Commit 80c74918 authored by Asad Kamal's avatar Asad Kamal Committed by Alex Deucher
Browse files

drm/amdkfd: Replace pr_err with dev_err 



Replace pr_err with dev_err to show the bus-id of
failing device with kfd queue errors

Signed-off-by: default avatarAsad Kamal <asad.kamal@amd.com>
Reviewed-by: default avatarLijo Lazar <lijo.lazar@amd.com>
Reviewed-by: default avatarFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
parent ee78ef04

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

+70 −46
Original line number Diff line number Diff line
@@ -232,7 +232,7 @@ static int add_queue_mes(struct device_queue_manager *dqm, struct queue *q, 


	queue_type = convert_to_mes_queue_type(q->properties.type);

	if (queue_type < 0) {

		pr_err("Queue type not supported with MES, queue:%d\n",

		dev_err(adev->dev, "Queue type not supported with MES, queue:%d\n",

			q->properties.type);

		return -EINVAL;

	}
@@ -244,9 +244,9 @@ static int add_queue_mes(struct device_queue_manager *dqm, struct queue *q, 
	r = adev->mes.funcs->add_hw_queue(&adev->mes, &queue_input);

	amdgpu_mes_unlock(&adev->mes);

	if (r) {

		pr_err("failed to add hardware queue to MES, doorbell=0x%x\n",

		dev_err(adev->dev, "failed to add hardware queue to MES, doorbell=0x%x\n",

			q->properties.doorbell_off);

		pr_err("MES might be in unrecoverable state, issue a GPU reset\n");

		dev_err(adev->dev, "MES might be in unrecoverable state, issue a GPU reset\n");

		kfd_hws_hang(dqm);

	}
@@ -272,9 +272,9 @@ static int remove_queue_mes(struct device_queue_manager *dqm, struct queue *q, 
	amdgpu_mes_unlock(&adev->mes);



	if (r) {

		pr_err("failed to remove hardware queue from MES, doorbell=0x%x\n",

		dev_err(adev->dev, "failed to remove hardware queue from MES, doorbell=0x%x\n",

			q->properties.doorbell_off);

		pr_err("MES might be in unrecoverable state, issue a GPU reset\n");

		dev_err(adev->dev, "MES might be in unrecoverable state, issue a GPU reset\n");

		kfd_hws_hang(dqm);

	}
@@ -284,6 +284,7 @@ static int remove_queue_mes(struct device_queue_manager *dqm, struct queue *q, 
static int remove_all_queues_mes(struct device_queue_manager *dqm)

{

	struct device_process_node *cur;

	struct device *dev = dqm->dev->adev->dev;

	struct qcm_process_device *qpd;

	struct queue *q;

	int retval = 0;
@@ -294,7 +295,7 @@ static int remove_all_queues_mes(struct device_queue_manager *dqm) 
			if (q->properties.is_active) {

				retval = remove_queue_mes(dqm, q, qpd);

				if (retval) {

					pr_err("%s: Failed to remove queue %d for dev %d",

					dev_err(dev, "%s: Failed to remove queue %d for dev %d",

						__func__,

						q->properties.queue_id,

						dqm->dev->id);
@@ -443,6 +444,7 @@ static int allocate_vmid(struct device_queue_manager *dqm, 
			struct qcm_process_device *qpd,

			struct queue *q)

{

	struct device *dev = dqm->dev->adev->dev;

	int allocated_vmid = -1, i;



	for (i = dqm->dev->vm_info.first_vmid_kfd;
@@ -454,7 +456,7 @@ static int allocate_vmid(struct device_queue_manager *dqm, 
	}



	if (allocated_vmid < 0) {

		pr_err("no more vmid to allocate\n");

		dev_err(dev, "no more vmid to allocate\n");

		return -ENOSPC;

	}
@@ -510,10 +512,12 @@ static void deallocate_vmid(struct device_queue_manager *dqm, 
				struct qcm_process_device *qpd,

				struct queue *q)

{

	struct device *dev = dqm->dev->adev->dev;



	/* On GFX v7, CP doesn't flush TC at dequeue */

	if (q->device->adev->asic_type == CHIP_HAWAII)

		if (flush_texture_cache_nocpsch(q->device, qpd))

			pr_err("Failed to flush TC\n");

			dev_err(dev, "Failed to flush TC\n");



	kfd_flush_tlb(qpd_to_pdd(qpd), TLB_FLUSH_LEGACY);
@@ -708,7 +712,7 @@ static int dbgdev_wave_reset_wavefronts(struct kfd_node *dev, struct kfd_process 
	pr_debug("Killing all process wavefronts\n");



	if (!dev->kfd2kgd->get_atc_vmid_pasid_mapping_info) {

		pr_err("no vmid pasid mapping supported \n");

		dev_err(dev->adev->dev, "no vmid pasid mapping supported\n");

		return -EOPNOTSUPP;

	}
@@ -729,7 +733,7 @@ static int dbgdev_wave_reset_wavefronts(struct kfd_node *dev, struct kfd_process 
	}



	if (vmid > last_vmid_to_scan) {

		pr_err("Didn't find vmid for pasid 0x%x\n", p->pasid);

		dev_err(dev->adev->dev, "Didn't find vmid for pasid 0x%x\n", p->pasid);

		return -EFAULT;

	}
@@ -821,6 +825,7 @@ static int destroy_queue_nocpsch(struct device_queue_manager *dqm, 
{

	int retval;

	uint64_t sdma_val = 0;

	struct device *dev = dqm->dev->adev->dev;

	struct kfd_process_device *pdd = qpd_to_pdd(qpd);

	struct mqd_manager *mqd_mgr =

		dqm->mqd_mgrs[get_mqd_type_from_queue_type(q->properties.type)];
@@ -831,7 +836,7 @@ static int destroy_queue_nocpsch(struct device_queue_manager *dqm, 
		retval = read_sdma_queue_counter((uint64_t __user *)q->properties.read_ptr,

							&sdma_val);

		if (retval)

			pr_err("Failed to read SDMA queue counter for queue: %d\n",

			dev_err(dev, "Failed to read SDMA queue counter for queue: %d\n",

				q->properties.queue_id);

	}
@@ -850,6 +855,7 @@ static int update_queue(struct device_queue_manager *dqm, struct queue *q, 
			struct mqd_update_info *minfo)

{

	int retval = 0;

	struct device *dev = dqm->dev->adev->dev;

	struct mqd_manager *mqd_mgr;

	struct kfd_process_device *pdd;

	bool prev_active = false;
@@ -875,7 +881,7 @@ static int update_queue(struct device_queue_manager *dqm, struct queue *q, 
			retval = remove_queue_mes(dqm, q, &pdd->qpd);



		if (retval) {

			pr_err("unmap queue failed\n");

			dev_err(dev, "unmap queue failed\n");

			goto out_unlock;

		}

	} else if (prev_active &&
@@ -894,7 +900,7 @@ static int update_queue(struct device_queue_manager *dqm, struct queue *q, 
				 KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN),

				KFD_UNMAP_LATENCY_MS, q->pipe, q->queue);

		if (retval) {

			pr_err("destroy mqd failed\n");

			dev_err(dev, "destroy mqd failed\n");

			goto out_unlock;

		}

	}
@@ -1088,6 +1094,7 @@ static int evict_process_queues_cpsch(struct device_queue_manager *dqm, 
				      struct qcm_process_device *qpd)

{

	struct queue *q;

	struct device *dev = dqm->dev->adev->dev;

	struct kfd_process_device *pdd;

	int retval = 0;
@@ -1121,7 +1128,7 @@ static int evict_process_queues_cpsch(struct device_queue_manager *dqm, 
		if (dqm->dev->kfd->shared_resources.enable_mes) {

			retval = remove_queue_mes(dqm, q, qpd);

			if (retval) {

				pr_err("Failed to evict queue %d\n",

				dev_err(dev, "Failed to evict queue %d\n",

					q->properties.queue_id);

				goto out;

			}
@@ -1225,6 +1232,7 @@ static int restore_process_queues_cpsch(struct device_queue_manager *dqm, 
					struct qcm_process_device *qpd)

{

	struct queue *q;

	struct device *dev = dqm->dev->adev->dev;

	struct kfd_process_device *pdd;

	uint64_t eviction_duration;

	int retval = 0;
@@ -1265,7 +1273,7 @@ static int restore_process_queues_cpsch(struct device_queue_manager *dqm, 
		if (dqm->dev->kfd->shared_resources.enable_mes) {

			retval = add_queue_mes(dqm, q, qpd);

			if (retval) {

				pr_err("Failed to restore queue %d\n",

				dev_err(dev, "Failed to restore queue %d\n",

					q->properties.queue_id);

				goto out;

			}
@@ -1474,18 +1482,19 @@ static void pre_reset(struct device_queue_manager *dqm) 
static int allocate_sdma_queue(struct device_queue_manager *dqm,

				struct queue *q, const uint32_t *restore_sdma_id)

{

	struct device *dev = dqm->dev->adev->dev;

	int bit;



	if (q->properties.type == KFD_QUEUE_TYPE_SDMA) {

		if (bitmap_empty(dqm->sdma_bitmap, KFD_MAX_SDMA_QUEUES)) {

			pr_err("No more SDMA queue to allocate\n");

			dev_err(dev, "No more SDMA queue to allocate\n");

			return -ENOMEM;

		}



		if (restore_sdma_id) {

			/* Re-use existing sdma_id */

			if (!test_bit(*restore_sdma_id, dqm->sdma_bitmap)) {

				pr_err("SDMA queue already in use\n");

				dev_err(dev, "SDMA queue already in use\n");

				return -EBUSY;

			}

			clear_bit(*restore_sdma_id, dqm->sdma_bitmap);
@@ -1504,13 +1513,13 @@ static int allocate_sdma_queue(struct device_queue_manager *dqm, 
				kfd_get_num_sdma_engines(dqm->dev);

	} else if (q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI) {

		if (bitmap_empty(dqm->xgmi_sdma_bitmap, KFD_MAX_SDMA_QUEUES)) {

			pr_err("No more XGMI SDMA queue to allocate\n");

			dev_err(dev, "No more XGMI SDMA queue to allocate\n");

			return -ENOMEM;

		}

		if (restore_sdma_id) {

			/* Re-use existing sdma_id */

			if (!test_bit(*restore_sdma_id, dqm->xgmi_sdma_bitmap)) {

				pr_err("SDMA queue already in use\n");

				dev_err(dev, "SDMA queue already in use\n");

				return -EBUSY;

			}

			clear_bit(*restore_sdma_id, dqm->xgmi_sdma_bitmap);
@@ -1562,6 +1571,7 @@ static int set_sched_resources(struct device_queue_manager *dqm) 
{

	int i, mec;

	struct scheduling_resources res;

	struct device *dev = dqm->dev->adev->dev;



	res.vmid_mask = dqm->dev->compute_vmid_bitmap;
@@ -1582,7 +1592,7 @@ static int set_sched_resources(struct device_queue_manager *dqm) 
		 * definition of res.queue_mask needs updating

		 */

		if (WARN_ON(i >= (sizeof(res.queue_mask)*8))) {

			pr_err("Invalid queue enabled by amdgpu: %d\n", i);

			dev_err(dev, "Invalid queue enabled by amdgpu: %d\n", i);

			break;

		}
@@ -1625,6 +1635,7 @@ static int initialize_cpsch(struct device_queue_manager *dqm) 


static int start_cpsch(struct device_queue_manager *dqm)

{

	struct device *dev = dqm->dev->adev->dev;

	int retval;



	retval = 0;
@@ -1671,7 +1682,7 @@ static int start_cpsch(struct device_queue_manager *dqm) 
		retval = pm_update_grace_period(&dqm->packet_mgr,

						grace_period);

		if (retval)

			pr_err("Setting grace timeout failed\n");

			dev_err(dev, "Setting grace timeout failed\n");

		else if (dqm->dev->kfd2kgd->build_grace_period_packet_info)

			/* Update dqm->wait_times maintained in software */

			dqm->dev->kfd2kgd->build_grace_period_packet_info(
@@ -1881,15 +1892,17 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q, 
	return retval;

}



int amdkfd_fence_wait_timeout(uint64_t *fence_addr,

int amdkfd_fence_wait_timeout(struct device_queue_manager *dqm,

			      uint64_t fence_value,

			      unsigned int timeout_ms)

{

	unsigned long end_jiffies = msecs_to_jiffies(timeout_ms) + jiffies;

	struct device *dev = dqm->dev->adev->dev;

	uint64_t *fence_addr =  dqm->fence_addr;



	while (*fence_addr != fence_value) {

		if (time_after(jiffies, end_jiffies)) {

			pr_err("qcm fence wait loop timeout expired\n");

			dev_err(dev, "qcm fence wait loop timeout expired\n");

			/* In HWS case, this is used to halt the driver thread

			 * in order not to mess up CP states before doing

			 * scandumps for FW debugging.
@@ -1908,6 +1921,7 @@ int amdkfd_fence_wait_timeout(uint64_t *fence_addr, 
/* dqm->lock mutex has to be locked before calling this function */

static int map_queues_cpsch(struct device_queue_manager *dqm)

{

	struct device *dev = dqm->dev->adev->dev;

	int retval;



	if (!dqm->sched_running)
@@ -1920,7 +1934,7 @@ static int map_queues_cpsch(struct device_queue_manager *dqm) 
	retval = pm_send_runlist(&dqm->packet_mgr, &dqm->queues);

	pr_debug("%s sent runlist\n", __func__);

	if (retval) {

		pr_err("failed to execute runlist\n");

		dev_err(dev, "failed to execute runlist\n");

		return retval;

	}

	dqm->active_runlist = true;
@@ -1935,8 +1949,9 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm, 
				uint32_t grace_period,

				bool reset)

{

	int retval = 0;

	struct device *dev = dqm->dev->adev->dev;

	struct mqd_manager *mqd_mgr;

	int retval = 0;



	if (!dqm->sched_running)

		return 0;
@@ -1959,10 +1974,10 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm, 
	pm_send_query_status(&dqm->packet_mgr, dqm->fence_gpu_addr,

				KFD_FENCE_COMPLETED);

	/* should be timed out */

	retval = amdkfd_fence_wait_timeout(dqm->fence_addr, KFD_FENCE_COMPLETED,

	retval = amdkfd_fence_wait_timeout(dqm, KFD_FENCE_COMPLETED,

					   queue_preemption_timeout_ms);

	if (retval) {

		pr_err("The cp might be in an unrecoverable state due to an unsuccessful queues preemption\n");

		dev_err(dev, "The cp might be in an unrecoverable state due to an unsuccessful queues preemption\n");

		kfd_hws_hang(dqm);

		return retval;

	}
@@ -1977,7 +1992,7 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm, 
	 */

	mqd_mgr = dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ];

	if (mqd_mgr->read_doorbell_id(dqm->packet_mgr.priv_queue->queue->mqd)) {

		pr_err("HIQ MQD's queue_doorbell_id0 is not 0, Queue preemption time out\n");

		dev_err(dev, "HIQ MQD's queue_doorbell_id0 is not 0, Queue preemption time out\n");

		while (halt_if_hws_hang)

			schedule();

		return -ETIME;
@@ -1987,7 +2002,7 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm, 
	if (grace_period != USE_DEFAULT_GRACE_PERIOD) {

		if (pm_update_grace_period(&dqm->packet_mgr,

					USE_DEFAULT_GRACE_PERIOD))

			pr_err("Failed to reset grace period\n");

			dev_err(dev, "Failed to reset grace period\n");

	}



	pm_release_ib(&dqm->packet_mgr);
@@ -2061,6 +2076,7 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm, 
	struct mqd_manager *mqd_mgr;

	uint64_t sdma_val = 0;

	struct kfd_process_device *pdd = qpd_to_pdd(qpd);

	struct device *dev = dqm->dev->adev->dev;



	/* Get the SDMA queue stats */

	if ((q->properties.type == KFD_QUEUE_TYPE_SDMA) ||
@@ -2068,7 +2084,7 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm, 
		retval = read_sdma_queue_counter((uint64_t __user *)q->properties.read_ptr,

							&sdma_val);

		if (retval)

			pr_err("Failed to read SDMA queue counter for queue: %d\n",

			dev_err(dev, "Failed to read SDMA queue counter for queue: %d\n",

				q->properties.queue_id);

	}
@@ -2349,6 +2365,7 @@ static int process_termination_cpsch(struct device_queue_manager *dqm, 
{

	int retval;

	struct queue *q;

	struct device *dev = dqm->dev->adev->dev;

	struct kernel_queue *kq, *kq_next;

	struct mqd_manager *mqd_mgr;

	struct device_process_node *cur, *next_dpn;
@@ -2382,7 +2399,7 @@ static int process_termination_cpsch(struct device_queue_manager *dqm, 
			if (dqm->dev->kfd->shared_resources.enable_mes) {

				retval = remove_queue_mes(dqm, q, qpd);

				if (retval)

					pr_err("Failed to remove queue %d\n",

					dev_err(dev, "Failed to remove queue %d\n",

						q->properties.queue_id);

			}

		}
@@ -2437,12 +2454,13 @@ static int process_termination_cpsch(struct device_queue_manager *dqm, 
static int init_mqd_managers(struct device_queue_manager *dqm)

{

	int i, j;

	struct device *dev = dqm->dev->adev->dev;

	struct mqd_manager *mqd_mgr;



	for (i = 0; i < KFD_MQD_TYPE_MAX; i++) {

		mqd_mgr = dqm->asic_ops.mqd_manager_init(i, dqm->dev);

		if (!mqd_mgr) {

			pr_err("mqd manager [%d] initialization failed\n", i);

			dev_err(dev, "mqd manager [%d] initialization failed\n", i);

			goto out_free;

		}

		dqm->mqd_mgrs[i] = mqd_mgr;
@@ -2552,7 +2570,7 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_node *dev) 
		dqm->ops.checkpoint_mqd = checkpoint_mqd;

		break;

	default:

		pr_err("Invalid scheduling policy %d\n", dqm->sched_policy);

		dev_err(dev->adev->dev, "Invalid scheduling policy %d\n", dqm->sched_policy);

		goto out_free;

	}
@@ -2590,7 +2608,7 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_node *dev) 
		goto out_free;



	if (!dev->kfd->shared_resources.enable_mes && allocate_hiq_sdma_mqd(dqm)) {

		pr_err("Failed to allocate hiq sdma mqd trunk buffer\n");

		dev_err(dev->adev->dev, "Failed to allocate hiq sdma mqd trunk buffer\n");

		goto out_free;

	}
@@ -2649,17 +2667,18 @@ int reserve_debug_trap_vmid(struct device_queue_manager *dqm, 
				struct qcm_process_device *qpd)

{

	int r;

	struct device *dev = dqm->dev->adev->dev;

	int updated_vmid_mask;



	if (dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) {

		pr_err("Unsupported on sched_policy: %i\n", dqm->sched_policy);

		dev_err(dev, "Unsupported on sched_policy: %i\n", dqm->sched_policy);

		return -EINVAL;

	}



	dqm_lock(dqm);



	if (dqm->trap_debug_vmid != 0) {

		pr_err("Trap debug id already reserved\n");

		dev_err(dev, "Trap debug id already reserved\n");

		r = -EBUSY;

		goto out_unlock;

	}
@@ -2695,19 +2714,20 @@ int reserve_debug_trap_vmid(struct device_queue_manager *dqm, 
int release_debug_trap_vmid(struct device_queue_manager *dqm,

			struct qcm_process_device *qpd)

{

	struct device *dev = dqm->dev->adev->dev;

	int r;

	int updated_vmid_mask;

	uint32_t trap_debug_vmid;



	if (dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) {

		pr_err("Unsupported on sched_policy: %i\n", dqm->sched_policy);

		dev_err(dev, "Unsupported on sched_policy: %i\n", dqm->sched_policy);

		return -EINVAL;

	}



	dqm_lock(dqm);

	trap_debug_vmid = dqm->trap_debug_vmid;

	if (dqm->trap_debug_vmid == 0) {

		pr_err("Trap debug id is not reserved\n");

		dev_err(dev, "Trap debug id is not reserved\n");

		r = -EINVAL;

		goto out_unlock;

	}
@@ -2844,6 +2864,7 @@ int resume_queues(struct kfd_process *p, 
	for (i = 0; i < p->n_pdds; i++) {

		struct kfd_process_device *pdd = p->pdds[i];

		struct device_queue_manager *dqm = pdd->dev->dqm;

		struct device *dev = dqm->dev->adev->dev;

		struct qcm_process_device *qpd = &pdd->qpd;

		struct queue *q;

		int r, per_device_resumed = 0;
@@ -2894,7 +2915,7 @@ int resume_queues(struct kfd_process *p, 
					0,

					USE_DEFAULT_GRACE_PERIOD);

		if (r) {

			pr_err("Failed to resume process queues\n");

			dev_err(dev, "Failed to resume process queues\n");

			if (queue_ids) {

				list_for_each_entry(q, &qpd->queues_list, list) {

					int q_idx = q_array_get_index(
@@ -2946,6 +2967,7 @@ int suspend_queues(struct kfd_process *p, 
	for (i = 0; i < p->n_pdds; i++) {

		struct kfd_process_device *pdd = p->pdds[i];

		struct device_queue_manager *dqm = pdd->dev->dqm;

		struct device *dev = dqm->dev->adev->dev;

		struct qcm_process_device *qpd = &pdd->qpd;

		struct queue *q;

		int r, per_device_suspended = 0;
@@ -2994,7 +3016,7 @@ int suspend_queues(struct kfd_process *p, 
			grace_period);



		if (r)

			pr_err("Failed to suspend process queues.\n");

			dev_err(dev, "Failed to suspend process queues.\n");

		else

			total_suspended += per_device_suspended;
@@ -3081,10 +3103,11 @@ void set_queue_snapshot_entry(struct queue *q, 


int debug_lock_and_unmap(struct device_queue_manager *dqm)

{

	struct device *dev = dqm->dev->adev->dev;

	int r;



	if (dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) {

		pr_err("Unsupported on sched_policy: %i\n", dqm->sched_policy);

		dev_err(dev, "Unsupported on sched_policy: %i\n", dqm->sched_policy);

		return -EINVAL;

	}
@@ -3102,10 +3125,11 @@ int debug_lock_and_unmap(struct device_queue_manager *dqm) 


int debug_map_and_unlock(struct device_queue_manager *dqm)

{

	struct device *dev = dqm->dev->adev->dev;

	int r;



	if (dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) {

		pr_err("Unsupported on sched_policy: %i\n", dqm->sched_policy);

		dev_err(dev, "Unsupported on sched_policy: %i\n", dqm->sched_policy);

		return -EINVAL;

	}

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

+1 −1
Original line number Diff line number Diff line
@@ -1343,7 +1343,7 @@ int pqm_get_queue_snapshot(struct process_queue_manager *pqm, 
			   int *num_qss_entries,

			   uint32_t *entry_size);



int amdkfd_fence_wait_timeout(uint64_t *fence_addr,

int amdkfd_fence_wait_timeout(struct device_queue_manager *dqm,

			      uint64_t fence_value,

			      unsigned int timeout_ms);