I’ve been experiencing weird crashes every day the since I upgraded my kernel from 5.18.19-3 to 5.19.7-1 a few days ago.
The crashes were of two types:
Video stays up, keyboard becomes unresponsive (cap-lock and num-lock do nothing, indicating kb driver is off-line), mouse pointer moves but clicking does nothing (mouse driver malfunctioning).
Video goes black, keyboard becomes unresponsive (cap-lock and num-lock do nothing, indicating kb driver is off-line), can’t tell what mouse is doing.
Sometimes after #2 above happens, video will come back after about 10min but I get thousands of boxes popping up, about 10 per second, saying “desktop manager has restarted after video reset”, “desktop manager has restarted after video reset”, “desktop manager has restarted after video reset”… (ad-infinitum, ad-nauseum)
After a crash like this, I generally have to hold my On/Off button down for 5 seconds to force a hard shutdown because I’m usually not able to use keyboard or mouse to “shut down” properly.
I just reverted to the earlier kernel a few minutes ago, and we’ll see if the crashes stop. (If it’s not a kernel thing, then my MB or CPU is dying and will need to be replaced.)
Anyone else notice similar problems with kernel 5.19.7-1?
For reference, here are some of my system specs:
Operating System: Manjaro Linux
KDE Plasma Version: 5.25.5
KDE Frameworks Version: 5.97.0
Qt Version: 5.15.5
Kernel Version: 5.19.7-1-MANJARO (64-bit) (currently reverted to 5.18)
Graphics Platform: X11
Processors: 4 × AMD A10-7800 Radeon R7, 12 Compute Cores 4C+8G
Memory: 14.6 GiB of RAM
Graphics Processor: AMD KAVERI
Manufacturer: LENOVO
Product Name: 90BG003KUS
System Version: Lenovo H50-55
What you are describing is not a kernel issue but a Plasma 5.25 issue. If you search the forum, then you will find ample threads and posts since the most recent update that all deal with Plasma 5.25.
The thing is that many of our users were loudly, vocally — and sometimes even angrily — complaining about Manjaro Stable remaining on Plasma 5.24 — which is an LTS release and very mature — while 5.25 was already in use in other distributions and was available in Manjaro only in the Unstable and Testing branches.
So the management took the decision to bring Manjaro Stable up to Plasma 5.25 as well, and as it turns out, the only people who have so far not experienced any problems with it are the ones who use Plasma as it comes installed out-of-the-box, without any customizations.
I’m afraid there’s very little we can do about it. The die has been cast, and the current 5.25.5 is the last of the Plasma 5.25 releases before 5.26 comes along, which means that there won’t be any more bug fixes until such time.
Trust me, I’m a staunch Plasma fan, and I’ve been using KDE ever since version 1, but I’m not happy with it either. Yesterday I had to roll back a timeshift update of my entire system and redo the update process all over again because Plasma 5.25 fatally crashed and left me with an unusable desktop. I have never had a situation like that happen to me before ever since Plasma was still in its 4.x.x stage, which was many years ago.
Interesting. Query: When did the Plasma 5.24 → 5.25 upgrade occur? I didn’t notice. Was it during the big upgrade (some hundreds of packages) a few days ago? I think that was also when my kernel got bumped from 5.18 to 5.19, which may confuse the issue.
That being said, I’ve been using my system heavily today, stressing the CPU, MB, video, and networks, watching YouTube videos, running complicated screensavers, running cross-network syncs between partitions on local machine and partitions on my notebook, etc. And yet, no crashes. And the only thing that’s changed is, today I’m running kernel 5.18, whereas the past few days (with daily crashes) i was running kernel 5.19. So that is evidence toward it being a kernel problem (rather than hardware or DE).
But we’ll see. I’ll report back in a couple days. (Unfortunately with these “sporadic” malfunctions, it takes days to troubleshoot because the malfunctions, while catastrophic, are random and typically happen once every 10-20 hours or so. Such malfunctions can be caused by sooooo many things: Hardware, firmware, BIOS, Grub, OS, kernel, DE, software, settings, etc. I’m betting “kernel” but we’ll see.
Well for what it’s worth, I’m using Kernel 5.19.7 and Plasma 5.25.5 since the update occurred and don’t have any problems at all.
Steam games, streaming, virtual machines, video encoding, remote control other machines etc. everything works fine.
For sure this doesn’t help you much with your problem, but it might be another indication that there is no general problem with Plasma, but a special problem with your computer’s hardware/kernel combination.
Ok, after a day and a half of no problems, this morning three crashes happened within 40 minutes.
Two were the “screen goes black and keyboard driver crashes (numlock and caplock keys fail to toggles their LEDs)” type.
One was the “dozens of popups, one per second, saying 'window decorations have reset due to graphics reset” type.
This was with kernel 5.18 and plasma 5.25. So its apparently not caused by kernel 5.18, as that was working fine until the big update.
But whatever it is appears to be clobbering the kernel. As I understand it, the cap-lock and num-lock keyboard buttons and LEDs are driver-driven, no? And hence, in Linux, part of kernel space rather than user space? So if this is a Plasma bug, it appears to be crossing over that boundary.
Query: Is there any way to revert Plasma from 5.25 to 5.24?
I looked in my Timeshift and it’s currently set to “daily snapshots only, keep last 3 only”, due to the vast amounts of disk space it requires (60GB per day). So I have snaps from 9/18, 9/19, 9/20. Not going to help something that started on 9/12. I’d need 9/11 for that. Perhaps I should have set it to weekly instead, but “hard-to-troubleshoot bugs that only happen every two days” never occurred to me. Live and learn.
Yep, it’s probably a combo of 7-year-old hardware (every time Linux launches I get multiple “lacking firmware to support module” warnings, which I interpret to mean that the UEFI on my MB is way behind current kernels), combined with the latest kernels, combined with Plasma 5.25. I’m guessing something in Plasma 5.25 is making unwarranted assumptions about hardware, and when those assumptions turn out to be false, the system crashes.
Well, that, or this computer is just slowly dying. It’s served its purpose, and I’m in the process of building a new one that will be about 8 times faster (CPU will upgrade from AMD-A10 to Ryzen-7-5700G). We’ll see how THAT works with the latest kernels and desktops. I suspect there will be fewer problems.
Timeshift uses RSYNC (or BTRFS on systems using that file system), so the snapshots are not of the complete system (except the oldest one), but the changes made to it. Unchanged parts are hard links. My 5 snapshots take less than 16 GiB.
Well, the X server does run part of the graphics rendering in the kernel — and Wayland all of it, even.
Only if you have a backup of your system when it still had 5.24.6 on it. In that case, you can restore the backup, add IgnoreGrp = plasma to /etc/pacman.conf and then redo the update.
It’s what I myself and several other people have done, but of course, this then leads to a “partial upgrade” situation, which means that it’s not supported in Manjaro, and that at the first sign of any trouble — whatever trouble it is — people will be telling you it’s your own fault.
Two days ago, I mistakenly wrote: Ok, I think I’ve found the cause and cure for these two bugs (“display blacks out and keyboard goes offline”, and “message boxes stating ‘Kwin window manager: window effects reset because of graphics reset’ repeated every 1 second”)…
[Edit, 2022-10-03: Nope, I spoke too soon. The very next day, after 9 days of no trouble, I experienced 3 crashes over a span of 15 minutes. Two “keyboard & mouse drivers go off-line” crashes, and one “repeated ‘kwin window manager: window effects reset due to graphics reset’ message boxes” crash.]
…The “shatter” window effect… was not working… “animation speed”… was set at “100% (instantaneous)” … So I changed it to about 50% … I haven’t had an instance of either of the two bugs I’ve been complaining about in this thread in the last 9 days…
[Edit, 2022-10-03: The problems recurred the very next day, so nope, it wasn’t animation speed.]
…I think I fixed it…
[Edit, 2022-10-03: Nope, problem recurred 1 day after I said that, so the cause remains unknown. About the only pattern I see is, when this bug happens, it’s usually a few seconds after starting to play a YouTube video on Firefox. Otherwise, it’s very unpredictable. So I’m UN-marking the “Solution” box.]
Huh! Surprisingly, it’s been 20 days and I’ve had zero repeats of either of these two bugs:
Bug1: video goes black and keyboard and mouse go off-line until a cold restart (pull plug, wait 20 seconds, plug-in, start computer)
Bug2: KWin “graphics reset” boxes pop up at 1 per second until I log off then back on.
Is it fixed? Maybe! If so, it must be due to something in either the 10/5 or 10/10 updates that fixed it, because I haven’t changed anything. I’ll give it another couple weeks then call it “fixed” if it doesn’t recur.
