diff options
author | Tom Caputi <[email protected]> | 2018-08-20 14:03:56 -0400 |
---|---|---|
committer | Brian Behlendorf <[email protected]> | 2018-08-20 11:03:56 -0700 |
commit | 149ce888bb14a220036ca3e5b301e3b56451bd71 (patch) | |
tree | 693a3b880bd10701fc13153bf866d4c37dded8a9 /module/zfs | |
parent | c962fd6c4e3836be223af93b0d75d92e8ee98fb4 (diff) |
Fix issues with raw receive_write_byref()
This patch fixes 2 issues with raw, deduplicated send streams. The
first is that datasets who had been completely received earlier in
the stream were not still marked as raw receives. This caused
problems when newly received datasets attempted to fetch raw data
from these datasets without this flag set.
The second problem was that the arc freeze checksum code was not
consistent about which locks needed to be held while performing
its asserts. The proper locking needed to run these asserts is
actually fairly nuanced, since the asserts touch the linked list
of buffers (requiring the header lock), the arc_state (requiring
the b_evict_lock), and the b_freeze_cksum (requiring the
b_freeze_lock). This seems like a large performance sacrifice and
a lot of unneeded complexity to verify that this relatively small
debug feature is working as intended, so this patch simply removes
these asserts instead.
Reviewed by: Matthew Ahrens <[email protected]>
Reviewed-by: Brian Behlendorf <[email protected]>
Signed-off-by: Tom Caputi <[email protected]>
Closes #7701
Diffstat (limited to 'module/zfs')
-rw-r--r-- | module/zfs/arc.c | 45 | ||||
-rw-r--r-- | module/zfs/dbuf.c | 1 | ||||
-rw-r--r-- | module/zfs/dmu.c | 9 | ||||
-rw-r--r-- | module/zfs/dmu_send.c | 34 |
4 files changed, 48 insertions, 41 deletions
diff --git a/module/zfs/arc.c b/module/zfs/arc.c index 7f2929c17..f9436d7c4 100644 --- a/module/zfs/arc.c +++ b/module/zfs/arc.c @@ -1561,6 +1561,9 @@ arc_cksum_free(arc_buf_hdr_t *hdr) static boolean_t arc_hdr_has_uncompressed_buf(arc_buf_hdr_t *hdr) { + ASSERT(hdr->b_l1hdr.b_state == arc_anon || + MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); + for (arc_buf_t *b = hdr->b_l1hdr.b_buf; b != NULL; b = b->b_next) { if (!ARC_BUF_COMPRESSED(b)) { return (B_TRUE); @@ -1584,15 +1587,13 @@ arc_cksum_verify(arc_buf_t *buf) if (!(zfs_flags & ZFS_DEBUG_MODIFY)) return; - if (ARC_BUF_COMPRESSED(buf)) { - ASSERT(hdr->b_l1hdr.b_freeze_cksum == NULL || - arc_hdr_has_uncompressed_buf(hdr)); + if (ARC_BUF_COMPRESSED(buf)) return; - } ASSERT(HDR_HAS_L1HDR(hdr)); mutex_enter(&hdr->b_l1hdr.b_freeze_lock); + if (hdr->b_l1hdr.b_freeze_cksum == NULL || HDR_IO_ERROR(hdr)) { mutex_exit(&hdr->b_l1hdr.b_freeze_lock); return; @@ -1650,11 +1651,7 @@ arc_cksum_compute(arc_buf_t *buf) ASSERT(HDR_HAS_L1HDR(hdr)); mutex_enter(&buf->b_hdr->b_l1hdr.b_freeze_lock); - if (hdr->b_l1hdr.b_freeze_cksum != NULL) { - ASSERT(arc_hdr_has_uncompressed_buf(hdr)); - mutex_exit(&hdr->b_l1hdr.b_freeze_lock); - return; - } else if (ARC_BUF_COMPRESSED(buf)) { + if (hdr->b_l1hdr.b_freeze_cksum != NULL || ARC_BUF_COMPRESSED(buf)) { mutex_exit(&hdr->b_l1hdr.b_freeze_lock); return; } @@ -1746,14 +1743,10 @@ arc_buf_thaw(arc_buf_t *buf) arc_cksum_verify(buf); /* - * Compressed buffers do not manipulate the b_freeze_cksum or - * allocate b_thawed. + * Compressed buffers do not manipulate the b_freeze_cksum. */ - if (ARC_BUF_COMPRESSED(buf)) { - ASSERT(hdr->b_l1hdr.b_freeze_cksum == NULL || - arc_hdr_has_uncompressed_buf(hdr)); + if (ARC_BUF_COMPRESSED(buf)) return; - } ASSERT(HDR_HAS_L1HDR(hdr)); arc_cksum_free(hdr); @@ -1763,26 +1756,14 @@ arc_buf_thaw(arc_buf_t *buf) void arc_buf_freeze(arc_buf_t *buf) { - arc_buf_hdr_t *hdr = buf->b_hdr; - kmutex_t *hash_lock; - if (!(zfs_flags & ZFS_DEBUG_MODIFY)) return; - if (ARC_BUF_COMPRESSED(buf)) { - ASSERT(hdr->b_l1hdr.b_freeze_cksum == NULL || - arc_hdr_has_uncompressed_buf(hdr)); + if (ARC_BUF_COMPRESSED(buf)) return; - } - - hash_lock = HDR_LOCK(hdr); - mutex_enter(hash_lock); - ASSERT(HDR_HAS_L1HDR(hdr)); - ASSERT(hdr->b_l1hdr.b_freeze_cksum != NULL || - hdr->b_l1hdr.b_state == arc_anon); + ASSERT(HDR_HAS_L1HDR(buf->b_hdr)); arc_cksum_compute(buf); - mutex_exit(hash_lock); } /* @@ -1901,7 +1882,7 @@ arc_hdr_authenticate(arc_buf_hdr_t *hdr, spa_t *spa, uint64_t dsobj) void *tmpbuf = NULL; abd_t *abd = hdr->b_l1hdr.b_pabd; - ASSERT(HDR_LOCK(hdr) == NULL || MUTEX_HELD(HDR_LOCK(hdr))); + ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); ASSERT(HDR_AUTHENTICATED(hdr)); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); @@ -1971,7 +1952,7 @@ arc_hdr_decrypt(arc_buf_hdr_t *hdr, spa_t *spa, const zbookmark_phys_t *zb) boolean_t no_crypt = B_FALSE; boolean_t bswap = (hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS); - ASSERT(HDR_LOCK(hdr) == NULL || MUTEX_HELD(HDR_LOCK(hdr))); + ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); ASSERT(HDR_ENCRYPTED(hdr)); arc_hdr_alloc_abd(hdr, B_FALSE); @@ -2090,7 +2071,7 @@ arc_buf_untransform_in_place(arc_buf_t *buf, kmutex_t *hash_lock) ASSERT(HDR_ENCRYPTED(hdr)); ASSERT3U(hdr->b_crypt_hdr.b_ot, ==, DMU_OT_DNODE); - ASSERT(HDR_LOCK(hdr) == NULL || MUTEX_HELD(HDR_LOCK(hdr))); + ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); zio_crypt_copy_dnode_bonus(hdr->b_l1hdr.b_pabd, buf->b_data, diff --git a/module/zfs/dbuf.c b/module/zfs/dbuf.c index f0bee6c2b..1af0d8010 100644 --- a/module/zfs/dbuf.c +++ b/module/zfs/dbuf.c @@ -1267,7 +1267,6 @@ dbuf_read_verify_dnode_crypt(dmu_buf_impl_t *db, uint32_t flags) !DMU_OT_IS_ENCRYPTED(dn->dn_bonustype)))) err = 0; - DB_DNODE_EXIT(db); return (err); diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c index 9d00098c8..88b574d4a 100644 --- a/module/zfs/dmu.c +++ b/module/zfs/dmu.c @@ -1622,6 +1622,7 @@ dmu_copy_from_buf(objset_t *os, uint64_t object, uint64_t offset, dmu_buf_t *dst_handle; dmu_buf_impl_t *dstdb; dmu_buf_impl_t *srcdb = (dmu_buf_impl_t *)handle; + dmu_object_type_t type; arc_buf_t *abuf; uint64_t datalen; boolean_t byteorder; @@ -1637,11 +1638,15 @@ dmu_copy_from_buf(objset_t *os, uint64_t object, uint64_t offset, dstdb = (dmu_buf_impl_t *)dst_handle; datalen = arc_buf_size(srcdb->db_buf); + DB_DNODE_ENTER(dstdb); + type = DB_DNODE(dstdb)->dn_type; + DB_DNODE_EXIT(dstdb); + /* allocated an arc buffer that matches the type of srcdb->db_buf */ if (arc_is_encrypted(srcdb->db_buf)) { arc_get_raw_params(srcdb->db_buf, &byteorder, salt, iv, mac); abuf = arc_loan_raw_buf(os->os_spa, dmu_objset_id(os), - byteorder, salt, iv, mac, DB_DNODE(dstdb)->dn_type, + byteorder, salt, iv, mac, type, datalen, arc_buf_lsize(srcdb->db_buf), arc_get_compression(srcdb->db_buf)); } else { @@ -1649,7 +1654,7 @@ dmu_copy_from_buf(objset_t *os, uint64_t object, uint64_t offset, ASSERT3U(arc_get_compression(srcdb->db_buf), ==, ZIO_COMPRESS_OFF); abuf = arc_loan_buf(os->os_spa, - DMU_OT_IS_METADATA(DB_DNODE(dstdb)->dn_type), datalen); + DMU_OT_IS_METADATA(type), datalen); } ASSERT3U(datalen, ==, arc_buf_size(abuf)); diff --git a/module/zfs/dmu_send.c b/module/zfs/dmu_send.c index 61d4ebd4f..1f4d3a104 100644 --- a/module/zfs/dmu_send.c +++ b/module/zfs/dmu_send.c @@ -2343,9 +2343,14 @@ free_guid_map_onexit(void *arg) guid_map_entry_t *gmep; while ((gmep = avl_destroy_nodes(ca, &cookie)) != NULL) { - dsl_dataset_long_rele(gmep->gme_ds, gmep); - dsl_dataset_rele_flags(gmep->gme_ds, - (gmep->raw) ? 0 : DS_HOLD_FLAG_DECRYPT, gmep); + ds_hold_flags_t dsflags = DS_HOLD_FLAG_DECRYPT; + + if (gmep->raw) { + gmep->gme_ds->ds_objset->os_raw_receive = B_FALSE; + dsflags &= ~DS_HOLD_FLAG_DECRYPT; + } + + dsl_dataset_disown(gmep->gme_ds, dsflags, gmep); kmem_free(gmep, sizeof (guid_map_entry_t)); } avl_destroy(ca); @@ -4237,6 +4242,7 @@ add_ds_to_guidmap(const char *name, avl_tree_t *guid_map, uint64_t snapobj, dsl_pool_t *dp; dsl_dataset_t *snapds; guid_map_entry_t *gmep; + objset_t *os; ds_hold_flags_t dsflags = (raw) ? 0 : DS_HOLD_FLAG_DECRYPT; int err; @@ -4246,13 +4252,29 @@ add_ds_to_guidmap(const char *name, avl_tree_t *guid_map, uint64_t snapobj, if (err != 0) return (err); gmep = kmem_alloc(sizeof (*gmep), KM_SLEEP); - err = dsl_dataset_hold_obj_flags(dp, snapobj, dsflags, gmep, &snapds); + err = dsl_dataset_own_obj(dp, snapobj, dsflags, gmep, &snapds); if (err == 0) { - gmep->guid = dsl_dataset_phys(snapds)->ds_guid; + /* + * If this is a deduplicated raw send stream, we need + * to make sure that we can still read raw blocks from + * earlier datasets in the stream, so we set the + * os_raw_receive flag now. + */ + if (raw) { + err = dmu_objset_from_ds(snapds, &os); + if (err != 0) { + dsl_dataset_disown(snapds, dsflags, FTAG); + dsl_pool_rele(dp, FTAG); + kmem_free(gmep, sizeof (*gmep)); + return (err); + } + os->os_raw_receive = B_TRUE; + } + gmep->raw = raw; + gmep->guid = dsl_dataset_phys(snapds)->ds_guid; gmep->gme_ds = snapds; avl_add(guid_map, gmep); - dsl_dataset_long_hold(snapds, gmep); } else { kmem_free(gmep, sizeof (*gmep)); } |