since aae4008096399a0e84abc7c016b35092caf9db25 of the 2D driver, VT switching back in with UXA triggers a kernel BUG. The BUG is legit, it seems -- objects got pinned while dev_priv->mm.suspended, which we overlooked when planning the GTT lookaside hack in userland. Switching back to EXA or reverting the commit works around it.
Marking it as a 2D blocker, as the kernel fix will likely result in userland needing fixing.
kernel half of the fix posted to intel-gfx.
That commit is only on master branch, not on xf86-video-intel-2.6-branch, so not in the release (candidate) yet.
Since Eric will be on vacation, I'm assigning it to Zhenyu.
For reference, aae4008096399a0e84abc7c016b35092caf9db25 fixed bug#19010,19014,19074.
Author: Zhenyu Wang <firstname.lastname@example.org>
Date: Wed Dec 31 22:56:57 2008 +0800
UXA: Fallback to dri_bo_map() if pin failed
This fixes VT switch issue with UXA after Eric's
aae4008096399a0e84abc7c016b35092caf9db25 on 2D side.
"after the patch, there are still bad wrong render results after VT switch" - zhenyu
reopen this bug...
Since the kernel BUG has been fixed by Eric, and the remaining issue is just rendering, I'm decreasing the severity so this won't block Q4 release.
Author: Eric Anholt <email@example.com>
Date: Mon Jan 5 23:21:07 2009 -0800
Fix pin leakage with EXA GTT-mapping shortcut, and crash with UXA on KMS.