My Lenovo Laptop cannot be left standing for any lengthy interval without requiring reboot

[stevecoh1]

Thanks, Alan!

@popey

1. There are repeated problems with the Logitech devices:

…

(Unless this was at the time you were manually disconnecting the devices?)
…

No, this was not when manually disconnecting the devices. But it is almost exactly around the time when I did the reboot.

@popey

2. The system is showing severe input lag with the Logitech mouse:

Feb 17 15:11:04 aigas /usr/libexec/gdm-x-session[2415]: (EE) event8  - Logitech USB Receiver Mouse: client bug: event processing lagging behind by 652ms, your system is too slow

(Do you notice lag?)

Did I notice lag: certainly. Symptom of the crash was failure to respond to mouse clicks or keyboard presses, which caused me to have to reboot. I suppose I should have tried these things on the laptop, but I typically turn off the touchpad, which I hate. Don’t remember if I pressed keys on the laptop.

@popey

3. There’s an RCU (Read-Copy Update) stall detected on multiple CPUs:

Feb 17 15:10:49 aigas kernel: rcu: INFO: rcu_preempt detected stalls on CPUs/tasks:

(Are there any “chunky” applications you use in virtual machines?)

Not sure what these might be but now, I think no. I haven’t really even gotten to use my astrophotography apps yet. Just web browsing for the most part.

@popey

4. The NVIDIA driver appears to be involved in the lockup, as shown in the stack trace:

Feb 17 15:10:49 aigas kernel:  os_acquire_spinlock+0x12/0x30 [nvidia]

(Have you changed any settings in the nvidia-settings application, or anywhere else, beyond just installing the driver?)

No. I haven’t done anything that explicitly even uses the GPU yet after that last brouhaha we had three weeks ago, let alone mess with any settings there. I am back to factory config of 22.04 LTS with nothing done to the GPU settings.

@popey

5. The rtkit-daemon is constantly having to demote real-time threads, indicating severe scheduling issues:

Feb 17 15:11:59 aigas rtkit-daemon[1571]: The canary thread is apparently starving. Taking action.

So that’s what that “canary” business is. I figured it might be.

@popey

Potential solutions to try:

1. Disable NVIDIA power management features in the GPU driver
2. Try using a different USB port for the Logitech receiver (preferably one on a different USB controller) - this might be hard, if you have a laptop with limited connectivity
3. Consider trying the (unsupported) Canonical kernel team “PPA” (not actually a PPA) of a newer upstream kernel
4. Consider updating or rolling back the NVIDIA driver
5. Add kernel boot parameters to disable CPU power states (like intel_idle.max_cstate=1)

I would pick one of these, and test it, and if it doesn’t work, undo it, and move to another, systematically.

It seems like I’ve already done #2: not just switched USB ports but eliminated solaar from the picture altogether, which felt like my leading suspect after I disproved the browser stuff. I had added solaar to the picture a few days ago so I could use just one usb connector for both peripherals, but that seems awfully dodgy to me now. I solved that by getting usb out of the picture entirely and using bluetooth instead.

#1 I have no idea of the pros or cons messing with NVIDIA power management
#3 I don’t feel like touching with a ten-foot pole.
#4 was what that whole “brouhaha” thread I linked to was about. Don’t feel like going back there until something changes on the OEM end.
#5 seems like maybe a good idea but I don’t know much about it. I would need help if I decide to do it.

Right now, things seem to be in a sane state, so I don’t want to mess around further unless I have to. But I suppose I should get the latest journalctl entries to see if any of these canaries or other beasts are in the coal mine. How would I do that?

And thanks again for your detailed suggestions.