On Tue, 2009-07-28 at 15:05 +0200, Thomas Hellström wrote: > Michel Dänzer wrote: > > Patches 1 and 2 are fixes for rather serious bugs, which made KMS pretty > > much > > unusable for me with AGP, and should probably go into 2.6.31 if possible. > > The > > others can use more testing and wait for 2.6.32. > > > > [PATCH 1/4] drm/radeon: Don't unreserve twice on failure to validate. > > [PATCH 2/4] drm/radeon: Fall back to evicting BOs with memcpy if necessary. > > [PATCH 3/4] drm/radeon: Pay more attention to object placement requested by > > userspace. > > [PATCH 4/4] drm/radeon: Don't always keep the fbcon BO pinned in VRAM. > > > Michel, > > Do these patches solve all your AGP problems?
It's looking that way :), at least the freezes I was getting look like they were due to the bugs fixed by the first two patches, resulting in a deadlock on a spinlock somewhere. By disabling CONFIG_PREEMPT_VOLUNTARY in the kernel configuration, I was 'only' getting the symptoms described in patch 1 instead. Thanks a lot Thomas for your hints leading in this direction. The only open AGP related issue is that only 1x is stable -- 2x and 4x lock up the GPU during KMS initialization -- whereas 4x is rock solid without KMS. I suspect we may be setting up some register(s) differently, but I haven't found anything yet. -- Earthling Michel Dänzer | http://www.vmware.com Libre software enthusiast | Debian, X and DRI developer ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july -- _______________________________________________ Dri-devel mailing list Dri-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dri-devel
