don't busy retry the inode on failed grab_super_passive()

This fixes a soft lockup on conditions

a) the flusher is working on a work by __bdi_start_writeback(), while

b) someone else calls writeback_inodes_sb*() or sync_inodes_sb(), which
   grab sb->s_umount and enqueue a new work for the flusher to execute

The s_umount grabbed by (b) will fail the grab_super_passive() in (a).
Then if the inode is requeued, wb_writeback() will busy retry on it.
As a result, wb_writeback() loops for ever without releasing
wb->list_lock, which further blocks other tasks.

Fix the busy loop by redirtying the inode. This may undesirably delay
the writeback of the inode, however most likely it will be picked up
soon by the queued work by writeback_inodes_sb*(), sync_inodes_sb() or
even writeback_inodes_wb().

bug url: http://www.spinics.net/lists/linux-fsdevel/msg47292.html

Reported-by: Christoph Hellwig <hch@infradead.org>
Tested-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>

diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c
index 1599aa985fe2accf6dcc9456a570ba942201685d..04cf3b91e5016a1f7e3ceef734c7d85574ec296d 100644 (file)
--- a/fs/fs-writeback.c
+++ b/fs/fs-writeback.c
@@ -618,7 +618,12 @@ static long __writeback_inodes_wb(struct bdi_writeback *wb,
                struct super_block *sb = inode->i_sb;
 
                if (!grab_super_passive(sb)) {
-                       requeue_io(inode, wb);
+                       /*
+                        * grab_super_passive() may fail consistently due to
+                        * s_umount being grabbed by someone else. Don't use
+                        * requeue_io() to avoid busy retrying the inode/sb.
+                        */
+                       redirty_tail(inode, wb);
                        continue;
                }
                wrote += writeback_sb_inodes(sb, wb, work);