Merge tag 'drm-vc4-fixes-2016-02-17' of github.com:anholt/linux into drm-fixes

This pull request fixes GPU reset (which was disabled shortly after
V3D integration due to build breakage) and waits for idle in the
presence of signals (which X likes to do a lot).

* tag 'drm-vc4-fixes-2016-02-17' of github.com:anholt/linux:
  drm/vc4: Use runtime PM to power cycle the device when the GPU hangs.
  drm/vc4: Enable runtime PM.
  drm/vc4: Fix spurious GPU resets due to BO reuse.
  drm/vc4: Drop error message on seqno wait timeouts.
  drm/vc4: Fix -ERESTARTSYS error return from BO waits.
  drm/vc4: Return an ERR_PTR from BO creation instead of NULL.
  drm/vc4: Fix the clear color for the first tile rendered.
  drm/vc4: Validate that WAIT_BO padding is cleared.

7 files changed, 118 insertions, 52 deletions

diff --git a/drivers/gpu/drm/vc4/vc4_bo.c b/drivers/gpu/drm/vc4/vc4_bo.c
index 18dfe3e..22278bc 100644
--- a/drivers/gpu/drm/vc4/vc4_bo.c
+++ b/drivers/gpu/drm/vc4/vc4_bo.c
@@ -215,7 +215,7 @@ struct vc4_bo *vc4_bo_create(struct drm_device *dev, size_t unaligned_size,
 	struct drm_gem_cma_object *cma_obj;
 
 	if (size == 0)
-		return NULL;
+		return ERR_PTR(-EINVAL);
 
 	/* First, try to get a vc4_bo from the kernel BO cache. */
 	if (from_cache) {
@@ -237,7 +237,7 @@ struct vc4_bo *vc4_bo_create(struct drm_device *dev, size_t unaligned_size,
 		if (IS_ERR(cma_obj)) {
 			DRM_ERROR("Failed to allocate from CMA:\n");
 			vc4_bo_stats_dump(vc4);
-			return NULL;
+			return ERR_PTR(-ENOMEM);
 		}
 	}
 
@@ -259,8 +259,8 @@ int vc4_dumb_create(struct drm_file *file_priv,
 		args->size = args->pitch * args->height;
 
 	bo = vc4_bo_create(dev, args->size, false);
-	if (!bo)
-		return -ENOMEM;
+	if (IS_ERR(bo))
+		return PTR_ERR(bo);
 
 	ret = drm_gem_handle_create(file_priv, &bo->base.base, &args->handle);
 	drm_gem_object_unreference_unlocked(&bo->base.base);
@@ -443,8 +443,8 @@ int vc4_create_bo_ioctl(struct drm_device *dev, void *data,
 	 * get zeroed, and that might leak data between users.
 	 */
 	bo = vc4_bo_create(dev, args->size, false);
-	if (!bo)
-		return -ENOMEM;
+	if (IS_ERR(bo))
+		return PTR_ERR(bo);
 
 	ret = drm_gem_handle_create(file_priv, &bo->base.base, &args->handle);
 	drm_gem_object_unreference_unlocked(&bo->base.base);
@@ -496,8 +496,8 @@ vc4_create_shader_bo_ioctl(struct drm_device *dev, void *data,
 	}
 
 	bo = vc4_bo_create(dev, args->size, true);
-	if (!bo)
-		return -ENOMEM;
+	if (IS_ERR(bo))
+		return PTR_ERR(bo);
 
 	ret = copy_from_user(bo->base.vaddr,
 			     (void __user *)(uintptr_t)args->data,
diff --git a/drivers/gpu/drm/vc4/vc4_drv.h b/drivers/gpu/drm/vc4/vc4_drv.h
index 080865e..51a6333 100644
--- a/drivers/gpu/drm/vc4/vc4_drv.h
+++ b/drivers/gpu/drm/vc4/vc4_drv.h
@@ -91,8 +91,12 @@ struct vc4_dev {
 	struct vc4_bo *overflow_mem;
 	struct work_struct overflow_mem_work;
 
+	int power_refcount;
+
+	/* Mutex controlling the power refcount. */
+	struct mutex power_lock;
+
 	struct {
-		uint32_t last_ct0ca, last_ct1ca;
 		struct timer_list timer;
 		struct work_struct reset_work;
 	} hangcheck;
@@ -142,6 +146,7 @@ struct vc4_seqno_cb {
 };
 
 struct vc4_v3d {
+	struct vc4_dev *vc4;
 	struct platform_device *pdev;
 	void __iomem *regs;
 };
@@ -192,6 +197,11 @@ struct vc4_exec_info {
 	/* Sequence number for this bin/render job. */
 	uint64_t seqno;
 
+	/* Last current addresses the hardware was processing when the
+	 * hangcheck timer checked on us.
+	 */
+	uint32_t last_ct0ca, last_ct1ca;
+
 	/* Kernel-space copy of the ioctl arguments */
 	struct drm_vc4_submit_cl *args;
 
@@ -434,7 +444,6 @@ void vc4_plane_async_set_fb(struct drm_plane *plane,
 extern struct platform_driver vc4_v3d_driver;
 int vc4_v3d_debugfs_ident(struct seq_file *m, void *unused);
 int vc4_v3d_debugfs_regs(struct seq_file *m, void *unused);
-int vc4_v3d_set_power(struct vc4_dev *vc4, bool on);
 
 /* vc4_validate.c */
 int
diff --git a/drivers/gpu/drm/vc4/vc4_gem.c b/drivers/gpu/drm/vc4/vc4_gem.c
index 48ce30a..202aa15 100644
--- a/drivers/gpu/drm/vc4/vc4_gem.c
+++ b/drivers/gpu/drm/vc4/vc4_gem.c
@@ -23,6 +23,7 @@
 
 #include <linux/module.h>
 #include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
 #include <linux/device.h>
 #include <linux/io.h>
 
@@ -228,8 +229,16 @@ vc4_reset(struct drm_device *dev)
 	struct vc4_dev *vc4 = to_vc4_dev(dev);
 
 	DRM_INFO("Resetting GPU.\n");
-	vc4_v3d_set_power(vc4, false);
-	vc4_v3d_set_power(vc4, true);
+
+	mutex_lock(&vc4->power_lock);
+	if (vc4->power_refcount) {
+		/* Power the device off and back on the by dropping the
+		 * reference on runtime PM.
+		 */
+		pm_runtime_put_sync_suspend(&vc4->v3d->pdev->dev);
+		pm_runtime_get_sync(&vc4->v3d->pdev->dev);
+	}
+	mutex_unlock(&vc4->power_lock);
 
 	vc4_irq_reset(dev);
 
@@ -257,10 +266,17 @@ vc4_hangcheck_elapsed(unsigned long data)
 	struct drm_device *dev = (struct drm_device *)data;
 	struct vc4_dev *vc4 = to_vc4_dev(dev);
 	uint32_t ct0ca, ct1ca;
+	unsigned long irqflags;
+	struct vc4_exec_info *exec;
+
+	spin_lock_irqsave(&vc4->job_lock, irqflags);
+	exec = vc4_first_job(vc4);
 
 	/* If idle, we can stop watching for hangs. */
-	if (list_empty(&vc4->job_list))
+	if (!exec) {
+		spin_unlock_irqrestore(&vc4->job_lock, irqflags);
 		return;
+	}
 
 	ct0ca = V3D_READ(V3D_CTNCA(0));
 	ct1ca = V3D_READ(V3D_CTNCA(1));
@@ -268,14 +284,16 @@ vc4_hangcheck_elapsed(unsigned long data)
 	/* If we've made any progress in execution, rearm the timer
 	 * and wait.
 	 */
-	if (ct0ca != vc4->hangcheck.last_ct0ca ||
-	    ct1ca != vc4->hangcheck.last_ct1ca) {
-		vc4->hangcheck.last_ct0ca = ct0ca;
-		vc4->hangcheck.last_ct1ca = ct1ca;
+	if (ct0ca != exec->last_ct0ca || ct1ca != exec->last_ct1ca) {
+		exec->last_ct0ca = ct0ca;
+		exec->last_ct1ca = ct1ca;
+		spin_unlock_irqrestore(&vc4->job_lock, irqflags);
 		vc4_queue_hangcheck(dev);
 		return;
 	}
 
+	spin_unlock_irqrestore(&vc4->job_lock, irqflags);
+
 	/* We've gone too long with no progress, reset.  This has to
 	 * be done from a work struct, since resetting can sleep and
 	 * this timer hook isn't allowed to.
@@ -340,12 +358,7 @@ vc4_wait_for_seqno(struct drm_device *dev, uint64_t seqno, uint64_t timeout_ns,
 	finish_wait(&vc4->job_wait_queue, &wait);
 	trace_vc4_wait_for_seqno_end(dev, seqno);
 
-	if (ret && ret != -ERESTARTSYS) {
-		DRM_ERROR("timeout waiting for render thread idle\n");
-		return ret;
-	}
-
-	return 0;
+	return ret;
 }
 
 static void
@@ -578,9 +591,9 @@ vc4_get_bcl(struct drm_device *dev, struct vc4_exec_info *exec)
 	}
 
 	bo = vc4_bo_create(dev, exec_size, true);
-	if (!bo) {
+	if (IS_ERR(bo)) {
 		DRM_ERROR("Couldn't allocate BO for binning\n");
-		ret = -ENOMEM;
+		ret = PTR_ERR(bo);
 		goto fail;
 	}
 	exec->exec_bo = &bo->base;
@@ -617,6 +630,7 @@ fail:
 static void
 vc4_complete_exec(struct drm_device *dev, struct vc4_exec_info *exec)
 {
+	struct vc4_dev *vc4 = to_vc4_dev(dev);
 	unsigned i;
 
 	/* Need the struct lock for drm_gem_object_unreference(). */
@@ -635,6 +649,11 @@ vc4_complete_exec(struct drm_device *dev, struct vc4_exec_info *exec)
 	}
 	mutex_unlock(&dev->struct_mutex);
 
+	mutex_lock(&vc4->power_lock);
+	if (--vc4->power_refcount == 0)
+		pm_runtime_put(&vc4->v3d->pdev->dev);
+	mutex_unlock(&vc4->power_lock);
+
 	kfree(exec);
 }
 
@@ -746,6 +765,9 @@ vc4_wait_bo_ioctl(struct drm_device *dev, void *data,
 	struct drm_gem_object *gem_obj;
 	struct vc4_bo *bo;
 
+	if (args->pad != 0)
+		return -EINVAL;
+
 	gem_obj = drm_gem_object_lookup(dev, file_priv, args->handle);
 	if (!gem_obj) {
 		DRM_ERROR("Failed to look up GEM BO %d\n", args->handle);
@@ -772,7 +794,7 @@ vc4_submit_cl_ioctl(struct drm_device *dev, void *data,
 	struct vc4_dev *vc4 = to_vc4_dev(dev);
 	struct drm_vc4_submit_cl *args = data;
 	struct vc4_exec_info *exec;
-	int ret;
+	int ret = 0;
 
 	if ((args->flags & ~VC4_SUBMIT_CL_USE_CLEAR_COLOR) != 0) {
 		DRM_ERROR("Unknown flags: 0x%02x\n", args->flags);
@@ -785,6 +807,15 @@ vc4_submit_cl_ioctl(struct drm_device *dev, void *data,
 		return -ENOMEM;
 	}
 
+	mutex_lock(&vc4->power_lock);
+	if (vc4->power_refcount++ == 0)
+		ret = pm_runtime_get_sync(&vc4->v3d->pdev->dev);
+	mutex_unlock(&vc4->power_lock);
+	if (ret < 0) {
+		kfree(exec);
+		return ret;
+	}
+
 	exec->args = args;
 	INIT_LIST_HEAD(&exec->unref_list);
 
@@ -839,6 +870,8 @@ vc4_gem_init(struct drm_device *dev)
 		    (unsigned long)dev);
 
 	INIT_WORK(&vc4->job_done_work, vc4_job_done_work);
+
+	mutex_init(&vc4->power_lock);
 }
 
 void
diff --git a/drivers/gpu/drm/vc4/vc4_irq.c b/drivers/gpu/drm/vc4/vc4_irq.c
index b68060e..78a2135 100644
--- a/drivers/gpu/drm/vc4/vc4_irq.c
+++ b/drivers/gpu/drm/vc4/vc4_irq.c
@@ -57,7 +57,7 @@ vc4_overflow_mem_work(struct work_struct *work)
 	struct vc4_bo *bo;
 
 	bo = vc4_bo_create(dev, 256 * 1024, true);
-	if (!bo) {
+	if (IS_ERR(bo)) {
 		DRM_ERROR("Couldn't allocate binner overflow mem\n");
 		return;
 	}
diff --git a/drivers/gpu/drm/vc4/vc4_render_cl.c b/drivers/gpu/drm/vc4/vc4_render_cl.c
index 8a2a312..0f12418 100644
--- a/drivers/gpu/drm/vc4/vc4_render_cl.c
+++ b/drivers/gpu/drm/vc4/vc4_render_cl.c
@@ -316,20 +316,11 @@ static int vc4_create_rcl_bo(struct drm_device *dev, struct vc4_exec_info *exec,
 	size += xtiles * ytiles * loop_body_size;
 
 	setup->rcl = &vc4_bo_create(dev, size, true)->base;
-	if (!setup->rcl)
-		return -ENOMEM;
+	if (IS_ERR(setup->rcl))
+		return PTR_ERR(setup->rcl);
 	list_add_tail(&to_vc4_bo(&setup->rcl->base)->unref_head,
 		      &exec->unref_list);
 
-	rcl_u8(setup, VC4_PACKET_TILE_RENDERING_MODE_CONFIG);
-	rcl_u32(setup,
-		(setup->color_write ? (setup->color_write->paddr +
-				       args->color_write.offset) :
-		 0));
-	rcl_u16(setup, args->width);
-	rcl_u16(setup, args->height);
-	rcl_u16(setup, args->color_write.bits);
-
 	/* The tile buffer gets cleared when the previous tile is stored.  If
 	 * the clear values changed between frames, then the tile buffer has
 	 * stale clear values in it, so we have to do a store in None mode (no
@@ -349,6 +340,15 @@ static int vc4_create_rcl_bo(struct drm_device *dev, struct vc4_exec_info *exec,
 		rcl_u32(setup, 0); /* no address, since we're in None mode */
 	}
 
+	rcl_u8(setup, VC4_PACKET_TILE_RENDERING_MODE_CONFIG);
+	rcl_u32(setup,
+		(setup->color_write ? (setup->color_write->paddr +
+				       args->color_write.offset) :
+		 0));
+	rcl_u16(setup, args->width);
+	rcl_u16(setup, args->height);
+	rcl_u16(setup, args->color_write.bits);
+
 	for (y = min_y_tile; y <= max_y_tile; y++) {
 		for (x = min_x_tile; x <= max_x_tile; x++) {
 			bool first = (x == min_x_tile && y == min_y_tile);
diff --git a/drivers/gpu/drm/vc4/vc4_v3d.c b/drivers/gpu/drm/vc4/vc4_v3d.c
index 314ff71..31de5d1 100644
--- a/drivers/gpu/drm/vc4/vc4_v3d.c
+++ b/drivers/gpu/drm/vc4/vc4_v3d.c
@@ -17,6 +17,7 @@
  */
 
 #include "linux/component.h"
+#include "linux/pm_runtime.h"
 #include "vc4_drv.h"
 #include "vc4_regs.h"
 
@@ -144,18 +145,6 @@ int vc4_v3d_debugfs_ident(struct seq_file *m, void *unused)
 }
 #endif /* CONFIG_DEBUG_FS */
 
-int
-vc4_v3d_set_power(struct vc4_dev *vc4, bool on)
-{
-	/* XXX: This interface is needed for GPU reset, and the way to
-	 * do it is to turn our power domain off and back on.  We
-	 * can't just reset from within the driver, because the reset
-	 * bits are in the power domain's register area, and get set
-	 * during the poweron process.
-	 */
-	return 0;
-}
-
 static void vc4_v3d_init_hw(struct drm_device *dev)
 {
 	struct vc4_dev *vc4 = to_vc4_dev(dev);
@@ -167,6 +156,29 @@ static void vc4_v3d_init_hw(struct drm_device *dev)
 	V3D_WRITE(V3D_VPMBASE, 0);
 }
 
+#ifdef CONFIG_PM
+static int vc4_v3d_runtime_suspend(struct device *dev)
+{
+	struct vc4_v3d *v3d = dev_get_drvdata(dev);
+	struct vc4_dev *vc4 = v3d->vc4;
+
+	vc4_irq_uninstall(vc4->dev);
+
+	return 0;
+}
+
+static int vc4_v3d_runtime_resume(struct device *dev)
+{
+	struct vc4_v3d *v3d = dev_get_drvdata(dev);
+	struct vc4_dev *vc4 = v3d->vc4;
+
+	vc4_v3d_init_hw(vc4->dev);
+	vc4_irq_postinstall(vc4->dev);
+
+	return 0;
+}
+#endif
+
 static int vc4_v3d_bind(struct device *dev, struct device *master, void *data)
 {
 	struct platform_device *pdev = to_platform_device(dev);
@@ -179,6 +191,8 @@ static int vc4_v3d_bind(struct device *dev, struct device *master, void *data)
 	if (!v3d)
 		return -ENOMEM;
 
+	dev_set_drvdata(dev, v3d);
+
 	v3d->pdev = pdev;
 
 	v3d->regs = vc4_ioremap_regs(pdev, 0);
@@ -186,6 +200,7 @@ static int vc4_v3d_bind(struct device *dev, struct device *master, void *data)
 		return PTR_ERR(v3d->regs);
 
 	vc4->v3d = v3d;
+	v3d->vc4 = vc4;
 
 	if (V3D_READ(V3D_IDENT0) != V3D_EXPECTED_IDENT0) {
 		DRM_ERROR("V3D_IDENT0 read 0x%08x instead of 0x%08x\n",
@@ -207,6 +222,8 @@ static int vc4_v3d_bind(struct device *dev, struct device *master, void *data)
 		return ret;
 	}
 
+	pm_runtime_enable(dev);
+
 	return 0;
 }
 
@@ -216,6 +233,8 @@ static void vc4_v3d_unbind(struct device *dev, struct device *master,
 	struct drm_device *drm = dev_get_drvdata(master);
 	struct vc4_dev *vc4 = to_vc4_dev(drm);
 
+	pm_runtime_disable(dev);
+
 	drm_irq_uninstall(drm);
 
 	/* Disable the binner's overflow memory address, so the next
@@ -228,6 +247,10 @@ static void vc4_v3d_unbind(struct device *dev, struct device *master,
 	vc4->v3d = NULL;
 }
 
+static const struct dev_pm_ops vc4_v3d_pm_ops = {
+	SET_RUNTIME_PM_OPS(vc4_v3d_runtime_suspend, vc4_v3d_runtime_resume, NULL)
+};
+
 static const struct component_ops vc4_v3d_ops = {
 	.bind   = vc4_v3d_bind,
 	.unbind = vc4_v3d_unbind,
@@ -255,5 +278,6 @@ struct platform_driver vc4_v3d_driver = {
 	.driver = {
 		.name = "vc4_v3d",
 		.of_match_table = vc4_v3d_dt_match,
+		.pm = &vc4_v3d_pm_ops,
 	},
 };
diff --git a/drivers/gpu/drm/vc4/vc4_validate.c b/drivers/gpu/drm/vc4/vc4_validate.c
index e26d9f6..24c2c74 100644
--- a/drivers/gpu/drm/vc4/vc4_validate.c
+++ b/drivers/gpu/drm/vc4/vc4_validate.c
@@ -401,8 +401,8 @@ validate_tile_binning_config(VALIDATE_ARGS)
 	tile_bo = vc4_bo_create(dev, exec->tile_alloc_offset + tile_alloc_size,
 				true);
 	exec->tile_bo = &tile_bo->base;
-	if (!exec->tile_bo)
-		return -ENOMEM;
+	if (IS_ERR(exec->tile_bo))
+		return PTR_ERR(exec->tile_bo);
 	list_add_tail(&tile_bo->unref_head, &exec->unref_list);
 
 	/* tile alloc address. */