Re: [PATCH 2/3] target/arm: Make IT insn undef when not present

On Fri, 21 Aug 2026 at 15:21, Richard Henderson
<[email protected]> wrote:
>
> On 8/21/26 06:25, Peter Maydell wrote:
> > The IT insn is introduced for A-profile starting in v6T2; for
> > M-profile it is present when the Main Extension is implemented (which
> > includes v7M and excludes v6M).  We were missing the feature-check,
> > so fail to UNDEF on earlier cores.
> >
> > Add the missing check.
> >
> > Cc: [email protected]
> > Fixes: 9ee6e8bb853bde ("ARMv7 support.")
> > Signed-off-by: Peter Maydell <[email protected]>
> > ---
> >   target/arm/tcg/translate.c | 10 ++++++++++
> >   1 file changed, 10 insertions(+)
> >
> > diff --git a/target/arm/tcg/translate.c b/target/arm/tcg/translate.c
> > index df98dc2e34d..103e2fe7c73 100644
> > --- a/target/arm/tcg/translate.c
> > +++ b/target/arm/tcg/translate.c
> > @@ -6038,6 +6038,16 @@ static bool trans_IT(DisasContext *s, arg_IT *a)
> >   {
> >       int cond_mask = a->cond_mask;
> >
> > +    /*
> > +     * IT insn introduced in v6T2 for A-profile; it is only present
> > +     * on M-profile if the Main Extension is implemented.
> > +     */
> > +    if (!arm_dc_feature(s, ARM_FEATURE_THUMB2) ||
> > +        (arm_dc_feature(s, ARM_FEATURE_M) &&
> > +         !arm_dc_feature(s, ARM_FEATURE_M_MAIN))) {
> > +        return false;
> > +    }
> In practice this is going to be the same as just thumb2, because v7M has v7 which implies
> thumb2, and all v7M have m_main.

v8M without the main extension exists, though we don't model any
of those CPUs at the moment. That will have ARM_FEATURE_THUMB2
(because FEATURE_V8 implies FEATURE_V7 and FEATURE_V7 implies
FEATURE_THUMB2).

> Is it clearer as
>
>         !(arm_dc_feature(s, ARM_FEATURE_M)
>            ? arm_dc_feature(s, ARM_FEATURE_M_MAIN)
>            : arm_dc_feature(s, ARM_FEATURE_THUMB2))
>
> ?

Yeah, I think so. (I actually wrote it first with nested if()s
roughly like that.)

-- PMM