From d2c8103a68e12581527e7d0e2a0c1429e2dcc9d1 Mon Sep 17 00:00:00 2001 From: Tim Chase Date: Mon, 27 Aug 2018 10:28:32 -0400 Subject: [PATCH] Fix problems receiving reallocated dnodes This is a port of 047116ac - Raw sends must be able to decrease nlevels, to the zfs-0.7-stable branch. It includes the various fixes to the problem of receiving incremental streams which include reallocated dnodes in which the number of dnode slots has changed but excludes the parts which are related to raw streams. From 047116ac: Currently, when a raw zfs send file includes a DRR_OBJECT record that would decrease the number of levels of an existing object, the object is reallocated with dmu_object_reclaim() which creates the new dnode using the old object's nlevels. For non-raw sends this doesn't really matter, but raw sends require that nlevels on the receive side match that of the send side so that the checksum-of-MAC tree can be properly maintained. This patch corrects the issue by freeing the object completely before allocating it again in this case. This patch also corrects several issues with dnode_hold_impl() and related functions that prevented dnodes (particularly multi-slot dnodes) from being reallocated properly due to the fact that existing dnodes were not being fully cleaned up when they were freed. This patch adds a test to make sure that zfs recv functions properly with incremental streams containing dnodes of different sizes. This also includes a one-liner fix from loli10K to fix a test failure: https://github.com/zfsonlinux/zfs/pull/7792#discussion_r212769264 Authored-by: Tom Caputi Reviewed by: Matthew Ahrens Reviewed-by: Jorgen Lundman Signed-off-by: Tom Caputi Signed-off-by: Brian Behlendorf Signed-off-by: Tim Chase Ported-by: Tim Chase Closes #6821 Closes #6864 NOTE: This is the first of the port of 3 related patches patches to the zfs-0.7-release branch of ZoL. The other two patches should immediately follow this one. --- cmd/ztest/ztest.c | 25 ++++- include/sys/dnode.h | 6 ++ lib/libzfs/libzfs_sendrecv.c | 1 + module/zfs/dmu_object.c | 1 - module/zfs/dmu_send.c | 51 +++++++++- module/zfs/dnode.c | 84 ++++++++++++++-- module/zfs/dnode_sync.c | 2 + tests/runfiles/linux.run | 2 +- .../tests/functional/rsend/Makefile.am | 3 +- .../rsend/send_realloc_dnode_size.ksh | 98 +++++++++++++++++++ 10 files changed, 258 insertions(+), 15 deletions(-) create mode 100644 tests/zfs-tests/tests/functional/rsend/send_realloc_dnode_size.ksh diff --git a/cmd/ztest/ztest.c b/cmd/ztest/ztest.c index 1a320b03a0..a410eeefba 100644 --- a/cmd/ztest/ztest.c +++ b/cmd/ztest/ztest.c @@ -197,7 +197,8 @@ extern uint64_t metaslab_gang_bang; extern uint64_t metaslab_df_alloc_threshold; extern int metaslab_preload_limit; extern boolean_t zfs_compressed_arc_enabled; -extern int zfs_abd_scatter_enabled; +extern int zfs_abd_scatter_enabled; +extern int dmu_object_alloc_chunk_shift; static ztest_shared_opts_t *ztest_shared_opts; static ztest_shared_opts_t ztest_opts; @@ -310,6 +311,7 @@ static ztest_shared_callstate_t *ztest_shared_callstate; ztest_func_t ztest_dmu_read_write; ztest_func_t ztest_dmu_write_parallel; ztest_func_t ztest_dmu_object_alloc_free; +ztest_func_t ztest_dmu_object_next_chunk; ztest_func_t ztest_dmu_commit_callbacks; ztest_func_t ztest_zap; ztest_func_t ztest_zap_parallel; @@ -357,6 +359,7 @@ ztest_info_t ztest_info[] = { ZTI_INIT(ztest_dmu_read_write, 1, &zopt_always), ZTI_INIT(ztest_dmu_write_parallel, 10, &zopt_always), ZTI_INIT(ztest_dmu_object_alloc_free, 1, &zopt_always), + ZTI_INIT(ztest_dmu_object_next_chunk, 1, &zopt_sometimes), ZTI_INIT(ztest_dmu_commit_callbacks, 1, &zopt_always), ZTI_INIT(ztest_zap, 30, &zopt_always), ZTI_INIT(ztest_zap_parallel, 100, &zopt_always), @@ -3927,6 +3930,26 @@ ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) umem_free(od, size); } +/* + * Rewind the global allocator to verify object allocation backfilling. + */ +void +ztest_dmu_object_next_chunk(ztest_ds_t *zd, uint64_t id) +{ + objset_t *os = zd->zd_os; + int dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift; + uint64_t object; + + /* + * Rewind the global allocator randomly back to a lower object number + * to force backfilling and reclamation of recently freed dnodes. + */ + mutex_enter(&os->os_obj_lock); + object = ztest_random(os->os_obj_next_chunk); + os->os_obj_next_chunk = P2ALIGN(object, dnodes_per_chunk); + mutex_exit(&os->os_obj_lock); +} + #undef OD_ARRAY_SIZE #define OD_ARRAY_SIZE 2 diff --git a/include/sys/dnode.h b/include/sys/dnode.h index c7efe55935..ea7defe19a 100644 --- a/include/sys/dnode.h +++ b/include/sys/dnode.h @@ -360,6 +360,7 @@ int dnode_next_offset(dnode_t *dn, int flags, uint64_t *off, int minlvl, uint64_t blkfill, uint64_t txg); void dnode_evict_dbufs(dnode_t *dn); void dnode_evict_bonus(dnode_t *dn); +void dnode_free_interior_slots(dnode_t *dn); #define DNODE_IS_CACHEABLE(_dn) \ ((_dn)->dn_objset->os_primary_cache == ZFS_CACHE_ALL || \ @@ -453,6 +454,11 @@ typedef struct dnode_stats { * which had already been unlinked in an earlier txg. */ kstat_named_t dnode_hold_free_txg; + /* + * Number of times dnode_free_interior_slots() needed to retry + * acquiring a slot zrl lock due to contention. + */ + kstat_named_t dnode_free_interior_lock_retry; /* * Number of new dnodes allocated by dnode_allocate(). */ diff --git a/lib/libzfs/libzfs_sendrecv.c b/lib/libzfs/libzfs_sendrecv.c index c5acd21a52..cadf16cce0 100644 --- a/lib/libzfs/libzfs_sendrecv.c +++ b/lib/libzfs/libzfs_sendrecv.c @@ -3577,6 +3577,7 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap, } newfs = B_TRUE; + *cp = '/'; } if (flags->verbose) { diff --git a/module/zfs/dmu_object.c b/module/zfs/dmu_object.c index e7412b7509..f53da407fd 100644 --- a/module/zfs/dmu_object.c +++ b/module/zfs/dmu_object.c @@ -275,7 +275,6 @@ dmu_object_reclaim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot, return (err); } - int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx) { diff --git a/module/zfs/dmu_send.c b/module/zfs/dmu_send.c index cdbc1cd1bc..148b5ff842 100644 --- a/module/zfs/dmu_send.c +++ b/module/zfs/dmu_send.c @@ -2156,10 +2156,8 @@ receive_object(struct receive_writer_arg *rwa, struct drr_object *drro, } err = dmu_object_info(rwa->os, drro->drr_object, &doi); - - if (err != 0 && err != ENOENT) + if (err != 0 && err != ENOENT && err != EEXIST) return (SET_ERROR(EINVAL)); - object = err == 0 ? drro->drr_object : DMU_NEW_OBJECT; if (drro->drr_object > rwa->max_object) rwa->max_object = drro->drr_object; @@ -2175,13 +2173,56 @@ receive_object(struct receive_writer_arg *rwa, struct drr_object *drro, nblkptr = deduce_nblkptr(drro->drr_bonustype, drro->drr_bonuslen); + object = drro->drr_object; + if (drro->drr_blksz != doi.doi_data_block_size || - nblkptr < doi.doi_nblkptr) { + nblkptr < doi.doi_nblkptr || + drro->drr_dn_slots != doi.doi_dnodesize >> DNODE_SHIFT) { err = dmu_free_long_range(rwa->os, drro->drr_object, 0, DMU_OBJECT_END); if (err != 0) return (SET_ERROR(EINVAL)); } + } else if (err == EEXIST) { + /* + * The object requested is currently an interior slot of a + * multi-slot dnode. This will be resolved when the next txg + * is synced out, since the send stream will have told us + * to free this slot when we freed the associated dnode + * earlier in the stream. + */ + txg_wait_synced(dmu_objset_pool(rwa->os), 0); + object = drro->drr_object; + } else { + /* object is free and we are about to allocate a new one */ + object = DMU_NEW_OBJECT; + } + + /* + * If this is a multi-slot dnode there is a chance that this + * object will expand into a slot that is already used by + * another object from the previous snapshot. We must free + * these objects before we attempt to allocate the new dnode. + */ + if (drro->drr_dn_slots > 1) { + for (uint64_t slot = drro->drr_object + 1; + slot < drro->drr_object + drro->drr_dn_slots; + slot++) { + dmu_object_info_t slot_doi; + + err = dmu_object_info(rwa->os, slot, &slot_doi); + if (err == ENOENT || err == EEXIST) + continue; + else if (err != 0) + return (err); + + err = dmu_free_long_object(rwa->os, slot); + + if (err != 0) + return (err); + } + + txg_wait_synced(dmu_objset_pool(rwa->os), 0); } tx = dmu_tx_create(rwa->os); @@ -2732,7 +2773,7 @@ receive_read_record(struct receive_arg *ra) * See receive_read_prefetch for an explanation why we're * storing this object in the ignore_obj_list. */ - if (err == ENOENT || + if (err == ENOENT || err == EEXIST || (err == 0 && doi.doi_data_block_size != drro->drr_blksz)) { objlist_insert(&ra->ignore_objlist, drro->drr_object); err = 0; diff --git a/module/zfs/dnode.c b/module/zfs/dnode.c index e05a4d0a55..df6a4872c9 100644 --- a/module/zfs/dnode.c +++ b/module/zfs/dnode.c @@ -55,6 +55,7 @@ dnode_stats_t dnode_stats = { { "dnode_hold_free_overflow", KSTAT_DATA_UINT64 }, { "dnode_hold_free_refcount", KSTAT_DATA_UINT64 }, { "dnode_hold_free_txg", KSTAT_DATA_UINT64 }, + { "dnode_free_interior_lock_retry", KSTAT_DATA_UINT64 }, { "dnode_allocate", KSTAT_DATA_UINT64 }, { "dnode_reallocate", KSTAT_DATA_UINT64 }, { "dnode_buf_evict", KSTAT_DATA_UINT64 }, @@ -516,7 +517,8 @@ dnode_destroy(dnode_t *dn) mutex_exit(&os->os_lock); /* the dnode can no longer move, so we can release the handle */ - zrl_remove(&dn->dn_handle->dnh_zrlock); + if (!zrl_is_locked(&dn->dn_handle->dnh_zrlock)) + zrl_remove(&dn->dn_handle->dnh_zrlock); dn->dn_allocated_txg = 0; dn->dn_free_txg = 0; @@ -662,6 +664,8 @@ dnode_reallocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, DN_BONUS_SIZE(spa_maxdnodesize(dmu_objset_spa(dn->dn_objset)))); dn_slots = dn_slots > 0 ? dn_slots : DNODE_MIN_SLOTS; + + dnode_free_interior_slots(dn); DNODE_STAT_BUMP(dnode_reallocate); /* clean up any unreferenced dbufs */ @@ -1062,19 +1066,73 @@ dnode_set_slots(dnode_children_t *children, int idx, int slots, void *ptr) } static boolean_t -dnode_check_slots(dnode_children_t *children, int idx, int slots, void *ptr) +dnode_check_slots_free(dnode_children_t *children, int idx, int slots) { ASSERT3S(idx + slots, <=, DNODES_PER_BLOCK); for (int i = idx; i < idx + slots; i++) { dnode_handle_t *dnh = &children->dnc_children[i]; - if (dnh->dnh_dnode != ptr) + dnode_t *dn = dnh->dnh_dnode; + + if (dn == DN_SLOT_FREE) { + continue; + } else if (DN_SLOT_IS_PTR(dn)) { + mutex_enter(&dn->dn_mtx); + dmu_object_type_t type = dn->dn_type; + mutex_exit(&dn->dn_mtx); + + if (type != DMU_OT_NONE) + return (B_FALSE); + + continue; + } else { return (B_FALSE); + } + + return (B_FALSE); } return (B_TRUE); } +static void +dnode_reclaim_slots(dnode_children_t *children, int idx, int slots) +{ + ASSERT3S(idx + slots, <=, DNODES_PER_BLOCK); + + for (int i = idx; i < idx + slots; i++) { + dnode_handle_t *dnh = &children->dnc_children[i]; + + ASSERT(zrl_is_locked(&dnh->dnh_zrlock)); + + if (DN_SLOT_IS_PTR(dnh->dnh_dnode)) { + ASSERT3S(dnh->dnh_dnode->dn_type, ==, DMU_OT_NONE); + dnode_destroy(dnh->dnh_dnode); + dnh->dnh_dnode = DN_SLOT_FREE; + } + } +} + +void +dnode_free_interior_slots(dnode_t *dn) +{ + dnode_children_t *children = dmu_buf_get_user(&dn->dn_dbuf->db); + int epb = dn->dn_dbuf->db.db_size >> DNODE_SHIFT; + int idx = (dn->dn_object & (epb - 1)) + 1; + int slots = dn->dn_num_slots - 1; + + if (slots == 0) + return; + + ASSERT3S(idx + slots, <=, DNODES_PER_BLOCK); + + while (!dnode_slots_tryenter(children, idx, slots)) + DNODE_STAT_BUMP(dnode_free_interior_lock_retry); + + dnode_set_slots(children, idx, slots, DN_SLOT_FREE); + dnode_slots_rele(children, idx, slots); +} + void dnode_special_close(dnode_handle_t *dnh) { @@ -1355,7 +1413,7 @@ dnode_hold_impl(objset_t *os, uint64_t object, int flag, int slots, while (dn == DN_SLOT_UNINIT) { dnode_slots_hold(dnc, idx, slots); - if (!dnode_check_slots(dnc, idx, slots, DN_SLOT_FREE)) { + if (!dnode_check_slots_free(dnc, idx, slots)) { DNODE_STAT_BUMP(dnode_hold_free_misses); dnode_slots_rele(dnc, idx, slots); dbuf_rele(db, FTAG); @@ -1368,15 +1426,29 @@ dnode_hold_impl(objset_t *os, uint64_t object, int flag, int slots, continue; } - if (!dnode_check_slots(dnc, idx, slots, DN_SLOT_FREE)) { + if (!dnode_check_slots_free(dnc, idx, slots)) { DNODE_STAT_BUMP(dnode_hold_free_lock_misses); dnode_slots_rele(dnc, idx, slots); dbuf_rele(db, FTAG); return (SET_ERROR(ENOSPC)); } + /* + * Allocated but otherwise free dnodes which would + * be in the interior of a multi-slot dnodes need + * to be freed. Single slot dnodes can be safely + * re-purposed as a performance optimization. + */ + if (slots > 1) + dnode_reclaim_slots(dnc, idx + 1, slots - 1); + dnh = &dnc->dnc_children[idx]; - dn = dnode_create(os, dn_block + idx, db, object, dnh); + if (DN_SLOT_IS_PTR(dnh->dnh_dnode)) { + dn = dnh->dnh_dnode; + } else { + dn = dnode_create(os, dn_block + idx, db, + object, dnh); + } } mutex_enter(&dn->dn_mtx); diff --git a/module/zfs/dnode_sync.c b/module/zfs/dnode_sync.c index 742d962bc2..8d65e38564 100644 --- a/module/zfs/dnode_sync.c +++ b/module/zfs/dnode_sync.c @@ -533,6 +533,7 @@ dnode_sync_free(dnode_t *dn, dmu_tx_t *tx) if (dn->dn_allocated_txg != dn->dn_free_txg) dmu_buf_will_dirty(&dn->dn_dbuf->db, tx); bzero(dn->dn_phys, sizeof (dnode_phys_t) * dn->dn_num_slots); + dnode_free_interior_slots(dn); mutex_enter(&dn->dn_mtx); dn->dn_type = DMU_OT_NONE; @@ -540,6 +541,7 @@ dnode_sync_free(dnode_t *dn, dmu_tx_t *tx) dn->dn_allocated_txg = 0; dn->dn_free_txg = 0; dn->dn_have_spill = B_FALSE; + dn->dn_num_slots = 1; mutex_exit(&dn->dn_mtx); ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); diff --git a/tests/runfiles/linux.run b/tests/runfiles/linux.run index 69e9eb267d..d8fe6f3ab0 100644 --- a/tests/runfiles/linux.run +++ b/tests/runfiles/linux.run @@ -605,7 +605,7 @@ tests = ['rsend_001_pos', 'rsend_002_pos', 'rsend_003_pos', 'rsend_004_pos', 'send-c_lz4_disabled', 'send-c_recv_lz4_disabled', 'send-c_mixed_compression', 'send-c_stream_size_estimate', 'send-cD', 'send-c_embedded_blocks', 'send-c_resume', 'send-cpL_varied_recsize', - 'send-c_recv_dedup', 'send_freeobjects'] + 'send-c_recv_dedup', 'send_freeobjects', 'send_realloc_dnode_size'] tags = ['functional', 'rsend'] [tests/functional/scrub_mirror] diff --git a/tests/zfs-tests/tests/functional/rsend/Makefile.am b/tests/zfs-tests/tests/functional/rsend/Makefile.am index 6b1aa8b320..a2837d1a63 100644 --- a/tests/zfs-tests/tests/functional/rsend/Makefile.am +++ b/tests/zfs-tests/tests/functional/rsend/Makefile.am @@ -36,7 +36,8 @@ dist_pkgdata_SCRIPTS = \ send-c_volume.ksh \ send-c_zstreamdump.ksh \ send-cpL_varied_recsize.ksh \ - send_freeobjects.ksh + send_freeobjects.ksh \ + send_realloc_dnode_size.ksh dist_pkgdata_DATA = \ rsend.cfg \ diff --git a/tests/zfs-tests/tests/functional/rsend/send_realloc_dnode_size.ksh b/tests/zfs-tests/tests/functional/rsend/send_realloc_dnode_size.ksh new file mode 100644 index 0000000000..206763949e --- /dev/null +++ b/tests/zfs-tests/tests/functional/rsend/send_realloc_dnode_size.ksh @@ -0,0 +1,98 @@ +#!/bin/ksh + +# +# This file and its contents are supplied under the terms of the +# Common Development and Distribution License ("CDDL"), version 1.0. +# You may only use this file in accordance with the terms of version +# 1.0 of the CDDL. +# +# A full copy of the text of the CDDL should have accompanied this +# source. A copy of the CDDL is also available via the Internet at +# http://www.illumos.org/license/CDDL. +# + +# +# Copyright (c) 2017 by Lawrence Livermore National Security, LLC. +# + +. $STF_SUITE/include/libtest.shlib +. $STF_SUITE/tests/functional/rsend/rsend.kshlib + +# +# Description: +# Verify incremental receive properly handles objects with changed +# dnode slot count. +# +# Strategy: +# 1. Populate a dataset with 1k byte dnodes and snapshot +# 2. Remove objects, set dnodesize=legacy, and remount dataset so new objects +# get recycled numbers and formerly "interior" dnode slots get assigned +# to new objects +# 3. Remove objects, set dnodesize=2k, and remount dataset so new objects +# overlap with recently recycled and formerly "normal" dnode slots get +# assigned to new objects +# 4. Generate initial and incremental streams +# 5. Verify initial and incremental streams can be received +# + +verify_runnable "both" + +log_assert "Verify incremental receive handles objects with changed dnode size" + +function cleanup +{ + rm -f $BACKDIR/fs-dn-legacy + rm -f $BACKDIR/fs-dn-1k + rm -f $BACKDIR/fs-dn-2k + + if datasetexists $POOL/fs ; then + log_must zfs destroy -rR $POOL/fs + fi + + if datasetexists $POOL/newfs ; then + log_must zfs destroy -rR $POOL/newfs + fi +} + +log_onexit cleanup + +# 1. Populate a dataset with 1k byte dnodes and snapshot +log_must zfs create -o dnodesize=1k $POOL/fs +log_must mk_files 200 262144 0 $POOL/fs +log_must zfs snapshot $POOL/fs@a + +# 2. Remove objects, set dnodesize=legacy, and remount dataset so new objects +# get recycled numbers and formerly "interior" dnode slots get assigned +# to new objects +rm /$POOL/fs/* + +log_must zfs unmount $POOL/fs +log_must zfs set dnodesize=legacy $POOL/fs +log_must zfs mount $POOL/fs + +log_must mk_files 200 262144 0 $POOL/fs +log_must zfs snapshot $POOL/fs@b + +# 3. Remove objects, set dnodesize=2k, and remount dataset so new objects +# overlap with recently recycled and formerly "normal" dnode slots get +# assigned to new objects +rm /$POOL/fs/* + +log_must zfs unmount $POOL/fs +log_must zfs set dnodesize=2k $POOL/fs +log_must zfs mount $POOL/fs + +mk_files 200 262144 0 $POOL/fs +log_must zfs snapshot $POOL/fs@c + +# 4. Generate initial and incremental streams +log_must eval "zfs send $POOL/fs@a > $BACKDIR/fs-dn-1k" +log_must eval "zfs send -i $POOL/fs@a $POOL/fs@b > $BACKDIR/fs-dn-legacy" +log_must eval "zfs send -i $POOL/fs@b $POOL/fs@c > $BACKDIR/fs-dn-2k" + +# 5. Verify initial and incremental streams can be received +log_must eval "zfs recv $POOL/newfs < $BACKDIR/fs-dn-1k" +log_must eval "zfs recv $POOL/newfs < $BACKDIR/fs-dn-legacy" +log_must eval "zfs recv $POOL/newfs < $BACKDIR/fs-dn-2k" + +log_pass "Verify incremental receive handles objects with changed dnode size"