[Intel-gfx] [PATCH 09/18] drm/i915: possibly invalidate TLB before context switch

[Ben Widawsky]

On Thu, 29 Mar 2012 21:25:49 +0200
Daniel Vetter <daniel at ffwll.ch> wrote:

> On Sun, Mar 18, 2012 at 01:39:49PM -0700, Ben Widawsky wrote:
> > From http://intellinuxgraphics.org/documentation/SNB/IHD_OS_Vol1_Part3.pdf
> > 
> > [DevSNB] If Flush TLB invalidation Mode is enabled it’s the driver’s
> > responsibility to invalidate the TLBs at least once after the previous
> > context switch after any GTT mappings changed (including new GTT
> > entries).  This can be done by a pipelined PIPE_CONTROL with TLB inv bit
> > set immediately before MI_SET_CONTEXT.
> > 
> > Signed-off-by: Ben Widawsky <ben at bwidawsk.net>
> 
> Hm, I've beend decently confused about the meaning of
> GFX_TLB_INVALIDATE_ALWAYS - it actually means that we flush tlbs on every
> full flush (i.e. always) when it's reset. And we don't set this so this
> workaround is pretty much just informational. I'm hence wondering whether
> a big comment wouldn't be better?
> 
> -Daniel

It's probably not much difference in terms of LOC with the plus that the
"comment" turns into correctly functioning code if we flip the bit.
Just from a quick glance at code though, it seems we don't explicitly
touch this bit for GEN6. Which was probably why I did it in the first
place.

if (IS_GEN7(dev))
	I915_WRITE(GFX_MODE_GEN7,
		   GFX_MODE_DISABLE(GFX_TLB_INVALIDATE_ALWAYS) |
		   GFX_MODE_ENABLE(GFX_REPLAY_MODE));


Anyway, I really don't care enough to argue, but I like the way it is,
and it's been tested the way it is; call the ball.

Ben