Strange dmessage upon upgrade to 6.2

[mefisto]

Greetings all,

after upgrading from 6.1 to 6.2 a message "vblank wait timed out on crtc 1", repeated five times appears during startup.

My search does not seem to result in any relevant topic or solution; consequently, any help would be appreciated.

Kindest regards,

M

[johnR]

That's something to do with your graphics hardware. Were there any changes that could affect your particular hardware (you didn't mention what you're using) between 6.1 and 6.2?

[LeFrettchen]

You can try to change the acceleration method of your graphic driver.

Depending on your graphic card, you've got several options for AccelMethod.

In xorg.conf :

Section "your_device"
  Identifier "your_dev_name"
  Driver "your_driver_name"
  AccelMethod "off" // or "SNA" or "UXA" or ...
EndSection

I.e. with a intel HD graphics, you can find options here : OpenBSD man's page - intel(4)

[jggimi]

The message is produced by sys/dev/pci/drm/drm_irq.c, which provides IRQ and vblank support for DRM in the kernel. It was added as part of revision 1.68, one of the two revisions to this module between OpenBSD-release 6.1 and 6.2.

----------------------------
revision 1.68
date: 2017/07/01 16:14:10;  author: kettenis;  state: Exp;  lines: +1076 -478;  commitid: KnwRPOZok9A30HI4;
Update inteldrm(4) to code based on Linux 4.4.70.  This brings us support for
Skylake and Cherryview and better support for Broadwell and Valleyview.  Also
adds MST support.  Some tweaks to the TTM code and radeondrm(4) to keep it
working with the updated generic DRM code needed for inteldrm(4).

Tested by many.
----------------------------

[mefisto]

Hi johnR,

thank you for your answer. It is the same hardware, just re-installation 6.2 instead of 6.1.

Hi LeFrettchen,

thank you for your suggestion. I will try that, but please also see my reply to jggimi.

Hi jggimi,

thank you for your answer. Since you noticed the source emitting the message, I wonder if there is not any relation with another message just received on a console: "error: [drmid81937:intel_pipe_update_start] *ERROR* Potential atomic update update failure on pipe A".

How to approach debugging?

Kindest regards,

M

[jggimi]

Originally Posted by mefisto

...I wonder if there is not any relation with another message just received on a console..

Yes. This is produced by src/sys/dev/pci/drm/i915/intel_sprite.c, which also had the same change log entry (revision 1.8) on July 1, and this message was also included as part of the code in this revision.

How to approach debugging?

It is not clear to me that there is a bug. These messages are produced when hardware is not behaving as expected.

• To understand the root cause of these messages, on your own, you will need to read the source code to determine why these warnings are produced.
• If you want assistance from the OpenBSD Project to understand why these messages are produced, and whether you should be concerned about them, you can submit an informal query to the Project's misc@ mailing list. At minimum, you must include your dmesg(8) so that your hardware environment and your specific kernel that is reporting the messages are made clear.
• Formal problem reports are sent to the bugs@ mailing list, most often through the use of the sendbug(1) tool.

You can do the same sort of research I did. It isn't debugging, and all it takes is a cvs(1) working directory of the kernel source. Example:

$ find /usr/src/sys -type f -exec grep "Potential atomic update failure" {} +

$ cd /usr/src/sys/dev/pci/drm/i915
$ cvs log intel_sprite.c | less
$ cvs blame intel_sprite.c | less

[mefisto]

Hi jggimi,

thank you for the pointers.

Kindest regards,

M

[ibara]

Additionally, if one does not want to or cannot check out the cvs tree, the site http://bxr.su can help with your search.
For example:
http://bxr.su/search?q=%22Potential+...roject=OpenBSD
You can get to the CVSweb from any source file, which does what log and blame do, effectively.

[jggimi]

Good suggestion!

It may help to know that cvs blame is a synonym for cvs annotate.