You are not logged in.
Hi, since I installed Arch Linux on bare metal a while ago I've been getting stuck for a minute or 2 on the "Starting version 250.4-2-arch" when attempting to shut down or reboot (during the shutdown phase) the system. Sometimes, this is accompanied by the following:
[ 1688.552382] watchdog: watchdog0: watchdog did not stop!
[ 1688.569199] sd-umount[5777]: Failed to umount /oldroot: Device or resource busy
[ 1688.569460] sd-umount[5777]: Failed to umount /oldroot/sys: Device or resource busy
[ 1688.569832] shutdown[1]: Failed to finalize file systems, ignoring.
Essentially, this error shows up once the "starting" line times out and triggers the error, then proceeding to reattempt and succeed on the second try, consistently. Do note that in no way does this prevent me from accessing my system, it just slows down the process of rebooting or shutting down the PC, and yet again, it doesn't seem to happen every time.
I use BTRFS and GRUB and only have a 128mb EFI and 930gb root partition (nvme0n1p4 and p5, since the system had originally had W10 installed as the main OS), without a Swap partition, on my 930 gig NVME M.2 SSD. Also note that this has been present since before I uninstalled W10 and modified the partitions.
Last edited by Gwynsav (2022-04-26 19:46:08)
Offline
what is mounted at "/oldroot" ? thats your problem
Online
what is mounted at "/oldroot" ? thats your problem
Well that would be the issue, I have no idea. I've been looking around the matter on the internet for a while now, and it seems most people suffering from pretty much this exact issue report that it's about the NVIDIA proprietary drivers pinning down that /oldroot/ directory until the very end of the shutdown process. It's also been brought to my attention that a lot of people have solved this by changing to the Nouveau drivers, but I have an RTX 3060ti and I've read that Nouveau is quite outdated and doesn't deal well with new cards.
Is Nouveau viable on newer cards? If not, does this help narrow down the issue a bit?
Edit: Nvm, I went to read Nouveau's website and it specifically states that the GA10X (30 series) cards are not supported.
Last edited by Gwynsav (2022-04-27 05:36:00)
Offline
This can be the nvidia driver, but it also can be random other stuff.
Remove the "quiet" parameter from the kernel commandline and also check the system journal for the previous boot (sudo journalctl -b -1) to see whether there're details available.
You could try to add a shutdown service that explicitly unloads the nvidia kernel modules, https://unix.stackexchange.com/a/41756
Offline
This can be the nvidia driver, but it also can be random other stuff.
Remove the "quiet" parameter from the kernel commandline and also check the system journal for the previous boot (sudo journalctl -b -1) to see whether there're details available.
You could try to add a shutdown service that explicitly unloads the nvidia kernel modules, https://unix.stackexchange.com/a/41756
Do you mean to remove the "quiet" parameters from the GRUB config file? If so, then I can tell you this issue was present even with those off. Neither "sudo journalctl -b -1" nor "-2" seem to show anything out of the ordinary. I will be trying that thing about unloading the NVIDIA modules on shutdown, though.
Last edited by Gwynsav (2022-04-27 16:46:04)
Offline
If so, then I can tell you this issue was present even with those off.
The point is to expose the stalling action - removing the "quiet" parameter won't fix anything, it just hides information.
Offline
Ok, so I disabled the quiet parameters but I saw nothing out of the ordinary in the boot sequence and I'm checking the journalctl boot log files and still nothing that (at least to me) seems to be failing. I'll be moving to making the service that shuts down the nvidia modules.
Last edited by Gwynsav (2022-04-28 09:09:42)
Offline