For an efi system using btrfs file systems, failure to properly mount /boot will cause boot failure – always.
You do not show /home as mounted, so I have to wonder where it is and why it would not have mounted properly when running mount -a
You should have then had the 4 virtual file systems already mounted plus 4 physical file systems mounted.
/ which is probably btrfs
/boot which is ext4
/boot/efi which is Fat32
/home which is also probably btrfs
My instructions were to mount the root file system (/) on /mnt and the /proc, /run, /sys, & /dev on the appropriate locations on the root file system then to do the chroot followed by mount -a for everything else. This would have mounted /boot, /boot/efi, and /home for you if /etc/fstab is properly configured.
Please boot to the live usb, and mount the root file system on /mnt. Then post exactly the output of lsblk -f and cat /mnt/etc/fstab. Please do not censor any of it so we can see the actual hardware file systems you have and what should normally be mounted when the system boots.
I have been hobbled in being able to give full and accurate commands since you have not shared any details on the system.
I’m not sure what details are needed excatly but here are some.
It’s an Acer laptop running intel processor and iGPU. It was dual booted with Fedora and Windows 10. It just occured to me that perhaps the system on laptop is a little bit on the older side. So here are the full specs that should be important.
Intel i5-4200u 1.6GHz
Intel HD Graphics 4400
8GB DDR3
/dev/sda2 should be mounted at /boot/efi. You can see that in the second line of the fstab file.
Your file system is ext4 and not btrfs. You have no separarate /boot or /home partitions.
For future posts, please always (if possible) use copy & paste of the text from the screen into your post as preformatted text using the </> button above.
This does a couple things. First an image cannot be searched so the info cannot be located by others that may have similar problems. Second, if we need to quote the data in the image it must be hand copied and typed into our response rather than copy&paste. Text is always best if possible.
I will modify the earlier posted instructions as follows.
boot from the live image and open a terminal
gain root permissions with su
Mount the root file system at /mnt with mount /dev/sda6 /mnt
mount the extra but necessary file systems. (all are virtual file systems) for dir in proc sys run dev ; do mount --bind /$dir /mnt/$dir ; done
mount the efi partition mount /dev/sda2 /mnt/boot/efi
now chroot to /mnt chroot /mnt
Verify that everything is properly mounted with mount to display all mounted file systems in this chroot environment. The result should look very similar to the result from your normally running system. You should see /, /boot/efi, /proc, /sys, /run, and /dev.
Now you can do all the needed recovery steps as if you had booted directly to the system. Those suggested steps are below.
Note that earlier you claimed to mount /dev/sda2 at /mnt/boot. That was definitely an error and we need to confirm that the data is in the proper places in /boot and /boot/efi.
You should be able to connect to the network properly when in the chroot environment so copy & paste into your posts should be quite easy.
I suspect that at least part of the problem is that you have /dev/sda1 as the efi partition for windows and /dev/sda2 as the efi partition for fedora. This makes things a little difficult when dual booting as it would have been easier to allow fedora to share the same efi partition
We need to see what the system indicates is the boot order.
Please, after doing the chroot again, run efibootmgr and let us see the result.
Also the next steps will be to ensure grub is properly configured for booting. That will require some repairs.
post the output of ls /boot/loader/entries
remove the 2 grub.cfg files rm /boot/grub2/grub.cfg rm /boot/efi/EFI/fedora/grub.cfg
repair grub booting and ensure the kernel is properly located with dnf reinstall grub2-common grub2-efi* kernel* which will recreate both the files just removed.
Depending upon what was shown with efibootmgr now we might be able to boot, but possibly not as a result of the earlier stated incorrect mounting of /boot/efi on /boot. Hopefully the reinstall done in step 3 will take care of that for you.
Sorry about posting an image. It didn’t occur to me till now that i could just log in to this account from live media and copy paste it so i used the phone. I don’t want to make empty promises, but I’ll try to document everything well when this is all fixed - really don’t want someone to have a lot of trouble as i did with this problem so i hope someone will find it useful.
I’m about to follow the instructions, then I’ll come back later to show results.
I understood from earlier posts that when i log out and reboot from live media, everything will unmount by itself. I just now booted at live media and i didn’t mount anything, but when i ran
mount /dev/sda6 /mnt
As root i got this notification
mount: /mnt: /dev/sda6 already mounted on /mnt
What do i do? Proceed with it normally? Why didn’t it unmount when i rebooted?
Edit: Unfortunately it seems i hit some kind of limit for replies and i have to wait a couple of hours. Here’s what i wanted to to write in the reply
I think i know what was wrong. I should have just rebooted again in live media, right?
What i did was pressed restart, then i went into uefi, changed boot order to fedora, and then tried to boot into fedora, later removing flashdrive.
Logging out and logging back in does not dismount anything. A power off or full reboot should dismount everything and start new.
That tells me you may not have actually done a full reboot since it seems the live media was still active.
How did you do the reboot? Did you use the icon in the upper right corner of the screen and tell it to reboot? Or did you by chance happen to click the logout which is the same location? A reboot or logout would have each given the popup in the middle of the screen for confirmation.
On my iMac the little white underscores stays there for some time before the GUI login appears. I asume it can take longer if houskeeping (rotating logs, checking filesystems) is needed. You may need to wait several minutes. I havn’t tried getting a console (<Ctrl+ALt 3>). If you have another computer, see if you get a ping response. while waiting If you can ssh into the system, you can try running dmesg -H tail:
% dmesg -H |tail
[ +0.003145] tg3 0000:04:00.0 enp4s0f0: EEE is disabled
[ +0.003169] IPv6: ADDRCONF(NETDEV_CHANGE): enp4s0f0: link becomes ready
[ +3.749957] IPv6: ADDRCONF(NETDEV_CHANGE): wlp3s0: link becomes ready
[Feb 8 11:44] rfkill: input handler disabled
[ +1.314878] Bluetooth: RFCOMM TTY layer initialized
[ +0.000018] Bluetooth: RFCOMM socket layer initialized
[ +0.000009] Bluetooth: RFCOMM ver 1.11
[ +13.197202] rfkill: input handler enabled
[ +3.186784] rfkill: input handler disabled
[ +0.262676] hfsplus: write access to a journaled filesystem is not supported, use the force option at your own risk, mounting read-only.
You can see that the last few lines took nearly 20 seconds.
Well, I tried rebooting everything again and got the same error. Good thing is Windows 10 is back in grub. I haven’t tried booting into it tho, but i will and when it hopefully works I can make a live inage of Fedora 37, since i’ve been using Fedora 34 live image untill now. Hopeful that wasn’t making any problem?
By the way, i couldn’t send a message for 4-5 hours because i’m a new member and i hit some kind of limit. I edited the newest message before this in hopes I’ll get some answers but i proceeded with doing the procedure anyway. I rebooted to the live media and everything was unmounted
First of all it shows that black screen white underscore for a second then it has Acer logo and it seems to be booting but then we get back to black screen white underscore.
I set a timer while booting and that final black screen white underscore was there for 27 minutes so i decided to force shut down with a pshysical button
I’m not sure if You got a notification on my reply(I edited a couple of times). Sorry for bothering You if You did get it, but I’m not sure what next step should be.
On my imac I sometimes get the grub menu or the grub prompt (probably something to do with the timeing when I press “Alt” after the startup chime. If I choose the latest Fedora kernel from the menu or type “exit” at the grub prompt I get the underscore screen. I tried switching to a console with . The underscore went away but I never get a login prompt.
The iMac is in an outbuilding so no easy way for me to try ssh from another system while the underscore is visible. I always remove the “rhgb quiet” from the kernel command line, so I see the messages until the graphics login come up. You should be able to edit the command line, either using the rescue procedure or using the grub menu. This might show you what is preventing the system from reaching the login screen.
I removed rhgb quiet and although everything was moving pretty fast i saw a couple of erros that were same as before.
However, i wanted to ask something else.
You know that string on letters and numbers after root=UUID=?
Is that string supposed to be the same as the newest kernel? Because if it is, in my case it isn’t.
I’m also not sure what ro after it stands for, but I guess I’ll find it in some kind of docs.
Well, that isn’t a problem then. But I’ve got some information that may be useful.
First of all ls /etc/modules-load.d is empty. It makes sense why the error is occuring, there are no modules at all. How do i get them back?
Second
I will modify the earlier posted instructions as follows.
boot from the live image and open a terminal
gain root permissions with su
Mount the root file system at /mnt with mount /dev/sda6 /mnt
mount the extra but necessary file systems. (all are virtual file systems) for dir in proc sys run dev ; do mount --bind /$dir /mnt/$dir ; done
mount the efi partition mount /dev/sda2 /mnt/boot/efi
now chroot to /mnt chroot /mnt
Verify that everything is properly mounted with mount to display all mounted file >systems in this chroot environment. The result should look very similar to the result from >your normally running system. You should see /, /boot/efi, /proc, /sys, /run, and /dev.
Now you can do all the needed recovery steps as if you had booted directly to the >system. Those suggested steps are below.
Shouldn’t we have mount -a after chrooting? I actually repeated chrooting procedure with mount -a and ran updates, dystro-syncs as well as the part about grub in original post.
dnf reinstall grub2-common grub2-efi* kernel*
This command is interesting because i noticed this, error, i guess:
Also, due to grub detecting Windows 10 again, I put Fedora 37 on USB so if nothing works, I’m fine with reinstalling. Ideally keeping all of the data, but let’s see if these additional detailed helped in any way first.
You will: v6.1 kernel parameters: with ro the root device is initially mounted read-only.
I have many years more experience booting from BIOS with init , so not sure how UEFI will behave, but you could try booting to runlevel 3 (multiuser). You used to be able to append the runlevel number to the kernel command line in grub, but now you may need to use the rescue system to change the systemd target from graphical to multiuser.
Adding some more interesting info. I get this error in booting live media of Fedora 37, that might as well be related to Fedora 37 upgrade, but I don’t know.
dracut-pre-udev[621]: sh: line 1: /sbin/sysctl: no such file or directory
Considering how a few posts here about this were about kernel and how i remember dracut was shown in some error in if i remember correctly systemctl logs i decided to add this info too. Also, on one of the forums online where they have a similar problem, one of the answers were to run
sudo update-initramfs -u
Which is apparently in /sbin. I tried running it and still couldn’t find it, even in /sbin.
Considering all of this I couldn’t but wonder that these might be connected.
Either way is fine. I put the extra mount command in step 5 and since you only have 2 partitions to mount in the fstab from the main drive step 3 & step 5 mounted them.
That is not an issue. You booted from the that kernel with the live media so it would not be found in the chroot environment. It was trying to remove the oldest kernel it knew about but the running kernel was not installed.
No, not bad. The 4th is the rescue kernel with 3 real kernels installed.
Errors in booting the live media do not necessarily carry over to the installed and updated system. After all that kernel is several updates behind, along with many other updates in the entire system.
What matters now is if you can actually boot the installed system and errors that may appear when attempting to boot.
Does the file /sbin/sysctl exist in the chroot environment? It is provided by the procps-ng package which should be installed by default. If that file does not exist then you may also need to reinstall it, but that should not be the issue here.
If it still will not boot then my last suggestion is to boot into the chroot environment as you have been doing. Then after verifying the proper partitions are mounted as shown in step 8 above try dracut --force and wait for that to complete before attempting to boot again.
dracut: cannot find module directory /lib/modules/6.0.7-301.fc37.x86_64/ dracut: and --no-kernel was not specified
I have the newest kernel in /lib/modules yet dracut doesn’t seem to care. I also have like 20 of them from f34 f35 and two from f37.
uname -r
Gave this kernel:
6.0.7-301.fc37.x86_64/
But the newest one is actually
6.1.9-200.fc37.x86_64
Both chroot and live enviornment have the same older kernel…It’s as if something is not detecting the new one. Which would then make sense why modules cannot be loaded.
Did you try adding the --no-kernel option to the dracut command as that message suggested?
Dracut is clearly telling you it cannot find the kernel files for the running kernel, which is expected in your situation.
You are still not telling us the exact messages you see when attempting to boot.
For the dracut issue I used man dracut and I see
A shortcut to generate the image at the default location for a specific kernel version is:
# dracut --kver 2.6.40-1.rc5.f20
then the use of dracut may not be required, but to make it work for that exact kernel the command should be dracut --force --kver 6.1.9-200.fc37.x86_64. Without using the --kver option dracut attempts to rebuild the images for all kernels, installed and running, which is giving you the error because the running kernel is not installed.
I see no messages attempting to boot. Nothing. Should i write down what i see when i remove rhgb quiet?
--no-kernel gives this:
dracut: Can't write to /boot/efi/734812ce8beb4231b9894eb577bbc3d8/6.0.7-301.fc37.x86_64: Directory /boot/efi/734812ce8beb4231b9894eb577bbc3d8/6.0.7-301.fc37.x86_64 does not exist or is not accessible.
ls /boot has all the files mentioned.
I think i already tried running this with --kver but it didn’t work. Going to try again but I don’t know I’m doing something wrong:
I run the command, exit chroot, reboot into the live usb again, then reboot into normal boot and see if it works?
I rebooted again ans interestingly, while booting it now states
Booting 'Fedora Linux (6.1.9-200.fc37.x86_64) 37 (Workstation Edition)
