static void
dsl_dataset_remove_clones_key(dsl_dataset_t *ds, uint64_t mintxg, dmu_tx_t *tx)
{
objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
zap_cursor_t zc;
zap_attribute_t za;
/*
* If it is the old version, dd_clones doesn't exist so we can't
* find the clones, but dsl_deadlist_remove_key() is a no-op so it
* doesn't matter.
*/
if (dsl_dir_phys(ds->ds_dir)->dd_clones == 0)
return;
for (zap_cursor_init(&zc, mos, dsl_dir_phys(ds->ds_dir)->dd_clones);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
dsl_dataset_t *clone;
VERIFY0(dsl_dataset_hold_obj(ds->ds_dir->dd_pool,
za.za_first_integer, FTAG, &clone));
if (clone->ds_dir->dd_origin_txg > mintxg) {
dsl_deadlist_remove_key(&clone->ds_deadlist,
mintxg, tx);
dsl_dataset_remove_clones_key(clone, mintxg, tx);
}
dsl_dataset_rele(clone, FTAG);
}
zap_cursor_fini(&zc);
}
/*
* Clean up any znodes that had no links when we either crashed or
* (force) umounted the file system.
*/
void
zfs_unlinked_drain(zfs_sb_t *zsb)
{
zap_cursor_t zc;
zap_attribute_t zap;
dmu_object_info_t doi;
znode_t *zp;
int error;
/*
* Interate over the contents of the unlinked set.
*/
for (zap_cursor_init(&zc, zsb->z_os, zsb->z_unlinkedobj);
zap_cursor_retrieve(&zc, &zap) == 0;
zap_cursor_advance(&zc)) {
/*
* See what kind of object we have in list
*/
error = dmu_object_info(zsb->z_os, zap.za_first_integer, &doi);
if (error != 0)
continue;
ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) ||
(doi.doi_type == DMU_OT_DIRECTORY_CONTENTS));
/*
* We need to re-mark these list entries for deletion,
* so we pull them back into core and set zp->z_unlinked.
*/
error = zfs_zget(zsb, zap.za_first_integer, &zp);
/*
* We may pick up znodes that are already marked for deletion.
* This could happen during the purge of an extended attribute
* directory. All we need to do is skip over them, since they
* are already in the system marked z_unlinked.
*/
if (error != 0)
continue;
zp->z_unlinked = B_TRUE;
/*
* If this is an attribute directory, purge its contents.
*/
if (S_ISDIR(ZTOI(zp)->i_mode) && (zp->z_pflags & ZFS_XATTR)) {
/*
* We don't need to check the return value of
* zfs_purgedir here, because zfs_rmnode will just
* return this xattr directory to the unlinked set
* until all of its xattrs are gone.
*/
(void) zfs_purgedir(zp);
}
iput(ZTOI(zp));
}
zap_cursor_fini(&zc);
}
/*
* Read as many directory entry names as will fit in to the provided buffer,
* or when no buffer is provided calculate the required buffer size.
*/
int
zpl_xattr_readdir(struct inode *dxip, xattr_filldir_t *xf)
{
zap_cursor_t zc;
zap_attribute_t zap;
int error;
zap_cursor_init(&zc, ITOZSB(dxip)->z_os, ITOZ(dxip)->z_id);
while ((error = -zap_cursor_retrieve(&zc, &zap)) == 0) {
if (zap.za_integer_length != 8 || zap.za_num_integers != 1) {
error = -ENXIO;
break;
}
error = zpl_xattr_filldir(xf, zap.za_name, strlen(zap.za_name));
if (error)
break;
zap_cursor_advance(&zc);
}
zap_cursor_fini(&zc);
if (error == -ENOENT)
error = 0;
return (error);
}
static void
dump_obj(objset_t *os, uint64_t obj, const char *name)
{
zap_cursor_t zc;
zap_attribute_t za;
(void) printf("%s_obj:\n", name);
for (zap_cursor_init(&zc, os, obj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
if (za.za_integer_length == 8) {
ASSERT(za.za_num_integers == 1);
(void) printf("\t%s = %llu\n",
za.za_name, (u_longlong_t)za.za_first_integer);
} else {
ASSERT(za.za_integer_length == 1);
char val[1024];
VERIFY(zap_lookup(os, obj, za.za_name,
1, sizeof (val), val) == 0);
(void) printf("\t%s = %s\n", za.za_name, val);
}
}
zap_cursor_fini(&zc);
}
/*
* Iterate over the sets specified in the specified zapobj
* and load them into the permsets avl tree.
*/
static int
dsl_load_sets(objset_t *mos, uint64_t zapobj,
char type, char checkflag, void *valp, avl_tree_t *avl)
{
zap_cursor_t zc;
zap_attribute_t za;
perm_set_t *permnode;
avl_index_t idx;
uint64_t jumpobj;
int error;
char whokey[ZFS_MAX_DELEG_NAME];
zfs_deleg_whokey(whokey, type, checkflag, valp);
error = zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj);
if (error != 0)
return (error);
for (zap_cursor_init(&zc, mos, jumpobj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
permnode = kmem_alloc(sizeof (perm_set_t), KM_SLEEP);
(void) strlcpy(permnode->p_setname, za.za_name,
sizeof (permnode->p_setname));
permnode->p_matched = B_FALSE;
if (avl_find(avl, permnode, &idx) == NULL) {
avl_insert(avl, permnode, idx);
} else {
kmem_free(permnode, sizeof (perm_set_t));
}
}
zap_cursor_fini(&zc);
return (0);
}
int
dsl_dataset_get_holds(const char *dsname, nvlist_t *nvl)
{
dsl_pool_t *dp;
dsl_dataset_t *ds;
int err;
err = dsl_pool_hold(dsname, FTAG, &dp);
if (err != 0)
return (err);
err = dsl_dataset_hold(dp, dsname, FTAG, &ds);
if (err != 0) {
dsl_pool_rele(dp, FTAG);
return (err);
}
if (ds->ds_phys->ds_userrefs_obj != 0) {
zap_attribute_t *za;
zap_cursor_t zc;
za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
for (zap_cursor_init(&zc, ds->ds_dir->dd_pool->dp_meta_objset,
ds->ds_phys->ds_userrefs_obj);
zap_cursor_retrieve(&zc, za) == 0;
zap_cursor_advance(&zc)) {
fnvlist_add_uint64(nvl, za->za_name,
za->za_first_integer);
}
zap_cursor_fini(&zc);
kmem_free(za, sizeof (zap_attribute_t));
}
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (0);
}
/*
* 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);
}
/*
* Special function to requeue the znodes for deletion that were
* in progress when we either crashed or umounted the file system.
*
* returns 1 if queue was drained.
*/
static int
zfs_drain_dq(zfsvfs_t *zfsvfs)
{
zap_cursor_t zc;
zap_attribute_t zap;
dmu_object_info_t doi;
znode_t *zp;
int error;
/*
* Interate over the contents of the delete queue.
*/
for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_dqueue);
zap_cursor_retrieve(&zc, &zap) == 0;
zap_cursor_advance(&zc)) {
/*
* Create more threads if necessary to balance the load.
* quit if the delete threads have been shut down.
*/
if (zfs_delete_thread_target(zfsvfs, -1) != 0)
return (0);
/*
* See what kind of object we have in queue
*/
error = dmu_object_info(zfsvfs->z_os,
zap.za_first_integer, &doi);
if (error != 0)
continue;
ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) ||
(doi.doi_type == DMU_OT_DIRECTORY_CONTENTS));
/*
* We need to re-mark these queue entries for reaping,
* so we pull them back into core and set zp->z_reap.
*/
error = zfs_zget(zfsvfs, zap.za_first_integer, &zp);
/*
* We may pick up znodes that are already marked for reaping.
* This could happen during the purge of an extended attribute
* directory. All we need to do is skip over them, since they
* are already in the system to be processed by the delete
* thread(s).
*/
if (error != 0) {
continue;
}
zp->z_reap = 1;
VN_RELE(ZTOV(zp));
}
zap_cursor_fini(&zc);
return (1);
}
/*
* 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);
}
/*
* 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);
}
/*
* 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);
}
/*
* Checks that the active features in the pool are supported by
* this software. Adds each unsupported feature (name -> description) to
* the supplied nvlist.
*/
boolean_t
spa_features_check(spa_t *spa, boolean_t for_write,
nvlist_t *unsup_feat, nvlist_t *enabled_feat)
{
objset_t *os = spa->spa_meta_objset;
boolean_t supported;
zap_cursor_t *zc;
zap_attribute_t *za;
uint64_t obj = for_write ?
spa->spa_feat_for_write_obj : spa->spa_feat_for_read_obj;
char *buf;
zc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP);
za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);
supported = B_TRUE;
for (zap_cursor_init(zc, os, obj);
zap_cursor_retrieve(zc, za) == 0;
zap_cursor_advance(zc)) {
ASSERT(za->za_integer_length == sizeof (uint64_t) &&
za->za_num_integers == 1);
if (NULL != enabled_feat) {
fnvlist_add_uint64(enabled_feat, za->za_name,
za->za_first_integer);
}
if (za->za_first_integer != 0 &&
!zfeature_is_supported(za->za_name)) {
supported = B_FALSE;
if (NULL != unsup_feat) {
char *desc = "";
if (zap_lookup(os, spa->spa_feat_desc_obj,
za->za_name, 1, MAXPATHLEN, buf) == 0)
desc = buf;
VERIFY(nvlist_add_string(unsup_feat,
za->za_name, desc) == 0);
}
}
}
zap_cursor_fini(zc);
kmem_free(buf, MAXPATHLEN);
kmem_free(za, sizeof (zap_attribute_t));
kmem_free(zc, sizeof (zap_cursor_t));
return (supported);
}
int
dsl_get_bookmarks_impl(dsl_dataset_t *ds, nvlist_t *props, nvlist_t *outnvl)
{
int err = 0;
zap_cursor_t zc;
zap_attribute_t attr;
dsl_pool_t *dp = ds->ds_dir->dd_pool;
uint64_t bmark_zapobj = ds->ds_bookmarks;
if (bmark_zapobj == 0)
return (0);
for (zap_cursor_init(&zc, dp->dp_meta_objset, bmark_zapobj);
zap_cursor_retrieve(&zc, &attr) == 0;
zap_cursor_advance(&zc)) {
nvlist_t *out_props;
char *bmark_name = attr.za_name;
zfs_bookmark_phys_t bmark_phys = { 0 };
err = dsl_dataset_bmark_lookup(ds, bmark_name, &bmark_phys);
ASSERT3U(err, !=, ENOENT);
if (err != 0)
break;
out_props = fnvlist_alloc();
if (nvlist_exists(props,
zfs_prop_to_name(ZFS_PROP_GUID))) {
dsl_prop_nvlist_add_uint64(out_props,
ZFS_PROP_GUID, bmark_phys.zbm_guid);
}
if (nvlist_exists(props,
zfs_prop_to_name(ZFS_PROP_CREATETXG))) {
dsl_prop_nvlist_add_uint64(out_props,
ZFS_PROP_CREATETXG, bmark_phys.zbm_creation_txg);
}
if (nvlist_exists(props,
zfs_prop_to_name(ZFS_PROP_CREATION))) {
dsl_prop_nvlist_add_uint64(out_props,
ZFS_PROP_CREATION, bmark_phys.zbm_creation_time);
}
if (nvlist_exists(props,
zfs_prop_to_name(ZFS_PROP_IVSET_GUID))) {
dsl_prop_nvlist_add_uint64(out_props,
ZFS_PROP_IVSET_GUID, bmark_phys.zbm_ivset_guid);
}
fnvlist_add_nvlist(outnvl, bmark_name, out_props);
fnvlist_free(out_props);
}
zap_cursor_fini(&zc);
return (err);
}
/*
* Clean up any znodes that had no links when we either crashed or
* (force) umounted the file system.
*/
void
zfs_unlinked_drain(zfsvfs_t *zfsvfs)
{
zap_cursor_t zc;
zap_attribute_t zap;
dmu_object_info_t doi;
znode_t *zp;
int error;
printf("ZFS: unlinked drain\n");
/*
* Interate over the contents of the unlinked set.
*/
for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj);
zap_cursor_retrieve(&zc, &zap) == 0;
zap_cursor_advance(&zc)) {
/*
* See what kind of object we have in list
*/
error = dmu_object_info(zfsvfs->z_os,
zap.za_first_integer, &doi);
if (error != 0)
continue;
ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) ||
(doi.doi_type == DMU_OT_DIRECTORY_CONTENTS));
/*
* We need to re-mark these list entries for deletion,
* so we pull them back into core and set zp->z_unlinked.
*/
error = zfs_zget(zfsvfs, zap.za_first_integer, &zp);
/*
* We may pick up znodes that are already marked for deletion.
* This could happen during the purge of an extended attribute
* directory. All we need to do is skip over them, since they
* are already in the system marked z_unlinked.
*/
if (error != 0)
continue;
zp->z_unlinked = B_TRUE;
VN_RELE(ZTOV(zp));
}
zap_cursor_fini(&zc);
printf("ZFS: unlinked drain completed.\n");
}
int
dsl_deleg_destroy(objset_t *mos, uint64_t zapobj, dmu_tx_t *tx)
{
zap_cursor_t zc;
zap_attribute_t za;
if (zapobj == 0)
return (0);
for (zap_cursor_init(&zc, mos, zapobj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
ASSERT(za.za_integer_length == 8 && za.za_num_integers == 1);
VERIFY(0 == zap_destroy(mos, za.za_first_integer, tx));
}
zap_cursor_fini(&zc);
VERIFY(0 == zap_destroy(mos, zapobj, tx));
return (0);
}
static void
copy_create_perms(dsl_dir_t *dd, uint64_t pzapobj,
boolean_t dosets, uint64_t uid, dmu_tx_t *tx)
{
objset_t *mos = dd->dd_pool->dp_meta_objset;
uint64_t jumpobj, pjumpobj;
uint64_t zapobj = dd->dd_phys->dd_deleg_zapobj;
zap_cursor_t zc;
zap_attribute_t za;
char whokey[ZFS_MAX_DELEG_NAME];
zfs_deleg_whokey(whokey,
dosets ? ZFS_DELEG_CREATE_SETS : ZFS_DELEG_CREATE,
ZFS_DELEG_LOCAL, NULL);
if (zap_lookup(mos, pzapobj, whokey, 8, 1, &pjumpobj) != 0)
return;
if (zapobj == 0) {
dmu_buf_will_dirty(dd->dd_dbuf, tx);
zapobj = dd->dd_phys->dd_deleg_zapobj = zap_create(mos,
DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx);
}
zfs_deleg_whokey(whokey,
dosets ? ZFS_DELEG_USER_SETS : ZFS_DELEG_USER,
ZFS_DELEG_LOCAL, &uid);
if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) == ENOENT) {
jumpobj = zap_create(mos, DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx);
VERIFY(zap_add(mos, zapobj, whokey, 8, 1, &jumpobj, tx) == 0);
}
for (zap_cursor_init(&zc, mos, pjumpobj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
uint64_t zero = 0;
ASSERT(za.za_integer_length == 8 && za.za_num_integers == 1);
VERIFY(zap_update(mos, jumpobj, za.za_name,
8, 1, &zero, tx) == 0);
}
zap_cursor_fini(&zc);
}
/*
* Checks that the features active in the specified object are supported by
* this software. Adds each unsupported feature (name -> description) to
* the supplied nvlist.
*/
boolean_t
feature_is_supported(objset_t *os, uint64_t obj, uint64_t desc_obj,
nvlist_t *unsup_feat, nvlist_t *enabled_feat)
{
boolean_t supported;
zap_cursor_t zc;
zap_attribute_t za;
supported = B_TRUE;
for (zap_cursor_init(&zc, os, obj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
ASSERT(za.za_integer_length == sizeof (uint64_t) &&
za.za_num_integers == 1);
if (NULL != enabled_feat) {
fnvlist_add_uint64(enabled_feat, za.za_name,
za.za_first_integer);
}
if (za.za_first_integer != 0 &&
!zfeature_is_supported(za.za_name)) {
supported = B_FALSE;
if (NULL != unsup_feat) {
char *desc = "";
char buf[MAXPATHLEN];
if (zap_lookup(os, desc_obj, za.za_name,
1, sizeof (buf), buf) == 0)
desc = buf;
VERIFY(nvlist_add_string(unsup_feat, za.za_name,
desc) == 0);
}
}
}
zap_cursor_fini(&zc);
return (supported);
}
static int osd_index_it_get(const struct lu_env *env, struct dt_it *di,
const struct dt_key *key)
{
struct osd_zap_it *it = (struct osd_zap_it *)di;
struct osd_object *obj = it->ozi_obj;
struct osd_device *osd = osd_obj2dev(obj);
ENTRY;
LASSERT(it);
LASSERT(it->ozi_zc);
/* XXX: API is broken at the moment */
LASSERT(*((const __u64 *)key) == 0);
zap_cursor_fini(it->ozi_zc);
memset(it->ozi_zc, 0, sizeof(*it->ozi_zc));
zap_cursor_init(it->ozi_zc, osd->od_objset.os, obj->oo_db->db_object);
it->ozi_reset = 1;
RETURN(+1);
}
static int osd_index_it_get(const struct lu_env *env, struct dt_it *di,
const struct dt_key *key)
{
struct osd_zap_it *it = (struct osd_zap_it *)di;
struct osd_object *obj = it->ozi_obj;
struct osd_device *osd = osd_obj2dev(obj);
ENTRY;
LASSERT(it);
LASSERT(it->ozi_zc);
/*
* XXX: we need a binary version of zap_cursor_move_to_key()
* to implement this API */
if (*((const __u64 *)key) != 0)
CERROR("NOT IMPLEMETED YET (move to %Lx)\n", *((__u64 *)key));
zap_cursor_fini(it->ozi_zc);
memset(it->ozi_zc, 0, sizeof(*it->ozi_zc));
zap_cursor_init(it->ozi_zc, osd->od_objset.os, obj->oo_db->db_object);
it->ozi_reset = 1;
RETURN(+1);
}
uint64_t
dmu_snapname_to_id(objset_t *os, const char *snapname)
{
dsl_dataset_t *ds = os->os->os_dsl_dataset;
zap_cursor_t cursor;
zap_attribute_t attr;
if (ds->ds_phys->ds_snapnames_zapobj == 0) {
return (ENOENT);
}
zap_cursor_init(&cursor,
ds->ds_dir->dd_pool->dp_meta_objset,
ds->ds_phys->ds_snapnames_zapobj);
if(!(zap_cursor_move_to_key(&cursor, snapname, MT_EXACT))) {
if (zap_cursor_retrieve(&cursor, &attr) != 0) {
zap_cursor_fini(&cursor);
/* not returning ENOENT as that might be the id no */
return 0; // no snapshot by name snapname
}
return attr.za_first_integer;
} else {
return 0; // no snapshot by name snapname
}
}
/*
* Find all 'allow' permissions from a given point and then continue
* traversing up to the root.
*
* This function constructs an nvlist of nvlists.
* each setpoint is an nvlist composed of an nvlist of an nvlist
* of the individual * users/groups/everyone/create
* permissions.
*
* The nvlist will look like this.
*
* { source fsname -> { whokeys { permissions,...}, ...}}
*
* The fsname nvpairs will be arranged in a bottom up order. For example,
* if we have the following structure a/b/c then the nvpairs for the fsnames
* will be ordered a/b/c, a/b, a.
*/
int
dsl_deleg_get(const char *ddname, nvlist_t **nvp)
{
dsl_dir_t *dd, *startdd;
dsl_pool_t *dp;
int error;
objset_t *mos;
zap_cursor_t *basezc, *zc;
zap_attribute_t *baseza, *za;
char *source;
error = dsl_pool_hold(ddname, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dir_hold(dp, ddname, FTAG, &startdd, NULL);
if (error != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
dp = startdd->dd_pool;
mos = dp->dp_meta_objset;
zc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP);
za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
basezc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP);
baseza = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
source = kmem_alloc(MAXNAMELEN + strlen(MOS_DIR_NAME) + 1, KM_SLEEP);
VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
for (dd = startdd; dd != NULL; dd = dd->dd_parent) {
nvlist_t *sp_nvp;
uint64_t n;
if (dd->dd_phys->dd_deleg_zapobj == 0 ||
zap_count(mos, dd->dd_phys->dd_deleg_zapobj, &n) != 0 ||
n == 0)
continue;
sp_nvp = fnvlist_alloc();
for (zap_cursor_init(basezc, mos,
dd->dd_phys->dd_deleg_zapobj);
zap_cursor_retrieve(basezc, baseza) == 0;
zap_cursor_advance(basezc)) {
nvlist_t *perms_nvp;
ASSERT(baseza->za_integer_length == 8);
ASSERT(baseza->za_num_integers == 1);
perms_nvp = fnvlist_alloc();
for (zap_cursor_init(zc, mos, baseza->za_first_integer);
zap_cursor_retrieve(zc, za) == 0;
zap_cursor_advance(zc)) {
fnvlist_add_boolean(perms_nvp, za->za_name);
}
zap_cursor_fini(zc);
fnvlist_add_nvlist(sp_nvp, baseza->za_name, perms_nvp);
fnvlist_free(perms_nvp);
}
zap_cursor_fini(basezc);
dsl_dir_name(dd, source);
fnvlist_add_nvlist(*nvp, source, sp_nvp);
nvlist_free(sp_nvp);
}
kmem_free(source, MAXNAMELEN + strlen(MOS_DIR_NAME) + 1);
kmem_free(baseza, sizeof (zap_attribute_t));
kmem_free(basezc, sizeof (zap_cursor_t));
kmem_free(za, sizeof (zap_attribute_t));
kmem_free(zc, sizeof (zap_cursor_t));
dsl_dir_rele(startdd, FTAG);
dsl_pool_rele(dp, FTAG);
return (0);
}
int
dsl_keychain_load_dd(dsl_dir_t *dd, uint64_t dsobj,
int crypt, zcrypt_key_t *wrappingkey)
{
zap_cursor_t zc;
zap_attribute_t za;
objset_t *mos = dd->dd_pool->dp_meta_objset;
uint64_t keychain_zapobj = dsl_dir_phys(dd)->dd_keychain_obj;
zcrypt_key_t *txgkey;
zcrypt_keystore_node_t *skn;
caddr_t wrappedkey;
size_t wkeylen;
spa_t *spa = dd->dd_pool->dp_spa;
int unwrapped = 0, entries = 0;
/*
* Basic algorithm is start with the ds_keychain_obj
* and iterate using zap_cursor_*() unwraping the
* values (the actual encryption keys) into zcrypt_key_t's
* and calling zcrypt_keychain_insert() to put them into the dsl AVL
* tree of keys.
*/
zcrypt_key_hold(wrappingkey, FTAG);
skn = zcrypt_keystore_insert(spa, dsobj, wrappingkey);
ASSERT(skn != NULL);
mutex_enter(&skn->skn_lock);
for (zap_cursor_init(&zc, mos, keychain_zapobj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
entries++;
wkeylen = za.za_num_integers;
wrappedkey = kmem_alloc(wkeylen, KM_SLEEP);
VERIFY3U(zap_lookup_uint64(mos, keychain_zapobj,
(uint64_t *)&za.za_name, 1, 1,
za.za_num_integers, wrappedkey), ==, 0);
if (zcrypt_unwrap_key(wrappingkey, crypt,
wrappedkey, wkeylen, &txgkey) != 0) {
kmem_free(wrappedkey, wkeylen);
continue;
}
unwrapped++;
kmem_free(wrappedkey, wkeylen);
zcrypt_keychain_insert(&skn->skn_keychain,
*(uint64_t *)za.za_name, txgkey);
}
zap_cursor_fini(&zc);
mutex_exit(&skn->skn_lock);
zcrypt_key_release(wrappingkey, FTAG);
if (entries > 0 && unwrapped == 0) {
/* Wrong wrapping key passed */
(void) zcrypt_keystore_remove(spa, dsobj);
return (EACCES);
}
/*
* If we didn't unwrap everything we have possible corruption.
* If an attempt is ever made to decrypt blocks either we won't
* find the key (ENOKEY) or we will use the wrong key which
* will result in the MAC failing to verify so ECKSUM will be
* set in zio->io_error which will result in an ereport being
* logged because the zio_read() failed.
* When we are running DEBUG lets ASSERT this instead.
*/
ASSERT3U(entries, ==, unwrapped);
return (0);
}
/*
* dsl_crypto_key_change
*
* The old key must already be present in memory since the user interface
* doesn't provide away to prompt or retrieve the old key.
*/
static void
dsl_crypto_key_change_sync(void *arg1, dmu_tx_t *tx)
{
struct wkey_change_arg *ca = arg1;
dsl_dataset_t *ds = ca->ca_ds;
size_t wkeylen;
char *wkeybuf = NULL;
zcrypt_key_t *txgkey;
zap_cursor_t zc;
zap_attribute_t za;
objset_t *mos;
uint64_t keychain_zapobj;
spa_t *spa;
zcrypt_keystore_node_t *zkn;
ASSERT(RRW_WRITE_HELD(&ds->ds_dir->dd_pool->dp_config_rwlock));
mos = ds->ds_dir->dd_pool->dp_meta_objset;
keychain_zapobj = dsl_dir_phys(ds->ds_dir)->dd_keychain_obj;
/*
* To allow for the case were the keychains of child datasets
* are not loaded (ie an explicit 'zfs key -u tank/fs/sub' had
* been done some time before doing 'zfs key -c tank/fs') we itterate
* over the zap objects on disk rather than copying from the
* in memory keystore node.
*/
for (zap_cursor_init(&zc, mos, keychain_zapobj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
wkeylen = za.za_num_integers;
wkeybuf = kmem_alloc(wkeylen, KM_SLEEP);
VERIFY(zap_lookup_uint64(mos, keychain_zapobj,
(uint64_t *)za.za_name, 1, 1, wkeylen, wkeybuf) == 0);
VERIFY(zcrypt_unwrap_key(ca->ca_old_key,
ds->ds_objset->os_crypt, wkeybuf, wkeylen, &txgkey) == 0);
kmem_free(wkeybuf, wkeylen);
VERIFY(zcrypt_wrap_key(ca->ca_new_key, txgkey,
&wkeybuf, &wkeylen,
zio_crypt_select_wrap(ds->ds_objset->os_crypt)) == 0);
zcrypt_key_free(txgkey);
VERIFY(zap_update_uint64(mos, keychain_zapobj,
(uint64_t *)za.za_name, 1, 1, wkeylen, wkeybuf, tx) == 0);
kmem_free(wkeybuf, wkeylen);
}
zap_cursor_fini(&zc);
spa = dsl_dataset_get_spa(ds);
/*
* If the wrapping key is loaded switch the in memory copy now.
*/
zkn = zcrypt_keystore_find_node(spa, ds->ds_object, B_FALSE);
if (zkn != NULL) {
zcrypt_key_free(zkn->skn_wrapkey);
zkn->skn_wrapkey = zcrypt_key_copy(ca->ca_new_key);
}
spa_history_log_internal(spa, "key change", tx,
"succeeded dataset = %llu", ds->ds_object);
}
/*
* Find all 'allow' permissions from a given point and then continue
* traversing up to the root.
*
* This function constructs an nvlist of nvlists.
* each setpoint is an nvlist composed of an nvlist of an nvlist
* of the individual * users/groups/everyone/create
* permissions.
*
* The nvlist will look like this.
*
* { source fsname -> { whokeys { permissions,...}, ...}}
*
* The fsname nvpairs will be arranged in a bottom up order. For example,
* if we have the following structure a/b/c then the nvpairs for the fsnames
* will be ordered a/b/c, a/b, a.
*/
int
dsl_deleg_get(const char *ddname, nvlist_t **nvp)
{
dsl_dir_t *dd, *startdd;
dsl_pool_t *dp;
int error;
objset_t *mos;
error = dsl_dir_open(ddname, FTAG, &startdd, NULL);
if (error)
return (error);
dp = startdd->dd_pool;
mos = dp->dp_meta_objset;
VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
rw_enter(&dp->dp_config_rwlock, RW_READER);
for (dd = startdd; dd != NULL; dd = dd->dd_parent) {
zap_cursor_t basezc;
zap_attribute_t baseza;
nvlist_t *sp_nvp;
uint64_t n;
char source[MAXNAMELEN];
if (dd->dd_phys->dd_deleg_zapobj &&
(zap_count(mos, dd->dd_phys->dd_deleg_zapobj,
&n) == 0) && n) {
VERIFY(nvlist_alloc(&sp_nvp,
NV_UNIQUE_NAME, KM_SLEEP) == 0);
} else {
continue;
}
for (zap_cursor_init(&basezc, mos,
dd->dd_phys->dd_deleg_zapobj);
zap_cursor_retrieve(&basezc, &baseza) == 0;
zap_cursor_advance(&basezc)) {
zap_cursor_t zc;
zap_attribute_t za;
nvlist_t *perms_nvp;
ASSERT(baseza.za_integer_length == 8);
ASSERT(baseza.za_num_integers == 1);
VERIFY(nvlist_alloc(&perms_nvp,
NV_UNIQUE_NAME, KM_SLEEP) == 0);
for (zap_cursor_init(&zc, mos, baseza.za_first_integer);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
VERIFY(nvlist_add_boolean(perms_nvp,
za.za_name) == 0);
}
zap_cursor_fini(&zc);
VERIFY(nvlist_add_nvlist(sp_nvp, baseza.za_name,
perms_nvp) == 0);
nvlist_free(perms_nvp);
}
zap_cursor_fini(&basezc);
dsl_dir_name(dd, source);
VERIFY(nvlist_add_nvlist(*nvp, source, sp_nvp) == 0);
nvlist_free(sp_nvp);
}
rw_exit(&dp->dp_config_rwlock);
dsl_dir_close(startdd, FTAG);
return (0);
}
/*
* Find all 'allow' permissions from a given point and then continue
* traversing up to the root.
*
* This function constructs an nvlist of nvlists.
* each setpoint is an nvlist composed of an nvlist of an nvlist
* of the individual * users/groups/everyone/create
* permissions.
*
* The nvlist will look like this.
*
* { source fsname -> { whokeys { permissions,...}, ...}}
*
* The fsname nvpairs will be arranged in a bottom up order. For example,
* if we have the following structure a/b/c then the nvpairs for the fsnames
* will be ordered a/b/c, a/b, a.
*/
int
dsl_deleg_get(const char *ddname, nvlist_t **nvp)
{
dsl_dir_t *dd, *startdd;
dsl_pool_t *dp;
int error;
objset_t *mos;
error = dsl_pool_hold(ddname, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dir_hold(dp, ddname, FTAG, &startdd, NULL);
if (error != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
dp = startdd->dd_pool;
mos = dp->dp_meta_objset;
VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
for (dd = startdd; dd != NULL; dd = dd->dd_parent) {
zap_cursor_t basezc;
zap_attribute_t baseza;
nvlist_t *sp_nvp;
uint64_t n;
char source[ZFS_MAX_DATASET_NAME_LEN];
if (dsl_dir_phys(dd)->dd_deleg_zapobj == 0 ||
zap_count(mos,
dsl_dir_phys(dd)->dd_deleg_zapobj, &n) != 0 || n == 0)
continue;
sp_nvp = fnvlist_alloc();
for (zap_cursor_init(&basezc, mos,
dsl_dir_phys(dd)->dd_deleg_zapobj);
zap_cursor_retrieve(&basezc, &baseza) == 0;
zap_cursor_advance(&basezc)) {
zap_cursor_t zc;
zap_attribute_t za;
nvlist_t *perms_nvp;
ASSERT(baseza.za_integer_length == 8);
ASSERT(baseza.za_num_integers == 1);
perms_nvp = fnvlist_alloc();
for (zap_cursor_init(&zc, mos, baseza.za_first_integer);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
fnvlist_add_boolean(perms_nvp, za.za_name);
}
zap_cursor_fini(&zc);
fnvlist_add_nvlist(sp_nvp, baseza.za_name, perms_nvp);
fnvlist_free(perms_nvp);
}
zap_cursor_fini(&basezc);
dsl_dir_name(dd, source);
fnvlist_add_nvlist(*nvp, source, sp_nvp);
nvlist_free(sp_nvp);
}
dsl_dir_rele(startdd, FTAG);
dsl_pool_rele(dp, FTAG);
return (0);
}
/*
* Delete the entire contents of a directory. Return a count
* of the number of entries that could not be deleted.
*
* 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)) {
#ifdef __APPLE__
error = zfs_zget_sans_vnode(zfsvfs, ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp);
ASSERT3U(error, ==, 0);
#else
error = zfs_zget(zfsvfs,
ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp);
ASSERT3U(error, ==, 0);
ASSERT((ZTOV(xzp)->v_type == VREG) ||
(ZTOV(xzp)->v_type == VLNK));
#endif /* __APPLE__ */
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);
#ifdef __APPLE__
if (ZTOV(xzp) == NULL) {
zfs_zinactive(xzp);
} else {
VN_RELE(ZTOV(xzp));
}
#else
VN_RELE(ZTOV(xzp));
#endif /* __APPLE__ */
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);
ASSERT3U(error, ==, 0);
dmu_tx_commit(tx);
#ifdef __APPLE__
if (ZTOV(xzp) == NULL) {
zfs_zinactive(xzp);
} else {
VN_RELE(ZTOV(xzp));
}
#else
VN_RELE(ZTOV(xzp));
#endif /* __APPLE__ */
}
zap_cursor_fini(&zc);
ASSERT(error == ENOENT);
return (skipped);
}
/*
* Find all objsets under name, call func on each
*/
int
dmu_objset_find_spa(spa_t *spa, const char *name,
int func(spa_t *, uint64_t, const char *, void *), void *arg, int flags)
{
dsl_dir_t *dd;
dsl_pool_t *dp;
dsl_dataset_t *ds;
zap_cursor_t zc;
zap_attribute_t *attr;
char *child;
uint64_t thisobj;
int err;
if (name == NULL)
name = spa_name(spa);
err = dsl_dir_open_spa(spa, name, FTAG, &dd, NULL);
if (err)
return (err);
/* Don't visit hidden ($MOS & $ORIGIN) objsets. */
if (dd->dd_myname[0] == '$') {
dsl_dir_close(dd, FTAG);
return (0);
}
thisobj = dd->dd_phys->dd_head_dataset_obj;
attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
dp = dd->dd_pool;
/*
* Iterate over all children.
*/
if (flags & DS_FIND_CHILDREN) {
for (zap_cursor_init(&zc, dp->dp_meta_objset,
dd->dd_phys->dd_child_dir_zapobj);
zap_cursor_retrieve(&zc, attr) == 0;
(void) zap_cursor_advance(&zc)) {
ASSERT(attr->za_integer_length == sizeof (uint64_t));
ASSERT(attr->za_num_integers == 1);
child = kmem_alloc(MAXPATHLEN, KM_SLEEP);
(void) strcpy(child, name);
(void) strcat(child, "/");
(void) strcat(child, attr->za_name);
err = dmu_objset_find_spa(spa, child, func, arg, flags);
kmem_free(child, MAXPATHLEN);
if (err)
break;
}
zap_cursor_fini(&zc);
if (err) {
dsl_dir_close(dd, FTAG);
kmem_free(attr, sizeof (zap_attribute_t));
return (err);
}
}
/*
* Iterate over all snapshots.
*/
if (flags & DS_FIND_SNAPSHOTS) {
if (!dsl_pool_sync_context(dp))
rw_enter(&dp->dp_config_rwlock, RW_READER);
err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds);
if (!dsl_pool_sync_context(dp))
rw_exit(&dp->dp_config_rwlock);
if (err == 0) {
uint64_t snapobj = ds->ds_phys->ds_snapnames_zapobj;
dsl_dataset_rele(ds, FTAG);
for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj);
zap_cursor_retrieve(&zc, attr) == 0;
(void) zap_cursor_advance(&zc)) {
ASSERT(attr->za_integer_length ==
sizeof (uint64_t));
ASSERT(attr->za_num_integers == 1);
child = kmem_alloc(MAXPATHLEN, KM_SLEEP);
(void) strcpy(child, name);
(void) strcat(child, "@");
(void) strcat(child, attr->za_name);
err = func(spa, attr->za_first_integer,
child, arg);
kmem_free(child, MAXPATHLEN);
if (err)
break;
}
zap_cursor_fini(&zc);
}
}
dsl_dir_close(dd, FTAG);
kmem_free(attr, sizeof (zap_attribute_t));
if (err)
return (err);
/*
* Apply to self if appropriate.
*/
err = func(spa, thisobj, name, arg);
return (err);
}
static int
sa_attr_table_setup(objset_t *os, sa_attr_reg_t *reg_attrs, int count)
{
sa_os_t *sa = os->os_sa;
uint64_t sa_attr_count = 0;
uint64_t sa_reg_count;
int error = 0;
uint64_t attr_value;
sa_attr_table_t *tb;
zap_cursor_t zc;
zap_attribute_t za;
int registered_count = 0;
int i;
dmu_objset_type_t ostype = dmu_objset_type(os);
sa->sa_user_table =
kmem_zalloc(count * sizeof (sa_attr_type_t), KM_SLEEP);
sa->sa_user_table_sz = count * sizeof (sa_attr_type_t);
if (sa->sa_reg_attr_obj != 0) {
error = zap_count(os, sa->sa_reg_attr_obj,
&sa_attr_count);
/*
* Make sure we retrieved a count and that it isn't zero
*/
if (error || (error == 0 && sa_attr_count == 0)) {
if (error == 0)
error = EINVAL;
goto bail;
}
sa_reg_count = sa_attr_count;
}
if (ostype == DMU_OST_ZFS && sa_attr_count == 0)
sa_attr_count += sa_legacy_attr_count;
/* Allocate attribute numbers for attributes that aren't registered */
for (i = 0; i != count; i++) {
boolean_t found = B_FALSE;
int j;
if (ostype == DMU_OST_ZFS) {
for (j = 0; j != sa_legacy_attr_count; j++) {
if (strcmp(reg_attrs[i].sa_name,
sa_legacy_attrs[j].sa_name) == 0) {
sa->sa_user_table[i] =
sa_legacy_attrs[j].sa_attr;
found = B_TRUE;
}
}
}
if (found)
continue;
if (sa->sa_reg_attr_obj)
error = zap_lookup(os, sa->sa_reg_attr_obj,
reg_attrs[i].sa_name, 8, 1, &attr_value);
else
error = ENOENT;
switch (error) {
case ENOENT:
sa->sa_user_table[i] = (sa_attr_type_t)sa_attr_count;
sa_attr_count++;
break;
case 0:
sa->sa_user_table[i] = ATTR_NUM(attr_value);
break;
default:
goto bail;
}
}
sa->sa_num_attrs = sa_attr_count;
tb = sa->sa_attr_table =
kmem_zalloc(sizeof (sa_attr_table_t) * sa_attr_count, KM_SLEEP);
/*
* Attribute table is constructed from requested attribute list,
* previously foreign registered attributes, and also the legacy
* ZPL set of attributes.
*/
if (sa->sa_reg_attr_obj) {
for (zap_cursor_init(&zc, os, sa->sa_reg_attr_obj);
(error = zap_cursor_retrieve(&zc, &za)) == 0;
zap_cursor_advance(&zc)) {
uint64_t value;
value = za.za_first_integer;
registered_count++;
tb[ATTR_NUM(value)].sa_attr = ATTR_NUM(value);
tb[ATTR_NUM(value)].sa_length = ATTR_LENGTH(value);
tb[ATTR_NUM(value)].sa_byteswap = ATTR_BSWAP(value);
tb[ATTR_NUM(value)].sa_registered = B_TRUE;
if (tb[ATTR_NUM(value)].sa_name) {
continue;
}
tb[ATTR_NUM(value)].sa_name =
kmem_zalloc(strlen(za.za_name) +1, KM_SLEEP);
(void) strlcpy(tb[ATTR_NUM(value)].sa_name, za.za_name,
strlen(za.za_name) +1);
}
zap_cursor_fini(&zc);
/*
* Make sure we processed the correct number of registered
* attributes
*/
if (registered_count != sa_reg_count) {
ASSERT(error != 0);
goto bail;
}
}
if (ostype == DMU_OST_ZFS) {
for (i = 0; i != sa_legacy_attr_count; i++) {
if (tb[i].sa_name)
continue;
tb[i].sa_attr = sa_legacy_attrs[i].sa_attr;
tb[i].sa_length = sa_legacy_attrs[i].sa_length;
tb[i].sa_byteswap = sa_legacy_attrs[i].sa_byteswap;
tb[i].sa_registered = B_FALSE;
tb[i].sa_name =
kmem_zalloc(strlen(sa_legacy_attrs[i].sa_name) +1,
KM_SLEEP);
(void) strlcpy(tb[i].sa_name,
sa_legacy_attrs[i].sa_name,
strlen(sa_legacy_attrs[i].sa_name) + 1);
}
}
for (i = 0; i != count; i++) {
sa_attr_type_t attr_id;
attr_id = sa->sa_user_table[i];
if (tb[attr_id].sa_name)
continue;
tb[attr_id].sa_length = reg_attrs[i].sa_length;
tb[attr_id].sa_byteswap = reg_attrs[i].sa_byteswap;
tb[attr_id].sa_attr = attr_id;
tb[attr_id].sa_name =
kmem_zalloc(strlen(reg_attrs[i].sa_name) + 1, KM_SLEEP);
(void) strlcpy(tb[attr_id].sa_name, reg_attrs[i].sa_name,
strlen(reg_attrs[i].sa_name) + 1);
}
sa->sa_need_attr_registration =
(sa_attr_count != registered_count);
return (0);
bail:
kmem_free(sa->sa_user_table, count * sizeof (sa_attr_type_t));
sa->sa_user_table = NULL;
sa_free_attr_table(sa);
return ((error != 0) ? error : EINVAL);
}
int
sa_setup(objset_t *os, uint64_t sa_obj, sa_attr_reg_t *reg_attrs, int count,
sa_attr_type_t **user_table)
{
zap_cursor_t zc;
zap_attribute_t za;
sa_os_t *sa;
dmu_objset_type_t ostype = dmu_objset_type(os);
sa_attr_type_t *tb;
int error;
mutex_enter(&os->os_lock);
if (os->os_sa) {
mutex_enter(&os->os_sa->sa_lock);
mutex_exit(&os->os_lock);
tb = os->os_sa->sa_user_table;
mutex_exit(&os->os_sa->sa_lock);
*user_table = tb;
return (0);
}
sa = kmem_zalloc(sizeof (sa_os_t), KM_SLEEP);
mutex_init(&sa->sa_lock, NULL, MUTEX_DEFAULT, NULL);
sa->sa_master_obj = sa_obj;
os->os_sa = sa;
mutex_enter(&sa->sa_lock);
mutex_exit(&os->os_lock);
avl_create(&sa->sa_layout_num_tree, layout_num_compare,
sizeof (sa_lot_t), offsetof(sa_lot_t, lot_num_node));
avl_create(&sa->sa_layout_hash_tree, layout_hash_compare,
sizeof (sa_lot_t), offsetof(sa_lot_t, lot_hash_node));
if (sa_obj) {
error = zap_lookup(os, sa_obj, SA_LAYOUTS,
8, 1, &sa->sa_layout_attr_obj);
if (error != 0 && error != ENOENT)
goto fail;
error = zap_lookup(os, sa_obj, SA_REGISTRY,
8, 1, &sa->sa_reg_attr_obj);
if (error != 0 && error != ENOENT)
goto fail;
}
if ((error = sa_attr_table_setup(os, reg_attrs, count)) != 0)
goto fail;
if (sa->sa_layout_attr_obj != 0) {
uint64_t layout_count;
error = zap_count(os, sa->sa_layout_attr_obj,
&layout_count);
/*
* Layout number count should be > 0
*/
if (error || (error == 0 && layout_count == 0)) {
if (error == 0)
error = EINVAL;
goto fail;
}
for (zap_cursor_init(&zc, os, sa->sa_layout_attr_obj);
(error = zap_cursor_retrieve(&zc, &za)) == 0;
zap_cursor_advance(&zc)) {
sa_attr_type_t *lot_attrs;
uint64_t lot_num;
lot_attrs = kmem_zalloc(sizeof (sa_attr_type_t) *
za.za_num_integers, KM_SLEEP);
if ((error = (zap_lookup(os, sa->sa_layout_attr_obj,
za.za_name, 2, za.za_num_integers,
lot_attrs))) != 0) {
kmem_free(lot_attrs, sizeof (sa_attr_type_t) *
za.za_num_integers);
break;
}
VERIFY(ddi_strtoull(za.za_name, NULL, 10,
(unsigned long long *)&lot_num) == 0);
(void) sa_add_layout_entry(os, lot_attrs,
za.za_num_integers, lot_num,
sa_layout_info_hash(lot_attrs,
za.za_num_integers), B_FALSE, NULL);
kmem_free(lot_attrs, sizeof (sa_attr_type_t) *
za.za_num_integers);
}
zap_cursor_fini(&zc);
/*
* Make sure layout count matches number of entries added
* to AVL tree
*/
if (avl_numnodes(&sa->sa_layout_num_tree) != layout_count) {
ASSERT(error != 0);
goto fail;
}
}
/* Add special layout number for old ZNODES */
if (ostype == DMU_OST_ZFS) {
(void) sa_add_layout_entry(os, sa_legacy_zpl_layout,
sa_legacy_attr_count, 0,
sa_layout_info_hash(sa_legacy_zpl_layout,
sa_legacy_attr_count), B_FALSE, NULL);
(void) sa_add_layout_entry(os, sa_dummy_zpl_layout, 0, 1,
0, B_FALSE, NULL);
}
*user_table = os->os_sa->sa_user_table;
mutex_exit(&sa->sa_lock);
return (0);
fail:
os->os_sa = NULL;
sa_free_attr_table(sa);
if (sa->sa_user_table)
kmem_free(sa->sa_user_table, sa->sa_user_table_sz);
mutex_exit(&sa->sa_lock);
kmem_free(sa, sizeof (sa_os_t));
return ((error == ECKSUM) ? EIO : error);
}
