search

ty2 / ideapad-laptop-tb2024g6plus

The IdeaPad ACPI Extras kernel modules for ThinkBook 2024 G6+
GNU General Public License v2.0
29 stars 4 forks source link

Problems with newer kernels >= 6.10 #5

Open mikamiel opened 2 months ago

mikamiel commented 2 months ago

Laptop: ThinkBook 16 G6+ IMH (Intel-based) OS: Ubuntu 24.04 LTS/Gnome

With some workarounds was able to build this kernel module for 6.8.x kernel(s). And it works pretty good.

But with newer kernels 6.10/6.11 quite noticeable UI-stuttering permanently occurs after entering standby mode and returning to awaked state (by closing-opening lid or via choosing standby menu option). I suppose it's not even this patch fault but some bug in the kernel, because I experience the same with this module disabled (but maybe doing smth wrong - not an expert in kernel modules tinkering ;) ).

CPU graph during normal operation: normal

and after awakening from standby mode (it's not some process activity, but kernel i suppose): stuttering

ty2 commented 2 months ago

I can’t reproduce the issue on my end.

Which version of the module did you install? Maybe try removing it or recompiling and reinstalling it from the main branch to see if the issue still happens.

ty2 commented 2 months ago

check dmesg and journalctl for any odd messages after waking from standby. The logs might help identify the cause.

mikamiel commented 2 months ago

Thanks a lot for reminding about dmesg)! I've found the source of this annoying UI lagging. And it seems to have nothing to do with your kernel module.

Looking through dmesg output: e1000e 0000:00:1f.6: PM: pci_pm_suspend(): e1000e_pm_suspend [e1000e] returns -2 and a lot of such strings: workqueue: e1000_watchdog_task [e1000e] hogged CPU for >10000us 35 times, consider switching to WQ_UNBOUND

Quick googling lead me to: Bug 218936 - e1000e won't enter suspend without a network connection

Most of the time i'm using wifi, so rmmod e1000e for now solves the problem.

mikamiel commented 2 months ago

@ty2 if you don't mind, let's not close the issue until kernel devs fix e1000e driver regression, so that users who face that problem could easily find it here.