USB EFI Boot/Install

Hi,
thanks for reporting the issue with much detail.

I will check for adding a UEFI boot option to the live ISOs in the next days. Currently you may create a UEFI bootable pendrive by using the “Install openmamba on USB” tool but of course you need to have a working openmamba desktop installation to be able to run this tool.

As for the SDDM startup error, recently this issue was caused by a missing /lib/systemd symlink pointing to /usr/lib/systemd, you may want to check if the link is missing and if so creating the symlink should fix the problem. I thought the problem would be fixed with the latest ISOs but I will check and eventually do a fix for this problem.

Hi Douglas,
by looking at the report you sent it seems that Xorg is unable to start because the kernel is started with the “nomodeset” option, so it would skip AMD/ATI drivers and try VESA/Framebuffer modes which fail too probably because VESA/Framebuffer modes are not available too. So for this part of the problem I would suggest to boot without the “nomodeset” option, then if you are able to access a console a report in this situation would be more interesting to see. If you can’t get to a console the boot could be blocking with some relevant message on screen instead (assuming that you are not booting with the bootsplash enabled, so without the ‘quiet splash’ kernel command line arguments.

Using the diskimg is a good idea to perform a EFI boot. The bootx64.efi from grub 2.04 has been the same for a long time and it works where I have tested it, BTW I’m working on Grub 2.06 update. The boot file has just to be inside a FAT partition and the PC has to be configured to support boot from the external device (i.e. things like secure boot need to be disabled). BTW I recently had a similar problem on a x86_64 tablet where bootx64.efi is no longer detected so I will be able to further investigate on this problem.

• This reply was modified 3 years, 2 months ago by Silvan.

Hi Douglas,
I’m glad that you could boot openmamba ISO and install it in CSM mode.

You could now try to create a USB pendrive from a ISO using the provided openmamba tool you find in the start menu by typing USB, then ideally boot in EFI mode from this pendrive and install (the calamares installer supports EFI boot).

My doubt is if it is possible (and whether other distribution do) to distribute a ISO image which you can write with dd into a pendrive and which can be booted in UEFI, but I don’t think so.

For this reason there is the openmamba diskimg download, an image that you already tried which contains the EFI ESP partition and should boot. But I think that in EFI boot mode there might be, as previously noticed, KMS video driver problems that should be investigated further…

Hi Douglas,
this means that the glibc-2.34-4mamba upgrade, which would make /lib[64] a symlink to /usr/lib[64], was a failure. The linux loader /lib64/ld-linux-x86-64.so.2 would disappear and no executable can be run, it is not a problem with the kernel, which will panic because it won’t be able to run /sbin/init after switching to the rootfs.

The fix is to either start from the most recent ISO livecd image or do a manual recovery by booting a live system and perform a series of terminal command such as:

wget https://cdn.openmamba.org/pub/openmamba/devel/RPMS.x86_64/glibc-2.34-4mamba.x86_64.rpm
sudo su
mkdir /mnt/sda1
mount /dev/sda1 /mnt/sda1
mount -o bind /dev /mnt/sda1/dev
mount -o bind /proc /mnt/sda1/proc
mount -o bind /sys /mnt/sda1/sys
rpm -r /mnt/sda1 -u glibc-2.34-4mamba.x86_64.rpm
chroot /mnt/sda1
dnf update

This would fix the main issue although repairing of any of the packages which were part of the upgrade may be needed as well.

• This reply was modified 3 years, 2 months ago by Silvan.
• This reply was modified 3 years, 2 months ago by Silvan.

Hi Douglas,
The instructions given are quick guidelines needing interpretation, as you did.

For the download URL, my mistake, the URL needs to start http instead of https.

Ok that you identified as /dev/sda2 and mounted the rootfs partition.

Next, the glibc recovery procedure (using rpm -r .. where -r is the option to specify the target rootfs path) was needed because I’ve assumed that during the upgrade in the previous post glibc broke and so all the system broke. This indeed is contradictory with the fact that you have been able to chroot to that rootfs with success and without fixing glibc before.

There are currently two different problems I would put the attention on:

1) glibc upgrade with migration from /lib64 to /usr/lib64 causes severe problems if it fails (i.e. the system is unbootable, and the kernel panics altough it’s not kernel’s fault);

2) sddm, the KDE Plasma login manager, fails to start the graphical desktop. This is the situation where the system enters graphical mode, you have the mouse pointer available on screen, but you get no more than a black screen. You can btw switch to a console by pressing CTRL-ALT-F2.

My previous post was about fixing the issue 1), it explains the guidelines for a “almost disaster” recovery procedure and is intended to be performed by a quite expert and conscious user. I’m not sure any longer that this applies to your last situation reported.

For the issue 2) I haven’t found a certain solution yet. There is a sddm upgrade which I’m testing and suggest to try to apply first (please note that I mean sddm, not sddm-kcm), while a workaround seems to be removing the file .Xauthority for user home and restart sddm:

rm -f ~/.Xauthority
systemctl restart sddm

By using the most recent livecd ISO the glibc upgrade problem is bypassed, but the sddm issue remains to fix.

Hi Douglas,
you say you could install the system and it was working until you did a system upgrade and reboot. So what ISO release did you install from, in order for me to limit which are the updates that caused the problem?

The nomodeset is an argument that is only used to support non-KMS drivers, like NVIDIA proprietary drivers (well, to be correct this is also not true anymore with recent drivers). In this case you use it as a workaround to get to a console but the system is not expected to start correctly in graphical mode with this option (it should with fb or vesa drivers but maybe not on UEFI btw we are not interested in starting the system with this option). So if the kernel fails to boot in default mode it should be started without the quiet splash arguments (and maybe also adding debug but this may cause excessive flooding of messages) and what appears on screen when it stops (a screenshot?) could be useful for debugging.

As for the startx command, as I told there is not interest in booting with nomodeset because the desktop is not expected to work correctly at all with standard video drivers.

Of course you can try to install other login managers but openmamba is designed to work (and did for years) with SDDM like any other KDE Plasma based distribution. BTW there is no lightdm package in openmamba, there is lxdm (which is used by openmamba-light flavour with LXQt).

As yours and other reports suggest I will try to package a kernel update, although I have three systems correctly running with this kernel (with Intel and AMD gpus) so I think the problem is elsewhere.

Hi Douglas,
as you installed starting from a june snapshot you should still have the previous kernel installed (5.10.56) and available in the Grub advanced options submenu. Did you try that? According to other reports problems seems to happen since 5.10.57 update and this would explain why your system worked before updating.

The kernel is built from the kernel source rpm that you can find in the sources folder.

• This reply was modified 3 years, 2 months ago by Silvan.
• This reply was modified 3 years, 2 months ago by Silvan.

Hi Douglas,
I’ve packaged a new kernel update which is under testing, if you want to test it against the problems you are having in your installation you may install it with the following command:

sudo dnf update kernel-mamba-x86_64 kernel-mamba-x86_64-headers --enablerepo=unstable-makedist

Hi Douglas,
thanks for reporting your attemps with openmamba.

Here are the news from my side: after deep research I identified the cause of Plasma Desktop not starting in the file /etc/xdg/kwinrc which would cause the window manager process kwin_x11 to run at 100% CPU and block the desktop startup. This only happens the first time a user logs in, because thereafter the file kwinrc is read from the local home directory. So the fix for this is in the desktop-base-kde update to version 5.0.

Other problems you may have after updating depend on the fact that you are starting from a three-months old live image and here the glibc update to the /usr flat filesystem is probably the cause of further problems, so I would suggest you to try the openmamba-diskimg-livecd (with EFI boot support) or openmamba-livecd that have been made available between yesterday and today with the more recent kernel and everything fixed from my point of view.

Hi Douglas,
I’m glad that you could finally have a working installation starting from the new downloadable image.
I can’t fully check the installation procedure every time new images are released and I did not receive any report before you so this was a problem that lasted for months and now I’m happy that it is resolved.
So thank you for your detailed reports and your patience.