Recently I built a new PC based on a proposal in the CT journal 1/15.
The proposal contains well designed hardware components. I slightly modified it, running it with a 256 GB SSD and 8 GB Ram. I use the built-in graphic of the celeron CPU.
Although it's just a 34 eur celeron dual core it's fast, silent and I am quite happy with it.
BUT: Recently I noticed, that there are frequent hangs. The desktop was freezing frequently (no mouse move, no keyboard input) for 2-5 seconds. This happened especially when working with firefox, scrolling or having google mail open.
Dmesg showed the following:
Jan 10 11:28:17 speedy kernel: [ 531.220050] [drm] stuck on render ring
Jan 10 11:28:17 speedy kernel: [ 531.220824] [drm] GPU HANG: ecode 0:0x87d3bffa, in Xorg , reason: Ring hung, action: reset
Jan 10 11:28:17 speedy kernel: [ 531.220825] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Jan 10 11:28:17 speedy kernel: [ 531.220826] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
Jan 10 11:28:17 speedy kernel: [ 531.220826] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Jan 10 11:28:17 speedy kernel: [ 531.220827] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
Jan 10 11:28:17 speedy kernel: [ 531.220828] [drm] GPU crash dump saved to /sys/class/drm/card0/error
Jan 10 11:28:19 speedy kernel: [ 533.221743] [drm] Enabling RC6 states: RC6 on, RC6p off, RC6pp off
Seems like something related to the intel 915 graphics driver crashed and was restarted...
[Funny side remark: the crash dump had a size of 0, but anyway I could open and read it... Copy it to another location and finally you have a size > 0. ]
So I started digging the internet, with the result that there are plenty of similar bug reports. This ubuntu bug report seemed to be a good point to start. It was closed (obviously a similar issue was fixed already in 14.04, june 2014) but re-opened. No big activity in it, but it lead me to https://bugs.freedesktop.org/show_bug.cgi?id=83677. It has a lot of duplicates, and the description matched my problem.
After some more time I figured out that - thanks to those guys - a kernel fix was submitted - brand new, on 16-Dec-14 (today it's the 10-Jan-15).
As I have now clue about applying a kernel patch (and the dropbox links there were already dead anyway), I searched kernel.org's git commit messages.
It turned out, that this fix found its way to the official kernel sources. Not included in the latest stable yet (3.18.2 as of today) but in the version 3.19 RC 3:
- drm/i915: Disable PSMI sleep messages on all rings around context switches
- drm/i915: Force the CS stall for invalidate flushes
As I was not willing to wait 2 or 3 months until this will get a stable release, I decided to give it a try and to install the 3.19 RC3 kernel.
This was easier than expected. Just follow this howto. Now I'm running:
user@system:~$ uname -r
So far: No freezes any more. The message is gone. Thanks to Chris Wilson and all the supporters who provided this fix.