X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Frecovery.c;h=9127d0e3ca2f6a3fd44e076b42f01ee6f7736427;hb=f3f005c76eb5636542a8f5b137bd1904d57e8f86;hp=460b1ba22c8e40190b1dd9bf10745e5952a2eb41;hpb=55e3496d06c9b112f93bb1dea942564f900c2f7d;p=bcachefs-tools-debian diff --git a/libbcachefs/recovery.c b/libbcachefs/recovery.c index 460b1ba..9127d0e 100644 --- a/libbcachefs/recovery.c +++ b/libbcachefs/recovery.c @@ -1,25 +1,34 @@ // SPDX-License-Identifier: GPL-2.0 #include "bcachefs.h" +#include "backpointers.h" #include "bkey_buf.h" #include "alloc_background.h" #include "btree_gc.h" +#include "btree_journal_iter.h" #include "btree_update.h" #include "btree_update_interior.h" #include "btree_io.h" #include "buckets.h" #include "dirent.h" #include "ec.h" +#include "errcode.h" #include "error.h" #include "fs-common.h" #include "fsck.h" #include "journal_io.h" #include "journal_reclaim.h" #include "journal_seq_blacklist.h" +#include "lru.h" +#include "logged_ops.h" #include "move.h" #include "quota.h" +#include "rebalance.h" #include "recovery.h" #include "replicas.h" +#include "sb-clean.h" +#include "sb-downgrade.h" +#include "snapshot.h" #include "subvolume.h" #include "super-io.h" @@ -28,13 +37,27 @@ #define QSTR(n) { { { .len = strlen(n) } }, .name = n } +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; + } +} + /* for -o reconstruct_alloc: */ static void drop_alloc_keys(struct journal_keys *keys) { size_t src, dst; for (src = 0, dst = 0; src < keys->nr; src++) - if (keys->d[src].btree_id != BTREE_ID_alloc) + if (!btree_id_is_alloc(keys->d[src].btree_id)) keys->d[dst++] = keys->d[src]; keys->nr = dst; @@ -54,595 +77,166 @@ static void zero_out_btree_mem_ptr(struct journal_keys *keys) bkey_i_to_btree_ptr_v2(i->k)->v.mem_ptr = 0; } -/* iterate over keys read from the journal: */ - -static int __journal_key_cmp(enum btree_id l_btree_id, - unsigned l_level, - struct bpos l_pos, - struct journal_key *r) -{ - return (cmp_int(l_btree_id, r->btree_id) ?: - cmp_int(l_level, r->level) ?: - bpos_cmp(l_pos, r->k->k.p)); -} - -static int journal_key_cmp(struct journal_key *l, struct journal_key *r) -{ - return (cmp_int(l->btree_id, r->btree_id) ?: - cmp_int(l->level, r->level) ?: - bpos_cmp(l->k->k.p, r->k->k.p)); -} - -static size_t journal_key_search(struct journal_keys *journal_keys, - enum btree_id id, unsigned level, - struct bpos pos) -{ - size_t l = 0, r = journal_keys->nr, m; - - while (l < r) { - m = l + ((r - l) >> 1); - if (__journal_key_cmp(id, level, pos, &journal_keys->d[m]) > 0) - l = m + 1; - else - r = m; - } - - BUG_ON(l < journal_keys->nr && - __journal_key_cmp(id, level, pos, &journal_keys->d[l]) > 0); - - BUG_ON(l && - __journal_key_cmp(id, level, pos, &journal_keys->d[l - 1]) <= 0); - - return l; -} - -static void journal_iter_fix(struct bch_fs *c, struct journal_iter *iter, unsigned idx) -{ - struct bkey_i *n = iter->keys->d[idx].k; - struct btree_and_journal_iter *biter = - container_of(iter, struct btree_and_journal_iter, journal); - - if (iter->idx > idx || - (iter->idx == idx && - biter->last && - bpos_cmp(n->k.p, biter->unpacked.p) <= 0)) - iter->idx++; -} - -int bch2_journal_key_insert(struct bch_fs *c, enum btree_id id, - unsigned level, struct bkey_i *k) -{ - struct journal_key n = { - .btree_id = id, - .level = level, - .k = k, - .allocated = true - }; - struct journal_keys *keys = &c->journal_keys; - struct journal_iter *iter; - unsigned idx = journal_key_search(keys, id, level, k->k.p); - - if (idx < keys->nr && - journal_key_cmp(&n, &keys->d[idx]) == 0) { - if (keys->d[idx].allocated) - kfree(keys->d[idx].k); - keys->d[idx] = n; - return 0; - } - - if (keys->nr == keys->size) { - struct journal_keys new_keys = { - .nr = keys->nr, - .size = keys->size * 2, - .journal_seq_base = keys->journal_seq_base, - }; - - new_keys.d = kvmalloc(sizeof(new_keys.d[0]) * new_keys.size, GFP_KERNEL); - if (!new_keys.d) { - bch_err(c, "%s: error allocating new key array (size %zu)", - __func__, new_keys.size); - return -ENOMEM; - } - - memcpy(new_keys.d, keys->d, sizeof(keys->d[0]) * keys->nr); - kvfree(keys->d); - *keys = new_keys; - } - - array_insert_item(keys->d, keys->nr, idx, n); - - list_for_each_entry(iter, &c->journal_iters, list) - journal_iter_fix(c, iter, idx); - - return 0; -} - -int bch2_journal_key_delete(struct bch_fs *c, enum btree_id id, - unsigned level, struct bpos pos) -{ - struct bkey_i *whiteout = - kmalloc(sizeof(struct bkey), GFP_KERNEL); - int ret; - - if (!whiteout) { - bch_err(c, "%s: error allocating new key", __func__); - return -ENOMEM; - } - - bkey_init(&whiteout->k); - whiteout->k.p = pos; - - ret = bch2_journal_key_insert(c, id, level, whiteout); - if (ret) - kfree(whiteout); - return ret; -} - -static struct bkey_i *bch2_journal_iter_peek(struct journal_iter *iter) -{ - struct journal_key *k = iter->idx - iter->keys->nr - ? iter->keys->d + iter->idx : NULL; - - if (k && - k->btree_id == iter->btree_id && - k->level == iter->level) - return k->k; - - iter->idx = iter->keys->nr; - return NULL; -} - -static void bch2_journal_iter_advance(struct journal_iter *iter) -{ - if (iter->idx < iter->keys->nr) - iter->idx++; -} - -static void bch2_journal_iter_exit(struct journal_iter *iter) -{ - list_del(&iter->list); -} - -static void bch2_journal_iter_init(struct bch_fs *c, - struct journal_iter *iter, - enum btree_id id, unsigned level, - struct bpos pos) -{ - iter->btree_id = id; - iter->level = level; - iter->keys = &c->journal_keys; - iter->idx = journal_key_search(&c->journal_keys, id, level, pos); - list_add(&iter->list, &c->journal_iters); -} - -static struct bkey_s_c bch2_journal_iter_peek_btree(struct btree_and_journal_iter *iter) -{ - return bch2_btree_node_iter_peek_unpack(&iter->node_iter, - iter->b, &iter->unpacked); -} - -static void bch2_journal_iter_advance_btree(struct btree_and_journal_iter *iter) -{ - bch2_btree_node_iter_advance(&iter->node_iter, iter->b); -} - -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; -} - -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)); - - if (btree_k.k && journal_k.k) { - int cmp = bpos_cmp(btree_k.k->p, journal_k.k->p); - - if (!cmp) - bch2_journal_iter_advance_btree(iter); - - 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 = iter->last == journal ? journal_k : btree_k; - - 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 (!bkey_deleted(ret.k)) - break; - - bch2_btree_and_journal_iter_advance(iter); - } - - 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); -} - -void bch2_btree_and_journal_iter_init_node_iter(struct btree_and_journal_iter *iter, - struct bch_fs *c, - struct btree *b) -{ - memset(iter, 0, sizeof(*iter)); - - iter->b = b; - bch2_btree_node_iter_init_from_start(&iter->node_iter, iter->b); - bch2_journal_iter_init(c, &iter->journal, - b->c.btree_id, b->c.level, b->data->min_key); -} - -/* Walk btree, overlaying keys from the journal: */ - -static void btree_and_journal_iter_prefetch(struct bch_fs *c, struct btree *b, - struct btree_and_journal_iter iter) -{ - unsigned i = 0, nr = b->c.level > 1 ? 2 : 16; - struct bkey_s_c k; - struct bkey_buf tmp; - - BUG_ON(!b->c.level); - - bch2_bkey_buf_init(&tmp); - - while (i < nr && - (k = bch2_btree_and_journal_iter_peek(&iter)).k) { - bch2_bkey_buf_reassemble(&tmp, c, k); - - bch2_btree_node_prefetch(c, NULL, NULL, tmp.k, - b->c.btree_id, b->c.level - 1); - - bch2_btree_and_journal_iter_advance(&iter); - i++; - } - - bch2_bkey_buf_exit(&tmp, c); -} - -static int bch2_btree_and_journal_walk_recurse(struct btree_trans *trans, struct btree *b, - enum btree_id btree_id, - btree_walk_key_fn key_fn) -{ - struct bch_fs *c = trans->c; - struct btree_and_journal_iter iter; - struct bkey_s_c k; - struct bkey_buf tmp; - struct btree *child; - int ret = 0; - - bch2_bkey_buf_init(&tmp); - bch2_btree_and_journal_iter_init_node_iter(&iter, c, b); - - while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) { - if (b->c.level) { - bch2_bkey_buf_reassemble(&tmp, c, k); - - child = bch2_btree_node_get_noiter(c, tmp.k, - b->c.btree_id, b->c.level - 1, - false); - - ret = PTR_ERR_OR_ZERO(child); - if (ret) - break; - - btree_and_journal_iter_prefetch(c, b, iter); - - ret = bch2_btree_and_journal_walk_recurse(trans, child, - btree_id, key_fn); - six_unlock_read(&child->c.lock); - } else { - ret = key_fn(trans, k); - } - - if (ret) - break; - - bch2_btree_and_journal_iter_advance(&iter); - } - - bch2_btree_and_journal_iter_exit(&iter); - bch2_bkey_buf_exit(&tmp, c); - return ret; -} - -int bch2_btree_and_journal_walk(struct btree_trans *trans, enum btree_id btree_id, - btree_walk_key_fn key_fn) -{ - struct bch_fs *c = trans->c; - struct btree *b = c->btree_roots[btree_id].b; - int ret = 0; - - if (btree_node_fake(b)) - return 0; - - six_lock_read(&b->c.lock, NULL, NULL); - ret = bch2_btree_and_journal_walk_recurse(trans, b, btree_id, key_fn); - six_unlock_read(&b->c.lock); - - return ret; -} - -/* sort and dedup all keys in the journal: */ - -void bch2_journal_entries_free(struct list_head *list) -{ - - while (!list_empty(list)) { - struct journal_replay *i = - list_first_entry(list, struct journal_replay, list); - list_del(&i->list); - kvpfree(i, offsetof(struct journal_replay, j) + - vstruct_bytes(&i->j)); - } -} - -/* - * When keys compare equal, oldest compares first: - */ -static int journal_sort_key_cmp(const void *_l, const void *_r) -{ - const struct journal_key *l = _l; - const struct journal_key *r = _r; - - return cmp_int(l->btree_id, r->btree_id) ?: - cmp_int(l->level, r->level) ?: - bpos_cmp(l->k->k.p, r->k->k.p) ?: - cmp_int(l->journal_seq, r->journal_seq) ?: - cmp_int(l->journal_offset, r->journal_offset); -} - -void bch2_journal_keys_free(struct journal_keys *keys) -{ - struct journal_key *i; - - for (i = keys->d; i < keys->d + keys->nr; i++) - if (i->allocated) - kfree(i->k); - - kvfree(keys->d); - keys->d = NULL; - keys->nr = 0; -} - -static struct journal_keys journal_keys_sort(struct list_head *journal_entries) -{ - struct journal_replay *i; - struct jset_entry *entry; - struct bkey_i *k, *_n; - struct journal_keys keys = { NULL }; - struct journal_key *src, *dst; - size_t nr_keys = 0; - - if (list_empty(journal_entries)) - return keys; - - list_for_each_entry(i, journal_entries, list) { - if (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++; - } - - keys.size = roundup_pow_of_two(nr_keys); - - keys.d = kvmalloc(sizeof(keys.d[0]) * keys.size, GFP_KERNEL); - if (!keys.d) - goto err; - - list_for_each_entry(i, journal_entries, list) { - if (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_offset = k->_data - i->j._data, - }; - } - - sort(keys.d, keys.nr, sizeof(keys.d[0]), journal_sort_key_cmp, NULL); - - src = dst = keys.d; - while (src < keys.d + keys.nr) { - 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)) - src++; - - *dst++ = *src++; - } - - keys.nr = dst - keys.d; -err: - return keys; -} - /* journal replay: */ 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++); } -static int __bch2_journal_replay_key(struct btree_trans *trans, - struct journal_key *k) +static int bch2_journal_replay_key(struct btree_trans *trans, + struct journal_key *k) { struct btree_iter iter; unsigned iter_flags = BTREE_ITER_INTENT| BTREE_ITER_NOT_EXTENTS; + unsigned update_flags = BTREE_TRIGGER_NORUN; int ret; + if (k->overwritten) + return 0; + + trans->journal_res.seq = k->journal_seq; + + /* + * BTREE_UPDATE_KEY_CACHE_RECLAIM disables key cache lookup/update to + * keep the key cache coherent with the underlying btree. Nothing + * besides the allocator is doing updates yet so we don't need key cache + * coherency for non-alloc btrees, and key cache fills for snapshots + * btrees use BTREE_ITER_FILTER_SNAPSHOTS, which isn't available until + * the snapshots recovery pass runs. + */ if (!k->level && k->btree_id == BTREE_ID_alloc) - iter_flags |= BTREE_ITER_CACHED|BTREE_ITER_CACHED_NOFILL; + iter_flags |= BTREE_ITER_CACHED; + else + update_flags |= BTREE_UPDATE_KEY_CACHE_RECLAIM; bch2_trans_node_iter_init(trans, &iter, k->btree_id, k->k->k.p, BTREE_MAX_DEPTH, k->level, iter_flags); - ret = bch2_btree_iter_traverse(&iter) ?: - bch2_trans_update(trans, &iter, k->k, BTREE_TRIGGER_NORUN); - bch2_trans_iter_exit(trans, &iter); - return ret; -} - -static int bch2_journal_replay_key(struct bch_fs *c, struct journal_key *k) -{ - unsigned commit_flags = - BTREE_INSERT_LAZY_RW| - BTREE_INSERT_NOFAIL| - BTREE_INSERT_JOURNAL_RESERVED; + ret = bch2_btree_iter_traverse(&iter); + if (ret) + goto out; - if (!k->allocated) - commit_flags |= BTREE_INSERT_JOURNAL_REPLAY; + /* Must be checked with btree locked: */ + if (k->overwritten) + goto out; - return bch2_trans_do(c, NULL, NULL, commit_flags, - __bch2_journal_replay_key(&trans, k)); + ret = bch2_trans_update(trans, &iter, k->k, update_flags); +out: + bch2_trans_iter_exit(trans, &iter); + return ret; } static int journal_sort_seq_cmp(const void *_l, const void *_r) { - const struct journal_key *l = _l; - const struct journal_key *r = _r; + const struct journal_key *l = *((const struct journal_key **)_l); + const struct journal_key *r = *((const struct journal_key **)_r); - return cmp_int(r->level, l->level) ?: - cmp_int(l->journal_seq, r->journal_seq) ?: - cmp_int(l->btree_id, r->btree_id) ?: - bpos_cmp(l->k->k.p, r->k->k.p); + return cmp_int(l->journal_seq, r->journal_seq); } -static int bch2_journal_replay(struct bch_fs *c, - struct journal_keys keys) +static int bch2_journal_replay(struct bch_fs *c) { + struct journal_keys *keys = &c->journal_keys; + DARRAY(struct journal_key *) keys_sorted = { 0 }; struct journal *j = &c->journal; - struct journal_key *i; - u64 seq; - int ret; - - sort(keys.d, keys.nr, sizeof(keys.d[0]), journal_sort_seq_cmp, NULL); + u64 start_seq = c->journal_replay_seq_start; + u64 end_seq = c->journal_replay_seq_start; + struct btree_trans *trans = bch2_trans_get(c); + int ret = 0; - if (keys.nr) - replay_now_at(j, keys.journal_seq_base); + if (keys->nr) { + ret = bch2_journal_log_msg(c, "Starting journal replay (%zu keys in entries %llu-%llu)", + keys->nr, start_seq, end_seq); + if (ret) + goto err; + } - seq = j->replay_journal_seq; + BUG_ON(!atomic_read(&keys->ref)); /* - * First replay updates to the alloc btree - these will only update the - * btree key cache: + * First, attempt to replay keys in sorted order. This is more + * efficient - better locality of btree access - but some might fail if + * that would cause a journal deadlock. */ - for_each_journal_key(keys, i) { + for (size_t i = 0; i < keys->nr; i++) { cond_resched(); - if (!i->level && i->btree_id == BTREE_ID_alloc) { - j->replay_journal_seq = keys.journal_seq_base + i->journal_seq; - ret = bch2_journal_replay_key(c, i); - if (ret) - goto err; - } - } - - /* - * Next replay updates to interior btree nodes: - */ - for_each_journal_key(keys, i) { - cond_resched(); + struct journal_key *k = keys->d + i; - if (i->level) { - j->replay_journal_seq = keys.journal_seq_base + i->journal_seq; - ret = bch2_journal_replay_key(c, i); + /* Skip fastpath if we're low on space in the journal */ + ret = c->journal.watermark ? -1 : + commit_do(trans, NULL, NULL, + BCH_TRANS_COMMIT_no_enospc| + BCH_TRANS_COMMIT_journal_reclaim| + (!k->allocated ? BCH_TRANS_COMMIT_no_journal_res : 0), + bch2_journal_replay_key(trans, k)); + BUG_ON(!ret && !k->overwritten); + if (ret) { + ret = darray_push(&keys_sorted, k); if (ret) goto err; } } /* - * Now that the btree is in a consistent state, we can start journal - * reclaim (which will be flushing entries from the btree key cache back - * to the btree: + * Now, replay any remaining keys in the order in which they appear in + * the journal, unpinning those journal entries as we go: */ - set_bit(BCH_FS_BTREE_INTERIOR_REPLAY_DONE, &c->flags); - set_bit(JOURNAL_RECLAIM_STARTED, &j->flags); - journal_reclaim_kick(j); + sort(keys_sorted.data, keys_sorted.nr, + sizeof(keys_sorted.data[0]), + journal_sort_seq_cmp, NULL); - j->replay_journal_seq = seq; - - /* - * Now replay leaf node updates: - */ - for_each_journal_key(keys, i) { + darray_for_each(keys_sorted, kp) { cond_resched(); - if (i->level || i->btree_id == BTREE_ID_alloc) - continue; + struct journal_key *k = *kp; - replay_now_at(j, keys.journal_seq_base + i->journal_seq); + replay_now_at(j, k->journal_seq); - ret = bch2_journal_replay_key(c, i); + ret = commit_do(trans, NULL, NULL, + BCH_TRANS_COMMIT_no_enospc| + (!k->allocated + ? BCH_TRANS_COMMIT_no_journal_res|BCH_WATERMARK_reclaim + : 0), + bch2_journal_replay_key(trans, k)); + bch_err_msg(c, ret, "while replaying key at btree %s level %u:", + bch2_btree_id_str(k->btree_id), k->level); if (ret) goto err; + + BUG_ON(!k->overwritten); } + /* + * We need to put our btree_trans before calling flush_all_pins(), since + * that will use a btree_trans internally + */ + bch2_trans_put(trans); + trans = NULL; + + if (!c->opts.keep_journal) + bch2_journal_keys_put_initial(c); + replay_now_at(j, j->replay_journal_seq_end); j->replay_journal_seq = 0; bch2_journal_set_replay_done(j); - bch2_journal_flush_all_pins(j); - return bch2_journal_error(j); + + if (keys->nr) + bch2_journal_log_msg(c, "journal replay finished"); err: - bch_err(c, "journal replay: error %d while replaying key at btree %s level %u", - ret, bch2_btree_ids[i->btree_id], i->level); + if (trans) + bch2_trans_put(trans); + darray_exit(&keys_sorted); + bch_err_fn(c, ret); return ret; } @@ -657,17 +251,17 @@ static int journal_replay_entry_early(struct bch_fs *c, case BCH_JSET_ENTRY_btree_root: { struct btree_root *r; - if (entry->btree_id >= BTREE_ID_NR) { - bch_err(c, "filesystem has unknown btree type %u", - entry->btree_id); - return -EINVAL; + while (entry->btree_id >= c->btree_roots_extra.nr + BTREE_ID_NR) { + ret = darray_push(&c->btree_roots_extra, (struct btree_root) { NULL }); + if (ret) + return ret; } - r = &c->btree_roots[entry->btree_id]; + r = bch2_btree_id_root(c, entry->btree_id); if (entry->u64s) { r->level = entry->level; - bkey_copy(&r->key, &entry->start[0]); + bkey_copy(&r->key, (struct bkey_i *) entry->start); r->error = 0; } else { r->error = -EIO; @@ -680,15 +274,15 @@ static int journal_replay_entry_early(struct bch_fs *c, container_of(entry, struct jset_entry_usage, entry); switch (entry->btree_id) { - case FS_USAGE_RESERVED: + case BCH_FS_USAGE_reserved: if (entry->level < BCH_REPLICAS_MAX) c->usage_base->persistent_reserved[entry->level] = le64_to_cpu(u->v); break; - case FS_USAGE_INODES: - c->usage_base->nr_inodes = le64_to_cpu(u->v); + case BCH_FS_USAGE_inodes: + c->usage_base->b.nr_inodes = le64_to_cpu(u->v); break; - case FS_USAGE_KEY_VERSION: + case BCH_FS_USAGE_key_version: atomic64_set(&c->key_version, le64_to_cpu(u->v)); break; @@ -708,13 +302,7 @@ static int journal_replay_entry_early(struct bch_fs *c, struct jset_entry_dev_usage *u = container_of(entry, struct jset_entry_dev_usage, entry); struct bch_dev *ca = bch_dev_bkey_exists(c, le32_to_cpu(u->dev)); - unsigned bytes = jset_u64s(le16_to_cpu(entry->u64s)) * sizeof(u64); - unsigned nr_types = (bytes - sizeof(struct jset_entry_dev_usage)) / - sizeof(struct jset_entry_dev_usage_type); - unsigned i; - - ca->usage_base->buckets_ec = le64_to_cpu(u->buckets_ec); - ca->usage_base->buckets_unavailable = le64_to_cpu(u->buckets_unavailable); + unsigned i, nr_types = jset_entry_dev_usage_nr_types(u); for (i = 0; i < min_t(unsigned, nr_types, BCH_DATA_NR); i++) { ca->usage_base->d[i].buckets = le64_to_cpu(u->d[i].buckets); @@ -754,251 +342,143 @@ static int journal_replay_entry_early(struct bch_fs *c, } static int journal_replay_early(struct bch_fs *c, - struct bch_sb_field_clean *clean, - struct list_head *journal) + struct bch_sb_field_clean *clean) { - struct journal_replay *i; - struct jset_entry *entry; - int ret; - if (clean) { - for (entry = clean->start; + for (struct jset_entry *entry = clean->start; entry != vstruct_end(&clean->field); entry = vstruct_next(entry)) { - ret = journal_replay_entry_early(c, entry); + int ret = journal_replay_entry_early(c, entry); if (ret) return ret; } } else { - list_for_each_entry(i, journal, list) { - if (i->ignore) + struct genradix_iter iter; + struct journal_replay *i, **_i; + + genradix_for_each(&c->journal_entries, iter, _i) { + i = *_i; + + if (!i || i->ignore) continue; vstruct_for_each(&i->j, entry) { - ret = journal_replay_entry_early(c, entry); + int ret = journal_replay_entry_early(c, entry); if (ret) return ret; } } } - bch2_fs_usage_initialize(c); - - return 0; -} - -/* sb clean section: */ - -static struct bkey_i *btree_root_find(struct bch_fs *c, - struct bch_sb_field_clean *clean, - struct jset *j, - enum btree_id id, unsigned *level) -{ - struct bkey_i *k; - struct jset_entry *entry, *start, *end; - - if (clean) { - start = clean->start; - end = vstruct_end(&clean->field); - } else { - start = j->start; - end = vstruct_last(j); - } - - for (entry = start; entry < end; entry = vstruct_next(entry)) - if (entry->type == BCH_JSET_ENTRY_btree_root && - entry->btree_id == id) - goto found; - - return NULL; -found: - if (!entry->u64s) - return ERR_PTR(-EINVAL); - - k = entry->start; - *level = entry->level; - return k; -} - -static int verify_superblock_clean(struct bch_fs *c, - struct bch_sb_field_clean **cleanp, - struct jset *j) -{ - unsigned i; - struct bch_sb_field_clean *clean = *cleanp; - int ret = 0; - - if (mustfix_fsck_err_on(j->seq != clean->journal_seq, c, - "superblock journal seq (%llu) doesn't match journal (%llu) after clean shutdown", - le64_to_cpu(clean->journal_seq), - le64_to_cpu(j->seq))) { - kfree(clean); - *cleanp = NULL; - return 0; - } - - for (i = 0; i < BTREE_ID_NR; i++) { - char buf1[200], buf2[200]; - struct bkey_i *k1, *k2; - unsigned l1 = 0, l2 = 0; - - k1 = btree_root_find(c, clean, NULL, i, &l1); - k2 = btree_root_find(c, NULL, j, i, &l2); - - if (!k1 && !k2) - continue; - - mustfix_fsck_err_on(!k1 || !k2 || - IS_ERR(k1) || - IS_ERR(k2) || - k1->k.u64s != k2->k.u64s || - memcmp(k1, k2, bkey_bytes(k1)) || - l1 != l2, 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)); - } -fsck_err: - return ret; -} - -static struct bch_sb_field_clean *read_superblock_clean(struct bch_fs *c) -{ - struct bch_sb_field_clean *clean, *sb_clean; - int ret; - - mutex_lock(&c->sb_lock); - sb_clean = bch2_sb_get_clean(c->disk_sb.sb); - - if (fsck_err_on(!sb_clean, c, - "superblock marked clean but clean section not present")) { - SET_BCH_SB_CLEAN(c->disk_sb.sb, false); - c->sb.clean = false; - mutex_unlock(&c->sb_lock); - return NULL; - } - - clean = kmemdup(sb_clean, vstruct_bytes(&sb_clean->field), - GFP_KERNEL); - if (!clean) { - mutex_unlock(&c->sb_lock); - return ERR_PTR(-ENOMEM); - } - - ret = bch2_sb_clean_validate(c, clean, READ); - if (ret) { - mutex_unlock(&c->sb_lock); - return ERR_PTR(ret); - } - - mutex_unlock(&c->sb_lock); + bch2_fs_usage_initialize(c); - return clean; -fsck_err: - mutex_unlock(&c->sb_lock); - return ERR_PTR(ret); + return 0; } +/* sb clean section: */ + static int read_btree_roots(struct bch_fs *c) { unsigned i; int ret = 0; - for (i = 0; i < BTREE_ID_NR; i++) { - struct btree_root *r = &c->btree_roots[i]; + for (i = 0; i < btree_id_nr_alive(c); i++) { + struct btree_root *r = bch2_btree_id_root(c, i); 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, + btree_root_bkey_invalid, "invalid btree root %s", - bch2_btree_ids[i]); + bch2_btree_id_str(i)); if (i == BTREE_ID_alloc) c->sb.compat &= ~(1ULL << BCH_COMPAT_alloc_info); } ret = bch2_btree_root_read(c, i, &r->key, r->level); if (ret) { - __fsck_err(c, i == BTREE_ID_alloc - ? FSCK_CAN_IGNORE : 0, - "error reading btree root %s", - bch2_btree_ids[i]); - if (i == BTREE_ID_alloc) + fsck_err(c, + btree_root_read_error, + "error reading btree root %s", + bch2_btree_id_str(i)); + if (btree_id_is_alloc(i)) c->sb.compat &= ~(1ULL << BCH_COMPAT_alloc_info); + ret = 0; } } - 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 = bch2_btree_id_root(c, i); + + if (!r->b) { + r->alive = false; + r->level = 0; bch2_btree_root_alloc(c, i); + } + } fsck_err: return ret; } -static int bch2_fs_initialize_subvolumes(struct bch_fs *c) +static int bch2_initialize_subvolumes(struct bch_fs *c) { - struct bkey_i_snapshot root_snapshot; - struct bkey_i_subvolume root_volume; + struct bkey_i_snapshot_tree root_tree; + struct bkey_i_snapshot root_snapshot; + struct bkey_i_subvolume root_volume; int ret; + bkey_snapshot_tree_init(&root_tree.k_i); + root_tree.k.p.offset = 1; + root_tree.v.master_subvol = cpu_to_le32(1); + root_tree.v.root_snapshot = cpu_to_le32(U32_MAX); + bkey_snapshot_init(&root_snapshot.k_i); root_snapshot.k.p.offset = U32_MAX; root_snapshot.v.flags = 0; root_snapshot.v.parent = 0; - root_snapshot.v.subvol = BCACHEFS_ROOT_SUBVOL; - root_snapshot.v.pad = 0; + root_snapshot.v.subvol = cpu_to_le32(BCACHEFS_ROOT_SUBVOL); + root_snapshot.v.tree = cpu_to_le32(1); SET_BCH_SNAPSHOT_SUBVOL(&root_snapshot.v, true); - ret = bch2_btree_insert(c, BTREE_ID_snapshots, - &root_snapshot.k_i, - NULL, NULL, 0); - if (ret) - return ret; - - bkey_subvolume_init(&root_volume.k_i); root_volume.k.p.offset = BCACHEFS_ROOT_SUBVOL; root_volume.v.flags = 0; root_volume.v.snapshot = cpu_to_le32(U32_MAX); root_volume.v.inode = cpu_to_le64(BCACHEFS_ROOT_INO); - ret = bch2_btree_insert(c, BTREE_ID_subvolumes, - &root_volume.k_i, - NULL, NULL, 0); - if (ret) - return ret; - - return 0; + ret = bch2_btree_insert(c, BTREE_ID_snapshot_trees, &root_tree.k_i, NULL, 0) ?: + bch2_btree_insert(c, BTREE_ID_snapshots, &root_snapshot.k_i, NULL, 0) ?: + bch2_btree_insert(c, BTREE_ID_subvolumes, &root_volume.k_i, NULL, 0); + bch_err_fn(c, ret); + return ret; } -static int bch2_fs_upgrade_for_subvolumes(struct btree_trans *trans) +static int __bch2_fs_upgrade_for_subvolumes(struct btree_trans *trans) { - struct bch_fs *c = trans->c; struct btree_iter iter; struct bkey_s_c k; struct bch_inode_unpacked inode; int ret; - bch2_trans_iter_init(trans, &iter, BTREE_ID_inodes, - SPOS(0, BCACHEFS_ROOT_INO, U32_MAX), 0); - k = bch2_btree_iter_peek_slot(&iter); + k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes, + SPOS(0, BCACHEFS_ROOT_INO, U32_MAX), 0); ret = bkey_err(k); if (ret) - goto err; + return ret; if (!bkey_is_inode(k.k)) { - bch_err(c, "root inode not found"); - ret = -ENOENT; + bch_err(trans->c, "root inode not found"); + ret = -BCH_ERR_ENOENT_inode; goto err; } @@ -1013,26 +493,267 @@ err: return ret; } +/* set bi_subvol on root inode */ +noinline_for_stack +static int bch2_fs_upgrade_for_subvolumes(struct bch_fs *c) +{ + int ret = bch2_trans_do(c, NULL, NULL, BCH_TRANS_COMMIT_lazy_rw, + __bch2_fs_upgrade_for_subvolumes(trans)); + bch_err_fn(c, ret); + return ret; +} + +const char * const bch2_recovery_passes[] = { +#define x(_fn, ...) #_fn, + BCH_RECOVERY_PASSES() +#undef x + NULL +}; + +static int bch2_check_allocations(struct bch_fs *c) +{ + return bch2_gc(c, true, c->opts.norecovery); +} + +static int bch2_set_may_go_rw(struct bch_fs *c) +{ + struct journal_keys *keys = &c->journal_keys; + + /* + * After we go RW, the journal keys buffer can't be modified (except for + * setting journal_key->overwritten: it will be accessed by multiple + * threads + */ + move_gap(keys->d, keys->nr, keys->size, keys->gap, keys->nr); + keys->gap = keys->nr; + + set_bit(BCH_FS_may_go_rw, &c->flags); + + if (keys->nr || c->opts.fsck || !c->sb.clean) + return bch2_fs_read_write_early(c); + return 0; +} + +struct recovery_pass_fn { + int (*fn)(struct bch_fs *); + unsigned when; +}; + +static struct recovery_pass_fn recovery_pass_fns[] = { +#define x(_fn, _id, _when) { .fn = bch2_##_fn, .when = _when }, + BCH_RECOVERY_PASSES() +#undef x +}; + +u64 bch2_recovery_passes_to_stable(u64 v) +{ + static const u8 map[] = { +#define x(n, id, ...) [BCH_RECOVERY_PASS_##n] = BCH_RECOVERY_PASS_STABLE_##n, + BCH_RECOVERY_PASSES() +#undef x + }; + + u64 ret = 0; + for (unsigned i = 0; i < ARRAY_SIZE(map); i++) + if (v & BIT_ULL(i)) + ret |= BIT_ULL(map[i]); + return ret; +} + +u64 bch2_recovery_passes_from_stable(u64 v) +{ + static const u8 map[] = { +#define x(n, id, ...) [BCH_RECOVERY_PASS_STABLE_##n] = BCH_RECOVERY_PASS_##n, + BCH_RECOVERY_PASSES() +#undef x + }; + + u64 ret = 0; + for (unsigned i = 0; i < ARRAY_SIZE(map); i++) + if (v & BIT_ULL(i)) + ret |= BIT_ULL(map[i]); + return ret; +} + +static bool check_version_upgrade(struct bch_fs *c) +{ + unsigned latest_compatible = bch2_latest_compatible_version(c->sb.version); + unsigned latest_version = bcachefs_metadata_version_current; + unsigned old_version = c->sb.version_upgrade_complete ?: c->sb.version; + unsigned new_version = 0; + + if (old_version < bcachefs_metadata_required_upgrade_below) { + if (c->opts.version_upgrade == BCH_VERSION_UPGRADE_incompatible || + latest_compatible < bcachefs_metadata_required_upgrade_below) + new_version = latest_version; + else + new_version = latest_compatible; + } else { + switch (c->opts.version_upgrade) { + case BCH_VERSION_UPGRADE_compatible: + new_version = latest_compatible; + break; + case BCH_VERSION_UPGRADE_incompatible: + new_version = latest_version; + break; + case BCH_VERSION_UPGRADE_none: + new_version = old_version; + break; + } + } + + if (new_version > old_version) { + struct printbuf buf = PRINTBUF; + + if (old_version < bcachefs_metadata_required_upgrade_below) + prt_str(&buf, "Version upgrade required:\n"); + + if (old_version != c->sb.version) { + prt_str(&buf, "Version upgrade from "); + bch2_version_to_text(&buf, c->sb.version_upgrade_complete); + prt_str(&buf, " to "); + bch2_version_to_text(&buf, c->sb.version); + prt_str(&buf, " incomplete\n"); + } + + prt_printf(&buf, "Doing %s version upgrade from ", + BCH_VERSION_MAJOR(old_version) != BCH_VERSION_MAJOR(new_version) + ? "incompatible" : "compatible"); + bch2_version_to_text(&buf, old_version); + prt_str(&buf, " to "); + bch2_version_to_text(&buf, new_version); + prt_newline(&buf); + + struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext); + __le64 passes = ext->recovery_passes_required[0]; + bch2_sb_set_upgrade(c, old_version, new_version); + passes = ext->recovery_passes_required[0] & ~passes; + + if (passes) { + prt_str(&buf, " running recovery passes: "); + prt_bitflags(&buf, bch2_recovery_passes, + bch2_recovery_passes_from_stable(le64_to_cpu(passes))); + } + + bch_info(c, "%s", buf.buf); + + bch2_sb_upgrade(c, new_version); + + printbuf_exit(&buf); + return true; + } + + return false; +} + +u64 bch2_fsck_recovery_passes(void) +{ + u64 ret = 0; + + for (unsigned i = 0; i < ARRAY_SIZE(recovery_pass_fns); i++) + if (recovery_pass_fns[i].when & PASS_FSCK) + ret |= BIT_ULL(i); + return ret; +} + +static bool should_run_recovery_pass(struct bch_fs *c, enum bch_recovery_pass pass) +{ + struct recovery_pass_fn *p = recovery_pass_fns + pass; + + if (c->opts.norecovery && pass > BCH_RECOVERY_PASS_snapshots_read) + return false; + if (c->recovery_passes_explicit & BIT_ULL(pass)) + return true; + if ((p->when & PASS_FSCK) && c->opts.fsck) + return true; + if ((p->when & PASS_UNCLEAN) && !c->sb.clean) + return true; + if (p->when & PASS_ALWAYS) + return true; + return false; +} + +static int bch2_run_recovery_pass(struct bch_fs *c, enum bch_recovery_pass pass) +{ + struct recovery_pass_fn *p = recovery_pass_fns + pass; + int ret; + + if (!(p->when & PASS_SILENT)) + bch2_print(c, KERN_INFO bch2_log_msg(c, "%s..."), + bch2_recovery_passes[pass]); + ret = p->fn(c); + if (ret) + return ret; + if (!(p->when & PASS_SILENT)) + bch2_print(c, KERN_CONT " done\n"); + + return 0; +} + +static int bch2_run_recovery_passes(struct bch_fs *c) +{ + int ret = 0; + + while (c->curr_recovery_pass < ARRAY_SIZE(recovery_pass_fns)) { + if (should_run_recovery_pass(c, c->curr_recovery_pass)) { + unsigned pass = c->curr_recovery_pass; + + ret = bch2_run_recovery_pass(c, c->curr_recovery_pass); + if (bch2_err_matches(ret, BCH_ERR_restart_recovery) || + (ret && c->curr_recovery_pass < pass)) + continue; + if (ret) + break; + + c->recovery_passes_complete |= BIT_ULL(c->curr_recovery_pass); + } + c->curr_recovery_pass++; + c->recovery_pass_done = max(c->recovery_pass_done, c->curr_recovery_pass); + } + + return ret; +} + +int bch2_run_online_recovery_passes(struct bch_fs *c) +{ + int ret = 0; + + for (unsigned i = 0; i < ARRAY_SIZE(recovery_pass_fns); i++) { + struct recovery_pass_fn *p = recovery_pass_fns + i; + + if (!(p->when & PASS_ONLINE)) + continue; + + ret = bch2_run_recovery_pass(c, i); + if (bch2_err_matches(ret, BCH_ERR_restart_recovery)) { + i = c->curr_recovery_pass; + continue; + } + if (ret) + break; + } + + return ret; +} + 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; - bool write_sb = false; + u64 last_seq = 0, blacklist_seq, journal_seq; int ret = 0; - if (c->sb.clean) - clean = read_superblock_clean(c); - ret = PTR_ERR_OR_ZERO(clean); - if (ret) - goto err; + if (c->sb.clean) { + clean = bch2_read_superblock_clean(c); + ret = PTR_ERR_OR_ZERO(clean); + if (ret) + goto err; - if (c->sb.clean) bch_info(c, "recovering from clean shutdown, journal seq %llu", le64_to_cpu(clean->journal_seq)); - else + } else { bch_info(c, "recovering from unclean shutdown"); + } if (!(c->sb.features & (1ULL << BCH_FEATURE_new_extent_overwrite))) { bch_err(c, "feature new_extent_overwrite not set, filesystem no longer supported"); @@ -1047,40 +768,76 @@ int bch2_fs_recovery(struct bch_fs *c) goto err; } - if (!(c->sb.compat & (1ULL << BCH_COMPAT_bformat_overflow_done))) { - bch_err(c, "filesystem may have incompatible bkey formats; run fsck from the compat branch to fix"); + if (c->opts.fsck && c->opts.norecovery) { + bch_err(c, "cannot select both norecovery and fsck"); ret = -EINVAL; goto err; } - if (!(c->sb.features & (1ULL << BCH_FEATURE_alloc_v2))) { - bch_info(c, "alloc_v2 feature bit not set, fsck required"); - c->opts.fsck = true; - c->opts.fix_errors = FSCK_OPT_YES; - } + if (!(c->opts.nochanges && c->opts.norecovery)) { + mutex_lock(&c->sb_lock); + bool write_sb = false; - if (!c->replicas.entries || - c->opts.rebuild_replicas) { - bch_info(c, "building replicas info"); - set_bit(BCH_FS_REBUILD_REPLICAS, &c->flags); - } + struct bch_sb_field_ext *ext = + bch2_sb_field_get_minsize(&c->disk_sb, ext, sizeof(*ext) / sizeof(u64)); + if (!ext) { + ret = -BCH_ERR_ENOSPC_sb; + mutex_unlock(&c->sb_lock); + goto err; + } + + if (BCH_SB_HAS_TOPOLOGY_ERRORS(c->disk_sb.sb)) { + ext->recovery_passes_required[0] |= + cpu_to_le64(bch2_recovery_passes_to_stable(BIT_ULL(BCH_RECOVERY_PASS_check_topology))); + write_sb = true; + } + + u64 sb_passes = bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0])); + if (sb_passes) { + struct printbuf buf = PRINTBUF; + prt_str(&buf, "superblock requires following recovery passes to be run:\n "); + prt_bitflags(&buf, bch2_recovery_passes, sb_passes); + bch_info(c, "%s", buf.buf); + printbuf_exit(&buf); + } + + if (bch2_check_version_downgrade(c)) { + struct printbuf buf = PRINTBUF; + + prt_str(&buf, "Version downgrade required:\n"); + + __le64 passes = ext->recovery_passes_required[0]; + bch2_sb_set_downgrade(c, + BCH_VERSION_MINOR(bcachefs_metadata_version_current), + BCH_VERSION_MINOR(c->sb.version)); + passes = ext->recovery_passes_required[0] & ~passes; + if (passes) { + prt_str(&buf, " running recovery passes: "); + prt_bitflags(&buf, bch2_recovery_passes, + bch2_recovery_passes_from_stable(le64_to_cpu(passes))); + } - if (!c->opts.nochanges) { - if (c->sb.version < bcachefs_metadata_version_inode_backpointers) { - bch_info(c, "version prior to inode 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_subvol_dirent) { - 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"); - c->opts.version_upgrade = true; + bch_info(c, "%s", buf.buf); + printbuf_exit(&buf); + write_sb = true; } + + if (check_version_upgrade(c)) + write_sb = true; + + if (write_sb) + bch2_write_super(c); + + c->recovery_passes_explicit |= bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0])); + mutex_unlock(&c->sb_lock); } + if (c->opts.fsck && IS_ENABLED(CONFIG_BCACHEFS_DEBUG)) + c->recovery_passes_explicit |= BIT_ULL(BCH_RECOVERY_PASS_check_topology); + + if (c->opts.fsck) + set_bit(BCH_FS_fsck_running, &c->flags); + ret = bch2_blacklist_table_initialize(c); if (ret) { bch_err(c, "error initializing blacklist table"); @@ -1088,22 +845,31 @@ int bch2_fs_recovery(struct bch_fs *c) } if (!c->sb.clean || c->opts.fsck || c->opts.keep_journal) { - struct journal_replay *i; + struct genradix_iter iter; + struct journal_replay **i; - ret = bch2_journal_read(c, &c->journal_entries, - &blacklist_seq, &journal_seq); + bch_verbose(c, "starting journal read"); + ret = bch2_journal_read(c, &last_seq, &blacklist_seq, &journal_seq); if (ret) goto err; - list_for_each_entry_reverse(i, &c->journal_entries, list) - if (!i->ignore) { - last_journal_entry = &i->j; + /* + * 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; break; } if (mustfix_fsck_err_on(c->sb.clean && last_journal_entry && !journal_entry_empty(last_journal_entry), c, + clean_but_journal_not_empty, "filesystem marked clean but journal not empty")) { c->sb.compat &= ~(1ULL << BCH_COMPAT_alloc_info); SET_BCH_SB_CLEAN(c->disk_sb.sb, false); @@ -1111,18 +877,33 @@ 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; + fsck_err_on(!c->sb.clean, c, + dirty_but_no_journal_entries, + "no journal entries found"); + if (clean) + goto use_clean; + + genradix_for_each_reverse(&c->journal_entries, iter, i) + if (*i) { + last_journal_entry = &(*i)->j; + (*i)->ignore = false; + /* + * This was probably a NO_FLUSH entry, + * so last_seq was garbage - but we know + * we're only using a single journal + * entry, set it here: + */ + (*i)->j.last_seq = (*i)->j.seq; + break; + } } - c->journal_keys = journal_keys_sort(&c->journal_entries); - if (!c->journal_keys.d) { - ret = -ENOMEM; + ret = bch2_journal_keys_sort(c); + if (ret) goto err; - } if (c->sb.clean && last_journal_entry) { - ret = verify_superblock_clean(c, &clean, + ret = bch2_verify_superblock_clean(c, &clean, last_journal_entry); if (ret) goto err; @@ -1131,13 +912,16 @@ 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; } + c->journal_replay_seq_start = last_seq; + c->journal_replay_seq_end = blacklist_seq - 1; + if (c->opts.reconstruct_alloc) { c->sb.compat &= ~(1ULL << BCH_COMPAT_alloc_info); drop_alloc_keys(&c->journal_keys); @@ -1145,7 +929,7 @@ use_clean: zero_out_btree_mem_ptr(&c->journal_keys); - ret = journal_replay_early(c, clean, &c->journal_entries); + ret = journal_replay_early(c, clean); if (ret) goto err; @@ -1160,7 +944,9 @@ use_clean: journal_seq += 8; if (blacklist_seq != journal_seq) { - ret = bch2_journal_seq_blacklist_add(c, + ret = bch2_journal_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"); @@ -1168,49 +954,14 @@ use_clean: } } - ret = bch2_fs_journal_start(&c->journal, journal_seq, - &c->journal_entries); - if (ret) - goto err; - - ret = read_btree_roots(c); - if (ret) - goto err; - - bch_verbose(c, "starting alloc read"); - err = "error reading allocation information"; - ret = bch2_alloc_read(c); + ret = bch2_journal_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; - bch_verbose(c, "alloc read done"); - - bch_verbose(c, "starting stripes_read"); - err = "error reading stripes"; - ret = bch2_stripes_read(c); - if (ret) - goto err; - bch_verbose(c, "stripes_read done"); - - set_bit(BCH_FS_ALLOC_READ_DONE, &c->flags); - 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)) { - bool metadata_only = c->opts.norecovery; - - bch_info(c, "starting mark and sweep"); - err = "error in mark and sweep"; - ret = bch2_gc(c, true, metadata_only); - if (ret) - goto err; - bch_verbose(c, "mark and sweep done"); - } - - bch2_stripes_heap_start(c); - - clear_bit(BCH_FS_REBUILD_REPLICAS, &c->flags); - set_bit(BCH_FS_INITIAL_GC_DONE, &c->flags); + if (c->opts.reconstruct_alloc) + bch2_journal_log_msg(c, "dropping alloc info"); /* * Skip past versions that might have possibly been used (as nonces), @@ -1219,72 +970,39 @@ use_clean: if (c->sb.encryption_type && !c->sb.clean) atomic64_add(1 << 16, &c->key_version); - if (c->opts.norecovery) - goto out; + ret = read_btree_roots(c); + if (ret) + goto err; - bch_verbose(c, "starting journal replay"); - err = "journal replay failed"; - ret = bch2_journal_replay(c, c->journal_keys); + ret = bch2_run_recovery_passes(c); if (ret) goto err; - bch_verbose(c, "journal replay done"); - if (test_bit(BCH_FS_NEED_ALLOC_WRITE, &c->flags) && - !c->opts.nochanges) { - /* - * note that even when filesystem was clean there might be work - * to do here, if we ran gc (because of fsck) which recalculated - * oldest_gen: - */ - bch_verbose(c, "writing allocation info"); - err = "error writing out alloc info"; - ret = bch2_alloc_write(c, BTREE_INSERT_LAZY_RW); - if (ret) { - bch_err(c, "error writing alloc info"); - goto err; - } - bch_verbose(c, "alloc write done"); - } + clear_bit(BCH_FS_fsck_running, &c->flags); - if (c->sb.version < bcachefs_metadata_version_snapshot_2) { - bch2_fs_lazy_rw(c); + /* If we fixed errors, verify that fs is actually clean now: */ + if (IS_ENABLED(CONFIG_BCACHEFS_DEBUG) && + test_bit(BCH_FS_errors_fixed, &c->flags) && + !test_bit(BCH_FS_errors_not_fixed, &c->flags) && + !test_bit(BCH_FS_error, &c->flags)) { + bch2_flush_fsck_errs(c); - err = "error creating root snapshot node"; - ret = bch2_fs_initialize_subvolumes(c); - if (ret) - goto err; - } + bch_info(c, "Fixed errors, running fsck a second time to verify fs is clean"); + clear_bit(BCH_FS_errors_fixed, &c->flags); - 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"); + c->curr_recovery_pass = BCH_RECOVERY_PASS_check_alloc_info; - if (c->sb.version < bcachefs_metadata_version_snapshot_2) { - /* set bi_subvol on root inode */ - err = "error upgrade root inode for subvolumes"; - ret = bch2_trans_do(c, NULL, NULL, BTREE_INSERT_LAZY_RW, - bch2_fs_upgrade_for_subvolumes(&trans)); + ret = bch2_run_recovery_passes(c); if (ret) goto err; - } - if (c->opts.fsck) { - bch_info(c, "starting fsck"); - err = "error in fsck"; - ret = bch2_fsck_full(c); - if (ret) - goto err; - bch_verbose(c, "fsck done"); - } else if (!c->sb.clean) { - bch_verbose(c, "checking for deleted inodes"); - err = "error in recovery"; - ret = bch2_fsck_walk_inodes_only(c); - if (ret) - goto err; - bch_verbose(c, "check inodes done"); + if (test_bit(BCH_FS_errors_fixed, &c->flags) || + test_bit(BCH_FS_errors_not_fixed, &c->flags)) { + bch_err(c, "Second fsck run was not clean"); + set_bit(BCH_FS_errors_not_fixed, &c->flags); + } + + set_bit(BCH_FS_errors_fixed, &c->flags); } if (enabled_qtypes(c)) { @@ -1295,38 +1013,34 @@ use_clean: bch_verbose(c, "quotas done"); } - if (!(c->sb.compat & (1ULL << BCH_COMPAT_extents_above_btree_updates_done)) || - !(c->sb.compat & (1ULL << BCH_COMPAT_bformat_overflow_done))) { - struct bch_move_stats stats; - - bch_move_stats_init(&stats, "recovery"); - - bch_info(c, "scanning for old btree nodes"); - ret = bch2_fs_read_write(c); - if (ret) - goto err; - - ret = bch2_scan_old_btree_nodes(c, &stats); - if (ret) - goto err; - bch_info(c, "scanning for old btree nodes done"); - } - mutex_lock(&c->sb_lock); - 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); + bool write_sb = false; + + if (BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb) != le16_to_cpu(c->disk_sb.sb->version)) { + SET_BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb, le16_to_cpu(c->disk_sb.sb->version)); write_sb = true; } - if (!test_bit(BCH_FS_ERROR, &c->flags)) { + if (!test_bit(BCH_FS_error, &c->flags) && + !(c->disk_sb.sb->compat[0] & cpu_to_le64(1ULL << BCH_COMPAT_alloc_info))) { c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_alloc_info); write_sb = true; } + if (!test_bit(BCH_FS_error, &c->flags)) { + struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext); + if (ext && + (!bch2_is_zero(ext->recovery_passes_required, sizeof(ext->recovery_passes_required)) || + !bch2_is_zero(ext->errors_silent, sizeof(ext->errors_silent)))) { + memset(ext->recovery_passes_required, 0, sizeof(ext->recovery_passes_required)); + memset(ext->errors_silent, 0, sizeof(ext->errors_silent)); + write_sb = true; + } + } + if (c->opts.fsck && - !test_bit(BCH_FS_ERROR, &c->flags) && - !test_bit(BCH_FS_ERRORS_NOT_FIXED, &c->flags)) { + !test_bit(BCH_FS_error, &c->flags) && + !test_bit(BCH_FS_errors_not_fixed, &c->flags)) { SET_BCH_SB_HAS_ERRORS(c->disk_sb.sb, 0); SET_BCH_SB_HAS_TOPOLOGY_ERRORS(c->disk_sb.sb, 0); write_sb = true; @@ -1336,24 +1050,45 @@ use_clean: bch2_write_super(c); mutex_unlock(&c->sb_lock); + if (!(c->sb.compat & (1ULL << BCH_COMPAT_extents_above_btree_updates_done)) || + c->sb.version_min < bcachefs_metadata_version_btree_ptr_sectors_written) { + struct bch_move_stats stats; + + bch2_move_stats_init(&stats, "recovery"); + + struct printbuf buf = PRINTBUF; + bch2_version_to_text(&buf, c->sb.version_min); + bch_info(c, "scanning for old btree nodes: min_version %s", buf.buf); + printbuf_exit(&buf); + + ret = bch2_fs_read_write_early(c) ?: + bch2_scan_old_btree_nodes(c, &stats); + if (ret) + goto err; + bch_info(c, "scanning for old btree nodes done"); + } + if (c->journal_seq_blacklist_table && c->journal_seq_blacklist_table->nr > 128) queue_work(system_long_wq, &c->journal_seq_blacklist_gc_work); ret = 0; out: - set_bit(BCH_FS_FSCK_DONE, &c->flags); bch2_flush_fsck_errs(c); - if (!c->opts.keep_journal) { - bch2_journal_keys_free(&c->journal_keys); - bch2_journal_entries_free(&c->journal_entries); - } + if (!c->opts.keep_journal && + test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags)) + bch2_journal_keys_put_initial(c); kfree(clean); - if (ret) - bch_err(c, "Error in recovery: %s (%i)", err, ret); - else - bch_verbose(c, "ret %i", ret); + + if (!ret && + test_bit(BCH_FS_need_delete_dead_snapshots, &c->flags) && + !c->opts.nochanges) { + bch2_fs_read_write_early(c); + bch2_delete_dead_snapshots_async(c); + } + + bch_err_fn(c, ret); return ret; err: fsck_err: @@ -1366,10 +1101,6 @@ int bch2_fs_initialize(struct bch_fs *c) struct bch_inode_unpacked root_inode, lostfound_inode; struct bkey_inode_buf packed_inode; struct qstr lostfound = QSTR("lost+found"); - const char *err = "cannot allocate memory"; - struct bch_dev *ca; - LIST_HEAD(journal); - unsigned i; int ret; bch_notice(c, "initializing new filesystem"); @@ -1378,42 +1109,35 @@ 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->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); + bch2_check_version_downgrade(c); + + if (c->opts.version_upgrade != BCH_VERSION_UPGRADE_none) { + bch2_sb_upgrade(c, bcachefs_metadata_version_current); + SET_BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb, bcachefs_metadata_version_current); bch2_write_super(c); } - - for_each_online_member(ca, c, i) - bch2_mark_dev_superblock(c, ca, 0); mutex_unlock(&c->sb_lock); - set_bit(BCH_FS_ALLOC_READ_DONE, &c->flags); - set_bit(BCH_FS_INITIAL_GC_DONE, &c->flags); + c->curr_recovery_pass = ARRAY_SIZE(recovery_pass_fns); + set_bit(BCH_FS_may_go_rw, &c->flags); - for (i = 0; i < BTREE_ID_NR; i++) + for (unsigned i = 0; i < BTREE_ID_NR; i++) bch2_btree_root_alloc(c, i); - set_bit(BCH_FS_BTREE_INTERIOR_REPLAY_DONE, &c->flags); - set_bit(JOURNAL_RECLAIM_STARTED, &c->journal.flags); + for_each_member_device(c, ca) + bch2_dev_usage_init(ca); - err = "unable to allocate journal buckets"; - for_each_online_member(ca, c, i) { - ret = bch2_dev_journal_alloc(ca); - if (ret) { - percpu_ref_put(&ca->io_ref); - goto err; - } - } + ret = bch2_fs_journal_alloc(c); + if (ret) + goto err; /* * journal_res_get() will crash if called before this has * set up the journal.pin FIFO and journal.cur pointer: */ - bch2_fs_journal_start(&c->journal, 1, &journal); + bch2_fs_journal_start(&c->journal, 1); bch2_journal_set_replay_done(&c->journal); - err = "error going read-write"; ret = bch2_fs_read_write_early(c); if (ret) goto err; @@ -1422,55 +1146,54 @@ int bch2_fs_initialize(struct bch_fs *c) * Write out the superblock and journal buckets, now that we can do * btree updates */ - err = "error marking superblock and journal"; - for_each_member_device(ca, c, i) { - ret = bch2_trans_mark_dev_sb(c, ca); - if (ret) { - percpu_ref_put(&ca->ref); - goto err; - } - } + bch_verbose(c, "marking superblocks"); + ret = bch2_trans_mark_dev_sbs(c); + bch_err_msg(c, ret, "marking superblocks"); + if (ret) + goto err; - err = "error creating root snapshot node"; - ret = bch2_fs_initialize_subvolumes(c); + for_each_online_member(c, ca) + ca->new_fs_bucket_idx = 0; + + ret = bch2_fs_freespace_init(c); + if (ret) + goto err; + + ret = bch2_initialize_subvolumes(c); if (ret) goto err; bch_verbose(c, "reading snapshots table"); - err = "error reading snapshots table"; - ret = bch2_fs_snapshots_start(c); + ret = bch2_snapshots_read(c); if (ret) 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"; - ret = bch2_btree_insert(c, BTREE_ID_inodes, - &packed_inode.inode.k_i, - NULL, NULL, 0); + ret = bch2_btree_insert(c, BTREE_ID_inodes, &packed_inode.inode.k_i, NULL, 0); + bch_err_msg(c, ret, "creating root directory"); if (ret) goto err; bch2_inode_init_early(c, &lostfound_inode); - err = "error creating lost+found"; ret = bch2_trans_do(c, NULL, NULL, 0, - bch2_create_trans(&trans, + bch2_create_trans(trans, BCACHEFS_ROOT_SUBVOL_INUM, &root_inode, &lostfound_inode, &lostfound, 0, 0, S_IFDIR|0700, 0, NULL, NULL, (subvol_inum) { 0 }, 0)); - if (ret) { - bch_err(c, "error creating lost+found"); + bch_err_msg(c, ret, "creating lost+found"); + if (ret) goto err; - } + + c->recovery_pass_done = ARRAY_SIZE(recovery_pass_fns) - 1; if (enabled_qtypes(c)) { ret = bch2_fs_quota_read(c); @@ -1478,8 +1201,8 @@ int bch2_fs_initialize(struct bch_fs *c) goto err; } - err = "error writing first journal entry"; ret = bch2_journal_flush(&c->journal); + bch_err_msg(c, ret, "writing first journal entry"); if (ret) goto err; @@ -1492,6 +1215,6 @@ int bch2_fs_initialize(struct bch_fs *c) return 0; err: - pr_err("Error initializing new filesystem: %s (%i)", err, ret); + bch_err_fn(c, ret); return ret; }