Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit 1679ae8f authored by Felix Kuehling's avatar Felix Kuehling Committed by Oded Gabbay
Browse files

drm/amdkfd: Use ordered workqueue to restore processes 



Restoring multiple processes concurrently can lead to live-locks
where each process prevents the other from validating all its BOs.

v2: fix duplicate check of same variable

Signed-off-by: default avatarFelix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: default avatarOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: default avatarOded Gabbay <oded.gabbay@gmail.com>
parent 810955ba

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

+5 −1
Original line number Original line Diff line number Diff line
@@ -133,7 +133,9 @@ static int __init kfd_module_init(void) 
	if (err < 0)

	if (err < 0)

		goto err_topology;

		goto err_topology;





	kfd_process_create_wq();

	err = kfd_process_create_wq();

	if (err < 0)

		goto err_create_wq;





	kfd_debugfs_init();

	kfd_debugfs_init();
@@ -143,6 +145,8 @@ static int __init kfd_module_init(void) 




	return 0;

	return 0;





err_create_wq:

	kfd_topology_shutdown();

err_topology:

err_topology:

	kfd_chardev_exit();

	kfd_chardev_exit();

err_ioctl:

err_ioctl:

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

+1 −1
Original line number Original line Diff line number Diff line
@@ -674,7 +674,7 @@ struct amdkfd_ioctl_desc { 
	const char *name;

	const char *name;

};

};





void kfd_process_create_wq(void);

int kfd_process_create_wq(void);

void kfd_process_destroy_wq(void);

void kfd_process_destroy_wq(void);

struct kfd_process *kfd_create_process(struct file *filep);

struct kfd_process *kfd_create_process(struct file *filep);

struct kfd_process *kfd_get_process(const struct task_struct *);

struct kfd_process *kfd_get_process(const struct task_struct *);

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

+26 −4
Original line number Original line Diff line number Diff line
@@ -48,8 +48,17 @@ static DEFINE_MUTEX(kfd_processes_mutex); 




DEFINE_SRCU(kfd_processes_srcu);

DEFINE_SRCU(kfd_processes_srcu);





/* For process termination handling */

static struct workqueue_struct *kfd_process_wq;

static struct workqueue_struct *kfd_process_wq;





/* Ordered, single-threaded workqueue for restoring evicted

 * processes. Restoring multiple processes concurrently under memory

 * pressure can lead to processes blocking each other from validating

 * their BOs and result in a live-lock situation where processes

 * remain evicted indefinitely.

 */

static struct workqueue_struct *kfd_restore_wq;



static struct kfd_process *find_process(const struct task_struct *thread);

static struct kfd_process *find_process(const struct task_struct *thread);

static void kfd_process_ref_release(struct kref *ref);

static void kfd_process_ref_release(struct kref *ref);

static struct kfd_process *create_process(const struct task_struct *thread,

static struct kfd_process *create_process(const struct task_struct *thread,
@@ -59,10 +68,19 @@ static void evict_process_worker(struct work_struct *work); 
static void restore_process_worker(struct work_struct *work);

static void restore_process_worker(struct work_struct *work);









void kfd_process_create_wq(void)

int kfd_process_create_wq(void)

{

{

	if (!kfd_process_wq)

	if (!kfd_process_wq)

		kfd_process_wq = alloc_workqueue("kfd_process_wq", 0, 0);

		kfd_process_wq = alloc_workqueue("kfd_process_wq", 0, 0);

	if (!kfd_restore_wq)

		kfd_restore_wq = alloc_ordered_workqueue("kfd_restore_wq", 0);



	if (!kfd_process_wq || !kfd_restore_wq) {

		kfd_process_destroy_wq();

		return -ENOMEM;

	}



	return 0;

}

}





void kfd_process_destroy_wq(void)

void kfd_process_destroy_wq(void)
@@ -71,6 +89,10 @@ void kfd_process_destroy_wq(void) 
		destroy_workqueue(kfd_process_wq);

		destroy_workqueue(kfd_process_wq);

		kfd_process_wq = NULL;

		kfd_process_wq = NULL;

	}

	}

	if (kfd_restore_wq) {

		destroy_workqueue(kfd_restore_wq);

		kfd_restore_wq = NULL;

	}

}

}





static void kfd_process_free_gpuvm(struct kgd_mem *mem,

static void kfd_process_free_gpuvm(struct kgd_mem *mem,
@@ -869,7 +891,7 @@ static void evict_process_worker(struct work_struct *work) 
		dma_fence_signal(p->ef);

		dma_fence_signal(p->ef);

		dma_fence_put(p->ef);

		dma_fence_put(p->ef);

		p->ef = NULL;

		p->ef = NULL;

		schedule_delayed_work(&p->restore_work,

		queue_delayed_work(kfd_restore_wq, &p->restore_work,

				msecs_to_jiffies(PROCESS_RESTORE_TIME_MS));

				msecs_to_jiffies(PROCESS_RESTORE_TIME_MS));





		pr_debug("Finished evicting pasid %d\n", p->pasid);

		pr_debug("Finished evicting pasid %d\n", p->pasid);
@@ -918,7 +940,7 @@ static void restore_process_worker(struct work_struct *work) 
	if (ret) {

	if (ret) {

		pr_debug("Failed to restore BOs of pasid %d, retry after %d ms\n",

		pr_debug("Failed to restore BOs of pasid %d, retry after %d ms\n",

			 p->pasid, PROCESS_BACK_OFF_TIME_MS);

			 p->pasid, PROCESS_BACK_OFF_TIME_MS);

		ret = schedule_delayed_work(&p->restore_work,

		ret = queue_delayed_work(kfd_restore_wq, &p->restore_work,

				msecs_to_jiffies(PROCESS_BACK_OFF_TIME_MS));

				msecs_to_jiffies(PROCESS_BACK_OFF_TIME_MS));

		WARN(!ret, "reschedule restore work failed\n");

		WARN(!ret, "reschedule restore work failed\n");

		return;

		return;
@@ -957,7 +979,7 @@ int kfd_resume_all_processes(void) 
	int ret = 0, idx = srcu_read_lock(&kfd_processes_srcu);

	int ret = 0, idx = srcu_read_lock(&kfd_processes_srcu);





	hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {

	hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {

		if (!schedule_delayed_work(&p->restore_work, 0)) {

		if (!queue_delayed_work(kfd_restore_wq, &p->restore_work, 0)) {

			pr_err("Restore process %d failed during resume\n",

			pr_err("Restore process %d failed during resume\n",

			       p->pasid);

			       p->pasid);

			ret = -EFAULT;

			ret = -EFAULT;