Re: p4-clockmod help

[Nigel Cunningham <ncunningham@cyclades.com>]

Hi again Pat et al.

On Sat, 2005-12-10 at 22:52, Pat Double wrote:
> On Friday 09 December 2005 19:14, Bojan Smojver wrote:
> > On Fri, 2005-12-09 at 11:07 -0600, Pat Double wrote:
> > > Has anyone been successful resuming with the p4-clockmod module loaded,
> > > or compiled in?
> >
> > It works on my notebook with FC Rawhide kernels and suspend2.
> 
> It does not work well here. I believe it is because I have a P4 HT, the SMP 
> part is causing problems. I get a BUG sometime after it starts writing the 
> image. Normally it happens while writing the image but can also happen during 
> the atomic copy. I tried to disable calls to change the freq during suspend 
> but I do not even get the suspend function called until the atomic copy, so 
> the bug happens before then. I do not know how to fix this.
> 
> Nigel, any help?
> 
> I get a BUG at p4-clockmod.c, line 142:
> BUG_ON(smp_processor_id() != i);
> 
> The error is an invalid operand 0000
> 
> Also, writing the image becomes VERY slow with p4-clockmod loaded. I do not 
> have any governer loaded except performance.

Since any fix should come from the cpufreq guys, and they'll understand
the issue better than me, I want to refer the problem to them...

Cqufreq guys, I've only had a superficial look at this, but this BUG_ON
appears to be triggering because hot unplugging of cpu1 is not being
taken into account. Can you please help?

Regards,

Nigel