int
dmu_objset_destroy(const char *name)
{
objset_t *os;
int error;
/*
* If it looks like we'll be able to destroy it, and there's
* an unplayed replay log sitting around, destroy the log.
* It would be nicer to do this in dsl_dataset_destroy_sync(),
* but the replay log objset is modified in open context.
*/
error = dmu_objset_open(name, DMU_OST_ANY,
DS_MODE_OWNER|DS_MODE_READONLY|DS_MODE_INCONSISTENT, &os);
if (error == 0) {
dsl_dataset_t *ds = os->os->os_dsl_dataset;
zil_destroy(dmu_objset_zil(os), B_FALSE);
error = dsl_dataset_destroy(ds, os);
/*
* dsl_dataset_destroy() closes the ds.
*/
kmem_free(os, sizeof (objset_t));
}
return (error);
}
/* ARGSUSED */
int
zil_clear_log_chain(char *osname, void *txarg)
{
zilog_t *zilog;
zil_header_t *zh;
objset_t *os;
dmu_tx_t *tx;
int error;
error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os);
if (error) {
cmn_err(CE_WARN, "can't open objset for %s", osname);
return (0);
}
zilog = dmu_objset_zil(os);
tx = dmu_tx_create(zilog->zl_os);
(void) dmu_tx_assign(tx, TXG_WAIT);
zh = zil_header_in_syncing_context(zilog);
BP_ZERO(&zh->zh_log);
dsl_dataset_dirty(dmu_objset_ds(os), tx);
dmu_tx_commit(tx);
dmu_objset_close(os);
return (0);
}
int
zil_claim(char *osname, void *txarg)
{
dmu_tx_t *tx = txarg;
uint64_t first_txg = dmu_tx_get_txg(tx);
zilog_t *zilog;
zil_header_t *zh;
objset_t *os;
int error;
error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os);
if (error) {
cmn_err(CE_WARN, "can't open objset for %s", osname);
return (0);
}
zilog = dmu_objset_zil(os);
zh = zil_header_in_syncing_context(zilog);
/*
* Record here whether the zil has any records to replay.
* If the header block pointer is null or the block points
* to the stubby then we know there are no valid log records.
* We use the header to store this state as the the zilog gets
* freed later in dmu_objset_close().
* The flags (and the rest of the header fields) are cleared in
* zil_sync() as a result of a zil_destroy(), after replaying the log.
*
* Note, the intent log can be empty but still need the
* stubby to be claimed.
*/
if (!zil_empty(zilog))
zh->zh_flags |= ZIL_REPLAY_NEEDED;
/*
* Claim all log blocks if we haven't already done so, and remember
* the highest claimed sequence number. This ensures that if we can
* read only part of the log now (e.g. due to a missing device),
* but we can read the entire log later, we will not try to replay
* or destroy beyond the last block we successfully claimed.
*/
ASSERT3U(zh->zh_claim_txg, <=, first_txg);
if (zh->zh_claim_txg == 0 && !BP_IS_HOLE(&zh->zh_log)) {
zh->zh_claim_txg = first_txg;
zh->zh_claim_seq = zil_parse(zilog, zil_claim_log_block,
zil_claim_log_record, tx, first_txg);
dsl_dataset_dirty(dmu_objset_ds(os), tx);
}
ASSERT3U(first_txg, ==, (spa_last_synced_txg(zilog->zl_spa) + 1));
dmu_objset_close(os);
return (0);
}
int
zfs_set_version(const char *name, uint64_t newvers)
{
int error;
objset_t *os;
dmu_tx_t *tx;
uint64_t curvers;
/*
* XXX for now, require that the filesystem be unmounted. Would
* be nice to find the zfsvfs_t and just update that if
* possible.
*/
if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION)
return (EINVAL);
error = dmu_objset_open(name, DMU_OST_ZFS, DS_MODE_PRIMARY, &os);
if (error)
return (error);
error = zap_lookup(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
8, 1, &curvers);
if (error)
goto out;
if (newvers < curvers) {
error = EINVAL;
goto out;
}
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, 0, ZPL_VERSION_STR);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
goto out;
}
error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, 8, 1,
&newvers, tx);
spa_history_internal_log(LOG_DS_UPGRADE,
dmu_objset_spa(os), tx, CRED(),
"oldver=%llu newver=%llu dataset = %llu", curvers, newvers,
dmu_objset_id(os));
dmu_tx_commit(tx);
out:
dmu_objset_close(os);
return (error);
}
static int
dmu_objset_snapshot_one(char *name, void *arg)
{
struct snaparg *sn = arg;
objset_t *os;
int err;
(void) strcpy(sn->failed, name);
/*
* Check permissions only when requested. This only applies when
* doing a recursive snapshot. The permission checks for the starting
* dataset have already been performed in zfs_secpolicy_snapshot()
*/
if (sn->checkperms == B_TRUE &&
(err = zfs_secpolicy_snapshot_perms(name, CRED())))
return (err);
err = dmu_objset_open(name, DMU_OST_ANY, DS_MODE_USER, &os);
if (err != 0)
return (err);
/* If the objset is in an inconsistent state, return busy */
if (os->os->os_dsl_dataset->ds_phys->ds_flags & DS_FLAG_INCONSISTENT) {
dmu_objset_close(os);
return (EBUSY);
}
/*
* NB: we need to wait for all in-flight changes to get to disk,
* so that we snapshot those changes. zil_suspend does this as
* a side effect.
*/
err = zil_suspend(dmu_objset_zil(os));
if (err == 0) {
struct osnode *osn;
dsl_sync_task_create(sn->dstg, dsl_dataset_snapshot_check,
dsl_dataset_snapshot_sync, os->os->os_dsl_dataset,
sn->snapname, 3);
osn = kmem_alloc(sizeof (struct osnode), KM_SLEEP);
osn->os = os;
list_insert_tail(&sn->objsets, osn);
} else {
dmu_objset_close(os);
}
return (err);
}
/*
* Reopen zfsvfs_t::z_os and release VOPs.
*/
int
zfs_resume_fs(zfsvfs_t *zfsvfs, const char *osname, int mode)
{
int err;
ASSERT(RRW_WRITE_HELD(&zfsvfs->z_teardown_lock));
ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock));
err = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os);
if (err) {
zfsvfs->z_os = NULL;
} else {
znode_t *zp;
VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0);
/*
* Attempt to re-establish all the active znodes with
* their dbufs. If a zfs_rezget() fails, then we'll let
* any potential callers discover that via ZFS_ENTER_VERIFY_VP
* when they try to use their znode.
*/
mutex_enter(&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);
}
/* release the VOPs */
rw_exit(&zfsvfs->z_teardown_inactive_lock);
rrw_exit(&zfsvfs->z_teardown_lock, FTAG);
if (err) {
/*
* Since we couldn't reopen zfsvfs::z_os, force
* unmount this file system.
*/
if (vn_vfswlock(zfsvfs->z_vfs->vfs_vnodecovered) == 0)
(void) dounmount(zfsvfs->z_vfs, MS_FORCE, curthread);
}
return (err);
}
/* ARGSUSED */
int
zil_vdev_offline(char *osname, void *arg)
{
objset_t *os;
zilog_t *zilog;
int error;
error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os);
if (error)
return (error);
zilog = dmu_objset_zil(os);
if (zil_suspend(zilog) != 0)
error = EEXIST;
else
zil_resume(zilog);
dmu_objset_close(os);
return (error);
}
/* ARGSUSED */
int
zil_check_log_chain(char *osname, void *txarg)
{
zilog_t *zilog;
zil_header_t *zh;
blkptr_t blk;
arc_buf_t *abuf;
objset_t *os;
char *lrbuf;
zil_trailer_t *ztp;
int error;
error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os);
if (error) {
cmn_err(CE_WARN, "can't open objset for %s", osname);
return (0);
}
zilog = dmu_objset_zil(os);
zh = zil_header_in_syncing_context(zilog);
blk = zh->zh_log;
if (BP_IS_HOLE(&blk)) {
dmu_objset_close(os);
return (0); /* no chain */
}
for (;;) {
error = zil_read_log_block(zilog, &blk, &abuf);
if (error)
break;
lrbuf = abuf->b_data;
ztp = (zil_trailer_t *)(lrbuf + BP_GET_LSIZE(&blk)) - 1;
blk = ztp->zit_next_blk;
VERIFY(arc_buf_remove_ref(abuf, &abuf) == 1);
}
dmu_objset_close(os);
if (error == ECKSUM)
return (0); /* normal end of chain */
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);
}
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);
}
/*
* Create a minor node for the specified volume.
*/
int
zvol_create_minor(zfs_cmd_t *zc)
{
char *name = zc->zc_name;
dev_t dev = zc->zc_dev;
zvol_state_t *zv;
objset_t *os;
uint64_t volsize;
minor_t minor = 0;
struct pathname linkpath;
int ds_mode = DS_MODE_PRIMARY;
vnode_t *vp = NULL;
char *devpath;
size_t devpathlen = strlen(ZVOL_FULL_DEV_DIR) + 1 + strlen(name) + 1;
char chrbuf[30], blkbuf[30];
int error;
mutex_enter(&zvol_state_lock);
if ((zv = zvol_minor_lookup(name)) != NULL) {
mutex_exit(&zvol_state_lock);
return (EEXIST);
}
if (strchr(name, '@') != 0)
ds_mode |= DS_MODE_READONLY;
error = dmu_objset_open(name, DMU_OST_ZVOL, ds_mode, &os);
if (error) {
mutex_exit(&zvol_state_lock);
return (error);
}
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error) {
dmu_objset_close(os);
mutex_exit(&zvol_state_lock);
return (error);
}
/*
* If there's an existing /dev/zvol symlink, try to use the
* same minor number we used last time.
*/
devpath = kmem_alloc(devpathlen, KM_SLEEP);
(void) sprintf(devpath, "%s/%s", ZVOL_FULL_DEV_DIR, name);
error = lookupname(devpath, UIO_SYSSPACE, NO_FOLLOW, NULL, &vp);
kmem_free(devpath, devpathlen);
if (error == 0 && vp->v_type != VLNK)
error = EINVAL;
if (error == 0) {
pn_alloc(&linkpath);
error = pn_getsymlink(vp, &linkpath, kcred);
if (error == 0) {
char *ms = strstr(linkpath.pn_path, ZVOL_PSEUDO_DEV);
if (ms != NULL) {
ms += strlen(ZVOL_PSEUDO_DEV);
minor = stoi(&ms);
}
}
pn_free(&linkpath);
}
if (vp != NULL)
VN_RELE(vp);
/*
* If we found a minor but it's already in use, we must pick a new one.
*/
if (minor != 0 && ddi_get_soft_state(zvol_state, minor) != NULL)
minor = 0;
if (minor == 0)
minor = zvol_minor_alloc();
if (minor == 0) {
dmu_objset_close(os);
mutex_exit(&zvol_state_lock);
return (ENXIO);
}
if (ddi_soft_state_zalloc(zvol_state, minor) != DDI_SUCCESS) {
dmu_objset_close(os);
mutex_exit(&zvol_state_lock);
return (EAGAIN);
}
(void) ddi_prop_update_string(minor, zfs_dip, ZVOL_PROP_NAME, name);
(void) sprintf(chrbuf, "%uc,raw", minor);
if (ddi_create_minor_node(zfs_dip, chrbuf, S_IFCHR,
minor, DDI_PSEUDO, 0) == DDI_FAILURE) {
ddi_soft_state_free(zvol_state, minor);
dmu_objset_close(os);
mutex_exit(&zvol_state_lock);
return (EAGAIN);
}
(void) sprintf(blkbuf, "%uc", minor);
if (ddi_create_minor_node(zfs_dip, blkbuf, S_IFBLK,
minor, DDI_PSEUDO, 0) == DDI_FAILURE) {
ddi_remove_minor_node(zfs_dip, chrbuf);
ddi_soft_state_free(zvol_state, minor);
dmu_objset_close(os);
mutex_exit(&zvol_state_lock);
return (EAGAIN);
}
zv = ddi_get_soft_state(zvol_state, minor);
(void) strcpy(zv->zv_name, name);
zv->zv_min_bs = DEV_BSHIFT;
zv->zv_minor = minor;
zv->zv_volsize = volsize;
zv->zv_objset = os;
zv->zv_mode = ds_mode;
zv->zv_zilog = zil_open(os, NULL);
rw_init(&zv->zv_dslock, NULL, RW_DEFAULT, NULL);
zil_replay(os, zv, &zv->zv_txg_assign, zvol_replay_vector, NULL);
zvol_size_changed(zv, dev);
/* XXX this should handle the possible i/o error */
VERIFY(dsl_prop_register(dmu_objset_ds(zv->zv_objset),
"readonly", zvol_readonly_changed_cb, zv) == 0);
zvol_minors++;
mutex_exit(&zvol_state_lock);
return (0);
}
/* ARGSUSED */
static int
zfsctl_snapdir_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
int *direntflags, pathname_t *realpnp)
{
zfsctl_snapdir_t *sdp = dvp->v_data;
objset_t *snap;
char snapname[MAXNAMELEN];
char real[MAXNAMELEN];
char *mountpoint;
zfs_snapentry_t *sep, search;
struct mounta margs;
vfs_t *vfsp;
size_t mountpoint_len;
avl_index_t where;
zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
int err;
/*
* No extended attributes allowed under .zfs
*/
if (flags & LOOKUP_XATTR)
return (EINVAL);
ASSERT(dvp->v_type == VDIR);
/*
* If we get a recursive call, that means we got called
* from the domount() code while it was trying to look up the
* spec (which looks like a local path for zfs). We need to
* add some flag to domount() to tell it not to do this lookup.
*/
if (MUTEX_HELD(&sdp->sd_lock))
return (ENOENT);
ZFS_ENTER(zfsvfs);
if (gfs_lookup_dot(vpp, dvp, zfsvfs->z_ctldir, nm) == 0) {
ZFS_EXIT(zfsvfs);
return (0);
}
if (flags & FIGNORECASE) {
boolean_t conflict = B_FALSE;
err = dmu_snapshot_realname(zfsvfs->z_os, nm, real,
MAXNAMELEN, &conflict);
if (err == 0) {
nm = real;
} else if (err != ENOTSUP) {
ZFS_EXIT(zfsvfs);
return (err);
}
if (realpnp)
(void) strlcpy(realpnp->pn_buf, nm,
realpnp->pn_bufsize);
if (conflict && direntflags)
*direntflags = ED_CASE_CONFLICT;
}
mutex_enter(&sdp->sd_lock);
search.se_name = (char *)nm;
if ((sep = avl_find(&sdp->sd_snaps, &search, &where)) != NULL) {
*vpp = sep->se_root;
VN_HOLD(*vpp);
err = traverse(vpp);
if (err) {
VN_RELE(*vpp);
*vpp = NULL;
} else if (*vpp == sep->se_root) {
/*
* The snapshot was unmounted behind our backs,
* try to remount it.
*/
goto domount;
} else {
/*
* VROOT was set during the traverse call. We need
* to clear it since we're pretending to be part
* of our parent's vfs.
*/
(*vpp)->v_flag &= ~VROOT;
}
mutex_exit(&sdp->sd_lock);
ZFS_EXIT(zfsvfs);
return (err);
}
/*
* The requested snapshot is not currently mounted, look it up.
*/
err = zfsctl_snapshot_zname(dvp, nm, MAXNAMELEN, snapname);
if (err) {
mutex_exit(&sdp->sd_lock);
ZFS_EXIT(zfsvfs);
/*
* handle "ls *" or "?" in a graceful manner,
* forcing EILSEQ to ENOENT.
* Since shell ultimately passes "*" or "?" as name to lookup
*/
return (err == EILSEQ ? ENOENT : err);
}
if (dmu_objset_open(snapname, DMU_OST_ZFS,
DS_MODE_USER | DS_MODE_READONLY, &snap) != 0) {
mutex_exit(&sdp->sd_lock);
ZFS_EXIT(zfsvfs);
return (ENOENT);
}
sep = kmem_alloc(sizeof (zfs_snapentry_t), KM_SLEEP);
sep->se_name = kmem_alloc(strlen(nm) + 1, KM_SLEEP);
(void) strcpy(sep->se_name, nm);
*vpp = sep->se_root = zfsctl_snapshot_mknode(dvp, dmu_objset_id(snap));
avl_insert(&sdp->sd_snaps, sep, where);
dmu_objset_close(snap);
domount:
mountpoint_len = strlen(refstr_value(dvp->v_vfsp->vfs_mntpt)) +
strlen("/.zfs/snapshot/") + strlen(nm) + 1;
mountpoint = kmem_alloc(mountpoint_len, KM_SLEEP);
(void) snprintf(mountpoint, mountpoint_len, "%s/.zfs/snapshot/%s",
refstr_value(dvp->v_vfsp->vfs_mntpt), nm);
margs.spec = snapname;
margs.dir = mountpoint;
margs.flags = MS_SYSSPACE | MS_NOMNTTAB;
margs.fstype = "zfs";
margs.dataptr = NULL;
margs.datalen = 0;
margs.optptr = NULL;
margs.optlen = 0;
err = domount("zfs", &margs, *vpp, kcred, &vfsp);
kmem_free(mountpoint, mountpoint_len);
if (err == 0) {
/*
* Return the mounted root rather than the covered mount point.
* Takes the GFS vnode at .zfs/snapshot/<snapname> and returns
* the ZFS vnode mounted on top of the GFS node. This ZFS
* vnode is the root of the newly created vfsp.
*/
VFS_RELE(vfsp);
err = traverse(vpp);
}
if (err == 0) {
/*
* Fix up the root vnode mounted on .zfs/snapshot/<snapname>.
*
* This is where we lie about our v_vfsp in order to
* make .zfs/snapshot/<snapname> accessible over NFS
* without requiring manual mounts of <snapname>.
*/
ASSERT(VTOZ(*vpp)->z_zfsvfs != zfsvfs);
VTOZ(*vpp)->z_zfsvfs->z_parent = zfsvfs;
(*vpp)->v_vfsp = zfsvfs->z_vfs;
(*vpp)->v_flag &= ~VROOT;
}
mutex_exit(&sdp->sd_lock);
ZFS_EXIT(zfsvfs);
/*
* If we had an error, drop our hold on the vnode and
* zfsctl_snapshot_inactive() will clean up.
*/
if (err) {
VN_RELE(*vpp);
*vpp = NULL;
}
return (err);
}
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);
}
