Legacy BIOS Support Remains Important For Some On Fedora, May Shift Responsibility To SIG

[dragon321]

That's not true. I have an HP laptop and it does not boot windows uefi by default and it is not hardcoded.

Then it was probably fixed in your machine but that doesn't mean it wasn't present in other machines. You can easily verify that by yourself, there are plenty of discussions mentioning this issue. It's also pointed in Arch Wiki in HP laptops page.

[Slartifartblast]

Microsoft being evil doesn't rule out Apple being evil.

I have no argument with that, I despise Apple and their ransomwares.

[willmore]

Have you read Bill Gates's 1999 infamous email about ACPI ?

http://antitrust.slated.org/www.iowaconsumercase.org/011607/3000/PX03020.pdf

Microsoft being evil doesn't rule out Apple being evil.

[Slartifartblast]

I'm going to place some of that blame on Apple because they pushed it early on and did things Their Way(tm). I'd love to see some major manufacturer get behind one of the open UEFI firmwares and use it for all of their products. Maybe that would push the industry towards the right direction.

Have you read Bill Gates's 1999 infamous email about ACPI ?

http://antitrust.slated.org/www.iowaconsumercase.org/011607/3000/PX03020.pdf

[zerothruster]

This sounds unlikely. As far as firmware goes, all that matters to an x86 CPU is that the EEPROM gets mapped to the right address. It is likely that no vendors of mobos with the right socket provide BIOS versions tho.

I gave you a 'like' because you are technically correct.

[sinepgib]

Unfortunately, ryzen2 and ryzen3 APUs do seem to require UEFI.

This sounds unlikely. As far as firmware goes, all that matters to an x86 CPU is that the EEPROM gets mapped to the right address. It is likely that no vendors of mobos with the right socket provide BIOS versions tho.

[zerothruster]

Its 2022, the last machine I have without EFI support is a i286 notebook. I wonder how large the user-base of systems older then 15 years really is. A high end system from back in those days is slower then a cheap notebook is today.

It has nothing to do with 15 years, only with systems that shipped with windows or macOS. I bought an athlon 200ge for my home server, and a ryzen 3400 (picasso) desktop system in the second half of 2019. Both use BIOS booting (why would I want the general inconvenience of UEFI if I can avoid it ?). Unfortunately, ryzen2 and ryzen3 APUs do seem to require UEFI.

Would these two systems of mine support being booted with UEFI ? Probably, but I'd have to throw away the current systems (plural) and go from a simple "this is how to fix grub when you screw up" to jumping through holes to find out what strangenesses the mobo manufacturers decided to come up with.

[kreijack]

I'm going to parrot my talking points from the last article about this:

All they have to do is adopt the Clover Bootloader for x86_64 platforms...

This was my first thought (not Clover specifically, but DUET and/or tiano). However this would not solved the problem that Fedora is trying to address: simplify the installer.
After a linux boot, it doesn't matter if it was started from a bios, from a firmware or from UEFI. In fact the kernel linux already abstracts the underling firmware/bios. The kernel image is the same, both for the bios that uefi.

The complexity is in the linux installer and very few specific packages that deal with the underling firmware.

What fedora is trying is to simplify the burner of supporting both the bios and the uefi; likely the first needs[*] a DOS partition scheme, where the latter mandates a GPT one. The bootloaders are different (grub vs grub-uefi), so the bootloader installer.

I am not saying that I fully agree with the Fedora idea, however using another bootloader is not a simplification of the installer. So likely it is not the solution.

I have to point out that the linux already installed will be supported (you can upgrade). It will be impossible to install a new Fedora to an old bios computer.

Finally I am curious about the system that are uefi but boots in legacy mode because thy saw a DOS partition....[*] I know that it is possible to use a GPT partition schema in a legacy bios environment. But covering this adds others complexity instead of a simplification.

[sinepgib]

I really hope we'll see a move to unified kernel images and systemd-boot soon. It's a breath of fresh air. Much simpler overall, you can easily realize a fully measured boot and you can get rid of the /boot partition among other things.

CloverBoot definitely seems like a good solution for legacy systems.

Yep, I love those little things from UEFI. Not needing GRUB, having proper metrics of boot time, the option of signing my kernels, etc.

Its 2022, the last machine I have without EFI support is a i286 notebook. I wonder how large the user-base of systems older then 15 years really is. A high end system from back in those days is slower then a cheap notebook is today.

Useful hardware is useful hardware.

That's not true. I have an HP laptop and it does not boot windows uefi by default and it is not hardcoded.

Bold to assume just because yours doesn't have that problem none have or had. Maybe it's a matter of revision. Mine was a 2012 Envy m6 (I think 1108dx or something like that).

Yeah, I had HP laptop with UEFI and remember this issue pretty well. What I did as a workaround was renaming Windows loader efi file, copying GRUB efi file into the same place with previous name and setting proper Windows loader in grub.cfg file. Effect was pretty satisfying- hardcoded UEFI would still try to boot Windows loader but since it was replaced then it started GRUB which let me choose between Linux and Windows. Windows was fine with renamed loader and would boot without issues. The only downside of that workaround was the fact that Windows sometimes updated boot files and it would replace GRUB with its own loader so replacing would need to be done again. Farily easy to automate with some shell script.

Yep, I did the same but with EFIstub and creating the new entry for Windows with efibootmgr. It worked well enough, I had no issue pressing F9 to boot Windows as it was a secondary OS for me.

[brent]

I really hope we'll see a move to unified kernel images and systemd-boot soon. It's a breath of fresh air. Much simpler overall, you can easily realize a fully measured boot and you can get rid of the /boot partition among other things.

CloverBoot definitely seems like a good solution for legacy systems.