Re: [PATCH v3 6/8] acpi/ghes: Use error_abort in acpi_ghes_memory_errors()

[Igor Mammedov <imammedo@redhat.com>]

On Tue, 11 Nov 2025 13:58:46 +1000
Gavin Shan <gshan@redhat.com> wrote:

> Hi Philippe,
> 
> On 11/11/25 12:54 AM, Philippe Mathieu-Daudé wrote:
> > On 5/11/25 12:44, Gavin Shan wrote:  
> >> Use error_abort in acpi_ghes_memory_errors() so that the caller needn't
> >> explicitly call abort() on errors. With this change, its return value
> >> isn't needed any more.
> >>
> >> Suggested-by: Igor Mammedov <imammedo@redhat.com>
> >> Signed-off-by: Gavin Shan <gshan@redhat.com>
> >> ---
> >>   hw/acpi/ghes-stub.c    |  6 +++---
> >>   hw/acpi/ghes.c         | 15 ++++-----------
> >>   include/hw/acpi/ghes.h |  5 +++--
> >>   target/arm/kvm.c       | 10 +++-------
> >>   4 files changed, 13 insertions(+), 23 deletions(-)  
> > 
> >   
> >> diff --git a/include/hw/acpi/ghes.h b/include/hw/acpi/ghes.h
> >> index f73908985d..35c7bbbb01 100644
> >> --- a/include/hw/acpi/ghes.h
> >> +++ b/include/hw/acpi/ghes.h
> >> @@ -98,8 +98,9 @@ void acpi_build_hest(AcpiGhesState *ags, GArray *table_data,
> >>                        const char *oem_id, const char *oem_table_id);
> >>   void acpi_ghes_add_fw_cfg(AcpiGhesState *vms, FWCfgState *s,
> >>                             GArray *hardware_errors);
> >> -int acpi_ghes_memory_errors(AcpiGhesState *ags, uint16_t source_id,
> >> -                            uint64_t *addresses, uint32_t num_of_addresses);
> >> +void acpi_ghes_memory_errors(AcpiGhesState *ags, uint16_t source_id,
> >> +                             uint64_t *addresses, uint32_t num_of_addresses,
> >> +                             Error **errp);  
> > 
> > This is an anti-pattern w.r.t. commit e3fe3988d78 ("error: Document
> > Error API usage rules"), because it can be called with an errp distinct
> > of &error_abort.
> > 
> > If you really want to abort(), remove the errp argument, directly call
> > abort() and rename as acpi_ghes_memory_abort_on_errors().
> >   
> 
> Thanks for pointing it out. I will improve this like below in next revision.
> 
> - Drop 'errp' argument from acpi_ghes_memory_errors(), but I prefer to keep
>    the function name.
> 
> - In acpi_ghes_memory_errors(), a local variable 'Error *err' is added and
>    pass it to ghes_record_cper_errors(), which is also called by QMP handler
>    qmp_inject_ghes_v2_error().
> 
> Please let me know if there are any more improvements needed.
> 
> >>   void ghes_record_cper_errors(AcpiGhesState *ags, const void *cper, size_t len,
> >>                                uint16_t source_id, Error **errp);
> >> diff --git a/target/arm/kvm.c b/target/arm/kvm.c
> >> index 459ca4a9b0..a889315606 100644
> >> --- a/target/arm/kvm.c
> >> +++ b/target/arm/kvm.c
> >> @@ -2458,13 +2458,9 @@ void kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void *addr)
> >>               addresses[0] = paddr;
> >>               if (code == BUS_MCEERR_AR) {
> >>                   kvm_cpu_synchronize_state(c);
> >> -                if (!acpi_ghes_memory_errors(ags, ACPI_HEST_SRC_ID_SYNC,
> >> -                                             addresses, 1)) {
> >> -                    kvm_inject_arm_sea(c);
> >> -                } else {
> >> -                    error_report("failed to record the error");
> >> -                    abort();
> >> -                }
> >> +                acpi_ghes_memory_errors(ags, ACPI_HEST_SRC_ID_SYNC,
> >> +                                        addresses, 1, &error_abort);

I guess that was my request,
with calling abort() directly you have to print error separately,
while using 'error_abort' is much more cleaner (as this hunk demonstrates)  

> >> +                kvm_inject_arm_sea(c);
> >>               }
> >>               return;
> >>           }  
> >   
> 
> Thanks,
> Gavin
>