Re: Patch to fixe Data Acess error in dup_fd

[Sharyathi Nagesh <sharyath@in.ibm.com>]

  I looked into a few memtests that were run in similar machine. There
are a few slab corruption issues but not while running memtest and no
other issues.
  Seems difficult to replicate.

On Wed, 2006-11-15 at 00:15 -0800, Vadim Lobanov wrote:
> On Wed, 2006-11-15 at 13:08 +0530, Sharyathi Nagesh wrote:
> >    This is very interesting: after reading through I am feeling there is high chance this 
> > could as well be a memory corruption issue. But if the issue is memory getting corrupted 
> > what could be the possible reasons.
> >    I had observed random slab corruption issues in the machine, could
> > that may have resulted in corruption, we may be opening up larger issues
> > here about which I am not much aware of, 
> 
> I'm guessing that you've already tried this, but it never hurts to be
> sure: does this machine pass memtest? :)
> 
> >    The kernel version on which it is tested is: 2.6.18-1 (Distro   
> >    variant)
> 
> Unless someone recognizes special magic values from the register dumps
> to point at any particular part of the kernel, the corruption could be
> coming from almost anywhere. If noone has any better guesses, then
> narrowing down the problem might be worthwhile: grab a vanilla
> non-distro 2.6.18-1 kernel (from kernel.org) and see if you can
> reproduce the problem with that, and then try to find the previous
> release where the problem disappears. Or use git instead, which folks
> say can do this bisection process rather well. :)
> 
> Thanks,
> -- Vadim Lobanov
>