PCI Config on Intel 855 chip after resume

[Jesse Barnes]

On Sunday, October 21, 2007 3:00 pm Peter Clifton wrote:
> I can still file a bug it you want, but the attached appears to fix /
> workaround the problem. Took _AGES_ to get to trying this, but it
> came about after some hunch that rendering state was involved, noting
> that the Xrandr Rotate code-paths touched that, and that rotating the
> screen would fix the textures.
>
> Finally tracked down the invariant state as a possible thing to test,
> and added it in the EnterVT function. (Now knowing what I was looking
> for, I see a similar line in the old i810 driver which did work
> across S3 on the 855GM).
>
> Any comments on the validity of the fix, or whether it should be
> pushed upstream?

I'm not sure if this is the correct fix... according to Eric, any 3D 
state should be updated according to the pI830->last_3d field, so maybe 
there's a call to IntelEmitInvarientState missing somewhere else?  The 
only place that looked at all suspicious was I830DRISwapContext, but 
since that sets up a new 3D state it should be safe, right?

Also, you should get an fdo account so you can check in fixes 
directly. :)  http://www.freedesktop.org/wiki/AccountRequests.

Jesse