[Intel-gfx] IVB page flipping fixes (hopefully final)

[Zhao, Jian J]

The issue(the patch may bring regression to piglit test) can't be reproduced now both with new code and old code, so maybe we can make the patch into our kernel tree. 
The details:  
I did some retest with newer commit in Mesa master and Xserver in 1.10 branch and run the piglit test without glean cases(because it will cost a long time, about 370 cases), the passrate were about 2101(288)/2643 with compiz enabled. (it's 2103(288)/2645 with compiz disabled). On ogles2 conformance side, there will be a lot of GPU hangs when enabled compiz.(compiz can't start without Jesse's patch) When disabled compiz, its passrate will return to 1194/1198 both with and without the patch in Kernel. 

Commit info:
Libdrm:         (master)2.4.26
Mesa:           (master)8875dd58719b978283e89acf04422a4eaf9b021d
Xserver:                (server-1.10-branch)xorg-server-1.10.2.901-2-g9fab8b475de99848866c07bde962a5f6ed01b987
Xf86_video_intel:               (master)2.15.0-125-g8fb98e22a63a4354b049ace7870abb6406541c28
Cairo:          (master)441f9c5037dd32464bd87e21108b702c0a3c508a
Libva:          (master)20574567f32a6ad9f46665a4ced5725df42355f1
Kernel:         (drm-intel-fixes)6a574b5b9b186e28abd3e571dfd1700c5220b510 + Jesse's patch 

> -----Original Message-----
> From: intel-gfx-bounces+jian.j.zhao=intel.com at lists.freedesktop.org
> [mailto:intel-gfx-bounces+jian.j.zhao=intel.com at lists.freedesktop.org] On
> Behalf Of Zhao, Jian J
> Sent: Monday, June 20, 2011 11:15 PM
> To: Jesse Barnes
> Cc: intel-gfx at lists.freedesktop.org
> Subject: Re: [Intel-gfx] IVB page flipping fixes (hopefully final)
> 
> Hi Jesse,
>  With your these two patches:
> http://lists.freedesktop.org/archives/intel-gfx/2011-June/011014.html It
> indeed fixed three bugs(#38362, #38392, #38393) on IvyBridge, but it also
> caused a regression(bug #38428) and another issue is with the patched
> kernel, the piglit cases' passrate dropped a lot, it changed from
> 2372(400)/3020 to 1682(397)/3020. The format is PASS(Not Supported)/Total.
> Though I find some crashed cases can't be reproduced when tested
> separately now. I will have an investigate on it. It is better if you could have
> some idea of the relation between the patch and piglit tests. Thanks.
> 
> Best regards
> zhaojian
> 
> > -----Original Message-----
> > From: intel-gfx-bounces+jian.j.zhao=intel.com at lists.freedesktop.org
> > [mailto:intel-gfx-bounces+jian.j.zhao=intel.com at lists.freedesktop.org] On
> > Behalf Of Jesse Barnes
> > Sent: Friday, June 17, 2011 12:19 AM
> > To: intel-gfx at lists.freedesktop.org
> > Subject: [Intel-gfx] IVB page flipping fixes (hopefully final)
> >
> > This set addresses comments from Dan and Chris, namely:
> >   - pin & fence the new fb on the correct ring for IVB
> >   - cleanup gen2/3 wait code by putting it in gen2/3 wait functions
> >   - remove dead code
> >   - fix a race in the split code with the interrupt handler and flip pending
> >
> > Thanks,
> > Jesse
> >
> > _______________________________________________
> > Intel-gfx mailing list
> > Intel-gfx at lists.freedesktop.org
> > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx