diff --git a/include/os/freebsd/zfs/sys/zfs_vfsops_os.h b/include/os/freebsd/zfs/sys/zfs_vfsops_os.h index f765d38dbac8..24bb03575f33 100644 --- a/include/os/freebsd/zfs/sys/zfs_vfsops_os.h +++ b/include/os/freebsd/zfs/sys/zfs_vfsops_os.h @@ -1,311 +1,310 @@ /* * 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 https://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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011 Pawel Jakub Dawidek . * All rights reserved. */ #ifndef _SYS_FS_ZFS_VFSOPS_H #define _SYS_FS_ZFS_VFSOPS_H #if __FreeBSD_version >= 1300125 #define TEARDOWN_RMS #endif #if __FreeBSD_version >= 1300109 #define TEARDOWN_INACTIVE_RMS #endif #include #include #include #include #include #include #ifdef TEARDOWN_INACTIVE_RMS #include #endif #include #ifdef __cplusplus extern "C" { #endif #ifdef TEARDOWN_RMS typedef struct rmslock zfs_teardown_lock_t; #else #define zfs_teardown_lock_t rrmlock_t #endif #ifdef TEARDOWN_INACTIVE_RMS typedef struct rmslock zfs_teardown_inactive_lock_t; #else #define zfs_teardown_inactive_lock_t krwlock_t #endif typedef struct zfsvfs zfsvfs_t; struct znode; struct zfsvfs { vfs_t *z_vfs; /* generic fs struct */ zfsvfs_t *z_parent; /* parent fs */ objset_t *z_os; /* objset reference */ uint64_t z_flags; /* super_block flags */ uint64_t z_root; /* id of root znode */ uint64_t z_unlinkedobj; /* id of unlinked zapobj */ uint64_t z_max_blksz; /* maximum block size for files */ uint64_t z_fuid_obj; /* fuid table object number */ uint64_t z_fuid_size; /* fuid table size */ avl_tree_t z_fuid_idx; /* fuid tree keyed by index */ avl_tree_t z_fuid_domain; /* fuid tree keyed by domain */ krwlock_t z_fuid_lock; /* fuid lock */ boolean_t z_fuid_loaded; /* fuid tables are loaded */ boolean_t z_fuid_dirty; /* need to sync fuid table ? */ struct zfs_fuid_info *z_fuid_replay; /* fuid info for replay */ zilog_t *z_log; /* intent log pointer */ uint_t z_acl_type; /* type of acl usable on this fs */ uint_t z_acl_mode; /* acl chmod/mode behavior */ uint_t z_acl_inherit; /* acl inheritance behavior */ zfs_case_t z_case; /* case-sense */ boolean_t z_utf8; /* utf8-only */ int z_norm; /* normalization flags */ boolean_t z_atime; /* enable atimes mount option */ boolean_t z_unmounted; /* unmounted */ zfs_teardown_lock_t z_teardown_lock; zfs_teardown_inactive_lock_t z_teardown_inactive_lock; list_t z_all_znodes; /* all vnodes in the fs */ - uint64_t z_nr_znodes; /* number of znodes in the fs */ kmutex_t z_znodes_lock; /* lock for z_all_znodes */ struct zfsctl_root *z_ctldir; /* .zfs directory pointer */ boolean_t z_show_ctldir; /* expose .zfs in the root dir */ boolean_t z_issnap; /* true if this is a snapshot */ boolean_t z_use_fuids; /* version allows fuids */ boolean_t z_replay; /* set during ZIL replay */ boolean_t z_use_sa; /* version allow system attributes */ boolean_t z_xattr_sa; /* allow xattrs to be stores as SA */ boolean_t z_use_namecache; /* make use of FreeBSD name cache */ uint8_t z_xattr; /* xattr type in use */ uint64_t z_version; /* ZPL version */ uint64_t z_shares_dir; /* hidden shares dir */ dataset_kstats_t z_kstat; /* fs kstats */ kmutex_t z_lock; uint64_t z_userquota_obj; uint64_t z_groupquota_obj; uint64_t z_userobjquota_obj; uint64_t z_groupobjquota_obj; uint64_t z_projectquota_obj; uint64_t z_projectobjquota_obj; uint64_t z_replay_eof; /* New end of file - replay only */ sa_attr_type_t *z_attr_table; /* SA attr mapping->id */ #define ZFS_OBJ_MTX_SZ 64 kmutex_t z_hold_mtx[ZFS_OBJ_MTX_SZ]; /* znode hold locks */ struct task z_unlinked_drain_task; }; #ifdef TEARDOWN_RMS #define ZFS_TEARDOWN_INIT(zfsvfs) \ rms_init(&(zfsvfs)->z_teardown_lock, "zfs teardown") #define ZFS_TEARDOWN_DESTROY(zfsvfs) \ rms_destroy(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_ENTER_READ(zfsvfs, tag) \ rms_rlock(&(zfsvfs)->z_teardown_lock); #define ZFS_TEARDOWN_EXIT_READ(zfsvfs, tag) \ rms_runlock(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_ENTER_WRITE(zfsvfs, tag) \ rms_wlock(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_EXIT_WRITE(zfsvfs) \ rms_wunlock(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_EXIT(zfsvfs, tag) \ rms_unlock(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_READ_HELD(zfsvfs) \ rms_rowned(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_WRITE_HELD(zfsvfs) \ rms_wowned(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_HELD(zfsvfs) \ rms_owned_any(&(zfsvfs)->z_teardown_lock) #else #define ZFS_TEARDOWN_INIT(zfsvfs) \ rrm_init(&(zfsvfs)->z_teardown_lock, B_FALSE) #define ZFS_TEARDOWN_DESTROY(zfsvfs) \ rrm_destroy(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_ENTER_READ(zfsvfs, tag) \ rrm_enter_read(&(zfsvfs)->z_teardown_lock, tag); #define ZFS_TEARDOWN_EXIT_READ(zfsvfs, tag) \ rrm_exit(&(zfsvfs)->z_teardown_lock, tag) #define ZFS_TEARDOWN_ENTER_WRITE(zfsvfs, tag) \ rrm_enter(&(zfsvfs)->z_teardown_lock, RW_WRITER, tag) #define ZFS_TEARDOWN_EXIT_WRITE(zfsvfs) \ rrm_exit(&(zfsvfs)->z_teardown_lock, tag) #define ZFS_TEARDOWN_EXIT(zfsvfs, tag) \ rrm_exit(&(zfsvfs)->z_teardown_lock, tag) #define ZFS_TEARDOWN_READ_HELD(zfsvfs) \ RRM_READ_HELD(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_WRITE_HELD(zfsvfs) \ RRM_WRITE_HELD(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_HELD(zfsvfs) \ RRM_LOCK_HELD(&(zfsvfs)->z_teardown_lock) #endif #ifdef TEARDOWN_INACTIVE_RMS #define ZFS_TEARDOWN_INACTIVE_INIT(zfsvfs) \ rms_init(&(zfsvfs)->z_teardown_inactive_lock, "zfs teardown inactive") #define ZFS_TEARDOWN_INACTIVE_DESTROY(zfsvfs) \ rms_destroy(&(zfsvfs)->z_teardown_inactive_lock) #define ZFS_TEARDOWN_INACTIVE_TRY_ENTER_READ(zfsvfs) \ rms_try_rlock(&(zfsvfs)->z_teardown_inactive_lock) #define ZFS_TEARDOWN_INACTIVE_ENTER_READ(zfsvfs) \ rms_rlock(&(zfsvfs)->z_teardown_inactive_lock) #define ZFS_TEARDOWN_INACTIVE_EXIT_READ(zfsvfs) \ rms_runlock(&(zfsvfs)->z_teardown_inactive_lock) #define ZFS_TEARDOWN_INACTIVE_ENTER_WRITE(zfsvfs) \ rms_wlock(&(zfsvfs)->z_teardown_inactive_lock) #define ZFS_TEARDOWN_INACTIVE_EXIT_WRITE(zfsvfs) \ rms_wunlock(&(zfsvfs)->z_teardown_inactive_lock) #define ZFS_TEARDOWN_INACTIVE_WRITE_HELD(zfsvfs) \ rms_wowned(&(zfsvfs)->z_teardown_inactive_lock) #else #define ZFS_TEARDOWN_INACTIVE_INIT(zfsvfs) \ rw_init(&(zfsvfs)->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL) #define ZFS_TEARDOWN_INACTIVE_DESTROY(zfsvfs) \ rw_destroy(&(zfsvfs)->z_teardown_inactive_lock) #define ZFS_TEARDOWN_INACTIVE_TRY_ENTER_READ(zfsvfs) \ rw_tryenter(&(zfsvfs)->z_teardown_inactive_lock, RW_READER) #define ZFS_TEARDOWN_INACTIVE_ENTER_READ(zfsvfs) \ rw_enter(&(zfsvfs)->z_teardown_inactive_lock, RW_READER) #define ZFS_TEARDOWN_INACTIVE_EXIT_READ(zfsvfs) \ rw_exit(&(zfsvfs)->z_teardown_inactive_lock) #define ZFS_TEARDOWN_INACTIVE_ENTER_WRITE(zfsvfs) \ rw_enter(&(zfsvfs)->z_teardown_inactive_lock, RW_WRITER) #define ZFS_TEARDOWN_INACTIVE_EXIT_WRITE(zfsvfs) \ rw_exit(&(zfsvfs)->z_teardown_inactive_lock) #define ZFS_TEARDOWN_INACTIVE_WRITE_HELD(zfsvfs) \ RW_WRITE_HELD(&(zfsvfs)->z_teardown_inactive_lock) #endif #define ZSB_XATTR 0x0001 /* Enable user xattrs */ /* * Normal filesystems (those not under .zfs/snapshot) have a total * file ID size limited to 12 bytes (including the length field) due to * NFSv2 protocol's limitation of 32 bytes for a filehandle. For historical * reasons, this same limit is being imposed by the Solaris NFSv3 implementation * (although the NFSv3 protocol actually permits a maximum of 64 bytes). It * is not possible to expand beyond 12 bytes without abandoning support * of NFSv2. * * For normal filesystems, we partition up the available space as follows: * 2 bytes fid length (required) * 6 bytes object number (48 bits) * 4 bytes generation number (32 bits) * * We reserve only 48 bits for the object number, as this is the limit * currently defined and imposed by the DMU. */ typedef struct zfid_short { uint16_t zf_len; uint8_t zf_object[6]; /* obj[i] = obj >> (8 * i) */ uint8_t zf_gen[4]; /* gen[i] = gen >> (8 * i) */ } zfid_short_t; /* * Filesystems under .zfs/snapshot have a total file ID size of 22[*] bytes * (including the length field). This makes files under .zfs/snapshot * accessible by NFSv3 and NFSv4, but not NFSv2. * * For files under .zfs/snapshot, we partition up the available space * as follows: * 2 bytes fid length (required) * 6 bytes object number (48 bits) * 4 bytes generation number (32 bits) * 6 bytes objset id (48 bits) * 4 bytes[**] currently just zero (32 bits) * * We reserve only 48 bits for the object number and objset id, as these are * the limits currently defined and imposed by the DMU. * * [*] 20 bytes on FreeBSD to fit into the size of struct fid. * [**] 2 bytes on FreeBSD for the above reason. */ typedef struct zfid_long { zfid_short_t z_fid; uint8_t zf_setid[6]; /* obj[i] = obj >> (8 * i) */ uint8_t zf_setgen[2]; /* gen[i] = gen >> (8 * i) */ } zfid_long_t; #define SHORT_FID_LEN (sizeof (zfid_short_t) - sizeof (uint16_t)) #define LONG_FID_LEN (sizeof (zfid_long_t) - sizeof (uint16_t)) extern uint_t zfs_fsyncer_key; extern int zfs_super_owner; extern void zfs_init(void); extern void zfs_fini(void); extern int zfs_suspend_fs(zfsvfs_t *zfsvfs); extern int zfs_resume_fs(zfsvfs_t *zfsvfs, struct dsl_dataset *ds); extern int zfs_end_fs(zfsvfs_t *zfsvfs, struct dsl_dataset *ds); extern int zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers); extern int zfsvfs_create(const char *name, boolean_t readonly, zfsvfs_t **zfvp); extern int zfsvfs_create_impl(zfsvfs_t **zfvp, zfsvfs_t *zfsvfs, objset_t *os); extern void zfsvfs_free(zfsvfs_t *zfsvfs); extern int zfs_check_global_label(const char *dsname, const char *hexsl); extern boolean_t zfs_is_readonly(zfsvfs_t *zfsvfs); extern int zfs_get_temporary_prop(struct dsl_dataset *ds, zfs_prop_t zfs_prop, uint64_t *val, char *setpoint); extern int zfs_busy(void); #ifdef __cplusplus } #endif #endif /* _SYS_FS_ZFS_VFSOPS_H */ diff --git a/include/os/linux/zfs/sys/zfs_vfsops_os.h b/include/os/linux/zfs/sys/zfs_vfsops_os.h index e320b8de4222..b4d5db21f5e5 100644 --- a/include/os/linux/zfs/sys/zfs_vfsops_os.h +++ b/include/os/linux/zfs/sys/zfs_vfsops_os.h @@ -1,256 +1,255 @@ /* * 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 https://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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2013, 2018 by Delphix. All rights reserved. */ #ifndef _SYS_FS_ZFS_VFSOPS_H #define _SYS_FS_ZFS_VFSOPS_H #include #include #include #include #include #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif typedef struct zfsvfs zfsvfs_t; struct znode; /* * This structure emulates the vfs_t from other platforms. It's purpose * is to facilitate the handling of mount options and minimize structural * differences between the platforms. */ typedef struct vfs { struct zfsvfs *vfs_data; char *vfs_mntpoint; /* Primary mount point */ uint64_t vfs_xattr; boolean_t vfs_readonly; boolean_t vfs_do_readonly; boolean_t vfs_setuid; boolean_t vfs_do_setuid; boolean_t vfs_exec; boolean_t vfs_do_exec; boolean_t vfs_devices; boolean_t vfs_do_devices; boolean_t vfs_do_xattr; boolean_t vfs_atime; boolean_t vfs_do_atime; boolean_t vfs_relatime; boolean_t vfs_do_relatime; boolean_t vfs_nbmand; boolean_t vfs_do_nbmand; } vfs_t; typedef struct zfs_mnt { const char *mnt_osname; /* Objset name */ char *mnt_data; /* Raw mount options */ } zfs_mnt_t; struct zfsvfs { vfs_t *z_vfs; /* generic fs struct */ struct super_block *z_sb; /* generic super_block */ struct zfsvfs *z_parent; /* parent fs */ objset_t *z_os; /* objset reference */ uint64_t z_flags; /* super_block flags */ uint64_t z_root; /* id of root znode */ uint64_t z_unlinkedobj; /* id of unlinked zapobj */ uint64_t z_max_blksz; /* maximum block size for files */ uint64_t z_fuid_obj; /* fuid table object number */ uint64_t z_fuid_size; /* fuid table size */ avl_tree_t z_fuid_idx; /* fuid tree keyed by index */ avl_tree_t z_fuid_domain; /* fuid tree keyed by domain */ krwlock_t z_fuid_lock; /* fuid lock */ boolean_t z_fuid_loaded; /* fuid tables are loaded */ boolean_t z_fuid_dirty; /* need to sync fuid table ? */ struct zfs_fuid_info *z_fuid_replay; /* fuid info for replay */ zilog_t *z_log; /* intent log pointer */ uint_t z_acl_mode; /* acl chmod/mode behavior */ uint_t z_acl_inherit; /* acl inheritance behavior */ uint_t z_acl_type; /* type of ACL usable on this FS */ zfs_case_t z_case; /* case-sense */ boolean_t z_utf8; /* utf8-only */ int z_norm; /* normalization flags */ boolean_t z_relatime; /* enable relatime mount option */ boolean_t z_unmounted; /* unmounted */ rrmlock_t z_teardown_lock; krwlock_t z_teardown_inactive_lock; list_t z_all_znodes; /* all znodes in the fs */ - uint64_t z_nr_znodes; /* number of znodes in the fs */ unsigned long z_rollback_time; /* last online rollback time */ unsigned long z_snap_defer_time; /* last snapshot unmount deferral */ kmutex_t z_znodes_lock; /* lock for z_all_znodes */ arc_prune_t *z_arc_prune; /* called by ARC to prune caches */ struct inode *z_ctldir; /* .zfs directory inode */ boolean_t z_show_ctldir; /* expose .zfs in the root dir */ boolean_t z_issnap; /* true if this is a snapshot */ boolean_t z_use_fuids; /* version allows fuids */ boolean_t z_replay; /* set during ZIL replay */ boolean_t z_use_sa; /* version allow system attributes */ boolean_t z_xattr_sa; /* allow xattrs to be stores as SA */ boolean_t z_draining; /* is true when drain is active */ boolean_t z_drain_cancel; /* signal the unlinked drain to stop */ uint64_t z_version; /* ZPL version */ uint64_t z_shares_dir; /* hidden shares dir */ dataset_kstats_t z_kstat; /* fs kstats */ kmutex_t z_lock; uint64_t z_userquota_obj; uint64_t z_groupquota_obj; uint64_t z_userobjquota_obj; uint64_t z_groupobjquota_obj; uint64_t z_projectquota_obj; uint64_t z_projectobjquota_obj; uint64_t z_replay_eof; /* New end of file - replay only */ sa_attr_type_t *z_attr_table; /* SA attr mapping->id */ uint64_t z_hold_size; /* znode hold array size */ avl_tree_t *z_hold_trees; /* znode hold trees */ kmutex_t *z_hold_locks; /* znode hold locks */ taskqid_t z_drain_task; /* task id for the unlink drain task */ }; #define ZFS_TEARDOWN_INIT(zfsvfs) \ rrm_init(&(zfsvfs)->z_teardown_lock, B_FALSE) #define ZFS_TEARDOWN_DESTROY(zfsvfs) \ rrm_destroy(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_ENTER_READ(zfsvfs, tag) \ rrm_enter_read(&(zfsvfs)->z_teardown_lock, tag); #define ZFS_TEARDOWN_EXIT_READ(zfsvfs, tag) \ rrm_exit(&(zfsvfs)->z_teardown_lock, tag) #define ZFS_TEARDOWN_ENTER_WRITE(zfsvfs, tag) \ rrm_enter(&(zfsvfs)->z_teardown_lock, RW_WRITER, tag) #define ZFS_TEARDOWN_EXIT_WRITE(zfsvfs) \ rrm_exit(&(zfsvfs)->z_teardown_lock, tag) #define ZFS_TEARDOWN_EXIT(zfsvfs, tag) \ rrm_exit(&(zfsvfs)->z_teardown_lock, tag) #define ZFS_TEARDOWN_READ_HELD(zfsvfs) \ RRM_READ_HELD(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_WRITE_HELD(zfsvfs) \ RRM_WRITE_HELD(&(zfsvfs)->z_teardown_lock) #define ZFS_TEARDOWN_HELD(zfsvfs) \ RRM_LOCK_HELD(&(zfsvfs)->z_teardown_lock) #define ZSB_XATTR 0x0001 /* Enable user xattrs */ /* * Allow a maximum number of links. While ZFS does not internally limit * this the inode->i_nlink member is defined as an unsigned int. To be * safe we use 2^31-1 as the limit. */ #define ZFS_LINK_MAX ((1U << 31) - 1U) /* * Normal filesystems (those not under .zfs/snapshot) have a total * file ID size limited to 12 bytes (including the length field) due to * NFSv2 protocol's limitation of 32 bytes for a filehandle. For historical * reasons, this same limit is being imposed by the Solaris NFSv3 implementation * (although the NFSv3 protocol actually permits a maximum of 64 bytes). It * is not possible to expand beyond 12 bytes without abandoning support * of NFSv2. * * For normal filesystems, we partition up the available space as follows: * 2 bytes fid length (required) * 6 bytes object number (48 bits) * 4 bytes generation number (32 bits) * * We reserve only 48 bits for the object number, as this is the limit * currently defined and imposed by the DMU. */ typedef struct zfid_short { uint16_t zf_len; uint8_t zf_object[6]; /* obj[i] = obj >> (8 * i) */ uint8_t zf_gen[4]; /* gen[i] = gen >> (8 * i) */ } zfid_short_t; /* * Filesystems under .zfs/snapshot have a total file ID size of 22 bytes * (including the length field). This makes files under .zfs/snapshot * accessible by NFSv3 and NFSv4, but not NFSv2. * * For files under .zfs/snapshot, we partition up the available space * as follows: * 2 bytes fid length (required) * 6 bytes object number (48 bits) * 4 bytes generation number (32 bits) * 6 bytes objset id (48 bits) * 4 bytes currently just zero (32 bits) * * We reserve only 48 bits for the object number and objset id, as these are * the limits currently defined and imposed by the DMU. */ typedef struct zfid_long { zfid_short_t z_fid; uint8_t zf_setid[6]; /* obj[i] = obj >> (8 * i) */ uint8_t zf_setgen[4]; /* gen[i] = gen >> (8 * i) */ } zfid_long_t; #define SHORT_FID_LEN (sizeof (zfid_short_t) - sizeof (uint16_t)) #define LONG_FID_LEN (sizeof (zfid_long_t) - sizeof (uint16_t)) extern void zfs_init(void); extern void zfs_fini(void); extern int zfs_suspend_fs(zfsvfs_t *zfsvfs); extern int zfs_resume_fs(zfsvfs_t *zfsvfs, struct dsl_dataset *ds); extern int zfs_end_fs(zfsvfs_t *zfsvfs, struct dsl_dataset *ds); extern void zfs_exit_fs(zfsvfs_t *zfsvfs); extern int zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers); extern int zfsvfs_create(const char *name, boolean_t readony, zfsvfs_t **zfvp); extern int zfsvfs_create_impl(zfsvfs_t **zfvp, zfsvfs_t *zfsvfs, objset_t *os); extern void zfsvfs_free(zfsvfs_t *zfsvfs); extern int zfs_check_global_label(const char *dsname, const char *hexsl); extern boolean_t zfs_is_readonly(zfsvfs_t *zfsvfs); extern int zfs_domount(struct super_block *sb, zfs_mnt_t *zm, int silent); extern void zfs_preumount(struct super_block *sb); extern int zfs_umount(struct super_block *sb); extern int zfs_remount(struct super_block *sb, int *flags, zfs_mnt_t *zm); extern int zfs_statvfs(struct inode *ip, struct kstatfs *statp); extern int zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp); extern int zfs_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects); extern int zfs_get_temporary_prop(dsl_dataset_t *ds, zfs_prop_t zfs_prop, uint64_t *val, char *setpoint); #ifdef __cplusplus } #endif #endif /* _SYS_FS_ZFS_VFSOPS_H */ diff --git a/module/os/freebsd/zfs/zfs_vfsops.c b/module/os/freebsd/zfs/zfs_vfsops.c index 33759fa26169..e8b9ada1316b 100644 --- a/module/os/freebsd/zfs/zfs_vfsops.c +++ b/module/os/freebsd/zfs/zfs_vfsops.c @@ -1,2517 +1,2515 @@ /* * 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 https://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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011 Pawel Jakub Dawidek . * All rights reserved. * Copyright (c) 2012, 2015 by Delphix. All rights reserved. * Copyright (c) 2014 Integros [integros.com] * Copyright 2016 Nexenta Systems, Inc. All rights reserved. */ /* Portions Copyright 2010 Robert Milkowski */ #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_comutil.h" #ifndef MNTK_VMSETSIZE_BUG #define MNTK_VMSETSIZE_BUG 0 #endif #ifndef MNTK_NOMSYNC #define MNTK_NOMSYNC 8 #endif struct mtx zfs_debug_mtx; MTX_SYSINIT(zfs_debug_mtx, &zfs_debug_mtx, "zfs_debug", MTX_DEF); SYSCTL_NODE(_vfs, OID_AUTO, zfs, CTLFLAG_RW, 0, "ZFS file system"); int zfs_super_owner; SYSCTL_INT(_vfs_zfs, OID_AUTO, super_owner, CTLFLAG_RW, &zfs_super_owner, 0, "File system owners can perform privileged operation on file systems"); int zfs_debug_level; SYSCTL_INT(_vfs_zfs, OID_AUTO, debug, CTLFLAG_RWTUN, &zfs_debug_level, 0, "Debug level"); struct zfs_jailparam { int mount_snapshot; }; static struct zfs_jailparam zfs_jailparam0 = { .mount_snapshot = 0, }; static int zfs_jailparam_slot; SYSCTL_JAIL_PARAM_SYS_NODE(zfs, CTLFLAG_RW, "Jail ZFS parameters"); SYSCTL_JAIL_PARAM(_zfs, mount_snapshot, CTLTYPE_INT | CTLFLAG_RW, "I", "Allow mounting snapshots in the .zfs directory for unjailed datasets"); SYSCTL_NODE(_vfs_zfs, OID_AUTO, version, CTLFLAG_RD, 0, "ZFS versions"); static int zfs_version_acl = ZFS_ACL_VERSION; SYSCTL_INT(_vfs_zfs_version, OID_AUTO, acl, CTLFLAG_RD, &zfs_version_acl, 0, "ZFS_ACL_VERSION"); static int zfs_version_spa = SPA_VERSION; SYSCTL_INT(_vfs_zfs_version, OID_AUTO, spa, CTLFLAG_RD, &zfs_version_spa, 0, "SPA_VERSION"); static int zfs_version_zpl = ZPL_VERSION; SYSCTL_INT(_vfs_zfs_version, OID_AUTO, zpl, CTLFLAG_RD, &zfs_version_zpl, 0, "ZPL_VERSION"); #if __FreeBSD_version >= 1400018 static int zfs_quotactl(vfs_t *vfsp, int cmds, uid_t id, void *arg, bool *mp_busy); #else static int zfs_quotactl(vfs_t *vfsp, int cmds, uid_t id, void *arg); #endif static int zfs_mount(vfs_t *vfsp); static int zfs_umount(vfs_t *vfsp, int fflag); static int zfs_root(vfs_t *vfsp, int flags, vnode_t **vpp); static int zfs_statfs(vfs_t *vfsp, struct statfs *statp); static int zfs_vget(vfs_t *vfsp, ino_t ino, int flags, vnode_t **vpp); static int zfs_sync(vfs_t *vfsp, int waitfor); #if __FreeBSD_version >= 1300098 static int zfs_checkexp(vfs_t *vfsp, struct sockaddr *nam, uint64_t *extflagsp, struct ucred **credanonp, int *numsecflavors, int *secflavors); #else static int zfs_checkexp(vfs_t *vfsp, struct sockaddr *nam, int *extflagsp, struct ucred **credanonp, int *numsecflavors, int **secflavors); #endif static int zfs_fhtovp(vfs_t *vfsp, fid_t *fidp, int flags, vnode_t **vpp); static void zfs_freevfs(vfs_t *vfsp); struct vfsops zfs_vfsops = { .vfs_mount = zfs_mount, .vfs_unmount = zfs_umount, #if __FreeBSD_version >= 1300049 .vfs_root = vfs_cache_root, .vfs_cachedroot = zfs_root, #else .vfs_root = zfs_root, #endif .vfs_statfs = zfs_statfs, .vfs_vget = zfs_vget, .vfs_sync = zfs_sync, .vfs_checkexp = zfs_checkexp, .vfs_fhtovp = zfs_fhtovp, .vfs_quotactl = zfs_quotactl, }; #ifdef VFCF_CROSS_COPY_FILE_RANGE VFS_SET(zfs_vfsops, zfs, VFCF_DELEGADMIN | VFCF_JAIL | VFCF_CROSS_COPY_FILE_RANGE); #else VFS_SET(zfs_vfsops, zfs, VFCF_DELEGADMIN | VFCF_JAIL); #endif /* * We need to keep a count of active fs's. * This is necessary to prevent our module * from being unloaded after a umount -f */ static uint32_t zfs_active_fs_count = 0; int zfs_get_temporary_prop(dsl_dataset_t *ds, zfs_prop_t zfs_prop, uint64_t *val, char *setpoint) { int error; zfsvfs_t *zfvp; vfs_t *vfsp; objset_t *os; uint64_t tmp = *val; error = dmu_objset_from_ds(ds, &os); if (error != 0) return (error); error = getzfsvfs_impl(os, &zfvp); if (error != 0) return (error); if (zfvp == NULL) return (ENOENT); vfsp = zfvp->z_vfs; switch (zfs_prop) { case ZFS_PROP_ATIME: if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL)) tmp = 0; if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL)) tmp = 1; break; case ZFS_PROP_DEVICES: if (vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL)) tmp = 0; if (vfs_optionisset(vfsp, MNTOPT_DEVICES, NULL)) tmp = 1; break; case ZFS_PROP_EXEC: if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) tmp = 0; if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) tmp = 1; break; case ZFS_PROP_SETUID: if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) tmp = 0; if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) tmp = 1; break; case ZFS_PROP_READONLY: if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) tmp = 0; if (vfs_optionisset(vfsp, MNTOPT_RO, NULL)) tmp = 1; break; case ZFS_PROP_XATTR: if (zfvp->z_flags & ZSB_XATTR) tmp = zfvp->z_xattr; break; case ZFS_PROP_NBMAND: if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL)) tmp = 0; if (vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL)) tmp = 1; break; default: vfs_unbusy(vfsp); return (ENOENT); } vfs_unbusy(vfsp); if (tmp != *val) { if (setpoint) (void) strcpy(setpoint, "temporary"); *val = tmp; } return (0); } static int zfs_getquota(zfsvfs_t *zfsvfs, uid_t id, int isgroup, struct dqblk64 *dqp) { int error = 0; char buf[32]; uint64_t usedobj, quotaobj; uint64_t quota, used = 0; timespec_t now; usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; if (quotaobj == 0 || zfsvfs->z_replay) { error = ENOENT; goto done; } (void) sprintf(buf, "%llx", (longlong_t)id); if ((error = zap_lookup(zfsvfs->z_os, quotaobj, buf, sizeof (quota), 1, "a)) != 0) { dprintf("%s(%d): quotaobj lookup failed\n", __FUNCTION__, __LINE__); goto done; } /* * quota(8) uses bsoftlimit as "quoota", and hardlimit as "limit". * So we set them to be the same. */ dqp->dqb_bsoftlimit = dqp->dqb_bhardlimit = btodb(quota); error = zap_lookup(zfsvfs->z_os, usedobj, buf, sizeof (used), 1, &used); if (error && error != ENOENT) { dprintf("%s(%d): usedobj failed; %d\n", __FUNCTION__, __LINE__, error); goto done; } dqp->dqb_curblocks = btodb(used); dqp->dqb_ihardlimit = dqp->dqb_isoftlimit = 0; vfs_timestamp(&now); /* * Setting this to 0 causes FreeBSD quota(8) to print * the number of days since the epoch, which isn't * particularly useful. */ dqp->dqb_btime = dqp->dqb_itime = now.tv_sec; done: return (error); } static int #if __FreeBSD_version >= 1400018 zfs_quotactl(vfs_t *vfsp, int cmds, uid_t id, void *arg, bool *mp_busy) #else zfs_quotactl(vfs_t *vfsp, int cmds, uid_t id, void *arg) #endif { zfsvfs_t *zfsvfs = vfsp->vfs_data; struct thread *td; int cmd, type, error = 0; int bitsize; zfs_userquota_prop_t quota_type; struct dqblk64 dqblk = { 0 }; td = curthread; cmd = cmds >> SUBCMDSHIFT; type = cmds & SUBCMDMASK; if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); if (id == -1) { switch (type) { case USRQUOTA: id = td->td_ucred->cr_ruid; break; case GRPQUOTA: id = td->td_ucred->cr_rgid; break; default: error = EINVAL; #if __FreeBSD_version < 1400018 if (cmd == Q_QUOTAON || cmd == Q_QUOTAOFF) vfs_unbusy(vfsp); #endif goto done; } } /* * Map BSD type to: * ZFS_PROP_USERUSED, * ZFS_PROP_USERQUOTA, * ZFS_PROP_GROUPUSED, * ZFS_PROP_GROUPQUOTA */ switch (cmd) { case Q_SETQUOTA: case Q_SETQUOTA32: if (type == USRQUOTA) quota_type = ZFS_PROP_USERQUOTA; else if (type == GRPQUOTA) quota_type = ZFS_PROP_GROUPQUOTA; else error = EINVAL; break; case Q_GETQUOTA: case Q_GETQUOTA32: if (type == USRQUOTA) quota_type = ZFS_PROP_USERUSED; else if (type == GRPQUOTA) quota_type = ZFS_PROP_GROUPUSED; else error = EINVAL; break; } /* * Depending on the cmd, we may need to get * the ruid and domain (see fuidstr_to_sid?), * the fuid (how?), or other information. * Create fuid using zfs_fuid_create(zfsvfs, id, * ZFS_OWNER or ZFS_GROUP, cr, &fuidp)? * I think I can use just the id? * * Look at zfs_id_overquota() to look up a quota. * zap_lookup(something, quotaobj, fuidstring, * sizeof (long long), 1, "a) * * See zfs_set_userquota() to set a quota. */ if ((uint32_t)type >= MAXQUOTAS) { error = EINVAL; goto done; } switch (cmd) { case Q_GETQUOTASIZE: bitsize = 64; error = copyout(&bitsize, arg, sizeof (int)); break; case Q_QUOTAON: // As far as I can tell, you can't turn quotas on or off on zfs error = 0; #if __FreeBSD_version < 1400018 vfs_unbusy(vfsp); #endif break; case Q_QUOTAOFF: error = ENOTSUP; #if __FreeBSD_version < 1400018 vfs_unbusy(vfsp); #endif break; case Q_SETQUOTA: error = copyin(arg, &dqblk, sizeof (dqblk)); if (error == 0) error = zfs_set_userquota(zfsvfs, quota_type, "", id, dbtob(dqblk.dqb_bhardlimit)); break; case Q_GETQUOTA: error = zfs_getquota(zfsvfs, id, type == GRPQUOTA, &dqblk); if (error == 0) error = copyout(&dqblk, arg, sizeof (dqblk)); break; default: error = EINVAL; break; } done: zfs_exit(zfsvfs, FTAG); return (error); } boolean_t zfs_is_readonly(zfsvfs_t *zfsvfs) { return (!!(zfsvfs->z_vfs->vfs_flag & VFS_RDONLY)); } static int zfs_sync(vfs_t *vfsp, int waitfor) { /* * Data integrity is job one. We don't want a compromised kernel * writing to the storage pool, so we never sync during panic. */ if (panicstr) return (0); /* * Ignore the system syncher. ZFS already commits async data * at zfs_txg_timeout intervals. */ if (waitfor == MNT_LAZY) return (0); if (vfsp != NULL) { /* * Sync a specific filesystem. */ zfsvfs_t *zfsvfs = vfsp->vfs_data; dsl_pool_t *dp; int error; if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); dp = dmu_objset_pool(zfsvfs->z_os); /* * If the system is shutting down, then skip any * filesystems which may exist on a suspended pool. */ if (rebooting && spa_suspended(dp->dp_spa)) { zfs_exit(zfsvfs, FTAG); return (0); } if (zfsvfs->z_log != NULL) zil_commit(zfsvfs->z_log, 0); zfs_exit(zfsvfs, FTAG); } else { /* * Sync all ZFS filesystems. This is what happens when you * run sync(8). Unlike other filesystems, ZFS honors the * request by waiting for all pools to commit all dirty data. */ spa_sync_allpools(); } return (0); } static void atime_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; if (newval == TRUE) { zfsvfs->z_atime = TRUE; zfsvfs->z_vfs->vfs_flag &= ~MNT_NOATIME; vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME); vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_ATIME, NULL, 0); } else { zfsvfs->z_atime = FALSE; zfsvfs->z_vfs->vfs_flag |= MNT_NOATIME; vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_ATIME); vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME, NULL, 0); } } static void xattr_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; if (newval == ZFS_XATTR_OFF) { zfsvfs->z_flags &= ~ZSB_XATTR; } else { zfsvfs->z_flags |= ZSB_XATTR; if (newval == ZFS_XATTR_SA) zfsvfs->z_xattr_sa = B_TRUE; else zfsvfs->z_xattr_sa = B_FALSE; } } static void blksz_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; ASSERT3U(newval, <=, spa_maxblocksize(dmu_objset_spa(zfsvfs->z_os))); ASSERT3U(newval, >=, SPA_MINBLOCKSIZE); ASSERT(ISP2(newval)); zfsvfs->z_max_blksz = newval; zfsvfs->z_vfs->mnt_stat.f_iosize = newval; } static void readonly_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; if (newval) { /* XXX locking on vfs_flag? */ zfsvfs->z_vfs->vfs_flag |= VFS_RDONLY; vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RW); vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RO, NULL, 0); } else { /* XXX locking on vfs_flag? */ zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY; vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RO); vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RW, NULL, 0); } } static void setuid_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; if (newval == FALSE) { zfsvfs->z_vfs->vfs_flag |= VFS_NOSETUID; vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_SETUID); vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID, NULL, 0); } else { zfsvfs->z_vfs->vfs_flag &= ~VFS_NOSETUID; vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID); vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_SETUID, NULL, 0); } } static void exec_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; if (newval == FALSE) { zfsvfs->z_vfs->vfs_flag |= VFS_NOEXEC; vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_EXEC); vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC, NULL, 0); } else { zfsvfs->z_vfs->vfs_flag &= ~VFS_NOEXEC; vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC); vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_EXEC, NULL, 0); } } /* * The nbmand mount option can be changed at mount time. * We can't allow it to be toggled on live file systems or incorrect * behavior may be seen from cifs clients * * This property isn't registered via dsl_prop_register(), but this callback * will be called when a file system is first mounted */ static void nbmand_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; if (newval == FALSE) { vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND); vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND, NULL, 0); } else { vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND); vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND, NULL, 0); } } static void snapdir_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; zfsvfs->z_show_ctldir = newval; } static void acl_mode_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; zfsvfs->z_acl_mode = newval; } static void acl_inherit_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; zfsvfs->z_acl_inherit = newval; } static void acl_type_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; zfsvfs->z_acl_type = newval; } static int zfs_register_callbacks(vfs_t *vfsp) { struct dsl_dataset *ds = NULL; objset_t *os = NULL; zfsvfs_t *zfsvfs = NULL; uint64_t nbmand; boolean_t readonly = B_FALSE; boolean_t do_readonly = B_FALSE; boolean_t setuid = B_FALSE; boolean_t do_setuid = B_FALSE; boolean_t exec = B_FALSE; boolean_t do_exec = B_FALSE; boolean_t xattr = B_FALSE; boolean_t atime = B_FALSE; boolean_t do_atime = B_FALSE; boolean_t do_xattr = B_FALSE; int error = 0; ASSERT3P(vfsp, !=, NULL); zfsvfs = vfsp->vfs_data; ASSERT3P(zfsvfs, !=, NULL); os = zfsvfs->z_os; /* * This function can be called for a snapshot when we update snapshot's * mount point, which isn't really supported. */ if (dmu_objset_is_snapshot(os)) return (EOPNOTSUPP); /* * The act of registering our callbacks will destroy any mount * options we may have. In order to enable temporary overrides * of mount options, we stash away the current values and * restore them after we register the callbacks. */ if (vfs_optionisset(vfsp, MNTOPT_RO, NULL) || !spa_writeable(dmu_objset_spa(os))) { readonly = B_TRUE; do_readonly = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) { readonly = B_FALSE; do_readonly = B_TRUE; } if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) { setuid = B_FALSE; do_setuid = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) { setuid = B_TRUE; do_setuid = B_TRUE; } if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) { exec = B_FALSE; do_exec = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) { exec = B_TRUE; do_exec = B_TRUE; } if (vfs_optionisset(vfsp, MNTOPT_NOXATTR, NULL)) { zfsvfs->z_xattr = xattr = ZFS_XATTR_OFF; do_xattr = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_XATTR, NULL)) { zfsvfs->z_xattr = xattr = ZFS_XATTR_DIR; do_xattr = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_DIRXATTR, NULL)) { zfsvfs->z_xattr = xattr = ZFS_XATTR_DIR; do_xattr = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_SAXATTR, NULL)) { zfsvfs->z_xattr = xattr = ZFS_XATTR_SA; do_xattr = B_TRUE; } if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL)) { atime = B_FALSE; do_atime = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL)) { atime = B_TRUE; do_atime = B_TRUE; } /* * We need to enter pool configuration here, so that we can use * dsl_prop_get_int_ds() to handle the special nbmand property below. * dsl_prop_get_integer() can not be used, because it has to acquire * spa_namespace_lock and we can not do that because we already hold * z_teardown_lock. The problem is that spa_write_cachefile() is called * with spa_namespace_lock held and the function calls ZFS vnode * operations to write the cache file and thus z_teardown_lock is * acquired after spa_namespace_lock. */ ds = dmu_objset_ds(os); dsl_pool_config_enter(dmu_objset_pool(os), FTAG); /* * nbmand is a special property. It can only be changed at * mount time. * * This is weird, but it is documented to only be changeable * at mount time. */ if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL)) { nbmand = B_FALSE; } else if (vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL)) { nbmand = B_TRUE; } else if ((error = dsl_prop_get_int_ds(ds, "nbmand", &nbmand)) != 0) { dsl_pool_config_exit(dmu_objset_pool(os), FTAG); return (error); } /* * 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... */ error = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ACLTYPE), acl_type_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ACLMODE), acl_mode_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zfsvfs); dsl_pool_config_exit(dmu_objset_pool(os), FTAG); if (error) goto unregister; /* * Invoke our callbacks to restore temporary mount options. */ if (do_readonly) readonly_changed_cb(zfsvfs, readonly); if (do_setuid) setuid_changed_cb(zfsvfs, setuid); if (do_exec) exec_changed_cb(zfsvfs, exec); if (do_xattr) xattr_changed_cb(zfsvfs, xattr); if (do_atime) atime_changed_cb(zfsvfs, atime); nbmand_changed_cb(zfsvfs, nbmand); return (0); unregister: dsl_prop_unregister_all(ds, zfsvfs); return (error); } /* * Associate this zfsvfs with the given objset, which must be owned. * This will cache a bunch of on-disk state from the objset in the * zfsvfs. */ static int zfsvfs_init(zfsvfs_t *zfsvfs, objset_t *os) { int error; uint64_t val; zfsvfs->z_max_blksz = SPA_OLD_MAXBLOCKSIZE; zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; zfsvfs->z_os = os; error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version); if (error != 0) return (error); if (zfsvfs->z_version > zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { (void) printf("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)zfsvfs->z_version, (u_longlong_t)spa_version(dmu_objset_spa(os))); return (SET_ERROR(ENOTSUP)); } error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &val); if (error != 0) return (error); zfsvfs->z_norm = (int)val; error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &val); if (error != 0) return (error); zfsvfs->z_utf8 = (val != 0); error = zfs_get_zplprop(os, ZFS_PROP_CASE, &val); if (error != 0) return (error); zfsvfs->z_case = (uint_t)val; error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &val); if (error != 0) return (error); zfsvfs->z_acl_type = (uint_t)val; /* * Fold case on file systems that are always or sometimes case * insensitive. */ if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE || zfsvfs->z_case == ZFS_CASE_MIXED) zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); uint64_t sa_obj = 0; if (zfsvfs->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 != 0) return (error); error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &val); if (error == 0 && val == ZFS_XATTR_SA) zfsvfs->z_xattr_sa = B_TRUE; } error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, &zfsvfs->z_attr_table); if (error != 0) return (error); if (zfsvfs->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, &zfsvfs->z_root); if (error != 0) return (error); ASSERT3U(zfsvfs->z_root, !=, 0); error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, &zfsvfs->z_unlinkedobj); if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], 8, 1, &zfsvfs->z_userquota_obj); if (error == ENOENT) zfsvfs->z_userquota_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], 8, 1, &zfsvfs->z_groupquota_obj); if (error == ENOENT) zfsvfs->z_groupquota_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA], 8, 1, &zfsvfs->z_projectquota_obj); if (error == ENOENT) zfsvfs->z_projectquota_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA], 8, 1, &zfsvfs->z_userobjquota_obj); if (error == ENOENT) zfsvfs->z_userobjquota_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA], 8, 1, &zfsvfs->z_groupobjquota_obj); if (error == ENOENT) zfsvfs->z_groupobjquota_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTOBJQUOTA], 8, 1, &zfsvfs->z_projectobjquota_obj); if (error == ENOENT) zfsvfs->z_projectobjquota_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, &zfsvfs->z_fuid_obj); if (error == ENOENT) zfsvfs->z_fuid_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, &zfsvfs->z_shares_dir); if (error == ENOENT) zfsvfs->z_shares_dir = 0; else if (error != 0) return (error); /* * Only use the name cache if we are looking for a * 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 (which is always the case on * FreeBSD). */ zfsvfs->z_use_namecache = !zfsvfs->z_norm || ((zfsvfs->z_case == ZFS_CASE_MIXED) && !(zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER)); return (0); } taskq_t *zfsvfs_taskq; static void zfsvfs_task_unlinked_drain(void *context, int pending __unused) { zfs_unlinked_drain((zfsvfs_t *)context); } int zfsvfs_create(const char *osname, boolean_t readonly, zfsvfs_t **zfvp) { objset_t *os; zfsvfs_t *zfsvfs; int error; boolean_t ro = (readonly || (strchr(osname, '@') != NULL)); /* * XXX: Fix struct statfs so this isn't necessary! * * The 'osname' is used as the filesystem's special node, which means * it must fit in statfs.f_mntfromname, or else it can't be * enumerated, so libzfs_mnttab_find() returns NULL, which causes * 'zfs unmount' to think it's not mounted when it is. */ if (strlen(osname) >= MNAMELEN) return (SET_ERROR(ENAMETOOLONG)); zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); error = dmu_objset_own(osname, DMU_OST_ZFS, ro, B_TRUE, zfsvfs, &os); if (error != 0) { kmem_free(zfsvfs, sizeof (zfsvfs_t)); return (error); } error = zfsvfs_create_impl(zfvp, zfsvfs, os); return (error); } int zfsvfs_create_impl(zfsvfs_t **zfvp, zfsvfs_t *zfsvfs, objset_t *os) { int error; zfsvfs->z_vfs = NULL; zfsvfs->z_parent = zfsvfs; mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL); list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), offsetof(znode_t, z_link_node)); TASK_INIT(&zfsvfs->z_unlinked_drain_task, 0, zfsvfs_task_unlinked_drain, zfsvfs); ZFS_TEARDOWN_INIT(zfsvfs); ZFS_TEARDOWN_INACTIVE_INIT(zfsvfs); rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL); for (int i = 0; i != ZFS_OBJ_MTX_SZ; i++) mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); error = zfsvfs_init(zfsvfs, os); if (error != 0) { dmu_objset_disown(os, B_TRUE, zfsvfs); *zfvp = NULL; kmem_free(zfsvfs, sizeof (zfsvfs_t)); return (error); } *zfvp = zfsvfs; return (0); } static int zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) { int error; /* * Check for a bad on-disk format version now since we * lied about owning the dataset readonly before. */ if (!(zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) && dmu_objset_incompatible_encryption_version(zfsvfs->z_os)) return (SET_ERROR(EROFS)); error = zfs_register_callbacks(zfsvfs->z_vfs); if (error) return (error); /* * 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; ASSERT3P(zfsvfs->z_kstat.dk_kstats, ==, NULL); error = dataset_kstats_create(&zfsvfs->z_kstat, zfsvfs->z_os); if (error) return (error); zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data, &zfsvfs->z_kstat.dk_zil_sums); /* * During replay we remove the read only flag to * allow replays to succeed. */ readonly = zfsvfs->z_vfs->vfs_flag & VFS_RDONLY; if (readonly != 0) { zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY; } else { dsl_dir_t *dd; zap_stats_t zs; if (zap_get_stats(zfsvfs->z_os, zfsvfs->z_unlinkedobj, &zs) == 0) { dataset_kstats_update_nunlinks_kstat( &zfsvfs->z_kstat, zs.zs_num_entries); dprintf_ds(zfsvfs->z_os->os_dsl_dataset, "num_entries in unlinked set: %llu", (u_longlong_t)zs.zs_num_entries); } zfs_unlinked_drain(zfsvfs); dd = zfsvfs->z_os->os_dsl_dataset->ds_dir; dd->dd_activity_cancelled = B_FALSE; } /* * 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(zfsvfs->z_os))) { if (zil_replay_disable) { zil_destroy(zfsvfs->z_log, B_FALSE); } else { boolean_t use_nc = zfsvfs->z_use_namecache; zfsvfs->z_use_namecache = B_FALSE; zfsvfs->z_replay = B_TRUE; zil_replay(zfsvfs->z_os, zfsvfs, zfs_replay_vector); zfsvfs->z_replay = B_FALSE; zfsvfs->z_use_namecache = use_nc; } } /* restore readonly bit */ if (readonly != 0) zfsvfs->z_vfs->vfs_flag |= VFS_RDONLY; } else { ASSERT3P(zfsvfs->z_kstat.dk_kstats, !=, NULL); zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data, &zfsvfs->z_kstat.dk_zil_sums); } /* * Set the objset user_ptr to track its zfsvfs. */ mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); dmu_objset_set_user(zfsvfs->z_os, zfsvfs); mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); return (0); } void zfsvfs_free(zfsvfs_t *zfsvfs) { int i; zfs_fuid_destroy(zfsvfs); mutex_destroy(&zfsvfs->z_znodes_lock); mutex_destroy(&zfsvfs->z_lock); - ASSERT3U(zfsvfs->z_nr_znodes, ==, 0); list_destroy(&zfsvfs->z_all_znodes); ZFS_TEARDOWN_DESTROY(zfsvfs); ZFS_TEARDOWN_INACTIVE_DESTROY(zfsvfs); rw_destroy(&zfsvfs->z_fuid_lock); for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) mutex_destroy(&zfsvfs->z_hold_mtx[i]); dataset_kstats_destroy(&zfsvfs->z_kstat); kmem_free(zfsvfs, sizeof (zfsvfs_t)); } static void zfs_set_fuid_feature(zfsvfs_t *zfsvfs) { zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); } static int zfs_domount(vfs_t *vfsp, char *osname) { uint64_t recordsize, fsid_guid; int error = 0; zfsvfs_t *zfsvfs; ASSERT3P(vfsp, !=, NULL); ASSERT3P(osname, !=, NULL); error = zfsvfs_create(osname, vfsp->mnt_flag & MNT_RDONLY, &zfsvfs); if (error) return (error); zfsvfs->z_vfs = vfsp; if ((error = dsl_prop_get_integer(osname, "recordsize", &recordsize, NULL))) goto out; zfsvfs->z_vfs->vfs_bsize = SPA_MINBLOCKSIZE; zfsvfs->z_vfs->mnt_stat.f_iosize = recordsize; vfsp->vfs_data = zfsvfs; vfsp->mnt_flag |= MNT_LOCAL; vfsp->mnt_kern_flag |= MNTK_LOOKUP_SHARED; vfsp->mnt_kern_flag |= MNTK_SHARED_WRITES; vfsp->mnt_kern_flag |= MNTK_EXTENDED_SHARED; /* * This can cause a loss of coherence between ARC and page cache * on ZoF - unclear if the problem is in FreeBSD or ZoF */ vfsp->mnt_kern_flag |= MNTK_NO_IOPF; /* vn_io_fault can be used */ vfsp->mnt_kern_flag |= MNTK_NOMSYNC; vfsp->mnt_kern_flag |= MNTK_VMSETSIZE_BUG; #if defined(_KERNEL) && !defined(KMEM_DEBUG) vfsp->mnt_kern_flag |= MNTK_FPLOOKUP; #endif /* * The fsid is 64 bits, composed of an 8-bit fs type, which * separates our fsid from any other filesystem types, and a * 56-bit objset unique ID. The objset unique ID is unique to * all objsets open on this system, provided by unique_create(). * The 8-bit fs type must be put in the low bits of fsid[1] * because that's where other Solaris filesystems put it. */ fsid_guid = dmu_objset_fsid_guid(zfsvfs->z_os); ASSERT3U((fsid_guid & ~((1ULL << 56) - 1)), ==, 0); vfsp->vfs_fsid.val[0] = fsid_guid; vfsp->vfs_fsid.val[1] = ((fsid_guid >> 32) << 8) | (vfsp->mnt_vfc->vfc_typenum & 0xFF); /* * Set features for file system. */ zfs_set_fuid_feature(zfsvfs); if (dmu_objset_is_snapshot(zfsvfs->z_os)) { uint64_t pval; atime_changed_cb(zfsvfs, B_FALSE); readonly_changed_cb(zfsvfs, B_TRUE); if ((error = dsl_prop_get_integer(osname, "xattr", &pval, NULL))) goto out; xattr_changed_cb(zfsvfs, pval); if ((error = dsl_prop_get_integer(osname, "acltype", &pval, NULL))) goto out; acl_type_changed_cb(zfsvfs, pval); zfsvfs->z_issnap = B_TRUE; zfsvfs->z_os->os_sync = ZFS_SYNC_DISABLED; mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); dmu_objset_set_user(zfsvfs->z_os, zfsvfs); mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); } else { if ((error = zfsvfs_setup(zfsvfs, B_TRUE))) goto out; } vfs_mountedfrom(vfsp, osname); if (!zfsvfs->z_issnap) zfsctl_create(zfsvfs); out: if (error) { dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs); zfsvfs_free(zfsvfs); } else { atomic_inc_32(&zfs_active_fs_count); } return (error); } static void zfs_unregister_callbacks(zfsvfs_t *zfsvfs) { objset_t *os = zfsvfs->z_os; if (!dmu_objset_is_snapshot(os)) dsl_prop_unregister_all(dmu_objset_ds(os), zfsvfs); } static int getpoolname(const char *osname, char *poolname) { char *p; p = strchr(osname, '/'); if (p == NULL) { if (strlen(osname) >= MAXNAMELEN) return (ENAMETOOLONG); (void) strcpy(poolname, osname); } else { if (p - osname >= MAXNAMELEN) return (ENAMETOOLONG); (void) strlcpy(poolname, osname, p - osname + 1); } return (0); } static void fetch_osname_options(char *name, bool *checkpointrewind) { if (name[0] == '!') { *checkpointrewind = true; memmove(name, name + 1, strlen(name)); } else { *checkpointrewind = false; } } static int zfs_mount(vfs_t *vfsp) { kthread_t *td = curthread; vnode_t *mvp = vfsp->mnt_vnodecovered; cred_t *cr = td->td_ucred; char *osname; int error = 0; int canwrite; bool checkpointrewind, isctlsnap = false; if (vfs_getopt(vfsp->mnt_optnew, "from", (void **)&osname, NULL)) return (SET_ERROR(EINVAL)); /* * If full-owner-access is enabled and delegated administration is * turned on, we must set nosuid. */ if (zfs_super_owner && dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr) != ECANCELED) { secpolicy_fs_mount_clearopts(cr, vfsp); } fetch_osname_options(osname, &checkpointrewind); isctlsnap = (mvp != NULL && zfsctl_is_node(mvp) && strchr(osname, '@') != NULL); /* * Check for mount privilege? * * If we don't have privilege then see if * we have local permission to allow it */ error = secpolicy_fs_mount(cr, mvp, vfsp); if (error && isctlsnap) { secpolicy_fs_mount_clearopts(cr, vfsp); } else if (error) { if (dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr) != 0) goto out; if (!(vfsp->vfs_flag & MS_REMOUNT)) { vattr_t vattr; /* * Make sure user is the owner of the mount point * or has sufficient privileges. */ vattr.va_mask = AT_UID; vn_lock(mvp, LK_SHARED | LK_RETRY); if (VOP_GETATTR(mvp, &vattr, cr)) { VOP_UNLOCK1(mvp); goto out; } if (secpolicy_vnode_owner(mvp, cr, vattr.va_uid) != 0 && VOP_ACCESS(mvp, VWRITE, cr, td) != 0) { VOP_UNLOCK1(mvp); goto out; } VOP_UNLOCK1(mvp); } secpolicy_fs_mount_clearopts(cr, vfsp); } /* * Refuse to mount a filesystem if we are in a local zone and the * dataset is not visible. */ if (!INGLOBALZONE(curproc) && (!zone_dataset_visible(osname, &canwrite) || !canwrite)) { boolean_t mount_snapshot = B_FALSE; /* * Snapshots may be mounted in .zfs for unjailed datasets * if allowed by the jail param zfs.mount_snapshot. */ if (isctlsnap) { struct prison *pr; struct zfs_jailparam *zjp; pr = curthread->td_ucred->cr_prison; mtx_lock(&pr->pr_mtx); zjp = osd_jail_get(pr, zfs_jailparam_slot); mtx_unlock(&pr->pr_mtx); if (zjp && zjp->mount_snapshot) mount_snapshot = B_TRUE; } if (!mount_snapshot) { error = SET_ERROR(EPERM); goto out; } } vfsp->vfs_flag |= MNT_NFS4ACLS; /* * When doing a remount, we simply refresh our temporary properties * according to those options set in the current VFS options. */ if (vfsp->vfs_flag & MS_REMOUNT) { zfsvfs_t *zfsvfs = vfsp->vfs_data; /* * Refresh mount options with z_teardown_lock blocking I/O while * the filesystem is in an inconsistent state. * The lock also serializes this code with filesystem * manipulations between entry to zfs_suspend_fs() and return * from zfs_resume_fs(). */ ZFS_TEARDOWN_ENTER_WRITE(zfsvfs, FTAG); zfs_unregister_callbacks(zfsvfs); error = zfs_register_callbacks(vfsp); ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); goto out; } /* Initial root mount: try hard to import the requested root pool. */ if ((vfsp->vfs_flag & MNT_ROOTFS) != 0 && (vfsp->vfs_flag & MNT_UPDATE) == 0) { char pname[MAXNAMELEN]; error = getpoolname(osname, pname); if (error == 0) error = spa_import_rootpool(pname, checkpointrewind); if (error) goto out; } DROP_GIANT(); error = zfs_domount(vfsp, osname); PICKUP_GIANT(); out: return (error); } static int zfs_statfs(vfs_t *vfsp, struct statfs *statp) { zfsvfs_t *zfsvfs = vfsp->vfs_data; uint64_t refdbytes, availbytes, usedobjs, availobjs; int error; statp->f_version = STATFS_VERSION; if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); dmu_objset_space(zfsvfs->z_os, &refdbytes, &availbytes, &usedobjs, &availobjs); /* * The underlying storage pool actually uses multiple block sizes. * We report the fragsize as the smallest block size we support, * and we report our blocksize as the filesystem's maximum blocksize. */ statp->f_bsize = SPA_MINBLOCKSIZE; statp->f_iosize = zfsvfs->z_vfs->mnt_stat.f_iosize; /* * The following report "total" blocks of various kinds in the * file system, but reported in terms of f_frsize - the * "fragment" size. */ statp->f_blocks = (refdbytes + availbytes) >> SPA_MINBLOCKSHIFT; statp->f_bfree = availbytes / statp->f_bsize; 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, statp->f_bfree); statp->f_files = statp->f_ffree + usedobjs; /* * We're a zfs filesystem. */ strlcpy(statp->f_fstypename, "zfs", sizeof (statp->f_fstypename)); strlcpy(statp->f_mntfromname, vfsp->mnt_stat.f_mntfromname, sizeof (statp->f_mntfromname)); strlcpy(statp->f_mntonname, vfsp->mnt_stat.f_mntonname, sizeof (statp->f_mntonname)); statp->f_namemax = MAXNAMELEN - 1; zfs_exit(zfsvfs, FTAG); return (0); } static int zfs_root(vfs_t *vfsp, int flags, vnode_t **vpp) { zfsvfs_t *zfsvfs = vfsp->vfs_data; znode_t *rootzp; int error; if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp); if (error == 0) *vpp = ZTOV(rootzp); zfs_exit(zfsvfs, FTAG); if (error == 0) { error = vn_lock(*vpp, flags); if (error != 0) { VN_RELE(*vpp); *vpp = NULL; } } return (error); } /* * Teardown the zfsvfs::z_os. * * Note, if 'unmounting' is FALSE, we return with the 'z_teardown_lock' * and 'z_teardown_inactive_lock' held. */ static int zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting) { znode_t *zp; dsl_dir_t *dd; /* * If someone has not already unmounted this file system, * drain the zrele_taskq to ensure all active references to the * zfsvfs_t have been handled only then can it be safely destroyed. */ if (zfsvfs->z_os) { /* * If we're unmounting we have to wait for the list to * drain completely. * * If we're not unmounting there's no guarantee the list * will drain completely, but zreles run from the taskq * may add the parents of dir-based xattrs to the taskq * so we want to wait for these. * - * We can safely read z_nr_znodes without locking because the - * VFS has already blocked operations which add to the - * z_all_znodes list and thus increment z_nr_znodes. + * We can safely check z_all_znodes for being empty because the + * VFS has already blocked operations which add to it. */ int round = 0; - while (zfsvfs->z_nr_znodes > 0) { + while (!list_is_empty(&zfsvfs->z_all_znodes)) { taskq_wait_outstanding(dsl_pool_zrele_taskq( dmu_objset_pool(zfsvfs->z_os)), 0); if (++round > 1 && !unmounting) break; } } ZFS_TEARDOWN_ENTER_WRITE(zfsvfs, FTAG); if (!unmounting) { /* * We purge the parent filesystem's vfsp as the parent * filesystem and all of its snapshots have their vnode's * v_vfsp set to the parent's filesystem's vfsp. Note, * 'z_parent' is self referential for non-snapshots. */ #ifdef FREEBSD_NAMECACHE #if __FreeBSD_version >= 1300117 cache_purgevfs(zfsvfs->z_parent->z_vfs); #else cache_purgevfs(zfsvfs->z_parent->z_vfs, true); #endif #endif } /* * Close the zil. NB: Can't close the zil while zfs_inactive * threads are blocked as zil_close can call zfs_inactive. */ if (zfsvfs->z_log) { zil_close(zfsvfs->z_log); zfsvfs->z_log = NULL; } ZFS_TEARDOWN_INACTIVE_ENTER_WRITE(zfsvfs); /* * 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 && (zfsvfs->z_unmounted || zfsvfs->z_os == NULL)) { ZFS_TEARDOWN_INACTIVE_EXIT_WRITE(zfsvfs); ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); return (SET_ERROR(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 * relevant for forced unmount). * * Release all holds on dbufs. */ mutex_enter(&zfsvfs->z_znodes_lock); for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL; zp = list_next(&zfsvfs->z_all_znodes, zp)) { if (zp->z_sa_hdl != NULL) { zfs_znode_dmu_fini(zp); } } mutex_exit(&zfsvfs->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) { zfsvfs->z_unmounted = B_TRUE; ZFS_TEARDOWN_INACTIVE_EXIT_WRITE(zfsvfs); ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); } /* * z_os will be NULL if there was an error in attempting to reopen * zfsvfs, so just return as the properties had already been * unregistered and cached data had been evicted before. */ if (zfsvfs->z_os == NULL) return (0); /* * Unregister properties. */ zfs_unregister_callbacks(zfsvfs); /* * Evict cached data */ if (!zfs_is_readonly(zfsvfs)) txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); dmu_objset_evict_dbufs(zfsvfs->z_os); dd = zfsvfs->z_os->os_dsl_dataset->ds_dir; dsl_dir_cancel_waiters(dd); return (0); } static int zfs_umount(vfs_t *vfsp, int fflag) { kthread_t *td = curthread; zfsvfs_t *zfsvfs = vfsp->vfs_data; objset_t *os; cred_t *cr = td->td_ucred; int ret; ret = secpolicy_fs_unmount(cr, vfsp); if (ret) { if (dsl_deleg_access((char *)vfsp->vfs_resource, ZFS_DELEG_PERM_MOUNT, cr)) return (ret); } /* * Unmount any snapshots mounted under .zfs before unmounting the * dataset itself. */ if (zfsvfs->z_ctldir != NULL) { if ((ret = zfsctl_umount_snapshots(vfsp, fflag, cr)) != 0) return (ret); } if (fflag & MS_FORCE) { /* * Mark file system as unmounted before calling * vflush(FORCECLOSE). This way we ensure no future vnops * will be called and risk operating on DOOMED vnodes. */ ZFS_TEARDOWN_ENTER_WRITE(zfsvfs, FTAG); zfsvfs->z_unmounted = B_TRUE; ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); } /* * Flush all the files. */ ret = vflush(vfsp, 0, (fflag & MS_FORCE) ? FORCECLOSE : 0, td); if (ret != 0) return (ret); while (taskqueue_cancel(zfsvfs_taskq->tq_queue, &zfsvfs->z_unlinked_drain_task, NULL) != 0) taskqueue_drain(zfsvfs_taskq->tq_queue, &zfsvfs->z_unlinked_drain_task); VERIFY0(zfsvfs_teardown(zfsvfs, B_TRUE)); os = zfsvfs->z_os; /* * z_os will be NULL if there was an error in * attempting to reopen zfsvfs. */ 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, B_TRUE, zfsvfs); } /* * We can now safely destroy the '.zfs' directory node. */ if (zfsvfs->z_ctldir != NULL) zfsctl_destroy(zfsvfs); zfs_freevfs(vfsp); return (0); } static int zfs_vget(vfs_t *vfsp, ino_t ino, int flags, vnode_t **vpp) { zfsvfs_t *zfsvfs = vfsp->vfs_data; znode_t *zp; int err; /* * zfs_zget() can't operate on virtual entries like .zfs/ or * .zfs/snapshot/ directories, that's why we return EOPNOTSUPP. * This will make NFS to switch to LOOKUP instead of using VGET. */ if (ino == ZFSCTL_INO_ROOT || ino == ZFSCTL_INO_SNAPDIR || (zfsvfs->z_shares_dir != 0 && ino == zfsvfs->z_shares_dir)) return (EOPNOTSUPP); if ((err = zfs_enter(zfsvfs, FTAG)) != 0) return (err); err = zfs_zget(zfsvfs, ino, &zp); if (err == 0 && zp->z_unlinked) { vrele(ZTOV(zp)); err = EINVAL; } if (err == 0) *vpp = ZTOV(zp); zfs_exit(zfsvfs, FTAG); if (err == 0) { err = vn_lock(*vpp, flags); if (err != 0) vrele(*vpp); } if (err != 0) *vpp = NULL; return (err); } static int #if __FreeBSD_version >= 1300098 zfs_checkexp(vfs_t *vfsp, struct sockaddr *nam, uint64_t *extflagsp, struct ucred **credanonp, int *numsecflavors, int *secflavors) #else zfs_checkexp(vfs_t *vfsp, struct sockaddr *nam, int *extflagsp, struct ucred **credanonp, int *numsecflavors, int **secflavors) #endif { zfsvfs_t *zfsvfs = vfsp->vfs_data; /* * If this is regular file system vfsp is the same as * zfsvfs->z_parent->z_vfs, but if it is snapshot, * zfsvfs->z_parent->z_vfs represents parent file system * which we have to use here, because only this file system * has mnt_export configured. */ return (vfs_stdcheckexp(zfsvfs->z_parent->z_vfs, nam, extflagsp, credanonp, numsecflavors, secflavors)); } _Static_assert(sizeof (struct fid) >= SHORT_FID_LEN, "struct fid bigger than SHORT_FID_LEN"); _Static_assert(sizeof (struct fid) >= LONG_FID_LEN, "struct fid bigger than LONG_FID_LEN"); static int zfs_fhtovp(vfs_t *vfsp, fid_t *fidp, int flags, vnode_t **vpp) { struct componentname cn; zfsvfs_t *zfsvfs = vfsp->vfs_data; znode_t *zp; vnode_t *dvp; uint64_t object = 0; uint64_t fid_gen = 0; uint64_t setgen = 0; uint64_t gen_mask; uint64_t zp_gen; int i, err; *vpp = NULL; if ((err = zfs_enter(zfsvfs, FTAG)) != 0) return (err); /* * On FreeBSD we can get snapshot's mount point or its parent file * system mount point depending if snapshot is already mounted or not. */ if (zfsvfs->z_parent == zfsvfs && fidp->fid_len == LONG_FID_LEN) { zfid_long_t *zlfid = (zfid_long_t *)fidp; uint64_t objsetid = 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(zfsvfs, FTAG); err = zfsctl_lookup_objset(vfsp, objsetid, &zfsvfs); if (err) return (SET_ERROR(EINVAL)); if ((err = zfs_enter(zfsvfs, FTAG)) != 0) return (err); } 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(zfsvfs, FTAG); return (SET_ERROR(EINVAL)); } if (fidp->fid_len == LONG_FID_LEN && setgen != 0) { zfs_exit(zfsvfs, FTAG); dprintf("snapdir fid: fid_gen (%llu) and setgen (%llu)\n", (u_longlong_t)fid_gen, (u_longlong_t)setgen); return (SET_ERROR(EINVAL)); } /* * A zero fid_gen means we are in .zfs or the .zfs/snapshot * directory tree. If the object == zfsvfs->z_shares_dir, then * we are in the .zfs/shares directory tree. */ if ((fid_gen == 0 && (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) || (zfsvfs->z_shares_dir != 0 && object == zfsvfs->z_shares_dir)) { zfs_exit(zfsvfs, FTAG); VERIFY0(zfsctl_root(zfsvfs, LK_SHARED, &dvp)); if (object == ZFSCTL_INO_SNAPDIR) { cn.cn_nameptr = "snapshot"; cn.cn_namelen = strlen(cn.cn_nameptr); cn.cn_nameiop = LOOKUP; cn.cn_flags = ISLASTCN | LOCKLEAF; cn.cn_lkflags = flags; VERIFY0(VOP_LOOKUP(dvp, vpp, &cn)); vput(dvp); } else if (object == zfsvfs->z_shares_dir) { /* * XXX This branch must not be taken, * if it is, then the lookup below will * explode. */ cn.cn_nameptr = "shares"; cn.cn_namelen = strlen(cn.cn_nameptr); cn.cn_nameiop = LOOKUP; cn.cn_flags = ISLASTCN; cn.cn_lkflags = flags; VERIFY0(VOP_LOOKUP(dvp, vpp, &cn)); vput(dvp); } else { *vpp = dvp; } return (err); } gen_mask = -1ULL >> (64 - 8 * i); dprintf("getting %llu [%llu mask %llx]\n", (u_longlong_t)object, (u_longlong_t)fid_gen, (u_longlong_t)gen_mask); if ((err = zfs_zget(zfsvfs, object, &zp))) { zfs_exit(zfsvfs, FTAG); return (err); } (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &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 (%llu) != fid gen (%llu)\n", (u_longlong_t)zp_gen, (u_longlong_t)fid_gen); vrele(ZTOV(zp)); zfs_exit(zfsvfs, FTAG); return (SET_ERROR(EINVAL)); } *vpp = ZTOV(zp); zfs_exit(zfsvfs, FTAG); err = vn_lock(*vpp, flags); if (err == 0) vnode_create_vobject(*vpp, zp->z_size, curthread); else *vpp = NULL; return (err); } /* * Block out VOPs and close zfsvfs_t::z_os * * Note, if successful, then we return with the 'z_teardown_lock' and * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying * dataset and objset intact so that they can be atomically handed off during * a subsequent rollback or recv operation and the resume thereafter. */ int zfs_suspend_fs(zfsvfs_t *zfsvfs) { int error; if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0) return (error); return (0); } /* * Rebuild SA and release VOPs. Note that ownership of the underlying dataset * is an invariant across any of the operations that can be performed while the * filesystem was suspended. Whether it succeeded or failed, the preconditions * are the same: the relevant objset and associated dataset are owned by * zfsvfs, held, and long held on entry. */ int zfs_resume_fs(zfsvfs_t *zfsvfs, dsl_dataset_t *ds) { int err; znode_t *zp; ASSERT(ZFS_TEARDOWN_WRITE_HELD(zfsvfs)); ASSERT(ZFS_TEARDOWN_INACTIVE_WRITE_HELD(zfsvfs)); /* * We already own this, so just update the objset_t, as the one we * had before may have been evicted. */ objset_t *os; VERIFY3P(ds->ds_owner, ==, zfsvfs); VERIFY(dsl_dataset_long_held(ds)); dsl_pool_t *dp = spa_get_dsl(dsl_dataset_get_spa(ds)); dsl_pool_config_enter(dp, FTAG); VERIFY0(dmu_objset_from_ds(ds, &os)); dsl_pool_config_exit(dp, FTAG); err = zfsvfs_init(zfsvfs, os); if (err != 0) goto bail; ds->ds_dir->dd_activity_cancelled = B_FALSE; VERIFY0(zfsvfs_setup(zfsvfs, B_FALSE)); zfs_set_fuid_feature(zfsvfs); /* * 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_zp * when they try to use their znode. */ mutex_enter(&zfsvfs->z_znodes_lock); for (zp = list_head(&zfsvfs->z_all_znodes); zp; zp = list_next(&zfsvfs->z_all_znodes, zp)) { (void) zfs_rezget(zp); } mutex_exit(&zfsvfs->z_znodes_lock); bail: /* release the VOPs */ ZFS_TEARDOWN_INACTIVE_EXIT_WRITE(zfsvfs); ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); if (err) { /* * Since we couldn't setup the sa framework, try to force * unmount this file system. */ if (vn_vfswlock(zfsvfs->z_vfs->vfs_vnodecovered) == 0) { vfs_ref(zfsvfs->z_vfs); (void) dounmount(zfsvfs->z_vfs, MS_FORCE, curthread); } } return (err); } static void zfs_freevfs(vfs_t *vfsp) { zfsvfs_t *zfsvfs = vfsp->vfs_data; zfsvfs_free(zfsvfs); atomic_dec_32(&zfs_active_fs_count); } #ifdef __i386__ static int desiredvnodes_backup; #include #include #include #include #include #endif static void zfs_vnodes_adjust(void) { #ifdef __i386__ int newdesiredvnodes; desiredvnodes_backup = desiredvnodes; /* * We calculate newdesiredvnodes the same way it is done in * vntblinit(). If it is equal to desiredvnodes, it means that * it wasn't tuned by the administrator and we can tune it down. */ newdesiredvnodes = min(maxproc + vm_cnt.v_page_count / 4, 2 * vm_kmem_size / (5 * (sizeof (struct vm_object) + sizeof (struct vnode)))); if (newdesiredvnodes == desiredvnodes) desiredvnodes = (3 * newdesiredvnodes) / 4; #endif } static void zfs_vnodes_adjust_back(void) { #ifdef __i386__ desiredvnodes = desiredvnodes_backup; #endif } void zfs_init(void) { printf("ZFS filesystem version: " ZPL_VERSION_STRING "\n"); /* * Initialize .zfs directory structures */ zfsctl_init(); /* * Initialize znode cache, vnode ops, etc... */ zfs_znode_init(); /* * Reduce number of vnodes. Originally number of vnodes is calculated * with UFS inode in mind. We reduce it here, because it's too big for * ZFS/i386. */ zfs_vnodes_adjust(); dmu_objset_register_type(DMU_OST_ZFS, zpl_get_file_info); zfsvfs_taskq = taskq_create("zfsvfs", 1, minclsyspri, 0, 0, 0); } void zfs_fini(void) { taskq_destroy(zfsvfs_taskq); zfsctl_fini(); zfs_znode_fini(); zfs_vnodes_adjust_back(); } int zfs_busy(void) { return (zfs_active_fs_count != 0); } /* * Release VOPs and unmount a suspended filesystem. */ int zfs_end_fs(zfsvfs_t *zfsvfs, dsl_dataset_t *ds) { ASSERT(ZFS_TEARDOWN_WRITE_HELD(zfsvfs)); ASSERT(ZFS_TEARDOWN_INACTIVE_WRITE_HELD(zfsvfs)); /* * We already own this, so just hold and rele it to update the * objset_t, as the one we had before may have been evicted. */ objset_t *os; VERIFY3P(ds->ds_owner, ==, zfsvfs); VERIFY(dsl_dataset_long_held(ds)); dsl_pool_t *dp = spa_get_dsl(dsl_dataset_get_spa(ds)); dsl_pool_config_enter(dp, FTAG); VERIFY0(dmu_objset_from_ds(ds, &os)); dsl_pool_config_exit(dp, FTAG); zfsvfs->z_os = os; /* release the VOPs */ ZFS_TEARDOWN_INACTIVE_EXIT_WRITE(zfsvfs); ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); /* * Try to force unmount this file system. */ (void) zfs_umount(zfsvfs->z_vfs, 0); zfsvfs->z_unmounted = B_TRUE; return (0); } int zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers) { int error; objset_t *os = zfsvfs->z_os; dmu_tx_t *tx; if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) return (SET_ERROR(EINVAL)); if (newvers < zfsvfs->z_version) return (SET_ERROR(EINVAL)); if (zfs_spa_version_map(newvers) > spa_version(dmu_objset_spa(zfsvfs->z_os))) return (SET_ERROR(ENOTSUP)); tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); if (newvers >= ZPL_VERSION_SA && !zfsvfs->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 && !zfsvfs->z_use_sa) { uint64_t sa_obj; ASSERT3U(spa_version(dmu_objset_spa(zfsvfs->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); ASSERT0(error); VERIFY0(sa_set_sa_object(os, sa_obj)); sa_register_update_callback(os, zfs_sa_upgrade); } spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx, "from %ju to %ju", (uintmax_t)zfsvfs->z_version, (uintmax_t)newvers); dmu_tx_commit(tx); zfsvfs->z_version = newvers; os->os_version = newvers; zfs_set_fuid_feature(zfsvfs); return (0); } /* * Return true if the corresponding vfs's unmounted flag is set. * Otherwise return false. * If this function returns true we know VFS unmount has been initiated. */ boolean_t zfs_get_vfs_flag_unmounted(objset_t *os) { zfsvfs_t *zfvp; boolean_t unmounted = B_FALSE; ASSERT3U(dmu_objset_type(os), ==, DMU_OST_ZFS); mutex_enter(&os->os_user_ptr_lock); zfvp = dmu_objset_get_user(os); if (zfvp != NULL && zfvp->z_vfs != NULL && (zfvp->z_vfs->mnt_kern_flag & MNTK_UNMOUNT)) unmounted = B_TRUE; mutex_exit(&os->os_user_ptr_lock); return (unmounted); } #ifdef _KERNEL void zfsvfs_update_fromname(const char *oldname, const char *newname) { char tmpbuf[MAXPATHLEN]; struct mount *mp; char *fromname; size_t oldlen; oldlen = strlen(oldname); mtx_lock(&mountlist_mtx); TAILQ_FOREACH(mp, &mountlist, mnt_list) { fromname = mp->mnt_stat.f_mntfromname; if (strcmp(fromname, oldname) == 0) { (void) strlcpy(fromname, newname, sizeof (mp->mnt_stat.f_mntfromname)); continue; } if (strncmp(fromname, oldname, oldlen) == 0 && (fromname[oldlen] == '/' || fromname[oldlen] == '@')) { (void) snprintf(tmpbuf, sizeof (tmpbuf), "%s%s", newname, fromname + oldlen); (void) strlcpy(fromname, tmpbuf, sizeof (mp->mnt_stat.f_mntfromname)); continue; } } mtx_unlock(&mountlist_mtx); } #endif /* * Find a prison with ZFS info. * Return the ZFS info and the (locked) prison. */ static struct zfs_jailparam * zfs_jailparam_find(struct prison *spr, struct prison **prp) { struct prison *pr; struct zfs_jailparam *zjp; for (pr = spr; ; pr = pr->pr_parent) { mtx_lock(&pr->pr_mtx); if (pr == &prison0) { zjp = &zfs_jailparam0; break; } zjp = osd_jail_get(pr, zfs_jailparam_slot); if (zjp != NULL) break; mtx_unlock(&pr->pr_mtx); } *prp = pr; return (zjp); } /* * Ensure a prison has its own ZFS info. If zjpp is non-null, point it to the * ZFS info and lock the prison. */ static void zfs_jailparam_alloc(struct prison *pr, struct zfs_jailparam **zjpp) { struct prison *ppr; struct zfs_jailparam *zjp, *nzjp; void **rsv; /* If this prison already has ZFS info, return that. */ zjp = zfs_jailparam_find(pr, &ppr); if (ppr == pr) goto done; /* * Allocate a new info record. Then check again, in case something * changed during the allocation. */ mtx_unlock(&ppr->pr_mtx); nzjp = malloc(sizeof (struct zfs_jailparam), M_PRISON, M_WAITOK); rsv = osd_reserve(zfs_jailparam_slot); zjp = zfs_jailparam_find(pr, &ppr); if (ppr == pr) { free(nzjp, M_PRISON); osd_free_reserved(rsv); goto done; } /* Inherit the initial values from the ancestor. */ mtx_lock(&pr->pr_mtx); (void) osd_jail_set_reserved(pr, zfs_jailparam_slot, rsv, nzjp); (void) memcpy(nzjp, zjp, sizeof (*zjp)); zjp = nzjp; mtx_unlock(&ppr->pr_mtx); done: if (zjpp != NULL) *zjpp = zjp; else mtx_unlock(&pr->pr_mtx); } /* * Jail OSD methods for ZFS VFS info. */ static int zfs_jailparam_create(void *obj, void *data) { struct prison *pr = obj; struct vfsoptlist *opts = data; int jsys; if (vfs_copyopt(opts, "zfs", &jsys, sizeof (jsys)) == 0 && jsys == JAIL_SYS_INHERIT) return (0); /* * Inherit a prison's initial values from its parent * (different from JAIL_SYS_INHERIT which also inherits changes). */ zfs_jailparam_alloc(pr, NULL); return (0); } static int zfs_jailparam_get(void *obj, void *data) { struct prison *ppr, *pr = obj; struct vfsoptlist *opts = data; struct zfs_jailparam *zjp; int jsys, error; zjp = zfs_jailparam_find(pr, &ppr); jsys = (ppr == pr) ? JAIL_SYS_NEW : JAIL_SYS_INHERIT; error = vfs_setopt(opts, "zfs", &jsys, sizeof (jsys)); if (error != 0 && error != ENOENT) goto done; if (jsys == JAIL_SYS_NEW) { error = vfs_setopt(opts, "zfs.mount_snapshot", &zjp->mount_snapshot, sizeof (zjp->mount_snapshot)); if (error != 0 && error != ENOENT) goto done; } else { /* * If this prison is inheriting its ZFS info, report * empty/zero parameters. */ static int mount_snapshot = 0; error = vfs_setopt(opts, "zfs.mount_snapshot", &mount_snapshot, sizeof (mount_snapshot)); if (error != 0 && error != ENOENT) goto done; } error = 0; done: mtx_unlock(&ppr->pr_mtx); return (error); } static int zfs_jailparam_set(void *obj, void *data) { struct prison *pr = obj; struct prison *ppr; struct vfsoptlist *opts = data; int error, jsys, mount_snapshot; /* Set the parameters, which should be correct. */ error = vfs_copyopt(opts, "zfs", &jsys, sizeof (jsys)); if (error == ENOENT) jsys = -1; error = vfs_copyopt(opts, "zfs.mount_snapshot", &mount_snapshot, sizeof (mount_snapshot)); if (error == ENOENT) mount_snapshot = -1; else jsys = JAIL_SYS_NEW; switch (jsys) { case JAIL_SYS_NEW: { /* "zfs=new" or "zfs.*": the prison gets its own ZFS info. */ struct zfs_jailparam *zjp; /* * A child jail cannot have more permissions than its parent */ if (pr->pr_parent != &prison0) { zjp = zfs_jailparam_find(pr->pr_parent, &ppr); mtx_unlock(&ppr->pr_mtx); if (zjp->mount_snapshot < mount_snapshot) { return (EPERM); } } zfs_jailparam_alloc(pr, &zjp); if (mount_snapshot != -1) zjp->mount_snapshot = mount_snapshot; mtx_unlock(&pr->pr_mtx); break; } case JAIL_SYS_INHERIT: /* "zfs=inherit": inherit the parent's ZFS info. */ mtx_lock(&pr->pr_mtx); osd_jail_del(pr, zfs_jailparam_slot); mtx_unlock(&pr->pr_mtx); break; case -1: /* * If the setting being changed is not ZFS related * then do nothing. */ break; } return (0); } static int zfs_jailparam_check(void *obj __unused, void *data) { struct vfsoptlist *opts = data; int error, jsys, mount_snapshot; /* Check that the parameters are correct. */ error = vfs_copyopt(opts, "zfs", &jsys, sizeof (jsys)); if (error != ENOENT) { if (error != 0) return (error); if (jsys != JAIL_SYS_NEW && jsys != JAIL_SYS_INHERIT) return (EINVAL); } error = vfs_copyopt(opts, "zfs.mount_snapshot", &mount_snapshot, sizeof (mount_snapshot)); if (error != ENOENT) { if (error != 0) return (error); if (mount_snapshot != 0 && mount_snapshot != 1) return (EINVAL); } return (0); } static void zfs_jailparam_destroy(void *data) { free(data, M_PRISON); } static void zfs_jailparam_sysinit(void *arg __unused) { struct prison *pr; osd_method_t methods[PR_MAXMETHOD] = { [PR_METHOD_CREATE] = zfs_jailparam_create, [PR_METHOD_GET] = zfs_jailparam_get, [PR_METHOD_SET] = zfs_jailparam_set, [PR_METHOD_CHECK] = zfs_jailparam_check, }; zfs_jailparam_slot = osd_jail_register(zfs_jailparam_destroy, methods); /* Copy the defaults to any existing prisons. */ sx_slock(&allprison_lock); TAILQ_FOREACH(pr, &allprison, pr_list) zfs_jailparam_alloc(pr, NULL); sx_sunlock(&allprison_lock); } static void zfs_jailparam_sysuninit(void *arg __unused) { osd_jail_deregister(zfs_jailparam_slot); } SYSINIT(zfs_jailparam_sysinit, SI_SUB_DRIVERS, SI_ORDER_ANY, zfs_jailparam_sysinit, NULL); SYSUNINIT(zfs_jailparam_sysuninit, SI_SUB_DRIVERS, SI_ORDER_ANY, zfs_jailparam_sysuninit, NULL); diff --git a/module/os/freebsd/zfs/zfs_znode.c b/module/os/freebsd/zfs/zfs_znode.c index c4f2b722ef4e..0d4c94555c6b 100644 --- a/module/os/freebsd/zfs/zfs_znode.c +++ b/module/os/freebsd/zfs/zfs_znode.c @@ -1,2228 +1,2226 @@ /* * 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 https://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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2014 by Delphix. All rights reserved. * Copyright (c) 2014 Integros [integros.com] */ /* Portions Copyright 2007 Jeremy Teo */ /* Portions Copyright 2011 Martin Matuska */ #ifdef _KERNEL #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #endif /* _KERNEL */ #include #include #include #include #include #include #include #include #include #include #include "zfs_prop.h" #include "zfs_comutil.h" /* Used by fstat(1). */ SYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD, SYSCTL_NULL_INT_PTR, sizeof (znode_t), "sizeof(znode_t)"); /* * Define ZNODE_STATS to turn on statistic gathering. By default, it is only * turned on when DEBUG is also defined. */ #ifdef ZFS_DEBUG #define ZNODE_STATS #endif /* DEBUG */ #ifdef ZNODE_STATS #define ZNODE_STAT_ADD(stat) ((stat)++) #else #define ZNODE_STAT_ADD(stat) /* nothing */ #endif /* ZNODE_STATS */ /* * Functions needed for userland (ie: libzpool) are not put under * #ifdef_KERNEL; the rest of the functions have dependencies * (such as VFS logic) that will not compile easily in userland. */ #ifdef _KERNEL #if !defined(KMEM_DEBUG) && __FreeBSD_version >= 1300102 #define _ZFS_USE_SMR static uma_zone_t znode_uma_zone; #else static kmem_cache_t *znode_cache = NULL; #endif extern struct vop_vector zfs_vnodeops; extern struct vop_vector zfs_fifoops; extern struct vop_vector zfs_shareops; /* * This callback is invoked when acquiring a RL_WRITER or RL_APPEND lock on * z_rangelock. It will modify the offset and length of the lock to reflect * znode-specific information, and convert RL_APPEND to RL_WRITER. This is * called with the rangelock_t's rl_lock held, which avoids races. */ static void zfs_rangelock_cb(zfs_locked_range_t *new, void *arg) { znode_t *zp = arg; /* * If in append mode, convert to writer and lock starting at the * current end of file. */ if (new->lr_type == RL_APPEND) { new->lr_offset = zp->z_size; new->lr_type = RL_WRITER; } /* * If we need to grow the block size then lock the whole file range. */ uint64_t end_size = MAX(zp->z_size, new->lr_offset + new->lr_length); if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) || zp->z_blksz < ZTOZSB(zp)->z_max_blksz)) { new->lr_offset = 0; new->lr_length = UINT64_MAX; } } static int zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) { znode_t *zp = buf; POINTER_INVALIDATE(&zp->z_zfsvfs); list_link_init(&zp->z_link_node); mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL); zfs_rangelock_init(&zp->z_rangelock, zfs_rangelock_cb, zp); zp->z_acl_cached = NULL; zp->z_xattr_cached = NULL; zp->z_xattr_parent = 0; zp->z_vnode = NULL; zp->z_sync_writes_cnt = 0; zp->z_async_writes_cnt = 0; return (0); } static void zfs_znode_cache_destructor(void *buf, void *arg) { (void) arg; znode_t *zp = buf; ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); ASSERT3P(zp->z_vnode, ==, NULL); ASSERT(!list_link_active(&zp->z_link_node)); mutex_destroy(&zp->z_lock); mutex_destroy(&zp->z_acl_lock); rw_destroy(&zp->z_xattr_lock); zfs_rangelock_fini(&zp->z_rangelock); ASSERT3P(zp->z_acl_cached, ==, NULL); ASSERT3P(zp->z_xattr_cached, ==, NULL); ASSERT0(atomic_load_32(&zp->z_sync_writes_cnt)); ASSERT0(atomic_load_32(&zp->z_async_writes_cnt)); } #ifdef _ZFS_USE_SMR VFS_SMR_DECLARE; static int zfs_znode_cache_constructor_smr(void *mem, int size __unused, void *private, int flags) { return (zfs_znode_cache_constructor(mem, private, flags)); } static void zfs_znode_cache_destructor_smr(void *mem, int size __unused, void *private) { zfs_znode_cache_destructor(mem, private); } void zfs_znode_init(void) { /* * Initialize zcache */ ASSERT3P(znode_uma_zone, ==, NULL); znode_uma_zone = uma_zcreate("zfs_znode_cache", sizeof (znode_t), zfs_znode_cache_constructor_smr, zfs_znode_cache_destructor_smr, NULL, NULL, 0, 0); VFS_SMR_ZONE_SET(znode_uma_zone); } static znode_t * zfs_znode_alloc_kmem(int flags) { return (uma_zalloc_smr(znode_uma_zone, flags)); } static void zfs_znode_free_kmem(znode_t *zp) { if (zp->z_xattr_cached) { nvlist_free(zp->z_xattr_cached); zp->z_xattr_cached = NULL; } uma_zfree_smr(znode_uma_zone, zp); } #else void zfs_znode_init(void) { /* * Initialize zcache */ ASSERT3P(znode_cache, ==, NULL); znode_cache = kmem_cache_create("zfs_znode_cache", sizeof (znode_t), 0, zfs_znode_cache_constructor, zfs_znode_cache_destructor, NULL, NULL, NULL, 0); } static znode_t * zfs_znode_alloc_kmem(int flags) { return (kmem_cache_alloc(znode_cache, flags)); } static void zfs_znode_free_kmem(znode_t *zp) { if (zp->z_xattr_cached) { nvlist_free(zp->z_xattr_cached); zp->z_xattr_cached = NULL; } kmem_cache_free(znode_cache, zp); } #endif void zfs_znode_fini(void) { /* * Cleanup zcache */ #ifdef _ZFS_USE_SMR if (znode_uma_zone) { uma_zdestroy(znode_uma_zone); znode_uma_zone = NULL; } #else if (znode_cache) { kmem_cache_destroy(znode_cache); znode_cache = NULL; } #endif } static int zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx) { zfs_acl_ids_t acl_ids; vattr_t vattr; znode_t *sharezp; znode_t *zp; int error; vattr.va_mask = AT_MODE|AT_UID|AT_GID; vattr.va_type = VDIR; vattr.va_mode = S_IFDIR|0555; vattr.va_uid = crgetuid(kcred); vattr.va_gid = crgetgid(kcred); sharezp = zfs_znode_alloc_kmem(KM_SLEEP); ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs)); sharezp->z_unlinked = 0; sharezp->z_atime_dirty = 0; sharezp->z_zfsvfs = zfsvfs; sharezp->z_is_sa = zfsvfs->z_use_sa; VERIFY0(zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr, kcred, NULL, &acl_ids, NULL)); zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids); ASSERT3P(zp, ==, sharezp); POINTER_INVALIDATE(&sharezp->z_zfsvfs); error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx); zfsvfs->z_shares_dir = sharezp->z_id; zfs_acl_ids_free(&acl_ids); sa_handle_destroy(sharezp->z_sa_hdl); zfs_znode_free_kmem(sharezp); return (error); } /* * define a couple of values we need available * for both 64 and 32 bit environments. */ #ifndef NBITSMINOR64 #define NBITSMINOR64 32 #endif #ifndef MAXMAJ64 #define MAXMAJ64 0xffffffffUL #endif #ifndef MAXMIN64 #define MAXMIN64 0xffffffffUL #endif /* * Create special expldev for ZFS private use. * Can't use standard expldev since it doesn't do * what we want. The standard expldev() takes a * dev32_t in LP64 and expands it to a long dev_t. * We need an interface that takes a dev32_t in ILP32 * and expands it to a long dev_t. */ static uint64_t zfs_expldev(dev_t dev) { return (((uint64_t)major(dev) << NBITSMINOR64) | minor(dev)); } /* * Special cmpldev for ZFS private use. * Can't use standard cmpldev since it takes * a long dev_t and compresses it to dev32_t in * LP64. We need to do a compaction of a long dev_t * to a dev32_t in ILP32. */ dev_t zfs_cmpldev(uint64_t dev) { return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64))); } static void zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp, dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) { ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs)); ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id))); ASSERT3P(zp->z_sa_hdl, ==, NULL); ASSERT3P(zp->z_acl_cached, ==, NULL); if (sa_hdl == NULL) { VERIFY0(sa_handle_get_from_db(zfsvfs->z_os, db, zp, SA_HDL_SHARED, &zp->z_sa_hdl)); } else { zp->z_sa_hdl = sa_hdl; sa_set_userp(sa_hdl, zp); } zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; /* * Slap on VROOT if we are the root znode unless we are the root * node of a snapshot mounted under .zfs. */ if (zp->z_id == zfsvfs->z_root && zfsvfs->z_parent == zfsvfs) ZTOV(zp)->v_flag |= VROOT; vn_exists(ZTOV(zp)); } void zfs_znode_dmu_fini(znode_t *zp) { ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) || ZFS_TEARDOWN_INACTIVE_WRITE_HELD(zp->z_zfsvfs)); sa_handle_destroy(zp->z_sa_hdl); zp->z_sa_hdl = NULL; } static void zfs_vnode_forget(vnode_t *vp) { /* copied from insmntque_stddtr */ vp->v_data = NULL; vp->v_op = &dead_vnodeops; vgone(vp); vput(vp); } /* * Construct a new znode/vnode and initialize. * * This does not do a call to dmu_set_user() that is * up to the caller to do, in case you don't want to * return the znode */ static znode_t * zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, dmu_object_type_t obj_type, sa_handle_t *hdl) { znode_t *zp; vnode_t *vp; uint64_t mode; uint64_t parent; #ifdef notyet uint64_t mtime[2], ctime[2]; #endif uint64_t projid = ZFS_DEFAULT_PROJID; sa_bulk_attr_t bulk[9]; int count = 0; int error; zp = zfs_znode_alloc_kmem(KM_SLEEP); #ifndef _ZFS_USE_SMR KASSERT((zfsvfs->z_parent->z_vfs->mnt_kern_flag & MNTK_FPLOOKUP) == 0, ("%s: fast path lookup enabled without smr", __func__)); #endif #if __FreeBSD_version >= 1300076 KASSERT(curthread->td_vp_reserved != NULL, ("zfs_znode_alloc: getnewvnode without any vnodes reserved")); #else KASSERT(curthread->td_vp_reserv > 0, ("zfs_znode_alloc: getnewvnode without any vnodes reserved")); #endif error = getnewvnode("zfs", zfsvfs->z_parent->z_vfs, &zfs_vnodeops, &vp); if (error != 0) { zfs_znode_free_kmem(zp); return (NULL); } zp->z_vnode = vp; vp->v_data = zp; /* * Acquire the vnode lock before any possible interaction with the * outside world. Specifically, there is an error path that calls * zfs_vnode_forget() and the vnode should be exclusively locked. */ vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); zp->z_sa_hdl = NULL; zp->z_unlinked = 0; zp->z_atime_dirty = 0; zp->z_mapcnt = 0; zp->z_id = db->db_object; zp->z_blksz = blksz; zp->z_seq = 0x7A4653; zp->z_sync_cnt = 0; zp->z_sync_writes_cnt = 0; zp->z_async_writes_cnt = 0; #if __FreeBSD_version >= 1300139 atomic_store_ptr(&zp->z_cached_symlink, NULL); #endif zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, &zp->z_size, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, &zp->z_links, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &zp->z_atime, 16); #ifdef notyet SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); #endif SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &zp->z_uid, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &zp->z_gid, 8); if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0 || (dmu_objset_projectquota_enabled(zfsvfs->z_os) && (zp->z_pflags & ZFS_PROJID) && sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), &projid, 8) != 0)) { if (hdl == NULL) sa_handle_destroy(zp->z_sa_hdl); zfs_vnode_forget(vp); zp->z_vnode = NULL; zfs_znode_free_kmem(zp); return (NULL); } zp->z_projid = projid; zp->z_mode = mode; /* Cache the xattr parent id */ if (zp->z_pflags & ZFS_XATTR) zp->z_xattr_parent = parent; vp->v_type = IFTOVT((mode_t)mode); switch (vp->v_type) { case VDIR: zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */ break; case VFIFO: vp->v_op = &zfs_fifoops; break; case VREG: if (parent == zfsvfs->z_shares_dir) { ASSERT0(zp->z_uid); ASSERT0(zp->z_gid); vp->v_op = &zfs_shareops; } break; default: break; } mutex_enter(&zfsvfs->z_znodes_lock); list_insert_tail(&zfsvfs->z_all_znodes, zp); - zfsvfs->z_nr_znodes++; zp->z_zfsvfs = zfsvfs; mutex_exit(&zfsvfs->z_znodes_lock); #if __FreeBSD_version >= 1400077 vn_set_state(vp, VSTATE_CONSTRUCTED); #endif VN_LOCK_AREC(vp); if (vp->v_type != VFIFO) VN_LOCK_ASHARE(vp); return (zp); } static uint64_t empty_xattr; static uint64_t pad[4]; static zfs_acl_phys_t acl_phys; /* * Create a new DMU object to hold a zfs znode. * * IN: dzp - parent directory for new znode * vap - file attributes for new znode * tx - dmu transaction id for zap operations * cr - credentials of caller * flag - flags: * IS_ROOT_NODE - new object will be root * IS_XATTR - new object is an attribute * bonuslen - length of bonus buffer * setaclp - File/Dir initial ACL * fuidp - Tracks fuid allocation. * * OUT: zpp - allocated znode * */ void zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) { uint64_t crtime[2], atime[2], mtime[2], ctime[2]; uint64_t mode, size, links, parent, pflags; uint64_t dzp_pflags = 0; uint64_t rdev = 0; zfsvfs_t *zfsvfs = dzp->z_zfsvfs; dmu_buf_t *db; timestruc_t now; uint64_t gen, obj; int bonuslen; int dnodesize; sa_handle_t *sa_hdl; dmu_object_type_t obj_type; sa_bulk_attr_t *sa_attrs; int cnt = 0; zfs_acl_locator_cb_t locate = { 0 }; ASSERT3P(vap, !=, NULL); ASSERT3U((vap->va_mask & AT_MODE), ==, AT_MODE); if (zfsvfs->z_replay) { obj = vap->va_nodeid; now = vap->va_ctime; /* see zfs_replay_create() */ gen = vap->va_nblocks; /* ditto */ dnodesize = vap->va_fsid; /* ditto */ } else { obj = 0; vfs_timestamp(&now); gen = dmu_tx_get_txg(tx); dnodesize = dmu_objset_dnodesize(zfsvfs->z_os); } if (dnodesize == 0) dnodesize = DNODE_MIN_SIZE; obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; bonuslen = (obj_type == DMU_OT_SA) ? DN_BONUS_SIZE(dnodesize) : ZFS_OLD_ZNODE_PHYS_SIZE; /* * Create a new DMU object. */ /* * There's currently no mechanism for pre-reading the blocks that will * be needed to allocate a new object, so we accept the small chance * that there will be an i/o error and we will fail one of the * assertions below. */ if (vap->va_type == VDIR) { if (zfsvfs->z_replay) { VERIFY0(zap_create_claim_norm_dnsize(zfsvfs->z_os, obj, zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, obj_type, bonuslen, dnodesize, tx)); } else { obj = zap_create_norm_dnsize(zfsvfs->z_os, zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, obj_type, bonuslen, dnodesize, tx); } } else { if (zfsvfs->z_replay) { VERIFY0(dmu_object_claim_dnsize(zfsvfs->z_os, obj, DMU_OT_PLAIN_FILE_CONTENTS, 0, obj_type, bonuslen, dnodesize, tx)); } else { obj = dmu_object_alloc_dnsize(zfsvfs->z_os, DMU_OT_PLAIN_FILE_CONTENTS, 0, obj_type, bonuslen, dnodesize, tx); } } ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); VERIFY0(sa_buf_hold(zfsvfs->z_os, obj, NULL, &db)); /* * If this is the root, fix up the half-initialized parent pointer * to reference the just-allocated physical data area. */ if (flag & IS_ROOT_NODE) { dzp->z_id = obj; } else { dzp_pflags = dzp->z_pflags; } /* * If parent is an xattr, so am I. */ if (dzp_pflags & ZFS_XATTR) { flag |= IS_XATTR; } if (zfsvfs->z_use_fuids) pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; else pflags = 0; if (vap->va_type == VDIR) { size = 2; /* contents ("." and "..") */ links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1; } else { size = links = 0; } if (vap->va_type == VBLK || vap->va_type == VCHR) { rdev = zfs_expldev(vap->va_rdev); } parent = dzp->z_id; mode = acl_ids->z_mode; if (flag & IS_XATTR) pflags |= ZFS_XATTR; /* * No execs denied will be determined when zfs_mode_compute() is called. */ pflags |= acl_ids->z_aclp->z_hints & (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); ZFS_TIME_ENCODE(&now, crtime); ZFS_TIME_ENCODE(&now, ctime); if (vap->va_mask & AT_ATIME) { ZFS_TIME_ENCODE(&vap->va_atime, atime); } else { ZFS_TIME_ENCODE(&now, atime); } if (vap->va_mask & AT_MTIME) { ZFS_TIME_ENCODE(&vap->va_mtime, mtime); } else { ZFS_TIME_ENCODE(&now, mtime); } /* Now add in all of the "SA" attributes */ VERIFY0(sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED, &sa_hdl)); /* * Setup the array of attributes to be replaced/set on the new file * * order for DMU_OT_ZNODE is critical since it needs to be constructed * in the old znode_phys_t format. Don't change this ordering */ sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); if (obj_type == DMU_OT_ZNODE) { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), NULL, &gen, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), NULL, &size, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8); } else { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), NULL, &size, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), NULL, &gen, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, &acl_ids->z_fuid, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, &acl_ids->z_fgid, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, &pflags, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16); } SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); if (obj_type == DMU_OT_ZNODE) { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL, &empty_xattr, 8); } if (obj_type == DMU_OT_ZNODE || (vap->va_type == VBLK || vap->va_type == VCHR)) { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs), NULL, &rdev, 8); } if (obj_type == DMU_OT_ZNODE) { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, &pflags, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, &acl_ids->z_fuid, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, &acl_ids->z_fgid, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad, sizeof (uint64_t) * 4); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL, &acl_phys, sizeof (zfs_acl_phys_t)); } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL, &acl_ids->z_aclp->z_acl_count, 8); locate.cb_aclp = acl_ids->z_aclp; SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs), zfs_acl_data_locator, &locate, acl_ids->z_aclp->z_acl_bytes); mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, acl_ids->z_fuid, acl_ids->z_fgid); } VERIFY0(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx)); if (!(flag & IS_ROOT_NODE)) { *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl); ASSERT3P(*zpp, !=, NULL); } else { /* * If we are creating the root node, the "parent" we * passed in is the znode for the root. */ *zpp = dzp; (*zpp)->z_sa_hdl = sa_hdl; } (*zpp)->z_pflags = pflags; (*zpp)->z_mode = mode; (*zpp)->z_dnodesize = dnodesize; if (vap->va_mask & AT_XVATTR) zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx); if (obj_type == DMU_OT_ZNODE || acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx)); } if (!(flag & IS_ROOT_NODE)) { vnode_t *vp = ZTOV(*zpp); vp->v_vflag |= VV_FORCEINSMQ; int err = insmntque(vp, zfsvfs->z_vfs); vp->v_vflag &= ~VV_FORCEINSMQ; (void) err; KASSERT(err == 0, ("insmntque() failed: error %d", err)); } kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END); ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); } /* * Update in-core attributes. It is assumed the caller will be doing an * sa_bulk_update to push the changes out. */ void zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) { xoptattr_t *xoap; xoap = xva_getxoptattr(xvap); ASSERT3P(xoap, !=, NULL); if (zp->z_zfsvfs->z_replay == B_FALSE) { ASSERT_VOP_IN_SEQC(ZTOV(zp)); } if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { uint64_t times[2]; ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs), ×, sizeof (times), tx); XVA_SET_RTN(xvap, XAT_CREATETIME); } if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_READONLY); } if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_HIDDEN); } if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_SYSTEM); } if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_ARCHIVE); } if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_IMMUTABLE); } if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_NOUNLINK); } if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_APPENDONLY); } if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_NODUMP); } if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_OPAQUE); } if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, xoap->xoa_av_quarantined, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); } if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_AV_MODIFIED); } if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { zfs_sa_set_scanstamp(zp, xvap, tx); XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); } if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_REPARSE); } if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_OFFLINE); } if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_SPARSE); } } int zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) { dmu_object_info_t doi; dmu_buf_t *db; znode_t *zp; vnode_t *vp; sa_handle_t *hdl; int locked; int err; getnewvnode_reserve_(); again: *zpp = NULL; ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); if (err) { ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); getnewvnode_drop_reserve(); return (err); } dmu_object_info_from_db(db, &doi); if (doi.doi_bonus_type != DMU_OT_SA && (doi.doi_bonus_type != DMU_OT_ZNODE || (doi.doi_bonus_type == DMU_OT_ZNODE && doi.doi_bonus_size < sizeof (znode_phys_t)))) { sa_buf_rele(db, NULL); ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); getnewvnode_drop_reserve(); return (SET_ERROR(EINVAL)); } hdl = dmu_buf_get_user(db); if (hdl != NULL) { zp = sa_get_userdata(hdl); /* * Since "SA" does immediate eviction we * should never find a sa handle that doesn't * know about the znode. */ ASSERT3P(zp, !=, NULL); ASSERT3U(zp->z_id, ==, obj_num); if (zp->z_unlinked) { err = SET_ERROR(ENOENT); } else { vp = ZTOV(zp); /* * Don't let the vnode disappear after * ZFS_OBJ_HOLD_EXIT. */ VN_HOLD(vp); *zpp = zp; err = 0; } sa_buf_rele(db, NULL); ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); if (err) { getnewvnode_drop_reserve(); return (err); } locked = VOP_ISLOCKED(vp); VI_LOCK(vp); if (VN_IS_DOOMED(vp) && locked != LK_EXCLUSIVE) { /* * The vnode is doomed and this thread doesn't * hold the exclusive lock on it, so the vnode * must be being reclaimed by another thread. * Otherwise the doomed vnode is being reclaimed * by this thread and zfs_zget is called from * ZIL internals. */ VI_UNLOCK(vp); /* * XXX vrele() locks the vnode when the last reference * is dropped. Although in this case the vnode is * doomed / dead and so no inactivation is required, * the vnode lock is still acquired. That could result * in a LOR with z_teardown_lock if another thread holds * the vnode's lock and tries to take z_teardown_lock. * But that is only possible if the other thread peforms * a ZFS vnode operation on the vnode. That either * should not happen if the vnode is dead or the thread * should also have a reference to the vnode and thus * our reference is not last. */ VN_RELE(vp); goto again; } VI_UNLOCK(vp); getnewvnode_drop_reserve(); return (err); } /* * Not found create new znode/vnode * but only if file exists. * * There is a small window where zfs_vget() could * find this object while a file create is still in * progress. This is checked for in zfs_znode_alloc() * * if zfs_znode_alloc() fails it will drop the hold on the * bonus buffer. */ zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size, doi.doi_bonus_type, NULL); if (zp == NULL) { err = SET_ERROR(ENOENT); } else { *zpp = zp; } if (err == 0) { vnode_t *vp = ZTOV(zp); err = insmntque(vp, zfsvfs->z_vfs); if (err == 0) { vp->v_hash = obj_num; VOP_UNLOCK1(vp); } else { zp->z_vnode = NULL; zfs_znode_dmu_fini(zp); zfs_znode_free(zp); *zpp = NULL; } } ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); getnewvnode_drop_reserve(); return (err); } int zfs_rezget(znode_t *zp) { zfsvfs_t *zfsvfs = zp->z_zfsvfs; dmu_object_info_t doi; dmu_buf_t *db; vnode_t *vp; uint64_t obj_num = zp->z_id; uint64_t mode, size; sa_bulk_attr_t bulk[8]; int err; int count = 0; uint64_t gen; /* * Remove cached pages before reloading the znode, so that they are not * lingering after we run into any error. Ideally, we should vgone() * the vnode in case of error, but currently we cannot do that * because of the LOR between the vnode lock and z_teardown_lock. * So, instead, we have to "doom" the znode in the illumos style. * * Ignore invalid pages during the scan. This is to avoid deadlocks * between page busying and the teardown lock, as pages are busied prior * to a VOP_GETPAGES operation, which acquires the teardown read lock. * Such pages will be invalid and can safely be skipped here. */ vp = ZTOV(zp); #if __FreeBSD_version >= 1400042 vn_pages_remove_valid(vp, 0, 0); #else vn_pages_remove(vp, 0, 0); #endif ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); mutex_enter(&zp->z_acl_lock); if (zp->z_acl_cached) { zfs_acl_free(zp->z_acl_cached); zp->z_acl_cached = NULL; } mutex_exit(&zp->z_acl_lock); rw_enter(&zp->z_xattr_lock, RW_WRITER); if (zp->z_xattr_cached) { nvlist_free(zp->z_xattr_cached); zp->z_xattr_cached = NULL; } rw_exit(&zp->z_xattr_lock); ASSERT3P(zp->z_sa_hdl, ==, NULL); err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); if (err) { ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); return (err); } dmu_object_info_from_db(db, &doi); if (doi.doi_bonus_type != DMU_OT_SA && (doi.doi_bonus_type != DMU_OT_ZNODE || (doi.doi_bonus_type == DMU_OT_ZNODE && doi.doi_bonus_size < sizeof (znode_phys_t)))) { sa_buf_rele(db, NULL); ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); return (SET_ERROR(EINVAL)); } zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL); size = zp->z_size; /* reload cached values */ SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &gen, sizeof (gen)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, &zp->z_size, sizeof (zp->z_size)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, &zp->z_links, sizeof (zp->z_links)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, sizeof (zp->z_pflags)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &zp->z_atime, sizeof (zp->z_atime)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &zp->z_uid, sizeof (zp->z_uid)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &zp->z_gid, sizeof (zp->z_gid)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, sizeof (mode)); if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { zfs_znode_dmu_fini(zp); ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); return (SET_ERROR(EIO)); } zp->z_mode = mode; if (gen != zp->z_gen) { zfs_znode_dmu_fini(zp); ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); return (SET_ERROR(EIO)); } /* * It is highly improbable but still quite possible that two * objects in different datasets are created with the same * object numbers and in transaction groups with the same * numbers. znodes corresponding to those objects would * have the same z_id and z_gen, but their other attributes * may be different. * zfs recv -F may replace one of such objects with the other. * As a result file properties recorded in the replaced * object's vnode may no longer match the received object's * properties. At present the only cached property is the * files type recorded in v_type. * So, handle this case by leaving the old vnode and znode * disassociated from the actual object. A new vnode and a * znode will be created if the object is accessed * (e.g. via a look-up). The old vnode and znode will be * recycled when the last vnode reference is dropped. */ if (vp->v_type != IFTOVT((mode_t)zp->z_mode)) { zfs_znode_dmu_fini(zp); ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); return (SET_ERROR(EIO)); } /* * If the file has zero links, then it has been unlinked on the send * side and it must be in the received unlinked set. * We call zfs_znode_dmu_fini() now to prevent any accesses to the * stale data and to prevent automatically removal of the file in * zfs_zinactive(). The file will be removed either when it is removed * on the send side and the next incremental stream is received or * when the unlinked set gets processed. */ zp->z_unlinked = (zp->z_links == 0); if (zp->z_unlinked) { zfs_znode_dmu_fini(zp); ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); return (0); } zp->z_blksz = doi.doi_data_block_size; if (zp->z_size != size) vnode_pager_setsize(vp, zp->z_size); ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); return (0); } void zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) { zfsvfs_t *zfsvfs = zp->z_zfsvfs; objset_t *os = zfsvfs->z_os; uint64_t obj = zp->z_id; uint64_t acl_obj = zfs_external_acl(zp); ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); if (acl_obj) { VERIFY(!zp->z_is_sa); VERIFY0(dmu_object_free(os, acl_obj, tx)); } VERIFY0(dmu_object_free(os, obj, tx)); zfs_znode_dmu_fini(zp); ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); zfs_znode_free(zp); } void zfs_zinactive(znode_t *zp) { zfsvfs_t *zfsvfs = zp->z_zfsvfs; uint64_t z_id = zp->z_id; ASSERT3P(zp->z_sa_hdl, !=, NULL); /* * Don't allow a zfs_zget() while were trying to release this znode */ ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id); /* * If this was the last reference to a file with no links, remove * the file from the file system unless the file system is mounted * read-only. That can happen, for example, if the file system was * originally read-write, the file was opened, then unlinked and * the file system was made read-only before the file was finally * closed. The file will remain in the unlinked set. */ if (zp->z_unlinked) { ASSERT(!zfsvfs->z_issnap); if ((zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) == 0) { ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); zfs_rmnode(zp); return; } } zfs_znode_dmu_fini(zp); ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); zfs_znode_free(zp); } void zfs_znode_free(znode_t *zp) { zfsvfs_t *zfsvfs = zp->z_zfsvfs; #if __FreeBSD_version >= 1300139 char *symlink; #endif ASSERT3P(zp->z_sa_hdl, ==, NULL); zp->z_vnode = NULL; mutex_enter(&zfsvfs->z_znodes_lock); POINTER_INVALIDATE(&zp->z_zfsvfs); list_remove(&zfsvfs->z_all_znodes, zp); - zfsvfs->z_nr_znodes--; mutex_exit(&zfsvfs->z_znodes_lock); #if __FreeBSD_version >= 1300139 symlink = atomic_load_ptr(&zp->z_cached_symlink); if (symlink != NULL) { atomic_store_rel_ptr((uintptr_t *)&zp->z_cached_symlink, (uintptr_t)NULL); cache_symlink_free(symlink, strlen(symlink) + 1); } #endif if (zp->z_acl_cached) { zfs_acl_free(zp->z_acl_cached); zp->z_acl_cached = NULL; } zfs_znode_free_kmem(zp); } void zfs_tstamp_update_setup_ext(znode_t *zp, uint_t flag, uint64_t mtime[2], uint64_t ctime[2], boolean_t have_tx) { timestruc_t now; vfs_timestamp(&now); if (have_tx) { /* will sa_bulk_update happen really soon? */ zp->z_atime_dirty = 0; zp->z_seq++; } else { zp->z_atime_dirty = 1; } if (flag & AT_ATIME) { ZFS_TIME_ENCODE(&now, zp->z_atime); } if (flag & AT_MTIME) { ZFS_TIME_ENCODE(&now, mtime); if (zp->z_zfsvfs->z_use_fuids) { zp->z_pflags |= (ZFS_ARCHIVE | ZFS_AV_MODIFIED); } } if (flag & AT_CTIME) { ZFS_TIME_ENCODE(&now, ctime); if (zp->z_zfsvfs->z_use_fuids) zp->z_pflags |= ZFS_ARCHIVE; } } void zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], uint64_t ctime[2]) { zfs_tstamp_update_setup_ext(zp, flag, mtime, ctime, B_TRUE); } /* * Grow the block size for a file. * * IN: zp - znode of file to free data in. * size - requested block size * tx - open transaction. * * NOTE: this function assumes that the znode is write locked. */ void zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) { int error; u_longlong_t dummy; if (size <= zp->z_blksz) return; /* * If the file size is already greater than the current blocksize, * we will not grow. If there is more than one block in a file, * the blocksize cannot change. */ if (zp->z_blksz && zp->z_size > zp->z_blksz) return; error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id, size, 0, tx); if (error == ENOTSUP) return; ASSERT0(error); /* What blocksize did we actually get? */ dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); } /* * Increase the file length * * IN: zp - znode of file to free data in. * end - new end-of-file * * RETURN: 0 on success, error code on failure */ static int zfs_extend(znode_t *zp, uint64_t end) { zfsvfs_t *zfsvfs = zp->z_zfsvfs; dmu_tx_t *tx; zfs_locked_range_t *lr; uint64_t newblksz; int error; /* * We will change zp_size, lock the whole file. */ lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER); /* * Nothing to do if file already at desired length. */ if (end <= zp->z_size) { zfs_rangelock_exit(lr); return (0); } tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); if (end > zp->z_blksz && (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { /* * We are growing the file past the current block size. */ if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) { /* * File's blocksize is already larger than the * "recordsize" property. Only let it grow to * the next power of 2. */ ASSERT(!ISP2(zp->z_blksz)); newblksz = MIN(end, 1 << highbit64(zp->z_blksz)); } else { newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz); } dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); } else { newblksz = 0; } error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); zfs_rangelock_exit(lr); return (error); } if (newblksz) zfs_grow_blocksize(zp, newblksz, tx); zp->z_size = end; VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs), &zp->z_size, sizeof (zp->z_size), tx)); vnode_pager_setsize(ZTOV(zp), end); zfs_rangelock_exit(lr); dmu_tx_commit(tx); return (0); } /* * Free space in a file. * * IN: zp - znode of file to free data in. * off - start of section to free. * len - length of section to free. * * RETURN: 0 on success, error code on failure */ static int zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) { zfsvfs_t *zfsvfs = zp->z_zfsvfs; zfs_locked_range_t *lr; int error; /* * Lock the range being freed. */ lr = zfs_rangelock_enter(&zp->z_rangelock, off, len, RL_WRITER); /* * Nothing to do if file already at desired length. */ if (off >= zp->z_size) { zfs_rangelock_exit(lr); return (0); } if (off + len > zp->z_size) len = zp->z_size - off; error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len); if (error == 0) { #if __FreeBSD_version >= 1400032 vnode_pager_purge_range(ZTOV(zp), off, off + len); #else /* * Before __FreeBSD_version 1400032 we cannot free block in the * middle of a file, but only at the end of a file, so this code * path should never happen. */ vnode_pager_setsize(ZTOV(zp), off); #endif } zfs_rangelock_exit(lr); return (error); } /* * Truncate a file * * IN: zp - znode of file to free data in. * end - new end-of-file. * * RETURN: 0 on success, error code on failure */ static int zfs_trunc(znode_t *zp, uint64_t end) { zfsvfs_t *zfsvfs = zp->z_zfsvfs; vnode_t *vp = ZTOV(zp); dmu_tx_t *tx; zfs_locked_range_t *lr; int error; sa_bulk_attr_t bulk[2]; int count = 0; /* * We will change zp_size, lock the whole file. */ lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER); /* * Nothing to do if file already at desired length. */ if (end >= zp->z_size) { zfs_rangelock_exit(lr); return (0); } error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, DMU_OBJECT_END); if (error) { zfs_rangelock_exit(lr); return (error); } tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); dmu_tx_mark_netfree(tx); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); zfs_rangelock_exit(lr); return (error); } zp->z_size = end; SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, &zp->z_size, sizeof (zp->z_size)); if (end == 0) { zp->z_pflags &= ~ZFS_SPARSE; SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, 8); } VERIFY0(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx)); dmu_tx_commit(tx); /* * Clear any mapped pages in the truncated region. This has to * happen outside of the transaction to avoid the possibility of * a deadlock with someone trying to push a page that we are * about to invalidate. */ vnode_pager_setsize(vp, end); zfs_rangelock_exit(lr); return (0); } /* * Free space in a file * * IN: zp - znode of file to free data in. * off - start of range * len - end of range (0 => EOF) * flag - current file open mode flags. * log - TRUE if this action should be logged * * RETURN: 0 on success, error code on failure */ int zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) { dmu_tx_t *tx; zfsvfs_t *zfsvfs = zp->z_zfsvfs; zilog_t *zilog = zfsvfs->z_log; uint64_t mode; uint64_t mtime[2], ctime[2]; sa_bulk_attr_t bulk[3]; int count = 0; int error; if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode, sizeof (mode))) != 0) return (error); if (off > zp->z_size) { error = zfs_extend(zp, off+len); if (error == 0 && log) goto log; else return (error); } if (len == 0) { error = zfs_trunc(zp, off); } else { if ((error = zfs_free_range(zp, off, len)) == 0 && off + len > zp->z_size) error = zfs_extend(zp, off+len); } if (error || !log) return (error); log: tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); return (error); } SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, 8); zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); ASSERT0(error); zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); dmu_tx_commit(tx); return (0); } void zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) { uint64_t moid, obj, sa_obj, version; uint64_t sense = ZFS_CASE_SENSITIVE; uint64_t norm = 0; nvpair_t *elem; int error; int i; znode_t *rootzp = NULL; zfsvfs_t *zfsvfs; vattr_t vattr; znode_t *zp; zfs_acl_ids_t acl_ids; /* * First attempt to create master node. */ /* * In an empty objset, there are no blocks to read and thus * there can be no i/o errors (which we assert below). */ moid = MASTER_NODE_OBJ; error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, DMU_OT_NONE, 0, tx); ASSERT0(error); /* * Set starting attributes. */ version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); elem = NULL; while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { /* For the moment we expect all zpl props to be uint64_ts */ uint64_t val; const char *name; ASSERT3S(nvpair_type(elem), ==, DATA_TYPE_UINT64); val = fnvpair_value_uint64(elem); name = nvpair_name(elem); if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { if (val < version) version = val; } else { error = zap_update(os, moid, name, 8, 1, &val, tx); } ASSERT0(error); if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) norm = val; else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) sense = val; } ASSERT3U(version, !=, 0); error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); ASSERT0(error); /* * Create zap object used for SA attribute registration */ if (version >= ZPL_VERSION_SA) { sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, DMU_OT_NONE, 0, tx); error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); ASSERT0(error); } else { sa_obj = 0; } /* * Create a delete queue. */ obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); ASSERT0(error); /* * Create root znode. Create minimal znode/vnode/zfsvfs * to allow zfs_mknode to work. */ VATTR_NULL(&vattr); vattr.va_mask = AT_MODE|AT_UID|AT_GID; vattr.va_type = VDIR; vattr.va_mode = S_IFDIR|0755; vattr.va_uid = crgetuid(cr); vattr.va_gid = crgetgid(cr); zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); rootzp = zfs_znode_alloc_kmem(KM_SLEEP); ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs)); rootzp->z_unlinked = 0; rootzp->z_atime_dirty = 0; rootzp->z_is_sa = USE_SA(version, os); zfsvfs->z_os = os; zfsvfs->z_parent = zfsvfs; zfsvfs->z_version = version; zfsvfs->z_use_fuids = USE_FUIDS(version, os); zfsvfs->z_use_sa = USE_SA(version, os); zfsvfs->z_norm = norm; error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, &zfsvfs->z_attr_table); ASSERT0(error); /* * Fold case on file systems that are always or sometimes case * insensitive. */ if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), offsetof(znode_t, z_link_node)); for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); rootzp->z_zfsvfs = zfsvfs; VERIFY0(zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, cr, NULL, &acl_ids, NULL)); zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); ASSERT3P(zp, ==, rootzp); error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); ASSERT0(error); zfs_acl_ids_free(&acl_ids); POINTER_INVALIDATE(&rootzp->z_zfsvfs); sa_handle_destroy(rootzp->z_sa_hdl); zfs_znode_free_kmem(rootzp); /* * Create shares directory */ error = zfs_create_share_dir(zfsvfs, tx); ASSERT0(error); for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) mutex_destroy(&zfsvfs->z_hold_mtx[i]); kmem_free(zfsvfs, sizeof (zfsvfs_t)); } #endif /* _KERNEL */ static int zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) { uint64_t sa_obj = 0; int error; error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); if (error != 0 && error != ENOENT) return (error); error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); return (error); } static int zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, dmu_buf_t **db, const void *tag) { dmu_object_info_t doi; int error; if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) return (error); dmu_object_info_from_db(*db, &doi); if ((doi.doi_bonus_type != DMU_OT_SA && doi.doi_bonus_type != DMU_OT_ZNODE) || (doi.doi_bonus_type == DMU_OT_ZNODE && doi.doi_bonus_size < sizeof (znode_phys_t))) { sa_buf_rele(*db, tag); return (SET_ERROR(ENOTSUP)); } error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); if (error != 0) { sa_buf_rele(*db, tag); return (error); } return (0); } static void zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, const void *tag) { sa_handle_destroy(hdl); sa_buf_rele(db, tag); } /* * Given an object number, return its parent object number and whether * or not the object is an extended attribute directory. */ static int zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp, int *is_xattrdir) { uint64_t parent; uint64_t pflags; uint64_t mode; uint64_t parent_mode; sa_bulk_attr_t bulk[3]; sa_handle_t *sa_hdl; dmu_buf_t *sa_db; int count = 0; int error; SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, &parent, sizeof (parent)); SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, &pflags, sizeof (pflags)); SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, &mode, sizeof (mode)); if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) return (error); /* * When a link is removed its parent pointer is not changed and will * be invalid. There are two cases where a link is removed but the * file stays around, when it goes to the delete queue and when there * are additional links. */ error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG); if (error != 0) return (error); error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode)); zfs_release_sa_handle(sa_hdl, sa_db, FTAG); if (error != 0) return (error); *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); /* * Extended attributes can be applied to files, directories, etc. * Otherwise the parent must be a directory. */ if (!*is_xattrdir && !S_ISDIR(parent_mode)) return (SET_ERROR(EINVAL)); *pobjp = parent; return (0); } /* * Given an object number, return some zpl level statistics */ static int zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, zfs_stat_t *sb) { sa_bulk_attr_t bulk[4]; int count = 0; SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, &sb->zs_mode, sizeof (sb->zs_mode)); SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, &sb->zs_gen, sizeof (sb->zs_gen)); SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, &sb->zs_links, sizeof (sb->zs_links)); SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, &sb->zs_ctime, sizeof (sb->zs_ctime)); return (sa_bulk_lookup(hdl, bulk, count)); } static int zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, sa_attr_type_t *sa_table, char *buf, int len) { sa_handle_t *sa_hdl; sa_handle_t *prevhdl = NULL; dmu_buf_t *prevdb = NULL; dmu_buf_t *sa_db = NULL; char *path = buf + len - 1; int error; *path = '\0'; sa_hdl = hdl; uint64_t deleteq_obj; VERIFY0(zap_lookup(osp, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, sizeof (uint64_t), 1, &deleteq_obj)); error = zap_lookup_int(osp, deleteq_obj, obj); if (error == 0) { return (ESTALE); } else if (error != ENOENT) { return (error); } for (;;) { uint64_t pobj; char component[MAXNAMELEN + 2]; size_t complen; int is_xattrdir; if (prevdb) { ASSERT3P(prevhdl, !=, NULL); zfs_release_sa_handle(prevhdl, prevdb, FTAG); } if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj, &is_xattrdir)) != 0) break; if (pobj == obj) { if (path[0] != '/') *--path = '/'; break; } component[0] = '/'; if (is_xattrdir) { (void) sprintf(component + 1, ""); } else { error = zap_value_search(osp, pobj, obj, ZFS_DIRENT_OBJ(-1ULL), component + 1); if (error != 0) break; } complen = strlen(component); path -= complen; ASSERT3P(path, >=, buf); memcpy(path, component, complen); obj = pobj; if (sa_hdl != hdl) { prevhdl = sa_hdl; prevdb = sa_db; } error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); if (error != 0) { sa_hdl = prevhdl; sa_db = prevdb; break; } } if (sa_hdl != NULL && sa_hdl != hdl) { ASSERT3P(sa_db, !=, NULL); zfs_release_sa_handle(sa_hdl, sa_db, FTAG); } if (error == 0) (void) memmove(buf, path, buf + len - path); return (error); } int zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) { sa_attr_type_t *sa_table; sa_handle_t *hdl; dmu_buf_t *db; int error; error = zfs_sa_setup(osp, &sa_table); if (error != 0) return (error); error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); if (error != 0) return (error); error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); zfs_release_sa_handle(hdl, db, FTAG); return (error); } int zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, char *buf, int len) { char *path = buf + len - 1; sa_attr_type_t *sa_table; sa_handle_t *hdl; dmu_buf_t *db; int error; *path = '\0'; error = zfs_sa_setup(osp, &sa_table); if (error != 0) return (error); error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); if (error != 0) return (error); error = zfs_obj_to_stats_impl(hdl, sa_table, sb); if (error != 0) { zfs_release_sa_handle(hdl, db, FTAG); return (error); } error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); zfs_release_sa_handle(hdl, db, FTAG); return (error); } /* * Read a property stored within the master node. */ int zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) { uint64_t *cached_copy = NULL; /* * Figure out where in the objset_t the cached copy would live, if it * is available for the requested property. */ if (os != NULL) { switch (prop) { case ZFS_PROP_VERSION: cached_copy = &os->os_version; break; case ZFS_PROP_NORMALIZE: cached_copy = &os->os_normalization; break; case ZFS_PROP_UTF8ONLY: cached_copy = &os->os_utf8only; break; case ZFS_PROP_CASE: cached_copy = &os->os_casesensitivity; break; default: break; } } if (cached_copy != NULL && *cached_copy != OBJSET_PROP_UNINITIALIZED) { *value = *cached_copy; return (0); } /* * If the property wasn't cached, look up the file system's value for * the property. For the version property, we look up a slightly * different string. */ const char *pname; int error = ENOENT; if (prop == ZFS_PROP_VERSION) { pname = ZPL_VERSION_STR; } else { pname = zfs_prop_to_name(prop); } if (os != NULL) { ASSERT3U(os->os_phys->os_type, ==, DMU_OST_ZFS); 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; case ZFS_PROP_ACLTYPE: *value = ZFS_ACLTYPE_NFSV4; break; default: return (error); } error = 0; } /* * If one of the methods for getting the property value above worked, * copy it into the objset_t's cache. */ if (error == 0 && cached_copy != NULL) { *cached_copy = *value; } return (error); } void zfs_znode_update_vfs(znode_t *zp) { vm_object_t object; if ((object = ZTOV(zp)->v_object) == NULL || zp->z_size == object->un_pager.vnp.vnp_size) return; vnode_pager_setsize(ZTOV(zp), zp->z_size); } #ifdef _KERNEL int zfs_znode_parent_and_name(znode_t *zp, znode_t **dzpp, char *buf) { zfsvfs_t *zfsvfs = zp->z_zfsvfs; uint64_t parent; int is_xattrdir; int err; /* Extended attributes should not be visible as regular files. */ if ((zp->z_pflags & ZFS_XATTR) != 0) return (SET_ERROR(EINVAL)); err = zfs_obj_to_pobj(zfsvfs->z_os, zp->z_sa_hdl, zfsvfs->z_attr_table, &parent, &is_xattrdir); if (err != 0) return (err); ASSERT0(is_xattrdir); /* No name as this is a root object. */ if (parent == zp->z_id) return (SET_ERROR(EINVAL)); err = zap_value_search(zfsvfs->z_os, parent, zp->z_id, ZFS_DIRENT_OBJ(-1ULL), buf); if (err != 0) return (err); err = zfs_zget(zfsvfs, parent, dzpp); return (err); } #endif /* _KERNEL */ #ifdef _KERNEL int zfs_rlimit_fsize(off_t fsize) { struct thread *td = curthread; off_t lim; if (td == NULL) return (0); lim = lim_cur(td, RLIMIT_FSIZE); if (__predict_true((uoff_t)fsize <= lim)) return (0); /* * The limit is reached. */ PROC_LOCK(td->td_proc); kern_psignal(td->td_proc, SIGXFSZ); PROC_UNLOCK(td->td_proc); return (EFBIG); } #endif /* _KERNEL */ diff --git a/module/os/linux/zfs/zfs_ctldir.c b/module/os/linux/zfs/zfs_ctldir.c index c45a3eb5a4eb..02cb379ea840 100644 --- a/module/os/linux/zfs/zfs_ctldir.c +++ b/module/os/linux/zfs/zfs_ctldir.c @@ -1,1318 +1,1317 @@ /* * 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 https://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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (C) 2011 Lawrence Livermore National Security, LLC. * Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER). * LLNL-CODE-403049. * Rewritten for Linux by: * Rohan Puri * Brian Behlendorf * Copyright (c) 2013 by Delphix. All rights reserved. * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved. * Copyright (c) 2018 George Melikov. All Rights Reserved. * Copyright (c) 2019 Datto, Inc. All rights reserved. * Copyright (c) 2020 The MathWorks, Inc. All rights reserved. */ /* * ZFS control directory (a.k.a. ".zfs") * * This directory provides a common location for all ZFS meta-objects. * Currently, this is only the 'snapshot' and 'shares' directory, but this may * expand in the future. The elements are built dynamically, as the hierarchy * does not actually exist on disk. * * For 'snapshot', we don't want to have all snapshots always mounted, because * this would take up a huge amount of space in /etc/mnttab. We have three * types of objects: * * ctldir ------> snapshotdir -------> snapshot * | * | * V * mounted fs * * The 'snapshot' node contains just enough information to lookup '..' and act * as a mountpoint for the snapshot. Whenever we lookup a specific snapshot, we * perform an automount of the underlying filesystem and return the * corresponding inode. * * All mounts are handled automatically by an user mode helper which invokes * the mount procedure. Unmounts are handled by allowing the mount * point to expire so the kernel may automatically unmount it. * * The '.zfs', '.zfs/snapshot', and all directories created under * '.zfs/snapshot' (ie: '.zfs/snapshot/') all share the same * zfsvfs_t as the head filesystem (what '.zfs' lives under). * * File systems mounted on top of the '.zfs/snapshot/' paths * (ie: snapshots) are complete ZFS filesystems and have their own unique * zfsvfs_t. However, the fsid reported by these mounts will be the same * as that used by the parent zfsvfs_t to make NFS happy. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "zfs_namecheck.h" /* * Two AVL trees are maintained which contain all currently automounted * snapshots. Every automounted snapshots maps to a single zfs_snapentry_t * entry which MUST: * * - be attached to both trees, and * - be unique, no duplicate entries are allowed. * * The zfs_snapshots_by_name tree is indexed by the full dataset name * while the zfs_snapshots_by_objsetid tree is indexed by the unique * objsetid. This allows for fast lookups either by name or objsetid. */ static avl_tree_t zfs_snapshots_by_name; static avl_tree_t zfs_snapshots_by_objsetid; static krwlock_t zfs_snapshot_lock; /* * Control Directory Tunables (.zfs) */ int zfs_expire_snapshot = ZFSCTL_EXPIRE_SNAPSHOT; static int zfs_admin_snapshot = 0; typedef struct { char *se_name; /* full snapshot name */ char *se_path; /* full mount path */ spa_t *se_spa; /* pool spa */ uint64_t se_objsetid; /* snapshot objset id */ struct dentry *se_root_dentry; /* snapshot root dentry */ krwlock_t se_taskqid_lock; /* scheduled unmount taskqid lock */ taskqid_t se_taskqid; /* scheduled unmount taskqid */ avl_node_t se_node_name; /* zfs_snapshots_by_name link */ avl_node_t se_node_objsetid; /* zfs_snapshots_by_objsetid link */ zfs_refcount_t se_refcount; /* reference count */ } zfs_snapentry_t; static void zfsctl_snapshot_unmount_delay_impl(zfs_snapentry_t *se, int delay); /* * Allocate a new zfs_snapentry_t being careful to make a copy of the * the snapshot name and provided mount point. No reference is taken. */ static zfs_snapentry_t * zfsctl_snapshot_alloc(const char *full_name, const char *full_path, spa_t *spa, uint64_t objsetid, struct dentry *root_dentry) { zfs_snapentry_t *se; se = kmem_zalloc(sizeof (zfs_snapentry_t), KM_SLEEP); se->se_name = kmem_strdup(full_name); se->se_path = kmem_strdup(full_path); se->se_spa = spa; se->se_objsetid = objsetid; se->se_root_dentry = root_dentry; se->se_taskqid = TASKQID_INVALID; rw_init(&se->se_taskqid_lock, NULL, RW_DEFAULT, NULL); zfs_refcount_create(&se->se_refcount); return (se); } /* * Free a zfs_snapentry_t the caller must ensure there are no active * references. */ static void zfsctl_snapshot_free(zfs_snapentry_t *se) { zfs_refcount_destroy(&se->se_refcount); kmem_strfree(se->se_name); kmem_strfree(se->se_path); rw_destroy(&se->se_taskqid_lock); kmem_free(se, sizeof (zfs_snapentry_t)); } /* * Hold a reference on the zfs_snapentry_t. */ static void zfsctl_snapshot_hold(zfs_snapentry_t *se) { zfs_refcount_add(&se->se_refcount, NULL); } /* * Release a reference on the zfs_snapentry_t. When the number of * references drops to zero the structure will be freed. */ static void zfsctl_snapshot_rele(zfs_snapentry_t *se) { if (zfs_refcount_remove(&se->se_refcount, NULL) == 0) zfsctl_snapshot_free(se); } /* * Add a zfs_snapentry_t to both the zfs_snapshots_by_name and * zfs_snapshots_by_objsetid trees. While the zfs_snapentry_t is part * of the trees a reference is held. */ static void zfsctl_snapshot_add(zfs_snapentry_t *se) { ASSERT(RW_WRITE_HELD(&zfs_snapshot_lock)); zfsctl_snapshot_hold(se); avl_add(&zfs_snapshots_by_name, se); avl_add(&zfs_snapshots_by_objsetid, se); } /* * Remove a zfs_snapentry_t from both the zfs_snapshots_by_name and * zfs_snapshots_by_objsetid trees. Upon removal a reference is dropped, * this can result in the structure being freed if that was the last * remaining reference. */ static void zfsctl_snapshot_remove(zfs_snapentry_t *se) { ASSERT(RW_WRITE_HELD(&zfs_snapshot_lock)); avl_remove(&zfs_snapshots_by_name, se); avl_remove(&zfs_snapshots_by_objsetid, se); zfsctl_snapshot_rele(se); } /* * Snapshot name comparison function for the zfs_snapshots_by_name. */ static int snapentry_compare_by_name(const void *a, const void *b) { const zfs_snapentry_t *se_a = a; const zfs_snapentry_t *se_b = b; int ret; ret = strcmp(se_a->se_name, se_b->se_name); if (ret < 0) return (-1); else if (ret > 0) return (1); else return (0); } /* * Snapshot name comparison function for the zfs_snapshots_by_objsetid. */ static int snapentry_compare_by_objsetid(const void *a, const void *b) { const zfs_snapentry_t *se_a = a; const zfs_snapentry_t *se_b = b; if (se_a->se_spa != se_b->se_spa) return ((ulong_t)se_a->se_spa < (ulong_t)se_b->se_spa ? -1 : 1); if (se_a->se_objsetid < se_b->se_objsetid) return (-1); else if (se_a->se_objsetid > se_b->se_objsetid) return (1); else return (0); } /* * Find a zfs_snapentry_t in zfs_snapshots_by_name. If the snapname * is found a pointer to the zfs_snapentry_t is returned and a reference * taken on the structure. The caller is responsible for dropping the * reference with zfsctl_snapshot_rele(). If the snapname is not found * NULL will be returned. */ static zfs_snapentry_t * zfsctl_snapshot_find_by_name(const char *snapname) { zfs_snapentry_t *se, search; ASSERT(RW_LOCK_HELD(&zfs_snapshot_lock)); search.se_name = (char *)snapname; se = avl_find(&zfs_snapshots_by_name, &search, NULL); if (se) zfsctl_snapshot_hold(se); return (se); } /* * Find a zfs_snapentry_t in zfs_snapshots_by_objsetid given the objset id * rather than the snapname. In all other respects it behaves the same * as zfsctl_snapshot_find_by_name(). */ static zfs_snapentry_t * zfsctl_snapshot_find_by_objsetid(spa_t *spa, uint64_t objsetid) { zfs_snapentry_t *se, search; ASSERT(RW_LOCK_HELD(&zfs_snapshot_lock)); search.se_spa = spa; search.se_objsetid = objsetid; se = avl_find(&zfs_snapshots_by_objsetid, &search, NULL); if (se) zfsctl_snapshot_hold(se); return (se); } /* * Rename a zfs_snapentry_t in the zfs_snapshots_by_name. The structure is * removed, renamed, and added back to the new correct location in the tree. */ static int zfsctl_snapshot_rename(const char *old_snapname, const char *new_snapname) { zfs_snapentry_t *se; ASSERT(RW_WRITE_HELD(&zfs_snapshot_lock)); se = zfsctl_snapshot_find_by_name(old_snapname); if (se == NULL) return (SET_ERROR(ENOENT)); zfsctl_snapshot_remove(se); kmem_strfree(se->se_name); se->se_name = kmem_strdup(new_snapname); zfsctl_snapshot_add(se); zfsctl_snapshot_rele(se); return (0); } /* * Delayed task responsible for unmounting an expired automounted snapshot. */ static void snapentry_expire(void *data) { zfs_snapentry_t *se = (zfs_snapentry_t *)data; spa_t *spa = se->se_spa; uint64_t objsetid = se->se_objsetid; if (zfs_expire_snapshot <= 0) { zfsctl_snapshot_rele(se); return; } rw_enter(&se->se_taskqid_lock, RW_WRITER); se->se_taskqid = TASKQID_INVALID; rw_exit(&se->se_taskqid_lock); (void) zfsctl_snapshot_unmount(se->se_name, MNT_EXPIRE); zfsctl_snapshot_rele(se); /* * Reschedule the unmount if the zfs_snapentry_t wasn't removed. * This can occur when the snapshot is busy. */ rw_enter(&zfs_snapshot_lock, RW_READER); if ((se = zfsctl_snapshot_find_by_objsetid(spa, objsetid)) != NULL) { zfsctl_snapshot_unmount_delay_impl(se, zfs_expire_snapshot); zfsctl_snapshot_rele(se); } rw_exit(&zfs_snapshot_lock); } /* * Cancel an automatic unmount of a snapname. This callback is responsible * for dropping the reference on the zfs_snapentry_t which was taken when * during dispatch. */ static void zfsctl_snapshot_unmount_cancel(zfs_snapentry_t *se) { int err = 0; rw_enter(&se->se_taskqid_lock, RW_WRITER); err = taskq_cancel_id(system_delay_taskq, se->se_taskqid); /* * if we get ENOENT, the taskq couldn't be found to be * canceled, so we can just mark it as invalid because * it's already gone. If we got EBUSY, then we already * blocked until it was gone _anyway_, so we don't care. */ se->se_taskqid = TASKQID_INVALID; rw_exit(&se->se_taskqid_lock); if (err == 0) { zfsctl_snapshot_rele(se); } } /* * Dispatch the unmount task for delayed handling with a hold protecting it. */ static void zfsctl_snapshot_unmount_delay_impl(zfs_snapentry_t *se, int delay) { if (delay <= 0) return; zfsctl_snapshot_hold(se); rw_enter(&se->se_taskqid_lock, RW_WRITER); /* * If this condition happens, we managed to: * - dispatch once * - want to dispatch _again_ before it returned * * So let's just return - if that task fails at unmounting, * we'll eventually dispatch again, and if it succeeds, * no problem. */ if (se->se_taskqid != TASKQID_INVALID) { rw_exit(&se->se_taskqid_lock); zfsctl_snapshot_rele(se); return; } se->se_taskqid = taskq_dispatch_delay(system_delay_taskq, snapentry_expire, se, TQ_SLEEP, ddi_get_lbolt() + delay * HZ); rw_exit(&se->se_taskqid_lock); } /* * Schedule an automatic unmount of objset id to occur in delay seconds from * now. Any previous delayed unmount will be cancelled in favor of the * updated deadline. A reference is taken by zfsctl_snapshot_find_by_name() * and held until the outstanding task is handled or cancelled. */ int zfsctl_snapshot_unmount_delay(spa_t *spa, uint64_t objsetid, int delay) { zfs_snapentry_t *se; int error = ENOENT; rw_enter(&zfs_snapshot_lock, RW_READER); if ((se = zfsctl_snapshot_find_by_objsetid(spa, objsetid)) != NULL) { zfsctl_snapshot_unmount_cancel(se); zfsctl_snapshot_unmount_delay_impl(se, delay); zfsctl_snapshot_rele(se); error = 0; } rw_exit(&zfs_snapshot_lock); return (error); } /* * Check if snapname is currently mounted. Returned non-zero when mounted * and zero when unmounted. */ static boolean_t zfsctl_snapshot_ismounted(const char *snapname) { zfs_snapentry_t *se; boolean_t ismounted = B_FALSE; rw_enter(&zfs_snapshot_lock, RW_READER); if ((se = zfsctl_snapshot_find_by_name(snapname)) != NULL) { zfsctl_snapshot_rele(se); ismounted = B_TRUE; } rw_exit(&zfs_snapshot_lock); return (ismounted); } /* * Check if the given inode is a part of the virtual .zfs directory. */ boolean_t zfsctl_is_node(struct inode *ip) { return (ITOZ(ip)->z_is_ctldir); } /* * Check if the given inode is a .zfs/snapshots/snapname directory. */ boolean_t zfsctl_is_snapdir(struct inode *ip) { return (zfsctl_is_node(ip) && (ip->i_ino <= ZFSCTL_INO_SNAPDIRS)); } /* * Allocate a new inode with the passed id and ops. */ static struct inode * zfsctl_inode_alloc(zfsvfs_t *zfsvfs, uint64_t id, const struct file_operations *fops, const struct inode_operations *ops, uint64_t creation) { struct inode *ip; znode_t *zp; inode_timespec_t now = {.tv_sec = creation}; ip = new_inode(zfsvfs->z_sb); if (ip == NULL) return (NULL); if (!creation) now = current_time(ip); zp = ITOZ(ip); ASSERT3P(zp->z_dirlocks, ==, NULL); ASSERT3P(zp->z_acl_cached, ==, NULL); ASSERT3P(zp->z_xattr_cached, ==, NULL); zp->z_id = id; zp->z_unlinked = B_FALSE; zp->z_atime_dirty = B_FALSE; zp->z_zn_prefetch = B_FALSE; zp->z_is_sa = B_FALSE; #if !defined(HAVE_FILEMAP_RANGE_HAS_PAGE) zp->z_is_mapped = B_FALSE; #endif zp->z_is_ctldir = B_TRUE; zp->z_sa_hdl = NULL; zp->z_blksz = 0; zp->z_seq = 0; zp->z_mapcnt = 0; zp->z_size = 0; zp->z_pflags = 0; zp->z_mode = 0; zp->z_sync_cnt = 0; zp->z_sync_writes_cnt = 0; zp->z_async_writes_cnt = 0; ip->i_generation = 0; ip->i_ino = id; ip->i_mode = (S_IFDIR | S_IRWXUGO); ip->i_uid = SUID_TO_KUID(0); ip->i_gid = SGID_TO_KGID(0); ip->i_blkbits = SPA_MINBLOCKSHIFT; ip->i_atime = now; ip->i_mtime = now; ip->i_ctime = now; ip->i_fop = fops; ip->i_op = ops; #if defined(IOP_XATTR) ip->i_opflags &= ~IOP_XATTR; #endif if (insert_inode_locked(ip)) { unlock_new_inode(ip); iput(ip); return (NULL); } mutex_enter(&zfsvfs->z_znodes_lock); list_insert_tail(&zfsvfs->z_all_znodes, zp); - zfsvfs->z_nr_znodes++; membar_producer(); mutex_exit(&zfsvfs->z_znodes_lock); unlock_new_inode(ip); return (ip); } /* * Lookup the inode with given id, it will be allocated if needed. */ static struct inode * zfsctl_inode_lookup(zfsvfs_t *zfsvfs, uint64_t id, const struct file_operations *fops, const struct inode_operations *ops) { struct inode *ip = NULL; uint64_t creation = 0; dsl_dataset_t *snap_ds; dsl_pool_t *pool; while (ip == NULL) { ip = ilookup(zfsvfs->z_sb, (unsigned long)id); if (ip) break; if (id <= ZFSCTL_INO_SNAPDIRS && !creation) { pool = dmu_objset_pool(zfsvfs->z_os); dsl_pool_config_enter(pool, FTAG); if (!dsl_dataset_hold_obj(pool, ZFSCTL_INO_SNAPDIRS - id, FTAG, &snap_ds)) { creation = dsl_get_creation(snap_ds); dsl_dataset_rele(snap_ds, FTAG); } dsl_pool_config_exit(pool, FTAG); } /* May fail due to concurrent zfsctl_inode_alloc() */ ip = zfsctl_inode_alloc(zfsvfs, id, fops, ops, creation); } return (ip); } /* * Create the '.zfs' directory. This directory is cached as part of the VFS * structure. This results in a hold on the zfsvfs_t. The code in zfs_umount() * therefore checks against a vfs_count of 2 instead of 1. This reference * is removed when the ctldir is destroyed in the unmount. All other entities * under the '.zfs' directory are created dynamically as needed. * * Because the dynamically created '.zfs' directory entries assume the use * of 64-bit inode numbers this support must be disabled on 32-bit systems. */ int zfsctl_create(zfsvfs_t *zfsvfs) { ASSERT(zfsvfs->z_ctldir == NULL); zfsvfs->z_ctldir = zfsctl_inode_alloc(zfsvfs, ZFSCTL_INO_ROOT, &zpl_fops_root, &zpl_ops_root, 0); if (zfsvfs->z_ctldir == NULL) return (SET_ERROR(ENOENT)); return (0); } /* * Destroy the '.zfs' directory or remove a snapshot from zfs_snapshots_by_name. * Only called when the filesystem is unmounted. */ void zfsctl_destroy(zfsvfs_t *zfsvfs) { if (zfsvfs->z_issnap) { zfs_snapentry_t *se; spa_t *spa = zfsvfs->z_os->os_spa; uint64_t objsetid = dmu_objset_id(zfsvfs->z_os); rw_enter(&zfs_snapshot_lock, RW_WRITER); se = zfsctl_snapshot_find_by_objsetid(spa, objsetid); if (se != NULL) zfsctl_snapshot_remove(se); rw_exit(&zfs_snapshot_lock); if (se != NULL) { zfsctl_snapshot_unmount_cancel(se); zfsctl_snapshot_rele(se); } } else if (zfsvfs->z_ctldir) { iput(zfsvfs->z_ctldir); zfsvfs->z_ctldir = NULL; } } /* * Given a root znode, retrieve the associated .zfs directory. * Add a hold to the vnode and return it. */ struct inode * zfsctl_root(znode_t *zp) { ASSERT(zfs_has_ctldir(zp)); /* Must have an existing ref, so igrab() cannot return NULL */ VERIFY3P(igrab(ZTOZSB(zp)->z_ctldir), !=, NULL); return (ZTOZSB(zp)->z_ctldir); } /* * Generate a long fid to indicate a snapdir. We encode whether snapdir is * already mounted in gen field. We do this because nfsd lookup will not * trigger automount. Next time the nfsd does fh_to_dentry, we will notice * this and do automount and return ESTALE to force nfsd revalidate and follow * mount. */ static int zfsctl_snapdir_fid(struct inode *ip, fid_t *fidp) { zfid_short_t *zfid = (zfid_short_t *)fidp; zfid_long_t *zlfid = (zfid_long_t *)fidp; uint32_t gen = 0; uint64_t object; uint64_t objsetid; int i; struct dentry *dentry; if (fidp->fid_len < LONG_FID_LEN) { fidp->fid_len = LONG_FID_LEN; return (SET_ERROR(ENOSPC)); } object = ip->i_ino; objsetid = ZFSCTL_INO_SNAPDIRS - ip->i_ino; zfid->zf_len = LONG_FID_LEN; dentry = d_obtain_alias(igrab(ip)); if (!IS_ERR(dentry)) { gen = !!d_mountpoint(dentry); dput(dentry); } for (i = 0; i < sizeof (zfid->zf_object); i++) zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); for (i = 0; i < sizeof (zfid->zf_gen); i++) zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); for (i = 0; i < sizeof (zlfid->zf_setid); i++) zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i)); for (i = 0; i < sizeof (zlfid->zf_setgen); i++) zlfid->zf_setgen[i] = 0; return (0); } /* * Generate an appropriate fid for an entry in the .zfs directory. */ int zfsctl_fid(struct inode *ip, fid_t *fidp) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ITOZSB(ip); uint64_t object = zp->z_id; zfid_short_t *zfid; int i; int error; if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); if (zfsctl_is_snapdir(ip)) { zfs_exit(zfsvfs, FTAG); return (zfsctl_snapdir_fid(ip, fidp)); } if (fidp->fid_len < SHORT_FID_LEN) { fidp->fid_len = SHORT_FID_LEN; zfs_exit(zfsvfs, FTAG); return (SET_ERROR(ENOSPC)); } zfid = (zfid_short_t *)fidp; zfid->zf_len = SHORT_FID_LEN; for (i = 0; i < sizeof (zfid->zf_object); i++) zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); /* .zfs znodes always have a generation number of 0 */ for (i = 0; i < sizeof (zfid->zf_gen); i++) zfid->zf_gen[i] = 0; zfs_exit(zfsvfs, FTAG); return (0); } /* * Construct a full dataset name in full_name: "pool/dataset@snap_name" */ static int zfsctl_snapshot_name(zfsvfs_t *zfsvfs, const char *snap_name, int len, char *full_name) { objset_t *os = zfsvfs->z_os; if (zfs_component_namecheck(snap_name, NULL, NULL) != 0) return (SET_ERROR(EILSEQ)); dmu_objset_name(os, full_name); if ((strlen(full_name) + 1 + strlen(snap_name)) >= len) return (SET_ERROR(ENAMETOOLONG)); (void) strcat(full_name, "@"); (void) strcat(full_name, snap_name); return (0); } /* * Returns full path in full_path: "/pool/dataset/.zfs/snapshot/snap_name/" */ static int zfsctl_snapshot_path_objset(zfsvfs_t *zfsvfs, uint64_t objsetid, int path_len, char *full_path) { objset_t *os = zfsvfs->z_os; fstrans_cookie_t cookie; char *snapname; boolean_t case_conflict; uint64_t id, pos = 0; int error = 0; if (zfsvfs->z_vfs->vfs_mntpoint == NULL) return (SET_ERROR(ENOENT)); cookie = spl_fstrans_mark(); snapname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); while (error == 0) { dsl_pool_config_enter(dmu_objset_pool(os), FTAG); error = dmu_snapshot_list_next(zfsvfs->z_os, ZFS_MAX_DATASET_NAME_LEN, snapname, &id, &pos, &case_conflict); dsl_pool_config_exit(dmu_objset_pool(os), FTAG); if (error) goto out; if (id == objsetid) break; } snprintf(full_path, path_len, "%s/.zfs/snapshot/%s", zfsvfs->z_vfs->vfs_mntpoint, snapname); out: kmem_free(snapname, ZFS_MAX_DATASET_NAME_LEN); spl_fstrans_unmark(cookie); return (error); } /* * Special case the handling of "..". */ int zfsctl_root_lookup(struct inode *dip, const char *name, struct inode **ipp, int flags, cred_t *cr, int *direntflags, pathname_t *realpnp) { zfsvfs_t *zfsvfs = ITOZSB(dip); int error = 0; if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); if (strcmp(name, "..") == 0) { *ipp = dip->i_sb->s_root->d_inode; } else if (strcmp(name, ZFS_SNAPDIR_NAME) == 0) { *ipp = zfsctl_inode_lookup(zfsvfs, ZFSCTL_INO_SNAPDIR, &zpl_fops_snapdir, &zpl_ops_snapdir); } else if (strcmp(name, ZFS_SHAREDIR_NAME) == 0) { *ipp = zfsctl_inode_lookup(zfsvfs, ZFSCTL_INO_SHARES, &zpl_fops_shares, &zpl_ops_shares); } else { *ipp = NULL; } if (*ipp == NULL) error = SET_ERROR(ENOENT); zfs_exit(zfsvfs, FTAG); return (error); } /* * Lookup entry point for the 'snapshot' directory. Try to open the * snapshot if it exist, creating the pseudo filesystem inode as necessary. */ int zfsctl_snapdir_lookup(struct inode *dip, const char *name, struct inode **ipp, int flags, cred_t *cr, int *direntflags, pathname_t *realpnp) { zfsvfs_t *zfsvfs = ITOZSB(dip); uint64_t id; int error; if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); error = dmu_snapshot_lookup(zfsvfs->z_os, name, &id); if (error) { zfs_exit(zfsvfs, FTAG); return (error); } *ipp = zfsctl_inode_lookup(zfsvfs, ZFSCTL_INO_SNAPDIRS - id, &simple_dir_operations, &simple_dir_inode_operations); if (*ipp == NULL) error = SET_ERROR(ENOENT); zfs_exit(zfsvfs, FTAG); return (error); } /* * Renaming a directory under '.zfs/snapshot' will automatically trigger * a rename of the snapshot to the new given name. The rename is confined * to the '.zfs/snapshot' directory snapshots cannot be moved elsewhere. */ int zfsctl_snapdir_rename(struct inode *sdip, const char *snm, struct inode *tdip, const char *tnm, cred_t *cr, int flags) { zfsvfs_t *zfsvfs = ITOZSB(sdip); char *to, *from, *real, *fsname; int error; if (!zfs_admin_snapshot) return (SET_ERROR(EACCES)); if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); to = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); from = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); real = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); fsname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) { error = dmu_snapshot_realname(zfsvfs->z_os, snm, real, ZFS_MAX_DATASET_NAME_LEN, NULL); if (error == 0) { snm = real; } else if (error != ENOTSUP) { goto out; } } dmu_objset_name(zfsvfs->z_os, fsname); error = zfsctl_snapshot_name(ITOZSB(sdip), snm, ZFS_MAX_DATASET_NAME_LEN, from); if (error == 0) error = zfsctl_snapshot_name(ITOZSB(tdip), tnm, ZFS_MAX_DATASET_NAME_LEN, to); if (error == 0) error = zfs_secpolicy_rename_perms(from, to, cr); if (error != 0) goto out; /* * Cannot move snapshots out of the snapdir. */ if (sdip != tdip) { error = SET_ERROR(EINVAL); goto out; } /* * No-op when names are identical. */ if (strcmp(snm, tnm) == 0) { error = 0; goto out; } rw_enter(&zfs_snapshot_lock, RW_WRITER); error = dsl_dataset_rename_snapshot(fsname, snm, tnm, B_FALSE); if (error == 0) (void) zfsctl_snapshot_rename(snm, tnm); rw_exit(&zfs_snapshot_lock); out: kmem_free(from, ZFS_MAX_DATASET_NAME_LEN); kmem_free(to, ZFS_MAX_DATASET_NAME_LEN); kmem_free(real, ZFS_MAX_DATASET_NAME_LEN); kmem_free(fsname, ZFS_MAX_DATASET_NAME_LEN); zfs_exit(zfsvfs, FTAG); return (error); } /* * Removing a directory under '.zfs/snapshot' will automatically trigger * the removal of the snapshot with the given name. */ int zfsctl_snapdir_remove(struct inode *dip, const char *name, cred_t *cr, int flags) { zfsvfs_t *zfsvfs = ITOZSB(dip); char *snapname, *real; int error; if (!zfs_admin_snapshot) return (SET_ERROR(EACCES)); if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); snapname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); real = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) { error = dmu_snapshot_realname(zfsvfs->z_os, name, real, ZFS_MAX_DATASET_NAME_LEN, NULL); if (error == 0) { name = real; } else if (error != ENOTSUP) { goto out; } } error = zfsctl_snapshot_name(ITOZSB(dip), name, ZFS_MAX_DATASET_NAME_LEN, snapname); if (error == 0) error = zfs_secpolicy_destroy_perms(snapname, cr); if (error != 0) goto out; error = zfsctl_snapshot_unmount(snapname, MNT_FORCE); if ((error == 0) || (error == ENOENT)) error = dsl_destroy_snapshot(snapname, B_FALSE); out: kmem_free(snapname, ZFS_MAX_DATASET_NAME_LEN); kmem_free(real, ZFS_MAX_DATASET_NAME_LEN); zfs_exit(zfsvfs, FTAG); return (error); } /* * Creating a directory under '.zfs/snapshot' will automatically trigger * the creation of a new snapshot with the given name. */ int zfsctl_snapdir_mkdir(struct inode *dip, const char *dirname, vattr_t *vap, struct inode **ipp, cred_t *cr, int flags) { zfsvfs_t *zfsvfs = ITOZSB(dip); char *dsname; int error; if (!zfs_admin_snapshot) return (SET_ERROR(EACCES)); dsname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); if (zfs_component_namecheck(dirname, NULL, NULL) != 0) { error = SET_ERROR(EILSEQ); goto out; } dmu_objset_name(zfsvfs->z_os, dsname); error = zfs_secpolicy_snapshot_perms(dsname, cr); if (error != 0) goto out; if (error == 0) { error = dmu_objset_snapshot_one(dsname, dirname); if (error != 0) goto out; error = zfsctl_snapdir_lookup(dip, dirname, ipp, 0, cr, NULL, NULL); } out: kmem_free(dsname, ZFS_MAX_DATASET_NAME_LEN); return (error); } /* * Flush everything out of the kernel's export table and such. * This is needed as once the snapshot is used over NFS, its * entries in svc_export and svc_expkey caches hold reference * to the snapshot mount point. There is no known way of flushing * only the entries related to the snapshot. */ static void exportfs_flush(void) { char *argv[] = { "/usr/sbin/exportfs", "-f", NULL }; char *envp[] = { NULL }; (void) call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC); } /* * Attempt to unmount a snapshot by making a call to user space. * There is no assurance that this can or will succeed, is just a * best effort. In the case where it does fail, perhaps because * it's in use, the unmount will fail harmlessly. */ int zfsctl_snapshot_unmount(const char *snapname, int flags) { char *argv[] = { "/usr/bin/env", "umount", "-t", "zfs", "-n", NULL, NULL }; char *envp[] = { NULL }; zfs_snapentry_t *se; int error; rw_enter(&zfs_snapshot_lock, RW_READER); if ((se = zfsctl_snapshot_find_by_name(snapname)) == NULL) { rw_exit(&zfs_snapshot_lock); return (SET_ERROR(ENOENT)); } rw_exit(&zfs_snapshot_lock); exportfs_flush(); if (flags & MNT_FORCE) argv[4] = "-fn"; argv[5] = se->se_path; dprintf("unmount; path=%s\n", se->se_path); error = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC); zfsctl_snapshot_rele(se); /* * The umount system utility will return 256 on error. We must * assume this error is because the file system is busy so it is * converted to the more sensible EBUSY. */ if (error) error = SET_ERROR(EBUSY); return (error); } int zfsctl_snapshot_mount(struct path *path, int flags) { struct dentry *dentry = path->dentry; struct inode *ip = dentry->d_inode; zfsvfs_t *zfsvfs; zfsvfs_t *snap_zfsvfs; zfs_snapentry_t *se; char *full_name, *full_path; char *argv[] = { "/usr/bin/env", "mount", "-t", "zfs", "-n", NULL, NULL, NULL }; char *envp[] = { NULL }; int error; struct path spath; if (ip == NULL) return (SET_ERROR(EISDIR)); zfsvfs = ITOZSB(ip); if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); full_name = kmem_zalloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); full_path = kmem_zalloc(MAXPATHLEN, KM_SLEEP); error = zfsctl_snapshot_name(zfsvfs, dname(dentry), ZFS_MAX_DATASET_NAME_LEN, full_name); if (error) goto error; /* * Construct a mount point path from sb of the ctldir inode and dirent * name, instead of from d_path(), so that chroot'd process doesn't fail * on mount.zfs(8). */ snprintf(full_path, MAXPATHLEN, "%s/.zfs/snapshot/%s", zfsvfs->z_vfs->vfs_mntpoint ? zfsvfs->z_vfs->vfs_mntpoint : "", dname(dentry)); /* * Multiple concurrent automounts of a snapshot are never allowed. * The snapshot may be manually mounted as many times as desired. */ if (zfsctl_snapshot_ismounted(full_name)) { error = 0; goto error; } /* * Attempt to mount the snapshot from user space. Normally this * would be done using the vfs_kern_mount() function, however that * function is marked GPL-only and cannot be used. On error we * careful to log the real error to the console and return EISDIR * to safely abort the automount. This should be very rare. * * If the user mode helper happens to return EBUSY, a concurrent * mount is already in progress in which case the error is ignored. * Take note that if the program was executed successfully the return * value from call_usermodehelper() will be (exitcode << 8 + signal). */ dprintf("mount; name=%s path=%s\n", full_name, full_path); argv[5] = full_name; argv[6] = full_path; error = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC); if (error) { if (!(error & MOUNT_BUSY << 8)) { zfs_dbgmsg("Unable to automount %s error=%d", full_path, error); error = SET_ERROR(EISDIR); } else { /* * EBUSY, this could mean a concurrent mount, or the * snapshot has already been mounted at completely * different place. We return 0 so VFS will retry. For * the latter case the VFS will retry several times * and return ELOOP, which is probably not a very good * behavior. */ error = 0; } goto error; } /* * Follow down in to the mounted snapshot and set MNT_SHRINKABLE * to identify this as an automounted filesystem. */ spath = *path; path_get(&spath); if (follow_down_one(&spath)) { snap_zfsvfs = ITOZSB(spath.dentry->d_inode); snap_zfsvfs->z_parent = zfsvfs; dentry = spath.dentry; spath.mnt->mnt_flags |= MNT_SHRINKABLE; rw_enter(&zfs_snapshot_lock, RW_WRITER); se = zfsctl_snapshot_alloc(full_name, full_path, snap_zfsvfs->z_os->os_spa, dmu_objset_id(snap_zfsvfs->z_os), dentry); zfsctl_snapshot_add(se); zfsctl_snapshot_unmount_delay_impl(se, zfs_expire_snapshot); rw_exit(&zfs_snapshot_lock); } path_put(&spath); error: kmem_free(full_name, ZFS_MAX_DATASET_NAME_LEN); kmem_free(full_path, MAXPATHLEN); zfs_exit(zfsvfs, FTAG); return (error); } /* * Get the snapdir inode from fid */ int zfsctl_snapdir_vget(struct super_block *sb, uint64_t objsetid, int gen, struct inode **ipp) { int error; struct path path; char *mnt; struct dentry *dentry; mnt = kmem_alloc(MAXPATHLEN, KM_SLEEP); error = zfsctl_snapshot_path_objset(sb->s_fs_info, objsetid, MAXPATHLEN, mnt); if (error) goto out; /* Trigger automount */ error = -kern_path(mnt, LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &path); if (error) goto out; path_put(&path); /* * Get the snapdir inode. Note, we don't want to use the above * path because it contains the root of the snapshot rather * than the snapdir. */ *ipp = ilookup(sb, ZFSCTL_INO_SNAPDIRS - objsetid); if (*ipp == NULL) { error = SET_ERROR(ENOENT); goto out; } /* check gen, see zfsctl_snapdir_fid */ dentry = d_obtain_alias(igrab(*ipp)); if (gen != (!IS_ERR(dentry) && d_mountpoint(dentry))) { iput(*ipp); *ipp = NULL; error = SET_ERROR(ENOENT); } if (!IS_ERR(dentry)) dput(dentry); out: kmem_free(mnt, MAXPATHLEN); return (error); } int zfsctl_shares_lookup(struct inode *dip, char *name, struct inode **ipp, int flags, cred_t *cr, int *direntflags, pathname_t *realpnp) { zfsvfs_t *zfsvfs = ITOZSB(dip); znode_t *zp; znode_t *dzp; int error; if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); if (zfsvfs->z_shares_dir == 0) { zfs_exit(zfsvfs, FTAG); return (SET_ERROR(ENOTSUP)); } if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) { error = zfs_lookup(dzp, name, &zp, 0, cr, NULL, NULL); zrele(dzp); } zfs_exit(zfsvfs, FTAG); return (error); } /* * Initialize the various pieces we'll need to create and manipulate .zfs * directories. Currently this is unused but available. */ void zfsctl_init(void) { avl_create(&zfs_snapshots_by_name, snapentry_compare_by_name, sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t, se_node_name)); avl_create(&zfs_snapshots_by_objsetid, snapentry_compare_by_objsetid, sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t, se_node_objsetid)); rw_init(&zfs_snapshot_lock, NULL, RW_DEFAULT, NULL); } /* * Cleanup the various pieces we needed for .zfs directories. In particular * ensure the expiry timer is canceled safely. */ void zfsctl_fini(void) { avl_destroy(&zfs_snapshots_by_name); avl_destroy(&zfs_snapshots_by_objsetid); rw_destroy(&zfs_snapshot_lock); } module_param(zfs_admin_snapshot, int, 0644); MODULE_PARM_DESC(zfs_admin_snapshot, "Enable mkdir/rmdir/mv in .zfs/snapshot"); module_param(zfs_expire_snapshot, int, 0644); MODULE_PARM_DESC(zfs_expire_snapshot, "Seconds to expire .zfs/snapshot"); diff --git a/module/os/linux/zfs/zfs_vfsops.c b/module/os/linux/zfs/zfs_vfsops.c index 464c12e1108d..a1db5c57c18b 100644 --- a/module/os/linux/zfs/zfs_vfsops.c +++ b/module/os/linux/zfs/zfs_vfsops.c @@ -1,2130 +1,2129 @@ /* * 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 https://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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2018 by Delphix. All rights reserved. */ /* Portions Copyright 2010 Robert Milkowski */ #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_comutil.h" enum { TOKEN_RO, TOKEN_RW, TOKEN_SETUID, TOKEN_NOSETUID, TOKEN_EXEC, TOKEN_NOEXEC, TOKEN_DEVICES, TOKEN_NODEVICES, TOKEN_DIRXATTR, TOKEN_SAXATTR, TOKEN_XATTR, TOKEN_NOXATTR, TOKEN_ATIME, TOKEN_NOATIME, TOKEN_RELATIME, TOKEN_NORELATIME, TOKEN_NBMAND, TOKEN_NONBMAND, TOKEN_MNTPOINT, TOKEN_LAST, }; static const match_table_t zpl_tokens = { { TOKEN_RO, MNTOPT_RO }, { TOKEN_RW, MNTOPT_RW }, { TOKEN_SETUID, MNTOPT_SETUID }, { TOKEN_NOSETUID, MNTOPT_NOSETUID }, { TOKEN_EXEC, MNTOPT_EXEC }, { TOKEN_NOEXEC, MNTOPT_NOEXEC }, { TOKEN_DEVICES, MNTOPT_DEVICES }, { TOKEN_NODEVICES, MNTOPT_NODEVICES }, { TOKEN_DIRXATTR, MNTOPT_DIRXATTR }, { TOKEN_SAXATTR, MNTOPT_SAXATTR }, { TOKEN_XATTR, MNTOPT_XATTR }, { TOKEN_NOXATTR, MNTOPT_NOXATTR }, { TOKEN_ATIME, MNTOPT_ATIME }, { TOKEN_NOATIME, MNTOPT_NOATIME }, { TOKEN_RELATIME, MNTOPT_RELATIME }, { TOKEN_NORELATIME, MNTOPT_NORELATIME }, { TOKEN_NBMAND, MNTOPT_NBMAND }, { TOKEN_NONBMAND, MNTOPT_NONBMAND }, { TOKEN_MNTPOINT, MNTOPT_MNTPOINT "=%s" }, { TOKEN_LAST, NULL }, }; static void zfsvfs_vfs_free(vfs_t *vfsp) { if (vfsp != NULL) { if (vfsp->vfs_mntpoint != NULL) kmem_strfree(vfsp->vfs_mntpoint); kmem_free(vfsp, sizeof (vfs_t)); } } static int zfsvfs_parse_option(char *option, int token, substring_t *args, vfs_t *vfsp) { switch (token) { case TOKEN_RO: vfsp->vfs_readonly = B_TRUE; vfsp->vfs_do_readonly = B_TRUE; break; case TOKEN_RW: vfsp->vfs_readonly = B_FALSE; vfsp->vfs_do_readonly = B_TRUE; break; case TOKEN_SETUID: vfsp->vfs_setuid = B_TRUE; vfsp->vfs_do_setuid = B_TRUE; break; case TOKEN_NOSETUID: vfsp->vfs_setuid = B_FALSE; vfsp->vfs_do_setuid = B_TRUE; break; case TOKEN_EXEC: vfsp->vfs_exec = B_TRUE; vfsp->vfs_do_exec = B_TRUE; break; case TOKEN_NOEXEC: vfsp->vfs_exec = B_FALSE; vfsp->vfs_do_exec = B_TRUE; break; case TOKEN_DEVICES: vfsp->vfs_devices = B_TRUE; vfsp->vfs_do_devices = B_TRUE; break; case TOKEN_NODEVICES: vfsp->vfs_devices = B_FALSE; vfsp->vfs_do_devices = B_TRUE; break; case TOKEN_DIRXATTR: vfsp->vfs_xattr = ZFS_XATTR_DIR; vfsp->vfs_do_xattr = B_TRUE; break; case TOKEN_SAXATTR: vfsp->vfs_xattr = ZFS_XATTR_SA; vfsp->vfs_do_xattr = B_TRUE; break; case TOKEN_XATTR: vfsp->vfs_xattr = ZFS_XATTR_DIR; vfsp->vfs_do_xattr = B_TRUE; break; case TOKEN_NOXATTR: vfsp->vfs_xattr = ZFS_XATTR_OFF; vfsp->vfs_do_xattr = B_TRUE; break; case TOKEN_ATIME: vfsp->vfs_atime = B_TRUE; vfsp->vfs_do_atime = B_TRUE; break; case TOKEN_NOATIME: vfsp->vfs_atime = B_FALSE; vfsp->vfs_do_atime = B_TRUE; break; case TOKEN_RELATIME: vfsp->vfs_relatime = B_TRUE; vfsp->vfs_do_relatime = B_TRUE; break; case TOKEN_NORELATIME: vfsp->vfs_relatime = B_FALSE; vfsp->vfs_do_relatime = B_TRUE; break; case TOKEN_NBMAND: vfsp->vfs_nbmand = B_TRUE; vfsp->vfs_do_nbmand = B_TRUE; break; case TOKEN_NONBMAND: vfsp->vfs_nbmand = B_FALSE; vfsp->vfs_do_nbmand = B_TRUE; break; case TOKEN_MNTPOINT: vfsp->vfs_mntpoint = match_strdup(&args[0]); if (vfsp->vfs_mntpoint == NULL) return (SET_ERROR(ENOMEM)); break; default: break; } return (0); } /* * Parse the raw mntopts and return a vfs_t describing the options. */ static int zfsvfs_parse_options(char *mntopts, vfs_t **vfsp) { vfs_t *tmp_vfsp; int error; tmp_vfsp = kmem_zalloc(sizeof (vfs_t), KM_SLEEP); if (mntopts != NULL) { substring_t args[MAX_OPT_ARGS]; char *tmp_mntopts, *p, *t; int token; tmp_mntopts = t = kmem_strdup(mntopts); if (tmp_mntopts == NULL) return (SET_ERROR(ENOMEM)); while ((p = strsep(&t, ",")) != NULL) { if (!*p) continue; args[0].to = args[0].from = NULL; token = match_token(p, zpl_tokens, args); error = zfsvfs_parse_option(p, token, args, tmp_vfsp); if (error) { kmem_strfree(tmp_mntopts); zfsvfs_vfs_free(tmp_vfsp); return (error); } } kmem_strfree(tmp_mntopts); } *vfsp = tmp_vfsp; return (0); } boolean_t zfs_is_readonly(zfsvfs_t *zfsvfs) { return (!!(zfsvfs->z_sb->s_flags & SB_RDONLY)); } int zfs_sync(struct super_block *sb, int wait, cred_t *cr) { (void) cr; zfsvfs_t *zfsvfs = sb->s_fs_info; /* * 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 (zfsvfs != NULL) { /* * Sync a specific filesystem. */ dsl_pool_t *dp; int error; if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); dp = dmu_objset_pool(zfsvfs->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(zfsvfs, FTAG); return (0); } if (zfsvfs->z_log != NULL) zil_commit(zfsvfs->z_log, 0); zfs_exit(zfsvfs, FTAG); } else { /* * Sync all ZFS filesystems. This is what happens when you * run sync(1). Unlike other filesystems, ZFS honors the * request by waiting for all pools to commit all dirty data. */ spa_sync_allpools(); } return (0); } static void atime_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; struct super_block *sb = zfsvfs->z_sb; if (sb == NULL) return; /* * Update SB_NOATIME bit in VFS super block. Since atime update is * determined by atime_needs_update(), atime_needs_update() needs to * return false if atime is turned off, and not unconditionally return * false if atime is turned on. */ if (newval) sb->s_flags &= ~SB_NOATIME; else sb->s_flags |= SB_NOATIME; } static void relatime_changed_cb(void *arg, uint64_t newval) { ((zfsvfs_t *)arg)->z_relatime = newval; } static void xattr_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; if (newval == ZFS_XATTR_OFF) { zfsvfs->z_flags &= ~ZSB_XATTR; } else { zfsvfs->z_flags |= ZSB_XATTR; if (newval == ZFS_XATTR_SA) zfsvfs->z_xattr_sa = B_TRUE; else zfsvfs->z_xattr_sa = B_FALSE; } } static void acltype_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; switch (newval) { case ZFS_ACLTYPE_NFSV4: case ZFS_ACLTYPE_OFF: zfsvfs->z_acl_type = ZFS_ACLTYPE_OFF; zfsvfs->z_sb->s_flags &= ~SB_POSIXACL; break; case ZFS_ACLTYPE_POSIX: #ifdef CONFIG_FS_POSIX_ACL zfsvfs->z_acl_type = ZFS_ACLTYPE_POSIX; zfsvfs->z_sb->s_flags |= SB_POSIXACL; #else zfsvfs->z_acl_type = ZFS_ACLTYPE_OFF; zfsvfs->z_sb->s_flags &= ~SB_POSIXACL; #endif /* CONFIG_FS_POSIX_ACL */ break; default: break; } } static void blksz_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; ASSERT3U(newval, <=, spa_maxblocksize(dmu_objset_spa(zfsvfs->z_os))); ASSERT3U(newval, >=, SPA_MINBLOCKSIZE); ASSERT(ISP2(newval)); zfsvfs->z_max_blksz = newval; } static void readonly_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; struct super_block *sb = zfsvfs->z_sb; if (sb == NULL) return; if (newval) sb->s_flags |= SB_RDONLY; else sb->s_flags &= ~SB_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) { zfsvfs_t *zfsvfs = arg; struct super_block *sb = zfsvfs->z_sb; if (sb == NULL) return; if (newval == TRUE) sb->s_flags |= SB_MANDLOCK; else sb->s_flags &= ~SB_MANDLOCK; } static void snapdir_changed_cb(void *arg, uint64_t newval) { ((zfsvfs_t *)arg)->z_show_ctldir = newval; } static void acl_mode_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; zfsvfs->z_acl_mode = newval; } static void acl_inherit_changed_cb(void *arg, uint64_t newval) { ((zfsvfs_t *)arg)->z_acl_inherit = newval; } static int zfs_register_callbacks(vfs_t *vfsp) { struct dsl_dataset *ds = NULL; objset_t *os = NULL; zfsvfs_t *zfsvfs = NULL; int error = 0; ASSERT(vfsp); zfsvfs = vfsp->vfs_data; ASSERT(zfsvfs); os = zfsvfs->z_os; /* * The act of registering our callbacks will destroy any mount * options we may have. In order to enable temporary overrides * of mount options, we stash away the current values and * restore them after we register the callbacks. */ if (zfs_is_readonly(zfsvfs) || !spa_writeable(dmu_objset_spa(os))) { vfsp->vfs_do_readonly = B_TRUE; vfsp->vfs_readonly = 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); dsl_pool_config_enter(dmu_objset_pool(os), FTAG); error = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ACLMODE), acl_mode_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zfsvfs); dsl_pool_config_exit(dmu_objset_pool(os), FTAG); if (error) goto unregister; /* * Invoke our callbacks to restore temporary mount options. */ if (vfsp->vfs_do_readonly) readonly_changed_cb(zfsvfs, vfsp->vfs_readonly); if (vfsp->vfs_do_setuid) setuid_changed_cb(zfsvfs, vfsp->vfs_setuid); if (vfsp->vfs_do_exec) exec_changed_cb(zfsvfs, vfsp->vfs_exec); if (vfsp->vfs_do_devices) devices_changed_cb(zfsvfs, vfsp->vfs_devices); if (vfsp->vfs_do_xattr) xattr_changed_cb(zfsvfs, vfsp->vfs_xattr); if (vfsp->vfs_do_atime) atime_changed_cb(zfsvfs, vfsp->vfs_atime); if (vfsp->vfs_do_relatime) relatime_changed_cb(zfsvfs, vfsp->vfs_relatime); if (vfsp->vfs_do_nbmand) nbmand_changed_cb(zfsvfs, vfsp->vfs_nbmand); return (0); unregister: dsl_prop_unregister_all(ds, zfsvfs); return (error); } /* * Takes a dataset, a property, a value and that value's setpoint as * found in the ZAP. Checks if the property has been changed in the vfs. * If so, val and setpoint will be overwritten with updated content. * Otherwise, they are left unchanged. */ int zfs_get_temporary_prop(dsl_dataset_t *ds, zfs_prop_t zfs_prop, uint64_t *val, char *setpoint) { int error; zfsvfs_t *zfvp; vfs_t *vfsp; objset_t *os; uint64_t tmp = *val; error = dmu_objset_from_ds(ds, &os); if (error != 0) return (error); if (dmu_objset_type(os) != DMU_OST_ZFS) return (EINVAL); mutex_enter(&os->os_user_ptr_lock); zfvp = dmu_objset_get_user(os); mutex_exit(&os->os_user_ptr_lock); if (zfvp == NULL) return (ESRCH); vfsp = zfvp->z_vfs; switch (zfs_prop) { case ZFS_PROP_ATIME: if (vfsp->vfs_do_atime) tmp = vfsp->vfs_atime; break; case ZFS_PROP_RELATIME: if (vfsp->vfs_do_relatime) tmp = vfsp->vfs_relatime; break; case ZFS_PROP_DEVICES: if (vfsp->vfs_do_devices) tmp = vfsp->vfs_devices; break; case ZFS_PROP_EXEC: if (vfsp->vfs_do_exec) tmp = vfsp->vfs_exec; break; case ZFS_PROP_SETUID: if (vfsp->vfs_do_setuid) tmp = vfsp->vfs_setuid; break; case ZFS_PROP_READONLY: if (vfsp->vfs_do_readonly) tmp = vfsp->vfs_readonly; break; case ZFS_PROP_XATTR: if (vfsp->vfs_do_xattr) tmp = vfsp->vfs_xattr; break; case ZFS_PROP_NBMAND: if (vfsp->vfs_do_nbmand) tmp = vfsp->vfs_nbmand; break; default: return (ENOENT); } if (tmp != *val) { if (setpoint) (void) strcpy(setpoint, "temporary"); *val = tmp; } return (0); } /* * Associate this zfsvfs with the given objset, which must be owned. * This will cache a bunch of on-disk state from the objset in the * zfsvfs. */ static int zfsvfs_init(zfsvfs_t *zfsvfs, objset_t *os) { int error; uint64_t val; zfsvfs->z_max_blksz = SPA_OLD_MAXBLOCKSIZE; zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; zfsvfs->z_os = os; error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version); if (error != 0) return (error); if (zfsvfs->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.\n", (u_longlong_t)zfsvfs->z_version, (u_longlong_t)spa_version(dmu_objset_spa(os))); return (SET_ERROR(ENOTSUP)); } error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &val); if (error != 0) return (error); zfsvfs->z_norm = (int)val; error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &val); if (error != 0) return (error); zfsvfs->z_utf8 = (val != 0); error = zfs_get_zplprop(os, ZFS_PROP_CASE, &val); if (error != 0) return (error); zfsvfs->z_case = (uint_t)val; if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &val)) != 0) return (error); zfsvfs->z_acl_type = (uint_t)val; /* * Fold case on file systems that are always or sometimes case * insensitive. */ if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE || zfsvfs->z_case == ZFS_CASE_MIXED) zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); uint64_t sa_obj = 0; if (zfsvfs->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 != 0) return (error); error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &val); if ((error == 0) && (val == ZFS_XATTR_SA)) zfsvfs->z_xattr_sa = B_TRUE; } error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, &zfsvfs->z_root); if (error != 0) return (error); ASSERT(zfsvfs->z_root != 0); error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, &zfsvfs->z_unlinkedobj); if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], 8, 1, &zfsvfs->z_userquota_obj); if (error == ENOENT) zfsvfs->z_userquota_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], 8, 1, &zfsvfs->z_groupquota_obj); if (error == ENOENT) zfsvfs->z_groupquota_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA], 8, 1, &zfsvfs->z_projectquota_obj); if (error == ENOENT) zfsvfs->z_projectquota_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA], 8, 1, &zfsvfs->z_userobjquota_obj); if (error == ENOENT) zfsvfs->z_userobjquota_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA], 8, 1, &zfsvfs->z_groupobjquota_obj); if (error == ENOENT) zfsvfs->z_groupobjquota_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTOBJQUOTA], 8, 1, &zfsvfs->z_projectobjquota_obj); if (error == ENOENT) zfsvfs->z_projectobjquota_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, &zfsvfs->z_fuid_obj); if (error == ENOENT) zfsvfs->z_fuid_obj = 0; else if (error != 0) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, &zfsvfs->z_shares_dir); if (error == ENOENT) zfsvfs->z_shares_dir = 0; else if (error != 0) return (error); error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, &zfsvfs->z_attr_table); if (error != 0) return (error); if (zfsvfs->z_version >= ZPL_VERSION_SA) sa_register_update_callback(os, zfs_sa_upgrade); return (0); } int zfsvfs_create(const char *osname, boolean_t readonly, zfsvfs_t **zfvp) { objset_t *os; zfsvfs_t *zfsvfs; int error; boolean_t ro = (readonly || (strchr(osname, '@') != NULL)); zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); error = dmu_objset_own(osname, DMU_OST_ZFS, ro, B_TRUE, zfsvfs, &os); if (error != 0) { kmem_free(zfsvfs, sizeof (zfsvfs_t)); return (error); } error = zfsvfs_create_impl(zfvp, zfsvfs, os); return (error); } /* * Note: zfsvfs is assumed to be malloc'd, and will be freed by this function * on a failure. Do not pass in a statically allocated zfsvfs. */ int zfsvfs_create_impl(zfsvfs_t **zfvp, zfsvfs_t *zfsvfs, objset_t *os) { int error; zfsvfs->z_vfs = NULL; zfsvfs->z_sb = NULL; zfsvfs->z_parent = zfsvfs; mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL); list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), offsetof(znode_t, z_link_node)); ZFS_TEARDOWN_INIT(zfsvfs); rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL); int size = MIN(1 << (highbit64(zfs_object_mutex_size) - 1), ZFS_OBJ_MTX_MAX); zfsvfs->z_hold_size = size; zfsvfs->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, KM_SLEEP); zfsvfs->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); for (int i = 0; i != size; i++) { avl_create(&zfsvfs->z_hold_trees[i], zfs_znode_hold_compare, sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); mutex_init(&zfsvfs->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); } error = zfsvfs_init(zfsvfs, os); if (error != 0) { dmu_objset_disown(os, B_TRUE, zfsvfs); *zfvp = NULL; zfsvfs_free(zfsvfs); return (error); } zfsvfs->z_drain_task = TASKQID_INVALID; zfsvfs->z_draining = B_FALSE; zfsvfs->z_drain_cancel = B_TRUE; *zfvp = zfsvfs; return (0); } static int zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) { int error; boolean_t readonly = zfs_is_readonly(zfsvfs); error = zfs_register_callbacks(zfsvfs->z_vfs); if (error) return (error); /* * 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) { ASSERT3P(zfsvfs->z_kstat.dk_kstats, ==, NULL); error = dataset_kstats_create(&zfsvfs->z_kstat, zfsvfs->z_os); if (error) return (error); zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data, &zfsvfs->z_kstat.dk_zil_sums); /* * During replay we remove the read only flag to * allow replays to succeed. */ if (readonly != 0) { readonly_changed_cb(zfsvfs, B_FALSE); } else { zap_stats_t zs; if (zap_get_stats(zfsvfs->z_os, zfsvfs->z_unlinkedobj, &zs) == 0) { dataset_kstats_update_nunlinks_kstat( &zfsvfs->z_kstat, zs.zs_num_entries); dprintf_ds(zfsvfs->z_os->os_dsl_dataset, "num_entries in unlinked set: %llu", zs.zs_num_entries); } zfs_unlinked_drain(zfsvfs); dsl_dir_t *dd = zfsvfs->z_os->os_dsl_dataset->ds_dir; dd->dd_activity_cancelled = B_FALSE; } /* * 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(zfsvfs->z_os))) { if (zil_replay_disable) { zil_destroy(zfsvfs->z_log, B_FALSE); } else { zfsvfs->z_replay = B_TRUE; zil_replay(zfsvfs->z_os, zfsvfs, zfs_replay_vector); zfsvfs->z_replay = B_FALSE; } } /* restore readonly bit */ if (readonly != 0) readonly_changed_cb(zfsvfs, B_TRUE); } else { ASSERT3P(zfsvfs->z_kstat.dk_kstats, !=, NULL); zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data, &zfsvfs->z_kstat.dk_zil_sums); } /* * Set the objset user_ptr to track its zfsvfs. */ mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); dmu_objset_set_user(zfsvfs->z_os, zfsvfs); mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); return (0); } void zfsvfs_free(zfsvfs_t *zfsvfs) { int i, size = zfsvfs->z_hold_size; zfs_fuid_destroy(zfsvfs); mutex_destroy(&zfsvfs->z_znodes_lock); mutex_destroy(&zfsvfs->z_lock); list_destroy(&zfsvfs->z_all_znodes); ZFS_TEARDOWN_DESTROY(zfsvfs); rw_destroy(&zfsvfs->z_teardown_inactive_lock); rw_destroy(&zfsvfs->z_fuid_lock); for (i = 0; i != size; i++) { avl_destroy(&zfsvfs->z_hold_trees[i]); mutex_destroy(&zfsvfs->z_hold_locks[i]); } vmem_free(zfsvfs->z_hold_trees, sizeof (avl_tree_t) * size); vmem_free(zfsvfs->z_hold_locks, sizeof (kmutex_t) * size); zfsvfs_vfs_free(zfsvfs->z_vfs); dataset_kstats_destroy(&zfsvfs->z_kstat); kmem_free(zfsvfs, sizeof (zfsvfs_t)); } static void zfs_set_fuid_feature(zfsvfs_t *zfsvfs) { zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); } static void zfs_unregister_callbacks(zfsvfs_t *zfsvfs) { objset_t *os = zfsvfs->z_os; if (!dmu_objset_is_snapshot(os)) dsl_prop_unregister_all(dmu_objset_ds(os), zfsvfs); } #ifdef HAVE_MLSLABEL /* * 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 (SET_ERROR(EACCES)); return (rdonly ? 0 : SET_ERROR(EACCES)); } return (SET_ERROR(EACCES)); } #endif /* HAVE_MLSLABEL */ static int zfs_statfs_project(zfsvfs_t *zfsvfs, znode_t *zp, struct kstatfs *statp, uint32_t bshift) { char buf[20 + DMU_OBJACCT_PREFIX_LEN]; uint64_t offset = DMU_OBJACCT_PREFIX_LEN; uint64_t quota; uint64_t used; int err; strlcpy(buf, DMU_OBJACCT_PREFIX, DMU_OBJACCT_PREFIX_LEN + 1); err = zfs_id_to_fuidstr(zfsvfs, NULL, zp->z_projid, buf + offset, sizeof (buf) - offset, B_FALSE); if (err) return (err); if (zfsvfs->z_projectquota_obj == 0) goto objs; err = zap_lookup(zfsvfs->z_os, zfsvfs->z_projectquota_obj, buf + offset, 8, 1, "a); if (err == ENOENT) goto objs; else if (err) return (err); err = zap_lookup(zfsvfs->z_os, DMU_PROJECTUSED_OBJECT, buf + offset, 8, 1, &used); if (unlikely(err == ENOENT)) { uint32_t blksize; u_longlong_t nblocks; /* * Quota accounting is async, so it is possible race case. * There is at least one object with the given project ID. */ sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); if (unlikely(zp->z_blksz == 0)) blksize = zfsvfs->z_max_blksz; used = blksize * nblocks; } else if (err) { return (err); } statp->f_blocks = quota >> bshift; statp->f_bfree = (quota > used) ? ((quota - used) >> bshift) : 0; statp->f_bavail = statp->f_bfree; objs: if (zfsvfs->z_projectobjquota_obj == 0) return (0); err = zap_lookup(zfsvfs->z_os, zfsvfs->z_projectobjquota_obj, buf + offset, 8, 1, "a); if (err == ENOENT) return (0); else if (err) return (err); err = zap_lookup(zfsvfs->z_os, DMU_PROJECTUSED_OBJECT, buf, 8, 1, &used); if (unlikely(err == ENOENT)) { /* * Quota accounting is async, so it is possible race case. * There is at least one object with the given project ID. */ used = 1; } else if (err) { return (err); } statp->f_files = quota; statp->f_ffree = (quota > used) ? (quota - used) : 0; return (0); } int zfs_statvfs(struct inode *ip, struct kstatfs *statp) { zfsvfs_t *zfsvfs = ITOZSB(ip); uint64_t refdbytes, availbytes, usedobjs, availobjs; int err = 0; if ((err = zfs_enter(zfsvfs, FTAG)) != 0) return (err); dmu_objset_space(zfsvfs->z_os, &refdbytes, &availbytes, &usedobjs, &availobjs); uint64_t fsid = dmu_objset_fsid_guid(zfsvfs->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 = zfsvfs->z_max_blksz; statp->f_bsize = zfsvfs->z_max_blksz; uint32_t 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. */ /* Round up so we never have a filesystem using 0 blocks. */ refdbytes = P2ROUNDUP(refdbytes, statp->f_bsize); 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 objects 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 = MAXNAMELEN - 1; /* * We have all of 40 characters to stuff a string here. * Is there anything useful we could/should provide? */ memset(statp->f_spare, 0, sizeof (statp->f_spare)); if (dmu_objset_projectquota_enabled(zfsvfs->z_os) && dmu_objset_projectquota_present(zfsvfs->z_os)) { znode_t *zp = ITOZ(ip); if (zp->z_pflags & ZFS_PROJINHERIT && zp->z_projid && zpl_is_valid_projid(zp->z_projid)) err = zfs_statfs_project(zfsvfs, zp, statp, bshift); } zfs_exit(zfsvfs, FTAG); return (err); } static int zfs_root(zfsvfs_t *zfsvfs, struct inode **ipp) { znode_t *rootzp; int error; if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp); if (error == 0) *ipp = ZTOI(rootzp); zfs_exit(zfsvfs, FTAG); return (error); } /* * Linux kernels older than 3.1 do not support a per-filesystem shrinker. * To accommodate this we must improvise and manually walk the list of znodes * attempting to prune dentries in order to be able to drop the inodes. * * To avoid scanning the same znodes multiple times they are always rotated * to the end of the z_all_znodes list. New znodes are inserted at the * end of the list so we're always scanning the oldest znodes first. */ static int zfs_prune_aliases(zfsvfs_t *zfsvfs, unsigned long nr_to_scan) { znode_t **zp_array, *zp; int max_array = MIN(nr_to_scan, PAGE_SIZE * 8 / sizeof (znode_t *)); int objects = 0; int i = 0, j = 0; zp_array = vmem_zalloc(max_array * sizeof (znode_t *), KM_SLEEP); mutex_enter(&zfsvfs->z_znodes_lock); while ((zp = list_head(&zfsvfs->z_all_znodes)) != NULL) { if ((i++ > nr_to_scan) || (j >= max_array)) break; ASSERT(list_link_active(&zp->z_link_node)); list_remove(&zfsvfs->z_all_znodes, zp); list_insert_tail(&zfsvfs->z_all_znodes, zp); /* Skip active znodes and .zfs entries */ if (MUTEX_HELD(&zp->z_lock) || zp->z_is_ctldir) continue; if (igrab(ZTOI(zp)) == NULL) continue; zp_array[j] = zp; j++; } mutex_exit(&zfsvfs->z_znodes_lock); for (i = 0; i < j; i++) { zp = zp_array[i]; ASSERT3P(zp, !=, NULL); d_prune_aliases(ZTOI(zp)); if (atomic_read(&ZTOI(zp)->i_count) == 1) objects++; zrele(zp); } vmem_free(zp_array, max_array * sizeof (znode_t *)); return (objects); } /* * The ARC has requested that the filesystem drop entries from the dentry * and inode caches. This can occur when the ARC needs to free meta data * blocks but can't because they are all pinned by entries in these caches. */ int zfs_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) { zfsvfs_t *zfsvfs = sb->s_fs_info; int error = 0; struct shrinker *shrinker = &sb->s_shrink; struct shrink_control sc = { .nr_to_scan = nr_to_scan, .gfp_mask = GFP_KERNEL, }; if ((error = zfs_enter(zfsvfs, FTAG)) != 0) return (error); #if defined(HAVE_SPLIT_SHRINKER_CALLBACK) && \ defined(SHRINK_CONTROL_HAS_NID) && \ defined(SHRINKER_NUMA_AWARE) if (sb->s_shrink.flags & SHRINKER_NUMA_AWARE) { *objects = 0; for_each_online_node(sc.nid) { *objects += (*shrinker->scan_objects)(shrinker, &sc); /* * reset sc.nr_to_scan, modified by * scan_objects == super_cache_scan */ sc.nr_to_scan = nr_to_scan; } } else { *objects = (*shrinker->scan_objects)(shrinker, &sc); } #elif defined(HAVE_SPLIT_SHRINKER_CALLBACK) *objects = (*shrinker->scan_objects)(shrinker, &sc); #elif defined(HAVE_SINGLE_SHRINKER_CALLBACK) *objects = (*shrinker->shrink)(shrinker, &sc); #elif defined(HAVE_D_PRUNE_ALIASES) #define D_PRUNE_ALIASES_IS_DEFAULT *objects = zfs_prune_aliases(zfsvfs, nr_to_scan); #else #error "No available dentry and inode cache pruning mechanism." #endif #if defined(HAVE_D_PRUNE_ALIASES) && !defined(D_PRUNE_ALIASES_IS_DEFAULT) #undef D_PRUNE_ALIASES_IS_DEFAULT /* * Fall back to zfs_prune_aliases if the kernel's per-superblock * shrinker couldn't free anything, possibly due to the inodes being * allocated in a different memcg. */ if (*objects == 0) *objects = zfs_prune_aliases(zfsvfs, nr_to_scan); #endif zfs_exit(zfsvfs, FTAG); dprintf_ds(zfsvfs->z_os->os_dsl_dataset, "pruning, nr_to_scan=%lu objects=%d error=%d\n", nr_to_scan, *objects, error); return (error); } /* * Teardown the zfsvfs_t. * * Note, if 'unmounting' is FALSE, we return with the 'z_teardown_lock' * and 'z_teardown_inactive_lock' held. */ static int zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting) { znode_t *zp; zfs_unlinked_drain_stop_wait(zfsvfs); /* * If someone has not already unmounted this file system, * drain the zrele_taskq to ensure all active references to the * zfsvfs_t have been handled only then can it be safely destroyed. */ if (zfsvfs->z_os) { /* * If we're unmounting we have to wait for the list to * drain completely. * * If we're not unmounting there's no guarantee the list * will drain completely, but iputs run from the taskq * may add the parents of dir-based xattrs to the taskq * so we want to wait for these. * - * We can safely read z_nr_znodes without locking because the - * VFS has already blocked operations which add to the - * z_all_znodes list and thus increment z_nr_znodes. + * We can safely check z_all_znodes for being empty because the + * VFS has already blocked operations which add to it. */ int round = 0; - while (zfsvfs->z_nr_znodes > 0) { + while (!list_is_empty(&zfsvfs->z_all_znodes)) { taskq_wait_outstanding(dsl_pool_zrele_taskq( dmu_objset_pool(zfsvfs->z_os)), 0); if (++round > 1 && !unmounting) break; } } ZFS_TEARDOWN_ENTER_WRITE(zfsvfs, 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(zfsvfs->z_parent->z_sb); } /* * Close the zil. NB: Can't close the zil while zfs_inactive * threads are blocked as zil_close can call zfs_inactive. */ if (zfsvfs->z_log) { zil_close(zfsvfs->z_log); zfsvfs->z_log = NULL; } rw_enter(&zfsvfs->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 && (zfsvfs->z_unmounted || zfsvfs->z_os == NULL)) { rw_exit(&zfsvfs->z_teardown_inactive_lock); ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); return (SET_ERROR(EIO)); } /* * At this point there are no VFS ops active, and any new VFS ops * will fail with EIO since we have z_teardown_lock for writer (only * relevant for forced unmount). * * Release all holds on dbufs. We also grab an extra reference to all * the remaining inodes so that the kernel does not attempt to free * any inodes of a suspended fs. This can cause deadlocks since the * zfs_resume_fs() process may involve starting threads, which might * attempt to free unreferenced inodes to free up memory for the new * thread. */ if (!unmounting) { mutex_enter(&zfsvfs->z_znodes_lock); for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL; zp = list_next(&zfsvfs->z_all_znodes, zp)) { if (zp->z_sa_hdl) zfs_znode_dmu_fini(zp); if (igrab(ZTOI(zp)) != NULL) zp->z_suspended = B_TRUE; } mutex_exit(&zfsvfs->z_znodes_lock); } /* * If we are unmounting, set the unmounted flag and let new VFS ops * unblock. zfs_inactive will have the unmounted behavior, and all * other VFS ops will fail with EIO. */ if (unmounting) { zfsvfs->z_unmounted = B_TRUE; rw_exit(&zfsvfs->z_teardown_inactive_lock); ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); } /* * z_os will be NULL if there was an error in attempting to reopen * zfsvfs, so just return as the properties had already been * * unregistered and cached data had been evicted before. */ if (zfsvfs->z_os == NULL) return (0); /* * Unregister properties. */ zfs_unregister_callbacks(zfsvfs); /* * Evict cached data. We must write out any dirty data before * disowning the dataset. */ objset_t *os = zfsvfs->z_os; boolean_t os_dirty = B_FALSE; for (int t = 0; t < TXG_SIZE; t++) { if (dmu_objset_is_dirty(os, t)) { os_dirty = B_TRUE; break; } } if (!zfs_is_readonly(zfsvfs) && os_dirty) { txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); } dmu_objset_evict_dbufs(zfsvfs->z_os); dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; dsl_dir_cancel_waiters(dd); return (0); } #if defined(HAVE_SUPER_SETUP_BDI_NAME) atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); #endif int zfs_domount(struct super_block *sb, zfs_mnt_t *zm, int silent) { const char *osname = zm->mnt_osname; struct inode *root_inode = NULL; uint64_t recordsize; int error = 0; zfsvfs_t *zfsvfs = NULL; vfs_t *vfs = NULL; int canwrite; int dataset_visible_zone; ASSERT(zm); ASSERT(osname); dataset_visible_zone = zone_dataset_visible(osname, &canwrite); /* * Refuse to mount a filesystem if we are in a namespace and the * dataset is not visible or writable in that namespace. */ if (!INGLOBALZONE(curproc) && (!dataset_visible_zone || !canwrite)) { return (SET_ERROR(EPERM)); } error = zfsvfs_parse_options(zm->mnt_data, &vfs); if (error) return (error); /* * If a non-writable filesystem is being mounted without the * read-only flag, pretend it was set, as done for snapshots. */ if (!canwrite) vfs->vfs_readonly = true; error = zfsvfs_create(osname, vfs->vfs_readonly, &zfsvfs); if (error) { zfsvfs_vfs_free(vfs); goto out; } if ((error = dsl_prop_get_integer(osname, "recordsize", &recordsize, NULL))) { zfsvfs_vfs_free(vfs); goto out; } vfs->vfs_data = zfsvfs; zfsvfs->z_vfs = vfs; zfsvfs->z_sb = sb; sb->s_fs_info = zfsvfs; 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); error = -zpl_bdi_setup(sb, "zfs"); if (error) goto out; sb->s_bdi->ra_pages = 0; /* 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(zfsvfs); if (dmu_objset_is_snapshot(zfsvfs->z_os)) { uint64_t pval; atime_changed_cb(zfsvfs, B_FALSE); readonly_changed_cb(zfsvfs, B_TRUE); if ((error = dsl_prop_get_integer(osname, "xattr", &pval, NULL))) goto out; xattr_changed_cb(zfsvfs, pval); if ((error = dsl_prop_get_integer(osname, "acltype", &pval, NULL))) goto out; acltype_changed_cb(zfsvfs, pval); zfsvfs->z_issnap = B_TRUE; zfsvfs->z_os->os_sync = ZFS_SYNC_DISABLED; zfsvfs->z_snap_defer_time = jiffies; mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); dmu_objset_set_user(zfsvfs->z_os, zfsvfs); mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); } else { if ((error = zfsvfs_setup(zfsvfs, B_TRUE))) goto out; } /* Allocate a root inode for the filesystem. */ error = zfs_root(zfsvfs, &root_inode); if (error) { (void) zfs_umount(sb); zfsvfs = NULL; /* avoid double-free; first in zfs_umount */ 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); zfsvfs = NULL; /* avoid double-free; first in zfs_umount */ error = SET_ERROR(ENOMEM); goto out; } if (!zfsvfs->z_issnap) zfsctl_create(zfsvfs); zfsvfs->z_arc_prune = arc_add_prune_callback(zpl_prune_sb, sb); out: if (error) { if (zfsvfs != NULL) { dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs); zfsvfs_free(zfsvfs); } /* * make sure we don't have dangling sb->s_fs_info which * zfs_preumount will use. */ sb->s_fs_info = NULL; } return (error); } /* * 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) { zfsvfs_t *zfsvfs = sb->s_fs_info; /* zfsvfs is NULL when zfs_domount fails during mount */ if (zfsvfs) { zfs_unlinked_drain_stop_wait(zfsvfs); zfsctl_destroy(sb->s_fs_info); /* * Wait for zrele_async before entering evict_inodes in * generic_shutdown_super. The reason we must finish before * evict_inodes is when lazytime is on, or when zfs_purgedir * calls zfs_zget, zrele would bump i_count from 0 to 1. This * would race with the i_count check in evict_inodes. This means * it could destroy the inode while we are still using it. * * We wait for two passes. xattr directories in the first pass * may add xattr entries in zfs_purgedir, so in the second pass * we wait for them. We don't use taskq_wait here because it is * a pool wide taskq. Other mounted filesystems can constantly * do zrele_async and there's no guarantee when taskq will be * empty. */ taskq_wait_outstanding(dsl_pool_zrele_taskq( dmu_objset_pool(zfsvfs->z_os)), 0); taskq_wait_outstanding(dsl_pool_zrele_taskq( dmu_objset_pool(zfsvfs->z_os)), 0); } } /* * Called once all other unmount released tear down has occurred. * It is our responsibility to release any remaining infrastructure. */ int zfs_umount(struct super_block *sb) { zfsvfs_t *zfsvfs = sb->s_fs_info; objset_t *os; if (zfsvfs->z_arc_prune != NULL) arc_remove_prune_callback(zfsvfs->z_arc_prune); VERIFY(zfsvfs_teardown(zfsvfs, B_TRUE) == 0); os = zfsvfs->z_os; zpl_bdi_destroy(sb); /* * z_os will be NULL if there was an error in * attempting to reopen zfsvfs. */ 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, B_TRUE, zfsvfs); } zfsvfs_free(zfsvfs); sb->s_fs_info = NULL; return (0); } int zfs_remount(struct super_block *sb, int *flags, zfs_mnt_t *zm) { zfsvfs_t *zfsvfs = sb->s_fs_info; vfs_t *vfsp; boolean_t issnap = dmu_objset_is_snapshot(zfsvfs->z_os); int error; if ((issnap || !spa_writeable(dmu_objset_spa(zfsvfs->z_os))) && !(*flags & SB_RDONLY)) { *flags |= SB_RDONLY; return (EROFS); } error = zfsvfs_parse_options(zm->mnt_data, &vfsp); if (error) return (error); if (!zfs_is_readonly(zfsvfs) && (*flags & SB_RDONLY)) txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); zfs_unregister_callbacks(zfsvfs); zfsvfs_vfs_free(zfsvfs->z_vfs); vfsp->vfs_data = zfsvfs; zfsvfs->z_vfs = vfsp; if (!issnap) (void) zfs_register_callbacks(vfsp); return (error); } int zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) { zfsvfs_t *zfsvfs = 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; 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 { return (SET_ERROR(EINVAL)); } /* LONG_FID_LEN means snapdirs */ 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); if (objsetid != ZFSCTL_INO_SNAPDIRS - object) { dprintf("snapdir fid: objsetid (%llu) != " "ZFSCTL_INO_SNAPDIRS (%llu) - object (%llu)\n", objsetid, ZFSCTL_INO_SNAPDIRS, object); return (SET_ERROR(EINVAL)); } if (fid_gen > 1 || setgen != 0) { dprintf("snapdir fid: fid_gen (%llu) and setgen " "(%llu)\n", fid_gen, setgen); return (SET_ERROR(EINVAL)); } return (zfsctl_snapdir_vget(sb, objsetid, fid_gen, ipp)); } if ((err = zfs_enter(zfsvfs, FTAG)) != 0) return (err); /* 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 = zfsvfs->z_ctldir; ASSERT(*ipp != NULL); if (object == ZFSCTL_INO_SNAPDIR) { VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, 0, kcred, NULL, NULL) == 0); } else { /* * Must have an existing ref, so igrab() * cannot return NULL */ VERIFY3P(igrab(*ipp), !=, NULL); } zfs_exit(zfsvfs, FTAG); return (0); } gen_mask = -1ULL >> (64 - 8 * i); dprintf("getting %llu [%llu mask %llx]\n", object, fid_gen, gen_mask); if ((err = zfs_zget(zfsvfs, object, &zp))) { zfs_exit(zfsvfs, FTAG); return (err); } /* Don't export xattr stuff */ if (zp->z_pflags & ZFS_XATTR) { zrele(zp); zfs_exit(zfsvfs, FTAG); return (SET_ERROR(ENOENT)); } (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &zp_gen, sizeof (uint64_t)); zp_gen = zp_gen & gen_mask; if (zp_gen == 0) zp_gen = 1; if ((fid_gen == 0) && (zfsvfs->z_root == object)) fid_gen = zp_gen; if (zp->z_unlinked || zp_gen != fid_gen) { dprintf("znode gen (%llu) != fid gen (%llu)\n", zp_gen, fid_gen); zrele(zp); zfs_exit(zfsvfs, FTAG); return (SET_ERROR(ENOENT)); } *ipp = ZTOI(zp); if (*ipp) zfs_znode_update_vfs(ITOZ(*ipp)); zfs_exit(zfsvfs, FTAG); return (0); } /* * Block out VFS ops and close zfsvfs_t * * Note, if successful, then we return with the 'z_teardown_lock' and * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying * dataset and objset intact so that they can be atomically handed off during * a subsequent rollback or recv operation and the resume thereafter. */ int zfs_suspend_fs(zfsvfs_t *zfsvfs) { int error; if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0) return (error); return (0); } /* * Rebuild SA and release VOPs. Note that ownership of the underlying dataset * is an invariant across any of the operations that can be performed while the * filesystem was suspended. Whether it succeeded or failed, the preconditions * are the same: the relevant objset and associated dataset are owned by * zfsvfs, held, and long held on entry. */ int zfs_resume_fs(zfsvfs_t *zfsvfs, dsl_dataset_t *ds) { int err, err2; znode_t *zp; ASSERT(ZFS_TEARDOWN_WRITE_HELD(zfsvfs)); ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)); /* * We already own this, so just update the objset_t, as the one we * had before may have been evicted. */ objset_t *os; VERIFY3P(ds->ds_owner, ==, zfsvfs); VERIFY(dsl_dataset_long_held(ds)); dsl_pool_t *dp = spa_get_dsl(dsl_dataset_get_spa(ds)); dsl_pool_config_enter(dp, FTAG); VERIFY0(dmu_objset_from_ds(ds, &os)); dsl_pool_config_exit(dp, FTAG); err = zfsvfs_init(zfsvfs, os); if (err != 0) goto bail; ds->ds_dir->dd_activity_cancelled = B_FALSE; VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0); zfs_set_fuid_feature(zfsvfs); zfsvfs->z_rollback_time = jiffies; /* * Attempt to re-establish all the active inodes with their * dbufs. If a zfs_rezget() fails, then we unhash the inode * and mark it stale. This prevents a collision if a new * inode/object is created which must use the same inode * number. The stale inode will be be released when the * VFS prunes the dentry holding the remaining references * on the stale inode. */ mutex_enter(&zfsvfs->z_znodes_lock); for (zp = list_head(&zfsvfs->z_all_znodes); zp; zp = list_next(&zfsvfs->z_all_znodes, zp)) { err2 = zfs_rezget(zp); if (err2) { zpl_d_drop_aliases(ZTOI(zp)); remove_inode_hash(ZTOI(zp)); } /* see comment in zfs_suspend_fs() */ if (zp->z_suspended) { zfs_zrele_async(zp); zp->z_suspended = B_FALSE; } } mutex_exit(&zfsvfs->z_znodes_lock); if (!zfs_is_readonly(zfsvfs) && !zfsvfs->z_unmounted) { /* * zfs_suspend_fs() could have interrupted freeing * of dnodes. We need to restart this freeing so * that we don't "leak" the space. */ zfs_unlinked_drain(zfsvfs); } /* * Most of the time zfs_suspend_fs is used for changing the contents * of the underlying dataset. ZFS rollback and receive operations * might create files for which negative dentries are present in * the cache. Since walking the dcache would require a lot of GPL-only * code duplication, it's much easier on these rather rare occasions * just to flush the whole dcache for the given dataset/filesystem. */ shrink_dcache_sb(zfsvfs->z_sb); bail: if (err != 0) zfsvfs->z_unmounted = B_TRUE; /* release the VFS ops */ rw_exit(&zfsvfs->z_teardown_inactive_lock); ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); if (err != 0) { /* * Since we couldn't setup the sa framework, try to force * unmount this file system. */ if (zfsvfs->z_os) (void) zfs_umount(zfsvfs->z_sb); } return (err); } /* * Release VOPs and unmount a suspended filesystem. */ int zfs_end_fs(zfsvfs_t *zfsvfs, dsl_dataset_t *ds) { ASSERT(ZFS_TEARDOWN_WRITE_HELD(zfsvfs)); ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)); /* * We already own this, so just hold and rele it to update the * objset_t, as the one we had before may have been evicted. */ objset_t *os; VERIFY3P(ds->ds_owner, ==, zfsvfs); VERIFY(dsl_dataset_long_held(ds)); dsl_pool_t *dp = spa_get_dsl(dsl_dataset_get_spa(ds)); dsl_pool_config_enter(dp, FTAG); VERIFY0(dmu_objset_from_ds(ds, &os)); dsl_pool_config_exit(dp, FTAG); zfsvfs->z_os = os; /* release the VOPs */ rw_exit(&zfsvfs->z_teardown_inactive_lock); ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); /* * Try to force unmount this file system. */ (void) zfs_umount(zfsvfs->z_sb); zfsvfs->z_unmounted = B_TRUE; return (0); } /* * Automounted snapshots rely on periodic revalidation * to defer snapshots from being automatically unmounted. */ inline void zfs_exit_fs(zfsvfs_t *zfsvfs) { if (!zfsvfs->z_issnap) return; if (time_after(jiffies, zfsvfs->z_snap_defer_time + MAX(zfs_expire_snapshot * HZ / 2, HZ))) { zfsvfs->z_snap_defer_time = jiffies; zfsctl_snapshot_unmount_delay(zfsvfs->z_os->os_spa, dmu_objset_id(zfsvfs->z_os), zfs_expire_snapshot); } } int zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers) { int error; objset_t *os = zfsvfs->z_os; dmu_tx_t *tx; if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) return (SET_ERROR(EINVAL)); if (newvers < zfsvfs->z_version) return (SET_ERROR(EINVAL)); if (zfs_spa_version_map(newvers) > spa_version(dmu_objset_spa(zfsvfs->z_os))) return (SET_ERROR(ENOTSUP)); tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); if (newvers >= ZPL_VERSION_SA && !zfsvfs->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 && !zfsvfs->z_use_sa) { uint64_t sa_obj; ASSERT3U(spa_version(dmu_objset_spa(zfsvfs->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); ASSERT0(error); VERIFY(0 == sa_set_sa_object(os, sa_obj)); sa_register_update_callback(os, zfs_sa_upgrade); } spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx, "from %llu to %llu", zfsvfs->z_version, newvers); dmu_tx_commit(tx); zfsvfs->z_version = newvers; os->os_version = newvers; zfs_set_fuid_feature(zfsvfs); return (0); } /* * Return true if the corresponding vfs's unmounted flag is set. * Otherwise return false. * If this function returns true we know VFS unmount has been initiated. */ boolean_t zfs_get_vfs_flag_unmounted(objset_t *os) { zfsvfs_t *zfvp; boolean_t unmounted = B_FALSE; ASSERT(dmu_objset_type(os) == DMU_OST_ZFS); mutex_enter(&os->os_user_ptr_lock); zfvp = dmu_objset_get_user(os); if (zfvp != NULL && zfvp->z_unmounted) unmounted = B_TRUE; mutex_exit(&os->os_user_ptr_lock); return (unmounted); } void zfsvfs_update_fromname(const char *oldname, const char *newname) { /* * We don't need to do anything here, the devname is always current by * virtue of zfsvfs->z_sb->s_op->show_devname. */ (void) oldname, (void) newname; } void zfs_init(void) { zfsctl_init(); zfs_znode_init(); dmu_objset_register_type(DMU_OST_ZFS, zpl_get_file_info); register_filesystem(&zpl_fs_type); #ifdef HAVE_VFS_FILE_OPERATIONS_EXTEND register_fo_extend(&zpl_file_operations); #endif } void zfs_fini(void) { /* * we don't use outstanding because zpl_posix_acl_free might add more. */ taskq_wait(system_delay_taskq); taskq_wait(system_taskq); #ifdef HAVE_VFS_FILE_OPERATIONS_EXTEND unregister_fo_extend(&zpl_file_operations); #endif unregister_filesystem(&zpl_fs_type); zfs_znode_fini(); zfsctl_fini(); } #if defined(_KERNEL) EXPORT_SYMBOL(zfs_suspend_fs); EXPORT_SYMBOL(zfs_resume_fs); EXPORT_SYMBOL(zfs_set_version); EXPORT_SYMBOL(zfsvfs_create); EXPORT_SYMBOL(zfsvfs_free); EXPORT_SYMBOL(zfs_is_readonly); EXPORT_SYMBOL(zfs_domount); EXPORT_SYMBOL(zfs_preumount); EXPORT_SYMBOL(zfs_umount); EXPORT_SYMBOL(zfs_remount); EXPORT_SYMBOL(zfs_statvfs); EXPORT_SYMBOL(zfs_vget); EXPORT_SYMBOL(zfs_prune); #endif diff --git a/module/os/linux/zfs/zfs_znode.c b/module/os/linux/zfs/zfs_znode.c index 335ae3460c58..52c8e51df659 100644 --- a/module/os/linux/zfs/zfs_znode.c +++ b/module/os/linux/zfs/zfs_znode.c @@ -1,2360 +1,2358 @@ /* * 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 https://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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2018 by Delphix. All rights reserved. */ /* Portions Copyright 2007 Jeremy Teo */ #ifdef _KERNEL #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #endif /* _KERNEL */ #include #include #include #include #include #include #include #include #include #include #include "zfs_prop.h" #include "zfs_comutil.h" /* * Functions needed for userland (ie: libzpool) are not put under * #ifdef_KERNEL; the rest of the functions have dependencies * (such as VFS logic) that will not compile easily in userland. */ #ifdef _KERNEL static kmem_cache_t *znode_cache = NULL; static kmem_cache_t *znode_hold_cache = NULL; unsigned int zfs_object_mutex_size = ZFS_OBJ_MTX_SZ; /* * This is used by the test suite so that it can delay znodes from being * freed in order to inspect the unlinked set. */ static int zfs_unlink_suspend_progress = 0; /* * This callback is invoked when acquiring a RL_WRITER or RL_APPEND lock on * z_rangelock. It will modify the offset and length of the lock to reflect * znode-specific information, and convert RL_APPEND to RL_WRITER. This is * called with the rangelock_t's rl_lock held, which avoids races. */ static void zfs_rangelock_cb(zfs_locked_range_t *new, void *arg) { znode_t *zp = arg; /* * If in append mode, convert to writer and lock starting at the * current end of file. */ if (new->lr_type == RL_APPEND) { new->lr_offset = zp->z_size; new->lr_type = RL_WRITER; } /* * If we need to grow the block size then lock the whole file range. */ uint64_t end_size = MAX(zp->z_size, new->lr_offset + new->lr_length); if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) || zp->z_blksz < ZTOZSB(zp)->z_max_blksz)) { new->lr_offset = 0; new->lr_length = UINT64_MAX; } } static int zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) { (void) arg, (void) kmflags; znode_t *zp = buf; inode_init_once(ZTOI(zp)); list_link_init(&zp->z_link_node); mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); rw_init(&zp->z_name_lock, NULL, RW_NOLOCKDEP, NULL); mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL); zfs_rangelock_init(&zp->z_rangelock, zfs_rangelock_cb, zp); zp->z_dirlocks = NULL; zp->z_acl_cached = NULL; zp->z_xattr_cached = NULL; zp->z_xattr_parent = 0; zp->z_sync_writes_cnt = 0; zp->z_async_writes_cnt = 0; return (0); } static void zfs_znode_cache_destructor(void *buf, void *arg) { (void) arg; znode_t *zp = buf; ASSERT(!list_link_active(&zp->z_link_node)); mutex_destroy(&zp->z_lock); rw_destroy(&zp->z_parent_lock); rw_destroy(&zp->z_name_lock); mutex_destroy(&zp->z_acl_lock); rw_destroy(&zp->z_xattr_lock); zfs_rangelock_fini(&zp->z_rangelock); ASSERT3P(zp->z_dirlocks, ==, NULL); ASSERT3P(zp->z_acl_cached, ==, NULL); ASSERT3P(zp->z_xattr_cached, ==, NULL); ASSERT0(atomic_load_32(&zp->z_sync_writes_cnt)); ASSERT0(atomic_load_32(&zp->z_async_writes_cnt)); } static int zfs_znode_hold_cache_constructor(void *buf, void *arg, int kmflags) { (void) arg, (void) kmflags; znode_hold_t *zh = buf; mutex_init(&zh->zh_lock, NULL, MUTEX_DEFAULT, NULL); zh->zh_refcount = 0; return (0); } static void zfs_znode_hold_cache_destructor(void *buf, void *arg) { (void) arg; znode_hold_t *zh = buf; mutex_destroy(&zh->zh_lock); } void zfs_znode_init(void) { /* * Initialize zcache. The KMC_SLAB hint is used in order that it be * backed by kmalloc() when on the Linux slab in order that any * wait_on_bit() operations on the related inode operate properly. */ ASSERT(znode_cache == NULL); znode_cache = kmem_cache_create("zfs_znode_cache", sizeof (znode_t), 0, zfs_znode_cache_constructor, zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_SLAB); ASSERT(znode_hold_cache == NULL); znode_hold_cache = kmem_cache_create("zfs_znode_hold_cache", sizeof (znode_hold_t), 0, zfs_znode_hold_cache_constructor, zfs_znode_hold_cache_destructor, NULL, NULL, NULL, 0); } void zfs_znode_fini(void) { /* * Cleanup zcache */ if (znode_cache) kmem_cache_destroy(znode_cache); znode_cache = NULL; if (znode_hold_cache) kmem_cache_destroy(znode_hold_cache); znode_hold_cache = NULL; } /* * The zfs_znode_hold_enter() / zfs_znode_hold_exit() functions are used to * serialize access to a znode and its SA buffer while the object is being * created or destroyed. This kind of locking would normally reside in the * znode itself but in this case that's impossible because the znode and SA * buffer may not yet exist. Therefore the locking is handled externally * with an array of mutexes and AVLs trees which contain per-object locks. * * In zfs_znode_hold_enter() a per-object lock is created as needed, inserted * in to the correct AVL tree and finally the per-object lock is held. In * zfs_znode_hold_exit() the process is reversed. The per-object lock is * released, removed from the AVL tree and destroyed if there are no waiters. * * This scheme has two important properties: * * 1) No memory allocations are performed while holding one of the z_hold_locks. * This ensures evict(), which can be called from direct memory reclaim, will * never block waiting on a z_hold_locks which just happens to have hashed * to the same index. * * 2) All locks used to serialize access to an object are per-object and never * shared. This minimizes lock contention without creating a large number * of dedicated locks. * * On the downside it does require znode_lock_t structures to be frequently * allocated and freed. However, because these are backed by a kmem cache * and very short lived this cost is minimal. */ int zfs_znode_hold_compare(const void *a, const void *b) { const znode_hold_t *zh_a = (const znode_hold_t *)a; const znode_hold_t *zh_b = (const znode_hold_t *)b; return (TREE_CMP(zh_a->zh_obj, zh_b->zh_obj)); } static boolean_t __maybe_unused zfs_znode_held(zfsvfs_t *zfsvfs, uint64_t obj) { znode_hold_t *zh, search; int i = ZFS_OBJ_HASH(zfsvfs, obj); boolean_t held; search.zh_obj = obj; mutex_enter(&zfsvfs->z_hold_locks[i]); zh = avl_find(&zfsvfs->z_hold_trees[i], &search, NULL); held = (zh && MUTEX_HELD(&zh->zh_lock)) ? B_TRUE : B_FALSE; mutex_exit(&zfsvfs->z_hold_locks[i]); return (held); } znode_hold_t * zfs_znode_hold_enter(zfsvfs_t *zfsvfs, uint64_t obj) { znode_hold_t *zh, *zh_new, search; int i = ZFS_OBJ_HASH(zfsvfs, obj); boolean_t found = B_FALSE; zh_new = kmem_cache_alloc(znode_hold_cache, KM_SLEEP); search.zh_obj = obj; mutex_enter(&zfsvfs->z_hold_locks[i]); zh = avl_find(&zfsvfs->z_hold_trees[i], &search, NULL); if (likely(zh == NULL)) { zh = zh_new; zh->zh_obj = obj; avl_add(&zfsvfs->z_hold_trees[i], zh); } else { ASSERT3U(zh->zh_obj, ==, obj); found = B_TRUE; } zh->zh_refcount++; ASSERT3S(zh->zh_refcount, >, 0); mutex_exit(&zfsvfs->z_hold_locks[i]); if (found == B_TRUE) kmem_cache_free(znode_hold_cache, zh_new); ASSERT(MUTEX_NOT_HELD(&zh->zh_lock)); mutex_enter(&zh->zh_lock); return (zh); } void zfs_znode_hold_exit(zfsvfs_t *zfsvfs, znode_hold_t *zh) { int i = ZFS_OBJ_HASH(zfsvfs, zh->zh_obj); boolean_t remove = B_FALSE; ASSERT(zfs_znode_held(zfsvfs, zh->zh_obj)); mutex_exit(&zh->zh_lock); mutex_enter(&zfsvfs->z_hold_locks[i]); ASSERT3S(zh->zh_refcount, >, 0); if (--zh->zh_refcount == 0) { avl_remove(&zfsvfs->z_hold_trees[i], zh); remove = B_TRUE; } mutex_exit(&zfsvfs->z_hold_locks[i]); if (remove == B_TRUE) kmem_cache_free(znode_hold_cache, zh); } dev_t zfs_cmpldev(uint64_t dev) { return (dev); } static void zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp, dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) { ASSERT(zfs_znode_held(zfsvfs, zp->z_id)); mutex_enter(&zp->z_lock); ASSERT(zp->z_sa_hdl == NULL); ASSERT(zp->z_acl_cached == NULL); if (sa_hdl == NULL) { VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp, SA_HDL_SHARED, &zp->z_sa_hdl)); } else { zp->z_sa_hdl = sa_hdl; sa_set_userp(sa_hdl, zp); } zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; mutex_exit(&zp->z_lock); } void zfs_znode_dmu_fini(znode_t *zp) { ASSERT(zfs_znode_held(ZTOZSB(zp), zp->z_id) || RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock)); sa_handle_destroy(zp->z_sa_hdl); zp->z_sa_hdl = NULL; } /* * Called by new_inode() to allocate a new inode. */ int zfs_inode_alloc(struct super_block *sb, struct inode **ip) { znode_t *zp; zp = kmem_cache_alloc(znode_cache, KM_SLEEP); *ip = ZTOI(zp); return (0); } /* * Called in multiple places when an inode should be destroyed. */ void zfs_inode_destroy(struct inode *ip) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ZTOZSB(zp); mutex_enter(&zfsvfs->z_znodes_lock); if (list_link_active(&zp->z_link_node)) { list_remove(&zfsvfs->z_all_znodes, zp); - zfsvfs->z_nr_znodes--; } mutex_exit(&zfsvfs->z_znodes_lock); if (zp->z_acl_cached) { zfs_acl_free(zp->z_acl_cached); zp->z_acl_cached = NULL; } if (zp->z_xattr_cached) { nvlist_free(zp->z_xattr_cached); zp->z_xattr_cached = NULL; } kmem_cache_free(znode_cache, zp); } static void zfs_inode_set_ops(zfsvfs_t *zfsvfs, struct inode *ip) { uint64_t rdev = 0; switch (ip->i_mode & S_IFMT) { case S_IFREG: ip->i_op = &zpl_inode_operations; #ifdef HAVE_VFS_FILE_OPERATIONS_EXTEND ip->i_fop = &zpl_file_operations.kabi_fops; #else ip->i_fop = &zpl_file_operations; #endif ip->i_mapping->a_ops = &zpl_address_space_operations; break; case S_IFDIR: #ifdef HAVE_RENAME2_OPERATIONS_WRAPPER ip->i_flags |= S_IOPS_WRAPPER; ip->i_op = &zpl_dir_inode_operations.ops; #else ip->i_op = &zpl_dir_inode_operations; #endif ip->i_fop = &zpl_dir_file_operations; ITOZ(ip)->z_zn_prefetch = B_TRUE; break; case S_IFLNK: ip->i_op = &zpl_symlink_inode_operations; break; /* * rdev is only stored in a SA only for device files. */ case S_IFCHR: case S_IFBLK: (void) sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zfsvfs), &rdev, sizeof (rdev)); zfs_fallthrough; case S_IFIFO: case S_IFSOCK: init_special_inode(ip, ip->i_mode, rdev); ip->i_op = &zpl_special_inode_operations; break; default: zfs_panic_recover("inode %llu has invalid mode: 0x%x\n", (u_longlong_t)ip->i_ino, ip->i_mode); /* Assume the inode is a file and attempt to continue */ ip->i_mode = S_IFREG | 0644; ip->i_op = &zpl_inode_operations; #ifdef HAVE_VFS_FILE_OPERATIONS_EXTEND ip->i_fop = &zpl_file_operations.kabi_fops; #else ip->i_fop = &zpl_file_operations; #endif ip->i_mapping->a_ops = &zpl_address_space_operations; break; } } static void zfs_set_inode_flags(znode_t *zp, struct inode *ip) { /* * Linux and Solaris have different sets of file attributes, so we * restrict this conversion to the intersection of the two. */ #ifdef HAVE_INODE_SET_FLAGS unsigned int flags = 0; if (zp->z_pflags & ZFS_IMMUTABLE) flags |= S_IMMUTABLE; if (zp->z_pflags & ZFS_APPENDONLY) flags |= S_APPEND; inode_set_flags(ip, flags, S_IMMUTABLE|S_APPEND); #else if (zp->z_pflags & ZFS_IMMUTABLE) ip->i_flags |= S_IMMUTABLE; else ip->i_flags &= ~S_IMMUTABLE; if (zp->z_pflags & ZFS_APPENDONLY) ip->i_flags |= S_APPEND; else ip->i_flags &= ~S_APPEND; #endif } /* * Update the embedded inode given the znode. */ void zfs_znode_update_vfs(znode_t *zp) { struct inode *ip; uint32_t blksize; u_longlong_t i_blocks; ASSERT(zp != NULL); ip = ZTOI(zp); /* Skip .zfs control nodes which do not exist on disk. */ if (zfsctl_is_node(ip)) return; dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize, &i_blocks); spin_lock(&ip->i_lock); ip->i_mode = zp->z_mode; ip->i_blocks = i_blocks; i_size_write(ip, zp->z_size); spin_unlock(&ip->i_lock); } /* * Construct a znode+inode and initialize. * * This does not do a call to dmu_set_user() that is * up to the caller to do, in case you don't want to * return the znode */ static znode_t * zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, dmu_object_type_t obj_type, sa_handle_t *hdl) { znode_t *zp; struct inode *ip; uint64_t mode; uint64_t parent; uint64_t tmp_gen; uint64_t links; uint64_t z_uid, z_gid; uint64_t atime[2], mtime[2], ctime[2], btime[2]; uint64_t projid = ZFS_DEFAULT_PROJID; sa_bulk_attr_t bulk[12]; int count = 0; ASSERT(zfsvfs != NULL); ip = new_inode(zfsvfs->z_sb); if (ip == NULL) return (NULL); zp = ITOZ(ip); ASSERT(zp->z_dirlocks == NULL); ASSERT3P(zp->z_acl_cached, ==, NULL); ASSERT3P(zp->z_xattr_cached, ==, NULL); zp->z_unlinked = B_FALSE; zp->z_atime_dirty = B_FALSE; #if !defined(HAVE_FILEMAP_RANGE_HAS_PAGE) zp->z_is_mapped = B_FALSE; #endif zp->z_is_ctldir = B_FALSE; zp->z_suspended = B_FALSE; zp->z_sa_hdl = NULL; zp->z_mapcnt = 0; zp->z_id = db->db_object; zp->z_blksz = blksz; zp->z_seq = 0x7A4653; zp->z_sync_cnt = 0; zp->z_sync_writes_cnt = 0; zp->z_async_writes_cnt = 0; zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &tmp_gen, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, &zp->z_size, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &z_uid, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &z_gid, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &btime, 16); if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || tmp_gen == 0 || (dmu_objset_projectquota_enabled(zfsvfs->z_os) && (zp->z_pflags & ZFS_PROJID) && sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), &projid, 8) != 0)) { if (hdl == NULL) sa_handle_destroy(zp->z_sa_hdl); zp->z_sa_hdl = NULL; goto error; } zp->z_projid = projid; zp->z_mode = ip->i_mode = mode; ip->i_generation = (uint32_t)tmp_gen; ip->i_blkbits = SPA_MINBLOCKSHIFT; set_nlink(ip, (uint32_t)links); zfs_uid_write(ip, z_uid); zfs_gid_write(ip, z_gid); zfs_set_inode_flags(zp, ip); /* Cache the xattr parent id */ if (zp->z_pflags & ZFS_XATTR) zp->z_xattr_parent = parent; ZFS_TIME_DECODE(&ip->i_atime, atime); ZFS_TIME_DECODE(&ip->i_mtime, mtime); ZFS_TIME_DECODE(&ip->i_ctime, ctime); ZFS_TIME_DECODE(&zp->z_btime, btime); ip->i_ino = zp->z_id; zfs_znode_update_vfs(zp); zfs_inode_set_ops(zfsvfs, ip); /* * The only way insert_inode_locked() can fail is if the ip->i_ino * number is already hashed for this super block. This can never * happen because the inode numbers map 1:1 with the object numbers. * * Exceptions include rolling back a mounted file system, either * from the zfs rollback or zfs recv command. * * Active inodes are unhashed during the rollback, but since zrele * can happen asynchronously, we can't guarantee they've been * unhashed. This can cause hash collisions in unlinked drain * processing so do not hash unlinked znodes. */ if (links > 0) VERIFY3S(insert_inode_locked(ip), ==, 0); mutex_enter(&zfsvfs->z_znodes_lock); list_insert_tail(&zfsvfs->z_all_znodes, zp); - zfsvfs->z_nr_znodes++; mutex_exit(&zfsvfs->z_znodes_lock); if (links > 0) unlock_new_inode(ip); return (zp); error: iput(ip); return (NULL); } /* * Safely mark an inode dirty. Inodes which are part of a read-only * file system or snapshot may not be dirtied. */ void zfs_mark_inode_dirty(struct inode *ip) { zfsvfs_t *zfsvfs = ITOZSB(ip); if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) return; mark_inode_dirty(ip); } static uint64_t empty_xattr; static uint64_t pad[4]; static zfs_acl_phys_t acl_phys; /* * Create a new DMU object to hold a zfs znode. * * IN: dzp - parent directory for new znode * vap - file attributes for new znode * tx - dmu transaction id for zap operations * cr - credentials of caller * flag - flags: * IS_ROOT_NODE - new object will be root * IS_TMPFILE - new object is of O_TMPFILE * IS_XATTR - new object is an attribute * acl_ids - ACL related attributes * * OUT: zpp - allocated znode (set to dzp if IS_ROOT_NODE) * */ void zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) { uint64_t crtime[2], atime[2], mtime[2], ctime[2]; uint64_t mode, size, links, parent, pflags; uint64_t projid = ZFS_DEFAULT_PROJID; uint64_t rdev = 0; zfsvfs_t *zfsvfs = ZTOZSB(dzp); dmu_buf_t *db; inode_timespec_t now; uint64_t gen, obj; int bonuslen; int dnodesize; sa_handle_t *sa_hdl; dmu_object_type_t obj_type; sa_bulk_attr_t *sa_attrs; int cnt = 0; zfs_acl_locator_cb_t locate = { 0 }; znode_hold_t *zh; if (zfsvfs->z_replay) { obj = vap->va_nodeid; now = vap->va_ctime; /* see zfs_replay_create() */ gen = vap->va_nblocks; /* ditto */ dnodesize = vap->va_fsid; /* ditto */ } else { obj = 0; gethrestime(&now); gen = dmu_tx_get_txg(tx); dnodesize = dmu_objset_dnodesize(zfsvfs->z_os); } if (dnodesize == 0) dnodesize = DNODE_MIN_SIZE; obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; bonuslen = (obj_type == DMU_OT_SA) ? DN_BONUS_SIZE(dnodesize) : ZFS_OLD_ZNODE_PHYS_SIZE; /* * Create a new DMU object. */ /* * There's currently no mechanism for pre-reading the blocks that will * be needed to allocate a new object, so we accept the small chance * that there will be an i/o error and we will fail one of the * assertions below. */ if (S_ISDIR(vap->va_mode)) { if (zfsvfs->z_replay) { VERIFY0(zap_create_claim_norm_dnsize(zfsvfs->z_os, obj, zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, obj_type, bonuslen, dnodesize, tx)); } else { obj = zap_create_norm_dnsize(zfsvfs->z_os, zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, obj_type, bonuslen, dnodesize, tx); } } else { if (zfsvfs->z_replay) { VERIFY0(dmu_object_claim_dnsize(zfsvfs->z_os, obj, DMU_OT_PLAIN_FILE_CONTENTS, 0, obj_type, bonuslen, dnodesize, tx)); } else { obj = dmu_object_alloc_dnsize(zfsvfs->z_os, DMU_OT_PLAIN_FILE_CONTENTS, 0, obj_type, bonuslen, dnodesize, tx); } } zh = zfs_znode_hold_enter(zfsvfs, obj); VERIFY0(sa_buf_hold(zfsvfs->z_os, obj, NULL, &db)); /* * If this is the root, fix up the half-initialized parent pointer * to reference the just-allocated physical data area. */ if (flag & IS_ROOT_NODE) { dzp->z_id = obj; } /* * If parent is an xattr, so am I. */ if (dzp->z_pflags & ZFS_XATTR) { flag |= IS_XATTR; } if (zfsvfs->z_use_fuids) pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; else pflags = 0; if (S_ISDIR(vap->va_mode)) { size = 2; /* contents ("." and "..") */ links = 2; } else { size = 0; links = (flag & IS_TMPFILE) ? 0 : 1; } if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode)) rdev = vap->va_rdev; parent = dzp->z_id; mode = acl_ids->z_mode; if (flag & IS_XATTR) pflags |= ZFS_XATTR; if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) { /* * With ZFS_PROJID flag, we can easily know whether there is * project ID stored on disk or not. See zfs_space_delta_cb(). */ if (obj_type != DMU_OT_ZNODE && dmu_objset_projectquota_enabled(zfsvfs->z_os)) pflags |= ZFS_PROJID; /* * Inherit project ID from parent if required. */ projid = zfs_inherit_projid(dzp); if (dzp->z_pflags & ZFS_PROJINHERIT) pflags |= ZFS_PROJINHERIT; } /* * No execs denied will be determined when zfs_mode_compute() is called. */ pflags |= acl_ids->z_aclp->z_hints & (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); ZFS_TIME_ENCODE(&now, crtime); ZFS_TIME_ENCODE(&now, ctime); if (vap->va_mask & ATTR_ATIME) { ZFS_TIME_ENCODE(&vap->va_atime, atime); } else { ZFS_TIME_ENCODE(&now, atime); } if (vap->va_mask & ATTR_MTIME) { ZFS_TIME_ENCODE(&vap->va_mtime, mtime); } else { ZFS_TIME_ENCODE(&now, mtime); } /* Now add in all of the "SA" attributes */ VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED, &sa_hdl)); /* * Setup the array of attributes to be replaced/set on the new file * * order for DMU_OT_ZNODE is critical since it needs to be constructed * in the old znode_phys_t format. Don't change this ordering */ sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); if (obj_type == DMU_OT_ZNODE) { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), NULL, &gen, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), NULL, &size, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8); } else { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), NULL, &size, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), NULL, &gen, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, &acl_ids->z_fuid, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, &acl_ids->z_fgid, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, &pflags, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16); } SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); if (obj_type == DMU_OT_ZNODE) { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL, &empty_xattr, 8); } else if (dmu_objset_projectquota_enabled(zfsvfs->z_os) && pflags & ZFS_PROJID) { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PROJID(zfsvfs), NULL, &projid, 8); } if (obj_type == DMU_OT_ZNODE || (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs), NULL, &rdev, 8); } if (obj_type == DMU_OT_ZNODE) { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, &pflags, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, &acl_ids->z_fuid, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, &acl_ids->z_fgid, 8); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad, sizeof (uint64_t) * 4); SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL, &acl_phys, sizeof (zfs_acl_phys_t)); } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL, &acl_ids->z_aclp->z_acl_count, 8); locate.cb_aclp = acl_ids->z_aclp; SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs), zfs_acl_data_locator, &locate, acl_ids->z_aclp->z_acl_bytes); mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, acl_ids->z_fuid, acl_ids->z_fgid); } VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0); if (!(flag & IS_ROOT_NODE)) { /* * The call to zfs_znode_alloc() may fail if memory is low * via the call path: alloc_inode() -> inode_init_always() -> * security_inode_alloc() -> inode_alloc_security(). Since * the existing code is written such that zfs_mknode() can * not fail retry until sufficient memory has been reclaimed. */ do { *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl); } while (*zpp == NULL); VERIFY(*zpp != NULL); VERIFY(dzp != NULL); } else { /* * If we are creating the root node, the "parent" we * passed in is the znode for the root. */ *zpp = dzp; (*zpp)->z_sa_hdl = sa_hdl; } (*zpp)->z_pflags = pflags; (*zpp)->z_mode = ZTOI(*zpp)->i_mode = mode; (*zpp)->z_dnodesize = dnodesize; (*zpp)->z_projid = projid; if (obj_type == DMU_OT_ZNODE || acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx)); } kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END); zfs_znode_hold_exit(zfsvfs, zh); } /* * Update in-core attributes. It is assumed the caller will be doing an * sa_bulk_update to push the changes out. */ void zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) { xoptattr_t *xoap; boolean_t update_inode = B_FALSE; xoap = xva_getxoptattr(xvap); ASSERT(xoap); if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { uint64_t times[2]; ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)), ×, sizeof (times), tx); XVA_SET_RTN(xvap, XAT_CREATETIME); } if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_READONLY); } if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_HIDDEN); } if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_SYSTEM); } if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_ARCHIVE); } if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_IMMUTABLE); update_inode = B_TRUE; } if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_NOUNLINK); } if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_APPENDONLY); update_inode = B_TRUE; } if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_NODUMP); } if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_OPAQUE); } if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, xoap->xoa_av_quarantined, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); } if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_AV_MODIFIED); } if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { zfs_sa_set_scanstamp(zp, xvap, tx); XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); } if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_REPARSE); } if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_OFFLINE); } if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_SPARSE); } if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { ZFS_ATTR_SET(zp, ZFS_PROJINHERIT, xoap->xoa_projinherit, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_PROJINHERIT); } if (update_inode) zfs_set_inode_flags(zp, ZTOI(zp)); } int zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) { dmu_object_info_t doi; dmu_buf_t *db; znode_t *zp; znode_hold_t *zh; int err; sa_handle_t *hdl; *zpp = NULL; again: zh = zfs_znode_hold_enter(zfsvfs, obj_num); err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); if (err) { zfs_znode_hold_exit(zfsvfs, zh); return (err); } dmu_object_info_from_db(db, &doi); if (doi.doi_bonus_type != DMU_OT_SA && (doi.doi_bonus_type != DMU_OT_ZNODE || (doi.doi_bonus_type == DMU_OT_ZNODE && doi.doi_bonus_size < sizeof (znode_phys_t)))) { sa_buf_rele(db, NULL); zfs_znode_hold_exit(zfsvfs, zh); return (SET_ERROR(EINVAL)); } hdl = dmu_buf_get_user(db); if (hdl != NULL) { zp = sa_get_userdata(hdl); /* * Since "SA" does immediate eviction we * should never find a sa handle that doesn't * know about the znode. */ ASSERT3P(zp, !=, NULL); mutex_enter(&zp->z_lock); ASSERT3U(zp->z_id, ==, obj_num); /* * If zp->z_unlinked is set, the znode is already marked * for deletion and should not be discovered. Check this * after checking igrab() due to fsetxattr() & O_TMPFILE. * * If igrab() returns NULL the VFS has independently * determined the inode should be evicted and has * called iput_final() to start the eviction process. * The SA handle is still valid but because the VFS * requires that the eviction succeed we must drop * our locks and references to allow the eviction to * complete. The zfs_zget() may then be retried. * * This unlikely case could be optimized by registering * a sops->drop_inode() callback. The callback would * need to detect the active SA hold thereby informing * the VFS that this inode should not be evicted. */ if (igrab(ZTOI(zp)) == NULL) { if (zp->z_unlinked) err = SET_ERROR(ENOENT); else err = SET_ERROR(EAGAIN); } else { *zpp = zp; err = 0; } mutex_exit(&zp->z_lock); sa_buf_rele(db, NULL); zfs_znode_hold_exit(zfsvfs, zh); if (err == EAGAIN) { /* inode might need this to finish evict */ cond_resched(); goto again; } return (err); } /* * Not found create new znode/vnode but only if file exists. * * There is a small window where zfs_vget() could * find this object while a file create is still in * progress. This is checked for in zfs_znode_alloc() * * if zfs_znode_alloc() fails it will drop the hold on the * bonus buffer. */ zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size, doi.doi_bonus_type, NULL); if (zp == NULL) { err = SET_ERROR(ENOENT); } else { *zpp = zp; } zfs_znode_hold_exit(zfsvfs, zh); return (err); } int zfs_rezget(znode_t *zp) { zfsvfs_t *zfsvfs = ZTOZSB(zp); dmu_object_info_t doi; dmu_buf_t *db; uint64_t obj_num = zp->z_id; uint64_t mode; uint64_t links; sa_bulk_attr_t bulk[11]; int err; int count = 0; uint64_t gen; uint64_t z_uid, z_gid; uint64_t atime[2], mtime[2], ctime[2], btime[2]; uint64_t projid = ZFS_DEFAULT_PROJID; znode_hold_t *zh; /* * skip ctldir, otherwise they will always get invalidated. This will * cause funny behaviour for the mounted snapdirs. Especially for * Linux >= 3.18, d_invalidate will detach the mountpoint and prevent * anyone automount it again as long as someone is still using the * detached mount. */ if (zp->z_is_ctldir) return (0); zh = zfs_znode_hold_enter(zfsvfs, obj_num); mutex_enter(&zp->z_acl_lock); if (zp->z_acl_cached) { zfs_acl_free(zp->z_acl_cached); zp->z_acl_cached = NULL; } mutex_exit(&zp->z_acl_lock); rw_enter(&zp->z_xattr_lock, RW_WRITER); if (zp->z_xattr_cached) { nvlist_free(zp->z_xattr_cached); zp->z_xattr_cached = NULL; } rw_exit(&zp->z_xattr_lock); ASSERT(zp->z_sa_hdl == NULL); err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); if (err) { zfs_znode_hold_exit(zfsvfs, zh); return (err); } dmu_object_info_from_db(db, &doi); if (doi.doi_bonus_type != DMU_OT_SA && (doi.doi_bonus_type != DMU_OT_ZNODE || (doi.doi_bonus_type == DMU_OT_ZNODE && doi.doi_bonus_size < sizeof (znode_phys_t)))) { sa_buf_rele(db, NULL); zfs_znode_hold_exit(zfsvfs, zh); return (SET_ERROR(EINVAL)); } zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL); /* reload cached values */ SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &gen, sizeof (gen)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, &zp->z_size, sizeof (zp->z_size)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, &links, sizeof (links)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, sizeof (zp->z_pflags)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &z_uid, sizeof (z_uid)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &z_gid, sizeof (z_gid)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, sizeof (mode)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &btime, 16); if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { zfs_znode_dmu_fini(zp); zfs_znode_hold_exit(zfsvfs, zh); return (SET_ERROR(EIO)); } if (dmu_objset_projectquota_enabled(zfsvfs->z_os)) { err = sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), &projid, 8); if (err != 0 && err != ENOENT) { zfs_znode_dmu_fini(zp); zfs_znode_hold_exit(zfsvfs, zh); return (SET_ERROR(err)); } } zp->z_projid = projid; zp->z_mode = ZTOI(zp)->i_mode = mode; zfs_uid_write(ZTOI(zp), z_uid); zfs_gid_write(ZTOI(zp), z_gid); ZFS_TIME_DECODE(&ZTOI(zp)->i_atime, atime); ZFS_TIME_DECODE(&ZTOI(zp)->i_mtime, mtime); ZFS_TIME_DECODE(&ZTOI(zp)->i_ctime, ctime); ZFS_TIME_DECODE(&zp->z_btime, btime); if ((uint32_t)gen != ZTOI(zp)->i_generation) { zfs_znode_dmu_fini(zp); zfs_znode_hold_exit(zfsvfs, zh); return (SET_ERROR(EIO)); } set_nlink(ZTOI(zp), (uint32_t)links); zfs_set_inode_flags(zp, ZTOI(zp)); zp->z_blksz = doi.doi_data_block_size; zp->z_atime_dirty = B_FALSE; zfs_znode_update_vfs(zp); /* * If the file has zero links, then it has been unlinked on the send * side and it must be in the received unlinked set. * We call zfs_znode_dmu_fini() now to prevent any accesses to the * stale data and to prevent automatic removal of the file in * zfs_zinactive(). The file will be removed either when it is removed * on the send side and the next incremental stream is received or * when the unlinked set gets processed. */ zp->z_unlinked = (ZTOI(zp)->i_nlink == 0); if (zp->z_unlinked) zfs_znode_dmu_fini(zp); zfs_znode_hold_exit(zfsvfs, zh); return (0); } void zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) { zfsvfs_t *zfsvfs = ZTOZSB(zp); objset_t *os = zfsvfs->z_os; uint64_t obj = zp->z_id; uint64_t acl_obj = zfs_external_acl(zp); znode_hold_t *zh; zh = zfs_znode_hold_enter(zfsvfs, obj); if (acl_obj) { VERIFY(!zp->z_is_sa); VERIFY(0 == dmu_object_free(os, acl_obj, tx)); } VERIFY(0 == dmu_object_free(os, obj, tx)); zfs_znode_dmu_fini(zp); zfs_znode_hold_exit(zfsvfs, zh); } void zfs_zinactive(znode_t *zp) { zfsvfs_t *zfsvfs = ZTOZSB(zp); uint64_t z_id = zp->z_id; znode_hold_t *zh; ASSERT(zp->z_sa_hdl); /* * Don't allow a zfs_zget() while were trying to release this znode. */ zh = zfs_znode_hold_enter(zfsvfs, z_id); mutex_enter(&zp->z_lock); /* * If this was the last reference to a file with no links, remove * the file from the file system unless the file system is mounted * read-only. That can happen, for example, if the file system was * originally read-write, the file was opened, then unlinked and * the file system was made read-only before the file was finally * closed. The file will remain in the unlinked set. */ if (zp->z_unlinked) { ASSERT(!zfsvfs->z_issnap); if (!zfs_is_readonly(zfsvfs) && !zfs_unlink_suspend_progress) { mutex_exit(&zp->z_lock); zfs_znode_hold_exit(zfsvfs, zh); zfs_rmnode(zp); return; } } mutex_exit(&zp->z_lock); zfs_znode_dmu_fini(zp); zfs_znode_hold_exit(zfsvfs, zh); } #if defined(HAVE_INODE_TIMESPEC64_TIMES) #define zfs_compare_timespec timespec64_compare #else #define zfs_compare_timespec timespec_compare #endif /* * Determine whether the znode's atime must be updated. The logic mostly * duplicates the Linux kernel's relatime_need_update() functionality. * This function is only called if the underlying filesystem actually has * atime updates enabled. */ boolean_t zfs_relatime_need_update(const struct inode *ip) { inode_timespec_t now; gethrestime(&now); /* * In relatime mode, only update the atime if the previous atime * is earlier than either the ctime or mtime or if at least a day * has passed since the last update of atime. */ if (zfs_compare_timespec(&ip->i_mtime, &ip->i_atime) >= 0) return (B_TRUE); if (zfs_compare_timespec(&ip->i_ctime, &ip->i_atime) >= 0) return (B_TRUE); if ((hrtime_t)now.tv_sec - (hrtime_t)ip->i_atime.tv_sec >= 24*60*60) return (B_TRUE); return (B_FALSE); } /* * Prepare to update znode time stamps. * * IN: zp - znode requiring timestamp update * flag - ATTR_MTIME, ATTR_CTIME flags * * OUT: zp - z_seq * mtime - new mtime * ctime - new ctime * * Note: We don't update atime here, because we rely on Linux VFS to do * atime updating. */ void zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], uint64_t ctime[2]) { inode_timespec_t now; gethrestime(&now); zp->z_seq++; if (flag & ATTR_MTIME) { ZFS_TIME_ENCODE(&now, mtime); ZFS_TIME_DECODE(&(ZTOI(zp)->i_mtime), mtime); if (ZTOZSB(zp)->z_use_fuids) { zp->z_pflags |= (ZFS_ARCHIVE | ZFS_AV_MODIFIED); } } if (flag & ATTR_CTIME) { ZFS_TIME_ENCODE(&now, ctime); ZFS_TIME_DECODE(&(ZTOI(zp)->i_ctime), ctime); if (ZTOZSB(zp)->z_use_fuids) zp->z_pflags |= ZFS_ARCHIVE; } } /* * Grow the block size for a file. * * IN: zp - znode of file to free data in. * size - requested block size * tx - open transaction. * * NOTE: this function assumes that the znode is write locked. */ void zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) { int error; u_longlong_t dummy; if (size <= zp->z_blksz) return; /* * If the file size is already greater than the current blocksize, * we will not grow. If there is more than one block in a file, * the blocksize cannot change. */ if (zp->z_blksz && zp->z_size > zp->z_blksz) return; error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id, size, 0, tx); if (error == ENOTSUP) return; ASSERT0(error); /* What blocksize did we actually get? */ dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); } /* * Increase the file length * * IN: zp - znode of file to free data in. * end - new end-of-file * * RETURN: 0 on success, error code on failure */ static int zfs_extend(znode_t *zp, uint64_t end) { zfsvfs_t *zfsvfs = ZTOZSB(zp); dmu_tx_t *tx; zfs_locked_range_t *lr; uint64_t newblksz; int error; /* * We will change zp_size, lock the whole file. */ lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER); /* * Nothing to do if file already at desired length. */ if (end <= zp->z_size) { zfs_rangelock_exit(lr); return (0); } tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); if (end > zp->z_blksz && (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { /* * We are growing the file past the current block size. */ if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) { /* * File's blocksize is already larger than the * "recordsize" property. Only let it grow to * the next power of 2. */ ASSERT(!ISP2(zp->z_blksz)); newblksz = MIN(end, 1 << highbit64(zp->z_blksz)); } else { newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz); } dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); } else { newblksz = 0; } error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); zfs_rangelock_exit(lr); return (error); } if (newblksz) zfs_grow_blocksize(zp, newblksz, tx); zp->z_size = end; VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)), &zp->z_size, sizeof (zp->z_size), tx)); zfs_rangelock_exit(lr); dmu_tx_commit(tx); return (0); } /* * zfs_zero_partial_page - Modeled after update_pages() but * with different arguments and semantics for use by zfs_freesp(). * * Zeroes a piece of a single page cache entry for zp at offset * start and length len. * * Caller must acquire a range lock on the file for the region * being zeroed in order that the ARC and page cache stay in sync. */ static void zfs_zero_partial_page(znode_t *zp, uint64_t start, uint64_t len) { struct address_space *mp = ZTOI(zp)->i_mapping; struct page *pp; int64_t off; void *pb; ASSERT((start & PAGE_MASK) == ((start + len - 1) & PAGE_MASK)); off = start & (PAGE_SIZE - 1); start &= PAGE_MASK; pp = find_lock_page(mp, start >> PAGE_SHIFT); if (pp) { if (mapping_writably_mapped(mp)) flush_dcache_page(pp); pb = kmap(pp); memset(pb + off, 0, len); kunmap(pp); if (mapping_writably_mapped(mp)) flush_dcache_page(pp); mark_page_accessed(pp); SetPageUptodate(pp); ClearPageError(pp); unlock_page(pp); put_page(pp); } } /* * Free space in a file. * * IN: zp - znode of file to free data in. * off - start of section to free. * len - length of section to free. * * RETURN: 0 on success, error code on failure */ static int zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) { zfsvfs_t *zfsvfs = ZTOZSB(zp); zfs_locked_range_t *lr; int error; /* * Lock the range being freed. */ lr = zfs_rangelock_enter(&zp->z_rangelock, off, len, RL_WRITER); /* * Nothing to do if file already at desired length. */ if (off >= zp->z_size) { zfs_rangelock_exit(lr); return (0); } if (off + len > zp->z_size) len = zp->z_size - off; error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len); /* * Zero partial page cache entries. This must be done under a * range lock in order to keep the ARC and page cache in sync. */ if (zn_has_cached_data(zp, off, off + len - 1)) { loff_t first_page, last_page, page_len; loff_t first_page_offset, last_page_offset; /* first possible full page in hole */ first_page = (off + PAGE_SIZE - 1) >> PAGE_SHIFT; /* last page of hole */ last_page = (off + len) >> PAGE_SHIFT; /* offset of first_page */ first_page_offset = first_page << PAGE_SHIFT; /* offset of last_page */ last_page_offset = last_page << PAGE_SHIFT; /* truncate whole pages */ if (last_page_offset > first_page_offset) { truncate_inode_pages_range(ZTOI(zp)->i_mapping, first_page_offset, last_page_offset - 1); } /* truncate sub-page ranges */ if (first_page > last_page) { /* entire punched area within a single page */ zfs_zero_partial_page(zp, off, len); } else { /* beginning of punched area at the end of a page */ page_len = first_page_offset - off; if (page_len > 0) zfs_zero_partial_page(zp, off, page_len); /* end of punched area at the beginning of a page */ page_len = off + len - last_page_offset; if (page_len > 0) zfs_zero_partial_page(zp, last_page_offset, page_len); } } zfs_rangelock_exit(lr); return (error); } /* * Truncate a file * * IN: zp - znode of file to free data in. * end - new end-of-file. * * RETURN: 0 on success, error code on failure */ static int zfs_trunc(znode_t *zp, uint64_t end) { zfsvfs_t *zfsvfs = ZTOZSB(zp); dmu_tx_t *tx; zfs_locked_range_t *lr; int error; sa_bulk_attr_t bulk[2]; int count = 0; /* * We will change zp_size, lock the whole file. */ lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER); /* * Nothing to do if file already at desired length. */ if (end >= zp->z_size) { zfs_rangelock_exit(lr); return (0); } error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, DMU_OBJECT_END); if (error) { zfs_rangelock_exit(lr); return (error); } tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); dmu_tx_mark_netfree(tx); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); zfs_rangelock_exit(lr); return (error); } zp->z_size = end; SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, &zp->z_size, sizeof (zp->z_size)); if (end == 0) { zp->z_pflags &= ~ZFS_SPARSE; SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, 8); } VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0); dmu_tx_commit(tx); zfs_rangelock_exit(lr); return (0); } /* * Free space in a file * * IN: zp - znode of file to free data in. * off - start of range * len - end of range (0 => EOF) * flag - current file open mode flags. * log - TRUE if this action should be logged * * RETURN: 0 on success, error code on failure */ int zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) { dmu_tx_t *tx; zfsvfs_t *zfsvfs = ZTOZSB(zp); zilog_t *zilog = zfsvfs->z_log; uint64_t mode; uint64_t mtime[2], ctime[2]; sa_bulk_attr_t bulk[3]; int count = 0; int error; if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode, sizeof (mode))) != 0) return (error); if (off > zp->z_size) { error = zfs_extend(zp, off+len); if (error == 0 && log) goto log; goto out; } if (len == 0) { error = zfs_trunc(zp, off); } else { if ((error = zfs_free_range(zp, off, len)) == 0 && off + len > zp->z_size) error = zfs_extend(zp, off+len); } if (error || !log) goto out; log: tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); goto out; } SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, 8); zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); ASSERT(error == 0); zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); dmu_tx_commit(tx); zfs_znode_update_vfs(zp); error = 0; out: /* * Truncate the page cache - for file truncate operations, use * the purpose-built API for truncations. For punching operations, * the truncation is handled under a range lock in zfs_free_range. */ if (len == 0) truncate_setsize(ZTOI(zp), off); return (error); } void zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) { struct super_block *sb; zfsvfs_t *zfsvfs; uint64_t moid, obj, sa_obj, version; uint64_t sense = ZFS_CASE_SENSITIVE; uint64_t norm = 0; nvpair_t *elem; int size; int error; int i; znode_t *rootzp = NULL; vattr_t vattr; znode_t *zp; zfs_acl_ids_t acl_ids; /* * First attempt to create master node. */ /* * In an empty objset, there are no blocks to read and thus * there can be no i/o errors (which we assert below). */ moid = MASTER_NODE_OBJ; error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, DMU_OT_NONE, 0, tx); ASSERT(error == 0); /* * Set starting attributes. */ version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); elem = NULL; while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { /* For the moment we expect all zpl props to be uint64_ts */ uint64_t val; const char *name; ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); VERIFY(nvpair_value_uint64(elem, &val) == 0); name = nvpair_name(elem); if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { if (val < version) version = val; } else { error = zap_update(os, moid, name, 8, 1, &val, tx); } ASSERT(error == 0); if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) norm = val; else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) sense = val; } ASSERT(version != 0); error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); ASSERT(error == 0); /* * Create zap object used for SA attribute registration */ if (version >= ZPL_VERSION_SA) { sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, DMU_OT_NONE, 0, tx); error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); ASSERT(error == 0); } else { sa_obj = 0; } /* * Create a delete queue. */ obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); ASSERT(error == 0); /* * Create root znode. Create minimal znode/inode/zfsvfs/sb * to allow zfs_mknode to work. */ vattr.va_mask = ATTR_MODE|ATTR_UID|ATTR_GID; vattr.va_mode = S_IFDIR|0755; vattr.va_uid = crgetuid(cr); vattr.va_gid = crgetgid(cr); rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); rootzp->z_unlinked = B_FALSE; rootzp->z_atime_dirty = B_FALSE; rootzp->z_is_sa = USE_SA(version, os); rootzp->z_pflags = 0; zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); zfsvfs->z_os = os; zfsvfs->z_parent = zfsvfs; zfsvfs->z_version = version; zfsvfs->z_use_fuids = USE_FUIDS(version, os); zfsvfs->z_use_sa = USE_SA(version, os); zfsvfs->z_norm = norm; sb = kmem_zalloc(sizeof (struct super_block), KM_SLEEP); sb->s_fs_info = zfsvfs; ZTOI(rootzp)->i_sb = sb; error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, &zfsvfs->z_attr_table); ASSERT(error == 0); /* * Fold case on file systems that are always or sometimes case * insensitive. */ if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), offsetof(znode_t, z_link_node)); size = MIN(1 << (highbit64(zfs_object_mutex_size)-1), ZFS_OBJ_MTX_MAX); zfsvfs->z_hold_size = size; zfsvfs->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, KM_SLEEP); zfsvfs->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); for (i = 0; i != size; i++) { avl_create(&zfsvfs->z_hold_trees[i], zfs_znode_hold_compare, sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); mutex_init(&zfsvfs->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); } VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, cr, NULL, &acl_ids, zfs_init_idmap)); zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); ASSERT3P(zp, ==, rootzp); error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); ASSERT(error == 0); zfs_acl_ids_free(&acl_ids); atomic_set(&ZTOI(rootzp)->i_count, 0); sa_handle_destroy(rootzp->z_sa_hdl); kmem_cache_free(znode_cache, rootzp); for (i = 0; i != size; i++) { avl_destroy(&zfsvfs->z_hold_trees[i]); mutex_destroy(&zfsvfs->z_hold_locks[i]); } mutex_destroy(&zfsvfs->z_znodes_lock); vmem_free(zfsvfs->z_hold_trees, sizeof (avl_tree_t) * size); vmem_free(zfsvfs->z_hold_locks, sizeof (kmutex_t) * size); kmem_free(sb, sizeof (struct super_block)); kmem_free(zfsvfs, sizeof (zfsvfs_t)); } #endif /* _KERNEL */ static int zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) { uint64_t sa_obj = 0; int error; error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); if (error != 0 && error != ENOENT) return (error); error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); return (error); } static int zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, dmu_buf_t **db, const void *tag) { dmu_object_info_t doi; int error; if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) return (error); dmu_object_info_from_db(*db, &doi); if ((doi.doi_bonus_type != DMU_OT_SA && doi.doi_bonus_type != DMU_OT_ZNODE) || (doi.doi_bonus_type == DMU_OT_ZNODE && doi.doi_bonus_size < sizeof (znode_phys_t))) { sa_buf_rele(*db, tag); return (SET_ERROR(ENOTSUP)); } error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); if (error != 0) { sa_buf_rele(*db, tag); return (error); } return (0); } static void zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, const void *tag) { sa_handle_destroy(hdl); sa_buf_rele(db, tag); } /* * Given an object number, return its parent object number and whether * or not the object is an extended attribute directory. */ static int zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp, int *is_xattrdir) { uint64_t parent; uint64_t pflags; uint64_t mode; uint64_t parent_mode; sa_bulk_attr_t bulk[3]; sa_handle_t *sa_hdl; dmu_buf_t *sa_db; int count = 0; int error; SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, &parent, sizeof (parent)); SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, &pflags, sizeof (pflags)); SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, &mode, sizeof (mode)); if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) return (error); /* * When a link is removed its parent pointer is not changed and will * be invalid. There are two cases where a link is removed but the * file stays around, when it goes to the delete queue and when there * are additional links. */ error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG); if (error != 0) return (error); error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode)); zfs_release_sa_handle(sa_hdl, sa_db, FTAG); if (error != 0) return (error); *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); /* * Extended attributes can be applied to files, directories, etc. * Otherwise the parent must be a directory. */ if (!*is_xattrdir && !S_ISDIR(parent_mode)) return (SET_ERROR(EINVAL)); *pobjp = parent; return (0); } /* * Given an object number, return some zpl level statistics */ static int zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, zfs_stat_t *sb) { sa_bulk_attr_t bulk[4]; int count = 0; SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, &sb->zs_mode, sizeof (sb->zs_mode)); SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, &sb->zs_gen, sizeof (sb->zs_gen)); SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, &sb->zs_links, sizeof (sb->zs_links)); SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, &sb->zs_ctime, sizeof (sb->zs_ctime)); return (sa_bulk_lookup(hdl, bulk, count)); } static int zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, sa_attr_type_t *sa_table, char *buf, int len) { sa_handle_t *sa_hdl; sa_handle_t *prevhdl = NULL; dmu_buf_t *prevdb = NULL; dmu_buf_t *sa_db = NULL; char *path = buf + len - 1; int error; *path = '\0'; sa_hdl = hdl; uint64_t deleteq_obj; VERIFY0(zap_lookup(osp, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, sizeof (uint64_t), 1, &deleteq_obj)); error = zap_lookup_int(osp, deleteq_obj, obj); if (error == 0) { return (ESTALE); } else if (error != ENOENT) { return (error); } for (;;) { uint64_t pobj = 0; char component[MAXNAMELEN + 2]; size_t complen; int is_xattrdir = 0; if (prevdb) { ASSERT(prevhdl != NULL); zfs_release_sa_handle(prevhdl, prevdb, FTAG); } if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj, &is_xattrdir)) != 0) break; if (pobj == obj) { if (path[0] != '/') *--path = '/'; break; } component[0] = '/'; if (is_xattrdir) { strcpy(component + 1, ""); } else { error = zap_value_search(osp, pobj, obj, ZFS_DIRENT_OBJ(-1ULL), component + 1); if (error != 0) break; } complen = strlen(component); path -= complen; ASSERT(path >= buf); memcpy(path, component, complen); obj = pobj; if (sa_hdl != hdl) { prevhdl = sa_hdl; prevdb = sa_db; } error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); if (error != 0) { sa_hdl = prevhdl; sa_db = prevdb; break; } } if (sa_hdl != NULL && sa_hdl != hdl) { ASSERT(sa_db != NULL); zfs_release_sa_handle(sa_hdl, sa_db, FTAG); } if (error == 0) (void) memmove(buf, path, buf + len - path); return (error); } int zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) { sa_attr_type_t *sa_table; sa_handle_t *hdl; dmu_buf_t *db; int error; error = zfs_sa_setup(osp, &sa_table); if (error != 0) return (error); error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); if (error != 0) return (error); error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); zfs_release_sa_handle(hdl, db, FTAG); return (error); } int zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, char *buf, int len) { char *path = buf + len - 1; sa_attr_type_t *sa_table; sa_handle_t *hdl; dmu_buf_t *db; int error; *path = '\0'; error = zfs_sa_setup(osp, &sa_table); if (error != 0) return (error); error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); if (error != 0) return (error); error = zfs_obj_to_stats_impl(hdl, sa_table, sb); if (error != 0) { zfs_release_sa_handle(hdl, db, FTAG); return (error); } error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); zfs_release_sa_handle(hdl, db, FTAG); return (error); } /* * Read a property stored within the master node. */ int zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) { uint64_t *cached_copy = NULL; /* * Figure out where in the objset_t the cached copy would live, if it * is available for the requested property. */ if (os != NULL) { switch (prop) { case ZFS_PROP_VERSION: cached_copy = &os->os_version; break; case ZFS_PROP_NORMALIZE: cached_copy = &os->os_normalization; break; case ZFS_PROP_UTF8ONLY: cached_copy = &os->os_utf8only; break; case ZFS_PROP_CASE: cached_copy = &os->os_casesensitivity; break; default: break; } } if (cached_copy != NULL && *cached_copy != OBJSET_PROP_UNINITIALIZED) { *value = *cached_copy; return (0); } /* * If the property wasn't cached, look up the file system's value for * the property. For the version property, we look up a slightly * different string. */ const char *pname; int error = ENOENT; if (prop == ZFS_PROP_VERSION) pname = ZPL_VERSION_STR; else pname = zfs_prop_to_name(prop); if (os != NULL) { ASSERT3U(os->os_phys->os_type, ==, DMU_OST_ZFS); 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; case ZFS_PROP_ACLTYPE: *value = ZFS_ACLTYPE_OFF; break; default: return (error); } error = 0; } /* * If one of the methods for getting the property value above worked, * copy it into the objset_t's cache. */ if (error == 0 && cached_copy != NULL) { *cached_copy = *value; } return (error); } #if defined(_KERNEL) EXPORT_SYMBOL(zfs_create_fs); EXPORT_SYMBOL(zfs_obj_to_path); /* CSTYLED */ module_param(zfs_object_mutex_size, uint, 0644); MODULE_PARM_DESC(zfs_object_mutex_size, "Size of znode hold array"); module_param(zfs_unlink_suspend_progress, int, 0644); MODULE_PARM_DESC(zfs_unlink_suspend_progress, "Set to prevent async unlinks " "(debug - leaks space into the unlinked set)"); #endif