Re: [Qemu-devel] [PATCH v9 3/6] target/ppc: Handle NMI guest exit

[Aravinda Prasad <aravinda@linux.vnet.ibm.com>]

On Thursday 06 June 2019 07:13 AM, David Gibson wrote:
> On Wed, May 29, 2019 at 11:10:32AM +0530, Aravinda Prasad wrote:
>> Memory error such as bit flips that cannot be corrected
>> by hardware are passed on to the kernel for handling.
>> If the memory address in error belongs to guest then
>> the guest kernel is responsible for taking suitable action.
>> Patch [1] enhances KVM to exit guest with exit reason
>> set to KVM_EXIT_NMI in such cases. This patch handles
>> KVM_EXIT_NMI exit.
>>
>> [1] https://www.spinics.net/lists/kvm-ppc/msg12637.html
>>     (e20bbd3d and related commits)
>>
>> Signed-off-by: Aravinda Prasad <aravinda@linux.vnet.ibm.com>
>> ---
>>  hw/ppc/spapr.c          |    1 +
>>  hw/ppc/spapr_events.c   |   23 +++++++++++++++++++++++
>>  hw/ppc/spapr_rtas.c     |    5 +++++
>>  include/hw/ppc/spapr.h  |    6 ++++++
>>  target/ppc/kvm.c        |   16 ++++++++++++++++
>>  target/ppc/kvm_ppc.h    |    2 ++
>>  target/ppc/trace-events |    1 +
>>  7 files changed, 54 insertions(+)
>>
>> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
>> index fae28a9..6b6c962 100644
>> --- a/hw/ppc/spapr.c
>> +++ b/hw/ppc/spapr.c
>> @@ -1809,6 +1809,7 @@ static void spapr_machine_reset(void)
>>  
>>      spapr->cas_reboot = false;
>>  
>> +    spapr->mc_status = -1;
>>      spapr->guest_machine_check_addr = -1;
>>  
>>      /* Signal all vCPUs waiting on this condition */
>> diff --git a/hw/ppc/spapr_events.c b/hw/ppc/spapr_events.c
>> index ae0f093..a18446b 100644
>> --- a/hw/ppc/spapr_events.c
>> +++ b/hw/ppc/spapr_events.c
>> @@ -620,6 +620,29 @@ void spapr_hotplug_req_remove_by_count_indexed(SpaprDrcType drc_type,
>>                              RTAS_LOG_V6_HP_ACTION_REMOVE, drc_type, &drc_id);
>>  }
>>  
>> +void spapr_mce_req_event(PowerPCCPU *cpu, bool recovered)
>> +{
>> +    SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
> 
> You ignore the 'recovered' parameter, is that right?

I use the "recovered" parameter, but in the next patch. This was left
out when the patch was split in one of the earlier versions. Will modify it.

> 
>> +    while (spapr->mc_status != -1) {
>> +        /*
>> +         * Check whether the same CPU got machine check error
>> +         * while still handling the mc error (i.e., before
>> +         * that CPU called "ibm,nmi-interlock")
>> +         */
>> +        if (spapr->mc_status == cpu->vcpu_id) {
>> +            qemu_system_guest_panicked(NULL);
>> +            return;
>> +        }
>> +        qemu_cond_wait_iothread(&spapr->mc_delivery_cond);
>> +        /* Meanwhile if the system is reset, then just return */
>> +        if (spapr->guest_machine_check_addr == -1) {
>> +            return;
>> +        }
>> +    }
>> +    spapr->mc_status = cpu->vcpu_id;
>> +}
>> +
>>  static void check_exception(PowerPCCPU *cpu, SpaprMachineState *spapr,
>>                              uint32_t token, uint32_t nargs,
>>                              target_ulong args,
>> diff --git a/hw/ppc/spapr_rtas.c b/hw/ppc/spapr_rtas.c
>> index e7509cf..e0bdfc8 100644
>> --- a/hw/ppc/spapr_rtas.c
>> +++ b/hw/ppc/spapr_rtas.c
>> @@ -379,6 +379,11 @@ static void rtas_ibm_nmi_interlock(PowerPCCPU *cpu,
>>          /* NMI register not called */
>>          rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
>>      } else {
>> +        /*
>> +         * vCPU issuing "ibm,nmi-interlock" is done with NMI handling,
>> +         * hence unset mc_status.
>> +         */
>> +        spapr->mc_status = -1;
>>          qemu_cond_signal(&spapr->mc_delivery_cond);
>>          rtas_st(rets, 0, RTAS_OUT_SUCCESS);
>>      }
>> diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
>> index 9dc5e30..fc3a776 100644
>> --- a/include/hw/ppc/spapr.h
>> +++ b/include/hw/ppc/spapr.h
>> @@ -190,6 +190,11 @@ struct SpaprMachineState {
>>  
>>      /* State related to "ibm,nmi-register" and "ibm,nmi-interlock" calls */
>>      target_ulong guest_machine_check_addr;
>> +    /*
>> +     * mc_status is set to -1 if mc is not in progress, else is set to the CPU
>> +     * handling the mc.
>> +     */
>> +    int mc_status;
>>      QemuCond mc_delivery_cond;
>>  
>>      /*< public >*/
>> @@ -793,6 +798,7 @@ void spapr_clear_pending_events(SpaprMachineState *spapr);
>>  int spapr_max_server_number(SpaprMachineState *spapr);
>>  void spapr_store_hpte(PowerPCCPU *cpu, hwaddr ptex,
>>                        uint64_t pte0, uint64_t pte1);
>> +void spapr_mce_req_event(PowerPCCPU *cpu, bool recovered);
>>  
>>  /* DRC callbacks. */
>>  void spapr_core_release(DeviceState *dev);
>> diff --git a/target/ppc/kvm.c b/target/ppc/kvm.c
>> index 3bf0a46..39f1a73 100644
>> --- a/target/ppc/kvm.c
>> +++ b/target/ppc/kvm.c
>> @@ -1761,6 +1761,11 @@ int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
>>          ret = 0;
>>          break;
>>  
>> +    case KVM_EXIT_NMI:
>> +        trace_kvm_handle_nmi_exception();
>> +        ret = kvm_handle_nmi(cpu, run);
>> +        break;
>> +
>>      default:
>>          fprintf(stderr, "KVM: unknown exit reason %d\n", run->exit_reason);
>>          ret = -1;
>> @@ -2844,6 +2849,17 @@ int kvm_arch_msi_data_to_gsi(uint32_t data)
>>      return data & 0xffff;
>>  }
>>  
>> +int kvm_handle_nmi(PowerPCCPU *cpu, struct kvm_run *run)
>> +{
>> +    bool recovered = run->flags & KVM_RUN_PPC_NMI_DISP_FULLY_RECOV;
>> +
>> +    cpu_synchronize_state(CPU(cpu));
>> +
>> +    spapr_mce_req_event(cpu, recovered);
> 
> Urgh.. KVM calling directly into spapr code isn't conceptually
> correct, although it's usually kind of ok in practice.  I guess we
> already do it for hypercalls, so this is no worse.  If we ever have
> NMI events for KVM PR or BookE KVM which could happen for non-PAPR
> guests, I guess we'll need to re-examine this.

ok.

> 
>> +    return 0;
>> +}
>> +
>>  int kvmppc_enable_hwrng(void)
>>  {
>>      if (!kvm_enabled() || !kvm_check_extension(kvm_state, KVM_CAP_PPC_HWRNG)) {
>> diff --git a/target/ppc/kvm_ppc.h b/target/ppc/kvm_ppc.h
>> index 45776ca..18693f1 100644
>> --- a/target/ppc/kvm_ppc.h
>> +++ b/target/ppc/kvm_ppc.h
>> @@ -81,6 +81,8 @@ bool kvmppc_hpt_needs_host_contiguous_pages(void);
>>  void kvm_check_mmu(PowerPCCPU *cpu, Error **errp);
>>  void kvmppc_set_reg_ppc_online(PowerPCCPU *cpu, unsigned int online);
>>  
>> +int kvm_handle_nmi(PowerPCCPU *cpu, struct kvm_run *run);
>> +
>>  #else
>>  
>>  static inline uint32_t kvmppc_get_tbfreq(void)
>> diff --git a/target/ppc/trace-events b/target/ppc/trace-events
>> index 3dc6740..6d15aa9 100644
>> --- a/target/ppc/trace-events
>> +++ b/target/ppc/trace-events
>> @@ -28,3 +28,4 @@ kvm_handle_papr_hcall(void) "handle PAPR hypercall"
>>  kvm_handle_epr(void) "handle epr"
>>  kvm_handle_watchdog_expiry(void) "handle watchdog expiry"
>>  kvm_handle_debug_exception(void) "handle debug exception"
>> +kvm_handle_nmi_exception(void) "handle NMI exception"
>>
> 

-- 
Regards,
Aravinda