Re: 2.6.28-rc9: oprofile regression

[Tim B. <ti...@kl...>]

On Thu, 2009-01-15 at 21:37 +0100, Thomas Gleixner wrote:
> On Thu, 15 Jan 2009, Tim Blechmann wrote:
> > On Thu, 2009-01-15 at 09:46 +0100, Thomas Gleixner wrote:
> > > Tim,
> > > 
> > > On Wed, 14 Jan 2009, Tim Blechmann wrote:
> > > > this code (line 81/82), changes counter_width from 32 to 40.
> > > > 
> > > > if (counter_width < eax.split.bit_width)
> > > >         counter_width = eax.split.bit_width;
> > > > 
> > > > however when removing these lines, and thus keeping the value 32 for
> > > > counter_width, doesn't change the behavior, only one NMI per cpu.
> > > 
> > > It would only help, when the reported bit_width would be bogus. We
> > > know that you get at least one NMI, so lets look at the results we get
> > > there.
> > 
> > it seems, that ppro_check_ctrs is never called:
> > [  982.238639] oprofile: using NMI interrupt.
> > 
> > hth, tim
> 
> Hmm, that confuses the hell out of me. Can you try the patch below,
> which restores the original code of writing the counter registers ?

no difference ... i must admit, i already spent some time, trying to
revert the specific changes of this commit ... i gave up to wait for
some help from someone, who actually knows how the code is working ...

possibly it makes sense to compile a full kernel of the first bad kommit
and introduce the different changes one by one ... i will try to find
some time for this during the next few days ...

but if you have more patches, that i can test, please let me know ...

cheers, tim
 
--
ti...@kl...
http://tim.klingt.org

Which is more musical, a truck passing by a factory or a truck passing
by a music school?
  John Cage