[PATCH] drm/amdkfd: refine the poison data consumption handling

[Felix Kuehling]

Am 2021-05-11 um 4:06 a.m. schrieb Dennis Li:
> The user applications maybe register the KFD_EVENT_TYPE_HW_EXCEPTION and

I guess the HW exception event is sent because the current handling of
poison consumption triggers a mode2 reset. If that can be removed in the
future, then we should not send a HW_EXCEPTION any more.


> KFD_EVENT_TYPE_MEMORY events, driver could notify them when poison data
> consumed. Beside that, some applications maybe register SIGBUS signal
> hander. These applications will handle poison data by themselves, exit
> or re-create context to re-dispatch works.
>
> Signed-off-by: Dennis Li <Dennis.Li at amd.com>

Reviewed-by: Felix Kuehling <Felix.Kuehling at amd.com>


>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_events.c
> index ba2c2ce0c55a..4d210f23c33c 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_events.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_events.c
> @@ -1050,3 +1050,42 @@ void kfd_signal_reset_event(struct kfd_dev *dev)
>  	}
>  	srcu_read_unlock(&kfd_processes_srcu, idx);
>  }
> +
> +void kfd_signal_poison_consumed_event(struct kfd_dev *dev, u32 pasid)
> +{
> +	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid);
> +	struct kfd_hsa_memory_exception_data memory_exception_data;
> +	struct kfd_hsa_hw_exception_data hw_exception_data;
> +	struct kfd_event *ev;
> +	uint32_t id = KFD_FIRST_NONSIGNAL_EVENT_ID;
> +
> +	if (!p)
> +		return; /* Presumably process exited. */
> +
> +	memset(&hw_exception_data, 0, sizeof(hw_exception_data));
> +	hw_exception_data.gpu_id = dev->id;
> +	hw_exception_data.memory_lost = 1;
> +	hw_exception_data.reset_cause = KFD_HW_EXCEPTION_ECC;
> +
> +	memset(&memory_exception_data, 0, sizeof(memory_exception_data));
> +	memory_exception_data.ErrorType = KFD_MEM_ERR_POISON_CONSUMED;
> +	memory_exception_data.gpu_id = dev->id;
> +	memory_exception_data.failure.imprecise = true;
> +
> +	mutex_lock(&p->event_mutex);
> +	idr_for_each_entry_continue(&p->event_idr, ev, id) {
> +		if (ev->type == KFD_EVENT_TYPE_HW_EXCEPTION) {
> +			ev->hw_exception_data = hw_exception_data;
> +			set_event(ev);
> +		}
> +
> +		if (ev->type == KFD_EVENT_TYPE_MEMORY) {
> +			ev->memory_exception_data = memory_exception_data;
> +			set_event(ev);
> +		}
> +	}
> +	mutex_unlock(&p->event_mutex);
> +
> +	/* user application will handle SIGBUS signal */
> +	send_sig(SIGBUS, p->lead_thread, 0);
> +}
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c
> index 97c36e3c8c80..9f9b1dfb9c37 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c
> @@ -230,7 +230,7 @@ static void event_interrupt_wq_v9(struct kfd_dev *dev,
>  					sq_intr_err);
>  				if (sq_intr_err != SQ_INTERRUPT_ERROR_TYPE_ILLEGAL_INST &&
>  					sq_intr_err != SQ_INTERRUPT_ERROR_TYPE_MEMVIOL) {
> -					kfd_signal_hw_exception_event(pasid);
> +					kfd_signal_poison_consumed_event(dev, pasid);
>  					amdgpu_amdkfd_gpu_reset(dev->kgd);
>  					return;
>  				}
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> index 64552f6b8ba4..daa9d47514c6 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> @@ -1144,6 +1144,8 @@ void kfd_signal_vm_fault_event(struct kfd_dev *dev, u32 pasid,
>  
>  void kfd_signal_reset_event(struct kfd_dev *dev);
>  
> +void kfd_signal_poison_consumed_event(struct kfd_dev *dev, u32 pasid);
> +
>  void kfd_flush_tlb(struct kfd_process_device *pdd);
>  
>  int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process *p);