Re: [RT PATCH v2] seqlock: serialize against writers

[Gregory Haskins <gregory.haskins@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1401 bytes --]

Gregory Haskins wrote:
> *Patch submitted for inclusion in PREEMPT_RT 26-rt4.  Applies to 2.6.26.3-rt3*
>
> Hi Ingo, Steven, Thomas,
>   Please consider for -rt4.  This fixes a nasty deadlock on my systems under
>   heavy load.
>
> [
> Changelog:
> 	v2: only touch seqlock_t because raw_seqlock_t doesn't require
> 	    serialization and userspace cannot modify data during a read
>
> 	v1: initial release
> ]
>   

Hi Andi,
  As it turns out, my distcc backend was an x86_64 machine running the
v1 patch and I started to notice sometime today that certain cc1 jobs
were sometimes (albeit rarely) segfaulting on me.  I noticed that before
I even published the first patch, but I chalked it up to a corrupt .o on
my NFS home.  Plus I was forgetting that the distcc machine was running
the patch, and I would probably have never made the userspace connection
had you not mentioned it.  In any case, I self-built this v2 patch with
v2 applied, and the segfaults have gone away.  So I think we know
several things:

1) You were right that this would cause an issue if the slow path is hit.
2) Steven was right that userspace must use raw_seqlock_t because it no
longer crashes with v2.
3) I am satisfied that my primary concern is still properly addressed.

Hopefully everyone is satisfied with this patch now.

Thanks for the help.  It's much appreciated!

-Greg



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 257 bytes --]