On Wed, Mar 11, 2015 at 05:49:39PM +0000, Peter Maydell wrote:
> On 11 March 2015 at 17:42, Andrew Jones <drjo...@redhat.com> wrote:
> > On Wed, Mar 11, 2015 at 05:02:00PM +0000, Peter Maydell wrote:
>
> >> > + if (is_aa64) {
> >> > + switch (regime_el(env, mmu_idx)) {
> >> > + case 1:
> >> > + if (is_user && !user_rw) {
> >> > + wxn = 0;
> >>
> >> I don't understand this. We ignore the WXN bit if this is
> >> a user access and the page is not readable ?
> >
> > Yup. If the page is not readable or writeable, AP[1]=0. I almost
> > submitted an errata to the ARM ARM when I saw this on the 2nd line
> > of table D4-32. I thought it must be a typo. However I tested it
> > on hardware, and it works this way. So at least the weirdness has
> > been implemented consistently...
>
> Still confused. If the page isn't readable or writable
> then WXN isn't going to kick in anyway because WXN only
> affects writable pages. I don't see what the case is
> where this bit of code will make a difference.
>
Ah, that is true. Too bad I didn't read this before sending v3,
as I could have removed it, if you prefer. I had it here to
be explicit about the ignoring of wxn - matching the spec, but
you're right, it's useless code. Should I send a v4?
drew
