> Since I raised this comment on X86, I think raising it for ARM for > discussion would be ok, > it might make sense to have consistent API experience for arm/x86 here. And indeed this is what I think is most important at this point. Considering that Aruna's series resets PKRU unconditionally (sigaltstack or not) and has already been pulled into mainline during 6.12-rc1 [2], I still believe that patch 1 is doing the right thing, i.e. aligning arm64 with x86. If the concern with performance is confirmed, and there is an agreement to reset POR/PKRU less eagerly on both architectures, this could potentially be revisited. - Kevin [2] https://lore.kernel.org/lkml/172656199227.2471820.13578261908219597067.tglx@xen13/ > [1] https://lore.kernel.org/lkml/CABi2SkWjF2Sicrr71=a6H8XJyf9q9L_Nd5FPp0CJ2mvB46Rrrg@xxxxxxxxxxxxxx/
