[Intel-gfx] [PATCH 00/15] Accurate frontbuffer tracking and psr conversion

[Chris Wilson]

On Mon, Jun 16, 2014 at 07:51:20PM +0200, Daniel Vetter wrote:
> Note that currently the gtt tracking has a bit a gap: We never exit it. Bunch of
> fixes are possible:
>
> - Wire up the core dirty_fb ioctl to flush framebuffers. This is used by generic
>   userspace to flush dummy buffers, which in our case means gtt mmaps. So maps
>   perfectly.
> 
> - Do a delayed gtt pte teardown and force-flush. Probably too ugly to care
>   really.

I don't consider that to be a gap, but accurate reflection of how
userspace is using the GTT. The most important usecase in this regard is
perhaps how the scanout is permenantly mapped for writing after a
terminal GPU hang. Trying to be smart in the kernel is more likely to
upset userspace, especially in this last resort effort at keeping
everything running.

> - Try to integrate hw gtt write tracking logic. Note that current psr code
>   doesn't rely on this - I've killed the X-tiled checks completely.

Also probably not worth it. In the normal sporadic use, we can rely on
the GTT being flushed as required.

Note, that the tracking as proposed will unfortunately get fbcon wrong
after it is touched by X. I think however there is some merit in
improving how we handle fbcon here.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre