Linus Torvalds Comments On STIBP & He's Not Happy - STIBP Default Will End Up Changing

[phoronix]

Phoronix: Linus Torvalds Comments On STIBP & He's Not Happy - STIBP Default Will End Up Changing

It turns out that Linus Torvalds himself was even taken by surprise with the performance hit we've outlined on Linux 4.20 as a result of STIBP "Single Thread Indirect Branch Predictors" introduction as well as back-porting already to stable series for cross-hyperthread Spectre V2 protection. He doesn't want this enabled in full by default...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-Torvalds-STIBP-Comment

[Shiba]

Oh, he's gonna explode... give him time XD

[Redi44]

Nah, he has been indoctrinated by PC people during the forced "vacation". We've lost boys.

[L_A_G]

On one hand this patch probably is necessary for people running systems that aren't getting microcode updates anymore, of which there are quite a few seeing how hardware vendors' cut-off points tend to have plenty serviceable systems falling on the wrong side of it and it could stop some undiscovered and yet-to-be-resolved issues. I'd be interested to see what this does to the new speculative execution vulnerabilities published as recently last week.

On the other hand the performance impact is unusually high for a vulnerability fix, particularly in a lot of server workloads and we've already seen what can happen when meltdown/spectre fixes kill the performance of your server workload in the server issues Fortnite experienced got the initial patches. They did fix the issue in a week or two, but I'm not sure if it was by adding new hardware or workarounds like disabling the mitigations.

[Brisse]

Nah, he has been indoctrinated by PC people during the forced "vacation". We've lost boys.

FOSS community is "boys" only? Perhaps that's why it's so immature.

[dungeon]

Harassment free for everybody without photosensitive seizure warning



I think anything with compositor, especially Wayland needs PSE warning on boot

[ALRBP]

If we can just disable it when compiling the kernel, then it should not be an issue for anyone compiling his kernel himself, or using a distro with sane maintainers.

"Warn once about it, and let the crazy people say "I'd rather take a 50% performance hit than worry about a theoretical issue"."
As a PhD student in computer science, I'm used to people who are essentially worried by theoretical questions, and actually seems to not care about practical considerations, to the point that a lot of their published work is in practice worthless (and lots of valuable work is not being given attention). I'm still surprised to see that engineers (kernel developers) are actually acting the same way. Maybe an effect of the excessive attention those issues where given after being reveled in research papers.
(Even in applied research, making a good article is considered more important than doing something this is actually useful)

At least, Torvalds and others are going to remove this from the default kernel config (hopefully, no distro maintainer will enable this when compiling their default kernel, except from security-centered distros ).

We talked a lot about Spectre, Meltdown or Heartbleed but, AFAIK, they were never actually exploited in practice, while tons of much less known issues where actually exploited to hack millions of systems. This kind of severity distortion, which has proved to be extremely dangerous when used to manipulate people's opinion for political purposes, seems to also be an issue in engendering.

Nah, he has been indoctrinated by PC people during the forced "vacation". We've lost boys.

I was sure that this subject would show up, but I'm surprised to see that Torvalds is the one pointed, and not those who created these security patchs. When reading his message, I find him perfectly sane from a technical pov, and maybe still a bit hard for public communication when using the word "crazy".

[wizard69]

It is a good thing I don't have to worry about server security, system performance and unknown exploits. However I'm still wondering who will make it to market first, with a processors that doens't have these issues.

Also is the real solution here to drop SMT completely and implement real cores? I ask because for personal use I usually find 8 threads to be good enough for mainstream use. Building software could always use more threads but that will likely always be the case.

[wizard69]

FOSS community is "boys" only? Perhaps that's why it's so immature.

That is reality. I've had the good pleasure of knowing a few good programmers in the business world but they are few and far between. Why I don't know, all I do know is the wash out rate in intro to comp-sci is huge for both males and females. Usually the reality of the job sinks in pretty fast, not a lot of people want to be anchored to a desk all day (which by the way is why I never pursued the industry).

[birdie]

Michael and it's me, sir, who made Linus notice the issue. I'm just saying. Despite tons of hatred that I receive here. Luckily I don't care.

It's kinda sad that when I'm saying something here people disregard me, but if it's Linus then, "Oh, God, he's so right".