Re: Latency data - 2.6.14-rc3-rt13

[Daniel Walker <dwalker@mvista.com>]

On Mon, 2005-10-10 at 15:09 -0700, Mark Knecht wrote:
> On 10/10/05, Daniel Walker <dwalker@mvista.com> wrote:
> > On Mon, 2005-10-10 at 14:28 -0700, Mark Knecht wrote:
> > > On 10/10/05, Mark Knecht <markknecht@gmail.com> wrote:
> > > <SNIP>
> > > > > then goto Kernel Hacking and select
> > > > > "Interrupts-off critical section latency timing"
> > > > > Then select "Latency tracing"
> > >
> > > Only had to turn on Latency Tracing. The others I had on...
> > >
> > > <SNIP>
> > > > >
> > > > > Daniel
> > > >
> > > > Will do. Building now. I'll be back later.
> > > >
> > >
> > > Unfortunately I didn't think I'd be back this fast. I built the new
> > > kernel and rebooted. The boot starts, gets down to the point where it
> > > tells me that the preempt debug stuff is on, and then jumps to an
> > > endlessly repeating error message:
> > >
> > > init[1]: segfault at ffffffff8010fce0 rip ffffffff8010fce0 rsp
> > > 00007fffffcb09b8 error 15
> > >
> > > This error repeasts endlessly until I reboot.
> > >
> > > Good thing I had another kernel I could boot back into! ;-)
> > >
> > > So, something isn't happy. Is this a -rt thing or a kernel issue?
> >
> > Hmm, it looks like latency tracing doesn't work on x86_64 .. I guess
> > you'll have to wait till someone fixes it .
> >
> > Another option is to turn off "Latency Tracing" then reboot, like it was
> > before but w/o the histogram. Then run,
> >
> > "echo 0 > /proc/sys/kernel/preempt_max_latency"
> >
> > Whenever a new maximum latency is observed it will log it with a stack
> > trace in the system logs. You can report that back here on LKML .
> >
> > You can view the system log with the command "dmesg" , I think .
> >
> > Daniel
> >
> >
> Yes, already that looks interesting. Do I have something going on with
> acpi? This is before running Jack. I'm in Gnome.There are a lot of
> these messages, but they've stopped. I suppose the 3997 is the delay?
> That would coorespond with the info I sent earlier.

I think this is a false reading . Sometimes when a system goes idle it
will cause interrupt disable latency that isn't real (due to process
halting) . You could try turning ACPI off if you can , and turn off
power management ..

Daniel

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