2015-06-01 06:23:15

by Bruno Prémont

[permalink] [raw]
Subject: Re: [Patch v3] apple-gmux: lock iGP IO to protect from vgaarb changes

On Fri, 29 May 2015 18:36:50 +0200 Darren Hart wrote:
> > Making sure to lock only the intel GPU when present and especially protecting
> > against nvidia driver will be hard if legacy-IO is being processed by a hidden
> > device!
>
> Ugh indeed. Worst case we can special case via dmi strings. Is this Apple device
> significantly different from others? Bruno, what are you testing on?

I only own a pretty old MacBook Air with just NVIDIA IGP and had to
rely on BUG reports and testing from affected users.

Not doing anything on apple-gmux when only a single GPU is visible
should be easy, but denying any vgaarb operation when Intel IGP is
hidden and just discrete GPU present is much harder (if one does not
want to risk opening the next can of worms).

DMI based special-casing would work but will it uncover the next issue
with the same device configured differently?

Bruno


2015-06-01 17:31:56

by Darren Hart

[permalink] [raw]
Subject: Re: [Patch v3] apple-gmux: lock iGP IO to protect from vgaarb changes

On Mon, Jun 01, 2015 at 08:22:27AM +0200, Bruno Pr?mont wrote:
> On Fri, 29 May 2015 18:36:50 +0200 Darren Hart wrote:
> > > Making sure to lock only the intel GPU when present and especially protecting
> > > against nvidia driver will be hard if legacy-IO is being processed by a hidden
> > > device!
> >
> > Ugh indeed. Worst case we can special case via dmi strings. Is this Apple device
> > significantly different from others? Bruno, what are you testing on?
>
> I only own a pretty old MacBook Air with just NVIDIA IGP and had to
> rely on BUG reports and testing from affected users.
>
> Not doing anything on apple-gmux when only a single GPU is visible
> should be easy, but denying any vgaarb operation when Intel IGP is
> hidden and just discrete GPU present is much harder (if one does not
> want to risk opening the next can of worms).
>
> DMI based special-casing would work but will it uncover the next issue
> with the same device configured differently?

No, we would need a combination. I presume "configured differently" would mean
the Intel GPU present - which would be detectable.

DMI + Nvidia = do A
DMI + Intel = do B

--
Darren Hart
Intel Open Source Technology Center