drm/amdkfd: Check preemption status on all XCDs

This patch adds the following functionality:
- Check the queue preemption status on all XCDs in a partition
for GFX 9.4.3.
- Update the queue preemption debug message to print the queue
doorbell id for which preemption failed.
- Change the signature of check preemption failed function to
return a bool instead of uint32_t and pass the MQD manager
as an argument.

Suggested-by: Jay Cornwall <[e-mail]>
Signed-off-by: Mukul Joshi <[e-mail]>
Reviewed-by: Felix Kuehling <[e-mail]>
Signed-off-by: Alex Deucher <[e-mail]>

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

···
 	 * check those fields
 	 */
 	mqd_mgr = dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ];
-	if (mqd_mgr->check_preemption_failed(dqm->packet_mgr.priv_queue->queue->mqd)) {
-		dev_err(dev, "HIQ MQD's queue_doorbell_id0 is not 0, Queue preemption time out\n");
+	if (mqd_mgr->check_preemption_failed(mqd_mgr, dqm->packet_mgr.priv_queue->queue->mqd)) {
 		while (halt_if_hws_hang)
 			schedule();
 		return -ETIME;

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

···
 {
 	return mm->mqd_size;
 }
+
+bool kfd_check_hiq_mqd_doorbell_id(struct kfd_node *node, uint32_t doorbell_id,
+				   uint32_t inst)
+{
+	if (doorbell_id) {
+		struct device *dev = node->adev->dev;
+
+		if (node->adev->xcp_mgr && node->adev->xcp_mgr->num_xcps > 0)
+			dev_err(dev, "XCC %d: Queue preemption failed for queue with doorbell_id: %x\n",
+							inst, doorbell_id);
+		else
+			dev_err(dev, "Queue preemption failed for queue with doorbell_id: %x\n",
+							doorbell_id);
+		return true;
+	}
+
+	return false;
+}

drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h

···
 #if defined(CONFIG_DEBUG_FS)
 	int	(*debugfs_show_mqd)(struct seq_file *m, void *data);
 #endif
-	uint32_t (*check_preemption_failed)(void *mqd);
+	bool (*check_preemption_failed)(struct mqd_manager *mm, void *mqd);
 	uint64_t (*mqd_stride)(struct mqd_manager *mm,
 				struct queue_properties *p);
 
···
 uint64_t kfd_hiq_mqd_stride(struct kfd_node *dev);
 uint64_t kfd_mqd_stride(struct mqd_manager *mm,
 			struct queue_properties *q);
+bool kfd_check_hiq_mqd_doorbell_id(struct kfd_node *node, uint32_t doorbell_id,
+				   uint32_t inst);
 #endif /* KFD_MQD_MANAGER_H_ */

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

···
 	q->is_active = QUEUE_IS_ACTIVE(*q);
 }
 
-static uint32_t check_preemption_failed(void *mqd)
+static bool check_preemption_failed(struct mqd_manager *mm, void *mqd)
 {
 	struct cik_mqd *m = (struct cik_mqd *)mqd;
 
-	return m->queue_doorbell_id0;
+	return kfd_check_hiq_mqd_doorbell_id(mm->dev, m->queue_doorbell_id0, 0);
 }
 
 static void update_mqd(struct mqd_manager *mm, void *mqd,

drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c

···
 	q->is_active = QUEUE_IS_ACTIVE(*q);
 }
 
-static uint32_t check_preemption_failed(void *mqd)
+static bool check_preemption_failed(struct mqd_manager *mm, void *mqd)
 {
 	struct v10_compute_mqd *m = (struct v10_compute_mqd *)mqd;
 
-	return m->queue_doorbell_id0;
+	return kfd_check_hiq_mqd_doorbell_id(mm->dev, m->queue_doorbell_id0, 0);
 }
 
 static int get_wave_state(struct mqd_manager *mm, void *mqd,

drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c

···
 	q->is_active = QUEUE_IS_ACTIVE(*q);
 }
 
-static uint32_t check_preemption_failed(void *mqd)
+static bool check_preemption_failed(struct mqd_manager *mm, void *mqd)
 {
 	struct v11_compute_mqd *m = (struct v11_compute_mqd *)mqd;
 
-	return m->queue_doorbell_id0;
+	return kfd_check_hiq_mqd_doorbell_id(mm->dev, m->queue_doorbell_id0, 0);
 }
 
 static int get_wave_state(struct mqd_manager *mm, void *mqd,

drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c

···
 }
 
 
-static uint32_t check_preemption_failed(void *mqd)
+static bool check_preemption_failed(struct mqd_manager *mm, void *mqd)
 {
 	struct v9_mqd *m = (struct v9_mqd *)mqd;
 
-	return m->queue_doorbell_id0;
+	return kfd_check_hiq_mqd_doorbell_id(mm->dev, m->queue_doorbell_id0, 0);
 }
 
 static int get_wave_state(struct mqd_manager *mm, void *mqd,
···
 	return err;
 }
 
+static bool check_preemption_failed_v9_4_3(struct mqd_manager *mm, void *mqd)
+{
+	uint64_t hiq_mqd_size = kfd_hiq_mqd_stride(mm->dev);
+	uint32_t xcc_mask = mm->dev->xcc_mask;
+	int inst = 0, xcc_id;
+	struct v9_mqd *m;
+	bool ret = false;
+
+	for_each_inst(xcc_id, xcc_mask) {
+		m = get_mqd(mqd + hiq_mqd_size * inst);
+		ret |= kfd_check_hiq_mqd_doorbell_id(mm->dev,
+					m->queue_doorbell_id0, inst);
+		++inst;
+	}
+
+	return ret;
+}
+
 static void get_xcc_mqd(struct kfd_mem_obj *mqd_mem_obj,
 			       struct kfd_mem_obj *xcc_mqd_mem_obj,
 			       uint64_t offset)
···
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
-		mqd->check_preemption_failed = check_preemption_failed;
 		if (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3)) {
 			mqd->init_mqd = init_mqd_hiq_v9_4_3;
 			mqd->load_mqd = hiq_load_mqd_kiq_v9_4_3;
 			mqd->destroy_mqd = destroy_hiq_mqd_v9_4_3;
+			mqd->check_preemption_failed = check_preemption_failed_v9_4_3;
 		} else {
 			mqd->init_mqd = init_mqd_hiq;
 			mqd->load_mqd = kfd_hiq_load_mqd_kiq;
 			mqd->destroy_mqd = destroy_hiq_mqd;
+			mqd->check_preemption_failed = check_preemption_failed;
 		}
 		break;
 	case KFD_MQD_TYPE_DIQ:

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

···
 	q->is_active = QUEUE_IS_ACTIVE(*q);
 }
 
-static uint32_t check_preemption_failed(void *mqd)
+static bool check_preemption_failed(struct mqd_manager *mm, void *mqd)
 {
 	struct vi_mqd *m = (struct vi_mqd *)mqd;
 
-	return m->queue_doorbell_id0;
+	return kfd_check_hiq_mqd_doorbell_id(mm->dev, m->queue_doorbell_id0, 0);
 }
 
 static void update_mqd(struct mqd_manager *mm, void *mqd,