On Sun, Jun 9, 2024 at 2:44 AM Walter Alejandro Iglesias <wai@roquesor.com>
wrote:

> On Wed Jun  5 16:00:04 2024 Theo de Raadt wrote:
> > Stuart Henderson <stu@spacehopper.org> wrote:
> > > On 2024/06/05 04:42, Crystal Kolipe wrote:
> > > > On Wed, Jun 05, 2024 at 09:06:41AM +0200, Walter Alejandro Iglesias
> wrote:
> > > > > Running halt -p the machine powers off perfectly but running
> reboot at
> > > > > the last moment in the halt process (screen already turned off) the
> > > > > machine does not turn off, it hangs indefinitely, I have to press
> the
> > > > > power button manually to turn it off.
> > > >
> > > > I've also seen this behaviour from time to time on a Thinkpad X260
> so the
> > > > underlying problem is probably not specific to your machine.
> > >
> > > X260 has always been problematic with reboots, not 100% of the time (in
> > > particular a reboot shortly after boot seems more likely to work) but
> > > very often. The only thing I've seen that improved things there is
> > > kettenis' "EFI runtime services on amd64" diff.
> >
> > On two machines with such a problem I've seen that if I disable the
> > umb(4) in the BIOS, it reboots fine.
> >
>
> I've tried changing settings on the bios, no effect.
>
> In any case, your advice does not apply to this case, I have been using
> this machine for many years without problems, in fact I bought it
> *especially* to be able to use OpenBSD, since until then I had not been
> able to run OpenBSD on any hardware.
>
> I have had little to no success lately on these lists on many issues
> that I have put a lot of time and effort into, I hope to be heard this
> time. :-)
>
> --
> Walter
>
>
If the machine was working before and stopped working at some moment I'd
recommend to
try a bisect with older snapshots, then maybe a bisect of kernel commits as
you get into shorter
time frames.

Oh, I read now that you already nailed it between June 2 and June 3, you
can try to build kernels
between that time frame to reproduce first (because afaik snapshots can
contain things out-of-tree)
then to isolate it to a commit

regards