Linux Begins Removing Support For Russia's Baikal CPUs

[-]

bunnythistle@reddit

TL;DR - the code hasn't been maintained and the CPUs never went to market, so they're quite rare and not worth maintaining.

[-]

Linux has the interesting approach that they aren't afraid to touch legacy code that is working. Anybody that changes internal kernel APIs are responsible for fixing breakage anywhere in the kernel. But fixing stuff in the Baikai port may be impossible since the chips virtually don't exist as a product. This means it will eventually rot

If Linux instead didn't touch code that is working, this would mean they would need to keep old APIs this code works. That is, this essentially means Linux would need stable internal kernel APIs. Which is a sound approach (it's been the approach of Windows for years, except when they break things anyway) but one that's rejected by the Linux devs

[-]

Masztufa@reddit

Unmaintained code is also untested (picking maintenancr back up starts with testing), and every bit of code is wrong until proven otherwise

[-]

mclipsco@reddit

Just fork it.

[-]

Southern-Yam1372@reddit

But why not keep it in the mainline. As I understand it it’s an optional module during compilation, yes? No harm in keeping it.

[-]

torsten_dev@reddit

If you need it maintain it. If nobody maintains it, nobody needs it.

That's how open source works.

[-]

mclipsco@reddit

Also legacy code can introduce security risks. Look at what AI tools are already uncovering: 20+ year old bugs.

[-]

AugustusLego@reddit

Wrong.

[-]

Southern-Yam1372@reddit

No your wrong ☝️

[-]

AugustusLego@reddit

You're*

[-]

Southern-Yam1372@reddit

You’re* wrong bozo 😰

[-]

spin81@reddit

Yes even if it works.

[-]

Existing-Tough-6517@reddit

Almost no hardware exists as it never became a finished product. Nobody can verify if it continues to work. It's cancelled and if it's brought back it will be a new thing that shares the same name.

[-]

KittensInc@reddit

So what are you going to do when (not "if"!) it breaks?

[-]

cheesegoat@reddit

In small enough time units all code is unmaintained.

[-]

6SixTy@reddit

If that's the case, shouldn't we also be looking at ISAs like SuperH, Alpha, PA-RISC, SPARC, and OpenRISC to be removed on the same criteria?

A lot of these CPU instruction sets have stopped their broad availability about 20 years ago, and they are unlikely to be fast enough to run modern Linux anyhow.

Though obviously removing an ISA not the same as removing some platform drivers that no one used, the justification used could equally be applicable here.

[-]

Dr_Hexagon@reddit

SuperH, Alpha, PA-RISC, SPARC, and OpenRISC

OpenRISC is still being developed mostly for academic use.

Oracle and Fujitsu are still selling SPARC based servers to existing customers.

You can still buy SuperH chips for embedded systems

Alpha and PA-RISC are indeed dead, but if theres a maintainer and hardware available then I guess no harm in keeping them

[-]

EODdoUbleU@reddit

SPARC is still used, though. Development was only stopped in 2017, but there's been newer CPUs based on it since then.

[-]

Business_Reindeer910@reddit

if they have maintainers, and still have deployable hardware, then it doesn't count for this logic. It'd count more for the i486 removal logic.

[-]

adamkex@reddit

True but how can they be expected to maintain it if Russian kernel maintainers were removed from the project?

[-]

MatchingTurret@reddit

The company went bankrupt in 2023. Russian maintainers were removed in 2024.

[-]

Jean_Luc_Lesmouches@reddit

I miss the old convention of underlining links. Nowadays you can't know if it's two links in a row or a single one. You have to actually hover different parts with the mouse or hold on mobile to peak at the URL (assuming they're long enough to be cut and it's the same domain so you can't see the difference).

[-]

MatchingTurret@reddit

Better now?

[-]

xplosm@reddit

👌

[-]

wandering_melissa@reddit

is this about web version? on mobile app they are underlined and blue

[-]

cubic_thought@reddit

Use the Stylus extension (or other custom css extensions) for your browser and add this as a global style, or add it to UserChrome.css in your firefox profile.

a, a * {
   text-decoration: underline !important;
}

[-]

Jean_Luc_Lesmouches@reddit

Or maybe designers should designs stuff that are functional instead of looking pretty on a powerpoint.

[-]

doublah@reddit

And to be clear, Russians weren't blocked, only those employed by companies subject to sanctions.

[-]

adamkex@reddit

True, but even then the point still stands

[-]

Vittulima@reddit

Wasn't it just Russian maintainers employed by sanctioned companies etc?

[-]

AtomicPeng@reddit

Yes, but this goes against the narrative of the poor, suppressed Russian.

[-]

tnoy@reddit

You don't need to be in the maintainers file to submit patches and you don't need to have your patches in the mainline kernel to have Linux support.

[-]

calibrono@reddit

Tough shit lol

[-]

Kartonrealista@reddit

Good

[-]

SirCharlesTupperBt@reddit

You're no fun, but you're right. If they're dropping support for i486, Baikal has no business being in the kernel. The world of CPUs are a lot less diverse today than the old days when anything that had a logic gate in it was being proposed as potential architecture for Linux.

If you want to run an exotic CPU with little practical use, there's always NetBSD -- hell, they still have BeBox ports -- and even they don't support Bailkal.

[-]

tjorben123@reddit

"than the old days when anything that had a logic gate in it was being proposed as potential architecture for Linux."

my toaster would like to have a word with you.

[-]

xplosm@reddit

Does it run NetBSD?

[-]

captain_zavec@reddit

Dang, they're making language-capable toaster models now too?

[-]

Kichigai@reddit

What could go wrong?

[-]

Mordiken@reddit

Could be worse.

[-]

Booty_Bumping@reddit

The calculus of dropping Baikal is completely different from dropping i486, though. i486 has serious technical debt that doesn't overlap with the needs of any other platform. Whereas the Baikal CPUs were all fairly standard MIPS and ARM CPUs. If history had gone a little differently it would have been fine to keep (LoongArch in comparison is highly supported), but of course lack of physical hardware is a big problem.

[-]

halfc00kie@reddit

dead code for dead silicon, makes sense

[-]

dgm9704@reddit

noooo don’t spoil it! it’s much more fun to start the ”what about the U.S. and Israel” when you don’t have the facts

[-]

lightwhite@reddit

Fair enough.

[-]

chemerys22@reddit

Send russia back to the Middle Ages!

[-]

creeper1074@reddit

I've been messing around with Linux on my PS4 way too much lately... My sleep-deprived brain thought this meant that there was a special Russian version of the Baikal southbridge that wouldn't be supported anymore.

[-]

fellipec@reddit

TIL Russia has a CPU called Baikal and Linux supports it.

[-]

Preisschild@reddit

And just like everything else in russia it sucks and is decades behind other countries

[-]

MidnightSunIdk@reddit

There are also older ones by company Elbrus, based on a VLIW - Very Long Instruction Word architecture... which isnt very effective

[-]

rz2k@reddit

[ Removed by Reddit ]

[-]

fellipec@reddit

The only thing I know about Russia and computers is that once they tried a trinary computer.

[-]

6SixTy@reddit

Soviet Union copied a lot of early western 8 bit microprocessors, but kind of stalled with the 8088.

[-]

Kichigai@reddit

Well, at least as far as consumer tech. Lord knows what kind of industrial monsters they had lurking in the dinosaur pens at Roscosmos.

[-]

Juma7C9@reddit

Well, it fell not long after, so it tracks.

But I guess when you're continuously trying to catch up and have little room to try out something new, it becomes extremely difficult to innovate.

[-]

Admirable-Safety1213@reddit

Mostly Ineffective taking the new Damage Dialogs into account

[-]

PraetorRU@reddit

The story is that company that produces Elbrus used to be a lab behind Sparcs and later Itaniums. Intel was very active in Russia until 2022, and significant drop in their linux drivers quality in the last few years is a result of breaking the ties. Right now most of the people who worked for Intel in Russia are working for Huawei on their new OS, hardware and software compatibility.

The main problem still is that sanctions by USA on lithography machines were never lifted since USSR times, and Russia can design and write software, but up to this day can't produce modern chips domestically.

[-]

MatchingTurret@reddit

ELBRUS is wider known and in actual use.

[-]

No_Article4254@reddit

They used liquid cooling (vodka)

[-]

spectralblade352@reddit

Same

[-]

theclosedeye@reddit

Минусы?

[-]

FlukyS@reddit

For those who don't follow the exciting world of CPU manufacturers the company that made them is bankrupt but they spun it out into another company who then also went bankrupt a few years back. They are kind of moving into RISCV apparently under a new company named the same.

[-]

regeya@reddit

Which is the smart move IMHO. RISC-V is the first time I've been hopeful about a post-x86 world in a while. I've been worried that we only had phones crammed into laptop clamshells to look forward to.

[-]

frankster@reddit

Out of interest what do you see as the advantage of risc-v over arm as an x86 replacement?

[-]

Possibly-Functional@reddit

License. ARM is tightly licensed. RISC-V is open source.

[-]

d32dasd@reddit

which means nothing in this case, as RISC-V is permissively license instead of copyleft. The companies have already learnt to close down everything by not repeating the standardization of BIOS in x86. Each RISC-V board will be a black box with its own kernel device tree that ties you to whatever Linux kernel fork the manufacturer has pooped for that board only.

[-]

Dr_Hexagon@reddit

Just like ARM.

ARM is promoting a UEFI standard for anything thats server, laptop or desktop aimed, and it is getting support.

https://www.arm.com/architecture/system-architectures/systemready-compliance-program

[-]

Possibly-Functional@reddit

It does help significantly for the hardware market, especially diversity and competition. As you say though, it doesn't help with open software execution which is a shame.

[-]

d32dasd@reddit

I remember when the run was between OpenRISC (copyleft) and RISC-V (permissive). Even in reddit, there was astroturfing for RISC-V. We never learn..

[-]

monocasa@reddit

OpenRISC was also just not as good of a design. Had a lot of the same baggage as MIPS that got in the way of both the very low end designs and the high end ones.

[-]

Natural_Night9957@reddit

Now it's Rust's time

[-]

regeya@reddit

Lack of licensing fees for one, and I guess the rest is wishful thinking. It seems like so far, RISC-V is more enthusiast-friendly whereas ARM devices tend to be black boxes. My hope is that open standards grow around RISC-V.

Like...as sexy as new Apples are, I do not like the notion that to get the performance they have, it needs to be a SoC. My first PC was in 1987 and every computer I've had since then, aside from a Raspberry Pi, has been customizable to some extent.

[-]

Indolent_Bard@reddit

For battery life and performance, soc is objectively better, so unless you can solder it won't be customizable. The strix halo chip is a good example. Framework had AMD's engineers try to make it modular and the ram only worked at like half the speed.

[-]

FlukyS@reddit

Well kind of, if x86 was pure x86 nowadays it would be true but both AMD and Intel cheat a good bit. Like E cores in Intel chips are basically just RISC cores that are stripped back a bit, there are quite a lot of x86 unused instructions that aren't implemented in hardware anymore and just emulated on chip for compatibility purposes but given they aren't used in 99.9% of cases no one notices. Also where RISCV fails slightly is it is missing features that actually do improve performance, they will come eventually into the ISA eventually for sure but for the time being the only way to compete for a RISCV chip maker would be just if you implemented it as extensions instead. So the answer here is it is very complicated.

[-]

Albos_Mum@reddit

That risc thing is pretty old as well, iirc the AMD K5 was not very far off being one of their 29k RISC chips with an x86 instruction decoder and 486 compatible FPU.

[-]

Novero95@reddit

Unfortunately, the general consumer has no idea what an SoC is, if RAM is upgradable, and most likely, doesn't even care. It sucks that technology is moving towards closed ecosystems and people is happy about it

[-]

FlukyS@reddit

I can answer this, ISAs are just ISAs the chip designs and how they are handled is the important part. ARM and RISCV have advantages over x86 because they are super tight in comparison. RISCV has an advantage over ARM in that it is an open ISA so anyone who is interested in a processor can make one that is compatible for free. From a desktop or server standpoint all ISAs in theory have the same capability ish but efficiency is the difference. RISCV the most excited people for it are ones who for instance make hard drives or embedded devices or chips like wifi or bluetooth modules...etc because before either you are running some custom thing or you had portions of it licensed but RISCV you can spin up basically a custom processor for that specific workflow. For desktop there needs to be quite a lot of work to get it ready and I don't see that happening any time soon really but I'd be happy if I'm surprised by some manufacturer really taking it seriously.

[-]

Kevin_Kofler@reddit

Unfortunately, the dystopian trend towards only locked-down Android/iOS-like setups is going to be enforced not by lack of open alternatives like RISC-V, but by governments (through their ID / "age verification" apps) and banks (through their banking apps) enforcing use of "secure" (as in "cryptographically remote-attestated to be unmodified code from Google or Apple on unmodified certified hardware") devices and operating systems.

The dystopias painted in some of RMS's essays, that RMS has been warning us from all this time, are unfortunately coming closer and closer. For smartphones, we are basically already there (and any smartphone that does not comply, such as my Librem 5, is excluded from large parts of the app ecosystem). Now they are coming for computers.

[-]

JaguarOrdinary1570@reddit

I'm really conflicted about it because a method for cryptographically verifying identity is ultimately going to be necessary for some things. Basically everyone's personal info has been made public through endless data breaches. Agentic AI systems are flooding the internet scraping every piece of data they can. Some time in the next few years (if not sooner), they'll have the data they need to look like you, sound like you, and know all of these "private" details about you. Essentially, perform impersonation at scale in a way that'll be really hard to deal with. And a lot of people will be victims of that, and there's basically no other solution.

It's just that it needs to be optional, not mandatory. I'm happy to have a zero-privacy locked down centrally governed and controlled authentication machine for certain things, as long as I can also have a machine that is truly mine and under my control, and that there's an "unauthenticated" part of the internet I can still participate in. But I don't think many companies/organizations/people who host servers would be willing to participate in that internet when an "authenticated" one exists. I don't know that I would if I were in their shoes- why subject myself to roving LLMs and botnets like that? It's why I kind of understand where Reddit is coming from when they talk about authentication. Who wants to be on a discussion forum just to interact with LLMs? But it still sucks. Even without the government stepping in like this, the existence of such a technology seems like it will eventually choke out free and open systems, just because all of the other players on the internet will refuse to interact with them.

[-]

xorgol@reddit

verifying identity

The only thing for which it is actually useful is accessing state-provided services, but that we've had for years.

[-]

primalbluewolf@reddit

a method for cryptographically verifying identity is ultimately going to be necessary for some things.

It is necessary that it doesn't ever exist.

[-]

FlyingBishop@reddit

The government just needs to provide cryptographically secure ID chips. They do it in Estonia. The current thing is outsourcing ID to Microsoft, Google, and Apple and the government doesn't actually provide ID systems.

[-]

pfp-disciple@reddit

I just want to say thank you for the accurate description of "secure" in this context.

[-]

Manic5PA@reddit

The final frontier has been domesticated and now it's just another playground for rent seekers.

[-]

plitk@reddit

Sun Microsystems called about their sparc and would like a word with you.

[-]

regeya@reddit

Yeah...that's a bummer, isn't it?

[-]

gnatinator@reddit

It's definitely not in free softwares' best interest to contribute to hardware which will most likely end up in Russian military hardware used against societies that maintain the kernel in the first place (ex: Linus is Finnish)- good riddance to dictatorship hardware.

[-]

lib20@reddit

And you think that your country is better?!

[-]

xorgol@reddit

Better than Russia's government? Sure.

[-]

Business_Reindeer910@reddit

please read the article.. that has nothing to do with anything here.

[-]

sothisismyalt1@reddit

Not the reason of ending support though, they're discontinued and code has Benn unmaintained for some time.

[-]

BodybuilderLong7849@reddit

But u can always fork the main repo and create a russian version of linux. That's what the chinese devs do right?

[-]

Business_Reindeer910@reddit

no, the chinese merge their CPU support right into linux like everybody else

[-]

dgm9704@reddit

… and the comments are full of people who didn’t read the article but somehow have an opinion about it’s contents 😁

[-]

EmperorOfAllCats@reddit

Click the links? In this economy?

[-]

ilep@reddit

Well, it links to phoronix so you are right about that.. You would get a bunch adverts or even worse - see phoronix itself.. /s

[-]

Wheeljack26@reddit

Read? What are we? Non jjk fans?

[-]

Maleficent_Celery_55@reddit

no its not?

[-]

dgm9704@reddit

Yes, they are. The comments of the article.

[-]

Maleficent_Celery_55@reddit

ah i thought you were referring to the comments here.

[-]

dgm9704@reddit

reddit comments sometimes get heated and/or sidetracked but phoronix comments are even worse ;)

[-]

earthman34@reddit

Good.

[-]

okktoplol@reddit

Am I the only one who didn't know Russia had domestic consumer processors?

[-]

6SixTy@reddit

Baikal isn't designed to be a domestic consumer processor. They received a bunch of funding from the Russian government and the Russian state owned military-industrial complex.

Their only "real" domestic processor are the MCST developed Elbrus processors. These use a bespoke VLIW ISA that I gather are supposed to offer an ISA level operating system isolation and translate ISA instructions like the Transmeta Crusoe.

[-]

mrquantumofficial@reddit

They aren't consumer processors

[-]

Electronic-BioRobot@reddit

And are basically made in China

[-]

erraticnods@reddit

russia does have local chip fabs (mainly in Zelenograd), but the technological processes aren't quite up to date. the chips they produce are of no interest to consumers, and of fairly limited interest to data centers and the military

decent enough for embedded hardware, though. you're not gonna care about the milliseconds of difference when a cash register communicates with the tax service's api

[-]

FarReachingConsense@reddit

you're not gonna care about the milliseconds of difference when a cash register communicates with the tax service's api

Highly likely that they end up flying into some Ukrainian poor souls livingroom abord a Kinzhal instead of something that would happen in a normal country though. I wish nothing good on russia

[-]

avg_php_dev@reddit

Oh comeone, following your logic, some intel or amd chips may end up in some poor iranian soul bedroom. Very normal.

[-]

FarReachingConsense@reddit

Sure, they do, and that's also shit.

[-]

DoubleOwl7777@reddit

or a cruise missile control system steers into the target...

[-]

Electronic-BioRobot@reddit

Yeah, I don’t doubt it.

It is just that the Baikal CPU was just designed in Russia and was produced in China, but in the end it got scrapped.

[-]

evmt@reddit

They were manufactured by TSMC in Taiwan.

[-]

Electronic-BioRobot@reddit

Oh, might be that I mixed up some information, but anyway still outside of Russia though

[-]

okktoplol@reddit

I mean that as "not made for a very specific military application"

[-]

MatchingTurret@reddit

List of Russian microprocessors

[-]

PraetorRU@reddit

Still has. But they're mostly for datacenters with broad range of applications, and military usage, not for consumers so far.

[-]

Wyciorek@reddit

They don't. What they had was yet another graft machine designed to suck up tax money without ever creating a viable product.

[-]

0riginal-Syn@reddit

I only knew of this thing. Through my friends from my old hardware engineering days, who are still in that arena. Not something most know much about. They originally started it using the MIPS architecture, not ARM, but then moved to licensing ARM and going in that direction. My friends initial interest in it was because of the MIPS architecture they started off with. He lost interest when they moved to ARM. But once they got hit with sanctions for the Ukraine war, TSMC froze all the shipments of what they needed and thus the bankruptcy of the parent company.

[-]