Bug 188833 - [suspend/resume] Suspend/resume with Intel GMA HD 4000: AIGLX fails to restart
Summary: [suspend/resume] Suspend/resume with Intel GMA HD 4000: AIGLX fails to restart
Status: Closed Overcome By Events
Alias: None
Product: Base System
Classification: Unclassified
Component: kern (show other bugs)
Version: 10.0-RELEASE
Hardware: Any Any
: Normal Affects Only Me
Assignee: freebsd-bugs (Nobody)
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-04-21 00:30 UTC by ivan.rokotov.bsd
Modified: 2018-01-14 21:02 UTC (History)
7 users (show)

See Also:
bahlgren: maintainer-feedback? (x11)


Attachments
xf86-video-intel batch buffer retry patch (849 bytes, patch)
2015-04-07 12:50 UTC, Bengt Ahlgren
no flags Details | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description ivan.rokotov.bsd 2014-04-21 00:30:00 UTC
My Intel GMA HD 4000 on Lenovo IdeaPad S400 with the latest VT, drm2 and i915kms from 10-STABLE can't resume properly because AIGLX fails to wake up. Of course, the graphics after it non-usable or horribly corrupted, like here:
http://i327.photobucket.com/albums/k477/ivan_rokotov_bsd/2_zpsd7564e4b.png

I found this patch, which solved the problem completely:
http://lists.freebsd.org/pipermail/freebsd-x11/2013-October/013727.html

Now, I can resome, and have this in X log after it:
[  4561.749] (WW) intel(0): retrying batchbuffer submit
[  4561.803] (WW) intel(0): retrying batchbuffer submit
[  4561.857] (WW) intel(0): retrying batchbuffer submit
[  4561.908] (WW) intel(0): retrying batchbuffer submit
[  4561.959] (WW) intel(0): retrying batchbuffer submit
[  4562.012] (WW) intel(0): retrying batchbuffer submit
[  4562.063] (WW) intel(0): retrying batchbuffer submit
[  4562.117] (WW) intel(0): retrying batchbuffer submit
[  4562.171] (WW) intel(0): retrying batchbuffer submit
[  4562.225] (WW) intel(0): retrying batchbuffer submit
[  4562.279] (WW) intel(0): retrying batchbuffer submit
[  4562.333] (WW) intel(0): retrying batchbuffer submit
[  4562.386] (WW) intel(0): retrying batchbuffer submit
[  4562.439] (WW) intel(0): retrying batchbuffer submit
[  4562.493] (WW) intel(0): retrying batchbuffer submit
[  4562.547] (WW) intel(0): retrying batchbuffer submit
[  4564.779] (II) AIGLX: Suspending AIGLX clients for VT switch
[  4564.779] (WW) intel(0): drmDropMaster failed: Unknown error: -22
[  4565.994] (II) AIGLX: Resuming AIGLX clients after VT switch
[  4566.096] (II) intel(0): EDID vendor "CMN", prod id 5239
[  4566.096] (II) intel(0): Printing DDC gathered Modelines:
[  4566.096] (II) intel(0): Modeline "1366x768"x0.0   71.59  1366 1410
1439 1512  768 771 775 789 -hsync -vsync (47.3 kHz eP)

but everything is alive: compiz restarts correctly, GL applications
work as they did before resume. This is very stable, I suspended/resumed my notebook more than ten times for 4-5 days (without shutting down) - and it works.

The ret value obtained with this code
xf86DrvMsg(scrn->scrnIndex, X_WARNING, "retrying batchbuffer submit,
ret=%d\n", ret);

is:

[  8572.583] (WW) intel(0): retrying batchbuffer submit, ret=-16

Fix: 

The retry hack proposed by Jan Kokemüller <jan.kokemueller@gmail.com> here
http://lists.freebsd.org/pipermail/freebsd-x11/2013-October/013727.html
solves the problem.
How-To-Repeat: Install FreeBSD with the latest VT code on Lenovo IdeaPad S400 and try to do suspend/resume from X.
Comment 1 Nikola Lečić 2014-06-22 11:26:32 UTC
I forgot to add here how the system fails to restart without the aforementioned patch. With AIGLX activated, I get this horrible corruption:
http://i327.photobucket.com/albums/k477/ivan_rokotov_bsd/2_zpsd7564e4b.png
and no moving or switching to console help. No GL application starts.
I also found this in the X log after resume:

[  1173.617] (EE) intel(0): Failed to submit batch buffer, expect
rendering corruption: Device busy.
 (II) AIGLX: Suspending AIGLX clients for VT switch
[  1173.663] (WW) intel(0): drmDropMaster failed: Unknown error: -22
[  1177.289] (II) AIGLX: Resuming AIGLX clients after VT switch
[  1177.340] (II) intel(0): EDID vendor "CMN", prod id 5239
[  1177.340] (II) intel(0): Printing DDC gathered Modelines:
[  1177.340] (II) intel(0): Modeline "1366x768"x0.0   71.59  1366 1410
1439 1512  768 771 775 789 -hsync -vsync (47.3 kHz eP)
Comment 2 Nikola Lečić 2014-06-22 11:29:07 UTC
(This is on Lenovo IdeaPad S400 as in the case of original poster.)
Comment 3 Bengt Ahlgren 2015-04-07 12:50:56 UTC
Created attachment 155303 [details]
xf86-video-intel batch buffer retry patch

It would be great if the simple patch by Jan Kokemüller that Ivan points to could be added to x11-drivers/xf86-video-intel/files!  I'm attaching that patch for easy access. I have been using this patch on a TP X201 for quite some time.
Comment 4 Graham Perrin freebsd_committer freebsd_triage 2015-12-30 16:11:50 UTC
https://lists.freebsd.org/pipermail/freebsd-x11/2015-December/017055.html suggests special attention to this bug. 

I can't comment on Intel GMA HD, 

https://forums.freebsd.org/threads/suspend-resume-problem.13311/#post-307411 describes a workaround that's applicable with PC-BSD 11.0-CURRENT with an Intel HD Graphics 5500, but the symptom there is blackness (not corruption) on resume.
Comment 5 Niclas Zeising freebsd_committer freebsd_triage 2018-01-12 22:12:24 UTC
Hi!

Is this still an issue with an updated xorg stack, possibly also using drm-next-kmod?
Comment 6 Bengt Ahlgren 2018-01-14 19:23:04 UTC
(In reply to Niclas Zeising from comment #5)
Not an issue for me since quite some time with a newer xorg stack (possibly also due to upgrading to 10.x from 9.x).
Comment 7 Niclas Zeising freebsd_committer freebsd_triage 2018-01-14 21:02:29 UTC
Closing this, submitter says the issue is fixed.