You are not logged in.
Pages: 1
Recently while running Arch, I've found that my PC suddenly reboots without any warning; my monitors go black and state that they can't find an input source all of a sudden and within seconds they're back at the boot screen. This has been happening to me 1-3 times every few days, across kernel updates. I find that they usually occur in relatively quick succession. Either within minutes of each other or so quickly that a reboot occurs while rebooting. I've checked my journal but I can't find anything obviously indicating an issue. I see a bunch of boot logs but very little of significance in between. I've had two shut downs in the last 15 minutes or so, so I'm attaching the output of journalctl covering that time period. Let me know if more information of any particular kind would be useful in troubleshooting this issue; I'd be happy to provide it. The first restart occurs around 14:26:56. As far as I can tell, there's nothing immediately preceding that indicating a reason for shutdown/reboot.
Here's the journal output: https://pastebin.com/e46aLtEa
Offline
The usual suspects to check:
CPU temperature going too high: lm_sensors to observe the temperatures. That is the primary suspect, since you report multiple consecutive events: that is consistent with CPU needing time to cool down.
PSU voltages going out of spec: can be done from firmware setup (“BIOS screen”) or using lm_sensors. The former will not detect load-related issues, but it is still worth checking: a voltage too low there is a sure indication of a problem.
Hardware being unstable: stress test CPU, test stability of RAM using memtest86+, memtester or a similar tool. While stressing CPU, monitor temperatures: the tests may destructively reveal faults that would go undetected under normal use.
Unfortunately the journal is unlikely to contain any useful information: even if anything of interest is written to it, power loss prevents the data from being written to the disk. You may SSH from another machine and use `journalctl -f` to get live monitoring, but that is not a sure way either.
Offline
I stress tested CPU and didn't find that the temperature was going abnormally high. No visible consequences to the stress test. I haven't run a memtest yet but today I found that my PC force rebooted even from within the BIOS menu, which seems to imply that this has little to do with my operating system at all. This is probably beyond the scope of this forum at this point, but further advice would still be appreciated. I think you might be right about the PSU voltage. I'm going to see what else I can find about that.
Edit: Just found https://bbs.archlinux.org/viewtopic.php?id=242384 and surely enough I have a ryzen CPU, so maybe this is the issue? Changed the Power Supply Idle Control setting to Typical Current Idle so hopefully this helps.
Last edited by rosalogia (2022-05-02 16:45:23)
Offline
Unfortunately I can’t help with that specific processor, as I’m not actively in the subject since 2015. So I can only give general advice.
If the reset happens in the BIOS menu, then load-related issues may be excluded. For the same reasons I doubt it would be an issue with voltages, as in this case I know two failure modes: they are consistently too low⁽¹⁾ (which you would see instantly) or they go out of spec under load (stress test is likely to reveal that). However, I would not completely skip the possibility that the power supply is faulty. I may be turning itself off. For that, however, I know no test that wouldn’t require a soldering iron and an oscilloscope. The only one I know is checking if the power goes down and, if it does, is it initiated by the motherboard.
____
⁽¹⁾ ± 5% of the nominal value.
Last edited by mpan (2022-05-02 19:44:39)
Offline
Thanks for your input. The BIOS option I mentioned setting didn't seem to help; my PC restarted while I was away. There seem to be some other pieces of advice for Ryzen users experiencing reboots, including the following from the gentoo wiki (https://wiki.gentoo.org/wiki/Ryzen#Rand … mce_events). However, it's worth noting that I don't see anything in my journal indicating that what I'm experiencing is an mce Hardware error. This seems to be the source of the issue for people experiencing this specific problem with Ryzen CPUs, so maybe none of this applies to me. For the record, my CPU is an AMD Ryzen 9 5950X 16-Core Processor. I'll keep looking and see if I can find anything else.
Offline
So the only idea I have is to have some other device keep logs. That may be done by connecting over SSH from another computer and calling `journalctl -f` or — if you can provide a serial port and something to observe its output — by enabling forwarding to serial console using the ForwardToConsole and TTYPath options, with the path pointing to the serial console.
But wait for other people — they may have better ideas.
Offline
Have to tried to limit c-states?
You're patching https://wiki.archlinux.org/title/Microcode ?
Did you rule out environmental issues? (No kernel or BIOS paramter can help if current/voltage of the wall outlet are unstable)
Online
Hi. I confirmed last night that this definitely doesn't occur, at least not with the same regularity, on Windows (left it running for ~7 hours and no reboots). Any reboot that's occurred earlier than proper boot has only followed a reboot that occurred while linux was booted, so I'm back to believing this is an arch/linux-side issue. I may try the logging suggestion by mpan; first I'll limit c-states as suggested. Thanks guys.
Edit: forgot to mention, yes I have amd-ucode installed and set up with grub.
Last edited by rosalogia (2022-05-03 18:08:42)
Offline
doesn't occur, at least not with the same regularity, on Windows (left it running for ~7 hours and no reboots)
3rd link below… it might very well be windows that's rebooting your system.
Online
Pages: 1