Re: [PATCH 06/11] drm/i915: plumb VM into object operations

[Daniel Vetter <daniel@ffwll.ch>]

On Mon, Jul 15, 2013 at 08:57:13PM -0700, Ben Widawsky wrote:
> So here is the plan summing up from this mail thread and the other
> one...
> 
> I'm calling the series done until we get feedback from Chris on the
> eviction/shrinker code. Basically whatever he signs off on, I am willing
> to implement (assuming you agree). I am confident enough in my rebase
> abilities to at least fix that up when we get to it.

I guess we could shovel this around after merging, shrinker gets frobbed
around once a while for anyway ;-)

> On top of this series, I am going smash the 3 patches which introduce
> bind/unbind function pointers. I am doing this because all this code
> already exists, plays nicely together, and has been tested.
> 
> Finally on top of this, I am going to have a [fairly large] patch to
> squash <obj,vm> pair into <vma> where I feel its appropriate. My plan
> for this is to basically look at a squashed version of all the patches
> where I introduce both arguments in one call, think about if it makes
> sense to me to think about the operation in terms of an object, or a
> vma, and then change as appropriate.
> 
> So the feedback I need from you:
> 1. Does that sound reasonable? If you want to call it weird, that's fine,
> but I think the diff churn is acceptable, and the series should be quite
> bisectable.

I guess I'll live ;-)

> 2. Are you aware of anything else I missed in other patch review which I
> haven't mentioned. I cannot find anything, and I think except for this
> patch I had updated everything locally.

Imo my review on the semantics of ->bind/unbind is fairly important, at
least for discussion. Atm you just add some new abstraction (and kill some
old one for which we've established some good use) without putting it to
real use imo. Haven't read any responses yet though, it's early.

> You can see the 14 patches (11 here + 3 from the others series):
> http://cgit.freedesktop.org/~bwidawsk/drm-intel/log/?h=ppgtt2
> 
> Hopefully this was all taken in good humor. I am not upset or anything,
> we just need to figure out how we can make some progress on this soon
> without making total crap, and without taking all my time (since I have
> a few other things to work on). I know you warned me about the latter a
> few weeks ago, but well, reality.

Imo the first five patches should get the small reviews addressed and then
merged. Atm that's what I'm kinda stalling on ... Then we can tackle the
next little bit of this major work.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch