[PATCH] drm/panfrost: Move the GPU reset bits outside the timeout handler

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



We've fixed many races in panfrost_job_timedout() but some remain.
Instead of trying to fix it again, let's simplify the logic and move
the reset bits to a separate work scheduled when one of the queue
reports a timeout.

Fixes: 1a11a88cfd9a ("drm/panfrost: Fix job timeout handling")
Cc: <stable@xxxxxxxxxxxxxxx>
Signed-off-by: Boris Brezillon <boris.brezillon@xxxxxxxxxxxxx>
---
 drivers/gpu/drm/panfrost/panfrost_device.c |   1 -
 drivers/gpu/drm/panfrost/panfrost_device.h |   6 +-
 drivers/gpu/drm/panfrost/panfrost_job.c    | 130 ++++++++++++---------
 3 files changed, 82 insertions(+), 55 deletions(-)

diff --git a/drivers/gpu/drm/panfrost/panfrost_device.c b/drivers/gpu/drm/panfrost/panfrost_device.c
index ea8d31863c50..a83b2ff5837a 100644
--- a/drivers/gpu/drm/panfrost/panfrost_device.c
+++ b/drivers/gpu/drm/panfrost/panfrost_device.c
@@ -200,7 +200,6 @@ int panfrost_device_init(struct panfrost_device *pfdev)
 	struct resource *res;
 
 	mutex_init(&pfdev->sched_lock);
-	mutex_init(&pfdev->reset_lock);
 	INIT_LIST_HEAD(&pfdev->scheduled_jobs);
 	INIT_LIST_HEAD(&pfdev->as_lru_list);
 
diff --git a/drivers/gpu/drm/panfrost/panfrost_device.h b/drivers/gpu/drm/panfrost/panfrost_device.h
index 2e9cbd1c4a58..67f9f66904be 100644
--- a/drivers/gpu/drm/panfrost/panfrost_device.h
+++ b/drivers/gpu/drm/panfrost/panfrost_device.h
@@ -105,7 +105,11 @@ struct panfrost_device {
 	struct panfrost_perfcnt *perfcnt;
 
 	struct mutex sched_lock;
-	struct mutex reset_lock;
+
+	struct {
+		struct work_struct work;
+		atomic_t pending;
+	} reset;
 
 	struct mutex shrinker_lock;
 	struct list_head shrinker_list;
diff --git a/drivers/gpu/drm/panfrost/panfrost_job.c b/drivers/gpu/drm/panfrost/panfrost_job.c
index d0469e944143..745ee9563a54 100644
--- a/drivers/gpu/drm/panfrost/panfrost_job.c
+++ b/drivers/gpu/drm/panfrost/panfrost_job.c
@@ -20,6 +20,8 @@
 #include "panfrost_gpu.h"
 #include "panfrost_mmu.h"
 
+#define JOB_TIMEOUT_MS 500
+
 #define job_write(dev, reg, data) writel(data, dev->iomem + (reg))
 #define job_read(dev, reg) readl(dev->iomem + (reg))
 
@@ -382,19 +384,37 @@ static bool panfrost_scheduler_stop(struct panfrost_queue_state *queue,
 			drm_sched_increase_karma(bad);
 		queue->stopped = true;
 		stopped = true;
+
+		/*
+		 * Set the timeout to max so the timer doesn't get started
+		 * when we return from the timeout handler (restored in
+		 * panfrost_scheduler_start()).
+		 */
+		queue->sched.timeout = MAX_SCHEDULE_TIMEOUT;
 	}
 	mutex_unlock(&queue->lock);
 
 	return stopped;
 }
 
+static void panfrost_scheduler_start(struct panfrost_queue_state *queue)
+{
+	if (WARN_ON(!queue->stopped))
+		return;
+
+	mutex_lock(&queue->lock);
+	/* Restore the original timeout before starting the scheduler. */
+	queue->sched.timeout = msecs_to_jiffies(JOB_TIMEOUT_MS);
+	drm_sched_start(&queue->sched, true);
+	queue->stopped = false;
+	mutex_unlock(&queue->lock);
+}
+
 static void panfrost_job_timedout(struct drm_sched_job *sched_job)
 {
 	struct panfrost_job *job = to_panfrost_job(sched_job);
 	struct panfrost_device *pfdev = job->pfdev;
 	int js = panfrost_job_get_slot(job);
-	unsigned long flags;
-	int i;
 
 	/*
 	 * If the GPU managed to complete this jobs fence, the timeout is
@@ -415,56 +435,9 @@ static void panfrost_job_timedout(struct drm_sched_job *sched_job)
 	if (!panfrost_scheduler_stop(&pfdev->js->queue[js], sched_job))
 		return;
 
-	if (!mutex_trylock(&pfdev->reset_lock))
-		return;
-
-	for (i = 0; i < NUM_JOB_SLOTS; i++) {
-		struct drm_gpu_scheduler *sched = &pfdev->js->queue[i].sched;
-
-		/*
-		 * If the queue is still active, make sure we wait for any
-		 * pending timeouts.
-		 */
-		if (!pfdev->js->queue[i].stopped)
-			cancel_delayed_work_sync(&sched->work_tdr);
-
-		/*
-		 * If the scheduler was not already stopped, there's a tiny
-		 * chance a timeout has expired just before we stopped it, and
-		 * drm_sched_stop() does not flush pending works. Let's flush
-		 * them now so the timeout handler doesn't get called in the
-		 * middle of a reset.
-		 */
-		if (panfrost_scheduler_stop(&pfdev->js->queue[i], NULL))
-			cancel_delayed_work_sync(&sched->work_tdr);
-
-		/*
-		 * Now that we cancelled the pending timeouts, we can safely
-		 * reset the stopped state.
-		 */
-		pfdev->js->queue[i].stopped = false;
-	}
-
-	spin_lock_irqsave(&pfdev->js->job_lock, flags);
-	for (i = 0; i < NUM_JOB_SLOTS; i++) {
-		if (pfdev->jobs[i]) {
-			pm_runtime_put_noidle(pfdev->dev);
-			panfrost_devfreq_record_idle(&pfdev->pfdevfreq);
-			pfdev->jobs[i] = NULL;
-		}
-	}
-	spin_unlock_irqrestore(&pfdev->js->job_lock, flags);
-
-	panfrost_device_reset(pfdev);
-
-	for (i = 0; i < NUM_JOB_SLOTS; i++)
-		drm_sched_resubmit_jobs(&pfdev->js->queue[i].sched);
-
-	mutex_unlock(&pfdev->reset_lock);
-
-	/* restart scheduler after GPU is usable again */
-	for (i = 0; i < NUM_JOB_SLOTS; i++)
-		drm_sched_start(&pfdev->js->queue[i].sched, true);
+	/* Schedule a reset. */
+	atomic_set(&pfdev->reset.pending, 1);
+	schedule_work(&pfdev->reset.work);
 }
 
 static const struct drm_sched_backend_ops panfrost_sched_ops = {
@@ -531,11 +504,62 @@ static irqreturn_t panfrost_job_irq_handler(int irq, void *data)
 	return IRQ_HANDLED;
 }
 
+static void panfrost_reset(struct work_struct *work)
+{
+	struct panfrost_device *pfdev = container_of(work,
+						     struct panfrost_device,
+						     reset.work);
+	unsigned long flags;
+	unsigned int i;
+
+	if (!atomic_read(&pfdev->reset.pending))
+		return;
+
+	for (i = 0; i < NUM_JOB_SLOTS; i++) {
+		/*
+		 * We want pending timeouts to be handled before we attempt
+		 * to stop the scheduler. If we don't do that and the timeout
+		 * handler is in flight, it might have removed the bad job
+		 * from the list, and we'll lose this job if the reset handler
+		 * enters the critical section in panfrost_scheduler_stop()
+		 * before the timeout handler.
+		 *
+		 * Timeout is set to max to make sure the timer is not
+		 * restarted after the cancellation.
+		 */
+		pfdev->js->queue[i].sched.timeout = MAX_SCHEDULE_TIMEOUT;
+		cancel_delayed_work_sync(&pfdev->js->queue[i].sched.work_tdr);
+		panfrost_scheduler_stop(&pfdev->js->queue[i], NULL);
+	}
+
+	/* All timers have been stopped, we can safely reset the pending state. */
+	atomic_set(&pfdev->reset.pending, 0);
+
+	spin_lock_irqsave(&pfdev->js->job_lock, flags);
+	for (i = 0; i < NUM_JOB_SLOTS; i++) {
+		if (pfdev->jobs[i]) {
+			pm_runtime_put_noidle(pfdev->dev);
+			panfrost_devfreq_record_idle(&pfdev->pfdevfreq);
+			pfdev->jobs[i] = NULL;
+		}
+	}
+	spin_unlock_irqrestore(&pfdev->js->job_lock, flags);
+
+	panfrost_device_reset(pfdev);
+
+	for (i = 0; i < NUM_JOB_SLOTS; i++) {
+		drm_sched_resubmit_jobs(&pfdev->js->queue[i].sched);
+		panfrost_scheduler_start(&pfdev->js->queue[i]);
+	}
+}
+
 int panfrost_job_init(struct panfrost_device *pfdev)
 {
 	struct panfrost_job_slot *js;
 	int ret, j, irq;
 
+	INIT_WORK(&pfdev->reset.work, panfrost_reset);
+
 	pfdev->js = js = devm_kzalloc(pfdev->dev, sizeof(*js), GFP_KERNEL);
 	if (!js)
 		return -ENOMEM;
@@ -558,7 +582,7 @@ int panfrost_job_init(struct panfrost_device *pfdev)
 
 		ret = drm_sched_init(&js->queue[j].sched,
 				     &panfrost_sched_ops,
-				     1, 0, msecs_to_jiffies(500),
+				     1, 0, msecs_to_jiffies(JOB_TIMEOUT_MS),
 				     "pan_js");
 		if (ret) {
 			dev_err(pfdev->dev, "Failed to create scheduler: %d.", ret);
-- 
2.26.2

_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/dri-devel



[Index of Archives]     [Linux DRI Users]     [Linux Intel Graphics]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]
  Powered by Linux