4 files changed, 66 insertions, 18 deletions

diff --git a/drivers/gpu/drm/v3d/v3d_drv.h b/drivers/gpu/drm/v3d/v3d_drv.h
index b74b1351bfc8..a366ea208787 100644
--- a/drivers/gpu/drm/v3d/v3d_drv.h
+++ b/drivers/gpu/drm/v3d/v3d_drv.h
@@ -62,6 +62,12 @@ struct v3d_perfmon {
 	u64 values[];
 };
 
+enum v3d_irq {
+	V3D_CORE_IRQ,
+	V3D_HUB_IRQ,
+	V3D_MAX_IRQS,
+};
+
 struct v3d_dev {
 	struct drm_device drm;
 
@@ -71,6 +77,8 @@ struct v3d_dev {
 	int ver;
 	bool single_irq_line;
 
+	int irq[V3D_MAX_IRQS];
+
 	void __iomem *hub_regs;
 	void __iomem *core_regs[3];
 	void __iomem *bridge_regs;
diff --git a/drivers/gpu/drm/v3d/v3d_gem.c b/drivers/gpu/drm/v3d/v3d_gem.c
index b8980440d137..8b6450a96ebc 100644
--- a/drivers/gpu/drm/v3d/v3d_gem.c
+++ b/drivers/gpu/drm/v3d/v3d_gem.c
@@ -119,6 +119,8 @@ v3d_reset(struct v3d_dev *v3d)
 	if (false)
 		v3d_idle_axi(v3d, 0);
 
+	v3d_irq_disable(v3d);
+
 	v3d_idle_gca(v3d);
 	v3d_reset_v3d(v3d);
 
diff --git a/drivers/gpu/drm/v3d/v3d_irq.c b/drivers/gpu/drm/v3d/v3d_irq.c
index b2d59a168697..641315dbee8b 100644
--- a/drivers/gpu/drm/v3d/v3d_irq.c
+++ b/drivers/gpu/drm/v3d/v3d_irq.c
@@ -215,7 +215,7 @@ v3d_hub_irq(int irq, void *arg)
 int
 v3d_irq_init(struct v3d_dev *v3d)
 {
-	int irq1, ret, core;
+	int irq, ret, core;
 
 	INIT_WORK(&v3d->overflow_mem_work, v3d_overflow_mem_work);
 
@@ -226,17 +226,24 @@ v3d_irq_init(struct v3d_dev *v3d)
 		V3D_CORE_WRITE(core, V3D_CTL_INT_CLR, V3D_CORE_IRQS);
 	V3D_WRITE(V3D_HUB_INT_CLR, V3D_HUB_IRQS);
 
-	irq1 = platform_get_irq_optional(v3d_to_pdev(v3d), 1);
-	if (irq1 == -EPROBE_DEFER)
-		return irq1;
-	if (irq1 > 0) {
-		ret = devm_request_irq(v3d->drm.dev, irq1,
+	irq = platform_get_irq_optional(v3d_to_pdev(v3d), 1);
+	if (irq == -EPROBE_DEFER)
+		return irq;
+	if (irq > 0) {
+		v3d->irq[V3D_CORE_IRQ] = irq;
+
+		ret = devm_request_irq(v3d->drm.dev, v3d->irq[V3D_CORE_IRQ],
 				       v3d_irq, IRQF_SHARED,
 				       "v3d_core0", v3d);
 		if (ret)
 			goto fail;
-		ret = devm_request_irq(v3d->drm.dev,
-				       platform_get_irq(v3d_to_pdev(v3d), 0),
+
+		irq = platform_get_irq(v3d_to_pdev(v3d), 0);
+		if (irq < 0)
+			return irq;
+		v3d->irq[V3D_HUB_IRQ] = irq;
+
+		ret = devm_request_irq(v3d->drm.dev, v3d->irq[V3D_HUB_IRQ],
 				       v3d_hub_irq, IRQF_SHARED,
 				       "v3d_hub", v3d);
 		if (ret)
@@ -244,8 +251,12 @@ v3d_irq_init(struct v3d_dev *v3d)
 	} else {
 		v3d->single_irq_line = true;
 
-		ret = devm_request_irq(v3d->drm.dev,
-				       platform_get_irq(v3d_to_pdev(v3d), 0),
+		irq = platform_get_irq(v3d_to_pdev(v3d), 0);
+		if (irq < 0)
+			return irq;
+		v3d->irq[V3D_CORE_IRQ] = irq;
+
+		ret = devm_request_irq(v3d->drm.dev, v3d->irq[V3D_CORE_IRQ],
 				       v3d_irq, IRQF_SHARED,
 				       "v3d", v3d);
 		if (ret)
@@ -286,6 +297,12 @@ v3d_irq_disable(struct v3d_dev *v3d)
 		V3D_CORE_WRITE(core, V3D_CTL_INT_MSK_SET, ~0);
 	V3D_WRITE(V3D_HUB_INT_MSK_SET, ~0);
 
+	/* Finish any interrupt handler still in flight. */
+	for (int i = 0; i < V3D_MAX_IRQS; i++) {
+		if (v3d->irq[i])
+			synchronize_irq(v3d->irq[i]);
+	}
+
 	/* Clear any pending interrupts we might have left. */
 	for (core = 0; core < v3d->cores; core++)
 		V3D_CORE_WRITE(core, V3D_CTL_INT_CLR, V3D_CORE_IRQS);
diff --git a/drivers/gpu/drm/v3d/v3d_sched.c b/drivers/gpu/drm/v3d/v3d_sched.c
index 06238e6d7f5c..41493cf3d03b 100644
--- a/drivers/gpu/drm/v3d/v3d_sched.c
+++ b/drivers/gpu/drm/v3d/v3d_sched.c
@@ -179,11 +179,15 @@ v3d_tfu_job_run(struct drm_sched_job *sched_job)
 	struct drm_device *dev = &v3d->drm;
 	struct dma_fence *fence;
 
+	if (unlikely(job->base.base.s_fence->finished.error))
+		return NULL;
+
+	v3d->tfu_job = job;
+
 	fence = v3d_fence_create(v3d, V3D_TFU);
 	if (IS_ERR(fence))
 		return NULL;
 
-	v3d->tfu_job = job;
 	if (job->base.irq_fence)
 		dma_fence_put(job->base.irq_fence);
 	job->base.irq_fence = dma_fence_get(fence);
@@ -217,6 +221,9 @@ v3d_csd_job_run(struct drm_sched_job *sched_job)
 	struct dma_fence *fence;
 	int i;
 
+	if (unlikely(job->base.base.s_fence->finished.error))
+		return NULL;
+
 	v3d->csd_job = job;
 
 	v3d_invalidate_caches(v3d);
@@ -282,11 +289,16 @@ v3d_gpu_reset_for_timeout(struct v3d_dev *v3d, struct drm_sched_job *sched_job)
 	return DRM_GPU_SCHED_STAT_NOMINAL;
 }
 
-/* If the current address or return address have changed, then the GPU
- * has probably made progress and we should delay the reset.  This
- * could fail if the GPU got in an infinite loop in the CL, but that
- * is pretty unlikely outside of an i-g-t testcase.
- */
+static void
+v3d_sched_skip_reset(struct drm_sched_job *sched_job)
+{
+	struct drm_gpu_scheduler *sched = sched_job->sched;
+
+	spin_lock(&sched->job_list_lock);
+	list_add(&sched_job->list, &sched->pending_list);
+	spin_unlock(&sched->job_list_lock);
+}
+
 static enum drm_gpu_sched_stat
 v3d_cl_job_timedout(struct drm_sched_job *sched_job, enum v3d_queue q,
 		    u32 *timedout_ctca, u32 *timedout_ctra)
@@ -296,9 +308,16 @@ v3d_cl_job_timedout(struct drm_sched_job *sched_job, enum v3d_queue q,
 	u32 ctca = V3D_CORE_READ(0, V3D_CLE_CTNCA(q));
 	u32 ctra = V3D_CORE_READ(0, V3D_CLE_CTNRA(q));
 
+	/* If the current address or return address have changed, then the GPU
+	 * has probably made progress and we should delay the reset. This
+	 * could fail if the GPU got in an infinite loop in the CL, but that
+	 * is pretty unlikely outside of an i-g-t testcase.
+	 */
 	if (*timedout_ctca != ctca || *timedout_ctra != ctra) {
 		*timedout_ctca = ctca;
 		*timedout_ctra = ctra;
+
+		v3d_sched_skip_reset(sched_job);
 		return DRM_GPU_SCHED_STAT_NOMINAL;
 	}
 
@@ -338,11 +357,13 @@ v3d_csd_job_timedout(struct drm_sched_job *sched_job)
 	struct v3d_dev *v3d = job->base.v3d;
 	u32 batches = V3D_CORE_READ(0, V3D_CSD_CURRENT_CFG4);
 
-	/* If we've made progress, skip reset and let the timer get
-	 * rearmed.
+	/* If we've made progress, skip reset, add the job to the pending
+	 * list, and let the timer get rearmed.
 	 */
 	if (job->timedout_batches != batches) {
 		job->timedout_batches = batches;
+
+		v3d_sched_skip_reset(sched_job);
 		return DRM_GPU_SCHED_STAT_NOMINAL;
 	}