Linux Fixes Indirect Branch Predictor Barrier "IBPB" Handling For Older AMD CPUs

[phoronix]

Phoronix: Linux Fixes Indirect Branch Predictor Barrier "IBPB" Handling For Older AMD CPUs

Merged today to Linux 6.12 Git were bug fixes to AMD's Indirect Branch Predictor Barrier (IBPB) handling that can be optionally used as part of the Retbleed and Speculative Return Stack Overflow (SRSO) mitigations on older AMD processors...

https://www.phoronix.com/news/Linux-AMD-Fixes-IBPB-Older-CPUs

[Danny3]

Good!
But I don't get it...
Is IBPB mitigation needed or not if one wants the best privacy and security possible?
Why is this not enabled by default in the kernel, does it have a high performance impact?

[blackshard]

Good!
But I don't get it...
Is IBPB mitigation needed or not if one wants the best privacy and security possible?
Why is this not enabled by default in the kernel, does it have a high performance impact?

Perhaps there is another better option with those processors

[F.Ultra]

Good!
But I don't get it...
Is IBPB mitigation needed or not if one wants the best privacy and security possible?
Why is this not enabled by default in the kernel, does it have a high performance impact?

zen3 uses full retpolines by default instead of IBPB. I guess the IBPB overhead on zen3 is substantial.

[MorrisS.]

so what to do to take benefit of this feature?