Created attachment 119967 [details] dmesg output with the message of a kernel hang Hi. I have recently (say, since upgrading to kernel 4.2, but not really sure of the exact period) having some GPU hangs with my Sandy Bridge notebook while using X and, during these events, all the user interface elements become frozen for a few moments. In one occasion, I even had my entire X session killed and I was back to the Linux console. I *believe* (but can't say with 100% confidence) that it happens when I play videos on Youtube with Debian's Iceweasel. I don't believe that I have experienced one of these hangs with "regular" video players like mpv, mplayer or vlc, but my memory is really weak these years. I am attaching some information and, if needed, I can try to provide anything else that you ask me to. Output of uname -a: * Linux zatz 4.2.0-1-amd64 #1 SMP Debian 4.2.6-1 (2015-11-10) x86_64 GNU/Linux Distribution: * Debian testing. Machine: * It's (I believe) a rebranded Clevo notebook with a Core i5-2410M and no other video card. I have a GPU crash dump that I'm attaching here. I will try to boot with special options listed on https://01.org/linuxgraphics/documentation/how-report-bugs, but I don't know if I will be able to reproduce this problem reliably. Once again, if there is anything else that I can provide, please just ask. Thanks, Rogério Brito.
Created attachment 119968 [details] Log from X
Created attachment 119969 [details] GPU dump from /sys/class/drm/card0/error
Yes, it looks like the libva hangs where the GPU stops executing commands.
Dear Chris, On Nov 20 2015, bugzilla-daemon@freedesktop.org wrote: > https://bugs.freedesktop.org/show_bug.cgi?id=93033 > > --- Comment #3 from Chris Wilson <chris@chris-wilson.co.uk> --- > Yes, it looks like the libva hangs where the GPU stops executing commands. Ah, that might explain why I don't see problems with, say, mpv, which is using OpenGL to show the videos. Is there anything else that I can send you that could help with the problem or do you already have an idea of what the problem may be? Thanks,
I do believe that libva is working on a fix (or at least I saw a patch in the last couple of days that seemed like it was going to address this bug).
Second look, and I finally found a pattern: When the ringbuffer wraps around at address 0, there is a chance that the GPU stops working. Marking as duplicate for tracking. *** This bug has been marked as a duplicate of bug 93262 ***
Dear Chris, (In reply to Chris Wilson from comment #6) > Second look, and I finally found a pattern: > > When the ringbuffer wraps around at address 0, there is a chance that the > GPU stops working. Marking as duplicate for tracking. > > *** This bug has been marked as a duplicate of bug 93262 *** Excellent that you found the pattern. Unfortunately, I won't be able to test this, as I don't have access to the computer anymore (it is now with my ex-partner). If, on the other hand, I could test it with a computer that has an hybrid Intel Sandybridge + an AMD Radeon board (I sincerely don't have any clue on how the switching between the cards happen or if I am only using one and not the other), then I can do whatever you want me to. On a second note, it does not seem to have been picked up by any of the released trees that I have seen so far... Should I probe someone to get this included, at least downstream? Thanks, Rogério.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.