Re: [linux-usb-devel] [4/4] 2.6.23-rc3: known regressions

[Linus Torvalds]

On Mon, 20 Aug 2007, Arjan van de Ven wrote:

> 
> On Mon, 2007-08-20 at 23:26 -0700, Linus Torvalds wrote:
> > 
> > On Mon, 20 Aug 2007, Arjan van de Ven wrote:
> > > 
> > > untested patch to add this to cpufreq; this is probably a good idea in
> > > general even if using the latency framework doesn't end up being used
> > > for fixing this regression...
> > > 
> > > 
> > > --- linux-2.6.23-rc2/drivers/cpufreq/cpufreq.c.org	2007-08-20 22:58:32.000000000 -0700
> > > +++ linux-2.6.23-rc2/drivers/cpufreq/cpufreq.c	2007-08-20 23:02:21.000000000 -0700
> > > @@ -1604,6 +1604,12 @@ static int __cpufreq_set_policy(struct c
> > >  	if (ret)
> > >  		goto error_out;
> > >  
> > > +
> > > +	if (system_latency_constraint() < policy->cpuinfo.transition_latency) {
> > 
> > That looks broken. "system_latency_constraint()" is in us, but 
> > transition_latency is in ns, afaik.
> > 
> > But adding a "/ 1000" to turn the ns into us, and it migth even work.
> 
> 
> eh woops yes indeed.
> Shows me for not testing; I'll do that tomorrow when I'm more awake

Side note: I think we migth want to also have some way of telling the user 
*why* we're not doing frequency changes. Maybe as simple as a rate-limited 
printk() or something.

Otherwise, we'll easily be in a situation where some poor sod ends up 
running constantly at lowest frequency, and no way of even seeing why. 
Which sounds like a debugging nightmare.

If the kernel spits out the occasional warning about the latency 
violation, at least we get notified about there being potential problems.

		Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/