Merge tag 'for-5.3-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:

 - tiny race window during 2 transactions aborting at the same time can
   accidentally lead to a commit

 - regression fix, possible deadlock during fiemap

 - fix for an old bug when incremental send can fail on a file that has
   been deduplicated in a special way

* tag 'for-5.3-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  Btrfs: fix deadlock between fiemap and transaction commits
  Btrfs: fix race leading to fs corruption after transaction abort
  Btrfs: fix incremental send failure after deduplication

fs/btrfs/backref.c

@@ -1483,7 +1483,7 @@ int btrfs_check_shared(struct btrfs_root *root, u64 inum, u64 bytenr,
 	ulist_init(roots);
 	ulist_init(tmp);
 
-	trans = btrfs_attach_transaction(root);
+	trans = btrfs_join_transaction_nostart(root);
 	if (IS_ERR(trans)) {
 		if (PTR_ERR(trans) != -ENOENT && PTR_ERR(trans) != -EROFS) {
 			ret = PTR_ERR(trans);

fs/btrfs/send.c

@@ -6322,68 +6322,21 @@ static int changed_extent(struct send_ctx *sctx,
 {
 	int ret = 0;
 
-	if (sctx->cur_ino != sctx->cmp_key->objectid) {
-
-		if (result == BTRFS_COMPARE_TREE_CHANGED) {
-			struct extent_buffer *leaf_l;
-			struct extent_buffer *leaf_r;
-			struct btrfs_file_extent_item *ei_l;
-			struct btrfs_file_extent_item *ei_r;
-
-			leaf_l = sctx->left_path->nodes[0];
-			leaf_r = sctx->right_path->nodes[0];
-			ei_l = btrfs_item_ptr(leaf_l,
-					      sctx->left_path->slots[0],
-					      struct btrfs_file_extent_item);
-			ei_r = btrfs_item_ptr(leaf_r,
-					      sctx->right_path->slots[0],
-					      struct btrfs_file_extent_item);
-
-			/*
-			 * We may have found an extent item that has changed
-			 * only its disk_bytenr field and the corresponding
-			 * inode item was not updated. This case happens due to
-			 * very specific timings during relocation when a leaf
-			 * that contains file extent items is COWed while
-			 * relocation is ongoing and its in the stage where it
-			 * updates data pointers. So when this happens we can
-			 * safely ignore it since we know it's the same extent,
-			 * but just at different logical and physical locations
-			 * (when an extent is fully replaced with a new one, we
-			 * know the generation number must have changed too,
-			 * since snapshot creation implies committing the current
-			 * transaction, and the inode item must have been updated
-			 * as well).
-			 * This replacement of the disk_bytenr happens at
-			 * relocation.c:replace_file_extents() through
-			 * relocation.c:btrfs_reloc_cow_block().
-			 */
-			if (btrfs_file_extent_generation(leaf_l, ei_l) ==
-			    btrfs_file_extent_generation(leaf_r, ei_r) &&
-			    btrfs_file_extent_ram_bytes(leaf_l, ei_l) ==
-			    btrfs_file_extent_ram_bytes(leaf_r, ei_r) &&
-			    btrfs_file_extent_compression(leaf_l, ei_l) ==
-			    btrfs_file_extent_compression(leaf_r, ei_r) &&
-			    btrfs_file_extent_encryption(leaf_l, ei_l) ==
-			    btrfs_file_extent_encryption(leaf_r, ei_r) &&
-			    btrfs_file_extent_other_encoding(leaf_l, ei_l) ==
-			    btrfs_file_extent_other_encoding(leaf_r, ei_r) &&
-			    btrfs_file_extent_type(leaf_l, ei_l) ==
-			    btrfs_file_extent_type(leaf_r, ei_r) &&
-			    btrfs_file_extent_disk_bytenr(leaf_l, ei_l) !=
-			    btrfs_file_extent_disk_bytenr(leaf_r, ei_r) &&
-			    btrfs_file_extent_disk_num_bytes(leaf_l, ei_l) ==
-			    btrfs_file_extent_disk_num_bytes(leaf_r, ei_r) &&
-			    btrfs_file_extent_offset(leaf_l, ei_l) ==
-			    btrfs_file_extent_offset(leaf_r, ei_r) &&
-			    btrfs_file_extent_num_bytes(leaf_l, ei_l) ==
-			    btrfs_file_extent_num_bytes(leaf_r, ei_r))
-				return 0;
-		}
-
-		inconsistent_snapshot_error(sctx, result, "extent");
-		return -EIO;
-	}
+	/*
+	 * We have found an extent item that changed without the inode item
+	 * having changed. This can happen either after relocation (where the
+	 * disk_bytenr of an extent item is replaced at
+	 * relocation.c:replace_file_extents()) or after deduplication into a
+	 * file in both the parent and send snapshots (where an extent item can
+	 * get modified or replaced with a new one). Note that deduplication
+	 * updates the inode item, but it only changes the iversion (sequence
+	 * field in the inode item) of the inode, so if a file is deduplicated
+	 * the same amount of times in both the parent and send snapshots, its
+	 * iversion becames the same in both snapshots, whence the inode item is
+	 * the same on both snapshots.
+	 */
+	if (sctx->cur_ino != sctx->cmp_key->objectid)
+		return 0;
 
 	if (!sctx->cur_inode_new_gen && !sctx->cur_inode_deleted) {
 		if (result != BTRFS_COMPARE_TREE_DELETED)

fs/btrfs/transaction.c

@@ -28,15 +28,18 @@ static const unsigned int btrfs_blocked_trans_types[TRANS_STATE_MAX] = {
 	[TRANS_STATE_COMMIT_START]	= (__TRANS_START | __TRANS_ATTACH),
 	[TRANS_STATE_COMMIT_DOING]	= (__TRANS_START |
 					   __TRANS_ATTACH |
-					   __TRANS_JOIN),
+					   __TRANS_JOIN |
+					   __TRANS_JOIN_NOSTART),
 	[TRANS_STATE_UNBLOCKED]		= (__TRANS_START |
 					   __TRANS_ATTACH |
 					   __TRANS_JOIN |
-					   __TRANS_JOIN_NOLOCK),
+					   __TRANS_JOIN_NOLOCK |
+					   __TRANS_JOIN_NOSTART),
 	[TRANS_STATE_COMPLETED]		= (__TRANS_START |
 					   __TRANS_ATTACH |
 					   __TRANS_JOIN |
-					   __TRANS_JOIN_NOLOCK),
+					   __TRANS_JOIN_NOLOCK |
+					   __TRANS_JOIN_NOSTART),
 };
 
 void btrfs_put_transaction(struct btrfs_transaction *transaction)
@@ -543,7 +546,8 @@ start_transaction(struct btrfs_root *root, unsigned int num_items,
 		ret = join_transaction(fs_info, type);
 		if (ret == -EBUSY) {
 			wait_current_trans(fs_info);
-			if (unlikely(type == TRANS_ATTACH))
+			if (unlikely(type == TRANS_ATTACH ||
+				     type == TRANS_JOIN_NOSTART))
 				ret = -ENOENT;
 		}
 	} while (ret == -EBUSY);
@@ -659,6 +663,16 @@ struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root
 				 BTRFS_RESERVE_NO_FLUSH, true);
 }
 
+/*
+ * Similar to regular join but it never starts a transaction when none is
+ * running or after waiting for the current one to finish.
+ */
+struct btrfs_trans_handle *btrfs_join_transaction_nostart(struct btrfs_root *root)
+{
+	return start_transaction(root, 0, TRANS_JOIN_NOSTART,
+				 BTRFS_RESERVE_NO_FLUSH, true);
+}
+
 /*
  * btrfs_attach_transaction() - catch the running transaction
  *
@@ -2037,6 +2051,16 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans)
 		}
 	} else {
 		spin_unlock(&fs_info->trans_lock);
+		/*
+		 * The previous transaction was aborted and was already removed
+		 * from the list of transactions at fs_info->trans_list. So we
+		 * abort to prevent writing a new superblock that reflects a
+		 * corrupt state (pointing to trees with unwritten nodes/leafs).
+		 */
+		if (test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state)) {
+			ret = -EROFS;
+			goto cleanup_transaction;
+		}
 	}
 
 	extwriter_counter_dec(cur_trans, trans->type);

fs/btrfs/transaction.h

@@ -94,11 +94,13 @@ struct btrfs_transaction {
 #define __TRANS_JOIN		(1U << 11)
 #define __TRANS_JOIN_NOLOCK	(1U << 12)
 #define __TRANS_DUMMY		(1U << 13)
+#define __TRANS_JOIN_NOSTART	(1U << 14)
 
 #define TRANS_START		(__TRANS_START | __TRANS_FREEZABLE)
 #define TRANS_ATTACH		(__TRANS_ATTACH)
 #define TRANS_JOIN		(__TRANS_JOIN | __TRANS_FREEZABLE)
 #define TRANS_JOIN_NOLOCK	(__TRANS_JOIN_NOLOCK)
+#define TRANS_JOIN_NOSTART	(__TRANS_JOIN_NOSTART)
 
 #define TRANS_EXTWRITERS	(__TRANS_START | __TRANS_ATTACH)
 
@@ -183,6 +185,7 @@ struct btrfs_trans_handle *btrfs_start_transaction_fallback_global_rsv(
 					int min_factor);
 struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root);
 struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root);
+struct btrfs_trans_handle *btrfs_join_transaction_nostart(struct btrfs_root *root);
 struct btrfs_trans_handle *btrfs_attach_transaction(struct btrfs_root *root);
 struct btrfs_trans_handle *btrfs_attach_transaction_barrier(
 					struct btrfs_root *root);