News for hackers. Stuff that matters.
Re:Spectre mitigations murder userspace performance by yyyk on Saturday June 19, 2021 @03:07PM (#27563407)
On the one hand, I haven't seen any real exploitation of Spectre, at least by a non-TLA. On the other hand, the mitigations aren't so slow anymore on a modern CPU:

https://www.phoronix.com/scan.php?page=article&item=3-years-...

IMHO, there's a good argument for turning off the mitigations on Skylake in some systems - but in modern CPUs the cost of leaving the mitigations on is bearable.

Re:Spectre mitigations murder userspace performance by staticassertion on Saturday June 19, 2021 @03:12PM (#27563459)
Yeah, kinda similar to rowhammer. I was just saying the other day that these exploit techniques are totally viable, but there's not much point - privesc is trivial on any desktop environment (linux or Windows, maybe less on osx), and usually not too hard on servers either. Fancy exploits aren't worth it when operating systems are so vulnerable to local attackers.

That said, the mitigations are assurances, and it also lets us point a finger at intel and say "fix your shit you just caused a global performance regression".

Re:Spectre mitigations murder userspace performance by lima on Saturday June 19, 2021 @07:41PM (#27565598)
> linux or Windows, maybe less on osx

Trivial on macOS, too.

Smartphone operating systems are better (and ChromeOS).