System seems locked while rebooting with Linux 5.2.1 and nvidia drivers 430.34 or 430.26

Graphics / Linux Linux
61

Can y’all try using this workaround?

$ cat /lib/systemd/system-shutdown/rmmod-nvidia-modules.sh
#! /bin/bash

export PATH=/usr/bin:/usr/sbin:/bin:/sbin

while :; do
    i=$((i+1))
    modules=`lsmod | awk 'BEGIN{ORS=" "}/^nvidia/{print $1}'`
    test -z "$modules" -o "$i" -gt 10 && break # 10 attempts maximum
    echo "Attempt $i to remove $modules... "
    rmmod $modules && echo OK && break
    sleep 1
done

$ chmod +x /lib/systemd/system-shutdown/rmmod-nvidia-modules.sh

It doesn’t solve the issue but if the NVIDIA driver is involved in the reboot system call, it might help.
rmmod-nvidia-modules.c (316 Bytes)

62

Would this work if xorg was still up?

63

You can’t have X.org running when you’re rebooting if your system is working properly.

If you have some weird custom configuration when it’s still running, then, no, it can’t help at all.

64

dodo.godlike and I have almost the same hardware: both Aorus Elite boards, nvidia card and intel processor. The error detail posted matches that of my own, I would say it seems like the same problem.

@dodo.godlike are you running arch or gentoo? Were you asking me to confirm I get the same results as you? I can do that when I get home. I think that linux-lts package in the main repos is version 4.19.12 [see subsequent post] (which is what I have installed at the moment as a work around).

I’ll also try birdie’s systemd script as a workaround.

65

@dodo.godlike

Actually, linux-lts package is version 4.19.80 - this is what I have installed at the moment and do not experience hang problems.

So what we can do is install the next kernel version after 4.19.80 (which looks like 4.20 https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/snapshot/linux-4.20.tar.gz) and see if the problem happens. The first version of 4.20.y that arch has is 4.20.1, so I guess it would make sense to try that one first before doing a build of 4.20

Make sense?
[EDIT]

I reread your chart - so you’ve already tried 4.20 as a custom build? So if linux-lts is configured the same way as linux, and 4.19.80 works, then change must be between 4.19.80 and 4.20 - can you verify if 4.19.80 works for you? Either custom build or linux-lts?

66

I tried this script with kernel version 5.3.7.arch1-1 and nvidia 435.21-13

My system does shutdown but it takes longer than normal. I also get stop jobs that go on for User Manager service, which I believe is related to lightdm, which based on the error message that occurs without this script is related the hanging problem. I’ve gone back to linux-lts for now and will try using the 4.20 linux kernel version from arch archives to see if the single point version difference is where the problem starts happening.

Thanks!

67

Good news here! Today I installed the latest BIOS update for my Gigabyte Z390 Aorus Elite, the F9 version [1] that came out 15th October, which seems to be a pretty big update.

Right now reboot/shutdown is working fine on linux! I tried LTS kernel 4.19.81 and latest 5.3.8 and everything seems to be working!

This means that (at least in my case) the issue may be related to the UEFI implementation of our chipset manufacturer (Gigabyte) as I suspected from the beginning. Please correct me if my assumption is wrong.

So to all of you still affected by the issue, check for BIOS upgrades!

[1] Z390 AORUS ELITE (rev. 1.0) Support | Motherboard - GIGABYTE Global

68

That’s awesome to hear - I just downloaded the update as well. Did you install it off a usb key? I notice that it’s an windows executable, which I would not enjoy attempting to run through wine haha!

69

Figured it out using the Q-Flash tool - I updated to F9 bios update (for Gigabyte Z390 Aorus Elite) and confirm that the problem is no longer using kernel 5.3.8-arch1-1 as well as the lts kernel 4.19.81.

FWIW, following the BIOS update, I had to reset the options in my BIOS to disable CSM Support and set the OS back to “Other” and then also boot into a live usb and reinstall GRUB to my EFI partition - go figure.

Everything is working as it should now.

70

It’s good to know that the problem is solved for you. Unfortunately, no BIOS update is out yet for my motherboard. Though I suppose they will release a newer version, it seems that this is an important fix.

71

I Update BIOS from F13 to F14 on my Gigabyte H370-HD3 (rev1.0) Motherboard.
I check reboot from Ubuntu 18.04 LTS with 5.0.0-32-generic kernel.
Reboot was be fast, no system hangs.
Also nvidia-smi -q work properly, not freeze/hangs.

I not check fresh install and latest kernels.
I write this for approve that problem maight be in motherboard from Gigabyte.

In BIOS Setup you can see options:
BIOS(Tab)->CSM Support - i do not know what is that but it Enabled for me.
BIOS(Tab)->Windows 8/10 Features - options where you can choose:
[ Windows 8/10; Windows 8/10 WHQL; Other ]
Choosen: Windows 8/10
Try change something may be it help solve your problem.

I write it message for any people who will search this thread later.

photo_2019-11-08_23-40-52.jpg
photo_2019-11-08_23-40-52.jpg1280×960 153 KB

photo_2019-11-08_23-40-51.jpg
photo_2019-11-08_23-40-51.jpg1280×960 144 KB

72

I also note, that on the Aorus Elite board, there used to be a red light on the motherboard next to “VGA” - after having done the BIOS update that light is no longer lit.

73

Unfortunately, there’s no updated BIOS for all H370 series motherboards, I checked each one of them. I created a support case with Gigabyte and asked them if they will release a newer BIOS version, like the one for Z390 Aorus. I’m on BIOS F12, the latest one is F13 which includes the updated microcode support for the Intel cpu flaw, which I include it anyway in my initramfs file while booting. Let’s hope that F13 solves the issue.

74

I installed the latest BIOS F13 and it seems to have solved the issue of not shutting down properly. I forgot to mention before, that after removing the Nvidia card and activated the intenal GPU of the processor, the system would either not shutdown or shutdown after 1-2 minutes. The reboot worked fine.

So, the F13 BIOS contained more fixes than mentioned. The system works perfectly now.

75

Good to hear, nice to know this is more or less resolved for everyone.

76

Well, I have to admit that I wonder if the previous BIOS was the source of all the problems with the Nvidia card…

77

Well for the Aorus Elite mobo, the BIOS update on gigabyte website states: “BIOS Structure refresh”, additionally, I saw a red light on the board next to the “VGA” which according to the manual indicates there is an “issue” with that component. After the bios update, the red light is gone. So I’m thinking maybe they fixed something in the BIOS that was preventing the operating system from successfully sending a shutdown signal…

I also note, that even though the newer kernel (5.x) was hanging on shutdown and spewing some errors, I was also seeing flashes of errors while using the 4.x (LTS) version of the kernel, even though it didn’t hang.

What board do you have?

78

I have the H370 Aorus Gaming Wifi 3.

I never bothered to install BIOS F13 (I had installed F12 a few months back), since the changelog included only the security fix for the Intel processors. But after reading here that upgrading the BIOS for other motherboards solved the issue for these users, I thought I should give it a try and it fortunately worked.

79

There must have been in a bug in their software… I don’t know much about how BIOS software is built, but if I could guess, I would guess there might be some common components of some sort that they use and then just build the software with some options for each boards particular features… but I’m not sure if that’s really the case. Glad to hear it’s working for you as well now.

80

It looks like issue has been fixed for almost every users after upgrading BIOS on Gigabyte motherboard.
If so, I think we are good to close bug which was filed with Nvidia.