X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Frecovery.c;h=e4983d144483072af1ece1e8dc69fc07b752f96a;hb=17d1c4f4fe29453fbb4087eb5de8aa6f9e6605c7;hp=8c1aa6d2ce45f02f8cea4c782a45be8aa2237b6e;hpb=71111771690f244d13650c73d52ff601ad914d95;p=bcachefs-tools-debian

diff --git a/libbcachefs/recovery.c b/libbcachefs/recovery.c
index 8c1aa6d..e4983d1 100644
--- a/libbcachefs/recovery.c
+++ b/libbcachefs/recovery.c
@@ -228,7 +228,7 @@ int bch2_journal_key_insert_take(struct bch_fs *c, enum btree_id id,
 		if (!new_keys.d) {
 			bch_err(c, "%s: error allocating new key array (size %zu)",
 				__func__, new_keys.size);
-			return -ENOMEM;
+			return -BCH_ERR_ENOMEM_journal_key_insert;
 		}
 
 		/* Since @keys was full, there was no gap: */
@@ -266,7 +266,7 @@ int bch2_journal_key_insert(struct bch_fs *c, enum btree_id id,
 
 	n = kmalloc(bkey_bytes(&k->k), GFP_KERNEL);
 	if (!n)
-		return -ENOMEM;
+		return -BCH_ERR_ENOMEM_journal_key_insert;
 
 	bkey_copy(n, k);
 	ret = bch2_journal_key_insert_take(c, id, level, n);
@@ -476,15 +476,34 @@ void bch2_journal_keys_free(struct journal_keys *keys)
 	keys->nr = keys->gap = keys->size = 0;
 }
 
+static void __journal_keys_sort(struct journal_keys *keys)
+{
+	struct journal_key *src, *dst;
+
+	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_eq(src[0].k->k.p, src[1].k->k.p))
+			src++;
+
+		*dst++ = *src++;
+	}
+
+	keys->nr = dst - keys->d;
+}
+
 static int journal_keys_sort(struct bch_fs *c)
 {
 	struct genradix_iter iter;
 	struct journal_replay *i, **_i;
 	struct jset_entry *entry;
-	struct bkey_i *k, *_n;
+	struct bkey_i *k;
 	struct journal_keys *keys = &c->journal_keys;
-	struct journal_key *src, *dst;
-	size_t nr_keys = 0;
+	size_t nr_keys = 0, nr_read = 0;
 
 	genradix_for_each(&c->journal_entries, iter, _i) {
 		i = *_i;
@@ -492,7 +511,7 @@ static int journal_keys_sort(struct bch_fs *c)
 		if (!i || i->ignore)
 			continue;
 
-		for_each_jset_key(k, _n, entry, &i->j)
+		for_each_jset_key(k, entry, &i->j)
 			nr_keys++;
 	}
 
@@ -502,8 +521,21 @@ static int journal_keys_sort(struct bch_fs *c)
 	keys->size = roundup_pow_of_two(nr_keys);
 
 	keys->d = kvmalloc_array(keys->size, sizeof(keys->d[0]), GFP_KERNEL);
-	if (!keys->d)
-		return -ENOMEM;
+	if (!keys->d) {
+		bch_err(c, "Failed to allocate buffer for sorted journal keys (%zu keys); trying slowpath",
+			nr_keys);
+
+		do {
+			keys->size >>= 1;
+			keys->d = kvmalloc_array(keys->size, sizeof(keys->d[0]), GFP_KERNEL);
+		} while (!keys->d && keys->size > nr_keys / 8);
+
+		if (!keys->d) {
+			bch_err(c, "Failed to allocate %zu size buffer for sorted journal keys; exiting",
+				keys->size);
+			return -BCH_ERR_ENOMEM_journal_keys_sort;
+		}
+	}
 
 	genradix_for_each(&c->journal_entries, iter, _i) {
 		i = *_i;
@@ -511,7 +543,19 @@ static int journal_keys_sort(struct bch_fs *c)
 		if (!i || i->ignore)
 			continue;
 
-		for_each_jset_key(k, _n, entry, &i->j)
+		cond_resched();
+
+		for_each_jset_key(k, entry, &i->j) {
+			if (keys->nr == keys->size) {
+				__journal_keys_sort(keys);
+
+				if (keys->nr > keys->size * 7 / 8) {
+					bch_err(c, "Too many journal keys for slowpath; have %zu compacted, buf size %zu, processed %zu/%zu",
+						keys->nr, keys->size, nr_read, nr_keys);
+					return -BCH_ERR_ENOMEM_journal_keys_sort;
+				}
+			}
+
 			keys->d[keys->nr++] = (struct journal_key) {
 				.btree_id	= entry->btree_id,
 				.level		= entry->level,
@@ -519,23 +563,15 @@ static int journal_keys_sort(struct bch_fs *c)
 				.journal_seq	= le64_to_cpu(i->j.seq),
 				.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_eq(src[0].k->k.p, src[1].k->k.p))
-			src++;
 
-		*dst++ = *src++;
+			nr_read++;
+		}
 	}
 
-	keys->nr = dst - keys->d;
+	__journal_keys_sort(keys);
 	keys->gap = keys->nr;
+
+	bch_verbose(c, "Journal keys: %zu read, %zu after sorting and compacting", nr_keys, keys->nr);
 	return 0;
 }
 
@@ -601,7 +637,7 @@ static int bch2_journal_replay(struct bch_fs *c, u64 start_seq, u64 end_seq)
 
 	keys_sorted = kvmalloc_array(sizeof(*keys_sorted), keys->nr, GFP_KERNEL);
 	if (!keys_sorted)
-		return -ENOMEM;
+		return -BCH_ERR_ENOMEM_journal_replay;
 
 	for (i = 0; i < keys->nr; i++)
 		keys_sorted[i] = &keys->d[i];
@@ -611,8 +647,8 @@ static int bch2_journal_replay(struct bch_fs *c, u64 start_seq, u64 end_seq)
 	     journal_sort_seq_cmp, NULL);
 
 	if (keys->nr) {
-		ret = bch2_fs_log_msg(c, "Starting journal replay (%zu keys in entries %llu-%llu)",
-				      keys->nr, start_seq, end_seq);
+		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;
 	}
@@ -646,7 +682,7 @@ static int bch2_journal_replay(struct bch_fs *c, u64 start_seq, u64 end_seq)
 	ret = bch2_journal_error(j);
 
 	if (keys->nr && !ret)
-		bch2_fs_log_msg(c, "journal replay finished");
+		bch2_journal_log_msg(c, "journal replay finished");
 err:
 	kvfree(keys_sorted);
 	return ret;
@@ -871,7 +907,7 @@ static int verify_superblock_clean(struct bch_fs *c,
 				    IS_ERR(k1) ||
 				    IS_ERR(k2) ||
 				    k1->k.u64s != k2->k.u64s ||
-				    memcmp(k1, k2, bkey_bytes(k1)) ||
+				    memcmp(k1, k2, bkey_bytes(&k1->k)) ||
 				    l1 != l2, c,
 			"superblock btree root %u doesn't match journal after clean shutdown\n"
 			"sb:      l=%u %s\n"
@@ -905,7 +941,7 @@ static struct bch_sb_field_clean *read_superblock_clean(struct bch_fs *c)
 			GFP_KERNEL);
 	if (!clean) {
 		mutex_unlock(&c->sb_lock);
-		return ERR_PTR(-ENOMEM);
+		return ERR_PTR(-BCH_ERR_ENOMEM_read_superblock_clean);
 	}
 
 	ret = bch2_sb_clean_validate_late(c, clean, READ);
@@ -969,30 +1005,45 @@ static int read_btree_roots(struct bch_fs *c)
 				   ? 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;
 }
 
 static int bch2_fs_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);
+	ret = bch2_btree_insert(c, BTREE_ID_snapshot_trees,
+				&root_tree.k_i,
+				NULL, NULL, 0);
+
 	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.tree	= cpu_to_le32(1);
 	SET_BCH_SNAPSHOT_SUBVOL(&root_snapshot.v, true);
 
 	ret = bch2_btree_insert(c, BTREE_ID_snapshots,
@@ -1023,16 +1074,15 @@ static int bch2_fs_upgrade_for_subvolumes(struct btree_trans *trans)
 	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(trans->c, "root inode not found");
-		ret = -ENOENT;
+		ret = -BCH_ERR_ENOENT_inode;
 		goto err;
 	}
 
@@ -1094,14 +1144,15 @@ int bch2_fs_recovery(struct bch_fs *c)
 	}
 
 	if (!c->opts.nochanges) {
-		if (c->sb.version < bcachefs_metadata_version_backpointers) {
-			bch_info(c, "version prior to backpointers, upgrade and fsck required");
+		if (c->sb.version < bcachefs_metadata_required_upgrade_below) {
+			bch_info(c, "version %s (%u) prior to %s (%u), upgrade and fsck required",
+				 bch2_metadata_versions[c->sb.version],
+				 c->sb.version,
+				 bch2_metadata_versions[bcachefs_metadata_required_upgrade_below],
+				 bcachefs_metadata_required_upgrade_below);
 			c->opts.version_upgrade	= true;
 			c->opts.fsck		= true;
 			c->opts.fix_errors	= FSCK_OPT_YES;
-		} else if (c->sb.version < bcachefs_metadata_version_inode_v3) {
-			bch_info(c, "version prior to inode_v3, upgrade required");
-			c->opts.version_upgrade	= true;
 		}
 	}
 
@@ -1204,8 +1255,8 @@ use_clean:
 		journal_seq += 8;
 
 	if (blacklist_seq != journal_seq) {
-		ret =   bch2_fs_log_msg(c, "blacklisting entries %llu-%llu",
-					blacklist_seq, journal_seq) ?:
+		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) {
@@ -1214,12 +1265,15 @@ use_clean:
 		}
 	}
 
-	ret =   bch2_fs_log_msg(c, "starting journal at entry %llu, replaying %llu-%llu",
-				journal_seq, last_seq, blacklist_seq - 1) ?:
+	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;
 
+	if (c->opts.reconstruct_alloc)
+		bch2_journal_log_msg(c, "dropping alloc info");
+
 	/*
 	 * Skip past versions that might have possibly been used (as nonces),
 	 * but hadn't had their pointers written:
@@ -1251,7 +1305,19 @@ use_clean:
 		goto err;
 	bch_verbose(c, "stripes_read done");
 
-	bch2_stripes_heap_start(c);
+	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) {
 		bool metadata_only = c->opts.norecovery;
@@ -1265,20 +1331,6 @@ use_clean:
 
 		set_bit(BCH_FS_INITIAL_GC_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");
-
 		set_bit(BCH_FS_MAY_GO_RW, &c->flags);
 
 		bch_info(c, "starting journal replay, %zu keys", c->journal_keys.nr);
@@ -1346,20 +1398,6 @@ use_clean:
 		if (c->opts.norecovery)
 			goto out;
 
-		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");
-
 		set_bit(BCH_FS_MAY_GO_RW, &c->flags);
 
 		bch_verbose(c, "starting journal replay, %zu keys", c->journal_keys.nr);
@@ -1635,6 +1673,6 @@ 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;
 }