* Michael Sammler: > As I understand it, the code you mention talks about setting a > protection key being slow[1], not reading the PKRU. Reading the PKRU > should be fast as it is an register in the processor, but I have not > conducted extensive benchmarks yet. > > [1] https://github.com/AndroidHardening/hardened_malloc/issues/62 This talks about the glibc implementation, which is curious in the Android context, but it's fixable by providing an inline implementation. I didn't want to do this because the interface is still evolving (see the discussion about PKEY_DISABLE_READ). Thanks, Florian
