Linus Torvalds is "fed up" with making kernel changes to address faulty hardware

Cal Jeffrey

Cal Jeffrey

Posts: 4,317   +1,499
Staff member
A hot potato: Linus Torvalds is sick and tired of trying to "fix" Linux to work on Intel's faulty processors. In his view, Linux isn't broken, so why should he (and other FOSS developers) fix anything? Intel is the one with the problems. He vented his ire in a thread recently.

Linux Creator Linus Torvalds recently expressed frustration with change proposals to the Linux kernel that are only "required" due to faulty hardware. Phoronix notes that Torvalds ranted on the Linux kernel mailing list over the last week, saying that buggy hardware should not be Linux's problem, suggesting kernel developers just say "no" and see what the "hardware people" say.

"I think this time we push back on the hardware people and tell them it's *THEIR* damn problem, and if they can't even be bothered to say yay-or-nay, we just sit tight," Torvalds said. "Because dammit, let's put the onus on where the blame lies, and not just take any random shit from bad hardware and say 'oh, but it *might* be a problem' [sic]."

He's also unhappy about kernel commits that address theoretical vulnerabilities, especially those that are difficult to implement and have never been seen in the wild.

"Honestly, I'm pretty damn fed up with buggy hardware and completely theoretical attacks that have never actually shown themselves to be used in practice," he said.

Part of the problem is that addressing hardware faults at the OS level complicates matters for developers and creates potential compatibility problems with other fault-free hardware. An example would be proposed kernel changes that will not work with Linear Address Masking (LAM) on Arrow Lake and Lunar Lake processors, which is what got Torvalds fired up in the first place. This feature allows user space to store metadata in some bits of pointers.

According to Intel Engineer Kirill Shitemov, Intel is replacing LAM with LASS (Linear Address Space Separation), which is more secure as it prevents hackers from accessing virtual address spaces across user/kernel mode. Shitemov indicated that there was supposed to be a patch that disabled LAM until LASS was ready for implementation, but the fix "never got applied for some reason."

Since earlier this year, Intel's 13th and 14th-gen issues have caused headaches on many levels with consumers, developers, and within Intel. The company has addressed the problems with recommended BIOS settings and microcode to varying degrees of success. Couple that with the long-running issue of Spectre/Meltdown vulnerabilities, which keep making headlines even six years after their discovery.

The problem, as Torvalds sees it, is that Intel can afford to pay its engineers to address its hardware issues. Meanwhile, FOSS developers are unpaid volunteers who are under increasing pressure to get Linux working on bad hardware while keeping it bug-free on good hardware. So, his frustration is understandable. Of course, Torvalds has never shied away from telling hardware OEMs where they can go.

Permalink to story:

Linus Torvalds is "fed up" with making kernel changes to address faulty hardware

 
You can't have security and performance-they're two different boats, and you can't be on both at the same time. The architecture of the operating system should prioritize security, but after that, if a new security fix is needed and affects performance, it should be optional, no matter how serious. Hardware security holes are not intentional (Intel is putting effort into Crystal Linux), they just happen in the chaotic environment.

I run w10 from 2018 (that's the date on the windows folder, version 1809 os build 17763.1) on the last codebase with the influence of the veteran programmers and before the spectre meltdown mitigations with almost no updates (control panel says 1 security update and 3 normal) and I never had any problems. The OS is very light like the win xp or w7 was and very responsive. I won't update anything even in 10 years, the only problem is they don't support directx agile (it's a shame they didn't make it available for this version of w10) but I can live without it.
 
I am sorry, but when you have different versions of OSes (Linus or Windows) gimping hardware performances, then I am having the opposite point of view.

Not to mention that MANY of SMEs in cybersecurity are calling the search for hardware flaws as a joke. When you need to have access to the metal to exploit hardware flaws, then it is just a way for the firm to get a 5 minutes of fame in the limelight. Not to mention that the naming of those hardware flaws are chosen to make them sounds like it is the end of the company making those chips.
 
  • Like
Reactions: BadThad, fb020997, arrowflash and 1 other person
NotYourITguy said:
The reason Apple ditched Intel is for money. Why pay someone else if you have hundred of billions sitting around in cash and can make even more? Bugs or not, Apple was going to leave Intel anyways.
Click to expand...
Apple buys lots of parts. It makes sense because as big as Apple is, the total market is larger - much larger. When major participants aren't screwing up, their larger cross-customer volume should enable higher quality at lower prices with less risk vs. needlessly trying to roll your own.

Of course, that's all out the window when the supplier who should be in the better position isn't interested in pursuing your top priorities and/or blows it with too many mistakes.
 
antiproduct said:
Isn't this pretty much why Apple ditched Intel?
Click to expand...
Nah. The reasons Apple ditched Intel are simple:
1. going with their own chips allows Apple to cut out the middle man and increase profit margins
2. Apple already had experience designing the A series chips, and is so wealthy, that it was low risk to do so
3. It allows Apple to exert far more control over their product, along with proprietary APIs like Metal
4. Intel's progress on making power efficiency and performance gains had totally stalled out 4+ years ago when Apple went to the M series. Lunar lake may finally break this trend, but Apple learned from the G5 powerPC fiasco.
 
  • Like
Reactions: ferrellsl, brunocasarini and toooooot
Bobbydpue said:
Linus should try making hardware before getting this annoyed. If making CPU's was easy there would be a lot more choice than we have now.
Click to expand...
You dont need to be a chef to criticize bad food, nor an engineer to criticize bad engineering.

I could say the same about you, that @Bobbydpue needs to write his own Unix kernel before complaining about Linus. I bet you'd think that was unreasonable.
 
  • Like
Reactions: ferrellsl, astralcyborg, axiomatic13 and 1 other person
Bobbydpue said:
Linus should try making hardware before getting this annoyed. If making CPU's was easy there would be a lot more choice than we have now.
Click to expand...
How I see it is: Linus is more annoyed by the fact that intel and amd do not own their mistakes and make their own patches. Otherwise, errors are unavoidable in software and in hardware. The problem is that most of the hardware errors are addressed in software. Hardware companies should make the patches and do it in a timely fashion. Adding to the frustration is the fact that most of the recent hardware exploits are theoretical, but they (the linux maintainers) still have to address them in order to inspire confidence.
 
  • Like
Reactions: tonyd
antiproduct said:
Isn't this pretty much why Apple ditched Intel?
Click to expand...
Nah, that'd be heat.

Apples endless goal of making devices stupidly thin was not a good match for Intels chip that run mighty hot. The almost instant throttling was disastrous for performance (and noise).

If you can't handle the heat, get out of the kitchen Intel contract.
 
  • Like
Reactions: scavengerspc
If Microsoft and Linux developers could get together and agree to stop fixing trash hardware
we would see a quick and rapid rise in hardware quality and reliability. They would have no choice but to actually work for a living..
 
Intel stuff has gotten bad.... right around the time thunderbolt came out it got a lot worse. Thunderbolt is glitchy as well and I've had so many issues with Intel based machines and docking stations. Not to mention all the CPU instability issues they've been having.
 
Whether or not he has a valid point, he is almost comically cranky. Apparently he's not happy unless he is writing or talking about being mad about something.
 
NotYourITguy said:
The reason Apple ditched Intel is for money. Why pay someone else if you have hundred of billions sitting around in cash and can make even more? Bugs or not, Apple was going to leave Intel anyways.
Click to expand...

The heat problem was much of the reason for the switch, intel chips became too hot for Apple's design parameters of thin and light.
 
Struggling to muster sympathy Linus, as you throw your unpaid engineers under the Intel bus as you look out the $150 million net worth mountain view of the Linux valley below. Cry me a river!! You intentionally designed Linux 33 years ago to be a largely free product, so unpaid developers had to be obvious even then and you made a considerable fortune riding the success of Linux being open source. Not a new idea as source code was freely shared in the early days of PCs, through magazines, BBSs, pre-AOL service like Prodigy and CompuServ. Looks as you could also pay your engineers!
 
You must log in or register to reply here.

Similar threads

S
Linus Torvalds worries RISC-V will repeat the same mistakes made with earlier architectures
Replies
8
Views
308
nintendo1889
N
A
Ubuntu will use the latest version of the Linux kernel for new releases moving forward
Replies
6
Views
229
Amamoh
A
zohaibahd
Real-time Linux has been integrated into the main kernel following two decades of development
Replies
3
Views
131
MrPractical
M

Latest posts

Back