/*
* Create minors for the specified dataset, including children and snapshots.
* Pay attention to the 'snapdev' property and iterate over the snapshots
* only if they are 'visible'. This approach allows one to assure that the
* snapshot metadata is read from disk only if it is needed.
*
* The name can represent a dataset to be recursively scanned for zvols and
* their snapshots, or a single zvol snapshot. If the name represents a
* dataset, the scan is performed in two nested stages:
* - scan the dataset for zvols, and
* - for each zvol, create a minor node, then check if the zvol's snapshots
* are 'visible', and only then iterate over the snapshots if needed
*
* If the name represents a snapshot, a check is perfromed if the snapshot is
* 'visible' (which also verifies that the parent is a zvol), and if so,
* a minor node for that snapshot is created.
*/
static int
zvol_create_minors_impl(const char *name)
{
int error = 0;
fstrans_cookie_t cookie;
char *atp, *parent;
if (zvol_inhibit_dev)
return (0);
parent = kmem_alloc(MAXPATHLEN, KM_SLEEP);
(void) strlcpy(parent, name, MAXPATHLEN);
if ((atp = strrchr(parent, '@')) != NULL) {
uint64_t snapdev;
*atp = '\0';
error = dsl_prop_get_integer(parent, "snapdev",
&snapdev, NULL);
if (error == 0 && snapdev == ZFS_SNAPDEV_VISIBLE)
error = zvol_create_minor_impl(name);
} else {
cookie = spl_fstrans_mark();
error = dmu_objset_find(parent, zvol_create_minors_cb,
NULL, DS_FIND_CHILDREN);
spl_fstrans_unmark(cookie);
}
kmem_free(parent, MAXPATHLEN);
return (SET_ERROR(error));
}
/*
* Set ZFS_PROP_VOLSIZE set entry point.
*/
int
zvol_set_volsize(const char *name, uint64_t volsize)
{
zvol_state_t *zv = NULL;
objset_t *os = NULL;
int error;
dmu_object_info_t *doi;
uint64_t readonly;
boolean_t owned = B_FALSE;
error = dsl_prop_get_integer(name,
zfs_prop_to_name(ZFS_PROP_READONLY), &readonly, NULL);
if (error != 0)
return (SET_ERROR(error));
if (readonly)
return (SET_ERROR(EROFS));
mutex_enter(&zvol_state_lock);
zv = zvol_find_by_name(name);
if (zv == NULL || zv->zv_objset == NULL) {
if ((error = dmu_objset_own(name, DMU_OST_ZVOL, B_FALSE,
FTAG, &os)) != 0) {
mutex_exit(&zvol_state_lock);
return (SET_ERROR(error));
}
owned = B_TRUE;
if (zv != NULL)
zv->zv_objset = os;
} else {
os = zv->zv_objset;
}
doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP);
if ((error = dmu_object_info(os, ZVOL_OBJ, doi)) ||
(error = zvol_check_volsize(volsize, doi->doi_data_block_size)))
goto out;
error = zvol_update_volsize(volsize, os);
kmem_free(doi, sizeof (dmu_object_info_t));
if (error == 0 && zv != NULL)
error = zvol_update_live_volsize(zv, volsize);
out:
if (owned) {
dmu_objset_disown(os, FTAG);
if (zv != NULL)
zv->zv_objset = NULL;
}
mutex_exit(&zvol_state_lock);
return (error);
}
/*
* Set ZFS_PROP_VOLSIZE set entry point.
*/
int
zvol_set_volsize(const char *name, uint64_t volsize)
{
zvol_state_t *zv;
dmu_object_info_t *doi;
objset_t *os = NULL;
uint64_t readonly;
int error;
mutex_enter(&zvol_state_lock);
zv = zvol_find_by_name(name);
if (zv == NULL) {
error = ENXIO;
goto out;
}
doi = kmem_alloc(sizeof(dmu_object_info_t), KM_SLEEP);
error = dmu_objset_hold(name, FTAG, &os);
if (error)
goto out_doi;
if ((error = dmu_object_info(os, ZVOL_OBJ, doi)) != 0 ||
(error = zvol_check_volsize(volsize,doi->doi_data_block_size)) != 0)
goto out_doi;
VERIFY(dsl_prop_get_integer(name, "readonly", &readonly, NULL) == 0);
if (readonly) {
error = EROFS;
goto out_doi;
}
if (get_disk_ro(zv->zv_disk) || (zv->zv_flags & ZVOL_RDONLY)) {
error = EROFS;
goto out_doi;
}
error = zvol_update_volsize(zv, volsize, os);
out_doi:
kmem_free(doi, sizeof(dmu_object_info_t));
out:
if (os)
dmu_objset_rele(os, FTAG);
mutex_exit(&zvol_state_lock);
return (error);
}
static int
zvol_first_open(zvol_state_t *zv)
{
objset_t *os;
uint64_t volsize;
int error;
uint64_t ro;
/* lie and say we're read-only */
error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, 1, zvol_tag, &os);
if (error)
return (SET_ERROR(-error));
zv->zv_objset = os;
error = dsl_prop_get_integer(zv->zv_name, "readonly", &ro, NULL);
if (error)
goto out_owned;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error)
goto out_owned;
error = dmu_bonus_hold(os, ZVOL_OBJ, zvol_tag, &zv->zv_dbuf);
if (error)
goto out_owned;
set_capacity(zv->zv_disk, volsize >> 9);
zv->zv_volsize = volsize;
zv->zv_zilog = zil_open(os, zvol_get_data);
if (ro || dmu_objset_is_snapshot(os) ||
!spa_writeable(dmu_objset_spa(os))) {
set_disk_ro(zv->zv_disk, 1);
zv->zv_flags |= ZVOL_RDONLY;
} else {
set_disk_ro(zv->zv_disk, 0);
zv->zv_flags &= ~ZVOL_RDONLY;
}
out_owned:
if (error) {
dmu_objset_disown(os, zvol_tag);
zv->zv_objset = NULL;
}
return (SET_ERROR(-error));
}
/*
* 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 : EACCES);
}
return (SET_ERROR(EACCES));
}
static int
__zvol_snapdev_hidden(const char *name)
{
uint64_t snapdev;
char *parent;
char *atp;
int error = 0;
parent = kmem_alloc(MAXPATHLEN, KM_SLEEP);
(void) strlcpy(parent, name, MAXPATHLEN);
if ((atp = strrchr(parent, '@')) != NULL) {
*atp = '\0';
error = dsl_prop_get_integer(parent, "snapdev", &snapdev, NULL);
if ((error == 0) && (snapdev == ZFS_SNAPDEV_HIDDEN))
error = ENODEV;
}
kmem_free(parent, MAXPATHLEN);
return (error);
}
/*
* Mask errors to continue dmu_objset_find() traversal
*/
static int
zvol_create_minors_cb(const char *dsname, void *arg)
{
uint64_t snapdev;
int error;
ASSERT0(MUTEX_HELD(&spa_namespace_lock));
error = dsl_prop_get_integer(dsname, "snapdev", &snapdev, NULL);
if (error)
return (0);
/*
* Given the name and the 'snapdev' property, create device minor nodes
* with the linkages to zvols/snapshots as needed.
* If the name represents a zvol, create a minor node for the zvol, then
* check if its snapshots are 'visible', and if so, iterate over the
* snapshots and create device minor nodes for those.
*/
if (strchr(dsname, '@') == 0) {
/* create minor for the 'dsname' explicitly */
error = zvol_create_minor_impl(dsname);
if ((error == 0 || error == EEXIST) &&
(snapdev == ZFS_SNAPDEV_VISIBLE)) {
fstrans_cookie_t cookie = spl_fstrans_mark();
/*
* traverse snapshots only, do not traverse children,
* and skip the 'dsname'
*/
error = dmu_objset_find((char *)dsname,
zvol_create_snap_minor_cb, (void *)dsname,
DS_FIND_SNAPSHOTS);
spl_fstrans_unmark(cookie);
}
} else {
dprintf("zvol_create_minors_cb(): %s is not a zvol name\n",
dsname);
}
return (0);
}
int
zfs_domount(struct super_block *sb, zfs_mntopts_t *zmo, int silent)
{
const char *osname = zmo->z_osname;
zfs_sb_t *zsb;
struct inode *root_inode;
uint64_t recordsize;
int error;
error = zfs_sb_create(osname, zmo, &zsb);
if (error)
return (error);
if ((error = dsl_prop_get_integer(osname, "recordsize",
&recordsize, NULL)))
goto out;
zsb->z_sb = sb;
sb->s_fs_info = zsb;
sb->s_magic = ZFS_SUPER_MAGIC;
sb->s_maxbytes = MAX_LFS_FILESIZE;
sb->s_time_gran = 1;
sb->s_blocksize = recordsize;
sb->s_blocksize_bits = ilog2(recordsize);
zsb->z_bdi.ra_pages = 0;
sb->s_bdi = &zsb->z_bdi;
error = -zpl_bdi_setup_and_register(&zsb->z_bdi, "zfs");
if (error)
goto out;
/* 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;
#ifdef HAVE_S_D_OP
sb->s_d_op = &zpl_dentry_operations;
#endif /* HAVE_S_D_OP */
/* Set features for file system. */
zfs_set_fuid_feature(zsb);
if (dmu_objset_is_snapshot(zsb->z_os)) {
uint64_t pval;
atime_changed_cb(zsb, B_FALSE);
readonly_changed_cb(zsb, B_TRUE);
if ((error = dsl_prop_get_integer(osname,
"xattr", &pval, NULL)))
goto out;
xattr_changed_cb(zsb, pval);
if ((error = dsl_prop_get_integer(osname,
"acltype", &pval, NULL)))
goto out;
acltype_changed_cb(zsb, pval);
zsb->z_issnap = B_TRUE;
zsb->z_os->os_sync = ZFS_SYNC_DISABLED;
zsb->z_snap_defer_time = jiffies;
mutex_enter(&zsb->z_os->os_user_ptr_lock);
dmu_objset_set_user(zsb->z_os, zsb);
mutex_exit(&zsb->z_os->os_user_ptr_lock);
} else {
error = zfs_sb_setup(zsb, B_TRUE);
}
/* Allocate a root inode for the filesystem. */
error = zfs_root(zsb, &root_inode);
if (error) {
(void) zfs_umount(sb);
goto out;
}
/* Allocate a root dentry for the filesystem */
sb->s_root = d_make_root(root_inode);
if (sb->s_root == NULL) {
(void) zfs_umount(sb);
error = SET_ERROR(ENOMEM);
goto out;
}
if (!zsb->z_issnap)
zfsctl_create(zsb);
zsb->z_arc_prune = arc_add_prune_callback(zpl_prune_sb, sb);
out:
if (error) {
dmu_objset_disown(zsb->z_os, zsb);
zfs_sb_free(zsb);
}
return (error);
}
static int
zfs_domount(struct mount *mp, dev_t mount_dev, char *osname, vfs_context_t ctx)
{
uint64_t readonly;
int error = 0;
int mode;
zfsvfs_t *zfsvfs;
znode_t *zp = NULL;
struct timeval tv;
ASSERT(mp);
ASSERT(osname);
/*
* Initialize the zfs-specific filesystem structure.
* Should probably make this a kmem cache, shuffle fields,
* and just bzero up to z_hold_mtx[].
*/
zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
zfsvfs->z_vfs = mp;
zfsvfs->z_parent = zfsvfs;
zfsvfs->z_assign = TXG_NOWAIT;
zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE;
zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
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));
rw_init(&zfsvfs->z_unmount_lock, NULL, RW_DEFAULT, NULL);
rw_init(&zfsvfs->z_unmount_inactive_lock, NULL, RW_DEFAULT, NULL);
#ifndef __APPLE__
/* Initialize the generic filesystem structure. */
vfsp->vfs_bcount = 0;
vfsp->vfs_data = NULL;
if (zfs_create_unique_device(&mount_dev) == -1) {
error = ENODEV;
goto out;
}
ASSERT(vfs_devismounted(mount_dev) == 0);
#endif
vfs_setfsprivate(mp, zfsvfs);
if (error = dsl_prop_get_integer(osname, "readonly", &readonly, NULL))
goto out;
if (readonly) {
mode = DS_MODE_PRIMARY | DS_MODE_READONLY;
vfs_setflags(mp, (u_int64_t)((unsigned int)MNT_RDONLY));
} else {
mode = DS_MODE_PRIMARY;
}
error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os);
if (error == EROFS) {
mode = DS_MODE_PRIMARY | DS_MODE_READONLY;
error = dmu_objset_open(osname, DMU_OST_ZFS, mode,
&zfsvfs->z_os);
}
if (error)
goto out;
if (error = zfs_init_fs(zfsvfs, &zp, (cred_t *) vfs_context_ucred(ctx)))
goto out;
/* The call to zfs_init_fs leaves the vnode held, release it here. */
vnode_put(ZTOV(zp));
if (dmu_objset_is_snapshot(zfsvfs->z_os)) {
uint64_t xattr;
ASSERT(mode & DS_MODE_READONLY);
#if 0
atime_changed_cb(zfsvfs, B_FALSE);
readonly_changed_cb(zfsvfs, B_TRUE);
if (error = dsl_prop_get_integer(osname, "xattr", &xattr, NULL))
goto out;
xattr_changed_cb(zfsvfs, xattr);
#endif
zfsvfs->z_issnap = B_TRUE;
} else {
if (!vfs_isrdonly(mp))
zfs_unlinked_drain(zfsvfs);
#ifndef __APPLE__
/*
* 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.
*/
zil_replay(zfsvfs->z_os, zfsvfs, &zfsvfs->z_assign,
zfs_replay_vector);
if (!zil_disable)
zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data);
#endif
}
#if 0
if (!zfsvfs->z_issnap)
zfsctl_create(zfsvfs);
#endif
/*
* Record the mount time (for Spotlight)
*/
microtime(&tv);
zfsvfs->z_mount_time = tv.tv_sec;
out:
if (error) {
if (zfsvfs->z_os)
dmu_objset_close(zfsvfs->z_os);
mutex_destroy(&zfsvfs->z_znodes_lock);
list_destroy(&zfsvfs->z_all_znodes);
rw_destroy(&zfsvfs->z_unmount_lock);
rw_destroy(&zfsvfs->z_unmount_inactive_lock);
kmem_free(zfsvfs, sizeof (zfsvfs_t));
} else {
OSIncrementAtomic(&zfs_active_fs_count);
(void) copystr(osname, vfs_statfs(mp)->f_mntfromname, MNAMELEN - 1, 0);
vfs_getnewfsid(mp);
}
return (error);
}
int
zfs_domount(struct super_block *sb, void *data, int silent)
{
zpl_mount_data_t *zmd = data;
const char *osname = zmd->z_osname;
zfs_sb_t *zsb;
struct inode *root_inode;
uint64_t recordsize;
int error;
error = zfs_sb_create(osname, &zsb);
if (error)
return (error);
if ((error = dsl_prop_get_integer(osname, "recordsize",
&recordsize, NULL)))
goto out;
zsb->z_sb = sb;
sb->s_fs_info = zsb;
sb->s_magic = ZFS_SUPER_MAGIC;
sb->s_maxbytes = MAX_LFS_FILESIZE;
sb->s_time_gran = 1;
sb->s_blocksize = recordsize;
sb->s_blocksize_bits = ilog2(recordsize);
#ifdef HAVE_BDI
/*
* 2.6.32 API change,
* Added backing_device_info (BDI) per super block interfaces. A BDI
* must be configured when using a non-device backed filesystem for
* proper writeback. This is not required for older pdflush kernels.
*
* NOTE: Linux read-ahead is disabled in favor of zfs read-ahead.
*/
zsb->z_bdi.ra_pages = 0;
sb->s_bdi = &zsb->z_bdi;
error = -bdi_setup_and_register(&zsb->z_bdi, "zfs", BDI_CAP_MAP_COPY);
if (error)
goto out;
#endif /* HAVE_BDI */
/* Set callback operations for the file system. */
sb->s_op = &zpl_super_operations;
sb->s_xattr = zpl_xattr_handlers;
sb->s_export_op = &zpl_export_operations;
#ifdef HAVE_S_D_OP
sb->s_d_op = &zpl_dentry_operations;
#endif /* HAVE_S_D_OP */
/* Set features for file system. */
zfs_set_fuid_feature(zsb);
if (dmu_objset_is_snapshot(zsb->z_os)) {
uint64_t pval;
atime_changed_cb(zsb, B_FALSE);
readonly_changed_cb(zsb, B_TRUE);
if ((error = dsl_prop_get_integer(osname,"xattr",&pval,NULL)))
goto out;
xattr_changed_cb(zsb, pval);
zsb->z_issnap = B_TRUE;
zsb->z_os->os_sync = ZFS_SYNC_DISABLED;
mutex_enter(&zsb->z_os->os_user_ptr_lock);
dmu_objset_set_user(zsb->z_os, zsb);
mutex_exit(&zsb->z_os->os_user_ptr_lock);
} else {
error = zfs_sb_setup(zsb, B_TRUE);
}
/* Allocate a root inode for the filesystem. */
error = zfs_root(zsb, &root_inode);
if (error) {
(void) zfs_umount(sb);
goto out;
}
/* Allocate a root dentry for the filesystem */
sb->s_root = d_make_root(root_inode);
if (sb->s_root == NULL) {
(void) zfs_umount(sb);
error = ENOMEM;
goto out;
}
if (!zsb->z_issnap)
zfsctl_create(zsb);
out:
if (error) {
dmu_objset_disown(zsb->z_os, zsb);
zfs_sb_free(zsb);
}
return (error);
}
static int
zvol_first_open(zvol_state_t *zv)
{
objset_t *os;
uint64_t volsize;
int locked = 0;
int error;
uint64_t ro;
/*
* In all other cases the spa_namespace_lock is taken before the
* bdev->bd_mutex lock. But in this case the Linux __blkdev_get()
* function calls fops->open() with the bdev->bd_mutex lock held.
*
* To avoid a potential lock inversion deadlock we preemptively
* try to take the spa_namespace_lock(). Normally it will not
* be contended and this is safe because spa_open_common() handles
* the case where the caller already holds the spa_namespace_lock.
*
* When it is contended we risk a lock inversion if we were to
* block waiting for the lock. Luckily, the __blkdev_get()
* function allows us to return -ERESTARTSYS which will result in
* bdev->bd_mutex being dropped, reacquired, and fops->open() being
* called again. This process can be repeated safely until both
* locks are acquired.
*/
if (!mutex_owned(&spa_namespace_lock)) {
locked = mutex_tryenter(&spa_namespace_lock);
if (!locked)
return (-ERESTARTSYS);
}
/* lie and say we're read-only */
error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, 1, zvol_tag, &os);
if (error)
goto out_mutex;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error) {
dmu_objset_disown(os, zvol_tag);
goto out_mutex;
}
zv->zv_objset = os;
error = dmu_bonus_hold(os, ZVOL_OBJ, zvol_tag, &zv->zv_dbuf);
if (error) {
dmu_objset_disown(os, zvol_tag);
goto out_mutex;
}
set_capacity(zv->zv_disk, volsize >> 9);
zv->zv_volsize = volsize;
zv->zv_zilog = zil_open(os, zvol_get_data);
VERIFY(dsl_prop_get_integer(zv->zv_name, "readonly", &ro, NULL) == 0);
if (ro || dmu_objset_is_snapshot(os) ||
!spa_writeable(dmu_objset_spa(os))) {
set_disk_ro(zv->zv_disk, 1);
zv->zv_flags |= ZVOL_RDONLY;
} else {
set_disk_ro(zv->zv_disk, 0);
zv->zv_flags &= ~ZVOL_RDONLY;
}
out_mutex:
if (locked)
mutex_exit(&spa_namespace_lock);
return (-error);
}
static int
zfs_domount(vfs_t *vfsp, char *osname, cred_t *cr)
{
dev_t mount_dev;
uint64_t recordsize, readonly;
int error = 0;
int mode;
zfsvfs_t *zfsvfs;
znode_t *zp = NULL;
ASSERT(vfsp);
ASSERT(osname);
/*
* Initialize the zfs-specific filesystem structure.
* Should probably make this a kmem cache, shuffle fields,
* and just bzero up to z_hold_mtx[].
*/
zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
zfsvfs->z_vfs = vfsp;
zfsvfs->z_parent = zfsvfs;
zfsvfs->z_assign = TXG_NOWAIT;
zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE;
zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
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));
rw_init(&zfsvfs->z_um_lock, NULL, RW_DEFAULT, NULL);
/* Initialize the generic filesystem structure. */
vfsp->vfs_bcount = 0;
vfsp->vfs_data = NULL;
if (zfs_create_unique_device(&mount_dev) == -1) {
error = ENODEV;
goto out;
}
ASSERT(vfs_devismounted(mount_dev) == 0);
if (error = dsl_prop_get_integer(osname, "recordsize", &recordsize,
NULL))
goto out;
vfsp->vfs_dev = mount_dev;
vfsp->vfs_fstype = zfsfstype;
vfsp->vfs_bsize = recordsize;
vfsp->vfs_flag |= VFS_NOTRUNC;
vfsp->vfs_data = zfsvfs;
if (error = dsl_prop_get_integer(osname, "readonly", &readonly, NULL))
goto out;
if (readonly)
mode = DS_MODE_PRIMARY | DS_MODE_READONLY;
else
mode = DS_MODE_PRIMARY;
error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os);
if (error == EROFS) {
mode = DS_MODE_PRIMARY | DS_MODE_READONLY;
error = dmu_objset_open(osname, DMU_OST_ZFS, mode,
&zfsvfs->z_os);
}
if (error)
goto out;
if (error = zfs_init_fs(zfsvfs, &zp, cr))
goto out;
/* The call to zfs_init_fs leaves the vnode held, release it here. */
VN_RELE(ZTOV(zp));
if (dmu_objset_is_snapshot(zfsvfs->z_os)) {
ASSERT(mode & DS_MODE_READONLY);
atime_changed_cb(zfsvfs, B_FALSE);
readonly_changed_cb(zfsvfs, B_TRUE);
zfsvfs->z_issnap = B_TRUE;
} else {
error = zfs_register_callbacks(vfsp);
if (error)
goto out;
/*
* Start a delete thread running.
*/
(void) zfs_delete_thread_target(zfsvfs, 1);
/*
* Parse and replay the intent log.
*/
zil_replay(zfsvfs->z_os, zfsvfs, &zfsvfs->z_assign,
zfs_replay_vector, (void (*)(void *))zfs_delete_wait_empty);
if (!zil_disable)
zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data);
}
if (!zfsvfs->z_issnap)
zfsctl_create(zfsvfs);
out:
if (error) {
if (zfsvfs->z_os)
dmu_objset_close(zfsvfs->z_os);
kmem_free(zfsvfs, sizeof (zfsvfs_t));
} else {
atomic_add_32(&zfs_active_fs_count, 1);
}
return (error);
}
static int
zfs_domount(vfs_t *vfsp, char *osname)
{
uint64_t recordsize, readonly;
int error = 0;
int mode;
zfsvfs_t *zfsvfs;
znode_t *zp = NULL;
ASSERT(vfsp);
ASSERT(osname);
/*
* Initialize the zfs-specific filesystem structure.
* Should probably make this a kmem cache, shuffle fields,
* and just bzero up to z_hold_mtx[].
*/
zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
zfsvfs->z_vfs = vfsp;
zfsvfs->z_parent = zfsvfs;
zfsvfs->z_assign = TXG_NOWAIT;
zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE;
zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&zfsvfs->z_online_recv_lock, NULL, MUTEX_DEFAULT, NULL);
list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
offsetof(znode_t, z_link_node));
rrw_init(&zfsvfs->z_teardown_lock);
rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL);
if (error = dsl_prop_get_integer(osname, "recordsize", &recordsize,
NULL))
goto out;
zfsvfs->z_vfs->vfs_bsize = recordsize;
vfsp->vfs_data = zfsvfs;
vfsp->mnt_flag |= MNT_LOCAL;
vfsp->mnt_kern_flag |= MNTK_MPSAFE;
vfsp->mnt_kern_flag |= MNTK_LOOKUP_SHARED;
vfsp->mnt_kern_flag |= MNTK_SHARED_WRITES;
if (error = dsl_prop_get_integer(osname, "readonly", &readonly, NULL))
goto out;
mode = DS_MODE_OWNER;
if (readonly)
mode |= DS_MODE_READONLY;
error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os);
if (error == EROFS) {
mode = DS_MODE_OWNER | DS_MODE_READONLY;
error = dmu_objset_open(osname, DMU_OST_ZFS, mode,
&zfsvfs->z_os);
}
if (error)
goto out;
if (error = zfs_init_fs(zfsvfs, &zp))
goto out;
/*
* Set features for file system.
*/
zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os);
if (zfsvfs->z_use_fuids) {
vfs_set_feature(vfsp, VFSFT_XVATTR);
vfs_set_feature(vfsp, VFSFT_SYSATTR_VIEWS);
vfs_set_feature(vfsp, VFSFT_ACEMASKONACCESS);
vfs_set_feature(vfsp, VFSFT_ACLONCREATE);
}
if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS);
vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE);
vfs_set_feature(vfsp, VFSFT_NOCASESENSITIVE);
} else if (zfsvfs->z_case == ZFS_CASE_MIXED) {
vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS);
vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE);
}
if (dmu_objset_is_snapshot(zfsvfs->z_os)) {
uint64_t pval;
ASSERT(mode & DS_MODE_READONLY);
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);
zfsvfs->z_issnap = B_TRUE;
} else {
error = zfsvfs_setup(zfsvfs, B_TRUE);
}
vfs_mountedfrom(vfsp, osname);
if (!zfsvfs->z_issnap)
zfsctl_create(zfsvfs);
out:
if (error) {
if (zfsvfs->z_os)
dmu_objset_close(zfsvfs->z_os);
zfs_freezfsvfs(zfsvfs);
} else {
atomic_add_32(&zfs_active_fs_count, 1);
}
return (error);
}
static int
zfs_register_callbacks(vfs_t *vfsp)
{
struct dsl_dataset *ds = NULL;
objset_t *os = NULL;
zfsvfs_t *zfsvfs = NULL;
uint64_t nbmand;
int readonly, do_readonly = FALSE;
int setuid, do_setuid = FALSE;
int exec, do_exec = FALSE;
int xattr, do_xattr = FALSE;
int atime, do_atime = FALSE;
int error = 0;
ASSERT(vfsp);
zfsvfs = vfsp->vfs_data;
ASSERT(zfsvfs);
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)) {
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_NOSUID, NULL)) {
setuid = B_FALSE;
do_setuid = B_TRUE;
} else {
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)) {
xattr = B_FALSE;
do_xattr = B_TRUE;
} else if (vfs_optionisset(vfsp, MNTOPT_XATTR, NULL)) {
xattr = B_TRUE;
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;
}
/*
* 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 {
char osname[MAXNAMELEN];
dmu_objset_name(os, osname);
if (error = dsl_prop_get_integer(osname, "nbmand", &nbmand,
NULL)) {
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...
*/
ds = dmu_objset_ds(os);
error = dsl_prop_register(ds, "atime", atime_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"xattr", xattr_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"recordsize", blksz_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"readonly", readonly_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"setuid", setuid_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"exec", exec_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"snapdir", snapdir_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"aclmode", acl_mode_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"aclinherit", acl_inherit_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"vscan", vscan_changed_cb, zfsvfs);
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:
/*
* We may attempt to unregister some callbacks that are not
* registered, but this is OK; it will simply return ENOMSG,
* which we will ignore.
*/
(void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb,
zfsvfs);
(void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zfsvfs);
return (error);
}
