Skip to contentSkip to conversation

2.6.38-rc8 regressions

DISCUSSION Jan Niehusmann 3 participants 9 messages Mar 14, 2011 – Jul 5, 2011

lore
Share & tools
lore
#0 lore raw

Hello,

just to provide some testing feedback. I didn't have time (and probably not even the necessary skills) to further diagnose these issues. But as I don't remember seeing these problems with 2.6.37, maybe the observations are interesting to you:

With 2.6.38-rc8 I see the following graphics related regressions (relative to 2.6.37) on a Thinkpad X61s with "Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller" (PCI ID 8086:2a02). Userspace is debian/squeeze.

1) Every now and then, terminal windows (urxvt) do not properly update their contents. After issueing a command like 'ls', which writes several lines of text at once, some lines are completely missing. It's not garbled glyphs, but full lines of text completely missing.

Interestingly, sometimes the correct contents of the full window become visible for a split second. So they seem to be 'somewhere' accessible to the GPU, just not shown on the screen.

When a single character in an affected line gets updated (e.g. by marking it with the cursor - or even by an update in a different console window next to the one affected) the correct contents of the full line become and stay visible.

When this problem occurs, it does so reproducibly: About every third command writing to the terminal shows the behaviour. In that situation, just closing and reopening the lid solves the problem: Console windows work as expected again, and AFAICT the problem doesn't reoccur until the next reboot or suspend/resume cycle.

2) I just had a full hang of the GPU (black screen) after opening a web page containing a video. kernel.log contains the following messages:

Mar 14 09:09:37 x61s kernel: [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
Mar 14 09:09:37 x61s kernel: [drm:i915_do_wait_request] *ERROR* i915_do_wait_request returns -11 (awaiting 10172 at 10171, next 10173)
Mar 14 09:09:37 x61s kernel: [drm:init_ring_common] *ERROR* render ring initialization failed ctl 00000000 head 00000000 tail 00000000 start 00000000
Mar 14 09:09:44 x61s kernel: [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
Mar 14 09:09:44 x61s kernel: [drm:init_ring_common] *ERROR* render ring initialization failed ctl 00000000 head 00000000 tail 00000000 start 00000000
Mar 14 09:09:44 x61s kernel: [drm:i915_do_wait_request] *ERROR* something (likely vbetool) disabled interrupts, re-enabling
Mar 14 09:09:44 x61s kernel: [drm:i915_do_wait_request] *ERROR* something (likely vbetool) disabled interrupts, re-enabling
Mar 14 09:09:44 x61s kernel: [drm:i915_do_wait_request] *ERROR* something (likely vbetool) disabled interrupts, re-enabling
Mar 14 09:09:44 x61s kernel: [drm:i915_do_wait_request] *ERROR* something (likely vbetool) disabled interrupts, re-enabling
Mar 14 09:09:44 x61s kernel: [drm:i915_do_wait_request] *ERROR* something (likely vbetool) disabled interrupts, re-enabling
[...]
Mar 14 09:09:45 x61s kernel: [drm:i915_do_wait_request] *ERROR* something (likely vbetool) disabled interrupts, re-enabling
Mar 14 09:09:45 x61s kernel: [drm:i915_do_wait_request] *ERROR* something (likely vbetool) disabled interrupts, re-enabling
Mar 14 09:09:45 x61s kernel: [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
Mar 14 09:09:45 x61s kernel: [drm:i915_do_wait_request] *ERROR* i915_do_wait_request returns -11 (awaiting 10180 at 10171, next 10181)
Mar 14 09:09:45 x61s kernel: [drm:i915_reset] *ERROR* GPU hanging too fast, declaring wedged!
Mar 14 09:09:45 x61s kernel: [drm:i915_reset] *ERROR* Failed to reset chip.

The same (probably) happened ten days ago with 2.6.38-rc6:

Mar  4 10:06:22 x61s kernel: [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
Mar  4 10:06:22 x61s kernel: [drm:i915_do_wait_request] *ERROR* i915_do_wait_request returns -11 (awaiting 50785 at 50784, next 50786)
Mar  4 10:06:22 x61s kernel: [drm:init_ring_common] *ERROR* render ring initialization failed ctl 00000000 head 00000000 tail 00000000 start 00000000
Mar  4 10:06:28 x61s kernel: [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
Mar  4 10:06:28 x61s kernel: [drm:init_ring_common] *ERROR* render ring initialization failed ctl 00000000 head 00000000 tail 00000000 start 00000000
Mar  4 10:06:31 x61s kernel: [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
Mar  4 10:06:31 x61s kernel: [drm:i915_reset] *ERROR* GPU hanging too fast, declaring wedged!
Mar  4 10:06:31 x61s kernel: [drm:i915_reset] *ERROR* Failed to reset chip.

So it doesn't happen particularly often, but still too often to just ignore it.

Regards,
Jan

#1 lore raw
Quoted 26 lines · Jan Niehusmann

Hello,

just to provide some testing feedback. I didn't have time (and probably not even the necessary skills) to further diagnose these issues. But as I don't remember seeing these problems with 2.6.37, maybe the observations are interesting to you:

With 2.6.38-rc8 I see the following graphics related regressions (relative to 2.6.37) on a Thinkpad X61s with "Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller" (PCI ID 8086:2a02). Userspace is debian/squeeze.

1) Every now and then, terminal windows (urxvt) do not properly update their contents. After issueing a command like 'ls', which writes several lines of text at once, some lines are completely missing. It's not garbled glyphs, but full lines of text completely missing.

Interestingly, sometimes the correct contents of the full window become visible for a split second. So they seem to be 'somewhere' accessible to the GPU, just not shown on the screen.

When a single character in an affected line gets updated (e.g. by marking it with the cursor - or even by an update in a different console window next to the one affected) the correct contents of the full line become and stay visible.

Quoted 5 lines

When this problem occurs, it does so reproducibly: About every third command writing to the terminal shows the behaviour. In that situation, just closing and reopening the lid solves the problem: Console windows work as expected again, and AFAICT the problem doesn't reoccur until the next reboot or suspend/resume cycle.

There was a bug in the DDX where we missing a flush (for precisely this style of bug):

commit 4a186a612376bdd6f86c026e8b8b442108868a0a

Author: Chris Wilson <[email protected]>
Date:   Tue Dec 7 16:56:57 2010 +0000

Always flush the batch before blocking for new X requests

This should prevent any lag when waiting upon user input, for example whilst logging in with gdm.

Quoted 2 lines

2) I just had a full hang of the GPU (black screen) after opening a web page containing a video. kernel.log contains the following messages:

The next kernel also includes the hint to look in /sys/kernel/debug/dri/0/i915_error_state, or at least to provide that file for us for debugging GPU hangs. Outside of initialisation, suspend and resume, and modesetting the cause of a GPU hang is usually an invalid batch buffer submitted by userspace. The i915_error_state should capture that erroneous batch buffer.
-Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
#2 lore raw

Hi Chris,

Quoted 6 lines · Chris Wilson
Quoted 4 lines · Jan Niehusmann

1) Every now and then, terminal windows (urxvt) do not properly update their contents. After issueing a command like 'ls', which writes several lines of text at once, some lines are completely missing. It's not garbled glyphs, but full lines of text completely missing.

[...]

Quoted 7 lines

There was a bug in the DDX where we missing a flush (for precisely this style of bug):

commit 4a186a612376bdd6f86c026e8b8b442108868a0a

Author: Chris Wilson <[email protected]>
Date:   Tue Dec 7 16:56:57 2010 +0000
Quoted 7 lines

Always flush the batch before blocking for new X requests

This should prevent any lag when waiting upon user input, for example whilst logging in with gdm.

I cherry-picked this patch into the source code of 2.13.0, as provided by the debian package xserver-xorg-video-intel version 2.13.0-5.

With that patch applied, I still observed the described behaviour. Additionally, some java application had display update problems. (But java generally has some problems because I'm using a non-reparenting window manager, 'awesome', which java doesn't like).

What do you think, would it be worthwhile to try a more recent version of xf86-video-intel?

Regards,
Jan

#3 lore raw
Quoted 7 lines · Jan Niehusmann

With that patch applied, I still observed the described behaviour. Additionally, some java application had display update problems. (But java generally has some problems because I'm using a non-reparenting window manager, 'awesome', which java doesn't like).

What do you think, would it be worthwhile to try a more recent version of xf86-video-intel?

Ok, that's more worrying. That bug certainly matched what you describe, and I don't offhand know of another commit since 2.13 that is relevant. It would be good to double-check first though.

Have you harvested any i915_error_states?
-Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
#4 lore raw
Quoted 6 lines · Chris Wilson
Quoted 5 lines · Jan Niehusmann

With that patch applied, I still observed the described behaviour. Additionally, some java application had display update problems. (But java generally has some problems because I'm using a non-reparenting window manager, 'awesome', which java doesn't like).

Quoted 6 lines
Quoted 2 lines

What do you think, would it be worthwhile to try a more recent version of xf86-video-intel?

Ok, that's more worrying. That bug certainly matched what you describe, and I don't offhand know of another commit since 2.13 that is relevant. It would be good to double-check first though.

Same with 2.14.901 from debian/experimental.

Jan

#5 lore raw
In reply to Chris Wilson
Quoted 6 lines · Chris Wilson
Quoted 5 lines · Jan Niehusmann

With that patch applied, I still observed the described behaviour. Additionally, some java application had display update problems. (But java generally has some problems because I'm using a non-reparenting window manager, 'awesome', which java doesn't like).

Quoted 6 lines
Quoted 2 lines

What do you think, would it be worthwhile to try a more recent version of xf86-video-intel?

Ok, that's more worrying. That bug certainly matched what you describe, and I don't offhand know of another commit since 2.13 that is relevant. It would be good to double-check first though.

I bisected the display update issues to start between 0b0b053a and c64f7ba5. fe669bf8 doesn't work at all (black screen after X startup), and I didn't try 1b6064d7 yet:

bad: c64f7ba agp/intel: Remove confusion of stolen entries not stolen memory
not tried: 1b6064d agp/intel: Remove the artificial cap on stolen size
crashes: fe669bf drm/i915: Compute physical addresses from base of stolen memory
good: 0b0b053 drm/i915/panel: Restore saved value of BLC_PWM_CTL

Another observation I made is that everything is fine while I am using a dual-screen setup. As dual-screen disables frame buffer compression, I tried to i915.powersave=0, and indeed, with this parameter I was unable to reproduce the issues, as well.

Is there anything else I could try? Unfortunately, because of conflicting changes, it's not easily possible to just revert these commits from 2.6.38.

Regards,
Jan

#6 lore raw
In reply to Jan Niehusmann
Quoted 4 lines · Jan Niehusmann

1) Every now and then, terminal windows (urxvt) do not properly update their contents. After issueing a command like 'ls', which writes several lines of text at once, some lines are completely missing. It's not garbled glyphs, but full lines of text completely missing.

Just a quick update on the issue I reported a while ago: While I still see this problem with 2.6.38.2, it seems like it's fixed with current 2.6.39 development tree at commit 40bd8ea1.

Out of curiosity, I reverted 36d527de ("Restore missing command flush before interrupt on BLT ring") from that, and was able to reproduce the bug, again.

So it looks like this is fixed now.

Jan

#7 lore raw
Quoted 3 lines · Jan Niehusmann

Just a quick update on the issue I reported a while ago: While I still see this problem with 2.6.38.2, it seems like it's fixed with current 2.6.39 development tree at commit 40bd8ea1.

[...]

Quoted 1 line

So it looks like this is fixed now.

Sorry, I have to take this back: I just observed the screen corruption with 2.6.39-rc2. So it may just be more difficult to trigger. (Or it's just coincidence that I wasn't able to reproduce it at first.)

Jan

#8 lore raw

Hello,

Quoted 3 lines · Jan Niehusmann <jan <at> gondor.com> writes:

Sorry, I have to take this back: I just observed the screen corruption with 2.6.39-rc2. So it may just be more difficult to trigger. (Or it's just coincidence that I wasn't able to reproduce it at first.)

I am also still seeing the exact same problem you describe with 2.6.39 (Arch Linux) and the xorg intel driver built from git master HEAD today (98f2e38), both with and without SNA. See this thread on the Arch Linux Forums for more information:
https://bbs.archlinux.org/viewtopic.php?id=117202

Moritz