/* ARGSUSED */
static int
object_from_path(const char *dataset, const char *path, struct stat64 *statbuf,
zinject_record_t *record)
{
objset_t *os;
int err;
/*
* Before doing any libzpool operations, call sync() to ensure that the
* on-disk state is consistent with the in-core state.
*/
sync();
err = dmu_objset_own(dataset, DMU_OST_ZFS, B_TRUE, FTAG, &os);
if (err != 0) {
(void) fprintf(stderr, "cannot open dataset '%s': %s\n",
dataset, strerror(err));
return (-1);
}
record->zi_objset = dmu_objset_id(os);
record->zi_object = statbuf->st_ino;
dmu_objset_disown(os, FTAG);
return (0);
}
/*ARGSUSED*/
int
zfs_umount(struct super_block *sb)
{
zfs_sb_t *zsb = sb->s_fs_info;
objset_t *os;
arc_remove_prune_callback(zsb->z_arc_prune);
VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0);
os = zsb->z_os;
bdi_destroy(sb->s_bdi);
/*
* z_os will be NULL if there was an error in
* attempting to reopen zsb.
*/
if (os != NULL) {
/*
* Unset the objset user_ptr.
*/
mutex_enter(&os->os_user_ptr_lock);
dmu_objset_set_user(os, NULL);
mutex_exit(&os->os_user_ptr_lock);
/*
* Finally release the objset
*/
dmu_objset_disown(os, zsb);
}
zfs_sb_free(zsb);
return (0);
}
static void
zvol_last_close(zvol_state_t *zv)
{
zil_close(zv->zv_zilog);
zv->zv_zilog = NULL;
dmu_buf_rele(zv->zv_dbuf, zvol_tag);
zv->zv_dbuf = NULL;
dmu_objset_disown(zv->zv_objset, zvol_tag);
zv->zv_objset = NULL;
}
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 (-error);
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error) {
dmu_objset_disown(os, zvol_tag);
return (-error);
}
zv->zv_objset = os;
error = dmu_bonus_hold(os, ZVOL_OBJ, zvol_tag, &zv->zv_dbuf);
if (error) {
dmu_objset_disown(os, zvol_tag);
return (-error);
}
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)) {
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;
}
return (-error);
}
/*
* Block out VFS ops and close zfs_sb_t
*
* Note, if successful, then we return with the 'z_teardown_lock' and
* 'z_teardown_inactive_lock' write held.
*/
int
zfs_suspend_fs(zfs_sb_t *zsb)
{
int error;
if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0)
return (error);
dmu_objset_disown(zsb->z_os, zsb);
return (0);
}
/*
* 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);
}
/* called from zpl */
int
dmu_objset_own(const char *name, dmu_objset_type_t type,
boolean_t readonly, void *tag, objset_t **osp)
{
dsl_dataset_t *ds;
int err;
err = dsl_dataset_own(name, B_FALSE, tag, &ds);
if (err)
return (err);
err = dmu_objset_from_ds(ds, osp);
if (err) {
dsl_dataset_disown(ds, tag);
} else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) {
dmu_objset_disown(*osp, tag);
return (EINVAL);
} else if (!readonly && dsl_dataset_is_snapshot(ds)) {
dmu_objset_disown(*osp, tag);
return (EROFS);
}
return (err);
}
static void
zpios_remove_objset(run_args_t *run_args)
{
zpios_time_t *t = &(run_args->stats.rm_time);
zpios_region_t *region;
char name[32];
int rc = 0, i;
(void)zpios_upcall(run_args->pre, PHASE_PRE_REMOVE, run_args, 0);
t->start = zpios_timespec_now();
(void)snprintf(name, 32, "%s/id_%d", run_args->pool, run_args->id);
if (run_args->flags & DMU_REMOVE) {
if (run_args->flags & DMU_FPP) {
for (i = 0; i < run_args->region_count; i++) {
region = &run_args->regions[i];
rc = zpios_dmu_object_free(run_args,
region->obj.os,
region->obj.obj);
if (rc)
zpios_print(run_args->file, "Error "
"removing object %d, %d\n",
(int)region->obj.obj, rc);
}
} else {
region = &run_args->regions[0];
rc = zpios_dmu_object_free(run_args,
region->obj.os,
region->obj.obj);
if (rc)
zpios_print(run_args->file, "Error "
"removing object %d, %d\n",
(int)region->obj.obj, rc);
}
}
dmu_objset_disown(run_args->os, zpios_tag);
if (run_args->flags & DMU_REMOVE) {
rc = dsl_destroy_head(name);
if (rc)
zpios_print(run_args->file, "Error dsl_destroy_head"
"(%s, ...) failed: %d\n", name, rc);
}
t->stop = zpios_timespec_now();
t->delta = zpios_timespec_sub(t->stop, t->start);
(void)zpios_upcall(run_args->post, PHASE_POST_REMOVE, run_args, rc);
}
static void
zvol_last_close(zvol_state_t *zv)
{
zil_close(zv->zv_zilog);
zv->zv_zilog = NULL;
dmu_buf_rele(zv->zv_dbuf, zvol_tag);
zv->zv_dbuf = NULL;
/*
* Evict cached data
*/
if (dsl_dataset_is_dirty(dmu_objset_ds(zv->zv_objset)) &&
!(zv->zv_flags & ZVOL_RDONLY))
txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0);
(void) dmu_objset_evict_dbufs(zv->zv_objset);
dmu_objset_disown(zv->zv_objset, zvol_tag);
zv->zv_objset = NULL;
}
/*
* When we are called, os MUST refer to an objset associated with a dataset
* that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner
* == tag. We will then release and reacquire ownership of the dataset while
* holding the pool config_rwlock to avoid intervening namespace or ownership
* changes may occur.
*
* This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to
* release the hold on its dataset and acquire a new one on the dataset of the
* same name so that it can be partially torn down and reconstructed.
*/
void
dmu_objset_refresh_ownership(objset_t *os, void *tag)
{
dsl_pool_t *dp;
dsl_dataset_t *ds, *newds;
char name[MAXNAMELEN];
ds = os->os_dsl_dataset;
VERIFY3P(ds, !=, NULL);
VERIFY3P(ds->ds_owner, ==, tag);
VERIFY(dsl_dataset_long_held(ds));
dsl_dataset_name(ds, name);
dp = dmu_objset_pool(os);
dsl_pool_config_enter(dp, FTAG);
dmu_objset_disown(os, tag);
VERIFY0(dsl_dataset_own(dp, name, tag, &newds));
VERIFY3P(newds, ==, os->os_dsl_dataset);
dsl_pool_config_exit(dp, FTAG);
}
/*
* Given a full path to a file, translate into a dataset name and a relative
* path within the dataset. 'dataset' must be at least MAXNAMELEN characters,
* and 'relpath' must be at least MAXPATHLEN characters. We also pass a stat
* buffer, which we need later to get the object ID.
*/
static int
parse_pathname(const char *inpath, char *dataset, char *relpath,
struct stat *statbuf)
{
struct extmnttab mp;
FILE *fp;
int match;
const char *rel;
char fullpath[MAXPATHLEN];
compress_slashes(inpath, fullpath);
if (fullpath[0] != '/') {
(void) fprintf(stderr, "invalid object '%s': must be full "
"path\n", fullpath);
usage();
return (-1);
}
if (strlen(fullpath) >= MAXPATHLEN) {
(void) fprintf(stderr, "invalid object; pathname too long\n");
return (-1);
}
if (stat(fullpath, statbuf) != 0) {
(void) fprintf(stderr, "cannot open '%s': %s\n",
fullpath, strerror(errno));
return (-1);
}
#ifdef HAVE_SETMNTENT
if ((fp = setmntent(MNTTAB, "r")) == NULL) {
#else
if ((fp = fopen(MNTTAB, "r")) == NULL) {
#endif
(void) fprintf(stderr, "cannot open /etc/mtab\n");
return (-1);
}
match = 0;
while (getextmntent(fp, &mp, sizeof (mp)) == 0) {
if (makedev(mp.mnt_major, mp.mnt_minor) == statbuf->st_dev) {
match = 1;
break;
}
}
if (!match) {
(void) fprintf(stderr, "cannot find mountpoint for '%s'\n",
fullpath);
return (-1);
}
if (strcmp(mp.mnt_fstype, MNTTYPE_ZFS) != 0) {
(void) fprintf(stderr, "invalid path '%s': not a ZFS "
"filesystem\n", fullpath);
return (-1);
}
if (strncmp(fullpath, mp.mnt_mountp, strlen(mp.mnt_mountp)) != 0) {
(void) fprintf(stderr, "invalid path '%s': mountpoint "
"doesn't match path\n", fullpath);
return (-1);
}
(void) strcpy(dataset, mp.mnt_special);
rel = fullpath + strlen(mp.mnt_mountp);
if (rel[0] == '/')
rel++;
(void) strcpy(relpath, rel);
return (0);
}
#endif
//From FreeBSD
static int
parse_pathname(const char *inpath, char *dataset, char *relpath,
struct stat *statbuf)
{
struct statfs sfs;
const char *rel;
char fullpath[MAXPATHLEN];
compress_slashes(inpath, fullpath);
if (fullpath[0] != '/') {
(void) fprintf(stderr, "invalid object '%s': must be full "
"path\n", fullpath);
usage();
return (-1);
}
if (strlen(fullpath) >= MAXPATHLEN) {
(void) fprintf(stderr, "invalid object; pathname too long\n");
return (-1);
}
if (stat(fullpath, statbuf) != 0) {
(void) fprintf(stderr, "cannot open '%s': %s\n",
fullpath, strerror(errno));
return (-1);
}
if (statfs(fullpath, &sfs) == -1) {
(void) fprintf(stderr, "cannot find mountpoint for '%s': %s\n",
fullpath, strerror(errno));
return (-1);
}
if (strcmp(sfs.f_fstypename, MNTTYPE_ZFS) != 0) {
(void) fprintf(stderr, "invalid path '%s': not a ZFS "
"filesystem\n", fullpath);
return (-1);
}
if (strncmp(fullpath, sfs.f_mntonname, strlen(sfs.f_mntonname)) != 0) {
(void) fprintf(stderr, "invalid path '%s': mountpoint "
"doesn't match path\n", fullpath);
return (-1);
}
(void) strcpy(dataset, sfs.f_mntfromname);
rel = fullpath + strlen(sfs.f_mntonname);
if (rel[0] == '/')
rel++;
(void) strcpy(relpath, rel);
return (0);
}
/*
* Convert from a (dataset, path) pair into a (objset, object) pair. Note that
* we grab the object number from the inode number, since looking this up via
* libzpool is a real pain.
*/
/* ARGSUSED */
static int
object_from_path(const char *dataset, const char *path, struct stat *statbuf,
zinject_record_t *record)
{
objset_t *os;
int err;
/*
* Before doing any libzpool operations, call sync() to ensure that the
* on-disk state is consistent with the in-core state.
*/
sync();
err = dmu_objset_own(dataset, DMU_OST_ZFS, B_TRUE, B_FALSE, FTAG, &os);
if (err != 0) {
(void) fprintf(stderr, "cannot open dataset '%s': %s\n",
dataset, strerror(err));
return (-1);
}
record->zi_objset = dmu_objset_id(os);
record->zi_object = statbuf->st_ino;
dmu_objset_disown(os, B_FALSE, FTAG);
return (0);
}
/*
* Create a block device minor node and setup the linkage between it
* and the specified volume. Once this function returns the block
* device is live and ready for use.
*/
static int
zvol_create_minor_impl(const char *name)
{
zvol_state_t *zv;
objset_t *os;
dmu_object_info_t *doi;
uint64_t volsize;
uint64_t len;
unsigned minor = 0;
int error = 0;
mutex_enter(&zvol_state_lock);
zv = zvol_find_by_name(name);
if (zv) {
error = SET_ERROR(EEXIST);
goto out;
}
doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP);
error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, zvol_tag, &os);
if (error)
goto out_doi;
error = dmu_object_info(os, ZVOL_OBJ, doi);
if (error)
goto out_dmu_objset_disown;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error)
goto out_dmu_objset_disown;
error = zvol_find_minor(&minor);
if (error)
goto out_dmu_objset_disown;
zv = zvol_alloc(MKDEV(zvol_major, minor), name);
if (zv == NULL) {
error = SET_ERROR(EAGAIN);
goto out_dmu_objset_disown;
}
if (dmu_objset_is_snapshot(os))
zv->zv_flags |= ZVOL_RDONLY;
zv->zv_volblocksize = doi->doi_data_block_size;
zv->zv_volsize = volsize;
zv->zv_objset = os;
set_capacity(zv->zv_disk, zv->zv_volsize >> 9);
blk_queue_max_hw_sectors(zv->zv_queue, (DMU_MAX_ACCESS / 4) >> 9);
blk_queue_max_segments(zv->zv_queue, UINT16_MAX);
blk_queue_max_segment_size(zv->zv_queue, UINT_MAX);
blk_queue_physical_block_size(zv->zv_queue, zv->zv_volblocksize);
blk_queue_io_opt(zv->zv_queue, zv->zv_volblocksize);
blk_queue_max_discard_sectors(zv->zv_queue,
(zvol_max_discard_blocks * zv->zv_volblocksize) >> 9);
blk_queue_discard_granularity(zv->zv_queue, zv->zv_volblocksize);
queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, zv->zv_queue);
#ifdef QUEUE_FLAG_NONROT
queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zv->zv_queue);
#endif
#ifdef QUEUE_FLAG_ADD_RANDOM
queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, zv->zv_queue);
#endif
if (spa_writeable(dmu_objset_spa(os))) {
if (zil_replay_disable)
zil_destroy(dmu_objset_zil(os), B_FALSE);
else
zil_replay(os, zv, zvol_replay_vector);
}
/*
* When udev detects the addition of the device it will immediately
* invoke blkid(8) to determine the type of content on the device.
* Prefetching the blocks commonly scanned by blkid(8) will speed
* up this process.
*/
len = MIN(MAX(zvol_prefetch_bytes, 0), SPA_MAXBLOCKSIZE);
if (len > 0) {
dmu_prefetch(os, ZVOL_OBJ, 0, 0, len, ZIO_PRIORITY_SYNC_READ);
dmu_prefetch(os, ZVOL_OBJ, 0, volsize - len, len,
ZIO_PRIORITY_SYNC_READ);
}
zv->zv_objset = NULL;
out_dmu_objset_disown:
dmu_objset_disown(os, zvol_tag);
out_doi:
kmem_free(doi, sizeof (dmu_object_info_t));
out:
if (error == 0) {
zvol_insert(zv);
/*
* Drop the lock to prevent deadlock with sys_open() ->
* zvol_open(), which first takes bd_disk->bd_mutex and then
* takes zvol_state_lock, whereas this code path first takes
* zvol_state_lock, and then takes bd_disk->bd_mutex.
*/
mutex_exit(&zvol_state_lock);
add_disk(zv->zv_disk);
} else {
mutex_exit(&zvol_state_lock);
}
return (SET_ERROR(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
__zvol_create_minor(const char *name)
{
zvol_state_t *zv;
objset_t *os;
dmu_object_info_t *doi;
uint64_t volsize;
unsigned minor = 0;
int error = 0;
ASSERT(MUTEX_HELD(&zvol_state_lock));
zv = zvol_find_by_name(name);
if (zv) {
error = EEXIST;
goto out;
}
doi = kmem_alloc(sizeof(dmu_object_info_t), KM_SLEEP);
error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, zvol_tag, &os);
if (error)
goto out_doi;
/* Make sure we have the key loaded if we need one. */
error = dsl_crypto_key_inherit(name);
if (error != 0 && error != EEXIST)
goto out_dmu_objset_disown;
error = dmu_object_info(os, ZVOL_OBJ, doi);
if (error)
goto out_dmu_objset_disown;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error)
goto out_dmu_objset_disown;
error = zvol_find_minor(&minor);
if (error)
goto out_dmu_objset_disown;
zv = zvol_alloc(MKDEV(zvol_major, minor), name);
if (zv == NULL) {
error = EAGAIN;
goto out_dmu_objset_disown;
}
if (dmu_objset_is_snapshot(os))
zv->zv_flags |= ZVOL_RDONLY;
zv->zv_volblocksize = doi->doi_data_block_size;
zv->zv_volsize = volsize;
zv->zv_objset = os;
set_capacity(zv->zv_disk, zv->zv_volsize >> 9);
blk_queue_max_hw_sectors(zv->zv_queue, UINT_MAX);
blk_queue_max_segments(zv->zv_queue, UINT16_MAX);
blk_queue_max_segment_size(zv->zv_queue, UINT_MAX);
blk_queue_physical_block_size(zv->zv_queue, zv->zv_volblocksize);
blk_queue_io_opt(zv->zv_queue, zv->zv_volblocksize);
#ifdef HAVE_BLK_QUEUE_DISCARD
blk_queue_max_discard_sectors(zv->zv_queue,
(zvol_max_discard_blocks * zv->zv_volblocksize) >> 9);
blk_queue_discard_granularity(zv->zv_queue, zv->zv_volblocksize);
queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, zv->zv_queue);
#endif
#ifdef HAVE_BLK_QUEUE_NONROT
queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zv->zv_queue);
#endif
if (zil_replay_disable)
zil_destroy(dmu_objset_zil(os), B_FALSE);
else
zil_replay(os, zv, zvol_replay_vector);
out_dmu_objset_disown:
dmu_objset_disown(os, zvol_tag);
zv->zv_objset = NULL;
out_doi:
kmem_free(doi, sizeof(dmu_object_info_t));
out:
if (error == 0) {
zvol_insert(zv);
add_disk(zv->zv_disk);
}
return (error);
}
int
zfs_sb_create(const char *osname, zfs_mntopts_t *zmo, zfs_sb_t **zsbp)
{
objset_t *os;
zfs_sb_t *zsb;
uint64_t zval;
int i, error;
uint64_t sa_obj;
zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP);
/*
* We claim to always be readonly so we can open snapshots;
* other ZPL code will prevent us from writing to snapshots.
*/
error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os);
if (error) {
kmem_free(zsb, sizeof (zfs_sb_t));
return (error);
}
/*
* Optional temporary mount options, free'd in zfs_sb_free().
*/
zsb->z_mntopts = (zmo ? zmo : zfs_mntopts_alloc());
/*
* Initialize the zfs-specific filesystem structure.
* Should probably make this a kmem cache, shuffle fields,
* and just bzero up to z_hold_mtx[].
*/
zsb->z_sb = NULL;
zsb->z_parent = zsb;
zsb->z_max_blksz = SPA_OLD_MAXBLOCKSIZE;
zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
zsb->z_os = os;
error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version);
if (error) {
goto out;
} else if (zsb->z_version > ZPL_VERSION) {
error = SET_ERROR(ENOTSUP);
goto out;
}
if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
goto out;
zsb->z_norm = (int)zval;
if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
goto out;
zsb->z_utf8 = (zval != 0);
if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0)
goto out;
zsb->z_case = (uint_t)zval;
if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &zval)) != 0)
goto out;
zsb->z_acl_type = (uint_t)zval;
/*
* Fold case on file systems that are always or sometimes case
* insensitive.
*/
if (zsb->z_case == ZFS_CASE_INSENSITIVE ||
zsb->z_case == ZFS_CASE_MIXED)
zsb->z_norm |= U8_TEXTPREP_TOUPPER;
zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os);
zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os);
if (zsb->z_use_sa) {
/* should either have both of these objects or none */
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1,
&sa_obj);
if (error)
goto out;
error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval);
if ((error == 0) && (zval == ZFS_XATTR_SA))
zsb->z_xattr_sa = B_TRUE;
} else {
/*
* Pre SA versions file systems should never touch
* either the attribute registration or layout objects.
*/
sa_obj = 0;
}
error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
&zsb->z_attr_table);
if (error)
goto out;
if (zsb->z_version >= ZPL_VERSION_SA)
sa_register_update_callback(os, zfs_sa_upgrade);
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
&zsb->z_root);
if (error)
goto out;
ASSERT(zsb->z_root != 0);
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
&zsb->z_unlinkedobj);
if (error)
goto out;
error = zap_lookup(os, MASTER_NODE_OBJ,
zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA],
8, 1, &zsb->z_userquota_obj);
if (error && error != ENOENT)
goto out;
error = zap_lookup(os, MASTER_NODE_OBJ,
zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA],
8, 1, &zsb->z_groupquota_obj);
if (error && error != ENOENT)
goto out;
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
&zsb->z_fuid_obj);
if (error && error != ENOENT)
goto out;
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1,
&zsb->z_shares_dir);
if (error && error != ENOENT)
goto out;
mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL);
list_create(&zsb->z_all_znodes, sizeof (znode_t),
offsetof(znode_t, z_link_node));
rrm_init(&zsb->z_teardown_lock, B_FALSE);
rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL);
zsb->z_hold_mtx = vmem_zalloc(sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ,
KM_SLEEP);
for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
*zsbp = zsb;
return (0);
out:
dmu_objset_disown(os, zsb);
*zsbp = NULL;
vmem_free(zsb->z_hold_mtx, sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ);
kmem_free(zsb, sizeof (zfs_sb_t));
return (error);
}
static int
__zvol_create_minor(const char *name)
{
zvol_state_t *zv;
objset_t *os;
dmu_object_info_t *doi;
uint64_t volsize;
unsigned minor = 0;
int error = 0;
ASSERT(MUTEX_HELD(&zvol_state_lock));
zv = zvol_find_by_name(name);
if (zv) {
error = EEXIST;
goto out;
}
doi = kmem_alloc(sizeof(dmu_object_info_t), KM_SLEEP);
error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, zvol_tag, &os);
if (error)
goto out_doi;
error = dmu_object_info(os, ZVOL_OBJ, doi);
if (error)
goto out_dmu_objset_disown;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error)
goto out_dmu_objset_disown;
error = zvol_find_minor(&minor);
if (error)
goto out_dmu_objset_disown;
zv = zvol_alloc(MKDEV(zvol_major, minor), name);
if (zv == NULL) {
error = EAGAIN;
goto out_dmu_objset_disown;
}
if (dmu_objset_is_snapshot(os))
zv->zv_flags |= ZVOL_RDONLY;
zv->zv_volblocksize = doi->doi_data_block_size;
zv->zv_volsize = volsize;
zv->zv_objset = os;
set_capacity(zv->zv_disk, zv->zv_volsize >> 9);
if (zil_replay_disable)
zil_destroy(dmu_objset_zil(os), B_FALSE);
else
zil_replay(os, zv, zvol_replay_vector);
out_dmu_objset_disown:
dmu_objset_disown(os, zvol_tag);
zv->zv_objset = NULL;
out_doi:
kmem_free(doi, sizeof(dmu_object_info_t));
out:
if (error == 0) {
zvol_insert(zv);
add_disk(zv->zv_disk);
}
return (error);
}
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);
}
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);
}