X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Frecovery.c;h=f5946b4dbce2b847dde77a1bb769d00957e3b1e5;hb=da6a35689518599b381c285cd9505ab8d58f7c73;hp=80acbc7d3efa90847fb32e95a6539742b625488d;hpb=7923c2ee0616ef35d984e292dd9e799babe96662;p=bcachefs-tools-debian diff --git a/libbcachefs/recovery.c b/libbcachefs/recovery.c index 80acbc7..f5946b4 100644 --- a/libbcachefs/recovery.c +++ b/libbcachefs/recovery.c @@ -1,6 +1,7 @@ // SPDX-License-Identifier: GPL-2.0 #include "bcachefs.h" +#include "backpointers.h" #include "bkey_buf.h" #include "alloc_background.h" #include "btree_gc.h" @@ -10,6 +11,7 @@ #include "buckets.h" #include "dirent.h" #include "ec.h" +#include "errcode.h" #include "error.h" #include "fs-common.h" #include "fsck.h" @@ -86,9 +88,9 @@ static inline struct journal_key *idx_to_key(struct journal_keys *keys, size_t i return keys->d + idx_to_pos(keys, idx); } -size_t bch2_journal_key_search(struct journal_keys *keys, - enum btree_id id, unsigned level, - struct bpos pos) +static size_t __bch2_journal_key_search(struct journal_keys *keys, + enum btree_id id, unsigned level, + struct bpos pos) { size_t l = 0, r = keys->nr, m; @@ -106,26 +108,41 @@ size_t bch2_journal_key_search(struct journal_keys *keys, BUG_ON(l && __journal_key_cmp(id, level, pos, idx_to_key(keys, l - 1)) <= 0); - return idx_to_pos(keys, l); + return l; +} + +static size_t bch2_journal_key_search(struct journal_keys *keys, + enum btree_id id, unsigned level, + struct bpos pos) +{ + return idx_to_pos(keys, __bch2_journal_key_search(keys, id, level, pos)); } struct bkey_i *bch2_journal_keys_peek_upto(struct bch_fs *c, enum btree_id btree_id, unsigned level, struct bpos pos, - struct bpos end_pos) + struct bpos end_pos, size_t *idx) { struct journal_keys *keys = &c->journal_keys; - size_t idx = bch2_journal_key_search(keys, btree_id, level, pos); - - while (idx < keys->size && - keys->d[idx].btree_id == btree_id && - keys->d[idx].level == level && - bpos_cmp(keys->d[idx].k->k.p, end_pos) <= 0) { - if (!keys->d[idx].overwritten) - return keys->d[idx].k; - - idx++; - if (idx == keys->gap) - idx += keys->size - keys->nr; + unsigned iters = 0; + struct journal_key *k; +search: + if (!*idx) + *idx = __bch2_journal_key_search(keys, btree_id, level, pos); + + while ((k = *idx < keys->nr ? idx_to_key(keys, *idx) : NULL)) { + if (__journal_key_cmp(btree_id, level, end_pos, k) < 0) + return NULL; + + if (__journal_key_cmp(btree_id, level, pos, k) <= 0 && + !k->overwritten) + return k->k; + + (*idx)++; + iters++; + if (iters == 10) { + *idx = 0; + goto search; + } } return NULL; @@ -134,28 +151,26 @@ struct bkey_i *bch2_journal_keys_peek_upto(struct bch_fs *c, enum btree_id btree struct bkey_i *bch2_journal_keys_peek_slot(struct bch_fs *c, enum btree_id btree_id, unsigned level, struct bpos pos) { - return bch2_journal_keys_peek_upto(c, btree_id, level, pos, pos); + size_t idx = 0; + + return bch2_journal_keys_peek_upto(c, btree_id, level, pos, pos, &idx); } static void journal_iters_fix(struct bch_fs *c) { struct journal_keys *keys = &c->journal_keys; /* The key we just inserted is immediately before the gap: */ - struct journal_key *n = &keys->d[keys->gap - 1]; size_t gap_end = keys->gap + (keys->size - keys->nr); struct btree_and_journal_iter *iter; /* - * If an iterator points one after the key we just inserted, - * and the key we just inserted compares >= the iterator's position, - * decrement the iterator so it points at the key we just inserted: + * If an iterator points one after the key we just inserted, decrement + * the iterator so it points at the key we just inserted - if the + * decrement was unnecessary, bch2_btree_and_journal_iter_peek() will + * handle that: */ list_for_each_entry(iter, &c->journal_iters, journal.list) - if (iter->journal.idx == gap_end && - iter->last && - iter->b->c.btree_id == n->btree_id && - iter->b->c.level == n->level && - bpos_cmp(n->k->k.p, iter->unpacked.p) >= 0) + if (iter->journal.idx == gap_end) iter->journal.idx = keys->gap - 1; } @@ -207,10 +222,9 @@ int bch2_journal_key_insert_take(struct bch_fs *c, enum btree_id id, struct journal_keys new_keys = { .nr = keys->nr, .size = max_t(size_t, keys->size, 8) * 2, - .journal_seq_base = keys->journal_seq_base, }; - new_keys.d = kvmalloc(sizeof(new_keys.d[0]) * new_keys.size, GFP_KERNEL); + new_keys.d = kvmalloc_array(new_keys.size, sizeof(new_keys.d[0]), GFP_KERNEL); if (!new_keys.d) { bch_err(c, "%s: error allocating new key array (size %zu)", __func__, new_keys.size); @@ -281,34 +295,34 @@ void bch2_journal_key_overwritten(struct bch_fs *c, enum btree_id btree, if (idx < keys->size && keys->d[idx].btree_id == btree && keys->d[idx].level == level && - !bpos_cmp(keys->d[idx].k->k.p, pos)) + bpos_eq(keys->d[idx].k->k.p, pos)) keys->d[idx].overwritten = true; } -static struct bkey_i *bch2_journal_iter_peek(struct journal_iter *iter) +static void bch2_journal_iter_advance(struct journal_iter *iter) +{ + if (iter->idx < iter->keys->size) { + iter->idx++; + if (iter->idx == iter->keys->gap) + iter->idx += iter->keys->size - iter->keys->nr; + } +} + +struct bkey_s_c bch2_journal_iter_peek(struct journal_iter *iter) { struct journal_key *k = iter->keys->d + iter->idx; - while (k < iter->keys->d + iter->keys->nr && + while (k < iter->keys->d + iter->keys->size && k->btree_id == iter->btree_id && k->level == iter->level) { if (!k->overwritten) - return k->k; + return bkey_i_to_s_c(k->k); - iter->idx++; + bch2_journal_iter_advance(iter); k = iter->keys->d + iter->idx; } - return NULL; -} - -static void bch2_journal_iter_advance(struct journal_iter *iter) -{ - if (iter->idx < iter->keys->size) { - iter->idx++; - if (iter->idx == iter->keys->gap) - iter->idx += iter->keys->size - iter->keys->nr; - } + return bkey_s_c_null; } static void bch2_journal_iter_exit(struct journal_iter *iter) @@ -340,71 +354,49 @@ static void bch2_journal_iter_advance_btree(struct btree_and_journal_iter *iter) void bch2_btree_and_journal_iter_advance(struct btree_and_journal_iter *iter) { - switch (iter->last) { - case none: - break; - case btree: - bch2_journal_iter_advance_btree(iter); - break; - case journal: - bch2_journal_iter_advance(&iter->journal); - break; - } - - iter->last = none; + if (bpos_eq(iter->pos, SPOS_MAX)) + iter->at_end = true; + else + iter->pos = bpos_successor(iter->pos); } struct bkey_s_c bch2_btree_and_journal_iter_peek(struct btree_and_journal_iter *iter) { - struct bkey_s_c ret; - - while (1) { - struct bkey_s_c btree_k = - bch2_journal_iter_peek_btree(iter); - struct bkey_s_c journal_k = - bkey_i_to_s_c(bch2_journal_iter_peek(&iter->journal)); + struct bkey_s_c btree_k, journal_k, ret; +again: + if (iter->at_end) + return bkey_s_c_null; - if (btree_k.k && journal_k.k) { - int cmp = bpos_cmp(btree_k.k->p, journal_k.k->p); + while ((btree_k = bch2_journal_iter_peek_btree(iter)).k && + bpos_lt(btree_k.k->p, iter->pos)) + bch2_journal_iter_advance_btree(iter); - if (!cmp) - bch2_journal_iter_advance_btree(iter); + while ((journal_k = bch2_journal_iter_peek(&iter->journal)).k && + bpos_lt(journal_k.k->p, iter->pos)) + bch2_journal_iter_advance(&iter->journal); - iter->last = cmp < 0 ? btree : journal; - } else if (btree_k.k) { - iter->last = btree; - } else if (journal_k.k) { - iter->last = journal; - } else { - iter->last = none; - return bkey_s_c_null; - } + ret = journal_k.k && + (!btree_k.k || bpos_le(journal_k.k->p, btree_k.k->p)) + ? journal_k + : btree_k; - ret = iter->last == journal ? journal_k : btree_k; + if (ret.k && iter->b && bpos_gt(ret.k->p, iter->b->data->max_key)) + ret = bkey_s_c_null; - if (iter->b && - bpos_cmp(ret.k->p, iter->b->data->max_key) > 0) { - iter->journal.idx = iter->journal.keys->nr; - iter->last = none; - return bkey_s_c_null; + if (ret.k) { + iter->pos = ret.k->p; + if (bkey_deleted(ret.k)) { + bch2_btree_and_journal_iter_advance(iter); + goto again; } - - if (!bkey_deleted(ret.k)) - break; - - bch2_btree_and_journal_iter_advance(iter); + } else { + iter->pos = SPOS_MAX; + iter->at_end = true; } return ret; } -struct bkey_s_c bch2_btree_and_journal_iter_next(struct btree_and_journal_iter *iter) -{ - bch2_btree_and_journal_iter_advance(iter); - - return bch2_btree_and_journal_iter_peek(iter); -} - void bch2_btree_and_journal_iter_exit(struct btree_and_journal_iter *iter) { bch2_journal_iter_exit(&iter->journal); @@ -422,6 +414,8 @@ void __bch2_btree_and_journal_iter_init_node_iter(struct btree_and_journal_iter iter->node_iter = node_iter; bch2_journal_iter_init(c, &iter->journal, b->c.btree_id, b->c.level, pos); INIT_LIST_HEAD(&iter->journal.list); + iter->pos = b->data->min_key; + iter->at_end = false; } /* @@ -498,9 +492,6 @@ static int journal_keys_sort(struct bch_fs *c) if (!i || i->ignore) continue; - if (!keys->journal_seq_base) - keys->journal_seq_base = le64_to_cpu(i->j.seq); - for_each_jset_key(k, _n, entry, &i->j) nr_keys++; } @@ -510,7 +501,7 @@ static int journal_keys_sort(struct bch_fs *c) keys->size = roundup_pow_of_two(nr_keys); - keys->d = kvmalloc(sizeof(keys->d[0]) * keys->size, GFP_KERNEL); + keys->d = kvmalloc_array(keys->size, sizeof(keys->d[0]), GFP_KERNEL); if (!keys->d) return -ENOMEM; @@ -520,15 +511,12 @@ static int journal_keys_sort(struct bch_fs *c) if (!i || i->ignore) continue; - BUG_ON(le64_to_cpu(i->j.seq) - keys->journal_seq_base > U32_MAX); - for_each_jset_key(k, _n, entry, &i->j) keys->d[keys->nr++] = (struct journal_key) { .btree_id = entry->btree_id, .level = entry->level, .k = k, - .journal_seq = le64_to_cpu(i->j.seq) - - keys->journal_seq_base, + .journal_seq = le64_to_cpu(i->j.seq), .journal_offset = k->_data - i->j._data, }; } @@ -540,7 +528,7 @@ static int journal_keys_sort(struct bch_fs *c) while (src + 1 < keys->d + keys->nr && src[0].btree_id == src[1].btree_id && src[0].level == src[1].level && - !bpos_cmp(src[0].k->k.p, src[1].k->k.p)) + bpos_eq(src[0].k->k.p, src[1].k->k.p)) src++; *dst++ = *src++; @@ -556,7 +544,8 @@ static int journal_keys_sort(struct bch_fs *c) static void replay_now_at(struct journal *j, u64 seq) { BUG_ON(seq < j->replay_journal_seq); - BUG_ON(seq > j->replay_journal_seq_end); + + seq = min(seq, j->replay_journal_seq_end); while (j->replay_journal_seq < seq) bch2_journal_pin_put(j, j->replay_journal_seq++); @@ -599,7 +588,7 @@ static int journal_sort_seq_cmp(const void *_l, const void *_r) return cmp_int(l->journal_seq, r->journal_seq); } -static int bch2_journal_replay(struct bch_fs *c) +static int bch2_journal_replay(struct bch_fs *c, u64 start_seq, u64 end_seq) { struct journal_keys *keys = &c->journal_keys; struct journal_key **keys_sorted, *k; @@ -621,16 +610,19 @@ static int bch2_journal_replay(struct bch_fs *c) sizeof(keys_sorted[0]), journal_sort_seq_cmp, NULL); - if (keys->nr) - replay_now_at(j, keys->journal_seq_base); + if (keys->nr) { + ret = bch2_fs_log_msg(c, "Starting journal replay (%zu keys in entries %llu-%llu)", + keys->nr, start_seq, end_seq); + if (ret) + goto err; + } for (i = 0; i < keys->nr; i++) { k = keys_sorted[i]; cond_resched(); - if (!k->allocated) - replay_now_at(j, keys->journal_seq_base + k->journal_seq); + replay_now_at(j, k->journal_seq); ret = bch2_trans_do(c, NULL, NULL, BTREE_INSERT_LAZY_RW| @@ -640,8 +632,8 @@ static int bch2_journal_replay(struct bch_fs *c) : 0), bch2_journal_replay_key(&trans, k)); if (ret) { - bch_err(c, "journal replay: error %d while replaying key at btree %s level %u", - ret, bch2_btree_ids[k->btree_id], k->level); + bch_err(c, "journal replay: error while replaying key at btree %s level %u: %s", + bch2_btree_ids[k->btree_id], k->level, bch2_err_str(ret)); goto err; } } @@ -654,7 +646,7 @@ static int bch2_journal_replay(struct bch_fs *c) ret = bch2_journal_error(j); if (keys->nr && !ret) - bch2_journal_log_msg(&c->journal, "journal replay finished"); + bch2_fs_log_msg(c, "journal replay finished"); err: kvfree(keys_sorted); return ret; @@ -868,12 +860,12 @@ static int verify_superblock_clean(struct bch_fs *c, if (k1) bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(k1)); else - pr_buf(&buf1, "(none)"); + prt_printf(&buf1, "(none)"); if (k2) bch2_bkey_val_to_text(&buf2, c, bkey_i_to_s_c(k2)); else - pr_buf(&buf2, "(none)"); + prt_printf(&buf2, "(none)"); mustfix_fsck_err_on(!k1 || !k2 || IS_ERR(k1) || @@ -930,6 +922,20 @@ fsck_err: return ERR_PTR(ret); } +static bool btree_id_is_alloc(enum btree_id id) +{ + switch (id) { + case BTREE_ID_alloc: + case BTREE_ID_backpointers: + case BTREE_ID_need_discard: + case BTREE_ID_freespace: + case BTREE_ID_bucket_gens: + return true; + default: + return false; + } +} + static int read_btree_roots(struct bch_fs *c) { unsigned i; @@ -941,14 +947,14 @@ static int read_btree_roots(struct bch_fs *c) if (!r->alive) continue; - if (i == BTREE_ID_alloc && + if (btree_id_is_alloc(i) && c->opts.reconstruct_alloc) { c->sb.compat &= ~(1ULL << BCH_COMPAT_alloc_info); continue; } if (r->error) { - __fsck_err(c, i == BTREE_ID_alloc + __fsck_err(c, btree_id_is_alloc(i) ? FSCK_CAN_IGNORE : 0, "invalid btree root %s", bch2_btree_ids[i]); @@ -958,18 +964,25 @@ static int read_btree_roots(struct bch_fs *c) ret = bch2_btree_root_read(c, i, &r->key, r->level); if (ret) { - __fsck_err(c, i == BTREE_ID_alloc + __fsck_err(c, + btree_id_is_alloc(i) ? FSCK_CAN_IGNORE : 0, "error reading btree root %s", bch2_btree_ids[i]); - if (i == BTREE_ID_alloc) + if (btree_id_is_alloc(i)) c->sb.compat &= ~(1ULL << BCH_COMPAT_alloc_info); } } - for (i = 0; i < BTREE_ID_NR; i++) - if (!c->btree_roots[i].b) + for (i = 0; i < BTREE_ID_NR; i++) { + struct btree_root *r = &c->btree_roots[i]; + + if (!r->b) { + r->alive = false; + r->level = 0; bch2_btree_root_alloc(c, i); + } + } fsck_err: return ret; } @@ -994,7 +1007,6 @@ static int bch2_fs_initialize_subvolumes(struct bch_fs *c) if (ret) return ret; - bkey_subvolume_init(&root_volume.k_i); root_volume.k.p.offset = BCACHEFS_ROOT_SUBVOL; root_volume.v.flags = 0; @@ -1046,7 +1058,7 @@ int bch2_fs_recovery(struct bch_fs *c) const char *err = "cannot allocate memory"; struct bch_sb_field_clean *clean = NULL; struct jset *last_journal_entry = NULL; - u64 blacklist_seq, journal_seq; + u64 last_seq, blacklist_seq, journal_seq; bool write_sb = false; int ret = 0; @@ -1087,21 +1099,24 @@ int bch2_fs_recovery(struct bch_fs *c) c->opts.fix_errors = FSCK_OPT_YES; } - if (!c->replicas.entries || - c->opts.rebuild_replicas) { - bch_info(c, "building replicas info"); - set_bit(BCH_FS_REBUILD_REPLICAS, &c->flags); - } - if (!c->opts.nochanges) { - if (c->sb.version < bcachefs_metadata_version_new_data_types) { - bch_info(c, "version prior to new_data_types, upgrade and fsck required"); + if (c->sb.version < bcachefs_metadata_version_lru_v2) { + bch_info(c, "version prior to backpointers, upgrade and fsck required"); c->opts.version_upgrade = true; c->opts.fsck = true; c->opts.fix_errors = FSCK_OPT_YES; + } else if (c->sb.version < bcachefs_metadata_version_fragmentation_lru) { + bch_info(c, "version prior to backpointers, upgrade required"); + c->opts.version_upgrade = true; } } + if (c->opts.fsck && c->opts.norecovery) { + bch_err(c, "cannot select both norecovery and fsck"); + ret = -EINVAL; + goto err; + } + ret = bch2_blacklist_table_initialize(c); if (ret) { bch_err(c, "error initializing blacklist table"); @@ -1113,10 +1128,17 @@ int bch2_fs_recovery(struct bch_fs *c) struct journal_replay **i; bch_verbose(c, "starting journal read"); - ret = bch2_journal_read(c, &blacklist_seq, &journal_seq); + ret = bch2_journal_read(c, &last_seq, &blacklist_seq, &journal_seq); if (ret) goto err; + /* + * note: cmd_list_journal needs the blacklist table fully up to date so + * it can asterisk ignored journal entries: + */ + if (c->opts.read_journal_only) + goto out; + genradix_for_each_reverse(&c->journal_entries, iter, i) if (*i && !(*i)->ignore) { last_journal_entry = &(*i)->j; @@ -1134,7 +1156,15 @@ int bch2_fs_recovery(struct bch_fs *c) if (!last_journal_entry) { fsck_err_on(!c->sb.clean, c, "no journal entries found"); - goto use_clean; + if (clean) + goto use_clean; + + genradix_for_each_reverse(&c->journal_entries, iter, i) + if (*i) { + last_journal_entry = &(*i)->j; + (*i)->ignore = false; + break; + } } ret = journal_keys_sort(c); @@ -1151,16 +1181,13 @@ int bch2_fs_recovery(struct bch_fs *c) use_clean: if (!clean) { bch_err(c, "no superblock clean section found"); - ret = BCH_FSCK_REPAIR_IMPOSSIBLE; + ret = -BCH_ERR_fsck_repair_impossible; goto err; } 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); @@ -1183,7 +1210,9 @@ use_clean: journal_seq += 8; if (blacklist_seq != journal_seq) { - ret = bch2_journal_seq_blacklist_add(c, + ret = bch2_fs_log_msg(c, "blacklisting entries %llu-%llu", + blacklist_seq, journal_seq) ?: + bch2_journal_seq_blacklist_add(c, blacklist_seq, journal_seq); if (ret) { bch_err(c, "error creating new journal seq blacklist entry"); @@ -1191,10 +1220,22 @@ use_clean: } } - ret = bch2_fs_journal_start(&c->journal, journal_seq); + ret = bch2_fs_log_msg(c, "starting journal at entry %llu, replaying %llu-%llu", + journal_seq, last_seq, blacklist_seq - 1) ?: + bch2_fs_journal_start(&c->journal, journal_seq); if (ret) goto err; + if (c->opts.reconstruct_alloc) + bch2_fs_log_msg(c, "dropping alloc info"); + + /* + * Skip past versions that might have possibly been used (as nonces), + * but hadn't had their pointers written: + */ + if (c->sb.encryption_type && !c->sb.clean) + atomic64_add(1 << 16, &c->key_version); + ret = read_btree_roots(c); if (ret) goto err; @@ -1203,7 +1244,9 @@ use_clean: err = "error reading allocation information"; down_read(&c->gc_lock); - ret = bch2_alloc_read(c); + ret = c->sb.version < bcachefs_metadata_version_bucket_gens + ? bch2_alloc_read(c) + : bch2_bucket_gens_read(c); up_read(&c->gc_lock); if (ret) @@ -1217,17 +1260,21 @@ use_clean: goto err; bch_verbose(c, "stripes_read done"); - /* - * If we're not running fsck, this ensures bch2_fsck_err() calls are - * instead interpreted as bch2_inconsistent_err() calls: - */ - if (!c->opts.fsck) - set_bit(BCH_FS_FSCK_DONE, &c->flags); + if (c->sb.version < bcachefs_metadata_version_snapshot_2) { + err = "error creating root snapshot node"; + ret = bch2_fs_initialize_subvolumes(c); + if (ret) + goto err; + } + + bch_verbose(c, "reading snapshots table"); + err = "error reading snapshots table"; + ret = bch2_fs_snapshots_start(c); + if (ret) + goto err; + bch_verbose(c, "reading snapshots done"); - if (c->opts.fsck || - !(c->sb.compat & (1ULL << BCH_COMPAT_alloc_info)) || - !(c->sb.compat & (1ULL << BCH_COMPAT_alloc_metadata)) || - test_bit(BCH_FS_REBUILD_REPLICAS, &c->flags)) { + if (c->opts.fsck) { bool metadata_only = c->opts.norecovery; bch_info(c, "checking allocations"); @@ -1236,78 +1283,101 @@ use_clean: if (ret) goto err; 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); + set_bit(BCH_FS_INITIAL_GC_DONE, &c->flags); + + set_bit(BCH_FS_MAY_GO_RW, &c->flags); + + bch_info(c, "starting journal replay, %zu keys", c->journal_keys.nr); + err = "journal replay failed"; + ret = bch2_journal_replay(c, last_seq, blacklist_seq - 1); if (ret) goto err; + if (c->opts.verbose || !c->sb.clean) + bch_info(c, "journal replay done"); - ret = bch2_check_lrus(c, true); + bch_info(c, "checking need_discard and freespace btrees"); + err = "error checking need_discard and freespace btrees"; + ret = bch2_check_alloc_info(c); 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); + set_bit(BCH_FS_CHECK_ALLOC_DONE, &c->flags); - /* - * Skip past versions that might have possibly been used (as nonces), - * but hadn't had their pointers written: - */ - if (c->sb.encryption_type && !c->sb.clean) - atomic64_add(1 << 16, &c->key_version); + bch_info(c, "checking lrus"); + err = "error checking lrus"; + ret = bch2_check_lrus(c); + if (ret) + goto err; + bch_verbose(c, "done checking lrus"); + set_bit(BCH_FS_CHECK_LRUS_DONE, &c->flags); - if (c->opts.norecovery) - goto out; + bch_info(c, "checking backpointers to alloc keys"); + err = "error checking backpointers to alloc keys"; + ret = bch2_check_btree_backpointers(c); + if (ret) + goto err; + bch_verbose(c, "done checking backpointers to alloc keys"); - bch_verbose(c, "starting journal replay, %zu keys", c->journal_keys.nr); - err = "journal replay failed"; - ret = bch2_journal_replay(c); - if (ret) - goto err; - if (c->opts.verbose || !c->sb.clean) - bch_info(c, "journal replay done"); + bch_info(c, "checking backpointers to extents"); + err = "error checking backpointers to extents"; + ret = bch2_check_backpointers_to_extents(c); + if (ret) + goto err; + bch_verbose(c, "done checking backpointers to extents"); - err = "error initializing freespace"; - ret = bch2_fs_freespace_init(c); - if (ret) - goto err; + bch_info(c, "checking extents to backpointers"); + err = "error checking extents to backpointers"; + ret = bch2_check_extents_to_backpointers(c); + if (ret) + goto err; + bch_verbose(c, "done checking extents to backpointers"); + set_bit(BCH_FS_CHECK_BACKPOINTERS_DONE, &c->flags); - if (c->opts.fsck) { bch_info(c, "checking alloc to lru refs"); err = "error checking alloc to lru refs"; ret = bch2_check_alloc_to_lru_refs(c); - if (ret) - goto err; - - ret = bch2_check_lrus(c, true); if (ret) goto err; bch_verbose(c, "done checking alloc to lru refs"); - } + set_bit(BCH_FS_CHECK_ALLOC_TO_LRU_REFS_DONE, &c->flags); + } else { + set_bit(BCH_FS_INITIAL_GC_DONE, &c->flags); + set_bit(BCH_FS_CHECK_ALLOC_DONE, &c->flags); + set_bit(BCH_FS_CHECK_LRUS_DONE, &c->flags); + set_bit(BCH_FS_CHECK_BACKPOINTERS_DONE, &c->flags); + set_bit(BCH_FS_CHECK_ALLOC_TO_LRU_REFS_DONE, &c->flags); + set_bit(BCH_FS_FSCK_DONE, &c->flags); - if (c->sb.version < bcachefs_metadata_version_snapshot_2) { - bch2_fs_lazy_rw(c); + if (c->opts.norecovery) + goto out; - err = "error creating root snapshot node"; - ret = bch2_fs_initialize_subvolumes(c); + set_bit(BCH_FS_MAY_GO_RW, &c->flags); + + bch_verbose(c, "starting journal replay, %zu keys", c->journal_keys.nr); + err = "journal replay failed"; + ret = bch2_journal_replay(c, last_seq, blacklist_seq - 1); if (ret) goto err; + if (c->opts.verbose || !c->sb.clean) + bch_info(c, "journal replay done"); } - bch_verbose(c, "reading snapshots table"); - err = "error reading snapshots table"; - ret = bch2_fs_snapshots_start(c); + err = "error initializing freespace"; + ret = bch2_fs_freespace_init(c); if (ret) goto err; - bch_verbose(c, "reading snapshots done"); + + if (c->sb.version < bcachefs_metadata_version_bucket_gens && + c->opts.version_upgrade) { + bch_info(c, "initializing bucket_gens"); + err = "error initializing bucket gens"; + ret = bch2_bucket_gens_init(c); + if (ret) + goto err; + bch_verbose(c, "bucket_gens init done"); + } if (c->sb.version < bcachefs_metadata_version_snapshot_2) { /* set bi_subvol on root inode */ @@ -1371,7 +1441,7 @@ use_clean: le16_to_cpu(c->sb.version_min) < bcachefs_metadata_version_btree_ptr_sectors_written) { struct bch_move_stats stats; - bch_move_stats_init(&stats, "recovery"); + bch2_move_stats_init(&stats, "recovery"); bch_info(c, "scanning for old btree nodes"); ret = bch2_fs_read_write(c); @@ -1393,15 +1463,22 @@ out: set_bit(BCH_FS_FSCK_DONE, &c->flags); bch2_flush_fsck_errs(c); - if (!c->opts.keep_journal) { + if (!c->opts.keep_journal && + test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags)) { bch2_journal_keys_free(&c->journal_keys); bch2_journal_entries_free(c); } kfree(clean); + + if (!ret && test_bit(BCH_FS_HAVE_DELETED_SNAPSHOTS, &c->flags)) { + bch2_fs_read_write_early(c); + bch2_delete_dead_snapshots_async(c); + } + if (ret) - bch_err(c, "Error in recovery: %s (%i)", err, ret); + bch_err(c, "Error in recovery: %s (%s)", err, bch2_err_str(ret)); else - bch_verbose(c, "ret %i", ret); + bch_verbose(c, "ret %s", bch2_err_str(ret)); return ret; err: fsck_err: @@ -1425,6 +1502,9 @@ int bch2_fs_initialize(struct bch_fs *c) c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_extents_above_btree_updates_done); c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_bformat_overflow_done); + if (c->sb.version < bcachefs_metadata_version_inode_v3) + c->opts.version_upgrade = true; + if (c->opts.version_upgrade) { c->disk_sb.sb->version = cpu_to_le16(bcachefs_metadata_version_current); c->disk_sb.sb->features[0] |= cpu_to_le64(BCH_SB_FEATURES_ALL); @@ -1433,6 +1513,9 @@ int bch2_fs_initialize(struct bch_fs *c) mutex_unlock(&c->sb_lock); set_bit(BCH_FS_INITIAL_GC_DONE, &c->flags); + set_bit(BCH_FS_CHECK_LRUS_DONE, &c->flags); + set_bit(BCH_FS_CHECK_BACKPOINTERS_DONE, &c->flags); + set_bit(BCH_FS_CHECK_ALLOC_TO_LRU_REFS_DONE, &c->flags); set_bit(BCH_FS_MAY_GO_RW, &c->flags); set_bit(BCH_FS_FSCK_DONE, &c->flags); @@ -1497,11 +1580,10 @@ int bch2_fs_initialize(struct bch_fs *c) goto err; bch_verbose(c, "reading snapshots done"); - bch2_inode_init(c, &root_inode, 0, 0, - S_IFDIR|S_IRWXU|S_IRUGO|S_IXUGO, 0, NULL); + bch2_inode_init(c, &root_inode, 0, 0, S_IFDIR|0755, 0, NULL); root_inode.bi_inum = BCACHEFS_ROOT_INO; root_inode.bi_subvol = BCACHEFS_ROOT_SUBVOL; - bch2_inode_pack(c, &packed_inode, &root_inode); + bch2_inode_pack(&packed_inode, &root_inode); packed_inode.inode.k.p.snapshot = U32_MAX; err = "error creating root directory"; @@ -1546,6 +1628,6 @@ int bch2_fs_initialize(struct bch_fs *c) return 0; err: - pr_err("Error initializing new filesystem: %s (%i)", err, ret); + pr_err("Error initializing new filesystem: %s (%s)", err, bch2_err_str(ret)); return ret; }