Crashing in tty

[prestige787]

I have installed gentoo a few times now and on my pc I always encounter the same issue: when in tty I will eventually get a black screen, and the system stops responding. The post code on the motherboard makes it seem like it's rebooting but it's not, a long hold of the power button is required. This always happens, sometimes after a few seconds, before I can even log in, sometimes after a few minutes. It doesn't happen once I get into my window manager, nor does it happen on any other linux distro I've used. EDIT: not true, have had it happen in debian aswell, but still not on the dist-kernel. It also doesn't happen on my laptop which is a very similar install. I'm assuming I'm missing something in my kernel config, everything else except network configs should be identical to my laptop which doesn't have the issue, but has a different kernel.

Hardware:

Amd 6800xt

ryzen 5900x

I followed the kernel configs on the pages for my gpu and cpu.

Kernel configs:

https://pastebin.com/zim5hC2N

For zen kernel (I know it's unsupported, should be pretty much the same config):

https://pastebin.com/ttB8U4yT

[mike155]

1) Which CPU does you PC have? The Ryzen 5900x?

2) How much RAM does it have?

3) Which graphics card do you have?

4) Do I understand correctly that you run the kernel "zim5hC2N" on your PC? What's the matter with the 'Zen' kernel? Where do you use that?

5) Please post the output of

[prestige787]

1) Yes, it's the ryzen 9 5900x

2) 64gb, had the same issue when running with 32gb

3) AMD radeon rx 6800xt

4) Yes, this is the kernel I'm using. Just to make sure I got the right one, here is the config again:

https://pastebin.com/7rF3cwmC

5)

lspci -knn: https://pastebin.com/E4qBv2kN

6)

dmesg: https://pastebin.com/87azYSZL

[mike155]

Thanks for the data.

There are a couple of error messages in dmesg:

[prestige787]

Ok, so I made all of those changes except the second one, I don't need wifi so instead I just disabled wireless lan, and the errors don't show up anymore. Unfortunately it's still crashing. I'm still getting the regulatory.db error, I think this is also just related to wireless but I'll try to get rid of it. I think all of the drivers are now being loaded, the random number generator is still at the very end of dmesg, but it's at about 12s instead of 79s. Is this still too late?

New config: https://pastebin.com/5n4Ut8AZ

dmesg: https://pastebin.com/Sgd0mTAv

lspci -knn: https://pastebin.com/TXpyf7f7

[mike155]

I compared your .config with my .config. Below are the most important differences:

1. enable CONFIG_RANDOM_TRUST_CPU (will push crng init time to < 1 sec)
   
2. enable CONFIG_POSIX_MQUEUE (recommended)
   
3. enable CONFIG_BSD_PROCESS_ACCT (recommended)
   
4. enable CONFIG_IKCONFIG_PROC (recommended)
   
5. enable CONFIG_SCHED_AUTOGROUP
   
6. I'm unsure about CONFIG_MK8=y. I have CONFIG_MCORE2=y
   
7. enable CONFIG_X86_REROUTE_FOR_BROKEN_BOOT_IRQS
   
8. enable CONFIG_AMD_MEM_ENCRYPT (I think your CPU supports it. unsure)
   
9. CONFIG_NUMA=y: is that the right choice? Does your CPU have a NUMA architecture? Probably not. unsure.
   
10. enable CONFIG_X86_PMEM_LEGACY
    
11. CONFIG_PM: unsure, this option is enabled on my machine
    
12. disable CONFIG_AGP
    
13. enable FB_MODE_HELPERS
    
14. enable AMD_PTDMA (unsure what it does, but it is enabled on my machine)
    
15. enable CONFIG_ACPI_WMI
    
16. enable CONFIG_FANOTIFY
    

[prestige787]

I made all the changes, unfortunately it's still crashing. Just to make 100% sure it's the kernel I downloaded the distribution kernel, and this does not have any problems, so it must be something with the config. I did notice that on the custom kernel, I get an 'FF' post code on my motherboard, as opposed to the 00 code on the distribution kernel. Post codes can be pretty useful so I checked what they are after it crashes. The post codes change very quickly, but the ones that it stops on for some time are:

FF->62->d2->Ad_ac->AA->00

FF is a reserved error code, some people say it mean fault free though. This code is also present while in the gentoo installer which never gave me any issues. Most people with this code had issues with windows software.

62 is a regular boot code meaning 'Installation of the PCH runtime services'. It stays on this code for a long time, has to do with the chipset, some people had this with graphics card issues.

D2 is a straight error code for 'PCH initialization error', should never show up during a normal boot. Some people solved this by removing non needed m.2 drives. I have 3 nvme m.2 drives, I might try to remove them later though it's an annoying thing to do. Starting to think that might be the root of the issue though, not sure what options there are for nvme settings in the kernel.

Ad is 'issue Ready To Boot event for OS Boot' wasn't able to find much info on this one, does stay on it for a long time. The rest are reserved except 00 which is what it should usually be in the os.

EDIT: I enabled nvme multipath support which unfortunately didn't help.

[mike155]

That's strange. I've never heard of such a problem before. We frequently have users that get a black screen immediately after booting. But that's because they failed to configure their graphics card.
A machine that freezes after a few minutes - and that stops freezing if you enter the graphics mode... That's highly unusual.

It's not necessary to use the Gentoo kernel. I use the vanilla kernel, for example. You could also use a kernel from a different distribution - if it works for you.

I think I would let memtest86-bin run for a couple of hours. Just to make sure it's not a memory error.

[prestige787]

Ran memtest for about 2 hours, would need to run it for probably around 10 to be sure with 64gb, might do that over night to be sure since memory errors can lead to some nasty problems. Also tried removing the two nvme's connected to the x570 chipset, which did not help. Loading amdgpu as a module seems to make it take longer to crash but hard to say with how random it is. I'll play around with some other kernels like you suggested, might try with a fresh config if I have time. Thanks for your time!

[prestige787]

Ok, so I decided to start from a working kernel (gentoo dist-kernel), and trim it down until I could replicate the issue. I believe it is one of the options in the "Networking Support" directory in the "menu config". This seems a bit odd, but the two configs:

https://pastebin.com/p6L9cejg

https://pastebin.com/1U3Bq6jA

Differ only in options from Networking support, the first one is the regular dist-kernel and it does not crash (there is a very occasional flicker), the second is the modified one and it does crash. Unfortunately there are way too many options to try one by one, but at least I have it narrowed down a little bit.

The only other difference is that when you emerge the dist-kernel, it automatically creates an initramfs, whereas as when I did it 'manually' I did it with dracut, but I don't think this would create any difference. (Also I'm pretty sure portage also just uses dracut but wasn't paying attention to the output).