Re: [PATCH] per-backing dev unplugging #2

[Chris Mason <mason@suse.com>]

Hmpf, one more.  If one proc does a wait_on_buffer while another does
discard_buffer, bh->b_bdev might be null by the time __wait_on_buffer
uses it.  

Someone hit this with reiserfs, but it should be possible to trigger
anywhere.

-chris

Index: linux.t/fs/buffer.c
===================================================================
--- linux.t.orig/fs/buffer.c	2004-03-15 15:05:38.000000000 -0500
+++ linux.t/fs/buffer.c	2004-03-15 16:49:17.000000000 -0500
@@ -132,7 +132,11 @@ void __wait_on_buffer(struct buffer_head
 	do {
 		prepare_to_wait(wqh, &wait, TASK_UNINTERRUPTIBLE);
 		if (buffer_locked(bh)) {
-			blk_run_address_space(bh->b_bdev->bd_inode->i_mapping);
+			struct block_device *bd;
+			smp_mb();
+			bd = bh->b_bdev;
+			if (bd)
+				blk_run_address_space(bd->bd_inode->i_mapping);
 			io_schedule();
 		}
 	} while (buffer_locked(bh));