Messages in this thread | | | Date | Thu, 16 Feb 2012 10:04:03 +0900 | From | "Jun'ichi Nomura" <> | Subject | Re: Yet another hot unplug NULL pointer dereference (was Re: status of oops in sd_revalidate_disk?) |
| |
Hi,
On 02/15/12 20:29, Bart Van Assche wrote: > On Wed, Feb 15, 2012 at 3:26 AM, Jun'ichi Nomura <j-nomura@ce.jp.nec.com> wrote: >> I think this patch is good, too. >> Since QUEUE_FLAG_DEAD is also racy, it is not much different >> from queuedata check, IMO. > > Are you sure ? As far as I know the block layer takes care of > synchronizing queue flag modifications against request_fn invocations.
QUEUE_FLAG_DEAD is set outside of queue_lock. So it's racy. However, it's single directional change and the race is benign if the driver, who sets QUEUE_FLAG_DEAD, is ready to reject requests. q->queuedata is same with that regard.
-- Jun'ichi Nomura, NEC Corporation
| |