On Tue, Mar 04, 2014 at 12:20:57PM +0000, Stuart Henderson wrote: > On 2014/03/04 11:58, Edd Barrett wrote: > > On Tue, Mar 04, 2014 at 10:22:47AM +0000, Stuart Henderson wrote: > > > On 2014/03/04 09:40, Edd Barrett wrote: > > > > This is since my last snapshot upgrade, but there again, I have not been > > > > updating all that frequently, so it could be an older issue. > > > > > > This report could be improved by including a date range (or even > > > better a bisection to figure out when it started..) > > > > Crikey. Well, IIRC, I last updated shortly before soft ports lock, but > > that is not a reliable estimate by any means. Sorry. > > > > -- > > Best Regards > > Edd Barrett > > > > http://www.theunixzoo.co.uk > > > > zgrep OpenBSD /var/log/messages/* > > For amd64 I have kernels from these dates which might speed up bisection > (and some older ones but I guess it's more recent?) > > Feb. 3, 4, 8, 10, 20, 21, 22, 23, 24, 25, 26, 27 > March 3 > > > I don't see this on my X220 (amd64 MP 8GB, HD 3000 on i7-2640M) with cwm.
I believe this was introduced with the intel driver update on Feb 3. I can reproduce the graphics corruption in KDE4 (in Konsole, the menu, and in the alt-tab switcher) after most (but not all) suspend/resume cycles on an X201. When it happens, this line is printed in Xorg.0.log: [ 106.316] (EE) intel(0): Failed to submit rendering commands, disabling acceleration. Restarting X makes the corruption go away. If I compile and use xf86-video-intel from -D"2014/02/03 00:00" with an otherwise -current system I don't see graphics corruption but still see the following lines on resume: [ 93.886] (EE) intel(0): Failed to submit batch buffer, expect rendering corruption or even a frozen display: Invalid argument. [ 93.887] (EE) intel(0): Failed to submit batch buffer, expect rendering corruption or even a frozen display: Invalid argument. I also checked the driver update from 2014/02/14 and it has the same broken behaviour as -current. It seems the old driver was a bit more resilient when faced with the batch buffer submission problem after resume? No idea where the root cause is.
