On 8/14/20 2:26 AM, Andrew Jones wrote: >> +static uint64_t __attribute__((noinline)) >> +pauth_computepac_impdef(uint64_t data, uint64_t modifier, ARMPACKey key) > > Out of curiosity, why do we need to make these computepac functions > noinline?
Oh, heh. Left over from profiling. Will remove. > I think this patch should come before the last one. As it stands, when > bisecting between the last one and this one a user could attempt to > enable pauth-imdef, but it wouldn't do anything, or it would potentially > break things. However, this patch shouldn't change anything if it comes > first. The current patch ordering would enable impdef but implement that with the architected algorithm. Which is ok. But you're right that the other ordering makes more sense. r~