I am desperate. Linux is not booting
SOLUTION BELOW
I have never been in a more confusing situation regarding Linux.
I have a Dell XPS 15 9560, which had a dual boot Windows 10 / EndeavourOS setup. It was running fine for months. 10 days ago I updated Linux and after restart it couldn't boot anymore. It got stuck at "A start job is running for /dev/disk/by-uuid/..." (which is the root partition).
First, with the help of a friend of mine who is quite knowledgeable about Linux (he runs vanilla Arch, etc), we spent 5 hours trying to fix it but had no luck.
Then I decided to back up everything and do a fresh install. Aaaand the same error happened again on the first boot. Then I though "ok, probably some problem with Arch, lets try Fedora". Nope. Some similar error about not finding the root partition. (Here I must say that the kernel which was shipped with the ISO was working fine, but after updating to the latest one, it failed.) Here I thought "ok, then it might be a problem with the latest kernel, let's install EndeavourOS with the LTS kernel." Nope, LTS kernel also didn't boot. Then I tried Ubuntu and it worked, but that's not solving the problem. Then I decided to put another nvme drive in the laptop and try there. The same error again.
Now the greatest part: If I put the nvme drive into an external usb case, EndeavourOS installs, updates, boots without any problem, no sign of the error.
So now I don't know how to proceed... Maybe there is something wrong with the pcie port in my laptop, but except for the booting problem, windows is working, I can also mount and access every partition in the ssd through a live usb. So no other signs of problem with the port whatsoever.
I would be grateful for any advice as I've lost several days trying to solve this and I am out of ideas...
Solution: The last working kernels are from 11. August 2023 (both linux and linux-lts) linux-6.4.10.arch1-1 and linux-lts-6.1.45-1. You can download them from here: linux / linux-lts and install them with
sudo pacman -U the_path_to_the_package
Thank you all for the help!
What is the output of
blkid
and the content of/etc/fstab
of the installed-but-not-booting system? You will need a live usb to get the those, probably.Please use tripe backticks to create codeblocks - much easier to read. Like this:
```
some text
```
results in
I'm also curious about the specific UUID it's looking for. Any chance you could take a photo of the screen and upload it somewhere?
https://postimg.cc/cgvdVrgR
What are the kernel parameters?
cat /proc/cmdline
EDIT actually that will show the live system config. Assuming you're using grub, what is the content of
/etc/default/grub
?Or
/boot/loader/loaders/something.conf
if you're on systemd-bootI'm on systemd-boot. Where isn't a directory
loaders
underloader
, but I found the parameters under/etc/kernel/cmdline
:nvme_load=YES nowatchdog rw root=UUID=9ae3c50f-be08-4594-ac30-2d094375868d
My bad, I think in your case it's in
/efi/loader/entries/something.conf
Since
/
is not mounted, yet, bootloader will not be able to read anything under/etc/
. Unless it's used to automatically populate the loader.conf.Also check
/efi/loader/loader.conf
.I found it!
I've never used
machine-id
with systemd-boot, but everything appears to be corrent. Presumably,/boot
contains a directory named6.4.12-arch1-1
, which contains fileslinux
andinitrd
, correct?You could try rebuilding the initramfs with
mkinitcpio --allpresets
while chrooted.they are under
/02ef85f9edc146d598502c1b296ff64a/6.4.12-arch1-1/
, but yes.EndeavourOS is using dracut by default.
Edit: we tried rebuilding initramfs before, but it didn't help
OK, I see nothing wrong. Let's try building a new config that's as minimal as possible. Copy
linux
andinitrd
files to/boot/
./efi/loader/entries/test.conf
I got another errors now
https://postimg.cc/Z9dyg1Vn
I think failure to change power states is a big issue, but this is out of my depth now. Sorry :(
It matches the observation with the external USB enclosure though. I think the ASPM / ACPI path would be the most promising.
If you know a last working and a first broken kernel version, maybe do a bisection.
Thank you for your time and patience!
I assume it should look like this
So this occurs after an update. Is it not possible to boot into the prior kernel?
If possible to boot into the prior kernel, can you inspect logs or the journal to see where your error is cropping up?
This issue sounds like a regression of sorts with a driver, but log/debug would help confirm. This would be one worth reporting to upstream if you can rescue some logs (I gather you can if you can boot the disk from another enclosure).
If you can boot into the machine, investigate note from the journal:
journalctl --list-boots
journalctl -b -1
,-- If you are booting into a live environment or are otherwise mounting the disk:
journalctl -D /var/log/journal/ID_GOES_HERE
/var/log/journal/2dff8304d5114c44bfb1311357a3cd87
-- Keep us posted.
If truly a driver regression, but you can boot from the prior kernel (if you don't have it, install it via livecd or so), definitely report this one and remain on the prior kernel until resolved. Bleeding edge things.
I have already wiped everything, so no logs... The only way to have it booting is to install EndeavourOS using the offline installer, which is using kernel 6.4.8. There is an option to install the LTS kernel alongside. So the system is booting with 6.4.8, but after updating, neither the new 6.4.12, or the LTS, which is 6.2, doesn't boot. I haven't tried booting with the LTS kernel before updating, to see if the same kernel is working before and after or not. I will try to reinstall it using the offline installer and then try to gather some logs after updating.
If you do go this route - try to update the kernel only, not the rest of the system. Yes, it's bad under normal circumstances, but I think we're way past those.
This would rule out some other part of system getting in the way - like systemd.
I updated only the kernel and I got the following:
(Continues in the second comment)
Did it boot or did it get stuck at the same spot again?
It got stuck at the same spot.
Implies it is the kernel. Mark it on hold and stick to the working kernel for the time being.
Does EndeavourOS use pacman?
You might consider modifying
/etc/pacman.conf
to include the optionIgnorePkg=linux linux-lts
until this is resolved.6.4.12.arch1-1
was added ~4 days ago. If you check the releases arch linux kernel releases here, they have nearly a weekly cadence. This may be a case you can ride out until a newer kernel is released that might solve your issue.If you need access to older releases, see the archive.
For root cause analysis -- is it possible for you to extract/view the journal logs now that you have upgraded the kernel causing the issue? --
/var/log/journal
is a good start. My time is limited, but I'm curious to see what's going on in there. In any case per your and Illecors' testing, it seems you have isolated the issue to thelinux
package.I just realized you're also on systemd-boot (I am too). I'll check into a way to revert back to prior kernels (for I also may run into a similar issue).
edit: updated IgnorePkg line to include both your mainline and LTS kernel (Pacman -Syu updates both) if you opt to hold them for updates.
This is the log https://ufile.io/p5wj1hu0
I'm not sure if this is relevant, but I added
pci=nommconf
to the kernel parameters as I had errors similar to the one in this topic, otherwise the log was over 40 000 lines long.Edit: I'm not sure if this is exactly a kernel problem, as I get the same error when booting with the LTS kernel, which is older than the one installed with the offline installer.
No worries. When checking that output, it is for the working 6.4.8-arch1-1 kernel. The broken kernel boot attempt would be most useful, but I don't want to make you suffer to get it, if you are back to a working system. I think at this point it is safe to say your laptop isn't a fan of the newer kernels.
I would :
/etc/pacman.conf
to ignore updates to packageslinux
andlinux-lts
--
Ideally: You could (from a working system) install a known working LTS image (pkg
linux-lts
), and exclude that from updates until you land on a working kernel release (keep an eye ontesting
andcore
repos once a week or so). in this way, you'll have a working LTS, and can upgrade/downgrade mainline kernels as you please, booting back into LTS to correct issues should they arise.edit: minor
I found the problem! It was actually the kernel, just both the latest and the lts versions released after 11th of August don't work. So for now I need to stick to the versions from 11th. Thank you very much for the help!
Edit: this is the bug
According to your logs when installing the kernel, your nvme modules are nowhere to be found, and basic system tools are unavailable. Your core system seems severely borked... can you do a memtest (broken ram can corrupt your storage and your attemps at finding the problem) and check your smartctl --all of that ssd?
when it gets stuck on something like that, it's because of a hardware related change; did you update or modify your bios recently or change hardware drivers?
Use a live USB to back up your home directory and a list of your packages, then reinstall. I don't think it's worth the trouble.
AHAHAHA that is a proper insane bug. One PCIe device shouldn't be able to slap others off of the "bus", "we're not on a bus all you did was mess up your own personal lanes mate".
I lost 3 days of my life + now I need reinstall everything 🥲
Try adding "PCIE_ASPM=off" to the kernel boot commandline.
This didn't help :/
That's a weird issue. Do you have encryption on by any chance? I had a similar error pop-up when I didn't have the correct systems hooks for the kernel so that after install update of kernel would make the system not boot.
No, I have never used encryption
Still might be worth it to check that you're not missing any kernel modules.
Share the errors you are experiencing, please
Could you please take a look at the comments from abrer and Illecors, I have shares a lot of info there.
Could you link to them or quote them, please?
The problem is already solved (At least I found the bug and a temporary solution) At the beginning of the post there is a link to the bug report.
I had se similar issue but with bluetooth, worked on windows not on linux. Fixed it by resetting the BIOSwith the CMOS battery thing
Shot in the dark: Did SecureBoot get re-enabled somehow? Could you check the UEFI?
Hard to not sound condescending in text, so forgive me.
If systemd is trying to mount the drives, we're at a point where the (most likely) ramdisk has already booted the kernel - secure boot would've prevented that.
I won't take it as condescending then! Thank you for the input :)
Contrary to popular belief using and maintaining arch is a novice exercise. The
btw
crowd likes to believe their actions are '1337' 'minimal' 'no bloat', etc. and they've sure got the neofetch receipts to prove it!The truth of it is that arch is easy, that's the whole allure. It is the most convenient distro for experienced users with specific (yet broad) set of needs.
The
btw
crowd are simply misled kooks deserving of pity and mercy, but not trust.https://en.m.wikipedia.org/wiki/Four_stages_of_competence
If you're so smart and competent, help OP
Note to future self: Don't dual boot any system EVER.
Dual (or triple or quad) booting isn't the problem, I did it for a long time, until I had a machine that could handle more than one virtual machine while leaving the base OS also usable.
I always dual boot and have never had an issue in many years. I wonder what I am doing differently.
The only way I could dual boot without Windows breaking things was to unplug the Windows drive, install Linux on another drive, plug the Windows drive back in, and boot to Windows by selecting it in the BIOS.
Trying to use grub would always have issues eventually after booting into Windows.
That's definitely not the norm. Used to be that installing Windows would wreck Grub, but you just needed to but a rescue disk and reinstall Grub one time to fix it. Most people dual booted for decades without any issue there.
To be fair this would have been decades ago. Ubuntu 8.xx and either XP or 7. It was my first dive into Linux so there's a good chance I did something wrong.
I was always able to repair grub when windows would wreck it thankfully.
My current setup is using the former, but I did the latter for so many years and the only issue I had was the clock would get fucked up after booting Windows. But I agree, it's much safer to have your Windows SSD physically removed when installing Linux so grub doesn't get installed on the same SSD as Windows. I wouldn't do it with all the horror stories I have read. My good experience was probably dumb luck.
Couldn't you use grub on one drive to point to the bootloader on a separate drive? Windows should leave that configuration alone, at least in theory.
Probably, but I just didn't want to give it the chance. It ended up being a non-issue in the end because I never booted back into Windows.