Failed to open \EFI\com.solus-project - Not Found

BuzzPCSOS

Axios it kind of looks like the bios has a maximum filename length threshold. Now I want to install windows again in UEFI because I know that worked, so that I can look at the efi boot address set in the bios for that. I'm guessing the address for the .efi file would be shorter.
It would be interesting to be able to rename the grub64.efi file to something shorter and then try pointing my bios at that. If it didn't wreck my computer it would certainly prove my theory one way or the other 😄. Unfortunately not sure I would be too confident working with the efi partition. These things are meant to be hard to access for our own protection.

Axios

BuzzPCSOS https://uefi.org/specs/UEFI/2.10/13_Protocols_Media_Access.html
The firmware automatically creates handles for any block device that supports the following file system formats:

FAT12

FAT16

FAT32

I would think file length would not be a problem
interesting

BuzzPCSOS

Axios Agreed. Although the Windows path is much simpler \EFI\Microsoft\Boot\bootmgfw.efi
I wonder if it could be that simple, just dumb down where Solus puts it's .efi loader.
Almost definitely a bios issue, as you say more modern UEFI machines have considerably less issues and they seem to enter UEFI without any issues. Although most boot so quickly that only eagle eyes would spot the Failed to open message.
Back in the days of BIOS/UEFI combined loaders there were considerably less universal standards so this sort of fallout figures.

BuzzPCSOS

Axios So a little more.
I compared the path for Windows UEFI boot with the path for Solus and finally with the path that my BIOS returns if I try to navigate to the .efi file manually
\EFI\Microsoft\Boot\bootmgfw.efi /EFI/com.solus-project/grubx64.efi /EFI/com.solus-project/grubx64.e

The number of characters in the Windows path = 32
The number of characters in the Solus path = 34
The number of characters that my BIOS truncates the path to = 32
This looks to me like this is the problem, whatever the install software sets the path to does not matter because my
BIOS only recognises the first 32 characters. If it is more than that it cannot find the .efi file and displays a "failed to open" message and drops back into a hybrid state, where the installer will go through the motions of creating a GPT install complete with an efi boot partition but the format of the drive will be MBR.

brent

BuzzPCSOS I compared the path for Windows UEFI boot with the path for Solus and finally with the path that my BIOS returns if I try to navigate to the .efi file manually

wow. fantastic work. where did you see the line/EFI/com.solus-project/grubx64.e ? In the file manager? Or in the bios? 32 can't be that common, can it, as far as bios limitations?
again, nice find.

Axios

BuzzPCSOS I see what your saying you got figure a process of elimination to really be sure

/EFI/com.solus-project/grubx64.efi
/EFI/com.solus-project/grubx64.e
If this is what you mean that it just deleted the (fi) am having hard time wrapping that around my head
something just doesnt ring right.
I say that because gubx64.efi is short filename why would it care how deep or what the directories are called?

(Got to ponder)
Note if grux was wildly long filename i can see it shorting it (at least with early windows kinda how it worked)

BuzzPCSOS

brent It seemed to fit. Trying to manually point my BIOS at the.efi file that failed to open on boot I noticed that it let me navigate in the BIOS to the .efi file but when I tried to save the path it was shortened to only 32 characters.
A little bit of research showed me that the path to Windows .efi file was exactly 32 characters long and that Windows will load faultlessly in UEFI mode on this machine.
It just felt like too much of a coincidence.
Suppose I should work out how to place this as an issue on the Bug Tracker without having very much proof for my theory.

brent

BuzzPCSOS Suppose I should work out how to place this as an issue on the Bug Tracker without having very much proof for my theory.

your theory seems convincing and/or 'has legs.'

BuzzPCSOS Trying to manually point my BIOS at the.efi file that failed to open on boot I noticed that it let me navigate in the BIOS to the .efi file but when I tried to save the path it was shortened to only 32 characters.

an unorthodox route to find out but that's where the best stuff gets discovered. Wonder if solus filename can be omitted 2 letters while preserving the efi in .efi? Anyway, it's interesting.

brent

Axios If this is what you mean that it just deleted the (fi) am having hard time wrapping that around my head
something just doesnt ring right.
I say that because gubx64.efi is short filename why would it care how deep or what the directories are called?

interesting, plot thickens

BuzzPCSOS

Axios Not entirely sure how to prove this to myself even. I do know that Ventoy will boot my laptop into UEFI mode and that UEFI only fails as I subsequently boot into Solus Live. Maybe I should look at the path to the .efi file in Ventoy.
Because this laptop is an early (2010) dual boot UEFI / BIOS machine it would make sense that some of the restrictions that necessitated the use of a BIOS boot loader were still being ironed out on the early UEFI boot mode.
Historically I have been able to navigate to the 32 character long Windows .efi boot file manually from the BIOS and fix a non-booting system. I can not do the same thing for the 34 character long Solus .efi boot file without the path being prematurely shortened when viewed from the BIOS.
The same boot media has no problem loading the .efi file on later computers.
Edit: A path length of 32 characters would represent a storage area of about 32 bytes which I think sounds sensible for early UEFI setups. BIOS chips did not have the room for extensive graphics and everything that we now expect to see. It seems very suspicious that the Windows path is the very same length that my BIOS cuts off at.

Harvey

Axios I say that because gubx64.efi is short filename why would it care how deep or what the directories are called?

Everything must have a defined max length, including the max characters this forum will allow in 1 post. I don't know what that limit is but I know it has to exist.

It is not that the firmware cares about filename length specifically it is the limit they set for storing the path+filename.

Developer 1: What is the max length we should support?
Developer 2: Windows requires 32, so we should support at least that.
Developer 1: 32 it is!

brent

BuzzPCSOS Edit: A path length of 32 characters would represent a storage area of about 32 bytes which I think sounds sensible for early UEFI setups. BIOS chips did not have the room for extensive graphics and everything that we now expect to see. It seems very suspicious that the Windows path is the very same length that my BIOS cuts off at.

Harvey It is not that the firmware cares about filename length specifically it is the limit they set for storing the path+filename.

Developer 1: What is the max length we should support?
Developer 2: Windows requires 32, so we should support at least that.
Developer 1: 32 it is!

So Buzz's hypothesis is right...go back to beginning..."The error displays for a few seconds and then the laptop continues to boot."

why does it continue to boot? When did (bios?) it recognize that 32-to-34 was a temporary hurdle?

Axios

Harvey It is not that the firmware cares about filename length specifically it is the limit they set for (storing) the path+filename.
That totally slipped my mind.

BuzzPCSOS

brent why does it continue to boot?

Because the Solus Live media never made it into UEFI in the first place. I see the "Failed to open" message flash up as it boots from the ISO.
I've got two computers that don't like Solus.
A Toshiba Satellite that requires Legacy but installs with a GPT formatted hard drive and cannot boot because of that and a Dell E5410 that can run with GPT but ends up with MBR formatted hard drive.
I fixed the Toshiba by manually setting the partitions with an MBR format, I suppose the manual fix option is available for my Dell but I don't have the skills to start editing EFI boot partitions. Yet.
Edit:
A quick wander round the internet shows all the UEFI boot paths that I could find for various distros (Fedora, Mint, OpenSuse, Ubuntu) were less than 32 characters. Next task is to try running one of those and see if I can load to a true UEFI environment.
Edit 2:
Well that did not span out as I expected. I formatted a USB stick with GPT format and ext4 file system to use as a simple test of being in UEFI environment. From a legacy boot a GPT drive cannot be mounted as readable media.
All OSs were run from another USB with Ventoy as the boot manager.
Linux Mint - Test drive wouldn't mount
Fedora Workstation - Test drive wouldn't mount
Solus Plasma - Test drive wouldn't mount
Spiral Linux -Drive mounted perfectly
I saw a few messages about MOK manager and secure boot not being present which makes sense because secure boot is not supported on my laptop.
I wonder what Spiral Linux is doing differently? It was also the only OS that I could find that would install onto my Toshiba laptop without tampering with the default install method.
More research needed.

Harvey

Your guess is as good as mine.

brent

Harvey maybe the storing of 32 was conditional or flexible somehow? Eh, lager time. We all know a lot more than when this started in July..BIOS works in mysterious ways..🙂

brent

BuzzPCSOS Well that did not span out as I expected. I formatted a USB stick with GPT format and ext4 file system to use as a simple test of being in UEFI environment. From a legacy boot a GPT drive cannot be mounted as readable media.
All OSs were run from another USB with Ventoy as the boot manager.
Linux Mint - Test drive wouldn't mount
Fedora Workstation - Test drive wouldn't mount
Solus Plasma - Test drive wouldn't mount
Spiral Linux -Drive mounted perfectly

I thought that would be a successful experiment too. then again Ventoy is batting about 50% for me as far as loading isos and exploring them...or installing them 🙂

Axios

BuzzPCSOS I dont have anything to add but watching thread.
Note it interesting that spiral works it uses same installer doesnt it?
Maybe need try check that out more
(Just thinking)

BuzzPCSOS

brent I have had mostly good experiences with Ventoy but just to be sure I also tried Etcher and burning the iso directly to DVD and still got the same result.
Maybe I need to follow the MOK manager notifications and see where that takes me.
I feel quite glad that I didn't end up posting a bug report before I had made a bit of an effort to prove my theory.

BuzzPCSOS

Axios it is on my to-do list.
At present fiddling with my Linx 12x64, upgraded to workstation 41 and then installed Fedora KDE Plasma Mobile on top. Lots of cleanup to do after that but I can now use it without a mouse or keyboard attached.
Strangely Plasma feels more fluid than the original Gnome and all of my programs work now.

« Previous Page Next Page »