X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Frecovery.c;h=ca92fe84c248a25f5863046bdc038f265274e20d;hb=70f2681838973f20a918b518f82d382b3556963d;hp=7e4400cc02a9249381e5c03b6b7ff7776c4ee910;hpb=931ed5a709c2afa239cbae2e13bc22f13e99713c;p=bcachefs-tools-debian diff --git a/libbcachefs/recovery.c b/libbcachefs/recovery.c index 7e4400c..ca92fe8 100644 --- a/libbcachefs/recovery.c +++ b/libbcachefs/recovery.c @@ -16,6 +16,7 @@ #include "journal_io.h" #include "journal_reclaim.h" #include "journal_seq_blacklist.h" +#include "lru.h" #include "move.h" #include "quota.h" #include "recovery.h" @@ -94,6 +95,24 @@ size_t bch2_journal_key_search(struct journal_keys *journal_keys, return l; } +struct bkey_i *bch2_journal_keys_peek(struct bch_fs *c, enum btree_id btree_id, + unsigned level, struct bpos pos) +{ + struct journal_keys *keys = &c->journal_keys; + struct journal_key *end = keys->d + keys->nr; + struct journal_key *k = keys->d + + bch2_journal_key_search(keys, btree_id, level, pos); + + while (k < end && k->overwritten) + k++; + + if (k < end && + k->btree_id == btree_id && + k->level == level) + return k->k; + return NULL; +} + static void journal_iter_fix(struct bch_fs *c, struct journal_iter *iter, unsigned idx) { struct bkey_i *n = iter->keys->d[idx].k; @@ -544,8 +563,9 @@ static int bch2_journal_replay(struct bch_fs *c) ret = bch2_trans_do(c, NULL, NULL, BTREE_INSERT_LAZY_RW| BTREE_INSERT_NOFAIL| - BTREE_INSERT_JOURNAL_RESERVED| - (!k->allocated ? BTREE_INSERT_JOURNAL_REPLAY : 0), + (!k->allocated + ? BTREE_INSERT_JOURNAL_REPLAY|JOURNAL_WATERMARK_reserved + : 0), bch2_journal_replay_key(&trans, k)); if (ret) { bch_err(c, "journal replay: error %d while replaying key at btree %s level %u", @@ -560,6 +580,9 @@ static int bch2_journal_replay(struct bch_fs *c) bch2_journal_set_replay_done(j); bch2_journal_flush_all_pins(j); ret = bch2_journal_error(j); + + if (keys->nr && !ret) + bch2_journal_log_msg(&c->journal, "journal replay finished"); err: kvfree(keys_sorted); return ret; @@ -742,6 +765,8 @@ static int verify_superblock_clean(struct bch_fs *c, { unsigned i; struct bch_sb_field_clean *clean = *cleanp; + struct printbuf buf1 = PRINTBUF; + struct printbuf buf2 = PRINTBUF; int ret = 0; if (mustfix_fsck_err_on(j->seq != clean->journal_seq, c, @@ -754,7 +779,6 @@ static int verify_superblock_clean(struct bch_fs *c, } for (i = 0; i < BTREE_ID_NR; i++) { - char buf1[200], buf2[200]; struct bkey_i *k1, *k2; unsigned l1 = 0, l2 = 0; @@ -764,6 +788,19 @@ static int verify_superblock_clean(struct bch_fs *c, if (!k1 && !k2) continue; + printbuf_reset(&buf1); + printbuf_reset(&buf2); + + if (k1) + bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(k1)); + else + pr_buf(&buf1, "(none)"); + + if (k2) + bch2_bkey_val_to_text(&buf2, c, bkey_i_to_s_c(k2)); + else + pr_buf(&buf2, "(none)"); + mustfix_fsck_err_on(!k1 || !k2 || IS_ERR(k1) || IS_ERR(k2) || @@ -773,10 +810,12 @@ static int verify_superblock_clean(struct bch_fs *c, "superblock btree root %u doesn't match journal after clean shutdown\n" "sb: l=%u %s\n" "journal: l=%u %s\n", i, - l1, (bch2_bkey_val_to_text(&PBUF(buf1), c, bkey_i_to_s_c(k1)), buf1), - l2, (bch2_bkey_val_to_text(&PBUF(buf2), c, bkey_i_to_s_c(k2)), buf2)); + l1, buf1.buf, + l2, buf2.buf); } fsck_err: + printbuf_exit(&buf2); + printbuf_exit(&buf1); return ret; } @@ -803,7 +842,7 @@ static struct bch_sb_field_clean *read_superblock_clean(struct bch_fs *c) return ERR_PTR(-ENOMEM); } - ret = bch2_sb_clean_validate(c, clean, READ); + ret = bch2_sb_clean_validate_late(c, clean, READ); if (ret) { mutex_unlock(&c->sb_lock); return ERR_PTR(ret); @@ -990,8 +1029,8 @@ int bch2_fs_recovery(struct bch_fs *c) bch_info(c, "filesystem version is prior to subvol_dirent - upgrading"); c->opts.version_upgrade = true; c->opts.fsck = true; - } else if (c->sb.version < bcachefs_metadata_version_inode_v2) { - bch_info(c, "filesystem version is prior to inode_v2 - upgrading"); + } else if (c->sb.version < bcachefs_metadata_version_alloc_v4) { + bch_info(c, "filesystem version is prior to alloc_v4 - upgrading"); c->opts.version_upgrade = true; } } @@ -1054,6 +1093,9 @@ use_clean: blacklist_seq = journal_seq = le64_to_cpu(clean->journal_seq) + 1; } + if (c->opts.read_journal_only) + goto out; + if (c->opts.reconstruct_alloc) { c->sb.compat &= ~(1ULL << BCH_COMPAT_alloc_info); drop_alloc_keys(&c->journal_keys); @@ -1097,7 +1139,7 @@ use_clean: err = "error reading allocation information"; down_read(&c->gc_lock); - ret = bch2_alloc_read(c, false, false); + ret = bch2_alloc_read(c); up_read(&c->gc_lock); if (ret) @@ -1111,8 +1153,6 @@ use_clean: goto err; bch_verbose(c, "stripes_read done"); - set_bit(BCH_FS_ALLOC_READ_DONE, &c->flags); - /* * If we're not running fsck, this ensures bch2_fsck_err() calls are * instead interpreted as bch2_inconsistent_err() calls: @@ -1126,18 +1166,32 @@ use_clean: test_bit(BCH_FS_REBUILD_REPLICAS, &c->flags)) { bool metadata_only = c->opts.norecovery; - bch_info(c, "starting mark and sweep"); - err = "error in mark and sweep"; + bch_info(c, "checking allocations"); + err = "error checking allocations"; ret = bch2_gc(c, true, metadata_only); if (ret) goto err; - bch_verbose(c, "mark and sweep done"); + bch_verbose(c, "done checking allocations"); + } + + if (c->opts.fsck) { + bch_info(c, "checking need_discard and freespace btrees"); + err = "error checking need_discard and freespace btrees"; + ret = bch2_check_alloc_info(c, true); + if (ret) + goto err; + + ret = bch2_check_lrus(c, true); + if (ret) + goto err; + bch_verbose(c, "done checking need_discard and freespace btrees"); } bch2_stripes_heap_start(c); clear_bit(BCH_FS_REBUILD_REPLICAS, &c->flags); set_bit(BCH_FS_INITIAL_GC_DONE, &c->flags); + set_bit(BCH_FS_MAY_GO_RW, &c->flags); /* * Skip past versions that might have possibly been used (as nonces), @@ -1157,6 +1211,11 @@ use_clean: if (c->opts.verbose || !c->sb.clean) bch_info(c, "journal replay done"); + err = "error initializing freespace"; + ret = bch2_fs_freespace_init(c); + if (ret) + goto err; + if (c->sb.version < bcachefs_metadata_version_snapshot_2) { bch2_fs_lazy_rw(c); @@ -1297,8 +1356,8 @@ int bch2_fs_initialize(struct bch_fs *c) } mutex_unlock(&c->sb_lock); - set_bit(BCH_FS_ALLOC_READ_DONE, &c->flags); set_bit(BCH_FS_INITIAL_GC_DONE, &c->flags); + set_bit(BCH_FS_MAY_GO_RW, &c->flags); set_bit(BCH_FS_FSCK_DONE, &c->flags); for (i = 0; i < BTREE_ID_NR; i++) @@ -1329,6 +1388,7 @@ int bch2_fs_initialize(struct bch_fs *c) * Write out the superblock and journal buckets, now that we can do * btree updates */ + bch_verbose(c, "marking superblocks"); err = "error marking superblock and journal"; for_each_member_device(ca, c, i) { ret = bch2_trans_mark_dev_sb(c, ca); @@ -1340,6 +1400,12 @@ int bch2_fs_initialize(struct bch_fs *c) ca->new_fs_bucket_idx = 0; } + bch_verbose(c, "initializing freespace"); + err = "error initializing freespace"; + ret = bch2_fs_freespace_init(c); + if (ret) + goto err; + err = "error creating root snapshot node"; ret = bch2_fs_initialize_subvolumes(c); if (ret)