On Sat May 18, 2024 at 9:05 PM AEST, Richard Henderson wrote:
> On 5/18/24 11:31, Nicholas Piggin wrote:
> > @@ -425,17 +426,17 @@ static void powerpc_set_excp_state(PowerPCCPU *cpu,
> > target_ulong vector,
> > env->reserve_addr = -1;
> > }
> >
> > -static void powerpc_mcheck_checkstop(CPUPPCState *env)
> > -{
> > - /* KVM guests always have MSR[ME] enabled */
> > #ifdef CONFIG_TCG
> > +/*
> > + * This stops the machine and logs CPU state without killing QEMU (like
> > + * cpu_abort()) because it is often a guest error as opposed to a QEMU
> > error,
> > + * so the machine can still be debugged.
> > + */
> > +static G_NORETURN void powerpc_checkstop(CPUPPCState *env, const char
> > *reason)
> > +{
> > CPUState *cs = env_cpu(env);
> > FILE *f;
> >
> > - if (FIELD_EX64(env->msr, MSR, ME)) {
> > - return;
> > - }
> > -
> > f = qemu_log_trylock();
> > if (f) {
> > fprintf(f, "Entering checkstop state: "
> > @@ -451,6 +452,30 @@ static void powerpc_mcheck_checkstop(CPUPPCState *env)
> > */
> > qemu_system_guest_panicked(NULL);
> > cpu_loop_exit_noexc(cs);
> > +}
>
> Looks like you're not using reason.
> Missed swapping out in the logging block?
Good catch thanks, I lost the hunk in a rebase.
Thanks,
Nick
