On Thu, 3 Aug 2023 at 15:42, Richard Henderson
<richard.hender...@linaro.org> wrote:
>
> On 8/3/23 06:10, Peter Maydell wrote:
> > On Thu, 27 Jul 2023 at 17:33, Richard Henderson
> >> - mop = MO_128;
> >> - if (s->align_mem) {
> >> - mop |= MO_ALIGN_8;
> >> - }
> >> - mop = finalize_memop_pair(s, mop);
> >> + mop = MO_128 | MO_ALIGN | MO_ATOM_IFALIGN_PAIR;
> >
> > So here we're implicitly assuming TAG_GRANULE is 16 and
> > then relying on the codegen for a MO_128 | MO_ALIGN
> > operation to give us the alignment fault if the guest
> > address isn't aligned to the tag granule, right ?
>
> Yes.
>
> >
> > Previously we also put s->be_data into the MemOp
> > (via finalize_memop_pair() calling finalize_memop_atom()).
> > Don't we still need to do that ?
>
> Whoops, yes.
I think you still need to respin this one, right?
(I re-found it because I remembered this "writes tags,
doesn't check them" thing and that it ought to apply to
the FEAT_MOPS SETG block-tag-set that I'm writing also.)
thanks
-- PMM
