diff options
Diffstat (limited to 'module')
-rw-r--r-- | module/zcommon/zfeature_common.c | 12 | ||||
-rw-r--r-- | module/zcommon/zfs_deleg.c | 4 | ||||
-rw-r--r-- | module/zcommon/zfs_prop.c | 6 | ||||
-rw-r--r-- | module/zfs/dbuf.c | 2 | ||||
-rw-r--r-- | module/zfs/dmu.c | 4 | ||||
-rw-r--r-- | module/zfs/dmu_objset.c | 181 | ||||
-rw-r--r-- | module/zfs/dmu_traverse.c | 26 | ||||
-rw-r--r-- | module/zfs/dnode.c | 22 | ||||
-rw-r--r-- | module/zfs/dsl_pool.c | 2 | ||||
-rw-r--r-- | module/zfs/dsl_scan.c | 6 | ||||
-rw-r--r-- | module/zfs/sa.c | 189 | ||||
-rw-r--r-- | module/zfs/spa.c | 2 | ||||
-rw-r--r-- | module/zfs/zfs_acl.c | 14 | ||||
-rw-r--r-- | module/zfs/zfs_dir.c | 2 | ||||
-rw-r--r-- | module/zfs/zfs_ioctl.c | 35 | ||||
-rw-r--r-- | module/zfs/zfs_log.c | 14 | ||||
-rw-r--r-- | module/zfs/zfs_replay.c | 13 | ||||
-rw-r--r-- | module/zfs/zfs_sa.c | 22 | ||||
-rw-r--r-- | module/zfs/zfs_vfsops.c | 322 | ||||
-rw-r--r-- | module/zfs/zfs_vnops.c | 327 | ||||
-rw-r--r-- | module/zfs/zfs_znode.c | 54 | ||||
-rw-r--r-- | module/zfs/zpl_file.c | 148 |
22 files changed, 1191 insertions, 216 deletions
diff --git a/module/zcommon/zfeature_common.c b/module/zcommon/zfeature_common.c index 7b782b45d..36d0d9613 100644 --- a/module/zcommon/zfeature_common.c +++ b/module/zcommon/zfeature_common.c @@ -321,6 +321,18 @@ zpool_feature_init(void) "Support for dataset level encryption", ZFEATURE_FLAG_PER_DATASET, encryption_deps); } + + { + static const spa_feature_t project_quota_deps[] = { + SPA_FEATURE_EXTENSIBLE_DATASET, + SPA_FEATURE_NONE + }; + zfeature_register(SPA_FEATURE_PROJECT_QUOTA, + "org.zfsonlinux:project_quota", "project_quota", + "space/object accounting based on project ID.", + ZFEATURE_FLAG_READONLY_COMPAT | ZFEATURE_FLAG_PER_DATASET, + project_quota_deps); + } } #if defined(_KERNEL) && defined(HAVE_SPL) diff --git a/module/zcommon/zfs_deleg.c b/module/zcommon/zfs_deleg.c index 18e5c11cc..3a51bc49a 100644 --- a/module/zcommon/zfs_deleg.c +++ b/module/zcommon/zfs_deleg.c @@ -71,6 +71,10 @@ zfs_deleg_perm_tab_t zfs_deleg_perm_tab[] = { {ZFS_DELEG_PERM_RELEASE}, {ZFS_DELEG_PERM_LOAD_KEY}, {ZFS_DELEG_PERM_CHANGE_KEY}, + {ZFS_DELEG_PERM_PROJECTUSED}, + {ZFS_DELEG_PERM_PROJECTQUOTA}, + {ZFS_DELEG_PERM_PROJECTOBJUSED}, + {ZFS_DELEG_PERM_PROJECTOBJQUOTA}, {NULL} }; diff --git a/module/zcommon/zfs_prop.c b/module/zcommon/zfs_prop.c index 42af9468c..0d44fd139 100644 --- a/module/zcommon/zfs_prop.c +++ b/module/zcommon/zfs_prop.c @@ -58,7 +58,11 @@ const char *zfs_userquota_prop_prefixes[] = { "userobjused@", "userobjquota@", "groupobjused@", - "groupobjquota@" + "groupobjquota@", + "projectused@", + "projectquota@", + "projectobjused@", + "projectobjquota@" }; zprop_desc_t * diff --git a/module/zfs/dbuf.c b/module/zfs/dbuf.c index 3668ea315..51cb0c982 100644 --- a/module/zfs/dbuf.c +++ b/module/zfs/dbuf.c @@ -2456,7 +2456,7 @@ dbuf_destroy(dmu_buf_impl_t *db) /* * Note: While bpp will always be updated if the function returns success, * parentp will not be updated if the dnode does not have dn_dbuf filled in; - * this happens when the dnode is the meta-dnode, or a userused or groupused + * this happens when the dnode is the meta-dnode, or {user|group|project}used * object. */ __attribute__((always_inline)) diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c index 20ed3ebff..cb86800f4 100644 --- a/module/zfs/dmu.c +++ b/module/zfs/dmu.c @@ -113,8 +113,8 @@ const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = { { DMU_BSWAP_UINT64, TRUE, FALSE, "FUID table size" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL dataset next clones"}, { DMU_BSWAP_ZAP, TRUE, FALSE, "scan work queue" }, - { DMU_BSWAP_ZAP, TRUE, TRUE, "ZFS user/group used" }, - { DMU_BSWAP_ZAP, TRUE, TRUE, "ZFS user/group quota" }, + { DMU_BSWAP_ZAP, TRUE, TRUE, "ZFS user/group/project used" }, + { DMU_BSWAP_ZAP, TRUE, TRUE, "ZFS user/group/project quota"}, { DMU_BSWAP_ZAP, TRUE, FALSE, "snapshot refcount tags"}, { DMU_BSWAP_ZAP, TRUE, FALSE, "DDT ZAP algorithm" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "DDT statistics" }, diff --git a/module/zfs/dmu_objset.c b/module/zfs/dmu_objset.c index befce9be6..0d9273fbb 100644 --- a/module/zfs/dmu_objset.c +++ b/module/zfs/dmu_objset.c @@ -58,6 +58,7 @@ #include <sys/policy.h> #include <sys/spa_impl.h> #include <sys/dmu_send.h> +#include <sys/zfs_project.h> /* * Needed to close a window in dnode_move() that allows the objset to be freed @@ -336,14 +337,17 @@ dmu_objset_byteswap(void *buf, size_t size) { objset_phys_t *osp = buf; - ASSERT(size == OBJSET_OLD_PHYS_SIZE || size == sizeof (objset_phys_t)); + ASSERT(size == OBJSET_PHYS_SIZE_V1 || size == OBJSET_PHYS_SIZE_V2 || + size == sizeof (objset_phys_t)); dnode_byteswap(&osp->os_meta_dnode); byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t)); osp->os_type = BSWAP_64(osp->os_type); osp->os_flags = BSWAP_64(osp->os_flags); - if (size == sizeof (objset_phys_t)) { + if (size >= OBJSET_PHYS_SIZE_V2) { dnode_byteswap(&osp->os_userused_dnode); dnode_byteswap(&osp->os_groupused_dnode); + if (size >= sizeof (objset_phys_t)) + dnode_byteswap(&osp->os_projectused_dnode); } } @@ -395,6 +399,7 @@ dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, if (!BP_IS_HOLE(os->os_rootbp)) { arc_flags_t aflags = ARC_FLAG_WAIT; zbookmark_phys_t zb; + int size; enum zio_flag zio_flags = ZIO_FLAG_CANFAIL; SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET, ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); @@ -420,12 +425,19 @@ dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, return (err); } + if (spa_version(spa) < SPA_VERSION_USERSPACE) + size = OBJSET_PHYS_SIZE_V1; + else if (!spa_feature_is_enabled(spa, + SPA_FEATURE_PROJECT_QUOTA)) + size = OBJSET_PHYS_SIZE_V2; + else + size = sizeof (objset_phys_t); + /* Increase the blocksize if we are permitted. */ - if (spa_version(spa) >= SPA_VERSION_USERSPACE && - arc_buf_size(os->os_phys_buf) < sizeof (objset_phys_t)) { + if (arc_buf_size(os->os_phys_buf) < size) { arc_buf_t *buf = arc_alloc_buf(spa, &os->os_phys_buf, - ARC_BUFC_METADATA, sizeof (objset_phys_t)); - bzero(buf->b_data, sizeof (objset_phys_t)); + ARC_BUFC_METADATA, size); + bzero(buf->b_data, size); bcopy(os->os_phys_buf->b_data, buf->b_data, arc_buf_size(os->os_phys_buf)); arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); @@ -436,7 +448,7 @@ dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, os->os_flags = os->os_phys->os_flags; } else { int size = spa_version(spa) >= SPA_VERSION_USERSPACE ? - sizeof (objset_phys_t) : OBJSET_OLD_PHYS_SIZE; + sizeof (objset_phys_t) : OBJSET_PHYS_SIZE_V1; os->os_phys_buf = arc_alloc_buf(spa, &os->os_phys_buf, ARC_BUFC_METADATA, size); os->os_phys = os->os_phys_buf->b_data; @@ -568,11 +580,15 @@ dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, dnode_special_open(os, &os->os_phys->os_meta_dnode, DMU_META_DNODE_OBJECT, &os->os_meta_dnode); - if (arc_buf_size(os->os_phys_buf) >= sizeof (objset_phys_t)) { + if (OBJSET_BUF_HAS_USERUSED(os->os_phys_buf)) { dnode_special_open(os, &os->os_phys->os_userused_dnode, DMU_USERUSED_OBJECT, &os->os_userused_dnode); dnode_special_open(os, &os->os_phys->os_groupused_dnode, DMU_GROUPUSED_OBJECT, &os->os_groupused_dnode); + if (OBJSET_BUF_HAS_PROJECTUSED(os->os_phys_buf)) + dnode_special_open(os, + &os->os_phys->os_projectused_dnode, + DMU_PROJECTUSED_OBJECT, &os->os_projectused_dnode); } mutex_init(&os->os_upgrade_lock, NULL, MUTEX_DEFAULT, NULL); @@ -711,9 +727,10 @@ dmu_objset_own(const char *name, dmu_objset_type_t type, } /* user accounting requires the dataset to be decrypted */ - if (dmu_objset_userobjspace_upgradable(*osp) && + if ((dmu_objset_userobjspace_upgradable(*osp) || + dmu_objset_projectquota_upgradable(*osp)) && (ds->ds_dir->dd_crypto_obj == 0 || decrypt)) - dmu_objset_userobjspace_upgrade(*osp); + dmu_objset_id_quota_upgrade(*osp); dsl_pool_rele(dp, FTAG); return (0); @@ -835,6 +852,8 @@ dmu_objset_evict_dbufs(objset_t *os) kmem_free(dn_marker, sizeof (dnode_t)); if (DMU_USERUSED_DNODE(os) != NULL) { + if (DMU_PROJECTUSED_DNODE(os) != NULL) + dnode_evict_dbufs(DMU_PROJECTUSED_DNODE(os)); dnode_evict_dbufs(DMU_GROUPUSED_DNODE(os)); dnode_evict_dbufs(DMU_USERUSED_DNODE(os)); } @@ -889,6 +908,8 @@ dmu_objset_evict_done(objset_t *os) dnode_special_close(&os->os_meta_dnode); if (DMU_USERUSED_DNODE(os)) { + if (DMU_PROJECTUSED_DNODE(os)) + dnode_special_close(&os->os_projectused_dnode); dnode_special_close(&os->os_userused_dnode); dnode_special_close(&os->os_groupused_dnode); } @@ -1004,6 +1025,12 @@ dmu_objset_create_impl_dnstats(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, os->os_phys->os_flags |= OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE; } + if (dmu_objset_projectquota_enabled(os)) { + ds->ds_feature_activation_needed[ + SPA_FEATURE_PROJECT_QUOTA] = B_TRUE; + os->os_phys->os_flags |= + OBJSET_FLAG_PROJECTQUOTA_COMPLETE; + } os->os_flags = os->os_phys->os_flags; } @@ -1408,7 +1435,7 @@ dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg) * Update rootbp fill count: it should be the number of objects * allocated in the object set (not counting the "special" * objects that are stored in the objset_phys_t -- the meta - * dnode and user/group accounting objects). + * dnode and user/group/project accounting objects). */ for (int i = 0; i < dnp->dn_nblkptr; i++) fill += BP_GET_FILL(&dnp->dn_blkptr[i]); @@ -1537,6 +1564,12 @@ dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx) dnode_sync(DMU_GROUPUSED_DNODE(os), tx); } + if (DMU_PROJECTUSED_DNODE(os) && + DMU_PROJECTUSED_DNODE(os)->dn_type != DMU_OT_NONE) { + DMU_PROJECTUSED_DNODE(os)->dn_zio = zio; + dnode_sync(DMU_PROJECTUSED_DNODE(os), tx); + } + txgoff = tx->tx_txg & TXG_MASK; if (dmu_objset_userused_enabled(os) && @@ -1620,6 +1653,14 @@ dmu_objset_userobjused_enabled(objset_t *os) spa_feature_is_enabled(os->os_spa, SPA_FEATURE_USEROBJ_ACCOUNTING)); } +boolean_t +dmu_objset_projectquota_enabled(objset_t *os) +{ + return (used_cbs[os->os_phys->os_type] != NULL && + DMU_PROJECTUSED_DNODE(os) != NULL && + spa_feature_is_enabled(os->os_spa, SPA_FEATURE_PROJECT_QUOTA)); +} + typedef struct userquota_node { /* must be in the first filed, see userquota_update_cache() */ char uqn_id[20 + DMU_OBJACCT_PREFIX_LEN]; @@ -1630,6 +1671,7 @@ typedef struct userquota_node { typedef struct userquota_cache { avl_tree_t uqc_user_deltas; avl_tree_t uqc_group_deltas; + avl_tree_t uqc_project_deltas; } userquota_cache_t; static int @@ -1682,6 +1724,19 @@ do_userquota_cacheflush(objset_t *os, userquota_cache_t *cache, dmu_tx_t *tx) kmem_free(uqn, sizeof (*uqn)); } avl_destroy(&cache->uqc_group_deltas); + + if (dmu_objset_projectquota_enabled(os)) { + cookie = NULL; + while ((uqn = avl_destroy_nodes(&cache->uqc_project_deltas, + &cookie)) != NULL) { + mutex_enter(&os->os_userused_lock); + VERIFY0(zap_increment(os, DMU_PROJECTUSED_OBJECT, + uqn->uqn_id, uqn->uqn_delta, tx)); + mutex_exit(&os->os_userused_lock); + kmem_free(uqn, sizeof (*uqn)); + } + avl_destroy(&cache->uqc_project_deltas); + } } static void @@ -1706,10 +1761,11 @@ userquota_update_cache(avl_tree_t *avl, const char *id, int64_t delta) } static void -do_userquota_update(userquota_cache_t *cache, uint64_t used, uint64_t flags, - uint64_t user, uint64_t group, boolean_t subtract) +do_userquota_update(objset_t *os, userquota_cache_t *cache, uint64_t used, + uint64_t flags, uint64_t user, uint64_t group, uint64_t project, + boolean_t subtract) { - if ((flags & DNODE_FLAG_USERUSED_ACCOUNTED)) { + if (flags & DNODE_FLAG_USERUSED_ACCOUNTED) { int64_t delta = DNODE_MIN_SIZE + used; char name[20]; @@ -1721,12 +1777,18 @@ do_userquota_update(userquota_cache_t *cache, uint64_t used, uint64_t flags, (void) sprintf(name, "%llx", (longlong_t)group); userquota_update_cache(&cache->uqc_group_deltas, name, delta); + + if (dmu_objset_projectquota_enabled(os)) { + (void) sprintf(name, "%llx", (longlong_t)project); + userquota_update_cache(&cache->uqc_project_deltas, + name, delta); + } } } static void -do_userobjquota_update(userquota_cache_t *cache, uint64_t flags, - uint64_t user, uint64_t group, boolean_t subtract) +do_userobjquota_update(objset_t *os, userquota_cache_t *cache, uint64_t flags, + uint64_t user, uint64_t group, uint64_t project, boolean_t subtract) { if (flags & DNODE_FLAG_USEROBJUSED_ACCOUNTED) { char name[20 + DMU_OBJACCT_PREFIX_LEN]; @@ -1739,6 +1801,13 @@ do_userobjquota_update(userquota_cache_t *cache, uint64_t flags, (void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx", (longlong_t)group); userquota_update_cache(&cache->uqc_group_deltas, name, delta); + + if (dmu_objset_projectquota_enabled(os)) { + (void) snprintf(name, sizeof (name), + DMU_OBJACCT_PREFIX "%llx", (longlong_t)project); + userquota_update_cache(&cache->uqc_project_deltas, + name, delta); + } } } @@ -1766,6 +1835,10 @@ userquota_updates_task(void *arg) sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node)); avl_create(&cache.uqc_group_deltas, userquota_compare, sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node)); + if (dmu_objset_projectquota_enabled(os)) + avl_create(&cache.uqc_project_deltas, userquota_compare, + sizeof (userquota_node_t), offsetof(userquota_node_t, + uqn_node)); while ((dn = multilist_sublist_head(list)) != NULL) { int flags; @@ -1777,18 +1850,21 @@ userquota_updates_task(void *arg) flags = dn->dn_id_flags; ASSERT(flags); if (flags & DN_ID_OLD_EXIST) { - do_userquota_update(&cache, - dn->dn_oldused, dn->dn_oldflags, - dn->dn_olduid, dn->dn_oldgid, B_TRUE); - do_userobjquota_update(&cache, dn->dn_oldflags, - dn->dn_olduid, dn->dn_oldgid, B_TRUE); + do_userquota_update(os, &cache, dn->dn_oldused, + dn->dn_oldflags, dn->dn_olduid, dn->dn_oldgid, + dn->dn_oldprojid, B_TRUE); + do_userobjquota_update(os, &cache, dn->dn_oldflags, + dn->dn_olduid, dn->dn_oldgid, + dn->dn_oldprojid, B_TRUE); } if (flags & DN_ID_NEW_EXIST) { - do_userquota_update(&cache, + do_userquota_update(os, &cache, DN_USED_BYTES(dn->dn_phys), dn->dn_phys->dn_flags, - dn->dn_newuid, dn->dn_newgid, B_FALSE); - do_userobjquota_update(&cache, dn->dn_phys->dn_flags, - dn->dn_newuid, dn->dn_newgid, B_FALSE); + dn->dn_newuid, dn->dn_newgid, + dn->dn_newprojid, B_FALSE); + do_userobjquota_update(os, &cache, + dn->dn_phys->dn_flags, dn->dn_newuid, dn->dn_newgid, + dn->dn_newprojid, B_FALSE); } mutex_enter(&dn->dn_mtx); @@ -1797,6 +1873,7 @@ userquota_updates_task(void *arg) if (dn->dn_id_flags & DN_ID_NEW_EXIST) { dn->dn_olduid = dn->dn_newuid; dn->dn_oldgid = dn->dn_newgid; + dn->dn_oldprojid = dn->dn_newprojid; dn->dn_id_flags |= DN_ID_OLD_EXIST; if (dn->dn_bonuslen == 0) dn->dn_id_flags |= DN_ID_CHKED_SPILL; @@ -1824,7 +1901,7 @@ dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx) if (os->os_encrypted && dmu_objset_is_receiving(os)) return; - /* Allocate the user/groupused objects if necessary. */ + /* Allocate the user/group/project used objects if necessary. */ if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) { VERIFY0(zap_create_claim(os, DMU_USERUSED_OBJECT, @@ -1834,6 +1911,12 @@ dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx) DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); } + if (dmu_objset_projectquota_enabled(os) && + DMU_PROJECTUSED_DNODE(os)->dn_type == DMU_OT_NONE) { + VERIFY0(zap_create_claim(os, DMU_PROJECTUSED_OBJECT, + DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); + } + for (int i = 0; i < multilist_get_num_sublists(os->os_synced_dnodes); i++) { userquota_updates_arg_t *uua = @@ -1896,6 +1979,7 @@ dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) dmu_buf_impl_t *db = NULL; uint64_t *user = NULL; uint64_t *group = NULL; + uint64_t *project = NULL; int flags = dn->dn_id_flags; int error; boolean_t have_spill = B_FALSE; @@ -1953,9 +2037,11 @@ dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) ASSERT(data); user = &dn->dn_olduid; group = &dn->dn_oldgid; + project = &dn->dn_oldprojid; } else if (data) { user = &dn->dn_newuid; group = &dn->dn_newgid; + project = &dn->dn_newprojid; } /* @@ -1963,7 +2049,7 @@ dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) * type has changed and that type isn't an object type to track */ error = used_cbs[os->os_phys->os_type](dn->dn_bonustype, data, - user, group); + user, group, project); /* * Preserve existing uid/gid when the callback can't determine @@ -1976,9 +2062,11 @@ dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) if (flags & DN_ID_OLD_EXIST) { dn->dn_newuid = dn->dn_olduid; dn->dn_newgid = dn->dn_oldgid; + dn->dn_newgid = dn->dn_oldprojid; } else { dn->dn_newuid = 0; dn->dn_newgid = 0; + dn->dn_newprojid = ZFS_DEFAULT_PROJID; } error = 0; } @@ -2016,6 +2104,13 @@ dmu_objset_userobjspace_present(objset_t *os) OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE); } +boolean_t +dmu_objset_projectquota_present(objset_t *os) +{ + return (os->os_phys->os_flags & + OBJSET_FLAG_PROJECTQUOTA_COMPLETE); +} + static int dmu_objset_space_upgrade(objset_t *os) { @@ -2085,33 +2180,43 @@ dmu_objset_userspace_upgrade(objset_t *os) } static int -dmu_objset_userobjspace_upgrade_cb(objset_t *os) +dmu_objset_id_quota_upgrade_cb(objset_t *os) { int err = 0; - if (dmu_objset_userobjspace_present(os)) + if (dmu_objset_userobjspace_present(os) && + dmu_objset_projectquota_present(os)) return (0); if (dmu_objset_is_snapshot(os)) return (SET_ERROR(EINVAL)); if (!dmu_objset_userobjused_enabled(os)) return (SET_ERROR(ENOTSUP)); + if (!dmu_objset_projectquota_enabled(os) && + dmu_objset_userobjspace_present(os)) + return (SET_ERROR(ENOTSUP)); dmu_objset_ds(os)->ds_feature_activation_needed[ SPA_FEATURE_USEROBJ_ACCOUNTING] = B_TRUE; + if (dmu_objset_projectquota_enabled(os)) + dmu_objset_ds(os)->ds_feature_activation_needed[ + SPA_FEATURE_PROJECT_QUOTA] = B_TRUE; err = dmu_objset_space_upgrade(os); if (err) return (err); os->os_flags |= OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE; + if (dmu_objset_projectquota_enabled(os)) + os->os_flags |= OBJSET_FLAG_PROJECTQUOTA_COMPLETE; + txg_wait_synced(dmu_objset_pool(os), 0); return (0); } void -dmu_objset_userobjspace_upgrade(objset_t *os) +dmu_objset_id_quota_upgrade(objset_t *os) { - dmu_objset_upgrade(os, dmu_objset_userobjspace_upgrade_cb); + dmu_objset_upgrade(os, dmu_objset_id_quota_upgrade_cb); } boolean_t @@ -2123,6 +2228,15 @@ dmu_objset_userobjspace_upgradable(objset_t *os) !dmu_objset_userobjspace_present(os)); } +boolean_t +dmu_objset_projectquota_upgradable(objset_t *os) +{ + return (dmu_objset_type(os) == DMU_OST_ZFS && + !dmu_objset_is_snapshot(os) && + dmu_objset_projectquota_enabled(os) && + !dmu_objset_projectquota_present(os)); +} + void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, uint64_t *usedobjsp, uint64_t *availobjsp) @@ -2731,7 +2845,10 @@ EXPORT_SYMBOL(dmu_objset_userused_enabled); EXPORT_SYMBOL(dmu_objset_userspace_upgrade); EXPORT_SYMBOL(dmu_objset_userspace_present); EXPORT_SYMBOL(dmu_objset_userobjused_enabled); -EXPORT_SYMBOL(dmu_objset_userobjspace_upgrade); EXPORT_SYMBOL(dmu_objset_userobjspace_upgradable); EXPORT_SYMBOL(dmu_objset_userobjspace_present); +EXPORT_SYMBOL(dmu_objset_projectquota_enabled); +EXPORT_SYMBOL(dmu_objset_projectquota_present); +EXPORT_SYMBOL(dmu_objset_projectquota_upgradable); +EXPORT_SYMBOL(dmu_objset_id_quota_upgrade); #endif diff --git a/module/zfs/dmu_traverse.c b/module/zfs/dmu_traverse.c index 15d29198f..5407e4817 100644 --- a/module/zfs/dmu_traverse.c +++ b/module/zfs/dmu_traverse.c @@ -386,7 +386,11 @@ traverse_visitbp(traverse_data_t *td, const dnode_phys_t *dnp, if (osp->os_meta_dnode.dn_maxblkid == 0) td->td_realloc_possible = B_FALSE; - if (arc_buf_size(buf) >= sizeof (objset_phys_t)) { + if (OBJSET_BUF_HAS_USERUSED(buf)) { + if (OBJSET_BUF_HAS_PROJECTUSED(buf)) + prefetch_dnode_metadata(td, + &osp->os_projectused_dnode, + zb->zb_objset, DMU_PROJECTUSED_OBJECT); prefetch_dnode_metadata(td, &osp->os_groupused_dnode, zb->zb_objset, DMU_GROUPUSED_OBJECT); prefetch_dnode_metadata(td, &osp->os_userused_dnode, @@ -395,13 +399,19 @@ traverse_visitbp(traverse_data_t *td, const dnode_phys_t *dnp, err = traverse_dnode(td, &osp->os_meta_dnode, zb->zb_objset, DMU_META_DNODE_OBJECT); - if (err == 0 && arc_buf_size(buf) >= sizeof (objset_phys_t)) { - err = traverse_dnode(td, &osp->os_groupused_dnode, - zb->zb_objset, DMU_GROUPUSED_OBJECT); - } - if (err == 0 && arc_buf_size(buf) >= sizeof (objset_phys_t)) { - err = traverse_dnode(td, &osp->os_userused_dnode, - zb->zb_objset, DMU_USERUSED_OBJECT); + if (err == 0 && OBJSET_BUF_HAS_USERUSED(buf)) { + if (OBJSET_BUF_HAS_PROJECTUSED(buf)) + err = traverse_dnode(td, + &osp->os_projectused_dnode, zb->zb_objset, + DMU_PROJECTUSED_OBJECT); + if (err == 0) + err = traverse_dnode(td, + &osp->os_groupused_dnode, zb->zb_objset, + DMU_GROUPUSED_OBJECT); + if (err == 0) + err = traverse_dnode(td, + &osp->os_userused_dnode, zb->zb_objset, + DMU_USERUSED_OBJECT); } } diff --git a/module/zfs/dnode.c b/module/zfs/dnode.c index b4c131e98..596983b47 100644 --- a/module/zfs/dnode.c +++ b/module/zfs/dnode.c @@ -38,6 +38,7 @@ #include <sys/dmu_zfetch.h> #include <sys/range_tree.h> #include <sys/trace_dnode.h> +#include <sys/zfs_project.h> dnode_stats_t dnode_stats = { { "dnode_hold_dbuf_hold", KSTAT_DATA_UINT64 }, @@ -157,8 +158,10 @@ dnode_cons(void *arg, void *unused, int kmflag) dn->dn_oldflags = 0; dn->dn_olduid = 0; dn->dn_oldgid = 0; + dn->dn_oldprojid = ZFS_DEFAULT_PROJID; dn->dn_newuid = 0; dn->dn_newgid = 0; + dn->dn_newprojid = ZFS_DEFAULT_PROJID; dn->dn_id_flags = 0; dn->dn_dbufs_count = 0; @@ -210,8 +213,10 @@ dnode_dest(void *arg, void *unused) ASSERT0(dn->dn_oldflags); ASSERT0(dn->dn_olduid); ASSERT0(dn->dn_oldgid); + ASSERT0(dn->dn_oldprojid); ASSERT0(dn->dn_newuid); ASSERT0(dn->dn_newgid); + ASSERT0(dn->dn_newprojid); ASSERT0(dn->dn_id_flags); ASSERT0(dn->dn_dbufs_count); @@ -543,8 +548,10 @@ dnode_destroy(dnode_t *dn) dn->dn_oldflags = 0; dn->dn_olduid = 0; dn->dn_oldgid = 0; + dn->dn_oldprojid = ZFS_DEFAULT_PROJID; dn->dn_newuid = 0; dn->dn_newgid = 0; + dn->dn_newprojid = ZFS_DEFAULT_PROJID; dn->dn_id_flags = 0; dmu_zfetch_fini(&dn->dn_zfetch); @@ -799,8 +806,10 @@ dnode_move_impl(dnode_t *odn, dnode_t *ndn) ndn->dn_oldflags = odn->dn_oldflags; ndn->dn_olduid = odn->dn_olduid; ndn->dn_oldgid = odn->dn_oldgid; + ndn->dn_oldprojid = odn->dn_oldprojid; ndn->dn_newuid = odn->dn_newuid; ndn->dn_newgid = odn->dn_newgid; + ndn->dn_newprojid = odn->dn_newprojid; ndn->dn_id_flags = odn->dn_id_flags; dmu_zfetch_init(&ndn->dn_zfetch, NULL); list_move_tail(&ndn->dn_zfetch.zf_stream, &odn->dn_zfetch.zf_stream); @@ -859,8 +868,10 @@ dnode_move_impl(dnode_t *odn, dnode_t *ndn) odn->dn_oldflags = 0; odn->dn_olduid = 0; odn->dn_oldgid = 0; + odn->dn_oldprojid = ZFS_DEFAULT_PROJID; odn->dn_newuid = 0; odn->dn_newgid = 0; + odn->dn_newprojid = ZFS_DEFAULT_PROJID; odn->dn_id_flags = 0; /* @@ -1265,9 +1276,14 @@ dnode_hold_impl(objset_t *os, uint64_t object, int flag, int slots, (spa_is_root(os->os_spa) && spa_config_held(os->os_spa, SCL_STATE, RW_WRITER))); - if (object == DMU_USERUSED_OBJECT || object == DMU_GROUPUSED_OBJECT) { - dn = (object == DMU_USERUSED_OBJECT) ? - DMU_USERUSED_DNODE(os) : DMU_GROUPUSED_DNODE(os); + if (object == DMU_USERUSED_OBJECT || object == DMU_GROUPUSED_OBJECT || + object == DMU_PROJECTUSED_OBJECT) { + if (object == DMU_USERUSED_OBJECT) + dn = DMU_USERUSED_DNODE(os); + else if (object == DMU_GROUPUSED_OBJECT) + dn = DMU_GROUPUSED_DNODE(os); + else + dn = DMU_PROJECTUSED_DNODE(os); if (dn == NULL) return (SET_ERROR(ENOENT)); type = dn->dn_type; diff --git a/module/zfs/dsl_pool.c b/module/zfs/dsl_pool.c index 86863fad8..db2e67742 100644 --- a/module/zfs/dsl_pool.c +++ b/module/zfs/dsl_pool.c @@ -580,7 +580,7 @@ dsl_pool_sync(dsl_pool_t *dp, uint64_t txg) /* * After the data blocks have been written (ensured by the zio_wait() - * above), update the user/group space accounting. This happens + * above), update the user/group/project space accounting. This happens * in tasks dispatched to dp_sync_taskq, so wait for them before * continuing. */ diff --git a/module/zfs/dsl_scan.c b/module/zfs/dsl_scan.c index fc0c24e1c..776f8f239 100644 --- a/module/zfs/dsl_scan.c +++ b/module/zfs/dsl_scan.c @@ -1684,11 +1684,15 @@ dsl_scan_recurse(dsl_scan_t *scn, dsl_dataset_t *ds, dmu_objset_type_t ostype, if (OBJSET_BUF_HAS_USERUSED(buf)) { /* - * We also always visit user/group accounting + * We also always visit user/group/project accounting * objects, and never skip them, even if we are * suspending. This is necessary so that the * space deltas from this txg get integrated. */ + if (OBJSET_BUF_HAS_PROJECTUSED(buf)) + dsl_scan_visitdnode(scn, ds, osp->os_type, + &osp->os_projectused_dnode, + DMU_PROJECTUSED_OBJECT, tx); dsl_scan_visitdnode(scn, ds, osp->os_type, &osp->os_groupused_dnode, DMU_GROUPUSED_OBJECT, tx); diff --git a/module/zfs/sa.c b/module/zfs/sa.c index f0a18bad8..4a863f9a5 100644 --- a/module/zfs/sa.c +++ b/module/zfs/sa.c @@ -44,6 +44,10 @@ #include <sys/errno.h> #include <sys/zfs_context.h> +#ifdef _KERNEL +#include <sys/zfs_znode.h> +#endif + /* * ZFS System attributes: * @@ -1456,8 +1460,9 @@ sa_lookup_impl(sa_handle_t *hdl, sa_bulk_attr_t *bulk, int count) return (sa_attr_op(hdl, bulk, count, SA_LOOKUP, NULL)); } -int -sa_lookup(sa_handle_t *hdl, sa_attr_type_t attr, void *buf, uint32_t buflen) +static int +sa_lookup_locked(sa_handle_t *hdl, sa_attr_type_t attr, void *buf, + uint32_t buflen) { int error; sa_bulk_attr_t bulk; @@ -1470,9 +1475,19 @@ sa_lookup(sa_handle_t *hdl, sa_attr_type_t attr, void *buf, uint32_t buflen) bulk.sa_data_func = NULL; ASSERT(hdl); - mutex_enter(&hdl->sa_lock); error = sa_lookup_impl(hdl, &bulk, 1); + return (error); +} + +int +sa_lookup(sa_handle_t *hdl, sa_attr_type_t attr, void *buf, uint32_t buflen) +{ + int error; + + mutex_enter(&hdl->sa_lock); + error = sa_lookup_locked(hdl, attr, buf, buflen); mutex_exit(&hdl->sa_lock); + return (error); } @@ -1497,6 +1512,173 @@ sa_lookup_uio(sa_handle_t *hdl, sa_attr_type_t attr, uio_t *uio) mutex_exit(&hdl->sa_lock); return (error); } + +/* + * For the existed object that is upgraded from old system, its ondisk layout + * has no slot for the project ID attribute. But quota accounting logic needs + * to access related slots by offset directly. So we need to adjust these old + * objects' layout to make the project ID to some unified and fixed offset. + */ +int +sa_add_projid(sa_handle_t *hdl, dmu_tx_t *tx, uint64_t projid) +{ + znode_t *zp = sa_get_userdata(hdl); + dmu_buf_t *db = sa_get_db(hdl); + zfsvfs_t *zfsvfs = ZTOZSB(zp); + int count = 0, err = 0; + sa_bulk_attr_t *bulk, *attrs; + zfs_acl_locator_cb_t locate = { 0 }; + uint64_t uid, gid, mode, rdev, xattr = 0, parent, gen, links; + uint64_t crtime[2], mtime[2], ctime[2], atime[2]; + zfs_acl_phys_t znode_acl = { 0 }; + char scanstamp[AV_SCANSTAMP_SZ]; + + if (zp->z_acl_cached == NULL) { + zfs_acl_t *aclp; + + mutex_enter(&zp->z_acl_lock); + err = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE); + mutex_exit(&zp->z_acl_lock); + if (err != 0 && err != ENOENT) + return (err); + } + + bulk = kmem_zalloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); + attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); + mutex_enter(&hdl->sa_lock); + mutex_enter(&zp->z_lock); + + err = sa_lookup_locked(hdl, SA_ZPL_PROJID(zfsvfs), &projid, + sizeof (uint64_t)); + if (unlikely(err == 0)) + /* Someone has added project ID attr by race. */ + err = EEXIST; + if (err != ENOENT) + goto out; + + /* First do a bulk query of the attributes that aren't cached */ + if (zp->z_is_sa) { + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, + &mode, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, + &gen, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, + &uid, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, + &gid, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, + &parent, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, + &atime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, + &mtime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, + &ctime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, + &crtime, 16); + if (S_ISBLK(ZTOI(zp)->i_mode) || S_ISCHR(ZTOI(zp)->i_mode)) + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL, + &rdev, 8); + } else { + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, + &atime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, + &mtime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, + &ctime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, + &crtime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, + &gen, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, + &mode, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, + &parent, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL, + &xattr, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL, + &rdev, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, + &uid, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, + &gid, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL, + &znode_acl, 88); + } + err = sa_bulk_lookup_locked(hdl, bulk, count); + if (err != 0) + goto out; + + err = sa_lookup_locked(hdl, SA_ZPL_XATTR(zfsvfs), &xattr, 8); + if (err != 0 && err != ENOENT) + goto out; + + zp->z_projid = projid; + zp->z_pflags |= ZFS_PROJID; + links = ZTOI(zp)->i_nlink; + count = 0; + err = 0; + + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_SIZE(zfsvfs), NULL, + &zp->z_size, 8); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_GEN(zfsvfs), NULL, &gen, 8); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL, + &zp->z_pflags, 8); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL, + &crtime, 16); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_PROJID(zfsvfs), NULL, &projid, 8); + + if (S_ISBLK(ZTOI(zp)->i_mode) || S_ISCHR(ZTOI(zp)->i_mode)) + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_RDEV(zfsvfs), NULL, + &rdev, 8); + + if (zp->z_acl_cached != NULL) { + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), NULL, + &zp->z_acl_cached->z_acl_count, 8); + if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID) + zfs_acl_xform(zp, zp->z_acl_cached, CRED()); + locate.cb_aclp = zp->z_acl_cached; + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_DACL_ACES(zfsvfs), + zfs_acl_data_locator, &locate, + zp->z_acl_cached->z_acl_bytes); + } + + if (xattr) + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_XATTR(zfsvfs), NULL, + &xattr, 8); + + if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) { + bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE, + scanstamp, AV_SCANSTAMP_SZ); + SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_SCANSTAMP(zfsvfs), NULL, + scanstamp, AV_SCANSTAMP_SZ); + zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP; + } + + VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0); + VERIFY(sa_replace_all_by_template_locked(hdl, attrs, count, tx) == 0); + if (znode_acl.z_acl_extern_obj) { + VERIFY(0 == dmu_object_free(zfsvfs->z_os, + znode_acl.z_acl_extern_obj, tx)); + } + + zp->z_is_sa = B_TRUE; + +out: + mutex_exit(&zp->z_lock); + mutex_exit(&hdl->sa_lock); + kmem_free(attrs, sizeof (sa_bulk_attr_t) * ZPL_END); + kmem_free(bulk, sizeof (sa_bulk_attr_t) * ZPL_END); + return (err); +} #endif static sa_idx_tab_t * @@ -2062,4 +2244,5 @@ EXPORT_SYMBOL(sa_hdrsize); EXPORT_SYMBOL(sa_handle_lock); EXPORT_SYMBOL(sa_handle_unlock); EXPORT_SYMBOL(sa_lookup_uio); +EXPORT_SYMBOL(sa_add_projid); #endif /* _KERNEL */ diff --git a/module/zfs/spa.c b/module/zfs/spa.c index dac042464..736b51fea 100644 --- a/module/zfs/spa.c +++ b/module/zfs/spa.c @@ -1188,7 +1188,7 @@ spa_activate(spa_t *spa, int mode) /* * The taskq to upgrade datasets in this pool. Currently used by - * feature SPA_FEATURE_USEROBJ_ACCOUNTING. + * feature SPA_FEATURE_USEROBJ_ACCOUNTING/SPA_FEATURE_PROJECT_QUOTA. */ spa->spa_upgrade_taskq = taskq_create("z_upgrade", boot_ncpus, defclsyspri, 1, INT_MAX, TASKQ_DYNAMIC); diff --git a/module/zfs/zfs_acl.c b/module/zfs/zfs_acl.c index 5ef20f088..b366e8f1c 100644 --- a/module/zfs/zfs_acl.c +++ b/module/zfs/zfs_acl.c @@ -1054,8 +1054,8 @@ zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp, * Read an external acl object. If the intent is to modify, always * create a new acl and leave any cached acl in place. */ -static int -zfs_acl_node_read(znode_t *zp, boolean_t have_lock, zfs_acl_t **aclpp, +int +zfs_acl_node_read(struct znode *zp, boolean_t have_lock, zfs_acl_t **aclpp, boolean_t will_modify) { zfs_acl_t *aclp; @@ -1883,12 +1883,12 @@ zfs_acl_ids_free(zfs_acl_ids_t *acl_ids) } boolean_t -zfs_acl_ids_overquota(zfsvfs_t *zfsvfs, zfs_acl_ids_t *acl_ids) +zfs_acl_ids_overquota(zfsvfs_t *zv, zfs_acl_ids_t *acl_ids, uint64_t projid) { - return (zfs_fuid_overquota(zfsvfs, B_FALSE, acl_ids->z_fuid) || - zfs_fuid_overquota(zfsvfs, B_TRUE, acl_ids->z_fgid) || - zfs_fuid_overobjquota(zfsvfs, B_FALSE, acl_ids->z_fuid) || - zfs_fuid_overobjquota(zfsvfs, B_TRUE, acl_ids->z_fgid)); + return (zfs_id_overquota(zv, DMU_USERUSED_OBJECT, acl_ids->z_fuid) || + zfs_id_overquota(zv, DMU_GROUPUSED_OBJECT, acl_ids->z_fgid) || + (projid != ZFS_DEFAULT_PROJID && projid != ZFS_INVALID_PROJID && + zfs_id_overquota(zv, DMU_PROJECTUSED_OBJECT, projid))); } /* diff --git a/module/zfs/zfs_dir.c b/module/zfs/zfs_dir.c index 6398a1d15..7eb426b78 100644 --- a/module/zfs/zfs_dir.c +++ b/module/zfs/zfs_dir.c @@ -1036,7 +1036,7 @@ zfs_make_xattrdir(znode_t *zp, vattr_t *vap, struct inode **xipp, cred_t *cr) 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)) { + if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zp->z_projid)) { zfs_acl_ids_free(&acl_ids); return (SET_ERROR(EDQUOT)); } diff --git a/module/zfs/zfs_ioctl.c b/module/zfs/zfs_ioctl.c index fcd2fca12..6bee042a0 100644 --- a/module/zfs/zfs_ioctl.c +++ b/module/zfs/zfs_ioctl.c @@ -260,10 +260,14 @@ static const char *userquota_perms[] = { ZFS_DELEG_PERM_USEROBJQUOTA, ZFS_DELEG_PERM_GROUPOBJUSED, ZFS_DELEG_PERM_GROUPOBJQUOTA, + ZFS_DELEG_PERM_PROJECTUSED, + ZFS_DELEG_PERM_PROJECTQUOTA, + ZFS_DELEG_PERM_PROJECTOBJUSED, + ZFS_DELEG_PERM_PROJECTOBJQUOTA, }; static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); -static int zfs_ioc_userobjspace_upgrade(zfs_cmd_t *zc); +static int zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc); static int zfs_check_settable(const char *name, nvpair_t *property, cred_t *cr); static int zfs_check_clearable(char *dataset, nvlist_t *props, @@ -1200,10 +1204,14 @@ zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) zc->zc_objset_type == ZFS_PROP_USEROBJQUOTA) { if (zc->zc_guid == crgetuid(cr)) return (0); - } else { + } else if (zc->zc_objset_type == ZFS_PROP_GROUPUSED || + zc->zc_objset_type == ZFS_PROP_GROUPQUOTA || + zc->zc_objset_type == ZFS_PROP_GROUPOBJUSED || + zc->zc_objset_type == ZFS_PROP_GROUPOBJQUOTA) { if (groupmember(zc->zc_guid, cr)) return (0); } + /* else is for project quota/used */ } return (zfs_secpolicy_write_perms(zc->zc_name, @@ -2516,7 +2524,7 @@ zfs_prop_set_special(const char *dsname, zprop_source_t source, zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); (void) strcpy(zc->zc_name, dsname); (void) zfs_ioc_userspace_upgrade(zc); - (void) zfs_ioc_userobjspace_upgrade(zc); + (void) zfs_ioc_id_quota_upgrade(zc); kmem_free(zc, sizeof (zfs_cmd_t)); } break; @@ -3897,6 +3905,10 @@ zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA]; const char *giq_prefix = zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA]; + const char *pq_prefix = + zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA]; + const char *piq_prefix = zfs_userquota_prop_prefixes[\ + ZFS_PROP_PROJECTOBJQUOTA]; if (strncmp(propname, uq_prefix, strlen(uq_prefix)) == 0) { @@ -3910,8 +3922,14 @@ zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) } else if (strncmp(propname, giq_prefix, strlen(giq_prefix)) == 0) { perm = ZFS_DELEG_PERM_GROUPOBJQUOTA; + } else if (strncmp(propname, pq_prefix, + strlen(pq_prefix)) == 0) { + perm = ZFS_DELEG_PERM_PROJECTQUOTA; + } else if (strncmp(propname, piq_prefix, + strlen(piq_prefix)) == 0) { + perm = ZFS_DELEG_PERM_PROJECTOBJQUOTA; } else { - /* USERUSED and GROUPUSED are read-only */ + /* {USER|GROUP|PROJECT}USED are read-only */ return (SET_ERROR(EINVAL)); } @@ -5180,7 +5198,7 @@ zfs_ioc_promote(zfs_cmd_t *zc) } /* - * Retrieve a single {user|group}{used|quota}@... property. + * Retrieve a single {user|group|project}{used|quota}@... property. * * inputs: * zc_name name of filesystem @@ -5306,7 +5324,7 @@ zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) * none */ static int -zfs_ioc_userobjspace_upgrade(zfs_cmd_t *zc) +zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc) { objset_t *os; int error; @@ -5315,14 +5333,15 @@ zfs_ioc_userobjspace_upgrade(zfs_cmd_t *zc) if (error != 0) return (error); - if (dmu_objset_userobjspace_upgradable(os)) { + if (dmu_objset_userobjspace_upgradable(os) || + dmu_objset_projectquota_upgradable(os)) { mutex_enter(&os->os_upgrade_lock); if (os->os_upgrade_id == 0) { /* clear potential error code and retry */ os->os_upgrade_status = 0; mutex_exit(&os->os_upgrade_lock); - dmu_objset_userobjspace_upgrade(os); + dmu_objset_id_quota_upgrade(os); } else { mutex_exit(&os->os_upgrade_lock); } diff --git a/module/zfs/zfs_log.c b/module/zfs/zfs_log.c index 8887f037a..ce7b84927 100644 --- a/module/zfs/zfs_log.c +++ b/module/zfs/zfs_log.c @@ -166,8 +166,17 @@ zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap) XAT0_AV_MODIFIED; if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) ZFS_TIME_ENCODE(&xoap->xoa_createtime, crtime); - if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) + if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { + ASSERT(!XVA_ISSET_REQ(xvap, XAT_PROJID)); + bcopy(xoap->xoa_av_scanstamp, scanstamp, AV_SCANSTAMP_SZ); + } else if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { + /* + * XAT_PROJID and XAT_AV_SCANSTAMP will never be valid + * at the same time, so we can share the same space. + */ + bcopy(&xoap->xoa_projid, scanstamp, sizeof (uint64_t)); + } if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) *attrs |= (xoap->xoa_reparse == 0) ? 0 : XAT0_REPARSE; @@ -177,6 +186,9 @@ zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap) if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) *attrs |= (xoap->xoa_sparse == 0) ? 0 : XAT0_SPARSE; + if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) + *attrs |= (xoap->xoa_projinherit == 0) ? 0 : + XAT0_PROJINHERIT; } static void * diff --git a/module/zfs/zfs_replay.c b/module/zfs/zfs_replay.c index bfba2fea0..e2ff00789 100644 --- a/module/zfs/zfs_replay.c +++ b/module/zfs/zfs_replay.c @@ -128,14 +128,25 @@ zfs_replay_xvattr(lr_attr_t *lrattr, xvattr_t *xvap) ((*attrs & XAT0_AV_QUARANTINED) != 0); if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) ZFS_TIME_DECODE(&xoap->xoa_createtime, crtime); - if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) + if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { + ASSERT(!XVA_ISSET_REQ(xvap, XAT_PROJID)); + bcopy(scanstamp, xoap->xoa_av_scanstamp, AV_SCANSTAMP_SZ); + } else if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { + /* + * XAT_PROJID and XAT_AV_SCANSTAMP will never be valid + * at the same time, so we can share the same space. + */ + bcopy(scanstamp, &xoap->xoa_projid, sizeof (uint64_t)); + } if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) xoap->xoa_reparse = ((*attrs & XAT0_REPARSE) != 0); if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) xoap->xoa_offline = ((*attrs & XAT0_OFFLINE) != 0); if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) xoap->xoa_sparse = ((*attrs & XAT0_SPARSE) != 0); + if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) + xoap->xoa_projinherit = ((*attrs & XAT0_PROJINHERIT) != 0); } static int diff --git a/module/zfs/zfs_sa.c b/module/zfs/zfs_sa.c index 3eff6acc6..bd21ba896 100644 --- a/module/zfs/zfs_sa.c +++ b/module/zfs/zfs_sa.c @@ -27,6 +27,8 @@ #include <sys/sa.h> #include <sys/zfs_acl.h> #include <sys/zfs_sa.h> +#include <sys/dmu_objset.h> +#include <sys/sa_impl.h> /* * ZPL attribute registration table. @@ -63,6 +65,7 @@ sa_attr_reg_t zfs_attr_table[ZPL_END+1] = { {"ZPL_SCANSTAMP", 32, SA_UINT8_ARRAY, 0}, {"ZPL_DACL_ACES", 0, SA_ACL, 0}, {"ZPL_DXATTR", 0, SA_UINT8_ARRAY, 0}, + {"ZPL_PROJID", sizeof (uint64_t), SA_UINT64_ARRAY, 0}, {NULL, 0, 0, 0} }; @@ -317,7 +320,7 @@ zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx) } /* First do a bulk query of the attributes that aren't cached */ - bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 20, KM_SLEEP); + bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); @@ -332,9 +335,13 @@ zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx) SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL, &znode_acl, 88); - if (sa_bulk_lookup_locked(hdl, bulk, count) != 0) { - kmem_free(bulk, sizeof (sa_bulk_attr_t) * 20); + if (sa_bulk_lookup_locked(hdl, bulk, count) != 0) goto done; + + if (dmu_objset_projectquota_enabled(hdl->sa_os) && + !(zp->z_pflags & ZFS_PROJID)) { + zp->z_pflags |= ZFS_PROJID; + zp->z_projid = ZFS_DEFAULT_PROJID; } /* @@ -342,7 +349,7 @@ zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx) * it is such a way to pick up an already existing layout number */ count = 0; - sa_attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * 20, KM_SLEEP); + sa_attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zfsvfs), NULL, &zp->z_size, 8); @@ -365,6 +372,9 @@ zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx) links = ZTOI(zp)->i_nlink; SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); + if (dmu_objset_projectquota_enabled(hdl->sa_os)) + SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PROJID(zfsvfs), NULL, + &zp->z_projid, 8); if (S_ISBLK(ZTOI(zp)->i_mode) || S_ISCHR(ZTOI(zp)->i_mode)) SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zfsvfs), NULL, &rdev, 8); @@ -400,9 +410,9 @@ zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx) znode_acl.z_acl_extern_obj, tx)); zp->z_is_sa = B_TRUE; - kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * 20); - kmem_free(bulk, sizeof (sa_bulk_attr_t) * 20); + kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END); done: + kmem_free(bulk, sizeof (sa_bulk_attr_t) * ZPL_END); if (drop_lock) mutex_exit(&zp->z_lock); } diff --git a/module/zfs/zfs_vfsops.c b/module/zfs/zfs_vfsops.c index bb380c920..971fd54bc 100644 --- a/module/zfs/zfs_vfsops.c +++ b/module/zfs/zfs_vfsops.c @@ -536,8 +536,14 @@ unregister: static int zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, - uint64_t *userp, uint64_t *groupp) + uint64_t *userp, uint64_t *groupp, uint64_t *projectp) { + sa_hdr_phys_t sa; + sa_hdr_phys_t *sap = data; + uint64_t flags; + int hdrsize; + boolean_t swap = B_FALSE; + /* * Is it a valid type of object to track? */ @@ -557,42 +563,49 @@ zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, znode_phys_t *znp = data; *userp = znp->zp_uid; *groupp = znp->zp_gid; + *projectp = ZFS_DEFAULT_PROJID; + return (0); + } + + if (sap->sa_magic == 0) { + /* + * This should only happen for newly created files + * that haven't had the znode data filled in yet. + */ + *userp = 0; + *groupp = 0; + *projectp = ZFS_DEFAULT_PROJID; + return (0); + } + + sa = *sap; + if (sa.sa_magic == BSWAP_32(SA_MAGIC)) { + sa.sa_magic = SA_MAGIC; + sa.sa_layout_info = BSWAP_16(sa.sa_layout_info); + swap = B_TRUE; } else { - int hdrsize; - sa_hdr_phys_t *sap = data; - sa_hdr_phys_t sa = *sap; - boolean_t swap = B_FALSE; - - ASSERT(bonustype == DMU_OT_SA); - - if (sa.sa_magic == 0) { - /* - * This should only happen for newly created - * files that haven't had the znode data filled - * in yet. - */ - *userp = 0; - *groupp = 0; - return (0); - } - if (sa.sa_magic == BSWAP_32(SA_MAGIC)) { - sa.sa_magic = SA_MAGIC; - sa.sa_layout_info = BSWAP_16(sa.sa_layout_info); - swap = B_TRUE; - } else { - VERIFY3U(sa.sa_magic, ==, SA_MAGIC); - } + VERIFY3U(sa.sa_magic, ==, SA_MAGIC); + } - hdrsize = sa_hdrsize(&sa); - VERIFY3U(hdrsize, >=, sizeof (sa_hdr_phys_t)); - *userp = *((uint64_t *)((uintptr_t)data + hdrsize + - SA_UID_OFFSET)); - *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + - SA_GID_OFFSET)); - if (swap) { - *userp = BSWAP_64(*userp); - *groupp = BSWAP_64(*groupp); - } + hdrsize = sa_hdrsize(&sa); + VERIFY3U(hdrsize, >=, sizeof (sa_hdr_phys_t)); + + *userp = *((uint64_t *)((uintptr_t)data + hdrsize + SA_UID_OFFSET)); + *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + SA_GID_OFFSET)); + flags = *((uint64_t *)((uintptr_t)data + hdrsize + SA_FLAGS_OFFSET)); + if (swap) + flags = BSWAP_64(flags); + + if (flags & ZFS_PROJID) + *projectp = *((uint64_t *)((uintptr_t)data + hdrsize + + SA_PROJID_OFFSET)); + else + *projectp = ZFS_DEFAULT_PROJID; + + if (swap) { + *userp = BSWAP_64(*userp); + *groupp = BSWAP_64(*groupp); + *projectp = BSWAP_64(*projectp); } return (0); } @@ -624,6 +637,9 @@ zfs_userquota_prop_to_obj(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type) case ZFS_PROP_GROUPUSED: case ZFS_PROP_GROUPOBJUSED: return (DMU_GROUPUSED_OBJECT); + case ZFS_PROP_PROJECTUSED: + case ZFS_PROP_PROJECTOBJUSED: + return (DMU_PROJECTUSED_OBJECT); case ZFS_PROP_USERQUOTA: return (zfsvfs->z_userquota_obj); case ZFS_PROP_GROUPQUOTA: @@ -632,6 +648,10 @@ zfs_userquota_prop_to_obj(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type) return (zfsvfs->z_userobjquota_obj); case ZFS_PROP_GROUPOBJQUOTA: return (zfsvfs->z_groupobjquota_obj); + case ZFS_PROP_PROJECTQUOTA: + return (zfsvfs->z_projectquota_obj); + case ZFS_PROP_PROJECTOBJQUOTA: + return (zfsvfs->z_projectobjquota_obj); default: return (ZFS_NO_OBJECT); } @@ -651,8 +671,16 @@ zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, if (!dmu_objset_userspace_present(zfsvfs->z_os)) return (SET_ERROR(ENOTSUP)); + if ((type == ZFS_PROP_PROJECTQUOTA || type == ZFS_PROP_PROJECTUSED || + type == ZFS_PROP_PROJECTOBJQUOTA || + type == ZFS_PROP_PROJECTOBJUSED) && + !dmu_objset_projectquota_present(zfsvfs->z_os)) + return (SET_ERROR(ENOTSUP)); + if ((type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED || - type == ZFS_PROP_USEROBJQUOTA || type == ZFS_PROP_GROUPOBJQUOTA) && + type == ZFS_PROP_USEROBJQUOTA || type == ZFS_PROP_GROUPOBJQUOTA || + type == ZFS_PROP_PROJECTOBJUSED || + type == ZFS_PROP_PROJECTOBJQUOTA) && !dmu_objset_userobjspace_present(zfsvfs->z_os)) return (SET_ERROR(ENOTSUP)); @@ -662,7 +690,8 @@ zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, return (0); } - if (type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED) + if (type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED || + type == ZFS_PROP_PROJECTOBJUSED) offset = DMU_OBJACCT_PREFIX_LEN; for (zap_cursor_init_serialized(&zc, zfsvfs->z_os, obj, *cookiep); @@ -731,15 +760,27 @@ zfs_userspace_one(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, return (SET_ERROR(ENOTSUP)); if ((type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED || - type == ZFS_PROP_USEROBJQUOTA || type == ZFS_PROP_GROUPOBJQUOTA) && + type == ZFS_PROP_USEROBJQUOTA || type == ZFS_PROP_GROUPOBJQUOTA || + type == ZFS_PROP_PROJECTOBJUSED || + type == ZFS_PROP_PROJECTOBJQUOTA) && !dmu_objset_userobjspace_present(zfsvfs->z_os)) return (SET_ERROR(ENOTSUP)); + if (type == ZFS_PROP_PROJECTQUOTA || type == ZFS_PROP_PROJECTUSED || + type == ZFS_PROP_PROJECTOBJQUOTA || + type == ZFS_PROP_PROJECTOBJUSED) { + if (!dmu_objset_projectquota_present(zfsvfs->z_os)) + return (SET_ERROR(ENOTSUP)); + if (!zpl_is_valid_projid(rid)) + return (SET_ERROR(EINVAL)); + } + obj = zfs_userquota_prop_to_obj(zfsvfs, type); if (obj == ZFS_NO_OBJECT) return (0); - if (type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED) { + if (type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED || + type == ZFS_PROP_PROJECTOBJUSED) { strlcpy(buf, DMU_OBJACCT_PREFIX, DMU_OBJACCT_PREFIX_LEN + 1); offset = DMU_OBJACCT_PREFIX_LEN; } @@ -780,6 +821,22 @@ zfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, case ZFS_PROP_GROUPOBJQUOTA: objp = &zfsvfs->z_groupobjquota_obj; break; + case ZFS_PROP_PROJECTQUOTA: + if (!dmu_objset_projectquota_enabled(zfsvfs->z_os)) + return (SET_ERROR(ENOTSUP)); + if (!zpl_is_valid_projid(rid)) + return (SET_ERROR(EINVAL)); + + objp = &zfsvfs->z_projectquota_obj; + break; + case ZFS_PROP_PROJECTOBJQUOTA: + if (!dmu_objset_projectquota_enabled(zfsvfs->z_os)) + return (SET_ERROR(ENOTSUP)); + if (!zpl_is_valid_projid(rid)) + return (SET_ERROR(EINVAL)); + + objp = &zfsvfs->z_projectobjquota_obj; + break; default: return (SET_ERROR(EINVAL)); } @@ -827,35 +884,51 @@ zfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, } boolean_t -zfs_fuid_overobjquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid) +zfs_id_overobjquota(zfsvfs_t *zfsvfs, uint64_t usedobj, uint64_t id) { char buf[20 + DMU_OBJACCT_PREFIX_LEN]; - uint64_t used, quota, usedobj, quotaobj; + uint64_t used, quota, quotaobj; int err; if (!dmu_objset_userobjspace_present(zfsvfs->z_os)) { if (dmu_objset_userobjspace_upgradable(zfsvfs->z_os)) { dsl_pool_config_enter( dmu_objset_pool(zfsvfs->z_os), FTAG); - dmu_objset_userobjspace_upgrade(zfsvfs->z_os); + dmu_objset_id_quota_upgrade(zfsvfs->z_os); dsl_pool_config_exit( dmu_objset_pool(zfsvfs->z_os), FTAG); } return (B_FALSE); } - usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; - quotaobj = isgroup ? zfsvfs->z_groupobjquota_obj : - zfsvfs->z_userobjquota_obj; + if (usedobj == DMU_PROJECTUSED_OBJECT) { + if (!dmu_objset_projectquota_present(zfsvfs->z_os)) { + if (dmu_objset_projectquota_upgradable(zfsvfs->z_os)) { + dsl_pool_config_enter( + dmu_objset_pool(zfsvfs->z_os), FTAG); + dmu_objset_id_quota_upgrade(zfsvfs->z_os); + dsl_pool_config_exit( + dmu_objset_pool(zfsvfs->z_os), FTAG); + } + return (B_FALSE); + } + quotaobj = zfsvfs->z_projectobjquota_obj; + } else if (usedobj == DMU_USERUSED_OBJECT) { + quotaobj = zfsvfs->z_userobjquota_obj; + } else if (usedobj == DMU_GROUPUSED_OBJECT) { + quotaobj = zfsvfs->z_groupobjquota_obj; + } else { + return (B_FALSE); + } if (quotaobj == 0 || zfsvfs->z_replay) return (B_FALSE); - (void) sprintf(buf, "%llx", (longlong_t)fuid); + (void) sprintf(buf, "%llx", (longlong_t)id); err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a); if (err != 0) return (B_FALSE); - (void) sprintf(buf, DMU_OBJACCT_PREFIX "%llx", (longlong_t)fuid); + (void) sprintf(buf, DMU_OBJACCT_PREFIX "%llx", (longlong_t)id); err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used); if (err != 0) return (B_FALSE); @@ -863,19 +936,35 @@ zfs_fuid_overobjquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid) } boolean_t -zfs_fuid_overquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid) +zfs_id_overblockquota(zfsvfs_t *zfsvfs, uint64_t usedobj, uint64_t id) { char buf[20]; - uint64_t used, quota, usedobj, quotaobj; + uint64_t used, quota, quotaobj; int err; - usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; - quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; - + if (usedobj == DMU_PROJECTUSED_OBJECT) { + if (!dmu_objset_projectquota_present(zfsvfs->z_os)) { + if (dmu_objset_projectquota_upgradable(zfsvfs->z_os)) { + dsl_pool_config_enter( + dmu_objset_pool(zfsvfs->z_os), FTAG); + dmu_objset_id_quota_upgrade(zfsvfs->z_os); + dsl_pool_config_exit( + dmu_objset_pool(zfsvfs->z_os), FTAG); + } + return (B_FALSE); + } + quotaobj = zfsvfs->z_projectquota_obj; + } else if (usedobj == DMU_USERUSED_OBJECT) { + quotaobj = zfsvfs->z_userquota_obj; + } else if (usedobj == DMU_GROUPUSED_OBJECT) { + quotaobj = zfsvfs->z_groupquota_obj; + } else { + return (B_FALSE); + } if (quotaobj == 0 || zfsvfs->z_replay) return (B_FALSE); - (void) sprintf(buf, "%llx", (longlong_t)fuid); + (void) sprintf(buf, "%llx", (longlong_t)id); err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a); if (err != 0) return (B_FALSE); @@ -887,20 +976,10 @@ zfs_fuid_overquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid) } boolean_t -zfs_owner_overquota(zfsvfs_t *zfsvfs, znode_t *zp, boolean_t isgroup) +zfs_id_overquota(zfsvfs_t *zfsvfs, uint64_t usedobj, uint64_t id) { - uint64_t fuid; - uint64_t quotaobj; - struct inode *ip = ZTOI(zp); - - quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; - - fuid = isgroup ? KGID_TO_SGID(ip->i_gid) : KUID_TO_SUID(ip->i_uid); - - if (quotaobj == 0 || zfsvfs->z_replay) - return (B_FALSE); - - return (zfs_fuid_overquota(zfsvfs, isgroup, fuid)); + return (zfs_id_overblockquota(zfsvfs, usedobj, id) || + zfs_id_overobjquota(zfsvfs, usedobj, id)); } /* @@ -1008,6 +1087,14 @@ zfsvfs_init(zfsvfs_t *zfsvfs, objset_t *os) return (error); error = zap_lookup(os, MASTER_NODE_OBJ, + zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA], + 8, 1, &zfsvfs->z_projectquota_obj); + if (error == ENOENT) + zfsvfs->z_projectquota_obj = 0; + else if (error != 0) + return (error); + + error = zap_lookup(os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA], 8, 1, &zfsvfs->z_userobjquota_obj); if (error == ENOENT) @@ -1023,6 +1110,14 @@ zfsvfs_init(zfsvfs_t *zfsvfs, objset_t *os) else if (error != 0) return (error); + error = zap_lookup(os, MASTER_NODE_OBJ, + zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTOBJQUOTA], + 8, 1, &zfsvfs->z_projectobjquota_obj); + if (error == ENOENT) + zfsvfs->z_projectobjquota_obj = 0; + else if (error != 0) + return (error); + error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, &zfsvfs->z_fuid_obj); if (error == ENOENT) @@ -1264,6 +1359,83 @@ zfs_check_global_label(const char *dsname, const char *hexsl) } #endif /* HAVE_MLSLABEL */ +static int +zfs_statfs_project(zfsvfs_t *zfsvfs, znode_t *zp, struct kstatfs *statp, + uint32_t bshift) +{ + char buf[20 + DMU_OBJACCT_PREFIX_LEN]; + uint64_t offset = DMU_OBJACCT_PREFIX_LEN; + uint64_t quota; + uint64_t used; + int err; + + strlcpy(buf, DMU_OBJACCT_PREFIX, DMU_OBJACCT_PREFIX_LEN + 1); + err = id_to_fuidstr(zfsvfs, NULL, zp->z_projid, buf + offset, B_FALSE); + if (err) + return (err); + + if (zfsvfs->z_projectquota_obj == 0) + goto objs; + + err = zap_lookup(zfsvfs->z_os, zfsvfs->z_projectquota_obj, + buf + offset, 8, 1, "a); + if (err == ENOENT) + goto objs; + else if (err) + return (err); + + err = zap_lookup(zfsvfs->z_os, DMU_PROJECTUSED_OBJECT, + buf + offset, 8, 1, &used); + if (unlikely(err == ENOENT)) { + uint32_t blksize; + u_longlong_t nblocks; + + /* + * Quota accounting is async, so it is possible race case. + * There is at least one object with the given project ID. + */ + sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); + if (unlikely(zp->z_blksz == 0)) + blksize = zfsvfs->z_max_blksz; + + used = blksize * nblocks; + } else if (err) { + return (err); + } + + statp->f_blocks = quota >> bshift; + statp->f_bfree = (quota > used) ? ((quota - used) >> bshift) : 0; + statp->f_bavail = statp->f_bfree; + +objs: + if (zfsvfs->z_projectobjquota_obj == 0) + return (0); + + err = zap_lookup(zfsvfs->z_os, zfsvfs->z_projectobjquota_obj, + buf + offset, 8, 1, "a); + if (err == ENOENT) + return (0); + else if (err) + return (err); + + err = zap_lookup(zfsvfs->z_os, DMU_PROJECTUSED_OBJECT, + buf, 8, 1, &used); + if (unlikely(err == ENOENT)) { + /* + * Quota accounting is async, so it is possible race case. + * There is at least one object with the given project ID. + */ + used = 1; + } else if (err) { + return (err); + } + + statp->f_files = quota; + statp->f_ffree = (quota > used) ? (quota - used) : 0; + + return (0); +} + int zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) { @@ -1271,6 +1443,7 @@ zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) uint64_t refdbytes, availbytes, usedobjs, availobjs; uint64_t fsid; uint32_t bshift; + int err = 0; ZFS_ENTER(zfsvfs); @@ -1322,8 +1495,17 @@ zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) */ bzero(statp->f_spare, sizeof (statp->f_spare)); + if (dmu_objset_projectquota_enabled(zfsvfs->z_os) && + dmu_objset_projectquota_present(zfsvfs->z_os)) { + znode_t *zp = ITOZ(dentry->d_inode); + + if (zp->z_pflags & ZFS_PROJINHERIT && zp->z_projid && + zpl_is_valid_projid(zp->z_projid)) + err = zfs_statfs_project(zfsvfs, zp, statp, bshift); + } + ZFS_EXIT(zfsvfs); - return (0); + return (err); } int @@ -2170,9 +2352,9 @@ EXPORT_SYMBOL(zfs_resume_fs); EXPORT_SYMBOL(zfs_userspace_one); EXPORT_SYMBOL(zfs_userspace_many); EXPORT_SYMBOL(zfs_set_userquota); -EXPORT_SYMBOL(zfs_owner_overquota); -EXPORT_SYMBOL(zfs_fuid_overquota); -EXPORT_SYMBOL(zfs_fuid_overobjquota); +EXPORT_SYMBOL(zfs_id_overblockquota); +EXPORT_SYMBOL(zfs_id_overobjquota); +EXPORT_SYMBOL(zfs_id_overquota); EXPORT_SYMBOL(zfs_set_version); EXPORT_SYMBOL(zfsvfs_create); EXPORT_SYMBOL(zfsvfs_free); diff --git a/module/zfs/zfs_vnops.c b/module/zfs/zfs_vnops.c index e21fe1aca..f35165de3 100644 --- a/module/zfs/zfs_vnops.c +++ b/module/zfs/zfs_vnops.c @@ -79,6 +79,7 @@ #include <sys/attr.h> #include <sys/zpl.h> #include <sys/zil.h> +#include <sys/sa_impl.h> /* * Programming rules. @@ -728,8 +729,13 @@ zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) while (n > 0) { abuf = NULL; woff = uio->uio_loffset; - if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || - zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { + if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, + KUID_TO_SUID(ip->i_uid)) || + zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, + KGID_TO_SGID(ip->i_gid)) || + (zp->z_projid != ZFS_DEFAULT_PROJID && + zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, + zp->z_projid))) { if (abuf != NULL) dmu_return_arcbuf(abuf); error = SET_ERROR(EDQUOT); @@ -1380,6 +1386,7 @@ top: if (zp == NULL) { uint64_t txtype; + uint64_t projid = ZFS_DEFAULT_PROJID; /* * Create a new file object and update the directory @@ -1408,7 +1415,9 @@ top: goto out; have_acl = B_TRUE; - if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { + if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) + projid = zfs_inherit_projid(dzp); + if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { zfs_acl_ids_free(&acl_ids); error = SET_ERROR(EDQUOT); goto out; @@ -1552,6 +1561,7 @@ zfs_tmpfile(struct inode *dip, vattr_t *vap, int excl, uid_t uid; gid_t gid; zfs_acl_ids_t acl_ids; + uint64_t projid = ZFS_DEFAULT_PROJID; boolean_t fuid_dirtied; boolean_t have_acl = B_FALSE; boolean_t waited = B_FALSE; @@ -1598,7 +1608,9 @@ top: goto out; have_acl = B_TRUE; - if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { + if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) + projid = zfs_inherit_projid(dzp); + if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { zfs_acl_ids_free(&acl_ids); error = SET_ERROR(EDQUOT); goto out; @@ -2009,7 +2021,7 @@ top: return (error); } - if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { + if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zfs_inherit_projid(dzp))) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); ZFS_EXIT(zfsvfs); @@ -2591,6 +2603,17 @@ zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) ((zp->z_pflags & ZFS_SPARSE) != 0); XVA_SET_RTN(xvap, XAT_SPARSE); } + + if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { + xoap->xoa_projinherit = + ((zp->z_pflags & ZFS_PROJINHERIT) != 0); + XVA_SET_RTN(xvap, XAT_PROJINHERIT); + } + + if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { + xoap->xoa_projid = zp->z_projid; + XVA_SET_RTN(xvap, XAT_PROJID); + } } ZFS_TIME_DECODE(&vap->va_atime, atime); @@ -2669,6 +2692,125 @@ zfs_getattr_fast(struct inode *ip, struct kstat *sp) } /* + * For the operation of changing file's user/group/project, we need to + * handle not only the main object that is assigned to the file directly, + * but also the ones that are used by the file via hidden xattr directory. + * + * Because the xattr directory may contains many EA entries, as to it may + * be impossible to change all of them via the transaction of changing the + * main object's user/group/project attributes. Then we have to change them + * via other multiple independent transactions one by one. It may be not good + * solution, but we have no better idea yet. + */ +static int +zfs_setattr_dir(znode_t *dzp) +{ + struct inode *dxip = ZTOI(dzp); + struct inode *xip = NULL; + zfsvfs_t *zfsvfs = ITOZSB(dxip); + objset_t *os = zfsvfs->z_os; + zap_cursor_t zc; + zap_attribute_t zap; + zfs_dirlock_t *dl; + znode_t *zp; + dmu_tx_t *tx = NULL; + uint64_t uid, gid; + sa_bulk_attr_t bulk[4]; + int count = 0; + int err; + + zap_cursor_init(&zc, os, dzp->z_id); + while ((err = zap_cursor_retrieve(&zc, &zap)) == 0) { + if (zap.za_integer_length != 8 || zap.za_num_integers != 1) { + err = ENXIO; + break; + } + + err = zfs_dirent_lock(&dl, dzp, (char *)zap.za_name, &zp, + ZEXISTS, NULL, NULL); + if (err == ENOENT) + goto next; + if (err) + break; + + xip = ZTOI(zp); + if (KUID_TO_SUID(xip->i_uid) == KUID_TO_SUID(dxip->i_uid) && + KGID_TO_SGID(xip->i_gid) == KGID_TO_SGID(dxip->i_gid) && + zp->z_projid == dzp->z_projid) + goto next; + + tx = dmu_tx_create(os); + if (!(zp->z_pflags & ZFS_PROJID)) + dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); + else + dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); + + err = dmu_tx_assign(tx, TXG_WAIT); + if (err) + break; + + mutex_enter(&dzp->z_lock); + + if (KUID_TO_SUID(xip->i_uid) != KUID_TO_SUID(dxip->i_uid)) { + xip->i_uid = dxip->i_uid; + uid = zfs_uid_read(dxip); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, + &uid, sizeof (uid)); + } + + if (KGID_TO_SGID(xip->i_gid) != KGID_TO_SGID(dxip->i_gid)) { + xip->i_gid = dxip->i_gid; + gid = zfs_gid_read(dxip); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, + &gid, sizeof (gid)); + } + + if (zp->z_projid != dzp->z_projid) { + if (!(zp->z_pflags & ZFS_PROJID)) { + zp->z_pflags |= ZFS_PROJID; + SA_ADD_BULK_ATTR(bulk, count, + SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, + sizeof (zp->z_pflags)); + } + + zp->z_projid = dzp->z_projid; + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PROJID(zfsvfs), + NULL, &zp->z_projid, sizeof (zp->z_projid)); + } + + mutex_exit(&dzp->z_lock); + + if (likely(count > 0)) { + err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); + dmu_tx_commit(tx); + } else { + dmu_tx_abort(tx); + } + tx = NULL; + if (err != 0 && err != ENOENT) + break; + +next: + if (xip) { + iput(xip); + xip = NULL; + zfs_dirent_unlock(dl); + } + zap_cursor_advance(&zc); + } + + if (tx) + dmu_tx_abort(tx); + if (xip) { + iput(xip); + zfs_dirent_unlock(dl); + } + zap_cursor_fini(&zc); + + return (err == ENOENT ? 0 : err); +} + +/* * Set the file attributes to the values contained in the * vattr structure. * @@ -2691,6 +2833,7 @@ zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ITOZSB(ip); + objset_t *os = zfsvfs->z_os; zilog_t *zilog; dmu_tx_t *tx; vattr_t oldva; @@ -2702,17 +2845,19 @@ zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid; uint64_t xattr_obj; uint64_t mtime[2], ctime[2], atime[2]; + uint64_t projid = ZFS_INVALID_PROJID; znode_t *attrzp; int need_policy = FALSE; - int err, err2; + int err, err2 = 0; zfs_fuid_info_t *fuidp = NULL; xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ xoptattr_t *xoap; zfs_acl_t *aclp; boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; boolean_t fuid_dirtied = B_FALSE; + boolean_t handle_eadir = B_FALSE; sa_bulk_attr_t *bulk, *xattr_bulk; - int count = 0, xattr_count = 0; + int count = 0, xattr_count = 0, bulks = 8; if (mask == 0) return (0); @@ -2720,6 +2865,39 @@ zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); + /* + * If this is a xvattr_t, then get a pointer to the structure of + * optional attributes. If this is NULL, then we have a vattr_t. + */ + xoap = xva_getxoptattr(xvap); + if (xoap != NULL && (mask & ATTR_XVATTR)) { + if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { + if (!dmu_objset_projectquota_enabled(os) || + (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode))) { + ZFS_EXIT(zfsvfs); + return (SET_ERROR(ENOTSUP)); + } + + projid = xoap->xoa_projid; + if (unlikely(projid == ZFS_INVALID_PROJID)) { + ZFS_EXIT(zfsvfs); + return (SET_ERROR(EINVAL)); + } + + if (projid == zp->z_projid && zp->z_pflags & ZFS_PROJID) + projid = ZFS_INVALID_PROJID; + else + need_policy = TRUE; + } + + if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT) && + (!dmu_objset_projectquota_enabled(os) || + (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode)))) { + ZFS_EXIT(zfsvfs); + return (SET_ERROR(ENOTSUP)); + } + } + zilog = zfsvfs->z_log; /* @@ -2745,17 +2923,11 @@ zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) return (SET_ERROR(EINVAL)); } - /* - * If this is an xvattr_t, then get a pointer to the structure of - * optional attributes. If this is NULL, then we have a vattr_t. - */ - xoap = xva_getxoptattr(xvap); - tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); xva_init(tmpxvattr); - bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); - xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); + bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); + xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); /* * Immutable files can only alter immutable bit and atime @@ -2901,6 +3073,16 @@ top: } } + if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { + if (xoap->xoa_projinherit != + ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) { + need_policy = TRUE; + } else { + XVA_CLR_REQ(xvap, XAT_PROJINHERIT); + XVA_SET_REQ(tmpxvattr, XAT_PROJINHERIT); + } + } + if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { if (xoap->xoa_nounlink != ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { @@ -3009,7 +3191,8 @@ top: */ mask = vap->va_mask; - if ((mask & (ATTR_UID | ATTR_GID))) { + if ((mask & (ATTR_UID | ATTR_GID)) || projid != ZFS_INVALID_PROJID) { + handle_eadir = B_TRUE; err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xattr_obj, sizeof (xattr_obj)); @@ -3022,7 +3205,8 @@ top: new_kuid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) && - zfs_fuid_overquota(zfsvfs, B_FALSE, new_kuid)) { + zfs_id_overquota(zfsvfs, DMU_USERUSED_OBJECT, + new_kuid)) { if (attrzp) iput(ZTOI(attrzp)); err = SET_ERROR(EDQUOT); @@ -3034,15 +3218,24 @@ top: new_kgid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp); if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) && - zfs_fuid_overquota(zfsvfs, B_TRUE, new_kgid)) { + zfs_id_overquota(zfsvfs, DMU_GROUPUSED_OBJECT, + new_kgid)) { if (attrzp) iput(ZTOI(attrzp)); err = SET_ERROR(EDQUOT); goto out2; } } + + if (projid != ZFS_INVALID_PROJID && + zfs_id_overquota(zfsvfs, DMU_PROJECTUSED_OBJECT, projid)) { + if (attrzp) + iput(ZTOI(attrzp)); + err = EDQUOT; + goto out2; + } } - tx = dmu_tx_create(zfsvfs->z_os); + tx = dmu_tx_create(os); if (mask & ATTR_MODE) { uint64_t pmode = zp->z_mode; @@ -3075,8 +3268,10 @@ top: mutex_exit(&zp->z_lock); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); } else { - if ((mask & ATTR_XVATTR) && - XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) + if (((mask & ATTR_XVATTR) && + XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) || + (projid != ZFS_INVALID_PROJID && + !(zp->z_pflags & ZFS_PROJID))) dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); else dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); @@ -3105,6 +3300,26 @@ top: * updated as a side-effect of calling this function. */ + if (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID)) { + /* + * For the existed object that is upgraded from old system, + * its on-disk layout has no slot for the project ID attribute. + * But quota accounting logic needs to access related slots by + * offset directly. So we need to adjust old objects' layout + * to make the project ID to some unified and fixed offset. + */ + if (attrzp) + err = sa_add_projid(attrzp->z_sa_hdl, tx, projid); + if (err == 0) + err = sa_add_projid(zp->z_sa_hdl, tx, projid); + + if (unlikely(err == EEXIST)) + err = 0; + else if (err != 0) + goto out; + else + projid = ZFS_INVALID_PROJID; + } if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) mutex_enter(&zp->z_acl_lock); @@ -3120,6 +3335,12 @@ top: SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, sizeof (attrzp->z_pflags)); + if (projid != ZFS_INVALID_PROJID) { + attrzp->z_projid = projid; + SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, + SA_ZPL_PROJID(zfsvfs), NULL, &attrzp->z_projid, + sizeof (attrzp->z_projid)); + } } if (mask & (ATTR_UID|ATTR_GID)) { @@ -3199,6 +3420,13 @@ top: ctime, sizeof (ctime)); } + if (projid != ZFS_INVALID_PROJID) { + zp->z_projid = projid; + SA_ADD_BULK_ATTR(bulk, count, + SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid, + sizeof (zp->z_projid)); + } + if (attrzp && mask) { SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, @@ -3235,6 +3463,9 @@ top: if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); } + if (XVA_ISSET_REQ(tmpxvattr, XAT_PROJINHERIT)) { + XVA_SET_REQ(xvap, XAT_PROJINHERIT); + } if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) ASSERT(S_ISREG(ip->i_mode)); @@ -3258,7 +3489,7 @@ top: mutex_exit(&attrzp->z_lock); } out: - if (err == 0 && attrzp) { + if (err == 0 && xattr_count > 0) { err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, xattr_count, tx); ASSERT(err2 == 0); @@ -3279,20 +3510,24 @@ out: if (err == ERESTART) goto top; } else { - err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); + if (count > 0) + err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); dmu_tx_commit(tx); - if (attrzp) + if (attrzp) { + if (err2 == 0 && handle_eadir) + err2 = zfs_setattr_dir(attrzp); iput(ZTOI(attrzp)); + } zfs_inode_update(zp); } out2: - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + if (os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); out3: - kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * 7); - kmem_free(bulk, sizeof (sa_bulk_attr_t) * 7); + kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * bulks); + kmem_free(bulk, sizeof (sa_bulk_attr_t) * bulks); kmem_free(tmpxvattr, sizeof (xvattr_t)); ZFS_EXIT(zfsvfs); return (err); @@ -3586,6 +3821,19 @@ top: } /* + * If we are using project inheritance, means if the directory has + * ZFS_PROJINHERIT set, then its descendant directories will inherit + * not only the project ID, but also the ZFS_PROJINHERIT flag. Under + * such case, we only allow renames into our tree when the project + * IDs are the same. + */ + if (tdzp->z_pflags & ZFS_PROJINHERIT && + tdzp->z_projid != szp->z_projid) { + error = SET_ERROR(EXDEV); + goto out; + } + + /* * Must have write access at the source to remove the old entry * and write access at the target to create the new entry. * Note that if target and source are the same, this can be @@ -3683,6 +3931,8 @@ top: error = zfs_link_create(tdl, szp, tx, ZRENAMING); if (error == 0) { szp->z_pflags |= ZFS_AV_MODIFIED; + if (tdzp->z_pflags & ZFS_PROJINHERIT) + szp->z_pflags |= ZFS_PROJINHERIT; error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), (void *)&szp->z_pflags, sizeof (uint64_t), tx); @@ -3829,7 +4079,7 @@ top: return (error); } - if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { + if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, ZFS_DEFAULT_PROJID)) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); ZFS_EXIT(zfsvfs); @@ -4013,6 +4263,18 @@ zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr, ZFS_VERIFY_ZP(szp); /* + * If we are using project inheritance, means if the directory has + * ZFS_PROJINHERIT set, then its descendant directories will inherit + * not only the project ID, but also the ZFS_PROJINHERIT flag. Under + * such case, we only allow hard link creation in our tree when the + * project IDs are the same. + */ + if (dzp->z_pflags & ZFS_PROJINHERIT && dzp->z_projid != szp->z_projid) { + ZFS_EXIT(zfsvfs); + return (SET_ERROR(EXDEV)); + } + + /* * We check i_sb because snapshots and the ctldir must have different * super blocks. */ @@ -4206,8 +4468,13 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc) * is to register a page_mkwrite() handler to count the page * against its quota when it is about to be dirtied. */ - if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || - zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { + if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, + KUID_TO_SUID(ip->i_uid)) || + zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, + KGID_TO_SGID(ip->i_gid)) || + (zp->z_projid != ZFS_DEFAULT_PROJID && + zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, + zp->z_projid))) { err = EDQUOT; } #endif diff --git a/module/zfs/zfs_znode.c b/module/zfs/zfs_znode.c index d3b68403b..5288c9c68 100644 --- a/module/zfs/zfs_znode.c +++ b/module/zfs/zfs_znode.c @@ -328,6 +328,7 @@ zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx) sharezp->z_atime_dirty = 0; sharezp->z_zfsvfs = zfsvfs; sharezp->z_is_sa = zfsvfs->z_use_sa; + sharezp->z_pflags = 0; vp = ZTOV(sharezp); vn_reinit(vp); @@ -558,6 +559,7 @@ zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, uint64_t links; uint64_t z_uid, z_gid; uint64_t atime[2], mtime[2], ctime[2]; + uint64_t projid = ZFS_DEFAULT_PROJID; sa_bulk_attr_t bulk[11]; int count = 0; @@ -604,13 +606,17 @@ zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); - if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || tmp_gen == 0) { + if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || tmp_gen == 0 || + (dmu_objset_projectquota_enabled(zfsvfs->z_os) && + (zp->z_pflags & ZFS_PROJID) && + sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), &projid, 8) != 0)) { if (hdl == NULL) sa_handle_destroy(zp->z_sa_hdl); zp->z_sa_hdl = NULL; goto error; } + zp->z_projid = projid; zp->z_mode = ip->i_mode = mode; ip->i_generation = (uint32_t)tmp_gen; ip->i_blkbits = SPA_MINBLOCKSHIFT; @@ -696,7 +702,7 @@ zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, { uint64_t crtime[2], atime[2], mtime[2], ctime[2]; uint64_t mode, size, links, parent, pflags; - uint64_t dzp_pflags = 0; + uint64_t projid = ZFS_DEFAULT_PROJID; uint64_t rdev = 0; zfsvfs_t *zfsvfs = ZTOZSB(dzp); dmu_buf_t *db; @@ -771,14 +777,12 @@ zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, */ if (flag & IS_ROOT_NODE) { dzp->z_id = obj; - } else { - dzp_pflags = dzp->z_pflags; } /* * If parent is an xattr, so am I. */ - if (dzp_pflags & ZFS_XATTR) { + if (dzp->z_pflags & ZFS_XATTR) { flag |= IS_XATTR; } @@ -803,6 +807,23 @@ zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, if (flag & IS_XATTR) pflags |= ZFS_XATTR; + if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) { + /* + * With ZFS_PROJID flag, we can easily know whether there is + * project ID stored on disk or not. See zfs_space_delta_cb(). + */ + if (obj_type != DMU_OT_ZNODE && + dmu_objset_projectquota_enabled(zfsvfs->z_os)) + pflags |= ZFS_PROJID; + + /* + * Inherit project ID from parent if required. + */ + projid = zfs_inherit_projid(dzp); + if (dzp->z_pflags & ZFS_PROJINHERIT) + pflags |= ZFS_PROJINHERIT; + } + /* * No execs denied will be deterimed when zfs_mode_compute() is called. */ @@ -884,6 +905,10 @@ zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, if (obj_type == DMU_OT_ZNODE) { SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL, &empty_xattr, 8); + } else if (dmu_objset_projectquota_enabled(zfsvfs->z_os) && + pflags & ZFS_PROJID) { + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PROJID(zfsvfs), + NULL, &projid, 8); } if (obj_type == DMU_OT_ZNODE || (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) { @@ -942,6 +967,7 @@ zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, (*zpp)->z_pflags = pflags; (*zpp)->z_mode = ZTOI(*zpp)->i_mode = mode; (*zpp)->z_dnodesize = dnodesize; + (*zpp)->z_projid = projid; if (obj_type == DMU_OT_ZNODE || acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { @@ -1049,6 +1075,11 @@ zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_SPARSE); } + if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { + ZFS_ATTR_SET(zp, ZFS_PROJINHERIT, xoap->xoa_projinherit, + zp->z_pflags, tx); + XVA_SET_RTN(xvap, XAT_PROJINHERIT); + } if (update_inode) zfs_set_inode_flags(zp, ZTOI(zp)); @@ -1166,6 +1197,7 @@ zfs_rezget(znode_t *zp) uint64_t gen; uint64_t z_uid, z_gid; uint64_t atime[2], mtime[2], ctime[2]; + uint64_t projid = ZFS_DEFAULT_PROJID; znode_hold_t *zh; /* @@ -1241,6 +1273,17 @@ zfs_rezget(znode_t *zp) return (SET_ERROR(EIO)); } + if (dmu_objset_projectquota_enabled(zfsvfs->z_os)) { + err = sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), + &projid, 8); + if (err != 0 && err != ENOENT) { + zfs_znode_dmu_fini(zp); + zfs_znode_hold_exit(zfsvfs, zh); + return (SET_ERROR(err)); + } + } + + zp->z_projid = projid; zp->z_mode = ZTOI(zp)->i_mode = mode; zfs_uid_write(ZTOI(zp), z_uid); zfs_gid_write(ZTOI(zp), z_gid); @@ -1861,6 +1904,7 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) rootzp->z_unlinked = 0; rootzp->z_atime_dirty = 0; rootzp->z_is_sa = USE_SA(version, os); + rootzp->z_pflags = 0; zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); zfsvfs->z_os = os; diff --git a/module/zfs/zpl_file.c b/module/zfs/zpl_file.c index 4dfa0dea4..1c5f5e409 100644 --- a/module/zfs/zpl_file.c +++ b/module/zfs/zpl_file.c @@ -31,7 +31,7 @@ #include <sys/zfs_vfsops.h> #include <sys/zfs_vnops.h> #include <sys/zfs_znode.h> -#include <sys/zpl.h> +#include <sys/zfs_project.h> static int @@ -720,17 +720,14 @@ zpl_fallocate(struct file *filp, int mode, loff_t offset, loff_t len) } #endif /* HAVE_FILE_FALLOCATE */ -/* - * Map zfs file z_pflags (xvattr_t) to linux file attributes. Only file - * attributes common to both Linux and Solaris are mapped. - */ -static int -zpl_ioctl_getflags(struct file *filp, void __user *arg) +#define ZFS_FL_USER_VISIBLE (FS_FL_USER_VISIBLE | ZFS_PROJINHERIT_FL) +#define ZFS_FL_USER_MODIFIABLE (FS_FL_USER_MODIFIABLE | ZFS_PROJINHERIT_FL) + +static uint32_t +__zpl_ioctl_getflags(struct inode *ip) { - struct inode *ip = file_inode(filp); - unsigned int ioctl_flags = 0; uint64_t zfs_flags = ITOZ(ip)->z_pflags; - int error; + uint32_t ioctl_flags = 0; if (zfs_flags & ZFS_IMMUTABLE) ioctl_flags |= FS_IMMUTABLE_FL; @@ -741,11 +738,26 @@ zpl_ioctl_getflags(struct file *filp, void __user *arg) if (zfs_flags & ZFS_NODUMP) ioctl_flags |= FS_NODUMP_FL; - ioctl_flags &= FS_FL_USER_VISIBLE; + if (zfs_flags & ZFS_PROJINHERIT) + ioctl_flags |= ZFS_PROJINHERIT_FL; - error = copy_to_user(arg, &ioctl_flags, sizeof (ioctl_flags)); + return (ioctl_flags & ZFS_FL_USER_VISIBLE); +} - return (error); +/* + * Map zfs file z_pflags (xvattr_t) to linux file attributes. Only file + * attributes common to both Linux and Solaris are mapped. + */ +static int +zpl_ioctl_getflags(struct file *filp, void __user *arg) +{ + uint32_t flags; + int err; + + flags = __zpl_ioctl_getflags(file_inode(filp)); + err = copy_to_user(arg, &flags, sizeof (flags)); + + return (err); } /* @@ -760,24 +772,16 @@ zpl_ioctl_getflags(struct file *filp, void __user *arg) #define fchange(f0, f1, b0, b1) (!((f0) & (b0)) != !((f1) & (b1))) static int -zpl_ioctl_setflags(struct file *filp, void __user *arg) +__zpl_ioctl_setflags(struct inode *ip, uint32_t ioctl_flags, xvattr_t *xva) { - struct inode *ip = file_inode(filp); - uint64_t zfs_flags = ITOZ(ip)->z_pflags; - unsigned int ioctl_flags; - cred_t *cr = CRED(); - xvattr_t xva; - xoptattr_t *xoap; - int error; - fstrans_cookie_t cookie; - - if (copy_from_user(&ioctl_flags, arg, sizeof (ioctl_flags))) - return (-EFAULT); + uint64_t zfs_flags = ITOZ(ip)->z_pflags; + xoptattr_t *xoap; - if ((ioctl_flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | FS_NODUMP_FL))) + if (ioctl_flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | FS_NODUMP_FL | + ZFS_PROJINHERIT_FL)) return (-EOPNOTSUPP); - if ((ioctl_flags & ~(FS_FL_USER_MODIFIABLE))) + if (ioctl_flags & ~ZFS_FL_USER_MODIFIABLE) return (-EACCES); if ((fchange(ioctl_flags, zfs_flags, FS_IMMUTABLE_FL, ZFS_IMMUTABLE) || @@ -788,28 +792,100 @@ zpl_ioctl_setflags(struct file *filp, void __user *arg) if (!zpl_inode_owner_or_capable(ip)) return (-EACCES); - xva_init(&xva); - xoap = xva_getxoptattr(&xva); + xva_init(xva); + xoap = xva_getxoptattr(xva); - XVA_SET_REQ(&xva, XAT_IMMUTABLE); + XVA_SET_REQ(xva, XAT_IMMUTABLE); if (ioctl_flags & FS_IMMUTABLE_FL) xoap->xoa_immutable = B_TRUE; - XVA_SET_REQ(&xva, XAT_APPENDONLY); + XVA_SET_REQ(xva, XAT_APPENDONLY); if (ioctl_flags & FS_APPEND_FL) xoap->xoa_appendonly = B_TRUE; - XVA_SET_REQ(&xva, XAT_NODUMP); + XVA_SET_REQ(xva, XAT_NODUMP); if (ioctl_flags & FS_NODUMP_FL) xoap->xoa_nodump = B_TRUE; + XVA_SET_REQ(xva, XAT_PROJINHERIT); + if (ioctl_flags & ZFS_PROJINHERIT_FL) + xoap->xoa_projinherit = B_TRUE; + + return (0); +} + +static int +zpl_ioctl_setflags(struct file *filp, void __user *arg) +{ + struct inode *ip = file_inode(filp); + uint32_t flags; + cred_t *cr = CRED(); + xvattr_t xva; + int err; + fstrans_cookie_t cookie; + + if (copy_from_user(&flags, arg, sizeof (flags))) + return (-EFAULT); + + err = __zpl_ioctl_setflags(ip, flags, &xva); + if (err) + return (err); + crhold(cr); cookie = spl_fstrans_mark(); - error = -zfs_setattr(ip, (vattr_t *)&xva, 0, cr); + err = -zfs_setattr(ip, (vattr_t *)&xva, 0, cr); spl_fstrans_unmark(cookie); crfree(cr); - return (error); + return (err); +} + +static int +zpl_ioctl_getxattr(struct file *filp, void __user *arg) +{ + zfsxattr_t fsx = { 0 }; + struct inode *ip = file_inode(filp); + int err; + + fsx.fsx_xflags = __zpl_ioctl_getflags(ip); + fsx.fsx_projid = ITOZ(ip)->z_projid; + err = copy_to_user(arg, &fsx, sizeof (fsx)); + + return (err); +} + +static int +zpl_ioctl_setxattr(struct file *filp, void __user *arg) +{ + struct inode *ip = file_inode(filp); + zfsxattr_t fsx; + cred_t *cr = CRED(); + xvattr_t xva; + xoptattr_t *xoap; + int err; + fstrans_cookie_t cookie; + + if (copy_from_user(&fsx, arg, sizeof (fsx))) + return (-EFAULT); + + if (!zpl_is_valid_projid(fsx.fsx_projid)) + return (-EINVAL); + + err = __zpl_ioctl_setflags(ip, fsx.fsx_xflags, &xva); + if (err) + return (err); + + xoap = xva_getxoptattr(&xva); + XVA_SET_REQ(&xva, XAT_PROJID); + xoap->xoa_projid = fsx.fsx_projid; + + crhold(cr); + cookie = spl_fstrans_mark(); + err = -zfs_setattr(ip, (vattr_t *)&xva, 0, cr); + spl_fstrans_unmark(cookie); + crfree(cr); + + return (err); } static long @@ -820,6 +896,10 @@ zpl_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) return (zpl_ioctl_getflags(filp, (void *)arg)); case FS_IOC_SETFLAGS: return (zpl_ioctl_setflags(filp, (void *)arg)); + case ZFS_IOC_FSGETXATTR: + return (zpl_ioctl_getxattr(filp, (void *)arg)); + case ZFS_IOC_FSSETXATTR: + return (zpl_ioctl_setxattr(filp, (void *)arg)); default: return (-ENOTTY); } |