Bug#643629: linux-image-2.6.39-bpo.2-amd64: BUG at fs/jbd2/transaction.c:1086

[Jonathan Nieder <jrnieder@gmail.com>]

Hi Nico,

nico wrote:

> kernel BUG at [...]/fs/jbd2/transaction.c:1086!
[...]
> Call Trace:
>  [<ffffffffa059b82d>] ? ocfs2_journal_dirty+0x47/0x58 [ocfs2]
>  [<ffffffffa057e3b1>] ? __ocfs2_add_entry+0x78d/0x88d [ocfs2]
>  [<ffffffffa03a1cb4>] ? jbd2_journal_put_journal_head+0x19/0x4e [jbd2]
>  [<ffffffffa039b9b4>] ? jbd2_journal_get_write_access+0x2d/0x38 [jbd2]
>  [<ffffffffa05a5e37>] ? ocfs2_link+0x55e/0x7ac [ocfs2]
>  [<ffffffffa058fd4e>] ? ocfs2_permission+0xec/0xf9 [ocfs2]
>  [<ffffffff81104701>] ? vfs_link+0xc3/0x126
>  [<ffffffff81107084>] ? sys_linkat+0x11d/0x179
>  [<ffffffff810ff09c>] ? sys_newstat+0x24/0x2d
>  [<ffffffff81339392>] ? system_call_fastpath+0x16/0x1b
> Code: 48 8b 13 f7 c2 00 00 10 00 75 f3 f0 0f ba 2b 14 19 d2 85 d2 75 e8 83 7d 10 00 75 19 c7 45 10 01 00 00 00 41 8b 55 08 85 d2 7f 04 <0f> 0b eb fe ff ca 41 89 55 08 48 39 45 28 75 11 83 7d 0c 01 75 
> RIP  [<ffffffffa039b05e>] jbd2_journal_dirty_metadata+0x5d/0xe3 [jbd2]

The tripped assertion is

	if (jh->b_modified == 0) {
		/*
		 * This buffer's got modified and becoming part
		 * of the transaction. This needs to be done
		 * once a transaction -bzzz
		 */
		jh->b_modified = 1;
		J_ASSERT_JH(jh, handle->h_buffer_credits > 0);
		handle->h_buffer_credits--;
	}

Is this reproducible?  If so, could you try a 3.x kernel from sid or
experimental?  (There have been some fixes for bugs like this in ext3
and ext4 but I don't see any similar change to ocfs2.  Even so, it
seems worth a try before contacting upstream about this.)

Thanks,
Jonathan