; if (zfeature_lookup_name(fname, NULL) != 0) { error = SET_ERROR(EINVAL); break; } has_feature = B_TRUE; break; case ZPOOL_PROP_VERSION: error = nvpair_value_uint64(elem, &intval); if (!error && (intval < spa_version(spa) || intval > SPA_VERSION_BEFORE_FEATURES || has_feature)) error = SET_ERROR(EINVAL); break; case ZPOOL_PROP_DELEGATION: case ZPOOL_PROP_AUTOREPLACE: case ZPOOL_PROP_LISTSNAPS: case ZPOOL_PROP_AUTOEXPAND: error = nvpair_value_uint64(elem, &intval); if (!error && intval > 1) error = SET_ERROR(EINVAL); break; case ZPOOL_PROP_BOOTFS: /* * If the pool version is less than SPA_VERSION_BOOTFS, * or the pool is still being created (version == 0), * the bootfs property cannot be set. */ if (spa_version(spa) < SPA_VERSION_BOOTFS) { error = SET_ERROR(ENOTSUP); break; } /* * Make sure the vdev config is bootable */ if (!vdev_is_bootable(spa->spa_root_vdev)) { error = SET_ERROR(ENOTSUP); break; } reset_bootfs = 1; error = nvpair_value_string(elem, &strval); if (!error) { objset_t *os; uint64_t propval; if (strval == NULL || strval[0] == '\0') { objnum = zpool_prop_default_numeric( ZPOOL_PROP_BOOTFS); break; } if (error = dmu_objset_hold(strval, FTAG, &os)) break; /* * Must be ZPL, and its property settings * must be supported by GRUB (compression * is not gzip, and large blocks are not used). */ if (dmu_objset_type(os) != DMU_OST_ZFS) { error = SET_ERROR(ENOTSUP); } else if ((error = dsl_prop_get_int_ds(dmu_objset_ds(os), zfs_prop_to_name(ZFS_PROP_COMPRESSION), &propval)) == 0 && !BOOTFS_COMPRESS_VALID(propval)) { error = SET_ERROR(ENOTSUP); } else if ((error = dsl_prop_get_int_ds(dmu_objset_ds(os), zfs_prop_to_name(ZFS_PROP_RECORDSIZE), &propval)) == 0 && propval > SPA_OLD_MAXBLOCKSIZE) { error = SET_ERROR(ENOTSUP); } else { objnum = dmu_objset_id(os); } dmu_objset_rele(os, FTAG); } break; case ZPOOL_PROP_FAILUREMODE: error = nvpair_value_uint64(elem, &intval); if (!error && (intval < ZIO_FAILURE_MODE_WAIT || intval > ZIO_FAILURE_MODE_PANIC)) error = SET_ERROR(EINVAL); /* * This is a special case which only occurs when * the pool has completely failed. This allows * the user to change the in-core failmode property * without syncing it out to disk (I/Os might * currently be blocked). We do this by returning * EIO to the caller (spa_prop_set) to trick it * into thinking we encountered a property validation * error. */ if (!error && spa_suspended(spa)) { spa->spa_failmode = intval; error = SET_ERROR(EIO); } break; case ZPOOL_PROP_CACHEFILE: if ((error = nvpair_value_string(elem, &strval)) != 0) break; if (strval[0] == '\0') break; if (strcmp(strval, "none") == 0) break; if (strval[0] != '/') { error = SET_ERROR(EINVAL); break; } slash = strrchr(strval, '/'); ASSERT(slash != NULL); if (slash[1] == '\0' || strcmp(slash, "/.") == 0 || strcmp(slash, "/..") == 0) error = SET_ERROR(EINVAL); break; case ZPOOL_PROP_COMMENT: if ((error = nvpair_value_string(elem, &strval)) != 0) break; for (check = strval; *check != '\0'; check++) { /* * The kernel doesn't have an easy isprint() * check. For this kernel check, we merely * check ASCII apart from DEL. Fix this if * there is an easy-to-use kernel isprint(). */ if (*check >= 0x7f) { error = SET_ERROR(EINVAL); break; } check++; } if (strlen(strval) > ZPROP_MAX_COMMENT) error = E2BIG; break; case ZPOOL_PROP_DEDUPDITTO: if (spa_version(spa) < SPA_VERSION_DEDUP) error = SET_ERROR(ENOTSUP); else error = nvpair_value_uint64(elem, &intval); if (error == 0 && intval != 0 && intval < ZIO_DEDUPDITTO_MIN) error = SET_ERROR(EINVAL); break; } if (error) break; } if (!error && reset_bootfs) { error = nvlist_remove(props, zpool_prop_to_name(ZPOOL_PROP_BOOTFS), DATA_TYPE_STRING); if (!error) { error = nvlist_add_uint64(props, zpool_prop_to_name(ZPOOL_PROP_BOOTFS), objnum); } } return (error); } void spa_configfile_set(spa_t *spa, nvlist_t *nvp, boolean_t need_sync) { char *cachefile; spa_config_dirent_t *dp; if (nvlist_lookup_string(nvp, zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), &cachefile) != 0) return; dp = kmem_alloc(sizeof (spa_config_dirent_t), KM_SLEEP); if (cachefile[0] == '\0') dp->scd_path = spa_strdup(spa_config_path); else if (strcmp(cachefile, "none") == 0) dp->scd_path = NULL; else dp->scd_path = spa_strdup(cachefile); list_insert_head(&spa->spa_config_list, dp); if (need_sync) spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE); } int spa_prop_set(spa_t *spa, nvlist_t *nvp) { int error; nvpair_t *elem = NULL; boolean_t need_sync = B_FALSE; if ((error = spa_prop_validate(spa, nvp)) != 0) return (error); while ((elem = nvlist_next_nvpair(nvp, elem)) != NULL) { zpool_prop_t prop = zpool_name_to_prop(nvpair_name(elem)); if (prop == ZPOOL_PROP_CACHEFILE || prop == ZPOOL_PROP_ALTROOT || prop == ZPOOL_PROP_READONLY) continue; if (prop == ZPOOL_PROP_VERSION || prop == ZPROP_INVAL) { uint64_t ver; if (prop == ZPOOL_PROP_VERSION) { VERIFY(nvpair_value_uint64(elem, &ver) == 0); } else { ASSERT(zpool_prop_feature(nvpair_name(elem))); ver = SPA_VERSION_FEATURES; need_sync = B_TRUE; } /* Save time if the version is already set. */ if (ver == spa_version(spa)) continue; /* * In addition to the pool directory object, we might * create the pool properties object, the features for * read object, the features for write object, or the * feature descriptions object. */ error = dsl_sync_task(spa->spa_name, NULL, spa_sync_version, &ver, 6, ZFS_SPACE_CHECK_RESERVED); if (error) return (error); continue; } need_sync = B_TRUE; break; } if (need_sync) { return (dsl_sync_task(spa->spa_name, NULL, spa_sync_props, nvp, 6, ZFS_SPACE_CHECK_RESERVED)); } return (0); } /* * If the bootfs property value is dsobj, clear it. */ void spa_prop_clear_bootfs(spa_t *spa, uint64_t dsobj, dmu_tx_t *tx) { if (spa->spa_bootfs == dsobj && spa->spa_pool_props_object != 0) { VERIFY(zap_remove(spa->spa_meta_objset, spa->spa_pool_props_object, zpool_prop_to_name(ZPOOL_PROP_BOOTFS), tx) == 0); spa->spa_bootfs = 0; } } /*ARGSUSED*/ static int spa_change_guid_check(void *arg, dmu_tx_t *tx) { uint64_t *newguid = arg; spa_t *spa = dmu_tx_pool(tx)->dp_spa; vdev_t *rvd = spa->spa_root_vdev; uint64_t vdev_state; spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); vdev_state = rvd->vdev_state; spa_config_exit(spa, SCL_STATE, FTAG); if (vdev_state != VDEV_STATE_HEALTHY) return (SET_ERROR(ENXIO)); ASSERT3U(spa_guid(spa), !=, *newguid); return (0); } static void spa_change_guid_sync(void *arg, dmu_tx_t *tx) { uint64_t *newguid = arg; spa_t *spa = dmu_tx_pool(tx)->dp_spa; uint64_t oldguid; vdev_t *rvd = spa->spa_root_vdev; oldguid = spa_guid(spa); spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); rvd->vdev_guid = *newguid; rvd->vdev_guid_sum += (*newguid - oldguid); vdev_config_dirty(rvd); spa_config_exit(spa, SCL_STATE, FTAG); spa_history_log_internal(spa, "guid change", tx, "old=%llu new=%llu", oldguid, *newguid); } /* * Change the GUID for the pool. This is done so that we can later * re-import a pool built from a clone of our own vdevs. We will modify * the root vdev's guid, our own pool guid, and then mark all of our * vdevs dirty. Note that we must make sure that all our vdevs are * online when we do this, or else any vdevs that weren't present * would be orphaned from our pool. We are also going to issue a * sysevent to update any watchers. */ int spa_change_guid(spa_t *spa) { int error; uint64_t guid; mutex_enter(&spa->spa_vdev_top_lock); mutex_enter(&spa_namespace_lock); guid = spa_generate_guid(NULL); error = dsl_sync_task(spa->spa_name, spa_change_guid_check, spa_change_guid_sync, &guid, 5, ZFS_SPACE_CHECK_RESERVED); if (error == 0) { spa_config_sync(spa, B_FALSE, B_TRUE); spa_event_notify(spa, NULL, ESC_ZFS_POOL_REGUID); } mutex_exit(&spa_namespace_lock); mutex_exit(&spa->spa_vdev_top_lock); return (error); } /* * ========================================================================== * SPA state manipulation (open/create/destroy/import/export) * ========================================================================== */ static int spa_error_entry_compare(const void *a, const void *b) { spa_error_entry_t *sa = (spa_error_entry_t *)a; spa_error_entry_t *sb = (spa_error_entry_t *)b; int ret; ret = bcmp(&sa->se_bookmark, &sb->se_bookmark, sizeof (zbookmark_phys_t)); if (ret < 0) return (-1); else if (ret > 0) return (1); else return (0); } /* * Utility function which retrieves copies of the current logs and * re-initializes them in the process. */ void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub) { ASSERT(MUTEX_HELD(&spa->spa_errlist_lock)); bcopy(&spa->spa_errlist_last, last, sizeof (avl_tree_t)); bcopy(&spa->spa_errlist_scrub, scrub, sizeof (avl_tree_t)); avl_create(&spa->spa_errlist_scrub, spa_error_entry_compare, sizeof (spa_error_entry_t), offsetof(spa_error_entry_t, se_avl)); avl_create(&spa->spa_errlist_last, spa_error_entry_compare, sizeof (spa_error_entry_t), offsetof(spa_error_entry_t, se_avl)); } static void spa_taskqs_init(spa_t *spa, zio_type_t t, zio_taskq_type_t q) { const zio_taskq_info_t *ztip = &zio_taskqs[t][q]; enum zti_modes mode = ztip->zti_mode; uint_t value = ztip->zti_value; uint_t count = ztip->zti_count; spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; char name[32]; uint_t flags = 0; boolean_t batch = B_FALSE; if (mode == ZTI_MODE_NULL) { tqs->stqs_count = 0; tqs->stqs_taskq = NULL; return; } ASSERT3U(count, >, 0); tqs->stqs_count = count; tqs->stqs_taskq = kmem_alloc(count * sizeof (taskq_t *), KM_SLEEP); switch (mode) { case ZTI_MODE_FIXED: ASSERT3U(value, >=, 1); value = MAX(value, 1); break; case ZTI_MODE_BATCH: batch = B_TRUE; flags |= TASKQ_THREADS_CPU_PCT; value = zio_taskq_batch_pct; break; default: panic("unrecognized mode for %s_%s taskq (%u:%u) in " "spa_activate()", zio_type_name[t], zio_taskq_types[q], mode, value); break; } for (uint_t i = 0; i < count; i++) { taskq_t *tq; if (count > 1) { (void) snprintf(name, sizeof (name), "%s_%s_%u", zio_type_name[t], zio_taskq_types[q], i); } else { (void) snprintf(name, sizeof (name), "%s_%s", zio_type_name[t], zio_taskq_types[q]); } #ifdef SYSDC if (zio_taskq_sysdc && spa->spa_proc != &p0) { if (batch) flags |= TASKQ_DC_BATCH; tq = taskq_create_sysdc(name, value, 50, INT_MAX, spa->spa_proc, zio_taskq_basedc, flags); } else { #endif pri_t pri = maxclsyspri; /* * The write issue taskq can be extremely CPU * intensive. Run it at slightly lower priority * than the other taskqs. */ if (t == ZIO_TYPE_WRITE && q == ZIO_TASKQ_ISSUE) pri--; tq = taskq_create_proc(name, value, pri, 50, INT_MAX, spa->spa_proc, flags); #ifdef SYSDC } #endif tqs->stqs_taskq[i] = tq; } } static void spa_taskqs_fini(spa_t *spa, zio_type_t t, zio_taskq_type_t q) { spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; if (tqs->stqs_taskq == NULL) { ASSERT0(tqs->stqs_count); return; } for (uint_t i = 0; i < tqs->stqs_count; i++) { ASSERT3P(tqs->stqs_taskq[i], !=, NULL); taskq_destroy(tqs->stqs_taskq[i]); } kmem_free(tqs->stqs_taskq, tqs->stqs_count * sizeof (taskq_t *)); tqs->stqs_taskq = NULL; } /* * Dispatch a task to the appropriate taskq for the ZFS I/O type and priority. * Note that a type may have multiple discrete taskqs to avoid lock contention * on the taskq itself. In that case we choose which taskq at random by using * the low bits of gethrtime(). */ void spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q, task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent) { spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; taskq_t *tq; ASSERT3P(tqs->stqs_taskq, !=, NULL); ASSERT3U(tqs->stqs_count, !=, 0); if (tqs->stqs_count == 1) { tq = tqs->stqs_taskq[0]; } else { #ifdef _KERNEL tq = tqs->stqs_taskq[cpu_ticks() % tqs->stqs_count]; #else tq = tqs->stqs_taskq[gethrtime() % tqs->stqs_count]; #endif } taskq_dispatch_ent(tq, func, arg, flags, ent); } static void spa_create_zio_taskqs(spa_t *spa) { for (int t = 0; t < ZIO_TYPES; t++) { for (int q = 0; q < ZIO_TASKQ_TYPES; q++) { spa_taskqs_init(spa, t, q); } } } #ifdef _KERNEL #ifdef SPA_PROCESS static void spa_thread(void *arg) { callb_cpr_t cprinfo; spa_t *spa = arg; user_t *pu = PTOU(curproc); CALLB_CPR_INIT(&cprinfo, &spa->spa_proc_lock, callb_generic_cpr, spa->spa_name); ASSERT(curproc != &p0); (void) snprintf(pu->u_psargs, sizeof (pu->u_psargs), "zpool-%s", spa->spa_name); (void) strlcpy(pu->u_comm, pu->u_psargs, sizeof (pu->u_comm)); #ifdef PSRSET_BIND /* bind this thread to the requested psrset */ if (zio_taskq_psrset_bind != PS_NONE) { pool_lock(); mutex_enter(&cpu_lock); mutex_enter(&pidlock); mutex_enter(&curproc->p_lock); if (cpupart_bind_thread(curthread, zio_taskq_psrset_bind, 0, NULL, NULL) == 0) { curthread->t_bind_pset = zio_taskq_psrset_bind; } else { cmn_err(CE_WARN, "Couldn't bind process for zfs pool \"%s\" to " "pset %d\n", spa->spa_name, zio_taskq_psrset_bind); } mutex_exit(&curproc->p_lock); mutex_exit(&pidlock); mutex_exit(&cpu_lock); pool_unlock(); } #endif #ifdef SYSDC if (zio_taskq_sysdc) { sysdc_thread_enter(curthread, 100, 0); } #endif spa->spa_proc = curproc; spa->spa_did = curthread->t_did; spa_create_zio_taskqs(spa); mutex_enter(&spa->spa_proc_lock); ASSERT(spa->spa_proc_state == SPA_PROC_CREATED); spa->spa_proc_state = SPA_PROC_ACTIVE; cv_broadcast(&spa->spa_proc_cv); CALLB_CPR_SAFE_BEGIN(&cprinfo); while (spa->spa_proc_state == SPA_PROC_ACTIVE) cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock); CALLB_CPR_SAFE_END(&cprinfo, &spa->spa_proc_lock); ASSERT(spa->spa_proc_state == SPA_PROC_DEACTIVATE); spa->spa_proc_state = SPA_PROC_GONE; spa->spa_proc = &p0; cv_broadcast(&spa->spa_proc_cv); CALLB_CPR_EXIT(&cprinfo); /* drops spa_proc_lock */ mutex_enter(&curproc->p_lock); lwp_exit(); } #endif /* SPA_PROCESS */ #endif /* * Activate an uninitialized pool. */ static void spa_activate(spa_t *spa, int mode) { ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); spa->spa_state = POOL_STATE_ACTIVE; spa->spa_mode = mode; spa->spa_normal_class = metaslab_class_create(spa, zfs_metaslab_ops); spa->spa_log_class = metaslab_class_create(spa, zfs_metaslab_ops); /* Try to create a covering process */ mutex_enter(&spa->spa_proc_lock); ASSERT(spa->spa_proc_state == SPA_PROC_NONE); ASSERT(spa->spa_proc == &p0); spa->spa_did = 0; #ifdef SPA_PROCESS /* Only create a process if we're going to be around a while. */ if (spa_create_process && strcmp(spa->spa_name, TRYIMPORT_NAME) != 0) { if (newproc(spa_thread, (caddr_t)spa, syscid, maxclsyspri, NULL, 0) == 0) { spa->spa_proc_state = SPA_PROC_CREATED; while (spa->spa_proc_state == SPA_PROC_CREATED) { cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock); } ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE); ASSERT(spa->spa_proc != &p0); ASSERT(spa->spa_did != 0); } else { #ifdef _KERNEL cmn_err(CE_WARN, "Couldn't create process for zfs pool \"%s\"\n", spa->spa_name); #endif } } #endif /* SPA_PROCESS */ mutex_exit(&spa->spa_proc_lock); /* If we didn't create a process, we need to create our taskqs. */ ASSERT(spa->spa_proc == &p0); if (spa->spa_proc == &p0) { spa_create_zio_taskqs(spa); } /* * Start TRIM thread. */ trim_thread_create(spa); list_create(&spa->spa_config_dirty_list, sizeof (vdev_t), offsetof(vdev_t, vdev_config_dirty_node)); list_create(&spa->spa_evicting_os_list, sizeof (objset_t), offsetof(objset_t, os_evicting_node)); list_create(&spa->spa_state_dirty_list, sizeof (vdev_t), offsetof(vdev_t, vdev_state_dirty_node)); txg_list_create(&spa->spa_vdev_txg_list, offsetof(struct vdev, vdev_txg_node)); avl_create(&spa->spa_errlist_scrub, spa_error_entry_compare, sizeof (spa_error_entry_t), offsetof(spa_error_entry_t, se_avl)); avl_create(&spa->spa_errlist_last, spa_error_entry_compare, sizeof (spa_error_entry_t), offsetof(spa_error_entry_t, se_avl)); } /* * Opposite of spa_activate(). */ static void spa_deactivate(spa_t *spa) { ASSERT(spa->spa_sync_on == B_FALSE); ASSERT(spa->spa_dsl_pool == NULL); ASSERT(spa->spa_root_vdev == NULL); ASSERT(spa->spa_async_zio_root == NULL); ASSERT(spa->spa_state != POOL_STATE_UNINITIALIZED); /* * Stop TRIM thread in case spa_unload() wasn't called directly * before spa_deactivate(). */ trim_thread_destroy(spa); spa_evicting_os_wait(spa); txg_list_destroy(&spa->spa_vdev_txg_list); list_destroy(&spa->spa_config_dirty_list); list_destroy(&spa->spa_evicting_os_list); list_destroy(&spa->spa_state_dirty_list); for (int t = 0; t < ZIO_TYPES; t++) { for (int q = 0; q < ZIO_TASKQ_TYPES; q++) { spa_taskqs_fini(spa, t, q); } } metaslab_class_destroy(spa->spa_normal_class); spa->spa_normal_class = NULL; metaslab_class_destroy(spa->spa_log_class); spa->spa_log_class = NULL; /* * If this was part of an import or the open otherwise failed, we may * still have errors left in the queues. Empty them just in case. */ spa_errlog_drain(spa); avl_destroy(&spa->spa_errlist_scrub); avl_destroy(&spa->spa_errlist_last); spa->spa_state = POOL_STATE_UNINITIALIZED; mutex_enter(&spa->spa_proc_lock); if (spa->spa_proc_state != SPA_PROC_NONE) { ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE); spa->spa_proc_state = SPA_PROC_DEACTIVATE; cv_broadcast(&spa->spa_proc_cv); while (spa->spa_proc_state == SPA_PROC_DEACTIVATE) { ASSERT(spa->spa_proc != &p0); cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock); } ASSERT(spa->spa_proc_state == SPA_PROC_GONE); spa->spa_proc_state = SPA_PROC_NONE; } ASSERT(spa->spa_proc == &p0); mutex_exit(&spa->spa_proc_lock); #ifdef SPA_PROCESS /* * We want to make sure spa_thread() has actually exited the ZFS * module, so that the module can't be unloaded out from underneath * it. */ if (spa->spa_did != 0) { thread_join(spa->spa_did); spa->spa_did = 0; } #endif /* SPA_PROCESS */ } /* * Verify a pool configuration, and construct the vdev tree appropriately. This * will create all the necessary vdevs in the appropriate layout, with each vdev * in the CLOSED state. This will prep the pool before open/creation/import. * All vdev validation is done by the vdev_alloc() routine. */ static int spa_config_parse(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id, int atype) { nvlist_t **child; uint_t children; int error; if ((error = vdev_alloc(spa, vdp, nv, parent, id, atype)) != 0) return (error); if ((*vdp)->vdev_ops->vdev_op_leaf) return (0); error = nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child, &children); if (error == ENOENT) return (0); if (error) { vdev_free(*vdp); *vdp = NULL; return (SET_ERROR(EINVAL)); } for (int c = 0; c < children; c++) { vdev_t *vd; if ((error = spa_config_parse(spa, &vd, child[c], *vdp, c, atype)) != 0) { vdev_free(*vdp); *vdp = NULL; return (error); } } ASSERT(*vdp != NULL); return (0); } /* * Opposite of spa_load(). */ static void spa_unload(spa_t *spa) { int i; ASSERT(MUTEX_HELD(&spa_namespace_lock)); /* * Stop TRIM thread. */ trim_thread_destroy(spa); /* * Stop async tasks. */ spa_async_suspend(spa); /* * Stop syncing. */ if (spa->spa_sync_on) { txg_sync_stop(spa->spa_dsl_pool); spa->spa_sync_on = B_FALSE; } /* * Wait for any outstanding async I/O to complete. */ if (spa->spa_async_zio_root != NULL) { for (int i = 0; i < max_ncpus; i++) (void) zio_wait(spa->spa_async_zio_root[i]); kmem_free(spa->spa_async_zio_root, max_ncpus * sizeof (void *)); spa->spa_async_zio_root = NULL; } bpobj_close(&spa->spa_deferred_bpobj); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); /* * Close all vdevs. */ if (spa->spa_root_vdev) vdev_free(spa->spa_root_vdev); ASSERT(spa->spa_root_vdev == NULL); /* * Close the dsl pool. */ if (spa->spa_dsl_pool) { dsl_pool_close(spa->spa_dsl_pool); spa->spa_dsl_pool = NULL; spa->spa_meta_objset = NULL; } ddt_unload(spa); /* * Drop and purge level 2 cache */ spa_l2cache_drop(spa); for (i = 0; i < spa->spa_spares.sav_count; i++) vdev_free(spa->spa_spares.sav_vdevs[i]); if (spa->spa_spares.sav_vdevs) { kmem_free(spa->spa_spares.sav_vdevs, spa->spa_spares.sav_count * sizeof (void *)); spa->spa_spares.sav_vdevs = NULL; } if (spa->spa_spares.sav_config) { nvlist_free(spa->spa_spares.sav_config); spa->spa_spares.sav_config = NULL; } spa->spa_spares.sav_count = 0; for (i = 0; i < spa->spa_l2cache.sav_count; i++) { vdev_clear_stats(spa->spa_l2cache.sav_vdevs[i]); vdev_free(spa->spa_l2cache.sav_vdevs[i]); } if (spa->spa_l2cache.sav_vdevs) { kmem_free(spa->spa_l2cache.sav_vdevs, spa->spa_l2cache.sav_count * sizeof (void *)); spa->spa_l2cache.sav_vdevs = NULL; } if (spa->spa_l2cache.sav_config) { nvlist_free(spa->spa_l2cache.sav_config); spa->spa_l2cache.sav_config = NULL; } spa->spa_l2cache.sav_count = 0; spa->spa_async_suspended = 0; if (spa->spa_comment != NULL) { spa_strfree(spa->spa_comment); spa->spa_comment = NULL; } spa_config_exit(spa, SCL_ALL, FTAG); } /* * Load (or re-load) the current list of vdevs describing the active spares for * this pool. When this is called, we have some form of basic information in * 'spa_spares.sav_config'. We parse this into vdevs, try to open them, and * then re-generate a more complete list including status information. */ static void spa_load_spares(spa_t *spa) { nvlist_t **spares; uint_t nspares; int i; vdev_t *vd, *tvd; ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); /* * First, close and free any existing spare vdevs. */ for (i = 0; i < spa->spa_spares.sav_count; i++) { vd = spa->spa_spares.sav_vdevs[i]; /* Undo the call to spa_activate() below */ if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid, B_FALSE)) != NULL && tvd->vdev_isspare) spa_spare_remove(tvd); vdev_close(vd); vdev_free(vd); } if (spa->spa_spares.sav_vdevs) kmem_free(spa->spa_spares.sav_vdevs, spa->spa_spares.sav_count * sizeof (void *)); if (spa->spa_spares.sav_config == NULL) nspares = 0; else VERIFY(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0); spa->spa_spares.sav_count = (int)nspares; spa->spa_spares.sav_vdevs = NULL; if (nspares == 0) return; /* * Construct the array of vdevs, opening them to get status in the * process. For each spare, there is potentially two different vdev_t * structures associated with it: one in the list of spares (used only * for basic validation purposes) and one in the active vdev * configuration (if it's spared in). During this phase we open and * validate each vdev on the spare list. If the vdev also exists in the * active configuration, then we also mark this vdev as an active spare. */ spa->spa_spares.sav_vdevs = kmem_alloc(nspares * sizeof (void *), KM_SLEEP); for (i = 0; i < spa->spa_spares.sav_count; i++) { VERIFY(spa_config_parse(spa, &vd, spares[i], NULL, 0, VDEV_ALLOC_SPARE) == 0); ASSERT(vd != NULL); spa->spa_spares.sav_vdevs[i] = vd; if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid, B_FALSE)) != NULL) { if (!tvd->vdev_isspare) spa_spare_add(tvd); /* * We only mark the spare active if we were successfully * able to load the vdev. Otherwise, importing a pool * with a bad active spare would result in strange * behavior, because multiple pool would think the spare * is actively in use. * * There is a vulnerability here to an equally bizarre * circumstance, where a dead active spare is later * brought back to life (onlined or otherwise). Given * the rarity of this scenario, and the extra complexity * it adds, we ignore the possibility. */ if (!vdev_is_dead(tvd)) spa_spare_activate(tvd); } vd->vdev_top = vd; vd->vdev_aux = &spa->spa_spares; if (vdev_open(vd) != 0) continue; if (vdev_validate_aux(vd) == 0) spa_spare_add(vd); } /* * Recompute the stashed list of spares, with status information * this time. */ VERIFY(nvlist_remove(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, DATA_TYPE_NVLIST_ARRAY) == 0); spares = kmem_alloc(spa->spa_spares.sav_count * sizeof (void *), KM_SLEEP); for (i = 0; i < spa->spa_spares.sav_count; i++) spares[i] = vdev_config_generate(spa, spa->spa_spares.sav_vdevs[i], B_TRUE, VDEV_CONFIG_SPARE); VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, spares, spa->spa_spares.sav_count) == 0); for (i = 0; i < spa->spa_spares.sav_count; i++) nvlist_free(spares[i]); kmem_free(spares, spa->spa_spares.sav_count * sizeof (void *)); } /* * Load (or re-load) the current list of vdevs describing the active l2cache for * this pool. When this is called, we have some form of basic information in * 'spa_l2cache.sav_config'. We parse this into vdevs, try to open them, and * then re-generate a more complete list including status information. * Devices which are already active have their details maintained, and are * not re-opened. */ static void spa_load_l2cache(spa_t *spa) { nvlist_t **l2cache; uint_t nl2cache; int i, j, oldnvdevs; uint64_t guid; vdev_t *vd, **oldvdevs, **newvdevs; spa_aux_vdev_t *sav = &spa->spa_l2cache; ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); if (sav->sav_config != NULL) { VERIFY(nvlist_lookup_nvlist_array(sav->sav_config, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0); newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_SLEEP); } else { nl2cache = 0; newvdevs = NULL; } oldvdevs = sav->sav_vdevs; oldnvdevs = sav->sav_count; sav->sav_vdevs = NULL; sav->sav_count = 0; /* * Process new nvlist of vdevs. */ for (i = 0; i < nl2cache; i++) { VERIFY(nvlist_lookup_uint64(l2cache[i], ZPOOL_CONFIG_GUID, &guid) == 0); newvdevs[i] = NULL; for (j = 0; j < oldnvdevs; j++) { vd = oldvdevs[j]; if (vd != NULL && guid == vd->vdev_guid) { /* * Retain previous vdev for add/remove ops. */ newvdevs[i] = vd; oldvdevs[j] = NULL; break; } } if (newvdevs[i] == NULL) { /* * Create new vdev */ VERIFY(spa_config_parse(spa, &vd, l2cache[i], NULL, 0, VDEV_ALLOC_L2CACHE) == 0); ASSERT(vd != NULL); newvdevs[i] = vd; /* * Commit this vdev as an l2cache device, * even if it fails to open. */ spa_l2cache_add(vd); vd->vdev_top = vd; vd->vdev_aux = sav; spa_l2cache_activate(vd); if (vdev_open(vd) != 0) continue; (void) vdev_validate_aux(vd); if (!vdev_is_dead(vd)) l2arc_add_vdev(spa, vd); } } /* * Purge vdevs that were dropped */ for (i = 0; i < oldnvdevs; i++) { uint64_t pool; vd = oldvdevs[i]; if (vd != NULL) { ASSERT(vd->vdev_isl2cache); if (spa_l2cache_exists(vd->vdev_guid, &pool) && pool != 0ULL && l2arc_vdev_present(vd)) l2arc_remove_vdev(vd); vdev_clear_stats(vd); vdev_free(vd); } } if (oldvdevs) kmem_free(oldvdevs, oldnvdevs * sizeof (void *)); if (sav->sav_config == NULL) goto out; sav->sav_vdevs = newvdevs; sav->sav_count = (int)nl2cache; /* * Recompute the stashed list of l2cache devices, with status * information this time. */ VERIFY(nvlist_remove(sav->sav_config, ZPOOL_CONFIG_L2CACHE, DATA_TYPE_NVLIST_ARRAY) == 0); l2cache = kmem_alloc(sav->sav_count * sizeof (void *), KM_SLEEP); for (i = 0; i < sav->sav_count; i++) l2cache[i] = vdev_config_generate(spa, sav->sav_vdevs[i], B_TRUE, VDEV_CONFIG_L2CACHE); VERIFY(nvlist_add_nvlist_array(sav->sav_config, ZPOOL_CONFIG_L2CACHE, l2cache, sav->sav_count) == 0); out: for (i = 0; i < sav->sav_count; i++) nvlist_free(l2cache[i]); if (sav->sav_count) kmem_free(l2cache, sav->sav_count * sizeof (void *)); } static int load_nvlist(spa_t *spa, uint64_t obj, nvlist_t **value) { dmu_buf_t *db; char *packed = NULL; size_t nvsize = 0; int error; *value = NULL; error = dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db); if (error != 0) return (error); nvsize = *(uint64_t *)db->db_data; dmu_buf_rele(db, FTAG); packed = kmem_alloc(nvsize, KM_SLEEP); error = dmu_read(spa->spa_meta_objset, obj, 0, nvsize, packed, DMU_READ_PREFETCH); if (error == 0) error = nvlist_unpack(packed, nvsize, value, 0); kmem_free(packed, nvsize); return (error); } /* * Checks to see if the given vdev could not be opened, in which case we post a * sysevent to notify the autoreplace code that the device has been removed. */ static void spa_check_removed(vdev_t *vd) { for (int c = 0; c < vd->vdev_children; c++) spa_check_removed(vd->vdev_child[c]); if (vd->vdev_ops->vdev_op_leaf && vdev_is_dead(vd) && !vd->vdev_ishole) { zfs_post_autoreplace(vd->vdev_spa, vd); spa_event_notify(vd->vdev_spa, vd, ESC_ZFS_VDEV_CHECK); } } /* * Validate the current config against the MOS config */ static boolean_t spa_config_valid(spa_t *spa, nvlist_t *config) { vdev_t *mrvd, *rvd = spa->spa_root_vdev; nvlist_t *nv; VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nv) == 0); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); VERIFY(spa_config_parse(spa, &mrvd, nv, NULL, 0, VDEV_ALLOC_LOAD) == 0); ASSERT3U(rvd->vdev_children, ==, mrvd->vdev_children); /* * If we're doing a normal import, then build up any additional * diagnostic information about missing devices in this config. * We'll pass this up to the user for further processing. */ if (!(spa->spa_import_flags & ZFS_IMPORT_MISSING_LOG)) { nvlist_t **child, *nv; uint64_t idx = 0; child = kmem_alloc(rvd->vdev_children * sizeof (nvlist_t **), KM_SLEEP); VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); for (int c = 0; c < rvd->vdev_children; c++) { vdev_t *tvd = rvd->vdev_child[c]; vdev_t *mtvd = mrvd->vdev_child[c]; if (tvd->vdev_ops == &vdev_missing_ops && mtvd->vdev_ops != &vdev_missing_ops && mtvd->vdev_islog) child[idx++] = vdev_config_generate(spa, mtvd, B_FALSE, 0); } if (idx) { VERIFY(nvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, child, idx) == 0); VERIFY(nvlist_add_nvlist(spa->spa_load_info, ZPOOL_CONFIG_MISSING_DEVICES, nv) == 0); for (int i = 0; i < idx; i++) nvlist_free(child[i]); } nvlist_free(nv); kmem_free(child, rvd->vdev_children * sizeof (char **)); } /* * Compare the root vdev tree with the information we have * from the MOS config (mrvd). Check each top-level vdev * with the corresponding MOS config top-level (mtvd). */ for (int c = 0; c < rvd->vdev_children; c++) { vdev_t *tvd = rvd->vdev_child[c]; vdev_t *mtvd = mrvd->vdev_child[c]; /* * Resolve any "missing" vdevs in the current configuration. * If we find that the MOS config has more accurate information * about the top-level vdev then use that vdev instead. */ if (tvd->vdev_ops == &vdev_missing_ops && mtvd->vdev_ops != &vdev_missing_ops) { if (!(spa->spa_import_flags & ZFS_IMPORT_MISSING_LOG)) continue; /* * Device specific actions. */ if (mtvd->vdev_islog) { spa_set_log_state(spa, SPA_LOG_CLEAR); } else { /* * XXX - once we have 'readonly' pool * support we should be able to handle * missing data devices by transitioning * the pool to readonly. */ continue; } /* * Swap the missing vdev with the data we were * able to obtain from the MOS config. */ vdev_remove_child(rvd, tvd); vdev_remove_child(mrvd, mtvd); vdev_add_child(rvd, mtvd); vdev_add_child(mrvd, tvd); spa_config_exit(spa, SCL_ALL, FTAG); vdev_load(mtvd); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); vdev_reopen(rvd); } else if (mtvd->vdev_islog) { /* * Load the slog device's state from the MOS config * since it's possible that the label does not * contain the most up-to-date information. */ vdev_load_log_state(tvd, mtvd); vdev_reopen(tvd); } } vdev_free(mrvd); spa_config_exit(spa, SCL_ALL, FTAG); /* * Ensure we were able to validate the config. */ return (rvd->vdev_guid_sum == spa->spa_uberblock.ub_guid_sum); } /* * Check for missing log devices */ static boolean_t spa_check_logs(spa_t *spa) { boolean_t rv = B_FALSE; dsl_pool_t *dp = spa_get_dsl(spa); switch (spa->spa_log_state) { case SPA_LOG_MISSING: /* need to recheck in case slog has been restored */ case SPA_LOG_UNKNOWN: rv = (dmu_objset_find_dp(dp, dp->dp_root_dir_obj, zil_check_log_chain, NULL, DS_FIND_CHILDREN) != 0); if (rv) spa_set_log_state(spa, SPA_LOG_MISSING); break; } return (rv); } static boolean_t spa_passivate_log(spa_t *spa) { vdev_t *rvd = spa->spa_root_vdev; boolean_t slog_found = B_FALSE; ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER)); if (!spa_has_slogs(spa)) return (B_FALSE); for (int c = 0; c < rvd->vdev_children; c++) { vdev_t *tvd = rvd->vdev_child[c]; metaslab_group_t *mg = tvd->vdev_mg; if (tvd->vdev_islog) { metaslab_group_passivate(mg); slog_found = B_TRUE; } } return (slog_found); } static void spa_activate_log(spa_t *spa) { vdev_t *rvd = spa->spa_root_vdev; ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER)); for (int c = 0; c < rvd->vdev_children; c++) { vdev_t *tvd = rvd->vdev_child[c]; metaslab_group_t *mg = tvd->vdev_mg; if (tvd->vdev_islog) metaslab_group_activate(mg); } } int spa_offline_log(spa_t *spa) { int error; error = dmu_objset_find(spa_name(spa), zil_vdev_offline, NULL, DS_FIND_CHILDREN); if (error == 0) { /* * We successfully offlined the log device, sync out the * current txg so that the "stubby" block can be removed * by zil_sync(). */ txg_wait_synced(spa->spa_dsl_pool, 0); } return (error); } static void spa_aux_check_removed(spa_aux_vdev_t *sav) { int i; for (i = 0; i < sav->sav_count; i++) spa_check_removed(sav->sav_vdevs[i]); } void spa_claim_notify(zio_t *zio) { spa_t *spa = zio->io_spa; if (zio->io_error) return; mutex_enter(&spa->spa_props_lock); /* any mutex will do */ if (spa->spa_claim_max_txg < zio->io_bp->blk_birth) spa->spa_claim_max_txg = zio->io_bp->blk_birth; mutex_exit(&spa->spa_props_lock); } typedef struct spa_load_error { uint64_t sle_meta_count; uint64_t sle_data_count; } spa_load_error_t; static void spa_load_verify_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; spa_load_error_t *sle = zio->io_private; dmu_object_type_t type = BP_GET_TYPE(bp); int error = zio->io_error; spa_t *spa = zio->io_spa; if (error) { if ((BP_GET_LEVEL(bp) != 0 || DMU_OT_IS_METADATA(type)) && type != DMU_OT_INTENT_LOG) atomic_inc_64(&sle->sle_meta_count); else atomic_inc_64(&sle->sle_data_count); } zio_data_buf_free(zio->io_data, zio->io_size); mutex_enter(&spa->spa_scrub_lock); spa->spa_scrub_inflight--; cv_broadcast(&spa->spa_scrub_io_cv); mutex_exit(&spa->spa_scrub_lock); } /* * Maximum number of concurrent scrub i/os to create while verifying * a pool while importing it. */ int spa_load_verify_maxinflight = 10000; boolean_t spa_load_verify_metadata = B_TRUE; boolean_t spa_load_verify_data = B_TRUE; SYSCTL_INT(_vfs_zfs, OID_AUTO, spa_load_verify_maxinflight, CTLFLAG_RWTUN, &spa_load_verify_maxinflight, 0, "Maximum number of concurrent scrub I/Os to create while verifying a " "pool while importing it"); SYSCTL_INT(_vfs_zfs, OID_AUTO, spa_load_verify_metadata, CTLFLAG_RWTUN, &spa_load_verify_metadata, 0, "Check metadata on import?"); SYSCTL_INT(_vfs_zfs, OID_AUTO, spa_load_verify_data, CTLFLAG_RWTUN, &spa_load_verify_data, 0, "Check user data on import?"); /*ARGSUSED*/ static int spa_load_verify_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, const zbookmark_phys_t *zb, const dnode_phys_t *dnp, void *arg) { if (bp == NULL || BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp)) return (0); /* * Note: normally this routine will not be called if * spa_load_verify_metadata is not set. However, it may be useful * to manually set the flag after the traversal has begun. */ if (!spa_load_verify_metadata) return (0); if (BP_GET_BUFC_TYPE(bp) == ARC_BUFC_DATA && !spa_load_verify_data) return (0); zio_t *rio = arg; size_t size = BP_GET_PSIZE(bp); void *data = zio_data_buf_alloc(size); mutex_enter(&spa->spa_scrub_lock); while (spa->spa_scrub_inflight >= spa_load_verify_maxinflight) cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock); spa->spa_scrub_inflight++; mutex_exit(&spa->spa_scrub_lock); zio_nowait(zio_read(rio, spa, bp, data, size, spa_load_verify_done, rio->io_private, ZIO_PRIORITY_SCRUB, ZIO_FLAG_SPECULATIVE | ZIO_FLAG_CANFAIL | ZIO_FLAG_SCRUB | ZIO_FLAG_RAW, zb)); return (0); } static int spa_load_verify(spa_t *spa) { zio_t *rio; spa_load_error_t sle = { 0 }; zpool_rewind_policy_t policy; boolean_t verify_ok = B_FALSE; int error = 0; zpool_get_rewind_policy(spa->spa_config, &policy); if (policy.zrp_request & ZPOOL_NEVER_REWIND) return (0); rio = zio_root(spa, NULL, &sle, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE); if (spa_load_verify_metadata) { error = traverse_pool(spa, spa->spa_verify_min_txg, TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA, spa_load_verify_cb, rio); } (void) zio_wait(rio); spa->spa_load_meta_errors = sle.sle_meta_count; spa->spa_load_data_errors = sle.sle_data_count; if (!error && sle.sle_meta_count <= policy.zrp_maxmeta && sle.sle_data_count <= policy.zrp_maxdata) { int64_t loss = 0; verify_ok = B_TRUE; spa->spa_load_txg = spa->spa_uberblock.ub_txg; spa->spa_load_txg_ts = spa->spa_uberblock.ub_timestamp; loss = spa->spa_last_ubsync_txg_ts - spa->spa_load_txg_ts; VERIFY(nvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_LOAD_TIME, spa->spa_load_txg_ts) == 0); VERIFY(nvlist_add_int64(spa->spa_load_info, ZPOOL_CONFIG_REWIND_TIME, loss) == 0); VERIFY(nvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_LOAD_DATA_ERRORS, sle.sle_data_count) == 0); } else { spa->spa_load_max_txg = spa->spa_uberblock.ub_txg; } if (error) { if (error != ENXIO && error != EIO) error = SET_ERROR(EIO); return (error); } return (verify_ok ? 0 : EIO); } /* * Find a value in the pool props object. */ static void spa_prop_find(spa_t *spa, zpool_prop_t prop, uint64_t *val) { (void) zap_lookup(spa->spa_meta_objset, spa->spa_pool_props_object, zpool_prop_to_name(prop), sizeof (uint64_t), 1, val); } /* * Find a value in the pool directory object. */ static int spa_dir_prop(spa_t *spa, const char *name, uint64_t *val) { return (zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, name, sizeof (uint64_t), 1, val)); } static int spa_vdev_err(vdev_t *vdev, vdev_aux_t aux, int err) { vdev_set_state(vdev, B_TRUE, VDEV_STATE_CANT_OPEN, aux); return (err); } /* * Fix up config after a partly-completed split. This is done with the * ZPOOL_CONFIG_SPLIT nvlist. Both the splitting pool and the split-off * pool have that entry in their config, but only the splitting one contains * a list of all the guids of the vdevs that are being split off. * * This function determines what to do with that list: either rejoin * all the disks to the pool, or complete the splitting process. To attempt * the rejoin, each disk that is offlined is marked online again, and * we do a reopen() call. If the vdev label for every disk that was * marked online indicates it was successfully split off (VDEV_AUX_SPLIT_POOL) * then we call vdev_split() on each disk, and complete the split. * * Otherwise we leave the config alone, with all the vdevs in place in * the original pool. */ static void spa_try_repair(spa_t *spa, nvlist_t *config) { uint_t extracted; uint64_t *glist; uint_t i, gcount; nvlist_t *nvl; vdev_t **vd; boolean_t attempt_reopen; if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) != 0) return; /* check that the config is complete */ if (nvlist_lookup_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST, &glist, &gcount) != 0) return; vd = kmem_zalloc(gcount * sizeof (vdev_t *), KM_SLEEP); /* attempt to online all the vdevs & validate */ attempt_reopen = B_TRUE; for (i = 0; i < gcount; i++) { if (glist[i] == 0) /* vdev is hole */ continue; vd[i] = spa_lookup_by_guid(spa, glist[i], B_FALSE); if (vd[i] == NULL) { /* * Don't bother attempting to reopen the disks; * just do the split. */ attempt_reopen = B_FALSE; } else { /* attempt to re-online it */ vd[i]->vdev_offline = B_FALSE; } } if (attempt_reopen) { vdev_reopen(spa->spa_root_vdev); /* check each device to see what state it's in */ for (extracted = 0, i = 0; i < gcount; i++) { if (vd[i] != NULL && vd[i]->vdev_stat.vs_aux != VDEV_AUX_SPLIT_POOL) break; ++extracted; } } /* * If every disk has been moved to the new pool, or if we never * even attempted to look at them, then we split them off for * good. */ if (!attempt_reopen || gcount == extracted) { for (i = 0; i < gcount; i++) if (vd[i] != NULL) vdev_split(vd[i]); vdev_reopen(spa->spa_root_vdev); } kmem_free(vd, gcount * sizeof (vdev_t *)); } static int spa_load(spa_t *spa, spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig) { nvlist_t *config = spa->spa_config; char *ereport = FM_EREPORT_ZFS_POOL; char *comment; int error; uint64_t pool_guid; nvlist_t *nvl; if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid)) return (SET_ERROR(EINVAL)); ASSERT(spa->spa_comment == NULL); if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0) spa->spa_comment = spa_strdup(comment); /* * Versioning wasn't explicitly added to the label until later, so if * it's not present treat it as the initial version. */ if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &spa->spa_ubsync.ub_version) != 0) spa->spa_ubsync.ub_version = SPA_VERSION_INITIAL; (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, &spa->spa_config_txg); if ((state == SPA_LOAD_IMPORT || state == SPA_LOAD_TRYIMPORT) && spa_guid_exists(pool_guid, 0)) { error = SET_ERROR(EEXIST); } else { spa->spa_config_guid = pool_guid; if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) == 0) { VERIFY(nvlist_dup(nvl, &spa->spa_config_splitting, KM_SLEEP) == 0); } nvlist_free(spa->spa_load_info); spa->spa_load_info = fnvlist_alloc(); gethrestime(&spa->spa_loaded_ts); error = spa_load_impl(spa, pool_guid, config, state, type, mosconfig, &ereport); } /* * Don't count references from objsets that are already closed * and are making their way through the eviction process. */ spa_evicting_os_wait(spa); spa->spa_minref = refcount_count(&spa->spa_refcount); if (error) { if (error != EEXIST) { spa->spa_loaded_ts.tv_sec = 0; spa->spa_loaded_ts.tv_nsec = 0; } if (error != EBADF) { zfs_ereport_post(ereport, spa, NULL, NULL, 0, 0); } } spa->spa_load_state = error ? SPA_LOAD_ERROR : SPA_LOAD_NONE; spa->spa_ena = 0; return (error); } /* * Load an existing storage pool, using the pool's builtin spa_config as a * source of configuration information. */ static int spa_load_impl(spa_t *spa, uint64_t pool_guid, nvlist_t *config, spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig, char **ereport) { int error = 0; nvlist_t *nvroot = NULL; nvlist_t *label; vdev_t *rvd; uberblock_t *ub = &spa->spa_uberblock; uint64_t children, config_cache_txg = spa->spa_config_txg; int orig_mode = spa->spa_mode; int parse; uint64_t obj; boolean_t missing_feat_write = B_FALSE; /* * If this is an untrusted config, access the pool in read-only mode. * This prevents things like resilvering recently removed devices. */ if (!mosconfig) spa->spa_mode = FREAD; ASSERT(MUTEX_HELD(&spa_namespace_lock)); spa->spa_load_state = state; if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot)) return (SET_ERROR(EINVAL)); parse = (type == SPA_IMPORT_EXISTING ? VDEV_ALLOC_LOAD : VDEV_ALLOC_SPLIT); /* * Create "The Godfather" zio to hold all async IOs */ spa->spa_async_zio_root = kmem_alloc(max_ncpus * sizeof (void *), KM_SLEEP); for (int i = 0; i < max_ncpus; i++) { spa->spa_async_zio_root[i] = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_GODFATHER); } /* * Parse the configuration into a vdev tree. We explicitly set the * value that will be returned by spa_version() since parsing the * configuration requires knowing the version number. */ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, parse); spa_config_exit(spa, SCL_ALL, FTAG); if (error != 0) return (error); ASSERT(spa->spa_root_vdev == rvd); ASSERT3U(spa->spa_min_ashift, >=, SPA_MINBLOCKSHIFT); ASSERT3U(spa->spa_max_ashift, <=, SPA_MAXBLOCKSHIFT); if (type != SPA_IMPORT_ASSEMBLE) { ASSERT(spa_guid(spa) == pool_guid); } /* * Try to open all vdevs, loading each label in the process. */ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); error = vdev_open(rvd); spa_config_exit(spa, SCL_ALL, FTAG); if (error != 0) return (error); /* * We need to validate the vdev labels against the configuration that * we have in hand, which is dependent on the setting of mosconfig. If * mosconfig is true then we're validating the vdev labels based on * that config. Otherwise, we're validating against the cached config * (zpool.cache) that was read when we loaded the zfs module, and then * later we will recursively call spa_load() and validate against * the vdev config. * * If we're assembling a new pool that's been split off from an * existing pool, the labels haven't yet been updated so we skip * validation for now. */ if (type != SPA_IMPORT_ASSEMBLE) { spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); error = vdev_validate(rvd, mosconfig); spa_config_exit(spa, SCL_ALL, FTAG); if (error != 0) return (error); if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) return (SET_ERROR(ENXIO)); } /* * Find the best uberblock. */ vdev_uberblock_load(rvd, ub, &label); /* * If we weren't able to find a single valid uberblock, return failure. */ if (ub->ub_txg == 0) { nvlist_free(label); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO)); } /* * If the pool has an unsupported version we can't open it. */ if (!SPA_VERSION_IS_SUPPORTED(ub->ub_version)) { nvlist_free(label); return (spa_vdev_err(rvd, VDEV_AUX_VERSION_NEWER, ENOTSUP)); } if (ub->ub_version >= SPA_VERSION_FEATURES) { nvlist_t *features; /* * If we weren't able to find what's necessary for reading the * MOS in the label, return failure. */ if (label == NULL || nvlist_lookup_nvlist(label, ZPOOL_CONFIG_FEATURES_FOR_READ, &features) != 0) { nvlist_free(label); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO)); } /* * Update our in-core representation with the definitive values * from the label. */ nvlist_free(spa->spa_label_features); VERIFY(nvlist_dup(features, &spa->spa_label_features, 0) == 0); } nvlist_free(label); /* * Look through entries in the label nvlist's features_for_read. If * there is a feature listed there which we don't understand then we * cannot open a pool. */ if (ub->ub_version >= SPA_VERSION_FEATURES) { nvlist_t *unsup_feat; VERIFY(nvlist_alloc(&unsup_feat, NV_UNIQUE_NAME, KM_SLEEP) == 0); for (nvpair_t *nvp = nvlist_next_nvpair(spa->spa_label_features, NULL); nvp != NULL; nvp = nvlist_next_nvpair(spa->spa_label_features, nvp)) { if (!zfeature_is_supported(nvpair_name(nvp))) { VERIFY(nvlist_add_string(unsup_feat, nvpair_name(nvp), "") == 0); } } if (!nvlist_empty(unsup_feat)) { VERIFY(nvlist_add_nvlist(spa->spa_load_info, ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat) == 0); nvlist_free(unsup_feat); return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT, ENOTSUP)); } nvlist_free(unsup_feat); } /* * If the vdev guid sum doesn't match the uberblock, we have an * incomplete configuration. We first check to see if the pool * is aware of the complete config (i.e ZPOOL_CONFIG_VDEV_CHILDREN). * If it is, defer the vdev_guid_sum check till later so we * can handle missing vdevs. */ if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN, &children) != 0 && mosconfig && type != SPA_IMPORT_ASSEMBLE && rvd->vdev_guid_sum != ub->ub_guid_sum) return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM, ENXIO)); if (type != SPA_IMPORT_ASSEMBLE && spa->spa_config_splitting) { spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_try_repair(spa, config); spa_config_exit(spa, SCL_ALL, FTAG); nvlist_free(spa->spa_config_splitting); spa->spa_config_splitting = NULL; } /* * Initialize internal SPA structures. */ spa->spa_state = POOL_STATE_ACTIVE; spa->spa_ubsync = spa->spa_uberblock; spa->spa_verify_min_txg = spa->spa_extreme_rewind ? TXG_INITIAL - 1 : spa_last_synced_txg(spa) - TXG_DEFER_SIZE - 1; spa->spa_first_txg = spa->spa_last_ubsync_txg ? spa->spa_last_ubsync_txg : spa_last_synced_txg(spa) + 1; spa->spa_claim_max_txg = spa->spa_first_txg; spa->spa_prev_software_version = ub->ub_software_version; error = dsl_pool_init(spa, spa->spa_first_txg, &spa->spa_dsl_pool); if (error) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); spa->spa_meta_objset = spa->spa_dsl_pool->dp_meta_objset; if (spa_dir_prop(spa, DMU_POOL_CONFIG, &spa->spa_config_object) != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); if (spa_version(spa) >= SPA_VERSION_FEATURES) { boolean_t missing_feat_read = B_FALSE; nvlist_t *unsup_feat, *enabled_feat; if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_READ, &spa->spa_feat_for_read_obj) != 0) { return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_WRITE, &spa->spa_feat_for_write_obj) != 0) { return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } if (spa_dir_prop(spa, DMU_POOL_FEATURE_DESCRIPTIONS, &spa->spa_feat_desc_obj) != 0) { return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } enabled_feat = fnvlist_alloc(); unsup_feat = fnvlist_alloc(); if (!spa_features_check(spa, B_FALSE, unsup_feat, enabled_feat)) missing_feat_read = B_TRUE; if (spa_writeable(spa) || state == SPA_LOAD_TRYIMPORT) { if (!spa_features_check(spa, B_TRUE, unsup_feat, enabled_feat)) { missing_feat_write = B_TRUE; } } fnvlist_add_nvlist(spa->spa_load_info, ZPOOL_CONFIG_ENABLED_FEAT, enabled_feat); if (!nvlist_empty(unsup_feat)) { fnvlist_add_nvlist(spa->spa_load_info, ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat); } fnvlist_free(enabled_feat); fnvlist_free(unsup_feat); if (!missing_feat_read) { fnvlist_add_boolean(spa->spa_load_info, ZPOOL_CONFIG_CAN_RDONLY); } /* * If the state is SPA_LOAD_TRYIMPORT, our objective is * twofold: to determine whether the pool is available for * import in read-write mode and (if it is not) whether the * pool is available for import in read-only mode. If the pool * is available for import in read-write mode, it is displayed * as available in userland; if it is not available for import * in read-only mode, it is displayed as unavailable in * userland. If the pool is available for import in read-only * mode but not read-write mode, it is displayed as unavailable * in userland with a special note that the pool is actually * available for open in read-only mode. * * As a result, if the state is SPA_LOAD_TRYIMPORT and we are * missing a feature for write, we must first determine whether * the pool can be opened read-only before returning to * userland in order to know whether to display the * abovementioned note. */ if (missing_feat_read || (missing_feat_write && spa_writeable(spa))) { return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT, ENOTSUP)); } /* * Load refcounts for ZFS features from disk into an in-memory * cache during SPA initialization. */ for (spa_feature_t i = 0; i < SPA_FEATURES; i++) { uint64_t refcount; error = feature_get_refcount_from_disk(spa, &spa_feature_table[i], &refcount); if (error == 0) { spa->spa_feat_refcount_cache[i] = refcount; } else if (error == ENOTSUP) { spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED; } else { return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } } } if (spa_feature_is_active(spa, SPA_FEATURE_ENABLED_TXG)) { if (spa_dir_prop(spa, DMU_POOL_FEATURE_ENABLED_TXG, &spa->spa_feat_enabled_txg_obj) != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); } spa->spa_is_initializing = B_TRUE; error = dsl_pool_open(spa->spa_dsl_pool); spa->spa_is_initializing = B_FALSE; if (error != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); if (!mosconfig) { uint64_t hostid; nvlist_t *policy = NULL, *nvconfig; if (load_nvlist(spa, spa->spa_config_object, &nvconfig) != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); if (!spa_is_root(spa) && nvlist_lookup_uint64(nvconfig, ZPOOL_CONFIG_HOSTID, &hostid) == 0) { char *hostname; unsigned long myhostid = 0; VERIFY(nvlist_lookup_string(nvconfig, ZPOOL_CONFIG_HOSTNAME, &hostname) == 0); #ifdef _KERNEL myhostid = zone_get_hostid(NULL); #else /* _KERNEL */ /* * We're emulating the system's hostid in userland, so * we can't use zone_get_hostid(). */ (void) ddi_strtoul(hw_serial, NULL, 10, &myhostid); #endif /* _KERNEL */ if (check_hostid && hostid != 0 && myhostid != 0 && hostid != myhostid) { nvlist_free(nvconfig); cmn_err(CE_WARN, "pool '%s' could not be " "loaded as it was last accessed by " "another system (host: %s hostid: 0x%lx). " "See: http://illumos.org/msg/ZFS-8000-EY", spa_name(spa), hostname, (unsigned long)hostid); return (SET_ERROR(EBADF)); } } if (nvlist_lookup_nvlist(spa->spa_config, ZPOOL_REWIND_POLICY, &policy) == 0) VERIFY(nvlist_add_nvlist(nvconfig, ZPOOL_REWIND_POLICY, policy) == 0); spa_config_set(spa, nvconfig); spa_unload(spa); spa_deactivate(spa); spa_activate(spa, orig_mode); return (spa_load(spa, state, SPA_IMPORT_EXISTING, B_TRUE)); } + /* Grab the secret checksum salt from the MOS. */ + error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, + DMU_POOL_CHECKSUM_SALT, 1, + sizeof (spa->spa_cksum_salt.zcs_bytes), + spa->spa_cksum_salt.zcs_bytes); + if (error == ENOENT) { + /* Generate a new salt for subsequent use */ + (void) random_get_pseudo_bytes(spa->spa_cksum_salt.zcs_bytes, + sizeof (spa->spa_cksum_salt.zcs_bytes)); + } else if (error != 0) { + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); + } + if (spa_dir_prop(spa, DMU_POOL_SYNC_BPOBJ, &obj) != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); error = bpobj_open(&spa->spa_deferred_bpobj, spa->spa_meta_objset, obj); if (error != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); /* * Load the bit that tells us to use the new accounting function * (raid-z deflation). If we have an older pool, this will not * be present. */ error = spa_dir_prop(spa, DMU_POOL_DEFLATE, &spa->spa_deflate); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); error = spa_dir_prop(spa, DMU_POOL_CREATION_VERSION, &spa->spa_creation_version); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); /* * Load the persistent error log. If we have an older pool, this will * not be present. */ error = spa_dir_prop(spa, DMU_POOL_ERRLOG_LAST, &spa->spa_errlog_last); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); error = spa_dir_prop(spa, DMU_POOL_ERRLOG_SCRUB, &spa->spa_errlog_scrub); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); /* * Load the history object. If we have an older pool, this * will not be present. */ error = spa_dir_prop(spa, DMU_POOL_HISTORY, &spa->spa_history); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); /* * If we're assembling the pool from the split-off vdevs of * an existing pool, we don't want to attach the spares & cache * devices. */ /* * Load any hot spares for this pool. */ error = spa_dir_prop(spa, DMU_POOL_SPARES, &spa->spa_spares.sav_object); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); if (error == 0 && type != SPA_IMPORT_ASSEMBLE) { ASSERT(spa_version(spa) >= SPA_VERSION_SPARES); if (load_nvlist(spa, spa->spa_spares.sav_object, &spa->spa_spares.sav_config) != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_spares(spa); spa_config_exit(spa, SCL_ALL, FTAG); } else if (error == 0) { spa->spa_spares.sav_sync = B_TRUE; } /* * Load any level 2 ARC devices for this pool. */ error = spa_dir_prop(spa, DMU_POOL_L2CACHE, &spa->spa_l2cache.sav_object); if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); if (error == 0 && type != SPA_IMPORT_ASSEMBLE) { ASSERT(spa_version(spa) >= SPA_VERSION_L2CACHE); if (load_nvlist(spa, spa->spa_l2cache.sav_object, &spa->spa_l2cache.sav_config) != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_l2cache(spa); spa_config_exit(spa, SCL_ALL, FTAG); } else if (error == 0) { spa->spa_l2cache.sav_sync = B_TRUE; } spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION); error = spa_dir_prop(spa, DMU_POOL_PROPS, &spa->spa_pool_props_object); if (error && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); if (error == 0) { uint64_t autoreplace; spa_prop_find(spa, ZPOOL_PROP_BOOTFS, &spa->spa_bootfs); spa_prop_find(spa, ZPOOL_PROP_AUTOREPLACE, &autoreplace); spa_prop_find(spa, ZPOOL_PROP_DELEGATION, &spa->spa_delegation); spa_prop_find(spa, ZPOOL_PROP_FAILUREMODE, &spa->spa_failmode); spa_prop_find(spa, ZPOOL_PROP_AUTOEXPAND, &spa->spa_autoexpand); spa_prop_find(spa, ZPOOL_PROP_DEDUPDITTO, &spa->spa_dedup_ditto); spa->spa_autoreplace = (autoreplace != 0); } /* * If the 'autoreplace' property is set, then post a resource notifying * the ZFS DE that it should not issue any faults for unopenable * devices. We also iterate over the vdevs, and post a sysevent for any * unopenable vdevs so that the normal autoreplace handler can take * over. */ if (spa->spa_autoreplace && state != SPA_LOAD_TRYIMPORT) { spa_check_removed(spa->spa_root_vdev); /* * For the import case, this is done in spa_import(), because * at this point we're using the spare definitions from * the MOS config, not necessarily from the userland config. */ if (state != SPA_LOAD_IMPORT) { spa_aux_check_removed(&spa->spa_spares); spa_aux_check_removed(&spa->spa_l2cache); } } /* * Load the vdev state for all toplevel vdevs. */ vdev_load(rvd); /* * Propagate the leaf DTLs we just loaded all the way up the tree. */ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); vdev_dtl_reassess(rvd, 0, 0, B_FALSE); spa_config_exit(spa, SCL_ALL, FTAG); /* * Load the DDTs (dedup tables). */ error = ddt_load(spa); if (error != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); spa_update_dspace(spa); /* * Validate the config, using the MOS config to fill in any * information which might be missing. If we fail to validate * the config then declare the pool unfit for use. If we're * assembling a pool from a split, the log is not transferred * over. */ if (type != SPA_IMPORT_ASSEMBLE) { nvlist_t *nvconfig; if (load_nvlist(spa, spa->spa_config_object, &nvconfig) != 0) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); if (!spa_config_valid(spa, nvconfig)) { nvlist_free(nvconfig); return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM, ENXIO)); } nvlist_free(nvconfig); /* * Now that we've validated the config, check the state of the * root vdev. If it can't be opened, it indicates one or * more toplevel vdevs are faulted. */ if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) return (SET_ERROR(ENXIO)); if (spa_writeable(spa) && spa_check_logs(spa)) { *ereport = FM_EREPORT_ZFS_LOG_REPLAY; return (spa_vdev_err(rvd, VDEV_AUX_BAD_LOG, ENXIO)); } } if (missing_feat_write) { ASSERT(state == SPA_LOAD_TRYIMPORT); /* * At this point, we know that we can open the pool in * read-only mode but not read-write mode. We now have enough * information and can return to userland. */ return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT, ENOTSUP)); } /* * We've successfully opened the pool, verify that we're ready * to start pushing transactions. */ if (state != SPA_LOAD_TRYIMPORT) { if (error = spa_load_verify(spa)) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error)); } if (spa_writeable(spa) && (state == SPA_LOAD_RECOVER || spa->spa_load_max_txg == UINT64_MAX)) { dmu_tx_t *tx; int need_update = B_FALSE; dsl_pool_t *dp = spa_get_dsl(spa); ASSERT(state != SPA_LOAD_TRYIMPORT); /* * Claim log blocks that haven't been committed yet. * This must all happen in a single txg. * Note: spa_claim_max_txg is updated by spa_claim_notify(), * invoked from zil_claim_log_block()'s i/o done callback. * Price of rollback is that we abandon the log. */ spa->spa_claiming = B_TRUE; tx = dmu_tx_create_assigned(dp, spa_first_txg(spa)); (void) dmu_objset_find_dp(dp, dp->dp_root_dir_obj, zil_claim, tx, DS_FIND_CHILDREN); dmu_tx_commit(tx); spa->spa_claiming = B_FALSE; spa_set_log_state(spa, SPA_LOG_GOOD); spa->spa_sync_on = B_TRUE; txg_sync_start(spa->spa_dsl_pool); /* * Wait for all claims to sync. We sync up to the highest * claimed log block birth time so that claimed log blocks * don't appear to be from the future. spa_claim_max_txg * will have been set for us by either zil_check_log_chain() * (invoked from spa_check_logs()) or zil_claim() above. */ txg_wait_synced(spa->spa_dsl_pool, spa->spa_claim_max_txg); /* * If the config cache is stale, or we have uninitialized * metaslabs (see spa_vdev_add()), then update the config. * * If this is a verbatim import, trust the current * in-core spa_config and update the disk labels. */ if (config_cache_txg != spa->spa_config_txg || state == SPA_LOAD_IMPORT || state == SPA_LOAD_RECOVER || (spa->spa_import_flags & ZFS_IMPORT_VERBATIM)) need_update = B_TRUE; for (int c = 0; c < rvd->vdev_children; c++) if (rvd->vdev_child[c]->vdev_ms_array == 0) need_update = B_TRUE; /* * Update the config cache asychronously in case we're the * root pool, in which case the config cache isn't writable yet. */ if (need_update) spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE); /* * Check all DTLs to see if anything needs resilvering. */ if (!dsl_scan_resilvering(spa->spa_dsl_pool) && vdev_resilver_needed(rvd, NULL, NULL)) spa_async_request(spa, SPA_ASYNC_RESILVER); /* * Log the fact that we booted up (so that we can detect if * we rebooted in the middle of an operation). */ spa_history_log_version(spa, "open"); /* * Delete any inconsistent datasets. */ (void) dmu_objset_find(spa_name(spa), dsl_destroy_inconsistent, NULL, DS_FIND_CHILDREN); /* * Clean up any stale temporary dataset userrefs. */ dsl_pool_clean_tmp_userrefs(spa->spa_dsl_pool); } return (0); } static int spa_load_retry(spa_t *spa, spa_load_state_t state, int mosconfig) { int mode = spa->spa_mode; spa_unload(spa); spa_deactivate(spa); spa->spa_load_max_txg = spa->spa_uberblock.ub_txg - 1; spa_activate(spa, mode); spa_async_suspend(spa); return (spa_load(spa, state, SPA_IMPORT_EXISTING, mosconfig)); } /* * If spa_load() fails this function will try loading prior txg's. If * 'state' is SPA_LOAD_RECOVER and one of these loads succeeds the pool * will be rewound to that txg. If 'state' is not SPA_LOAD_RECOVER this * function will not rewind the pool and will return the same error as * spa_load(). */ static int spa_load_best(spa_t *spa, spa_load_state_t state, int mosconfig, uint64_t max_request, int rewind_flags) { nvlist_t *loadinfo = NULL; nvlist_t *config = NULL; int load_error, rewind_error; uint64_t safe_rewind_txg; uint64_t min_txg; if (spa->spa_load_txg && state == SPA_LOAD_RECOVER) { spa->spa_load_max_txg = spa->spa_load_txg; spa_set_log_state(spa, SPA_LOG_CLEAR); } else { spa->spa_load_max_txg = max_request; if (max_request != UINT64_MAX) spa->spa_extreme_rewind = B_TRUE; } load_error = rewind_error = spa_load(spa, state, SPA_IMPORT_EXISTING, mosconfig); if (load_error == 0) return (0); if (spa->spa_root_vdev != NULL) config = spa_config_generate(spa, NULL, -1ULL, B_TRUE); spa->spa_last_ubsync_txg = spa->spa_uberblock.ub_txg; spa->spa_last_ubsync_txg_ts = spa->spa_uberblock.ub_timestamp; if (rewind_flags & ZPOOL_NEVER_REWIND) { nvlist_free(config); return (load_error); } if (state == SPA_LOAD_RECOVER) { /* Price of rolling back is discarding txgs, including log */ spa_set_log_state(spa, SPA_LOG_CLEAR); } else { /* * If we aren't rolling back save the load info from our first * import attempt so that we can restore it after attempting * to rewind. */ loadinfo = spa->spa_load_info; spa->spa_load_info = fnvlist_alloc(); } spa->spa_load_max_txg = spa->spa_last_ubsync_txg; safe_rewind_txg = spa->spa_last_ubsync_txg - TXG_DEFER_SIZE; min_txg = (rewind_flags & ZPOOL_EXTREME_REWIND) ? TXG_INITIAL : safe_rewind_txg; /* * Continue as long as we're finding errors, we're still within * the acceptable rewind range, and we're still finding uberblocks */ while (rewind_error && spa->spa_uberblock.ub_txg >= min_txg && spa->spa_uberblock.ub_txg <= spa->spa_load_max_txg) { if (spa->spa_load_max_txg < safe_rewind_txg) spa->spa_extreme_rewind = B_TRUE; rewind_error = spa_load_retry(spa, state, mosconfig); } spa->spa_extreme_rewind = B_FALSE; spa->spa_load_max_txg = UINT64_MAX; if (config && (rewind_error || state != SPA_LOAD_RECOVER)) spa_config_set(spa, config); if (state == SPA_LOAD_RECOVER) { ASSERT3P(loadinfo, ==, NULL); return (rewind_error); } else { /* Store the rewind info as part of the initial load info */ fnvlist_add_nvlist(loadinfo, ZPOOL_CONFIG_REWIND_INFO, spa->spa_load_info); /* Restore the initial load info */ fnvlist_free(spa->spa_load_info); spa->spa_load_info = loadinfo; return (load_error); } } /* * Pool Open/Import * * The import case is identical to an open except that the configuration is sent * down from userland, instead of grabbed from the configuration cache. For the * case of an open, the pool configuration will exist in the * POOL_STATE_UNINITIALIZED state. * * The stats information (gen/count/ustats) is used to gather vdev statistics at * the same time open the pool, without having to keep around the spa_t in some * ambiguous state. */ static int spa_open_common(const char *pool, spa_t **spapp, void *tag, nvlist_t *nvpolicy, nvlist_t **config) { spa_t *spa; spa_load_state_t state = SPA_LOAD_OPEN; int error; int locked = B_FALSE; int firstopen = B_FALSE; *spapp = NULL; /* * As disgusting as this is, we need to support recursive calls to this * function because dsl_dir_open() is called during spa_load(), and ends * up calling spa_open() again. The real fix is to figure out how to * avoid dsl_dir_open() calling this in the first place. */ if (mutex_owner(&spa_namespace_lock) != curthread) { mutex_enter(&spa_namespace_lock); locked = B_TRUE; } if ((spa = spa_lookup(pool)) == NULL) { if (locked) mutex_exit(&spa_namespace_lock); return (SET_ERROR(ENOENT)); } if (spa->spa_state == POOL_STATE_UNINITIALIZED) { zpool_rewind_policy_t policy; firstopen = B_TRUE; zpool_get_rewind_policy(nvpolicy ? nvpolicy : spa->spa_config, &policy); if (policy.zrp_request & ZPOOL_DO_REWIND) state = SPA_LOAD_RECOVER; spa_activate(spa, spa_mode_global); if (state != SPA_LOAD_RECOVER) spa->spa_last_ubsync_txg = spa->spa_load_txg = 0; error = spa_load_best(spa, state, B_FALSE, policy.zrp_txg, policy.zrp_request); if (error == EBADF) { /* * If vdev_validate() returns failure (indicated by * EBADF), it indicates that one of the vdevs indicates * that the pool has been exported or destroyed. If * this is the case, the config cache is out of sync and * we should remove the pool from the namespace. */ spa_unload(spa); spa_deactivate(spa); spa_config_sync(spa, B_TRUE, B_TRUE); spa_remove(spa); if (locked) mutex_exit(&spa_namespace_lock); return (SET_ERROR(ENOENT)); } if (error) { /* * We can't open the pool, but we still have useful * information: the state of each vdev after the * attempted vdev_open(). Return this to the user. */ if (config != NULL && spa->spa_config) { VERIFY(nvlist_dup(spa->spa_config, config, KM_SLEEP) == 0); VERIFY(nvlist_add_nvlist(*config, ZPOOL_CONFIG_LOAD_INFO, spa->spa_load_info) == 0); } spa_unload(spa); spa_deactivate(spa); spa->spa_last_open_failed = error; if (locked) mutex_exit(&spa_namespace_lock); *spapp = NULL; return (error); } } spa_open_ref(spa, tag); if (config != NULL) *config = spa_config_generate(spa, NULL, -1ULL, B_TRUE); /* * If we've recovered the pool, pass back any information we * gathered while doing the load. */ if (state == SPA_LOAD_RECOVER) { VERIFY(nvlist_add_nvlist(*config, ZPOOL_CONFIG_LOAD_INFO, spa->spa_load_info) == 0); } if (locked) { spa->spa_last_open_failed = 0; spa->spa_last_ubsync_txg = 0; spa->spa_load_txg = 0; mutex_exit(&spa_namespace_lock); #ifdef __FreeBSD__ #ifdef _KERNEL if (firstopen) zvol_create_minors(spa->spa_name); #endif #endif } *spapp = spa; return (0); } int spa_open_rewind(const char *name, spa_t **spapp, void *tag, nvlist_t *policy, nvlist_t **config) { return (spa_open_common(name, spapp, tag, policy, config)); } int spa_open(const char *name, spa_t **spapp, void *tag) { return (spa_open_common(name, spapp, tag, NULL, NULL)); } /* * Lookup the given spa_t, incrementing the inject count in the process, * preventing it from being exported or destroyed. */ spa_t * spa_inject_addref(char *name) { spa_t *spa; mutex_enter(&spa_namespace_lock); if ((spa = spa_lookup(name)) == NULL) { mutex_exit(&spa_namespace_lock); return (NULL); } spa->spa_inject_ref++; mutex_exit(&spa_namespace_lock); return (spa); } void spa_inject_delref(spa_t *spa) { mutex_enter(&spa_namespace_lock); spa->spa_inject_ref--; mutex_exit(&spa_namespace_lock); } /* * Add spares device information to the nvlist. */ static void spa_add_spares(spa_t *spa, nvlist_t *config) { nvlist_t **spares; uint_t i, nspares; nvlist_t *nvroot; uint64_t guid; vdev_stat_t *vs; uint_t vsc; uint64_t pool; ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER)); if (spa->spa_spares.sav_count == 0) return; VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); VERIFY(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0); if (nspares != 0) { VERIFY(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, spares, nspares) == 0); VERIFY(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0); /* * Go through and find any spares which have since been * repurposed as an active spare. If this is the case, update * their status appropriately. */ for (i = 0; i < nspares; i++) { VERIFY(nvlist_lookup_uint64(spares[i], ZPOOL_CONFIG_GUID, &guid) == 0); if (spa_spare_exists(guid, &pool, NULL) && pool != 0ULL) { VERIFY(nvlist_lookup_uint64_array( spares[i], ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) == 0); vs->vs_state = VDEV_STATE_CANT_OPEN; vs->vs_aux = VDEV_AUX_SPARED; } } } } /* * Add l2cache device information to the nvlist, including vdev stats. */ static void spa_add_l2cache(spa_t *spa, nvlist_t *config) { nvlist_t **l2cache; uint_t i, j, nl2cache; nvlist_t *nvroot; uint64_t guid; vdev_t *vd; vdev_stat_t *vs; uint_t vsc; ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER)); if (spa->spa_l2cache.sav_count == 0) return; VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); VERIFY(nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0); if (nl2cache != 0) { VERIFY(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0); VERIFY(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0); /* * Update level 2 cache device stats. */ for (i = 0; i < nl2cache; i++) { VERIFY(nvlist_lookup_uint64(l2cache[i], ZPOOL_CONFIG_GUID, &guid) == 0); vd = NULL; for (j = 0; j < spa->spa_l2cache.sav_count; j++) { if (guid == spa->spa_l2cache.sav_vdevs[j]->vdev_guid) { vd = spa->spa_l2cache.sav_vdevs[j]; break; } } ASSERT(vd != NULL); VERIFY(nvlist_lookup_uint64_array(l2cache[i], ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) == 0); vdev_get_stats(vd, vs); } } } static void spa_add_feature_stats(spa_t *spa, nvlist_t *config) { nvlist_t *features; zap_cursor_t zc; zap_attribute_t za; ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER)); VERIFY(nvlist_alloc(&features, NV_UNIQUE_NAME, KM_SLEEP) == 0); /* We may be unable to read features if pool is suspended. */ if (spa_suspended(spa)) goto out; if (spa->spa_feat_for_read_obj != 0) { for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_feat_for_read_obj); zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { ASSERT(za.za_integer_length == sizeof (uint64_t) && za.za_num_integers == 1); VERIFY3U(0, ==, nvlist_add_uint64(features, za.za_name, za.za_first_integer)); } zap_cursor_fini(&zc); } if (spa->spa_feat_for_write_obj != 0) { for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_feat_for_write_obj); zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { ASSERT(za.za_integer_length == sizeof (uint64_t) && za.za_num_integers == 1); VERIFY3U(0, ==, nvlist_add_uint64(features, za.za_name, za.za_first_integer)); } zap_cursor_fini(&zc); } out: VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURE_STATS, features) == 0); nvlist_free(features); } int spa_get_stats(const char *name, nvlist_t **config, char *altroot, size_t buflen) { int error; spa_t *spa; *config = NULL; error = spa_open_common(name, &spa, FTAG, NULL, config); if (spa != NULL) { /* * This still leaves a window of inconsistency where the spares * or l2cache devices could change and the config would be * self-inconsistent. */ spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); if (*config != NULL) { uint64_t loadtimes[2]; loadtimes[0] = spa->spa_loaded_ts.tv_sec; loadtimes[1] = spa->spa_loaded_ts.tv_nsec; VERIFY(nvlist_add_uint64_array(*config, ZPOOL_CONFIG_LOADED_TIME, loadtimes, 2) == 0); VERIFY(nvlist_add_uint64(*config, ZPOOL_CONFIG_ERRCOUNT, spa_get_errlog_size(spa)) == 0); if (spa_suspended(spa)) VERIFY(nvlist_add_uint64(*config, ZPOOL_CONFIG_SUSPENDED, spa->spa_failmode) == 0); spa_add_spares(spa, *config); spa_add_l2cache(spa, *config); spa_add_feature_stats(spa, *config); } } /* * We want to get the alternate root even for faulted pools, so we cheat * and call spa_lookup() directly. */ if (altroot) { if (spa == NULL) { mutex_enter(&spa_namespace_lock); spa = spa_lookup(name); if (spa) spa_altroot(spa, altroot, buflen); else altroot[0] = '\0'; spa = NULL; mutex_exit(&spa_namespace_lock); } else { spa_altroot(spa, altroot, buflen); } } if (spa != NULL) { spa_config_exit(spa, SCL_CONFIG, FTAG); spa_close(spa, FTAG); } return (error); } /* * Validate that the auxiliary device array is well formed. We must have an * array of nvlists, each which describes a valid leaf vdev. If this is an * import (mode is VDEV_ALLOC_SPARE), then we allow corrupted spares to be * specified, as long as they are well-formed. */ static int spa_validate_aux_devs(spa_t *spa, nvlist_t *nvroot, uint64_t crtxg, int mode, spa_aux_vdev_t *sav, const char *config, uint64_t version, vdev_labeltype_t label) { nvlist_t **dev; uint_t i, ndev; vdev_t *vd; int error; ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); /* * It's acceptable to have no devs specified. */ if (nvlist_lookup_nvlist_array(nvroot, config, &dev, &ndev) != 0) return (0); if (ndev == 0) return (SET_ERROR(EINVAL)); /* * Make sure the pool is formatted with a version that supports this * device type. */ if (spa_version(spa) < version) return (SET_ERROR(ENOTSUP)); /* * Set the pending device list so we correctly handle device in-use * checking. */ sav->sav_pending = dev; sav->sav_npending = ndev; for (i = 0; i < ndev; i++) { if ((error = spa_config_parse(spa, &vd, dev[i], NULL, 0, mode)) != 0) goto out; if (!vd->vdev_ops->vdev_op_leaf) { vdev_free(vd); error = SET_ERROR(EINVAL); goto out; } /* * The L2ARC currently only supports disk devices in * kernel context. For user-level testing, we allow it. */ #ifdef _KERNEL if ((strcmp(config, ZPOOL_CONFIG_L2CACHE) == 0) && strcmp(vd->vdev_ops->vdev_op_type, VDEV_TYPE_DISK) != 0) { error = SET_ERROR(ENOTBLK); vdev_free(vd); goto out; } #endif vd->vdev_top = vd; if ((error = vdev_open(vd)) == 0 && (error = vdev_label_init(vd, crtxg, label)) == 0) { VERIFY(nvlist_add_uint64(dev[i], ZPOOL_CONFIG_GUID, vd->vdev_guid) == 0); } vdev_free(vd); if (error && (mode != VDEV_ALLOC_SPARE && mode != VDEV_ALLOC_L2CACHE)) goto out; else error = 0; } out: sav->sav_pending = NULL; sav->sav_npending = 0; return (error); } static int spa_validate_aux(spa_t *spa, nvlist_t *nvroot, uint64_t crtxg, int mode) { int error; ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); if ((error = spa_validate_aux_devs(spa, nvroot, crtxg, mode, &spa->spa_spares, ZPOOL_CONFIG_SPARES, SPA_VERSION_SPARES, VDEV_LABEL_SPARE)) != 0) { return (error); } return (spa_validate_aux_devs(spa, nvroot, crtxg, mode, &spa->spa_l2cache, ZPOOL_CONFIG_L2CACHE, SPA_VERSION_L2CACHE, VDEV_LABEL_L2CACHE)); } static void spa_set_aux_vdevs(spa_aux_vdev_t *sav, nvlist_t **devs, int ndevs, const char *config) { int i; if (sav->sav_config != NULL) { nvlist_t **olddevs; uint_t oldndevs; nvlist_t **newdevs; /* * Generate new dev list by concatentating with the * current dev list. */ VERIFY(nvlist_lookup_nvlist_array(sav->sav_config, config, &olddevs, &oldndevs) == 0); newdevs = kmem_alloc(sizeof (void *) * (ndevs + oldndevs), KM_SLEEP); for (i = 0; i < oldndevs; i++) VERIFY(nvlist_dup(olddevs[i], &newdevs[i], KM_SLEEP) == 0); for (i = 0; i < ndevs; i++) VERIFY(nvlist_dup(devs[i], &newdevs[i + oldndevs], KM_SLEEP) == 0); VERIFY(nvlist_remove(sav->sav_config, config, DATA_TYPE_NVLIST_ARRAY) == 0); VERIFY(nvlist_add_nvlist_array(sav->sav_config, config, newdevs, ndevs + oldndevs) == 0); for (i = 0; i < oldndevs + ndevs; i++) nvlist_free(newdevs[i]); kmem_free(newdevs, (oldndevs + ndevs) * sizeof (void *)); } else { /* * Generate a new dev list. */ VERIFY(nvlist_alloc(&sav->sav_config, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_nvlist_array(sav->sav_config, config, devs, ndevs) == 0); } } /* * Stop and drop level 2 ARC devices */ void spa_l2cache_drop(spa_t *spa) { vdev_t *vd; int i; spa_aux_vdev_t *sav = &spa->spa_l2cache; for (i = 0; i < sav->sav_count; i++) { uint64_t pool; vd = sav->sav_vdevs[i]; ASSERT(vd != NULL); if (spa_l2cache_exists(vd->vdev_guid, &pool) && pool != 0ULL && l2arc_vdev_present(vd)) l2arc_remove_vdev(vd); } } /* * Pool Creation */ int spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, nvlist_t *zplprops) { spa_t *spa; char *altroot = NULL; vdev_t *rvd; dsl_pool_t *dp; dmu_tx_t *tx; int error = 0; uint64_t txg = TXG_INITIAL; nvlist_t **spares, **l2cache; uint_t nspares, nl2cache; uint64_t version, obj; boolean_t has_features; /* * If this pool already exists, return failure. */ mutex_enter(&spa_namespace_lock); if (spa_lookup(pool) != NULL) { mutex_exit(&spa_namespace_lock); return (SET_ERROR(EEXIST)); } /* * Allocate a new spa_t structure. */ (void) nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); spa = spa_add(pool, NULL, altroot); spa_activate(spa, spa_mode_global); if (props && (error = spa_prop_validate(spa, props))) { spa_deactivate(spa); spa_remove(spa); mutex_exit(&spa_namespace_lock); return (error); } has_features = B_FALSE; for (nvpair_t *elem = nvlist_next_nvpair(props, NULL); elem != NULL; elem = nvlist_next_nvpair(props, elem)) { if (zpool_prop_feature(nvpair_name(elem))) has_features = B_TRUE; } if (has_features || nvlist_lookup_uint64(props, zpool_prop_to_name(ZPOOL_PROP_VERSION), &version) != 0) { version = SPA_VERSION; } ASSERT(SPA_VERSION_IS_SUPPORTED(version)); spa->spa_first_txg = txg; spa->spa_uberblock.ub_txg = txg - 1; spa->spa_uberblock.ub_version = version; spa->spa_ubsync = spa->spa_uberblock; /* * Create "The Godfather" zio to hold all async IOs */ spa->spa_async_zio_root = kmem_alloc(max_ncpus * sizeof (void *), KM_SLEEP); for (int i = 0; i < max_ncpus; i++) { spa->spa_async_zio_root[i] = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_GODFATHER); } /* * Create the root vdev. */ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, VDEV_ALLOC_ADD); ASSERT(error != 0 || rvd != NULL); ASSERT(error != 0 || spa->spa_root_vdev == rvd); if (error == 0 && !zfs_allocatable_devs(nvroot)) error = SET_ERROR(EINVAL); if (error == 0 && (error = vdev_create(rvd, txg, B_FALSE)) == 0 && (error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) == 0) { for (int c = 0; c < rvd->vdev_children; c++) { vdev_ashift_optimize(rvd->vdev_child[c]); vdev_metaslab_set_size(rvd->vdev_child[c]); vdev_expand(rvd->vdev_child[c], txg); } } spa_config_exit(spa, SCL_ALL, FTAG); if (error != 0) { spa_unload(spa); spa_deactivate(spa); spa_remove(spa); mutex_exit(&spa_namespace_lock); return (error); } /* * Get the list of spares, if specified. */ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) { VERIFY(nvlist_alloc(&spa->spa_spares.sav_config, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, spares, nspares) == 0); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_spares(spa); spa_config_exit(spa, SCL_ALL, FTAG); spa->spa_spares.sav_sync = B_TRUE; } /* * Get the list of level 2 cache devices, if specified. */ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0) { VERIFY(nvlist_alloc(&spa->spa_l2cache.sav_config, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_nvlist_array(spa->spa_l2cache.sav_config, ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_l2cache(spa); spa_config_exit(spa, SCL_ALL, FTAG); spa->spa_l2cache.sav_sync = B_TRUE; } spa->spa_is_initializing = B_TRUE; spa->spa_dsl_pool = dp = dsl_pool_create(spa, zplprops, txg); spa->spa_meta_objset = dp->dp_meta_objset; spa->spa_is_initializing = B_FALSE; /* * Create DDTs (dedup tables). */ ddt_create(spa); spa_update_dspace(spa); tx = dmu_tx_create_assigned(dp, txg); /* * Create the pool config object. */ spa->spa_config_object = dmu_object_alloc(spa->spa_meta_objset, DMU_OT_PACKED_NVLIST, SPA_CONFIG_BLOCKSIZE, DMU_OT_PACKED_NVLIST_SIZE, sizeof (uint64_t), tx); if (zap_add(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CONFIG, sizeof (uint64_t), 1, &spa->spa_config_object, tx) != 0) { cmn_err(CE_PANIC, "failed to add pool config"); } if (spa_version(spa) >= SPA_VERSION_FEATURES) spa_feature_create_zap_objects(spa, tx); if (zap_add(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CREATION_VERSION, sizeof (uint64_t), 1, &version, tx) != 0) { cmn_err(CE_PANIC, "failed to add pool version"); } /* Newly created pools with the right version are always deflated. */ if (version >= SPA_VERSION_RAIDZ_DEFLATE) { spa->spa_deflate = TRUE; if (zap_add(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE, sizeof (uint64_t), 1, &spa->spa_deflate, tx) != 0) { cmn_err(CE_PANIC, "failed to add deflate"); } } /* * Create the deferred-free bpobj. Turn off compression * because sync-to-convergence takes longer if the blocksize * keeps changing. */ obj = bpobj_alloc(spa->spa_meta_objset, 1 << 14, tx); dmu_object_set_compress(spa->spa_meta_objset, obj, ZIO_COMPRESS_OFF, tx); if (zap_add(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPOBJ, sizeof (uint64_t), 1, &obj, tx) != 0) { cmn_err(CE_PANIC, "failed to add bpobj"); } VERIFY3U(0, ==, bpobj_open(&spa->spa_deferred_bpobj, spa->spa_meta_objset, obj)); /* * Create the pool's history object. */ if (version >= SPA_VERSION_ZPOOL_HISTORY) spa_history_create_obj(spa, tx); /* + * Generate some random noise for salted checksums to operate on. + */ + (void) random_get_pseudo_bytes(spa->spa_cksum_salt.zcs_bytes, + sizeof (spa->spa_cksum_salt.zcs_bytes)); + + /* * Set pool properties. */ spa->spa_bootfs = zpool_prop_default_numeric(ZPOOL_PROP_BOOTFS); spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION); spa->spa_failmode = zpool_prop_default_numeric(ZPOOL_PROP_FAILUREMODE); spa->spa_autoexpand = zpool_prop_default_numeric(ZPOOL_PROP_AUTOEXPAND); if (props != NULL) { spa_configfile_set(spa, props, B_FALSE); spa_sync_props(props, tx); } dmu_tx_commit(tx); spa->spa_sync_on = B_TRUE; txg_sync_start(spa->spa_dsl_pool); /* * We explicitly wait for the first transaction to complete so that our * bean counters are appropriately updated. */ txg_wait_synced(spa->spa_dsl_pool, txg); spa_config_sync(spa, B_FALSE, B_TRUE); spa_event_notify(spa, NULL, ESC_ZFS_POOL_CREATE); spa_history_log_version(spa, "create"); /* * Don't count references from objsets that are already closed * and are making their way through the eviction process. */ spa_evicting_os_wait(spa); spa->spa_minref = refcount_count(&spa->spa_refcount); mutex_exit(&spa_namespace_lock); return (0); } #ifdef _KERNEL #if defined(sun) /* * Get the root pool information from the root disk, then import the root pool * during the system boot up time. */ extern int vdev_disk_read_rootlabel(char *, char *, nvlist_t **); static nvlist_t * spa_generate_rootconf(char *devpath, char *devid, uint64_t *guid) { nvlist_t *config; nvlist_t *nvtop, *nvroot; uint64_t pgid; if (vdev_disk_read_rootlabel(devpath, devid, &config) != 0) return (NULL); /* * Add this top-level vdev to the child array. */ VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0); VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pgid) == 0); VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, guid) == 0); /* * Put this pool's top-level vdevs into a root vdev. */ VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) == 0); VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, pgid) == 0); VERIFY(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &nvtop, 1) == 0); /* * Replace the existing vdev_tree with the new root vdev in * this pool's configuration (remove the old, add the new). */ VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0); nvlist_free(nvroot); return (config); } /* * Walk the vdev tree and see if we can find a device with "better" * configuration. A configuration is "better" if the label on that * device has a more recent txg. */ static void spa_alt_rootvdev(vdev_t *vd, vdev_t **avd, uint64_t *txg) { for (int c = 0; c < vd->vdev_children; c++) spa_alt_rootvdev(vd->vdev_child[c], avd, txg); if (vd->vdev_ops->vdev_op_leaf) { nvlist_t *label; uint64_t label_txg; if (vdev_disk_read_rootlabel(vd->vdev_physpath, vd->vdev_devid, &label) != 0) return; VERIFY(nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG, &label_txg) == 0); /* * Do we have a better boot device? */ if (label_txg > *txg) { *txg = label_txg; *avd = vd; } nvlist_free(label); } } /* * Import a root pool. * * For x86. devpath_list will consist of devid and/or physpath name of * the vdev (e.g. "id1,sd@SSEAGATE..." or "/pci@1f,0/ide@d/disk@0,0:a"). * The GRUB "findroot" command will return the vdev we should boot. * * For Sparc, devpath_list consists the physpath name of the booting device * no matter the rootpool is a single device pool or a mirrored pool. * e.g. * "/pci@1f,0/ide@d/disk@0,0:a" */ int spa_import_rootpool(char *devpath, char *devid) { spa_t *spa; vdev_t *rvd, *bvd, *avd = NULL; nvlist_t *config, *nvtop; uint64_t guid, txg; char *pname; int error; /* * Read the label from the boot device and generate a configuration. */ config = spa_generate_rootconf(devpath, devid, &guid); #if defined(_OBP) && defined(_KERNEL) if (config == NULL) { if (strstr(devpath, "/iscsi/ssd") != NULL) { /* iscsi boot */ get_iscsi_bootpath_phy(devpath); config = spa_generate_rootconf(devpath, devid, &guid); } } #endif if (config == NULL) { cmn_err(CE_NOTE, "Cannot read the pool label from '%s'", devpath); return (SET_ERROR(EIO)); } VERIFY(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, &pname) == 0); VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, &txg) == 0); mutex_enter(&spa_namespace_lock); if ((spa = spa_lookup(pname)) != NULL) { /* * Remove the existing root pool from the namespace so that we * can replace it with the correct config we just read in. */ spa_remove(spa); } spa = spa_add(pname, config, NULL); spa->spa_is_root = B_TRUE; spa->spa_import_flags = ZFS_IMPORT_VERBATIM; /* * Build up a vdev tree based on the boot device's label config. */ VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); error = spa_config_parse(spa, &rvd, nvtop, NULL, 0, VDEV_ALLOC_ROOTPOOL); spa_config_exit(spa, SCL_ALL, FTAG); if (error) { mutex_exit(&spa_namespace_lock); nvlist_free(config); cmn_err(CE_NOTE, "Can not parse the config for pool '%s'", pname); return (error); } /* * Get the boot vdev. */ if ((bvd = vdev_lookup_by_guid(rvd, guid)) == NULL) { cmn_err(CE_NOTE, "Can not find the boot vdev for guid %llu", (u_longlong_t)guid); error = SET_ERROR(ENOENT); goto out; } /* * Determine if there is a better boot device. */ avd = bvd; spa_alt_rootvdev(rvd, &avd, &txg); if (avd != bvd) { cmn_err(CE_NOTE, "The boot device is 'degraded'. Please " "try booting from '%s'", avd->vdev_path); error = SET_ERROR(EINVAL); goto out; } /* * If the boot device is part of a spare vdev then ensure that * we're booting off the active spare. */ if (bvd->vdev_parent->vdev_ops == &vdev_spare_ops && !bvd->vdev_isspare) { cmn_err(CE_NOTE, "The boot device is currently spared. Please " "try booting from '%s'", bvd->vdev_parent-> vdev_child[bvd->vdev_parent->vdev_children - 1]->vdev_path); error = SET_ERROR(EINVAL); goto out; } error = 0; out: spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); vdev_free(rvd); spa_config_exit(spa, SCL_ALL, FTAG); mutex_exit(&spa_namespace_lock); nvlist_free(config); return (error); } #else extern int vdev_geom_read_pool_label(const char *name, nvlist_t ***configs, uint64_t *count); static nvlist_t * spa_generate_rootconf(const char *name) { nvlist_t **configs, **tops; nvlist_t *config; nvlist_t *best_cfg, *nvtop, *nvroot; uint64_t *holes; uint64_t best_txg; uint64_t nchildren; uint64_t pgid; uint64_t count; uint64_t i; uint_t nholes; if (vdev_geom_read_pool_label(name, &configs, &count) != 0) return (NULL); ASSERT3U(count, !=, 0); best_txg = 0; for (i = 0; i < count; i++) { uint64_t txg; VERIFY(nvlist_lookup_uint64(configs[i], ZPOOL_CONFIG_POOL_TXG, &txg) == 0); if (txg > best_txg) { best_txg = txg; best_cfg = configs[i]; } } /* * Multi-vdev root pool configuration discovery is not supported yet. */ nchildren = 1; nvlist_lookup_uint64(best_cfg, ZPOOL_CONFIG_VDEV_CHILDREN, &nchildren); holes = NULL; nvlist_lookup_uint64_array(best_cfg, ZPOOL_CONFIG_HOLE_ARRAY, &holes, &nholes); tops = kmem_zalloc(nchildren * sizeof(void *), KM_SLEEP); for (i = 0; i < nchildren; i++) { if (i >= count) break; if (configs[i] == NULL) continue; VERIFY(nvlist_lookup_nvlist(configs[i], ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0); nvlist_dup(nvtop, &tops[i], KM_SLEEP); } for (i = 0; holes != NULL && i < nholes; i++) { if (i >= nchildren) continue; if (tops[holes[i]] != NULL) continue; nvlist_alloc(&tops[holes[i]], NV_UNIQUE_NAME, KM_SLEEP); VERIFY(nvlist_add_string(tops[holes[i]], ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE) == 0); VERIFY(nvlist_add_uint64(tops[holes[i]], ZPOOL_CONFIG_ID, holes[i]) == 0); VERIFY(nvlist_add_uint64(tops[holes[i]], ZPOOL_CONFIG_GUID, 0) == 0); } for (i = 0; i < nchildren; i++) { if (tops[i] != NULL) continue; nvlist_alloc(&tops[i], NV_UNIQUE_NAME, KM_SLEEP); VERIFY(nvlist_add_string(tops[i], ZPOOL_CONFIG_TYPE, VDEV_TYPE_MISSING) == 0); VERIFY(nvlist_add_uint64(tops[i], ZPOOL_CONFIG_ID, i) == 0); VERIFY(nvlist_add_uint64(tops[i], ZPOOL_CONFIG_GUID, 0) == 0); } /* * Create pool config based on the best vdev config. */ nvlist_dup(best_cfg, &config, KM_SLEEP); /* * Put this pool's top-level vdevs into a root vdev. */ VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pgid) == 0); VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) == 0); VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, pgid) == 0); VERIFY(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, tops, nchildren) == 0); /* * Replace the existing vdev_tree with the new root vdev in * this pool's configuration (remove the old, add the new). */ VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0); /* * Drop vdev config elements that should not be present at pool level. */ nvlist_remove(config, ZPOOL_CONFIG_GUID, DATA_TYPE_UINT64); nvlist_remove(config, ZPOOL_CONFIG_TOP_GUID, DATA_TYPE_UINT64); for (i = 0; i < count; i++) nvlist_free(configs[i]); kmem_free(configs, count * sizeof(void *)); for (i = 0; i < nchildren; i++) nvlist_free(tops[i]); kmem_free(tops, nchildren * sizeof(void *)); nvlist_free(nvroot); return (config); } int spa_import_rootpool(const char *name) { spa_t *spa; vdev_t *rvd, *bvd, *avd = NULL; nvlist_t *config, *nvtop; uint64_t txg; char *pname; int error; /* * Read the label from the boot device and generate a configuration. */ config = spa_generate_rootconf(name); mutex_enter(&spa_namespace_lock); if (config != NULL) { VERIFY(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, &pname) == 0 && strcmp(name, pname) == 0); VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, &txg) == 0); if ((spa = spa_lookup(pname)) != NULL) { /* * Remove the existing root pool from the namespace so * that we can replace it with the correct config * we just read in. */ spa_remove(spa); } spa = spa_add(pname, config, NULL); /* * Set spa_ubsync.ub_version as it can be used in vdev_alloc() * via spa_version(). */ if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &spa->spa_ubsync.ub_version) != 0) spa->spa_ubsync.ub_version = SPA_VERSION_INITIAL; } else if ((spa = spa_lookup(name)) == NULL) { mutex_exit(&spa_namespace_lock); nvlist_free(config); cmn_err(CE_NOTE, "Cannot find the pool label for '%s'", name); return (EIO); } else { VERIFY(nvlist_dup(spa->spa_config, &config, KM_SLEEP) == 0); } spa->spa_is_root = B_TRUE; spa->spa_import_flags = ZFS_IMPORT_VERBATIM; /* * Build up a vdev tree based on the boot device's label config. */ VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); error = spa_config_parse(spa, &rvd, nvtop, NULL, 0, VDEV_ALLOC_ROOTPOOL); spa_config_exit(spa, SCL_ALL, FTAG); if (error) { mutex_exit(&spa_namespace_lock); nvlist_free(config); cmn_err(CE_NOTE, "Can not parse the config for pool '%s'", pname); return (error); } spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); vdev_free(rvd); spa_config_exit(spa, SCL_ALL, FTAG); mutex_exit(&spa_namespace_lock); nvlist_free(config); return (0); } #endif /* sun */ #endif /* * Import a non-root pool into the system. */ int spa_import(const char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags) { spa_t *spa; char *altroot = NULL; spa_load_state_t state = SPA_LOAD_IMPORT; zpool_rewind_policy_t policy; uint64_t mode = spa_mode_global; uint64_t readonly = B_FALSE; int error; nvlist_t *nvroot; nvlist_t **spares, **l2cache; uint_t nspares, nl2cache; /* * If a pool with this name exists, return failure. */ mutex_enter(&spa_namespace_lock); if (spa_lookup(pool) != NULL) { mutex_exit(&spa_namespace_lock); return (SET_ERROR(EEXIST)); } /* * Create and initialize the spa structure. */ (void) nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); (void) nvlist_lookup_uint64(props, zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly); if (readonly) mode = FREAD; spa = spa_add(pool, config, altroot); spa->spa_import_flags = flags; /* * Verbatim import - Take a pool and insert it into the namespace * as if it had been loaded at boot. */ if (spa->spa_import_flags & ZFS_IMPORT_VERBATIM) { if (props != NULL) spa_configfile_set(spa, props, B_FALSE); spa_config_sync(spa, B_FALSE, B_TRUE); spa_event_notify(spa, NULL, ESC_ZFS_POOL_IMPORT); mutex_exit(&spa_namespace_lock); return (0); } spa_activate(spa, mode); /* * Don't start async tasks until we know everything is healthy. */ spa_async_suspend(spa); zpool_get_rewind_policy(config, &policy); if (policy.zrp_request & ZPOOL_DO_REWIND) state = SPA_LOAD_RECOVER; /* * Pass off the heavy lifting to spa_load(). Pass TRUE for mosconfig * because the user-supplied config is actually the one to trust when * doing an import. */ if (state != SPA_LOAD_RECOVER) spa->spa_last_ubsync_txg = spa->spa_load_txg = 0; error = spa_load_best(spa, state, B_TRUE, policy.zrp_txg, policy.zrp_request); /* * Propagate anything learned while loading the pool and pass it * back to caller (i.e. rewind info, missing devices, etc). */ VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, spa->spa_load_info) == 0); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); /* * Toss any existing sparelist, as it doesn't have any validity * anymore, and conflicts with spa_has_spare(). */ if (spa->spa_spares.sav_config) { nvlist_free(spa->spa_spares.sav_config); spa->spa_spares.sav_config = NULL; spa_load_spares(spa); } if (spa->spa_l2cache.sav_config) { nvlist_free(spa->spa_l2cache.sav_config); spa->spa_l2cache.sav_config = NULL; spa_load_l2cache(spa); } VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); if (error == 0) error = spa_validate_aux(spa, nvroot, -1ULL, VDEV_ALLOC_SPARE); if (error == 0) error = spa_validate_aux(spa, nvroot, -1ULL, VDEV_ALLOC_L2CACHE); spa_config_exit(spa, SCL_ALL, FTAG); if (props != NULL) spa_configfile_set(spa, props, B_FALSE); if (error != 0 || (props && spa_writeable(spa) && (error = spa_prop_set(spa, props)))) { spa_unload(spa); spa_deactivate(spa); spa_remove(spa); mutex_exit(&spa_namespace_lock); return (error); } spa_async_resume(spa); /* * Override any spares and level 2 cache devices as specified by * the user, as these may have correct device names/devids, etc. */ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) { if (spa->spa_spares.sav_config) VERIFY(nvlist_remove(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, DATA_TYPE_NVLIST_ARRAY) == 0); else VERIFY(nvlist_alloc(&spa->spa_spares.sav_config, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, spares, nspares) == 0); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_spares(spa); spa_config_exit(spa, SCL_ALL, FTAG); spa->spa_spares.sav_sync = B_TRUE; } if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0) { if (spa->spa_l2cache.sav_config) VERIFY(nvlist_remove(spa->spa_l2cache.sav_config, ZPOOL_CONFIG_L2CACHE, DATA_TYPE_NVLIST_ARRAY) == 0); else VERIFY(nvlist_alloc(&spa->spa_l2cache.sav_config, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_nvlist_array(spa->spa_l2cache.sav_config, ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_l2cache(spa); spa_config_exit(spa, SCL_ALL, FTAG); spa->spa_l2cache.sav_sync = B_TRUE; } /* * Check for any removed devices. */ if (spa->spa_autoreplace) { spa_aux_check_removed(&spa->spa_spares); spa_aux_check_removed(&spa->spa_l2cache); } if (spa_writeable(spa)) { /* * Update the config cache to include the newly-imported pool. */ spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); } /* * It's possible that the pool was expanded while it was exported. * We kick off an async task to handle this for us. */ spa_async_request(spa, SPA_ASYNC_AUTOEXPAND); spa_history_log_version(spa, "import"); spa_event_notify(spa, NULL, ESC_ZFS_POOL_IMPORT); mutex_exit(&spa_namespace_lock); #ifdef __FreeBSD__ #ifdef _KERNEL zvol_create_minors(pool); #endif #endif return (0); } nvlist_t * spa_tryimport(nvlist_t *tryconfig) { nvlist_t *config = NULL; char *poolname; spa_t *spa; uint64_t state; int error; if (nvlist_lookup_string(tryconfig, ZPOOL_CONFIG_POOL_NAME, &poolname)) return (NULL); if (nvlist_lookup_uint64(tryconfig, ZPOOL_CONFIG_POOL_STATE, &state)) return (NULL); /* * Create and initialize the spa structure. */ mutex_enter(&spa_namespace_lock); spa = spa_add(TRYIMPORT_NAME, tryconfig, NULL); spa_activate(spa, FREAD); /* * Pass off the heavy lifting to spa_load(). * Pass TRUE for mosconfig because the user-supplied config * is actually the one to trust when doing an import. */ error = spa_load(spa, SPA_LOAD_TRYIMPORT, SPA_IMPORT_EXISTING, B_TRUE); /* * If 'tryconfig' was at least parsable, return the current config. */ if (spa->spa_root_vdev != NULL) { config = spa_config_generate(spa, NULL, -1ULL, B_TRUE); VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, poolname) == 0); VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, state) == 0); VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TIMESTAMP, spa->spa_uberblock.ub_timestamp) == 0); VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, spa->spa_load_info) == 0); /* * If the bootfs property exists on this pool then we * copy it out so that external consumers can tell which * pools are bootable. */ if ((!error || error == EEXIST) && spa->spa_bootfs) { char *tmpname = kmem_alloc(MAXPATHLEN, KM_SLEEP); /* * We have to play games with the name since the * pool was opened as TRYIMPORT_NAME. */ if (dsl_dsobj_to_dsname(spa_name(spa), spa->spa_bootfs, tmpname) == 0) { char *cp; char *dsname = kmem_alloc(MAXPATHLEN, KM_SLEEP); cp = strchr(tmpname, '/'); if (cp == NULL) { (void) strlcpy(dsname, tmpname, MAXPATHLEN); } else { (void) snprintf(dsname, MAXPATHLEN, "%s/%s", poolname, ++cp); } VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_BOOTFS, dsname) == 0); kmem_free(dsname, MAXPATHLEN); } kmem_free(tmpname, MAXPATHLEN); } /* * Add the list of hot spares and level 2 cache devices. */ spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); spa_add_spares(spa, config); spa_add_l2cache(spa, config); spa_config_exit(spa, SCL_CONFIG, FTAG); } spa_unload(spa); spa_deactivate(spa); spa_remove(spa); mutex_exit(&spa_namespace_lock); return (config); } /* * Pool export/destroy * * The act of destroying or exporting a pool is very simple. We make sure there * is no more pending I/O and any references to the pool are gone. Then, we * update the pool state and sync all the labels to disk, removing the * configuration from the cache afterwards. If the 'hardforce' flag is set, then * we don't sync the labels or remove the configuration cache. */ static int spa_export_common(char *pool, int new_state, nvlist_t **oldconfig, boolean_t force, boolean_t hardforce) { spa_t *spa; if (oldconfig) *oldconfig = NULL; if (!(spa_mode_global & FWRITE)) return (SET_ERROR(EROFS)); mutex_enter(&spa_namespace_lock); if ((spa = spa_lookup(pool)) == NULL) { mutex_exit(&spa_namespace_lock); return (SET_ERROR(ENOENT)); } /* * Put a hold on the pool, drop the namespace lock, stop async tasks, * reacquire the namespace lock, and see if we can export. */ spa_open_ref(spa, FTAG); mutex_exit(&spa_namespace_lock); spa_async_suspend(spa); mutex_enter(&spa_namespace_lock); spa_close(spa, FTAG); /* * The pool will be in core if it's openable, * in which case we can modify its state. */ if (spa->spa_state != POOL_STATE_UNINITIALIZED && spa->spa_sync_on) { /* * Objsets may be open only because they're dirty, so we * have to force it to sync before checking spa_refcnt. */ txg_wait_synced(spa->spa_dsl_pool, 0); spa_evicting_os_wait(spa); /* * A pool cannot be exported or destroyed if there are active * references. If we are resetting a pool, allow references by * fault injection handlers. */ if (!spa_refcount_zero(spa) || (spa->spa_inject_ref != 0 && new_state != POOL_STATE_UNINITIALIZED)) { spa_async_resume(spa); mutex_exit(&spa_namespace_lock); return (SET_ERROR(EBUSY)); } /* * A pool cannot be exported if it has an active shared spare. * This is to prevent other pools stealing the active spare * from an exported pool. At user's own will, such pool can * be forcedly exported. */ if (!force && new_state == POOL_STATE_EXPORTED && spa_has_active_shared_spare(spa)) { spa_async_resume(spa); mutex_exit(&spa_namespace_lock); return (SET_ERROR(EXDEV)); } /* * We want this to be reflected on every label, * so mark them all dirty. spa_unload() will do the * final sync that pushes these changes out. */ if (new_state != POOL_STATE_UNINITIALIZED && !hardforce) { spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa->spa_state = new_state; spa->spa_final_txg = spa_last_synced_txg(spa) + TXG_DEFER_SIZE + 1; vdev_config_dirty(spa->spa_root_vdev); spa_config_exit(spa, SCL_ALL, FTAG); } } spa_event_notify(spa, NULL, ESC_ZFS_POOL_DESTROY); if (spa->spa_state != POOL_STATE_UNINITIALIZED) { spa_unload(spa); spa_deactivate(spa); } if (oldconfig && spa->spa_config) VERIFY(nvlist_dup(spa->spa_config, oldconfig, 0) == 0); if (new_state != POOL_STATE_UNINITIALIZED) { if (!hardforce) spa_config_sync(spa, B_TRUE, B_TRUE); spa_remove(spa); } mutex_exit(&spa_namespace_lock); return (0); } /* * Destroy a storage pool. */ int spa_destroy(char *pool) { return (spa_export_common(pool, POOL_STATE_DESTROYED, NULL, B_FALSE, B_FALSE)); } /* * Export a storage pool. */ int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, boolean_t hardforce) { return (spa_export_common(pool, POOL_STATE_EXPORTED, oldconfig, force, hardforce)); } /* * Similar to spa_export(), this unloads the spa_t without actually removing it * from the namespace in any way. */ int spa_reset(char *pool) { return (spa_export_common(pool, POOL_STATE_UNINITIALIZED, NULL, B_FALSE, B_FALSE)); } /* * ========================================================================== * Device manipulation * ========================================================================== */ /* * Add a device to a storage pool. */ int spa_vdev_add(spa_t *spa, nvlist_t *nvroot) { uint64_t txg, id; int error; vdev_t *rvd = spa->spa_root_vdev; vdev_t *vd, *tvd; nvlist_t **spares, **l2cache; uint_t nspares, nl2cache; ASSERT(spa_writeable(spa)); txg = spa_vdev_enter(spa); if ((error = spa_config_parse(spa, &vd, nvroot, NULL, 0, VDEV_ALLOC_ADD)) != 0) return (spa_vdev_exit(spa, NULL, txg, error)); spa->spa_pending_vdev = vd; /* spa_vdev_exit() will clear this */ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares, &nspares) != 0) nspares = 0; if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) != 0) nl2cache = 0; if (vd->vdev_children == 0 && nspares == 0 && nl2cache == 0) return (spa_vdev_exit(spa, vd, txg, EINVAL)); if (vd->vdev_children != 0 && (error = vdev_create(vd, txg, B_FALSE)) != 0) return (spa_vdev_exit(spa, vd, txg, error)); /* * We must validate the spares and l2cache devices after checking the * children. Otherwise, vdev_inuse() will blindly overwrite the spare. */ if ((error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) != 0) return (spa_vdev_exit(spa, vd, txg, error)); /* * Transfer each new top-level vdev from vd to rvd. */ for (int c = 0; c < vd->vdev_children; c++) { /* * Set the vdev id to the first hole, if one exists. */ for (id = 0; id < rvd->vdev_children; id++) { if (rvd->vdev_child[id]->vdev_ishole) { vdev_free(rvd->vdev_child[id]); break; } } tvd = vd->vdev_child[c]; vdev_remove_child(vd, tvd); tvd->vdev_id = id; vdev_add_child(rvd, tvd); vdev_config_dirty(tvd); } if (nspares != 0) { spa_set_aux_vdevs(&spa->spa_spares, spares, nspares, ZPOOL_CONFIG_SPARES); spa_load_spares(spa); spa->spa_spares.sav_sync = B_TRUE; } if (nl2cache != 0) { spa_set_aux_vdevs(&spa->spa_l2cache, l2cache, nl2cache, ZPOOL_CONFIG_L2CACHE); spa_load_l2cache(spa); spa->spa_l2cache.sav_sync = B_TRUE; } /* * We have to be careful when adding new vdevs to an existing pool. * If other threads start allocating from these vdevs before we * sync the config cache, and we lose power, then upon reboot we may * fail to open the pool because there are DVAs that the config cache * can't translate. Therefore, we first add the vdevs without * initializing metaslabs; sync the config cache (via spa_vdev_exit()); * and then let spa_config_update() initialize the new metaslabs. * * spa_load() checks for added-but-not-initialized vdevs, so that * if we lose power at any point in this sequence, the remaining * steps will be completed the next time we load the pool. */ (void) spa_vdev_exit(spa, vd, txg, 0); mutex_enter(&spa_namespace_lock); spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); spa_event_notify(spa, NULL, ESC_ZFS_VDEV_ADD); mutex_exit(&spa_namespace_lock); return (0); } /* * Attach a device to a mirror. The arguments are the path to any device * in the mirror, and the nvroot for the new device. If the path specifies * a device that is not mirrored, we automatically insert the mirror vdev. * * If 'replacing' is specified, the new device is intended to replace the * existing device; in this case the two devices are made into their own * mirror using the 'replacing' vdev, which is functionally identical to * the mirror vdev (it actually reuses all the same ops) but has a few * extra rules: you can't attach to it after it's been created, and upon * completion of resilvering, the first disk (the one being replaced) * is automatically detached. */ int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing) { uint64_t txg, dtl_max_txg; vdev_t *rvd = spa->spa_root_vdev; vdev_t *oldvd, *newvd, *newrootvd, *pvd, *tvd; vdev_ops_t *pvops; char *oldvdpath, *newvdpath; int newvd_isspare; int error; ASSERT(spa_writeable(spa)); txg = spa_vdev_enter(spa); oldvd = spa_lookup_by_guid(spa, guid, B_FALSE); if (oldvd == NULL) return (spa_vdev_exit(spa, NULL, txg, ENODEV)); if (!oldvd->vdev_ops->vdev_op_leaf) return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); pvd = oldvd->vdev_parent; if ((error = spa_config_parse(spa, &newrootvd, nvroot, NULL, 0, VDEV_ALLOC_ATTACH)) != 0) return (spa_vdev_exit(spa, NULL, txg, EINVAL)); if (newrootvd->vdev_children != 1) return (spa_vdev_exit(spa, newrootvd, txg, EINVAL)); newvd = newrootvd->vdev_child[0]; if (!newvd->vdev_ops->vdev_op_leaf) return (spa_vdev_exit(spa, newrootvd, txg, EINVAL)); if ((error = vdev_create(newrootvd, txg, replacing)) != 0) return (spa_vdev_exit(spa, newrootvd, txg, error)); /* * Spares can't replace logs */ if (oldvd->vdev_top->vdev_islog && newvd->vdev_isspare) return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); if (!replacing) { /* * For attach, the only allowable parent is a mirror or the root * vdev. */ if (pvd->vdev_ops != &vdev_mirror_ops && pvd->vdev_ops != &vdev_root_ops) return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); pvops = &vdev_mirror_ops; } else { /* * Active hot spares can only be replaced by inactive hot * spares. */ if (pvd->vdev_ops == &vdev_spare_ops && oldvd->vdev_isspare && !spa_has_spare(spa, newvd->vdev_guid)) return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); /* * If the source is a hot spare, and the parent isn't already a * spare, then we want to create a new hot spare. Otherwise, we * want to create a replacing vdev. The user is not allowed to * attach to a spared vdev child unless the 'isspare' state is * the same (spare replaces spare, non-spare replaces * non-spare). */ if (pvd->vdev_ops == &vdev_replacing_ops && spa_version(spa) < SPA_VERSION_MULTI_REPLACE) { return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); } else if (pvd->vdev_ops == &vdev_spare_ops && newvd->vdev_isspare != oldvd->vdev_isspare) { return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); } if (newvd->vdev_isspare) pvops = &vdev_spare_ops; else pvops = &vdev_replacing_ops; } /* * Make sure the new device is big enough. */ if (newvd->vdev_asize < vdev_get_min_asize(oldvd)) return (spa_vdev_exit(spa, newrootvd, txg, EOVERFLOW)); /* * The new device cannot have a higher alignment requirement * than the top-level vdev. */ if (newvd->vdev_ashift > oldvd->vdev_top->vdev_ashift) return (spa_vdev_exit(spa, newrootvd, txg, EDOM)); /* * If this is an in-place replacement, update oldvd's path and devid * to make it distinguishable from newvd, and unopenable from now on. */ if (strcmp(oldvd->vdev_path, newvd->vdev_path) == 0) { spa_strfree(oldvd->vdev_path); oldvd->vdev_path = kmem_alloc(strlen(newvd->vdev_path) + 5, KM_SLEEP); (void) sprintf(oldvd->vdev_path, "%s/%s", newvd->vdev_path, "old"); if (oldvd->vdev_devid != NULL) { spa_strfree(oldvd->vdev_devid); oldvd->vdev_devid = NULL; } } /* mark the device being resilvered */ newvd->vdev_resilver_txg = txg; /* * If the parent is not a mirror, or if we're replacing, insert the new * mirror/replacing/spare vdev above oldvd. */ if (pvd->vdev_ops != pvops) pvd = vdev_add_parent(oldvd, pvops); ASSERT(pvd->vdev_top->vdev_parent == rvd); ASSERT(pvd->vdev_ops == pvops); ASSERT(oldvd->vdev_parent == pvd); /* * Extract the new device from its root and add it to pvd. */ vdev_remove_child(newrootvd, newvd); newvd->vdev_id = pvd->vdev_children; newvd->vdev_crtxg = oldvd->vdev_crtxg; vdev_add_child(pvd, newvd); tvd = newvd->vdev_top; ASSERT(pvd->vdev_top == tvd); ASSERT(tvd->vdev_parent == rvd); vdev_config_dirty(tvd); /* * Set newvd's DTL to [TXG_INITIAL, dtl_max_txg) so that we account * for any dmu_sync-ed blocks. It will propagate upward when * spa_vdev_exit() calls vdev_dtl_reassess(). */ dtl_max_txg = txg + TXG_CONCURRENT_STATES; vdev_dtl_dirty(newvd, DTL_MISSING, TXG_INITIAL, dtl_max_txg - TXG_INITIAL); if (newvd->vdev_isspare) { spa_spare_activate(newvd); spa_event_notify(spa, newvd, ESC_ZFS_VDEV_SPARE); } oldvdpath = spa_strdup(oldvd->vdev_path); newvdpath = spa_strdup(newvd->vdev_path); newvd_isspare = newvd->vdev_isspare; /* * Mark newvd's DTL dirty in this txg. */ vdev_dirty(tvd, VDD_DTL, newvd, txg); /* * Schedule the resilver to restart in the future. We do this to * ensure that dmu_sync-ed blocks have been stitched into the * respective datasets. */ dsl_resilver_restart(spa->spa_dsl_pool, dtl_max_txg); if (spa->spa_bootfs) spa_event_notify(spa, newvd, ESC_ZFS_BOOTFS_VDEV_ATTACH); spa_event_notify(spa, newvd, ESC_ZFS_VDEV_ATTACH); /* * Commit the config */ (void) spa_vdev_exit(spa, newrootvd, dtl_max_txg, 0); spa_history_log_internal(spa, "vdev attach", NULL, "%s vdev=%s %s vdev=%s", replacing && newvd_isspare ? "spare in" : replacing ? "replace" : "attach", newvdpath, replacing ? "for" : "to", oldvdpath); spa_strfree(oldvdpath); spa_strfree(newvdpath); return (0); } /* * Detach a device from a mirror or replacing vdev. * * If 'replace_done' is specified, only detach if the parent * is a replacing vdev. */ int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done) { uint64_t txg; int error; vdev_t *rvd = spa->spa_root_vdev; vdev_t *vd, *pvd, *cvd, *tvd; boolean_t unspare = B_FALSE; uint64_t unspare_guid = 0; char *vdpath; ASSERT(spa_writeable(spa)); txg = spa_vdev_enter(spa); vd = spa_lookup_by_guid(spa, guid, B_FALSE); if (vd == NULL) return (spa_vdev_exit(spa, NULL, txg, ENODEV)); if (!vd->vdev_ops->vdev_op_leaf) return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); pvd = vd->vdev_parent; /* * If the parent/child relationship is not as expected, don't do it. * Consider M(A,R(B,C)) -- that is, a mirror of A with a replacing * vdev that's replacing B with C. The user's intent in replacing * is to go from M(A,B) to M(A,C). If the user decides to cancel * the replace by detaching C, the expected behavior is to end up * M(A,B). But suppose that right after deciding to detach C, * the replacement of B completes. We would have M(A,C), and then * ask to detach C, which would leave us with just A -- not what * the user wanted. To prevent this, we make sure that the * parent/child relationship hasn't changed -- in this example, * that C's parent is still the replacing vdev R. */ if (pvd->vdev_guid != pguid && pguid != 0) return (spa_vdev_exit(spa, NULL, txg, EBUSY)); /* * Only 'replacing' or 'spare' vdevs can be replaced. */ if (replace_done && pvd->vdev_ops != &vdev_replacing_ops && pvd->vdev_ops != &vdev_spare_ops) return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); ASSERT(pvd->vdev_ops != &vdev_spare_ops || spa_version(spa) >= SPA_VERSION_SPARES); /* * Only mirror, replacing, and spare vdevs support detach. */ if (pvd->vdev_ops != &vdev_replacing_ops && pvd->vdev_ops != &vdev_mirror_ops && pvd->vdev_ops != &vdev_spare_ops) return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); /* * If this device has the only valid copy of some data, * we cannot safely detach it. */ if (vdev_dtl_required(vd)) return (spa_vdev_exit(spa, NULL, txg, EBUSY)); ASSERT(pvd->vdev_children >= 2); /* * If we are detaching the second disk from a replacing vdev, then * check to see if we changed the original vdev's path to have "/old" * at the end in spa_vdev_attach(). If so, undo that change now. */ if (pvd->vdev_ops == &vdev_replacing_ops && vd->vdev_id > 0 && vd->vdev_path != NULL) { size_t len = strlen(vd->vdev_path); for (int c = 0; c < pvd->vdev_children; c++) { cvd = pvd->vdev_child[c]; if (cvd == vd || cvd->vdev_path == NULL) continue; if (strncmp(cvd->vdev_path, vd->vdev_path, len) == 0 && strcmp(cvd->vdev_path + len, "/old") == 0) { spa_strfree(cvd->vdev_path); cvd->vdev_path = spa_strdup(vd->vdev_path); break; } } } /* * If we are detaching the original disk from a spare, then it implies * that the spare should become a real disk, and be removed from the * active spare list for the pool. */ if (pvd->vdev_ops == &vdev_spare_ops && vd->vdev_id == 0 && pvd->vdev_child[pvd->vdev_children - 1]->vdev_isspare) unspare = B_TRUE; /* * Erase the disk labels so the disk can be used for other things. * This must be done after all other error cases are handled, * but before we disembowel vd (so we can still do I/O to it). * But if we can't do it, don't treat the error as fatal -- * it may be that the unwritability of the disk is the reason * it's being detached! */ error = vdev_label_init(vd, 0, VDEV_LABEL_REMOVE); /* * Remove vd from its parent and compact the parent's children. */ vdev_remove_child(pvd, vd); vdev_compact_children(pvd); /* * Remember one of the remaining children so we can get tvd below. */ cvd = pvd->vdev_child[pvd->vdev_children - 1]; /* * If we need to remove the remaining child from the list of hot spares, * do it now, marking the vdev as no longer a spare in the process. * We must do this before vdev_remove_parent(), because that can * change the GUID if it creates a new toplevel GUID. For a similar * reason, we must remove the spare now, in the same txg as the detach; * otherwise someone could attach a new sibling, change the GUID, and * the subsequent attempt to spa_vdev_remove(unspare_guid) would fail. */ if (unspare) { ASSERT(cvd->vdev_isspare); spa_spare_remove(cvd); unspare_guid = cvd->vdev_guid; (void) spa_vdev_remove(spa, unspare_guid, B_TRUE); cvd->vdev_unspare = B_TRUE; } /* * If the parent mirror/replacing vdev only has one child, * the parent is no longer needed. Remove it from the tree. */ if (pvd->vdev_children == 1) { if (pvd->vdev_ops == &vdev_spare_ops) cvd->vdev_unspare = B_FALSE; vdev_remove_parent(cvd); } /* * We don't set tvd until now because the parent we just removed * may have been the previous top-level vdev. */ tvd = cvd->vdev_top; ASSERT(tvd->vdev_parent == rvd); /* * Reevaluate the parent vdev state. */ vdev_propagate_state(cvd); /* * If the 'autoexpand' property is set on the pool then automatically * try to expand the size of the pool. For example if the device we * just detached was smaller than the others, it may be possible to * add metaslabs (i.e. grow the pool). We need to reopen the vdev * first so that we can obtain the updated sizes of the leaf vdevs. */ if (spa->spa_autoexpand) { vdev_reopen(tvd); vdev_expand(tvd, txg); } vdev_config_dirty(tvd); /* * Mark vd's DTL as dirty in this txg. vdev_dtl_sync() will see that * vd->vdev_detached is set and free vd's DTL object in syncing context. * But first make sure we're not on any *other* txg's DTL list, to * prevent vd from being accessed after it's freed. */ vdpath = spa_strdup(vd->vdev_path); for (int t = 0; t < TXG_SIZE; t++) (void) txg_list_remove_this(&tvd->vdev_dtl_list, vd, t); vd->vdev_detached = B_TRUE; vdev_dirty(tvd, VDD_DTL, vd, txg); spa_event_notify(spa, vd, ESC_ZFS_VDEV_REMOVE); /* hang on to the spa before we release the lock */ spa_open_ref(spa, FTAG); error = spa_vdev_exit(spa, vd, txg, 0); spa_history_log_internal(spa, "detach", NULL, "vdev=%s", vdpath); spa_strfree(vdpath); /* * If this was the removal of the original device in a hot spare vdev, * then we want to go through and remove the device from the hot spare * list of every other pool. */ if (unspare) { spa_t *altspa = NULL; mutex_enter(&spa_namespace_lock); while ((altspa = spa_next(altspa)) != NULL) { if (altspa->spa_state != POOL_STATE_ACTIVE || altspa == spa) continue; spa_open_ref(altspa, FTAG); mutex_exit(&spa_namespace_lock); (void) spa_vdev_remove(altspa, unspare_guid, B_TRUE); mutex_enter(&spa_namespace_lock); spa_close(altspa, FTAG); } mutex_exit(&spa_namespace_lock); /* search the rest of the vdevs for spares to remove */ spa_vdev_resilver_done(spa); } /* all done with the spa; OK to release */ mutex_enter(&spa_namespace_lock); spa_close(spa, FTAG); mutex_exit(&spa_namespace_lock); return (error); } /* * Split a set of devices from their mirrors, and create a new pool from them. */ int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, nvlist_t *props, boolean_t exp) { int error = 0; uint64_t txg, *glist; spa_t *newspa; uint_t c, children, lastlog; nvlist_t **child, *nvl, *tmp; dmu_tx_t *tx; char *altroot = NULL; vdev_t *rvd, **vml = NULL; /* vdev modify list */ boolean_t activate_slog; ASSERT(spa_writeable(spa)); txg = spa_vdev_enter(spa); /* clear the log and flush everything up to now */ activate_slog = spa_passivate_log(spa); (void) spa_vdev_config_exit(spa, NULL, txg, 0, FTAG); error = spa_offline_log(spa); txg = spa_vdev_config_enter(spa); if (activate_slog) spa_activate_log(spa); if (error != 0) return (spa_vdev_exit(spa, NULL, txg, error)); /* check new spa name before going any further */ if (spa_lookup(newname) != NULL) return (spa_vdev_exit(spa, NULL, txg, EEXIST)); /* * scan through all the children to ensure they're all mirrors */ if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvl) != 0 || nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) return (spa_vdev_exit(spa, NULL, txg, EINVAL)); /* first, check to ensure we've got the right child count */ rvd = spa->spa_root_vdev; lastlog = 0; for (c = 0; c < rvd->vdev_children; c++) { vdev_t *vd = rvd->vdev_child[c]; /* don't count the holes & logs as children */ if (vd->vdev_islog || vd->vdev_ishole) { if (lastlog == 0) lastlog = c; continue; } lastlog = 0; } if (children != (lastlog != 0 ? lastlog : rvd->vdev_children)) return (spa_vdev_exit(spa, NULL, txg, EINVAL)); /* next, ensure no spare or cache devices are part of the split */ if (nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_SPARES, &tmp) == 0 || nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_L2CACHE, &tmp) == 0) return (spa_vdev_exit(spa, NULL, txg, EINVAL)); vml = kmem_zalloc(children * sizeof (vdev_t *), KM_SLEEP); glist = kmem_zalloc(children * sizeof (uint64_t), KM_SLEEP); /* then, loop over each vdev and validate it */ for (c = 0; c < children; c++) { uint64_t is_hole = 0; (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, &is_hole); if (is_hole != 0) { if (spa->spa_root_vdev->vdev_child[c]->vdev_ishole || spa->spa_root_vdev->vdev_child[c]->vdev_islog) { continue; } else { error = SET_ERROR(EINVAL); break; } } /* which disk is going to be split? */ if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_GUID, &glist[c]) != 0) { error = SET_ERROR(EINVAL); break; } /* look it up in the spa */ vml[c] = spa_lookup_by_guid(spa, glist[c], B_FALSE); if (vml[c] == NULL) { error = SET_ERROR(ENODEV); break; } /* make sure there's nothing stopping the split */ if (vml[c]->vdev_parent->vdev_ops != &vdev_mirror_ops || vml[c]->vdev_islog || vml[c]->vdev_ishole || vml[c]->vdev_isspare || vml[c]->vdev_isl2cache || !vdev_writeable(vml[c]) || vml[c]->vdev_children != 0 || vml[c]->vdev_state != VDEV_STATE_HEALTHY || c != spa->spa_root_vdev->vdev_child[c]->vdev_id) { error = SET_ERROR(EINVAL); break; } if (vdev_dtl_required(vml[c])) { error = SET_ERROR(EBUSY); break; } /* we need certain info from the top level */ VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_ARRAY, vml[c]->vdev_top->vdev_ms_array) == 0); VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_SHIFT, vml[c]->vdev_top->vdev_ms_shift) == 0); VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_ASIZE, vml[c]->vdev_top->vdev_asize) == 0); VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_ASHIFT, vml[c]->vdev_top->vdev_ashift) == 0); } if (error != 0) { kmem_free(vml, children * sizeof (vdev_t *)); kmem_free(glist, children * sizeof (uint64_t)); return (spa_vdev_exit(spa, NULL, txg, error)); } /* stop writers from using the disks */ for (c = 0; c < children; c++) { if (vml[c] != NULL) vml[c]->vdev_offline = B_TRUE; } vdev_reopen(spa->spa_root_vdev); /* * Temporarily record the splitting vdevs in the spa config. This * will disappear once the config is regenerated. */ VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST, glist, children) == 0); kmem_free(glist, children * sizeof (uint64_t)); mutex_enter(&spa->spa_props_lock); VERIFY(nvlist_add_nvlist(spa->spa_config, ZPOOL_CONFIG_SPLIT, nvl) == 0); mutex_exit(&spa->spa_props_lock); spa->spa_config_splitting = nvl; vdev_config_dirty(spa->spa_root_vdev); /* configure and create the new pool */ VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, newname) == 0); VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, exp ? POOL_STATE_EXPORTED : POOL_STATE_ACTIVE) == 0); VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, spa_version(spa)) == 0); VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, spa->spa_config_txg) == 0); VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID, spa_generate_guid(NULL)) == 0); (void) nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); /* add the new pool to the namespace */ newspa = spa_add(newname, config, altroot); newspa->spa_config_txg = spa->spa_config_txg; spa_set_log_state(newspa, SPA_LOG_CLEAR); /* release the spa config lock, retaining the namespace lock */ spa_vdev_config_exit(spa, NULL, txg, 0, FTAG); if (zio_injection_enabled) zio_handle_panic_injection(spa, FTAG, 1); spa_activate(newspa, spa_mode_global); spa_async_suspend(newspa); #ifndef sun /* mark that we are creating new spa by splitting */ newspa->spa_splitting_newspa = B_TRUE; #endif /* create the new pool from the disks of the original pool */ error = spa_load(newspa, SPA_LOAD_IMPORT, SPA_IMPORT_ASSEMBLE, B_TRUE); #ifndef sun newspa->spa_splitting_newspa = B_FALSE; #endif if (error) goto out; /* if that worked, generate a real config for the new pool */ if (newspa->spa_root_vdev != NULL) { VERIFY(nvlist_alloc(&newspa->spa_config_splitting, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_uint64(newspa->spa_config_splitting, ZPOOL_CONFIG_SPLIT_GUID, spa_guid(spa)) == 0); spa_config_set(newspa, spa_config_generate(newspa, NULL, -1ULL, B_TRUE)); } /* set the props */ if (props != NULL) { spa_configfile_set(newspa, props, B_FALSE); error = spa_prop_set(newspa, props); if (error) goto out; } /* flush everything */ txg = spa_vdev_config_enter(newspa); vdev_config_dirty(newspa->spa_root_vdev); (void) spa_vdev_config_exit(newspa, NULL, txg, 0, FTAG); if (zio_injection_enabled) zio_handle_panic_injection(spa, FTAG, 2); spa_async_resume(newspa); /* finally, update the original pool's config */ txg = spa_vdev_config_enter(spa); tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); error = dmu_tx_assign(tx, TXG_WAIT); if (error != 0) dmu_tx_abort(tx); for (c = 0; c < children; c++) { if (vml[c] != NULL) { vdev_split(vml[c]); if (error == 0) spa_history_log_internal(spa, "detach", tx, "vdev=%s", vml[c]->vdev_path); vdev_free(vml[c]); } } vdev_config_dirty(spa->spa_root_vdev); spa->spa_config_splitting = NULL; nvlist_free(nvl); if (error == 0) dmu_tx_commit(tx); (void) spa_vdev_exit(spa, NULL, txg, 0); if (zio_injection_enabled) zio_handle_panic_injection(spa, FTAG, 3); /* split is complete; log a history record */ spa_history_log_internal(newspa, "split", NULL, "from pool %s", spa_name(spa)); kmem_free(vml, children * sizeof (vdev_t *)); /* if we're not going to mount the filesystems in userland, export */ if (exp) error = spa_export_common(newname, POOL_STATE_EXPORTED, NULL, B_FALSE, B_FALSE); return (error); out: spa_unload(newspa); spa_deactivate(newspa); spa_remove(newspa); txg = spa_vdev_config_enter(spa); /* re-online all offlined disks */ for (c = 0; c < children; c++) { if (vml[c] != NULL) vml[c]->vdev_offline = B_FALSE; } vdev_reopen(spa->spa_root_vdev); nvlist_free(spa->spa_config_splitting); spa->spa_config_splitting = NULL; (void) spa_vdev_exit(spa, NULL, txg, error); kmem_free(vml, children * sizeof (vdev_t *)); return (error); } static nvlist_t * spa_nvlist_lookup_by_guid(nvlist_t **nvpp, int count, uint64_t target_guid) { for (int i = 0; i < count; i++) { uint64_t guid; VERIFY(nvlist_lookup_uint64(nvpp[i], ZPOOL_CONFIG_GUID, &guid) == 0); if (guid == target_guid) return (nvpp[i]); } return (NULL); } static void spa_vdev_remove_aux(nvlist_t *config, char *name, nvlist_t **dev, int count, nvlist_t *dev_to_remove) { nvlist_t **newdev = NULL; if (count > 1) newdev = kmem_alloc((count - 1) * sizeof (void *), KM_SLEEP); for (int i = 0, j = 0; i < count; i++) { if (dev[i] == dev_to_remove) continue; VERIFY(nvlist_dup(dev[i], &newdev[j++], KM_SLEEP) == 0); } VERIFY(nvlist_remove(config, name, DATA_TYPE_NVLIST_ARRAY) == 0); VERIFY(nvlist_add_nvlist_array(config, name, newdev, count - 1) == 0); for (int i = 0; i < count - 1; i++) nvlist_free(newdev[i]); if (count > 1) kmem_free(newdev, (count - 1) * sizeof (void *)); } /* * Evacuate the device. */ static int spa_vdev_remove_evacuate(spa_t *spa, vdev_t *vd) { uint64_t txg; int error = 0; ASSERT(MUTEX_HELD(&spa_namespace_lock)); ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); ASSERT(vd == vd->vdev_top); /* * Evacuate the device. We don't hold the config lock as writer * since we need to do I/O but we do keep the * spa_namespace_lock held. Once this completes the device * should no longer have any blocks allocated on it. */ if (vd->vdev_islog) { if (vd->vdev_stat.vs_alloc != 0) error = spa_offline_log(spa); } else { error = SET_ERROR(ENOTSUP); } if (error) return (error); /* * The evacuation succeeded. Remove any remaining MOS metadata * associated with this vdev, and wait for these changes to sync. */ ASSERT0(vd->vdev_stat.vs_alloc); txg = spa_vdev_config_enter(spa); vd->vdev_removing = B_TRUE; vdev_dirty_leaves(vd, VDD_DTL, txg); vdev_config_dirty(vd); spa_vdev_config_exit(spa, NULL, txg, 0, FTAG); return (0); } /* * Complete the removal by cleaning up the namespace. */ static void spa_vdev_remove_from_namespace(spa_t *spa, vdev_t *vd) { vdev_t *rvd = spa->spa_root_vdev; uint64_t id = vd->vdev_id; boolean_t last_vdev = (id == (rvd->vdev_children - 1)); ASSERT(MUTEX_HELD(&spa_namespace_lock)); ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); ASSERT(vd == vd->vdev_top); /* * Only remove any devices which are empty. */ if (vd->vdev_stat.vs_alloc != 0) return; (void) vdev_label_init(vd, 0, VDEV_LABEL_REMOVE); if (list_link_active(&vd->vdev_state_dirty_node)) vdev_state_clean(vd); if (list_link_active(&vd->vdev_config_dirty_node)) vdev_config_clean(vd); vdev_free(vd); if (last_vdev) { vdev_compact_children(rvd); } else { vd = vdev_alloc_common(spa, id, 0, &vdev_hole_ops); vdev_add_child(rvd, vd); } vdev_config_dirty(rvd); /* * Reassess the health of our root vdev. */ vdev_reopen(rvd); } /* * Remove a device from the pool - * * Removing a device from the vdev namespace requires several steps * and can take a significant amount of time. As a result we use * the spa_vdev_config_[enter/exit] functions which allow us to * grab and release the spa_config_lock while still holding the namespace * lock. During each step the configuration is synced out. * * Currently, this supports removing only hot spares, slogs, and level 2 ARC * devices. */ int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare) { vdev_t *vd; metaslab_group_t *mg; nvlist_t **spares, **l2cache, *nv; uint64_t txg = 0; uint_t nspares, nl2cache; int error = 0; boolean_t locked = MUTEX_HELD(&spa_namespace_lock); ASSERT(spa_writeable(spa)); if (!locked) txg = spa_vdev_enter(spa); vd = spa_lookup_by_guid(spa, guid, B_FALSE); if (spa->spa_spares.sav_vdevs != NULL && nvlist_lookup_nvlist_array(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0 && (nv = spa_nvlist_lookup_by_guid(spares, nspares, guid)) != NULL) { /* * Only remove the hot spare if it's not currently in use * in this pool. */ if (vd == NULL || unspare) { spa_vdev_remove_aux(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, spares, nspares, nv); spa_load_spares(spa); spa->spa_spares.sav_sync = B_TRUE; } else { error = SET_ERROR(EBUSY); } } else if (spa->spa_l2cache.sav_vdevs != NULL && nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0 && (nv = spa_nvlist_lookup_by_guid(l2cache, nl2cache, guid)) != NULL) { /* * Cache devices can always be removed. */ spa_vdev_remove_aux(spa->spa_l2cache.sav_config, ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache, nv); spa_load_l2cache(spa); spa->spa_l2cache.sav_sync = B_TRUE; } else if (vd != NULL && vd->vdev_islog) { ASSERT(!locked); ASSERT(vd == vd->vdev_top); mg = vd->vdev_mg; /* * Stop allocating from this vdev. */ metaslab_group_passivate(mg); /* * Wait for the youngest allocations and frees to sync, * and then wait for the deferral of those frees to finish. */ spa_vdev_config_exit(spa, NULL, txg + TXG_CONCURRENT_STATES + TXG_DEFER_SIZE, 0, FTAG); /* * Attempt to evacuate the vdev. */ error = spa_vdev_remove_evacuate(spa, vd); txg = spa_vdev_config_enter(spa); /* * If we couldn't evacuate the vdev, unwind. */ if (error) { metaslab_group_activate(mg); return (spa_vdev_exit(spa, NULL, txg, error)); } /* * Clean up the vdev namespace. */ spa_vdev_remove_from_namespace(spa, vd); } else if (vd != NULL) { /* * Normal vdevs cannot be removed (yet). */ error = SET_ERROR(ENOTSUP); } else { /* * There is no vdev of any kind with the specified guid. */ error = SET_ERROR(ENOENT); } if (!locked) return (spa_vdev_exit(spa, NULL, txg, error)); return (error); } /* * Find any device that's done replacing, or a vdev marked 'unspare' that's * currently spared, so we can detach it. */ static vdev_t * spa_vdev_resilver_done_hunt(vdev_t *vd) { vdev_t *newvd, *oldvd; for (int c = 0; c < vd->vdev_children; c++) { oldvd = spa_vdev_resilver_done_hunt(vd->vdev_child[c]); if (oldvd != NULL) return (oldvd); } /* * Check for a completed replacement. We always consider the first * vdev in the list to be the oldest vdev, and the last one to be * the newest (see spa_vdev_attach() for how that works). In * the case where the newest vdev is faulted, we will not automatically * remove it after a resilver completes. This is OK as it will require * user intervention to determine which disk the admin wishes to keep. */ if (vd->vdev_ops == &vdev_replacing_ops) { ASSERT(vd->vdev_children > 1); newvd = vd->vdev_child[vd->vdev_children - 1]; oldvd = vd->vdev_child[0]; if (vdev_dtl_empty(newvd, DTL_MISSING) && vdev_dtl_empty(newvd, DTL_OUTAGE) && !vdev_dtl_required(oldvd)) return (oldvd); } /* * Check for a completed resilver with the 'unspare' flag set. */ if (vd->vdev_ops == &vdev_spare_ops) { vdev_t *first = vd->vdev_child[0]; vdev_t *last = vd->vdev_child[vd->vdev_children - 1]; if (last->vdev_unspare) { oldvd = first; newvd = last; } else if (first->vdev_unspare) { oldvd = last; newvd = first; } else { oldvd = NULL; } if (oldvd != NULL && vdev_dtl_empty(newvd, DTL_MISSING) && vdev_dtl_empty(newvd, DTL_OUTAGE) && !vdev_dtl_required(oldvd)) return (oldvd); /* * If there are more than two spares attached to a disk, * and those spares are not required, then we want to * attempt to free them up now so that they can be used * by other pools. Once we're back down to a single * disk+spare, we stop removing them. */ if (vd->vdev_children > 2) { newvd = vd->vdev_child[1]; if (newvd->vdev_isspare && last->vdev_isspare && vdev_dtl_empty(last, DTL_MISSING) && vdev_dtl_empty(last, DTL_OUTAGE) && !vdev_dtl_required(newvd)) return (newvd); } } return (NULL); } static void spa_vdev_resilver_done(spa_t *spa) { vdev_t *vd, *pvd, *ppvd; uint64_t guid, sguid, pguid, ppguid; spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); while ((vd = spa_vdev_resilver_done_hunt(spa->spa_root_vdev)) != NULL) { pvd = vd->vdev_parent; ppvd = pvd->vdev_parent; guid = vd->vdev_guid; pguid = pvd->vdev_guid; ppguid = ppvd->vdev_guid; sguid = 0; /* * If we have just finished replacing a hot spared device, then * we need to detach the parent's first child (the original hot * spare) as well. */ if (ppvd->vdev_ops == &vdev_spare_ops && pvd->vdev_id == 0 && ppvd->vdev_children == 2) { ASSERT(pvd->vdev_ops == &vdev_replacing_ops); sguid = ppvd->vdev_child[1]->vdev_guid; } ASSERT(vd->vdev_resilver_txg == 0 || !vdev_dtl_required(vd)); spa_config_exit(spa, SCL_ALL, FTAG); if (spa_vdev_detach(spa, guid, pguid, B_TRUE) != 0) return; if (sguid && spa_vdev_detach(spa, sguid, ppguid, B_TRUE) != 0) return; spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); } spa_config_exit(spa, SCL_ALL, FTAG); } /* * Update the stored path or FRU for this vdev. */ int spa_vdev_set_common(spa_t *spa, uint64_t guid, const char *value, boolean_t ispath) { vdev_t *vd; boolean_t sync = B_FALSE; ASSERT(spa_writeable(spa)); spa_vdev_state_enter(spa, SCL_ALL); if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL) return (spa_vdev_state_exit(spa, NULL, ENOENT)); if (!vd->vdev_ops->vdev_op_leaf) return (spa_vdev_state_exit(spa, NULL, ENOTSUP)); if (ispath) { if (strcmp(value, vd->vdev_path) != 0) { spa_strfree(vd->vdev_path); vd->vdev_path = spa_strdup(value); sync = B_TRUE; } } else { if (vd->vdev_fru == NULL) { vd->vdev_fru = spa_strdup(value); sync = B_TRUE; } else if (strcmp(value, vd->vdev_fru) != 0) { spa_strfree(vd->vdev_fru); vd->vdev_fru = spa_strdup(value); sync = B_TRUE; } } return (spa_vdev_state_exit(spa, sync ? vd : NULL, 0)); } int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath) { return (spa_vdev_set_common(spa, guid, newpath, B_TRUE)); } int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru) { return (spa_vdev_set_common(spa, guid, newfru, B_FALSE)); } /* * ========================================================================== * SPA Scanning * ========================================================================== */ int spa_scan_stop(spa_t *spa) { ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); if (dsl_scan_resilvering(spa->spa_dsl_pool)) return (SET_ERROR(EBUSY)); return (dsl_scan_cancel(spa->spa_dsl_pool)); } int spa_scan(spa_t *spa, pool_scan_func_t func) { ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); if (func >= POOL_SCAN_FUNCS || func == POOL_SCAN_NONE) return (SET_ERROR(ENOTSUP)); /* * If a resilver was requested, but there is no DTL on a * writeable leaf device, we have nothing to do. */ if (func == POOL_SCAN_RESILVER && !vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) { spa_async_request(spa, SPA_ASYNC_RESILVER_DONE); return (0); } return (dsl_scan(spa->spa_dsl_pool, func)); } /* * ========================================================================== * SPA async task processing * ========================================================================== */ static void spa_async_remove(spa_t *spa, vdev_t *vd) { if (vd->vdev_remove_wanted) { vd->vdev_remove_wanted = B_FALSE; vd->vdev_delayed_close = B_FALSE; vdev_set_state(vd, B_FALSE, VDEV_STATE_REMOVED, VDEV_AUX_NONE); /* * We want to clear the stats, but we don't want to do a full * vdev_clear() as that will cause us to throw away * degraded/faulted state as well as attempt to reopen the * device, all of which is a waste. */ vd->vdev_stat.vs_read_errors = 0; vd->vdev_stat.vs_write_errors = 0; vd->vdev_stat.vs_checksum_errors = 0; vdev_state_dirty(vd->vdev_top); } for (int c = 0; c < vd->vdev_children; c++) spa_async_remove(spa, vd->vdev_child[c]); } static void spa_async_probe(spa_t *spa, vdev_t *vd) { if (vd->vdev_probe_wanted) { vd->vdev_probe_wanted = B_FALSE; vdev_reopen(vd); /* vdev_open() does the actual probe */ } for (int c = 0; c < vd->vdev_children; c++) spa_async_probe(spa, vd->vdev_child[c]); } static void spa_async_autoexpand(spa_t *spa, vdev_t *vd) { sysevent_id_t eid; nvlist_t *attr; char *physpath; if (!spa->spa_autoexpand) return; for (int c = 0; c < vd->vdev_children; c++) { vdev_t *cvd = vd->vdev_child[c]; spa_async_autoexpand(spa, cvd); } if (!vd->vdev_ops->vdev_op_leaf || vd->vdev_physpath == NULL) return; physpath = kmem_zalloc(MAXPATHLEN, KM_SLEEP); (void) snprintf(physpath, MAXPATHLEN, "/devices%s", vd->vdev_physpath); VERIFY(nvlist_alloc(&attr, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_string(attr, DEV_PHYS_PATH, physpath) == 0); (void) ddi_log_sysevent(zfs_dip, SUNW_VENDOR, EC_DEV_STATUS, ESC_ZFS_VDEV_AUTOEXPAND, attr, &eid, DDI_SLEEP); nvlist_free(attr); kmem_free(physpath, MAXPATHLEN); } static void spa_async_thread(void *arg) { spa_t *spa = arg; int tasks; ASSERT(spa->spa_sync_on); mutex_enter(&spa->spa_async_lock); tasks = spa->spa_async_tasks; spa->spa_async_tasks &= SPA_ASYNC_REMOVE; mutex_exit(&spa->spa_async_lock); /* * See if the config needs to be updated. */ if (tasks & SPA_ASYNC_CONFIG_UPDATE) { uint64_t old_space, new_space; mutex_enter(&spa_namespace_lock); old_space = metaslab_class_get_space(spa_normal_class(spa)); spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); new_space = metaslab_class_get_space(spa_normal_class(spa)); mutex_exit(&spa_namespace_lock); /* * If the pool grew as a result of the config update, * then log an internal history event. */ if (new_space != old_space) { spa_history_log_internal(spa, "vdev online", NULL, "pool '%s' size: %llu(+%llu)", spa_name(spa), new_space, new_space - old_space); } } if ((tasks & SPA_ASYNC_AUTOEXPAND) && !spa_suspended(spa)) { spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); spa_async_autoexpand(spa, spa->spa_root_vdev); spa_config_exit(spa, SCL_CONFIG, FTAG); } /* * See if any devices need to be probed. */ if (tasks & SPA_ASYNC_PROBE) { spa_vdev_state_enter(spa, SCL_NONE); spa_async_probe(spa, spa->spa_root_vdev); (void) spa_vdev_state_exit(spa, NULL, 0); } /* * If any devices are done replacing, detach them. */ if (tasks & SPA_ASYNC_RESILVER_DONE) spa_vdev_resilver_done(spa); /* * Kick off a resilver. */ if (tasks & SPA_ASYNC_RESILVER) dsl_resilver_restart(spa->spa_dsl_pool, 0); /* * Let the world know that we're done. */ mutex_enter(&spa->spa_async_lock); spa->spa_async_thread = NULL; cv_broadcast(&spa->spa_async_cv); mutex_exit(&spa->spa_async_lock); thread_exit(); } static void spa_async_thread_vd(void *arg) { spa_t *spa = arg; int tasks; ASSERT(spa->spa_sync_on); mutex_enter(&spa->spa_async_lock); tasks = spa->spa_async_tasks; retry: spa->spa_async_tasks &= ~SPA_ASYNC_REMOVE; mutex_exit(&spa->spa_async_lock); /* * See if any devices need to be marked REMOVED. */ if (tasks & SPA_ASYNC_REMOVE) { spa_vdev_state_enter(spa, SCL_NONE); spa_async_remove(spa, spa->spa_root_vdev); for (int i = 0; i < spa->spa_l2cache.sav_count; i++) spa_async_remove(spa, spa->spa_l2cache.sav_vdevs[i]); for (int i = 0; i < spa->spa_spares.sav_count; i++) spa_async_remove(spa, spa->spa_spares.sav_vdevs[i]); (void) spa_vdev_state_exit(spa, NULL, 0); } /* * Let the world know that we're done. */ mutex_enter(&spa->spa_async_lock); tasks = spa->spa_async_tasks; if ((tasks & SPA_ASYNC_REMOVE) != 0) goto retry; spa->spa_async_thread_vd = NULL; cv_broadcast(&spa->spa_async_cv); mutex_exit(&spa->spa_async_lock); thread_exit(); } void spa_async_suspend(spa_t *spa) { mutex_enter(&spa->spa_async_lock); spa->spa_async_suspended++; while (spa->spa_async_thread != NULL && spa->spa_async_thread_vd != NULL) cv_wait(&spa->spa_async_cv, &spa->spa_async_lock); mutex_exit(&spa->spa_async_lock); } void spa_async_resume(spa_t *spa) { mutex_enter(&spa->spa_async_lock); ASSERT(spa->spa_async_suspended != 0); spa->spa_async_suspended--; mutex_exit(&spa->spa_async_lock); } static boolean_t spa_async_tasks_pending(spa_t *spa) { uint_t non_config_tasks; uint_t config_task; boolean_t config_task_suspended; non_config_tasks = spa->spa_async_tasks & ~(SPA_ASYNC_CONFIG_UPDATE | SPA_ASYNC_REMOVE); config_task = spa->spa_async_tasks & SPA_ASYNC_CONFIG_UPDATE; if (spa->spa_ccw_fail_time == 0) { config_task_suspended = B_FALSE; } else { config_task_suspended = (gethrtime() - spa->spa_ccw_fail_time) < (zfs_ccw_retry_interval * NANOSEC); } return (non_config_tasks || (config_task && !config_task_suspended)); } static void spa_async_dispatch(spa_t *spa) { mutex_enter(&spa->spa_async_lock); if (spa_async_tasks_pending(spa) && !spa->spa_async_suspended && spa->spa_async_thread == NULL && rootdir != NULL) spa->spa_async_thread = thread_create(NULL, 0, spa_async_thread, spa, 0, &p0, TS_RUN, maxclsyspri); mutex_exit(&spa->spa_async_lock); } static void spa_async_dispatch_vd(spa_t *spa) { mutex_enter(&spa->spa_async_lock); if ((spa->spa_async_tasks & SPA_ASYNC_REMOVE) != 0 && !spa->spa_async_suspended && spa->spa_async_thread_vd == NULL && rootdir != NULL) spa->spa_async_thread_vd = thread_create(NULL, 0, spa_async_thread_vd, spa, 0, &p0, TS_RUN, maxclsyspri); mutex_exit(&spa->spa_async_lock); } void spa_async_request(spa_t *spa, int task) { zfs_dbgmsg("spa=%s async request task=%u", spa->spa_name, task); mutex_enter(&spa->spa_async_lock); spa->spa_async_tasks |= task; mutex_exit(&spa->spa_async_lock); spa_async_dispatch_vd(spa); } /* * ========================================================================== * SPA syncing routines * ========================================================================== */ static int bpobj_enqueue_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx) { bpobj_t *bpo = arg; bpobj_enqueue(bpo, bp, tx); return (0); } static int spa_free_sync_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx) { zio_t *zio = arg; zio_nowait(zio_free_sync(zio, zio->io_spa, dmu_tx_get_txg(tx), bp, BP_GET_PSIZE(bp), zio->io_flags)); return (0); } /* * Note: this simple function is not inlined to make it easier to dtrace the * amount of time spent syncing frees. */ static void spa_sync_frees(spa_t *spa, bplist_t *bpl, dmu_tx_t *tx) { zio_t *zio = zio_root(spa, NULL, NULL, 0); bplist_iterate(bpl, spa_free_sync_cb, zio, tx); VERIFY(zio_wait(zio) == 0); } /* * Note: this simple function is not inlined to make it easier to dtrace the * amount of time spent syncing deferred frees. */ static void spa_sync_deferred_frees(spa_t *spa, dmu_tx_t *tx) { zio_t *zio = zio_root(spa, NULL, NULL, 0); VERIFY3U(bpobj_iterate(&spa->spa_deferred_bpobj, spa_free_sync_cb, zio, tx), ==, 0); VERIFY0(zio_wait(zio)); } static void spa_sync_nvlist(spa_t *spa, uint64_t obj, nvlist_t *nv, dmu_tx_t *tx) { char *packed = NULL; size_t bufsize; size_t nvsize = 0; dmu_buf_t *db; VERIFY(nvlist_size(nv, &nvsize, NV_ENCODE_XDR) == 0); /* * Write full (SPA_CONFIG_BLOCKSIZE) blocks of configuration * information. This avoids the dmu_buf_will_dirty() path and * saves us a pre-read to get data we don't actually care about. */ bufsize = P2ROUNDUP((uint64_t)nvsize, SPA_CONFIG_BLOCKSIZE); packed = kmem_alloc(bufsize, KM_SLEEP); VERIFY(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR, KM_SLEEP) == 0); bzero(packed + nvsize, bufsize - nvsize); dmu_write(spa->spa_meta_objset, obj, 0, bufsize, packed, tx); kmem_free(packed, bufsize); VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db)); dmu_buf_will_dirty(db, tx); *(uint64_t *)db->db_data = nvsize; dmu_buf_rele(db, FTAG); } static void spa_sync_aux_dev(spa_t *spa, spa_aux_vdev_t *sav, dmu_tx_t *tx, const char *config, const char *entry) { nvlist_t *nvroot; nvlist_t **list; int i; if (!sav->sav_sync) return; /* * Update the MOS nvlist describing the list of available devices. * spa_validate_aux() will have already made sure this nvlist is * valid and the vdevs are labeled appropriately. */ if (sav->sav_object == 0) { sav->sav_object = dmu_object_alloc(spa->spa_meta_objset, DMU_OT_PACKED_NVLIST, 1 << 14, DMU_OT_PACKED_NVLIST_SIZE, sizeof (uint64_t), tx); VERIFY(zap_update(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, entry, sizeof (uint64_t), 1, &sav->sav_object, tx) == 0); } VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_SLEEP) == 0); if (sav->sav_count == 0) { VERIFY(nvlist_add_nvlist_array(nvroot, config, NULL, 0) == 0); } else { list = kmem_alloc(sav->sav_count * sizeof (void *), KM_SLEEP); for (i = 0; i < sav->sav_count; i++) list[i] = vdev_config_generate(spa, sav->sav_vdevs[i], B_FALSE, VDEV_CONFIG_L2CACHE); VERIFY(nvlist_add_nvlist_array(nvroot, config, list, sav->sav_count) == 0); for (i = 0; i < sav->sav_count; i++) nvlist_free(list[i]); kmem_free(list, sav->sav_count * sizeof (void *)); } spa_sync_nvlist(spa, sav->sav_object, nvroot, tx); nvlist_free(nvroot); sav->sav_sync = B_FALSE; } static void spa_sync_config_object(spa_t *spa, dmu_tx_t *tx) { nvlist_t *config; if (list_is_empty(&spa->spa_config_dirty_list)) return; spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); config = spa_config_generate(spa, spa->spa_root_vdev, dmu_tx_get_txg(tx), B_FALSE); /* * If we're upgrading the spa version then make sure that * the config object gets updated with the correct version. */ if (spa->spa_ubsync.ub_version < spa->spa_uberblock.ub_version) fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, spa->spa_uberblock.ub_version); spa_config_exit(spa, SCL_STATE, FTAG); if (spa->spa_config_syncing) nvlist_free(spa->spa_config_syncing); spa->spa_config_syncing = config; spa_sync_nvlist(spa, spa->spa_config_object, config, tx); } static void spa_sync_version(void *arg, dmu_tx_t *tx) { uint64_t *versionp = arg; uint64_t version = *versionp; spa_t *spa = dmu_tx_pool(tx)->dp_spa; /* * Setting the version is special cased when first creating the pool. */ ASSERT(tx->tx_txg != TXG_INITIAL); ASSERT(SPA_VERSION_IS_SUPPORTED(version)); ASSERT(version >= spa_version(spa)); spa->spa_uberblock.ub_version = version; vdev_config_dirty(spa->spa_root_vdev); spa_history_log_internal(spa, "set", tx, "version=%lld", version); } /* * Set zpool properties. */ static void spa_sync_props(void *arg, dmu_tx_t *tx) { nvlist_t *nvp = arg; spa_t *spa = dmu_tx_pool(tx)->dp_spa; objset_t *mos = spa->spa_meta_objset; nvpair_t *elem = NULL; mutex_enter(&spa->spa_props_lock); while ((elem = nvlist_next_nvpair(nvp, elem))) { uint64_t intval; char *strval, *fname; zpool_prop_t prop; const char *propname; zprop_type_t proptype; spa_feature_t fid; switch (prop = zpool_name_to_prop(nvpair_name(elem))) { case ZPROP_INVAL: /* * We checked this earlier in spa_prop_validate(). */ ASSERT(zpool_prop_feature(nvpair_name(elem))); fname = strchr(nvpair_name(elem), '@') + 1; VERIFY0(zfeature_lookup_name(fname, &fid)); spa_feature_enable(spa, fid, tx); spa_history_log_internal(spa, "set", tx, "%s=enabled", nvpair_name(elem)); break; case ZPOOL_PROP_VERSION: intval = fnvpair_value_uint64(elem); /* * The version is synced seperatly before other * properties and should be correct by now. */ ASSERT3U(spa_version(spa), >=, intval); break; case ZPOOL_PROP_ALTROOT: /* * 'altroot' is a non-persistent property. It should * have been set temporarily at creation or import time. */ ASSERT(spa->spa_root != NULL); break; case ZPOOL_PROP_READONLY: case ZPOOL_PROP_CACHEFILE: /* * 'readonly' and 'cachefile' are also non-persisitent * properties. */ break; case ZPOOL_PROP_COMMENT: strval = fnvpair_value_string(elem); if (spa->spa_comment != NULL) spa_strfree(spa->spa_comment); spa->spa_comment = spa_strdup(strval); /* * We need to dirty the configuration on all the vdevs * so that their labels get updated. It's unnecessary * to do this for pool creation since the vdev's * configuratoin has already been dirtied. */ if (tx->tx_txg != TXG_INITIAL) vdev_config_dirty(spa->spa_root_vdev); spa_history_log_internal(spa, "set", tx, "%s=%s", nvpair_name(elem), strval); break; default: /* * Set pool property values in the poolprops mos object. */ if (spa->spa_pool_props_object == 0) { spa->spa_pool_props_object = zap_create_link(mos, DMU_OT_POOL_PROPS, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_PROPS, tx); } /* normalize the property name */ propname = zpool_prop_to_name(prop); proptype = zpool_prop_get_type(prop); if (nvpair_type(elem) == DATA_TYPE_STRING) { ASSERT(proptype == PROP_TYPE_STRING); strval = fnvpair_value_string(elem); VERIFY0(zap_update(mos, spa->spa_pool_props_object, propname, 1, strlen(strval) + 1, strval, tx)); spa_history_log_internal(spa, "set", tx, "%s=%s", nvpair_name(elem), strval); } else if (nvpair_type(elem) == DATA_TYPE_UINT64) { intval = fnvpair_value_uint64(elem); if (proptype == PROP_TYPE_INDEX) { const char *unused; VERIFY0(zpool_prop_index_to_string( prop, intval, &unused)); } VERIFY0(zap_update(mos, spa->spa_pool_props_object, propname, 8, 1, &intval, tx)); spa_history_log_internal(spa, "set", tx, "%s=%lld", nvpair_name(elem), intval); } else { ASSERT(0); /* not allowed */ } switch (prop) { case ZPOOL_PROP_DELEGATION: spa->spa_delegation = intval; break; case ZPOOL_PROP_BOOTFS: spa->spa_bootfs = intval; break; case ZPOOL_PROP_FAILUREMODE: spa->spa_failmode = intval; break; case ZPOOL_PROP_AUTOEXPAND: spa->spa_autoexpand = intval; if (tx->tx_txg != TXG_INITIAL) spa_async_request(spa, SPA_ASYNC_AUTOEXPAND); break; case ZPOOL_PROP_DEDUPDITTO: spa->spa_dedup_ditto = intval; break; default: break; } } } mutex_exit(&spa->spa_props_lock); } /* * Perform one-time upgrade on-disk changes. spa_version() does not * reflect the new version this txg, so there must be no changes this * txg to anything that the upgrade code depends on after it executes. * Therefore this must be called after dsl_pool_sync() does the sync * tasks. */ static void spa_sync_upgrades(spa_t *spa, dmu_tx_t *tx) { dsl_pool_t *dp = spa->spa_dsl_pool; ASSERT(spa->spa_sync_pass == 1); rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG); if (spa->spa_ubsync.ub_version < SPA_VERSION_ORIGIN && spa->spa_uberblock.ub_version >= SPA_VERSION_ORIGIN) { dsl_pool_create_origin(dp, tx); /* Keeping the origin open increases spa_minref */ spa->spa_minref += 3; } if (spa->spa_ubsync.ub_version < SPA_VERSION_NEXT_CLONES && spa->spa_uberblock.ub_version >= SPA_VERSION_NEXT_CLONES) { dsl_pool_upgrade_clones(dp, tx); } if (spa->spa_ubsync.ub_version < SPA_VERSION_DIR_CLONES && spa->spa_uberblock.ub_version >= SPA_VERSION_DIR_CLONES) { dsl_pool_upgrade_dir_clones(dp, tx); /* Keeping the freedir open increases spa_minref */ spa->spa_minref += 3; } if (spa->spa_ubsync.ub_version < SPA_VERSION_FEATURES && spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) { spa_feature_create_zap_objects(spa, tx); } /* * LZ4_COMPRESS feature's behaviour was changed to activate_on_enable * when possibility to use lz4 compression for metadata was added * Old pools that have this feature enabled must be upgraded to have * this feature active */ if (spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) { boolean_t lz4_en = spa_feature_is_enabled(spa, SPA_FEATURE_LZ4_COMPRESS); boolean_t lz4_ac = spa_feature_is_active(spa, SPA_FEATURE_LZ4_COMPRESS); if (lz4_en && !lz4_ac) spa_feature_incr(spa, SPA_FEATURE_LZ4_COMPRESS, tx); } + + /* + * If we haven't written the salt, do so now. Note that the + * feature may not be activated yet, but that's fine since + * the presence of this ZAP entry is backwards compatible. + */ + if (zap_contains(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, + DMU_POOL_CHECKSUM_SALT) == ENOENT) { + VERIFY0(zap_add(spa->spa_meta_objset, + DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CHECKSUM_SALT, 1, + sizeof (spa->spa_cksum_salt.zcs_bytes), + spa->spa_cksum_salt.zcs_bytes, tx)); + } + rrw_exit(&dp->dp_config_rwlock, FTAG); } /* * Sync the specified transaction group. New blocks may be dirtied as * part of the process, so we iterate until it converges. */ void spa_sync(spa_t *spa, uint64_t txg) { dsl_pool_t *dp = spa->spa_dsl_pool; objset_t *mos = spa->spa_meta_objset; bplist_t *free_bpl = &spa->spa_free_bplist[txg & TXG_MASK]; vdev_t *rvd = spa->spa_root_vdev; vdev_t *vd; dmu_tx_t *tx; int error; VERIFY(spa_writeable(spa)); /* * Lock out configuration changes. */ spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); spa->spa_syncing_txg = txg; spa->spa_sync_pass = 0; /* * If there are any pending vdev state changes, convert them * into config changes that go out with this transaction group. */ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); while (list_head(&spa->spa_state_dirty_list) != NULL) { /* * We need the write lock here because, for aux vdevs, * calling vdev_config_dirty() modifies sav_config. * This is ugly and will become unnecessary when we * eliminate the aux vdev wart by integrating all vdevs * into the root vdev tree. */ spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_WRITER); while ((vd = list_head(&spa->spa_state_dirty_list)) != NULL) { vdev_state_clean(vd); vdev_config_dirty(vd); } spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER); } spa_config_exit(spa, SCL_STATE, FTAG); tx = dmu_tx_create_assigned(dp, txg); spa->spa_sync_starttime = gethrtime(); #ifdef illumos VERIFY(cyclic_reprogram(spa->spa_deadman_cycid, spa->spa_sync_starttime + spa->spa_deadman_synctime)); #else /* FreeBSD */ #ifdef _KERNEL callout_reset(&spa->spa_deadman_cycid, hz * spa->spa_deadman_synctime / NANOSEC, spa_deadman, spa); #endif #endif /* * If we are upgrading to SPA_VERSION_RAIDZ_DEFLATE this txg, * set spa_deflate if we have no raid-z vdevs. */ if (spa->spa_ubsync.ub_version < SPA_VERSION_RAIDZ_DEFLATE && spa->spa_uberblock.ub_version >= SPA_VERSION_RAIDZ_DEFLATE) { int i; for (i = 0; i < rvd->vdev_children; i++) { vd = rvd->vdev_child[i]; if (vd->vdev_deflate_ratio != SPA_MINBLOCKSIZE) break; } if (i == rvd->vdev_children) { spa->spa_deflate = TRUE; VERIFY(0 == zap_add(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE, sizeof (uint64_t), 1, &spa->spa_deflate, tx)); } } /* * Iterate to convergence. */ do { int pass = ++spa->spa_sync_pass; spa_sync_config_object(spa, tx); spa_sync_aux_dev(spa, &spa->spa_spares, tx, ZPOOL_CONFIG_SPARES, DMU_POOL_SPARES); spa_sync_aux_dev(spa, &spa->spa_l2cache, tx, ZPOOL_CONFIG_L2CACHE, DMU_POOL_L2CACHE); spa_errlog_sync(spa, txg); dsl_pool_sync(dp, txg); if (pass < zfs_sync_pass_deferred_free) { spa_sync_frees(spa, free_bpl, tx); } else { /* * We can not defer frees in pass 1, because * we sync the deferred frees later in pass 1. */ ASSERT3U(pass, >, 1); bplist_iterate(free_bpl, bpobj_enqueue_cb, &spa->spa_deferred_bpobj, tx); } ddt_sync(spa, txg); dsl_scan_sync(dp, tx); while (vd = txg_list_remove(&spa->spa_vdev_txg_list, txg)) vdev_sync(vd, txg); if (pass == 1) { spa_sync_upgrades(spa, tx); ASSERT3U(txg, >=, spa->spa_uberblock.ub_rootbp.blk_birth); /* * Note: We need to check if the MOS is dirty * because we could have marked the MOS dirty * without updating the uberblock (e.g. if we * have sync tasks but no dirty user data). We * need to check the uberblock's rootbp because * it is updated if we have synced out dirty * data (though in this case the MOS will most * likely also be dirty due to second order * effects, we don't want to rely on that here). */ if (spa->spa_uberblock.ub_rootbp.blk_birth < txg && !dmu_objset_is_dirty(mos, txg)) { /* * Nothing changed on the first pass, * therefore this TXG is a no-op. Avoid * syncing deferred frees, so that we * can keep this TXG as a no-op. */ ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg)); ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg)); ASSERT(txg_list_empty(&dp->dp_sync_tasks, txg)); break; } spa_sync_deferred_frees(spa, tx); } } while (dmu_objset_is_dirty(mos, txg)); /* * Rewrite the vdev configuration (which includes the uberblock) * to commit the transaction group. * * If there are no dirty vdevs, we sync the uberblock to a few * random top-level vdevs that are known to be visible in the * config cache (see spa_vdev_add() for a complete description). * If there *are* dirty vdevs, sync the uberblock to all vdevs. */ for (;;) { /* * We hold SCL_STATE to prevent vdev open/close/etc. * while we're attempting to write the vdev labels. */ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); if (list_is_empty(&spa->spa_config_dirty_list)) { vdev_t *svd[SPA_DVAS_PER_BP]; int svdcount = 0; int children = rvd->vdev_children; int c0 = spa_get_random(children); for (int c = 0; c < children; c++) { vd = rvd->vdev_child[(c0 + c) % children]; if (vd->vdev_ms_array == 0 || vd->vdev_islog) continue; svd[svdcount++] = vd; if (svdcount == SPA_DVAS_PER_BP) break; } error = vdev_config_sync(svd, svdcount, txg, B_FALSE); if (error != 0) error = vdev_config_sync(svd, svdcount, txg, B_TRUE); } else { error = vdev_config_sync(rvd->vdev_child, rvd->vdev_children, txg, B_FALSE); if (error != 0) error = vdev_config_sync(rvd->vdev_child, rvd->vdev_children, txg, B_TRUE); } if (error == 0) spa->spa_last_synced_guid = rvd->vdev_guid; spa_config_exit(spa, SCL_STATE, FTAG); if (error == 0) break; zio_suspend(spa, NULL); zio_resume_wait(spa); } dmu_tx_commit(tx); #ifdef illumos VERIFY(cyclic_reprogram(spa->spa_deadman_cycid, CY_INFINITY)); #else /* FreeBSD */ #ifdef _KERNEL callout_drain(&spa->spa_deadman_cycid); #endif #endif /* * Clear the dirty config list. */ while ((vd = list_head(&spa->spa_config_dirty_list)) != NULL) vdev_config_clean(vd); /* * Now that the new config has synced transactionally, * let it become visible to the config cache. */ if (spa->spa_config_syncing != NULL) { spa_config_set(spa, spa->spa_config_syncing); spa->spa_config_txg = txg; spa->spa_config_syncing = NULL; } spa->spa_ubsync = spa->spa_uberblock; dsl_pool_sync_done(dp, txg); /* * Update usable space statistics. */ while (vd = txg_list_remove(&spa->spa_vdev_txg_list, TXG_CLEAN(txg))) vdev_sync_done(vd, txg); spa_update_dspace(spa); /* * It had better be the case that we didn't dirty anything * since vdev_config_sync(). */ ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg)); ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg)); ASSERT(txg_list_empty(&spa->spa_vdev_txg_list, txg)); spa->spa_sync_pass = 0; spa_config_exit(spa, SCL_CONFIG, FTAG); spa_handle_ignored_writes(spa); /* * If any async tasks have been requested, kick them off. */ spa_async_dispatch(spa); spa_async_dispatch_vd(spa); } /* * Sync all pools. We don't want to hold the namespace lock across these * operations, so we take a reference on the spa_t and drop the lock during the * sync. */ void spa_sync_allpools(void) { spa_t *spa = NULL; mutex_enter(&spa_namespace_lock); while ((spa = spa_next(spa)) != NULL) { if (spa_state(spa) != POOL_STATE_ACTIVE || !spa_writeable(spa) || spa_suspended(spa)) continue; spa_open_ref(spa, FTAG); mutex_exit(&spa_namespace_lock); txg_wait_synced(spa_get_dsl(spa), 0); mutex_enter(&spa_namespace_lock); spa_close(spa, FTAG); } mutex_exit(&spa_namespace_lock); } /* * ========================================================================== * Miscellaneous routines * ========================================================================== */ /* * Remove all pools in the system. */ void spa_evict_all(void) { spa_t *spa; /* * Remove all cached state. All pools should be closed now, * so every spa in the AVL tree should be unreferenced. */ mutex_enter(&spa_namespace_lock); while ((spa = spa_next(NULL)) != NULL) { /* * Stop async tasks. The async thread may need to detach * a device that's been replaced, which requires grabbing * spa_namespace_lock, so we must drop it here. */ spa_open_ref(spa, FTAG); mutex_exit(&spa_namespace_lock); spa_async_suspend(spa); mutex_enter(&spa_namespace_lock); spa_close(spa, FTAG); if (spa->spa_state != POOL_STATE_UNINITIALIZED) { spa_unload(spa); spa_deactivate(spa); } spa_remove(spa); } mutex_exit(&spa_namespace_lock); } vdev_t * spa_lookup_by_guid(spa_t *spa, uint64_t guid, boolean_t aux) { vdev_t *vd; int i; if ((vd = vdev_lookup_by_guid(spa->spa_root_vdev, guid)) != NULL) return (vd); if (aux) { for (i = 0; i < spa->spa_l2cache.sav_count; i++) { vd = spa->spa_l2cache.sav_vdevs[i]; if (vd->vdev_guid == guid) return (vd); } for (i = 0; i < spa->spa_spares.sav_count; i++) { vd = spa->spa_spares.sav_vdevs[i]; if (vd->vdev_guid == guid) return (vd); } } return (NULL); } void spa_upgrade(spa_t *spa, uint64_t version) { ASSERT(spa_writeable(spa)); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); /* * This should only be called for a non-faulted pool, and since a * future version would result in an unopenable pool, this shouldn't be * possible. */ ASSERT(SPA_VERSION_IS_SUPPORTED(spa->spa_uberblock.ub_version)); ASSERT3U(version, >=, spa->spa_uberblock.ub_version); spa->spa_uberblock.ub_version = version; vdev_config_dirty(spa->spa_root_vdev); spa_config_exit(spa, SCL_ALL, FTAG); txg_wait_synced(spa_get_dsl(spa), 0); } boolean_t spa_has_spare(spa_t *spa, uint64_t guid) { int i; uint64_t spareguid; spa_aux_vdev_t *sav = &spa->spa_spares; for (i = 0; i < sav->sav_count; i++) if (sav->sav_vdevs[i]->vdev_guid == guid) return (B_TRUE); for (i = 0; i < sav->sav_npending; i++) { if (nvlist_lookup_uint64(sav->sav_pending[i], ZPOOL_CONFIG_GUID, &spareguid) == 0 && spareguid == guid) return (B_TRUE); } return (B_FALSE); } /* * Check if a pool has an active shared spare device. * Note: reference count of an active spare is 2, as a spare and as a replace */ static boolean_t spa_has_active_shared_spare(spa_t *spa) { int i, refcnt; uint64_t pool; spa_aux_vdev_t *sav = &spa->spa_spares; for (i = 0; i < sav->sav_count; i++) { if (spa_spare_exists(sav->sav_vdevs[i]->vdev_guid, &pool, &refcnt) && pool != 0ULL && pool == spa_guid(spa) && refcnt > 2) return (B_TRUE); } return (B_FALSE); } /* * Post a sysevent corresponding to the given event. The 'name' must be one of * the event definitions in sys/sysevent/eventdefs.h. The payload will be * filled in from the spa and (optionally) the vdev. This doesn't do anything * in the userland libzpool, as we don't want consumers to misinterpret ztest * or zdb as real changes. */ void spa_event_notify(spa_t *spa, vdev_t *vd, const char *name) { #ifdef _KERNEL sysevent_t *ev; sysevent_attr_list_t *attr = NULL; sysevent_value_t value; sysevent_id_t eid; ev = sysevent_alloc(EC_ZFS, (char *)name, SUNW_KERN_PUB "zfs", SE_SLEEP); value.value_type = SE_DATA_TYPE_STRING; value.value.sv_string = spa_name(spa); if (sysevent_add_attr(&attr, ZFS_EV_POOL_NAME, &value, SE_SLEEP) != 0) goto done; value.value_type = SE_DATA_TYPE_UINT64; value.value.sv_uint64 = spa_guid(spa); if (sysevent_add_attr(&attr, ZFS_EV_POOL_GUID, &value, SE_SLEEP) != 0) goto done; if (vd) { value.value_type = SE_DATA_TYPE_UINT64; value.value.sv_uint64 = vd->vdev_guid; if (sysevent_add_attr(&attr, ZFS_EV_VDEV_GUID, &value, SE_SLEEP) != 0) goto done; if (vd->vdev_path) { value.value_type = SE_DATA_TYPE_STRING; value.value.sv_string = vd->vdev_path; if (sysevent_add_attr(&attr, ZFS_EV_VDEV_PATH, &value, SE_SLEEP) != 0) goto done; } } if (sysevent_attach_attributes(ev, attr) != 0) goto done; attr = NULL; (void) log_sysevent(ev, SE_SLEEP, &eid); done: if (attr) sysevent_free_attr(attr); sysevent_free(ev); #endif } Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa_misc.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa_misc.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa_misc.c (revision 290757) @@ -1,2096 +1,2101 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2015 by Delphix. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright 2013 Martin Matuska . All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. + * Copyright 2013 Saso Kiselkov. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "zfs_prop.h" -#include "zfeature_common.h" +#include /* * SPA locking * * There are four basic locks for managing spa_t structures: * * spa_namespace_lock (global mutex) * * This lock must be acquired to do any of the following: * * - Lookup a spa_t by name * - Add or remove a spa_t from the namespace * - Increase spa_refcount from non-zero * - Check if spa_refcount is zero * - Rename a spa_t * - add/remove/attach/detach devices * - Held for the duration of create/destroy/import/export * * It does not need to handle recursion. A create or destroy may * reference objects (files or zvols) in other pools, but by * definition they must have an existing reference, and will never need * to lookup a spa_t by name. * * spa_refcount (per-spa refcount_t protected by mutex) * * This reference count keep track of any active users of the spa_t. The * spa_t cannot be destroyed or freed while this is non-zero. Internally, * the refcount is never really 'zero' - opening a pool implicitly keeps * some references in the DMU. Internally we check against spa_minref, but * present the image of a zero/non-zero value to consumers. * * spa_config_lock[] (per-spa array of rwlocks) * * This protects the spa_t from config changes, and must be held in * the following circumstances: * * - RW_READER to perform I/O to the spa * - RW_WRITER to change the vdev config * * The locking order is fairly straightforward: * * spa_namespace_lock -> spa_refcount * * The namespace lock must be acquired to increase the refcount from 0 * or to check if it is zero. * * spa_refcount -> spa_config_lock[] * * There must be at least one valid reference on the spa_t to acquire * the config lock. * * spa_namespace_lock -> spa_config_lock[] * * The namespace lock must always be taken before the config lock. * * * The spa_namespace_lock can be acquired directly and is globally visible. * * The namespace is manipulated using the following functions, all of which * require the spa_namespace_lock to be held. * * spa_lookup() Lookup a spa_t by name. * * spa_add() Create a new spa_t in the namespace. * * spa_remove() Remove a spa_t from the namespace. This also * frees up any memory associated with the spa_t. * * spa_next() Returns the next spa_t in the system, or the * first if NULL is passed. * * spa_evict_all() Shutdown and remove all spa_t structures in * the system. * * spa_guid_exists() Determine whether a pool/device guid exists. * * The spa_refcount is manipulated using the following functions: * * spa_open_ref() Adds a reference to the given spa_t. Must be * called with spa_namespace_lock held if the * refcount is currently zero. * * spa_close() Remove a reference from the spa_t. This will * not free the spa_t or remove it from the * namespace. No locking is required. * * spa_refcount_zero() Returns true if the refcount is currently * zero. Must be called with spa_namespace_lock * held. * * The spa_config_lock[] is an array of rwlocks, ordered as follows: * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). * * To read the configuration, it suffices to hold one of these locks as reader. * To modify the configuration, you must hold all locks as writer. To modify * vdev state without altering the vdev tree's topology (e.g. online/offline), * you must hold SCL_STATE and SCL_ZIO as writer. * * We use these distinct config locks to avoid recursive lock entry. * For example, spa_sync() (which holds SCL_CONFIG as reader) induces * block allocations (SCL_ALLOC), which may require reading space maps * from disk (dmu_read() -> zio_read() -> SCL_ZIO). * * The spa config locks cannot be normal rwlocks because we need the * ability to hand off ownership. For example, SCL_ZIO is acquired * by the issuing thread and later released by an interrupt thread. * They do, however, obey the usual write-wanted semantics to prevent * writer (i.e. system administrator) starvation. * * The lock acquisition rules are as follows: * * SCL_CONFIG * Protects changes to the vdev tree topology, such as vdev * add/remove/attach/detach. Protects the dirty config list * (spa_config_dirty_list) and the set of spares and l2arc devices. * * SCL_STATE * Protects changes to pool state and vdev state, such as vdev * online/offline/fault/degrade/clear. Protects the dirty state list * (spa_state_dirty_list) and global pool state (spa_state). * * SCL_ALLOC * Protects changes to metaslab groups and classes. * Held as reader by metaslab_alloc() and metaslab_claim(). * * SCL_ZIO * Held by bp-level zios (those which have no io_vd upon entry) * to prevent changes to the vdev tree. The bp-level zio implicitly * protects all of its vdev child zios, which do not hold SCL_ZIO. * * SCL_FREE * Protects changes to metaslab groups and classes. * Held as reader by metaslab_free(). SCL_FREE is distinct from * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free * blocks in zio_done() while another i/o that holds either * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. * * SCL_VDEV * Held as reader to prevent changes to the vdev tree during trivial * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the * other locks, and lower than all of them, to ensure that it's safe * to acquire regardless of caller context. * * In addition, the following rules apply: * * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. * The lock ordering is SCL_CONFIG > spa_props_lock. * * (b) I/O operations on leaf vdevs. For any zio operation that takes * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), * or zio_write_phys() -- the caller must ensure that the config cannot * cannot change in the interim, and that the vdev cannot be reopened. * SCL_STATE as reader suffices for both. * * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). * * spa_vdev_enter() Acquire the namespace lock and the config lock * for writing. * * spa_vdev_exit() Release the config lock, wait for all I/O * to complete, sync the updated configs to the * cache, and release the namespace lock. * * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual * locking is, always, based on spa_namespace_lock and spa_config_lock[]. * * spa_rename() is also implemented within this file since it requires * manipulation of the namespace. */ static avl_tree_t spa_namespace_avl; kmutex_t spa_namespace_lock; static kcondvar_t spa_namespace_cv; static int spa_active_count; int spa_max_replication_override = SPA_DVAS_PER_BP; static kmutex_t spa_spare_lock; static avl_tree_t spa_spare_avl; static kmutex_t spa_l2cache_lock; static avl_tree_t spa_l2cache_avl; kmem_cache_t *spa_buffer_pool; int spa_mode_global; #ifdef ZFS_DEBUG /* Everything except dprintf and spa is on by default in debug builds */ int zfs_flags = ~(ZFS_DEBUG_DPRINTF | ZFS_DEBUG_SPA); #else int zfs_flags = 0; #endif SYSCTL_DECL(_debug); TUNABLE_INT("debug.zfs_flags", &zfs_flags); SYSCTL_INT(_debug, OID_AUTO, zfs_flags, CTLFLAG_RWTUN, &zfs_flags, 0, "ZFS debug flags."); /* * zfs_recover can be set to nonzero to attempt to recover from * otherwise-fatal errors, typically caused by on-disk corruption. When * set, calls to zfs_panic_recover() will turn into warning messages. * This should only be used as a last resort, as it typically results * in leaked space, or worse. */ boolean_t zfs_recover = B_FALSE; SYSCTL_DECL(_vfs_zfs); TUNABLE_INT("vfs.zfs.recover", &zfs_recover); SYSCTL_INT(_vfs_zfs, OID_AUTO, recover, CTLFLAG_RWTUN, &zfs_recover, 0, "Try to recover from otherwise-fatal errors."); /* * If destroy encounters an EIO while reading metadata (e.g. indirect * blocks), space referenced by the missing metadata can not be freed. * Normally this causes the background destroy to become "stalled", as * it is unable to make forward progress. While in this stalled state, * all remaining space to free from the error-encountering filesystem is * "temporarily leaked". Set this flag to cause it to ignore the EIO, * permanently leak the space from indirect blocks that can not be read, * and continue to free everything else that it can. * * The default, "stalling" behavior is useful if the storage partially * fails (i.e. some but not all i/os fail), and then later recovers. In * this case, we will be able to continue pool operations while it is * partially failed, and when it recovers, we can continue to free the * space, with no leaks. However, note that this case is actually * fairly rare. * * Typically pools either (a) fail completely (but perhaps temporarily, * e.g. a top-level vdev going offline), or (b) have localized, * permanent errors (e.g. disk returns the wrong data due to bit flip or * firmware bug). In case (a), this setting does not matter because the * pool will be suspended and the sync thread will not be able to make * forward progress regardless. In case (b), because the error is * permanent, the best we can do is leak the minimum amount of space, * which is what setting this flag will do. Therefore, it is reasonable * for this flag to normally be set, but we chose the more conservative * approach of not setting it, so that there is no possibility of * leaking space in the "partial temporary" failure case. */ boolean_t zfs_free_leak_on_eio = B_FALSE; /* * Expiration time in milliseconds. This value has two meanings. First it is * used to determine when the spa_deadman() logic should fire. By default the * spa_deadman() will fire if spa_sync() has not completed in 1000 seconds. * Secondly, the value determines if an I/O is considered "hung". Any I/O that * has not completed in zfs_deadman_synctime_ms is considered "hung" resulting * in a system panic. */ uint64_t zfs_deadman_synctime_ms = 1000000ULL; TUNABLE_QUAD("vfs.zfs.deadman_synctime_ms", &zfs_deadman_synctime_ms); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, deadman_synctime_ms, CTLFLAG_RDTUN, &zfs_deadman_synctime_ms, 0, "Stalled ZFS I/O expiration time in milliseconds"); /* * Check time in milliseconds. This defines the frequency at which we check * for hung I/O. */ uint64_t zfs_deadman_checktime_ms = 5000ULL; TUNABLE_QUAD("vfs.zfs.deadman_checktime_ms", &zfs_deadman_checktime_ms); SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, deadman_checktime_ms, CTLFLAG_RDTUN, &zfs_deadman_checktime_ms, 0, "Period of checks for stalled ZFS I/O in milliseconds"); /* * Default value of -1 for zfs_deadman_enabled is resolved in * zfs_deadman_init() */ int zfs_deadman_enabled = -1; TUNABLE_INT("vfs.zfs.deadman_enabled", &zfs_deadman_enabled); SYSCTL_INT(_vfs_zfs, OID_AUTO, deadman_enabled, CTLFLAG_RDTUN, &zfs_deadman_enabled, 0, "Kernel panic on stalled ZFS I/O"); /* * The worst case is single-sector max-parity RAID-Z blocks, in which * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) * times the size; so just assume that. Add to this the fact that * we can have up to 3 DVAs per bp, and one more factor of 2 because * the block may be dittoed with up to 3 DVAs by ddt_sync(). All together, * the worst case is: * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24 */ int spa_asize_inflation = 24; TUNABLE_INT("vfs.zfs.spa_asize_inflation", &spa_asize_inflation); SYSCTL_INT(_vfs_zfs, OID_AUTO, spa_asize_inflation, CTLFLAG_RWTUN, &spa_asize_inflation, 0, "Worst case inflation factor for single sector writes"); #ifndef illumos #ifdef _KERNEL static void zfs_deadman_init() { /* * If we are not i386 or amd64 or in a virtual machine, * disable ZFS deadman thread by default */ if (zfs_deadman_enabled == -1) { #if defined(__amd64__) || defined(__i386__) zfs_deadman_enabled = (vm_guest == VM_GUEST_NO) ? 1 : 0; #else zfs_deadman_enabled = 0; #endif } } #endif /* _KERNEL */ #endif /* !illumos */ /* * Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in * the pool to be consumed. This ensures that we don't run the pool * completely out of space, due to unaccounted changes (e.g. to the MOS). * It also limits the worst-case time to allocate space. If we have * less than this amount of free space, most ZPL operations (e.g. write, * create) will return ENOSPC. * * Certain operations (e.g. file removal, most administrative actions) can * use half the slop space. They will only return ENOSPC if less than half * the slop space is free. Typically, once the pool has less than the slop * space free, the user will use these operations to free up space in the pool. * These are the operations that call dsl_pool_adjustedsize() with the netfree * argument set to TRUE. * * A very restricted set of operations are always permitted, regardless of * the amount of free space. These are the operations that call * dsl_sync_task(ZFS_SPACE_CHECK_NONE), e.g. "zfs destroy". If these * operations result in a net increase in the amount of space used, * it is possible to run the pool completely out of space, causing it to * be permanently read-only. * * See also the comments in zfs_space_check_t. */ int spa_slop_shift = 5; SYSCTL_INT(_vfs_zfs, OID_AUTO, spa_slop_shift, CTLFLAG_RWTUN, &spa_slop_shift, 0, "Shift value of reserved space (1/(2^spa_slop_shift))."); /* * ========================================================================== * SPA config locking * ========================================================================== */ static void spa_config_lock_init(spa_t *spa) { for (int i = 0; i < SCL_LOCKS; i++) { spa_config_lock_t *scl = &spa->spa_config_lock[i]; mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); refcount_create_untracked(&scl->scl_count); scl->scl_writer = NULL; scl->scl_write_wanted = 0; } } static void spa_config_lock_destroy(spa_t *spa) { for (int i = 0; i < SCL_LOCKS; i++) { spa_config_lock_t *scl = &spa->spa_config_lock[i]; mutex_destroy(&scl->scl_lock); cv_destroy(&scl->scl_cv); refcount_destroy(&scl->scl_count); ASSERT(scl->scl_writer == NULL); ASSERT(scl->scl_write_wanted == 0); } } int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw) { for (int i = 0; i < SCL_LOCKS; i++) { spa_config_lock_t *scl = &spa->spa_config_lock[i]; if (!(locks & (1 << i))) continue; mutex_enter(&scl->scl_lock); if (rw == RW_READER) { if (scl->scl_writer || scl->scl_write_wanted) { mutex_exit(&scl->scl_lock); spa_config_exit(spa, locks ^ (1 << i), tag); return (0); } } else { ASSERT(scl->scl_writer != curthread); if (!refcount_is_zero(&scl->scl_count)) { mutex_exit(&scl->scl_lock); spa_config_exit(spa, locks ^ (1 << i), tag); return (0); } scl->scl_writer = curthread; } (void) refcount_add(&scl->scl_count, tag); mutex_exit(&scl->scl_lock); } return (1); } void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw) { int wlocks_held = 0; ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY); for (int i = 0; i < SCL_LOCKS; i++) { spa_config_lock_t *scl = &spa->spa_config_lock[i]; if (scl->scl_writer == curthread) wlocks_held |= (1 << i); if (!(locks & (1 << i))) continue; mutex_enter(&scl->scl_lock); if (rw == RW_READER) { while (scl->scl_writer || scl->scl_write_wanted) { cv_wait(&scl->scl_cv, &scl->scl_lock); } } else { ASSERT(scl->scl_writer != curthread); while (!refcount_is_zero(&scl->scl_count)) { scl->scl_write_wanted++; cv_wait(&scl->scl_cv, &scl->scl_lock); scl->scl_write_wanted--; } scl->scl_writer = curthread; } (void) refcount_add(&scl->scl_count, tag); mutex_exit(&scl->scl_lock); } ASSERT(wlocks_held <= locks); } void spa_config_exit(spa_t *spa, int locks, void *tag) { for (int i = SCL_LOCKS - 1; i >= 0; i--) { spa_config_lock_t *scl = &spa->spa_config_lock[i]; if (!(locks & (1 << i))) continue; mutex_enter(&scl->scl_lock); ASSERT(!refcount_is_zero(&scl->scl_count)); if (refcount_remove(&scl->scl_count, tag) == 0) { ASSERT(scl->scl_writer == NULL || scl->scl_writer == curthread); scl->scl_writer = NULL; /* OK in either case */ cv_broadcast(&scl->scl_cv); } mutex_exit(&scl->scl_lock); } } int spa_config_held(spa_t *spa, int locks, krw_t rw) { int locks_held = 0; for (int i = 0; i < SCL_LOCKS; i++) { spa_config_lock_t *scl = &spa->spa_config_lock[i]; if (!(locks & (1 << i))) continue; if ((rw == RW_READER && !refcount_is_zero(&scl->scl_count)) || (rw == RW_WRITER && scl->scl_writer == curthread)) locks_held |= 1 << i; } return (locks_held); } /* * ========================================================================== * SPA namespace functions * ========================================================================== */ /* * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. * Returns NULL if no matching spa_t is found. */ spa_t * spa_lookup(const char *name) { static spa_t search; /* spa_t is large; don't allocate on stack */ spa_t *spa; avl_index_t where; char *cp; ASSERT(MUTEX_HELD(&spa_namespace_lock)); (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); /* * If it's a full dataset name, figure out the pool name and * just use that. */ cp = strpbrk(search.spa_name, "/@#"); if (cp != NULL) *cp = '\0'; spa = avl_find(&spa_namespace_avl, &search, &where); return (spa); } /* * Fires when spa_sync has not completed within zfs_deadman_synctime_ms. * If the zfs_deadman_enabled flag is set then it inspects all vdev queues * looking for potentially hung I/Os. */ void spa_deadman(void *arg) { spa_t *spa = arg; /* * Disable the deadman timer if the pool is suspended. */ if (spa_suspended(spa)) { #ifdef illumos VERIFY(cyclic_reprogram(spa->spa_deadman_cycid, CY_INFINITY)); #else /* Nothing. just don't schedule any future callouts. */ #endif return; } zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu", (gethrtime() - spa->spa_sync_starttime) / NANOSEC, ++spa->spa_deadman_calls); if (zfs_deadman_enabled) vdev_deadman(spa->spa_root_vdev); } /* * Create an uninitialized spa_t with the given name. Requires * spa_namespace_lock. The caller must ensure that the spa_t doesn't already * exist by calling spa_lookup() first. */ spa_t * spa_add(const char *name, nvlist_t *config, const char *altroot) { spa_t *spa; spa_config_dirent_t *dp; #ifdef illumos cyc_handler_t hdlr; cyc_time_t when; #endif ASSERT(MUTEX_HELD(&spa_namespace_lock)); spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_evicting_os_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); + mutex_init(&spa->spa_cksum_tmpls_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL); cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); for (int t = 0; t < TXG_SIZE; t++) bplist_create(&spa->spa_free_bplist[t]); (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); spa->spa_state = POOL_STATE_UNINITIALIZED; spa->spa_freeze_txg = UINT64_MAX; spa->spa_final_txg = UINT64_MAX; spa->spa_load_max_txg = UINT64_MAX; spa->spa_proc = &p0; spa->spa_proc_state = SPA_PROC_NONE; #ifdef illumos hdlr.cyh_func = spa_deadman; hdlr.cyh_arg = spa; hdlr.cyh_level = CY_LOW_LEVEL; #endif spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms); #ifdef illumos /* * This determines how often we need to check for hung I/Os after * the cyclic has already fired. Since checking for hung I/Os is * an expensive operation we don't want to check too frequently. * Instead wait for 5 seconds before checking again. */ when.cyt_interval = MSEC2NSEC(zfs_deadman_checktime_ms); when.cyt_when = CY_INFINITY; mutex_enter(&cpu_lock); spa->spa_deadman_cycid = cyclic_add(&hdlr, &when); mutex_exit(&cpu_lock); #else /* !illumos */ #ifdef _KERNEL callout_init(&spa->spa_deadman_cycid, CALLOUT_MPSAFE); #endif #endif refcount_create(&spa->spa_refcount); spa_config_lock_init(spa); avl_add(&spa_namespace_avl, spa); /* * Set the alternate root, if there is one. */ if (altroot) { spa->spa_root = spa_strdup(altroot); spa_active_count++; } /* * Every pool starts with the default cachefile */ list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), offsetof(spa_config_dirent_t, scd_link)); dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); list_insert_head(&spa->spa_config_list, dp); VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, KM_SLEEP) == 0); if (config != NULL) { nvlist_t *features; if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, &features) == 0) { VERIFY(nvlist_dup(features, &spa->spa_label_features, 0) == 0); } VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); } if (spa->spa_label_features == NULL) { VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, KM_SLEEP) == 0); } spa->spa_debug = ((zfs_flags & ZFS_DEBUG_SPA) != 0); spa->spa_min_ashift = INT_MAX; spa->spa_max_ashift = 0; /* * As a pool is being created, treat all features as disabled by * setting SPA_FEATURE_DISABLED for all entries in the feature * refcount cache. */ for (int i = 0; i < SPA_FEATURES; i++) { spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED; } return (spa); } /* * Removes a spa_t from the namespace, freeing up any memory used. Requires * spa_namespace_lock. This is called only after the spa_t has been closed and * deactivated. */ void spa_remove(spa_t *spa) { spa_config_dirent_t *dp; ASSERT(MUTEX_HELD(&spa_namespace_lock)); ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); ASSERT3U(refcount_count(&spa->spa_refcount), ==, 0); nvlist_free(spa->spa_config_splitting); avl_remove(&spa_namespace_avl, spa); cv_broadcast(&spa_namespace_cv); if (spa->spa_root) { spa_strfree(spa->spa_root); spa_active_count--; } while ((dp = list_head(&spa->spa_config_list)) != NULL) { list_remove(&spa->spa_config_list, dp); if (dp->scd_path != NULL) spa_strfree(dp->scd_path); kmem_free(dp, sizeof (spa_config_dirent_t)); } list_destroy(&spa->spa_config_list); nvlist_free(spa->spa_label_features); nvlist_free(spa->spa_load_info); spa_config_set(spa, NULL); #ifdef illumos mutex_enter(&cpu_lock); if (spa->spa_deadman_cycid != CYCLIC_NONE) cyclic_remove(spa->spa_deadman_cycid); mutex_exit(&cpu_lock); spa->spa_deadman_cycid = CYCLIC_NONE; #else /* !illumos */ #ifdef _KERNEL callout_drain(&spa->spa_deadman_cycid); #endif #endif refcount_destroy(&spa->spa_refcount); spa_config_lock_destroy(spa); for (int t = 0; t < TXG_SIZE; t++) bplist_destroy(&spa->spa_free_bplist[t]); + zio_checksum_templates_free(spa); + cv_destroy(&spa->spa_async_cv); cv_destroy(&spa->spa_evicting_os_cv); cv_destroy(&spa->spa_proc_cv); cv_destroy(&spa->spa_scrub_io_cv); cv_destroy(&spa->spa_suspend_cv); mutex_destroy(&spa->spa_async_lock); mutex_destroy(&spa->spa_errlist_lock); mutex_destroy(&spa->spa_errlog_lock); mutex_destroy(&spa->spa_evicting_os_lock); mutex_destroy(&spa->spa_history_lock); mutex_destroy(&spa->spa_proc_lock); mutex_destroy(&spa->spa_props_lock); + mutex_destroy(&spa->spa_cksum_tmpls_lock); mutex_destroy(&spa->spa_scrub_lock); mutex_destroy(&spa->spa_suspend_lock); mutex_destroy(&spa->spa_vdev_top_lock); kmem_free(spa, sizeof (spa_t)); } /* * Given a pool, return the next pool in the namespace, or NULL if there is * none. If 'prev' is NULL, return the first pool. */ spa_t * spa_next(spa_t *prev) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); if (prev) return (AVL_NEXT(&spa_namespace_avl, prev)); else return (avl_first(&spa_namespace_avl)); } /* * ========================================================================== * SPA refcount functions * ========================================================================== */ /* * Add a reference to the given spa_t. Must have at least one reference, or * have the namespace lock held. */ void spa_open_ref(spa_t *spa, void *tag) { ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref || MUTEX_HELD(&spa_namespace_lock)); (void) refcount_add(&spa->spa_refcount, tag); } /* * Remove a reference to the given spa_t. Must have at least one reference, or * have the namespace lock held. */ void spa_close(spa_t *spa, void *tag) { ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref || MUTEX_HELD(&spa_namespace_lock)); (void) refcount_remove(&spa->spa_refcount, tag); } /* * Remove a reference to the given spa_t held by a dsl dir that is * being asynchronously released. Async releases occur from a taskq * performing eviction of dsl datasets and dirs. The namespace lock * isn't held and the hold by the object being evicted may contribute to * spa_minref (e.g. dataset or directory released during pool export), * so the asserts in spa_close() do not apply. */ void spa_async_close(spa_t *spa, void *tag) { (void) refcount_remove(&spa->spa_refcount, tag); } /* * Check to see if the spa refcount is zero. Must be called with * spa_namespace_lock held. We really compare against spa_minref, which is the * number of references acquired when opening a pool */ boolean_t spa_refcount_zero(spa_t *spa) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); return (refcount_count(&spa->spa_refcount) == spa->spa_minref); } /* * ========================================================================== * SPA spare and l2cache tracking * ========================================================================== */ /* * Hot spares and cache devices are tracked using the same code below, * for 'auxiliary' devices. */ typedef struct spa_aux { uint64_t aux_guid; uint64_t aux_pool; avl_node_t aux_avl; int aux_count; } spa_aux_t; static int spa_aux_compare(const void *a, const void *b) { const spa_aux_t *sa = a; const spa_aux_t *sb = b; if (sa->aux_guid < sb->aux_guid) return (-1); else if (sa->aux_guid > sb->aux_guid) return (1); else return (0); } void spa_aux_add(vdev_t *vd, avl_tree_t *avl) { avl_index_t where; spa_aux_t search; spa_aux_t *aux; search.aux_guid = vd->vdev_guid; if ((aux = avl_find(avl, &search, &where)) != NULL) { aux->aux_count++; } else { aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); aux->aux_guid = vd->vdev_guid; aux->aux_count = 1; avl_insert(avl, aux, where); } } void spa_aux_remove(vdev_t *vd, avl_tree_t *avl) { spa_aux_t search; spa_aux_t *aux; avl_index_t where; search.aux_guid = vd->vdev_guid; aux = avl_find(avl, &search, &where); ASSERT(aux != NULL); if (--aux->aux_count == 0) { avl_remove(avl, aux); kmem_free(aux, sizeof (spa_aux_t)); } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { aux->aux_pool = 0ULL; } } boolean_t spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) { spa_aux_t search, *found; search.aux_guid = guid; found = avl_find(avl, &search, NULL); if (pool) { if (found) *pool = found->aux_pool; else *pool = 0ULL; } if (refcnt) { if (found) *refcnt = found->aux_count; else *refcnt = 0; } return (found != NULL); } void spa_aux_activate(vdev_t *vd, avl_tree_t *avl) { spa_aux_t search, *found; avl_index_t where; search.aux_guid = vd->vdev_guid; found = avl_find(avl, &search, &where); ASSERT(found != NULL); ASSERT(found->aux_pool == 0ULL); found->aux_pool = spa_guid(vd->vdev_spa); } /* * Spares are tracked globally due to the following constraints: * * - A spare may be part of multiple pools. * - A spare may be added to a pool even if it's actively in use within * another pool. * - A spare in use in any pool can only be the source of a replacement if * the target is a spare in the same pool. * * We keep track of all spares on the system through the use of a reference * counted AVL tree. When a vdev is added as a spare, or used as a replacement * spare, then we bump the reference count in the AVL tree. In addition, we set * the 'vdev_isspare' member to indicate that the device is a spare (active or * inactive). When a spare is made active (used to replace a device in the * pool), we also keep track of which pool its been made a part of. * * The 'spa_spare_lock' protects the AVL tree. These functions are normally * called under the spa_namespace lock as part of vdev reconfiguration. The * separate spare lock exists for the status query path, which does not need to * be completely consistent with respect to other vdev configuration changes. */ static int spa_spare_compare(const void *a, const void *b) { return (spa_aux_compare(a, b)); } void spa_spare_add(vdev_t *vd) { mutex_enter(&spa_spare_lock); ASSERT(!vd->vdev_isspare); spa_aux_add(vd, &spa_spare_avl); vd->vdev_isspare = B_TRUE; mutex_exit(&spa_spare_lock); } void spa_spare_remove(vdev_t *vd) { mutex_enter(&spa_spare_lock); ASSERT(vd->vdev_isspare); spa_aux_remove(vd, &spa_spare_avl); vd->vdev_isspare = B_FALSE; mutex_exit(&spa_spare_lock); } boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) { boolean_t found; mutex_enter(&spa_spare_lock); found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); mutex_exit(&spa_spare_lock); return (found); } void spa_spare_activate(vdev_t *vd) { mutex_enter(&spa_spare_lock); ASSERT(vd->vdev_isspare); spa_aux_activate(vd, &spa_spare_avl); mutex_exit(&spa_spare_lock); } /* * Level 2 ARC devices are tracked globally for the same reasons as spares. * Cache devices currently only support one pool per cache device, and so * for these devices the aux reference count is currently unused beyond 1. */ static int spa_l2cache_compare(const void *a, const void *b) { return (spa_aux_compare(a, b)); } void spa_l2cache_add(vdev_t *vd) { mutex_enter(&spa_l2cache_lock); ASSERT(!vd->vdev_isl2cache); spa_aux_add(vd, &spa_l2cache_avl); vd->vdev_isl2cache = B_TRUE; mutex_exit(&spa_l2cache_lock); } void spa_l2cache_remove(vdev_t *vd) { mutex_enter(&spa_l2cache_lock); ASSERT(vd->vdev_isl2cache); spa_aux_remove(vd, &spa_l2cache_avl); vd->vdev_isl2cache = B_FALSE; mutex_exit(&spa_l2cache_lock); } boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool) { boolean_t found; mutex_enter(&spa_l2cache_lock); found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); mutex_exit(&spa_l2cache_lock); return (found); } void spa_l2cache_activate(vdev_t *vd) { mutex_enter(&spa_l2cache_lock); ASSERT(vd->vdev_isl2cache); spa_aux_activate(vd, &spa_l2cache_avl); mutex_exit(&spa_l2cache_lock); } /* * ========================================================================== * SPA vdev locking * ========================================================================== */ /* * Lock the given spa_t for the purpose of adding or removing a vdev. * Grabs the global spa_namespace_lock plus the spa config lock for writing. * It returns the next transaction group for the spa_t. */ uint64_t spa_vdev_enter(spa_t *spa) { mutex_enter(&spa->spa_vdev_top_lock); mutex_enter(&spa_namespace_lock); return (spa_vdev_config_enter(spa)); } /* * Internal implementation for spa_vdev_enter(). Used when a vdev * operation requires multiple syncs (i.e. removing a device) while * keeping the spa_namespace_lock held. */ uint64_t spa_vdev_config_enter(spa_t *spa) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); return (spa_last_synced_txg(spa) + 1); } /* * Used in combination with spa_vdev_config_enter() to allow the syncing * of multiple transactions without releasing the spa_namespace_lock. */ void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); int config_changed = B_FALSE; ASSERT(txg > spa_last_synced_txg(spa)); spa->spa_pending_vdev = NULL; /* * Reassess the DTLs. */ vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { config_changed = B_TRUE; spa->spa_config_generation++; } /* * Verify the metaslab classes. */ ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); spa_config_exit(spa, SCL_ALL, spa); /* * Panic the system if the specified tag requires it. This * is useful for ensuring that configurations are updated * transactionally. */ if (zio_injection_enabled) zio_handle_panic_injection(spa, tag, 0); /* * Note: this txg_wait_synced() is important because it ensures * that there won't be more than one config change per txg. * This allows us to use the txg as the generation number. */ if (error == 0) txg_wait_synced(spa->spa_dsl_pool, txg); if (vd != NULL) { ASSERT(!vd->vdev_detached || vd->vdev_dtl_sm == NULL); spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); vdev_free(vd); spa_config_exit(spa, SCL_ALL, spa); } /* * If the config changed, update the config cache. */ if (config_changed) spa_config_sync(spa, B_FALSE, B_TRUE); } /* * Unlock the spa_t after adding or removing a vdev. Besides undoing the * locking of spa_vdev_enter(), we also want make sure the transactions have * synced to disk, and then update the global configuration cache with the new * information. */ int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) { spa_vdev_config_exit(spa, vd, txg, error, FTAG); mutex_exit(&spa_namespace_lock); mutex_exit(&spa->spa_vdev_top_lock); return (error); } /* * Lock the given spa_t for the purpose of changing vdev state. */ void spa_vdev_state_enter(spa_t *spa, int oplocks) { int locks = SCL_STATE_ALL | oplocks; /* * Root pools may need to read of the underlying devfs filesystem * when opening up a vdev. Unfortunately if we're holding the * SCL_ZIO lock it will result in a deadlock when we try to issue * the read from the root filesystem. Instead we "prefetch" * the associated vnodes that we need prior to opening the * underlying devices and cache them so that we can prevent * any I/O when we are doing the actual open. */ if (spa_is_root(spa)) { int low = locks & ~(SCL_ZIO - 1); int high = locks & ~low; spa_config_enter(spa, high, spa, RW_WRITER); vdev_hold(spa->spa_root_vdev); spa_config_enter(spa, low, spa, RW_WRITER); } else { spa_config_enter(spa, locks, spa, RW_WRITER); } spa->spa_vdev_locks = locks; } int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) { boolean_t config_changed = B_FALSE; if (vd != NULL || error == 0) vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev, 0, 0, B_FALSE); if (vd != NULL) { vdev_state_dirty(vd->vdev_top); config_changed = B_TRUE; spa->spa_config_generation++; } if (spa_is_root(spa)) vdev_rele(spa->spa_root_vdev); ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); spa_config_exit(spa, spa->spa_vdev_locks, spa); /* * If anything changed, wait for it to sync. This ensures that, * from the system administrator's perspective, zpool(1M) commands * are synchronous. This is important for things like zpool offline: * when the command completes, you expect no further I/O from ZFS. */ if (vd != NULL) txg_wait_synced(spa->spa_dsl_pool, 0); /* * If the config changed, update the config cache. */ if (config_changed) { mutex_enter(&spa_namespace_lock); spa_config_sync(spa, B_FALSE, B_TRUE); mutex_exit(&spa_namespace_lock); } return (error); } /* * ========================================================================== * Miscellaneous functions * ========================================================================== */ void spa_activate_mos_feature(spa_t *spa, const char *feature, dmu_tx_t *tx) { if (!nvlist_exists(spa->spa_label_features, feature)) { fnvlist_add_boolean(spa->spa_label_features, feature); /* * When we are creating the pool (tx_txg==TXG_INITIAL), we can't * dirty the vdev config because lock SCL_CONFIG is not held. * Thankfully, in this case we don't need to dirty the config * because it will be written out anyway when we finish * creating the pool. */ if (tx->tx_txg != TXG_INITIAL) vdev_config_dirty(spa->spa_root_vdev); } } void spa_deactivate_mos_feature(spa_t *spa, const char *feature) { if (nvlist_remove_all(spa->spa_label_features, feature) == 0) vdev_config_dirty(spa->spa_root_vdev); } /* * Rename a spa_t. */ int spa_rename(const char *name, const char *newname) { spa_t *spa; int err; /* * Lookup the spa_t and grab the config lock for writing. We need to * actually open the pool so that we can sync out the necessary labels. * It's OK to call spa_open() with the namespace lock held because we * allow recursive calls for other reasons. */ mutex_enter(&spa_namespace_lock); if ((err = spa_open(name, &spa, FTAG)) != 0) { mutex_exit(&spa_namespace_lock); return (err); } spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); avl_remove(&spa_namespace_avl, spa); (void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name)); avl_add(&spa_namespace_avl, spa); /* * Sync all labels to disk with the new names by marking the root vdev * dirty and waiting for it to sync. It will pick up the new pool name * during the sync. */ vdev_config_dirty(spa->spa_root_vdev); spa_config_exit(spa, SCL_ALL, FTAG); txg_wait_synced(spa->spa_dsl_pool, 0); /* * Sync the updated config cache. */ spa_config_sync(spa, B_FALSE, B_TRUE); spa_close(spa, FTAG); mutex_exit(&spa_namespace_lock); return (0); } /* * Return the spa_t associated with given pool_guid, if it exists. If * device_guid is non-zero, determine whether the pool exists *and* contains * a device with the specified device_guid. */ spa_t * spa_by_guid(uint64_t pool_guid, uint64_t device_guid) { spa_t *spa; avl_tree_t *t = &spa_namespace_avl; ASSERT(MUTEX_HELD(&spa_namespace_lock)); for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { if (spa->spa_state == POOL_STATE_UNINITIALIZED) continue; if (spa->spa_root_vdev == NULL) continue; if (spa_guid(spa) == pool_guid) { if (device_guid == 0) break; if (vdev_lookup_by_guid(spa->spa_root_vdev, device_guid) != NULL) break; /* * Check any devices we may be in the process of adding. */ if (spa->spa_pending_vdev) { if (vdev_lookup_by_guid(spa->spa_pending_vdev, device_guid) != NULL) break; } } } return (spa); } /* * Determine whether a pool with the given pool_guid exists. */ boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) { return (spa_by_guid(pool_guid, device_guid) != NULL); } char * spa_strdup(const char *s) { size_t len; char *new; len = strlen(s); new = kmem_alloc(len + 1, KM_SLEEP); bcopy(s, new, len); new[len] = '\0'; return (new); } void spa_strfree(char *s) { kmem_free(s, strlen(s) + 1); } uint64_t spa_get_random(uint64_t range) { uint64_t r; ASSERT(range != 0); (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); return (r % range); } uint64_t spa_generate_guid(spa_t *spa) { uint64_t guid = spa_get_random(-1ULL); if (spa != NULL) { while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)) guid = spa_get_random(-1ULL); } else { while (guid == 0 || spa_guid_exists(guid, 0)) guid = spa_get_random(-1ULL); } return (guid); } void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp) { char type[256]; char *checksum = NULL; char *compress = NULL; if (bp != NULL) { if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { dmu_object_byteswap_t bswap = DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); (void) snprintf(type, sizeof (type), "bswap %s %s", DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? "metadata" : "data", dmu_ot_byteswap[bswap].ob_name); } else { (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, sizeof (type)); } if (!BP_IS_EMBEDDED(bp)) { checksum = zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; } compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; } SNPRINTF_BLKPTR(snprintf, ' ', buf, buflen, bp, type, checksum, compress); } void spa_freeze(spa_t *spa) { uint64_t freeze_txg = 0; spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); if (spa->spa_freeze_txg == UINT64_MAX) { freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; spa->spa_freeze_txg = freeze_txg; } spa_config_exit(spa, SCL_ALL, FTAG); if (freeze_txg != 0) txg_wait_synced(spa_get_dsl(spa), freeze_txg); } void zfs_panic_recover(const char *fmt, ...) { va_list adx; va_start(adx, fmt); vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); va_end(adx); } /* * This is a stripped-down version of strtoull, suitable only for converting * lowercase hexadecimal numbers that don't overflow. */ uint64_t zfs_strtonum(const char *str, char **nptr) { uint64_t val = 0; char c; int digit; while ((c = *str) != '\0') { if (c >= '0' && c <= '9') digit = c - '0'; else if (c >= 'a' && c <= 'f') digit = 10 + c - 'a'; else break; val *= 16; val += digit; str++; } if (nptr) *nptr = (char *)str; return (val); } /* * ========================================================================== * Accessor functions * ========================================================================== */ boolean_t spa_shutting_down(spa_t *spa) { return (spa->spa_async_suspended); } dsl_pool_t * spa_get_dsl(spa_t *spa) { return (spa->spa_dsl_pool); } boolean_t spa_is_initializing(spa_t *spa) { return (spa->spa_is_initializing); } blkptr_t * spa_get_rootblkptr(spa_t *spa) { return (&spa->spa_ubsync.ub_rootbp); } void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) { spa->spa_uberblock.ub_rootbp = *bp; } void spa_altroot(spa_t *spa, char *buf, size_t buflen) { if (spa->spa_root == NULL) buf[0] = '\0'; else (void) strncpy(buf, spa->spa_root, buflen); } int spa_sync_pass(spa_t *spa) { return (spa->spa_sync_pass); } char * spa_name(spa_t *spa) { return (spa->spa_name); } uint64_t spa_guid(spa_t *spa) { dsl_pool_t *dp = spa_get_dsl(spa); uint64_t guid; /* * If we fail to parse the config during spa_load(), we can go through * the error path (which posts an ereport) and end up here with no root * vdev. We stash the original pool guid in 'spa_config_guid' to handle * this case. */ if (spa->spa_root_vdev == NULL) return (spa->spa_config_guid); guid = spa->spa_last_synced_guid != 0 ? spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid; /* * Return the most recently synced out guid unless we're * in syncing context. */ if (dp && dsl_pool_sync_context(dp)) return (spa->spa_root_vdev->vdev_guid); else return (guid); } uint64_t spa_load_guid(spa_t *spa) { /* * This is a GUID that exists solely as a reference for the * purposes of the arc. It is generated at load time, and * is never written to persistent storage. */ return (spa->spa_load_guid); } uint64_t spa_last_synced_txg(spa_t *spa) { return (spa->spa_ubsync.ub_txg); } uint64_t spa_first_txg(spa_t *spa) { return (spa->spa_first_txg); } uint64_t spa_syncing_txg(spa_t *spa) { return (spa->spa_syncing_txg); } pool_state_t spa_state(spa_t *spa) { return (spa->spa_state); } spa_load_state_t spa_load_state(spa_t *spa) { return (spa->spa_load_state); } uint64_t spa_freeze_txg(spa_t *spa) { return (spa->spa_freeze_txg); } /* ARGSUSED */ uint64_t spa_get_asize(spa_t *spa, uint64_t lsize) { return (lsize * spa_asize_inflation); } /* * Return the amount of slop space in bytes. It is 1/32 of the pool (3.2%), * or at least 32MB. * * See the comment above spa_slop_shift for details. */ uint64_t spa_get_slop_space(spa_t *spa) { uint64_t space = spa_get_dspace(spa); return (MAX(space >> spa_slop_shift, SPA_MINDEVSIZE >> 1)); } uint64_t spa_get_dspace(spa_t *spa) { return (spa->spa_dspace); } void spa_update_dspace(spa_t *spa) { spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + ddt_get_dedup_dspace(spa); } /* * Return the failure mode that has been set to this pool. The default * behavior will be to block all I/Os when a complete failure occurs. */ uint8_t spa_get_failmode(spa_t *spa) { return (spa->spa_failmode); } boolean_t spa_suspended(spa_t *spa) { return (spa->spa_suspended); } uint64_t spa_version(spa_t *spa) { return (spa->spa_ubsync.ub_version); } boolean_t spa_deflate(spa_t *spa) { return (spa->spa_deflate); } metaslab_class_t * spa_normal_class(spa_t *spa) { return (spa->spa_normal_class); } metaslab_class_t * spa_log_class(spa_t *spa) { return (spa->spa_log_class); } void spa_evicting_os_register(spa_t *spa, objset_t *os) { mutex_enter(&spa->spa_evicting_os_lock); list_insert_head(&spa->spa_evicting_os_list, os); mutex_exit(&spa->spa_evicting_os_lock); } void spa_evicting_os_deregister(spa_t *spa, objset_t *os) { mutex_enter(&spa->spa_evicting_os_lock); list_remove(&spa->spa_evicting_os_list, os); cv_broadcast(&spa->spa_evicting_os_cv); mutex_exit(&spa->spa_evicting_os_lock); } void spa_evicting_os_wait(spa_t *spa) { mutex_enter(&spa->spa_evicting_os_lock); while (!list_is_empty(&spa->spa_evicting_os_list)) cv_wait(&spa->spa_evicting_os_cv, &spa->spa_evicting_os_lock); mutex_exit(&spa->spa_evicting_os_lock); dmu_buf_user_evict_wait(); } int spa_max_replication(spa_t *spa) { /* * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to * handle BPs with more than one DVA allocated. Set our max * replication level accordingly. */ if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) return (1); return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); } int spa_prev_software_version(spa_t *spa) { return (spa->spa_prev_software_version); } uint64_t spa_deadman_synctime(spa_t *spa) { return (spa->spa_deadman_synctime); } uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva) { uint64_t asize = DVA_GET_ASIZE(dva); uint64_t dsize = asize; ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); if (asize != 0 && spa->spa_deflate) { uint64_t vdev = DVA_GET_VDEV(dva); vdev_t *vd = vdev_lookup_top(spa, vdev); if (vd == NULL) { panic( "dva_get_dsize_sync(): bad DVA %llu:%llu", (u_longlong_t)vdev, (u_longlong_t)asize); } dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; } return (dsize); } uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) { uint64_t dsize = 0; for (int d = 0; d < BP_GET_NDVAS(bp); d++) dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); return (dsize); } uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp) { uint64_t dsize = 0; spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); for (int d = 0; d < BP_GET_NDVAS(bp); d++) dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); spa_config_exit(spa, SCL_VDEV, FTAG); return (dsize); } /* * ========================================================================== * Initialization and Termination * ========================================================================== */ static int spa_name_compare(const void *a1, const void *a2) { const spa_t *s1 = a1; const spa_t *s2 = a2; int s; s = strcmp(s1->spa_name, s2->spa_name); if (s > 0) return (1); if (s < 0) return (-1); return (0); } int spa_busy(void) { return (spa_active_count); } void spa_boot_init() { spa_config_load(); } #ifdef _KERNEL EVENTHANDLER_DEFINE(mountroot, spa_boot_init, NULL, 0); #endif void spa_init(int mode) { mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), offsetof(spa_t, spa_avl)); avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), offsetof(spa_aux_t, aux_avl)); avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), offsetof(spa_aux_t, aux_avl)); spa_mode_global = mode; #ifdef illumos #ifdef _KERNEL spa_arch_init(); #else if (spa_mode_global != FREAD && dprintf_find_string("watch")) { arc_procfd = open("/proc/self/ctl", O_WRONLY); if (arc_procfd == -1) { perror("could not enable watchpoints: " "opening /proc/self/ctl failed: "); } else { arc_watch = B_TRUE; } } #endif #endif /* illumos */ refcount_sysinit(); unique_init(); range_tree_init(); zio_init(); lz4_init(); dmu_init(); zil_init(); vdev_cache_stat_init(); zfs_prop_init(); zpool_prop_init(); zpool_feature_init(); spa_config_load(); l2arc_start(); #ifndef illumos #ifdef _KERNEL zfs_deadman_init(); #endif #endif /* !illumos */ } void spa_fini(void) { l2arc_stop(); spa_evict_all(); vdev_cache_stat_fini(); zil_fini(); dmu_fini(); lz4_fini(); zio_fini(); range_tree_fini(); unique_fini(); refcount_fini(); avl_destroy(&spa_namespace_avl); avl_destroy(&spa_spare_avl); avl_destroy(&spa_l2cache_avl); cv_destroy(&spa_namespace_cv); mutex_destroy(&spa_namespace_lock); mutex_destroy(&spa_spare_lock); mutex_destroy(&spa_l2cache_lock); } /* * Return whether this pool has slogs. No locking needed. * It's not a problem if the wrong answer is returned as it's only for * performance and not correctness */ boolean_t spa_has_slogs(spa_t *spa) { return (spa->spa_log_class->mc_rotor != NULL); } spa_log_state_t spa_get_log_state(spa_t *spa) { return (spa->spa_log_state); } void spa_set_log_state(spa_t *spa, spa_log_state_t state) { spa->spa_log_state = state; } boolean_t spa_is_root(spa_t *spa) { return (spa->spa_is_root); } boolean_t spa_writeable(spa_t *spa) { return (!!(spa->spa_mode & FWRITE)); } /* * Returns true if there is a pending sync task in any of the current * syncing txg, the current quiescing txg, or the current open txg. */ boolean_t spa_has_pending_synctask(spa_t *spa) { return (!txg_all_lists_empty(&spa->spa_dsl_pool->dp_sync_tasks)); } int spa_mode(spa_t *spa) { return (spa->spa_mode); } uint64_t spa_bootfs(spa_t *spa) { return (spa->spa_bootfs); } uint64_t spa_delegation(spa_t *spa) { return (spa->spa_delegation); } objset_t * spa_meta_objset(spa_t *spa) { return (spa->spa_meta_objset); } enum zio_checksum spa_dedup_checksum(spa_t *spa) { return (spa->spa_dedup_checksum); } /* * Reset pool scan stat per scan pass (or reboot). */ void spa_scan_stat_init(spa_t *spa) { /* data not stored on disk */ spa->spa_scan_pass_start = gethrestime_sec(); spa->spa_scan_pass_exam = 0; vdev_scan_stat_init(spa->spa_root_vdev); } /* * Get scan stats for zpool status reports */ int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) { dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE) return (SET_ERROR(ENOENT)); bzero(ps, sizeof (pool_scan_stat_t)); /* data stored on disk */ ps->pss_func = scn->scn_phys.scn_func; ps->pss_start_time = scn->scn_phys.scn_start_time; ps->pss_end_time = scn->scn_phys.scn_end_time; ps->pss_to_examine = scn->scn_phys.scn_to_examine; ps->pss_examined = scn->scn_phys.scn_examined; ps->pss_to_process = scn->scn_phys.scn_to_process; ps->pss_processed = scn->scn_phys.scn_processed; ps->pss_errors = scn->scn_phys.scn_errors; ps->pss_state = scn->scn_phys.scn_state; /* data not stored on disk */ ps->pss_pass_start = spa->spa_scan_pass_start; ps->pss_pass_exam = spa->spa_scan_pass_exam; return (0); } boolean_t spa_debug_enabled(spa_t *spa) { return (spa->spa_debug); } int spa_maxblocksize(spa_t *spa) { if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) return (SPA_MAXBLOCKSIZE); else return (SPA_OLD_MAXBLOCKSIZE); } Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/dmu.h =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/dmu.h (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/dmu.h (revision 290757) @@ -1,950 +1,952 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2014 by Delphix. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2012, Joyent, Inc. All rights reserved. * Copyright 2013 DEY Storage Systems, Inc. * Copyright 2014 HybridCluster. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. + * Copyright 2013 Saso Kiselkov. All rights reserved. */ /* Portions Copyright 2010 Robert Milkowski */ #ifndef _SYS_DMU_H #define _SYS_DMU_H /* * This file describes the interface that the DMU provides for its * consumers. * * The DMU also interacts with the SPA. That interface is described in * dmu_spa.h. */ #include #include #include #include #ifdef __cplusplus extern "C" { #endif struct uio; struct xuio; struct page; struct vnode; struct spa; struct zilog; struct zio; struct blkptr; struct zap_cursor; struct dsl_dataset; struct dsl_pool; struct dnode; struct drr_begin; struct drr_end; struct zbookmark_phys; struct spa; struct nvlist; struct arc_buf; struct zio_prop; struct sa_handle; struct file; typedef struct objset objset_t; typedef struct dmu_tx dmu_tx_t; typedef struct dsl_dir dsl_dir_t; typedef enum dmu_object_byteswap { DMU_BSWAP_UINT8, DMU_BSWAP_UINT16, DMU_BSWAP_UINT32, DMU_BSWAP_UINT64, DMU_BSWAP_ZAP, DMU_BSWAP_DNODE, DMU_BSWAP_OBJSET, DMU_BSWAP_ZNODE, DMU_BSWAP_OLDACL, DMU_BSWAP_ACL, /* * Allocating a new byteswap type number makes the on-disk format * incompatible with any other format that uses the same number. * * Data can usually be structured to work with one of the * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. */ DMU_BSWAP_NUMFUNCS } dmu_object_byteswap_t; #define DMU_OT_NEWTYPE 0x80 #define DMU_OT_METADATA 0x40 #define DMU_OT_BYTESWAP_MASK 0x3f /* * Defines a uint8_t object type. Object types specify if the data * in the object is metadata (boolean) and how to byteswap the data * (dmu_object_byteswap_t). */ #define DMU_OT(byteswap, metadata) \ (DMU_OT_NEWTYPE | \ ((metadata) ? DMU_OT_METADATA : 0) | \ ((byteswap) & DMU_OT_BYTESWAP_MASK)) #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \ ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \ (ot) < DMU_OT_NUMTYPES) #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \ ((ot) & DMU_OT_METADATA) : \ dmu_ot[(ot)].ot_metadata) /* * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill * is repurposed for embedded BPs. */ #define DMU_OT_HAS_FILL(ot) \ ((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET) #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \ ((ot) & DMU_OT_BYTESWAP_MASK) : \ dmu_ot[(ot)].ot_byteswap) typedef enum dmu_object_type { DMU_OT_NONE, /* general: */ DMU_OT_OBJECT_DIRECTORY, /* ZAP */ DMU_OT_OBJECT_ARRAY, /* UINT64 */ DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ DMU_OT_BPOBJ, /* UINT64 */ DMU_OT_BPOBJ_HDR, /* UINT64 */ /* spa: */ DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ DMU_OT_SPACE_MAP, /* UINT64 */ /* zil: */ DMU_OT_INTENT_LOG, /* UINT64 */ /* dmu: */ DMU_OT_DNODE, /* DNODE */ DMU_OT_OBJSET, /* OBJSET */ /* dsl: */ DMU_OT_DSL_DIR, /* UINT64 */ DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ DMU_OT_DSL_PROPS, /* ZAP */ DMU_OT_DSL_DATASET, /* UINT64 */ /* zpl: */ DMU_OT_ZNODE, /* ZNODE */ DMU_OT_OLDACL, /* Old ACL */ DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ DMU_OT_MASTER_NODE, /* ZAP */ DMU_OT_UNLINKED_SET, /* ZAP */ /* zvol: */ DMU_OT_ZVOL, /* UINT8 */ DMU_OT_ZVOL_PROP, /* ZAP */ /* other; for testing only! */ DMU_OT_PLAIN_OTHER, /* UINT8 */ DMU_OT_UINT64_OTHER, /* UINT64 */ DMU_OT_ZAP_OTHER, /* ZAP */ /* new object types: */ DMU_OT_ERROR_LOG, /* ZAP */ DMU_OT_SPA_HISTORY, /* UINT8 */ DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ DMU_OT_POOL_PROPS, /* ZAP */ DMU_OT_DSL_PERMS, /* ZAP */ DMU_OT_ACL, /* ACL */ DMU_OT_SYSACL, /* SYSACL */ DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ DMU_OT_NEXT_CLONES, /* ZAP */ DMU_OT_SCAN_QUEUE, /* ZAP */ DMU_OT_USERGROUP_USED, /* ZAP */ DMU_OT_USERGROUP_QUOTA, /* ZAP */ DMU_OT_USERREFS, /* ZAP */ DMU_OT_DDT_ZAP, /* ZAP */ DMU_OT_DDT_STATS, /* ZAP */ DMU_OT_SA, /* System attr */ DMU_OT_SA_MASTER_NODE, /* ZAP */ DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ DMU_OT_SCAN_XLATE, /* ZAP */ DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ DMU_OT_DEADLIST, /* ZAP */ DMU_OT_DEADLIST_HDR, /* UINT64 */ DMU_OT_DSL_CLONES, /* ZAP */ DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ /* * Do not allocate new object types here. Doing so makes the on-disk * format incompatible with any other format that uses the same object * type number. * * When creating an object which does not have one of the above types * use the DMU_OTN_* type with the correct byteswap and metadata * values. * * The DMU_OTN_* types do not have entries in the dmu_ot table, * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead * of indexing into dmu_ot directly (this works for both DMU_OT_* types * and DMU_OTN_* types). */ DMU_OT_NUMTYPES, /* * Names for valid types declared with DMU_OT(). */ DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE), } dmu_object_type_t; typedef enum txg_how { TXG_WAIT = 1, TXG_NOWAIT, TXG_WAITED, } txg_how_t; void byteswap_uint64_array(void *buf, size_t size); void byteswap_uint32_array(void *buf, size_t size); void byteswap_uint16_array(void *buf, size_t size); void byteswap_uint8_array(void *buf, size_t size); void zap_byteswap(void *buf, size_t size); void zfs_oldacl_byteswap(void *buf, size_t size); void zfs_acl_byteswap(void *buf, size_t size); void zfs_znode_byteswap(void *buf, size_t size); #define DS_FIND_SNAPSHOTS (1<<0) #define DS_FIND_CHILDREN (1<<1) #define DS_FIND_SERIALIZE (1<<2) /* * The maximum number of bytes that can be accessed as part of one * operation, including metadata. */ #define DMU_MAX_ACCESS (32 * 1024 * 1024) /* 32MB */ #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ #define DMU_USERUSED_OBJECT (-1ULL) #define DMU_GROUPUSED_OBJECT (-2ULL) /* * artificial blkids for bonus buffer and spill blocks */ #define DMU_BONUS_BLKID (-1ULL) #define DMU_SPILL_BLKID (-2ULL) /* * Public routines to create, destroy, open, and close objsets. */ int dmu_objset_hold(const char *name, void *tag, objset_t **osp); int dmu_objset_own(const char *name, dmu_objset_type_t type, boolean_t readonly, void *tag, objset_t **osp); void dmu_objset_rele(objset_t *os, void *tag); void dmu_objset_disown(objset_t *os, void *tag); int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); void dmu_objset_evict_dbufs(objset_t *os); int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg); int dmu_get_recursive_snaps_nvl(char *fsname, const char *snapname, struct nvlist *snaps); int dmu_objset_clone(const char *name, const char *origin); int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer, struct nvlist *errlist); int dmu_objset_snapshot_one(const char *fsname, const char *snapname); int dmu_objset_snapshot_tmp(const char *, const char *, int); int dmu_objset_find(char *name, int func(const char *, void *), void *arg, int flags); void dmu_objset_byteswap(void *buf, size_t size); int dsl_dataset_rename_snapshot(const char *fsname, const char *oldsnapname, const char *newsnapname, boolean_t recursive); typedef struct dmu_buf { uint64_t db_object; /* object that this buffer is part of */ uint64_t db_offset; /* byte offset in this object */ uint64_t db_size; /* size of buffer in bytes */ void *db_data; /* data in buffer */ } dmu_buf_t; /* * The names of zap entries in the DIRECTORY_OBJECT of the MOS. */ #define DMU_POOL_DIRECTORY_OBJECT 1 #define DMU_POOL_CONFIG "config" #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" #define DMU_POOL_FEATURES_FOR_READ "features_for_read" #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" #define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg" #define DMU_POOL_ROOT_DATASET "root_dataset" #define DMU_POOL_SYNC_BPOBJ "sync_bplist" #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" #define DMU_POOL_ERRLOG_LAST "errlog_last" #define DMU_POOL_SPARES "spares" #define DMU_POOL_DEFLATE "deflate" #define DMU_POOL_HISTORY "history" #define DMU_POOL_PROPS "pool_props" #define DMU_POOL_L2CACHE "l2cache" #define DMU_POOL_TMP_USERREFS "tmp_userrefs" #define DMU_POOL_DDT "DDT-%s-%s-%s" #define DMU_POOL_DDT_STATS "DDT-statistics" #define DMU_POOL_CREATION_VERSION "creation_version" #define DMU_POOL_SCAN "scan" #define DMU_POOL_FREE_BPOBJ "free_bpobj" #define DMU_POOL_BPTREE_OBJ "bptree_obj" #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" +#define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" /* * Allocate an object from this objset. The range of object numbers * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. * * The transaction must be assigned to a txg. The newly allocated * object will be "held" in the transaction (ie. you can modify the * newly allocated object in this transaction). * * dmu_object_alloc() chooses an object and returns it in *objectp. * * dmu_object_claim() allocates a specific object number. If that * number is already allocated, it fails and returns EEXIST. * * Return 0 on success, or ENOSPC or EEXIST as specified above. */ uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp); /* * Free an object from this objset. * * The object's data will be freed as well (ie. you don't need to call * dmu_free(object, 0, -1, tx)). * * The object need not be held in the transaction. * * If there are any holds on this object's buffers (via dmu_buf_hold()), * or tx holds on the object (via dmu_tx_hold_object()), you can not * free it; it fails and returns EBUSY. * * If the object is not allocated, it fails and returns ENOENT. * * Return 0 on success, or EBUSY or ENOENT as specified above. */ int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); /* * Find the next allocated or free object. * * The objectp parameter is in-out. It will be updated to be the next * object which is allocated. Ignore objects which have not been * modified since txg. * * XXX Can only be called on a objset with no dirty data. * * Returns 0 on success, or ENOENT if there are no more objects. */ int dmu_object_next(objset_t *os, uint64_t *objectp, boolean_t hole, uint64_t txg); /* * Set the data blocksize for an object. * * The object cannot have any blocks allcated beyond the first. If * the first block is allocated already, the new size must be greater * than the current block size. If these conditions are not met, * ENOTSUP will be returned. * * Returns 0 on success, or EBUSY if there are any holds on the object * contents, or ENOTSUP as described above. */ int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, int ibs, dmu_tx_t *tx); /* * Set the checksum property on a dnode. The new checksum algorithm will * apply to all newly written blocks; existing blocks will not be affected. */ void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, dmu_tx_t *tx); /* * Set the compress property on a dnode. The new compression algorithm will * apply to all newly written blocks; existing blocks will not be affected. */ void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, dmu_tx_t *tx); void dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset, void *data, uint8_t etype, uint8_t comp, int uncompressed_size, int compressed_size, int byteorder, dmu_tx_t *tx); /* * Decide how to write a block: checksum, compression, number of copies, etc. */ #define WP_NOFILL 0x1 #define WP_DMU_SYNC 0x2 #define WP_SPILL 0x4 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp, struct zio_prop *zp); /* * The bonus data is accessed more or less like a regular buffer. * You must dmu_bonus_hold() to get the buffer, which will give you a * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus * data. As with any normal buffer, you must call dmu_buf_read() to * read db_data, dmu_buf_will_dirty() before modifying it, and the * object must be held in an assigned transaction before calling * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus * buffer as well. You must release your hold with dmu_buf_rele(). * * Returns ENOENT, EIO, or 0. */ int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **); int dmu_bonus_max(void); int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); dmu_object_type_t dmu_get_bonustype(dmu_buf_t *); int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); /* * Special spill buffer support used by "SA" framework */ int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags, void *tag, dmu_buf_t **dbp); int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); /* * Obtain the DMU buffer from the specified object which contains the * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so * that it will remain in memory. You must release the hold with * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your * hold. You must have a hold on any dmu_buf_t* you pass to the DMU. * * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill * on the returned buffer before reading or writing the buffer's * db_data. The comments for those routines describe what particular * operations are valid after calling them. * * The object number must be a valid, allocated object number. */ int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, void *tag, dmu_buf_t **, int flags); /* * Add a reference to a dmu buffer that has already been held via * dmu_buf_hold() in the current context. */ void dmu_buf_add_ref(dmu_buf_t *db, void* tag); /* * Attempt to add a reference to a dmu buffer that is in an unknown state, * using a pointer that may have been invalidated by eviction processing. * The request will succeed if the passed in dbuf still represents the * same os/object/blkid, is ineligible for eviction, and has at least * one hold by a user other than the syncer. */ boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object, uint64_t blkid, void *tag); void dmu_buf_rele(dmu_buf_t *db, void *tag); uint64_t dmu_buf_refcount(dmu_buf_t *db); /* * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a * range of an object. A pointer to an array of dmu_buf_t*'s is * returned (in *dbpp). * * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and * frees the array. The hold on the array of buffers MUST be released * with dmu_buf_rele_array. You can NOT release the hold on each buffer * individually with dmu_buf_rele. */ int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, uint64_t length, boolean_t read, void *tag, int *numbufsp, dmu_buf_t ***dbpp); void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag); typedef void dmu_buf_evict_func_t(void *user_ptr); /* * A DMU buffer user object may be associated with a dbuf for the * duration of its lifetime. This allows the user of a dbuf (client) * to attach private data to a dbuf (e.g. in-core only data such as a * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified * when that dbuf has been evicted. Clients typically respond to the * eviction notification by freeing their private data, thus ensuring * the same lifetime for both dbuf and private data. * * The mapping from a dmu_buf_user_t to any client private data is the * client's responsibility. All current consumers of the API with private * data embed a dmu_buf_user_t as the first member of the structure for * their private data. This allows conversions between the two types * with a simple cast. Since the DMU buf user API never needs access * to the private data, other strategies can be employed if necessary * or convenient for the client (e.g. using container_of() to do the * conversion for private data that cannot have the dmu_buf_user_t as * its first member). * * Eviction callbacks are executed without the dbuf mutex held or any * other type of mechanism to guarantee that the dbuf is still available. * For this reason, users must assume the dbuf has already been freed * and not reference the dbuf from the callback context. * * Users requesting "immediate eviction" are notified as soon as the dbuf * is only referenced by dirty records (dirties == holds). Otherwise the * notification occurs after eviction processing for the dbuf begins. */ typedef struct dmu_buf_user { /* * Asynchronous user eviction callback state. */ taskq_ent_t dbu_tqent; /* This instance's eviction function pointer. */ dmu_buf_evict_func_t *dbu_evict_func; #ifdef ZFS_DEBUG /* * Pointer to user's dbuf pointer. NULL for clients that do * not associate a dbuf with their user data. * * The dbuf pointer is cleared upon eviction so as to catch * use-after-evict bugs in clients. */ dmu_buf_t **dbu_clear_on_evict_dbufp; #endif } dmu_buf_user_t; /* * Initialize the given dmu_buf_user_t instance with the eviction function * evict_func, to be called when the user is evicted. * * NOTE: This function should only be called once on a given dmu_buf_user_t. * To allow enforcement of this, dbu must already be zeroed on entry. */ #ifdef __lint /* Very ugly, but it beats issuing suppression directives in many Makefiles. */ extern void dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func, dmu_buf_t **clear_on_evict_dbufp); #else /* __lint */ static inline void dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func, dmu_buf_t **clear_on_evict_dbufp) { ASSERT(dbu->dbu_evict_func == NULL); ASSERT(evict_func != NULL); dbu->dbu_evict_func = evict_func; #ifdef ZFS_DEBUG dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp; #endif } #endif /* __lint */ /* * Attach user data to a dbuf and mark it for normal (when the dbuf's * data is cleared or its reference count goes to zero) eviction processing. * * Returns NULL on success, or the existing user if another user currently * owns the buffer. */ void *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user); /* * Attach user data to a dbuf and mark it for immediate (its dirty and * reference counts are equal) eviction processing. * * Returns NULL on success, or the existing user if another user currently * owns the buffer. */ void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user); /* * Replace the current user of a dbuf. * * If given the current user of a dbuf, replaces the dbuf's user with * "new_user" and returns the user data pointer that was replaced. * Otherwise returns the current, and unmodified, dbuf user pointer. */ void *dmu_buf_replace_user(dmu_buf_t *db, dmu_buf_user_t *old_user, dmu_buf_user_t *new_user); /* * Remove the specified user data for a DMU buffer. * * Returns the user that was removed on success, or the current user if * another user currently owns the buffer. */ void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user); /* * Returns the user data (dmu_buf_user_t *) associated with this dbuf. */ void *dmu_buf_get_user(dmu_buf_t *db); /* Block until any in-progress dmu buf user evictions complete. */ void dmu_buf_user_evict_wait(void); /* * Returns the blkptr associated with this dbuf, or NULL if not set. */ struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db); /* * Indicate that you are going to modify the buffer's data (db_data). * * The transaction (tx) must be assigned to a txg (ie. you've called * dmu_tx_assign()). The buffer's object must be held in the tx * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). */ void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); /* * Tells if the given dbuf is freeable. */ boolean_t dmu_buf_freeable(dmu_buf_t *); /* * You must create a transaction, then hold the objects which you will * (or might) modify as part of this transaction. Then you must assign * the transaction to a transaction group. Once the transaction has * been assigned, you can modify buffers which belong to held objects as * part of this transaction. You can't modify buffers before the * transaction has been assigned; you can't modify buffers which don't * belong to objects which this transaction holds; you can't hold * objects once the transaction has been assigned. You may hold an * object which you are going to free (with dmu_object_free()), but you * don't have to. * * You can abort the transaction before it has been assigned. * * Note that you may hold buffers (with dmu_buf_hold) at any time, * regardless of transaction state. */ #define DMU_NEW_OBJECT (-1ULL) #define DMU_OBJECT_END (-1ULL) dmu_tx_t *dmu_tx_create(objset_t *os); void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, uint64_t len); void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); void dmu_tx_abort(dmu_tx_t *tx); int dmu_tx_assign(dmu_tx_t *tx, enum txg_how txg_how); void dmu_tx_wait(dmu_tx_t *tx); void dmu_tx_commit(dmu_tx_t *tx); void dmu_tx_mark_netfree(dmu_tx_t *tx); /* * To register a commit callback, dmu_tx_callback_register() must be called. * * dcb_data is a pointer to caller private data that is passed on as a * callback parameter. The caller is responsible for properly allocating and * freeing it. * * When registering a callback, the transaction must be already created, but * it cannot be committed or aborted. It can be assigned to a txg or not. * * The callback will be called after the transaction has been safely written * to stable storage and will also be called if the dmu_tx is aborted. * If there is any error which prevents the transaction from being committed to * disk, the callback will be called with a value of error != 0. */ typedef void dmu_tx_callback_func_t(void *dcb_data, int error); void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, void *dcb_data); /* * Free up the data blocks for a defined range of a file. If size is * -1, the range from offset to end-of-file is freed. */ int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, dmu_tx_t *tx); int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, uint64_t size); int dmu_free_long_object(objset_t *os, uint64_t object); /* * Convenience functions. * * Canfail routines will return 0 on success, or an errno if there is a * nonrecoverable I/O error. */ #define DMU_READ_PREFETCH 0 /* prefetch */ #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, void *buf, uint32_t flags); void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, const void *buf, dmu_tx_t *tx); void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, dmu_tx_t *tx); int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size); int dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size); int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size, dmu_tx_t *tx); int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size, dmu_tx_t *tx); #ifdef _KERNEL #ifdef sun int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, struct page *pp, dmu_tx_t *tx); #else int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, struct vm_page **ppa, dmu_tx_t *tx); #endif #endif struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); void dmu_return_arcbuf(struct arc_buf *buf); void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf, dmu_tx_t *tx); int dmu_xuio_init(struct xuio *uio, int niov); void dmu_xuio_fini(struct xuio *uio); int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off, size_t n); int dmu_xuio_cnt(struct xuio *uio); struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i); void dmu_xuio_clear(struct xuio *uio, int i); void xuio_stat_wbuf_copied(); void xuio_stat_wbuf_nocopy(); extern boolean_t zfs_prefetch_disable; extern int zfs_max_recordsize; /* * Asynchronously try to read in the data. */ void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset, uint64_t len, enum zio_priority pri); typedef struct dmu_object_info { /* All sizes are in bytes unless otherwise indicated. */ uint32_t doi_data_block_size; uint32_t doi_metadata_block_size; dmu_object_type_t doi_type; dmu_object_type_t doi_bonus_type; uint64_t doi_bonus_size; uint8_t doi_indirection; /* 2 = dnode->indirect->data */ uint8_t doi_checksum; uint8_t doi_compress; uint8_t doi_nblkptr; uint8_t doi_pad[4]; uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ uint64_t doi_max_offset; uint64_t doi_fill_count; /* number of non-empty blocks */ } dmu_object_info_t; typedef void arc_byteswap_func_t(void *buf, size_t size); typedef struct dmu_object_type_info { dmu_object_byteswap_t ot_byteswap; boolean_t ot_metadata; char *ot_name; } dmu_object_type_info_t; typedef struct dmu_object_byteswap_info { arc_byteswap_func_t *ob_func; char *ob_name; } dmu_object_byteswap_info_t; extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS]; /* * Get information on a DMU object. * * Return 0 on success or ENOENT if object is not allocated. * * If doi is NULL, just indicates whether the object exists. */ int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); /* Like dmu_object_info, but faster if you have a held dnode in hand. */ void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi); /* Like dmu_object_info, but faster if you have a held dbuf in hand. */ void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); /* * Like dmu_object_info_from_db, but faster still when you only care about * the size. This is specifically optimized for zfs_getattr(). */ void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, u_longlong_t *nblk512); typedef struct dmu_objset_stats { uint64_t dds_num_clones; /* number of clones of this */ uint64_t dds_creation_txg; uint64_t dds_guid; dmu_objset_type_t dds_type; uint8_t dds_is_snapshot; uint8_t dds_inconsistent; char dds_origin[MAXNAMELEN]; } dmu_objset_stats_t; /* * Get stats on a dataset. */ void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); /* * Add entries to the nvlist for all the objset's properties. See * zfs_prop_table[] and zfs(1m) for details on the properties. */ void dmu_objset_stats(objset_t *os, struct nvlist *nv); /* * Get the space usage statistics for statvfs(). * * refdbytes is the amount of space "referenced" by this objset. * availbytes is the amount of space available to this objset, taking * into account quotas & reservations, assuming that no other objsets * use the space first. These values correspond to the 'referenced' and * 'available' properties, described in the zfs(1m) manpage. * * usedobjs and availobjs are the number of objects currently allocated, * and available. */ void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, uint64_t *usedobjsp, uint64_t *availobjsp); /* * The fsid_guid is a 56-bit ID that can change to avoid collisions. * (Contrast with the ds_guid which is a 64-bit ID that will never * change, so there is a small probability that it will collide.) */ uint64_t dmu_objset_fsid_guid(objset_t *os); /* * Get the [cm]time for an objset's snapshot dir */ timestruc_t dmu_objset_snap_cmtime(objset_t *os); int dmu_objset_is_snapshot(objset_t *os); extern struct spa *dmu_objset_spa(objset_t *os); extern struct zilog *dmu_objset_zil(objset_t *os); extern struct dsl_pool *dmu_objset_pool(objset_t *os); extern struct dsl_dataset *dmu_objset_ds(objset_t *os); extern void dmu_objset_name(objset_t *os, char *buf); extern dmu_objset_type_t dmu_objset_type(objset_t *os); extern uint64_t dmu_objset_id(objset_t *os); extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os); extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os); extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, uint64_t *id, uint64_t *offp, boolean_t *case_conflict); extern int dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, boolean_t *conflict); extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, uint64_t *idp, uint64_t *offp); typedef int objset_used_cb_t(dmu_object_type_t bonustype, void *bonus, uint64_t *userp, uint64_t *groupp); extern void dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb); extern void dmu_objset_set_user(objset_t *os, void *user_ptr); extern void *dmu_objset_get_user(objset_t *os); /* * Return the txg number for the given assigned transaction. */ uint64_t dmu_tx_get_txg(dmu_tx_t *tx); /* * Synchronous write. * If a parent zio is provided this function initiates a write on the * provided buffer as a child of the parent zio. * In the absence of a parent zio, the write is completed synchronously. * At write completion, blk is filled with the bp of the written block. * Note that while the data covered by this function will be on stable * storage when the write completes this new data does not become a * permanent part of the file until the associated transaction commits. */ /* * {zfs,zvol,ztest}_get_done() args */ typedef struct zgd { struct zilog *zgd_zilog; struct blkptr *zgd_bp; dmu_buf_t *zgd_db; struct rl *zgd_rl; void *zgd_private; } zgd_t; typedef void dmu_sync_cb_t(zgd_t *arg, int error); int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); /* * Find the next hole or data block in file starting at *off * Return found offset in *off. Return ESRCH for end of file. */ int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, uint64_t *off); /* * Check if a DMU object has any dirty blocks. If so, sync out * all pending transaction groups. Otherwise, this function * does not alter DMU state. This could be improved to only sync * out the necessary transaction groups for this particular * object. */ int dmu_object_wait_synced(objset_t *os, uint64_t object); /* * Initial setup and final teardown. */ extern void dmu_init(void); extern void dmu_fini(void); typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, uint64_t object, uint64_t offset, int len); void dmu_traverse_objset(objset_t *os, uint64_t txg_start, dmu_traverse_cb_t cb, void *arg); int dmu_diff(const char *tosnap_name, const char *fromsnap_name, struct file *fp, offset_t *offp); /* CRC64 table */ #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ extern uint64_t zfs_crc64_table[256]; extern int zfs_mdcomp_disable; #ifdef __cplusplus } #endif #endif /* _SYS_DMU_H */ Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/spa.h =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/spa.h (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/spa.h (revision 290757) @@ -1,913 +1,922 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2014 by Delphix. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. + * Copyright 2013 Saso Kiselkov. All rights reserved. */ #ifndef _SYS_SPA_H #define _SYS_SPA_H #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif /* * Forward references that lots of things need. */ typedef struct spa spa_t; typedef struct vdev vdev_t; typedef struct metaslab metaslab_t; typedef struct metaslab_group metaslab_group_t; typedef struct metaslab_class metaslab_class_t; typedef struct zio zio_t; typedef struct zilog zilog_t; typedef struct spa_aux_vdev spa_aux_vdev_t; typedef struct ddt ddt_t; typedef struct ddt_entry ddt_entry_t; struct dsl_pool; struct dsl_dataset; /* * General-purpose 32-bit and 64-bit bitfield encodings. */ #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) #define BF32_GET(x, low, len) BF32_DECODE(x, low, len) #define BF64_GET(x, low, len) BF64_DECODE(x, low, len) #define BF32_SET(x, low, len, val) do { \ ASSERT3U(val, <, 1U << (len)); \ ASSERT3U(low + len, <=, 32); \ (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \ _NOTE(CONSTCOND) } while (0) #define BF64_SET(x, low, len, val) do { \ ASSERT3U(val, <, 1ULL << (len)); \ ASSERT3U(low + len, <=, 64); \ ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \ _NOTE(CONSTCOND) } while (0) #define BF32_GET_SB(x, low, len, shift, bias) \ ((BF32_GET(x, low, len) + (bias)) << (shift)) #define BF64_GET_SB(x, low, len, shift, bias) \ ((BF64_GET(x, low, len) + (bias)) << (shift)) #define BF32_SET_SB(x, low, len, shift, bias, val) do { \ ASSERT(IS_P2ALIGNED(val, 1U << shift)); \ ASSERT3S((val) >> (shift), >=, bias); \ BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \ _NOTE(CONSTCOND) } while (0) #define BF64_SET_SB(x, low, len, shift, bias, val) do { \ ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \ ASSERT3S((val) >> (shift), >=, bias); \ BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \ _NOTE(CONSTCOND) } while (0) /* * We currently support block sizes from 512 bytes to 16MB. * The benefits of larger blocks, and thus larger IO, need to be weighed * against the cost of COWing a giant block to modify one byte, and the * large latency of reading or writing a large block. * * Note that although blocks up to 16MB are supported, the recordsize * property can not be set larger than zfs_max_recordsize (default 1MB). * See the comment near zfs_max_recordsize in dsl_dataset.c for details. * * Note that although the LSIZE field of the blkptr_t can store sizes up * to 32MB, the dnode's dn_datablkszsec can only store sizes up to * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB. */ #define SPA_MINBLOCKSHIFT 9 #define SPA_OLD_MAXBLOCKSHIFT 17 #define SPA_MAXBLOCKSHIFT 24 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) #define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT) #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) /* * Default maximum supported logical ashift. * * The current 8k allocation block size limit is due to the 8k * aligned/sized operations performed by vdev_probe() on * vdev_label->vl_pad2. Using another "safe region" for these tests * would allow the limit to be raised to 16k, at the expense of * only having 8 available uberblocks in the label area. */ #define SPA_MAXASHIFT 13 /* * Default minimum supported logical ashift. */ #define SPA_MINASHIFT SPA_MINBLOCKSHIFT /* * Size of block to hold the configuration data (a packed nvlist) */ #define SPA_CONFIG_BLOCKSIZE (1ULL << 14) /* * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. * The ASIZE encoding should be at least 64 times larger (6 more bits) * to support up to 4-way RAID-Z mirror mode with worst-case gang block * overhead, three DVAs per bp, plus one more bit in case we do anything * else that expands the ASIZE. */ #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ /* * All SPA data is represented by 128-bit data virtual addresses (DVAs). * The members of the dva_t should be considered opaque outside the SPA. */ typedef struct dva { uint64_t dva_word[2]; } dva_t; /* * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. */ typedef struct zio_cksum { uint64_t zc_word[4]; } zio_cksum_t; + +/* + * Some checksums/hashes need a 256-bit initialization salt. This salt is kept + * secret and is suitable for use in MAC algorithms as the key. + */ +typedef struct zio_cksum_salt { + uint8_t zcs_bytes[32]; +} zio_cksum_salt_t; /* * Each block is described by its DVAs, time of birth, checksum, etc. * The word-by-word, bit-by-bit layout of the blkptr is as follows: * * 64 56 48 40 32 24 16 8 0 * +-------+-------+-------+-------+-------+-------+-------+-------+ * 0 | vdev1 | GRID | ASIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 1 |G| offset1 | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 2 | vdev2 | GRID | ASIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 3 |G| offset2 | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 4 | vdev3 | GRID | ASIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 5 |G| offset3 | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 7 | padding | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 8 | padding | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 9 | physical birth txg | * +-------+-------+-------+-------+-------+-------+-------+-------+ * a | logical birth txg | * +-------+-------+-------+-------+-------+-------+-------+-------+ * b | fill count | * +-------+-------+-------+-------+-------+-------+-------+-------+ * c | checksum[0] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * d | checksum[1] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * e | checksum[2] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * f | checksum[3] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * * Legend: * * vdev virtual device ID * offset offset into virtual device * LSIZE logical size * PSIZE physical size (after compression) * ASIZE allocated size (including RAID-Z parity and gang block headers) * GRID RAID-Z layout information (reserved for future use) * cksum checksum function * comp compression function * G gang block indicator * B byteorder (endianness) * D dedup * X encryption (on version 30, which is not supported) * E blkptr_t contains embedded data (see below) * lvl level of indirection * type DMU object type * phys birth txg of block allocation; zero if same as logical birth txg * log. birth transaction group in which the block was logically born * fill count number of non-zero blocks under this bp * checksum[4] 256-bit checksum of the data this bp describes */ /* * "Embedded" blkptr_t's don't actually point to a block, instead they * have a data payload embedded in the blkptr_t itself. See the comment * in blkptr.c for more details. * * The blkptr_t is laid out as follows: * * 64 56 48 40 32 24 16 8 0 * +-------+-------+-------+-------+-------+-------+-------+-------+ * 0 | payload | * 1 | payload | * 2 | payload | * 3 | payload | * 4 | payload | * 5 | payload | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 7 | payload | * 8 | payload | * 9 | payload | * +-------+-------+-------+-------+-------+-------+-------+-------+ * a | logical birth txg | * +-------+-------+-------+-------+-------+-------+-------+-------+ * b | payload | * c | payload | * d | payload | * e | payload | * f | payload | * +-------+-------+-------+-------+-------+-------+-------+-------+ * * Legend: * * payload contains the embedded data * B (byteorder) byteorder (endianness) * D (dedup) padding (set to zero) * X encryption (set to zero; see above) * E (embedded) set to one * lvl indirection level * type DMU object type * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) * comp compression function of payload * PSIZE size of payload after compression, in bytes * LSIZE logical size of payload, in bytes * note that 25 bits is enough to store the largest * "normal" BP's LSIZE (2^16 * 2^9) in bytes * log. birth transaction group in which the block was logically born * * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded * bp's they are stored in units of SPA_MINBLOCKSHIFT. * Generally, the generic BP_GET_*() macros can be used on embedded BP's. * The B, D, X, lvl, type, and comp fields are stored the same as with normal * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before * other macros, as they assert that they are only used on BP's of the correct * "embedded-ness". */ #define BPE_GET_ETYPE(bp) \ (ASSERT(BP_IS_EMBEDDED(bp)), \ BF64_GET((bp)->blk_prop, 40, 8)) #define BPE_SET_ETYPE(bp, t) do { \ ASSERT(BP_IS_EMBEDDED(bp)); \ BF64_SET((bp)->blk_prop, 40, 8, t); \ _NOTE(CONSTCOND) } while (0) #define BPE_GET_LSIZE(bp) \ (ASSERT(BP_IS_EMBEDDED(bp)), \ BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) #define BPE_SET_LSIZE(bp, x) do { \ ASSERT(BP_IS_EMBEDDED(bp)); \ BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ _NOTE(CONSTCOND) } while (0) #define BPE_GET_PSIZE(bp) \ (ASSERT(BP_IS_EMBEDDED(bp)), \ BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) #define BPE_SET_PSIZE(bp, x) do { \ ASSERT(BP_IS_EMBEDDED(bp)); \ BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ _NOTE(CONSTCOND) } while (0) typedef enum bp_embedded_type { BP_EMBEDDED_TYPE_DATA, BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED } bp_embedded_type_t; #define BPE_NUM_WORDS 14 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) #define BPE_IS_PAYLOADWORD(bp, wp) \ ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ /* * A block is a hole when it has either 1) never been written to, or * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads * without physically allocating disk space. Holes are represented in the * blkptr_t structure by zeroed blk_dva. Correct checking for holes is * done through the BP_IS_HOLE macro. For holes, the logical size, level, * DMU object type, and birth times are all also stored for holes that * were written to at some point (i.e. were punched after having been filled). */ typedef struct blkptr { dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ uint64_t blk_prop; /* size, compression, type, etc */ uint64_t blk_pad[2]; /* Extra space for the future */ uint64_t blk_phys_birth; /* txg when block was allocated */ uint64_t blk_birth; /* transaction group at birth */ uint64_t blk_fill; /* fill count */ zio_cksum_t blk_cksum; /* 256-bit checksum */ } blkptr_t; /* * Macros to get and set fields in a bp or DVA. */ #define DVA_GET_ASIZE(dva) \ BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) #define DVA_SET_ASIZE(dva, x) \ BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ SPA_MINBLOCKSHIFT, 0, x) #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) #define DVA_GET_OFFSET(dva) \ BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) #define DVA_SET_OFFSET(dva, x) \ BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) #define BP_GET_LSIZE(bp) \ (BP_IS_EMBEDDED(bp) ? \ (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) #define BP_SET_LSIZE(bp, x) do { \ ASSERT(!BP_IS_EMBEDDED(bp)); \ BF64_SET_SB((bp)->blk_prop, \ 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ _NOTE(CONSTCOND) } while (0) #define BP_GET_PSIZE(bp) \ (BP_IS_EMBEDDED(bp) ? 0 : \ BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) #define BP_SET_PSIZE(bp, x) do { \ ASSERT(!BP_IS_EMBEDDED(bp)); \ BF64_SET_SB((bp)->blk_prop, \ 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ _NOTE(CONSTCOND) } while (0) #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) #define BP_GET_CHECKSUM(bp) \ (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ BF64_GET((bp)->blk_prop, 40, 8)) #define BP_SET_CHECKSUM(bp, x) do { \ ASSERT(!BP_IS_EMBEDDED(bp)); \ BF64_SET((bp)->blk_prop, 40, 8, x); \ _NOTE(CONSTCOND) } while (0) #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) #define BP_PHYSICAL_BIRTH(bp) \ (BP_IS_EMBEDDED(bp) ? 0 : \ (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) #define BP_SET_BIRTH(bp, logical, physical) \ { \ ASSERT(!BP_IS_EMBEDDED(bp)); \ (bp)->blk_birth = (logical); \ (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ } #define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill) #define BP_GET_ASIZE(bp) \ (BP_IS_EMBEDDED(bp) ? 0 : \ DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ DVA_GET_ASIZE(&(bp)->blk_dva[2])) #define BP_GET_UCSIZE(bp) \ ((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \ BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) #define BP_GET_NDVAS(bp) \ (BP_IS_EMBEDDED(bp) ? 0 : \ !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) #define BP_COUNT_GANG(bp) \ (BP_IS_EMBEDDED(bp) ? 0 : \ (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ DVA_GET_GANG(&(bp)->blk_dva[1]) + \ DVA_GET_GANG(&(bp)->blk_dva[2]))) #define DVA_EQUAL(dva1, dva2) \ ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ (dva1)->dva_word[0] == (dva2)->dva_word[0]) #define BP_EQUAL(bp1, bp2) \ (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \ (bp1)->blk_birth == (bp2)->blk_birth && \ DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ ((zc1).zc_word[3] - (zc2).zc_word[3]))) #define ZIO_CHECKSUM_IS_ZERO(zc) \ (0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \ (zc)->zc_word[2] | (zc)->zc_word[3])) #define ZIO_CHECKSUM_BSWAP(zcp) \ { \ (zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]); \ (zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]); \ (zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]); \ (zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]); \ } #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ { \ (zcp)->zc_word[0] = w0; \ (zcp)->zc_word[1] = w1; \ (zcp)->zc_word[2] = w2; \ (zcp)->zc_word[3] = w3; \ } #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0]) #define BP_IS_GANG(bp) \ (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp))) #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ (dva)->dva_word[1] == 0ULL) #define BP_IS_HOLE(bp) \ (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp))) /* BP_IS_RAIDZ(bp) assumes no block compression */ #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ BP_GET_PSIZE(bp)) #define BP_ZERO(bp) \ { \ (bp)->blk_dva[0].dva_word[0] = 0; \ (bp)->blk_dva[0].dva_word[1] = 0; \ (bp)->blk_dva[1].dva_word[0] = 0; \ (bp)->blk_dva[1].dva_word[1] = 0; \ (bp)->blk_dva[2].dva_word[0] = 0; \ (bp)->blk_dva[2].dva_word[1] = 0; \ (bp)->blk_prop = 0; \ (bp)->blk_pad[0] = 0; \ (bp)->blk_pad[1] = 0; \ (bp)->blk_phys_birth = 0; \ (bp)->blk_birth = 0; \ (bp)->blk_fill = 0; \ ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ } #if BYTE_ORDER == _BIG_ENDIAN #define ZFS_HOST_BYTEORDER (0ULL) #else #define ZFS_HOST_BYTEORDER (1ULL) #endif #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) #define BP_SPRINTF_LEN 320 /* * This macro allows code sharing between zfs, libzpool, and mdb. * 'func' is either snprintf() or mdb_snprintf(). * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line. */ #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \ { \ static const char *copyname[] = \ { "zero", "single", "double", "triple" }; \ int len = 0; \ int copies = 0; \ \ if (bp == NULL) { \ len += func(buf + len, size - len, ""); \ } else if (BP_IS_HOLE(bp)) { \ len += func(buf + len, size - len, \ "HOLE [L%llu %s] " \ "size=%llxL birth=%lluL", \ (u_longlong_t)BP_GET_LEVEL(bp), \ type, \ (u_longlong_t)BP_GET_LSIZE(bp), \ (u_longlong_t)bp->blk_birth); \ } else if (BP_IS_EMBEDDED(bp)) { \ len = func(buf + len, size - len, \ "EMBEDDED [L%llu %s] et=%u %s " \ "size=%llxL/%llxP birth=%lluL", \ (u_longlong_t)BP_GET_LEVEL(bp), \ type, \ (int)BPE_GET_ETYPE(bp), \ compress, \ (u_longlong_t)BPE_GET_LSIZE(bp), \ (u_longlong_t)BPE_GET_PSIZE(bp), \ (u_longlong_t)bp->blk_birth); \ } else { \ for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ const dva_t *dva = &bp->blk_dva[d]; \ if (DVA_IS_VALID(dva)) \ copies++; \ len += func(buf + len, size - len, \ "DVA[%d]=<%llu:%llx:%llx>%c", d, \ (u_longlong_t)DVA_GET_VDEV(dva), \ (u_longlong_t)DVA_GET_OFFSET(dva), \ (u_longlong_t)DVA_GET_ASIZE(dva), \ ws); \ } \ if (BP_IS_GANG(bp) && \ DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ copies--; \ len += func(buf + len, size - len, \ "[L%llu %s] %s %s %s %s %s %s%c" \ "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ "cksum=%llx:%llx:%llx:%llx", \ (u_longlong_t)BP_GET_LEVEL(bp), \ type, \ checksum, \ compress, \ BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ BP_IS_GANG(bp) ? "gang" : "contiguous", \ BP_GET_DEDUP(bp) ? "dedup" : "unique", \ copyname[copies], \ ws, \ (u_longlong_t)BP_GET_LSIZE(bp), \ (u_longlong_t)BP_GET_PSIZE(bp), \ (u_longlong_t)bp->blk_birth, \ (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \ (u_longlong_t)BP_GET_FILL(bp), \ ws, \ (u_longlong_t)bp->blk_cksum.zc_word[0], \ (u_longlong_t)bp->blk_cksum.zc_word[1], \ (u_longlong_t)bp->blk_cksum.zc_word[2], \ (u_longlong_t)bp->blk_cksum.zc_word[3]); \ } \ ASSERT(len < size); \ } #include #define BP_GET_BUFC_TYPE(bp) \ (((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \ ARC_BUFC_METADATA : ARC_BUFC_DATA) typedef enum spa_import_type { SPA_IMPORT_EXISTING, SPA_IMPORT_ASSEMBLE } spa_import_type_t; /* state manipulation functions */ extern int spa_open(const char *pool, spa_t **, void *tag); extern int spa_open_rewind(const char *pool, spa_t **, void *tag, nvlist_t *policy, nvlist_t **config); extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot, size_t buflen); extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props, nvlist_t *zplprops); #if defined(sun) extern int spa_import_rootpool(char *devpath, char *devid); #else extern int spa_import_rootpool(const char *name); #endif extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags); extern nvlist_t *spa_tryimport(nvlist_t *tryconfig); extern int spa_destroy(char *pool); extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, boolean_t hardforce); extern int spa_reset(char *pool); extern void spa_async_request(spa_t *spa, int flag); extern void spa_async_unrequest(spa_t *spa, int flag); extern void spa_async_suspend(spa_t *spa); extern void spa_async_resume(spa_t *spa); extern spa_t *spa_inject_addref(char *pool); extern void spa_inject_delref(spa_t *spa); extern void spa_scan_stat_init(spa_t *spa); extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps); #define SPA_ASYNC_CONFIG_UPDATE 0x01 #define SPA_ASYNC_REMOVE 0x02 #define SPA_ASYNC_PROBE 0x04 #define SPA_ASYNC_RESILVER_DONE 0x08 #define SPA_ASYNC_RESILVER 0x10 #define SPA_ASYNC_AUTOEXPAND 0x20 #define SPA_ASYNC_REMOVE_DONE 0x40 #define SPA_ASYNC_REMOVE_STOP 0x80 /* * Controls the behavior of spa_vdev_remove(). */ #define SPA_REMOVE_UNSPARE 0x01 #define SPA_REMOVE_DONE 0x02 /* device manipulation */ extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot); extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing); extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done); extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare); extern boolean_t spa_vdev_remove_active(spa_t *spa); extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, nvlist_t *props, boolean_t exp); /* spare state (which is global across all pools) */ extern void spa_spare_add(vdev_t *vd); extern void spa_spare_remove(vdev_t *vd); extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); extern void spa_spare_activate(vdev_t *vd); /* L2ARC state (which is global across all pools) */ extern void spa_l2cache_add(vdev_t *vd); extern void spa_l2cache_remove(vdev_t *vd); extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); extern void spa_l2cache_activate(vdev_t *vd); extern void spa_l2cache_drop(spa_t *spa); /* scanning */ extern int spa_scan(spa_t *spa, pool_scan_func_t func); extern int spa_scan_stop(spa_t *spa); /* spa syncing */ extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ extern void spa_sync_allpools(void); /* spa namespace global mutex */ extern kmutex_t spa_namespace_lock; /* * SPA configuration functions in spa_config.c */ #define SPA_CONFIG_UPDATE_POOL 0 #define SPA_CONFIG_UPDATE_VDEVS 1 extern void spa_config_sync(spa_t *, boolean_t, boolean_t); extern void spa_config_load(void); extern nvlist_t *spa_all_configs(uint64_t *); extern void spa_config_set(spa_t *spa, nvlist_t *config); extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats); extern void spa_config_update(spa_t *spa, int what); /* * Miscellaneous SPA routines in spa_misc.c */ /* Namespace manipulation */ extern spa_t *spa_lookup(const char *name); extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); extern void spa_remove(spa_t *spa); extern spa_t *spa_next(spa_t *prev); /* Refcount functions */ extern void spa_open_ref(spa_t *spa, void *tag); extern void spa_close(spa_t *spa, void *tag); extern void spa_async_close(spa_t *spa, void *tag); extern boolean_t spa_refcount_zero(spa_t *spa); #define SCL_NONE 0x00 #define SCL_CONFIG 0x01 #define SCL_STATE 0x02 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */ #define SCL_ALLOC 0x08 #define SCL_ZIO 0x10 #define SCL_FREE 0x20 #define SCL_VDEV 0x40 #define SCL_LOCKS 7 #define SCL_ALL ((1 << SCL_LOCKS) - 1) #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO) /* Pool configuration locks */ extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw); extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw); extern void spa_config_exit(spa_t *spa, int locks, void *tag); extern int spa_config_held(spa_t *spa, int locks, krw_t rw); /* Pool vdev add/remove lock */ extern uint64_t spa_vdev_enter(spa_t *spa); extern uint64_t spa_vdev_config_enter(spa_t *spa); extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag); extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); /* Pool vdev state change lock */ extern void spa_vdev_state_enter(spa_t *spa, int oplock); extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); /* Log state */ typedef enum spa_log_state { SPA_LOG_UNKNOWN = 0, /* unknown log state */ SPA_LOG_MISSING, /* missing log(s) */ SPA_LOG_CLEAR, /* clear the log(s) */ SPA_LOG_GOOD, /* log(s) are good */ } spa_log_state_t; extern spa_log_state_t spa_get_log_state(spa_t *spa); extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); extern int spa_offline_log(spa_t *spa); /* Log claim callback */ extern void spa_claim_notify(zio_t *zio); /* Accessor functions */ extern boolean_t spa_shutting_down(spa_t *spa); extern struct dsl_pool *spa_get_dsl(spa_t *spa); extern boolean_t spa_is_initializing(spa_t *spa); extern blkptr_t *spa_get_rootblkptr(spa_t *spa); extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); extern void spa_altroot(spa_t *, char *, size_t); extern int spa_sync_pass(spa_t *spa); extern char *spa_name(spa_t *spa); extern uint64_t spa_guid(spa_t *spa); extern uint64_t spa_load_guid(spa_t *spa); extern uint64_t spa_last_synced_txg(spa_t *spa); extern uint64_t spa_first_txg(spa_t *spa); extern uint64_t spa_syncing_txg(spa_t *spa); extern uint64_t spa_version(spa_t *spa); extern pool_state_t spa_state(spa_t *spa); extern spa_load_state_t spa_load_state(spa_t *spa); extern uint64_t spa_freeze_txg(spa_t *spa); extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize); extern uint64_t spa_get_dspace(spa_t *spa); extern uint64_t spa_get_slop_space(spa_t *spa); extern void spa_update_dspace(spa_t *spa); extern uint64_t spa_version(spa_t *spa); extern boolean_t spa_deflate(spa_t *spa); extern metaslab_class_t *spa_normal_class(spa_t *spa); extern metaslab_class_t *spa_log_class(spa_t *spa); extern void spa_evicting_os_register(spa_t *, objset_t *os); extern void spa_evicting_os_deregister(spa_t *, objset_t *os); extern void spa_evicting_os_wait(spa_t *spa); extern int spa_max_replication(spa_t *spa); extern int spa_prev_software_version(spa_t *spa); extern int spa_busy(void); extern uint8_t spa_get_failmode(spa_t *spa); extern boolean_t spa_suspended(spa_t *spa); extern uint64_t spa_bootfs(spa_t *spa); extern uint64_t spa_delegation(spa_t *spa); extern objset_t *spa_meta_objset(spa_t *spa); extern uint64_t spa_deadman_synctime(spa_t *spa); /* Miscellaneous support routines */ extern void spa_activate_mos_feature(spa_t *spa, const char *feature, dmu_tx_t *tx); extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); extern int spa_rename(const char *oldname, const char *newname); extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); extern char *spa_strdup(const char *); extern void spa_strfree(char *); extern uint64_t spa_get_random(uint64_t range); extern uint64_t spa_generate_guid(spa_t *spa); extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); extern void spa_freeze(spa_t *spa); extern int spa_change_guid(spa_t *spa); extern void spa_upgrade(spa_t *spa, uint64_t version); extern void spa_evict_all(void); extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, boolean_t l2cache); extern boolean_t spa_has_spare(spa_t *, uint64_t guid); extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); extern boolean_t spa_has_slogs(spa_t *spa); extern boolean_t spa_is_root(spa_t *spa); extern boolean_t spa_writeable(spa_t *spa); extern boolean_t spa_has_pending_synctask(spa_t *spa); extern int spa_maxblocksize(spa_t *spa); extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp); extern int spa_mode(spa_t *spa); extern uint64_t zfs_strtonum(const char *str, char **nptr); #define strtonum(str, nptr) zfs_strtonum((str), (nptr)) extern char *spa_his_ievent_table[]; extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, char *his_buf); extern int spa_history_log(spa_t *spa, const char *his_buf); extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); extern void spa_history_log_version(spa_t *spa, const char *operation); extern void spa_history_log_internal(spa_t *spa, const char *operation, dmu_tx_t *tx, const char *fmt, ...); extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, dmu_tx_t *tx, const char *fmt, ...); extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, dmu_tx_t *tx, const char *fmt, ...); /* error handling */ struct zbookmark_phys; extern void spa_log_error(spa_t *spa, zio_t *zio); extern void zfs_ereport_post(const char *cls, spa_t *spa, vdev_t *vd, zio_t *zio, uint64_t stateoroffset, uint64_t length); extern void zfs_post_remove(spa_t *spa, vdev_t *vd); extern void zfs_post_state_change(spa_t *spa, vdev_t *vd); extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); extern uint64_t spa_get_errlog_size(spa_t *spa); extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count); extern void spa_errlog_rotate(spa_t *spa); extern void spa_errlog_drain(spa_t *spa); extern void spa_errlog_sync(spa_t *spa, uint64_t txg); extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); /* vdev cache */ extern void vdev_cache_stat_init(void); extern void vdev_cache_stat_fini(void); /* Initialization and termination */ extern void spa_init(int flags); extern void spa_fini(void); extern void spa_boot_init(); /* properties */ extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); extern int spa_prop_get(spa_t *spa, nvlist_t **nvp); extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); /* asynchronous event notification */ extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name); #ifdef ZFS_DEBUG #define dprintf_bp(bp, fmt, ...) do { \ if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ kmem_free(__blkbuf, BP_SPRINTF_LEN); \ } \ _NOTE(CONSTCOND) } while (0) #else #define dprintf_bp(bp, fmt, ...) #endif extern boolean_t spa_debug_enabled(spa_t *spa); #define spa_dbgmsg(spa, ...) \ { \ if (spa_debug_enabled(spa)) \ zfs_dbgmsg(__VA_ARGS__); \ } extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */ #ifdef __cplusplus } #endif #endif /* _SYS_SPA_H */ Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/spa_impl.h =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/spa_impl.h (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/spa_impl.h (revision 290757) @@ -1,298 +1,303 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2015 by Delphix. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright 2013 Martin Matuska . All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. + * Copyright 2013 Saso Kiselkov. All rights reserved. */ #ifndef _SYS_SPA_IMPL_H #define _SYS_SPA_IMPL_H #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif typedef struct spa_error_entry { zbookmark_phys_t se_bookmark; char *se_name; avl_node_t se_avl; } spa_error_entry_t; typedef struct spa_history_phys { uint64_t sh_pool_create_len; /* ending offset of zpool create */ uint64_t sh_phys_max_off; /* physical EOF */ uint64_t sh_bof; /* logical BOF */ uint64_t sh_eof; /* logical EOF */ uint64_t sh_records_lost; /* num of records overwritten */ } spa_history_phys_t; struct spa_aux_vdev { uint64_t sav_object; /* MOS object for device list */ nvlist_t *sav_config; /* cached device config */ vdev_t **sav_vdevs; /* devices */ int sav_count; /* number devices */ boolean_t sav_sync; /* sync the device list */ nvlist_t **sav_pending; /* pending device additions */ uint_t sav_npending; /* # pending devices */ }; typedef struct spa_config_lock { kmutex_t scl_lock; kthread_t *scl_writer; int scl_write_wanted; kcondvar_t scl_cv; refcount_t scl_count; } spa_config_lock_t; typedef struct spa_config_dirent { list_node_t scd_link; char *scd_path; } spa_config_dirent_t; typedef enum zio_taskq_type { ZIO_TASKQ_ISSUE = 0, ZIO_TASKQ_ISSUE_HIGH, ZIO_TASKQ_INTERRUPT, ZIO_TASKQ_INTERRUPT_HIGH, ZIO_TASKQ_TYPES } zio_taskq_type_t; /* * State machine for the zpool-poolname process. The states transitions * are done as follows: * * From To Routine * PROC_NONE -> PROC_CREATED spa_activate() * PROC_CREATED -> PROC_ACTIVE spa_thread() * PROC_ACTIVE -> PROC_DEACTIVATE spa_deactivate() * PROC_DEACTIVATE -> PROC_GONE spa_thread() * PROC_GONE -> PROC_NONE spa_deactivate() */ typedef enum spa_proc_state { SPA_PROC_NONE, /* spa_proc = &p0, no process created */ SPA_PROC_CREATED, /* spa_activate() has proc, is waiting */ SPA_PROC_ACTIVE, /* taskqs created, spa_proc set */ SPA_PROC_DEACTIVATE, /* spa_deactivate() requests process exit */ SPA_PROC_GONE /* spa_thread() is exiting, spa_proc = &p0 */ } spa_proc_state_t; typedef struct spa_taskqs { uint_t stqs_count; taskq_t **stqs_taskq; } spa_taskqs_t; struct spa { /* * Fields protected by spa_namespace_lock. */ char spa_name[MAXNAMELEN]; /* pool name */ char *spa_comment; /* comment */ avl_node_t spa_avl; /* node in spa_namespace_avl */ nvlist_t *spa_config; /* last synced config */ nvlist_t *spa_config_syncing; /* currently syncing config */ nvlist_t *spa_config_splitting; /* config for splitting */ nvlist_t *spa_load_info; /* info and errors from load */ uint64_t spa_config_txg; /* txg of last config change */ int spa_sync_pass; /* iterate-to-convergence */ pool_state_t spa_state; /* pool state */ int spa_inject_ref; /* injection references */ uint8_t spa_sync_on; /* sync threads are running */ spa_load_state_t spa_load_state; /* current load operation */ uint64_t spa_import_flags; /* import specific flags */ spa_taskqs_t spa_zio_taskq[ZIO_TYPES][ZIO_TASKQ_TYPES]; dsl_pool_t *spa_dsl_pool; boolean_t spa_is_initializing; /* true while opening pool */ metaslab_class_t *spa_normal_class; /* normal data class */ metaslab_class_t *spa_log_class; /* intent log data class */ uint64_t spa_first_txg; /* first txg after spa_open() */ uint64_t spa_final_txg; /* txg of export/destroy */ uint64_t spa_freeze_txg; /* freeze pool at this txg */ uint64_t spa_load_max_txg; /* best initial ub_txg */ uint64_t spa_claim_max_txg; /* highest claimed birth txg */ timespec_t spa_loaded_ts; /* 1st successful open time */ objset_t *spa_meta_objset; /* copy of dp->dp_meta_objset */ kmutex_t spa_evicting_os_lock; /* Evicting objset list lock */ list_t spa_evicting_os_list; /* Objsets being evicted. */ kcondvar_t spa_evicting_os_cv; /* Objset Eviction Completion */ txg_list_t spa_vdev_txg_list; /* per-txg dirty vdev list */ vdev_t *spa_root_vdev; /* top-level vdev container */ int spa_min_ashift; /* of vdevs in normal class */ int spa_max_ashift; /* of vdevs in normal class */ uint64_t spa_config_guid; /* config pool guid */ uint64_t spa_load_guid; /* spa_load initialized guid */ uint64_t spa_last_synced_guid; /* last synced guid */ list_t spa_config_dirty_list; /* vdevs with dirty config */ list_t spa_state_dirty_list; /* vdevs with dirty state */ spa_aux_vdev_t spa_spares; /* hot spares */ spa_aux_vdev_t spa_l2cache; /* L2ARC cache devices */ nvlist_t *spa_label_features; /* Features for reading MOS */ uint64_t spa_config_object; /* MOS object for pool config */ uint64_t spa_config_generation; /* config generation number */ uint64_t spa_syncing_txg; /* txg currently syncing */ bpobj_t spa_deferred_bpobj; /* deferred-free bplist */ bplist_t spa_free_bplist[TXG_SIZE]; /* bplist of stuff to free */ + zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */ + /* checksum context templates */ + kmutex_t spa_cksum_tmpls_lock; + void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS]; uberblock_t spa_ubsync; /* last synced uberblock */ uberblock_t spa_uberblock; /* current uberblock */ boolean_t spa_extreme_rewind; /* rewind past deferred frees */ uint64_t spa_last_io; /* lbolt of last non-scan I/O */ kmutex_t spa_scrub_lock; /* resilver/scrub lock */ uint64_t spa_scrub_inflight; /* in-flight scrub I/Os */ kcondvar_t spa_scrub_io_cv; /* scrub I/O completion */ uint8_t spa_scrub_active; /* active or suspended? */ uint8_t spa_scrub_type; /* type of scrub we're doing */ uint8_t spa_scrub_finished; /* indicator to rotate logs */ uint8_t spa_scrub_started; /* started since last boot */ uint8_t spa_scrub_reopen; /* scrub doing vdev_reopen */ uint64_t spa_scan_pass_start; /* start time per pass/reboot */ uint64_t spa_scan_pass_exam; /* examined bytes per pass */ kmutex_t spa_async_lock; /* protect async state */ kthread_t *spa_async_thread; /* thread doing async task */ kthread_t *spa_async_thread_vd; /* thread doing vd async task */ int spa_async_suspended; /* async tasks suspended */ kcondvar_t spa_async_cv; /* wait for thread_exit() */ uint16_t spa_async_tasks; /* async task mask */ char *spa_root; /* alternate root directory */ uint64_t spa_ena; /* spa-wide ereport ENA */ int spa_last_open_failed; /* error if last open failed */ uint64_t spa_last_ubsync_txg; /* "best" uberblock txg */ uint64_t spa_last_ubsync_txg_ts; /* timestamp from that ub */ uint64_t spa_load_txg; /* ub txg that loaded */ uint64_t spa_load_txg_ts; /* timestamp from that ub */ uint64_t spa_load_meta_errors; /* verify metadata err count */ uint64_t spa_load_data_errors; /* verify data err count */ uint64_t spa_verify_min_txg; /* start txg of verify scrub */ kmutex_t spa_errlog_lock; /* error log lock */ uint64_t spa_errlog_last; /* last error log object */ uint64_t spa_errlog_scrub; /* scrub error log object */ kmutex_t spa_errlist_lock; /* error list/ereport lock */ avl_tree_t spa_errlist_last; /* last error list */ avl_tree_t spa_errlist_scrub; /* scrub error list */ uint64_t spa_deflate; /* should we deflate? */ uint64_t spa_history; /* history object */ kmutex_t spa_history_lock; /* history lock */ vdev_t *spa_pending_vdev; /* pending vdev additions */ kmutex_t spa_props_lock; /* property lock */ uint64_t spa_pool_props_object; /* object for properties */ uint64_t spa_bootfs; /* default boot filesystem */ uint64_t spa_failmode; /* failure mode for the pool */ uint64_t spa_delegation; /* delegation on/off */ list_t spa_config_list; /* previous cache file(s) */ /* per-CPU array of root of async I/O: */ zio_t **spa_async_zio_root; zio_t *spa_suspend_zio_root; /* root of all suspended I/O */ kmutex_t spa_suspend_lock; /* protects suspend_zio_root */ kcondvar_t spa_suspend_cv; /* notification of resume */ uint8_t spa_suspended; /* pool is suspended */ uint8_t spa_claiming; /* pool is doing zil_claim() */ boolean_t spa_debug; /* debug enabled? */ boolean_t spa_is_root; /* pool is root */ int spa_minref; /* num refs when first opened */ int spa_mode; /* FREAD | FWRITE */ spa_log_state_t spa_log_state; /* log state */ uint64_t spa_autoexpand; /* lun expansion on/off */ ddt_t *spa_ddt[ZIO_CHECKSUM_FUNCTIONS]; /* in-core DDTs */ uint64_t spa_ddt_stat_object; /* DDT statistics */ uint64_t spa_dedup_ditto; /* dedup ditto threshold */ uint64_t spa_dedup_checksum; /* default dedup checksum */ uint64_t spa_dspace; /* dspace in normal class */ kmutex_t spa_vdev_top_lock; /* dueling offline/remove */ kmutex_t spa_proc_lock; /* protects spa_proc* */ kcondvar_t spa_proc_cv; /* spa_proc_state transitions */ spa_proc_state_t spa_proc_state; /* see definition */ struct proc *spa_proc; /* "zpool-poolname" process */ uint64_t spa_did; /* if procp != p0, did of t1 */ kthread_t *spa_trim_thread; /* thread sending TRIM I/Os */ kmutex_t spa_trim_lock; /* protects spa_trim_cv */ kcondvar_t spa_trim_cv; /* used to notify TRIM thread */ boolean_t spa_autoreplace; /* autoreplace set in open */ int spa_vdev_locks; /* locks grabbed */ uint64_t spa_creation_version; /* version at pool creation */ uint64_t spa_prev_software_version; /* See ub_software_version */ uint64_t spa_feat_for_write_obj; /* required to write to pool */ uint64_t spa_feat_for_read_obj; /* required to read from pool */ uint64_t spa_feat_desc_obj; /* Feature descriptions */ uint64_t spa_feat_enabled_txg_obj; /* Feature enabled txg */ /* cache feature refcounts */ uint64_t spa_feat_refcount_cache[SPA_FEATURES]; #ifdef illumos cyclic_id_t spa_deadman_cycid; /* cyclic id */ #else /* FreeBSD */ #ifdef _KERNEL struct callout spa_deadman_cycid; /* callout id */ #endif #endif /* illumos */ uint64_t spa_deadman_calls; /* number of deadman calls */ hrtime_t spa_sync_starttime; /* starting time fo spa_sync */ uint64_t spa_deadman_synctime; /* deadman expiration timer */ #ifdef illumos /* * spa_iokstat_lock protects spa_iokstat and * spa_queue_stats[]. */ kmutex_t spa_iokstat_lock; struct kstat *spa_iokstat; /* kstat of io to this pool */ struct { int spa_active; int spa_queued; } spa_queue_stats[ZIO_PRIORITY_NUM_QUEUEABLE]; #endif hrtime_t spa_ccw_fail_time; /* Conf cache write fail time */ /* * spa_refcount & spa_config_lock must be the last elements * because refcount_t changes size based on compilation options. * In order for the MDB module to function correctly, the other * fields must remain in the same location. */ spa_config_lock_t spa_config_lock[SCL_LOCKS]; /* config changes */ refcount_t spa_refcount; /* number of opens */ #ifndef sun boolean_t spa_splitting_newspa; /* creating new spa in split */ #endif }; extern const char *spa_config_path; extern void spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q, task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent); #ifdef __cplusplus } #endif #endif /* _SYS_SPA_IMPL_H */ Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zio.h =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zio.h (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zio.h (revision 290757) @@ -1,641 +1,646 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2012, 2014 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. */ #ifndef _ZIO_H #define _ZIO_H #include #include #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif /* * Embedded checksum */ #define ZEC_MAGIC 0x210da7ab10c7a11ULL typedef struct zio_eck { uint64_t zec_magic; /* for validation, endianness */ zio_cksum_t zec_cksum; /* 256-bit checksum */ } zio_eck_t; /* * Gang block headers are self-checksumming and contain an array * of block pointers. */ #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \ sizeof (zio_eck_t)) / sizeof (blkptr_t)) #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \ sizeof (zio_eck_t) - \ (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\ sizeof (uint64_t)) typedef struct zio_gbh { blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS]; uint64_t zg_filler[SPA_GBH_FILLER]; zio_eck_t zg_tail; } zio_gbh_phys_t; enum zio_checksum { ZIO_CHECKSUM_INHERIT = 0, ZIO_CHECKSUM_ON, ZIO_CHECKSUM_OFF, ZIO_CHECKSUM_LABEL, ZIO_CHECKSUM_GANG_HEADER, ZIO_CHECKSUM_ZILOG, ZIO_CHECKSUM_FLETCHER_2, ZIO_CHECKSUM_FLETCHER_4, ZIO_CHECKSUM_SHA256, ZIO_CHECKSUM_ZILOG2, ZIO_CHECKSUM_NOPARITY, +#ifdef illumos + ZIO_CHECKSUM_SHA512, + ZIO_CHECKSUM_SKEIN, + ZIO_CHECKSUM_EDONR, +#endif ZIO_CHECKSUM_FUNCTIONS }; /* * The number of "legacy" compression functions which can be set on individual * objects. */ #define ZIO_CHECKSUM_LEGACY_FUNCTIONS ZIO_CHECKSUM_ZILOG2 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON #define ZIO_CHECKSUM_MASK 0xffULL #define ZIO_CHECKSUM_VERIFY (1 << 8) #define ZIO_DEDUPCHECKSUM ZIO_CHECKSUM_SHA256 #define ZIO_DEDUPDITTO_MIN 100 enum zio_compress { ZIO_COMPRESS_INHERIT = 0, ZIO_COMPRESS_ON, ZIO_COMPRESS_OFF, ZIO_COMPRESS_LZJB, ZIO_COMPRESS_EMPTY, ZIO_COMPRESS_GZIP_1, ZIO_COMPRESS_GZIP_2, ZIO_COMPRESS_GZIP_3, ZIO_COMPRESS_GZIP_4, ZIO_COMPRESS_GZIP_5, ZIO_COMPRESS_GZIP_6, ZIO_COMPRESS_GZIP_7, ZIO_COMPRESS_GZIP_8, ZIO_COMPRESS_GZIP_9, ZIO_COMPRESS_ZLE, ZIO_COMPRESS_LZ4, ZIO_COMPRESS_FUNCTIONS }; /* * The number of "legacy" compression functions which can be set on individual * objects. */ #define ZIO_COMPRESS_LEGACY_FUNCTIONS ZIO_COMPRESS_LZ4 /* * The meaning of "compress = on" selected by the compression features enabled * on a given pool. */ #define ZIO_COMPRESS_LEGACY_ON_VALUE ZIO_COMPRESS_LZJB #define ZIO_COMPRESS_LZ4_ON_VALUE ZIO_COMPRESS_LZ4 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF #define BOOTFS_COMPRESS_VALID(compress) \ ((compress) == ZIO_COMPRESS_LZJB || \ (compress) == ZIO_COMPRESS_LZ4 || \ (compress) == ZIO_COMPRESS_ON || \ (compress) == ZIO_COMPRESS_OFF) #define ZIO_FAILURE_MODE_WAIT 0 #define ZIO_FAILURE_MODE_CONTINUE 1 #define ZIO_FAILURE_MODE_PANIC 2 #define ZIO_PIPELINE_CONTINUE 0x100 #define ZIO_PIPELINE_STOP 0x101 enum zio_flag { /* * Flags inherited by gang, ddt, and vdev children, * and that must be equal for two zios to aggregate */ ZIO_FLAG_DONT_AGGREGATE = 1 << 0, ZIO_FLAG_IO_REPAIR = 1 << 1, ZIO_FLAG_SELF_HEAL = 1 << 2, ZIO_FLAG_RESILVER = 1 << 3, ZIO_FLAG_SCRUB = 1 << 4, ZIO_FLAG_SCAN_THREAD = 1 << 5, ZIO_FLAG_PHYSICAL = 1 << 6, #define ZIO_FLAG_AGG_INHERIT (ZIO_FLAG_CANFAIL - 1) /* * Flags inherited by ddt, gang, and vdev children. */ ZIO_FLAG_CANFAIL = 1 << 7, /* must be first for INHERIT */ ZIO_FLAG_SPECULATIVE = 1 << 8, ZIO_FLAG_CONFIG_WRITER = 1 << 9, ZIO_FLAG_DONT_RETRY = 1 << 10, ZIO_FLAG_DONT_CACHE = 1 << 11, ZIO_FLAG_NODATA = 1 << 12, ZIO_FLAG_INDUCE_DAMAGE = 1 << 13, #define ZIO_FLAG_DDT_INHERIT (ZIO_FLAG_IO_RETRY - 1) #define ZIO_FLAG_GANG_INHERIT (ZIO_FLAG_IO_RETRY - 1) /* * Flags inherited by vdev children. */ ZIO_FLAG_IO_RETRY = 1 << 14, /* must be first for INHERIT */ ZIO_FLAG_PROBE = 1 << 15, ZIO_FLAG_TRYHARD = 1 << 16, ZIO_FLAG_OPTIONAL = 1 << 17, #define ZIO_FLAG_VDEV_INHERIT (ZIO_FLAG_DONT_QUEUE - 1) /* * Flags not inherited by any children. */ ZIO_FLAG_DONT_QUEUE = 1 << 18, /* must be first for INHERIT */ ZIO_FLAG_DONT_PROPAGATE = 1 << 19, ZIO_FLAG_IO_BYPASS = 1 << 20, ZIO_FLAG_IO_REWRITE = 1 << 21, ZIO_FLAG_RAW = 1 << 22, ZIO_FLAG_GANG_CHILD = 1 << 23, ZIO_FLAG_DDT_CHILD = 1 << 24, ZIO_FLAG_GODFATHER = 1 << 25, ZIO_FLAG_NOPWRITE = 1 << 26, ZIO_FLAG_REEXECUTED = 1 << 27, ZIO_FLAG_DELEGATED = 1 << 28, }; #define ZIO_FLAG_MUSTSUCCEED 0 #define ZIO_DDT_CHILD_FLAGS(zio) \ (((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) | \ ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL) #define ZIO_GANG_CHILD_FLAGS(zio) \ (((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) | \ ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL) #define ZIO_VDEV_CHILD_FLAGS(zio) \ (((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) | \ ZIO_FLAG_CANFAIL) enum zio_child { ZIO_CHILD_VDEV = 0, ZIO_CHILD_GANG, ZIO_CHILD_DDT, ZIO_CHILD_LOGICAL, ZIO_CHILD_TYPES }; enum zio_wait_type { ZIO_WAIT_READY = 0, ZIO_WAIT_DONE, ZIO_WAIT_TYPES }; /* * We'll take the number 122 and 123 to indicate checksum errors and * fragmentation. Those doesn't collide with any errno values as they * are greater than ELAST. */ #define ECKSUM 122 #define EFRAGS 123 typedef void zio_done_func_t(zio_t *zio); extern const char *zio_type_name[ZIO_TYPES]; /* * A bookmark is a four-tuple that uniquely * identifies any block in the pool. By convention, the meta-objset (MOS) * is objset 0, and the meta-dnode is object 0. This covers all blocks * except root blocks and ZIL blocks, which are defined as follows: * * Root blocks (objset_phys_t) are object 0, level -1: . * ZIL blocks are bookmarked . * dmu_sync()ed ZIL data blocks are bookmarked . * dnode visit bookmarks are . * * Note: this structure is called a bookmark because its original purpose * was to remember where to resume a pool-wide traverse. * * Note: this structure is passed between userland and the kernel, and is * stored on disk (by virtue of being incorporated into other on-disk * structures, e.g. dsl_scan_phys_t). */ typedef struct zbookmark_phys { uint64_t zb_objset; uint64_t zb_object; int64_t zb_level; uint64_t zb_blkid; } zbookmark_phys_t; #define SET_BOOKMARK(zb, objset, object, level, blkid) \ { \ (zb)->zb_objset = objset; \ (zb)->zb_object = object; \ (zb)->zb_level = level; \ (zb)->zb_blkid = blkid; \ } #define ZB_DESTROYED_OBJSET (-1ULL) #define ZB_ROOT_OBJECT (0ULL) #define ZB_ROOT_LEVEL (-1LL) #define ZB_ROOT_BLKID (0ULL) #define ZB_ZIL_OBJECT (0ULL) #define ZB_ZIL_LEVEL (-2LL) #define ZB_DNODE_LEVEL (-3LL) #define ZB_DNODE_BLKID (0ULL) #define ZB_IS_ZERO(zb) \ ((zb)->zb_objset == 0 && (zb)->zb_object == 0 && \ (zb)->zb_level == 0 && (zb)->zb_blkid == 0) #define ZB_IS_ROOT(zb) \ ((zb)->zb_object == ZB_ROOT_OBJECT && \ (zb)->zb_level == ZB_ROOT_LEVEL && \ (zb)->zb_blkid == ZB_ROOT_BLKID) typedef struct zio_prop { enum zio_checksum zp_checksum; enum zio_compress zp_compress; dmu_object_type_t zp_type; uint8_t zp_level; uint8_t zp_copies; boolean_t zp_dedup; boolean_t zp_dedup_verify; boolean_t zp_nopwrite; } zio_prop_t; typedef struct zio_cksum_report zio_cksum_report_t; typedef void zio_cksum_finish_f(zio_cksum_report_t *rep, const void *good_data); typedef void zio_cksum_free_f(void *cbdata, size_t size); struct zio_bad_cksum; /* defined in zio_checksum.h */ struct dnode_phys; struct zio_cksum_report { struct zio_cksum_report *zcr_next; nvlist_t *zcr_ereport; nvlist_t *zcr_detector; void *zcr_cbdata; size_t zcr_cbinfo; /* passed to zcr_free() */ uint64_t zcr_align; uint64_t zcr_length; zio_cksum_finish_f *zcr_finish; zio_cksum_free_f *zcr_free; /* internal use only */ struct zio_bad_cksum *zcr_ckinfo; /* information from failure */ }; typedef void zio_vsd_cksum_report_f(zio_t *zio, zio_cksum_report_t *zcr, void *arg); zio_vsd_cksum_report_f zio_vsd_default_cksum_report; typedef struct zio_vsd_ops { zio_done_func_t *vsd_free; zio_vsd_cksum_report_f *vsd_cksum_report; } zio_vsd_ops_t; typedef struct zio_gang_node { zio_gbh_phys_t *gn_gbh; struct zio_gang_node *gn_child[SPA_GBH_NBLKPTRS]; } zio_gang_node_t; typedef zio_t *zio_gang_issue_func_t(zio_t *zio, blkptr_t *bp, zio_gang_node_t *gn, void *data); typedef void zio_transform_func_t(zio_t *zio, void *data, uint64_t size); typedef struct zio_transform { void *zt_orig_data; uint64_t zt_orig_size; uint64_t zt_bufsize; zio_transform_func_t *zt_transform; struct zio_transform *zt_next; } zio_transform_t; typedef int zio_pipe_stage_t(zio_t *zio); /* * The io_reexecute flags are distinct from io_flags because the child must * be able to propagate them to the parent. The normal io_flags are local * to the zio, not protected by any lock, and not modifiable by children; * the reexecute flags are protected by io_lock, modifiable by children, * and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set. */ #define ZIO_REEXECUTE_NOW 0x01 #define ZIO_REEXECUTE_SUSPEND 0x02 typedef struct zio_link { zio_t *zl_parent; zio_t *zl_child; list_node_t zl_parent_node; list_node_t zl_child_node; } zio_link_t; /* * Used for TRIM kstat. */ typedef struct zio_trim_stats { /* * Number of bytes successfully TRIMmed. */ kstat_named_t bytes; /* * Number of successful TRIM requests. */ kstat_named_t success; /* * Number of TRIM requests that failed because TRIM is not * supported. */ kstat_named_t unsupported; /* * Number of TRIM requests that failed for other reasons. */ kstat_named_t failed; } zio_trim_stats_t; extern zio_trim_stats_t zio_trim_stats; #define ZIO_TRIM_STAT_INCR(stat, val) \ atomic_add_64(&zio_trim_stats.stat.value.ui64, (val)); #define ZIO_TRIM_STAT_BUMP(stat) \ ZIO_TRIM_STAT_INCR(stat, 1); struct zio { /* Core information about this I/O */ zbookmark_phys_t io_bookmark; zio_prop_t io_prop; zio_type_t io_type; enum zio_child io_child_type; int io_cmd; zio_priority_t io_priority; uint8_t io_reexecute; uint8_t io_state[ZIO_WAIT_TYPES]; uint64_t io_txg; spa_t *io_spa; blkptr_t *io_bp; blkptr_t *io_bp_override; blkptr_t io_bp_copy; list_t io_parent_list; list_t io_child_list; zio_link_t *io_walk_link; zio_t *io_logical; zio_transform_t *io_transform_stack; /* Callback info */ zio_done_func_t *io_ready; zio_done_func_t *io_physdone; zio_done_func_t *io_done; void *io_private; int64_t io_prev_space_delta; /* DMU private */ blkptr_t io_bp_orig; /* Data represented by this I/O */ void *io_data; void *io_orig_data; uint64_t io_size; uint64_t io_orig_size; /* Stuff for the vdev stack */ vdev_t *io_vd; void *io_vsd; const zio_vsd_ops_t *io_vsd_ops; uint64_t io_offset; hrtime_t io_timestamp; avl_node_t io_queue_node; avl_node_t io_offset_node; /* Internal pipeline state */ enum zio_flag io_flags; enum zio_stage io_stage; enum zio_stage io_pipeline; enum zio_flag io_orig_flags; enum zio_stage io_orig_stage; enum zio_stage io_orig_pipeline; int io_error; int io_child_error[ZIO_CHILD_TYPES]; uint64_t io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES]; uint64_t io_child_count; uint64_t io_phys_children; uint64_t io_parent_count; uint64_t *io_stall; zio_t *io_gang_leader; zio_gang_node_t *io_gang_tree; void *io_executor; void *io_waiter; kmutex_t io_lock; kcondvar_t io_cv; /* FMA state */ zio_cksum_report_t *io_cksum_report; uint64_t io_ena; /* Taskq dispatching state */ taskq_ent_t io_tqent; avl_node_t io_trim_node; list_node_t io_trim_link; }; extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd, zio_done_func_t *done, void *priv, enum zio_flag flags); extern zio_t *zio_root(spa_t *spa, zio_done_func_t *done, void *priv, enum zio_flag flags); extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, void *data, uint64_t size, zio_done_func_t *done, void *priv, zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb); extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, void *data, uint64_t size, const zio_prop_t *zp, zio_done_func_t *ready, zio_done_func_t *physdone, zio_done_func_t *done, void *priv, zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb); extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, void *data, uint64_t size, zio_done_func_t *done, void *priv, zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb); extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies, boolean_t nopwrite); extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp); extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, zio_done_func_t *done, void *priv, enum zio_flag flags); extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd, uint64_t offset, uint64_t size, zio_done_func_t *done, void *priv, zio_priority_t priority, enum zio_flag flags); extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, void *data, int checksum, zio_done_func_t *done, void *priv, zio_priority_t priority, enum zio_flag flags, boolean_t labels); extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, void *data, int checksum, zio_done_func_t *done, void *priv, zio_priority_t priority, enum zio_flag flags, boolean_t labels); extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, uint64_t size, enum zio_flag flags); extern int zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp, blkptr_t *old_bp, uint64_t size, boolean_t use_slog); extern void zio_free_zil(spa_t *spa, uint64_t txg, blkptr_t *bp); extern void zio_flush(zio_t *zio, vdev_t *vd); extern zio_t *zio_trim(zio_t *zio, spa_t *spa, vdev_t *vd, uint64_t offset, uint64_t size); extern void zio_shrink(zio_t *zio, uint64_t size); extern int zio_wait(zio_t *zio); extern void zio_nowait(zio_t *zio); extern void zio_execute(zio_t *zio); extern void zio_interrupt(zio_t *zio); extern zio_t *zio_walk_parents(zio_t *cio); extern zio_t *zio_walk_children(zio_t *pio); extern zio_t *zio_unique_parent(zio_t *cio); extern void zio_add_child(zio_t *pio, zio_t *cio); extern void *zio_buf_alloc(size_t size); extern void zio_buf_free(void *buf, size_t size); extern void *zio_data_buf_alloc(size_t size); extern void zio_data_buf_free(void *buf, size_t size); extern void zio_resubmit_stage_async(void *); extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd, uint64_t offset, void *data, uint64_t size, int type, zio_priority_t priority, enum zio_flag flags, zio_done_func_t *done, void *priv); extern zio_t *zio_vdev_delegated_io(vdev_t *vd, uint64_t offset, void *data, uint64_t size, int type, zio_priority_t priority, enum zio_flag flags, zio_done_func_t *done, void *priv); extern void zio_vdev_io_bypass(zio_t *zio); extern void zio_vdev_io_reissue(zio_t *zio); extern void zio_vdev_io_redone(zio_t *zio); extern void zio_checksum_verified(zio_t *zio); extern int zio_worst_error(int e1, int e2); extern enum zio_checksum zio_checksum_select(enum zio_checksum child, enum zio_checksum parent); extern enum zio_checksum zio_checksum_dedup_select(spa_t *spa, enum zio_checksum child, enum zio_checksum parent); extern enum zio_compress zio_compress_select(spa_t *spa, enum zio_compress child, enum zio_compress parent); extern void zio_suspend(spa_t *spa, zio_t *zio); extern int zio_resume(spa_t *spa); extern void zio_resume_wait(spa_t *spa); /* * Initial setup and teardown. */ extern void zio_init(void); extern void zio_fini(void); /* * Fault injection */ struct zinject_record; extern uint32_t zio_injection_enabled; extern int zio_inject_fault(char *name, int flags, int *id, struct zinject_record *record); extern int zio_inject_list_next(int *id, char *name, size_t buflen, struct zinject_record *record); extern int zio_clear_fault(int id); extern void zio_handle_panic_injection(spa_t *spa, char *tag, uint64_t type); extern int zio_handle_fault_injection(zio_t *zio, int error); extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error); extern int zio_handle_label_injection(zio_t *zio, int error); extern void zio_handle_ignored_writes(zio_t *zio); extern uint64_t zio_handle_io_delay(zio_t *zio); /* * Checksum ereport functions */ extern void zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd, struct zio *zio, uint64_t offset, uint64_t length, void *arg, struct zio_bad_cksum *info); extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report, const void *good_data, const void *bad_data, boolean_t drop_if_identical); extern void zfs_ereport_send_interim_checksum(zio_cksum_report_t *report); extern void zfs_ereport_free_checksum(zio_cksum_report_t *report); /* If we have the good data in hand, this function can be used */ extern void zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd, struct zio *zio, uint64_t offset, uint64_t length, const void *good_data, const void *bad_data, struct zio_bad_cksum *info); /* Called from spa_sync(), but primarily an injection handler */ extern void spa_handle_ignored_writes(spa_t *spa); /* zbookmark_phys functions */ boolean_t zbookmark_subtree_completed(const struct dnode_phys *dnp, const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block); int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2, uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2); #ifdef __cplusplus } #endif #endif /* _ZIO_H */ Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zio_checksum.h =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zio_checksum.h (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zio_checksum.h (revision 290757) @@ -1,76 +1,113 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. - * Copyright (c) 2014 by Delphix. All rights reserved. + * Copyright (c) 2014, 2015 by Delphix. All rights reserved. + * Copyright Saso Kiselkov 2013, All rights reserved. */ #ifndef _SYS_ZIO_CHECKSUM_H #define _SYS_ZIO_CHECKSUM_H #include +#include #ifdef __cplusplus extern "C" { #endif /* * Signature for checksum functions. */ -typedef void zio_checksum_func_t(const void *, uint64_t, zio_cksum_t *); +typedef void zio_checksum_t(const void *data, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp); +typedef void *zio_checksum_tmpl_init_t(const zio_cksum_salt_t *salt); +typedef void zio_checksum_tmpl_free_t(void *ctx_template); +typedef enum zio_checksum_flags { + /* Strong enough for metadata? */ + ZCHECKSUM_FLAG_METADATA = (1 << 1), + /* ZIO embedded checksum */ + ZCHECKSUM_FLAG_EMBEDDED = (1 << 2), + /* Strong enough for dedup (without verification)? */ + ZCHECKSUM_FLAG_DEDUP = (1 << 3), + /* Uses salt value */ + ZCHECKSUM_FLAG_SALTED = (1 << 4), + /* Strong enough for nopwrite? */ + ZCHECKSUM_FLAG_NOPWRITE = (1 << 5) +} zio_checksum_flags_t; + /* * Information about each checksum function. */ typedef struct zio_checksum_info { - zio_checksum_func_t *ci_func[2]; /* checksum function per byteorder */ - int ci_correctable; /* number of correctable bits */ - int ci_eck; /* uses zio embedded checksum? */ - boolean_t ci_dedup; /* strong enough for dedup? */ - char *ci_name; /* descriptive name */ + /* checksum function for each byteorder */ + zio_checksum_t *ci_func[2]; + zio_checksum_tmpl_init_t *ci_tmpl_init; + zio_checksum_tmpl_free_t *ci_tmpl_free; + zio_checksum_flags_t ci_flags; + char *ci_name; /* descriptive name */ } zio_checksum_info_t; typedef struct zio_bad_cksum { zio_cksum_t zbc_expected; zio_cksum_t zbc_actual; const char *zbc_checksum_name; uint8_t zbc_byteswapped; uint8_t zbc_injected; uint8_t zbc_has_cksum; /* expected/actual valid */ } zio_bad_cksum_t; extern zio_checksum_info_t zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS]; /* * Checksum routines. */ -extern zio_checksum_func_t zio_checksum_SHA256; +extern zio_checksum_t zio_checksum_SHA256; +#ifdef illumos +extern zio_checksum_t zio_checksum_SHA512_native; +extern zio_checksum_t zio_checksum_SHA512_byteswap; +/* Skein */ +extern zio_checksum_t zio_checksum_skein_native; +extern zio_checksum_t zio_checksum_skein_byteswap; +extern zio_checksum_tmpl_init_t zio_checksum_skein_tmpl_init; +extern zio_checksum_tmpl_free_t zio_checksum_skein_tmpl_free; + +/* Edon-R */ +extern zio_checksum_t zio_checksum_edonr_native; +extern zio_checksum_t zio_checksum_edonr_byteswap; +extern zio_checksum_tmpl_init_t zio_checksum_edonr_tmpl_init; +extern zio_checksum_tmpl_free_t zio_checksum_edonr_tmpl_free; +#endif + extern void zio_checksum_compute(zio_t *zio, enum zio_checksum checksum, void *data, uint64_t size); extern int zio_checksum_error(zio_t *zio, zio_bad_cksum_t *out); extern enum zio_checksum spa_dedup_checksum(spa_t *spa); +extern void zio_checksum_templates_free(spa_t *spa); +extern spa_feature_t zio_checksum_to_feature(enum zio_checksum cksum); #ifdef __cplusplus } #endif #endif /* _SYS_ZIO_CHECKSUM_H */ Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_ioctl.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_ioctl.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_ioctl.c (revision 290757) @@ -1,6644 +1,6679 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011-2012 Pawel Jakub Dawidek . * All rights reserved. * Copyright 2013 Martin Matuska . All rights reserved. * Copyright 2014 Xin Li . All rights reserved. * Copyright 2015 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2014, Joyent, Inc. All rights reserved. * Copyright (c) 2011, 2015 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2013 Steven Hartland. All rights reserved. */ /* * ZFS ioctls. * * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. * * There are two ways that we handle ioctls: the legacy way where almost * all of the logic is in the ioctl callback, and the new way where most * of the marshalling is handled in the common entry point, zfsdev_ioctl(). * * Non-legacy ioctls should be registered by calling * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked * from userland by lzc_ioctl(). * * The registration arguments are as follows: * * const char *name * The name of the ioctl. This is used for history logging. If the * ioctl returns successfully (the callback returns 0), and allow_log * is true, then a history log entry will be recorded with the input & * output nvlists. The log entry can be printed with "zpool history -i". * * zfs_ioc_t ioc * The ioctl request number, which userland will pass to ioctl(2). * The ioctl numbers can change from release to release, because * the caller (libzfs) must be matched to the kernel. * * zfs_secpolicy_func_t *secpolicy * This function will be called before the zfs_ioc_func_t, to * determine if this operation is permitted. It should return EPERM * on failure, and 0 on success. Checks include determining if the * dataset is visible in this zone, and if the user has either all * zfs privileges in the zone (SYS_MOUNT), or has been granted permission * to do this operation on this dataset with "zfs allow". * * zfs_ioc_namecheck_t namecheck * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool * name, a dataset name, or nothing. If the name is not well-formed, * the ioctl will fail and the callback will not be called. * Therefore, the callback can assume that the name is well-formed * (e.g. is null-terminated, doesn't have more than one '@' character, * doesn't have invalid characters). * * zfs_ioc_poolcheck_t pool_check * This specifies requirements on the pool state. If the pool does * not meet them (is suspended or is readonly), the ioctl will fail * and the callback will not be called. If any checks are specified * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | * POOL_CHECK_READONLY). * * boolean_t smush_outnvlist * If smush_outnvlist is true, then the output is presumed to be a * list of errors, and it will be "smushed" down to fit into the * caller's buffer, by removing some entries and replacing them with a * single "N_MORE_ERRORS" entry indicating how many were removed. See * nvlist_smush() for details. If smush_outnvlist is false, and the * outnvlist does not fit into the userland-provided buffer, then the * ioctl will fail with ENOMEM. * * zfs_ioc_func_t *func * The callback function that will perform the operation. * * The callback should return 0 on success, or an error number on * failure. If the function fails, the userland ioctl will return -1, * and errno will be set to the callback's return value. The callback * will be called with the following arguments: * * const char *name * The name of the pool or dataset to operate on, from * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the * expected type (pool, dataset, or none). * * nvlist_t *innvl * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or * NULL if no input nvlist was provided. Changes to this nvlist are * ignored. If the input nvlist could not be deserialized, the * ioctl will fail and the callback will not be called. * * nvlist_t *outnvl * The output nvlist, initially empty. The callback can fill it in, * and it will be returned to userland by serializing it into * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization * fails (e.g. because the caller didn't supply a large enough * buffer), then the overall ioctl will fail. See the * 'smush_nvlist' argument above for additional behaviors. * * There are two typical uses of the output nvlist: * - To return state, e.g. property values. In this case, * smush_outnvlist should be false. If the buffer was not large * enough, the caller will reallocate a larger buffer and try * the ioctl again. * * - To return multiple errors from an ioctl which makes on-disk * changes. In this case, smush_outnvlist should be true. * Ioctls which make on-disk modifications should generally not * use the outnvl if they succeed, because the caller can not * distinguish between the operation failing, and * deserialization failing. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include "zfs_namecheck.h" #include "zfs_prop.h" #include "zfs_deleg.h" #include "zfs_comutil.h" #include "zfs_ioctl_compat.h" CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX); static struct cdev *zfsdev; extern void zfs_init(void); extern void zfs_fini(void); uint_t zfs_fsyncer_key; extern uint_t rrw_tsd_key; static uint_t zfs_allow_log_key; typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); typedef enum { NO_NAME, POOL_NAME, DATASET_NAME } zfs_ioc_namecheck_t; typedef enum { POOL_CHECK_NONE = 1 << 0, POOL_CHECK_SUSPENDED = 1 << 1, POOL_CHECK_READONLY = 1 << 2, } zfs_ioc_poolcheck_t; typedef struct zfs_ioc_vec { zfs_ioc_legacy_func_t *zvec_legacy_func; zfs_ioc_func_t *zvec_func; zfs_secpolicy_func_t *zvec_secpolicy; zfs_ioc_namecheck_t zvec_namecheck; boolean_t zvec_allow_log; zfs_ioc_poolcheck_t zvec_pool_check; boolean_t zvec_smush_outnvlist; const char *zvec_name; } zfs_ioc_vec_t; /* This array is indexed by zfs_userquota_prop_t */ static const char *userquota_perms[] = { ZFS_DELEG_PERM_USERUSED, ZFS_DELEG_PERM_USERQUOTA, ZFS_DELEG_PERM_GROUPUSED, ZFS_DELEG_PERM_GROUPQUOTA, }; static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); static int zfs_check_settable(const char *name, nvpair_t *property, cred_t *cr); static int zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errors); static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, boolean_t *); int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); static void zfsdev_close(void *data); static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ void __dprintf(const char *file, const char *func, int line, const char *fmt, ...) { const char *newfile; char buf[512]; va_list adx; /* * Get rid of annoying "../common/" prefix to filename. */ newfile = strrchr(file, '/'); if (newfile != NULL) { newfile = newfile + 1; /* Get rid of leading / */ } else { newfile = file; } va_start(adx, fmt); (void) vsnprintf(buf, sizeof (buf), fmt, adx); va_end(adx); /* * To get this data, use the zfs-dprintf probe as so: * dtrace -q -n 'zfs-dprintf \ * /stringof(arg0) == "dbuf.c"/ \ * {printf("%s: %s", stringof(arg1), stringof(arg3))}' * arg0 = file name * arg1 = function name * arg2 = line number * arg3 = message */ DTRACE_PROBE4(zfs__dprintf, char *, newfile, char *, func, int, line, char *, buf); } static void history_str_free(char *buf) { kmem_free(buf, HIS_MAX_RECORD_LEN); } static char * history_str_get(zfs_cmd_t *zc) { char *buf; if (zc->zc_history == 0) return (NULL); buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); if (copyinstr((void *)(uintptr_t)zc->zc_history, buf, HIS_MAX_RECORD_LEN, NULL) != 0) { history_str_free(buf); return (NULL); } buf[HIS_MAX_RECORD_LEN -1] = '\0'; return (buf); } /* * Check to see if the named dataset is currently defined as bootable */ static boolean_t zfs_is_bootfs(const char *name) { objset_t *os; if (dmu_objset_hold(name, FTAG, &os) == 0) { boolean_t ret; ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); dmu_objset_rele(os, FTAG); return (ret); } return (B_FALSE); } /* * Return non-zero if the spa version is less than requested version. */ static int zfs_earlier_version(const char *name, int version) { spa_t *spa; if (spa_open(name, &spa, FTAG) == 0) { if (spa_version(spa) < version) { spa_close(spa, FTAG); return (1); } spa_close(spa, FTAG); } return (0); } /* * Return TRUE if the ZPL version is less than requested version. */ static boolean_t zpl_earlier_version(const char *name, int version) { objset_t *os; boolean_t rc = B_TRUE; if (dmu_objset_hold(name, FTAG, &os) == 0) { uint64_t zplversion; if (dmu_objset_type(os) != DMU_OST_ZFS) { dmu_objset_rele(os, FTAG); return (B_TRUE); } /* XXX reading from non-owned objset */ if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) rc = zplversion < version; dmu_objset_rele(os, FTAG); } return (rc); } static void zfs_log_history(zfs_cmd_t *zc) { spa_t *spa; char *buf; if ((buf = history_str_get(zc)) == NULL) return; if (spa_open(zc->zc_name, &spa, FTAG) == 0) { if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) (void) spa_history_log(spa, buf); spa_close(spa, FTAG); } history_str_free(buf); } /* * Policy for top-level read operations (list pools). Requires no privileges, * and can be used in the local zone, as there is no associated dataset. */ /* ARGSUSED */ static int zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { return (0); } /* * Policy for dataset read operations (list children, get statistics). Requires * no privileges, but must be visible in the local zone. */ /* ARGSUSED */ static int zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { if (INGLOBALZONE(curthread) || zone_dataset_visible(zc->zc_name, NULL)) return (0); return (SET_ERROR(ENOENT)); } static int zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) { int writable = 1; /* * The dataset must be visible by this zone -- check this first * so they don't see EPERM on something they shouldn't know about. */ if (!INGLOBALZONE(curthread) && !zone_dataset_visible(dataset, &writable)) return (SET_ERROR(ENOENT)); if (INGLOBALZONE(curthread)) { /* * If the fs is zoned, only root can access it from the * global zone. */ if (secpolicy_zfs(cr) && zoned) return (SET_ERROR(EPERM)); } else { /* * If we are in a local zone, the 'zoned' property must be set. */ if (!zoned) return (SET_ERROR(EPERM)); /* must be writable by this zone */ if (!writable) return (SET_ERROR(EPERM)); } return (0); } static int zfs_dozonecheck(const char *dataset, cred_t *cr) { uint64_t zoned; if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL)) return (SET_ERROR(ENOENT)); return (zfs_dozonecheck_impl(dataset, zoned, cr)); } static int zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) { uint64_t zoned; if (dsl_prop_get_int_ds(ds, "jailed", &zoned)) return (SET_ERROR(ENOENT)); return (zfs_dozonecheck_impl(dataset, zoned, cr)); } static int zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, const char *perm, cred_t *cr) { int error; error = zfs_dozonecheck_ds(name, ds, cr); if (error == 0) { error = secpolicy_zfs(cr); if (error != 0) error = dsl_deleg_access_impl(ds, perm, cr); } return (error); } static int zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) { int error; dsl_dataset_t *ds; dsl_pool_t *dp; error = dsl_pool_hold(name, FTAG, &dp); if (error != 0) return (error); error = dsl_dataset_hold(dp, name, FTAG, &ds); if (error != 0) { dsl_pool_rele(dp, FTAG); return (error); } error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); dsl_dataset_rele(ds, FTAG); dsl_pool_rele(dp, FTAG); return (error); } #ifdef SECLABEL /* * Policy for setting the security label property. * * Returns 0 for success, non-zero for access and other errors. */ static int zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) { char ds_hexsl[MAXNAMELEN]; bslabel_t ds_sl, new_sl; boolean_t new_default = FALSE; uint64_t zoned; int needed_priv = -1; int error; /* First get the existing dataset label. */ error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 1, sizeof (ds_hexsl), &ds_hexsl, NULL); if (error != 0) return (SET_ERROR(EPERM)); if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) new_default = TRUE; /* The label must be translatable */ if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) return (SET_ERROR(EINVAL)); /* * In a non-global zone, disallow attempts to set a label that * doesn't match that of the zone; otherwise no other checks * are needed. */ if (!INGLOBALZONE(curproc)) { if (new_default || !blequal(&new_sl, CR_SL(CRED()))) return (SET_ERROR(EPERM)); return (0); } /* * For global-zone datasets (i.e., those whose zoned property is * "off", verify that the specified new label is valid for the * global zone. */ if (dsl_prop_get_integer(name, zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) return (SET_ERROR(EPERM)); if (!zoned) { if (zfs_check_global_label(name, strval) != 0) return (SET_ERROR(EPERM)); } /* * If the existing dataset label is nondefault, check if the * dataset is mounted (label cannot be changed while mounted). * Get the zfsvfs; if there isn't one, then the dataset isn't * mounted (or isn't a dataset, doesn't exist, ...). */ if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { objset_t *os; static char *setsl_tag = "setsl_tag"; /* * Try to own the dataset; abort if there is any error, * (e.g., already mounted, in use, or other error). */ error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, setsl_tag, &os); if (error != 0) return (SET_ERROR(EPERM)); dmu_objset_disown(os, setsl_tag); if (new_default) { needed_priv = PRIV_FILE_DOWNGRADE_SL; goto out_check; } if (hexstr_to_label(strval, &new_sl) != 0) return (SET_ERROR(EPERM)); if (blstrictdom(&ds_sl, &new_sl)) needed_priv = PRIV_FILE_DOWNGRADE_SL; else if (blstrictdom(&new_sl, &ds_sl)) needed_priv = PRIV_FILE_UPGRADE_SL; } else { /* dataset currently has a default label */ if (!new_default) needed_priv = PRIV_FILE_UPGRADE_SL; } out_check: if (needed_priv != -1) return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); return (0); } #endif /* SECLABEL */ static int zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, cred_t *cr) { char *strval; /* * Check permissions for special properties. */ switch (prop) { case ZFS_PROP_ZONED: /* * Disallow setting of 'zoned' from within a local zone. */ if (!INGLOBALZONE(curthread)) return (SET_ERROR(EPERM)); break; case ZFS_PROP_QUOTA: case ZFS_PROP_FILESYSTEM_LIMIT: case ZFS_PROP_SNAPSHOT_LIMIT: if (!INGLOBALZONE(curthread)) { uint64_t zoned; char setpoint[MAXNAMELEN]; /* * Unprivileged users are allowed to modify the * limit on things *under* (ie. contained by) * the thing they own. */ if (dsl_prop_get_integer(dsname, "jailed", &zoned, setpoint)) return (SET_ERROR(EPERM)); if (!zoned || strlen(dsname) <= strlen(setpoint)) return (SET_ERROR(EPERM)); } break; case ZFS_PROP_MLSLABEL: #ifdef SECLABEL if (!is_system_labeled()) return (SET_ERROR(EPERM)); if (nvpair_value_string(propval, &strval) == 0) { int err; err = zfs_set_slabel_policy(dsname, strval, CRED()); if (err != 0) return (err); } #else return (EOPNOTSUPP); #endif break; } return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); } /* ARGSUSED */ static int zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { int error; error = zfs_dozonecheck(zc->zc_name, cr); if (error != 0) return (error); /* * permission to set permissions will be evaluated later in * dsl_deleg_can_allow() */ return (0); } /* ARGSUSED */ static int zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { return (zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_ROLLBACK, cr)); } /* ARGSUSED */ static int zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { dsl_pool_t *dp; dsl_dataset_t *ds; char *cp; int error; /* * Generate the current snapshot name from the given objsetid, then * use that name for the secpolicy/zone checks. */ cp = strchr(zc->zc_name, '@'); if (cp == NULL) return (SET_ERROR(EINVAL)); error = dsl_pool_hold(zc->zc_name, FTAG, &dp); if (error != 0) return (error); error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); if (error != 0) { dsl_pool_rele(dp, FTAG); return (error); } dsl_dataset_name(ds, zc->zc_name); error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, ZFS_DELEG_PERM_SEND, cr); dsl_dataset_rele(ds, FTAG); dsl_pool_rele(dp, FTAG); return (error); } /* ARGSUSED */ static int zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { return (zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_SEND, cr)); } /* ARGSUSED */ static int zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { vnode_t *vp; int error; if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, NO_FOLLOW, NULL, &vp)) != 0) return (error); /* Now make sure mntpnt and dataset are ZFS */ if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), zc->zc_name) != 0)) { VN_RELE(vp); return (SET_ERROR(EPERM)); } VN_RELE(vp); return (dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_SHARE, cr)); } int zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { if (!INGLOBALZONE(curthread)) return (SET_ERROR(EPERM)); if (secpolicy_nfs(cr) == 0) { return (0); } else { return (zfs_secpolicy_deleg_share(zc, innvl, cr)); } } int zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { if (!INGLOBALZONE(curthread)) return (SET_ERROR(EPERM)); if (secpolicy_smb(cr) == 0) { return (0); } else { return (zfs_secpolicy_deleg_share(zc, innvl, cr)); } } static int zfs_get_parent(const char *datasetname, char *parent, int parentsize) { char *cp; /* * Remove the @bla or /bla from the end of the name to get the parent. */ (void) strncpy(parent, datasetname, parentsize); cp = strrchr(parent, '@'); if (cp != NULL) { cp[0] = '\0'; } else { cp = strrchr(parent, '/'); if (cp == NULL) return (SET_ERROR(ENOENT)); cp[0] = '\0'; } return (0); } int zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) { int error; if ((error = zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_MOUNT, cr)) != 0) return (error); return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); } /* ARGSUSED */ static int zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); } /* * Destroying snapshots with delegated permissions requires * descendant mount and destroy permissions. */ /* ARGSUSED */ static int zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { nvlist_t *snaps; nvpair_t *pair, *nextpair; int error = 0; if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) return (SET_ERROR(EINVAL)); for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; pair = nextpair) { nextpair = nvlist_next_nvpair(snaps, pair); error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); if (error == ENOENT) { /* * Ignore any snapshots that don't exist (we consider * them "already destroyed"). Remove the name from the * nvl here in case the snapshot is created between * now and when we try to destroy it (in which case * we don't want to destroy it since we haven't * checked for permission). */ fnvlist_remove_nvpair(snaps, pair); error = 0; } if (error != 0) break; } return (error); } int zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) { char parentname[MAXNAMELEN]; int error; if ((error = zfs_secpolicy_write_perms(from, ZFS_DELEG_PERM_RENAME, cr)) != 0) return (error); if ((error = zfs_secpolicy_write_perms(from, ZFS_DELEG_PERM_MOUNT, cr)) != 0) return (error); if ((error = zfs_get_parent(to, parentname, sizeof (parentname))) != 0) return (error); if ((error = zfs_secpolicy_write_perms(parentname, ZFS_DELEG_PERM_CREATE, cr)) != 0) return (error); if ((error = zfs_secpolicy_write_perms(parentname, ZFS_DELEG_PERM_MOUNT, cr)) != 0) return (error); return (error); } /* ARGSUSED */ static int zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { char *at = NULL; int error; if ((zc->zc_cookie & 1) != 0) { /* * This is recursive rename, so the starting snapshot might * not exist. Check file system or volume permission instead. */ at = strchr(zc->zc_name, '@'); if (at == NULL) return (EINVAL); *at = '\0'; } error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr); if (at != NULL) *at = '@'; return (error); } /* ARGSUSED */ static int zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { dsl_pool_t *dp; dsl_dataset_t *clone; int error; error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_PROMOTE, cr); if (error != 0) return (error); error = dsl_pool_hold(zc->zc_name, FTAG, &dp); if (error != 0) return (error); error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); if (error == 0) { char parentname[MAXNAMELEN]; dsl_dataset_t *origin = NULL; dsl_dir_t *dd; dd = clone->ds_dir; error = dsl_dataset_hold_obj(dd->dd_pool, dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin); if (error != 0) { dsl_dataset_rele(clone, FTAG); dsl_pool_rele(dp, FTAG); return (error); } error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, ZFS_DELEG_PERM_MOUNT, cr); dsl_dataset_name(origin, parentname); if (error == 0) { error = zfs_secpolicy_write_perms_ds(parentname, origin, ZFS_DELEG_PERM_PROMOTE, cr); } dsl_dataset_rele(clone, FTAG); dsl_dataset_rele(origin, FTAG); } dsl_pool_rele(dp, FTAG); return (error); } /* ARGSUSED */ static int zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { int error; if ((error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_RECEIVE, cr)) != 0) return (error); if ((error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr)) != 0) return (error); return (zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_CREATE, cr)); } int zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) { return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_SNAPSHOT, cr)); } /* * Check for permission to create each snapshot in the nvlist. */ /* ARGSUSED */ static int zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { nvlist_t *snaps; int error; nvpair_t *pair; if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) return (SET_ERROR(EINVAL)); for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; pair = nvlist_next_nvpair(snaps, pair)) { char *name = nvpair_name(pair); char *atp = strchr(name, '@'); if (atp == NULL) { error = SET_ERROR(EINVAL); break; } *atp = '\0'; error = zfs_secpolicy_snapshot_perms(name, cr); *atp = '@'; if (error != 0) break; } return (error); } /* * Check for permission to create each snapshot in the nvlist. */ /* ARGSUSED */ static int zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { int error = 0; for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { char *name = nvpair_name(pair); char *hashp = strchr(name, '#'); if (hashp == NULL) { error = SET_ERROR(EINVAL); break; } *hashp = '\0'; error = zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_BOOKMARK, cr); *hashp = '#'; if (error != 0) break; } return (error); } /* ARGSUSED */ static int zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { nvpair_t *pair, *nextpair; int error = 0; for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; pair = nextpair) { char *name = nvpair_name(pair); char *hashp = strchr(name, '#'); nextpair = nvlist_next_nvpair(innvl, pair); if (hashp == NULL) { error = SET_ERROR(EINVAL); break; } *hashp = '\0'; error = zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr); *hashp = '#'; if (error == ENOENT) { /* * Ignore any filesystems that don't exist (we consider * their bookmarks "already destroyed"). Remove * the name from the nvl here in case the filesystem * is created between now and when we try to destroy * the bookmark (in which case we don't want to * destroy it since we haven't checked for permission). */ fnvlist_remove_nvpair(innvl, pair); error = 0; } if (error != 0) break; } return (error); } /* ARGSUSED */ static int zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { /* * Even root must have a proper TSD so that we know what pool * to log to. */ if (tsd_get(zfs_allow_log_key) == NULL) return (SET_ERROR(EPERM)); return (0); } static int zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { char parentname[MAXNAMELEN]; int error; char *origin; if ((error = zfs_get_parent(zc->zc_name, parentname, sizeof (parentname))) != 0) return (error); if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && (error = zfs_secpolicy_write_perms(origin, ZFS_DELEG_PERM_CLONE, cr)) != 0) return (error); if ((error = zfs_secpolicy_write_perms(parentname, ZFS_DELEG_PERM_CREATE, cr)) != 0) return (error); return (zfs_secpolicy_write_perms(parentname, ZFS_DELEG_PERM_MOUNT, cr)); } /* * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires * SYS_CONFIG privilege, which is not available in a local zone. */ /* ARGSUSED */ static int zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { if (secpolicy_sys_config(cr, B_FALSE) != 0) return (SET_ERROR(EPERM)); return (0); } /* * Policy for object to name lookups. */ /* ARGSUSED */ static int zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { int error; if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) return (0); error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); return (error); } /* * Policy for fault injection. Requires all privileges. */ /* ARGSUSED */ static int zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { return (secpolicy_zinject(cr)); } /* ARGSUSED */ static int zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); if (prop == ZPROP_INVAL) { if (!zfs_prop_user(zc->zc_value)) return (SET_ERROR(EINVAL)); return (zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_USERPROP, cr)); } else { return (zfs_secpolicy_setprop(zc->zc_name, prop, NULL, cr)); } } static int zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { int err = zfs_secpolicy_read(zc, innvl, cr); if (err) return (err); if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) return (SET_ERROR(EINVAL)); if (zc->zc_value[0] == 0) { /* * They are asking about a posix uid/gid. If it's * themself, allow it. */ if (zc->zc_objset_type == ZFS_PROP_USERUSED || zc->zc_objset_type == ZFS_PROP_USERQUOTA) { if (zc->zc_guid == crgetuid(cr)) return (0); } else { if (groupmember(zc->zc_guid, cr)) return (0); } } return (zfs_secpolicy_write_perms(zc->zc_name, userquota_perms[zc->zc_objset_type], cr)); } static int zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { int err = zfs_secpolicy_read(zc, innvl, cr); if (err) return (err); if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) return (SET_ERROR(EINVAL)); return (zfs_secpolicy_write_perms(zc->zc_name, userquota_perms[zc->zc_objset_type], cr)); } /* ARGSUSED */ static int zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, NULL, cr)); } /* ARGSUSED */ static int zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { nvpair_t *pair; nvlist_t *holds; int error; error = nvlist_lookup_nvlist(innvl, "holds", &holds); if (error != 0) return (SET_ERROR(EINVAL)); for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; pair = nvlist_next_nvpair(holds, pair)) { char fsname[MAXNAMELEN]; error = dmu_fsname(nvpair_name(pair), fsname); if (error != 0) return (error); error = zfs_secpolicy_write_perms(fsname, ZFS_DELEG_PERM_HOLD, cr); if (error != 0) return (error); } return (0); } /* ARGSUSED */ static int zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { nvpair_t *pair; int error; for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { char fsname[MAXNAMELEN]; error = dmu_fsname(nvpair_name(pair), fsname); if (error != 0) return (error); error = zfs_secpolicy_write_perms(fsname, ZFS_DELEG_PERM_RELEASE, cr); if (error != 0) return (error); } return (0); } /* * Policy for allowing temporary snapshots to be taken or released */ static int zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) { /* * A temporary snapshot is the same as a snapshot, * hold, destroy and release all rolled into one. * Delegated diff alone is sufficient that we allow this. */ int error; if ((error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr)) == 0) return (0); error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); if (error == 0) error = zfs_secpolicy_hold(zc, innvl, cr); if (error == 0) error = zfs_secpolicy_release(zc, innvl, cr); if (error == 0) error = zfs_secpolicy_destroy(zc, innvl, cr); return (error); } /* * Returns the nvlist as specified by the user in the zfs_cmd_t. */ static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) { char *packed; int error; nvlist_t *list = NULL; /* * Read in and unpack the user-supplied nvlist. */ if (size == 0) return (SET_ERROR(EINVAL)); packed = kmem_alloc(size, KM_SLEEP); if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, iflag)) != 0) { kmem_free(packed, size); return (error); } if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { kmem_free(packed, size); return (error); } kmem_free(packed, size); *nvp = list; return (0); } /* * Reduce the size of this nvlist until it can be serialized in 'max' bytes. * Entries will be removed from the end of the nvlist, and one int32 entry * named "N_MORE_ERRORS" will be added indicating how many entries were * removed. */ static int nvlist_smush(nvlist_t *errors, size_t max) { size_t size; size = fnvlist_size(errors); if (size > max) { nvpair_t *more_errors; int n = 0; if (max < 1024) return (SET_ERROR(ENOMEM)); fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); more_errors = nvlist_prev_nvpair(errors, NULL); do { nvpair_t *pair = nvlist_prev_nvpair(errors, more_errors); fnvlist_remove_nvpair(errors, pair); n++; size = fnvlist_size(errors); } while (size > max); fnvlist_remove_nvpair(errors, more_errors); fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); ASSERT3U(fnvlist_size(errors), <=, max); } return (0); } static int put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) { char *packed = NULL; int error = 0; size_t size; size = fnvlist_size(nvl); if (size > zc->zc_nvlist_dst_size) { /* * Solaris returns ENOMEM here, because even if an error is * returned from an ioctl(2), new zc_nvlist_dst_size will be * passed to the userland. This is not the case for FreeBSD. * We need to return 0, so the kernel will copy the * zc_nvlist_dst_size back and the userland can discover that a * bigger buffer is needed. */ error = 0; } else { packed = fnvlist_pack(nvl, &size); if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, size, zc->zc_iflags) != 0) error = SET_ERROR(EFAULT); fnvlist_pack_free(packed, size); } zc->zc_nvlist_dst_size = size; zc->zc_nvlist_dst_filled = B_TRUE; return (error); } static int getzfsvfs(const char *dsname, zfsvfs_t **zfvp) { objset_t *os; int error; error = dmu_objset_hold(dsname, FTAG, &os); if (error != 0) return (error); if (dmu_objset_type(os) != DMU_OST_ZFS) { dmu_objset_rele(os, FTAG); return (SET_ERROR(EINVAL)); } mutex_enter(&os->os_user_ptr_lock); *zfvp = dmu_objset_get_user(os); if (*zfvp) { VFS_HOLD((*zfvp)->z_vfs); } else { error = SET_ERROR(ESRCH); } mutex_exit(&os->os_user_ptr_lock); dmu_objset_rele(os, FTAG); return (error); } /* * Find a zfsvfs_t for a mounted filesystem, or create our own, in which * case its z_vfs will be NULL, and it will be opened as the owner. * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, * which prevents all vnode ops from running. */ static int zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) { int error = 0; if (getzfsvfs(name, zfvp) != 0) error = zfsvfs_create(name, zfvp); if (error == 0) { rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : RW_READER, tag); if ((*zfvp)->z_unmounted) { /* * XXX we could probably try again, since the unmounting * thread should be just about to disassociate the * objset from the zfsvfs. */ rrm_exit(&(*zfvp)->z_teardown_lock, tag); return (SET_ERROR(EBUSY)); } } return (error); } static void zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) { rrm_exit(&zfsvfs->z_teardown_lock, tag); if (zfsvfs->z_vfs) { VFS_RELE(zfsvfs->z_vfs); } else { dmu_objset_disown(zfsvfs->z_os, zfsvfs); zfsvfs_free(zfsvfs); } } static int zfs_ioc_pool_create(zfs_cmd_t *zc) { int error; nvlist_t *config, *props = NULL; nvlist_t *rootprops = NULL; nvlist_t *zplprops = NULL; if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, zc->zc_iflags, &config)) return (error); if (zc->zc_nvlist_src_size != 0 && (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, zc->zc_iflags, &props))) { nvlist_free(config); return (error); } if (props) { nvlist_t *nvl = NULL; uint64_t version = SPA_VERSION; (void) nvlist_lookup_uint64(props, zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); if (!SPA_VERSION_IS_SUPPORTED(version)) { error = SET_ERROR(EINVAL); goto pool_props_bad; } (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); if (nvl) { error = nvlist_dup(nvl, &rootprops, KM_SLEEP); if (error != 0) { nvlist_free(config); nvlist_free(props); return (error); } (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); } VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); error = zfs_fill_zplprops_root(version, rootprops, zplprops, NULL); if (error != 0) goto pool_props_bad; } error = spa_create(zc->zc_name, config, props, zplprops); /* * Set the remaining root properties */ if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) (void) spa_destroy(zc->zc_name); pool_props_bad: nvlist_free(rootprops); nvlist_free(zplprops); nvlist_free(config); nvlist_free(props); return (error); } static int zfs_ioc_pool_destroy(zfs_cmd_t *zc) { int error; zfs_log_history(zc); error = spa_destroy(zc->zc_name); if (error == 0) zvol_remove_minors(zc->zc_name); return (error); } static int zfs_ioc_pool_import(zfs_cmd_t *zc) { nvlist_t *config, *props = NULL; uint64_t guid; int error; if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, zc->zc_iflags, &config)) != 0) return (error); if (zc->zc_nvlist_src_size != 0 && (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, zc->zc_iflags, &props))) { nvlist_free(config); return (error); } if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || guid != zc->zc_guid) error = SET_ERROR(EINVAL); else error = spa_import(zc->zc_name, config, props, zc->zc_cookie); if (zc->zc_nvlist_dst != 0) { int err; if ((err = put_nvlist(zc, config)) != 0) error = err; } nvlist_free(config); if (props) nvlist_free(props); return (error); } static int zfs_ioc_pool_export(zfs_cmd_t *zc) { int error; boolean_t force = (boolean_t)zc->zc_cookie; boolean_t hardforce = (boolean_t)zc->zc_guid; zfs_log_history(zc); error = spa_export(zc->zc_name, NULL, force, hardforce); if (error == 0) zvol_remove_minors(zc->zc_name); return (error); } static int zfs_ioc_pool_configs(zfs_cmd_t *zc) { nvlist_t *configs; int error; if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) return (SET_ERROR(EEXIST)); error = put_nvlist(zc, configs); nvlist_free(configs); return (error); } /* * inputs: * zc_name name of the pool * * outputs: * zc_cookie real errno * zc_nvlist_dst config nvlist * zc_nvlist_dst_size size of config nvlist */ static int zfs_ioc_pool_stats(zfs_cmd_t *zc) { nvlist_t *config; int error; int ret = 0; error = spa_get_stats(zc->zc_name, &config, zc->zc_value, sizeof (zc->zc_value)); if (config != NULL) { ret = put_nvlist(zc, config); nvlist_free(config); /* * The config may be present even if 'error' is non-zero. * In this case we return success, and preserve the real errno * in 'zc_cookie'. */ zc->zc_cookie = error; } else { ret = error; } return (ret); } /* * Try to import the given pool, returning pool stats as appropriate so that * user land knows which devices are available and overall pool health. */ static int zfs_ioc_pool_tryimport(zfs_cmd_t *zc) { nvlist_t *tryconfig, *config; int error; if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, zc->zc_iflags, &tryconfig)) != 0) return (error); config = spa_tryimport(tryconfig); nvlist_free(tryconfig); if (config == NULL) return (SET_ERROR(EINVAL)); error = put_nvlist(zc, config); nvlist_free(config); return (error); } /* * inputs: * zc_name name of the pool * zc_cookie scan func (pool_scan_func_t) */ static int zfs_ioc_pool_scan(zfs_cmd_t *zc) { spa_t *spa; int error; if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) return (error); if (zc->zc_cookie == POOL_SCAN_NONE) error = spa_scan_stop(spa); else error = spa_scan(spa, zc->zc_cookie); spa_close(spa, FTAG); return (error); } static int zfs_ioc_pool_freeze(zfs_cmd_t *zc) { spa_t *spa; int error; error = spa_open(zc->zc_name, &spa, FTAG); if (error == 0) { spa_freeze(spa); spa_close(spa, FTAG); } return (error); } static int zfs_ioc_pool_upgrade(zfs_cmd_t *zc) { spa_t *spa; int error; if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) return (error); if (zc->zc_cookie < spa_version(spa) || !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { spa_close(spa, FTAG); return (SET_ERROR(EINVAL)); } spa_upgrade(spa, zc->zc_cookie); spa_close(spa, FTAG); return (error); } static int zfs_ioc_pool_get_history(zfs_cmd_t *zc) { spa_t *spa; char *hist_buf; uint64_t size; int error; if ((size = zc->zc_history_len) == 0) return (SET_ERROR(EINVAL)); if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) return (error); if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { spa_close(spa, FTAG); return (SET_ERROR(ENOTSUP)); } hist_buf = kmem_alloc(size, KM_SLEEP); if ((error = spa_history_get(spa, &zc->zc_history_offset, &zc->zc_history_len, hist_buf)) == 0) { error = ddi_copyout(hist_buf, (void *)(uintptr_t)zc->zc_history, zc->zc_history_len, zc->zc_iflags); } spa_close(spa, FTAG); kmem_free(hist_buf, size); return (error); } static int zfs_ioc_pool_reguid(zfs_cmd_t *zc) { spa_t *spa; int error; error = spa_open(zc->zc_name, &spa, FTAG); if (error == 0) { error = spa_change_guid(spa); spa_close(spa, FTAG); } return (error); } static int zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) { return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); } /* * inputs: * zc_name name of filesystem * zc_obj object to find * * outputs: * zc_value name of object */ static int zfs_ioc_obj_to_path(zfs_cmd_t *zc) { objset_t *os; int error; /* XXX reading from objset not owned */ if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) return (error); if (dmu_objset_type(os) != DMU_OST_ZFS) { dmu_objset_rele(os, FTAG); return (SET_ERROR(EINVAL)); } error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, sizeof (zc->zc_value)); dmu_objset_rele(os, FTAG); return (error); } /* * inputs: * zc_name name of filesystem * zc_obj object to find * * outputs: * zc_stat stats on object * zc_value path to object */ static int zfs_ioc_obj_to_stats(zfs_cmd_t *zc) { objset_t *os; int error; /* XXX reading from objset not owned */ if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) return (error); if (dmu_objset_type(os) != DMU_OST_ZFS) { dmu_objset_rele(os, FTAG); return (SET_ERROR(EINVAL)); } error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, sizeof (zc->zc_value)); dmu_objset_rele(os, FTAG); return (error); } static int zfs_ioc_vdev_add(zfs_cmd_t *zc) { spa_t *spa; int error; nvlist_t *config, **l2cache, **spares; uint_t nl2cache = 0, nspares = 0; error = spa_open(zc->zc_name, &spa, FTAG); if (error != 0) return (error); error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, zc->zc_iflags, &config); (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache); (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, &spares, &nspares); #ifdef illumos /* * A root pool with concatenated devices is not supported. * Thus, can not add a device to a root pool. * * Intent log device can not be added to a rootpool because * during mountroot, zil is replayed, a seperated log device * can not be accessed during the mountroot time. * * l2cache and spare devices are ok to be added to a rootpool. */ if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { nvlist_free(config); spa_close(spa, FTAG); return (SET_ERROR(EDOM)); } #endif /* illumos */ if (error == 0) { error = spa_vdev_add(spa, config); nvlist_free(config); } spa_close(spa, FTAG); return (error); } /* * inputs: * zc_name name of the pool * zc_nvlist_conf nvlist of devices to remove * zc_cookie to stop the remove? */ static int zfs_ioc_vdev_remove(zfs_cmd_t *zc) { spa_t *spa; int error; error = spa_open(zc->zc_name, &spa, FTAG); if (error != 0) return (error); error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); spa_close(spa, FTAG); return (error); } static int zfs_ioc_vdev_set_state(zfs_cmd_t *zc) { spa_t *spa; int error; vdev_state_t newstate = VDEV_STATE_UNKNOWN; if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) return (error); switch (zc->zc_cookie) { case VDEV_STATE_ONLINE: error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); break; case VDEV_STATE_OFFLINE: error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); break; case VDEV_STATE_FAULTED: if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && zc->zc_obj != VDEV_AUX_EXTERNAL) zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); break; case VDEV_STATE_DEGRADED: if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && zc->zc_obj != VDEV_AUX_EXTERNAL) zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); break; default: error = SET_ERROR(EINVAL); } zc->zc_cookie = newstate; spa_close(spa, FTAG); return (error); } static int zfs_ioc_vdev_attach(zfs_cmd_t *zc) { spa_t *spa; int replacing = zc->zc_cookie; nvlist_t *config; int error; if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) return (error); if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, zc->zc_iflags, &config)) == 0) { error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); nvlist_free(config); } spa_close(spa, FTAG); return (error); } static int zfs_ioc_vdev_detach(zfs_cmd_t *zc) { spa_t *spa; int error; if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) return (error); error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); spa_close(spa, FTAG); return (error); } static int zfs_ioc_vdev_split(zfs_cmd_t *zc) { spa_t *spa; nvlist_t *config, *props = NULL; int error; boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) return (error); if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, zc->zc_iflags, &config)) { spa_close(spa, FTAG); return (error); } if (zc->zc_nvlist_src_size != 0 && (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, zc->zc_iflags, &props))) { spa_close(spa, FTAG); nvlist_free(config); return (error); } error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); spa_close(spa, FTAG); nvlist_free(config); nvlist_free(props); return (error); } static int zfs_ioc_vdev_setpath(zfs_cmd_t *zc) { spa_t *spa; char *path = zc->zc_value; uint64_t guid = zc->zc_guid; int error; error = spa_open(zc->zc_name, &spa, FTAG); if (error != 0) return (error); error = spa_vdev_setpath(spa, guid, path); spa_close(spa, FTAG); return (error); } static int zfs_ioc_vdev_setfru(zfs_cmd_t *zc) { spa_t *spa; char *fru = zc->zc_value; uint64_t guid = zc->zc_guid; int error; error = spa_open(zc->zc_name, &spa, FTAG); if (error != 0) return (error); error = spa_vdev_setfru(spa, guid, fru); spa_close(spa, FTAG); return (error); } static int zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) { int error = 0; nvlist_t *nv; dmu_objset_fast_stat(os, &zc->zc_objset_stats); if (zc->zc_nvlist_dst != 0 && (error = dsl_prop_get_all(os, &nv)) == 0) { dmu_objset_stats(os, nv); /* * NB: zvol_get_stats() will read the objset contents, * which we aren't supposed to do with a * DS_MODE_USER hold, because it could be * inconsistent. So this is a bit of a workaround... * XXX reading with out owning */ if (!zc->zc_objset_stats.dds_inconsistent && dmu_objset_type(os) == DMU_OST_ZVOL) { error = zvol_get_stats(os, nv); if (error == EIO) return (error); VERIFY0(error); } error = put_nvlist(zc, nv); nvlist_free(nv); } return (error); } /* * inputs: * zc_name name of filesystem * zc_nvlist_dst_size size of buffer for property nvlist * * outputs: * zc_objset_stats stats * zc_nvlist_dst property nvlist * zc_nvlist_dst_size size of property nvlist */ static int zfs_ioc_objset_stats(zfs_cmd_t *zc) { objset_t *os; int error; error = dmu_objset_hold(zc->zc_name, FTAG, &os); if (error == 0) { error = zfs_ioc_objset_stats_impl(zc, os); dmu_objset_rele(os, FTAG); } if (error == ENOMEM) error = 0; return (error); } /* * inputs: * zc_name name of filesystem * zc_nvlist_dst_size size of buffer for property nvlist * * outputs: * zc_nvlist_dst received property nvlist * zc_nvlist_dst_size size of received property nvlist * * Gets received properties (distinct from local properties on or after * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from * local property values. */ static int zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) { int error = 0; nvlist_t *nv; /* * Without this check, we would return local property values if the * caller has not already received properties on or after * SPA_VERSION_RECVD_PROPS. */ if (!dsl_prop_get_hasrecvd(zc->zc_name)) return (SET_ERROR(ENOTSUP)); if (zc->zc_nvlist_dst != 0 && (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { error = put_nvlist(zc, nv); nvlist_free(nv); } return (error); } static int nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) { uint64_t value; int error; /* * zfs_get_zplprop() will either find a value or give us * the default value (if there is one). */ if ((error = zfs_get_zplprop(os, prop, &value)) != 0) return (error); VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); return (0); } /* * inputs: * zc_name name of filesystem * zc_nvlist_dst_size size of buffer for zpl property nvlist * * outputs: * zc_nvlist_dst zpl property nvlist * zc_nvlist_dst_size size of zpl property nvlist */ static int zfs_ioc_objset_zplprops(zfs_cmd_t *zc) { objset_t *os; int err; /* XXX reading without owning */ if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) return (err); dmu_objset_fast_stat(os, &zc->zc_objset_stats); /* * NB: nvl_add_zplprop() will read the objset contents, * which we aren't supposed to do with a DS_MODE_USER * hold, because it could be inconsistent. */ if (zc->zc_nvlist_dst != 0 && !zc->zc_objset_stats.dds_inconsistent && dmu_objset_type(os) == DMU_OST_ZFS) { nvlist_t *nv; VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) err = put_nvlist(zc, nv); nvlist_free(nv); } else { err = SET_ERROR(ENOENT); } dmu_objset_rele(os, FTAG); return (err); } boolean_t dataset_name_hidden(const char *name) { /* * Skip over datasets that are not visible in this zone, * internal datasets (which have a $ in their name), and * temporary datasets (which have a % in their name). */ if (strchr(name, '$') != NULL) return (B_TRUE); if (strchr(name, '%') != NULL) return (B_TRUE); if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL)) return (B_TRUE); return (B_FALSE); } /* * inputs: * zc_name name of filesystem * zc_cookie zap cursor * zc_nvlist_dst_size size of buffer for property nvlist * * outputs: * zc_name name of next filesystem * zc_cookie zap cursor * zc_objset_stats stats * zc_nvlist_dst property nvlist * zc_nvlist_dst_size size of property nvlist */ static int zfs_ioc_dataset_list_next(zfs_cmd_t *zc) { objset_t *os; int error; char *p; size_t orig_len = strlen(zc->zc_name); top: if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { if (error == ENOENT) error = SET_ERROR(ESRCH); return (error); } p = strrchr(zc->zc_name, '/'); if (p == NULL || p[1] != '\0') (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); p = zc->zc_name + strlen(zc->zc_name); do { error = dmu_dir_list_next(os, sizeof (zc->zc_name) - (p - zc->zc_name), p, NULL, &zc->zc_cookie); if (error == ENOENT) error = SET_ERROR(ESRCH); } while (error == 0 && dataset_name_hidden(zc->zc_name)); dmu_objset_rele(os, FTAG); /* * If it's an internal dataset (ie. with a '$' in its name), * don't try to get stats for it, otherwise we'll return ENOENT. */ if (error == 0 && strchr(zc->zc_name, '$') == NULL) { error = zfs_ioc_objset_stats(zc); /* fill in the stats */ if (error == ENOENT) { /* We lost a race with destroy, get the next one. */ zc->zc_name[orig_len] = '\0'; goto top; } } return (error); } /* * inputs: * zc_name name of filesystem * zc_cookie zap cursor * zc_nvlist_dst_size size of buffer for property nvlist * zc_simple when set, only name is requested * * outputs: * zc_name name of next snapshot * zc_objset_stats stats * zc_nvlist_dst property nvlist * zc_nvlist_dst_size size of property nvlist */ static int zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) { objset_t *os; int error; error = dmu_objset_hold(zc->zc_name, FTAG, &os); if (error != 0) { return (error == ENOENT ? ESRCH : error); } /* * A dataset name of maximum length cannot have any snapshots, * so exit immediately. */ if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) { dmu_objset_rele(os, FTAG); return (SET_ERROR(ESRCH)); } error = dmu_snapshot_list_next(os, sizeof (zc->zc_name) - strlen(zc->zc_name), zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, NULL); if (error == 0 && !zc->zc_simple) { dsl_dataset_t *ds; dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); if (error == 0) { objset_t *ossnap; error = dmu_objset_from_ds(ds, &ossnap); if (error == 0) error = zfs_ioc_objset_stats_impl(zc, ossnap); dsl_dataset_rele(ds, FTAG); } } else if (error == ENOENT) { error = SET_ERROR(ESRCH); } dmu_objset_rele(os, FTAG); /* if we failed, undo the @ that we tacked on to zc_name */ if (error != 0) *strchr(zc->zc_name, '@') = '\0'; return (error); } static int zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) { const char *propname = nvpair_name(pair); uint64_t *valary; unsigned int vallen; const char *domain; char *dash; zfs_userquota_prop_t type; uint64_t rid; uint64_t quota; zfsvfs_t *zfsvfs; int err; if (nvpair_type(pair) == DATA_TYPE_NVLIST) { nvlist_t *attrs; VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, &pair) != 0) return (SET_ERROR(EINVAL)); } /* * A correctly constructed propname is encoded as * userquota@-. */ if ((dash = strchr(propname, '-')) == NULL || nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || vallen != 3) return (SET_ERROR(EINVAL)); domain = dash + 1; type = valary[0]; rid = valary[1]; quota = valary[2]; err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); if (err == 0) { err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); zfsvfs_rele(zfsvfs, FTAG); } return (err); } /* * If the named property is one that has a special function to set its value, * return 0 on success and a positive error code on failure; otherwise if it is * not one of the special properties handled by this function, return -1. * * XXX: It would be better for callers of the property interface if we handled * these special cases in dsl_prop.c (in the dsl layer). */ static int zfs_prop_set_special(const char *dsname, zprop_source_t source, nvpair_t *pair) { const char *propname = nvpair_name(pair); zfs_prop_t prop = zfs_name_to_prop(propname); uint64_t intval; int err = -1; if (prop == ZPROP_INVAL) { if (zfs_prop_userquota(propname)) return (zfs_prop_set_userquota(dsname, pair)); return (-1); } if (nvpair_type(pair) == DATA_TYPE_NVLIST) { nvlist_t *attrs; VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, &pair) == 0); } if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) return (-1); VERIFY(0 == nvpair_value_uint64(pair, &intval)); switch (prop) { case ZFS_PROP_QUOTA: err = dsl_dir_set_quota(dsname, source, intval); break; case ZFS_PROP_REFQUOTA: err = dsl_dataset_set_refquota(dsname, source, intval); break; case ZFS_PROP_FILESYSTEM_LIMIT: case ZFS_PROP_SNAPSHOT_LIMIT: if (intval == UINT64_MAX) { /* clearing the limit, just do it */ err = 0; } else { err = dsl_dir_activate_fs_ss_limit(dsname); } /* * Set err to -1 to force the zfs_set_prop_nvlist code down the * default path to set the value in the nvlist. */ if (err == 0) err = -1; break; case ZFS_PROP_RESERVATION: err = dsl_dir_set_reservation(dsname, source, intval); break; case ZFS_PROP_REFRESERVATION: err = dsl_dataset_set_refreservation(dsname, source, intval); break; case ZFS_PROP_VOLSIZE: err = zvol_set_volsize(dsname, intval); break; case ZFS_PROP_VERSION: { zfsvfs_t *zfsvfs; if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) break; err = zfs_set_version(zfsvfs, intval); zfsvfs_rele(zfsvfs, FTAG); if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { zfs_cmd_t *zc; zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); (void) strcpy(zc->zc_name, dsname); (void) zfs_ioc_userspace_upgrade(zc); kmem_free(zc, sizeof (zfs_cmd_t)); } break; } default: err = -1; } return (err); } /* * This function is best effort. If it fails to set any of the given properties, * it continues to set as many as it can and returns the last error * encountered. If the caller provides a non-NULL errlist, it will be filled in * with the list of names of all the properties that failed along with the * corresponding error numbers. * * If every property is set successfully, zero is returned and errlist is not * modified. */ int zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, nvlist_t *errlist) { nvpair_t *pair; nvpair_t *propval; int rv = 0; uint64_t intval; char *strval; nvlist_t *genericnvl = fnvlist_alloc(); nvlist_t *retrynvl = fnvlist_alloc(); retry: pair = NULL; while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { const char *propname = nvpair_name(pair); zfs_prop_t prop = zfs_name_to_prop(propname); int err = 0; /* decode the property value */ propval = pair; if (nvpair_type(pair) == DATA_TYPE_NVLIST) { nvlist_t *attrs; attrs = fnvpair_value_nvlist(pair); if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, &propval) != 0) err = SET_ERROR(EINVAL); } /* Validate value type */ if (err == 0 && prop == ZPROP_INVAL) { if (zfs_prop_user(propname)) { if (nvpair_type(propval) != DATA_TYPE_STRING) err = SET_ERROR(EINVAL); } else if (zfs_prop_userquota(propname)) { if (nvpair_type(propval) != DATA_TYPE_UINT64_ARRAY) err = SET_ERROR(EINVAL); } else { err = SET_ERROR(EINVAL); } } else if (err == 0) { if (nvpair_type(propval) == DATA_TYPE_STRING) { if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) err = SET_ERROR(EINVAL); } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { const char *unused; intval = fnvpair_value_uint64(propval); switch (zfs_prop_get_type(prop)) { case PROP_TYPE_NUMBER: break; case PROP_TYPE_STRING: err = SET_ERROR(EINVAL); break; case PROP_TYPE_INDEX: if (zfs_prop_index_to_string(prop, intval, &unused) != 0) err = SET_ERROR(EINVAL); break; default: cmn_err(CE_PANIC, "unknown property type"); } } else { err = SET_ERROR(EINVAL); } } /* Validate permissions */ if (err == 0) err = zfs_check_settable(dsname, pair, CRED()); if (err == 0) { err = zfs_prop_set_special(dsname, source, pair); if (err == -1) { /* * For better performance we build up a list of * properties to set in a single transaction. */ err = nvlist_add_nvpair(genericnvl, pair); } else if (err != 0 && nvl != retrynvl) { /* * This may be a spurious error caused by * receiving quota and reservation out of order. * Try again in a second pass. */ err = nvlist_add_nvpair(retrynvl, pair); } } if (err != 0) { if (errlist != NULL) fnvlist_add_int32(errlist, propname, err); rv = err; } } if (nvl != retrynvl && !nvlist_empty(retrynvl)) { nvl = retrynvl; goto retry; } if (!nvlist_empty(genericnvl) && dsl_props_set(dsname, source, genericnvl) != 0) { /* * If this fails, we still want to set as many properties as we * can, so try setting them individually. */ pair = NULL; while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { const char *propname = nvpair_name(pair); int err = 0; propval = pair; if (nvpair_type(pair) == DATA_TYPE_NVLIST) { nvlist_t *attrs; attrs = fnvpair_value_nvlist(pair); propval = fnvlist_lookup_nvpair(attrs, ZPROP_VALUE); } if (nvpair_type(propval) == DATA_TYPE_STRING) { strval = fnvpair_value_string(propval); err = dsl_prop_set_string(dsname, propname, source, strval); } else { intval = fnvpair_value_uint64(propval); err = dsl_prop_set_int(dsname, propname, source, intval); } if (err != 0) { if (errlist != NULL) { fnvlist_add_int32(errlist, propname, err); } rv = err; } } } nvlist_free(genericnvl); nvlist_free(retrynvl); return (rv); } /* * Check that all the properties are valid user properties. */ static int zfs_check_userprops(const char *fsname, nvlist_t *nvl) { nvpair_t *pair = NULL; int error = 0; while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { const char *propname = nvpair_name(pair); if (!zfs_prop_user(propname) || nvpair_type(pair) != DATA_TYPE_STRING) return (SET_ERROR(EINVAL)); if (error = zfs_secpolicy_write_perms(fsname, ZFS_DELEG_PERM_USERPROP, CRED())) return (error); if (strlen(propname) >= ZAP_MAXNAMELEN) return (SET_ERROR(ENAMETOOLONG)); if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) return (E2BIG); } return (0); } static void props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) { nvpair_t *pair; VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); pair = NULL; while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { if (nvlist_exists(skipped, nvpair_name(pair))) continue; VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); } } static int clear_received_props(const char *dsname, nvlist_t *props, nvlist_t *skipped) { int err = 0; nvlist_t *cleared_props = NULL; props_skip(props, skipped, &cleared_props); if (!nvlist_empty(cleared_props)) { /* * Acts on local properties until the dataset has received * properties at least once on or after SPA_VERSION_RECVD_PROPS. */ zprop_source_t flags = (ZPROP_SRC_NONE | (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); } nvlist_free(cleared_props); return (err); } /* * inputs: * zc_name name of filesystem * zc_value name of property to set * zc_nvlist_src{_size} nvlist of properties to apply * zc_cookie received properties flag * * outputs: * zc_nvlist_dst{_size} error for each unapplied received property */ static int zfs_ioc_set_prop(zfs_cmd_t *zc) { nvlist_t *nvl; boolean_t received = zc->zc_cookie; zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : ZPROP_SRC_LOCAL); nvlist_t *errors; int error; if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, zc->zc_iflags, &nvl)) != 0) return (error); if (received) { nvlist_t *origprops; if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { (void) clear_received_props(zc->zc_name, origprops, nvl); nvlist_free(origprops); } error = dsl_prop_set_hasrecvd(zc->zc_name); } errors = fnvlist_alloc(); if (error == 0) error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); if (zc->zc_nvlist_dst != 0 && errors != NULL) { (void) put_nvlist(zc, errors); } nvlist_free(errors); nvlist_free(nvl); return (error); } /* * inputs: * zc_name name of filesystem * zc_value name of property to inherit * zc_cookie revert to received value if TRUE * * outputs: none */ static int zfs_ioc_inherit_prop(zfs_cmd_t *zc) { const char *propname = zc->zc_value; zfs_prop_t prop = zfs_name_to_prop(propname); boolean_t received = zc->zc_cookie; zprop_source_t source = (received ? ZPROP_SRC_NONE /* revert to received value, if any */ : ZPROP_SRC_INHERITED); /* explicitly inherit */ if (received) { nvlist_t *dummy; nvpair_t *pair; zprop_type_t type; int err; /* * zfs_prop_set_special() expects properties in the form of an * nvpair with type info. */ if (prop == ZPROP_INVAL) { if (!zfs_prop_user(propname)) return (SET_ERROR(EINVAL)); type = PROP_TYPE_STRING; } else if (prop == ZFS_PROP_VOLSIZE || prop == ZFS_PROP_VERSION) { return (SET_ERROR(EINVAL)); } else { type = zfs_prop_get_type(prop); } VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); switch (type) { case PROP_TYPE_STRING: VERIFY(0 == nvlist_add_string(dummy, propname, "")); break; case PROP_TYPE_NUMBER: case PROP_TYPE_INDEX: VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); break; default: nvlist_free(dummy); return (SET_ERROR(EINVAL)); } pair = nvlist_next_nvpair(dummy, NULL); err = zfs_prop_set_special(zc->zc_name, source, pair); nvlist_free(dummy); if (err != -1) return (err); /* special property already handled */ } else { /* * Only check this in the non-received case. We want to allow * 'inherit -S' to revert non-inheritable properties like quota * and reservation to the received or default values even though * they are not considered inheritable. */ if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) return (SET_ERROR(EINVAL)); } /* property name has been validated by zfs_secpolicy_inherit_prop() */ return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); } static int zfs_ioc_pool_set_props(zfs_cmd_t *zc) { nvlist_t *props; spa_t *spa; int error; nvpair_t *pair; if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, zc->zc_iflags, &props)) return (error); /* * If the only property is the configfile, then just do a spa_lookup() * to handle the faulted case. */ pair = nvlist_next_nvpair(props, NULL); if (pair != NULL && strcmp(nvpair_name(pair), zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && nvlist_next_nvpair(props, pair) == NULL) { mutex_enter(&spa_namespace_lock); if ((spa = spa_lookup(zc->zc_name)) != NULL) { spa_configfile_set(spa, props, B_FALSE); spa_config_sync(spa, B_FALSE, B_TRUE); } mutex_exit(&spa_namespace_lock); if (spa != NULL) { nvlist_free(props); return (0); } } if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { nvlist_free(props); return (error); } error = spa_prop_set(spa, props); nvlist_free(props); spa_close(spa, FTAG); return (error); } static int zfs_ioc_pool_get_props(zfs_cmd_t *zc) { spa_t *spa; int error; nvlist_t *nvp = NULL; if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { /* * If the pool is faulted, there may be properties we can still * get (such as altroot and cachefile), so attempt to get them * anyway. */ mutex_enter(&spa_namespace_lock); if ((spa = spa_lookup(zc->zc_name)) != NULL) error = spa_prop_get(spa, &nvp); mutex_exit(&spa_namespace_lock); } else { error = spa_prop_get(spa, &nvp); spa_close(spa, FTAG); } if (error == 0 && zc->zc_nvlist_dst != 0) error = put_nvlist(zc, nvp); else error = SET_ERROR(EFAULT); nvlist_free(nvp); return (error); } /* * inputs: * zc_name name of filesystem * zc_nvlist_src{_size} nvlist of delegated permissions * zc_perm_action allow/unallow flag * * outputs: none */ static int zfs_ioc_set_fsacl(zfs_cmd_t *zc) { int error; nvlist_t *fsaclnv = NULL; if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, zc->zc_iflags, &fsaclnv)) != 0) return (error); /* * Verify nvlist is constructed correctly */ if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { nvlist_free(fsaclnv); return (SET_ERROR(EINVAL)); } /* * If we don't have PRIV_SYS_MOUNT, then validate * that user is allowed to hand out each permission in * the nvlist(s) */ error = secpolicy_zfs(CRED()); if (error != 0) { if (zc->zc_perm_action == B_FALSE) { error = dsl_deleg_can_allow(zc->zc_name, fsaclnv, CRED()); } else { error = dsl_deleg_can_unallow(zc->zc_name, fsaclnv, CRED()); } } if (error == 0) error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); nvlist_free(fsaclnv); return (error); } /* * inputs: * zc_name name of filesystem * * outputs: * zc_nvlist_src{_size} nvlist of delegated permissions */ static int zfs_ioc_get_fsacl(zfs_cmd_t *zc) { nvlist_t *nvp; int error; if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { error = put_nvlist(zc, nvp); nvlist_free(nvp); } return (error); } /* * Search the vfs list for a specified resource. Returns a pointer to it * or NULL if no suitable entry is found. The caller of this routine * is responsible for releasing the returned vfs pointer. */ static vfs_t * zfs_get_vfs(const char *resource) { vfs_t *vfsp; mtx_lock(&mountlist_mtx); TAILQ_FOREACH(vfsp, &mountlist, mnt_list) { if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { VFS_HOLD(vfsp); break; } } mtx_unlock(&mountlist_mtx); return (vfsp); } /* ARGSUSED */ static void zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) { zfs_creat_t *zct = arg; zfs_create_fs(os, cr, zct->zct_zplprops, tx); } #define ZFS_PROP_UNDEFINED ((uint64_t)-1) /* * inputs: * os parent objset pointer (NULL if root fs) * fuids_ok fuids allowed in this version of the spa? * sa_ok SAs allowed in this version of the spa? * createprops list of properties requested by creator * * outputs: * zplprops values for the zplprops we attach to the master node object * is_ci true if requested file system will be purely case-insensitive * * Determine the settings for utf8only, normalization and * casesensitivity. Specific values may have been requested by the * creator and/or we can inherit values from the parent dataset. If * the file system is of too early a vintage, a creator can not * request settings for these properties, even if the requested * setting is the default value. We don't actually want to create dsl * properties for these, so remove them from the source nvlist after * processing. */ static int zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, nvlist_t *zplprops, boolean_t *is_ci) { uint64_t sense = ZFS_PROP_UNDEFINED; uint64_t norm = ZFS_PROP_UNDEFINED; uint64_t u8 = ZFS_PROP_UNDEFINED; ASSERT(zplprops != NULL); /* * Pull out creator prop choices, if any. */ if (createprops) { (void) nvlist_lookup_uint64(createprops, zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); (void) nvlist_lookup_uint64(createprops, zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); (void) nvlist_remove_all(createprops, zfs_prop_to_name(ZFS_PROP_NORMALIZE)); (void) nvlist_lookup_uint64(createprops, zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); (void) nvlist_remove_all(createprops, zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); (void) nvlist_lookup_uint64(createprops, zfs_prop_to_name(ZFS_PROP_CASE), &sense); (void) nvlist_remove_all(createprops, zfs_prop_to_name(ZFS_PROP_CASE)); } /* * If the zpl version requested is whacky or the file system * or pool is version is too "young" to support normalization * and the creator tried to set a value for one of the props, * error out. */ if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || (zplver >= ZPL_VERSION_FUID && !fuids_ok) || (zplver >= ZPL_VERSION_SA && !sa_ok) || (zplver < ZPL_VERSION_NORMALIZATION && (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || sense != ZFS_PROP_UNDEFINED))) return (SET_ERROR(ENOTSUP)); /* * Put the version in the zplprops */ VERIFY(nvlist_add_uint64(zplprops, zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); if (norm == ZFS_PROP_UNDEFINED) VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); VERIFY(nvlist_add_uint64(zplprops, zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); /* * If we're normalizing, names must always be valid UTF-8 strings. */ if (norm) u8 = 1; if (u8 == ZFS_PROP_UNDEFINED) VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); VERIFY(nvlist_add_uint64(zplprops, zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); if (sense == ZFS_PROP_UNDEFINED) VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); VERIFY(nvlist_add_uint64(zplprops, zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); if (is_ci) *is_ci = (sense == ZFS_CASE_INSENSITIVE); return (0); } static int zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, nvlist_t *zplprops, boolean_t *is_ci) { boolean_t fuids_ok, sa_ok; uint64_t zplver = ZPL_VERSION; objset_t *os = NULL; char parentname[MAXNAMELEN]; char *cp; spa_t *spa; uint64_t spa_vers; int error; (void) strlcpy(parentname, dataset, sizeof (parentname)); cp = strrchr(parentname, '/'); ASSERT(cp != NULL); cp[0] = '\0'; if ((error = spa_open(dataset, &spa, FTAG)) != 0) return (error); spa_vers = spa_version(spa); spa_close(spa, FTAG); zplver = zfs_zpl_version_map(spa_vers); fuids_ok = (zplver >= ZPL_VERSION_FUID); sa_ok = (zplver >= ZPL_VERSION_SA); /* * Open parent object set so we can inherit zplprop values. */ if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) return (error); error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, zplprops, is_ci); dmu_objset_rele(os, FTAG); return (error); } static int zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, nvlist_t *zplprops, boolean_t *is_ci) { boolean_t fuids_ok; boolean_t sa_ok; uint64_t zplver = ZPL_VERSION; int error; zplver = zfs_zpl_version_map(spa_vers); fuids_ok = (zplver >= ZPL_VERSION_FUID); sa_ok = (zplver >= ZPL_VERSION_SA); error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, createprops, zplprops, is_ci); return (error); } /* * innvl: { * "type" -> dmu_objset_type_t (int32) * (optional) "props" -> { prop -> value } * } * * outnvl: propname -> error code (int32) */ static int zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) { int error = 0; zfs_creat_t zct = { 0 }; nvlist_t *nvprops = NULL; void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); int32_t type32; dmu_objset_type_t type; boolean_t is_insensitive = B_FALSE; if (nvlist_lookup_int32(innvl, "type", &type32) != 0) return (SET_ERROR(EINVAL)); type = type32; (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); switch (type) { case DMU_OST_ZFS: cbfunc = zfs_create_cb; break; case DMU_OST_ZVOL: cbfunc = zvol_create_cb; break; default: cbfunc = NULL; break; } if (strchr(fsname, '@') || strchr(fsname, '%')) return (SET_ERROR(EINVAL)); zct.zct_props = nvprops; if (cbfunc == NULL) return (SET_ERROR(EINVAL)); if (type == DMU_OST_ZVOL) { uint64_t volsize, volblocksize; if (nvprops == NULL) return (SET_ERROR(EINVAL)); if (nvlist_lookup_uint64(nvprops, zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) return (SET_ERROR(EINVAL)); if ((error = nvlist_lookup_uint64(nvprops, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &volblocksize)) != 0 && error != ENOENT) return (SET_ERROR(EINVAL)); if (error != 0) volblocksize = zfs_prop_default_numeric( ZFS_PROP_VOLBLOCKSIZE); if ((error = zvol_check_volblocksize( volblocksize)) != 0 || (error = zvol_check_volsize(volsize, volblocksize)) != 0) return (error); } else if (type == DMU_OST_ZFS) { int error; /* * We have to have normalization and * case-folding flags correct when we do the * file system creation, so go figure them out * now. */ VERIFY(nvlist_alloc(&zct.zct_zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); error = zfs_fill_zplprops(fsname, nvprops, zct.zct_zplprops, &is_insensitive); if (error != 0) { nvlist_free(zct.zct_zplprops); return (error); } } error = dmu_objset_create(fsname, type, is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); nvlist_free(zct.zct_zplprops); /* * It would be nice to do this atomically. */ if (error == 0) { error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, nvprops, outnvl); if (error != 0) (void) dsl_destroy_head(fsname); } #ifdef __FreeBSD__ if (error == 0 && type == DMU_OST_ZVOL) zvol_create_minors(fsname); #endif return (error); } /* * innvl: { * "origin" -> name of origin snapshot * (optional) "props" -> { prop -> value } * } * * outnvl: propname -> error code (int32) */ static int zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) { int error = 0; nvlist_t *nvprops = NULL; char *origin_name; if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) return (SET_ERROR(EINVAL)); (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); if (strchr(fsname, '@') || strchr(fsname, '%')) return (SET_ERROR(EINVAL)); if (dataset_namecheck(origin_name, NULL, NULL) != 0) return (SET_ERROR(EINVAL)); error = dmu_objset_clone(fsname, origin_name); if (error != 0) return (error); /* * It would be nice to do this atomically. */ if (error == 0) { error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, nvprops, outnvl); if (error != 0) (void) dsl_destroy_head(fsname); } #ifdef __FreeBSD__ if (error == 0) zvol_create_minors(fsname); #endif return (error); } /* * innvl: { * "snaps" -> { snapshot1, snapshot2 } * (optional) "props" -> { prop -> value (string) } * } * * outnvl: snapshot -> error code (int32) */ static int zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) { nvlist_t *snaps; nvlist_t *props = NULL; int error, poollen; nvpair_t *pair; (void) nvlist_lookup_nvlist(innvl, "props", &props); if ((error = zfs_check_userprops(poolname, props)) != 0) return (error); if (!nvlist_empty(props) && zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) return (SET_ERROR(ENOTSUP)); if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) return (SET_ERROR(EINVAL)); poollen = strlen(poolname); for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; pair = nvlist_next_nvpair(snaps, pair)) { const char *name = nvpair_name(pair); const char *cp = strchr(name, '@'); /* * The snap name must contain an @, and the part after it must * contain only valid characters. */ if (cp == NULL || zfs_component_namecheck(cp + 1, NULL, NULL) != 0) return (SET_ERROR(EINVAL)); /* * The snap must be in the specified pool. */ if (strncmp(name, poolname, poollen) != 0 || (name[poollen] != '/' && name[poollen] != '@')) return (SET_ERROR(EXDEV)); /* This must be the only snap of this fs. */ for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { if (strncmp(name, nvpair_name(pair2), cp - name + 1) == 0) { return (SET_ERROR(EXDEV)); } } } error = dsl_dataset_snapshot(snaps, props, outnvl); return (error); } /* * innvl: "message" -> string */ /* ARGSUSED */ static int zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) { char *message; spa_t *spa; int error; char *poolname; /* * The poolname in the ioctl is not set, we get it from the TSD, * which was set at the end of the last successful ioctl that allows * logging. The secpolicy func already checked that it is set. * Only one log ioctl is allowed after each successful ioctl, so * we clear the TSD here. */ poolname = tsd_get(zfs_allow_log_key); (void) tsd_set(zfs_allow_log_key, NULL); error = spa_open(poolname, &spa, FTAG); strfree(poolname); if (error != 0) return (error); if (nvlist_lookup_string(innvl, "message", &message) != 0) { spa_close(spa, FTAG); return (SET_ERROR(EINVAL)); } if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { spa_close(spa, FTAG); return (SET_ERROR(ENOTSUP)); } error = spa_history_log(spa, message); spa_close(spa, FTAG); return (error); } /* * The dp_config_rwlock must not be held when calling this, because the * unmount may need to write out data. * * This function is best-effort. Callers must deal gracefully if it * remains mounted (or is remounted after this call). * * Returns 0 if the argument is not a snapshot, or it is not currently a * filesystem, or we were able to unmount it. Returns error code otherwise. */ int zfs_unmount_snap(const char *snapname) { vfs_t *vfsp; zfsvfs_t *zfsvfs; int err; if (strchr(snapname, '@') == NULL) return (0); vfsp = zfs_get_vfs(snapname); if (vfsp == NULL) return (0); zfsvfs = vfsp->vfs_data; ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); err = vn_vfswlock(vfsp->vfs_vnodecovered); VFS_RELE(vfsp); if (err != 0) return (SET_ERROR(err)); /* * Always force the unmount for snapshots. */ #ifdef illumos (void) dounmount(vfsp, MS_FORCE, kcred); #else vfs_ref(vfsp); (void) dounmount(vfsp, MS_FORCE, curthread); #endif return (0); } /* ARGSUSED */ static int zfs_unmount_snap_cb(const char *snapname, void *arg) { return (zfs_unmount_snap(snapname)); } /* * When a clone is destroyed, its origin may also need to be destroyed, * in which case it must be unmounted. This routine will do that unmount * if necessary. */ void zfs_destroy_unmount_origin(const char *fsname) { int error; objset_t *os; dsl_dataset_t *ds; error = dmu_objset_hold(fsname, FTAG, &os); if (error != 0) return; ds = dmu_objset_ds(os); if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { char originname[MAXNAMELEN]; dsl_dataset_name(ds->ds_prev, originname); dmu_objset_rele(os, FTAG); (void) zfs_unmount_snap(originname); } else { dmu_objset_rele(os, FTAG); } } /* * innvl: { * "snaps" -> { snapshot1, snapshot2 } * (optional boolean) "defer" * } * * outnvl: snapshot -> error code (int32) * */ /* ARGSUSED */ static int zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) { int error, poollen; nvlist_t *snaps; nvpair_t *pair; boolean_t defer; if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) return (SET_ERROR(EINVAL)); defer = nvlist_exists(innvl, "defer"); poollen = strlen(poolname); for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; pair = nvlist_next_nvpair(snaps, pair)) { const char *name = nvpair_name(pair); /* * The snap must be in the specified pool to prevent the * invalid removal of zvol minors below. */ if (strncmp(name, poolname, poollen) != 0 || (name[poollen] != '/' && name[poollen] != '@')) return (SET_ERROR(EXDEV)); error = zfs_unmount_snap(name); if (error != 0) return (error); #if defined(__FreeBSD__) zvol_remove_minors(name); #endif } return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); } /* * Create bookmarks. Bookmark names are of the form #. * All bookmarks must be in the same pool. * * innvl: { * bookmark1 -> snapshot1, bookmark2 -> snapshot2 * } * * outnvl: bookmark -> error code (int32) * */ /* ARGSUSED */ static int zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) { for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { char *snap_name; /* * Verify the snapshot argument. */ if (nvpair_value_string(pair, &snap_name) != 0) return (SET_ERROR(EINVAL)); /* Verify that the keys (bookmarks) are unique */ for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) return (SET_ERROR(EINVAL)); } } return (dsl_bookmark_create(innvl, outnvl)); } /* * innvl: { * property 1, property 2, ... * } * * outnvl: { * bookmark name 1 -> { property 1, property 2, ... }, * bookmark name 2 -> { property 1, property 2, ... } * } * */ static int zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) { return (dsl_get_bookmarks(fsname, innvl, outnvl)); } /* * innvl: { * bookmark name 1, bookmark name 2 * } * * outnvl: bookmark -> error code (int32) * */ static int zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) { int error, poollen; poollen = strlen(poolname); for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { const char *name = nvpair_name(pair); const char *cp = strchr(name, '#'); /* * The bookmark name must contain an #, and the part after it * must contain only valid characters. */ if (cp == NULL || zfs_component_namecheck(cp + 1, NULL, NULL) != 0) return (SET_ERROR(EINVAL)); /* * The bookmark must be in the specified pool. */ if (strncmp(name, poolname, poollen) != 0 || (name[poollen] != '/' && name[poollen] != '#')) return (SET_ERROR(EXDEV)); } error = dsl_bookmark_destroy(innvl, outnvl); return (error); } /* * inputs: * zc_name name of dataset to destroy * zc_objset_type type of objset * zc_defer_destroy mark for deferred destroy * * outputs: none */ static int zfs_ioc_destroy(zfs_cmd_t *zc) { int err; if (zc->zc_objset_type == DMU_OST_ZFS) { err = zfs_unmount_snap(zc->zc_name); if (err != 0) return (err); } if (strchr(zc->zc_name, '@')) err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); else err = dsl_destroy_head(zc->zc_name); if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) #ifdef __FreeBSD__ zvol_remove_minors(zc->zc_name); #else (void) zvol_remove_minor(zc->zc_name); #endif return (err); } /* * fsname is name of dataset to rollback (to most recent snapshot) * * innvl is not used. * * outnvl: "target" -> name of most recent snapshot * } */ /* ARGSUSED */ static int zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) { zfsvfs_t *zfsvfs; int error; if (getzfsvfs(fsname, &zfsvfs) == 0) { error = zfs_suspend_fs(zfsvfs); if (error == 0) { int resume_err; error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); resume_err = zfs_resume_fs(zfsvfs, fsname); error = error ? error : resume_err; } VFS_RELE(zfsvfs->z_vfs); } else { error = dsl_dataset_rollback(fsname, NULL, outnvl); } return (error); } static int recursive_unmount(const char *fsname, void *arg) { const char *snapname = arg; char fullname[MAXNAMELEN]; (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); return (zfs_unmount_snap(fullname)); } /* * inputs: * zc_name old name of dataset * zc_value new name of dataset * zc_cookie recursive flag (only valid for snapshots) * * outputs: none */ static int zfs_ioc_rename(zfs_cmd_t *zc) { boolean_t recursive = zc->zc_cookie & 1; #ifdef __FreeBSD__ boolean_t allow_mounted = zc->zc_cookie & 2; #endif char *at; zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || strchr(zc->zc_value, '%')) return (SET_ERROR(EINVAL)); at = strchr(zc->zc_name, '@'); if (at != NULL) { /* snaps must be in same fs */ int error; if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) return (SET_ERROR(EXDEV)); *at = '\0'; #ifdef illumos if (zc->zc_objset_type == DMU_OST_ZFS) { #else if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) { #endif error = dmu_objset_find(zc->zc_name, recursive_unmount, at + 1, recursive ? DS_FIND_CHILDREN : 0); if (error != 0) { *at = '@'; return (error); } } error = dsl_dataset_rename_snapshot(zc->zc_name, at + 1, strchr(zc->zc_value, '@') + 1, recursive); *at = '@'; return (error); } else { #ifdef illumos if (zc->zc_objset_type == DMU_OST_ZVOL) (void) zvol_remove_minor(zc->zc_name); #endif return (dsl_dir_rename(zc->zc_name, zc->zc_value)); } } static int zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) { const char *propname = nvpair_name(pair); boolean_t issnap = (strchr(dsname, '@') != NULL); zfs_prop_t prop = zfs_name_to_prop(propname); uint64_t intval; int err; if (prop == ZPROP_INVAL) { if (zfs_prop_user(propname)) { if (err = zfs_secpolicy_write_perms(dsname, ZFS_DELEG_PERM_USERPROP, cr)) return (err); return (0); } if (!issnap && zfs_prop_userquota(propname)) { const char *perm = NULL; const char *uq_prefix = zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; const char *gq_prefix = zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; if (strncmp(propname, uq_prefix, strlen(uq_prefix)) == 0) { perm = ZFS_DELEG_PERM_USERQUOTA; } else if (strncmp(propname, gq_prefix, strlen(gq_prefix)) == 0) { perm = ZFS_DELEG_PERM_GROUPQUOTA; } else { /* USERUSED and GROUPUSED are read-only */ return (SET_ERROR(EINVAL)); } if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) return (err); return (0); } return (SET_ERROR(EINVAL)); } if (issnap) return (SET_ERROR(EINVAL)); if (nvpair_type(pair) == DATA_TYPE_NVLIST) { /* * dsl_prop_get_all_impl() returns properties in this * format. */ nvlist_t *attrs; VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, &pair) == 0); } /* * Check that this value is valid for this pool version */ switch (prop) { case ZFS_PROP_COMPRESSION: /* * If the user specified gzip compression, make sure * the SPA supports it. We ignore any errors here since * we'll catch them later. */ if (nvpair_value_uint64(pair, &intval) == 0) { if (intval >= ZIO_COMPRESS_GZIP_1 && intval <= ZIO_COMPRESS_GZIP_9 && zfs_earlier_version(dsname, SPA_VERSION_GZIP_COMPRESSION)) { return (SET_ERROR(ENOTSUP)); } if (intval == ZIO_COMPRESS_ZLE && zfs_earlier_version(dsname, SPA_VERSION_ZLE_COMPRESSION)) return (SET_ERROR(ENOTSUP)); if (intval == ZIO_COMPRESS_LZ4) { spa_t *spa; if ((err = spa_open(dsname, &spa, FTAG)) != 0) return (err); if (!spa_feature_is_enabled(spa, SPA_FEATURE_LZ4_COMPRESS)) { spa_close(spa, FTAG); return (SET_ERROR(ENOTSUP)); } spa_close(spa, FTAG); } /* * If this is a bootable dataset then * verify that the compression algorithm * is supported for booting. We must return * something other than ENOTSUP since it * implies a downrev pool version. */ if (zfs_is_bootfs(dsname) && !BOOTFS_COMPRESS_VALID(intval)) { return (SET_ERROR(ERANGE)); } } break; case ZFS_PROP_COPIES: if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) return (SET_ERROR(ENOTSUP)); break; - case ZFS_PROP_DEDUP: - if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) - return (SET_ERROR(ENOTSUP)); - break; - case ZFS_PROP_RECORDSIZE: /* Record sizes above 128k need the feature to be enabled */ if (nvpair_value_uint64(pair, &intval) == 0 && intval > SPA_OLD_MAXBLOCKSIZE) { spa_t *spa; /* * If this is a bootable dataset then * the we don't allow large (>128K) blocks, * because GRUB doesn't support them. */ if (zfs_is_bootfs(dsname) && intval > SPA_OLD_MAXBLOCKSIZE) { return (SET_ERROR(ERANGE)); } /* * We don't allow setting the property above 1MB, * unless the tunable has been changed. */ if (intval > zfs_max_recordsize || intval > SPA_MAXBLOCKSIZE) return (SET_ERROR(ERANGE)); if ((err = spa_open(dsname, &spa, FTAG)) != 0) return (err); if (!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) { spa_close(spa, FTAG); return (SET_ERROR(ENOTSUP)); } spa_close(spa, FTAG); } break; case ZFS_PROP_SHARESMB: if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) return (SET_ERROR(ENOTSUP)); break; case ZFS_PROP_ACLINHERIT: if (nvpair_type(pair) == DATA_TYPE_UINT64 && nvpair_value_uint64(pair, &intval) == 0) { if (intval == ZFS_ACL_PASSTHROUGH_X && zfs_earlier_version(dsname, SPA_VERSION_PASSTHROUGH_X)) return (SET_ERROR(ENOTSUP)); } break; + + case ZFS_PROP_CHECKSUM: + case ZFS_PROP_DEDUP: + { + spa_feature_t feature; + spa_t *spa; + + /* dedup feature version checks */ + if (prop == ZFS_PROP_DEDUP && + zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) + return (SET_ERROR(ENOTSUP)); + + if (nvpair_value_uint64(pair, &intval) != 0) + return (SET_ERROR(EINVAL)); + + /* check prop value is enabled in features */ + feature = zio_checksum_to_feature(intval); + if (feature == SPA_FEATURE_NONE) + break; + + if ((err = spa_open(dsname, &spa, FTAG)) != 0) + return (err); + /* + * Salted checksums are not supported on root pools. + */ + if (spa_bootfs(spa) != 0 && + intval < ZIO_CHECKSUM_FUNCTIONS && + (zio_checksum_table[intval].ci_flags & + ZCHECKSUM_FLAG_SALTED)) { + spa_close(spa, FTAG); + return (SET_ERROR(ERANGE)); + } + if (!spa_feature_is_enabled(spa, feature)) { + spa_close(spa, FTAG); + return (SET_ERROR(ENOTSUP)); + } + spa_close(spa, FTAG); + break; + } } return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); } /* * Checks for a race condition to make sure we don't increment a feature flag * multiple times. */ static int zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) { spa_t *spa = dmu_tx_pool(tx)->dp_spa; spa_feature_t *featurep = arg; if (!spa_feature_is_active(spa, *featurep)) return (0); else return (SET_ERROR(EBUSY)); } /* * The callback invoked on feature activation in the sync task caused by * zfs_prop_activate_feature. */ static void zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) { spa_t *spa = dmu_tx_pool(tx)->dp_spa; spa_feature_t *featurep = arg; spa_feature_incr(spa, *featurep, tx); } /* * Activates a feature on a pool in response to a property setting. This * creates a new sync task which modifies the pool to reflect the feature * as being active. */ static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) { int err; /* EBUSY here indicates that the feature is already active */ err = dsl_sync_task(spa_name(spa), zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, &feature, 2, ZFS_SPACE_CHECK_RESERVED); if (err != 0 && err != EBUSY) return (err); else return (0); } /* * Removes properties from the given props list that fail permission checks * needed to clear them and to restore them in case of a receive error. For each * property, make sure we have both set and inherit permissions. * * Returns the first error encountered if any permission checks fail. If the * caller provides a non-NULL errlist, it also gives the complete list of names * of all the properties that failed a permission check along with the * corresponding error numbers. The caller is responsible for freeing the * returned errlist. * * If every property checks out successfully, zero is returned and the list * pointed at by errlist is NULL. */ static int zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) { zfs_cmd_t *zc; nvpair_t *pair, *next_pair; nvlist_t *errors; int err, rv = 0; if (props == NULL) return (0); VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); (void) strcpy(zc->zc_name, dataset); pair = nvlist_next_nvpair(props, NULL); while (pair != NULL) { next_pair = nvlist_next_nvpair(props, pair); (void) strcpy(zc->zc_value, nvpair_name(pair)); if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { VERIFY(nvlist_remove_nvpair(props, pair) == 0); VERIFY(nvlist_add_int32(errors, zc->zc_value, err) == 0); } pair = next_pair; } kmem_free(zc, sizeof (zfs_cmd_t)); if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { nvlist_free(errors); errors = NULL; } else { VERIFY(nvpair_value_int32(pair, &rv) == 0); } if (errlist == NULL) nvlist_free(errors); else *errlist = errors; return (rv); } static boolean_t propval_equals(nvpair_t *p1, nvpair_t *p2) { if (nvpair_type(p1) == DATA_TYPE_NVLIST) { /* dsl_prop_get_all_impl() format */ nvlist_t *attrs; VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, &p1) == 0); } if (nvpair_type(p2) == DATA_TYPE_NVLIST) { nvlist_t *attrs; VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, &p2) == 0); } if (nvpair_type(p1) != nvpair_type(p2)) return (B_FALSE); if (nvpair_type(p1) == DATA_TYPE_STRING) { char *valstr1, *valstr2; VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); return (strcmp(valstr1, valstr2) == 0); } else { uint64_t intval1, intval2; VERIFY(nvpair_value_uint64(p1, &intval1) == 0); VERIFY(nvpair_value_uint64(p2, &intval2) == 0); return (intval1 == intval2); } } /* * Remove properties from props if they are not going to change (as determined * by comparison with origprops). Remove them from origprops as well, since we * do not need to clear or restore properties that won't change. */ static void props_reduce(nvlist_t *props, nvlist_t *origprops) { nvpair_t *pair, *next_pair; if (origprops == NULL) return; /* all props need to be received */ pair = nvlist_next_nvpair(props, NULL); while (pair != NULL) { const char *propname = nvpair_name(pair); nvpair_t *match; next_pair = nvlist_next_nvpair(props, pair); if ((nvlist_lookup_nvpair(origprops, propname, &match) != 0) || !propval_equals(pair, match)) goto next; /* need to set received value */ /* don't clear the existing received value */ (void) nvlist_remove_nvpair(origprops, match); /* don't bother receiving the property */ (void) nvlist_remove_nvpair(props, pair); next: pair = next_pair; } } #ifdef DEBUG static boolean_t zfs_ioc_recv_inject_err; #endif /* * inputs: * zc_name name of containing filesystem * zc_nvlist_src{_size} nvlist of properties to apply * zc_value name of snapshot to create * zc_string name of clone origin (if DRR_FLAG_CLONE) * zc_cookie file descriptor to recv from * zc_begin_record the BEGIN record of the stream (not byteswapped) * zc_guid force flag * zc_cleanup_fd cleanup-on-exit file descriptor * zc_action_handle handle for this guid/ds mapping (or zero on first call) * zc_resumable if data is incomplete assume sender will resume * * outputs: * zc_cookie number of bytes read * zc_nvlist_dst{_size} error for each unapplied received property * zc_obj zprop_errflags_t * zc_action_handle handle for this guid/ds mapping */ static int zfs_ioc_recv(zfs_cmd_t *zc) { file_t *fp; dmu_recv_cookie_t drc; boolean_t force = (boolean_t)zc->zc_guid; int fd; int error = 0; int props_error = 0; nvlist_t *errors; offset_t off; nvlist_t *props = NULL; /* sent properties */ nvlist_t *origprops = NULL; /* existing properties */ char *origin = NULL; char *tosnap; char tofs[ZFS_MAXNAMELEN]; cap_rights_t rights; boolean_t first_recvd_props = B_FALSE; if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || strchr(zc->zc_value, '@') == NULL || strchr(zc->zc_value, '%')) return (SET_ERROR(EINVAL)); (void) strcpy(tofs, zc->zc_value); tosnap = strchr(tofs, '@'); *tosnap++ = '\0'; if (zc->zc_nvlist_src != 0 && (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, zc->zc_iflags, &props)) != 0) return (error); fd = zc->zc_cookie; #ifdef illumos fp = getf(fd); #else fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp); #endif if (fp == NULL) { nvlist_free(props); return (SET_ERROR(EBADF)); } errors = fnvlist_alloc(); if (zc->zc_string[0]) origin = zc->zc_string; error = dmu_recv_begin(tofs, tosnap, &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc); if (error != 0) goto out; /* * Set properties before we receive the stream so that they are applied * to the new data. Note that we must call dmu_recv_stream() if * dmu_recv_begin() succeeds. */ if (props != NULL && !drc.drc_newfs) { if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= SPA_VERSION_RECVD_PROPS && !dsl_prop_get_hasrecvd(tofs)) first_recvd_props = B_TRUE; /* * If new received properties are supplied, they are to * completely replace the existing received properties, so stash * away the existing ones. */ if (dsl_prop_get_received(tofs, &origprops) == 0) { nvlist_t *errlist = NULL; /* * Don't bother writing a property if its value won't * change (and avoid the unnecessary security checks). * * The first receive after SPA_VERSION_RECVD_PROPS is a * special case where we blow away all local properties * regardless. */ if (!first_recvd_props) props_reduce(props, origprops); if (zfs_check_clearable(tofs, origprops, &errlist) != 0) (void) nvlist_merge(errors, errlist, 0); nvlist_free(errlist); if (clear_received_props(tofs, origprops, first_recvd_props ? NULL : props) != 0) zc->zc_obj |= ZPROP_ERR_NOCLEAR; } else { zc->zc_obj |= ZPROP_ERR_NOCLEAR; } } if (props != NULL) { props_error = dsl_prop_set_hasrecvd(tofs); if (props_error == 0) { (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, props, errors); } } if (zc->zc_nvlist_dst_size != 0 && (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || put_nvlist(zc, errors) != 0)) { /* * Caller made zc->zc_nvlist_dst less than the minimum expected * size or supplied an invalid address. */ props_error = SET_ERROR(EINVAL); } off = fp->f_offset; error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, &zc->zc_action_handle); if (error == 0) { zfsvfs_t *zfsvfs = NULL; if (getzfsvfs(tofs, &zfsvfs) == 0) { /* online recv */ int end_err; error = zfs_suspend_fs(zfsvfs); /* * If the suspend fails, then the recv_end will * likely also fail, and clean up after itself. */ end_err = dmu_recv_end(&drc, zfsvfs); if (error == 0) error = zfs_resume_fs(zfsvfs, tofs); error = error ? error : end_err; VFS_RELE(zfsvfs->z_vfs); } else { error = dmu_recv_end(&drc, NULL); } } zc->zc_cookie = off - fp->f_offset; if (off >= 0 && off <= MAXOFFSET_T) fp->f_offset = off; #ifdef DEBUG if (zfs_ioc_recv_inject_err) { zfs_ioc_recv_inject_err = B_FALSE; error = 1; } #endif #ifdef __FreeBSD__ if (error == 0) zvol_create_minors(tofs); #endif /* * On error, restore the original props. */ if (error != 0 && props != NULL && !drc.drc_newfs) { if (clear_received_props(tofs, props, NULL) != 0) { /* * We failed to clear the received properties. * Since we may have left a $recvd value on the * system, we can't clear the $hasrecvd flag. */ zc->zc_obj |= ZPROP_ERR_NORESTORE; } else if (first_recvd_props) { dsl_prop_unset_hasrecvd(tofs); } if (origprops == NULL && !drc.drc_newfs) { /* We failed to stash the original properties. */ zc->zc_obj |= ZPROP_ERR_NORESTORE; } /* * dsl_props_set() will not convert RECEIVED to LOCAL on or * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL * explictly if we're restoring local properties cleared in the * first new-style receive. */ if (origprops != NULL && zfs_set_prop_nvlist(tofs, (first_recvd_props ? ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), origprops, NULL) != 0) { /* * We stashed the original properties but failed to * restore them. */ zc->zc_obj |= ZPROP_ERR_NORESTORE; } } out: nvlist_free(props); nvlist_free(origprops); nvlist_free(errors); releasef(fd); if (error == 0) error = props_error; return (error); } /* * inputs: * zc_name name of snapshot to send * zc_cookie file descriptor to send stream to * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) * zc_sendobj objsetid of snapshot to send * zc_fromobj objsetid of incremental fromsnap (may be zero) * zc_guid if set, estimate size of stream only. zc_cookie is ignored. * output size in zc_objset_type. * zc_flags lzc_send_flags * * outputs: * zc_objset_type estimated size, if zc_guid is set */ static int zfs_ioc_send(zfs_cmd_t *zc) { int error; offset_t off; boolean_t estimate = (zc->zc_guid != 0); boolean_t embedok = (zc->zc_flags & 0x1); boolean_t large_block_ok = (zc->zc_flags & 0x2); if (zc->zc_obj != 0) { dsl_pool_t *dp; dsl_dataset_t *tosnap; error = dsl_pool_hold(zc->zc_name, FTAG, &dp); if (error != 0) return (error); error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); if (error != 0) { dsl_pool_rele(dp, FTAG); return (error); } if (dsl_dir_is_clone(tosnap->ds_dir)) zc->zc_fromobj = dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj; dsl_dataset_rele(tosnap, FTAG); dsl_pool_rele(dp, FTAG); } if (estimate) { dsl_pool_t *dp; dsl_dataset_t *tosnap; dsl_dataset_t *fromsnap = NULL; error = dsl_pool_hold(zc->zc_name, FTAG, &dp); if (error != 0) return (error); error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); if (error != 0) { dsl_pool_rele(dp, FTAG); return (error); } if (zc->zc_fromobj != 0) { error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, FTAG, &fromsnap); if (error != 0) { dsl_dataset_rele(tosnap, FTAG); dsl_pool_rele(dp, FTAG); return (error); } } error = dmu_send_estimate(tosnap, fromsnap, &zc->zc_objset_type); if (fromsnap != NULL) dsl_dataset_rele(fromsnap, FTAG); dsl_dataset_rele(tosnap, FTAG); dsl_pool_rele(dp, FTAG); } else { file_t *fp; cap_rights_t rights; #ifdef illumos fp = getf(zc->zc_cookie); #else fget_write(curthread, zc->zc_cookie, cap_rights_init(&rights, CAP_WRITE), &fp); #endif if (fp == NULL) return (SET_ERROR(EBADF)); off = fp->f_offset; error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, zc->zc_fromobj, embedok, large_block_ok, #ifdef illumos zc->zc_cookie, fp->f_vnode, &off); #else zc->zc_cookie, fp, &off); #endif if (off >= 0 && off <= MAXOFFSET_T) fp->f_offset = off; releasef(zc->zc_cookie); } return (error); } /* * inputs: * zc_name name of snapshot on which to report progress * zc_cookie file descriptor of send stream * * outputs: * zc_cookie number of bytes written in send stream thus far */ static int zfs_ioc_send_progress(zfs_cmd_t *zc) { dsl_pool_t *dp; dsl_dataset_t *ds; dmu_sendarg_t *dsp = NULL; int error; error = dsl_pool_hold(zc->zc_name, FTAG, &dp); if (error != 0) return (error); error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); if (error != 0) { dsl_pool_rele(dp, FTAG); return (error); } mutex_enter(&ds->ds_sendstream_lock); /* * Iterate over all the send streams currently active on this dataset. * If there's one which matches the specified file descriptor _and_ the * stream was started by the current process, return the progress of * that stream. */ for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; dsp = list_next(&ds->ds_sendstreams, dsp)) { if (dsp->dsa_outfd == zc->zc_cookie && dsp->dsa_proc == curproc) break; } if (dsp != NULL) zc->zc_cookie = *(dsp->dsa_off); else error = SET_ERROR(ENOENT); mutex_exit(&ds->ds_sendstream_lock); dsl_dataset_rele(ds, FTAG); dsl_pool_rele(dp, FTAG); return (error); } static int zfs_ioc_inject_fault(zfs_cmd_t *zc) { int id, error; error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, &zc->zc_inject_record); if (error == 0) zc->zc_guid = (uint64_t)id; return (error); } static int zfs_ioc_clear_fault(zfs_cmd_t *zc) { return (zio_clear_fault((int)zc->zc_guid)); } static int zfs_ioc_inject_list_next(zfs_cmd_t *zc) { int id = (int)zc->zc_guid; int error; error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), &zc->zc_inject_record); zc->zc_guid = id; return (error); } static int zfs_ioc_error_log(zfs_cmd_t *zc) { spa_t *spa; int error; size_t count = (size_t)zc->zc_nvlist_dst_size; if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) return (error); error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, &count); if (error == 0) zc->zc_nvlist_dst_size = count; else zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); spa_close(spa, FTAG); return (error); } static int zfs_ioc_clear(zfs_cmd_t *zc) { spa_t *spa; vdev_t *vd; int error; /* * On zpool clear we also fix up missing slogs */ mutex_enter(&spa_namespace_lock); spa = spa_lookup(zc->zc_name); if (spa == NULL) { mutex_exit(&spa_namespace_lock); return (SET_ERROR(EIO)); } if (spa_get_log_state(spa) == SPA_LOG_MISSING) { /* we need to let spa_open/spa_load clear the chains */ spa_set_log_state(spa, SPA_LOG_CLEAR); } spa->spa_last_open_failed = 0; mutex_exit(&spa_namespace_lock); if (zc->zc_cookie & ZPOOL_NO_REWIND) { error = spa_open(zc->zc_name, &spa, FTAG); } else { nvlist_t *policy; nvlist_t *config = NULL; if (zc->zc_nvlist_src == 0) return (SET_ERROR(EINVAL)); if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { error = spa_open_rewind(zc->zc_name, &spa, FTAG, policy, &config); if (config != NULL) { int err; if ((err = put_nvlist(zc, config)) != 0) error = err; nvlist_free(config); } nvlist_free(policy); } } if (error != 0) return (error); spa_vdev_state_enter(spa, SCL_NONE); if (zc->zc_guid == 0) { vd = NULL; } else { vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); if (vd == NULL) { (void) spa_vdev_state_exit(spa, NULL, ENODEV); spa_close(spa, FTAG); return (SET_ERROR(ENODEV)); } } vdev_clear(spa, vd); (void) spa_vdev_state_exit(spa, NULL, 0); /* * Resume any suspended I/Os. */ if (zio_resume(spa) != 0) error = SET_ERROR(EIO); spa_close(spa, FTAG); return (error); } static int zfs_ioc_pool_reopen(zfs_cmd_t *zc) { spa_t *spa; int error; error = spa_open(zc->zc_name, &spa, FTAG); if (error != 0) return (error); spa_vdev_state_enter(spa, SCL_NONE); /* * If a resilver is already in progress then set the * spa_scrub_reopen flag to B_TRUE so that we don't restart * the scan as a side effect of the reopen. Otherwise, let * vdev_open() decided if a resilver is required. */ spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); vdev_reopen(spa->spa_root_vdev); spa->spa_scrub_reopen = B_FALSE; (void) spa_vdev_state_exit(spa, NULL, 0); spa_close(spa, FTAG); return (0); } /* * inputs: * zc_name name of filesystem * zc_value name of origin snapshot * * outputs: * zc_string name of conflicting snapshot, if there is one */ static int zfs_ioc_promote(zfs_cmd_t *zc) { char *cp; /* * We don't need to unmount *all* the origin fs's snapshots, but * it's easier. */ cp = strchr(zc->zc_value, '@'); if (cp) *cp = '\0'; (void) dmu_objset_find(zc->zc_value, zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); } /* * Retrieve a single {user|group}{used|quota}@... property. * * inputs: * zc_name name of filesystem * zc_objset_type zfs_userquota_prop_t * zc_value domain name (eg. "S-1-234-567-89") * zc_guid RID/UID/GID * * outputs: * zc_cookie property value */ static int zfs_ioc_userspace_one(zfs_cmd_t *zc) { zfsvfs_t *zfsvfs; int error; if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) return (SET_ERROR(EINVAL)); error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); if (error != 0) return (error); error = zfs_userspace_one(zfsvfs, zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); zfsvfs_rele(zfsvfs, FTAG); return (error); } /* * inputs: * zc_name name of filesystem * zc_cookie zap cursor * zc_objset_type zfs_userquota_prop_t * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) * * outputs: * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) * zc_cookie zap cursor */ static int zfs_ioc_userspace_many(zfs_cmd_t *zc) { zfsvfs_t *zfsvfs; int bufsize = zc->zc_nvlist_dst_size; if (bufsize <= 0) return (SET_ERROR(ENOMEM)); int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); if (error != 0) return (error); void *buf = kmem_alloc(bufsize, KM_SLEEP); error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, buf, &zc->zc_nvlist_dst_size); if (error == 0) { error = ddi_copyout(buf, (void *)(uintptr_t)zc->zc_nvlist_dst, zc->zc_nvlist_dst_size, zc->zc_iflags); } kmem_free(buf, bufsize); zfsvfs_rele(zfsvfs, FTAG); return (error); } /* * inputs: * zc_name name of filesystem * * outputs: * none */ static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) { objset_t *os; int error = 0; zfsvfs_t *zfsvfs; if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { /* * If userused is not enabled, it may be because the * objset needs to be closed & reopened (to grow the * objset_phys_t). Suspend/resume the fs will do that. */ error = zfs_suspend_fs(zfsvfs); if (error == 0) { dmu_objset_refresh_ownership(zfsvfs->z_os, zfsvfs); error = zfs_resume_fs(zfsvfs, zc->zc_name); } } if (error == 0) error = dmu_objset_userspace_upgrade(zfsvfs->z_os); VFS_RELE(zfsvfs->z_vfs); } else { /* XXX kind of reading contents without owning */ error = dmu_objset_hold(zc->zc_name, FTAG, &os); if (error != 0) return (error); error = dmu_objset_userspace_upgrade(os); dmu_objset_rele(os, FTAG); } return (error); } #ifdef sun /* * We don't want to have a hard dependency * against some special symbols in sharefs * nfs, and smbsrv. Determine them if needed when * the first file system is shared. * Neither sharefs, nfs or smbsrv are unloadable modules. */ int (*znfsexport_fs)(void *arg); int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); int (*zsmbexport_fs)(void *arg, boolean_t add_share); int zfs_nfsshare_inited; int zfs_smbshare_inited; ddi_modhandle_t nfs_mod; ddi_modhandle_t sharefs_mod; ddi_modhandle_t smbsrv_mod; #endif /* sun */ kmutex_t zfs_share_lock; #ifdef sun static int zfs_init_sharefs() { int error; ASSERT(MUTEX_HELD(&zfs_share_lock)); /* Both NFS and SMB shares also require sharetab support. */ if (sharefs_mod == NULL && ((sharefs_mod = ddi_modopen("fs/sharefs", KRTLD_MODE_FIRST, &error)) == NULL)) { return (SET_ERROR(ENOSYS)); } if (zshare_fs == NULL && ((zshare_fs = (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { return (SET_ERROR(ENOSYS)); } return (0); } #endif /* sun */ static int zfs_ioc_share(zfs_cmd_t *zc) { #ifdef sun int error; int opcode; switch (zc->zc_share.z_sharetype) { case ZFS_SHARE_NFS: case ZFS_UNSHARE_NFS: if (zfs_nfsshare_inited == 0) { mutex_enter(&zfs_share_lock); if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", KRTLD_MODE_FIRST, &error)) == NULL)) { mutex_exit(&zfs_share_lock); return (SET_ERROR(ENOSYS)); } if (znfsexport_fs == NULL && ((znfsexport_fs = (int (*)(void *)) ddi_modsym(nfs_mod, "nfs_export", &error)) == NULL)) { mutex_exit(&zfs_share_lock); return (SET_ERROR(ENOSYS)); } error = zfs_init_sharefs(); if (error != 0) { mutex_exit(&zfs_share_lock); return (SET_ERROR(ENOSYS)); } zfs_nfsshare_inited = 1; mutex_exit(&zfs_share_lock); } break; case ZFS_SHARE_SMB: case ZFS_UNSHARE_SMB: if (zfs_smbshare_inited == 0) { mutex_enter(&zfs_share_lock); if (smbsrv_mod == NULL && ((smbsrv_mod = ddi_modopen("drv/smbsrv", KRTLD_MODE_FIRST, &error)) == NULL)) { mutex_exit(&zfs_share_lock); return (SET_ERROR(ENOSYS)); } if (zsmbexport_fs == NULL && ((zsmbexport_fs = (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, "smb_server_share", &error)) == NULL)) { mutex_exit(&zfs_share_lock); return (SET_ERROR(ENOSYS)); } error = zfs_init_sharefs(); if (error != 0) { mutex_exit(&zfs_share_lock); return (SET_ERROR(ENOSYS)); } zfs_smbshare_inited = 1; mutex_exit(&zfs_share_lock); } break; default: return (SET_ERROR(EINVAL)); } switch (zc->zc_share.z_sharetype) { case ZFS_SHARE_NFS: case ZFS_UNSHARE_NFS: if (error = znfsexport_fs((void *) (uintptr_t)zc->zc_share.z_exportdata)) return (error); break; case ZFS_SHARE_SMB: case ZFS_UNSHARE_SMB: if (error = zsmbexport_fs((void *) (uintptr_t)zc->zc_share.z_exportdata, zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? B_TRUE: B_FALSE)) { return (error); } break; } opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? SHAREFS_ADD : SHAREFS_REMOVE; /* * Add or remove share from sharetab */ error = zshare_fs(opcode, (void *)(uintptr_t)zc->zc_share.z_sharedata, zc->zc_share.z_sharemax); return (error); #else /* !sun */ return (ENOSYS); #endif /* !sun */ } ace_t full_access[] = { {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} }; /* * inputs: * zc_name name of containing filesystem * zc_obj object # beyond which we want next in-use object # * * outputs: * zc_obj next in-use object # */ static int zfs_ioc_next_obj(zfs_cmd_t *zc) { objset_t *os = NULL; int error; error = dmu_objset_hold(zc->zc_name, FTAG, &os); if (error != 0) return (error); error = dmu_object_next(os, &zc->zc_obj, B_FALSE, dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg); dmu_objset_rele(os, FTAG); return (error); } /* * inputs: * zc_name name of filesystem * zc_value prefix name for snapshot * zc_cleanup_fd cleanup-on-exit file descriptor for calling process * * outputs: * zc_value short name of new snapshot */ static int zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) { char *snap_name; char *hold_name; int error; minor_t minor; error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); if (error != 0) return (error); snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, (u_longlong_t)ddi_get_lbolt64()); hold_name = kmem_asprintf("%%%s", zc->zc_value); error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, hold_name); if (error == 0) (void) strcpy(zc->zc_value, snap_name); strfree(snap_name); strfree(hold_name); zfs_onexit_fd_rele(zc->zc_cleanup_fd); return (error); } /* * inputs: * zc_name name of "to" snapshot * zc_value name of "from" snapshot * zc_cookie file descriptor to write diff data on * * outputs: * dmu_diff_record_t's to the file descriptor */ static int zfs_ioc_diff(zfs_cmd_t *zc) { file_t *fp; cap_rights_t rights; offset_t off; int error; #ifdef illumos fp = getf(zc->zc_cookie); #else fget_write(curthread, zc->zc_cookie, cap_rights_init(&rights, CAP_WRITE), &fp); #endif if (fp == NULL) return (SET_ERROR(EBADF)); off = fp->f_offset; #ifdef illumos error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); #else error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); #endif if (off >= 0 && off <= MAXOFFSET_T) fp->f_offset = off; releasef(zc->zc_cookie); return (error); } #ifdef sun /* * Remove all ACL files in shares dir */ static int zfs_smb_acl_purge(znode_t *dzp) { zap_cursor_t zc; zap_attribute_t zap; zfsvfs_t *zfsvfs = dzp->z_zfsvfs; int error; for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); (error = zap_cursor_retrieve(&zc, &zap)) == 0; zap_cursor_advance(&zc)) { if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, NULL, 0)) != 0) break; } zap_cursor_fini(&zc); return (error); } #endif /* sun */ static int zfs_ioc_smb_acl(zfs_cmd_t *zc) { #ifdef sun vnode_t *vp; znode_t *dzp; vnode_t *resourcevp = NULL; znode_t *sharedir; zfsvfs_t *zfsvfs; nvlist_t *nvlist; char *src, *target; vattr_t vattr; vsecattr_t vsec; int error = 0; if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, NO_FOLLOW, NULL, &vp)) != 0) return (error); /* Now make sure mntpnt and dataset are ZFS */ if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), zc->zc_name) != 0)) { VN_RELE(vp); return (SET_ERROR(EINVAL)); } dzp = VTOZ(vp); zfsvfs = dzp->z_zfsvfs; ZFS_ENTER(zfsvfs); /* * Create share dir if its missing. */ mutex_enter(&zfsvfs->z_lock); if (zfsvfs->z_shares_dir == 0) { dmu_tx_t *tx; tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, ZFS_SHARES_DIR); dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); error = dmu_tx_assign(tx, TXG_WAIT); if (error != 0) { dmu_tx_abort(tx); } else { error = zfs_create_share_dir(zfsvfs, tx); dmu_tx_commit(tx); } if (error != 0) { mutex_exit(&zfsvfs->z_lock); VN_RELE(vp); ZFS_EXIT(zfsvfs); return (error); } } mutex_exit(&zfsvfs->z_lock); ASSERT(zfsvfs->z_shares_dir); if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { VN_RELE(vp); ZFS_EXIT(zfsvfs); return (error); } switch (zc->zc_cookie) { case ZFS_SMB_ACL_ADD: vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; vattr.va_type = VREG; vattr.va_mode = S_IFREG|0777; vattr.va_uid = 0; vattr.va_gid = 0; vsec.vsa_mask = VSA_ACE; vsec.vsa_aclentp = &full_access; vsec.vsa_aclentsz = sizeof (full_access); vsec.vsa_aclcnt = 1; error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); if (resourcevp) VN_RELE(resourcevp); break; case ZFS_SMB_ACL_REMOVE: error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, NULL, 0); break; case ZFS_SMB_ACL_RENAME: if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { VN_RELE(vp); VN_RELE(ZTOV(sharedir)); ZFS_EXIT(zfsvfs); return (error); } if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, &target)) { VN_RELE(vp); VN_RELE(ZTOV(sharedir)); ZFS_EXIT(zfsvfs); nvlist_free(nvlist); return (error); } error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, kcred, NULL, 0); nvlist_free(nvlist); break; case ZFS_SMB_ACL_PURGE: error = zfs_smb_acl_purge(sharedir); break; default: error = SET_ERROR(EINVAL); break; } VN_RELE(vp); VN_RELE(ZTOV(sharedir)); ZFS_EXIT(zfsvfs); return (error); #else /* !sun */ return (EOPNOTSUPP); #endif /* !sun */ } /* * innvl: { * "holds" -> { snapname -> holdname (string), ... } * (optional) "cleanup_fd" -> fd (int32) * } * * outnvl: { * snapname -> error value (int32) * ... * } */ /* ARGSUSED */ static int zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) { nvpair_t *pair; nvlist_t *holds; int cleanup_fd = -1; int error; minor_t minor = 0; error = nvlist_lookup_nvlist(args, "holds", &holds); if (error != 0) return (SET_ERROR(EINVAL)); /* make sure the user didn't pass us any invalid (empty) tags */ for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; pair = nvlist_next_nvpair(holds, pair)) { char *htag; error = nvpair_value_string(pair, &htag); if (error != 0) return (SET_ERROR(error)); if (strlen(htag) == 0) return (SET_ERROR(EINVAL)); } if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { error = zfs_onexit_fd_hold(cleanup_fd, &minor); if (error != 0) return (error); } error = dsl_dataset_user_hold(holds, minor, errlist); if (minor != 0) zfs_onexit_fd_rele(cleanup_fd); return (error); } /* * innvl is not used. * * outnvl: { * holdname -> time added (uint64 seconds since epoch) * ... * } */ /* ARGSUSED */ static int zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) { return (dsl_dataset_get_holds(snapname, outnvl)); } /* * innvl: { * snapname -> { holdname, ... } * ... * } * * outnvl: { * snapname -> error value (int32) * ... * } */ /* ARGSUSED */ static int zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) { return (dsl_dataset_user_release(holds, errlist)); } /* * inputs: * zc_name name of new filesystem or snapshot * zc_value full name of old snapshot * * outputs: * zc_cookie space in bytes * zc_objset_type compressed space in bytes * zc_perm_action uncompressed space in bytes */ static int zfs_ioc_space_written(zfs_cmd_t *zc) { int error; dsl_pool_t *dp; dsl_dataset_t *new, *old; error = dsl_pool_hold(zc->zc_name, FTAG, &dp); if (error != 0) return (error); error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); if (error != 0) { dsl_pool_rele(dp, FTAG); return (error); } error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); if (error != 0) { dsl_dataset_rele(new, FTAG); dsl_pool_rele(dp, FTAG); return (error); } error = dsl_dataset_space_written(old, new, &zc->zc_cookie, &zc->zc_objset_type, &zc->zc_perm_action); dsl_dataset_rele(old, FTAG); dsl_dataset_rele(new, FTAG); dsl_pool_rele(dp, FTAG); return (error); } /* * innvl: { * "firstsnap" -> snapshot name * } * * outnvl: { * "used" -> space in bytes * "compressed" -> compressed space in bytes * "uncompressed" -> uncompressed space in bytes * } */ static int zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) { int error; dsl_pool_t *dp; dsl_dataset_t *new, *old; char *firstsnap; uint64_t used, comp, uncomp; if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) return (SET_ERROR(EINVAL)); error = dsl_pool_hold(lastsnap, FTAG, &dp); if (error != 0) return (error); error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); if (error == 0 && !new->ds_is_snapshot) { dsl_dataset_rele(new, FTAG); error = SET_ERROR(EINVAL); } if (error != 0) { dsl_pool_rele(dp, FTAG); return (error); } error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); if (error == 0 && !old->ds_is_snapshot) { dsl_dataset_rele(old, FTAG); error = SET_ERROR(EINVAL); } if (error != 0) { dsl_dataset_rele(new, FTAG); dsl_pool_rele(dp, FTAG); return (error); } error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); dsl_dataset_rele(old, FTAG); dsl_dataset_rele(new, FTAG); dsl_pool_rele(dp, FTAG); fnvlist_add_uint64(outnvl, "used", used); fnvlist_add_uint64(outnvl, "compressed", comp); fnvlist_add_uint64(outnvl, "uncompressed", uncomp); return (error); } static int zfs_ioc_jail(zfs_cmd_t *zc) { return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, (int)zc->zc_jailid)); } static int zfs_ioc_unjail(zfs_cmd_t *zc) { return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, (int)zc->zc_jailid)); } /* * innvl: { * "fd" -> file descriptor to write stream to (int32) * (optional) "fromsnap" -> full snap name to send an incremental from * (optional) "largeblockok" -> (value ignored) * indicates that blocks > 128KB are permitted * (optional) "embedok" -> (value ignored) * presence indicates DRR_WRITE_EMBEDDED records are permitted * (optional) "resume_object" and "resume_offset" -> (uint64) * if present, resume send stream from specified object and offset. * } * * outnvl is unused */ /* ARGSUSED */ static int zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) { cap_rights_t rights; file_t *fp; int error; offset_t off; char *fromname = NULL; int fd; boolean_t largeblockok; boolean_t embedok; uint64_t resumeobj = 0; uint64_t resumeoff = 0; error = nvlist_lookup_int32(innvl, "fd", &fd); if (error != 0) return (SET_ERROR(EINVAL)); (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); largeblockok = nvlist_exists(innvl, "largeblockok"); embedok = nvlist_exists(innvl, "embedok"); (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj); (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff); #ifdef illumos file_t *fp = getf(fd); #else fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp); #endif if (fp == NULL) return (SET_ERROR(EBADF)); off = fp->f_offset; error = dmu_send(snapname, fromname, embedok, largeblockok, fd, #ifdef illumos resumeobj, resumeoff, fp->f_vnode, &off); #else resumeobj, resumeoff, fp, &off); #endif #ifdef illumos if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) fp->f_offset = off; #else fp->f_offset = off; #endif releasef(fd); return (error); } /* * Determine approximately how large a zfs send stream will be -- the number * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). * * innvl: { * (optional) "from" -> full snap or bookmark name to send an incremental * from * } * * outnvl: { * "space" -> bytes of space (uint64) * } */ static int zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) { dsl_pool_t *dp; dsl_dataset_t *tosnap; int error; char *fromname; uint64_t space; error = dsl_pool_hold(snapname, FTAG, &dp); if (error != 0) return (error); error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); if (error != 0) { dsl_pool_rele(dp, FTAG); return (error); } error = nvlist_lookup_string(innvl, "from", &fromname); if (error == 0) { if (strchr(fromname, '@') != NULL) { /* * If from is a snapshot, hold it and use the more * efficient dmu_send_estimate to estimate send space * size using deadlists. */ dsl_dataset_t *fromsnap; error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); if (error != 0) goto out; error = dmu_send_estimate(tosnap, fromsnap, &space); dsl_dataset_rele(fromsnap, FTAG); } else if (strchr(fromname, '#') != NULL) { /* * If from is a bookmark, fetch the creation TXG of the * snapshot it was created from and use that to find * blocks that were born after it. */ zfs_bookmark_phys_t frombm; error = dsl_bookmark_lookup(dp, fromname, tosnap, &frombm); if (error != 0) goto out; error = dmu_send_estimate_from_txg(tosnap, frombm.zbm_creation_txg, &space); } else { /* * from is not properly formatted as a snapshot or * bookmark */ error = SET_ERROR(EINVAL); goto out; } } else { // If estimating the size of a full send, use dmu_send_estimate error = dmu_send_estimate(tosnap, NULL, &space); } fnvlist_add_uint64(outnvl, "space", space); out: dsl_dataset_rele(tosnap, FTAG); dsl_pool_rele(dp, FTAG); return (error); } static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; static void zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, boolean_t log_history, zfs_ioc_poolcheck_t pool_check) { zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; ASSERT3U(ioc, >=, ZFS_IOC_FIRST); ASSERT3U(ioc, <, ZFS_IOC_LAST); ASSERT3P(vec->zvec_legacy_func, ==, NULL); ASSERT3P(vec->zvec_func, ==, NULL); vec->zvec_legacy_func = func; vec->zvec_secpolicy = secpolicy; vec->zvec_namecheck = namecheck; vec->zvec_allow_log = log_history; vec->zvec_pool_check = pool_check; } /* * See the block comment at the beginning of this file for details on * each argument to this function. */ static void zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, boolean_t allow_log) { zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; ASSERT3U(ioc, >=, ZFS_IOC_FIRST); ASSERT3U(ioc, <, ZFS_IOC_LAST); ASSERT3P(vec->zvec_legacy_func, ==, NULL); ASSERT3P(vec->zvec_func, ==, NULL); /* if we are logging, the name must be valid */ ASSERT(!allow_log || namecheck != NO_NAME); vec->zvec_name = name; vec->zvec_func = func; vec->zvec_secpolicy = secpolicy; vec->zvec_namecheck = namecheck; vec->zvec_pool_check = pool_check; vec->zvec_smush_outnvlist = smush_outnvlist; vec->zvec_allow_log = allow_log; } static void zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy, boolean_t log_history, zfs_ioc_poolcheck_t pool_check) { zfs_ioctl_register_legacy(ioc, func, secpolicy, POOL_NAME, log_history, pool_check); } static void zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) { zfs_ioctl_register_legacy(ioc, func, secpolicy, DATASET_NAME, B_FALSE, pool_check); } static void zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) { zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); } static void zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) { zfs_ioctl_register_legacy(ioc, func, secpolicy, NO_NAME, B_FALSE, POOL_CHECK_NONE); } static void zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) { zfs_ioctl_register_legacy(ioc, func, secpolicy, DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); } static void zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) { zfs_ioctl_register_dataset_read_secpolicy(ioc, func, zfs_secpolicy_read); } static void zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) { zfs_ioctl_register_legacy(ioc, func, secpolicy, DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); } static void zfs_ioctl_init(void) { zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); zfs_ioctl_register("create", ZFS_IOC_CREATE, zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); zfs_ioctl_register("clone", ZFS_IOC_CLONE, zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); zfs_ioctl_register("hold", ZFS_IOC_HOLD, zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); zfs_ioctl_register("release", ZFS_IOC_RELEASE, zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); /* IOCTLS that use the legacy function signature */ zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, zfs_ioc_pool_scan); zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, zfs_ioc_pool_upgrade); zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, zfs_ioc_vdev_add); zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, zfs_ioc_vdev_remove); zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, zfs_ioc_vdev_set_state); zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, zfs_ioc_vdev_attach); zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, zfs_ioc_vdev_detach); zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, zfs_ioc_vdev_setpath); zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, zfs_ioc_vdev_setfru); zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, zfs_ioc_pool_set_props); zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, zfs_ioc_vdev_split); zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, zfs_ioc_pool_reguid); zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, zfs_ioc_pool_configs, zfs_secpolicy_none); zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, zfs_ioc_pool_tryimport, zfs_secpolicy_config); zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, zfs_ioc_inject_fault, zfs_secpolicy_inject); zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, zfs_ioc_clear_fault, zfs_secpolicy_inject); zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, zfs_ioc_inject_list_next, zfs_secpolicy_inject); /* * pool destroy, and export don't log the history as part of * zfsdev_ioctl, but rather zfs_ioc_pool_export * does the logging of those commands. */ zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, zfs_ioc_dsobj_to_dsname, zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, zfs_ioc_pool_get_history, zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, zfs_ioc_space_written); zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, zfs_ioc_objset_recvd_props); zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, zfs_ioc_next_obj); zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, zfs_ioc_get_fsacl); zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, zfs_ioc_objset_stats); zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, zfs_ioc_objset_zplprops); zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, zfs_ioc_dataset_list_next); zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, zfs_ioc_snapshot_list_next); zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, zfs_ioc_send_progress); zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, zfs_ioc_diff, zfs_secpolicy_diff); zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, zfs_ioc_obj_to_stats, zfs_secpolicy_diff); zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, zfs_ioc_obj_to_path, zfs_secpolicy_diff); zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, zfs_ioc_send, zfs_secpolicy_send); zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, zfs_secpolicy_none); zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, zfs_secpolicy_destroy); zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, zfs_secpolicy_rename); zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, zfs_secpolicy_recv); zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, zfs_secpolicy_promote); zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, zfs_secpolicy_set_fsacl); zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, zfs_secpolicy_share, POOL_CHECK_NONE); zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, POOL_CHECK_NONE); zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); #ifdef __FreeBSD__ zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, zfs_secpolicy_config, POOL_CHECK_NONE); zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, zfs_secpolicy_config, POOL_CHECK_NONE); #endif } int pool_status_check(const char *name, zfs_ioc_namecheck_t type, zfs_ioc_poolcheck_t check) { spa_t *spa; int error; ASSERT(type == POOL_NAME || type == DATASET_NAME); if (check & POOL_CHECK_NONE) return (0); error = spa_open(name, &spa, FTAG); if (error == 0) { if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) error = SET_ERROR(EAGAIN); else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) error = SET_ERROR(EROFS); spa_close(spa, FTAG); } return (error); } /* * Find a free minor number. */ minor_t zfsdev_minor_alloc(void) { static minor_t last_minor; minor_t m; ASSERT(MUTEX_HELD(&spa_namespace_lock)); for (m = last_minor + 1; m != last_minor; m++) { if (m > ZFSDEV_MAX_MINOR) m = 1; if (ddi_get_soft_state(zfsdev_state, m) == NULL) { last_minor = m; return (m); } } return (0); } static int zfs_ctldev_init(struct cdev *devp) { minor_t minor; zfs_soft_state_t *zs; ASSERT(MUTEX_HELD(&spa_namespace_lock)); minor = zfsdev_minor_alloc(); if (minor == 0) return (SET_ERROR(ENXIO)); if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) return (SET_ERROR(EAGAIN)); devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); zs = ddi_get_soft_state(zfsdev_state, minor); zs->zss_type = ZSST_CTLDEV; zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); return (0); } static void zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) { ASSERT(MUTEX_HELD(&spa_namespace_lock)); zfs_onexit_destroy(zo); ddi_soft_state_free(zfsdev_state, minor); } void * zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) { zfs_soft_state_t *zp; zp = ddi_get_soft_state(zfsdev_state, minor); if (zp == NULL || zp->zss_type != which) return (NULL); return (zp->zss_data); } static int zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) { int error = 0; #ifdef sun if (getminor(*devp) != 0) return (zvol_open(devp, flag, otyp, cr)); #endif /* This is the control device. Allocate a new minor if requested. */ if (flag & FEXCL) { mutex_enter(&spa_namespace_lock); error = zfs_ctldev_init(devp); mutex_exit(&spa_namespace_lock); } return (error); } static void zfsdev_close(void *data) { zfs_onexit_t *zo; minor_t minor = (minor_t)(uintptr_t)data; if (minor == 0) return; mutex_enter(&spa_namespace_lock); zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); if (zo == NULL) { mutex_exit(&spa_namespace_lock); return; } zfs_ctldev_destroy(zo, minor); mutex_exit(&spa_namespace_lock); } static int zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, struct thread *td) { zfs_cmd_t *zc; uint_t vecnum; int error, rc, len; #ifdef illumos minor_t minor = getminor(dev); #else zfs_iocparm_t *zc_iocparm; int cflag, cmd, oldvecnum; boolean_t newioc, compat; void *compat_zc = NULL; cred_t *cr = td->td_ucred; #endif const zfs_ioc_vec_t *vec; char *saved_poolname = NULL; nvlist_t *innvl = NULL; cflag = ZFS_CMD_COMPAT_NONE; compat = B_FALSE; newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */ len = IOCPARM_LEN(zcmd); vecnum = cmd = zcmd & 0xff; /* * Check if we are talking to supported older binaries * and translate zfs_cmd if necessary */ if (len != sizeof(zfs_iocparm_t)) { newioc = B_FALSE; compat = B_TRUE; vecnum = cmd; switch (len) { case sizeof(zfs_cmd_zcmd_t): cflag = ZFS_CMD_COMPAT_LZC; break; case sizeof(zfs_cmd_deadman_t): cflag = ZFS_CMD_COMPAT_DEADMAN; break; case sizeof(zfs_cmd_v28_t): cflag = ZFS_CMD_COMPAT_V28; break; case sizeof(zfs_cmd_v15_t): cflag = ZFS_CMD_COMPAT_V15; vecnum = zfs_ioctl_v15_to_v28[cmd]; /* * Return without further handling * if the command is blacklisted. */ if (vecnum == ZFS_IOC_COMPAT_PASS) return (0); else if (vecnum == ZFS_IOC_COMPAT_FAIL) return (ENOTSUP); break; default: return (EINVAL); } } #ifdef illumos vecnum = cmd - ZFS_IOC_FIRST; ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); #endif if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) return (SET_ERROR(EINVAL)); vec = &zfs_ioc_vec[vecnum]; zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); #ifdef illumos error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); if (error != 0) { error = SET_ERROR(EFAULT); goto out; } #else /* !illumos */ bzero(zc, sizeof(zfs_cmd_t)); if (newioc) { zc_iocparm = (void *)arg; switch (zc_iocparm->zfs_ioctl_version) { case ZFS_IOCVER_CURRENT: if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { error = SET_ERROR(EINVAL); goto out; } break; case ZFS_IOCVER_EDBP: if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_edbp_t)) { error = SET_ERROR(EFAULT); goto out; } compat = B_TRUE; cflag = ZFS_CMD_COMPAT_EDBP; break; case ZFS_IOCVER_ZCMD: if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) || zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) { error = SET_ERROR(EFAULT); goto out; } compat = B_TRUE; cflag = ZFS_CMD_COMPAT_ZCMD; break; default: error = SET_ERROR(EINVAL); goto out; /* NOTREACHED */ } if (compat) { ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); bzero(compat_zc, sizeof(zfs_cmd_t)); error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, compat_zc, zc_iocparm->zfs_cmd_size, flag); if (error != 0) { error = SET_ERROR(EFAULT); goto out; } } else { error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, zc, zc_iocparm->zfs_cmd_size, flag); if (error != 0) { error = SET_ERROR(EFAULT); goto out; } } } if (compat) { if (newioc) { ASSERT(compat_zc != NULL); zfs_cmd_compat_get(zc, compat_zc, cflag); } else { ASSERT(compat_zc == NULL); zfs_cmd_compat_get(zc, arg, cflag); } oldvecnum = vecnum; error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); if (error != 0) goto out; if (oldvecnum != vecnum) vec = &zfs_ioc_vec[vecnum]; } #endif /* !illumos */ zc->zc_iflags = flag & FKIOCTL; if (zc->zc_nvlist_src_size != 0) { error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, zc->zc_iflags, &innvl); if (error != 0) goto out; } /* rewrite innvl for backwards compatibility */ if (compat) innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); /* * Ensure that all pool/dataset names are valid before we pass down to * the lower layers. */ zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; switch (vec->zvec_namecheck) { case POOL_NAME: if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) error = SET_ERROR(EINVAL); else error = pool_status_check(zc->zc_name, vec->zvec_namecheck, vec->zvec_pool_check); break; case DATASET_NAME: if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) error = SET_ERROR(EINVAL); else error = pool_status_check(zc->zc_name, vec->zvec_namecheck, vec->zvec_pool_check); break; case NO_NAME: break; } if (error == 0 && !(flag & FKIOCTL)) error = vec->zvec_secpolicy(zc, innvl, cr); if (error != 0) goto out; /* legacy ioctls can modify zc_name */ len = strcspn(zc->zc_name, "/@#") + 1; saved_poolname = kmem_alloc(len, KM_SLEEP); (void) strlcpy(saved_poolname, zc->zc_name, len); if (vec->zvec_func != NULL) { nvlist_t *outnvl; int puterror = 0; spa_t *spa; nvlist_t *lognv = NULL; ASSERT(vec->zvec_legacy_func == NULL); /* * Add the innvl to the lognv before calling the func, * in case the func changes the innvl. */ if (vec->zvec_allow_log) { lognv = fnvlist_alloc(); fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, vec->zvec_name); if (!nvlist_empty(innvl)) { fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, innvl); } } outnvl = fnvlist_alloc(); error = vec->zvec_func(zc->zc_name, innvl, outnvl); if (error == 0 && vec->zvec_allow_log && spa_open(zc->zc_name, &spa, FTAG) == 0) { if (!nvlist_empty(outnvl)) { fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, outnvl); } (void) spa_history_log_nvl(spa, lognv); spa_close(spa, FTAG); } fnvlist_free(lognv); /* rewrite outnvl for backwards compatibility */ if (compat) outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, cflag); if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { int smusherror = 0; if (vec->zvec_smush_outnvlist) { smusherror = nvlist_smush(outnvl, zc->zc_nvlist_dst_size); } if (smusherror == 0) puterror = put_nvlist(zc, outnvl); } if (puterror != 0) error = puterror; nvlist_free(outnvl); } else { error = vec->zvec_legacy_func(zc); } out: nvlist_free(innvl); #ifdef illumos rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); if (error == 0 && rc != 0) error = SET_ERROR(EFAULT); #else if (compat) { zfs_ioctl_compat_post(zc, cmd, cflag); if (newioc) { ASSERT(compat_zc != NULL); ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag); rc = ddi_copyout(compat_zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, zc_iocparm->zfs_cmd_size, flag); if (error == 0 && rc != 0) error = SET_ERROR(EFAULT); kmem_free(compat_zc, sizeof (zfs_cmd_t)); } else { zfs_cmd_compat_put(zc, arg, vecnum, cflag); } } else { ASSERT(newioc); rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, sizeof (zfs_cmd_t), flag); if (error == 0 && rc != 0) error = SET_ERROR(EFAULT); } #endif if (error == 0 && vec->zvec_allow_log) { char *s = tsd_get(zfs_allow_log_key); if (s != NULL) strfree(s); (void) tsd_set(zfs_allow_log_key, saved_poolname); } else { if (saved_poolname != NULL) strfree(saved_poolname); } kmem_free(zc, sizeof (zfs_cmd_t)); return (error); } #ifdef sun static int zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) { if (cmd != DDI_ATTACH) return (DDI_FAILURE); if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, DDI_PSEUDO, 0) == DDI_FAILURE) return (DDI_FAILURE); zfs_dip = dip; ddi_report_dev(dip); return (DDI_SUCCESS); } static int zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) { if (spa_busy() || zfs_busy() || zvol_busy()) return (DDI_FAILURE); if (cmd != DDI_DETACH) return (DDI_FAILURE); zfs_dip = NULL; ddi_prop_remove_all(dip); ddi_remove_minor_node(dip, NULL); return (DDI_SUCCESS); } /*ARGSUSED*/ static int zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) { switch (infocmd) { case DDI_INFO_DEVT2DEVINFO: *result = zfs_dip; return (DDI_SUCCESS); case DDI_INFO_DEVT2INSTANCE: *result = (void *)0; return (DDI_SUCCESS); } return (DDI_FAILURE); } #endif /* sun */ /* * OK, so this is a little weird. * * /dev/zfs is the control node, i.e. minor 0. * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. * * /dev/zfs has basically nothing to do except serve up ioctls, * so most of the standard driver entry points are in zvol.c. */ #ifdef sun static struct cb_ops zfs_cb_ops = { zfsdev_open, /* open */ zfsdev_close, /* close */ zvol_strategy, /* strategy */ nodev, /* print */ zvol_dump, /* dump */ zvol_read, /* read */ zvol_write, /* write */ zfsdev_ioctl, /* ioctl */ nodev, /* devmap */ nodev, /* mmap */ nodev, /* segmap */ nochpoll, /* poll */ ddi_prop_op, /* prop_op */ NULL, /* streamtab */ D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ CB_REV, /* version */ nodev, /* async read */ nodev, /* async write */ }; static struct dev_ops zfs_dev_ops = { DEVO_REV, /* version */ 0, /* refcnt */ zfs_info, /* info */ nulldev, /* identify */ nulldev, /* probe */ zfs_attach, /* attach */ zfs_detach, /* detach */ nodev, /* reset */ &zfs_cb_ops, /* driver operations */ NULL, /* no bus operations */ NULL, /* power */ ddi_quiesce_not_needed, /* quiesce */ }; static struct modldrv zfs_modldrv = { &mod_driverops, "ZFS storage pool", &zfs_dev_ops }; static struct modlinkage modlinkage = { MODREV_1, (void *)&zfs_modlfs, (void *)&zfs_modldrv, NULL }; #endif /* sun */ static struct cdevsw zfs_cdevsw = { .d_version = D_VERSION, .d_open = zfsdev_open, .d_ioctl = zfsdev_ioctl, .d_name = ZFS_DEV_NAME }; static void zfs_allow_log_destroy(void *arg) { char *poolname = arg; strfree(poolname); } static void zfsdev_init(void) { zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, ZFS_DEV_NAME); } static void zfsdev_fini(void) { if (zfsdev != NULL) destroy_dev(zfsdev); } static struct root_hold_token *zfs_root_token; struct proc *zfsproc; #ifdef sun int _init(void) { int error; spa_init(FREAD | FWRITE); zfs_init(); zvol_init(); zfs_ioctl_init(); if ((error = mod_install(&modlinkage)) != 0) { zvol_fini(); zfs_fini(); spa_fini(); return (error); } tsd_create(&zfs_fsyncer_key, NULL); tsd_create(&rrw_tsd_key, rrw_tsd_destroy); tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); error = ldi_ident_from_mod(&modlinkage, &zfs_li); ASSERT(error == 0); mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); return (0); } int _fini(void) { int error; if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) return (SET_ERROR(EBUSY)); if ((error = mod_remove(&modlinkage)) != 0) return (error); zvol_fini(); zfs_fini(); spa_fini(); if (zfs_nfsshare_inited) (void) ddi_modclose(nfs_mod); if (zfs_smbshare_inited) (void) ddi_modclose(smbsrv_mod); if (zfs_nfsshare_inited || zfs_smbshare_inited) (void) ddi_modclose(sharefs_mod); tsd_destroy(&zfs_fsyncer_key); ldi_ident_release(zfs_li); zfs_li = NULL; mutex_destroy(&zfs_share_lock); return (error); } int _info(struct modinfo *modinfop) { return (mod_info(&modlinkage, modinfop)); } #endif /* sun */ static int zfs__init(void); static int zfs__fini(void); static void zfs_shutdown(void *, int); static eventhandler_tag zfs_shutdown_event_tag; #define ZFS_MIN_KSTACK_PAGES 4 int zfs__init(void) { #if KSTACK_PAGES < ZFS_MIN_KSTACK_PAGES printf("ZFS NOTICE: KSTACK_PAGES is %d which could result in stack " "overflow panic!\nPlease consider adding " "'options KSTACK_PAGES=%d' to your kernel config\n", KSTACK_PAGES, ZFS_MIN_KSTACK_PAGES); #endif zfs_root_token = root_mount_hold("ZFS"); mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); spa_init(FREAD | FWRITE); zfs_init(); zvol_init(); zfs_ioctl_init(); tsd_create(&zfs_fsyncer_key, NULL); tsd_create(&rrw_tsd_key, rrw_tsd_destroy); tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); root_mount_rel(zfs_root_token); zfsdev_init(); return (0); } int zfs__fini(void) { if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) { return (EBUSY); } zfsdev_fini(); zvol_fini(); zfs_fini(); spa_fini(); tsd_destroy(&zfs_fsyncer_key); tsd_destroy(&rrw_tsd_key); tsd_destroy(&zfs_allow_log_key); mutex_destroy(&zfs_share_lock); return (0); } static void zfs_shutdown(void *arg __unused, int howto __unused) { /* * ZFS fini routines can not properly work in a panic-ed system. */ if (panicstr == NULL) (void)zfs__fini(); } static int zfs_modevent(module_t mod, int type, void *unused __unused) { int err; switch (type) { case MOD_LOAD: err = zfs__init(); if (err == 0) zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( shutdown_post_sync, zfs_shutdown, NULL, SHUTDOWN_PRI_FIRST); return (err); case MOD_UNLOAD: err = zfs__fini(); if (err == 0 && zfs_shutdown_event_tag != NULL) EVENTHANDLER_DEREGISTER(shutdown_post_sync, zfs_shutdown_event_tag); return (err); case MOD_SHUTDOWN: return (0); default: break; } return (EOPNOTSUPP); } static moduledata_t zfs_mod = { "zfsctrl", zfs_modevent, 0 }; DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); MODULE_VERSION(zfsctrl, 1); MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zio.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zio.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zio.c (revision 290757) @@ -1,3582 +1,3595 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2015 by Delphix. All rights reserved. * Copyright (c) 2011 Nexenta Systems, Inc. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include SYSCTL_DECL(_vfs_zfs); SYSCTL_NODE(_vfs_zfs, OID_AUTO, zio, CTLFLAG_RW, 0, "ZFS ZIO"); #if defined(__amd64__) static int zio_use_uma = 1; #else static int zio_use_uma = 0; #endif TUNABLE_INT("vfs.zfs.zio.use_uma", &zio_use_uma); SYSCTL_INT(_vfs_zfs_zio, OID_AUTO, use_uma, CTLFLAG_RDTUN, &zio_use_uma, 0, "Use uma(9) for ZIO allocations"); static int zio_exclude_metadata = 0; TUNABLE_INT("vfs.zfs.zio.exclude_metadata", &zio_exclude_metadata); SYSCTL_INT(_vfs_zfs_zio, OID_AUTO, exclude_metadata, CTLFLAG_RDTUN, &zio_exclude_metadata, 0, "Exclude metadata buffers from dumps as well"); zio_trim_stats_t zio_trim_stats = { { "bytes", KSTAT_DATA_UINT64, "Number of bytes successfully TRIMmed" }, { "success", KSTAT_DATA_UINT64, "Number of successful TRIM requests" }, { "unsupported", KSTAT_DATA_UINT64, "Number of TRIM requests that failed because TRIM is not supported" }, { "failed", KSTAT_DATA_UINT64, "Number of TRIM requests that failed for reasons other than not supported" }, }; static kstat_t *zio_trim_ksp; /* * ========================================================================== * I/O type descriptions * ========================================================================== */ const char *zio_type_name[ZIO_TYPES] = { "zio_null", "zio_read", "zio_write", "zio_free", "zio_claim", "zio_ioctl" }; /* * ========================================================================== * I/O kmem caches * ========================================================================== */ kmem_cache_t *zio_cache; kmem_cache_t *zio_link_cache; kmem_cache_t *zio_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT]; kmem_cache_t *zio_data_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT]; #ifdef _KERNEL extern vmem_t *zio_alloc_arena; #endif #define BP_SPANB(indblkshift, level) \ (((uint64_t)1) << ((level) * ((indblkshift) - SPA_BLKPTRSHIFT))) #define COMPARE_META_LEVEL 0x80000000ul /* * The following actions directly effect the spa's sync-to-convergence logic. * The values below define the sync pass when we start performing the action. * Care should be taken when changing these values as they directly impact * spa_sync() performance. Tuning these values may introduce subtle performance * pathologies and should only be done in the context of performance analysis. * These tunables will eventually be removed and replaced with #defines once * enough analysis has been done to determine optimal values. * * The 'zfs_sync_pass_deferred_free' pass must be greater than 1 to ensure that * regular blocks are not deferred. */ int zfs_sync_pass_deferred_free = 2; /* defer frees starting in this pass */ TUNABLE_INT("vfs.zfs.sync_pass_deferred_free", &zfs_sync_pass_deferred_free); SYSCTL_INT(_vfs_zfs, OID_AUTO, sync_pass_deferred_free, CTLFLAG_RDTUN, &zfs_sync_pass_deferred_free, 0, "defer frees starting in this pass"); int zfs_sync_pass_dont_compress = 5; /* don't compress starting in this pass */ TUNABLE_INT("vfs.zfs.sync_pass_dont_compress", &zfs_sync_pass_dont_compress); SYSCTL_INT(_vfs_zfs, OID_AUTO, sync_pass_dont_compress, CTLFLAG_RDTUN, &zfs_sync_pass_dont_compress, 0, "don't compress starting in this pass"); int zfs_sync_pass_rewrite = 2; /* rewrite new bps starting in this pass */ TUNABLE_INT("vfs.zfs.sync_pass_rewrite", &zfs_sync_pass_rewrite); SYSCTL_INT(_vfs_zfs, OID_AUTO, sync_pass_rewrite, CTLFLAG_RDTUN, &zfs_sync_pass_rewrite, 0, "rewrite new bps starting in this pass"); /* * An allocating zio is one that either currently has the DVA allocate * stage set or will have it later in its lifetime. */ #define IO_IS_ALLOCATING(zio) ((zio)->io_orig_pipeline & ZIO_STAGE_DVA_ALLOCATE) boolean_t zio_requeue_io_start_cut_in_line = B_TRUE; #ifdef ZFS_DEBUG int zio_buf_debug_limit = 16384; #else int zio_buf_debug_limit = 0; #endif void zio_init(void) { size_t c; zio_cache = kmem_cache_create("zio_cache", sizeof (zio_t), 0, NULL, NULL, NULL, NULL, NULL, 0); zio_link_cache = kmem_cache_create("zio_link_cache", sizeof (zio_link_t), 0, NULL, NULL, NULL, NULL, NULL, 0); if (!zio_use_uma) goto out; /* * For small buffers, we want a cache for each multiple of * SPA_MINBLOCKSIZE. For larger buffers, we want a cache * for each quarter-power of 2. */ for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) { size_t size = (c + 1) << SPA_MINBLOCKSHIFT; size_t p2 = size; size_t align = 0; size_t cflags = (size > zio_buf_debug_limit) ? KMC_NODEBUG : 0; while (!ISP2(p2)) p2 &= p2 - 1; #ifdef illumos #ifndef _KERNEL /* * If we are using watchpoints, put each buffer on its own page, * to eliminate the performance overhead of trapping to the * kernel when modifying a non-watched buffer that shares the * page with a watched buffer. */ if (arc_watch && !IS_P2ALIGNED(size, PAGESIZE)) continue; #endif #endif /* illumos */ if (size <= 4 * SPA_MINBLOCKSIZE) { align = SPA_MINBLOCKSIZE; } else if (IS_P2ALIGNED(size, p2 >> 2)) { align = MIN(p2 >> 2, PAGESIZE); } if (align != 0) { char name[36]; (void) sprintf(name, "zio_buf_%lu", (ulong_t)size); zio_buf_cache[c] = kmem_cache_create(name, size, align, NULL, NULL, NULL, NULL, NULL, cflags); /* * Since zio_data bufs do not appear in crash dumps, we * pass KMC_NOTOUCH so that no allocator metadata is * stored with the buffers. */ (void) sprintf(name, "zio_data_buf_%lu", (ulong_t)size); zio_data_buf_cache[c] = kmem_cache_create(name, size, align, NULL, NULL, NULL, NULL, NULL, cflags | KMC_NOTOUCH | KMC_NODEBUG); } } while (--c != 0) { ASSERT(zio_buf_cache[c] != NULL); if (zio_buf_cache[c - 1] == NULL) zio_buf_cache[c - 1] = zio_buf_cache[c]; ASSERT(zio_data_buf_cache[c] != NULL); if (zio_data_buf_cache[c - 1] == NULL) zio_data_buf_cache[c - 1] = zio_data_buf_cache[c]; } out: zio_inject_init(); zio_trim_ksp = kstat_create("zfs", 0, "zio_trim", "misc", KSTAT_TYPE_NAMED, sizeof(zio_trim_stats) / sizeof(kstat_named_t), KSTAT_FLAG_VIRTUAL); if (zio_trim_ksp != NULL) { zio_trim_ksp->ks_data = &zio_trim_stats; kstat_install(zio_trim_ksp); } } void zio_fini(void) { size_t c; kmem_cache_t *last_cache = NULL; kmem_cache_t *last_data_cache = NULL; for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) { if (zio_buf_cache[c] != last_cache) { last_cache = zio_buf_cache[c]; kmem_cache_destroy(zio_buf_cache[c]); } zio_buf_cache[c] = NULL; if (zio_data_buf_cache[c] != last_data_cache) { last_data_cache = zio_data_buf_cache[c]; kmem_cache_destroy(zio_data_buf_cache[c]); } zio_data_buf_cache[c] = NULL; } kmem_cache_destroy(zio_link_cache); kmem_cache_destroy(zio_cache); zio_inject_fini(); if (zio_trim_ksp != NULL) { kstat_delete(zio_trim_ksp); zio_trim_ksp = NULL; } } /* * ========================================================================== * Allocate and free I/O buffers * ========================================================================== */ /* * Use zio_buf_alloc to allocate ZFS metadata. This data will appear in a * crashdump if the kernel panics, so use it judiciously. Obviously, it's * useful to inspect ZFS metadata, but if possible, we should avoid keeping * excess / transient data in-core during a crashdump. */ void * zio_buf_alloc(size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; int flags = zio_exclude_metadata ? KM_NODEBUG : 0; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); if (zio_use_uma) return (kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE)); else return (kmem_alloc(size, KM_SLEEP|flags)); } /* * Use zio_data_buf_alloc to allocate data. The data will not appear in a * crashdump if the kernel panics. This exists so that we will limit the amount * of ZFS data that shows up in a kernel crashdump. (Thus reducing the amount * of kernel heap dumped to disk when the kernel panics) */ void * zio_data_buf_alloc(size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); if (zio_use_uma) return (kmem_cache_alloc(zio_data_buf_cache[c], KM_PUSHPAGE)); else return (kmem_alloc(size, KM_SLEEP | KM_NODEBUG)); } void zio_buf_free(void *buf, size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); if (zio_use_uma) kmem_cache_free(zio_buf_cache[c], buf); else kmem_free(buf, size); } void zio_data_buf_free(void *buf, size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); if (zio_use_uma) kmem_cache_free(zio_data_buf_cache[c], buf); else kmem_free(buf, size); } /* * ========================================================================== * Push and pop I/O transform buffers * ========================================================================== */ static void zio_push_transform(zio_t *zio, void *data, uint64_t size, uint64_t bufsize, zio_transform_func_t *transform) { zio_transform_t *zt = kmem_alloc(sizeof (zio_transform_t), KM_SLEEP); zt->zt_orig_data = zio->io_data; zt->zt_orig_size = zio->io_size; zt->zt_bufsize = bufsize; zt->zt_transform = transform; zt->zt_next = zio->io_transform_stack; zio->io_transform_stack = zt; zio->io_data = data; zio->io_size = size; } static void zio_pop_transforms(zio_t *zio) { zio_transform_t *zt; while ((zt = zio->io_transform_stack) != NULL) { if (zt->zt_transform != NULL) zt->zt_transform(zio, zt->zt_orig_data, zt->zt_orig_size); if (zt->zt_bufsize != 0) zio_buf_free(zio->io_data, zt->zt_bufsize); zio->io_data = zt->zt_orig_data; zio->io_size = zt->zt_orig_size; zio->io_transform_stack = zt->zt_next; kmem_free(zt, sizeof (zio_transform_t)); } } /* * ========================================================================== * I/O transform callbacks for subblocks and decompression * ========================================================================== */ static void zio_subblock(zio_t *zio, void *data, uint64_t size) { ASSERT(zio->io_size > size); if (zio->io_type == ZIO_TYPE_READ) bcopy(zio->io_data, data, size); } static void zio_decompress(zio_t *zio, void *data, uint64_t size) { if (zio->io_error == 0 && zio_decompress_data(BP_GET_COMPRESS(zio->io_bp), zio->io_data, data, zio->io_size, size) != 0) zio->io_error = SET_ERROR(EIO); } /* * ========================================================================== * I/O parent/child relationships and pipeline interlocks * ========================================================================== */ /* * NOTE - Callers to zio_walk_parents() and zio_walk_children must * continue calling these functions until they return NULL. * Otherwise, the next caller will pick up the list walk in * some indeterminate state. (Otherwise every caller would * have to pass in a cookie to keep the state represented by * io_walk_link, which gets annoying.) */ zio_t * zio_walk_parents(zio_t *cio) { zio_link_t *zl = cio->io_walk_link; list_t *pl = &cio->io_parent_list; zl = (zl == NULL) ? list_head(pl) : list_next(pl, zl); cio->io_walk_link = zl; if (zl == NULL) return (NULL); ASSERT(zl->zl_child == cio); return (zl->zl_parent); } zio_t * zio_walk_children(zio_t *pio) { zio_link_t *zl = pio->io_walk_link; list_t *cl = &pio->io_child_list; zl = (zl == NULL) ? list_head(cl) : list_next(cl, zl); pio->io_walk_link = zl; if (zl == NULL) return (NULL); ASSERT(zl->zl_parent == pio); return (zl->zl_child); } zio_t * zio_unique_parent(zio_t *cio) { zio_t *pio = zio_walk_parents(cio); VERIFY(zio_walk_parents(cio) == NULL); return (pio); } void zio_add_child(zio_t *pio, zio_t *cio) { zio_link_t *zl = kmem_cache_alloc(zio_link_cache, KM_SLEEP); /* * Logical I/Os can have logical, gang, or vdev children. * Gang I/Os can have gang or vdev children. * Vdev I/Os can only have vdev children. * The following ASSERT captures all of these constraints. */ ASSERT(cio->io_child_type <= pio->io_child_type); zl->zl_parent = pio; zl->zl_child = cio; mutex_enter(&cio->io_lock); mutex_enter(&pio->io_lock); ASSERT(pio->io_state[ZIO_WAIT_DONE] == 0); for (int w = 0; w < ZIO_WAIT_TYPES; w++) pio->io_children[cio->io_child_type][w] += !cio->io_state[w]; list_insert_head(&pio->io_child_list, zl); list_insert_head(&cio->io_parent_list, zl); pio->io_child_count++; cio->io_parent_count++; mutex_exit(&pio->io_lock); mutex_exit(&cio->io_lock); } static void zio_remove_child(zio_t *pio, zio_t *cio, zio_link_t *zl) { ASSERT(zl->zl_parent == pio); ASSERT(zl->zl_child == cio); mutex_enter(&cio->io_lock); mutex_enter(&pio->io_lock); list_remove(&pio->io_child_list, zl); list_remove(&cio->io_parent_list, zl); pio->io_child_count--; cio->io_parent_count--; mutex_exit(&pio->io_lock); mutex_exit(&cio->io_lock); kmem_cache_free(zio_link_cache, zl); } static boolean_t zio_wait_for_children(zio_t *zio, enum zio_child child, enum zio_wait_type wait) { uint64_t *countp = &zio->io_children[child][wait]; boolean_t waiting = B_FALSE; mutex_enter(&zio->io_lock); ASSERT(zio->io_stall == NULL); if (*countp != 0) { zio->io_stage >>= 1; zio->io_stall = countp; waiting = B_TRUE; } mutex_exit(&zio->io_lock); return (waiting); } static void zio_notify_parent(zio_t *pio, zio_t *zio, enum zio_wait_type wait) { uint64_t *countp = &pio->io_children[zio->io_child_type][wait]; int *errorp = &pio->io_child_error[zio->io_child_type]; mutex_enter(&pio->io_lock); if (zio->io_error && !(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE)) *errorp = zio_worst_error(*errorp, zio->io_error); pio->io_reexecute |= zio->io_reexecute; ASSERT3U(*countp, >, 0); (*countp)--; if (*countp == 0 && pio->io_stall == countp) { pio->io_stall = NULL; mutex_exit(&pio->io_lock); zio_execute(pio); } else { mutex_exit(&pio->io_lock); } } static void zio_inherit_child_errors(zio_t *zio, enum zio_child c) { if (zio->io_child_error[c] != 0 && zio->io_error == 0) zio->io_error = zio->io_child_error[c]; } /* * ========================================================================== * Create the various types of I/O (read, write, free, etc) * ========================================================================== */ static zio_t * zio_create(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, void *data, uint64_t size, zio_done_func_t *done, void *private, zio_type_t type, zio_priority_t priority, enum zio_flag flags, vdev_t *vd, uint64_t offset, const zbookmark_phys_t *zb, enum zio_stage stage, enum zio_stage pipeline) { zio_t *zio; ASSERT3U(type == ZIO_TYPE_FREE || size, <=, SPA_MAXBLOCKSIZE); ASSERT(P2PHASE(size, SPA_MINBLOCKSIZE) == 0); ASSERT(P2PHASE(offset, SPA_MINBLOCKSIZE) == 0); ASSERT(!vd || spa_config_held(spa, SCL_STATE_ALL, RW_READER)); ASSERT(!bp || !(flags & ZIO_FLAG_CONFIG_WRITER)); ASSERT(vd || stage == ZIO_STAGE_OPEN); zio = kmem_cache_alloc(zio_cache, KM_SLEEP); bzero(zio, sizeof (zio_t)); mutex_init(&zio->io_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&zio->io_cv, NULL, CV_DEFAULT, NULL); list_create(&zio->io_parent_list, sizeof (zio_link_t), offsetof(zio_link_t, zl_parent_node)); list_create(&zio->io_child_list, sizeof (zio_link_t), offsetof(zio_link_t, zl_child_node)); if (vd != NULL) zio->io_child_type = ZIO_CHILD_VDEV; else if (flags & ZIO_FLAG_GANG_CHILD) zio->io_child_type = ZIO_CHILD_GANG; else if (flags & ZIO_FLAG_DDT_CHILD) zio->io_child_type = ZIO_CHILD_DDT; else zio->io_child_type = ZIO_CHILD_LOGICAL; if (bp != NULL) { zio->io_bp = (blkptr_t *)bp; zio->io_bp_copy = *bp; zio->io_bp_orig = *bp; if (type != ZIO_TYPE_WRITE || zio->io_child_type == ZIO_CHILD_DDT) zio->io_bp = &zio->io_bp_copy; /* so caller can free */ if (zio->io_child_type == ZIO_CHILD_LOGICAL) zio->io_logical = zio; if (zio->io_child_type > ZIO_CHILD_GANG && BP_IS_GANG(bp)) pipeline |= ZIO_GANG_STAGES; } zio->io_spa = spa; zio->io_txg = txg; zio->io_done = done; zio->io_private = private; zio->io_type = type; zio->io_priority = priority; zio->io_vd = vd; zio->io_offset = offset; zio->io_orig_data = zio->io_data = data; zio->io_orig_size = zio->io_size = size; zio->io_orig_flags = zio->io_flags = flags; zio->io_orig_stage = zio->io_stage = stage; zio->io_orig_pipeline = zio->io_pipeline = pipeline; zio->io_state[ZIO_WAIT_READY] = (stage >= ZIO_STAGE_READY); zio->io_state[ZIO_WAIT_DONE] = (stage >= ZIO_STAGE_DONE); if (zb != NULL) zio->io_bookmark = *zb; if (pio != NULL) { if (zio->io_logical == NULL) zio->io_logical = pio->io_logical; if (zio->io_child_type == ZIO_CHILD_GANG) zio->io_gang_leader = pio->io_gang_leader; zio_add_child(pio, zio); } return (zio); } static void zio_destroy(zio_t *zio) { list_destroy(&zio->io_parent_list); list_destroy(&zio->io_child_list); mutex_destroy(&zio->io_lock); cv_destroy(&zio->io_cv); kmem_cache_free(zio_cache, zio); } zio_t * zio_null(zio_t *pio, spa_t *spa, vdev_t *vd, zio_done_func_t *done, void *private, enum zio_flag flags) { zio_t *zio; zio = zio_create(pio, spa, 0, NULL, NULL, 0, done, private, ZIO_TYPE_NULL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL, ZIO_STAGE_OPEN, ZIO_INTERLOCK_PIPELINE); return (zio); } zio_t * zio_root(spa_t *spa, zio_done_func_t *done, void *private, enum zio_flag flags) { return (zio_null(NULL, spa, NULL, done, private, flags)); } void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp) { if (!DMU_OT_IS_VALID(BP_GET_TYPE(bp))) { zfs_panic_recover("blkptr at %p has invalid TYPE %llu", bp, (longlong_t)BP_GET_TYPE(bp)); } if (BP_GET_CHECKSUM(bp) >= ZIO_CHECKSUM_FUNCTIONS || BP_GET_CHECKSUM(bp) <= ZIO_CHECKSUM_ON) { zfs_panic_recover("blkptr at %p has invalid CHECKSUM %llu", bp, (longlong_t)BP_GET_CHECKSUM(bp)); } if (BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_FUNCTIONS || BP_GET_COMPRESS(bp) <= ZIO_COMPRESS_ON) { zfs_panic_recover("blkptr at %p has invalid COMPRESS %llu", bp, (longlong_t)BP_GET_COMPRESS(bp)); } if (BP_GET_LSIZE(bp) > SPA_MAXBLOCKSIZE) { zfs_panic_recover("blkptr at %p has invalid LSIZE %llu", bp, (longlong_t)BP_GET_LSIZE(bp)); } if (BP_GET_PSIZE(bp) > SPA_MAXBLOCKSIZE) { zfs_panic_recover("blkptr at %p has invalid PSIZE %llu", bp, (longlong_t)BP_GET_PSIZE(bp)); } if (BP_IS_EMBEDDED(bp)) { if (BPE_GET_ETYPE(bp) > NUM_BP_EMBEDDED_TYPES) { zfs_panic_recover("blkptr at %p has invalid ETYPE %llu", bp, (longlong_t)BPE_GET_ETYPE(bp)); } } /* * Pool-specific checks. * * Note: it would be nice to verify that the blk_birth and * BP_PHYSICAL_BIRTH() are not too large. However, spa_freeze() * allows the birth time of log blocks (and dmu_sync()-ed blocks * that are in the log) to be arbitrarily large. */ for (int i = 0; i < BP_GET_NDVAS(bp); i++) { uint64_t vdevid = DVA_GET_VDEV(&bp->blk_dva[i]); if (vdevid >= spa->spa_root_vdev->vdev_children) { zfs_panic_recover("blkptr at %p DVA %u has invalid " "VDEV %llu", bp, i, (longlong_t)vdevid); continue; } vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid]; if (vd == NULL) { zfs_panic_recover("blkptr at %p DVA %u has invalid " "VDEV %llu", bp, i, (longlong_t)vdevid); continue; } if (vd->vdev_ops == &vdev_hole_ops) { zfs_panic_recover("blkptr at %p DVA %u has hole " "VDEV %llu", bp, i, (longlong_t)vdevid); continue; } if (vd->vdev_ops == &vdev_missing_ops) { /* * "missing" vdevs are valid during import, but we * don't have their detailed info (e.g. asize), so * we can't perform any more checks on them. */ continue; } uint64_t offset = DVA_GET_OFFSET(&bp->blk_dva[i]); uint64_t asize = DVA_GET_ASIZE(&bp->blk_dva[i]); if (BP_IS_GANG(bp)) asize = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE); if (offset + asize > vd->vdev_asize) { zfs_panic_recover("blkptr at %p DVA %u has invalid " "OFFSET %llu", bp, i, (longlong_t)offset); } } } zio_t * zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, void *data, uint64_t size, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb) { zio_t *zio; zfs_blkptr_verify(spa, bp); zio = zio_create(pio, spa, BP_PHYSICAL_BIRTH(bp), bp, data, size, done, private, ZIO_TYPE_READ, priority, flags, NULL, 0, zb, ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ? ZIO_DDT_CHILD_READ_PIPELINE : ZIO_READ_PIPELINE); return (zio); } zio_t * zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, void *data, uint64_t size, const zio_prop_t *zp, zio_done_func_t *ready, zio_done_func_t *physdone, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb) { zio_t *zio; ASSERT(zp->zp_checksum >= ZIO_CHECKSUM_OFF && zp->zp_checksum < ZIO_CHECKSUM_FUNCTIONS && zp->zp_compress >= ZIO_COMPRESS_OFF && zp->zp_compress < ZIO_COMPRESS_FUNCTIONS && DMU_OT_IS_VALID(zp->zp_type) && zp->zp_level < 32 && zp->zp_copies > 0 && zp->zp_copies <= spa_max_replication(spa)); zio = zio_create(pio, spa, txg, bp, data, size, done, private, ZIO_TYPE_WRITE, priority, flags, NULL, 0, zb, ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ? ZIO_DDT_CHILD_WRITE_PIPELINE : ZIO_WRITE_PIPELINE); zio->io_ready = ready; zio->io_physdone = physdone; zio->io_prop = *zp; /* * Data can be NULL if we are going to call zio_write_override() to * provide the already-allocated BP. But we may need the data to * verify a dedup hit (if requested). In this case, don't try to * dedup (just take the already-allocated BP verbatim). */ if (data == NULL && zio->io_prop.zp_dedup_verify) { zio->io_prop.zp_dedup = zio->io_prop.zp_dedup_verify = B_FALSE; } return (zio); } zio_t * zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, void *data, uint64_t size, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb) { zio_t *zio; zio = zio_create(pio, spa, txg, bp, data, size, done, private, ZIO_TYPE_WRITE, priority, flags, NULL, 0, zb, ZIO_STAGE_OPEN, ZIO_REWRITE_PIPELINE); return (zio); } void zio_write_override(zio_t *zio, blkptr_t *bp, int copies, boolean_t nopwrite) { ASSERT(zio->io_type == ZIO_TYPE_WRITE); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(zio->io_stage == ZIO_STAGE_OPEN); ASSERT(zio->io_txg == spa_syncing_txg(zio->io_spa)); /* * We must reset the io_prop to match the values that existed * when the bp was first written by dmu_sync() keeping in mind * that nopwrite and dedup are mutually exclusive. */ zio->io_prop.zp_dedup = nopwrite ? B_FALSE : zio->io_prop.zp_dedup; zio->io_prop.zp_nopwrite = nopwrite; zio->io_prop.zp_copies = copies; zio->io_bp_override = bp; } void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp) { /* * The check for EMBEDDED is a performance optimization. We * process the free here (by ignoring it) rather than * putting it on the list and then processing it in zio_free_sync(). */ if (BP_IS_EMBEDDED(bp)) return; metaslab_check_free(spa, bp); /* * Frees that are for the currently-syncing txg, are not going to be * deferred, and which will not need to do a read (i.e. not GANG or * DEDUP), can be processed immediately. Otherwise, put them on the * in-memory list for later processing. */ if (zfs_trim_enabled || BP_IS_GANG(bp) || BP_GET_DEDUP(bp) || txg != spa->spa_syncing_txg || spa_sync_pass(spa) >= zfs_sync_pass_deferred_free) { bplist_append(&spa->spa_free_bplist[txg & TXG_MASK], bp); } else { VERIFY0(zio_wait(zio_free_sync(NULL, spa, txg, bp, BP_GET_PSIZE(bp), 0))); } } zio_t * zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, uint64_t size, enum zio_flag flags) { zio_t *zio; enum zio_stage stage = ZIO_FREE_PIPELINE; ASSERT(!BP_IS_HOLE(bp)); ASSERT(spa_syncing_txg(spa) == txg); ASSERT(spa_sync_pass(spa) < zfs_sync_pass_deferred_free); if (BP_IS_EMBEDDED(bp)) return (zio_null(pio, spa, NULL, NULL, NULL, 0)); metaslab_check_free(spa, bp); arc_freed(spa, bp); if (zfs_trim_enabled) stage |= ZIO_STAGE_ISSUE_ASYNC | ZIO_STAGE_VDEV_IO_START | ZIO_STAGE_VDEV_IO_ASSESS; /* * GANG and DEDUP blocks can induce a read (for the gang block header, * or the DDT), so issue them asynchronously so that this thread is * not tied up. */ else if (BP_IS_GANG(bp) || BP_GET_DEDUP(bp)) stage |= ZIO_STAGE_ISSUE_ASYNC; flags |= ZIO_FLAG_DONT_QUEUE; zio = zio_create(pio, spa, txg, bp, NULL, size, NULL, NULL, ZIO_TYPE_FREE, ZIO_PRIORITY_NOW, flags, NULL, 0, NULL, ZIO_STAGE_OPEN, stage); return (zio); } zio_t * zio_claim(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, zio_done_func_t *done, void *private, enum zio_flag flags) { zio_t *zio; dprintf_bp(bp, "claiming in txg %llu", txg); if (BP_IS_EMBEDDED(bp)) return (zio_null(pio, spa, NULL, NULL, NULL, 0)); /* * A claim is an allocation of a specific block. Claims are needed * to support immediate writes in the intent log. The issue is that * immediate writes contain committed data, but in a txg that was * *not* committed. Upon opening the pool after an unclean shutdown, * the intent log claims all blocks that contain immediate write data * so that the SPA knows they're in use. * * All claims *must* be resolved in the first txg -- before the SPA * starts allocating blocks -- so that nothing is allocated twice. * If txg == 0 we just verify that the block is claimable. */ ASSERT3U(spa->spa_uberblock.ub_rootbp.blk_birth, <, spa_first_txg(spa)); ASSERT(txg == spa_first_txg(spa) || txg == 0); ASSERT(!BP_GET_DEDUP(bp) || !spa_writeable(spa)); /* zdb(1M) */ zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp), done, private, ZIO_TYPE_CLAIM, ZIO_PRIORITY_NOW, flags, NULL, 0, NULL, ZIO_STAGE_OPEN, ZIO_CLAIM_PIPELINE); return (zio); } zio_t * zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd, uint64_t offset, uint64_t size, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags) { zio_t *zio; int c; if (vd->vdev_children == 0) { zio = zio_create(pio, spa, 0, NULL, NULL, size, done, private, ZIO_TYPE_IOCTL, priority, flags, vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_IOCTL_PIPELINE); zio->io_cmd = cmd; } else { zio = zio_null(pio, spa, NULL, NULL, NULL, flags); for (c = 0; c < vd->vdev_children; c++) zio_nowait(zio_ioctl(zio, spa, vd->vdev_child[c], cmd, offset, size, done, private, priority, flags)); } return (zio); } zio_t * zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, void *data, int checksum, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, boolean_t labels) { zio_t *zio; ASSERT(vd->vdev_children == 0); ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE || offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE); ASSERT3U(offset + size, <=, vd->vdev_psize); zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, done, private, ZIO_TYPE_READ, priority, flags | ZIO_FLAG_PHYSICAL, vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_READ_PHYS_PIPELINE); zio->io_prop.zp_checksum = checksum; return (zio); } zio_t * zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, void *data, int checksum, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, boolean_t labels) { zio_t *zio; ASSERT(vd->vdev_children == 0); ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE || offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE); ASSERT3U(offset + size, <=, vd->vdev_psize); zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, done, private, ZIO_TYPE_WRITE, priority, flags | ZIO_FLAG_PHYSICAL, vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_WRITE_PHYS_PIPELINE); zio->io_prop.zp_checksum = checksum; - if (zio_checksum_table[checksum].ci_eck) { + if (zio_checksum_table[checksum].ci_flags & ZCHECKSUM_FLAG_EMBEDDED) { /* * zec checksums are necessarily destructive -- they modify * the end of the write buffer to hold the verifier/checksum. * Therefore, we must make a local copy in case the data is * being written to multiple places in parallel. */ void *wbuf = zio_buf_alloc(size); bcopy(data, wbuf, size); zio_push_transform(zio, wbuf, size, size, NULL); } return (zio); } /* * Create a child I/O to do some work for us. */ zio_t * zio_vdev_child_io(zio_t *pio, blkptr_t *bp, vdev_t *vd, uint64_t offset, void *data, uint64_t size, int type, zio_priority_t priority, enum zio_flag flags, zio_done_func_t *done, void *private) { enum zio_stage pipeline = ZIO_VDEV_CHILD_PIPELINE; zio_t *zio; ASSERT(vd->vdev_parent == (pio->io_vd ? pio->io_vd : pio->io_spa->spa_root_vdev)); if (type == ZIO_TYPE_READ && bp != NULL) { /* * If we have the bp, then the child should perform the * checksum and the parent need not. This pushes error * detection as close to the leaves as possible and * eliminates redundant checksums in the interior nodes. */ pipeline |= ZIO_STAGE_CHECKSUM_VERIFY; pio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY; } /* Not all IO types require vdev io done stage e.g. free */ if (!(pio->io_pipeline & ZIO_STAGE_VDEV_IO_DONE)) pipeline &= ~ZIO_STAGE_VDEV_IO_DONE; if (vd->vdev_children == 0) offset += VDEV_LABEL_START_SIZE; flags |= ZIO_VDEV_CHILD_FLAGS(pio) | ZIO_FLAG_DONT_PROPAGATE; /* * If we've decided to do a repair, the write is not speculative -- * even if the original read was. */ if (flags & ZIO_FLAG_IO_REPAIR) flags &= ~ZIO_FLAG_SPECULATIVE; zio = zio_create(pio, pio->io_spa, pio->io_txg, bp, data, size, done, private, type, priority, flags, vd, offset, &pio->io_bookmark, ZIO_STAGE_VDEV_IO_START >> 1, pipeline); zio->io_physdone = pio->io_physdone; if (vd->vdev_ops->vdev_op_leaf && zio->io_logical != NULL) zio->io_logical->io_phys_children++; return (zio); } zio_t * zio_vdev_delegated_io(vdev_t *vd, uint64_t offset, void *data, uint64_t size, int type, zio_priority_t priority, enum zio_flag flags, zio_done_func_t *done, void *private) { zio_t *zio; ASSERT(vd->vdev_ops->vdev_op_leaf); zio = zio_create(NULL, vd->vdev_spa, 0, NULL, data, size, done, private, type, priority, flags | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_RETRY | ZIO_FLAG_DELEGATED, vd, offset, NULL, ZIO_STAGE_VDEV_IO_START >> 1, ZIO_VDEV_CHILD_PIPELINE); return (zio); } void zio_flush(zio_t *zio, vdev_t *vd) { zio_nowait(zio_ioctl(zio, zio->io_spa, vd, DKIOCFLUSHWRITECACHE, 0, 0, NULL, NULL, ZIO_PRIORITY_NOW, ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY)); } zio_t * zio_trim(zio_t *zio, spa_t *spa, vdev_t *vd, uint64_t offset, uint64_t size) { ASSERT(vd->vdev_ops->vdev_op_leaf); return (zio_create(zio, spa, 0, NULL, NULL, size, NULL, NULL, ZIO_TYPE_FREE, ZIO_PRIORITY_TRIM, ZIO_FLAG_DONT_AGGREGATE | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY, vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_FREE_PHYS_PIPELINE)); } void zio_shrink(zio_t *zio, uint64_t size) { ASSERT(zio->io_executor == NULL); ASSERT(zio->io_orig_size == zio->io_size); ASSERT(size <= zio->io_size); /* * We don't shrink for raidz because of problems with the * reconstruction when reading back less than the block size. * Note, BP_IS_RAIDZ() assumes no compression. */ ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF); if (!BP_IS_RAIDZ(zio->io_bp)) zio->io_orig_size = zio->io_size = size; } /* * ========================================================================== * Prepare to read and write logical blocks * ========================================================================== */ static int zio_read_bp_init(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF && zio->io_child_type == ZIO_CHILD_LOGICAL && !(zio->io_flags & ZIO_FLAG_RAW)) { uint64_t psize = BP_IS_EMBEDDED(bp) ? BPE_GET_PSIZE(bp) : BP_GET_PSIZE(bp); void *cbuf = zio_buf_alloc(psize); zio_push_transform(zio, cbuf, psize, psize, zio_decompress); } if (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA) { zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; decode_embedded_bp_compressed(bp, zio->io_data); } else { ASSERT(!BP_IS_EMBEDDED(bp)); } if (!DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) && BP_GET_LEVEL(bp) == 0) zio->io_flags |= ZIO_FLAG_DONT_CACHE; if (BP_GET_TYPE(bp) == DMU_OT_DDT_ZAP) zio->io_flags |= ZIO_FLAG_DONT_CACHE; if (BP_GET_DEDUP(bp) && zio->io_child_type == ZIO_CHILD_LOGICAL) zio->io_pipeline = ZIO_DDT_READ_PIPELINE; return (ZIO_PIPELINE_CONTINUE); } static int zio_write_bp_init(zio_t *zio) { spa_t *spa = zio->io_spa; zio_prop_t *zp = &zio->io_prop; enum zio_compress compress = zp->zp_compress; blkptr_t *bp = zio->io_bp; uint64_t lsize = zio->io_size; uint64_t psize = lsize; int pass = 1; /* * If our children haven't all reached the ready stage, * wait for them and then repeat this pipeline stage. */ if (zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_READY) || zio_wait_for_children(zio, ZIO_CHILD_LOGICAL, ZIO_WAIT_READY)) return (ZIO_PIPELINE_STOP); if (!IO_IS_ALLOCATING(zio)) return (ZIO_PIPELINE_CONTINUE); ASSERT(zio->io_child_type != ZIO_CHILD_DDT); if (zio->io_bp_override) { ASSERT(bp->blk_birth != zio->io_txg); ASSERT(BP_GET_DEDUP(zio->io_bp_override) == 0); *bp = *zio->io_bp_override; zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; if (BP_IS_EMBEDDED(bp)) return (ZIO_PIPELINE_CONTINUE); /* * If we've been overridden and nopwrite is set then * set the flag accordingly to indicate that a nopwrite * has already occurred. */ if (!BP_IS_HOLE(bp) && zp->zp_nopwrite) { ASSERT(!zp->zp_dedup); zio->io_flags |= ZIO_FLAG_NOPWRITE; return (ZIO_PIPELINE_CONTINUE); } ASSERT(!zp->zp_nopwrite); if (BP_IS_HOLE(bp) || !zp->zp_dedup) return (ZIO_PIPELINE_CONTINUE); - ASSERT(zio_checksum_table[zp->zp_checksum].ci_dedup || - zp->zp_dedup_verify); + ASSERT((zio_checksum_table[zp->zp_checksum].ci_flags & + ZCHECKSUM_FLAG_DEDUP) || zp->zp_dedup_verify); if (BP_GET_CHECKSUM(bp) == zp->zp_checksum) { BP_SET_DEDUP(bp, 1); zio->io_pipeline |= ZIO_STAGE_DDT_WRITE; return (ZIO_PIPELINE_CONTINUE); } } if (!BP_IS_HOLE(bp) && bp->blk_birth == zio->io_txg) { /* * We're rewriting an existing block, which means we're * working on behalf of spa_sync(). For spa_sync() to * converge, it must eventually be the case that we don't * have to allocate new blocks. But compression changes * the blocksize, which forces a reallocate, and makes * convergence take longer. Therefore, after the first * few passes, stop compressing to ensure convergence. */ pass = spa_sync_pass(spa); ASSERT(zio->io_txg == spa_syncing_txg(spa)); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(!BP_GET_DEDUP(bp)); if (pass >= zfs_sync_pass_dont_compress) compress = ZIO_COMPRESS_OFF; /* Make sure someone doesn't change their mind on overwrites */ ASSERT(BP_IS_EMBEDDED(bp) || MIN(zp->zp_copies + BP_IS_GANG(bp), spa_max_replication(spa)) == BP_GET_NDVAS(bp)); } if (compress != ZIO_COMPRESS_OFF) { void *cbuf = zio_buf_alloc(lsize); psize = zio_compress_data(compress, zio->io_data, cbuf, lsize); if (psize == 0 || psize == lsize) { compress = ZIO_COMPRESS_OFF; zio_buf_free(cbuf, lsize); } else if (!zp->zp_dedup && psize <= BPE_PAYLOAD_SIZE && zp->zp_level == 0 && !DMU_OT_HAS_FILL(zp->zp_type) && spa_feature_is_enabled(spa, SPA_FEATURE_EMBEDDED_DATA)) { encode_embedded_bp_compressed(bp, cbuf, compress, lsize, psize); BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_DATA); BP_SET_TYPE(bp, zio->io_prop.zp_type); BP_SET_LEVEL(bp, zio->io_prop.zp_level); zio_buf_free(cbuf, lsize); bp->blk_birth = zio->io_txg; zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; ASSERT(spa_feature_is_active(spa, SPA_FEATURE_EMBEDDED_DATA)); return (ZIO_PIPELINE_CONTINUE); } else { /* * Round up compressed size up to the ashift * of the smallest-ashift device, and zero the tail. * This ensures that the compressed size of the BP * (and thus compressratio property) are correct, * in that we charge for the padding used to fill out * the last sector. */ ASSERT3U(spa->spa_min_ashift, >=, SPA_MINBLOCKSHIFT); size_t rounded = (size_t)P2ROUNDUP(psize, 1ULL << spa->spa_min_ashift); if (rounded >= lsize) { compress = ZIO_COMPRESS_OFF; zio_buf_free(cbuf, lsize); psize = lsize; } else { bzero((char *)cbuf + psize, rounded - psize); psize = rounded; zio_push_transform(zio, cbuf, psize, lsize, NULL); } } } /* * The final pass of spa_sync() must be all rewrites, but the first * few passes offer a trade-off: allocating blocks defers convergence, * but newly allocated blocks are sequential, so they can be written * to disk faster. Therefore, we allow the first few passes of * spa_sync() to allocate new blocks, but force rewrites after that. * There should only be a handful of blocks after pass 1 in any case. */ if (!BP_IS_HOLE(bp) && bp->blk_birth == zio->io_txg && BP_GET_PSIZE(bp) == psize && pass >= zfs_sync_pass_rewrite) { ASSERT(psize != 0); enum zio_stage gang_stages = zio->io_pipeline & ZIO_GANG_STAGES; zio->io_pipeline = ZIO_REWRITE_PIPELINE | gang_stages; zio->io_flags |= ZIO_FLAG_IO_REWRITE; } else { BP_ZERO(bp); zio->io_pipeline = ZIO_WRITE_PIPELINE; } if (psize == 0) { if (zio->io_bp_orig.blk_birth != 0 && spa_feature_is_active(spa, SPA_FEATURE_HOLE_BIRTH)) { BP_SET_LSIZE(bp, lsize); BP_SET_TYPE(bp, zp->zp_type); BP_SET_LEVEL(bp, zp->zp_level); BP_SET_BIRTH(bp, zio->io_txg, 0); } zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; } else { ASSERT(zp->zp_checksum != ZIO_CHECKSUM_GANG_HEADER); BP_SET_LSIZE(bp, lsize); BP_SET_TYPE(bp, zp->zp_type); BP_SET_LEVEL(bp, zp->zp_level); BP_SET_PSIZE(bp, psize); BP_SET_COMPRESS(bp, compress); BP_SET_CHECKSUM(bp, zp->zp_checksum); BP_SET_DEDUP(bp, zp->zp_dedup); BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER); if (zp->zp_dedup) { ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); zio->io_pipeline = ZIO_DDT_WRITE_PIPELINE; } if (zp->zp_nopwrite) { ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); zio->io_pipeline |= ZIO_STAGE_NOP_WRITE; } } return (ZIO_PIPELINE_CONTINUE); } static int zio_free_bp_init(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (zio->io_child_type == ZIO_CHILD_LOGICAL) { if (BP_GET_DEDUP(bp)) zio->io_pipeline = ZIO_DDT_FREE_PIPELINE; } return (ZIO_PIPELINE_CONTINUE); } /* * ========================================================================== * Execute the I/O pipeline * ========================================================================== */ static void zio_taskq_dispatch(zio_t *zio, zio_taskq_type_t q, boolean_t cutinline) { spa_t *spa = zio->io_spa; zio_type_t t = zio->io_type; int flags = (cutinline ? TQ_FRONT : 0); ASSERT(q == ZIO_TASKQ_ISSUE || q == ZIO_TASKQ_INTERRUPT); /* * If we're a config writer or a probe, the normal issue and * interrupt threads may all be blocked waiting for the config lock. * In this case, select the otherwise-unused taskq for ZIO_TYPE_NULL. */ if (zio->io_flags & (ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_PROBE)) t = ZIO_TYPE_NULL; /* * A similar issue exists for the L2ARC write thread until L2ARC 2.0. */ if (t == ZIO_TYPE_WRITE && zio->io_vd && zio->io_vd->vdev_aux) t = ZIO_TYPE_NULL; /* * If this is a high priority I/O, then use the high priority taskq if * available. */ if (zio->io_priority == ZIO_PRIORITY_NOW && spa->spa_zio_taskq[t][q + 1].stqs_count != 0) q++; ASSERT3U(q, <, ZIO_TASKQ_TYPES); /* * NB: We are assuming that the zio can only be dispatched * to a single taskq at a time. It would be a grievous error * to dispatch the zio to another taskq at the same time. */ #if defined(illumos) || !defined(_KERNEL) ASSERT(zio->io_tqent.tqent_next == NULL); #else ASSERT(zio->io_tqent.tqent_task.ta_pending == 0); #endif spa_taskq_dispatch_ent(spa, t, q, (task_func_t *)zio_execute, zio, flags, &zio->io_tqent); } static boolean_t zio_taskq_member(zio_t *zio, zio_taskq_type_t q) { kthread_t *executor = zio->io_executor; spa_t *spa = zio->io_spa; for (zio_type_t t = 0; t < ZIO_TYPES; t++) { spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; uint_t i; for (i = 0; i < tqs->stqs_count; i++) { if (taskq_member(tqs->stqs_taskq[i], executor)) return (B_TRUE); } } return (B_FALSE); } static int zio_issue_async(zio_t *zio) { zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE); return (ZIO_PIPELINE_STOP); } void zio_interrupt(zio_t *zio) { zio_taskq_dispatch(zio, ZIO_TASKQ_INTERRUPT, B_FALSE); } /* * Execute the I/O pipeline until one of the following occurs: * * (1) the I/O completes * (2) the pipeline stalls waiting for dependent child I/Os * (3) the I/O issues, so we're waiting for an I/O completion interrupt * (4) the I/O is delegated by vdev-level caching or aggregation * (5) the I/O is deferred due to vdev-level queueing * (6) the I/O is handed off to another thread. * * In all cases, the pipeline stops whenever there's no CPU work; it never * burns a thread in cv_wait(). * * There's no locking on io_stage because there's no legitimate way * for multiple threads to be attempting to process the same I/O. */ static zio_pipe_stage_t *zio_pipeline[]; void zio_execute(zio_t *zio) { zio->io_executor = curthread; while (zio->io_stage < ZIO_STAGE_DONE) { enum zio_stage pipeline = zio->io_pipeline; enum zio_stage stage = zio->io_stage; int rv; ASSERT(!MUTEX_HELD(&zio->io_lock)); ASSERT(ISP2(stage)); ASSERT(zio->io_stall == NULL); do { stage <<= 1; } while ((stage & pipeline) == 0); ASSERT(stage <= ZIO_STAGE_DONE); /* * If we are in interrupt context and this pipeline stage * will grab a config lock that is held across I/O, * or may wait for an I/O that needs an interrupt thread * to complete, issue async to avoid deadlock. * * For VDEV_IO_START, we cut in line so that the io will * be sent to disk promptly. */ if ((stage & ZIO_BLOCKING_STAGES) && zio->io_vd == NULL && zio_taskq_member(zio, ZIO_TASKQ_INTERRUPT)) { boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ? zio_requeue_io_start_cut_in_line : B_FALSE; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut); return; } zio->io_stage = stage; rv = zio_pipeline[highbit64(stage) - 1](zio); if (rv == ZIO_PIPELINE_STOP) return; ASSERT(rv == ZIO_PIPELINE_CONTINUE); } } /* * ========================================================================== * Initiate I/O, either sync or async * ========================================================================== */ int zio_wait(zio_t *zio) { int error; ASSERT(zio->io_stage == ZIO_STAGE_OPEN); ASSERT(zio->io_executor == NULL); zio->io_waiter = curthread; zio_execute(zio); mutex_enter(&zio->io_lock); while (zio->io_executor != NULL) cv_wait(&zio->io_cv, &zio->io_lock); mutex_exit(&zio->io_lock); error = zio->io_error; zio_destroy(zio); return (error); } void zio_nowait(zio_t *zio) { ASSERT(zio->io_executor == NULL); if (zio->io_child_type == ZIO_CHILD_LOGICAL && zio_unique_parent(zio) == NULL) { /* * This is a logical async I/O with no parent to wait for it. * We add it to the spa_async_root_zio "Godfather" I/O which * will ensure they complete prior to unloading the pool. */ spa_t *spa = zio->io_spa; zio_add_child(spa->spa_async_zio_root[CPU_SEQID], zio); } zio_execute(zio); } /* * ========================================================================== * Reexecute or suspend/resume failed I/O * ========================================================================== */ static void zio_reexecute(zio_t *pio) { zio_t *cio, *cio_next; ASSERT(pio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(pio->io_orig_stage == ZIO_STAGE_OPEN); ASSERT(pio->io_gang_leader == NULL); ASSERT(pio->io_gang_tree == NULL); pio->io_flags = pio->io_orig_flags; pio->io_stage = pio->io_orig_stage; pio->io_pipeline = pio->io_orig_pipeline; pio->io_reexecute = 0; pio->io_flags |= ZIO_FLAG_REEXECUTED; pio->io_error = 0; for (int w = 0; w < ZIO_WAIT_TYPES; w++) pio->io_state[w] = 0; for (int c = 0; c < ZIO_CHILD_TYPES; c++) pio->io_child_error[c] = 0; if (IO_IS_ALLOCATING(pio)) BP_ZERO(pio->io_bp); /* * As we reexecute pio's children, new children could be created. * New children go to the head of pio's io_child_list, however, * so we will (correctly) not reexecute them. The key is that * the remainder of pio's io_child_list, from 'cio_next' onward, * cannot be affected by any side effects of reexecuting 'cio'. */ for (cio = zio_walk_children(pio); cio != NULL; cio = cio_next) { cio_next = zio_walk_children(pio); mutex_enter(&pio->io_lock); for (int w = 0; w < ZIO_WAIT_TYPES; w++) pio->io_children[cio->io_child_type][w]++; mutex_exit(&pio->io_lock); zio_reexecute(cio); } /* * Now that all children have been reexecuted, execute the parent. * We don't reexecute "The Godfather" I/O here as it's the * responsibility of the caller to wait on him. */ if (!(pio->io_flags & ZIO_FLAG_GODFATHER)) zio_execute(pio); } void zio_suspend(spa_t *spa, zio_t *zio) { if (spa_get_failmode(spa) == ZIO_FAILURE_MODE_PANIC) fm_panic("Pool '%s' has encountered an uncorrectable I/O " "failure and the failure mode property for this pool " "is set to panic.", spa_name(spa)); zfs_ereport_post(FM_EREPORT_ZFS_IO_FAILURE, spa, NULL, NULL, 0, 0); mutex_enter(&spa->spa_suspend_lock); if (spa->spa_suspend_zio_root == NULL) spa->spa_suspend_zio_root = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_GODFATHER); spa->spa_suspended = B_TRUE; if (zio != NULL) { ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER)); ASSERT(zio != spa->spa_suspend_zio_root); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(zio_unique_parent(zio) == NULL); ASSERT(zio->io_stage == ZIO_STAGE_DONE); zio_add_child(spa->spa_suspend_zio_root, zio); } mutex_exit(&spa->spa_suspend_lock); } int zio_resume(spa_t *spa) { zio_t *pio; /* * Reexecute all previously suspended i/o. */ mutex_enter(&spa->spa_suspend_lock); spa->spa_suspended = B_FALSE; cv_broadcast(&spa->spa_suspend_cv); pio = spa->spa_suspend_zio_root; spa->spa_suspend_zio_root = NULL; mutex_exit(&spa->spa_suspend_lock); if (pio == NULL) return (0); zio_reexecute(pio); return (zio_wait(pio)); } void zio_resume_wait(spa_t *spa) { mutex_enter(&spa->spa_suspend_lock); while (spa_suspended(spa)) cv_wait(&spa->spa_suspend_cv, &spa->spa_suspend_lock); mutex_exit(&spa->spa_suspend_lock); } /* * ========================================================================== * Gang blocks. * * A gang block is a collection of small blocks that looks to the DMU * like one large block. When zio_dva_allocate() cannot find a block * of the requested size, due to either severe fragmentation or the pool * being nearly full, it calls zio_write_gang_block() to construct the * block from smaller fragments. * * A gang block consists of a gang header (zio_gbh_phys_t) and up to * three (SPA_GBH_NBLKPTRS) gang members. The gang header is just like * an indirect block: it's an array of block pointers. It consumes * only one sector and hence is allocatable regardless of fragmentation. * The gang header's bps point to its gang members, which hold the data. * * Gang blocks are self-checksumming, using the bp's * as the verifier to ensure uniqueness of the SHA256 checksum. * Critically, the gang block bp's blk_cksum is the checksum of the data, * not the gang header. This ensures that data block signatures (needed for * deduplication) are independent of how the block is physically stored. * * Gang blocks can be nested: a gang member may itself be a gang block. * Thus every gang block is a tree in which root and all interior nodes are * gang headers, and the leaves are normal blocks that contain user data. * The root of the gang tree is called the gang leader. * * To perform any operation (read, rewrite, free, claim) on a gang block, * zio_gang_assemble() first assembles the gang tree (minus data leaves) * in the io_gang_tree field of the original logical i/o by recursively * reading the gang leader and all gang headers below it. This yields * an in-core tree containing the contents of every gang header and the * bps for every constituent of the gang block. * * With the gang tree now assembled, zio_gang_issue() just walks the gang tree * and invokes a callback on each bp. To free a gang block, zio_gang_issue() * calls zio_free_gang() -- a trivial wrapper around zio_free() -- for each bp. * zio_claim_gang() provides a similarly trivial wrapper for zio_claim(). * zio_read_gang() is a wrapper around zio_read() that omits reading gang * headers, since we already have those in io_gang_tree. zio_rewrite_gang() * performs a zio_rewrite() of the data or, for gang headers, a zio_rewrite() * of the gang header plus zio_checksum_compute() of the data to update the * gang header's blk_cksum as described above. * * The two-phase assemble/issue model solves the problem of partial failure -- * what if you'd freed part of a gang block but then couldn't read the * gang header for another part? Assembling the entire gang tree first * ensures that all the necessary gang header I/O has succeeded before * starting the actual work of free, claim, or write. Once the gang tree * is assembled, free and claim are in-memory operations that cannot fail. * * In the event that a gang write fails, zio_dva_unallocate() walks the * gang tree to immediately free (i.e. insert back into the space map) * everything we've allocated. This ensures that we don't get ENOSPC * errors during repeated suspend/resume cycles due to a flaky device. * * Gang rewrites only happen during sync-to-convergence. If we can't assemble * the gang tree, we won't modify the block, so we can safely defer the free * (knowing that the block is still intact). If we *can* assemble the gang * tree, then even if some of the rewrites fail, zio_dva_unallocate() will free * each constituent bp and we can allocate a new block on the next sync pass. * * In all cases, the gang tree allows complete recovery from partial failure. * ========================================================================== */ static zio_t * zio_read_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, void *data) { if (gn != NULL) return (pio); return (zio_read(pio, pio->io_spa, bp, data, BP_GET_PSIZE(bp), NULL, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark)); } zio_t * zio_rewrite_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, void *data) { zio_t *zio; if (gn != NULL) { zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp, gn->gn_gbh, SPA_GANGBLOCKSIZE, NULL, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); /* * As we rewrite each gang header, the pipeline will compute * a new gang block header checksum for it; but no one will * compute a new data checksum, so we do that here. The one * exception is the gang leader: the pipeline already computed * its data checksum because that stage precedes gang assembly. * (Presently, nothing actually uses interior data checksums; * this is just good hygiene.) */ if (gn != pio->io_gang_leader->io_gang_tree) { zio_checksum_compute(zio, BP_GET_CHECKSUM(bp), data, BP_GET_PSIZE(bp)); } /* * If we are here to damage data for testing purposes, * leave the GBH alone so that we can detect the damage. */ if (pio->io_gang_leader->io_flags & ZIO_FLAG_INDUCE_DAMAGE) zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES; } else { zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp, data, BP_GET_PSIZE(bp), NULL, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); } return (zio); } /* ARGSUSED */ zio_t * zio_free_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, void *data) { return (zio_free_sync(pio, pio->io_spa, pio->io_txg, bp, BP_IS_GANG(bp) ? SPA_GANGBLOCKSIZE : BP_GET_PSIZE(bp), ZIO_GANG_CHILD_FLAGS(pio))); } /* ARGSUSED */ zio_t * zio_claim_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, void *data) { return (zio_claim(pio, pio->io_spa, pio->io_txg, bp, NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio))); } static zio_gang_issue_func_t *zio_gang_issue_func[ZIO_TYPES] = { NULL, zio_read_gang, zio_rewrite_gang, zio_free_gang, zio_claim_gang, NULL }; static void zio_gang_tree_assemble_done(zio_t *zio); static zio_gang_node_t * zio_gang_node_alloc(zio_gang_node_t **gnpp) { zio_gang_node_t *gn; ASSERT(*gnpp == NULL); gn = kmem_zalloc(sizeof (*gn), KM_SLEEP); gn->gn_gbh = zio_buf_alloc(SPA_GANGBLOCKSIZE); *gnpp = gn; return (gn); } static void zio_gang_node_free(zio_gang_node_t **gnpp) { zio_gang_node_t *gn = *gnpp; for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) ASSERT(gn->gn_child[g] == NULL); zio_buf_free(gn->gn_gbh, SPA_GANGBLOCKSIZE); kmem_free(gn, sizeof (*gn)); *gnpp = NULL; } static void zio_gang_tree_free(zio_gang_node_t **gnpp) { zio_gang_node_t *gn = *gnpp; if (gn == NULL) return; for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) zio_gang_tree_free(&gn->gn_child[g]); zio_gang_node_free(gnpp); } static void zio_gang_tree_assemble(zio_t *gio, blkptr_t *bp, zio_gang_node_t **gnpp) { zio_gang_node_t *gn = zio_gang_node_alloc(gnpp); ASSERT(gio->io_gang_leader == gio); ASSERT(BP_IS_GANG(bp)); zio_nowait(zio_read(gio, gio->io_spa, bp, gn->gn_gbh, SPA_GANGBLOCKSIZE, zio_gang_tree_assemble_done, gn, gio->io_priority, ZIO_GANG_CHILD_FLAGS(gio), &gio->io_bookmark)); } static void zio_gang_tree_assemble_done(zio_t *zio) { zio_t *gio = zio->io_gang_leader; zio_gang_node_t *gn = zio->io_private; blkptr_t *bp = zio->io_bp; ASSERT(gio == zio_unique_parent(zio)); ASSERT(zio->io_child_count == 0); if (zio->io_error) return; if (BP_SHOULD_BYTESWAP(bp)) byteswap_uint64_array(zio->io_data, zio->io_size); ASSERT(zio->io_data == gn->gn_gbh); ASSERT(zio->io_size == SPA_GANGBLOCKSIZE); ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC); for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g]; if (!BP_IS_GANG(gbp)) continue; zio_gang_tree_assemble(gio, gbp, &gn->gn_child[g]); } } static void zio_gang_tree_issue(zio_t *pio, zio_gang_node_t *gn, blkptr_t *bp, void *data) { zio_t *gio = pio->io_gang_leader; zio_t *zio; ASSERT(BP_IS_GANG(bp) == !!gn); ASSERT(BP_GET_CHECKSUM(bp) == BP_GET_CHECKSUM(gio->io_bp)); ASSERT(BP_GET_LSIZE(bp) == BP_GET_PSIZE(bp) || gn == gio->io_gang_tree); /* * If you're a gang header, your data is in gn->gn_gbh. * If you're a gang member, your data is in 'data' and gn == NULL. */ zio = zio_gang_issue_func[gio->io_type](pio, bp, gn, data); if (gn != NULL) { ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC); for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g]; if (BP_IS_HOLE(gbp)) continue; zio_gang_tree_issue(zio, gn->gn_child[g], gbp, data); data = (char *)data + BP_GET_PSIZE(gbp); } } if (gn == gio->io_gang_tree && gio->io_data != NULL) ASSERT3P((char *)gio->io_data + gio->io_size, ==, data); if (zio != pio) zio_nowait(zio); } static int zio_gang_assemble(zio_t *zio) { blkptr_t *bp = zio->io_bp; ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == NULL); ASSERT(zio->io_child_type > ZIO_CHILD_GANG); zio->io_gang_leader = zio; zio_gang_tree_assemble(zio, bp, &zio->io_gang_tree); return (ZIO_PIPELINE_CONTINUE); } static int zio_gang_issue(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_DONE)) return (ZIO_PIPELINE_STOP); ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == zio); ASSERT(zio->io_child_type > ZIO_CHILD_GANG); if (zio->io_child_error[ZIO_CHILD_GANG] == 0) zio_gang_tree_issue(zio, zio->io_gang_tree, bp, zio->io_data); else zio_gang_tree_free(&zio->io_gang_tree); zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; return (ZIO_PIPELINE_CONTINUE); } static void zio_write_gang_member_ready(zio_t *zio) { zio_t *pio = zio_unique_parent(zio); zio_t *gio = zio->io_gang_leader; dva_t *cdva = zio->io_bp->blk_dva; dva_t *pdva = pio->io_bp->blk_dva; uint64_t asize; if (BP_IS_HOLE(zio->io_bp)) return; ASSERT(BP_IS_HOLE(&zio->io_bp_orig)); ASSERT(zio->io_child_type == ZIO_CHILD_GANG); ASSERT3U(zio->io_prop.zp_copies, ==, gio->io_prop.zp_copies); ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(zio->io_bp)); ASSERT3U(pio->io_prop.zp_copies, <=, BP_GET_NDVAS(pio->io_bp)); ASSERT3U(BP_GET_NDVAS(zio->io_bp), <=, BP_GET_NDVAS(pio->io_bp)); mutex_enter(&pio->io_lock); for (int d = 0; d < BP_GET_NDVAS(zio->io_bp); d++) { ASSERT(DVA_GET_GANG(&pdva[d])); asize = DVA_GET_ASIZE(&pdva[d]); asize += DVA_GET_ASIZE(&cdva[d]); DVA_SET_ASIZE(&pdva[d], asize); } mutex_exit(&pio->io_lock); } static int zio_write_gang_block(zio_t *pio) { spa_t *spa = pio->io_spa; blkptr_t *bp = pio->io_bp; zio_t *gio = pio->io_gang_leader; zio_t *zio; zio_gang_node_t *gn, **gnpp; zio_gbh_phys_t *gbh; uint64_t txg = pio->io_txg; uint64_t resid = pio->io_size; uint64_t lsize; int copies = gio->io_prop.zp_copies; int gbh_copies = MIN(copies + 1, spa_max_replication(spa)); zio_prop_t zp; int error; error = metaslab_alloc(spa, spa_normal_class(spa), SPA_GANGBLOCKSIZE, bp, gbh_copies, txg, pio == gio ? NULL : gio->io_bp, METASLAB_HINTBP_FAVOR | METASLAB_GANG_HEADER); if (error) { pio->io_error = error; return (ZIO_PIPELINE_CONTINUE); } if (pio == gio) { gnpp = &gio->io_gang_tree; } else { gnpp = pio->io_private; ASSERT(pio->io_ready == zio_write_gang_member_ready); } gn = zio_gang_node_alloc(gnpp); gbh = gn->gn_gbh; bzero(gbh, SPA_GANGBLOCKSIZE); /* * Create the gang header. */ zio = zio_rewrite(pio, spa, txg, bp, gbh, SPA_GANGBLOCKSIZE, NULL, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); /* * Create and nowait the gang children. */ for (int g = 0; resid != 0; resid -= lsize, g++) { lsize = P2ROUNDUP(resid / (SPA_GBH_NBLKPTRS - g), SPA_MINBLOCKSIZE); ASSERT(lsize >= SPA_MINBLOCKSIZE && lsize <= resid); zp.zp_checksum = gio->io_prop.zp_checksum; zp.zp_compress = ZIO_COMPRESS_OFF; zp.zp_type = DMU_OT_NONE; zp.zp_level = 0; zp.zp_copies = gio->io_prop.zp_copies; zp.zp_dedup = B_FALSE; zp.zp_dedup_verify = B_FALSE; zp.zp_nopwrite = B_FALSE; zio_nowait(zio_write(zio, spa, txg, &gbh->zg_blkptr[g], (char *)pio->io_data + (pio->io_size - resid), lsize, &zp, zio_write_gang_member_ready, NULL, NULL, &gn->gn_child[g], pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark)); } /* * Set pio's pipeline to just wait for zio to finish. */ pio->io_pipeline = ZIO_INTERLOCK_PIPELINE; zio_nowait(zio); return (ZIO_PIPELINE_CONTINUE); } /* - * The zio_nop_write stage in the pipeline determines if allocating - * a new bp is necessary. By leveraging a cryptographically secure checksum, - * such as SHA256, we can compare the checksums of the new data and the old - * to determine if allocating a new block is required. The nopwrite - * feature can handle writes in either syncing or open context (i.e. zil - * writes) and as a result is mutually exclusive with dedup. + * The zio_nop_write stage in the pipeline determines if allocating a + * new bp is necessary. The nopwrite feature can handle writes in + * either syncing or open context (i.e. zil writes) and as a result is + * mutually exclusive with dedup. + * + * By leveraging a cryptographically secure checksum, such as SHA256, we + * can compare the checksums of the new data and the old to determine if + * allocating a new block is required. Note that our requirements for + * cryptographic strength are fairly weak: there can't be any accidental + * hash collisions, but we don't need to be secure against intentional + * (malicious) collisions. To trigger a nopwrite, you have to be able + * to write the file to begin with, and triggering an incorrect (hash + * collision) nopwrite is no worse than simply writing to the file. + * That said, there are no known attacks against the checksum algorithms + * used for nopwrite, assuming that the salt and the checksums + * themselves remain secret. */ static int zio_nop_write(zio_t *zio) { blkptr_t *bp = zio->io_bp; blkptr_t *bp_orig = &zio->io_bp_orig; zio_prop_t *zp = &zio->io_prop; ASSERT(BP_GET_LEVEL(bp) == 0); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); ASSERT(zp->zp_nopwrite); ASSERT(!zp->zp_dedup); ASSERT(zio->io_bp_override == NULL); ASSERT(IO_IS_ALLOCATING(zio)); /* * Check to see if the original bp and the new bp have matching * characteristics (i.e. same checksum, compression algorithms, etc). * If they don't then just continue with the pipeline which will * allocate a new bp. */ if (BP_IS_HOLE(bp_orig) || - !zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_dedup || + !(zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_flags & + ZCHECKSUM_FLAG_NOPWRITE) || BP_GET_CHECKSUM(bp) != BP_GET_CHECKSUM(bp_orig) || BP_GET_COMPRESS(bp) != BP_GET_COMPRESS(bp_orig) || BP_GET_DEDUP(bp) != BP_GET_DEDUP(bp_orig) || zp->zp_copies != BP_GET_NDVAS(bp_orig)) return (ZIO_PIPELINE_CONTINUE); /* * If the checksums match then reset the pipeline so that we * avoid allocating a new bp and issuing any I/O. */ if (ZIO_CHECKSUM_EQUAL(bp->blk_cksum, bp_orig->blk_cksum)) { - ASSERT(zio_checksum_table[zp->zp_checksum].ci_dedup); + ASSERT(zio_checksum_table[zp->zp_checksum].ci_flags & + ZCHECKSUM_FLAG_NOPWRITE); ASSERT3U(BP_GET_PSIZE(bp), ==, BP_GET_PSIZE(bp_orig)); ASSERT3U(BP_GET_LSIZE(bp), ==, BP_GET_LSIZE(bp_orig)); ASSERT(zp->zp_compress != ZIO_COMPRESS_OFF); ASSERT(bcmp(&bp->blk_prop, &bp_orig->blk_prop, sizeof (uint64_t)) == 0); *bp = *bp_orig; zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; zio->io_flags |= ZIO_FLAG_NOPWRITE; } return (ZIO_PIPELINE_CONTINUE); } /* * ========================================================================== * Dedup * ========================================================================== */ static void zio_ddt_child_read_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp; zio_t *pio = zio_unique_parent(zio); mutex_enter(&pio->io_lock); ddp = ddt_phys_select(dde, bp); if (zio->io_error == 0) ddt_phys_clear(ddp); /* this ddp doesn't need repair */ if (zio->io_error == 0 && dde->dde_repair_data == NULL) dde->dde_repair_data = zio->io_data; else zio_buf_free(zio->io_data, zio->io_size); mutex_exit(&pio->io_lock); } static int zio_ddt_read_start(zio_t *zio) { blkptr_t *bp = zio->io_bp; ASSERT(BP_GET_DEDUP(bp)); ASSERT(BP_GET_PSIZE(bp) == zio->io_size); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); if (zio->io_child_error[ZIO_CHILD_DDT]) { ddt_t *ddt = ddt_select(zio->io_spa, bp); ddt_entry_t *dde = ddt_repair_start(ddt, bp); ddt_phys_t *ddp = dde->dde_phys; ddt_phys_t *ddp_self = ddt_phys_select(dde, bp); blkptr_t blk; ASSERT(zio->io_vsd == NULL); zio->io_vsd = dde; if (ddp_self == NULL) return (ZIO_PIPELINE_CONTINUE); for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) { if (ddp->ddp_phys_birth == 0 || ddp == ddp_self) continue; ddt_bp_create(ddt->ddt_checksum, &dde->dde_key, ddp, &blk); zio_nowait(zio_read(zio, zio->io_spa, &blk, zio_buf_alloc(zio->io_size), zio->io_size, zio_ddt_child_read_done, dde, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio) | ZIO_FLAG_DONT_PROPAGATE, &zio->io_bookmark)); } return (ZIO_PIPELINE_CONTINUE); } zio_nowait(zio_read(zio, zio->io_spa, bp, zio->io_data, zio->io_size, NULL, NULL, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark)); return (ZIO_PIPELINE_CONTINUE); } static int zio_ddt_read_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (zio_wait_for_children(zio, ZIO_CHILD_DDT, ZIO_WAIT_DONE)) return (ZIO_PIPELINE_STOP); ASSERT(BP_GET_DEDUP(bp)); ASSERT(BP_GET_PSIZE(bp) == zio->io_size); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); if (zio->io_child_error[ZIO_CHILD_DDT]) { ddt_t *ddt = ddt_select(zio->io_spa, bp); ddt_entry_t *dde = zio->io_vsd; if (ddt == NULL) { ASSERT(spa_load_state(zio->io_spa) != SPA_LOAD_NONE); return (ZIO_PIPELINE_CONTINUE); } if (dde == NULL) { zio->io_stage = ZIO_STAGE_DDT_READ_START >> 1; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE); return (ZIO_PIPELINE_STOP); } if (dde->dde_repair_data != NULL) { bcopy(dde->dde_repair_data, zio->io_data, zio->io_size); zio->io_child_error[ZIO_CHILD_DDT] = 0; } ddt_repair_done(ddt, dde); zio->io_vsd = NULL; } ASSERT(zio->io_vsd == NULL); return (ZIO_PIPELINE_CONTINUE); } static boolean_t zio_ddt_collision(zio_t *zio, ddt_t *ddt, ddt_entry_t *dde) { spa_t *spa = zio->io_spa; /* * Note: we compare the original data, not the transformed data, * because when zio->io_bp is an override bp, we will not have * pushed the I/O transforms. That's an important optimization * because otherwise we'd compress/encrypt all dmu_sync() data twice. */ for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) { zio_t *lio = dde->dde_lead_zio[p]; if (lio != NULL) { return (lio->io_orig_size != zio->io_orig_size || bcmp(zio->io_orig_data, lio->io_orig_data, zio->io_orig_size) != 0); } } for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) { ddt_phys_t *ddp = &dde->dde_phys[p]; if (ddp->ddp_phys_birth != 0) { arc_buf_t *abuf = NULL; arc_flags_t aflags = ARC_FLAG_WAIT; blkptr_t blk = *zio->io_bp; int error; ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth); ddt_exit(ddt); error = arc_read(NULL, spa, &blk, arc_getbuf_func, &abuf, ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &aflags, &zio->io_bookmark); if (error == 0) { if (arc_buf_size(abuf) != zio->io_orig_size || bcmp(abuf->b_data, zio->io_orig_data, zio->io_orig_size) != 0) error = SET_ERROR(EEXIST); VERIFY(arc_buf_remove_ref(abuf, &abuf)); } ddt_enter(ddt); return (error != 0); } } return (B_FALSE); } static void zio_ddt_child_write_ready(zio_t *zio) { int p = zio->io_prop.zp_copies; ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp); ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp = &dde->dde_phys[p]; zio_t *pio; if (zio->io_error) return; ddt_enter(ddt); ASSERT(dde->dde_lead_zio[p] == zio); ddt_phys_fill(ddp, zio->io_bp); while ((pio = zio_walk_parents(zio)) != NULL) ddt_bp_fill(ddp, pio->io_bp, zio->io_txg); ddt_exit(ddt); } static void zio_ddt_child_write_done(zio_t *zio) { int p = zio->io_prop.zp_copies; ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp); ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp = &dde->dde_phys[p]; ddt_enter(ddt); ASSERT(ddp->ddp_refcnt == 0); ASSERT(dde->dde_lead_zio[p] == zio); dde->dde_lead_zio[p] = NULL; if (zio->io_error == 0) { while (zio_walk_parents(zio) != NULL) ddt_phys_addref(ddp); } else { ddt_phys_clear(ddp); } ddt_exit(ddt); } static void zio_ddt_ditto_write_done(zio_t *zio) { int p = DDT_PHYS_DITTO; zio_prop_t *zp = &zio->io_prop; blkptr_t *bp = zio->io_bp; ddt_t *ddt = ddt_select(zio->io_spa, bp); ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp = &dde->dde_phys[p]; ddt_key_t *ddk = &dde->dde_key; ddt_enter(ddt); ASSERT(ddp->ddp_refcnt == 0); ASSERT(dde->dde_lead_zio[p] == zio); dde->dde_lead_zio[p] = NULL; if (zio->io_error == 0) { ASSERT(ZIO_CHECKSUM_EQUAL(bp->blk_cksum, ddk->ddk_cksum)); ASSERT(zp->zp_copies < SPA_DVAS_PER_BP); ASSERT(zp->zp_copies == BP_GET_NDVAS(bp) - BP_IS_GANG(bp)); if (ddp->ddp_phys_birth != 0) ddt_phys_free(ddt, ddk, ddp, zio->io_txg); ddt_phys_fill(ddp, bp); } ddt_exit(ddt); } static int zio_ddt_write(zio_t *zio) { spa_t *spa = zio->io_spa; blkptr_t *bp = zio->io_bp; uint64_t txg = zio->io_txg; zio_prop_t *zp = &zio->io_prop; int p = zp->zp_copies; int ditto_copies; zio_t *cio = NULL; zio_t *dio = NULL; ddt_t *ddt = ddt_select(spa, bp); ddt_entry_t *dde; ddt_phys_t *ddp; ASSERT(BP_GET_DEDUP(bp)); ASSERT(BP_GET_CHECKSUM(bp) == zp->zp_checksum); ASSERT(BP_IS_HOLE(bp) || zio->io_bp_override); ddt_enter(ddt); dde = ddt_lookup(ddt, bp, B_TRUE); ddp = &dde->dde_phys[p]; if (zp->zp_dedup_verify && zio_ddt_collision(zio, ddt, dde)) { /* * If we're using a weak checksum, upgrade to a strong checksum * and try again. If we're already using a strong checksum, * we can't resolve it, so just convert to an ordinary write. * (And automatically e-mail a paper to Nature?) */ - if (!zio_checksum_table[zp->zp_checksum].ci_dedup) { + if (!(zio_checksum_table[zp->zp_checksum].ci_flags & + ZCHECKSUM_FLAG_DEDUP)) { zp->zp_checksum = spa_dedup_checksum(spa); zio_pop_transforms(zio); zio->io_stage = ZIO_STAGE_OPEN; BP_ZERO(bp); } else { zp->zp_dedup = B_FALSE; } zio->io_pipeline = ZIO_WRITE_PIPELINE; ddt_exit(ddt); return (ZIO_PIPELINE_CONTINUE); } ditto_copies = ddt_ditto_copies_needed(ddt, dde, ddp); ASSERT(ditto_copies < SPA_DVAS_PER_BP); if (ditto_copies > ddt_ditto_copies_present(dde) && dde->dde_lead_zio[DDT_PHYS_DITTO] == NULL) { zio_prop_t czp = *zp; czp.zp_copies = ditto_copies; /* * If we arrived here with an override bp, we won't have run * the transform stack, so we won't have the data we need to * generate a child i/o. So, toss the override bp and restart. * This is safe, because using the override bp is just an * optimization; and it's rare, so the cost doesn't matter. */ if (zio->io_bp_override) { zio_pop_transforms(zio); zio->io_stage = ZIO_STAGE_OPEN; zio->io_pipeline = ZIO_WRITE_PIPELINE; zio->io_bp_override = NULL; BP_ZERO(bp); ddt_exit(ddt); return (ZIO_PIPELINE_CONTINUE); } dio = zio_write(zio, spa, txg, bp, zio->io_orig_data, zio->io_orig_size, &czp, NULL, NULL, zio_ddt_ditto_write_done, dde, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark); zio_push_transform(dio, zio->io_data, zio->io_size, 0, NULL); dde->dde_lead_zio[DDT_PHYS_DITTO] = dio; } if (ddp->ddp_phys_birth != 0 || dde->dde_lead_zio[p] != NULL) { if (ddp->ddp_phys_birth != 0) ddt_bp_fill(ddp, bp, txg); if (dde->dde_lead_zio[p] != NULL) zio_add_child(zio, dde->dde_lead_zio[p]); else ddt_phys_addref(ddp); } else if (zio->io_bp_override) { ASSERT(bp->blk_birth == txg); ASSERT(BP_EQUAL(bp, zio->io_bp_override)); ddt_phys_fill(ddp, bp); ddt_phys_addref(ddp); } else { cio = zio_write(zio, spa, txg, bp, zio->io_orig_data, zio->io_orig_size, zp, zio_ddt_child_write_ready, NULL, zio_ddt_child_write_done, dde, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark); zio_push_transform(cio, zio->io_data, zio->io_size, 0, NULL); dde->dde_lead_zio[p] = cio; } ddt_exit(ddt); if (cio) zio_nowait(cio); if (dio) zio_nowait(dio); return (ZIO_PIPELINE_CONTINUE); } ddt_entry_t *freedde; /* for debugging */ static int zio_ddt_free(zio_t *zio) { spa_t *spa = zio->io_spa; blkptr_t *bp = zio->io_bp; ddt_t *ddt = ddt_select(spa, bp); ddt_entry_t *dde; ddt_phys_t *ddp; ASSERT(BP_GET_DEDUP(bp)); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ddt_enter(ddt); freedde = dde = ddt_lookup(ddt, bp, B_TRUE); ddp = ddt_phys_select(dde, bp); ddt_phys_decref(ddp); ddt_exit(ddt); return (ZIO_PIPELINE_CONTINUE); } /* * ========================================================================== * Allocate and free blocks * ========================================================================== */ static int zio_dva_allocate(zio_t *zio) { spa_t *spa = zio->io_spa; metaslab_class_t *mc = spa_normal_class(spa); blkptr_t *bp = zio->io_bp; int error; int flags = 0; if (zio->io_gang_leader == NULL) { ASSERT(zio->io_child_type > ZIO_CHILD_GANG); zio->io_gang_leader = zio; } ASSERT(BP_IS_HOLE(bp)); ASSERT0(BP_GET_NDVAS(bp)); ASSERT3U(zio->io_prop.zp_copies, >, 0); ASSERT3U(zio->io_prop.zp_copies, <=, spa_max_replication(spa)); ASSERT3U(zio->io_size, ==, BP_GET_PSIZE(bp)); /* * The dump device does not support gang blocks so allocation on * behalf of the dump device (i.e. ZIO_FLAG_NODATA) must avoid * the "fast" gang feature. */ flags |= (zio->io_flags & ZIO_FLAG_NODATA) ? METASLAB_GANG_AVOID : 0; flags |= (zio->io_flags & ZIO_FLAG_GANG_CHILD) ? METASLAB_GANG_CHILD : 0; error = metaslab_alloc(spa, mc, zio->io_size, bp, zio->io_prop.zp_copies, zio->io_txg, NULL, flags); if (error) { spa_dbgmsg(spa, "%s: metaslab allocation failure: zio %p, " "size %llu, error %d", spa_name(spa), zio, zio->io_size, error); if (error == ENOSPC && zio->io_size > SPA_MINBLOCKSIZE) return (zio_write_gang_block(zio)); zio->io_error = error; } return (ZIO_PIPELINE_CONTINUE); } static int zio_dva_free(zio_t *zio) { metaslab_free(zio->io_spa, zio->io_bp, zio->io_txg, B_FALSE); return (ZIO_PIPELINE_CONTINUE); } static int zio_dva_claim(zio_t *zio) { int error; error = metaslab_claim(zio->io_spa, zio->io_bp, zio->io_txg); if (error) zio->io_error = error; return (ZIO_PIPELINE_CONTINUE); } /* * Undo an allocation. This is used by zio_done() when an I/O fails * and we want to give back the block we just allocated. * This handles both normal blocks and gang blocks. */ static void zio_dva_unallocate(zio_t *zio, zio_gang_node_t *gn, blkptr_t *bp) { ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp)); ASSERT(zio->io_bp_override == NULL); if (!BP_IS_HOLE(bp)) metaslab_free(zio->io_spa, bp, bp->blk_birth, B_TRUE); if (gn != NULL) { for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { zio_dva_unallocate(zio, gn->gn_child[g], &gn->gn_gbh->zg_blkptr[g]); } } } /* * Try to allocate an intent log block. Return 0 on success, errno on failure. */ int zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp, blkptr_t *old_bp, uint64_t size, boolean_t use_slog) { int error = 1; ASSERT(txg > spa_syncing_txg(spa)); /* * ZIL blocks are always contiguous (i.e. not gang blocks) so we * set the METASLAB_GANG_AVOID flag so that they don't "fast gang" * when allocating them. */ if (use_slog) { error = metaslab_alloc(spa, spa_log_class(spa), size, new_bp, 1, txg, old_bp, METASLAB_HINTBP_AVOID | METASLAB_GANG_AVOID); } if (error) { error = metaslab_alloc(spa, spa_normal_class(spa), size, new_bp, 1, txg, old_bp, METASLAB_HINTBP_AVOID); } if (error == 0) { BP_SET_LSIZE(new_bp, size); BP_SET_PSIZE(new_bp, size); BP_SET_COMPRESS(new_bp, ZIO_COMPRESS_OFF); BP_SET_CHECKSUM(new_bp, spa_version(spa) >= SPA_VERSION_SLIM_ZIL ? ZIO_CHECKSUM_ZILOG2 : ZIO_CHECKSUM_ZILOG); BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG); BP_SET_LEVEL(new_bp, 0); BP_SET_DEDUP(new_bp, 0); BP_SET_BYTEORDER(new_bp, ZFS_HOST_BYTEORDER); } return (error); } /* * Free an intent log block. */ void zio_free_zil(spa_t *spa, uint64_t txg, blkptr_t *bp) { ASSERT(BP_GET_TYPE(bp) == DMU_OT_INTENT_LOG); ASSERT(!BP_IS_GANG(bp)); zio_free(spa, txg, bp); } /* * ========================================================================== * Read, write and delete to physical devices * ========================================================================== */ static int zio_vdev_io_start(zio_t *zio) { vdev_t *vd = zio->io_vd; uint64_t align; spa_t *spa = zio->io_spa; int ret; ASSERT(zio->io_error == 0); ASSERT(zio->io_child_error[ZIO_CHILD_VDEV] == 0); if (vd == NULL) { if (!(zio->io_flags & ZIO_FLAG_CONFIG_WRITER)) spa_config_enter(spa, SCL_ZIO, zio, RW_READER); /* * The mirror_ops handle multiple DVAs in a single BP. */ return (vdev_mirror_ops.vdev_op_io_start(zio)); } if (vd->vdev_ops->vdev_op_leaf && zio->io_type == ZIO_TYPE_FREE && zio->io_priority == ZIO_PRIORITY_NOW) { trim_map_free(vd, zio->io_offset, zio->io_size, zio->io_txg); return (ZIO_PIPELINE_CONTINUE); } /* * We keep track of time-sensitive I/Os so that the scan thread * can quickly react to certain workloads. In particular, we care * about non-scrubbing, top-level reads and writes with the following * characteristics: * - synchronous writes of user data to non-slog devices * - any reads of user data * When these conditions are met, adjust the timestamp of spa_last_io * which allows the scan thread to adjust its workload accordingly. */ if (!(zio->io_flags & ZIO_FLAG_SCAN_THREAD) && zio->io_bp != NULL && vd == vd->vdev_top && !vd->vdev_islog && zio->io_bookmark.zb_objset != DMU_META_OBJSET && zio->io_txg != spa_syncing_txg(spa)) { uint64_t old = spa->spa_last_io; uint64_t new = ddi_get_lbolt64(); if (old != new) (void) atomic_cas_64(&spa->spa_last_io, old, new); } align = 1ULL << vd->vdev_top->vdev_ashift; if ((!(zio->io_flags & ZIO_FLAG_PHYSICAL) || (vd->vdev_top->vdev_physical_ashift > SPA_MINBLOCKSHIFT)) && P2PHASE(zio->io_size, align) != 0) { /* Transform logical writes to be a full physical block size. */ uint64_t asize = P2ROUNDUP(zio->io_size, align); char *abuf = NULL; if (zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE) abuf = zio_buf_alloc(asize); ASSERT(vd == vd->vdev_top); if (zio->io_type == ZIO_TYPE_WRITE) { bcopy(zio->io_data, abuf, zio->io_size); bzero(abuf + zio->io_size, asize - zio->io_size); } zio_push_transform(zio, abuf, asize, abuf ? asize : 0, zio_subblock); } /* * If this is not a physical io, make sure that it is properly aligned * before proceeding. */ if (!(zio->io_flags & ZIO_FLAG_PHYSICAL)) { ASSERT0(P2PHASE(zio->io_offset, align)); ASSERT0(P2PHASE(zio->io_size, align)); } else { /* * For physical writes, we allow 512b aligned writes and assume * the device will perform a read-modify-write as necessary. */ ASSERT0(P2PHASE(zio->io_offset, SPA_MINBLOCKSIZE)); ASSERT0(P2PHASE(zio->io_size, SPA_MINBLOCKSIZE)); } VERIFY(zio->io_type == ZIO_TYPE_READ || spa_writeable(spa)); /* * If this is a repair I/O, and there's no self-healing involved -- * that is, we're just resilvering what we expect to resilver -- * then don't do the I/O unless zio's txg is actually in vd's DTL. * This prevents spurious resilvering with nested replication. * For example, given a mirror of mirrors, (A+B)+(C+D), if only * A is out of date, we'll read from C+D, then use the data to * resilver A+B -- but we don't actually want to resilver B, just A. * The top-level mirror has no way to know this, so instead we just * discard unnecessary repairs as we work our way down the vdev tree. * The same logic applies to any form of nested replication: * ditto + mirror, RAID-Z + replacing, etc. This covers them all. */ if ((zio->io_flags & ZIO_FLAG_IO_REPAIR) && !(zio->io_flags & ZIO_FLAG_SELF_HEAL) && zio->io_txg != 0 && /* not a delegated i/o */ !vdev_dtl_contains(vd, DTL_PARTIAL, zio->io_txg, 1)) { ASSERT(zio->io_type == ZIO_TYPE_WRITE); zio_vdev_io_bypass(zio); return (ZIO_PIPELINE_CONTINUE); } if (vd->vdev_ops->vdev_op_leaf) { switch (zio->io_type) { case ZIO_TYPE_READ: if (vdev_cache_read(zio)) return (ZIO_PIPELINE_CONTINUE); /* FALLTHROUGH */ case ZIO_TYPE_WRITE: case ZIO_TYPE_FREE: if ((zio = vdev_queue_io(zio)) == NULL) return (ZIO_PIPELINE_STOP); if (!vdev_accessible(vd, zio)) { zio->io_error = SET_ERROR(ENXIO); zio_interrupt(zio); return (ZIO_PIPELINE_STOP); } break; } /* * Note that we ignore repair writes for TRIM because they can * conflict with normal writes. This isn't an issue because, by * definition, we only repair blocks that aren't freed. */ if (zio->io_type == ZIO_TYPE_WRITE && !(zio->io_flags & ZIO_FLAG_IO_REPAIR) && !trim_map_write_start(zio)) return (ZIO_PIPELINE_STOP); } ret = vd->vdev_ops->vdev_op_io_start(zio); ASSERT(ret == ZIO_PIPELINE_STOP); return (ret); } static int zio_vdev_io_done(zio_t *zio) { vdev_t *vd = zio->io_vd; vdev_ops_t *ops = vd ? vd->vdev_ops : &vdev_mirror_ops; boolean_t unexpected_error = B_FALSE; if (zio_wait_for_children(zio, ZIO_CHILD_VDEV, ZIO_WAIT_DONE)) return (ZIO_PIPELINE_STOP); ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE || zio->io_type == ZIO_TYPE_FREE); if (vd != NULL && vd->vdev_ops->vdev_op_leaf && (zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE || zio->io_type == ZIO_TYPE_FREE)) { if (zio->io_type == ZIO_TYPE_WRITE && !(zio->io_flags & ZIO_FLAG_IO_REPAIR)) trim_map_write_done(zio); vdev_queue_io_done(zio); if (zio->io_type == ZIO_TYPE_WRITE) vdev_cache_write(zio); if (zio_injection_enabled && zio->io_error == 0) zio->io_error = zio_handle_device_injection(vd, zio, EIO); if (zio_injection_enabled && zio->io_error == 0) zio->io_error = zio_handle_label_injection(zio, EIO); if (zio->io_error) { if (zio->io_error == ENOTSUP && zio->io_type == ZIO_TYPE_FREE) { /* Not all devices support TRIM. */ } else if (!vdev_accessible(vd, zio)) { zio->io_error = SET_ERROR(ENXIO); } else { unexpected_error = B_TRUE; } } } ops->vdev_op_io_done(zio); if (unexpected_error) VERIFY(vdev_probe(vd, zio) == NULL); return (ZIO_PIPELINE_CONTINUE); } /* * For non-raidz ZIOs, we can just copy aside the bad data read from the * disk, and use that to finish the checksum ereport later. */ static void zio_vsd_default_cksum_finish(zio_cksum_report_t *zcr, const void *good_buf) { /* no processing needed */ zfs_ereport_finish_checksum(zcr, good_buf, zcr->zcr_cbdata, B_FALSE); } /*ARGSUSED*/ void zio_vsd_default_cksum_report(zio_t *zio, zio_cksum_report_t *zcr, void *ignored) { void *buf = zio_buf_alloc(zio->io_size); bcopy(zio->io_data, buf, zio->io_size); zcr->zcr_cbinfo = zio->io_size; zcr->zcr_cbdata = buf; zcr->zcr_finish = zio_vsd_default_cksum_finish; zcr->zcr_free = zio_buf_free; } static int zio_vdev_io_assess(zio_t *zio) { vdev_t *vd = zio->io_vd; if (zio_wait_for_children(zio, ZIO_CHILD_VDEV, ZIO_WAIT_DONE)) return (ZIO_PIPELINE_STOP); if (vd == NULL && !(zio->io_flags & ZIO_FLAG_CONFIG_WRITER)) spa_config_exit(zio->io_spa, SCL_ZIO, zio); if (zio->io_vsd != NULL) { zio->io_vsd_ops->vsd_free(zio); zio->io_vsd = NULL; } if (zio_injection_enabled && zio->io_error == 0) zio->io_error = zio_handle_fault_injection(zio, EIO); if (zio->io_type == ZIO_TYPE_FREE && zio->io_priority != ZIO_PRIORITY_NOW) { switch (zio->io_error) { case 0: ZIO_TRIM_STAT_INCR(bytes, zio->io_size); ZIO_TRIM_STAT_BUMP(success); break; case EOPNOTSUPP: ZIO_TRIM_STAT_BUMP(unsupported); break; default: ZIO_TRIM_STAT_BUMP(failed); break; } } /* * If the I/O failed, determine whether we should attempt to retry it. * * On retry, we cut in line in the issue queue, since we don't want * compression/checksumming/etc. work to prevent our (cheap) IO reissue. */ if (zio->io_error && vd == NULL && !(zio->io_flags & (ZIO_FLAG_DONT_RETRY | ZIO_FLAG_IO_RETRY))) { ASSERT(!(zio->io_flags & ZIO_FLAG_DONT_QUEUE)); /* not a leaf */ ASSERT(!(zio->io_flags & ZIO_FLAG_IO_BYPASS)); /* not a leaf */ zio->io_error = 0; zio->io_flags |= ZIO_FLAG_IO_RETRY | ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_AGGREGATE; zio->io_stage = ZIO_STAGE_VDEV_IO_START >> 1; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, zio_requeue_io_start_cut_in_line); return (ZIO_PIPELINE_STOP); } /* * If we got an error on a leaf device, convert it to ENXIO * if the device is not accessible at all. */ if (zio->io_error && vd != NULL && vd->vdev_ops->vdev_op_leaf && !vdev_accessible(vd, zio)) zio->io_error = SET_ERROR(ENXIO); /* * If we can't write to an interior vdev (mirror or RAID-Z), * set vdev_cant_write so that we stop trying to allocate from it. */ if (zio->io_error == ENXIO && zio->io_type == ZIO_TYPE_WRITE && vd != NULL && !vd->vdev_ops->vdev_op_leaf) { vd->vdev_cant_write = B_TRUE; } if (zio->io_error) zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; if (vd != NULL && vd->vdev_ops->vdev_op_leaf && zio->io_physdone != NULL) { ASSERT(!(zio->io_flags & ZIO_FLAG_DELEGATED)); ASSERT(zio->io_child_type == ZIO_CHILD_VDEV); zio->io_physdone(zio->io_logical); } return (ZIO_PIPELINE_CONTINUE); } void zio_vdev_io_reissue(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START); ASSERT(zio->io_error == 0); zio->io_stage >>= 1; } void zio_vdev_io_redone(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_DONE); zio->io_stage >>= 1; } void zio_vdev_io_bypass(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START); ASSERT(zio->io_error == 0); zio->io_flags |= ZIO_FLAG_IO_BYPASS; zio->io_stage = ZIO_STAGE_VDEV_IO_ASSESS >> 1; } /* * ========================================================================== * Generate and verify checksums * ========================================================================== */ static int zio_checksum_generate(zio_t *zio) { blkptr_t *bp = zio->io_bp; enum zio_checksum checksum; if (bp == NULL) { /* * This is zio_write_phys(). * We're either generating a label checksum, or none at all. */ checksum = zio->io_prop.zp_checksum; if (checksum == ZIO_CHECKSUM_OFF) return (ZIO_PIPELINE_CONTINUE); ASSERT(checksum == ZIO_CHECKSUM_LABEL); } else { if (BP_IS_GANG(bp) && zio->io_child_type == ZIO_CHILD_GANG) { ASSERT(!IO_IS_ALLOCATING(zio)); checksum = ZIO_CHECKSUM_GANG_HEADER; } else { checksum = BP_GET_CHECKSUM(bp); } } zio_checksum_compute(zio, checksum, zio->io_data, zio->io_size); return (ZIO_PIPELINE_CONTINUE); } static int zio_checksum_verify(zio_t *zio) { zio_bad_cksum_t info; blkptr_t *bp = zio->io_bp; int error; ASSERT(zio->io_vd != NULL); if (bp == NULL) { /* * This is zio_read_phys(). * We're either verifying a label checksum, or nothing at all. */ if (zio->io_prop.zp_checksum == ZIO_CHECKSUM_OFF) return (ZIO_PIPELINE_CONTINUE); ASSERT(zio->io_prop.zp_checksum == ZIO_CHECKSUM_LABEL); } if ((error = zio_checksum_error(zio, &info)) != 0) { zio->io_error = error; if (error == ECKSUM && !(zio->io_flags & ZIO_FLAG_SPECULATIVE)) { zfs_ereport_start_checksum(zio->io_spa, zio->io_vd, zio, zio->io_offset, zio->io_size, NULL, &info); } } return (ZIO_PIPELINE_CONTINUE); } /* * Called by RAID-Z to ensure we don't compute the checksum twice. */ void zio_checksum_verified(zio_t *zio) { zio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY; } /* * ========================================================================== * Error rank. Error are ranked in the order 0, ENXIO, ECKSUM, EIO, other. * An error of 0 indicates success. ENXIO indicates whole-device failure, * which may be transient (e.g. unplugged) or permament. ECKSUM and EIO * indicate errors that are specific to one I/O, and most likely permanent. * Any other error is presumed to be worse because we weren't expecting it. * ========================================================================== */ int zio_worst_error(int e1, int e2) { static int zio_error_rank[] = { 0, ENXIO, ECKSUM, EIO }; int r1, r2; for (r1 = 0; r1 < sizeof (zio_error_rank) / sizeof (int); r1++) if (e1 == zio_error_rank[r1]) break; for (r2 = 0; r2 < sizeof (zio_error_rank) / sizeof (int); r2++) if (e2 == zio_error_rank[r2]) break; return (r1 > r2 ? e1 : e2); } /* * ========================================================================== * I/O completion * ========================================================================== */ static int zio_ready(zio_t *zio) { blkptr_t *bp = zio->io_bp; zio_t *pio, *pio_next; if (zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_READY) || zio_wait_for_children(zio, ZIO_CHILD_DDT, ZIO_WAIT_READY)) return (ZIO_PIPELINE_STOP); if (zio->io_ready) { ASSERT(IO_IS_ALLOCATING(zio)); ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp) || (zio->io_flags & ZIO_FLAG_NOPWRITE)); ASSERT(zio->io_children[ZIO_CHILD_GANG][ZIO_WAIT_READY] == 0); zio->io_ready(zio); } if (bp != NULL && bp != &zio->io_bp_copy) zio->io_bp_copy = *bp; if (zio->io_error) zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; mutex_enter(&zio->io_lock); zio->io_state[ZIO_WAIT_READY] = 1; pio = zio_walk_parents(zio); mutex_exit(&zio->io_lock); /* * As we notify zio's parents, new parents could be added. * New parents go to the head of zio's io_parent_list, however, * so we will (correctly) not notify them. The remainder of zio's * io_parent_list, from 'pio_next' onward, cannot change because * all parents must wait for us to be done before they can be done. */ for (; pio != NULL; pio = pio_next) { pio_next = zio_walk_parents(zio); zio_notify_parent(pio, zio, ZIO_WAIT_READY); } if (zio->io_flags & ZIO_FLAG_NODATA) { if (BP_IS_GANG(bp)) { zio->io_flags &= ~ZIO_FLAG_NODATA; } else { ASSERT((uintptr_t)zio->io_data < SPA_MAXBLOCKSIZE); zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES; } } if (zio_injection_enabled && zio->io_spa->spa_syncing_txg == zio->io_txg) zio_handle_ignored_writes(zio); return (ZIO_PIPELINE_CONTINUE); } static int zio_done(zio_t *zio) { spa_t *spa = zio->io_spa; zio_t *lio = zio->io_logical; blkptr_t *bp = zio->io_bp; vdev_t *vd = zio->io_vd; uint64_t psize = zio->io_size; zio_t *pio, *pio_next; /* * If our children haven't all completed, * wait for them and then repeat this pipeline stage. */ if (zio_wait_for_children(zio, ZIO_CHILD_VDEV, ZIO_WAIT_DONE) || zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_DONE) || zio_wait_for_children(zio, ZIO_CHILD_DDT, ZIO_WAIT_DONE) || zio_wait_for_children(zio, ZIO_CHILD_LOGICAL, ZIO_WAIT_DONE)) return (ZIO_PIPELINE_STOP); for (int c = 0; c < ZIO_CHILD_TYPES; c++) for (int w = 0; w < ZIO_WAIT_TYPES; w++) ASSERT(zio->io_children[c][w] == 0); if (bp != NULL && !BP_IS_EMBEDDED(bp)) { ASSERT(bp->blk_pad[0] == 0); ASSERT(bp->blk_pad[1] == 0); ASSERT(bcmp(bp, &zio->io_bp_copy, sizeof (blkptr_t)) == 0 || (bp == zio_unique_parent(zio)->io_bp)); if (zio->io_type == ZIO_TYPE_WRITE && !BP_IS_HOLE(bp) && zio->io_bp_override == NULL && !(zio->io_flags & ZIO_FLAG_IO_REPAIR)) { ASSERT(!BP_SHOULD_BYTESWAP(bp)); ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(bp)); ASSERT(BP_COUNT_GANG(bp) == 0 || (BP_COUNT_GANG(bp) == BP_GET_NDVAS(bp))); } if (zio->io_flags & ZIO_FLAG_NOPWRITE) VERIFY(BP_EQUAL(bp, &zio->io_bp_orig)); } /* * If there were child vdev/gang/ddt errors, they apply to us now. */ zio_inherit_child_errors(zio, ZIO_CHILD_VDEV); zio_inherit_child_errors(zio, ZIO_CHILD_GANG); zio_inherit_child_errors(zio, ZIO_CHILD_DDT); /* * If the I/O on the transformed data was successful, generate any * checksum reports now while we still have the transformed data. */ if (zio->io_error == 0) { while (zio->io_cksum_report != NULL) { zio_cksum_report_t *zcr = zio->io_cksum_report; uint64_t align = zcr->zcr_align; uint64_t asize = P2ROUNDUP(psize, align); char *abuf = zio->io_data; if (asize != psize) { abuf = zio_buf_alloc(asize); bcopy(zio->io_data, abuf, psize); bzero(abuf + psize, asize - psize); } zio->io_cksum_report = zcr->zcr_next; zcr->zcr_next = NULL; zcr->zcr_finish(zcr, abuf); zfs_ereport_free_checksum(zcr); if (asize != psize) zio_buf_free(abuf, asize); } } zio_pop_transforms(zio); /* note: may set zio->io_error */ vdev_stat_update(zio, psize); if (zio->io_error) { /* * If this I/O is attached to a particular vdev, * generate an error message describing the I/O failure * at the block level. We ignore these errors if the * device is currently unavailable. */ if (zio->io_error != ECKSUM && vd != NULL && !vdev_is_dead(vd)) zfs_ereport_post(FM_EREPORT_ZFS_IO, spa, vd, zio, 0, 0); if ((zio->io_error == EIO || !(zio->io_flags & (ZIO_FLAG_SPECULATIVE | ZIO_FLAG_DONT_PROPAGATE))) && zio == lio) { /* * For logical I/O requests, tell the SPA to log the * error and generate a logical data ereport. */ spa_log_error(spa, zio); zfs_ereport_post(FM_EREPORT_ZFS_DATA, spa, NULL, zio, 0, 0); } } if (zio->io_error && zio == lio) { /* * Determine whether zio should be reexecuted. This will * propagate all the way to the root via zio_notify_parent(). */ ASSERT(vd == NULL && bp != NULL); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); if (IO_IS_ALLOCATING(zio) && !(zio->io_flags & ZIO_FLAG_CANFAIL)) { if (zio->io_error != ENOSPC) zio->io_reexecute |= ZIO_REEXECUTE_NOW; else zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; } if ((zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_FREE) && !(zio->io_flags & ZIO_FLAG_SCAN_THREAD) && zio->io_error == ENXIO && spa_load_state(spa) == SPA_LOAD_NONE && spa_get_failmode(spa) != ZIO_FAILURE_MODE_CONTINUE) zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; if (!(zio->io_flags & ZIO_FLAG_CANFAIL) && !zio->io_reexecute) zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; /* * Here is a possibly good place to attempt to do * either combinatorial reconstruction or error correction * based on checksums. It also might be a good place * to send out preliminary ereports before we suspend * processing. */ } /* * If there were logical child errors, they apply to us now. * We defer this until now to avoid conflating logical child * errors with errors that happened to the zio itself when * updating vdev stats and reporting FMA events above. */ zio_inherit_child_errors(zio, ZIO_CHILD_LOGICAL); if ((zio->io_error || zio->io_reexecute) && IO_IS_ALLOCATING(zio) && zio->io_gang_leader == zio && !(zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE))) zio_dva_unallocate(zio, zio->io_gang_tree, bp); zio_gang_tree_free(&zio->io_gang_tree); /* * Godfather I/Os should never suspend. */ if ((zio->io_flags & ZIO_FLAG_GODFATHER) && (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) zio->io_reexecute = 0; if (zio->io_reexecute) { /* * This is a logical I/O that wants to reexecute. * * Reexecute is top-down. When an i/o fails, if it's not * the root, it simply notifies its parent and sticks around. * The parent, seeing that it still has children in zio_done(), * does the same. This percolates all the way up to the root. * The root i/o will reexecute or suspend the entire tree. * * This approach ensures that zio_reexecute() honors * all the original i/o dependency relationships, e.g. * parents not executing until children are ready. */ ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); zio->io_gang_leader = NULL; mutex_enter(&zio->io_lock); zio->io_state[ZIO_WAIT_DONE] = 1; mutex_exit(&zio->io_lock); /* * "The Godfather" I/O monitors its children but is * not a true parent to them. It will track them through * the pipeline but severs its ties whenever they get into * trouble (e.g. suspended). This allows "The Godfather" * I/O to return status without blocking. */ for (pio = zio_walk_parents(zio); pio != NULL; pio = pio_next) { zio_link_t *zl = zio->io_walk_link; pio_next = zio_walk_parents(zio); if ((pio->io_flags & ZIO_FLAG_GODFATHER) && (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) { zio_remove_child(pio, zio, zl); zio_notify_parent(pio, zio, ZIO_WAIT_DONE); } } if ((pio = zio_unique_parent(zio)) != NULL) { /* * We're not a root i/o, so there's nothing to do * but notify our parent. Don't propagate errors * upward since we haven't permanently failed yet. */ ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER)); zio->io_flags |= ZIO_FLAG_DONT_PROPAGATE; zio_notify_parent(pio, zio, ZIO_WAIT_DONE); } else if (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND) { /* * We'd fail again if we reexecuted now, so suspend * until conditions improve (e.g. device comes online). */ zio_suspend(spa, zio); } else { /* * Reexecution is potentially a huge amount of work. * Hand it off to the otherwise-unused claim taskq. */ #if defined(illumos) || !defined(_KERNEL) ASSERT(zio->io_tqent.tqent_next == NULL); #else ASSERT(zio->io_tqent.tqent_task.ta_pending == 0); #endif spa_taskq_dispatch_ent(spa, ZIO_TYPE_CLAIM, ZIO_TASKQ_ISSUE, (task_func_t *)zio_reexecute, zio, 0, &zio->io_tqent); } return (ZIO_PIPELINE_STOP); } ASSERT(zio->io_child_count == 0); ASSERT(zio->io_reexecute == 0); ASSERT(zio->io_error == 0 || (zio->io_flags & ZIO_FLAG_CANFAIL)); /* * Report any checksum errors, since the I/O is complete. */ while (zio->io_cksum_report != NULL) { zio_cksum_report_t *zcr = zio->io_cksum_report; zio->io_cksum_report = zcr->zcr_next; zcr->zcr_next = NULL; zcr->zcr_finish(zcr, NULL); zfs_ereport_free_checksum(zcr); } /* * It is the responsibility of the done callback to ensure that this * particular zio is no longer discoverable for adoption, and as * such, cannot acquire any new parents. */ if (zio->io_done) zio->io_done(zio); mutex_enter(&zio->io_lock); zio->io_state[ZIO_WAIT_DONE] = 1; mutex_exit(&zio->io_lock); for (pio = zio_walk_parents(zio); pio != NULL; pio = pio_next) { zio_link_t *zl = zio->io_walk_link; pio_next = zio_walk_parents(zio); zio_remove_child(pio, zio, zl); zio_notify_parent(pio, zio, ZIO_WAIT_DONE); } if (zio->io_waiter != NULL) { mutex_enter(&zio->io_lock); zio->io_executor = NULL; cv_broadcast(&zio->io_cv); mutex_exit(&zio->io_lock); } else { zio_destroy(zio); } return (ZIO_PIPELINE_STOP); } /* * ========================================================================== * I/O pipeline definition * ========================================================================== */ static zio_pipe_stage_t *zio_pipeline[] = { NULL, zio_read_bp_init, zio_free_bp_init, zio_issue_async, zio_write_bp_init, zio_checksum_generate, zio_nop_write, zio_ddt_read_start, zio_ddt_read_done, zio_ddt_write, zio_ddt_free, zio_gang_assemble, zio_gang_issue, zio_dva_allocate, zio_dva_free, zio_dva_claim, zio_ready, zio_vdev_io_start, zio_vdev_io_done, zio_vdev_io_assess, zio_checksum_verify, zio_done }; /* * Compare two zbookmark_phys_t's to see which we would reach first in a * pre-order traversal of the object tree. * * This is simple in every case aside from the meta-dnode object. For all other * objects, we traverse them in order (object 1 before object 2, and so on). * However, all of these objects are traversed while traversing object 0, since * the data it points to is the list of objects. Thus, we need to convert to a * canonical representation so we can compare meta-dnode bookmarks to * non-meta-dnode bookmarks. * * We do this by calculating "equivalents" for each field of the zbookmark. * zbookmarks outside of the meta-dnode use their own object and level, and * calculate the level 0 equivalent (the first L0 blkid that is contained in the * blocks this bookmark refers to) by multiplying their blkid by their span * (the number of L0 blocks contained within one block at their level). * zbookmarks inside the meta-dnode calculate their object equivalent * (which is L0equiv * dnodes per data block), use 0 for their L0equiv, and use * level + 1<<31 (any value larger than a level could ever be) for their level. * This causes them to always compare before a bookmark in their object * equivalent, compare appropriately to bookmarks in other objects, and to * compare appropriately to other bookmarks in the meta-dnode. */ int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2, uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2) { /* * These variables represent the "equivalent" values for the zbookmark, * after converting zbookmarks inside the meta dnode to their * normal-object equivalents. */ uint64_t zb1obj, zb2obj; uint64_t zb1L0, zb2L0; uint64_t zb1level, zb2level; if (zb1->zb_object == zb2->zb_object && zb1->zb_level == zb2->zb_level && zb1->zb_blkid == zb2->zb_blkid) return (0); /* * BP_SPANB calculates the span in blocks. */ zb1L0 = (zb1->zb_blkid) * BP_SPANB(ibs1, zb1->zb_level); zb2L0 = (zb2->zb_blkid) * BP_SPANB(ibs2, zb2->zb_level); if (zb1->zb_object == DMU_META_DNODE_OBJECT) { zb1obj = zb1L0 * (dbss1 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT)); zb1L0 = 0; zb1level = zb1->zb_level + COMPARE_META_LEVEL; } else { zb1obj = zb1->zb_object; zb1level = zb1->zb_level; } if (zb2->zb_object == DMU_META_DNODE_OBJECT) { zb2obj = zb2L0 * (dbss2 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT)); zb2L0 = 0; zb2level = zb2->zb_level + COMPARE_META_LEVEL; } else { zb2obj = zb2->zb_object; zb2level = zb2->zb_level; } /* Now that we have a canonical representation, do the comparison. */ if (zb1obj != zb2obj) return (zb1obj < zb2obj ? -1 : 1); else if (zb1L0 != zb2L0) return (zb1L0 < zb2L0 ? -1 : 1); else if (zb1level != zb2level) return (zb1level > zb2level ? -1 : 1); /* * This can (theoretically) happen if the bookmarks have the same object * and level, but different blkids, if the block sizes are not the same. * There is presently no way to change the indirect block sizes */ return (0); } /* * This function checks the following: given that last_block is the place that * our traversal stopped last time, does that guarantee that we've visited * every node under subtree_root? Therefore, we can't just use the raw output * of zbookmark_compare. We have to pass in a modified version of * subtree_root; by incrementing the block id, and then checking whether * last_block is before or equal to that, we can tell whether or not having * visited last_block implies that all of subtree_root's children have been * visited. */ boolean_t zbookmark_subtree_completed(const dnode_phys_t *dnp, const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block) { zbookmark_phys_t mod_zb = *subtree_root; mod_zb.zb_blkid++; ASSERT(last_block->zb_level == 0); /* The objset_phys_t isn't before anything. */ if (dnp == NULL) return (B_FALSE); /* * We pass in 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT) for the * data block size in sectors, because that variable is only used if * the bookmark refers to a block in the meta-dnode. Since we don't * know without examining it what object it refers to, and there's no * harm in passing in this value in other cases, we always pass it in. * * We pass in 0 for the indirect block size shift because zb2 must be * level 0. The indirect block size is only used to calculate the span * of the bookmark, but since the bookmark must be level 0, the span is * always 1, so the math works out. * * If you make changes to how the zbookmark_compare code works, be sure * to make sure that this code still works afterwards. */ return (zbookmark_compare(dnp->dn_datablkszsec, dnp->dn_indblkshift, 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT), 0, &mod_zb, last_block) <= 0); } Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zio_checksum.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zio_checksum.c (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zio_checksum.c (revision 290757) @@ -1,277 +1,404 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2013 by Delphix. All rights reserved. * Copyright (c) 2013, Joyent, Inc. All rights reserved. + * Copyright 2013 Saso Kiselkov. All rights reserved. */ #include #include +#include #include #include #include #include /* * Checksum vectors. * * In the SPA, everything is checksummed. We support checksum vectors * for three distinct reasons: * * 1. Different kinds of data need different levels of protection. * For SPA metadata, we always want a very strong checksum. * For user data, we let users make the trade-off between speed * and checksum strength. * * 2. Cryptographic hash and MAC algorithms are an area of active research. * It is likely that in future hash functions will be at least as strong * as current best-of-breed, and may be substantially faster as well. * We want the ability to take advantage of these new hashes as soon as * they become available. * * 3. If someone develops hardware that can compute a strong hash quickly, * we want the ability to take advantage of that hardware. * * Of course, we don't want a checksum upgrade to invalidate existing * data, so we store the checksum *function* in eight bits of the bp. * This gives us room for up to 256 different checksum functions. * * When writing a block, we always checksum it with the latest-and-greatest * checksum function of the appropriate strength. When reading a block, * we compare the expected checksum against the actual checksum, which we * compute via the checksum function specified by BP_GET_CHECKSUM(bp). + * + * SALTED CHECKSUMS + * + * To enable the use of less secure hash algorithms with dedup, we + * introduce the notion of salted checksums (MACs, really). A salted + * checksum is fed both a random 256-bit value (the salt) and the data + * to be checksummed. This salt is kept secret (stored on the pool, but + * never shown to the user). Thus even if an attacker knew of collision + * weaknesses in the hash algorithm, they won't be able to mount a known + * plaintext attack on the DDT, since the actual hash value cannot be + * known ahead of time. How the salt is used is algorithm-specific + * (some might simply prefix it to the data block, others might need to + * utilize a full-blown HMAC). On disk the salt is stored in a ZAP + * object in the MOS (DMU_POOL_CHECKSUM_SALT). + * + * CONTEXT TEMPLATES + * + * Some hashing algorithms need to perform a substantial amount of + * initialization work (e.g. salted checksums above may need to pre-hash + * the salt) before being able to process data. Performing this + * redundant work for each block would be wasteful, so we instead allow + * a checksum algorithm to do the work once (the first time it's used) + * and then keep this pre-initialized context as a template inside the + * spa_t (spa_cksum_tmpls). If the zio_checksum_info_t contains + * non-NULL ci_tmpl_init and ci_tmpl_free callbacks, they are used to + * construct and destruct the pre-initialized checksum context. The + * pre-initialized context is then reused during each checksum + * invocation and passed to the checksum function. */ /*ARGSUSED*/ static void -zio_checksum_off(const void *buf, uint64_t size, zio_cksum_t *zcp) +zio_checksum_off(const void *buf, uint64_t size, + const void *ctx_template, zio_cksum_t *zcp) { ZIO_SET_CHECKSUM(zcp, 0, 0, 0, 0); } zio_checksum_info_t zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS] = { - {{NULL, NULL}, 0, 0, 0, "inherit"}, - {{NULL, NULL}, 0, 0, 0, "on"}, - {{zio_checksum_off, zio_checksum_off}, 0, 0, 0, "off"}, - {{zio_checksum_SHA256, zio_checksum_SHA256}, 1, 1, 0, "label"}, - {{zio_checksum_SHA256, zio_checksum_SHA256}, 1, 1, 0, "gang_header"}, - {{fletcher_2_native, fletcher_2_byteswap}, 0, 1, 0, "zilog"}, - {{fletcher_2_native, fletcher_2_byteswap}, 0, 0, 0, "fletcher2"}, - {{fletcher_4_native, fletcher_4_byteswap}, 1, 0, 0, "fletcher4"}, - {{zio_checksum_SHA256, zio_checksum_SHA256}, 1, 0, 1, "sha256"}, - {{fletcher_4_native, fletcher_4_byteswap}, 0, 1, 0, "zilog2"}, - {{zio_checksum_off, zio_checksum_off}, 0, 0, 0, "noparity"}, + {{NULL, NULL}, NULL, NULL, 0, "inherit"}, + {{NULL, NULL}, NULL, NULL, 0, "on"}, + {{zio_checksum_off, zio_checksum_off}, + NULL, NULL, 0, "off"}, + {{zio_checksum_SHA256, zio_checksum_SHA256}, + NULL, NULL, ZCHECKSUM_FLAG_METADATA | ZCHECKSUM_FLAG_EMBEDDED, + "label"}, + {{zio_checksum_SHA256, zio_checksum_SHA256}, + NULL, NULL, ZCHECKSUM_FLAG_METADATA | ZCHECKSUM_FLAG_EMBEDDED, + "gang_header"}, + {{fletcher_2_native, fletcher_2_byteswap}, + NULL, NULL, ZCHECKSUM_FLAG_EMBEDDED, "zilog"}, + {{fletcher_2_native, fletcher_2_byteswap}, + NULL, NULL, 0, "fletcher2"}, + {{fletcher_4_native, fletcher_4_byteswap}, + NULL, NULL, ZCHECKSUM_FLAG_METADATA, "fletcher4"}, + {{zio_checksum_SHA256, zio_checksum_SHA256}, + NULL, NULL, ZCHECKSUM_FLAG_METADATA | ZCHECKSUM_FLAG_DEDUP | + ZCHECKSUM_FLAG_NOPWRITE, "sha256"}, + {{fletcher_4_native, fletcher_4_byteswap}, + NULL, NULL, ZCHECKSUM_FLAG_EMBEDDED, "zilog2"}, + {{zio_checksum_off, zio_checksum_off}, + NULL, NULL, 0, "noparity"}, +#ifdef illumos + {{zio_checksum_SHA512_native, zio_checksum_SHA512_byteswap}, + NULL, NULL, ZCHECKSUM_FLAG_METADATA | ZCHECKSUM_FLAG_DEDUP | + ZCHECKSUM_FLAG_NOPWRITE, "sha512"}, + {{zio_checksum_skein_native, zio_checksum_skein_byteswap}, + zio_checksum_skein_tmpl_init, zio_checksum_skein_tmpl_free, + ZCHECKSUM_FLAG_METADATA | ZCHECKSUM_FLAG_DEDUP | + ZCHECKSUM_FLAG_SALTED | ZCHECKSUM_FLAG_NOPWRITE, "skein"}, + {{zio_checksum_edonr_native, zio_checksum_edonr_byteswap}, + zio_checksum_edonr_tmpl_init, zio_checksum_edonr_tmpl_free, + ZCHECKSUM_FLAG_METADATA | ZCHECKSUM_FLAG_SALTED | + ZCHECKSUM_FLAG_NOPWRITE, "edonr"}, +#endif }; +spa_feature_t +zio_checksum_to_feature(enum zio_checksum cksum) +{ +#ifdef illumos + switch (cksum) { + case ZIO_CHECKSUM_SHA512: + return (SPA_FEATURE_SHA512); + case ZIO_CHECKSUM_SKEIN: + return (SPA_FEATURE_SKEIN); + case ZIO_CHECKSUM_EDONR: + return (SPA_FEATURE_EDONR); + } +#endif + return (SPA_FEATURE_NONE); +} + enum zio_checksum zio_checksum_select(enum zio_checksum child, enum zio_checksum parent) { ASSERT(child < ZIO_CHECKSUM_FUNCTIONS); ASSERT(parent < ZIO_CHECKSUM_FUNCTIONS); ASSERT(parent != ZIO_CHECKSUM_INHERIT && parent != ZIO_CHECKSUM_ON); if (child == ZIO_CHECKSUM_INHERIT) return (parent); if (child == ZIO_CHECKSUM_ON) return (ZIO_CHECKSUM_ON_VALUE); return (child); } enum zio_checksum zio_checksum_dedup_select(spa_t *spa, enum zio_checksum child, enum zio_checksum parent) { ASSERT((child & ZIO_CHECKSUM_MASK) < ZIO_CHECKSUM_FUNCTIONS); ASSERT((parent & ZIO_CHECKSUM_MASK) < ZIO_CHECKSUM_FUNCTIONS); ASSERT(parent != ZIO_CHECKSUM_INHERIT && parent != ZIO_CHECKSUM_ON); if (child == ZIO_CHECKSUM_INHERIT) return (parent); if (child == ZIO_CHECKSUM_ON) return (spa_dedup_checksum(spa)); if (child == (ZIO_CHECKSUM_ON | ZIO_CHECKSUM_VERIFY)) return (spa_dedup_checksum(spa) | ZIO_CHECKSUM_VERIFY); - ASSERT(zio_checksum_table[child & ZIO_CHECKSUM_MASK].ci_dedup || + ASSERT((zio_checksum_table[child & ZIO_CHECKSUM_MASK].ci_flags & + ZCHECKSUM_FLAG_DEDUP) || (child & ZIO_CHECKSUM_VERIFY) || child == ZIO_CHECKSUM_OFF); return (child); } /* * Set the external verifier for a gang block based on , * a tuple which is guaranteed to be unique for the life of the pool. */ static void zio_checksum_gang_verifier(zio_cksum_t *zcp, blkptr_t *bp) { dva_t *dva = BP_IDENTITY(bp); uint64_t txg = BP_PHYSICAL_BIRTH(bp); ASSERT(BP_IS_GANG(bp)); ZIO_SET_CHECKSUM(zcp, DVA_GET_VDEV(dva), DVA_GET_OFFSET(dva), txg, 0); } /* * Set the external verifier for a label block based on its offset. * The vdev is implicit, and the txg is unknowable at pool open time -- * hence the logic in vdev_uberblock_load() to find the most recent copy. */ static void zio_checksum_label_verifier(zio_cksum_t *zcp, uint64_t offset) { ZIO_SET_CHECKSUM(zcp, offset, 0, 0, 0); } /* + * Calls the template init function of a checksum which supports context + * templates and installs the template into the spa_t. + */ +static void +zio_checksum_template_init(enum zio_checksum checksum, spa_t *spa) +{ + zio_checksum_info_t *ci = &zio_checksum_table[checksum]; + + if (ci->ci_tmpl_init == NULL) + return; + if (spa->spa_cksum_tmpls[checksum] != NULL) + return; + + VERIFY(ci->ci_tmpl_free != NULL); + mutex_enter(&spa->spa_cksum_tmpls_lock); + if (spa->spa_cksum_tmpls[checksum] == NULL) { + spa->spa_cksum_tmpls[checksum] = + ci->ci_tmpl_init(&spa->spa_cksum_salt); + VERIFY(spa->spa_cksum_tmpls[checksum] != NULL); + } + mutex_exit(&spa->spa_cksum_tmpls_lock); +} + +/* * Generate the checksum. */ void zio_checksum_compute(zio_t *zio, enum zio_checksum checksum, void *data, uint64_t size) { blkptr_t *bp = zio->io_bp; uint64_t offset = zio->io_offset; zio_checksum_info_t *ci = &zio_checksum_table[checksum]; zio_cksum_t cksum; + spa_t *spa = zio->io_spa; ASSERT((uint_t)checksum < ZIO_CHECKSUM_FUNCTIONS); ASSERT(ci->ci_func[0] != NULL); - if (ci->ci_eck) { + zio_checksum_template_init(checksum, spa); + + if (ci->ci_flags & ZCHECKSUM_FLAG_EMBEDDED) { zio_eck_t *eck; if (checksum == ZIO_CHECKSUM_ZILOG2) { zil_chain_t *zilc = data; size = P2ROUNDUP_TYPED(zilc->zc_nused, ZIL_MIN_BLKSZ, uint64_t); eck = &zilc->zc_eck; } else { eck = (zio_eck_t *)((char *)data + size) - 1; } if (checksum == ZIO_CHECKSUM_GANG_HEADER) zio_checksum_gang_verifier(&eck->zec_cksum, bp); else if (checksum == ZIO_CHECKSUM_LABEL) zio_checksum_label_verifier(&eck->zec_cksum, offset); else bp->blk_cksum = eck->zec_cksum; eck->zec_magic = ZEC_MAGIC; - ci->ci_func[0](data, size, &cksum); + ci->ci_func[0](data, size, spa->spa_cksum_tmpls[checksum], + &cksum); eck->zec_cksum = cksum; } else { - ci->ci_func[0](data, size, &bp->blk_cksum); + ci->ci_func[0](data, size, spa->spa_cksum_tmpls[checksum], + &bp->blk_cksum); } } int zio_checksum_error(zio_t *zio, zio_bad_cksum_t *info) { blkptr_t *bp = zio->io_bp; uint_t checksum = (bp == NULL ? zio->io_prop.zp_checksum : (BP_IS_GANG(bp) ? ZIO_CHECKSUM_GANG_HEADER : BP_GET_CHECKSUM(bp))); int byteswap; int error; uint64_t size = (bp == NULL ? zio->io_size : (BP_IS_GANG(bp) ? SPA_GANGBLOCKSIZE : BP_GET_PSIZE(bp))); uint64_t offset = zio->io_offset; void *data = zio->io_data; zio_checksum_info_t *ci = &zio_checksum_table[checksum]; zio_cksum_t actual_cksum, expected_cksum, verifier; + spa_t *spa = zio->io_spa; if (checksum >= ZIO_CHECKSUM_FUNCTIONS || ci->ci_func[0] == NULL) return (SET_ERROR(EINVAL)); - if (ci->ci_eck) { + zio_checksum_template_init(checksum, spa); + + if (ci->ci_flags & ZCHECKSUM_FLAG_EMBEDDED) { zio_eck_t *eck; if (checksum == ZIO_CHECKSUM_ZILOG2) { zil_chain_t *zilc = data; uint64_t nused; eck = &zilc->zc_eck; if (eck->zec_magic == ZEC_MAGIC) nused = zilc->zc_nused; else if (eck->zec_magic == BSWAP_64(ZEC_MAGIC)) nused = BSWAP_64(zilc->zc_nused); else return (SET_ERROR(ECKSUM)); if (nused > size) return (SET_ERROR(ECKSUM)); size = P2ROUNDUP_TYPED(nused, ZIL_MIN_BLKSZ, uint64_t); } else { eck = (zio_eck_t *)((char *)data + size) - 1; } if (checksum == ZIO_CHECKSUM_GANG_HEADER) zio_checksum_gang_verifier(&verifier, bp); else if (checksum == ZIO_CHECKSUM_LABEL) zio_checksum_label_verifier(&verifier, offset); else verifier = bp->blk_cksum; byteswap = (eck->zec_magic == BSWAP_64(ZEC_MAGIC)); if (byteswap) byteswap_uint64_array(&verifier, sizeof (zio_cksum_t)); expected_cksum = eck->zec_cksum; eck->zec_cksum = verifier; - ci->ci_func[byteswap](data, size, &actual_cksum); + ci->ci_func[byteswap](data, size, + spa->spa_cksum_tmpls[checksum], &actual_cksum); eck->zec_cksum = expected_cksum; if (byteswap) byteswap_uint64_array(&expected_cksum, sizeof (zio_cksum_t)); } else { ASSERT(!BP_IS_GANG(bp)); byteswap = BP_SHOULD_BYTESWAP(bp); expected_cksum = bp->blk_cksum; - ci->ci_func[byteswap](data, size, &actual_cksum); + ci->ci_func[byteswap](data, size, + spa->spa_cksum_tmpls[checksum], &actual_cksum); } info->zbc_expected = expected_cksum; info->zbc_actual = actual_cksum; info->zbc_checksum_name = ci->ci_name; info->zbc_byteswapped = byteswap; info->zbc_injected = 0; info->zbc_has_cksum = 1; if (!ZIO_CHECKSUM_EQUAL(actual_cksum, expected_cksum)) return (SET_ERROR(ECKSUM)); if (zio_injection_enabled && !zio->io_error && (error = zio_handle_fault_injection(zio, ECKSUM)) != 0) { info->zbc_injected = 1; return (error); } return (0); +} + +/* + * Called by a spa_t that's about to be deallocated. This steps through + * all of the checksum context templates and deallocates any that were + * initialized using the algorithm-specific template init function. + */ +void +zio_checksum_templates_free(spa_t *spa) +{ + for (enum zio_checksum checksum = 0; + checksum < ZIO_CHECKSUM_FUNCTIONS; checksum++) { + if (spa->spa_cksum_tmpls[checksum] != NULL) { + zio_checksum_info_t *ci = &zio_checksum_table[checksum]; + + VERIFY(ci->ci_tmpl_free != NULL); + ci->ci_tmpl_free(spa->spa_cksum_tmpls[checksum]); + spa->spa_cksum_tmpls[checksum] = NULL; + } + } } Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/sys/debug.h =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/sys/debug.h (revision 290756) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/sys/debug.h (revision 290757) @@ -1,153 +1,164 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright 2010 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. */ /* * Copyright (c) 2012 by Delphix. All rights reserved. + * Copyright 2013 Saso Kiselkov. All rights reserved. */ /* Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T */ /* All Rights Reserved */ #ifndef _SYS_DEBUG_H #define _SYS_DEBUG_H #include #include #ifdef __cplusplus extern "C" { #endif /* * ASSERT(ex) causes a panic or debugger entry if expression ex is not * true. ASSERT() is included only for debugging, and is a no-op in * production kernels. VERIFY(ex), on the other hand, behaves like * ASSERT and is evaluated on both debug and non-debug kernels. */ #if defined(__STDC__) extern int assfail(const char *, const char *, int); #define VERIFY(EX) ((void)((EX) || assfail(#EX, __FILE__, __LINE__))) #ifdef DEBUG #define ASSERT(EX) ((void)((EX) || assfail(#EX, __FILE__, __LINE__))) #else #define ASSERT(x) ((void)0) #endif #else /* defined(__STDC__) */ extern int assfail(); #define VERIFY(EX) ((void)((EX) || assfail("EX", __FILE__, __LINE__))) #ifdef DEBUG #define ASSERT(EX) ((void)((EX) || assfail("EX", __FILE__, __LINE__))) #else #define ASSERT(x) ((void)0) #endif #endif /* defined(__STDC__) */ /* * Assertion variants sensitive to the compilation data model */ #if defined(_LP64) #define ASSERT64(x) ASSERT(x) #define ASSERT32(x) #else #define ASSERT64(x) #define ASSERT32(x) ASSERT(x) #endif /* * IMPLY and EQUIV are assertions of the form: * * if (a) then (b) * and * if (a) then (b) *AND* if (b) then (a) */ #ifdef DEBUG #define IMPLY(A, B) \ ((void)(((!(A)) || (B)) || \ assfail("(" #A ") implies (" #B ")", __FILE__, __LINE__))) #define EQUIV(A, B) \ ((void)((!!(A) == !!(B)) || \ assfail("(" #A ") is equivalent to (" #B ")", __FILE__, __LINE__))) #else #define IMPLY(A, B) ((void)0) #define EQUIV(A, B) ((void)0) #endif /* * ASSERT3() behaves like ASSERT() except that it is an explicit conditional, * and prints out the values of the left and right hand expressions as part of * the panic message to ease debugging. The three variants imply the type * of their arguments. ASSERT3S() is for signed data types, ASSERT3U() is * for unsigned, and ASSERT3P() is for pointers. The VERIFY3*() macros * have the same relationship as above. */ extern void assfail3(const char *, uintmax_t, const char *, uintmax_t, const char *, int); #define VERIFY3_IMPL(LEFT, OP, RIGHT, TYPE) do { \ const TYPE __left = (TYPE)(LEFT); \ const TYPE __right = (TYPE)(RIGHT); \ if (!(__left OP __right)) \ assfail3(#LEFT " " #OP " " #RIGHT, \ (uintmax_t)__left, #OP, (uintmax_t)__right, \ __FILE__, __LINE__); \ _NOTE(CONSTCOND) } while (0) #define VERIFY3S(x, y, z) VERIFY3_IMPL(x, y, z, int64_t) #define VERIFY3U(x, y, z) VERIFY3_IMPL(x, y, z, uint64_t) #define VERIFY3P(x, y, z) VERIFY3_IMPL(x, y, z, uintptr_t) #define VERIFY0(x) VERIFY3_IMPL(x, ==, 0, uintmax_t) #ifdef DEBUG #define ASSERT3S(x, y, z) VERIFY3_IMPL(x, y, z, int64_t) #define ASSERT3U(x, y, z) VERIFY3_IMPL(x, y, z, uint64_t) #define ASSERT3P(x, y, z) VERIFY3_IMPL(x, y, z, uintptr_t) #define ASSERT0(x) VERIFY3_IMPL(x, ==, 0, uintmax_t) #else #define ASSERT3S(x, y, z) ((void)0) #define ASSERT3U(x, y, z) ((void)0) #define ASSERT3P(x, y, z) ((void)0) #define ASSERT0(x) ((void)0) +#endif + +/* + * Compile-time assertion. The condition 'x' must be constant. + */ +#ifndef CTASSERT +#define CTASSERT(x) _CTASSERT(x, __LINE__) +#define _CTASSERT(x, y) __CTASSERT(x, y) +#define __CTASSERT(x, y) \ + typedef char __compile_time_assertion__ ## y [(x) ? 1 : -1] #endif #ifdef _KERNEL extern void abort_sequence_enter(char *); extern void debug_enter(char *); #endif /* _KERNEL */ #if defined(DEBUG) && !defined(__sun) /* CSTYLED */ #define STATIC #else /* CSTYLED */ #define STATIC static #endif #ifdef __cplusplus } #endif #endif /* _SYS_DEBUG_H */ Index: stable/10 =================================================================== --- stable/10 (revision 290756) +++ stable/10 (revision 290757) Property changes on: stable/10 ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head:r289422