[PATCH] drm/amdgpu: Remove GC HW IP 9.3.0 from noretry=1

[Tim Van Patten]

On Fri, May 17, 2024 at 12:35 AM Christian König
<christian.koenig at amd.com> wrote:
>
> Am 16.05.24 um 19:57 schrieb Tim Van Patten:
> > From: Tim Van Patten <timvp at google.com>
> >
> > The following commit updated gmc->noretry from 0 to 1 for GC HW IP
> > 9.3.0:
> >
> >      commit 5f3854f1f4e2 ("drm/amdgpu: add more cases to noretry=1")
> >
> > This causes the device to hang when a page fault occurs, until the
> > device is rebooted. Instead, revert back to gmc->noretry=0 so the device
> > is still responsive.
>
> Wait a second. Why does the device hang on a page fault? That shouldn't
> happen independent of noretry.

No idea, but hopefully someone within AMD can help answer that.

I'm not an expert in this area, I was just able to bisect to the CL
causing the change in behavior. There are other reports of people
bisecting to the same CL, so this issue appears to extend beyond
ChromeOS:
https://gitlab.freedesktop.org/mesa/mesa/-/issues/9728#note_2063879

> So that strongly sounds like this is just hiding a bug elsewhere.

That's entirely possible, bringing the number of real issues up to (at
least) two:
1. Why the page faults are occurring to begin with.
  - Any video of size 66x2158 seems to trigger the issue.
2. Why the page faults result in the device hanging with gmc->noretry=1.

I've asked prathyushi.nangia at amd (chris.kuruts at amd may be helping as
well) to look into the page faults further, since they can't hang the
device if they don't exist. She should be able to provide any further
details if you're interested, but please feel free to reach out to me
as well if you have any other questions.