Suspend/Resumes fails early with r600(RV710) in "unpin not necessary"

[Michel Dänzer]

On Sam, 2012-03-17 at 20:48 +0100, Manuel Krause wrote: 
> 
> EE opt_core.c:2986 r600_shader_optimize - shader optimization: 
> unsupported chip

WTH is this? I can't find any opt_core.c file or r600_shader_optimize
function in upstream Mesa Git, and Google doesn't know much about it
either.

If your X session uses OpenGL when you try to suspend/resume (e.g. if
you're using an OpenGL compositing manager such as kwin, gnome-shell or
compiz), you might want to investigate this.


> Mar 14 01:50:51 COOLHEAD kernel: agpgart-via 0000:00:00.0: 
> Refused to change power state, currently in D0

Does radeon.agpmode=-1 help at all?


> Mar 14 01:50:51 COOLHEAD kernel: PM: Creating hibernation image:
> Mar 14 01:50:51 COOLHEAD kernel: PM: Need to copy 270062 pages
> Mar 14 01:50:51 COOLHEAD kernel: PM: Not enough free memory
> Mar 14 01:50:51 COOLHEAD kernel: PM: Error -12 creating 
> hibernation image
> Mar 14 01:50:51 COOLHEAD kernel: ACPI: Waking up from system 
> sleep state S4

This looks like it fails to enter hibernation in the first place.


> Mar 14 01:51:23 COOLHEAD kernel: radeon 0000:01:00.0: GPU lockup 
> CP stall for more than 10000msec
> Mar 14 01:51:23 COOLHEAD kernel: GPU lockup (waiting for 
> 0x004EBE11 last fence id 0x004EBE10)
> Mar 14 01:51:23 COOLHEAD kernel: radeon 0000:01:00.0: f7227600 
> unpin not necessary

FWIW, the 'unpin not necessary' message itself is usually harmless, the
problem is the GPU lockups.


> After eight times I switched to tty1 and was able to reboot in 
> blindness.

After eight times of what? GPU lockup/reset?


-- 
Earthling Michel Dänzer           |                   http://www.amd.com
Libre software enthusiast         |          Debian, X and DRI developer