Random log out back to login screen with "nouveau: kernel rejected pushbuf: No such device"

meriops commented 3 months ago

Sway Version: 1.10-dev-4d4c88f0 (Jul 23 2024, branch 'master')
Debug Log: sway.log
Configuration File: config file
Stack Trace: stack trace
Description: Basically, sway randomly logs me out of the session and displays the login screen (greetd + greetd-tuigreet). I cannot reproduce in a consistent way, as it really seems random with respect to both when and how often it happens. The sway debug file indicates an error mentioning wezterm, but the crash happens too with foot or alacritty as the only running terminal. It does not happen when I ssh to the machine, which makes me think it is sway / wlroots related. It might as well be something I messed up somewhere in a configuration file since I have a similar configuration on two other machines (desktop, laptop) with different hardware, where the crash does not happen (although it did happen at some point on the laptop, but 'went away' after a while...). I would greatly appreciate some pointers to try and figure out what I'm doing wrong here, since this crash really hinders my work... Please let me know where to post this if here is not the right place.

emersion commented 2 months ago

The backtrace you provided doesn't contain debug symbols. This most likely happens because the Sway binary you're using doesn't have debug information bundled.

Can you try again with a manually compiled Sway binary? See https://github.com/swaywm/sway/wiki/Development-Setup#compiling-as-a-subproject

meriops commented 2 months ago

Thank your for your reply. I am in a busy period right now, so it will likely take some days/weeks before I can get back to you with the proper debug information.

meriops commented 2 months ago

I managed to try with a manually compiled sway... I added here the log file, the coredump, and the output of gdb... hopefully, there's enough information now. Please let me know if it is not the case. Thanks again for your help!

emersion commented 2 months ago

Sounds like this a nouveau bug:


nouveau: kernel rejected pushbuf: No such device
nouveau: ch16: krec 0 pushes 1 bufs 4 relocs 0
nouveau: ch16: buf 00000000 0000000e 00000004 00000004 00000000 0x7b946e870000 0x7f0000 0x80000
nouveau: ch16: buf 00000001 00000006 00000004 00000000 00000004 0x7b94835fa000 0x21c000 0x1000
nouveau: ch16: buf 00000002 0000002c 00000002 00000000 00000002 (nil) 0xf21000 0x1000
nouveau: ch16: buf 00000003 0000002e 00000004 00000004 00000000 0x7b94835d9000 0xf22000 0x1000
nouveau: ch16: psh 00000000 000002a214 000002a264
nouveau:    0x200181c2
nouveau:    0x03303210
nouveau:    0x200681c3
nouveau:    0x00001020
nouveau:    0x00000018
nouveau:    0x00000018
nouveau:    0x00000001
nouveau:    0x00000000
nouveau:    0x00000000
nouveau:    0x20088100
nouveau:    0x00000000
nouveau:    0x00f22000
nouveau:    0x00000000
nouveau:    0x00f21000
nouveau:    0x00000060
nouveau:    0x00000080
nouveau:    0x00000018
nouveau:    0x00000018
nouveau:    0x200180c0
nouveau:    0x00000686
nouveau: kernel rejected pushbuf: No such device
nouveau: ch16: krec 0 pushes 1 bufs 0 relocs 0

atk91 commented 2 months ago

Happens to me as well and this is pretty frustrating to be logged out unexpectedly.

I have noticed that is not so random, it happens when you move mouse cursor so it changes shape (e.g. from textbox to non-textbox, from and out firefox address bar, etc.) or from window to background if gaps are enabled. It does not happen every time, like 1 out of 100, approximately, so you should move mouse quickly in order to reproduce.

BlakeOrth commented 1 month ago

I had initially thought @atk91's mouse cursor theory was a bit far fetched, but moving my mouse in a way that caused rapid cursor changes indeed caused a crash with this signature relatively quickly.

If anyone else is running into this issue, I had been crashing at least daily, and certain web tools would often crash within minutes of usage. The workaround seems to be setting WLR_NO_HARDWARE_CURSORS=1, after which I have yet to see a crash. If I end up crashing with the variable set I'll update this comment to adjust expectations appropriately.

Sway+nouveau info for posterity:

local/linux 6.11.1.arch1-1
local/mesa 1:24.2.3-1
local/sway 1:1.9-5

swaywm / sway

Random log out back to login screen with "nouveau: kernel rejected pushbuf: No such device" #8279