Re: [Qemu-devel] [Qemu-ppc] [PATCH 8/8] target-ppc Disentangle ppc64 hash mmu path for cpu_ppc_handle_mmu_fault

[David Gibson]

On Fri, Feb 22, 2013 at 05:13:56PM +0100, Alexander Graf wrote:
> On 12.02.2013, at 03:00, David Gibson wrote:
[snip]
> > +    } else if (ret < 0) {
> > +        LOG_MMU_STATE(env);
> > +        if (access_type == ACCESS_CODE) {
> > +            switch (ret) {
> > +            case -1:
> > +                env->exception_index = POWERPC_EXCP_ISI;
> > +                env->error_code = 0x40000000;
> > +                break;
> > +            case -2:
> > +                /* Access rights violation */
> 
> These really should become an enum or defines. Something where the
> thing we check on tells us what we're checking for without a comment
> :).

So, I didn't want to introduce an emum or whatever at the same time as
I was just moving the code around - makes for more chance of
mistakes.  And by the end of the series, I've gotten rid of these
magic error values anyway, by reporting the exceptions in-line in mmu
fault handle, so I think it's kind of moot.

[snip]
> > +            }
> > +        }
> > +#if 0
> 
> #ifdef DEBUG_MMU?

Again, this is just moved code.


> > +        printf("%s: set exception to %d %02x\n", __func__,
> > +               env->exception, env->error_code);
> > +#endif
> > +        ret = 1;
> > +    }
> > +
> > +    return ret;
> > +}
> > diff --git a/target-ppc/mmu_helper.c b/target-ppc/mmu_helper.c
> > index 98143dd..f8f213b 100644
> > --- a/target-ppc/mmu_helper.c
> > +++ b/target-ppc/mmu_helper.c
> > @@ -1533,6 +1533,18 @@ int cpu_ppc_handle_mmu_fault(CPUPPCState *env, 
> > target_ulong address, int rw,
> >     int access_type;
> >     int ret = 0;
> > 
> > +    switch (env->mmu_model) {
> > +#if defined(TARGET_PPC64)
> > +    case POWERPC_MMU_64B:
> > +    case POWERPC_MMU_2_06:
> > +    case POWERPC_MMU_2_06d:
> > +        return ppc_hash64_handle_mmu_fault(env, address, rw, mmu_idx);
> > +#endif
> > +
> > +    default:
> > +        ; /* Otherwise fall through to the general code below */
> > +    }
> 
> Sigh. This too should become a class function. Maybe it is within
> the scope of this patch set after all :). You could keep the
> "legacy" handling code alive and just point the non-converted
> targets to the existing function. Then override the handler for
> book3s_64 families to ppc_hash64_handle_mmu_fault.

Well, as I said, I think it makes sense to move the switch to one
place first, then change it to a method dispatch.  I can look at
adding patches to do that conversion to the end of the series, but
I'll need to pull down the cpu qoming stuff furst.

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

signature.asc
Description: Digital signature