On Sun, Jul 1, 2012 at 12:37 AM, Rusty Russell <rusty.russell at linaro.org> wrote: > On Sat, 30 Jun 2012 21:57:00 -0400, Christoffer Dall <c.dall at virtualopensystems.com> wrote: >> On Sat, Jun 30, 2012 at 7:42 PM, Peter Maydell <peter.maydell at linaro.org> wrote: >> > On 1 July 2012 00:17, Christoffer Dall <c.dall at virtualopensystems.com> wrote: >> >> I'm writing test code for the Thumb/ARM load-store decoding and MMIO >> >> code, which led me to a couple of thoughts questions: >> >> >> >> First, Are all memory access instructions valid for MMIO access, >> >> including load/store multiple, SIMD instructions, etc.? >> > >> > Yes, although some of them have interesting caveats that mean >> > that nobody in their right mind would actually use them for >> > accessing device registers, eg: >> > >> > # The architecture permits an Advanced SIMD element or structure load >> > # instruction to access bytes in Device or Strongly-ordered memory that >> > # are not explicitly accessed by the instruction, provided the bytes >> > # accessed are in a 16-byte window, aligned to 16-bytes, that contains >> > # at least one byte that is explicitly accessed by the instruction >> > >> >> ok, I'll probably punt on SIMD instructions for now then... > > Agreed. It'd be nice to inject some kind of trap into the guest for > this though, rather than silent fail. > an abort? We already have the undefined injection in place. -Christoffer
