drm/amdkfd: Track SDMA utilization per process

Track SDMA usage on a per process basis and report it through sysfs.
The value in the sysfs file indicates the amount of time SDMA has
been in-use by this process since the creation of the process.
This value is in microsecond granularity.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

1 files changed, 125 insertions, 12 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index d27221ddcdeb..db010c5da144 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -25,6 +25,7 @@
 #include <linux/sched.h>
 #include <linux/sched/mm.h>
 #include <linux/sched/task.h>
+#include <linux/mmu_context.h>
 #include <linux/slab.h>
 #include <linux/amd-iommu.h>
 #include <linux/notifier.h>
@@ -76,6 +77,74 @@ struct kfd_procfs_tree {
 
 static struct kfd_procfs_tree procfs;
 
+/*
+ * Structure for SDMA activity tracking
+ */
+struct kfd_sdma_activity_handler_workarea {
+	struct work_struct sdma_activity_work;
+	struct kfd_process_device *pdd;
+	uint64_t sdma_activity_counter;
+};
+
+static void kfd_sdma_activity_worker(struct work_struct *work)
+{
+	struct kfd_sdma_activity_handler_workarea *workarea;
+	struct kfd_process_device *pdd;
+	uint64_t val;
+	struct mm_struct *mm;
+	struct queue *q;
+	struct qcm_process_device *qpd;
+	struct device_queue_manager *dqm;
+	int ret = 0;
+
+	workarea = container_of(work, struct kfd_sdma_activity_handler_workarea,
+				sdma_activity_work);
+	if (!workarea)
+		return;
+
+	pdd = workarea->pdd;
+	dqm = pdd->dev->dqm;
+	qpd = &pdd->qpd;
+
+	if (!pdd || !dqm || !qpd)
+		return;
+
+	mm = get_task_mm(pdd->process->lead_thread);
+	if (!mm) {
+		return;
+	}
+
+	use_mm(mm);
+
+	dqm_lock(dqm);
+
+	/*
+	 * Total SDMA activity is current SDMA activity + past SDMA activity
+	 */
+	workarea->sdma_activity_counter = pdd->sdma_past_activity_counter;
+
+	/*
+	 * Get the current activity counters for all active SDMA queues
+	 */
+	list_for_each_entry(q, &qpd->queues_list, list) {
+		if ((q->properties.type == KFD_QUEUE_TYPE_SDMA) ||
+		    (q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI)) {
+			val = 0;
+			ret = read_sdma_queue_counter(q, &val);
+			if (ret)
+				pr_debug("Failed to read SDMA queue active "
+					 "counter for queue id: %d",
+					 q->properties.queue_id);
+			else
+				workarea->sdma_activity_counter += val;
+		}
+	}
+
+	dqm_unlock(dqm);
+	unuse_mm(mm);
+	mmput(mm);
+}
+
 static ssize_t kfd_procfs_show(struct kobject *kobj, struct attribute *attr,
 			       char *buffer)
 {
@@ -87,8 +156,24 @@ static ssize_t kfd_procfs_show(struct kobject *kobj, struct attribute *attr,
 	} else if (strncmp(attr->name, "vram_", 5) == 0) {
 		struct kfd_process_device *pdd = container_of(attr, struct kfd_process_device,
 							      attr_vram);
-		if (pdd)
-			return snprintf(buffer, PAGE_SIZE, "%llu\n", READ_ONCE(pdd->vram_usage));
+		return snprintf(buffer, PAGE_SIZE, "%llu\n", READ_ONCE(pdd->vram_usage));
+	} else if (strncmp(attr->name, "sdma_", 5) == 0) {
+		struct kfd_process_device *pdd = container_of(attr, struct kfd_process_device,
+							      attr_sdma);
+		struct kfd_sdma_activity_handler_workarea sdma_activity_work_handler;
+
+		INIT_WORK(&sdma_activity_work_handler.sdma_activity_work,
+					kfd_sdma_activity_worker);
+
+		sdma_activity_work_handler.pdd = pdd;
+
+		schedule_work(&sdma_activity_work_handler.sdma_activity_work);
+
+		flush_work(&sdma_activity_work_handler.sdma_activity_work);
+
+		return snprintf(buffer, PAGE_SIZE, "%llu\n",
+				(sdma_activity_work_handler.sdma_activity_counter)/
+				 SDMA_ACTIVITY_DIVISOR);
 	} else {
 		pr_err("Invalid attribute");
 		return -EINVAL;
@@ -210,7 +295,24 @@ int kfd_procfs_add_queue(struct queue *q)
 	return 0;
 }
 
-int kfd_procfs_add_vram_usage(struct kfd_process *p)
+static int kfd_sysfs_create_file(struct kfd_process *p, struct attribute *attr,
+				 char *name)
+{
+	int ret = 0;
+
+	if (!p || !attr || !name)
+		return -EINVAL;
+
+	attr->name = name;
+	attr->mode = KFD_SYSFS_FILE_MODE;
+	sysfs_attr_init(attr);
+
+	ret = sysfs_create_file(p->kobj, attr);
+
+	return ret;
+}
+
+int kfd_procfs_add_sysfs_files(struct kfd_process *p)
 {
 	int ret = 0;
 	struct kfd_process_device *pdd;
@@ -221,17 +323,25 @@ int kfd_procfs_add_vram_usage(struct kfd_process *p)
 	if (!p->kobj)
 		return -EFAULT;
 
-	/* Create proc/<pid>/vram_<gpuid> file for each GPU */
+	/*
+	 * Create sysfs files for each GPU:
+	 * - proc/<pid>/vram_<gpuid>
+	 * - proc/<pid>/sdma_<gpuid>
+	 */
 	list_for_each_entry(pdd, &p->per_device_data, per_device_list) {
-		snprintf(pdd->vram_filename, MAX_VRAM_FILENAME_LEN, "vram_%u",
+		snprintf(pdd->vram_filename, MAX_SYSFS_FILENAME_LEN, "vram_%u",
 			 pdd->dev->id);
-		pdd->attr_vram.name = pdd->vram_filename;
-		pdd->attr_vram.mode = KFD_SYSFS_FILE_MODE;
-		sysfs_attr_init(&pdd->attr_vram);
-		ret = sysfs_create_file(p->kobj, &pdd->attr_vram);
+		ret = kfd_sysfs_create_file(p, &pdd->attr_vram, pdd->vram_filename);
 		if (ret)
 			pr_warn("Creating vram usage for gpu id %d failed",
 				(int)pdd->dev->id);
+
+		snprintf(pdd->sdma_filename, MAX_SYSFS_FILENAME_LEN, "sdma_%u",
+			 pdd->dev->id);
+		ret = kfd_sysfs_create_file(p, &pdd->attr_sdma, pdd->sdma_filename);
+		if (ret)
+			pr_warn("Creating sdma usage for gpu id %d failed",
+				(int)pdd->dev->id);
 	}
 
 	return ret;
@@ -444,9 +554,9 @@ struct kfd_process *kfd_create_process(struct file *filep)
 		if (!process->kobj_queues)
 			pr_warn("Creating KFD proc/queues folder failed");
 
-		ret = kfd_procfs_add_vram_usage(process);
+		ret = kfd_procfs_add_sysfs_files(process);
 		if (ret)
-			pr_warn("Creating vram usage file for pid %d failed",
+			pr_warn("Creating sysfs usage file for pid %d failed",
 				(int)process->lead_thread->pid);
 	}
 out:
@@ -597,8 +707,10 @@ static void kfd_process_wq_release(struct work_struct *work)
 		kobject_put(p->kobj_queues);
 		p->kobj_queues = NULL;
 
-		list_for_each_entry(pdd, &p->per_device_data, per_device_list)
+		list_for_each_entry(pdd, &p->per_device_data, per_device_list) {
 			sysfs_remove_file(p->kobj, &pdd->attr_vram);
+			sysfs_remove_file(p->kobj, &pdd->attr_sdma);
+		}
 
 		kobject_del(p->kobj);
 		kobject_put(p->kobj);
@@ -906,6 +1018,7 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev,
 	pdd->already_dequeued = false;
 	pdd->runtime_inuse = false;
 	pdd->vram_usage = 0;
+	pdd->sdma_past_activity_counter = 0;
 	list_add(&pdd->per_device_list, &p->per_device_data);
 
 	/* Init idr used for memory handle translation */