gpu driver crashes system during suspend and lock screen

The i915 kernel module seems to crash my system during suspend and lock screen. I tested on multiple kernel versions, using Fedora 32. The lock screen consistenly crashes my system for all tested kernels from 5.4 to 5.8. Crash means that after locking the screen goes black and after about 15 seconds never comes back, regardless of which keys and buttons I press. I also, ssh sessions are interrupted after about 15 seconds.

Furthermore, suspend crashes my system from kernel 5.7.6 onward. 5.7.5 still works fine. Suspend crashes are also indicated by the power button light not turning off.

Now what makes me suspect that this is an issue with the i915 kernel module? I added i915.modeset=0 to grub and the problems somewhat disappear. Lock screen works on all tested kernels: It successfully locks the screen and remains responsive to ssh for an extended period of time (tried up to 30 minutes). However, the screen does not turn back on. But I can type my password and turn off the system without the screen displaying anything.

Similar story for suspend, it now works on all tested kernels, but the screen stays black after wakeup. However, I can use the computer without screen to unlock and shutdown and ssh works after wakeup as well.

Now my hardware: Dell Latitude 5590 Intel(R) Core(TM) i5-8350U CPU @ 1.70GHz

I do not know how to further debug this issue. journalctl output is not helpful. Please tell me how I can help fix this issue.

Hi Reporter, can you try to reproduce this issue on latest drm-tip.

https://cgit.freedesktop.org/drm-tip

Please post your observations on it and any steps followed to reproduce the issue.

Hi @ysainan, I am afraid I cannot simply install the latest drm-tip. I am not much of a developer and have no clue about the whole drm stack. I see that Fedora ships a libdrm package which I assume I could simply build with a more recent version, but I suppose that this is note quite related to drm-tip.

However, if it helps, the problem persists with Fedora 33 and kernel 5.9.11

@laolux, The updates for Fedora 33 should include a 5.11 kernel now, does that fix things?

If that does not fix, try installing the latest kernel from rawhide (that is pretty close to drm-tip), take the latest successful kernel build from here:

https://koji.fedoraproject.org/koji/packageinfo?packageID=8

And then follow these instructions for installing it: https://fedorapeople.org/~jwrdegoede/kernel-test-instructions.txt

@jwrdegoede thanks for the instructions. A couple days ago I upgraded my system to Fedora 34 beta, in the hope that it would improve the situation, but to no avail. Kernel is 5.11.11-300.fc34.x86_64. However one time out of about 15 tries within the last couple of days my system did suspend correctly and woke up again. And I was immediately greeted by abrt that xorg-x11-drv-intel had encountered a problem and recovered. abrt created a new bug for it, along with all the automatically collected information, if that is of help: https://bugzilla.redhat.com/show_bug.cgi?id=1945807

Now, just for "fun" I decided to install the latest kernel from koji, 5.12.0-0.rc5.180.fc35.x86_64. This does not fix all my problems, but I can now successfully suspend about 50% of my reboots. I rebooted my computer 6 times and on tries 1, 4 and 6 I was able to suspend and wake up. The other times my computer crashed just like before.

Any more information I can help with?

@laolux one more thing which you can try is removing the xorg-x11-drv-intel package, Fedora no longer install that by default, since the modesetting drivers works just as well and the modesetting driver tends to have less issues.

To remove it run: sudo dnf remove xorg-x11-drv-intel, if you want you can later re-install it (not recommended) by doing: sudo dnf install xorg-x11-drv-intel

@jwrdegoede thanks, I tried that, but to no avail. I also tried with live usb of Fedora 31 and Fedora 34 beta, but both crash upon suspend about 50% of the tries.

I also tried Windows 10, which came preinstalled on this laptop and which I need to keep around on a separate partition (workplace rules). This was also hit and miss. Sometimes suspend (sleep) would work, sometimes freeze the computer, sometimes the computer would turn off power a couple of seconds after getting the sleep command.

If things don't work under Windows too, this sounds like it is not a Linux problem, but rather a firmware or hardware issue.

If the laptop is still under warranty I would try installing the latest BIOS under Windows and if that does not solve the issue under Windows, then IMHO it is time to file a warranty claim with your vendor.

Thanks, I tried updating to the latest BIOS, directly from Windows, as lvfs does not have the most recent versions. To no avail. So maybe it is a hardware issue. But warranty has long expired, so cannot go that route. Maybe I will find some time to completely reset the BIOS and freshly install the OS, but currently I cannot afford to break/brick my device.

Ok, I reset everything and tried again. This time Windows worked reliably across ~14 reboots and more then 20 sleep wake cycles, with Fedora booting and failing to sleep in between. Last try on Windows led to some hardware crash with the indicator LED (not power button) flashing amber twice and white once. I haven't been able to figure out what that indicates. Dell has a very nice user manual for my Latitude 5590, but not a word on the led error codes. So I assume that I do have a hardware issue, but strangely Linux does not trigger the corresponding led error code and after reseting and updating Windows runs in to the problem significantly less than Linux. Not sure what to make of this and if it is worth continuing to try to fix this issue.

I decided to give up on it. I've spent enough time just trying to understand the cause of the issue that I could have easily invested into simply restarting my computer multiple times daily. For whenever I need to take my laptop somewhere to present, I will have to ensure not to close the lid and leave it running and simply ignore my coworker's comment that linux is not a desktop OS. ysainan and @jwrdegoede, thanks for your time and effort trying to help, I really appreciate it.

closed

Hi @laolux and thank you for your report. I am in the same situation as you and your fix (i915.modeset=0) solved the crash. Thank you !

The laptop worked fine with Debian 9 for a couple of years. I switched to F34 (5.12.9-300.fc34.x86_64) and the crash started as you precisely described. However there is a difference : as soon as I type or click the screen wakes up I am able to type the password (not blindly).

The laptop model specs are :

DELL PC portable Inspiron 13-7370 - 13,3" FHD - Core i7-8550U - RAM 16Go - Stockage 512 Go SSD

I really hope a fix comes around. Is there anything that I could try to help you guys, @ysainan and @jwrdegoede, to fix it ?