[PATCH 2/2] drm/amdgpu: Add modeset module option

[Takashi Iwai]

On Tue, 03 Apr 2018 11:18:34 +0200,
Michel D4nzer wrote:
> 
> On 2018-04-03 11:02 AM, Takashi Iwai wrote:
> > On Tue, 03 Apr 2018 10:57:56 +0200,
> > Christian K6nig wrote:
> >>
> >> Am 03.04.2018 um 10:36 schrieb Michel Dänzer:
> >>> On 2018-04-01 07:45 PM, Ilia Mirkin wrote:
> >>>> On Sun, Apr 1, 2018 at 1:39 PM, Christian König
> >>>> <christian.koenig at amd.com> wrote:
> >>>>> Am 30.03.2018 um 22:45 schrieb Takashi Iwai:
> >>>>>> amdgpu driver lacks of modeset module option other drm drivers provide
> >>>>>> for enforcing or disabling the driver load.  Interestingly, the
> >>>>>> amdgpu_mode variable declaration is already found in the header file,
> >>>>>> but the actual implementation seems to have been forgotten.
> >>>>>>
> >>>>>> This patch adds the missing piece.
> >>>>>
> >>>>> NAK, modesetting is mandatory for amdgpu and we should probably remove the
> >>>>> option to disable it from other DRM drivers without UMS support as well
> >>>>> (pretty much all of them now).
> >>>>>
> >>>>> If you want to prevent a driver from loading I think the correct way to do
> >>>>> so is to give modprobe.blacklist=amdgpu on the kernel commandline.
> >>>>>
> >>>>> That would remove the possibility to prevent the driver from loading when it
> >>>>> is compiled in, but I don't see much of a problem with that.
> >>>> Having a way to kill the graphics driver is a very useful debugging
> >>>> tool, and also a quick and easy way to get out of an unpleasant
> >>>> situation where graphics are messed up / system hangs / etc. The
> >>>> modprobe blacklist kernel arg only works in certain environments (and
> >>>> only if it's a module).
> >>> Building amdgpu into the kernel isn't feasible for a generic kernel such
> >>> as a distro one, because it would require including all microcode into
> >>> the kernel as well (12M right now, and growing).
> >>>
> >>> If a user decides to build amdgpu into their custom kernel and runs into
> >>> trouble due to that, that's "doctor, it hurts if I do this" territory.
> >>
> >> Correct, but I agree that even in this situation it would be very
> >> helpful to prevent the gfx drivers from loading and fallback to
> >> efifb/vesafd (or whatever the platform provides).
> >>
> >> It's just that the "nomodeset" and "amdgpu.modeset=0" options are
> >> really not well named for this task.
> > 
> > Agreed with the naming mess.  But OTOH, it's already a thing that is
> > too popular to kill.  You can add a more suitable option name, but you
> > cannot drop these existing ones easily.  It's already in a gray zone
> > of the golden "don't break user-space" rule.
> 
> That's quite a stretch argument, given that amdgpu has never supported
> the modeset parameter.

Oh I don't mean about my own patch but the foreseen action Christian
mentioned.

> Also, module parameters aren't UAPI.

Right, but we care not only about UAPI.  If the kernel breaks
something, it's a regression.  It's what Linus suggested many times.
The same argument has been applied to proc or sysfs files in the past,
and we had to correct back again.

Don't get me wrong: I'm not always objecting to the removal of any
module existing options.  But if it break a major usage pattern, it's
a problem.  And the removal of nomodeset option would be a kind of
such things, IMO, that's why I mentioned as "a gray zone" in the
above.


thanks,

Takashi