Re: Latency data - 2.6.14-rc3-rt13

[Daniel Walker]

On Mon, 2005-10-10 at 14:12 -0700, Mark Knecht wrote:
> On 10/10/05, Daniel Walker <dwalker@xxxxxxxxxx> wrote:
> > On Mon, 2005-10-10 at 13:16 -0700, Mark Knecht wrote:
> > > Hi,
> >
> > >
> > > How can I get data that would be more useful in terms of real debug?
> >
> > The IRQ off times look like the worst . If you do "make menuconfig"
> >
> > then goto Kernel Hacking and select
> > "Interrupts-off critical section latency timing"
> > Then select "Latency tracing"
> >
> > Then when you boot the system before run the following,
> >
> > "echo 0 > /proc/sys/kernel/preempt_max_latency"
> 
> So this disables the tracing of preempt times but keeps IRQ times on? Cool.

Yes, that what the compile options are doing.

> >
> > That will record a trace for every maximum latency observed for IRQ
> > latency . You can view the trace with this command
> > "cat /proc/latency_trace" , and you can attach the trace to an email to
> > LKML so we can review it (compress it if it's big though) .
> >
> > Daniel
> 
> Will do. Building now. I'll be back later.
> 
> Is there anything specific I should look for in the traces myself?
> Anyway to help narrow it down?

You want the trace to represent the largest latency. For instance in the
histogram it showed IRQ latency was ~3000us or greater, so you want the
trace to show at least that size latency .

Daniel 

-
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/