[PATCH] drm/amd/amdgpu: disable GFXOFF around debugfs access to MMIO

[Huang Rui]

On Fri, Feb 21, 2020 at 10:35:33PM +0800, StDenis, Tom wrote:
> 
> On 2020-02-21 9:34 a.m., Huang Rui wrote:
> > On Wed, Feb 19, 2020 at 10:09:46AM -0500, Tom St Denis wrote:
> >> I got some messages after a while:
> >>
> >> [  741.788564] Failed to send Message 8.
> >> [  746.671509] Failed to send Message 8.
> >> [  748.749673] Failed to send Message 2b.
> >> [  759.245414] Failed to send Message 7.
> >> [  763.216902] Failed to send Message 2a.
> >>
> >> Are there any additional locks that should be held?  Because some commands
> >> like --top or --waves can do a lot of distinct read operations (causing a
> >> lot of enable/disable calls).
> >>
> >> I'm going to sit on this a bit since I don't think the patch is ready for
> >> pushing out.
> >>
> > How about use RREG32_KIQ and WREG32_KIQ?
> 
> 
> For all register accesses (in the debugfs read/write method)? Can we use 
> those on all ASICs?

It can be used for all register access, but using KIQ is not as fast as MMIO.

So we can check if GFXOFF enabled, then go with KIQ path. Because KIQ can
wake up the GFX to "on" state at runtime.

Thanks,
Ray

> 
> 
> Tom
>