Re: 2.6.28-rc9: oprofile regression

[Robert Richter <robert.richter@amd.com>]

On 17.01.09 16:09:23, Tim Blechmann wrote:
> > > however, trying to apply this patch to 2.6.28, the behavior is the same
> > > as before (one NMI) ... so possibly, it is a combination of two bugs,
> > > with similar symptoms ... 
> > 
> > Tim, could you revert 7c64ade53a6f977d73f16243865c42ceae999aea too?
> > 
> > If this not helps, last chance is
> > 59512900baab03c5629f2ff5efad1d5d4e682ece, but this seems to be save.
> 
> i tried to revert both commits, however the behavior doesn't seem to
> change. will try to apply the working patch to the child commits, maybe
> i can find something interesting ...

Hmm, strange. Actually 7c64ade53a6f977d73f16243865c42ceae999aea fixed
a similiar bug, see here:
http://bugzilla.kernel.org/show_bug.cgi?id=11908

Your patch with 2.6.28, does:

 grep NMI /proc/interrupts

returns exactly 1 NMI per core or some more?

> 
> best, tim
> 
> btw, i am not very familiar with kernel programming, but is it safe to
> have `static u64 *reset_value' uninitialized, or should it be
> initialized to NULL?

External and static variables should be gaaranteed to be initialized
to zero. Only local variables are uninitialized.

-Robert

-- 
Advanced Micro Devices, Inc.
Operating System Research Center
email: robert.richter@amd.com