[Intel-gfx] [PATCH] drm/i915: Always wake the device to flush the GTT

[Daniel Vetter]

On Wed, Aug 30, 2017 at 01:56:40PM +0100, Chris Wilson wrote:
> Quoting Daniel Vetter (2017-08-30 13:23:56)
> > Or just the need to add a pile more tests to pm_rpm?
> 
> No. It's just your regular combinatorial explosion. The approach I would
> take here would be to register a sysenter callback that attempted to do a
> rpm suspend (i.e. so ~every ioctl would start from idle, and controlled
> via the faultinjection framework) and then run the minimal test set that
> exercises all ioctl paths, and then expand to all driver branches.
> 
> First we need coverage feedback.

What I meant to imply: As long as any display is on we will never rpm
suspend. Mostly this is the case for CI machines.

The new testcases I've had in mind would explicitly dpms off the display
before running a set of gem testcases. We don't want to do that everywhere
though, because a dpms on/off is very costly.

I guess once we switch (eventually, hopefully) to a binary-at-time model
for full igt CI we could amortize that over a pile of substests and do it
almost everywhere.

So I think adding a force rpm suspend won't help, because under normal CI
runs we can't ever get there becaus of the active display. And that's why
we're not hitting all these tons of problems anywhere else.

This might also be good reason for more server chips, or at least fake
server chips, where we disable the display entirely.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch