Bug 97515

Summary: System lockups with 4.7 kernel with PSR enabled, general unstability since 4.5
Product: DRI Reporter: Ruud van Asseldonk <freedesktop>
Component: DRM/IntelAssignee: Jim Bride <jim.bride>
Status: CLOSED INVALID QA Contact: Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity: normal    
Priority: medium CC: intel-gfx-bugs, przanoni
Version: XOrg git   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: SKL i915 features: display/PSR
Attachments:
Description Flags
Dmesg output with drm.debug=0x1e
none
Dmesg output with drm.debug=0x06
none
Xorg.0.log (same run as drm.debug=0x06 dmesg) none

Description Ruud van Asseldonk 2016-08-28 12:21:36 UTC
First of all, my apologies for the fuzzy bug report. I’m not even sure whether this is an i915 issue or something else in the kernel. Please let me know if there is anything I can do to help diagnose this issue.

My hardware:

A Dell XPS 15 9550, with i7-6700HQ and GeForce GTX 960M. As far as I am aware, I am not using the discrete gpu. I don’t have Nouveau installed.

Software:

Kernel 4.4.5-1-ARCH (Arch Linux) was the last kernel without issues, after that several regressions occurred. I have enabled panel self-refresh by appending `i915.enable_psr=1` to my kernel parameters. With this kernel I experience no lockups or freezes, and an external monitor works as expected.

In kernel 4.5.1, the display would freeze regularly for a few seconds. This happened every five minutes or so, and it would happen consistently every session. The cursor was not movable either. It was annoying, but the system was still usable apart from freezes. When plugging in an external display I experienced issues too. It is a few months ago and I don’t recall exactly what happened, but it did freeze my machine (I don’t remember whether I closed the lid before this happened) and I had to power it off. I downgraded my kernel to 4.4.5 until 4.6.2 became available.

4.6 included some fixes that looked promising, but I still experienced the freezes. I had the feeling that they were less frequent, but I have no hard data on this. I did not try using an external monitor. I tried 4.6.2, 4.6.3, and 4.6.4 when they became available, but the issue persisted, so I downgraded to 4.4.5 every time.

With 4.7 a worse regression happened: I can boot into my desktop environment (Gnome) perfectly fine, and I see the login animation with my wallpaper appearing. I can move the cursor. But otherwise the desktop is frozen completely; it does not respond to clicks and nothing except for the cursor is updated. I can switch to a terminal using ctrl+alt+f2. It takes a few seconds to respond, and even in the terminal there are frequent lockups of a few seconds where the system does not respond to input at all. But at least I was able to log in and downgrade my kernel. Sometimes I also succeeded in switching back to my desktop environment, and it would work for a few seconds, and then freeze again. I experience this issue with kernel 4.7.1 and 4.7.2.

Throughout the process I’ve kept my xf86-video-intel package up to date. The Arch package version I have installed currently is 1:2.99.917+697+g12c14de-1.

System architecture: x86_64
Linux distribution: Arch Linux

I have just updated my UEFI to the latest version (1.2.13).

I’ll do another run with 4.7.2 and attach the dmesg debug output and xorg log in a moment.
Comment 1 Ruud van Asseldonk 2016-08-28 12:52:44 UTC
Created attachment 126083 [details]
Dmesg output with drm.debug=0x1e
Comment 2 Ruud van Asseldonk 2016-08-28 12:53:10 UTC
Created attachment 126084 [details]
Dmesg output with drm.debug=0x06
Comment 3 Ruud van Asseldonk 2016-08-28 12:54:14 UTC
Created attachment 126085 [details]
Xorg.0.log (same run as drm.debug=0x06 dmesg)
Comment 4 Ruud van Asseldonk 2016-08-28 12:58:07 UTC
I booted 4.7.2 twice, once with drm.debug=0x1e and once with drm.debug=0x06, see attachments.

Then I removed i915.enable_psr=1 from my kernel parameters, and now everything appears to work smoothly; I haven’t experienced a single freeze since boot. So it seems that PSR causes all of these issues. (I haven’t tried connecting an external monitor yet.) I’ll keep PSR disabled for now.
Comment 5 Peter Y. Chuang 2016-10-21 20:54:10 UTC
I believe I have been having the same problem as well. Disabling PSR prevents those freezes.

My machine is Dell XPS 13 (9350). As Ruud described, the problem began since kernel 4.5, and I can confirm that the freezes still exist in kernel 4.8.

I've tried removing xf86-video-intel entirely, and the freezes seem to happen slightly less frequently, but they are still there.
Comment 6 Paulo Zanoni 2016-12-13 20:44:15 UTC
PSR is disabled by default on SKL exactly because it's known to cause bugs. As a general rule, we really don't recommend setting any i915 options to enable or disable features. But I see you enabled PSR to work around another bug, so you do have a valid reason there, but it seems we've already fixed those problems.

Due to that, I'm closing this bug report. If you still experience issues with the default Kernel parameters, please feel free to reopen the bug report and provide the appropriate log files.

Thanks,
Paulo
Comment 7 Ruud van Asseldonk 2016-12-13 22:49:43 UTC
To summarise: I had i915.enable_psr=1 enabled on 4.4.5-1-ARCH, and that worked fine. On later kernels this resulted in instabilities, so I disabled PSR. My system is now stable again, but now I don’t have PSR.

I did not enable PSR to work around another bug, I enabled it to save power.

Using PSR with later kernels is still an issue (or maybe this has been fixed, frankly I have not tried enabling PSR again after reporting this), but if you are aware of this and if you have more specific reports for what is happening there then feel free to close this one.
Comment 8 Paulo Zanoni 2016-12-14 12:32:55 UTC
(In reply to Ruud van Asseldonk from comment #7)
> To summarise: I had i915.enable_psr=1 enabled on 4.4.5-1-ARCH, and that
> worked fine. On later kernels this resulted in instabilities, so I disabled
> PSR. My system is now stable again, but now I don’t have PSR.
> 
> I did not enable PSR to work around another bug, I enabled it to save power.

Looks like I misunderstood you. Thanks for the clarification.

> 
> Using PSR with later kernels is still an issue (or maybe this has been
> fixed, frankly I have not tried enabling PSR again after reporting this),
> but if you are aware of this and if you have more specific reports for what
> is happening there then feel free to close this one.

Yes, PSR is known to cause problems on all platforms, unfortunately, so it's staying disabled by default for now.
Comment 9 Dhinakaran Pandiyan 2018-08-16 07:53:53 UTC
(In reply to Ruud van Asseldonk from comment #7)
> To summarise: I had i915.enable_psr=1 enabled on 4.4.5-1-ARCH, and that
> worked fine. On later kernels this resulted in instabilities, so I disabled
> PSR. My system is now stable again, but now I don’t have PSR.
> 
> I did not enable PSR to work around another bug, I enabled it to save power.
> 
> Using PSR with later kernels is still an issue (or maybe this has been
> fixed, frankly I have not tried enabling PSR again after reporting this),
> but if you are aware of this and if you have more specific reports for what
> is happening there then feel free to close this one.


Do you mind giving i915.enable_psr=1 another try now?
Comment 10 Ruud van Asseldonk 2018-08-18 14:28:55 UTC
> Do you mind giving i915.enable_psr=1 another try now?

Sure, I enabled it now. In the ten minutes since boot I have not experienced any issues. I will report back in a few days. My current kernel version is 4.18.1-arch1-1-ARCH.
Comment 11 Ruud van Asseldonk 2018-09-02 14:12:25 UTC
PSR appears to be working nicely, I haven’t experienced any issues in the two weeks after enabling it.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.