btrfs: properly split extent_map for REQ_OP_ZONE_APPEND

Damien reported a test failure with btrfs/209. The test itself ran fine,
but the fsck ran afterwards reported a corrupted filesystem.

The filesystem corruption happens because we're splitting an extent and
then writing the extent twice. We have to split the extent though, because
we're creating too large extents for a REQ_OP_ZONE_APPEND operation.

When dumping the extent tree, we can see two EXTENT_ITEMs at the same
start address but different lengths.

$ btrfs inspect dump-tree /dev/nullb1 -t extent
...
   item 19 key (269484032 EXTENT_ITEM 126976) itemoff 15470 itemsize 53
           refs 1 gen 7 flags DATA
           extent data backref root FS_TREE objectid 257 offset 786432 count 1
   item 20 key (269484032 EXTENT_ITEM 262144) itemoff 15417 itemsize 53
           refs 1 gen 7 flags DATA
           extent data backref root FS_TREE objectid 257 offset 786432 count 1

The duplicated EXTENT_ITEMs originally come from wrongly split extent_map in
extract_ordered_extent(). Since extract_ordered_extent() uses
create_io_em() to split an existing extent_map, we will have
split->orig_start != split->start. Then, it will be logged with non-zero
"extent data offset". Finally, the logged entries are replayed into
a duplicated EXTENT_ITEM.

Introduce and use proper splitting function for extent_map. The function is
intended to be simple and specific usage for extract_ordered_extent() e.g.
not supporting compression case (we do not allow splitting compressed
extent_map anyway).

There was a question raised by Qu, in summary why we want to split the
extent map (and not the bio):

The problem is not the limit on the zone end, which as you mention is
the same as the block group end. The problem is that data write use zone
append (ZA) operations. ZA BIOs cannot be split so a large extent may
need to be processed with multiple ZA BIOs, While that is also true for
regular writes, the major difference is that ZA are "nameless" write
operation giving back the written sectors on completion. And ZA
operations may be reordered by the block layer (not intentionally
though). Combine both of these characteristics and you can see that the
data for a large extent may end up being shuffled when written resulting
in data corruption and the impossibility to map the extent to some start
sector.

To avoid this problem, zoned btrfs uses the principle "one data extent
== one ZA BIO". So large extents need to be split. This is unfortunate,
but we can revisit this later and optimize, e.g. merge back together the
fragments of an extent once written if they actually were written
sequentially in the zone.
Reported-by: Damien Le Moal <damien.lemoal@wdc.com>
Fixes: d22002fd ("btrfs: zoned: split ordered extent when bio is sent")
CC: stable@vger.kernel.org # 5.12+
CC: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>

fs/btrfs/inode.c

@@ -2271,13 +2271,127 @@ static blk_status_t btrfs_submit_bio_start(struct inode *inode, struct bio *bio,
 	return btrfs_csum_one_bio(BTRFS_I(inode), bio, 0, 0);
 }
 
+/*
+ * Split an extent_map at [start, start + len]
+ *
+ * This function is intended to be used only for extract_ordered_extent().
+ */
+static int split_zoned_em(struct btrfs_inode *inode, u64 start, u64 len,
+			  u64 pre, u64 post)
+{
+	struct extent_map_tree *em_tree = &inode->extent_tree;
+	struct extent_map *em;
+	struct extent_map *split_pre = NULL;
+	struct extent_map *split_mid = NULL;
+	struct extent_map *split_post = NULL;
+	int ret = 0;
+	int modified;
+	unsigned long flags;
+
+	/* Sanity check */
+	if (pre == 0 && post == 0)
+		return 0;
+
+	split_pre = alloc_extent_map();
+	if (pre)
+		split_mid = alloc_extent_map();
+	if (post)
+		split_post = alloc_extent_map();
+	if (!split_pre || (pre && !split_mid) || (post && !split_post)) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	ASSERT(pre + post < len);
+
+	lock_extent(&inode->io_tree, start, start + len - 1);
+	write_lock(&em_tree->lock);
+	em = lookup_extent_mapping(em_tree, start, len);
+	if (!em) {
+		ret = -EIO;
+		goto out_unlock;
+	}
+
+	ASSERT(em->len == len);
+	ASSERT(!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags));
+	ASSERT(em->block_start < EXTENT_MAP_LAST_BYTE);
+
+	flags = em->flags;
+	clear_bit(EXTENT_FLAG_PINNED, &em->flags);
+	clear_bit(EXTENT_FLAG_LOGGING, &flags);
+	modified = !list_empty(&em->list);
+
+	/* First, replace the em with a new extent_map starting from * em->start */
+	split_pre->start = em->start;
+	split_pre->len = (pre ? pre : em->len - post);
+	split_pre->orig_start = split_pre->start;
+	split_pre->block_start = em->block_start;
+	split_pre->block_len = split_pre->len;
+	split_pre->orig_block_len = split_pre->block_len;
+	split_pre->ram_bytes = split_pre->len;
+	split_pre->flags = flags;
+	split_pre->compress_type = em->compress_type;
+	split_pre->generation = em->generation;
+
+	replace_extent_mapping(em_tree, em, split_pre, modified);
+
+	/*
+	 * Now we only have an extent_map at:
+	 *     [em->start, em->start + pre] if pre != 0
+	 *     [em->start, em->start + em->len - post] if pre == 0
+	 */
+
+	if (pre) {
+		/* Insert the middle extent_map */
+		split_mid->start = em->start + pre;
+		split_mid->len = em->len - pre - post;
+		split_mid->orig_start = split_mid->start;
+		split_mid->block_start = em->block_start + pre;
+		split_mid->block_len = split_mid->len;
+		split_mid->orig_block_len = split_mid->block_len;
+		split_mid->ram_bytes = split_mid->len;
+		split_mid->flags = flags;
+		split_mid->compress_type = em->compress_type;
+		split_mid->generation = em->generation;
+		add_extent_mapping(em_tree, split_mid, modified);
+	}
+
+	if (post) {
+		split_post->start = em->start + em->len - post;
+		split_post->len = post;
+		split_post->orig_start = split_post->start;
+		split_post->block_start = em->block_start + em->len - post;
+		split_post->block_len = split_post->len;
+		split_post->orig_block_len = split_post->block_len;
+		split_post->ram_bytes = split_post->len;
+		split_post->flags = flags;
+		split_post->compress_type = em->compress_type;
+		split_post->generation = em->generation;
+		add_extent_mapping(em_tree, split_post, modified);
+	}
+
+	/* Once for us */
+	free_extent_map(em);
+	/* Once for the tree */
+	free_extent_map(em);
+
+out_unlock:
+	write_unlock(&em_tree->lock);
+	unlock_extent(&inode->io_tree, start, start + len - 1);
+out:
+	free_extent_map(split_pre);
+	free_extent_map(split_mid);
+	free_extent_map(split_post);
+
+	return ret;
+}
+
 static blk_status_t extract_ordered_extent(struct btrfs_inode *inode,
 					   struct bio *bio, loff_t file_offset)
 {
 	struct btrfs_ordered_extent *ordered;
-	struct extent_map *em = NULL, *em_new = NULL;
-	struct extent_map_tree *em_tree = &inode->extent_tree;
 	u64 start = (u64)bio->bi_iter.bi_sector << SECTOR_SHIFT;
+	u64 file_len;
 	u64 len = bio->bi_iter.bi_size;
 	u64 end = start + len;
 	u64 ordered_end;
@@ -2317,41 +2431,16 @@ static blk_status_t extract_ordered_extent(struct btrfs_inode *inode,
 		goto out;
 	}
 
+	file_len = ordered->num_bytes;
 	pre = start - ordered->disk_bytenr;
 	post = ordered_end - end;
 
 	ret = btrfs_split_ordered_extent(ordered, pre, post);
 	if (ret)
 		goto out;
-
-	read_lock(&em_tree->lock);
-	em = lookup_extent_mapping(em_tree, ordered->file_offset, len);
-	if (!em) {
-		read_unlock(&em_tree->lock);
-		ret = -EIO;
-		goto out;
-	}
-	read_unlock(&em_tree->lock);
-
-	ASSERT(!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags));
-	/*
-	 * We cannot reuse em_new here but have to create a new one, as
-	 * unpin_extent_cache() expects the start of the extent map to be the
-	 * logical offset of the file, which does not hold true anymore after
-	 * splitting.
-	 */
-	em_new = create_io_em(inode, em->start + pre, len,
-			      em->start + pre, em->block_start + pre, len,
-			      len, len, BTRFS_COMPRESS_NONE,
-			      BTRFS_ORDERED_REGULAR);
-	if (IS_ERR(em_new)) {
-		ret = PTR_ERR(em_new);
-		goto out;
-	}
-	free_extent_map(em_new);
+	ret = split_zoned_em(inode, file_offset, file_len, pre, post);
 
 out:
-	free_extent_map(em);
 	btrfs_put_ordered_extent(ordered);
 
 	return errno_to_blk_status(ret);