On Fri, Jul 03, 2020 at 02:10:08PM -0700, Linus Torvalds wrote: > Yeah, the "stac" instruction isn't hugely fast, and serializes the > pipeline, so it's a nasty 20 cycles or something. > > But for chissake, this > (a) happens approximately never > (b) is after a fault that took a thousand cycles > > so the trivial thing to do is to just say "yeah, you need to add the > STAC when your optimistic thing failed and you have to fall back to > the byte-at-a-time tail case".
Not the problem I'm concerned about, really. However, I would really like to lift stac/clac into the *callers* of raw_copy_from_user() et.al. and fold them into user_access_begin/user_access_end there. And that's where the rules become very interesting - raw_copy_from_user() is not "succeed or fail" thing, it's "tell me how much has been left to copy" one. Put it that way - here we really do have outputs on fault. PS: I hope to kill __copy_from_user()/__copy_to_user() outside of arch/* this cycle; not much is left by now. So I'm not talking about lifting stac/clac out into the wild - it will merge with access_ok into user_access_begin/end.
