Broken NFS (perhaps Cache invalidation bug ?)

Broken NFS (perhaps Cache invalidation bug ?)

[L. D. Marks]

I have a problem which appears to be very similar to a cache invalidation 
bug previously reported: 
http://www.ussg.iu.edu/hypermail/linux/kernel/0510.1/0582.html

With a conventional nfs mount (not automount), c0-0 a client node, 
running on the nfs server:

echo 10 > Probe
ssh -x c0-0 cat Probe
echo 11 > Probe
ssh -x c0-0 cat Probe

Will give, most times, 10 & 10 rather than 10 & 11.

I've tried a wide range of things (including consulting local experts), 
so far nothing. I'm not a kernel developer, so please be gentle. This 
problem has occurred ever since we moved to the 4.X releases of rocks 
(http://www.rocksclusters.org), X=0 or 1 although it's hidden by their 
default use of automounting and it's taken some time to reduce it to 
something simple.

-----------------------------------------------
Laurence Marks
Department of Materials Science and Engineering
MSE Rm 2036 Cook Hall
2220 N Campus Drive
Northwestern University
Evanston, IL 60201, USA
Tel: (847) 491-3996 Fax: (847) 491-7820
email: L-marks at northwestern dot edu
http://www.numis.northwestern.edu
-----------------------------------------------

Re: Broken NFS (perhaps Cache invalidation bug ?)

[L. D. Marks]

I should have added the kernel: 2.6.9-22.0.2.ELsmp #1 SMP with 
nfs-utils-1.0.6-65.EL4


On Mon, 6 Feb 2006, L. D. Marks wrote:

> I have a problem which appears to be very similar to a cache invalidation bug 
> previously reported: 
> http://www.ussg.iu.edu/hypermail/linux/kernel/0510.1/0582.html
>
> With a conventional nfs mount (not automount), c0-0 a client node, running on 
> the nfs server:
>
> echo 10 > Probe
> ssh -x c0-0 cat Probe
> echo 11 > Probe
> ssh -x c0-0 cat Probe
>
> Will give, most times, 10 & 10 rather than 10 & 11.
>
> I've tried a wide range of things (including consulting local experts), so 
> far nothing. I'm not a kernel developer, so please be gentle. This problem 
> has occurred ever since we moved to the 4.X releases of rocks 
> (http://www.rocksclusters.org), X=0 or 1 although it's hidden by their 
> default use of automounting and it's taken some time to reduce it to 
> something simple.
>
> -----------------------------------------------
> Laurence Marks
> Department of Materials Science and Engineering
> MSE Rm 2036 Cook Hall
> 2220 N Campus Drive
> Northwestern University
> Evanston, IL 60201, USA
> Tel: (847) 491-3996 Fax: (847) 491-7820
> email: L-marks at northwestern dot edu
> http://www.numis.northwestern.edu
> -----------------------------------------------
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>

Note: if you have an old email address for me, please note that "nwu" has
been changed to "northwestern".
-----------------------------------------------
Laurence Marks
Department of Materials Science and Engineering
MSE Rm 2036 Cook Hall
2220 N Campus Drive
Northwestern University
Evanston, IL 60201, USA
Tel: (847) 491-3996 Fax: (847) 491-7820
email: L-marks at northwestern dot edu
http://www.numis.northwestern.edu
-----------------------------------------------

Re: Broken NFS (perhaps Cache invalidation bug ?)

[Alan Cox]

On Llu, 2006-02-06 at 20:52 -0600, L. D. Marks wrote:
> I should have added the kernel: 2.6.9-22.0.2.ELsmp #1 SMP with 
> nfs-utils-1.0.6-65.EL4

Vendor kernels diverge quite a bit from the base tree (because they
freeze a kernel and backport the critical fixes from thereon). So really
you are best advised to ask your vendor (Red Hat or Centos or whoever)
rather than ask here for that kernel.

Alan

Re: Broken NFS (perhaps Cache invalidation bug ?)

[Trond Myklebust]

On Mon, 2006-02-06 at 20:49 -0600, L. D. Marks wrote:
> I have a problem which appears to be very similar to a cache invalidation 
> bug previously reported: 
> http://www.ussg.iu.edu/hypermail/linux/kernel/0510.1/0582.html
> 

Which should be fixed.

Cheers,
  Trond