[Intel-gfx] [PATCH] drm/i915: RMW register cycles considered evil

[Daniel Vetter]

On Mon, Jul 06, 2015 at 04:23:23PM -0300, Paulo Zanoni wrote:
> 2015-07-06 12:04 GMT-03:00 Daniel Vetter <daniel at ffwll.ch>:
> > On Mon, Jul 06, 2015 at 03:50:49PM +0300, Ville Syrjälä wrote:
> >> On Mon, Jul 06, 2015 at 02:42:02PM +0200, Daniel Vetter wrote:
> >> > Especially for workarounds which is stuff that's almost impossible to
> >> > verify: The initial state from the firmware on boot-up and after
> >> > resume could be different, which will hide bugs when we do an RMW
> >> > cycle.
> >>
> >> If you're really worried about that then we should then explicitly
> >> initialize all the registers that might affect stuff.
> >>
> >> For a bunch of GT registers we could just do a GPU reset at driver
> >> init. That that won't help with UCGCTL and such.
> >>
> >> I'm also worried that if we don't use RMWs for early parts, the hardware
> >> folks may still change the default for some ofhte other bits, and then
> >> we end up clobbering those.
> >
> > The point is that we'll at least consistently clobber them, which is the
> > important part. Chasing a bug which only happens when you freshly boot but
> > not after the first gpu reset (or first resume or the other way round or
> > whatever) is not fun at all.
> 
> I think there are possibly other ways to be consistent. We could, for
> example, save the values which we think are correct at boot - even if
> we rely on the BIOS - and then (optionally) check them at every resume
> or runtime resume. Maybe there are even other ideas for that.
> 
> But I do see the value in what you're doing, even though I'm also
> afraid of the possible bugs brought by it, and I like the idea of
> starting this with new gens only. If you decide to keep this strategy,
> can you please print on debugfs when the RMW value is different from
> the non-RMW value?

We have a few testcases somewhere which try to make sure the w/a stick
after the usual suspects. Unfortunately the kernel instrumentation is very
weak for that (atm it only covers the render ring init w/a we do). We
could try to extend that a bit.

Or we could start screaming into dmesg if the rmw has an effect, like
you're suggesting.
-Daneil
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch