appiah4 wrote on Yesterday, 11:22:

Jo22 wrote on 2025-05-08, 20:40:

If Linux becomes an OS that targets modern hardware all the time, then what's the purpose of it? 😟

I'm trying not to start sounding like that "Um Ahktually" linux nerd, but this question tells me you don't really 'get' linux..

Hey, that's the kind of Linux I grew up with! 🙂
Linux used to be the savior for old and forgotten hardware.
Hardware being abandoned by society. It used to be the unix for the poor.
It was this aspect of Linux that I respected.

Since Linux Foundation has no proper roadline for abandoning architectures,
it's to no surprise if someone like me has to wonder where it eventually leads to.
Now the i486 architecture is being dropped, what is going to be dropped next year(s)?

And how does this removal affect, say, external modules that can be loaded?
Are kernel entry points being removed for classic "drivers" that could be loaded via modprobe?

Such changes have the bad habbit of affecting parts of the ecosystem that are not being taken into account first.

Also, is the whole thing related to Rust?
Are parts of the kernel being "thrown out" to make transitioning easier?
It's unlikely, but who knows? I don’t see a clear roadmap here.

To me, the tower PC in the sample picture used to be a typical Linux era PC,
noticeable by the AT case and "good" network card with BNC connector.
Sample picture is borrowed from Wikipedia, for license reaons.

Edit: The one thing that wasn't "cheap" or period-correct about Linux used to be memory consumption.
Linux required three or four time the RAM expansion of a DOS, Windows or OS/2 installation.
It probably gave birth to the first hot-rod PCs among private users.
Say, a i486 with 64 MB of RAM or a Pentium 75 with 256MB of RAM.
Unusual configurations for desktop OSes of the time, but not exactly weird for server systems.

Edit: I'm not making things up, btw. The reputation of Linux used to be like that where I live.
Here for example, is an quote of an older news article, titled "rescue from the junkyard".

Linux is actually the only solution for users who have older computers and don't want to scrap them or upgrade them technically - unless they stop using their device to access the Internet in the future and only use it as a better typewriter.

[Machine translation]
Source: https://www.fr.de/kultur/rettung-schrottplatz-11228248.html

Jo22 wrote on 2025-05-08, 20:40:

Maybe. But it doesn't solve the (my) "problem".
Namely, that current Linux was useful to keep old hardware running.
On the internet, not just isolated on an old PC in the corner in the cellar.

Linux being open source the solution to your "problem" is running

1git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git && checkout -b preserving-486

and preparing to backport future changes.

This will cover your kernel needs, as for userland you most probably need to patch other stuff at some point too.

This is a lot of effort but I am sure fellow vogon members would appreciate this and surely offer help testing /porting!

Cheers,
Flow

Thank you! ^^ It's definitely helpful!

Linux being open source the solution to your "problem" is running

Hi, I meant it more in the way of "my issue with it is: [..]" 😅

Jo22 wrote on Yesterday, 15:33:

And how does this removal affect, say, external modules that can be loaded?
Are kernel entry points being removed for classic "drivers" that could be loaded via modprobe?

It doesn't and no. Linux does not have a stable kernel ABI so any kernel modules need to be built for the running kernel anyway.

Jo22 wrote on Yesterday, 15:33:

Also, is the whole thing related to Rust?
Are parts of the kernel being "thrown out" to make transitioning easier?
It's unlikely, but who knows? I don’t see a clear roadmap here.

No need to speculate, this is the original e-mail where a CI runner tripped over a corner case with a kernel not configured with TSC but TSC is available in the (emulated) hardware. Linus suggested not to "waste one second of development effort on this kind of issue" so someone else sent a patch series to make RDTSC and CMPXCHG8B mandatory.

Linux simply supports hardware that people show up to write support for and maintain. The m68k architecture is still being maintained even though it's older than the i386 (also helped by the fact that it's a dead architecture, unlike x86 which is still under active development). OTOH, support for Itanium (IA64) was removed in Linux 6.7 because no-one wanted to maintain it any more.

Also consider that based on this thread, kernel 6.7-rc1 (released November 2023) and later are currently not bootable on an early i486 without the CPUID instruction. That was only reported in October 2024, 11 months later. The fix was only committed a few days ago to the x86/urgent branch, might be another week or two before it lands in mainline, and then another week after that before it's backported to supported releases, which means 6.12 and 6.14.

jtchip wrote on Yesterday, 22:17:

Linux simply supports hardware that people show up to write support for and maintain. The m68k architecture is still being maintained even though it's older than the i386 (also helped by the fact that it's a dead architecture, unlike x86 which is still under active development). OTOH, support for Itanium (IA64) was removed in Linux 6.7 because no-one wanted to maintain it any more.

It's kind of funny how that happens. Because 386/486 support was intermixed with more modern x86 support through #ifdefs and conditionals, it's "in the way" of the maintenance and testing of that code, while if it were cordoned off elsewhere (like m68k) the maintainers would be able to just leave it alone.