Btrfs: make sure to flush queued bios if write_cache_pages waits

write_cache_pages tries to build up a large bio to stuff down the pipe.
But if it needs to wait for a page lock, it needs to make sure and send
down any pending writes so we don't deadlock with anyone who has the
page lock and is waiting for writeback of things inside the bio.

Dave Sterba triggered this as a deadlock between the autodefrag code and
the extent write_cache_pages

Signed-off-by: Chris Mason <chris.mason@oracle.com>

1 files changed, 10 insertions, 4 deletions

diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index f284d4e5f44..b40ba75f448 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -2613,10 +2613,16 @@ retry:
 			 * swizzled back from swapper_space to tmpfs file
 			 * mapping
 			 */
-			if (tree->ops && tree->ops->write_cache_pages_lock_hook)
-				tree->ops->write_cache_pages_lock_hook(page);
-			else
-				lock_page(page);
+			if (tree->ops &&
+			    tree->ops->write_cache_pages_lock_hook) {
+				tree->ops->write_cache_pages_lock_hook(page,
+							       data, flush_fn);
+			} else {
+				if (!trylock_page(page)) {
+					flush_fn(data);
+					lock_page(page);
+				}
+			}
 
 			if (unlikely(page->mapping != mapping)) {
 				unlock_page(page);