ubifs: Limit the number of pages in shrink_liability

If the number of dirty pages to be written back is large,
then writeback_inodes_sb will block waiting for a long time,
causing hung task detection alarm. Therefore, we should limit
the maximum number of pages written back this time, which let
the budget be completed faster. The remaining dirty pages
tend to rely on the writeback mechanism to complete the
synchronization.

Fixes: b6e5131 ("writeback: separate starting of sync vs opportunistic writeback")
Signed-off-by: Liu Song <liu.song11@zte.com.cn>
Signed-off-by: Richard Weinberger <richard@nod.at>

fs/ubifs/budget.c

@@ -51,7 +51,7 @@
 static void shrink_liability(struct ubifs_info *c, int nr_to_write)
 {
 	down_read(&c->vfs_sb->s_umount);
-	writeback_inodes_sb(c->vfs_sb, WB_REASON_FS_FREE_SPACE);
+	writeback_inodes_sb_nr(c->vfs_sb, nr_to_write, WB_REASON_FS_FREE_SPACE);
 	up_read(&c->vfs_sb->s_umount);
 }