Update bcachefs sources to 9b4ab159abcd bcachefs: Read retries are after checksum errors now REQ_FUA

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

6 files changed, 82 insertions, 39 deletions

diff --git a/.bcachefs_revision b/.bcachefs_revision
index d1c1f3ca..f08feb4a 100644
--- a/.bcachefs_revision
+++ b/.bcachefs_revision
@@ -1 +1 @@
-f0ebca18293c398beddfd80e2ec3575401e7abc4
+9b4ab159abcd84cf0c25ee851dda8c40baffecc8
diff --git a/libbcachefs/btree_gc.c b/libbcachefs/btree_gc.c
index b4332338..92ae3173 100644
--- a/libbcachefs/btree_gc.c
+++ b/libbcachefs/btree_gc.c
@@ -48,6 +48,27 @@
 #define DROP_PREV_NODE		11
 #define DID_FILL_FROM_SCAN	12
 
+/*
+ * Returns true if it's a btree we can easily reconstruct, or otherwise won't
+ * cause data loss if it's missing:
+ */
+static bool btree_id_important(enum btree_id btree)
+{
+	if (btree_id_is_alloc(btree))
+		return false;
+
+	switch (btree) {
+	case BTREE_ID_quotas:
+	case BTREE_ID_snapshot_trees:
+	case BTREE_ID_logged_ops:
+	case BTREE_ID_rebalance_work:
+	case BTREE_ID_subvolume_children:
+		return false;
+	default:
+		return true;
+	}
+}
+
 static const char * const bch2_gc_phase_strs[] = {
 #define x(n)	#n,
 	GC_PHASES()
@@ -525,8 +546,10 @@ reconstruct_root:
 			r->error = 0;
 
 			if (!bch2_btree_has_scanned_nodes(c, i)) {
-				mustfix_fsck_err(trans, btree_root_unreadable_and_scan_found_nothing,
-						 "no nodes found for btree %s, continue?", buf.buf);
+				__fsck_err(trans,
+					   FSCK_CAN_FIX|(!btree_id_important(i) ? FSCK_AUTOFIX : 0),
+					   btree_root_unreadable_and_scan_found_nothing,
+					   "no nodes found for btree %s, continue?", buf.buf);
 				bch2_btree_root_alloc_fake_trans(trans, i, 0);
 			} else {
 				bch2_btree_root_alloc_fake_trans(trans, i, 1);
diff --git a/libbcachefs/btree_journal_iter.c b/libbcachefs/btree_journal_iter.c
index 7d6c971d..ade3b5ad 100644
--- a/libbcachefs/btree_journal_iter.c
+++ b/libbcachefs/btree_journal_iter.c
@@ -288,7 +288,7 @@ int bch2_journal_key_insert_take(struct bch_fs *c, enum btree_id id,
 			.size			= max_t(size_t, keys->size, 8) * 2,
 		};
 
-		new_keys.data = kvmalloc_array(new_keys.size, sizeof(new_keys.data[0]), GFP_KERNEL);
+		new_keys.data = bch2_kvmalloc(new_keys.size * sizeof(new_keys.data[0]), GFP_KERNEL);
 		if (!new_keys.data) {
 			bch_err(c, "%s: error allocating new key array (size %zu)",
 				__func__, new_keys.size);
diff --git a/libbcachefs/btree_update_interior.c b/libbcachefs/btree_update_interior.c
index fcca8b01..3d25c2be 100644
--- a/libbcachefs/btree_update_interior.c
+++ b/libbcachefs/btree_update_interior.c
@@ -1396,7 +1396,7 @@ static void bch2_insert_fixup_btree_ptr(struct btree_update *as,
 	printbuf_exit(&buf);
 }
 
-static void
+static int
 bch2_btree_insert_keys_interior(struct btree_update *as,
 				struct btree_trans *trans,
 				struct btree_path *path,
@@ -1418,7 +1418,8 @@ bch2_btree_insert_keys_interior(struct btree_update *as,
 	     insert = bkey_next(insert))
 		bch2_insert_fixup_btree_ptr(as, trans, path, b, &node_iter, insert);
 
-	if (bch2_btree_node_check_topology(trans, b)) {
+	int ret = bch2_btree_node_check_topology(trans, b);
+	if (ret) {
 		struct printbuf buf = PRINTBUF;
 
 		for (struct bkey_i *k = keys->keys;
@@ -1428,11 +1429,15 @@ bch2_btree_insert_keys_interior(struct btree_update *as,
 			prt_newline(&buf);
 		}
 
-		panic("%s(): check_topology error: inserted keys\n%s", __func__, buf.buf);
+		bch2_fs_fatal_error(as->c, "%ps -> %s(): check_topology error %s: inserted keys\n%s",
+				    (void *) _RET_IP_, __func__, bch2_err_str(ret), buf.buf);
+		dump_stack();
+		return ret;
 	}
 
 	memmove_u64s_down(keys->keys, insert, keys->top_p - insert->_data);
 	keys->top_p -= insert->_data - keys->keys_p;
+	return 0;
 }
 
 static bool key_deleted_in_insert(struct keylist *insert_keys, struct bpos pos)
@@ -1566,11 +1571,11 @@ static void __btree_split_node(struct btree_update *as,
  * nodes that were coalesced, and thus in the middle of a child node post
  * coalescing:
  */
-static void btree_split_insert_keys(struct btree_update *as,
-				    struct btree_trans *trans,
-				    btree_path_idx_t path_idx,
-				    struct btree *b,
-				    struct keylist *keys)
+static int btree_split_insert_keys(struct btree_update *as,
+				   struct btree_trans *trans,
+				   btree_path_idx_t path_idx,
+				   struct btree *b,
+				   struct keylist *keys)
 {
 	struct btree_path *path = trans->paths + path_idx;
 
@@ -1580,8 +1585,12 @@ static void btree_split_insert_keys(struct btree_update *as,
 
 		bch2_btree_node_iter_init(&node_iter, b, &bch2_keylist_front(keys)->k.p);
 
-		bch2_btree_insert_keys_interior(as, trans, path, b, node_iter, keys);
+		int ret = bch2_btree_insert_keys_interior(as, trans, path, b, node_iter, keys);
+		if (ret)
+			return ret;
 	}
+
+	return 0;
 }
 
 static int btree_split(struct btree_update *as, struct btree_trans *trans,
@@ -1614,8 +1623,10 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans,
 		__btree_split_node(as, trans, b, n, keys);
 
 		if (keys) {
-			btree_split_insert_keys(as, trans, path, n1, keys);
-			btree_split_insert_keys(as, trans, path, n2, keys);
+			ret =   btree_split_insert_keys(as, trans, path, n1, keys) ?:
+				btree_split_insert_keys(as, trans, path, n2, keys);
+			if (ret)
+				goto err;
 			BUG_ON(!bch2_keylist_empty(keys));
 		}
 
@@ -1661,7 +1672,9 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans,
 			n3->sib_u64s[0] = U16_MAX;
 			n3->sib_u64s[1] = U16_MAX;
 
-			btree_split_insert_keys(as, trans, path, n3, &as->parent_keys);
+			ret = btree_split_insert_keys(as, trans, path, n3, &as->parent_keys);
+			if (ret)
+				goto err;
 		}
 	} else {
 		trace_and_count(c, btree_node_compact, trans, b);
@@ -1669,7 +1682,9 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans,
 		n1 = bch2_btree_node_alloc_replacement(as, trans, b);
 
 		if (keys) {
-			btree_split_insert_keys(as, trans, path, n1, keys);
+			ret = btree_split_insert_keys(as, trans, path, n1, keys);
+			if (ret)
+				goto err;
 			BUG_ON(!bch2_keylist_empty(keys));
 		}
 
@@ -1816,15 +1831,15 @@ static int bch2_btree_insert_node(struct btree_update *as, struct btree_trans *t
 		goto split;
 	}
 
-	ret = bch2_btree_node_check_topology(trans, b);
+
+	ret =   bch2_btree_node_check_topology(trans, b) ?:
+		bch2_btree_insert_keys_interior(as, trans, path, b,
+					path->l[b->c.level].iter, keys);
 	if (ret) {
 		bch2_btree_node_unlock_write(trans, path, b);
 		return ret;
 	}
 
-	bch2_btree_insert_keys_interior(as, trans, path, b,
-					path->l[b->c.level].iter, keys);
-
 	trans_for_each_path_with_node(trans, b, linked, i)
 		bch2_btree_node_iter_peek(&linked->l[b->c.level].iter, b);
 
diff --git a/libbcachefs/error.c b/libbcachefs/error.c
index dc3f99dc..d7bc70fd 100644
--- a/libbcachefs/error.c
+++ b/libbcachefs/error.c
@@ -526,7 +526,9 @@ int __bch2_fsck_err(struct bch_fs *c,
 	} else if (!test_bit(BCH_FS_fsck_running, &c->flags)) {
 		if (c->opts.errors != BCH_ON_ERROR_continue ||
 		    !(flags & (FSCK_CAN_FIX|FSCK_CAN_IGNORE))) {
-			prt_str(out, ", shutting down");
+			prt_str_indented(out, ", shutting down\n"
+					 "error not marked as autofix and not in fsck\n"
+					 "run fsck, and forward to devs so error can be marked for self-healing");
 			inconsistent = true;
 			print = true;
 			ret = -BCH_ERR_fsck_errors_not_fixed;
diff --git a/libbcachefs/super.c b/libbcachefs/super.c
index 7102a2bf..0f9c38f8 100644
--- a/libbcachefs/super.c
+++ b/libbcachefs/super.c
@@ -1894,6 +1894,7 @@ int bch2_dev_add(struct bch_fs *c, const char *path)
 		goto err_unlock;
 	}
 	unsigned dev_idx = ret;
+	ret = 0;
 
 	/* success: */
 
@@ -1913,27 +1914,29 @@ int bch2_dev_add(struct bch_fs *c, const char *path)
 	bch2_write_super(c);
 	mutex_unlock(&c->sb_lock);
 
-	ret = bch2_dev_usage_init(ca, false);
-	if (ret)
-		goto err_late;
+	if (test_bit(BCH_FS_started, &c->flags)) {
+		ret = bch2_dev_usage_init(ca, false);
+		if (ret)
+			goto err_late;
 
-	ret = bch2_trans_mark_dev_sb(c, ca, BTREE_TRIGGER_transactional);
-	bch_err_msg(ca, ret, "marking new superblock");
-	if (ret)
-		goto err_late;
+		ret = bch2_trans_mark_dev_sb(c, ca, BTREE_TRIGGER_transactional);
+		bch_err_msg(ca, ret, "marking new superblock");
+		if (ret)
+			goto err_late;
 
-	ret = bch2_fs_freespace_init(c);
-	bch_err_msg(ca, ret, "initializing free space");
-	if (ret)
-		goto err_late;
+		ret = bch2_fs_freespace_init(c);
+		bch_err_msg(ca, ret, "initializing free space");
+		if (ret)
+			goto err_late;
 
-	if (ca->mi.state == BCH_MEMBER_STATE_rw)
-		__bch2_dev_read_write(c, ca);
+		if (ca->mi.state == BCH_MEMBER_STATE_rw)
+			__bch2_dev_read_write(c, ca);
 
-	ret = bch2_dev_journal_alloc(ca, false);
-	bch_err_msg(c, ret, "allocating journal");
-	if (ret)
-		goto err_late;
+		ret = bch2_dev_journal_alloc(ca, false);
+		bch_err_msg(c, ret, "allocating journal");
+		if (ret)
+			goto err_late;
+	}
 
 	up_write(&c->state_lock);
 out: