On Thu, Oct 6, 2022 at 10:04 AM Igor Petruk <[email protected]> wrote:
>
> On Wed, Oct 5, 2022 at 9:18 PM Mike Larkin <[email protected]> wrote:
> >
> > On Wed, Oct 05, 2022 at 12:55:33PM -0700, Mike Larkin wrote:
> > > On Wed, Oct 05, 2022 at 01:20:47PM +0100, [email protected] wrote:
> > > > >Synopsis:  Intel NUC 11 is very slow, acpi0 kernel thread always take 
> > > > >100%
> > > > >Category:  system, kernel, amd64
> > > > >Environment:
> > > >     System      : OpenBSD 7.2
> > > >     Details     : OpenBSD 7.2-current (GENERIC.MP) #767: Tue Oct  4 
> > > > 23:38:08 MDT 2022
> > > >                      
> > > > [email protected]:/usr/src/sys/arch/amd64/compile/GENERIC.MP
> > > >
> > > >     Architecture: OpenBSD.amd64
> > > >     Machine     : amd64
> > > > >Description:
> > > >     The system is impossible to use, UI rendering lags in seconds. The 
> > > > acpi0 is spinning even if no XOrg is running.
> > > > >How-To-Repeat:
> > > >         Install OpenBSD on Intel NUC Gen 11, Core i7, Iris Xe Graphics.
> > > >         Working in console is tolerable, but any UI, including fvwm is 
> > > > already lagging.
> > > >         Gnome and XFce4 are almost frozen. Stable vs snapshot, 
> > > > hyperthreading vs no, noatime - nothing helps much.
> > > > >Fix:
> > > >         Not aware of any
> > > >
> > >
> > > Third time we've seen stuck GPEs in the last few weeks.
> > >
> > > Try changing this line in acpi_gpe_task:
> > >
> > >         dnprintf(10, "handle gpe: %x\n", gpe);
> > >
> > > to
> > >       printf("handle gpe: %x\n", gpe);
> > >
> > > and see what gpe is firing over and over. It will be in dmesg.
> > >
> > > Then maybe we can track it down.
> > >
> > > -ml
> >
> > After comparing the previous report and this, I'm certain they are the same
> > and you're going to see 6F as the culprit.
>
> Thanks for looking into this.
>
> I've done it. You are right, it prints a bunch of 6Fs. The system is
> even slower, which is expected due to printing in the logs.

I've looked at the previous threads on this issue. One explanation was
that it was a buggy hardware, e.g. cheap motherboard. In my case this
is Intel NUC. I've just upgraded the BIOS from Feb 2022 version to Aug
16th, 2022 version from
https://www.intel.com/content/www/us/en/download/19698/bios-update-tntgl357.html.

The problem persists. I've seen in other OSs there are indeed some
mitigations available, whether via writing to `/sys`, sysctl or kernel
boot options to block the interrupt. It might be a useful feature for
OpenBSD to mitigate issues like this until a proper fix on either side
(hardware or kernel) is implemented.

Thanks,
Igor.

Reply via email to