On Thu, Oct 27, 2022 at 11:58:35AM -0700, Kees Cook wrote: > I wandered around attempting to deal with in_nmi(), etc. And in > the end just drop the attempt to cover it. It's worth noting that > copy_from_user_nmi() exists on 1 architecture and has exactly 1 > call-site... Yeah, back when I wrote it, it was a lot more complicated because we could not reliably take #PF from NMI context; it did manual page-walks, kmap_atomic()s and mempcy(). That's all fixed now and it's really mostly a rudiment -- except for these instrumentation issues it seems.
