Good call out there bro! Yeah, after 5 days of believing my problems were totally solved with the 5.11.14 kernel, I experienced a light crash (KDE died and my session got logged out, but I could log back in easily), so I can’t state that was the fix.
I hope you’re right and the experimental kernel brings up our fix Theory is on our side, since that commit looks like fixing the errors we saw when experiencing it. I don’t wanna claim victory early, since this issue has been one tough boss to fight, but I’m putting all my hope on the solution. I’ll try this in a couple of days.
Please, keep us updated if you have any further experience on this! And thanks for your contribution
@elektropepi@B007C0DE So i installed the experimental 5.12-rc7 kernel as you suggested, but this time I wasn’t even able to log in to my system After inputting my credentials, the screen went totally black and system would not respond. Here’s what got logged at journalctl:
Along with some of the typical page fault errors. Would you please tell me which version of mesa drivers are you using? Maybe there’s some incompatibility between the kernel and the driver itself. I’m using the latter’s latest version.
UPDATE: I tried booting into that kernel again, and this time I could do so. Idk what may have happened the first time, but I hope it was some really random and isolated error, and that 5.12 carries the real fix for this
Yeah, I would dare to state that each kernel update gives us better results (5.11.14 made a huge improvement for me, and hopefully 5.12 does the same), but still not there yet. I had never experienced them either! All of this began with the April 9th update for me.
Wish there was a way some kernel/GPU drivers dev sees this so we could get some authorized opinion… Let’s keep on updating future findings on this! We’ll get through it
Ok that’s shit. Since I switched kernel, I’ve not had the freeze once. Did you choose the correct kernel on startup (maybe your grub defaults to a LTS kernel)?
I’ll keep you guys posted if that error occurs again.
Just updated kernel to 5.12.rc7. (Lastest available on Stable, for me)
Rebooted.
System is finally responding quickly as it used to before experiencing many of the above mentioned problems others have been having. My system would crash, freeze, run astonishingly sluggishly when it did run, for the past month or so.
Not sure if this is related to my problem with freezing recently about 20 minutes after boot. man.db is hogging resources for the first 20 minutes - about 21% cpu time alongside rsynch at around 20%. It seems that any app I run after 20 minutes the entire system freezes. Initially I thought is was teams, but then noticed it happened to thunderbird, firefox and chrome and vlc. The main journal reports an EXT4 error on my main linux disk and insists that I run efsock -D (or something similar). Currently I am in windows to grab the latest ISO to burn onto a USB so I can check and try and fix that. Can’t recall any graphics issues in the logs though they may have been present. The last update is when I started experiencing freezes - not even able to drop in to tty. My setup also have AMD gpu for graphics. Will update this post later.
No more errors regarding the disk in my logs. Ran the e2fsck command and used disks to check the root drive from a live usb and all seems fine 2 hours in and no freezing issues so far.
Ohhh, I’m sorry to hear that. I’d have thought that is some disk issue, but if your scan showed no faulty devices then that may not be it. Are you running on the experimental kernel?
Same problem also using an AMD 3400G, since update a few days ago system is unstable and will page fault and crash, sometimes if i hit ctrl-alt-f1 it will reset to login screen but sometimes it’s just total hard freeze.
Journalctl shows long list of errors in red
My mistake really as my post is somewhat related - but only because I was experiencing random system freezes since the last update - as I have an AMD GPU - I thought your situation was perhaps connected. Though I saw no error log connected to my graphics and only a disk error, sometimes though, I find you fix one error and then the other is revealed - however my system is now stable and no serious issues visible in any logs. My kernel is the latest but not experimental. But thanks for taking the time out to reply and I hope you find a solution soon.
Ohh, sorry to hear that you got affected by this too. It seems there’s a big bunch of us suffering this already
Have you tried running the 5.12 experimental kernel? I wouldn’t call it a definitive solution, but it has proved to enhance the experience of many user here (mine as well), since it introduced a fix related to GPU memory overflows (to put it in a simple way; a better and detailed explanation can be found some comments above, with a reference to the commit regarding that fix).
You can simply do so by opening the Kernel application (if you’re running KDE) or by running sudo pacman -S linux512 at your terminal. Both will install the latest kernel, and then you’ll have to reboot your machine to start it (GRUB should have it set as the default kernel after installing it). You can verify what kernel you’re running after logging in by typing uname -r in the terminal.
Nice! I hadn’t thought nor seeked through those principles as a possible cause for this. I’ll give it a deeper read later and probably get in touch with that comment’s author to see what we can intersect
I’ll definitely try that! Did you add it to your GRUB setup or are you adding it manually before booting?
I’ve been experiencing less crashes than before with the newest kernel, but they still happen some times. I’ve just had a system freeze right now after some screen tearing effect I had never seen before (maybe it’s related to the GPU power issue that @happyxhw mentioned), but was able to softly stop the system by TTY-ing and executing a shutdown now.
UPDATE: I hope that kernel parameter is really a solution, but per its documentation I’d bet it’s a different thing. Here it says that it’s for removing certain CPU threads from the candidates list for RCU callbacks (Read-Copy-Update); maybe it has some influence on GPU processes