Intel Graphics - Lags and glitches since linux 3.6?

robotangel · 2012-10-20 21:11:32

Hi there,

since the upgrade which included the linux-3.6 package (among others, unfourtunately I have no idea which packages were upgraded exactly, only downgrading the kernel was no fix), I experience graphical glitches when I use my laptop. There's a Ivy Bridge graphics chip in it and I use the intel driver together with SNA. When I boot up the computer it says

drm:__gen6_gt_force_wake_mt_get] *ERROR* Force wake wait timed out

and I can find this in my journal from time to time

Oct 20 21:14:35 marco-thinkpad kernel: [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
Oct 20 21:14:35 marco-thinkpad kernel: [drm] capturing error event; look for more information in /debug/dri/0/i915_error_state
Oct 20 21:14:41 marco-thinkpad kernel: [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
Oct 20 21:14:47 marco-thinkpad kernel: [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
Oct 20 21:14:53 marco-thinkpad kernel: [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
Oct 20 21:14:59 marco-thinkpad kernel: [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
Oct 20 21:15:05 marco-thinkpad kernel: [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
Oct 20 21:15:11 marco-thinkpad kernel: [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
Oct 20 21:15:17 marco-thinkpad kernel: [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung

The i915_error_state file is very big, so I had to xz and upload it: http://www30.zippyshare.com/v/33532322/file.html

Additionally, everything just feels slow somehow, for example KDE's Expose effect gives me approximately 5-6fps while it was perfectly smooth before.

Here is an example of how the graphical glitches look like: http://www.abload.de/img/intelbug1yostj.png

I hope there's someone here who can help me Thank you very much in advance!

wes · 2012-10-21 06:32:34

Hello,

I recently noticed a performance regression on my machine as well, but
fortunately not so many things changed at once and I was able to find the
cause. For me, downgrading xf86-video-intel back to 2.20.9-2 was the answer.
Hope that helps. I've yet to search for an existing bug report, but I'll look
into it if you and others can confirm.

WonderWoofy · 2012-10-21 06:51:16

Yeah, so I thought I had dodged this, but after my update today, my sh*t went crazy. What I am experiencing though is pretty different though. The best way to describe it is that it looks like my text is all redacted like a decassified government document. Black bars everywhere. I can see the text if I highlight it, but it isn't pretty. Also, my pseudo transparency is borked in urxvt.

I made a post here before I had determined that this is what it was.
https://bbs.archlinux.org/viewtopic.php?id=151118

I thought it might be something else since I have not read of the same symptoms that my system is experiencing.

wes · 2012-10-21 07:05:45

Funny -- I was just reading that other thread of yours as you were posting
here. Anyhow, that sounds pretty insane @_@ I didn't notice anything like
that, but I also didn't hang around very long. The only application I ran was
xterm before I noticed the issue and downgraded. Transparency is handled by
compiz for me, and it seemed to work ok.

kicka · 2012-10-21 08:31:22

wonderwoofy i had the same exact problem... i solved it downgrading:

xf86-video-intel to 2.20.9-1
xf86-input-evdev to 2.7.2.1
xorg-server to 1.12.4.1
xorg-server-common to 1.12.4.1

I downgraded using the aur package https://aur.archlinux.org/packages.php?ID=31937, and put in /etc/pacman.conf

IgnorePkg = xorg-server xorg-server-common xf86-video-intel xf86-input-evdev

To do so it's better you hit CTRL-Alt-F1 to see what you are writing at least...

It's not a solution but a temporary solution hoping they will fix it soon...

mynis01 · 2012-10-31 00:46:03

I started noticing this problem after upgrading from kernel 3.6 to 3.7 on my laptop which uses the i915 driver. It only seemed to happen when using a web browser, so I disabled hardware acceleration in firefox and switched back to the plain old arch kernel, and it seems to be gone. There's a few posts about this on lkml and gmane and kernel development places so I don't think it's an xorg thing. Mostly it seems to be related to workarounds they put in the kernel for the sake of older intel chips.

cfr · 2012-10-31 01:31:10

Using the default (rather than sna) has improved graphical performance hugely on my system.

Doesn't help with the power regressions, unfortunately...

WonderWoofy · 2012-10-31 01:42:22

That is interesting cfr. Makes me think I should give it a try...

ChALkeR · 2012-11-01 14:30:01

This patch fixed simular problems for me:
https://projects.archlinux.org/svntogit … a9861de2cc

Just upgrade kdebase-workspace to 4.9.2-6

PL_kolek · 2012-11-05 21:34:40

kicka wrote:

xf86-video-intel to 2.20.9-1
xf86-input-evdev to 2.7.2.1
xorg-server to 1.12.4.1
xorg-server-common to 1.12.4.1
I downgraded using the aur package https://aur.archlinux.org/packages.php?ID=31937, and put in /etc/pacman.conf
IgnorePkg = xorg-server xorg-server-common xf86-video-intel xf86-input-evdev

I experienced the same behaviour - GPU hung. However, I have only encountered it playing Team Fortress 2 using wine and bumblebee. Strange, because the game runs on nvidia card. I had to kill hl2.exe, and restart the game.

Today, I did exactly what you said, and so far, after 2 hours of gameplay just one crash, but most propably unrelated to the intel GPU. Has anyone opened of found a bug report related to that problem? Downgrading is only a short-term solution.

About performance regressions, I haven't experienced any using the system normally (but cairo-compmgr isn't heavy for a gpu). On the other hand, Team Fortress 2 has bigger fps drops than before, but that might be my subjective feeling.

WonderWoofy · 2012-11-06 00:36:16

This bug was fixed within hours (maybe a day) or the update, and what some half a month ago. So I think you may be a bit late to this party. It is probably the amazing bumblebee setup you have that is causing issues, or even your nvidia card.

PL_kolek · 2012-11-06 22:57:21

I am not sure if that's true what you're saying. I played today once again for about 2-3 hours. Still no crashes, no GPU hang journal lines. Playing with the newest Intel drivers is nearly impossible. I have to kill hl2.exe three times an hour due to GPU hung.

Of course there is a possibility, that bumblebee or nvidia might be the cause, strange thigs happen in software all the time, and this wouldn't be the least expected one. But... why does it end with i915 error and only on the newer intel drivers?

I have one more guess where the cause may lie. Maybe some kernel options I added for power saving started to mess things up after one of the updates (switching to systemd?). I'll try that tomorrow.

bassu · 2012-11-08 13:47:42

robotangel wrote:

There's a Ivy Bridge graphics chip in it and I use the intel driver together with SNA. When I boot up the computer it says

On my Ivy bridge with i5-321M, I have no power regression, no hiccups or hangs, just this gen6 related error on boot and rest of dmesg is clean from any further drm issues.

[    0.529967] [drm:__gen6_gt_force_wake_mt_get] *ERROR* Force wake wait timed out

WonderWoofy wrote:

Black bars everywhere. I can see the text if I highlight it, but it isn't pretty.

That only happens for me if I downgrade the kernel to 3.2, but not on 3.5+.

kicka wrote:

i solved it downgrading:
xf86-video-intel to 2.20.9-1
xf86-input-evdev to 2.7.2.1
xorg-server to 1.12.4.1
xorg-server-common to 1.12.4.1

Right now, I'm just seeing the error and no crashes. But I shall downgrade if anything crashes on 3.6+.

MetaMan · 2012-12-05 00:10:19

I have been having the errors described in the first post, but they only occur when I have a graphic-intensive game open, like TF2 (Linux Beta) or Minecraft. Sometimes I'll be working on my server and switching between Minecraft and Firefox, and I will have frequent pauses where everything justs freezes. It becomes almost impossible to get work done.

Also, and this has only happened with TF2 so far, everything will "spaz out" and everything on-screen becomes a series of oddly colored geometric shapes with warped text on top, even after closing the game.

Arch Linux

#1 2012-10-20 21:11:32

Intel Graphics - Lags and glitches since linux 3.6?

#2 2012-10-21 06:32:34

Re: Intel Graphics - Lags and glitches since linux 3.6?

#3 2012-10-21 06:51:16

Re: Intel Graphics - Lags and glitches since linux 3.6?

#4 2012-10-21 07:05:45

Re: Intel Graphics - Lags and glitches since linux 3.6?

#5 2012-10-21 08:31:22

Re: Intel Graphics - Lags and glitches since linux 3.6?

#6 2012-10-31 00:46:03

Re: Intel Graphics - Lags and glitches since linux 3.6?

#7 2012-10-31 01:31:10

Re: Intel Graphics - Lags and glitches since linux 3.6?

#8 2012-10-31 01:42:22

Re: Intel Graphics - Lags and glitches since linux 3.6?

#9 2012-11-01 14:30:01

Re: Intel Graphics - Lags and glitches since linux 3.6?

#10 2012-11-05 21:34:40

Re: Intel Graphics - Lags and glitches since linux 3.6?

#11 2012-11-06 00:36:16

Re: Intel Graphics - Lags and glitches since linux 3.6?

#12 2012-11-06 22:57:21

Re: Intel Graphics - Lags and glitches since linux 3.6?

#13 2012-11-08 13:47:42

Re: Intel Graphics - Lags and glitches since linux 3.6?

#14 2012-12-05 00:10:19

Re: Intel Graphics - Lags and glitches since linux 3.6?

Board footer