Re: [PATCH 24/24] target/ppc: Restrict ppc_cpu_tlb_fill to TCG

[Bruno Piazera Larsen]

On 20/05/2021 11:52, Richard Henderson wrote:

> On
>       5/20/21 8:18 AM, Bruno Piazera Larsen wrote:
>       
> > > +#ifdef CONFIG_TCG
> > >           
> > >             bool ppc_cpu_tlb_fill(CPUState *cs, vaddr eaddr, int size,
> > >           
> > >                                   MMUAccessType access_type, int
> > >           mmu_idx,
> > >           
> > >                                   bool probe, uintptr_t retaddr)
> > >           
> > >           @@ -2984,3 +2985,4 @@ bool ppc_cpu_tlb_fill(CPUState *cs,
> > >           vaddr eaddr, int size,
> > >           
> > >                 raise_exception_err_ra(&cpu->env,
> > >           cs->exception_index,
> > >           
> > >                                        cpu->env.error_code, retaddr);
> > >           
> > >             }
> > >           
> > >           +#endif
> > >           
> > >         
> >         
> >         This patch makes it look like we would compile mmu_helper.c
> >         after all. Is that it? That looks like the simplest solution
> >         (ifdef'ing away all helpers) but I thought mmu_helper was
> >         supposed to have all TCG-only code relating to MMU.
> >         
> >       
>       
>       Yes, we will always compile mmu_helper.c. -- it was always going
>       to have the stuff for gdbstub.
>       
>       
>       It was exc_helper which I was hoping for tcg-only (but then we
>       decided that ifdefs were going to be the short-term solution).
>       
>       
>

Ah, it was just me misremembering then. Well, with this patch and the final bits that I hadn't sent with v2 (I didn't want to change meson.build without fully supporting disable-tcg) everything is compiling, but tests are... spotty, at best

The check-acceptance group of tests only had one test that was passed, the tests/acceptance/machine_ppc.py:PpcMachine.test_ppc64_pseries, all the others either timed out or failed, and make check got up to prom-env-test, which failed and left the whole thing hanging. The fail was: ../../tests/qtest/prom-env-test.c:43:check_guest_memory: assertion failed (signature == MAGIC): (0x00000000 == 0xcafec0de)

I bring this up because I don't know how much shoul dbe working and how much is TCG-only testing which shouldn't be running and we should disable for the final patch

--
Bruno Piazera Larsen
Instituto de Pesquisas ELDORADO
Departamento Computação Embarcada
Analista de Software Trainee
Aviso Legal - Disclaimer