Numerous Linux/X11 Display Drivers Can No Longer Even Properly Build

[oiaohm]

The kernel has been too large to fit on a floppy disc for years now,

That is in fact a false statement.

https://bits.p1x.in/floppinux-an-embedded-linux-on-a-single-floppy/

A modern Linux kernel optimised for old systems will and does fit on a floppy. There are civil infrastructure thing still running on 486dx chips that have network connection that cannot be running with a old known flawed Linux kernel. Yes a majority of you Linux distributions have kernels optimised for the most hardware support this end up with a kernel bigger than a floppy. Yes this include the initramfs having a stack of firmware files old systems will not need.

August 24, 1995 with the release of Windows 95 OEM that came on CD that was to be installed from cd means systems after that data normally have a cdrom drive of course lot of those system in 1995 that came with Windows 95 OEM did not support booting from cdrom directly instead you had to use a boot floppy. Yes instructions how to put grub on a floppy disc still exist.

Plus, even the "lightweight" distros designed for older hardware mostly require 512 MB or 1 GB of memory as a minimum

http://tinycorelinux.net/

No you were not looking at the true lightweight distributions designed for the 486 time frame hardware that is still maintained. Tinycore Linux limit graphical only 12 megs of ram required.

running a modern distro on 20 year old hardware is not really a viable option, and there's no good reason to attempt it.

This line is basically not true. For many reasons. There are good reasons to attempt running modern distribution like Tinycore on a very old system as in a 486 from 1989 if you work in civil infrastructure. 486 from 1989 only 33 year old hardware and civil infrastructure and the replacement cycle can be 50 years so it has 17 more years of service to still do. Remember this old system could be in control of a bridge/power.... items that if hacked could kill people and also scary enough can be internet connected. The logic of civil infrastructure is brain bending at times due to running these absolute old systems in charge of life endangering things that are internet connected that starts you thinking are they trying the make the most dangerous thing ever. Yes if these system fail by either hardware or being hacked people can die and you don't have the budget to replace the hardware so the best you can do is update the software.

Of course the civil infrastructure use case does not care that you functionality is limited like not being able to use a modern day feature rich web browser or office suite. Yes you can still use modern day web browsers like dillo and so on quite few of these ultra lightweight web browsers the reasons why they get updates is the civil infrastructure usage..

I would say running a modern distribution on 20+ year old hardware you have to serous limit you expectations and you most likely only do it like in the civil infrastructure case. Yes that internet connection is normally need to the companies chat program for the management to talk to their staff like for staff checking in that they are at work and the like. Yes many of those control rooms don't have room for another PC and due to what they are made from mobile phones mostly don't work. I would call the civil infrastructure case as equal to someone holding a gun to your head where people are absolutely dead if you do nothing and people still might be dead if you do something. Yes so true no win location here.

Basically just because you can do something does not mean you will enjoy doing it. Just because something is horrible on end users does not mean its not due to over a 1 million users every day of the week. There is over 1 million 486dx clones(this include fpga replacements) and originals still in use in Clivil infrastructure running modern Linux kernels.

Yes some of where I know 486dx are still in use I wish I did not know due to how much trouble its going to cause when the powersupply motherboard, cpu or ram finally dies and these are systems that have been running almost 24/7 for the past 33 years.

[NateHubbard]

People should be able to run modern Linux on 30 years old computers if they choose.

Unless your name is Linus, I don't think you can make that statement.
Anyway, this doesn't mean you can't run Linux necessarily, but you won't be able to run X.

[torsionbar28]

This line is basically not true. For many reasons. There are good reasons to attempt running modern distribution like Tinycore on a very old system as in a 486 from 1989 if you work in civil infrastructure. 486 from 1989 only 33 year old hardware and civil infrastructure and the replacement cycle can be 50 years so it has 17 more years of service to still do. Remember this old system could be in control of a bridge/power.... items that if hacked could kill people and also scary enough can be internet connected. The logic of civil infrastructure is brain bending at times due to running these absolute old systems in charge of life endangering things that are internet connected that starts you thinking are they trying the make the most dangerous thing ever. Yes if these system fail by either hardware or being hacked people can die and you don't have the budget to replace the hardware so the best you can do is update the software.

In which country is this? I have yet to see a case where a modern Linux kernel is running on ancient hardware in civil infrastructure. I have worked quite a lot with civil infrastructure and embedded devices. All of it was running various RTOS like QNX, VxWorks, OS-9, INtime, etc. None of it ran on Linux. And the larger stuff like power generation (hydro, nuclear) pretty much all was Tandem NonStop NSK. Very little of the civil stuff was x86 based, and none of it was running a modern Linux kernel on obsolete hardware. This is my experience from decades in the industry in the US, so I'm guessing you're located elsewhere.

I do appreciate you sharing TinyCore, I had not heard of that distro before so I will check it out.

[JanC]

As much as I'm for supporting old hardware and fighting planned obsolescence, it's important to note VESA drivers should still support these cards if all you care is having a display. What these drivers are about is exploiting the full capabilities of cards, i.e. 3D and sometimes 2D acceleration. But display you'll have forever thanks to VESA drivers.

Unfortunately, some systems with SiS graphics come with a broken VESA BIOS Extension that crashes the system when you try to set the video mode using the VBE API, so that's not always a good option… (And I wouldn’t be surprised if some other hardware either has no VBE or has a similarly broken one.)

[oiaohm]

In which country is this? I have yet to see a case where a modern Linux kernel is running on ancient hardware in civil infrastructure. I have worked quite a lot with civil infrastructure and embedded devices. All of it was running various RTOS like QNX, VxWorks, OS-9, INtime, etc. None of it ran on Linux. And the larger stuff like power generation (hydro, nuclear) pretty much all was Tandem NonStop NSK. Very little of the civil stuff was x86 based, and none of it was running a modern Linux kernel on obsolete hardware. This is my experience from decades in the industry in the US, so I'm guessing you're located elsewhere.

Great fun here is everything you just listed in OS is not certified for 486 based systems.
http://www.elcus.ru/english/e_boards.php?ID=zfx86-pc104 These are in fact after 2006 not that far in the past. That is a Cyrix 486DX at 100 mhz with 32 megs of ram in a soc chip. Yes on built on many different industrial board.

The reality is the only up to date OS kernel certified to go on to most of this 486dx horrible hardware is a 32 bit Linux kernel.

Yes I am in Australia but its not just Australia with 486s or 486 clones or ao486(a fpga implementation) floating around in civil infrastructure.

I would guess you would be working in places where the hardware you are using hardware that fall inside "Civil Infrastructure Platform." This is where your x86 are atoms and xeons that support 64 bit and all you 32 bit be arm based.

So you would not be coming across ZFx86 that are 486s or Vortex86DX that are i586 or i686. Some the Vortex86DX are not support by the Linux kernel because they don't have a FPU and yes they are a newer chip than the ZFx86. Yes Vortex86DX were still be deployed new in 2010.

Yes why these later clones have PC/104 support is the originals using Intel 486dx like https://resources.winsystems.com/dat...sat-dx4-ds.pdf is using PC/104. Yes PC/104 is from 1992 and they were making replacement parts for PC/104 in 2010 with the Vortex86DX stuff.

Tandem NonStop hardware means you were not dealing with mass volumes of PC/104 hardware. Just to be horrible yes Non-x86 PC/104 CPU board do exist but most software from the PC/104 eco system is 32 bit x86 to be exact large sections of 32 bit 486 machine code. Yes you do find early prototype of the PC/104 in use as in before the standard from 1989. Plugging a a PC/104 prototype into a PC/104 standard board nicely causes hardware virus due to pin alignments being different. So when you find a 1989 version version of this stuff you absolutely want to replace it straight away before it causes more problems.

torsionbar28 basically be happy you are not dealing with the PC/104 headache.

[Spring]

Is it correct that those are drivers which use non-standard APIs? Does modern hardware require such drivers?

[oiaohm]

Is it correct that those are drivers which use non-standard APIs? Does modern hardware require such drivers?

Non standard APIs is not exactly right. Legacy/Deprecated APIs so modern hardware with properly maintained drivers(that most of it) does not require these drivers. These legacy drivers that don't build are UMS (user mode setting) not the modern KMS (kernel mode setting) drivers. Items have been removed from the X11 server and items have been removed from most kernels. Like old UMS drivers are known to use /dev/mem to direct access system memory ramdac bit that gone is linked to this. Yes some things were Deprecated so removed because they were insecure.

Most of this is the drivers have not been build in long time and the drivers have not had person doing proper maintenance on them either. One of the failures to build is was such a simple fix a value got renamed from bool to boolean now compiler complains that does not match. So lack of continuous integration testing and lack of remaining aligned with upstream. Yes some of the fixes to fix the problems found by this build run had been sitting on the mailing list for over 5 years as a merge request.

So lack of personal, Lack of funding and Lack of Care leading to bitrot of the old drivers.

[torsionbar28]

torsionbar28 basically be happy you are not dealing with the PC/104 headache.

Ha, good to know, yes I remember ads for PC-104 boards in Linux Journal Magazine in the late 90's, but I've never worked with that hardware.

[oiaohm]

Ha, good to know, yes I remember ads for PC-104 boards in Linux Journal Magazine in the late 90's, but I've never worked with that hardware.

Please note fpga 486 PC-104 is not the most cursed PC-104 board you can get new.

https://www.iwavesystems.com/wp-content/uploads/2020/11/iW-80186xl-daughter-board-R1.0.pdf

Yes a 16 bit x86 CPU done in fpga and they are made new by many different vendors.
PC-104 when you have to deal with systems like that you have 16 bit x86, 32 bit x86 and 64 bit x86.

Working with PC-104 comes horrible example of how the saying "if it ain't broke, don't fix it" can go horrible wrong. Yes it takes a real turn for the worse with the define if it breaks replace it with exactly like what was their before even if it means using fpga to emulate the old processor because we could not rewrite the software.

There are cursed platforms in civil infrastructure where most of the rules of what people would class as common sense is basically non existant. PC-104 is one of the really cursed ones. There is a horrible amount of PC-104 stuff still in use that either really old or has been repaired with fpga parts to emulate really old.

Yes systems deployed in the late 90's using PC-104 are still in use in one form or another yes parts have been replaced. The question always is how much is modern fpga and how much is really old legacy boards. There are a lot PC-104 systems are real world "The Ship of Theseus" where every part has been replaced with modern fpga and like yet they are still 486dx systems or 80186 system with the cpu now in fpga instead of asic. Of course in that mess of PC-104 systems there are still systems using original boards from the early 90s.

I am not exactly sure its good to know about the PC-104 stuff that still around as it more a pure nightmare of having to deal with very old cpu designs with very limited amounts of ram yes 32 to 64 megs of memory is still common with the new 486 done in fpga PC-104 boards.

torsionbar28 not working with PC-104 boards you have fairly much dodged a bullet. There are some people in the USA who also have the PC-104 nightmare.

[sinepgib]

Working with PC-104 comes horrible example of how the saying "if it ain't broke, don't fix it" can go horrible wrong. Yes it takes a real turn for the worse with the define if it breaks replace it with exactly like what was their before even if it means using fpga to emulate the old processor because we could not rewrite the software.

I'll take this as my cue to also vent my disgraces. At my first job I had to work with a bare metal (which actually did make things easier, I admit) 80C186 emulated by an FPGA. The fun thing is they made that rather than use a more modern ARM with Linux because they thought it would be easier for us programmers because it meant modifying fewer code, as we used that same model but properly implemented in hardware by commodity factories, all the while we had a portable Linux version on x86 for about a decade already. They made an ARM model a few years later and migration took a single line to convert an unaligned read (at the time ARM required all access to be aligned, I don't know if that's still true) to aligned.

So you would not be coming across ZFx86 that are 486s or Vortex86DX that are i586 or i686. Some the Vortex86DX are not support by the Linux kernel because they don't have a FPU and yes they are a newer chip than the ZFx86.

Why does the Linux kernel require an FPU for x86 if it doesn't actually use it itself? Is it a matter of how it initializes it? Surely that should be easy to fix?

Unfortunately, some systems with SiS graphics come with a broken VESA BIOS Extension that crashes the system when you try to set the video mode using the VBE API, so that's not always a good option… (And I wouldn’t be surprised if some other hardware either has no VBE or has a similarly broken one.)

Hmmmm, that's sad. I guess a basic framebuffer driver could be enough for those tho. Not ideal because of maintenance, but better than nothing and it doesn't need harder to maintain accelerated drivers on the X side.