All new kernels blackscreen almost immediately

Laachax · n00b Joined: 03 Jun 2017 Posts: 31 Location: Null island

I installed gentoo not too long ago, all went well; took a long time on my system, but it went rather smoothly considering how much of a newbie I am. Anyways, now that I've gotten the base setup, I decided I want to use the ck-sources kernel. I do the usual configure, make, install, etc; but then it immediately blackscreens not too long after I select it in grub, I frankly don't even think it loaded the kernel, it definitely didn't unpack the raminits.

I figured there's a bug in the ck-sources, so I go back to the generic gentoo kernel that the handbook recommends. I do the whole procedure again, but it does the exact same thing. In fact every single kernel I try does it, regardless of config; even genkernel fails! I managed to save the only kernel that can boot, which is the kernel I first compiled. I'm honestly not sure how it boots either, because if I compile another kernel with the exact same config it doesn't boot either(I'm not sure if it's the same config that the safe kernel uses).

Here's lspci -k ; I still need help configuring this whole thing, my ultrawide isn't recognized correctly, and my second monitor isn't recognized at all. And my sound doesn't work. But of course the most important thing is that no new kernels work at all.

NeddySeagoon · Posted: Sat Jun 03, 2017 8:26 pm Post subject:

Laachax,

Make friends with wgetpaste.
Its a command line tool to put files and command output on the web.
Post the URLs it gives you.

One oddity. Your working config has

Laachax · n00b Joined: 03 Jun 2017 Posts: 31 Location: Null island

Thanks for the tool! Will become very handy.

working config https://paste.pound-python.org/show/gNiqKyUIq77QxaScIuKz/

non working config https://paste.pound-python.org/show/FG0c1wEWZtgGe1SmrKcv/

If you need the copy of any other files on here, let me know. I'm currently compiling the new kernel with that option, and I'll see if it changes anything. I'll edit this post to report back.

edit: changed nothing.

NeddySeagoon · Posted: Sat Jun 03, 2017 9:21 pm Post subject:

Laachax,

Post the content of

Laachax · n00b Joined: 03 Jun 2017 Posts: 31 Location: Null island

NeddySeagoon · Posted: Sat Jun 03, 2017 9:40 pm Post subject:

Laachax,

That's interesting ...

Your running kernel was made on Thu May 25 20:16:44 CDT 2017, and it was your #3 attempt.
However, there are no kernels in your /boot dated before May 30.

That suggests that you are not running the kernel you think you are.

What is in /boot when /boot is not mounted?
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

Laachax · n00b Joined: 03 Jun 2017 Posts: 31 Location: Null island

Ah, I renamed that kernel on the 30th to -safe, that's why it's dated from the 30th and not the 25th. I also cleaned up the non-working kernels.

Laachax · n00b Joined: 03 Jun 2017 Posts: 31 Location: Null island

A friend recommend me try to set amdgpu as a module and load it via the etc/conf.d/modules modules="amdgpu". Is this right? And do I add more by leaving spaces or commas between them?

Anyways, doing that greatly changed when the system crashes, in fact a I see a few lines of the familiar blue text of the start up text, but then it goes black. Numlock is not stuck, and I can change tty's and verify this by the numlock state changing. Otherwise the screen is black. I'll see if I can grab the dmesg tomorrow from a rescue CD and see what it gets hung up upon.

edit: I also edited the main post with my boot situation for a bit more info. I'm using EFI in legacy mode with mbr SSD and mbr HDD dualbooting with windows, the motherboard is rather old. First gen asus m5a97 no revision version.

NeddySeagoon · Posted: Sun Jun 04, 2017 7:58 am Post subject:

Laachax,

amdgpu, either as a module, or built in provides your framebuffer console.
When its a kernel built in, it loads earlier than when its a module, so you see boot messages sooner.

You should not need the /etc/conf.d/modules modules="amdgpu". The auto kernel module loading should do it for you.
Exactly which file that goes in depends on the version of openRC you are using.

That you get a few lines of boot text suggests that you still have Simple Framebuffer enabled.
That works until the amdgpu framebuffer console kicks in, it starts later now.

That's a pointer to something amdgpu related.
Wild guess, the firmware and new kernels do not play nicely together. When did you update linux-firmware last?

I suspect your system boots, you just can't see it ... can you ssh into it and get dmesg ?
That would be very useful.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

Laachax · n00b Joined: 03 Jun 2017 Posts: 31 Location: Null island

Sorry for the lack of updates. Been really busy the last few days. Anyways, I tried to SSH into it and no dice, cannot connect. I suppose I could try a rescue CD and see if it writes to dmesg. But if I can't ssh into it, and ssh is running and default level, I'm not sure if it writes anything.

I updated the firmware not too long ago, probably around the 28th.

shrike · Posted: Thu Jun 08, 2017 12:33 pm Post subject:

WRT to dmesg you could simply mount the drive:

Laachax · n00b Joined: 03 Jun 2017 Posts: 31 Location: Null island

Hu · Moderator Joined: 06 Mar 2007 Posts: 21586

Your prior kernel configuration suggests, and the latest dmesg confirms, that you have included the graphics driver in the core kernel. If you do this, you must also embed the firmware in the kernel because the graphics driver will need it before your root filesystem is mounted. Your edits suggest that you probably understand this, but it is easy to miss. The other problem that I have seen for people in this situation is that they embed the firmware in a way that gives it a path other than the one the kernel uses, so the kernel fails to find the firmware even though it is present. Please pastebin the config that corresponds to your most recent dmesg.

Laachax · n00b Joined: 03 Jun 2017 Posts: 31 Location: Null island

Hu · Moderator Joined: 06 Mar 2007 Posts: 21586

Although slight variations might be permissible, I try to have the path in .config match exactly the path shown in dmesg. Yours do not. You have a leading slash on each of the files in CONFIG_EXTRA_FIRMWARE, relative to the output from the kernel. Please remove those slashes and try again.

Laachax · n00b Joined: 03 Jun 2017 Posts: 31 Location: Null island

Done, nothing changed.

Here's the new dmesg with a new kernel with the new config https://paste.pound-python.org/show/8QlDB4SoffTFKxxpRHCW/

the config https://paste.pound-python.org/show/M5VchBMOEI2L7yfUkQJB/

Hu · Moderator Joined: 06 Mar 2007 Posts: 21586

Laachax · n00b Joined: 03 Jun 2017 Posts: 31 Location: Null island

That's strange. I can assure you I'm trying to boot 4.11 and not 4.9. It must have overwritten my dmesg.bak when I booted back into the working kernel to send the wgetpaste.

Laachax · n00b Joined: 03 Jun 2017 Posts: 31 Location: Null island

Been very busy again, sorry for the lack of updates...

Anyways, thanks to the help of a very nice redditor we have pinned it down to mesa/amdgpu misbehaving. I blacklisted amdgpu from loading and the kernel boots! Then I modprobe it in and it black screens... At this point I'm not sure what to do as this is incredibly beyond my knowledge. At least all kernels boot now that I blacklisted the amdgpu module from loading.

Check the reddit thread to see what we have done so far https://www.reddit.com/r/Gentoo/comments/6gd4xh/having_kernel_and_xorg_issues_on_new_install_can/

Any help is appreciated. If I should make a new thread, I will. Thanks

Laachax · n00b Joined: 03 Jun 2017 Posts: 31 Location: Null island

So on a hunch I changed slots my graphics card was in...

it works now. No problems whatsoever.

I don't even. Sorry everyone for the wild ride. I apparently need a new motherboard or something...