Re: [PATCH for-6.2 07/43] target/ppc: Set fault address in ppc_cpu_do_unaligned_access

[Richard Henderson]

On 7/29/21 3:44 AM, Peter Maydell wrote:
> On Thu, 29 Jul 2021 at 01:51, Richard Henderson
> <richard.henderson@linaro.org> wrote:
> > We ought to have been recording the virtual address for reporting
> > to the guest trap handler.
> >
> > Cc: qemu-ppc@nongnu.org
> > Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> > ---
> >   target/ppc/excp_helper.c | 2 ++
> >   1 file changed, 2 insertions(+)
> >
> > diff --git a/target/ppc/excp_helper.c b/target/ppc/excp_helper.c
> > index a79a0ed465..0b2c6de442 100644
> > --- a/target/ppc/excp_helper.c
> > +++ b/target/ppc/excp_helper.c
> > @@ -1503,6 +1503,8 @@ void ppc_cpu_do_unaligned_access(CPUState *cs, vaddr 
> > vaddr,
> >       CPUPPCState *env = cs->env_ptr;
> >       uint32_t insn;
> >
> > +    env->spr[SPR_DAR] = vaddr;
> > +
>
> Is this the right SPR for all PPC variants? For instance the
> kernel's code in arch/powerpc/kernel/exceptions-64e.S looks
> in SPRN_DEAR, which is our SPR_BOOKE_DEAR or SPR_40x_DEAR.

I have no idea.  I glanced through a handful of the mmu's, and looked at the current BookS 
docs, but that's certainly not all.

I'll note that if we do need to set different regs for different mmus, we'll probably want 
to standardize on this one for user-only, like we did for the user-only copy of 
ppc_cpu_tlb_fill.


r~