> Date: Sun, 13 Nov 2022 16:05:36 +0300
> From: Mikhail <mp39...@gmail.com>
>
> On Sun, Nov 13, 2022 at 04:25:00PM +1100, ja...@tubnor.net wrote:
> >
> >
> > > -----Original Message-----
> > > From: Mark Kettenis <mark.kette...@xs4all.nl>
> > > Sent: Saturday, 12 November 2022 11:00 PM
> > > To: ja...@tubnor.net
> > > Cc: bugs@openbsd.org
> > > Subject: Re: ACPI 6.4 Could not convert 1 to 4 panic
> > >
> > >
> > > Could you boot a normal kernel (i.e. not a ramdisk kernel) without
> > > Mikhail's diff and show me the output? A screen image is ok.
> >
> > No problems. This was a sysupgrade to the latest -current from the patched
> > to unpatched GENERIC.MP kernel.
>
> Offending code is in SSDT.8:
>
> Scope (\_SB.PC00.PEG0) {
>
> [...]
> Method (_STA, 0, NotSerialized) // _STA: Status
> {
> If ((PG0E == One))
> {
> Return (0x0F)
> }
>
> Return (Zero)
> }
> [...]
> }
>
> PG0E is defined as External/UnknownObj, the problem is that in
> DSDT we have two definition of PG0E - one is the field unit under
> "DefinitionBlock" root, another one as a package under "Scope (_SB)". My
> suspicion is that they meant "\PG0E" and simply forgot to define scope
> explicitly, because comparing package to One doesn't make sense.
Thanks. Yes I agree with that analysis.
> So the patch I sent works, but whole situation looks like not as a
> bug or not implemented functionality in OpenBSD, but as a bug in the
> vendor's ASL, and I am not sure what is the policy for including such
> workarounds into the tree.
Not sure we have a policy. But it does seem to indicate that adding
an implicit conversion from Package to Integer isn't the right
approach. Need to think a bit more about this, but maybe the right
thing to do is having _STA() return 0 if an unexpected AML failure
occurs.
