Re: [PATCH v4 10/17] target/m68k: Implement TRAPcc

[Richard Henderson]

On 5/25/22 17:15, Laurent Vivier wrote:
> Le 26/05/2022 à 00:26, Richard Henderson a écrit :
> > On 5/25/22 14:40, Laurent Vivier wrote:
> > > > +DISAS_INSN(trapcc)
> > > > +{
> > > > +    DisasCompare c;
> > > > +
> > > > +    /* Consume and discard the immediate operand. */
> > > > +    switch (extract32(insn, 0, 3)) {
> > > > +    case 2: /* trapcc.w */
> > > > +        (void)read_im16(env, s);
> > > > +        break;
> > > > +    case 3: /* trapcc.l */
> > > > +        (void)read_im32(env, s);
> > > > +        break;
> > >
> > > Do we really need to read the data or do we only need to increment s->pc (as the data 
> > > are only here to be available for the trap handler)?
> >
> > We need to read the data to (1) trigger sigsegv when this insn crosses a page and (2) 
> > passing to tcg plugins.
>
> For (1) I was wondering if the real CPU is actually doing it.
>
> Nothing is said about it in the instruction definition.

Surely the cpu reads cachelines at a time, so of course it would.

r~