drm/amdgpu: prepare to handle pasid poison consumption

Prepare to handle pasid poison consumption.

Signed-off-by: YiPeng Chai <[email protected]>
Reviewed-by: Tao Zhou <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

Author: YiPeng Chai

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c

@@ -747,10 +747,17 @@ bool amdgpu_amdkfd_is_fed(struct amdgpu_device *adev)
 	return amdgpu_ras_get_fed_status(adev);
 }
 
+void amdgpu_amdkfd_ras_pasid_poison_consumption_handler(struct amdgpu_device *adev,
+				enum amdgpu_ras_block block, uint16_t pasid,
+				pasid_notify pasid_fn, void *data, uint32_t reset)
+{
+	amdgpu_umc_pasid_poison_handler(adev, block, pasid, pasid_fn, data, reset);
+}
+
 void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev,
 	enum amdgpu_ras_block block, uint32_t reset)
 {
-	amdgpu_umc_poison_handler(adev, block, reset);
+	amdgpu_umc_pasid_poison_handler(adev, block, 0, NULL, NULL, reset);
 }
 
 int amdgpu_amdkfd_send_close_event_drain_irq(struct amdgpu_device *adev,

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h

@@ -337,6 +337,11 @@ int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev,
 				struct tile_config *config);
 void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev,
 			enum amdgpu_ras_block block, uint32_t reset);
+
+void amdgpu_amdkfd_ras_pasid_poison_consumption_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint16_t pasid,
+			pasid_notify pasid_fn, void *data, uint32_t reset);
+
 bool amdgpu_amdkfd_is_fed(struct amdgpu_device *adev);
 bool amdgpu_amdkfd_bo_mapped_to_dev(struct amdgpu_device *adev, struct kgd_mem *mem);
 void amdgpu_amdkfd_block_mmu_notifications(void *p);

drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c

@@ -252,8 +252,9 @@ int amdgpu_umc_bad_page_polling_timeout(struct amdgpu_device *adev,
 	return 0;
 }
 
-int amdgpu_umc_poison_handler(struct amdgpu_device *adev,
-			enum amdgpu_ras_block block, uint32_t reset)
+int amdgpu_umc_pasid_poison_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint16_t pasid,
+			pasid_notify pasid_fn, void *data, uint32_t reset)
 {
 	int ret = AMDGPU_RAS_SUCCESS;
 
@@ -291,16 +292,14 @@ int amdgpu_umc_poison_handler(struct amdgpu_device *adev,
 
 			amdgpu_ras_error_data_fini(&err_data);
 		} else {
-			if (reset) {
-				amdgpu_umc_bad_page_polling_timeout(adev,
-							reset, MAX_UMC_POISON_POLLING_TIME_SYNC);
-			} else {
 				struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 
+				amdgpu_ras_put_poison_req(adev,
+					block, pasid, pasid_fn, data, reset);
+
 				atomic_inc(&con->page_retirement_req_cnt);
 
 				wake_up(&con->page_retirement_wq);
-			}
 		}
 	} else {
 		if (adev->virt.ops && adev->virt.ops->ras_poison_handler)
@@ -313,6 +312,13 @@ int amdgpu_umc_poison_handler(struct amdgpu_device *adev,
 	return ret;
 }
 
+int amdgpu_umc_poison_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint32_t reset)
+{
+	return amdgpu_umc_pasid_poison_handler(adev,
+				block, 0, NULL, NULL, reset);
+}
+
 int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev,
 		void *ras_error_status,
 		struct amdgpu_iv_entry *entry)

drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h

@@ -106,6 +106,9 @@ int amdgpu_umc_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_umc_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block);
 int amdgpu_umc_poison_handler(struct amdgpu_device *adev,
 			enum amdgpu_ras_block block, uint32_t reset);
+int amdgpu_umc_pasid_poison_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint16_t pasid,
+			pasid_notify pasid_fn, void *data, uint32_t reset);
 int amdgpu_umc_process_ecc_irq(struct amdgpu_device *adev,
 		struct amdgpu_irq_src *source,
 		struct amdgpu_iv_entry *entry);

drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c

@@ -190,7 +190,8 @@ static void event_interrupt_poison_consumption_v9(struct kfd_node *dev,
 	dev_warn(dev->adev->dev,
 		 "poison is consumed by client %d, kick off gpu reset flow\n", client_id);
 
-	amdgpu_amdkfd_ras_poison_consumption_handler(dev->adev, block, reset);
+	amdgpu_amdkfd_ras_pasid_poison_consumption_handler(dev->adev,
+		block, pasid, NULL, NULL, reset);
 }
 
 static bool context_id_expected(struct kfd_dev *dev)