Messages in this thread | | | Subject | Re: Strange block/scsi/workqueue issue | From | James Bottomley <> | Date | Fri, 22 Apr 2011 13:06:18 -0500 |
| |
On Fri, 2011-04-22 at 20:01 +0200, Tejun Heo wrote: > Hello, guys. > > On Wed, Apr 13, 2011 at 03:17:46PM -0500, James Bottomley wrote: > > On Wed, 2011-04-13 at 22:12 +0200, Jens Axboe wrote: > > > This patch looks pretty clean. Shouldn't you serialize that ->queuedata > > > = NULL assignment with the queue lock, though? > > > > pointer assignment is atomic, isn't it? As in on a 32 bit arch, it > > definitely is, and I thought on 64 bits it also was. So a simultaneous > > reader either sees previous value or NULL. > > What was the conclusion of this thread? James' fix seems minimal and > good for this cycle but I don't see it there yet. Jens, James?
Sorry, forgot to cc you.
I've proposed the minimal fixes for this cycle here:
http://marc.info/?l=linux-scsi&m=130348665328233 http://marc.info/?l=linux-scsi&m=130348673628282 http://marc.info/?l=linux-scsi&m=130348680328356
And I'll merge at least the scsi one.
But there's been no long term conclusion yet.
James
| |