Kernel 5.9 Unable to boot into desktop

[jiibus]

Installed the below updates today and after reboot, I get no video whatsoever. I have startup set to noisy, and once it reaches SDDM launching, my monitors shut off as if they’re getting no input. Trying to go into a TTY doesn’t turn them on.

Updates

:: Repository packages will be installed:
libvirt                               6.8.0-3              -> 1:6.5.0-2
linux                                 5.8.14.arch1-1       -> 5.9.1.arch1-1
linux-headers                         5.8.14.arch1-1       -> 5.9.1.arch1-1
linux-zen                             5.8.14.zen1-1        -> 5.9.1.zen2-1
linux-zen-headers                     5.8.14.zen1-1        -> 5.9.1.zen2-1
python-lxml                           4.5.2-1              -> 4.6.1-1
python-setuptools                     1:50.3.1-1           -> 1:50.3.2-1
kactivitymanagerd                     5.20.0-1             -> 5.20.0-2
libjpeg-turbo                         2.0.5-2              -> 2.0.5-3

:: Third-party repository packages will be installed:
chaotic-aur/chaotic-mirrorlist        20200916-1           -> 20201018-1
chaotic-aur/bpytop                    1.0.42-1             -> 1.0.43-1

I’ve tried booting into both linux & linux-zen with no luck and had to revert snapshots. Using a 5700XT for the GPU. I’m not even sure how to access the logs to troubleshoot. Any ideas?

[anon3337769]

There are some problems with nVidia and kernel 5.9; you might want to try the LTS kernel for now.

You can read about it starting from this post forward:

Linux 5.9 kernel released

Yep—I see that. Installed 5.9 & nVidia is broken–rebooted into the LTS kernel for now.

[jiibus]

I’m using an AMD RX 5700XT.

[anon3337769]

Oops, sorry - I thought that was an nvidia card.

[jiibus]

No worries!

[jiibus]

Now that I think about it, I’d be able to access the logs by upgrading then booting into a live environment and chrooting into it, wouldn’t I?

[axt]

Btw., tardy, there has been Nvidia GF 5600XT and 5900XT. The 5700 series has had other additions like LE.

They wouldn’t run with nvidia anymore, but theoretically with nouveau. nouveau has no issues with kernel 5.9.

[axt]

@jiibus

chrooting

First I would set boot options (xforcevesa).

[jiibus]

Does it matter where in the order that is?

[jiibus]

Setting xforcevesa didn’t work for me, so I went ahead and booted into a live environment and chrooted into my install.

This is the output of journalctl --since "15 minutes ago" -p 3:

Summary

-- Reboot --
Oct 18 22:31:50 xxx kernel: sp5100-tco sp5100-tco: Watchdog hardware is disabled
Oct 18 22:31:51 xxx kernel: Bluetooth: hci0: BCM: firmware Patch file not found, tried:
Oct 18 22:31:51 xxx kernel: Bluetooth: hci0: BCM: 'brcm/BCM20702A1-0a5c-21e8.hcd'
Oct 18 22:31:51 xxx kernel: Bluetooth: hci0: BCM: 'brcm/BCM-0a5c-21e8.hcd'
Oct 18 22:31:54 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: failed send message: SetHardMinByFreq (28)         param: 0x000600f5 response 0xffffffc2
Oct 18 22:31:56 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: Msg issuing pre-check failed and SMU may be not in the right state!
Oct 18 22:31:56 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: Failed to power gate JPEG!
Oct 18 22:31:56 xxx kernel: [drm:jpeg_v2_0_set_powergating_state [amdgpu]] *ERROR* Dpm disable jpeg failed, ret = -62. 
Oct 18 22:31:58 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: Msg issuing pre-check failed and SMU may be not in the right state!
Oct 18 22:31:58 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: Failed to power gate VCN!
Oct 18 22:31:58 xxx kernel: [drm:amdgpu_dpm_enable_uvd [amdgpu]] *ERROR* Dpm disable uvd failed, ret = -62. 
Oct 18 22:32:01 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: Msg issuing pre-check failed and SMU may be not in the right state!
Oct 18 22:32:03 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: Msg issuing pre-check failed and SMU may be not in the right state!
Oct 18 22:32:05 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: Msg issuing pre-check failed and SMU may be not in the right state!
Oct 18 22:32:07 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: Msg issuing pre-check failed and SMU may be not in the right state!
Oct 18 22:32:09 xxx libvirtd[1232]: cannot open directory '/run/media/jiibus/samsung_usb/ISOs': No such file or directory
Oct 18 22:32:09 xxx libvirtd[1232]: internal error: Failed to autostart storage pool 'USB_ISOs': cannot open directory '/run/media/jiibus/samsung_usb/ISOs': No such file or directory
Oct 18 22:32:10 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: Msg issuing pre-check failed and SMU may be not in the right state!
Oct 18 22:32:10 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: Failed to get overdrive table!
Oct 18 22:32:10 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: Failed to setup default OD settings!
Oct 18 22:32:10 xxx kernel: [drm:amdgpu_device_ip_late_init [amdgpu]] *ERROR* late_init of IP block <smu> failed -62
Oct 18 22:32:10 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: amdgpu_device_ip_late_init failed
Oct 18 22:32:10 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: Fatal error during GPU init
Oct 18 22:32:14 xxx systemd-coredump[2125]: Process 2123 (ffdetect-emby) of user 1001 dumped core.
                                                       
                                                       Stack trace of thread 2123:
                                                       #0  0x00000000004041ca n/a (ffdetect-emby + 0x41ca)
                                                       #1  0x00000000004048e0 n/a (ffdetect-emby + 0x48e0)
                                                       #2  0x000000000040ef59 n/a (ffdetect-emby + 0xef59)
                                                       #3  0x000000000040f15f n/a (ffdetect-emby + 0xf15f)
                                                       #4  0x00000000004033c7 n/a (ffdetect-emby + 0x33c7)
                                                       #5  0x00007fe753134152 __libc_start_main (libc.so.6 + 0x28152)
                                                       #6  0x0000000000403808 n/a (ffdetect-emby + 0x3808)
Oct 18 22:32:15 xxx systemd-coredump[2133]: Process 2131 (ffdetect-emby) of user 1001 dumped core.
                                                       
                                                       Stack trace of thread 2131:
                                                       #0  0x00000000004041ca n/a (ffdetect-emby + 0x41ca)
                                                       #1  0x00000000004048e0 n/a (ffdetect-emby + 0x48e0)
                                                       #2  0x000000000040ef59 n/a (ffdetect-emby + 0xef59)
                                                       #3  0x000000000040bbfc n/a (ffdetect-emby + 0xbbfc)
                                                       #4  0x0000000000403793 n/a (ffdetect-emby + 0x3793)
                                                       #5  0x00007f06acf9e152 __libc_start_main (libc.so.6 + 0x28152)
                                                       #6  0x0000000000403808 n/a (ffdetect-emby + 0x3808)
-- Reboot --

Is emby causing my GPU to freak out?

[ringo]

personal would have linux-lts beside linux-zen or linux , if that works could be a kernel thing with mesa probably linux 5.9 with my rx490 did not to bad…

[Krimkerre]

@jiibus

my monitors shut off as if they’re getting no input

So you are using dual monitors then? Perhaps both of them through displayport at 60hz?

When looking through your log, the errors start here:

@jiibus

Oct 18 22:31:54 xxx kernel: amdgpu 0000:2f:00.0: amdgpu: failed send message: SetHardMinByFreq (28) param: 0x000600f5 response 0xffffffc2

When putting SetHardMinByFreq in a search engine, the first thing I came across was this:

AMD Navi10 GPU powerplay issues when using two DisplayPort connectors (#929)...

It’s an older post, so it may not be relevant, but it might be worth reading nonetheless

I have an RX 5600XT, but sadly I don’t have 2 display port monitors, so there’s no way for me to try to replicate the problem

[anon31687413]

Does it work when you disconnect (or turn off) the secondary monitor before booting?
EDIT: the link from Krimkerre above contains a workaround which involves setting the modes on boot. You might want to try that.

[jiibus]

@Krimkerre

When putting SetHardMinByFreq in a search engine, the first thing I came across was this:

I have been getting this error for a while now, it’s probably as old as the post you reference. I do have 2 1440p monitors using display port, but I have them set to 144hz.

[jiibus]

@anon31687413

Does it work when you disconnect (or turn off) the secondary monitor before booting?

I have not tried that, but I will today after updating again after work.

[jiibus]

Got it working! Reading the post @Krimkerre linked leads to this reply from yesterday:

@Joe Kowalski

I still see this issue with 5.9.0 and firmware package dated 20200918 with just plain default kernel boot flags.

That said I’ve reliably worked around this bug completely by setting the KMS video mode on boot with video=DP-1:2560x1440@144 video=DP-2:2560x1440@144 on my kernel flags as described here - https://wiki.archlinux.org/index.php/Kernel_mode_setting#Forcing_modes

Not sure why, but I would guess there is some kind of race condition in the modesetting concurrent with the firmware loading?

And once I set that, I could successfully boot into my desktop.

Thanks @Krimkerre & @anon31687413, I’ve visited that thread quite a few times before and probably would have ignored it in my searching.