[PATCH 1/2] btrfs: don't try to flush data write bio if we hit error preparing it

[Qu Wenruo <wqu@suse.com>]

[BUG]
When running with 64K page size and 4K sectorsize (aka, subpage), there
is a chance that generic/475 would crash with the following BUG_ON() in
btrfs_csum_one_bio() triggered:

		if (!ordered) {
			ordered = btrfs_lookup_ordered_extent(inode, offset);
			BUG_ON(!ordered); /* Logic error */
		}

[CAUSE]
Test case generic/475 is doing dm-error testing, while for subpage case
we could hit error caused by dm-errors.

In that case, we will proper call end_extent_writepage() with @err = 1
to do the cleanup, including finishing the ordered extent.

In that case, the assembled bio still needs to be finished, by
end_write_bio() function.

But there are call sites that doesn't properly call end_write_bio(), but
go flush_write_bio() to submit the assembled bio.

In that case, we will call btrfs_csum_one_bio() even the ordered extent
is already cleaned up, and trigger the BUG_ON().

[FIX]
There are two call sites where we still try to call flush_write_bio()
even if we hit error:

- extent_write_cache_pages()
- extent_write_locked_range()

Fix both of them by calling end_write_bio() if we hit error.

Signed-off-by: Qu Wenruo <wqu@suse.com>
---
 fs/btrfs/extent_io.c | 17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)

diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index e82328bcb281..49f7dbbb2d73 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -5035,9 +5035,13 @@ static int extent_write_cache_pages(struct address_space *mapping,
 		 * page in our current bio, and thus deadlock, so flush the
 		 * write bio here.
 		 */
-		ret = flush_write_bio(epd);
-		if (!ret)
-			goto retry;
+		if (ret < 0) {
+			end_write_bio(epd, ret);
+		} else {
+			ret = flush_write_bio(epd);
+			if (!ret)
+				goto retry;
+		}
 	}
 
 	if (wbc->range_cyclic || (wbc->nr_to_write > 0 && range_whole))
@@ -5095,9 +5099,11 @@ int extent_write_locked_range(struct inode *inode, u64 start, u64 end,
 	wbc_attach_fdatawrite_inode(&wbc_writepages, inode);
 	while (start <= end) {
 		page = find_get_page(mapping, start >> PAGE_SHIFT);
-		if (clear_page_dirty_for_io(page))
+		if (clear_page_dirty_for_io(page)) {
 			ret = __extent_writepage(page, &wbc_writepages, &epd);
-		else {
+			if (ret < 0)
+				goto out;
+		} else {
 			btrfs_writepage_endio_finish_ordered(BTRFS_I(inode),
 					page, start, start + PAGE_SIZE - 1, true);
 			unlock_page(page);
@@ -5106,6 +5112,7 @@ int extent_write_locked_range(struct inode *inode, u64 start, u64 end,
 		start += PAGE_SIZE;
 	}
 
+out:
 	ASSERT(ret <= 0);
 	if (ret == 0)
 		ret = flush_write_bio(&epd);
-- 
2.32.0