static uint64_t
zpios_dmu_object_create(run_args_t *run_args, objset_t *os)
{
struct dmu_tx *tx;
uint64_t obj = 0ULL;
int rc;
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, OBJ_SIZE);
rc = dmu_tx_assign(tx, TXG_WAIT);
if (rc) {
zpios_print(run_args->file,
"dmu_tx_assign() failed: %d\n", rc);
dmu_tx_abort(tx);
return obj;
}
obj = dmu_object_alloc(os, DMU_OT_UINT64_OTHER, 0,
DMU_OT_NONE, 0, tx);
rc = dmu_object_set_blocksize(os, obj, 128ULL << 10, 0, tx);
if (rc) {
zpios_print(run_args->file,
"dmu_object_set_blocksize() failed: %d\n", rc);
dmu_tx_abort(tx);
return obj;
}
dmu_tx_commit(tx);
return obj;
}
static int
zpios_dmu_object_free(run_args_t *run_args, objset_t *os, uint64_t obj)
{
struct dmu_tx *tx;
int rc;
tx = dmu_tx_create(os);
dmu_tx_hold_free(tx, obj, 0, DMU_OBJECT_END);
rc = dmu_tx_assign(tx, TXG_WAIT);
if (rc) {
zpios_print(run_args->file,
"dmu_tx_assign() failed: %d\n", rc);
dmu_tx_abort(tx);
return (rc);
}
rc = dmu_object_free(os, obj, tx);
if (rc) {
zpios_print(run_args->file,
"dmu_object_free() failed: %d\n", rc);
dmu_tx_abort(tx);
return (rc);
}
dmu_tx_commit(tx);
return (0);
}
static int
zfs_vfs_sync(struct mount *mp, __unused int waitfor, __unused vfs_context_t context)
{
zfsvfs_t *zfsvfs = vfs_fsprivate(mp);
ZFS_ENTER(zfsvfs);
/*
* Mac OS X needs a file system modify time
*
* We use the mtime of the "com.apple.system.mtime"
* extended attribute, which is associated with the
* file system root directory.
*
* Here we sync any mtime changes to this attribute.
*/
if (zfsvfs->z_mtime_vp != NULL) {
timestruc_t mtime;
znode_t *zp;
top:
zp = VTOZ(zfsvfs->z_mtime_vp);
ZFS_TIME_DECODE(&mtime, zp->z_phys->zp_mtime);
if (zfsvfs->z_last_mtime_synced < mtime.tv_sec) {
dmu_tx_t *tx;
int error;
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_bonus(tx, zp->z_id);
error = dmu_tx_assign(tx, zfsvfs->z_assign);
if (error) {
if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
dmu_tx_abort(tx);
} else {
dmu_buf_will_dirty(zp->z_dbuf, tx);
dmu_tx_commit(tx);
zfsvfs->z_last_mtime_synced = mtime.tv_sec;
}
}
}
if (zfsvfs->z_log != NULL)
zil_commit(zfsvfs->z_log, UINT64_MAX, 0);
else
txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
ZFS_EXIT(zfsvfs);
return (0);
}
/*
* TX_WRITE2 are only generated when dmu_sync() returns EALREADY
* meaning the pool block is already being synced. So now that we always write
* out full blocks, all we have to do is expand the eof if
* the file is grown.
*/
static int
zfs_replay_write2(zfsvfs_t *zfsvfs, void *data, boolean_t byteswap)
{
#ifndef TODO_OSV
kprintf("TX_WRITE2\n");
return EOPNOTSUPP;
#else
lr_write_t *lr = data;
znode_t *zp;
int error;
uint64_t end;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
return (error);
top:
end = lr->lr_offset + lr->lr_length;
if (end > zp->z_size) {
dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
zp->z_size = end;
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
VN_RELE(ZTOV(zp));
if (error == ERESTART) {
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
dmu_tx_abort(tx);
return (error);
}
(void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
(void *)&zp->z_size, sizeof (uint64_t), tx);
/* Ensure the replayed seq is updated */
(void) zil_replaying(zfsvfs->z_log, tx);
dmu_tx_commit(tx);
}
VN_RELE(ZTOV(zp));
return (error);
#endif
}
int
zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type,
const char *domain, uint64_t rid, uint64_t quota)
{
char buf[32];
int err;
dmu_tx_t *tx;
uint64_t *objp;
boolean_t fuid_dirtied;
if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA)
return (SET_ERROR(EINVAL));
if (zsb->z_version < ZPL_VERSION_USERSPACE)
return (SET_ERROR(ENOTSUP));
objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj :
&zsb->z_groupquota_obj;
err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE);
if (err)
return (err);
fuid_dirtied = zsb->z_fuid_dirty;
tx = dmu_tx_create(zsb->z_os);
dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL);
if (*objp == 0) {
dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
zfs_userquota_prop_prefixes[type]);
}
if (fuid_dirtied)
zfs_fuid_txhold(zsb, tx);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err) {
dmu_tx_abort(tx);
return (err);
}
mutex_enter(&zsb->z_lock);
if (*objp == 0) {
*objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA,
DMU_OT_NONE, 0, tx);
VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ,
zfs_userquota_prop_prefixes[type], 8, 1, objp, tx));
}
mutex_exit(&zsb->z_lock);
if (quota == 0) {
err = zap_remove(zsb->z_os, *objp, buf, tx);
if (err == ENOENT)
err = 0;
} else {
err = zap_update(zsb->z_os, *objp, buf, 8, 1, "a, tx);
}
ASSERT(err == 0);
if (fuid_dirtied)
zfs_fuid_sync(zsb, tx);
dmu_tx_commit(tx);
return (err);
}
int
spa_history_log_nvl(spa_t *spa, nvlist_t *nvl)
{
int err = 0;
dmu_tx_t *tx;
nvlist_t *nvarg;
if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY || !spa_writeable(spa))
return (SET_ERROR(EINVAL));
tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err) {
dmu_tx_abort(tx);
return (err);
}
VERIFY0(nvlist_dup(nvl, &nvarg, KM_SLEEP));
if (spa_history_zone() != NULL) {
fnvlist_add_string(nvarg, ZPOOL_HIST_ZONE,
spa_history_zone());
}
fnvlist_add_uint64(nvarg, ZPOOL_HIST_WHO, crgetruid(CRED()));
/* Kick this off asynchronously; errors are ignored. */
dsl_sync_task_nowait(spa_get_dsl(spa), spa_history_log_sync,
nvarg, 0, ZFS_SPACE_CHECK_NONE, tx);
dmu_tx_commit(tx);
/* spa_history_log_sync will free nvl */
return (err);
}
void
spa_history_log_internal(spa_t *spa, const char *operation,
dmu_tx_t *tx, const char *fmt, ...)
{
dmu_tx_t *htx = tx;
va_list adx;
nvlist_t *nvl;
/* create a tx if we didn't get one */
if (tx == NULL) {
htx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
if (dmu_tx_assign(htx, TXG_WAIT) != 0) {
dmu_tx_abort(htx);
return;
}
}
va_start(adx, fmt);
VERIFY0(nvlist_alloc(&nvl, NV_UNIQUE_NAME, KM_PUSHPAGE));
log_internal(nvl, operation, spa, htx, fmt, adx);
va_end(adx);
/* if we didn't get a tx from the caller, commit the one we made */
if (tx == NULL)
dmu_tx_commit(htx);
}
/*
* Ensure the zap is flushed then inform the VFS of the capacity change.
*/
static int
zvol_update_volsize(uint64_t volsize, objset_t *os)
{
dmu_tx_t *tx;
int error;
uint64_t txg;
ASSERT(MUTEX_HELD(&zvol_state_lock));
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
return (SET_ERROR(error));
}
txg = dmu_tx_get_txg(tx);
error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1,
&volsize, tx);
dmu_tx_commit(tx);
txg_wait_synced(dmu_objset_pool(os), txg);
if (error == 0)
error = dmu_free_long_range(os,
ZVOL_OBJ, volsize, DMU_OBJECT_END);
return (error);
}
/*
* Replay a TX_WRITE ZIL transaction that didn't get committed
* after a system failure
*/
static int
zvol_replay_write(zvol_state_t *zv, lr_write_t *lr, boolean_t byteswap)
{
objset_t *os = zv->zv_objset;
char *data = (char *)(lr + 1); /* data follows lr_write_t */
uint64_t off = lr->lr_offset;
uint64_t len = lr->lr_length;
dmu_tx_t *tx;
int error;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, ZVOL_OBJ, off, len);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
} else {
dmu_write(os, ZVOL_OBJ, off, len, data, tx);
dmu_tx_commit(tx);
}
return (SET_ERROR(error));
}
/*
* Write out a history event.
*/
int
spa_history_log(spa_t *spa, const char *history_str, history_log_type_t what)
{
history_arg_t *ha;
int err = 0;
dmu_tx_t *tx;
ASSERT(what != LOG_INTERNAL);
tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err) {
dmu_tx_abort(tx);
return (err);
}
ha = kmem_alloc(sizeof (history_arg_t), KM_SLEEP);
ha->ha_history_str = strdup(history_str);
ha->ha_zone = strdup(spa_history_zone());
ha->ha_log_type = what;
ha->ha_uid = crgetuid(CRED());
/* Kick this off asynchronously; errors are ignored. */
dsl_sync_task_do_nowait(spa_get_dsl(spa), NULL,
spa_history_log_sync, spa, ha, 0, tx);
dmu_tx_commit(tx);
/* spa_history_log_sync will free ha and strings */
return (err);
}
/*
* Delete the entire contents of a directory. Return a count
* of the number of entries that could not be deleted. If we encounter
* an error, return a count of at least one so that the directory stays
* in the unlinked set.
*
* NOTE: this function assumes that the directory is inactive,
* so there is no need to lock its entries before deletion.
* Also, it assumes the directory contents is *only* regular
* files.
*/
static int
zfs_purgedir(znode_t *dzp)
{
zap_cursor_t zc;
zap_attribute_t zap;
znode_t *xzp;
dmu_tx_t *tx;
zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
zfs_dirlock_t dl;
int skipped = 0;
int error;
for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
(error = zap_cursor_retrieve(&zc, &zap)) == 0;
zap_cursor_advance(&zc)) {
error = zfs_zget(zfsvfs,
ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp);
if (error) {
skipped += 1;
continue;
}
/*
ASSERT((ZTOV(xzp)->v_type == VREG) ||
(ZTOV(xzp)->v_type == VLNK));
*/
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name);
dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
/* Is this really needed ? */
zfs_sa_upgrade_txholds(tx, xzp);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
//VN_RELE(ZTOV(xzp)); // async
VN_RELE_ASYNC(ZTOV(xzp), dsl_pool_vnrele_taskq(dmu_objset_pool(zfsvfs->z_os)));
skipped += 1;
continue;
}
bzero(&dl, sizeof (dl));
dl.dl_dzp = dzp;
dl.dl_name = zap.za_name;
error = zfs_link_destroy(&dl, xzp, tx, 0, NULL);
if (error)
skipped += 1;
dmu_tx_commit(tx);
//VN_RELE(ZTOV(xzp)); // async
VN_RELE_ASYNC(ZTOV(xzp), dsl_pool_vnrele_taskq(dmu_objset_pool(zfsvfs->z_os)));
}
zap_cursor_fini(&zc);
if (error != ENOENT)
skipped += 1;
return (skipped);
}
/*
* TX_WRITE2 are only generated when dmu_sync() returns EALREADY
* meaning the pool block is already being synced. So now that we always write
* out full blocks, all we have to do is expand the eof if
* the file is grown.
*/
static int
zfs_replay_write2(void *arg1, void *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = arg1;
lr_write_t *lr = arg2;
znode_t *zp;
int error;
uint64_t end;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
return (error);
top:
end = lr->lr_offset + lr->lr_length;
if (end > zp->z_size) {
dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
zp->z_size = end;
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
iput(ZTOI(zp));
if (error == ERESTART) {
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
dmu_tx_abort(tx);
return (error);
}
(void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
(void *)&zp->z_size, sizeof (uint64_t), tx);
/* Ensure the replayed seq is updated */
(void) zil_replaying(zfsvfs->z_log, tx);
dmu_tx_commit(tx);
}
iput(ZTOI(zp));
return (error);
}
int
zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
znode_t *xzp;
dmu_tx_t *tx;
int error;
zfs_fuid_info_t *fuidp = NULL;
*xvpp = NULL;
/*
* In FreeBSD, access checking for creating an EA is being done
* in zfs_setextattr(),
*/
#ifndef __FreeBSD__
if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))
return (error);
#endif
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_bonus(tx, zp->z_id);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
if (IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))) {
if (zfsvfs->z_fuid_obj == 0) {
dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
FUID_SIZE_ESTIMATE(zfsvfs));
dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
} else {
dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
FUID_SIZE_ESTIMATE(zfsvfs));
}
}
error = dmu_tx_assign(tx, zfsvfs->z_assign);
if (error) {
if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT)
dmu_tx_wait(tx);
dmu_tx_abort(tx);
return (error);
}
zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, 0, NULL, &fuidp);
ASSERT(xzp->z_phys->zp_parent == zp->z_id);
dmu_buf_will_dirty(zp->z_dbuf, tx);
zp->z_phys->zp_xattr = xzp->z_id;
(void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp,
xzp, "", NULL, fuidp, vap);
if (fuidp)
zfs_fuid_info_free(fuidp);
dmu_tx_commit(tx);
*xvpp = ZTOV(xzp);
return (0);
}
int
zfs_sa_set_xattr(znode_t *zp)
{
zfs_sb_t *zsb = ZTOZSB(zp);
dmu_tx_t *tx;
char *obj;
size_t size;
int error;
ASSERT(RW_WRITE_HELD(&zp->z_xattr_lock));
ASSERT(zp->z_xattr_cached);
ASSERT(zp->z_is_sa);
error = nvlist_size(zp->z_xattr_cached, &size, NV_ENCODE_XDR);
if (error)
goto out;
obj = zio_buf_alloc(size);
error = nvlist_pack(zp->z_xattr_cached, &obj, &size,
NV_ENCODE_XDR, KM_SLEEP);
if (error)
goto out_free;
tx = dmu_tx_create(zsb->z_os);
dmu_tx_hold_sa_create(tx, size);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
} else {
error = sa_update(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb),
obj, size, tx);
if (error)
dmu_tx_abort(tx);
else
dmu_tx_commit(tx);
}
out_free:
zio_buf_free(obj, size);
out:
return (error);
}
static int
zvol_write(struct bio *bio)
{
zvol_state_t *zv = bio->bi_bdev->bd_disk->private_data;
uint64_t offset = BIO_BI_SECTOR(bio) << 9;
uint64_t size = BIO_BI_SIZE(bio);
int error = 0;
dmu_tx_t *tx;
rl_t *rl;
uio_t uio;
if (bio->bi_rw & VDEV_REQ_FLUSH)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
/*
* Some requests are just for flush and nothing else.
*/
if (size == 0)
goto out;
uio.uio_bvec = &bio->bi_io_vec[BIO_BI_IDX(bio)];
uio.uio_skip = BIO_BI_SKIP(bio);
uio.uio_resid = size;
uio.uio_iovcnt = bio->bi_vcnt - BIO_BI_IDX(bio);
uio.uio_loffset = offset;
uio.uio_limit = MAXOFFSET_T;
uio.uio_segflg = UIO_BVEC;
rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_WRITER);
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_hold_write(tx, ZVOL_OBJ, offset, size);
/* This will only fail for ENOSPC */
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
zfs_range_unlock(rl);
goto out;
}
error = dmu_write_uio(zv->zv_objset, ZVOL_OBJ, &uio, size, tx);
if (error == 0)
zvol_log_write(zv, tx, offset, size,
!!(bio->bi_rw & VDEV_REQ_FUA));
dmu_tx_commit(tx);
zfs_range_unlock(rl);
if ((bio->bi_rw & VDEV_REQ_FUA) ||
zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
out:
return (error);
}
/*
* Delete the entire contents of a directory. Return a count
* of the number of entries that could not be deleted. If we encounter
* an error, return a count of at least one so that the directory stays
* in the unlinked set.
*
* NOTE: this function assumes that the directory is inactive,
* so there is no need to lock its entries before deletion.
* Also, it assumes the directory contents is *only* regular
* files.
*/
static int
zfs_purgedir(znode_t *dzp)
{
zap_cursor_t zc;
zap_attribute_t zap;
znode_t *xzp;
dmu_tx_t *tx;
zfsvfs_t *zfsvfs = ZTOZSB(dzp);
zfs_dirlock_t dl;
int skipped = 0;
int error;
for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
(error = zap_cursor_retrieve(&zc, &zap)) == 0;
zap_cursor_advance(&zc)) {
error = zfs_zget(zfsvfs,
ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp);
if (error) {
skipped += 1;
continue;
}
ASSERT(S_ISREG(ZTOI(xzp)->i_mode) ||
S_ISLNK(ZTOI(xzp)->i_mode));
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name);
dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
/* Is this really needed ? */
zfs_sa_upgrade_txholds(tx, xzp);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
zfs_iput_async(ZTOI(xzp));
skipped += 1;
continue;
}
bzero(&dl, sizeof (dl));
dl.dl_dzp = dzp;
dl.dl_name = zap.za_name;
error = zfs_link_destroy(&dl, xzp, tx, 0, NULL);
if (error)
skipped += 1;
dmu_tx_commit(tx);
zfs_iput_async(ZTOI(xzp));
}
zap_cursor_fini(&zc);
if (error != ENOENT)
skipped += 1;
return (skipped);
}
int
zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
znode_t *xzp;
dmu_tx_t *tx;
int error;
zfs_acl_ids_t acl_ids;
boolean_t fuid_dirtied;
*xvpp = NULL;
if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))
return (error);
if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL,
&acl_ids)) != 0)
return (error);
if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
zfs_acl_ids_free(&acl_ids);
return (EDQUOT);
}
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_bonus(tx, zp->z_id);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
fuid_dirtied = zfsvfs->z_fuid_dirty;
if (fuid_dirtied)
zfs_fuid_txhold(zfsvfs, tx);
error = dmu_tx_assign(tx, TXG_NOWAIT);
if (error) {
zfs_acl_ids_free(&acl_ids);
if (error == ERESTART)
dmu_tx_wait(tx);
dmu_tx_abort(tx);
return (error);
}
zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, 0, &acl_ids);
if (fuid_dirtied)
zfs_fuid_sync(zfsvfs, tx);
ASSERT(xzp->z_phys->zp_parent == zp->z_id);
dmu_buf_will_dirty(zp->z_dbuf, tx);
zp->z_phys->zp_xattr = xzp->z_id;
(void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp,
xzp, "", NULL, acl_ids.z_fuidp, vap);
zfs_acl_ids_free(&acl_ids);
dmu_tx_commit(tx);
*xvpp = ZTOV(xzp);
return (0);
}
int
zvol_set_volsize(zfs_cmd_t *zc)
{
zvol_state_t *zv;
dev_t dev = zc->zc_dev;
dmu_tx_t *tx;
int error;
dmu_object_info_t doi;
mutex_enter(&zvol_state_lock);
if ((zv = zvol_minor_lookup(zc->zc_name)) == NULL) {
mutex_exit(&zvol_state_lock);
return (ENXIO);
}
if ((error = dmu_object_info(zv->zv_objset, ZVOL_OBJ, &doi)) != 0 ||
(error = zvol_check_volsize(zc, doi.doi_data_block_size)) != 0) {
mutex_exit(&zvol_state_lock);
return (error);
}
if (zv->zv_readonly || (zv->zv_mode & DS_MODE_READONLY)) {
mutex_exit(&zvol_state_lock);
return (EROFS);
}
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
dmu_tx_hold_free(tx, ZVOL_OBJ, zc->zc_volsize, DMU_OBJECT_END);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
mutex_exit(&zvol_state_lock);
return (error);
}
error = zap_update(zv->zv_objset, ZVOL_ZAP_OBJ, "size", 8, 1,
&zc->zc_volsize, tx);
if (error == 0) {
error = dmu_free_range(zv->zv_objset, ZVOL_OBJ, zc->zc_volsize,
DMU_OBJECT_END, tx);
}
dmu_tx_commit(tx);
if (error == 0) {
zv->zv_volsize = zc->zc_volsize;
zvol_size_changed(zv, dev);
}
mutex_exit(&zvol_state_lock);
return (error);
}
/*
* Delete the entire contents of a directory. Return a count
* of the number of entries that could not be deleted. If we encounter
* an error, return a count of at least one so that the directory stays
* in the unlinked set.
*
* NOTE: this function assumes that the directory is inactive,
* so there is no need to lock its entries before deletion.
* Also, it assumes the directory contents is *only* regular
* files.
*/
static int
zfs_purgedir(znode_t *dzp)
{
zap_cursor_t zc;
zap_attribute_t zap;
znode_t *xzp;
dmu_tx_t *tx;
zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
zfs_dirlock_t dl;
int skipped = 0;
int error;
for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
(error = zap_cursor_retrieve(&zc, &zap)) == 0;
zap_cursor_advance(&zc)) {
error = zfs_zget(zfsvfs,
ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp);
if (error) {
skipped += 1;
continue;
}
ASSERT((ZTOV(xzp)->v_type == VREG) ||
(ZTOV(xzp)->v_type == VLNK));
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_bonus(tx, dzp->z_id);
dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name);
dmu_tx_hold_bonus(tx, xzp->z_id);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
VN_RELE(ZTOV(xzp));
skipped += 1;
continue;
}
bzero(&dl, sizeof (dl));
dl.dl_dzp = dzp;
dl.dl_name = zap.za_name;
error = zfs_link_destroy(&dl, xzp, tx, 0, NULL);
if (error)
skipped += 1;
dmu_tx_commit(tx);
VN_RELE(ZTOV(xzp));
}
zap_cursor_fini(&zc);
if (error != ENOENT)
skipped += 1;
return (skipped);
}
/*
* Common write path running under the zvol taskq context. This function
* is responsible for copying the request structure data in to the DMU and
* signaling the request queue with the result of the copy.
*/
static void
zvol_write(void *arg)
{
struct request *req = (struct request *)arg;
struct request_queue *q = req->q;
zvol_state_t *zv = q->queuedata;
fstrans_cookie_t cookie = spl_fstrans_mark();
uint64_t offset = blk_rq_pos(req) << 9;
uint64_t size = blk_rq_bytes(req);
int error = 0;
dmu_tx_t *tx;
rl_t *rl;
if (req->cmd_flags & VDEV_REQ_FLUSH)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
/*
* Some requests are just for flush and nothing else.
*/
if (size == 0) {
error = 0;
goto out;
}
rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_WRITER);
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_hold_write(tx, ZVOL_OBJ, offset, size);
/* This will only fail for ENOSPC */
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
zfs_range_unlock(rl);
goto out;
}
error = dmu_write_req(zv->zv_objset, ZVOL_OBJ, req, tx);
if (error == 0)
zvol_log_write(zv, tx, offset, size,
req->cmd_flags & VDEV_REQ_FUA);
dmu_tx_commit(tx);
zfs_range_unlock(rl);
if ((req->cmd_flags & VDEV_REQ_FUA) ||
zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
out:
blk_end_request(req, -error, size);
spl_fstrans_unmark(cookie);
}
int
zfs_set_version(zfs_sb_t *zsb, uint64_t newvers)
{
int error;
objset_t *os = zsb->z_os;
dmu_tx_t *tx;
if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION)
return (SET_ERROR(EINVAL));
if (newvers < zsb->z_version)
return (SET_ERROR(EINVAL));
if (zfs_spa_version_map(newvers) >
spa_version(dmu_objset_spa(zsb->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 && !zsb->z_use_sa) {
dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
ZFS_SA_ATTRS);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
}
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
return (error);
}
error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
8, 1, &newvers, tx);
if (error) {
dmu_tx_commit(tx);
return (error);
}
if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) {
uint64_t sa_obj;
ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=,
SPA_VERSION_SA);
sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
DMU_OT_NONE, 0, tx);
error = zap_add(os, MASTER_NODE_OBJ,
ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
ASSERT0(error);
VERIFY(0 == sa_set_sa_object(os, sa_obj));
sa_register_update_callback(os, zfs_sa_upgrade);
}
/*
* Delete the entire contents of a directory. Return a count
* of the number of entries that could not be deleted. If we encounter
* an error, return a count of at least one so that the directory stays
* in the unlinked set.
*
* NOTE: this function assumes that the directory is inactive,
* so there is no need to lock its entries before deletion.
* Also, it assumes the directory contents is *only* regular
* files.
*/
static int
zfs_purgedir(znode_t *dzp)
{
zap_cursor_t zc;
zap_attribute_t zap;
znode_t *xzp;
dmu_tx_t *tx;
zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
int skipped = 0;
int error;
for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
(error = zap_cursor_retrieve(&zc, &zap)) == 0;
zap_cursor_advance(&zc)) {
error = zfs_zget(zfsvfs,
ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp);
if (error) {
skipped += 1;
continue;
}
vn_lock(ZTOV(xzp), LK_EXCLUSIVE | LK_RETRY);
ASSERT((ZTOV(xzp)->v_type == VREG) ||
(ZTOV(xzp)->v_type == VLNK));
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name);
dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
/* Is this really needed ? */
zfs_sa_upgrade_txholds(tx, xzp);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
vput(ZTOV(xzp));
skipped += 1;
continue;
}
error = zfs_link_destroy(dzp, zap.za_name, xzp, tx, 0, NULL);
if (error)
skipped += 1;
dmu_tx_commit(tx);
vput(ZTOV(xzp));
}
zap_cursor_fini(&zc);
if (error != ENOENT)
skipped += 1;
return (skipped);
}
static uint64_t
zpios_dmu_object_create(run_args_t *run_args, objset_t *os)
{
struct dmu_tx *tx;
uint64_t obj = 0ULL;
uint64_t blksize = run_args->block_size;
int rc;
if (blksize < SPA_MINBLOCKSIZE ||
blksize > spa_maxblocksize(dmu_objset_spa(os)) ||
!ISP2(blksize)) {
zpios_print(run_args->file,
"invalid block size for pool: %d\n", (int)blksize);
return (obj);
}
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, OBJ_SIZE);
rc = dmu_tx_assign(tx, TXG_WAIT);
if (rc) {
zpios_print(run_args->file,
"dmu_tx_assign() failed: %d\n", rc);
dmu_tx_abort(tx);
return (obj);
}
obj = dmu_object_alloc(os, DMU_OT_UINT64_OTHER, 0, DMU_OT_NONE, 0, tx);
rc = dmu_object_set_blocksize(os, obj, blksize, 0, tx);
if (rc) {
zpios_print(run_args->file,
"dmu_object_set_blocksize to %d failed: %d\n",
(int)blksize, rc);
dmu_tx_abort(tx);
return (obj);
}
dmu_tx_commit(tx);
return (obj);
}
static int
zvol_discard(struct bio *bio)
{
zvol_state_t *zv = bio->bi_bdev->bd_disk->private_data;
uint64_t start = BIO_BI_SECTOR(bio) << 9;
uint64_t size = BIO_BI_SIZE(bio);
uint64_t end = start + size;
int error;
rl_t *rl;
dmu_tx_t *tx;
ASSERT(zv && zv->zv_open_count > 0);
if (end > zv->zv_volsize)
return (SET_ERROR(EIO));
/*
* Align the request to volume block boundaries when REQ_SECURE is
* available, but not requested. If we don't, then this will force
* dnode_free_range() to zero out the unaligned parts, which is slow
* (read-modify-write) and useless since we are not freeing any space
* by doing so. Kernels that do not support REQ_SECURE (2.6.32 through
* 2.6.35) will not receive this optimization.
*/
#ifdef REQ_SECURE
if (!(bio->bi_rw & REQ_SECURE)) {
start = P2ROUNDUP(start, zv->zv_volblocksize);
end = P2ALIGN(end, zv->zv_volblocksize);
size = end - start;
}
#endif
if (start >= end)
return (0);
rl = zfs_range_lock(&zv->zv_znode, start, size, RL_WRITER);
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error != 0) {
dmu_tx_abort(tx);
} else {
zvol_log_truncate(zv, tx, start, size, B_TRUE);
dmu_tx_commit(tx);
error = dmu_free_long_range(zv->zv_objset,
ZVOL_OBJ, start, size);
}
zfs_range_unlock(rl);
return (error);
}
static int
zpios_dmu_write(run_args_t *run_args, objset_t *os, uint64_t object,
uint64_t offset, uint64_t size, const void *buf)
{
struct dmu_tx *tx;
int rc, how = TXG_WAIT;
// int flags = 0;
if (run_args->flags & DMU_WRITE_NOWAIT)
how = TXG_NOWAIT;
while (1) {
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, object, offset, size);
rc = dmu_tx_assign(tx, how);
if (rc) {
if (rc == ERESTART && how == TXG_NOWAIT) {
dmu_tx_wait(tx);
dmu_tx_abort(tx);
continue;
}
zpios_print(run_args->file,
"Error in dmu_tx_assign(), %d", rc);
dmu_tx_abort(tx);
return (rc);
}
break;
}
// if (run_args->flags & DMU_WRITE_ZC)
// flags |= DMU_WRITE_ZEROCOPY;
dmu_write(os, object, offset, size, buf, tx);
dmu_tx_commit(tx);
return (0);
}
/*
* Ensure the zap is flushed then inform the VFS of the capacity change.
*/
static int
zvol_update_volsize(zvol_state_t *zv, uint64_t volsize, objset_t *os)
{
struct block_device *bdev;
dmu_tx_t *tx;
int error;
ASSERT(MUTEX_HELD(&zvol_state_lock));
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
return (error);
}
error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1,
&volsize, tx);
dmu_tx_commit(tx);
if (error)
return (error);
error = dmu_free_long_range(os,
ZVOL_OBJ, volsize, DMU_OBJECT_END);
if (error)
return (error);
bdev = bdget_disk(zv->zv_disk, 0);
if (!bdev)
return (EIO);
/*
* 2.6.28 API change
* Added check_disk_size_change() helper function.
*/
#ifdef HAVE_CHECK_DISK_SIZE_CHANGE
set_capacity(zv->zv_disk, volsize >> 9);
zv->zv_volsize = volsize;
check_disk_size_change(zv->zv_disk, bdev);
#else
zv->zv_volsize = volsize;
zv->zv_changed = 1;
(void) check_disk_change(bdev);
#endif /* HAVE_CHECK_DISK_SIZE_CHANGE */
bdput(bdev);
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
zvol_discard(struct bio *bio)
{
zvol_state_t *zv = bio->bi_bdev->bd_disk->private_data;
uint64_t start = BIO_BI_SECTOR(bio) << 9;
uint64_t size = BIO_BI_SIZE(bio);
uint64_t end = start + size;
int error;
rl_t *rl;
dmu_tx_t *tx;
ASSERT(zv && zv->zv_open_count > 0);
if (end > zv->zv_volsize)
return (SET_ERROR(EIO));
/*
* Align the request to volume block boundaries when a secure erase is
* not required. This will prevent dnode_free_range() from zeroing out
* the unaligned parts which is slow (read-modify-write) and useless
* since we are not freeing any space by doing so.
*/
if (!bio_is_secure_erase(bio)) {
start = P2ROUNDUP(start, zv->zv_volblocksize);
end = P2ALIGN(end, zv->zv_volblocksize);
size = end - start;
}
if (start >= end)
return (0);
rl = zfs_range_lock(&zv->zv_range_lock, start, size, RL_WRITER);
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error != 0) {
dmu_tx_abort(tx);
} else {
zvol_log_truncate(zv, tx, start, size, B_TRUE);
dmu_tx_commit(tx);
error = dmu_free_long_range(zv->zv_objset,
ZVOL_OBJ, start, size);
}
zfs_range_unlock(rl);
return (error);
}
int
dmu_objset_userspace_upgrade(objset_t *os)
{
uint64_t obj;
int err = 0;
if (dmu_objset_userspace_present(os))
return (0);
if (!dmu_objset_userused_enabled(os->os))
return (ENOTSUP);
if (dmu_objset_is_snapshot(os))
return (EINVAL);
/*
* We simply need to mark every object dirty, so that it will be
* synced out and now accounted. If this is called
* concurrently, or if we already did some work before crashing,
* that's fine, since we track each object's accounted state
* independently.
*/
for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) {
dmu_tx_t *tx;
dmu_buf_t *db;
int objerr;
if (issig(JUSTLOOKING) && issig(FORREAL))
return (EINTR);
objerr = dmu_bonus_hold(os, obj, FTAG, &db);
if (objerr)
continue;
tx = dmu_tx_create(os);
dmu_tx_hold_bonus(tx, obj);
objerr = dmu_tx_assign(tx, TXG_WAIT);
if (objerr) {
dmu_tx_abort(tx);
continue;
}
dmu_buf_will_dirty(db, tx);
dmu_buf_rele(db, FTAG);
dmu_tx_commit(tx);
}
os->os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE;
txg_wait_synced(dmu_objset_pool(os), 0);
return (0);
}
int
zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
znode_t *xzp;
dmu_tx_t *tx;
int error;
zfs_fuid_info_t *fuidp = NULL;
*xvpp = NULL;
if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))
return (error);
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_bonus(tx, zp->z_id);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
if (zfsvfs->z_fuid_obj == 0) {
dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE);
dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
} else {
dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0, SPA_MAXBLOCKSIZE);
}
error = dmu_tx_assign(tx, zfsvfs->z_assign);
if (error) {
if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT)
dmu_tx_wait(tx);
dmu_tx_abort(tx);
return (error);
}
zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, 0, NULL, &fuidp);
ASSERT(xzp->z_phys->zp_parent == zp->z_id);
dmu_buf_will_dirty(zp->z_dbuf, tx);
zp->z_phys->zp_xattr = xzp->z_id;
(void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp,
xzp, "", NULL, fuidp, vap);
if (fuidp)
zfs_fuid_info_free(fuidp);
dmu_tx_commit(tx);
*xvpp = ZTOV(xzp);
return (0);
}
