refind + intel-ucode failing to boot, but without ucode, boots fine.

Topic: refind + intel-ucode failing to boot, but without ucode, boots fine. (Read 1012 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.

refind + intel-ucode failing to boot, but without ucode, boots fine.

21 August 2021, 10:31:33

so decided to swap out arch for artix on another device to minimize the amount of distros i'm depending on.

anyway long story short..

its a chinese tablet (teclast x80 power) is an x86_64 with intel hardware (atom x5-z8350). its not intended to be used as a normal tablet that why i used arch the last like 5 years on it.

anyway for some reason it just wont boot from grub black screen after "welcome grub!" flashes. old issue been such since i installed arch on it like 5 years ago.

it boots fine with whatever system www.system-rescue.org uses to boot, arch-iso does not boot from grub, and artix-iso does either. just tested from the isos to see if grub had fixed the issue that it apparently had with the hardware, it has not, so i'm stuck with refind for now. (or the bootloader system-rescue usb uses, but i think it looks horrible so i choose refind..)

now after the usual install the base system, install refind, setup refind.conf (so it can use linux-lts as a "version") and fix the regular stuff in '/boot/refind_linux.conf' file

Code: [Select]

"Boot using standard options"   "root=PARTUUID=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx rw add_efi_memmap initrd=boot\intel-ucode.img initrd=boot\initramfs-%v.img"
"Boot using fallback initramfs" "root=PARTUUID=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx rw add_efi_memmap initrd=boot\intel-ucode.img initrd=boot\initramfs-%v-fallback.img"

ofcourse xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx is substituted for my real PARTUUID..

now comes the wierd part..

if i update the layout

Code: [Select]

"Boot using standard options + intel-ucode"   "root=PARTUUID=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx rw add_efi_memmap initrd=boot\intel-ucode.img initrd=boot\initramfs-%v.img"
"Boot using fallback initramfs + intel-ucode" "root=PARTUUID=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx rw add_efi_memmap initrd=boot\intel-ucode.img initrd=boot\initramfs-%v-fallback.img"
"Boot using standard options"   "root=PARTUUID=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx rw add_efi_memmap initrd=boot\initramfs-%v.img"
"Boot using fallback initramfs" "root=PARTUUID=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx rw add_efi_memmap initrd=boot\initramfs-%v-fallback.img"

now both the non-ucode items boot fine, and the ones with ucode in them crashes during fsck of the local filesystem..
handwritten from screen.. might be typos..

Quote

* Checking local filesystem ....
fsck.fat 4.2 (2021-01-31)
open: no such file or directory
* Filesystem couldn't be fixed
[ !! ]
* rc: Aborting!
Starting Default runlevel
* Checking local filesystem ....
fsck.fat 4.2 (2021-01-31)
open: no such file or directory
* Filesystem couldn't be fixed [ !! ]
* rc: Aborting!

so the question is, can i get more info out of the startup with some kernel parameter?
cause i guess it tries to open the efi drive for fsck check, but it cant find it.. but since it does not tell what it tries to open and fail, its hard to diagnose..

usually i diagnose stuff with a running system below, so openrc and kernel stuff well that was a decade ago i messed with it..

any bright ideas? for now i'll have to run without the intel-ucode patches applied but i would prefer to have them for obvious reasons

Re: refind + intel-ucode failing to boot, but without ucode, boots fine.

Reply #1 – 21 August 2021, 11:22:48

just to be thorough i tried to create a manual stanza too.. same issue..

until i used a soft-reset (ctl + alt + delete after the filesystem failure) and it just booted fine.. both from cold boot and hot restart.
after that it has been booting cleanly a few iterations (but the screen does not show anything until after late kmod..) so i guess i'll start there by adding the intel kmod to the initcpio..

anyway there seems to be gremlins somewhere around refind + intel-ucode and the hardware..

so not marking this as solved yet, cause i think it might be something to keep an eye on..

Re: refind + intel-ucode failing to boot, but without ucode, boots fine.

Reply #2 – 21 August 2021, 12:24:18

and we're back to normal setting does not boot with intel-ucode (on manual stanza) but fallback initramfs (on manual stanza) does boot with intel-ucode loaded...

it seems like i should create a manual stanza without intel-ucode just to have as a fallback in case shit goes south with loading the ucode..

hot-reboot trick does not work every time..

Re: refind + intel-ucode failing to boot, but without ucode, boots fine.

Reply #3 – 22 August 2021, 03:05:51

It sounds a neat setup to get that to run at all!
See the mkinitcpio man page for rd.debug and rd.log kernel command line parameters, and possibly rd.break if you get that far into the boot.
Also more ideas here:
https://wiki.archlinux.org/title/General_troubleshooting#Boot_problems
Perhaps some info about microcode might help too:
https://wiki.archlinux.org/title/Microcode
You can potentially load it on a running kernel (early vs late loading) which might offer a workaround if all else fails.

Re: refind + intel-ucode failing to boot, but without ucode, boots fine.

Reply #4 – 22 August 2021, 10:20:24

thanks for the input hashes! forgot about mkinitcpio docs should contain info solve the debug output on boots..

i read the microcode page like 5 times yesterday, it goes to a point where you cant see the forest for all the trees..

but yea late loading with a openrc script is the "last" type of workaround but that just feels like cheating, its supposed to work..

Re: refind + intel-ucode failing to boot, but without ucode, boots fine.

Reply #5 – 22 August 2021, 11:11:57

well the debug info was not to much use, only notes that fsck is terminated by a sigterm signal..

if i recall like 10 years ago usually when something broke boot, it dropped to a busybox-shell so i had access to the busybox on the initramfs..

the big question is how to access the initramfs logfile to store it on a usb or something?

8" screen and debug logs that run off screen are not much useful for debug..

well thats it for this weekend, other more pressing matters then getting it to boot intel-ucode..

might have another go at it at another time.

Re: refind + intel-ucode failing to boot, but without ucode, boots fine.

Reply #6 – 23 August 2021, 03:35:05

I did some early logging with dracut a few years ago, and you could save the logs somewhere, it was very impressive actually, I guess mkinitcpio has similar features now but haven't tried with that. It can be used instead of mkinitcpio and was going to be the new default but that didn't happen, so far anyway. One very easy way to save onscreen output is video it and replay in slo-mo!

Re: refind + intel-ucode failing to boot, but without ucode, boots fine.

Reply #7 – 23 August 2021, 10:57:55

yea the video and slowmo was the first idea i had..
big files to keep around as logs was the first known downside..
and transcribing from video well not my cup of tea..

hmm dracut, yea used it a few times on fedora installations mostly to test it out, i think centos 7 used dracut too..

might be worth a go next weekend if i got the spare time...

Re: refind + intel-ucode failing to boot, but without ucode, boots fine.

Reply #8 – 23 August 2021, 15:52:31

https://github.com/archlinux/mkinitcpio/commit/b5ae503d85f242ad37c6a32640ca550f0b62c73d
Looks like some recent changes have been made re microcode in mkinitcpio, (and UEFI in other commits) whether anything fixes or causes your problem who knows. Perhaps downgrading it or upgrading to a git AUR version might help?
mkinitcpio is no longer being replaced by dracut, but it went unmaintained for a while, and has new maintainers:
https://lists.archlinux.org/pipermail/arch-general/2021-January/048549.html