diff --git a/module/zfs/zfs_dir.c b/module/zfs/zfs_dir.c index 670e31398d80..a373537dd4e0 100644 --- a/module/zfs/zfs_dir.c +++ b/module/zfs/zfs_dir.c @@ -1,1124 +1,1095 @@ /* * 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. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "fs/fs_subr.h" #include #include #include #include #include #include #include #include #include /* * zfs_match_find() is used by zfs_dirent_lock() to peform zap lookups * of names after deciding which is the appropriate lookup interface. */ static int zfs_match_find(zfs_sb_t *zsb, znode_t *dzp, char *name, boolean_t exact, boolean_t update, int *deflags, pathname_t *rpnp, uint64_t *zoid) { boolean_t conflict = B_FALSE; int error; if (zsb->z_norm) { matchtype_t mt = MT_FIRST; size_t bufsz = 0; char *buf = NULL; if (rpnp) { buf = rpnp->pn_buf; bufsz = rpnp->pn_bufsize; } if (exact) mt = MT_EXACT; /* * In the non-mixed case we only expect there would ever * be one match, but we need to use the normalizing lookup. */ error = zap_lookup_norm(zsb->z_os, dzp->z_id, name, 8, 1, zoid, mt, buf, bufsz, &conflict); } else { error = zap_lookup(zsb->z_os, dzp->z_id, name, 8, 1, zoid); } /* * Allow multiple entries provided the first entry is * the object id. Non-zpl consumers may safely make * use of the additional space. * * XXX: This should be a feature flag for compatibility */ if (error == EOVERFLOW) error = 0; if (zsb->z_norm && !error && deflags) *deflags = conflict ? ED_CASE_CONFLICT : 0; *zoid = ZFS_DIRENT_OBJ(*zoid); #ifdef HAVE_DNLC if (error == ENOENT && update) dnlc_update(ZTOI(dzp), name, DNLC_NO_VNODE); #endif /* HAVE_DNLC */ return (error); } /* * Lock a directory entry. A dirlock on protects that name * in dzp's directory zap object. As long as you hold a dirlock, you can * assume two things: (1) dzp cannot be reaped, and (2) no other thread * can change the zap entry for (i.e. link or unlink) this name. * * Input arguments: * dzp - znode for directory * name - name of entry to lock * flag - ZNEW: if the entry already exists, fail with EEXIST. * ZEXISTS: if the entry does not exist, fail with ENOENT. * ZSHARED: allow concurrent access with other ZSHARED callers. * ZXATTR: we want dzp's xattr directory * ZCILOOK: On a mixed sensitivity file system, * this lookup should be case-insensitive. * ZCIEXACT: On a purely case-insensitive file system, * this lookup should be case-sensitive. * ZRENAMING: we are locking for renaming, force narrow locks * ZHAVELOCK: Don't grab the z_name_lock for this call. The * current thread already holds it. * * Output arguments: * zpp - pointer to the znode for the entry (NULL if there isn't one) * dlpp - pointer to the dirlock for this entry (NULL on error) * direntflags - (case-insensitive lookup only) * flags if multiple case-sensitive matches exist in directory * realpnp - (case-insensitive lookup only) * actual name matched within the directory * * Return value: 0 on success or errno on failure. * * NOTE: Always checks for, and rejects, '.' and '..'. * NOTE: For case-insensitive file systems we take wide locks (see below), * but return znode pointers to a single match. */ int zfs_dirent_lock(zfs_dirlock_t **dlpp, znode_t *dzp, char *name, znode_t **zpp, int flag, int *direntflags, pathname_t *realpnp) { zfs_sb_t *zsb = ZTOZSB(dzp); zfs_dirlock_t *dl; boolean_t update; boolean_t exact; uint64_t zoid; #ifdef HAVE_DNLC vnode_t *vp = NULL; #endif /* HAVE_DNLC */ int error = 0; int cmpflags; *zpp = NULL; *dlpp = NULL; /* * Verify that we are not trying to lock '.', '..', or '.zfs' */ if ((name[0] == '.' && (name[1] == '\0' || (name[1] == '.' && name[2] == '\0'))) || (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0)) return (EEXIST); /* * Case sensitivity and normalization preferences are set when * the file system is created. These are stored in the * zsb->z_case and zsb->z_norm fields. These choices * affect what vnodes can be cached in the DNLC, how we * perform zap lookups, and the "width" of our dirlocks. * * A normal dirlock locks a single name. Note that with * normalization a name can be composed multiple ways, but * when normalized, these names all compare equal. A wide * dirlock locks multiple names. We need these when the file * system is supporting mixed-mode access. It is sometimes * necessary to lock all case permutations of file name at * once so that simultaneous case-insensitive/case-sensitive * behaves as rationally as possible. */ /* * Decide if exact matches should be requested when performing * a zap lookup on file systems supporting case-insensitive * access. */ exact = ((zsb->z_case == ZFS_CASE_INSENSITIVE) && (flag & ZCIEXACT)) || ((zsb->z_case == ZFS_CASE_MIXED) && !(flag & ZCILOOK)); /* * Only look in or update the DNLC if we are looking for the * name on a file system that does not require normalization * or case folding. We can also look there if we happen to be * on a non-normalizing, mixed sensitivity file system IF we * are looking for the exact name. * * Maybe can add TO-UPPERed version of name to dnlc in ci-only * case for performance improvement? */ update = !zsb->z_norm || ((zsb->z_case == ZFS_CASE_MIXED) && !(zsb->z_norm & ~U8_TEXTPREP_TOUPPER) && !(flag & ZCILOOK)); /* * ZRENAMING indicates we are in a situation where we should * take narrow locks regardless of the file system's * preferences for normalizing and case folding. This will * prevent us deadlocking trying to grab the same wide lock * twice if the two names happen to be case-insensitive * matches. */ if (flag & ZRENAMING) cmpflags = 0; else cmpflags = zsb->z_norm; /* * Wait until there are no locks on this name. * * Don't grab the the lock if it is already held. However, cannot * have both ZSHARED and ZHAVELOCK together. */ ASSERT(!(flag & ZSHARED) || !(flag & ZHAVELOCK)); if (!(flag & ZHAVELOCK)) rw_enter(&dzp->z_name_lock, RW_READER); mutex_enter(&dzp->z_lock); for (;;) { if (dzp->z_unlinked) { mutex_exit(&dzp->z_lock); if (!(flag & ZHAVELOCK)) rw_exit(&dzp->z_name_lock); return (ENOENT); } for (dl = dzp->z_dirlocks; dl != NULL; dl = dl->dl_next) { if ((u8_strcmp(name, dl->dl_name, 0, cmpflags, U8_UNICODE_LATEST, &error) == 0) || error != 0) break; } if (error != 0) { mutex_exit(&dzp->z_lock); if (!(flag & ZHAVELOCK)) rw_exit(&dzp->z_name_lock); return (ENOENT); } if (dl == NULL) { /* * Allocate a new dirlock and add it to the list. */ dl = kmem_alloc(sizeof (zfs_dirlock_t), KM_SLEEP); cv_init(&dl->dl_cv, NULL, CV_DEFAULT, NULL); dl->dl_name = name; dl->dl_sharecnt = 0; dl->dl_namelock = 0; dl->dl_namesize = 0; dl->dl_dzp = dzp; dl->dl_next = dzp->z_dirlocks; dzp->z_dirlocks = dl; break; } if ((flag & ZSHARED) && dl->dl_sharecnt != 0) break; cv_wait(&dl->dl_cv, &dzp->z_lock); } /* * If the z_name_lock was NOT held for this dirlock record it. */ if (flag & ZHAVELOCK) dl->dl_namelock = 1; if ((flag & ZSHARED) && ++dl->dl_sharecnt > 1 && dl->dl_namesize == 0) { /* * We're the second shared reference to dl. Make a copy of * dl_name in case the first thread goes away before we do. * Note that we initialize the new name before storing its * pointer into dl_name, because the first thread may load * dl->dl_name at any time. He'll either see the old value, * which is his, or the new shared copy; either is OK. */ dl->dl_namesize = strlen(dl->dl_name) + 1; name = kmem_alloc(dl->dl_namesize, KM_SLEEP); bcopy(dl->dl_name, name, dl->dl_namesize); dl->dl_name = name; } mutex_exit(&dzp->z_lock); /* * We have a dirlock on the name. (Note that it is the dirlock, * not the dzp's z_lock, that protects the name in the zap object.) * See if there's an object by this name; if so, put a hold on it. */ if (flag & ZXATTR) { error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zsb), &zoid, sizeof (zoid)); if (error == 0) error = (zoid == 0 ? ENOENT : 0); } else { #ifdef HAVE_DNLC if (update) vp = dnlc_lookup(ZTOI(dzp), name); if (vp == DNLC_NO_VNODE) { iput(vp); error = ENOENT; } else if (vp) { if (flag & ZNEW) { zfs_dirent_unlock(dl); iput(vp); return (EEXIST); } *dlpp = dl; *zpp = VTOZ(vp); return (0); } else { error = zfs_match_find(zsb, dzp, name, exact, update, direntflags, realpnp, &zoid); } #else error = zfs_match_find(zsb, dzp, name, exact, update, direntflags, realpnp, &zoid); #endif /* HAVE_DNLC */ } if (error) { if (error != ENOENT || (flag & ZEXISTS)) { zfs_dirent_unlock(dl); return (error); } } else { if (flag & ZNEW) { zfs_dirent_unlock(dl); return (EEXIST); } error = zfs_zget(zsb, zoid, zpp); if (error) { zfs_dirent_unlock(dl); return (error); } #ifdef HAVE_DNLC if (!(flag & ZXATTR) && update) dnlc_update(ZTOI(dzp), name, ZTOI(*zpp)); #endif /* HAVE_DNLC */ } *dlpp = dl; return (0); } /* * Unlock this directory entry and wake anyone who was waiting for it. */ void zfs_dirent_unlock(zfs_dirlock_t *dl) { znode_t *dzp = dl->dl_dzp; zfs_dirlock_t **prev_dl, *cur_dl; mutex_enter(&dzp->z_lock); if (!dl->dl_namelock) rw_exit(&dzp->z_name_lock); if (dl->dl_sharecnt > 1) { dl->dl_sharecnt--; mutex_exit(&dzp->z_lock); return; } prev_dl = &dzp->z_dirlocks; while ((cur_dl = *prev_dl) != dl) prev_dl = &cur_dl->dl_next; *prev_dl = dl->dl_next; cv_broadcast(&dl->dl_cv); mutex_exit(&dzp->z_lock); if (dl->dl_namesize != 0) kmem_free(dl->dl_name, dl->dl_namesize); cv_destroy(&dl->dl_cv); kmem_free(dl, sizeof (*dl)); } /* * Look up an entry in a directory. * * NOTE: '.' and '..' are handled as special cases because * no directory entries are actually stored for them. If this is * the root of a filesystem, then '.zfs' is also treated as a * special pseudo-directory. */ int zfs_dirlook(znode_t *dzp, char *name, struct inode **ipp, int flags, int *deflg, pathname_t *rpnp) { zfs_dirlock_t *dl; znode_t *zp; int error = 0; uint64_t parent; if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) { *ipp = ZTOI(dzp); igrab(*ipp); } else if (name[0] == '.' && name[1] == '.' && name[2] == 0) { zfs_sb_t *zsb = ZTOZSB(dzp); /* * If we are a snapshot mounted under .zfs, return * the inode pointer for the snapshot directory. */ if ((error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_PARENT(zsb), &parent, sizeof (parent))) != 0) return (error); if (parent == dzp->z_id && zsb->z_parent != zsb) { error = zfsctl_root_lookup(zsb->z_parent->z_ctldir, "snapshot", ipp, 0, kcred, NULL, NULL); return (error); } rw_enter(&dzp->z_parent_lock, RW_READER); error = zfs_zget(zsb, parent, &zp); if (error == 0) *ipp = ZTOI(zp); rw_exit(&dzp->z_parent_lock); } else if (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0) { *ipp = zfsctl_root(dzp); } else { int zf; zf = ZEXISTS | ZSHARED; if (flags & FIGNORECASE) zf |= ZCILOOK; error = zfs_dirent_lock(&dl, dzp, name, &zp, zf, deflg, rpnp); if (error == 0) { *ipp = ZTOI(zp); zfs_dirent_unlock(dl); dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */ } rpnp = NULL; } if ((flags & FIGNORECASE) && rpnp && !error) (void) strlcpy(rpnp->pn_buf, name, rpnp->pn_bufsize); return (error); } /* * unlinked Set (formerly known as the "delete queue") Error Handling * * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we * don't specify the name of the entry that we will be manipulating. We * also fib and say that we won't be adding any new entries to the * unlinked set, even though we might (this is to lower the minimum file * size that can be deleted in a full filesystem). So on the small * chance that the nlink list is using a fat zap (ie. has more than * 2000 entries), we *may* not pre-read a block that's needed. * Therefore it is remotely possible for some of the assertions * regarding the unlinked set below to fail due to i/o error. On a * nondebug system, this will result in the space being leaked. */ void zfs_unlinked_add(znode_t *zp, dmu_tx_t *tx) { zfs_sb_t *zsb = ZTOZSB(zp); ASSERT(zp->z_unlinked); ASSERT(zp->z_links == 0); VERIFY3U(0, ==, zap_add_int(zsb->z_os, zsb->z_unlinkedobj, zp->z_id, tx)); } +/* + * Clean up any znodes that had no links when we either crashed or + * (force) umounted the file system. + */ +void +zfs_unlinked_drain(zfs_sb_t *zsb) +{ + zap_cursor_t zc; + zap_attribute_t zap; + dmu_object_info_t doi; + znode_t *zp; + int error; + + /* + * Interate over the contents of the unlinked set. + */ + for (zap_cursor_init(&zc, zsb->z_os, zsb->z_unlinkedobj); + zap_cursor_retrieve(&zc, &zap) == 0; + zap_cursor_advance(&zc)) { + + /* + * See what kind of object we have in list + */ + + error = dmu_object_info(zsb->z_os, zap.za_first_integer, &doi); + if (error != 0) + continue; + + ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) || + (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS)); + /* + * We need to re-mark these list entries for deletion, + * so we pull them back into core and set zp->z_unlinked. + */ + error = zfs_zget(zsb, zap.za_first_integer, &zp); + + /* + * We may pick up znodes that are already marked for deletion. + * This could happen during the purge of an extended attribute + * directory. All we need to do is skip over them, since they + * are already in the system marked z_unlinked. + */ + if (error != 0) + continue; + + zp->z_unlinked = B_TRUE; + iput(ZTOI(zp)); + } + zap_cursor_fini(&zc); +} + /* * Delete the entire contents of a directory. Return a count * of the number of entries that could not be deleted. If we encounter * an error, return a count of at least one so that the directory stays * in the unlinked set. * * NOTE: this function assumes that the directory is inactive, * so there is no need to lock its entries before deletion. * Also, it assumes the directory contents is *only* regular * files. */ static int zfs_purgedir(znode_t *dzp) { zap_cursor_t zc; zap_attribute_t zap; znode_t *xzp; dmu_tx_t *tx; zfs_sb_t *zsb = ZTOZSB(dzp); zfs_dirlock_t dl; int skipped = 0; int error; for (zap_cursor_init(&zc, zsb->z_os, dzp->z_id); (error = zap_cursor_retrieve(&zc, &zap)) == 0; zap_cursor_advance(&zc)) { error = zfs_zget(zsb, ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp); if (error) { skipped += 1; continue; } ASSERT(S_ISREG(ZTOI(xzp)->i_mode)||S_ISLNK(ZTOI(xzp)->i_mode)); tx = dmu_tx_create(zsb->z_os); dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name); dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); /* Is this really needed ? */ zfs_sa_upgrade_txholds(tx, xzp); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); iput(ZTOI(xzp)); skipped += 1; continue; } bzero(&dl, sizeof (dl)); dl.dl_dzp = dzp; dl.dl_name = zap.za_name; error = zfs_link_destroy(&dl, xzp, tx, 0, NULL); if (error) skipped += 1; dmu_tx_commit(tx); iput(ZTOI(xzp)); } zap_cursor_fini(&zc); if (error != ENOENT) skipped += 1; return (skipped); } -/* - * Clean up any znodes that had no links when we either crashed or - * (force) umounted the file system. - */ -void -zfs_unlinked_drain(zfs_sb_t *zsb) -{ - zap_cursor_t zc; - zap_attribute_t zap; - dmu_object_info_t doi; - znode_t *zp; - int error; - - /* - * Interate over the contents of the unlinked set. - */ - for (zap_cursor_init(&zc, zsb->z_os, zsb->z_unlinkedobj); - zap_cursor_retrieve(&zc, &zap) == 0; - zap_cursor_advance(&zc)) { - - /* - * See what kind of object we have in list - */ - - error = dmu_object_info(zsb->z_os, zap.za_first_integer, &doi); - if (error != 0) - continue; - - ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) || - (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS)); - /* - * We need to re-mark these list entries for deletion, - * so we pull them back into core and set zp->z_unlinked. - */ - error = zfs_zget(zsb, zap.za_first_integer, &zp); - - /* - * We may pick up znodes that are already marked for deletion. - * This could happen during the purge of an extended attribute - * directory. All we need to do is skip over them, since they - * are already in the system marked z_unlinked. - */ - if (error != 0) - continue; - - zp->z_unlinked = B_TRUE; - - /* - * If this is an attribute directory, purge its contents. - */ - if (S_ISDIR(ZTOI(zp)->i_mode) && (zp->z_pflags & ZFS_XATTR)) { - /* - * We don't need to check the return value of - * zfs_purgedir here, because zfs_rmnode will just - * return this xattr directory to the unlinked set - * until all of its xattrs are gone. - */ - (void) zfs_purgedir(zp); - } - - iput(ZTOI(zp)); - } - zap_cursor_fini(&zc); -} - void zfs_rmnode(znode_t *zp) { zfs_sb_t *zsb = ZTOZSB(zp); objset_t *os = zsb->z_os; znode_t *xzp = NULL; dmu_tx_t *tx; uint64_t acl_obj; uint64_t xattr_obj; - uint64_t count; int error; ASSERT(zp->z_links == 0); ASSERT(atomic_read(&ZTOI(zp)->i_count) == 0); /* * If this is an attribute directory, purge its contents. */ if (S_ISDIR(ZTOI(zp)->i_mode) && (zp->z_pflags & ZFS_XATTR)) { - error = zap_count(os, zp->z_id, &count); - if (error) { - zfs_znode_dmu_fini(zp); - return; - } - - if (count > 0) { - taskq_t *taskq; - + if (zfs_purgedir(zp) != 0) { /* - * There are still directory entries in this xattr - * directory. Let zfs_unlinked_drain() deal with - * them to avoid deadlocking this process in the - * zfs_purgedir()->zfs_zget()->ilookup() callpath - * on the xattr inode's I_FREEING bit. + * Not enough space to delete some xattrs. + * Leave it in the unlinked set. */ - taskq = dsl_pool_iput_taskq(dmu_objset_pool(os)); - taskq_dispatch(taskq, (task_func_t *) - zfs_unlinked_drain, zsb, TQ_SLEEP); - zfs_znode_dmu_fini(zp); + return; } } /* * Free up all the data in the file. */ error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END); if (error) { /* * Not enough space. Leave the file in the unlinked set. */ zfs_znode_dmu_fini(zp); return; } /* * If the file has extended attributes, we're going to unlink * the xattr dir. */ error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), &xattr_obj, sizeof (xattr_obj)); if (error == 0 && xattr_obj) { error = zfs_zget(zsb, xattr_obj, &xzp); ASSERT(error == 0); } acl_obj = zfs_external_acl(zp); /* * Set up the final transaction. */ tx = dmu_tx_create(os); dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END); dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); if (xzp) { dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, TRUE, NULL); dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); } if (acl_obj) dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); zfs_sa_upgrade_txholds(tx, zp); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { /* * Not enough space to delete the file. Leave it in the * unlinked set, leaking it until the fs is remounted (at * which point we'll call zfs_unlinked_drain() to process it). */ dmu_tx_abort(tx); zfs_znode_dmu_fini(zp); goto out; } if (xzp) { ASSERT(error == 0); mutex_enter(&xzp->z_lock); xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */ xzp->z_links = 0; /* no more links to it */ VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zsb), &xzp->z_links, sizeof (xzp->z_links), tx)); mutex_exit(&xzp->z_lock); zfs_unlinked_add(xzp, tx); } /* Remove this znode from the unlinked set */ VERIFY3U(0, ==, zap_remove_int(zsb->z_os, zsb->z_unlinkedobj, zp->z_id, tx)); zfs_znode_delete(zp, tx); dmu_tx_commit(tx); out: if (xzp) iput(ZTOI(xzp)); } static uint64_t zfs_dirent(znode_t *zp, uint64_t mode) { uint64_t de = zp->z_id; if (ZTOZSB(zp)->z_version >= ZPL_VERSION_DIRENT_TYPE) de |= IFTODT(mode) << 60; return (de); } /* * Link zp into dl. Can only fail if zp has been unlinked. */ int zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag) { znode_t *dzp = dl->dl_dzp; zfs_sb_t *zsb = ZTOZSB(zp); uint64_t value; int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode); sa_bulk_attr_t bulk[5]; uint64_t mtime[2], ctime[2]; int count = 0; int error; mutex_enter(&zp->z_lock); if (!(flag & ZRENAMING)) { if (zp->z_unlinked) { /* no new links to unlinked zp */ ASSERT(!(flag & (ZNEW | ZEXISTS))); mutex_exit(&zp->z_lock); return (ENOENT); } zp->z_links++; SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, sizeof (zp->z_links)); } SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL, &dzp->z_id, sizeof (dzp->z_id)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, &zp->z_pflags, sizeof (zp->z_pflags)); if (!(flag & ZNEW)) { SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, sizeof (ctime)); zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime, B_TRUE); } error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); ASSERT(error == 0); mutex_exit(&zp->z_lock); mutex_enter(&dzp->z_lock); dzp->z_size++; dzp->z_links += zp_is_dir; count = 0; SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &dzp->z_size, sizeof (dzp->z_size)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &dzp->z_links, sizeof (dzp->z_links)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, sizeof (mtime)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, sizeof (ctime)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, &dzp->z_pflags, sizeof (dzp->z_pflags)); zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); ASSERT(error == 0); mutex_exit(&dzp->z_lock); value = zfs_dirent(zp, zp->z_mode); error = zap_add(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name, 8, 1, &value, tx); ASSERT(error == 0); return (0); } static int zfs_dropname(zfs_dirlock_t *dl, znode_t *zp, znode_t *dzp, dmu_tx_t *tx, int flag) { int error; if (ZTOZSB(zp)->z_norm) { if (((ZTOZSB(zp)->z_case == ZFS_CASE_INSENSITIVE) && (flag & ZCIEXACT)) || ((ZTOZSB(zp)->z_case == ZFS_CASE_MIXED) && !(flag & ZCILOOK))) error = zap_remove_norm(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name, MT_EXACT, tx); else error = zap_remove_norm(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name, MT_FIRST, tx); } else { error = zap_remove(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name, tx); } return (error); } /* * Unlink zp from dl, and mark zp for deletion if this was the last link. Can * fail if zp is a mount point (EBUSY) or a non-empty directory (ENOTEMPTY). * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list. * If it's non-NULL, we use it to indicate whether the znode needs deletion, * and it's the caller's job to do it. */ int zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag, boolean_t *unlinkedp) { znode_t *dzp = dl->dl_dzp; zfs_sb_t *zsb = ZTOZSB(dzp); int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode); boolean_t unlinked = B_FALSE; sa_bulk_attr_t bulk[5]; uint64_t mtime[2], ctime[2]; int count = 0; int error; #ifdef HAVE_DNLC dnlc_remove(ZTOI(dzp), dl->dl_name); #endif /* HAVE_DNLC */ if (!(flag & ZRENAMING)) { mutex_enter(&zp->z_lock); if (zp_is_dir && !zfs_dirempty(zp)) { mutex_exit(&zp->z_lock); return (ENOTEMPTY); } /* * If we get here, we are going to try to remove the object. * First try removing the name from the directory; if that * fails, return the error. */ error = zfs_dropname(dl, zp, dzp, tx, flag); if (error != 0) { mutex_exit(&zp->z_lock); return (error); } if (zp->z_links <= zp_is_dir) { zfs_panic_recover("zfs: link count on %lu is %u, " "should be at least %u", zp->z_id, (int)zp->z_links, zp_is_dir + 1); zp->z_links = zp_is_dir + 1; } if (--zp->z_links == zp_is_dir) { zp->z_unlinked = B_TRUE; zp->z_links = 0; unlinked = B_TRUE; } else { SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, sizeof (ctime)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, &zp->z_pflags, sizeof (zp->z_pflags)); zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime, B_TRUE); } SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, sizeof (zp->z_links)); error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); count = 0; ASSERT(error == 0); mutex_exit(&zp->z_lock); } else { error = zfs_dropname(dl, zp, dzp, tx, flag); if (error != 0) return (error); } mutex_enter(&dzp->z_lock); dzp->z_size--; /* one dirent removed */ dzp->z_links -= zp_is_dir; /* ".." link from zp */ SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &dzp->z_links, sizeof (dzp->z_links)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &dzp->z_size, sizeof (dzp->z_size)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, sizeof (ctime)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, sizeof (mtime)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, &dzp->z_pflags, sizeof (dzp->z_pflags)); zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); ASSERT(error == 0); mutex_exit(&dzp->z_lock); if (unlinkedp != NULL) *unlinkedp = unlinked; else if (unlinked) zfs_unlinked_add(zp, tx); return (0); } /* * Indicate whether the directory is empty. Works with or without z_lock * held, but can only be consider a hint in the latter case. Returns true * if only "." and ".." remain and there's no work in progress. */ boolean_t zfs_dirempty(znode_t *dzp) { return (dzp->z_size == 2 && dzp->z_dirlocks == 0); } int zfs_make_xattrdir(znode_t *zp, vattr_t *vap, struct inode **xipp, cred_t *cr) { zfs_sb_t *zsb = ZTOZSB(zp); znode_t *xzp; dmu_tx_t *tx; int error; zfs_acl_ids_t acl_ids; boolean_t fuid_dirtied; #ifdef DEBUG uint64_t parent; #endif *xipp = NULL; if ((error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))) return (error); if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL, &acl_ids)) != 0) return (error); if (zfs_acl_ids_overquota(zsb, &acl_ids)) { zfs_acl_ids_free(&acl_ids); return (EDQUOT); } top: tx = dmu_tx_create(zsb->z_os); dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + ZFS_SA_BASE_ATTR_SIZE); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); fuid_dirtied = zsb->z_fuid_dirty; if (fuid_dirtied) zfs_fuid_txhold(zsb, tx); error = dmu_tx_assign(tx, TXG_NOWAIT); if (error) { if (error == ERESTART) { dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } zfs_acl_ids_free(&acl_ids); dmu_tx_abort(tx); return (error); } zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids); if (fuid_dirtied) zfs_fuid_sync(zsb, tx); #ifdef DEBUG error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zsb), &parent, sizeof (parent)); ASSERT(error == 0 && parent == zp->z_id); #endif VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), &xzp->z_id, sizeof (xzp->z_id), tx)); (void) zfs_log_create(zsb->z_log, tx, TX_MKXATTR, zp, xzp, "", NULL, acl_ids.z_fuidp, vap); zfs_acl_ids_free(&acl_ids); dmu_tx_commit(tx); *xipp = ZTOI(xzp); return (0); } /* * Return a znode for the extended attribute directory for zp. * ** If the directory does not already exist, it is created ** * * IN: zp - znode to obtain attribute directory from * cr - credentials of caller * flags - flags from the VOP_LOOKUP call * * OUT: xipp - pointer to extended attribute znode * * RETURN: 0 on success * error number on failure */ int zfs_get_xattrdir(znode_t *zp, struct inode **xipp, cred_t *cr, int flags) { zfs_sb_t *zsb = ZTOZSB(zp); znode_t *xzp; zfs_dirlock_t *dl; vattr_t va; int error; top: error = zfs_dirent_lock(&dl, zp, "", &xzp, ZXATTR, NULL, NULL); if (error) return (error); if (xzp != NULL) { *xipp = ZTOI(xzp); zfs_dirent_unlock(dl); return (0); } if (!(flags & CREATE_XATTR_DIR)) { zfs_dirent_unlock(dl); return (ENOENT); } if (zfs_is_readonly(zsb)) { zfs_dirent_unlock(dl); return (EROFS); } /* * The ability to 'create' files in an attribute * directory comes from the write_xattr permission on the base file. * * The ability to 'search' an attribute directory requires * read_xattr permission on the base file. * * Once in a directory the ability to read/write attributes * is controlled by the permissions on the attribute file. */ va.va_mask = ATTR_MODE | ATTR_UID | ATTR_GID; va.va_mode = S_IFDIR | S_ISVTX | 0777; zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid); va.va_dentry = NULL; error = zfs_make_xattrdir(zp, &va, xipp, cr); zfs_dirent_unlock(dl); if (error == ERESTART) { /* NB: we already did dmu_tx_wait() if necessary */ goto top; } return (error); } /* * Decide whether it is okay to remove within a sticky directory. * * In sticky directories, write access is not sufficient; * you can remove entries from a directory only if: * * you own the directory, * you own the entry, * the entry is a plain file and you have write access, * or you are privileged (checked in secpolicy...). * * The function returns 0 if remove access is granted. */ int zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr) { uid_t uid; uid_t downer; uid_t fowner; zfs_sb_t *zsb = ZTOZSB(zdp); if (zsb->z_replay) return (0); if ((zdp->z_mode & S_ISVTX) == 0) return (0); downer = zfs_fuid_map_id(zsb, zdp->z_uid, cr, ZFS_OWNER); fowner = zfs_fuid_map_id(zsb, zp->z_uid, cr, ZFS_OWNER); if ((uid = crgetuid(cr)) == downer || uid == fowner || (S_ISDIR(ZTOI(zp)->i_mode) && zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0)) return (0); else return (secpolicy_vnode_remove(cr)); } diff --git a/module/zfs/zfs_vfsops.c b/module/zfs/zfs_vfsops.c index 90f9055afdc5..175dca821462 100644 --- a/module/zfs/zfs_vfsops.c +++ b/module/zfs/zfs_vfsops.c @@ -1,1593 +1,1587 @@ /* * 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 2010 Robert Milkowski */ #include #include #include #include #include #include #include #include #include #include #include #include #include "fs/fs_subr.h" #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_comutil.h" /*ARGSUSED*/ int zfs_sync(struct super_block *sb, int wait, cred_t *cr) { zfs_sb_t *zsb = sb->s_fs_info; /* * Data integrity is job one. We don't want a compromised kernel * writing to the storage pool, so we never sync during panic. */ if (unlikely(oops_in_progress)) return (0); /* * Semantically, the only requirement is that the sync be initiated. * The DMU syncs out txgs frequently, so there's nothing to do. */ if (!wait) return (0); if (zsb != NULL) { /* * Sync a specific filesystem. */ dsl_pool_t *dp; ZFS_ENTER(zsb); dp = dmu_objset_pool(zsb->z_os); /* * If the system is shutting down, then skip any * filesystems which may exist on a suspended pool. */ if (spa_suspended(dp->dp_spa)) { ZFS_EXIT(zsb); return (0); } if (zsb->z_log != NULL) zil_commit(zsb->z_log, 0); ZFS_EXIT(zsb); } else { /* * Sync all ZFS filesystems. This is what happens when you * run sync(1M). Unlike other filesystems, ZFS honors the * request by waiting for all pools to commit all dirty data. */ spa_sync_allpools(); } return (0); } EXPORT_SYMBOL(zfs_sync); boolean_t zfs_is_readonly(zfs_sb_t *zsb) { return (!!(zsb->z_sb->s_flags & MS_RDONLY)); } EXPORT_SYMBOL(zfs_is_readonly); static void atime_changed_cb(void *arg, uint64_t newval) { ((zfs_sb_t *)arg)->z_atime = newval; } static void xattr_changed_cb(void *arg, uint64_t newval) { zfs_sb_t *zsb = arg; if (newval == ZFS_XATTR_OFF) { zsb->z_flags &= ~ZSB_XATTR; } else { zsb->z_flags |= ZSB_XATTR; if (newval == ZFS_XATTR_SA) zsb->z_xattr_sa = B_TRUE; else zsb->z_xattr_sa = B_FALSE; } } static void blksz_changed_cb(void *arg, uint64_t newval) { zfs_sb_t *zsb = arg; if (newval < SPA_MINBLOCKSIZE || newval > SPA_MAXBLOCKSIZE || !ISP2(newval)) newval = SPA_MAXBLOCKSIZE; zsb->z_max_blksz = newval; } static void readonly_changed_cb(void *arg, uint64_t newval) { zfs_sb_t *zsb = arg; struct super_block *sb = zsb->z_sb; if (sb == NULL) return; if (newval) sb->s_flags |= MS_RDONLY; else sb->s_flags &= ~MS_RDONLY; } static void devices_changed_cb(void *arg, uint64_t newval) { } static void setuid_changed_cb(void *arg, uint64_t newval) { } static void exec_changed_cb(void *arg, uint64_t newval) { } static void nbmand_changed_cb(void *arg, uint64_t newval) { zfs_sb_t *zsb = arg; struct super_block *sb = zsb->z_sb; if (sb == NULL) return; if (newval == TRUE) sb->s_flags |= MS_MANDLOCK; else sb->s_flags &= ~MS_MANDLOCK; } static void snapdir_changed_cb(void *arg, uint64_t newval) { ((zfs_sb_t *)arg)->z_show_ctldir = newval; } static void vscan_changed_cb(void *arg, uint64_t newval) { ((zfs_sb_t *)arg)->z_vscan = newval; } static void acl_inherit_changed_cb(void *arg, uint64_t newval) { ((zfs_sb_t *)arg)->z_acl_inherit = newval; } int zfs_register_callbacks(zfs_sb_t *zsb) { struct dsl_dataset *ds = NULL; objset_t *os = zsb->z_os; int error = 0; if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) readonly_changed_cb(zsb, B_TRUE); /* * Register property callbacks. * * It would probably be fine to just check for i/o error from * the first prop_register(), but I guess I like to go * overboard... */ ds = dmu_objset_ds(os); error = dsl_prop_register(ds, "atime", atime_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "xattr", xattr_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "recordsize", blksz_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "readonly", readonly_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "devices", devices_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "setuid", setuid_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "exec", exec_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "snapdir", snapdir_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "aclinherit", acl_inherit_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "vscan", vscan_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "nbmand", nbmand_changed_cb, zsb); if (error) goto unregister; return (0); unregister: /* * We may attempt to unregister some callbacks that are not * registered, but this is OK; it will simply return ENOMSG, * which we will ignore. */ (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zsb); (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zsb); (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zsb); (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zsb); (void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zsb); (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zsb); (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zsb); (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zsb); (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb, zsb); (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zsb); (void) dsl_prop_unregister(ds, "nbmand", nbmand_changed_cb, zsb); return (error); } EXPORT_SYMBOL(zfs_register_callbacks); static int zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, uint64_t *userp, uint64_t *groupp) { znode_phys_t *znp = data; int error = 0; /* * Is it a valid type of object to track? */ if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) return (ENOENT); /* * If we have a NULL data pointer * then assume the id's aren't changing and * return EEXIST to the dmu to let it know to * use the same ids */ if (data == NULL) return (EEXIST); if (bonustype == DMU_OT_ZNODE) { *userp = znp->zp_uid; *groupp = znp->zp_gid; } else { int hdrsize; ASSERT(bonustype == DMU_OT_SA); hdrsize = sa_hdrsize(data); if (hdrsize != 0) { *userp = *((uint64_t *)((uintptr_t)data + hdrsize + SA_UID_OFFSET)); *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + SA_GID_OFFSET)); } else { /* * This should only happen for newly created * files that haven't had the znode data filled * in yet. */ *userp = 0; *groupp = 0; } } return (error); } static void fuidstr_to_sid(zfs_sb_t *zsb, const char *fuidstr, char *domainbuf, int buflen, uid_t *ridp) { uint64_t fuid; const char *domain; fuid = strtonum(fuidstr, NULL); domain = zfs_fuid_find_by_idx(zsb, FUID_INDEX(fuid)); if (domain) (void) strlcpy(domainbuf, domain, buflen); else domainbuf[0] = '\0'; *ridp = FUID_RID(fuid); } static uint64_t zfs_userquota_prop_to_obj(zfs_sb_t *zsb, zfs_userquota_prop_t type) { switch (type) { case ZFS_PROP_USERUSED: return (DMU_USERUSED_OBJECT); case ZFS_PROP_GROUPUSED: return (DMU_GROUPUSED_OBJECT); case ZFS_PROP_USERQUOTA: return (zsb->z_userquota_obj); case ZFS_PROP_GROUPQUOTA: return (zsb->z_groupquota_obj); default: return (ENOTSUP); } return (0); } int zfs_userspace_many(zfs_sb_t *zsb, zfs_userquota_prop_t type, uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) { int error; zap_cursor_t zc; zap_attribute_t za; zfs_useracct_t *buf = vbuf; uint64_t obj; if (!dmu_objset_userspace_present(zsb->z_os)) return (ENOTSUP); obj = zfs_userquota_prop_to_obj(zsb, type); if (obj == 0) { *bufsizep = 0; return (0); } for (zap_cursor_init_serialized(&zc, zsb->z_os, obj, *cookiep); (error = zap_cursor_retrieve(&zc, &za)) == 0; zap_cursor_advance(&zc)) { if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > *bufsizep) break; fuidstr_to_sid(zsb, za.za_name, buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); buf->zu_space = za.za_first_integer; buf++; } if (error == ENOENT) error = 0; ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; *cookiep = zap_cursor_serialize(&zc); zap_cursor_fini(&zc); return (error); } EXPORT_SYMBOL(zfs_userspace_many); /* * buf must be big enough (eg, 32 bytes) */ static int id_to_fuidstr(zfs_sb_t *zsb, const char *domain, uid_t rid, char *buf, boolean_t addok) { uint64_t fuid; int domainid = 0; if (domain && domain[0]) { domainid = zfs_fuid_find_by_domain(zsb, domain, NULL, addok); if (domainid == -1) return (ENOENT); } fuid = FUID_ENCODE(domainid, rid); (void) sprintf(buf, "%llx", (longlong_t)fuid); return (0); } int zfs_userspace_one(zfs_sb_t *zsb, zfs_userquota_prop_t type, const char *domain, uint64_t rid, uint64_t *valp) { char buf[32]; int err; uint64_t obj; *valp = 0; if (!dmu_objset_userspace_present(zsb->z_os)) return (ENOTSUP); obj = zfs_userquota_prop_to_obj(zsb, type); if (obj == 0) return (0); err = id_to_fuidstr(zsb, domain, rid, buf, B_FALSE); if (err) return (err); err = zap_lookup(zsb->z_os, obj, buf, 8, 1, valp); if (err == ENOENT) err = 0; return (err); } EXPORT_SYMBOL(zfs_userspace_one); int zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type, const char *domain, uint64_t rid, uint64_t quota) { char buf[32]; int err; dmu_tx_t *tx; uint64_t *objp; boolean_t fuid_dirtied; if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) return (EINVAL); if (zsb->z_version < ZPL_VERSION_USERSPACE) return (ENOTSUP); objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj : &zsb->z_groupquota_obj; err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE); if (err) return (err); fuid_dirtied = zsb->z_fuid_dirty; tx = dmu_tx_create(zsb->z_os); dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); if (*objp == 0) { dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, zfs_userquota_prop_prefixes[type]); } if (fuid_dirtied) zfs_fuid_txhold(zsb, tx); err = dmu_tx_assign(tx, TXG_WAIT); if (err) { dmu_tx_abort(tx); return (err); } mutex_enter(&zsb->z_lock); if (*objp == 0) { *objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA, DMU_OT_NONE, 0, tx); VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); } mutex_exit(&zsb->z_lock); if (quota == 0) { err = zap_remove(zsb->z_os, *objp, buf, tx); if (err == ENOENT) err = 0; } else { err = zap_update(zsb->z_os, *objp, buf, 8, 1, "a, tx); } ASSERT(err == 0); if (fuid_dirtied) zfs_fuid_sync(zsb, tx); dmu_tx_commit(tx); return (err); } EXPORT_SYMBOL(zfs_set_userquota); boolean_t zfs_fuid_overquota(zfs_sb_t *zsb, boolean_t isgroup, uint64_t fuid) { char buf[32]; uint64_t used, quota, usedobj, quotaobj; int err; usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; if (quotaobj == 0 || zsb->z_replay) return (B_FALSE); (void) sprintf(buf, "%llx", (longlong_t)fuid); err = zap_lookup(zsb->z_os, quotaobj, buf, 8, 1, "a); if (err != 0) return (B_FALSE); err = zap_lookup(zsb->z_os, usedobj, buf, 8, 1, &used); if (err != 0) return (B_FALSE); return (used >= quota); } EXPORT_SYMBOL(zfs_fuid_overquota); boolean_t zfs_owner_overquota(zfs_sb_t *zsb, znode_t *zp, boolean_t isgroup) { uint64_t fuid; uint64_t quotaobj; quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; fuid = isgroup ? zp->z_gid : zp->z_uid; if (quotaobj == 0 || zsb->z_replay) return (B_FALSE); return (zfs_fuid_overquota(zsb, isgroup, fuid)); } EXPORT_SYMBOL(zfs_owner_overquota); int zfs_sb_create(const char *osname, zfs_sb_t **zsbp) { objset_t *os; zfs_sb_t *zsb; uint64_t zval; int i, error; uint64_t sa_obj; zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP | KM_NODEBUG); /* * We claim to always be readonly so we can open snapshots; * other ZPL code will prevent us from writing to snapshots. */ error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os); if (error) { kmem_free(zsb, sizeof (zfs_sb_t)); return (error); } /* * Initialize the zfs-specific filesystem structure. * Should probably make this a kmem cache, shuffle fields, * and just bzero up to z_hold_mtx[]. */ zsb->z_sb = NULL; zsb->z_parent = zsb; zsb->z_max_blksz = SPA_MAXBLOCKSIZE; zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; zsb->z_os = os; error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version); if (error) { goto out; } else if (zsb->z_version > zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { (void) printk("Can't mount a version %lld file system " "on a version %lld pool\n. Pool must be upgraded to mount " "this file system.", (u_longlong_t)zsb->z_version, (u_longlong_t)spa_version(dmu_objset_spa(os))); error = ENOTSUP; goto out; } if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) goto out; zsb->z_norm = (int)zval; if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) goto out; zsb->z_utf8 = (zval != 0); if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) goto out; zsb->z_case = (uint_t)zval; /* * Fold case on file systems that are always or sometimes case * insensitive. */ if (zsb->z_case == ZFS_CASE_INSENSITIVE || zsb->z_case == ZFS_CASE_MIXED) zsb->z_norm |= U8_TEXTPREP_TOUPPER; zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); if (zsb->z_use_sa) { /* should either have both of these objects or none */ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); if (error) goto out; error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval); if ((error == 0) && (zval == ZFS_XATTR_SA)) zsb->z_xattr_sa = B_TRUE; } else { /* * Pre SA versions file systems should never touch * either the attribute registration or layout objects. */ sa_obj = 0; } error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, &zsb->z_attr_table); if (error) goto out; if (zsb->z_version >= ZPL_VERSION_SA) sa_register_update_callback(os, zfs_sa_upgrade); error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, &zsb->z_root); if (error) goto out; ASSERT(zsb->z_root != 0); error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, &zsb->z_unlinkedobj); if (error) goto out; error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], 8, 1, &zsb->z_userquota_obj); if (error && error != ENOENT) goto out; error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], 8, 1, &zsb->z_groupquota_obj); if (error && error != ENOENT) goto out; error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, &zsb->z_fuid_obj); if (error && error != ENOENT) goto out; error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, &zsb->z_shares_dir); if (error && error != ENOENT) goto out; mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL); list_create(&zsb->z_all_znodes, sizeof (znode_t), offsetof(znode_t, z_link_node)); rrw_init(&zsb->z_teardown_lock); rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL); for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); avl_create(&zsb->z_ctldir_snaps, snapentry_compare, sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t, se_node)); mutex_init(&zsb->z_ctldir_lock, NULL, MUTEX_DEFAULT, NULL); *zsbp = zsb; return (0); out: dmu_objset_disown(os, zsb); *zsbp = NULL; kmem_free(zsb, sizeof (zfs_sb_t)); return (error); } EXPORT_SYMBOL(zfs_sb_create); int zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting) { int error; error = zfs_register_callbacks(zsb); if (error) return (error); /* * Set the objset user_ptr to track its zsb. */ mutex_enter(&zsb->z_os->os_user_ptr_lock); dmu_objset_set_user(zsb->z_os, zsb); mutex_exit(&zsb->z_os->os_user_ptr_lock); zsb->z_log = zil_open(zsb->z_os, zfs_get_data); /* * If we are not mounting (ie: online recv), then we don't * have to worry about replaying the log as we blocked all * operations out since we closed the ZIL. */ if (mounting) { boolean_t readonly; /* * During replay we remove the read only flag to * allow replays to succeed. */ readonly = zfs_is_readonly(zsb); if (readonly != 0) readonly_changed_cb(zsb, B_FALSE); else zfs_unlinked_drain(zsb); /* * Parse and replay the intent log. * * Because of ziltest, this must be done after * zfs_unlinked_drain(). (Further note: ziltest * doesn't use readonly mounts, where * zfs_unlinked_drain() isn't called.) This is because * ziltest causes spa_sync() to think it's committed, * but actually it is not, so the intent log contains * many txg's worth of changes. * * In particular, if object N is in the unlinked set in * the last txg to actually sync, then it could be * actually freed in a later txg and then reallocated * in a yet later txg. This would write a "create * object N" record to the intent log. Normally, this * would be fine because the spa_sync() would have * written out the fact that object N is free, before * we could write the "create object N" intent log * record. * * But when we are in ziltest mode, we advance the "open * txg" without actually spa_sync()-ing the changes to * disk. So we would see that object N is still * allocated and in the unlinked set, and there is an * intent log record saying to allocate it. */ if (spa_writeable(dmu_objset_spa(zsb->z_os))) { if (zil_replay_disable) { zil_destroy(zsb->z_log, B_FALSE); } else { zsb->z_replay = B_TRUE; zil_replay(zsb->z_os, zsb, zfs_replay_vector); zsb->z_replay = B_FALSE; } } /* restore readonly bit */ if (readonly != 0) readonly_changed_cb(zsb, B_TRUE); } return (0); } EXPORT_SYMBOL(zfs_sb_setup); void zfs_sb_free(zfs_sb_t *zsb) { int i; zfs_fuid_destroy(zsb); mutex_destroy(&zsb->z_znodes_lock); mutex_destroy(&zsb->z_lock); list_destroy(&zsb->z_all_znodes); rrw_destroy(&zsb->z_teardown_lock); rw_destroy(&zsb->z_teardown_inactive_lock); rw_destroy(&zsb->z_fuid_lock); for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) mutex_destroy(&zsb->z_hold_mtx[i]); mutex_destroy(&zsb->z_ctldir_lock); avl_destroy(&zsb->z_ctldir_snaps); kmem_free(zsb, sizeof (zfs_sb_t)); } EXPORT_SYMBOL(zfs_sb_free); static void zfs_set_fuid_feature(zfs_sb_t *zsb) { zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); } void zfs_unregister_callbacks(zfs_sb_t *zsb) { objset_t *os = zsb->z_os; struct dsl_dataset *ds; /* * Unregister properties. */ if (!dmu_objset_is_snapshot(os)) { ds = dmu_objset_ds(os); VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb, zsb) == 0); VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zsb) == 0); VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zsb) == 0); VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zsb) == 0); VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb, zsb) == 0); VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zsb) == 0); VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb, zsb) == 0); VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zsb) == 0); VERIFY(dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb, zsb) == 0); VERIFY(dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zsb) == 0); VERIFY(dsl_prop_unregister(ds, "nbmand", nbmand_changed_cb, zsb) == 0); } } EXPORT_SYMBOL(zfs_unregister_callbacks); #ifdef HAVE_MLSLABEL /* * zfs_check_global_label: * Check that the hex label string is appropriate for the dataset * being mounted into the global_zone proper. * * Return an error if the hex label string is not default or * admin_low/admin_high. For admin_low labels, the corresponding * dataset must be readonly. */ int zfs_check_global_label(const char *dsname, const char *hexsl) { if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) return (0); if (strcasecmp(hexsl, ADMIN_HIGH) == 0) return (0); if (strcasecmp(hexsl, ADMIN_LOW) == 0) { /* must be readonly */ uint64_t rdonly; if (dsl_prop_get_integer(dsname, zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) return (EACCES); return (rdonly ? 0 : EACCES); } return (EACCES); } EXPORT_SYMBOL(zfs_check_global_label); #endif /* HAVE_MLSLABEL */ int zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) { zfs_sb_t *zsb = dentry->d_sb->s_fs_info; uint64_t refdbytes, availbytes, usedobjs, availobjs; uint64_t fsid; uint32_t bshift; ZFS_ENTER(zsb); dmu_objset_space(zsb->z_os, &refdbytes, &availbytes, &usedobjs, &availobjs); fsid = dmu_objset_fsid_guid(zsb->z_os); /* * The underlying storage pool actually uses multiple block * size. Under Solaris frsize (fragment size) is reported as * the smallest block size we support, and bsize (block size) * as the filesystem's maximum block size. Unfortunately, * under Linux the fragment size and block size are often used * interchangeably. Thus we are forced to report both of them * as the filesystem's maximum block size. */ statp->f_frsize = zsb->z_max_blksz; statp->f_bsize = zsb->z_max_blksz; bshift = fls(statp->f_bsize) - 1; /* * The following report "total" blocks of various kinds in * the file system, but reported in terms of f_bsize - the * "preferred" size. */ statp->f_blocks = (refdbytes + availbytes) >> bshift; statp->f_bfree = availbytes >> bshift; statp->f_bavail = statp->f_bfree; /* no root reservation */ /* * statvfs() should really be called statufs(), because it assumes * static metadata. ZFS doesn't preallocate files, so the best * we can do is report the max that could possibly fit in f_files, * and that minus the number actually used in f_ffree. * For f_ffree, report the smaller of the number of object available * and the number of blocks (each object will take at least a block). */ statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); statp->f_files = statp->f_ffree + usedobjs; statp->f_fsid.val[0] = (uint32_t)fsid; statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); statp->f_type = ZFS_SUPER_MAGIC; statp->f_namelen = ZFS_MAXNAMELEN; /* * We have all of 40 characters to stuff a string here. * Is there anything useful we could/should provide? */ bzero(statp->f_spare, sizeof (statp->f_spare)); ZFS_EXIT(zsb); return (0); } EXPORT_SYMBOL(zfs_statvfs); int zfs_root(zfs_sb_t *zsb, struct inode **ipp) { znode_t *rootzp; int error; ZFS_ENTER(zsb); error = zfs_zget(zsb, zsb->z_root, &rootzp); if (error == 0) *ipp = ZTOI(rootzp); ZFS_EXIT(zsb); return (error); } EXPORT_SYMBOL(zfs_root); #ifdef HAVE_SHRINK int zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) { zfs_sb_t *zsb = sb->s_fs_info; struct shrinker *shrinker = &sb->s_shrink; struct shrink_control sc = { .nr_to_scan = nr_to_scan, .gfp_mask = GFP_KERNEL, }; ZFS_ENTER(zsb); *objects = (*shrinker->shrink)(shrinker, &sc); ZFS_EXIT(zsb); return (0); } EXPORT_SYMBOL(zfs_sb_prune); #endif /* HAVE_SHRINK */ /* * Teardown the zfs_sb_t::z_os. * * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' * and 'z_teardown_inactive_lock' held. */ int zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting) { znode_t *zp; rrw_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG); if (!unmounting) { /* * We purge the parent filesystem's super block as the * parent filesystem and all of its snapshots have their * inode's super block set to the parent's filesystem's * super block. Note, 'z_parent' is self referential * for non-snapshots. */ shrink_dcache_sb(zsb->z_parent->z_sb); (void) spl_invalidate_inodes(zsb->z_parent->z_sb, 0); } - /* - * Drain the iput_taskq to ensure all active references to the - * zfs_sb_t have been handled only then can it be safely destroyed. - */ - taskq_wait(dsl_pool_iput_taskq(dmu_objset_pool(zsb->z_os))); - /* * Close the zil. NB: Can't close the zil while zfs_inactive * threads are blocked as zil_close can call zfs_inactive. */ if (zsb->z_log) { zil_close(zsb->z_log); zsb->z_log = NULL; } rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER); /* * If we are not unmounting (ie: online recv) and someone already * unmounted this file system while we were doing the switcheroo, * or a reopen of z_os failed then just bail out now. */ if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) { rw_exit(&zsb->z_teardown_inactive_lock); rrw_exit(&zsb->z_teardown_lock, FTAG); return (EIO); } /* * At this point there are no vops active, and any new vops will * fail with EIO since we have z_teardown_lock for writer (only * relavent for forced unmount). * * Release all holds on dbufs. */ mutex_enter(&zsb->z_znodes_lock); for (zp = list_head(&zsb->z_all_znodes); zp != NULL; zp = list_next(&zsb->z_all_znodes, zp)) if (zp->z_sa_hdl) { ASSERT(atomic_read(&ZTOI(zp)->i_count) > 0); zfs_znode_dmu_fini(zp); } mutex_exit(&zsb->z_znodes_lock); /* * If we are unmounting, set the unmounted flag and let new vops * unblock. zfs_inactive will have the unmounted behavior, and all * other vops will fail with EIO. */ if (unmounting) { zsb->z_unmounted = B_TRUE; rrw_exit(&zsb->z_teardown_lock, FTAG); rw_exit(&zsb->z_teardown_inactive_lock); } /* * z_os will be NULL if there was an error in attempting to reopen * zsb, so just return as the properties had already been * * unregistered and cached data had been evicted before. */ if (zsb->z_os == NULL) return (0); /* * Unregister properties. */ zfs_unregister_callbacks(zsb); /* * Evict cached data */ if (dsl_dataset_is_dirty(dmu_objset_ds(zsb->z_os)) && !zfs_is_readonly(zsb)) txg_wait_synced(dmu_objset_pool(zsb->z_os), 0); (void) dmu_objset_evict_dbufs(zsb->z_os); return (0); } EXPORT_SYMBOL(zfs_sb_teardown); #if defined(HAVE_BDI) && !defined(HAVE_BDI_SETUP_AND_REGISTER) atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); #endif /* HAVE_BDI && !HAVE_BDI_SETUP_AND_REGISTER */ int zfs_domount(struct super_block *sb, void *data, int silent) { zpl_mount_data_t *zmd = data; const char *osname = zmd->z_osname; zfs_sb_t *zsb; struct inode *root_inode; uint64_t recordsize; int error; error = zfs_sb_create(osname, &zsb); if (error) return (error); if ((error = dsl_prop_get_integer(osname, "recordsize", &recordsize, NULL))) goto out; zsb->z_sb = sb; sb->s_fs_info = zsb; sb->s_magic = ZFS_SUPER_MAGIC; sb->s_maxbytes = MAX_LFS_FILESIZE; sb->s_time_gran = 1; sb->s_blocksize = recordsize; sb->s_blocksize_bits = ilog2(recordsize); #ifdef HAVE_BDI /* * 2.6.32 API change, * Added backing_device_info (BDI) per super block interfaces. A BDI * must be configured when using a non-device backed filesystem for * proper writeback. This is not required for older pdflush kernels. * * NOTE: Linux read-ahead is disabled in favor of zfs read-ahead. */ zsb->z_bdi.ra_pages = 0; sb->s_bdi = &zsb->z_bdi; error = -bdi_setup_and_register(&zsb->z_bdi, "zfs", BDI_CAP_MAP_COPY); if (error) goto out; #endif /* HAVE_BDI */ /* Set callback operations for the file system. */ sb->s_op = &zpl_super_operations; sb->s_xattr = zpl_xattr_handlers; sb->s_export_op = &zpl_export_operations; /* Set features for file system. */ zfs_set_fuid_feature(zsb); if (dmu_objset_is_snapshot(zsb->z_os)) { uint64_t pval; atime_changed_cb(zsb, B_FALSE); readonly_changed_cb(zsb, B_TRUE); if ((error = dsl_prop_get_integer(osname,"xattr",&pval,NULL))) goto out; xattr_changed_cb(zsb, pval); zsb->z_issnap = B_TRUE; zsb->z_os->os_sync = ZFS_SYNC_DISABLED; mutex_enter(&zsb->z_os->os_user_ptr_lock); dmu_objset_set_user(zsb->z_os, zsb); mutex_exit(&zsb->z_os->os_user_ptr_lock); } else { error = zfs_sb_setup(zsb, B_TRUE); } /* Allocate a root inode for the filesystem. */ error = zfs_root(zsb, &root_inode); if (error) { (void) zfs_umount(sb); goto out; } /* Allocate a root dentry for the filesystem */ sb->s_root = d_make_root(root_inode); if (sb->s_root == NULL) { (void) zfs_umount(sb); error = ENOMEM; goto out; } if (!zsb->z_issnap) zfsctl_create(zsb); out: if (error) { dmu_objset_disown(zsb->z_os, zsb); zfs_sb_free(zsb); } return (error); } EXPORT_SYMBOL(zfs_domount); /* * Called when an unmount is requested and certain sanity checks have * already passed. At this point no dentries or inodes have been reclaimed * from their respective caches. We drop the extra reference on the .zfs * control directory to allow everything to be reclaimed. All snapshots * must already have been unmounted to reach this point. */ void zfs_preumount(struct super_block *sb) { zfs_sb_t *zsb = sb->s_fs_info; if (zsb != NULL && zsb->z_ctldir != NULL) zfsctl_destroy(zsb); } EXPORT_SYMBOL(zfs_preumount); /* * Called once all other unmount released tear down has occurred. * It is our responsibility to release any remaining infrastructure. */ /*ARGSUSED*/ int zfs_umount(struct super_block *sb) { zfs_sb_t *zsb = sb->s_fs_info; objset_t *os; VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0); os = zsb->z_os; #ifdef HAVE_BDI bdi_destroy(sb->s_bdi); #endif /* HAVE_BDI */ /* * z_os will be NULL if there was an error in * attempting to reopen zsb. */ if (os != NULL) { /* * Unset the objset user_ptr. */ mutex_enter(&os->os_user_ptr_lock); dmu_objset_set_user(os, NULL); mutex_exit(&os->os_user_ptr_lock); /* * Finally release the objset */ dmu_objset_disown(os, zsb); } zfs_sb_free(zsb); return (0); } EXPORT_SYMBOL(zfs_umount); int zfs_remount(struct super_block *sb, int *flags, char *data) { /* * All namespace flags (MNT_*) and super block flags (MS_*) will * be handled by the Linux VFS. Only handle custom options here. */ return (0); } EXPORT_SYMBOL(zfs_remount); int zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) { zfs_sb_t *zsb = sb->s_fs_info; znode_t *zp; uint64_t object = 0; uint64_t fid_gen = 0; uint64_t gen_mask; uint64_t zp_gen; int i, err; *ipp = NULL; ZFS_ENTER(zsb); if (fidp->fid_len == LONG_FID_LEN) { zfid_long_t *zlfid = (zfid_long_t *)fidp; uint64_t objsetid = 0; uint64_t setgen = 0; for (i = 0; i < sizeof (zlfid->zf_setid); i++) objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); for (i = 0; i < sizeof (zlfid->zf_setgen); i++) setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); ZFS_EXIT(zsb); err = zfsctl_lookup_objset(sb, objsetid, &zsb); if (err) return (EINVAL); ZFS_ENTER(zsb); } if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { zfid_short_t *zfid = (zfid_short_t *)fidp; for (i = 0; i < sizeof (zfid->zf_object); i++) object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); for (i = 0; i < sizeof (zfid->zf_gen); i++) fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); } else { ZFS_EXIT(zsb); return (EINVAL); } /* A zero fid_gen means we are in the .zfs control directories */ if (fid_gen == 0 && (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { *ipp = zsb->z_ctldir; ASSERT(*ipp != NULL); if (object == ZFSCTL_INO_SNAPDIR) { VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, 0, kcred, NULL, NULL) == 0); } else { igrab(*ipp); } ZFS_EXIT(zsb); return (0); } gen_mask = -1ULL >> (64 - 8 * i); dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask); if ((err = zfs_zget(zsb, object, &zp))) { ZFS_EXIT(zsb); return (err); } (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen, sizeof (uint64_t)); zp_gen = zp_gen & gen_mask; if (zp_gen == 0) zp_gen = 1; if (zp->z_unlinked || zp_gen != fid_gen) { dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen, fid_gen); iput(ZTOI(zp)); ZFS_EXIT(zsb); return (EINVAL); } *ipp = ZTOI(zp); if (*ipp) zfs_inode_update(ITOZ(*ipp)); ZFS_EXIT(zsb); return (0); } EXPORT_SYMBOL(zfs_vget); /* * Block out VOPs and close zfs_sb_t::z_os * * Note, if successful, then we return with the 'z_teardown_lock' and * 'z_teardown_inactive_lock' write held. */ int zfs_suspend_fs(zfs_sb_t *zsb) { int error; if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0) return (error); dmu_objset_disown(zsb->z_os, zsb); return (0); } EXPORT_SYMBOL(zfs_suspend_fs); /* * Reopen zfs_sb_t::z_os and release VOPs. */ int zfs_resume_fs(zfs_sb_t *zsb, const char *osname) { int err, err2; ASSERT(RRW_WRITE_HELD(&zsb->z_teardown_lock)); ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)); err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zsb, &zsb->z_os); if (err) { zsb->z_os = NULL; } else { znode_t *zp; uint64_t sa_obj = 0; err2 = zap_lookup(zsb->z_os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); if ((err || err2) && zsb->z_version >= ZPL_VERSION_SA) goto bail; if ((err = sa_setup(zsb->z_os, sa_obj, zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0) goto bail; VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0); /* * Attempt to re-establish all the active znodes with * their dbufs. If a zfs_rezget() fails, then we'll let * any potential callers discover that via ZFS_ENTER_VERIFY_VP * when they try to use their znode. */ mutex_enter(&zsb->z_znodes_lock); for (zp = list_head(&zsb->z_all_znodes); zp; zp = list_next(&zsb->z_all_znodes, zp)) { (void) zfs_rezget(zp); } mutex_exit(&zsb->z_znodes_lock); } bail: /* release the VOPs */ rw_exit(&zsb->z_teardown_inactive_lock); rrw_exit(&zsb->z_teardown_lock, FTAG); if (err) { /* * Since we couldn't reopen zfs_sb_t::z_os, force * unmount this file system. */ (void) zfs_umount(zsb->z_sb); } return (err); } EXPORT_SYMBOL(zfs_resume_fs); int zfs_set_version(zfs_sb_t *zsb, uint64_t newvers) { int error; objset_t *os = zsb->z_os; dmu_tx_t *tx; if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) return (EINVAL); if (newvers < zsb->z_version) return (EINVAL); if (zfs_spa_version_map(newvers) > spa_version(dmu_objset_spa(zsb->z_os))) return (ENOTSUP); tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, ZFS_SA_ATTRS); dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); } error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); return (error); } error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, 8, 1, &newvers, tx); if (error) { dmu_tx_commit(tx); return (error); } if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { uint64_t sa_obj; ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=, SPA_VERSION_SA); sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, DMU_OT_NONE, 0, tx); error = zap_add(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); ASSERT3U(error, ==, 0); VERIFY(0 == sa_set_sa_object(os, sa_obj)); sa_register_update_callback(os, zfs_sa_upgrade); } spa_history_log_internal(LOG_DS_UPGRADE, dmu_objset_spa(os), tx, "oldver=%llu newver=%llu dataset = %llu", zsb->z_version, newvers, dmu_objset_id(os)); dmu_tx_commit(tx); zsb->z_version = newvers; if (zsb->z_version >= ZPL_VERSION_FUID) zfs_set_fuid_feature(zsb); return (0); } EXPORT_SYMBOL(zfs_set_version); /* * Read a property stored within the master node. */ int zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) { const char *pname; int error = ENOENT; /* * Look up the file system's value for the property. For the * version property, we look up a slightly different string. */ if (prop == ZFS_PROP_VERSION) pname = ZPL_VERSION_STR; else pname = zfs_prop_to_name(prop); if (os != NULL) error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); if (error == ENOENT) { /* No value set, use the default value */ switch (prop) { case ZFS_PROP_VERSION: *value = ZPL_VERSION; break; case ZFS_PROP_NORMALIZE: case ZFS_PROP_UTF8ONLY: *value = 0; break; case ZFS_PROP_CASE: *value = ZFS_CASE_SENSITIVE; break; default: return (error); } error = 0; } return (error); } EXPORT_SYMBOL(zfs_get_zplprop); void zfs_init(void) { zfsctl_init(); zfs_znode_init(); dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); register_filesystem(&zpl_fs_type); (void) arc_add_prune_callback(zpl_prune_sbs, NULL); } void zfs_fini(void) { unregister_filesystem(&zpl_fs_type); zfs_znode_fini(); zfsctl_fini(); }