Merge tag 'drm/tegra/for-5.17-rc1' of https://gitlab.freedesktop.org/drm/tegra into drm-next

drm/tegra: Changes for v5.17-rc1

This contains a fairly large rework that makes the buffer objects behave
more according to what the DMA-BUF infrastructure expects. A buffer
object cache is implemented on top of that to make certain operations
such as page-flipping more efficient by avoiding needless map/unmap
operations. This in turn is useful to implement asynchronous commits to
support legacy cursor updates.

Another fairly big addition is the NVDEC driver. This uses the updated
UABI introduced in v5.15-rc1 to provide access to the video decode
engines found on Tegra210 and later.

This also includes some power management improvements that are useful on
older devices in particular because they, together with a bunch of other
changes across the kernel, allow the system to scale down frequency and
voltages when mostly idle and prevent these devices from becoming
excessively hot.

The remainder of these changes is an assortment of cleanups and minor
fixes.
Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Thierry Reding <thierry.reding@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211217142912.558095-1-thierry.reding@gmail.com

drivers/gpu/drm/tegra/Kconfig

@@ -12,6 +12,9 @@ config DRM_TEGRA
 	select INTERCONNECT
 	select IOMMU_IOVA
 	select CEC_CORE if CEC_NOTIFIER
+	select SND_SIMPLE_CARD if SND_SOC_TEGRA20_SPDIF
+	select SND_SOC_HDMI_CODEC if SND_SOC_TEGRA20_SPDIF
+	select SND_AUDIO_GRAPH_CARD if SND_SOC_TEGRA20_SPDIF
 	help
 	  Choose this option if you have an NVIDIA Tegra SoC.
 

drivers/gpu/drm/tegra/Makefile

@@ -23,7 +23,8 @@ tegra-drm-y := \
 	gr2d.o \
 	gr3d.o \
 	falcon.o \
-	vic.o
+	vic.o \
+	nvdec.o
 
 tegra-drm-y += trace.o
 

drivers/gpu/drm/tegra/dc.h

@@ -76,6 +76,7 @@ struct tegra_dc_soc_info {
 	bool has_win_b_vfilter_mem_client;
 	bool has_win_c_without_vert_filter;
 	bool plane_tiled_memory_bandwidth_x2;
+	bool has_pll_d2_out0;
 };
 
 struct tegra_dc {
@@ -100,6 +101,8 @@ struct tegra_dc {
 	struct drm_info_list *debugfs_files;
 
 	const struct tegra_dc_soc_info *soc;
+
+	bool has_opp_table;
 };
 
 static inline struct tegra_dc *

drivers/gpu/drm/tegra/drm.c

@@ -10,6 +10,7 @@
 #include <linux/iommu.h>
 #include <linux/module.h>
 #include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
 
 #include <drm/drm_aperture.h>
 #include <drm/drm_atomic.h>
@@ -21,6 +22,10 @@
 #include <drm/drm_prime.h>
 #include <drm/drm_vblank.h>
 
+#if IS_ENABLED(CONFIG_ARM_DMA_USE_IOMMU)
+#include <asm/dma-iommu.h>
+#endif
+
 #include "dc.h"
 #include "drm.h"
 #include "gem.h"
@@ -116,6 +121,7 @@ static int tegra_drm_open(struct drm_device *drm, struct drm_file *filp)
 static void tegra_drm_context_free(struct tegra_drm_context *context)
 {
 	context->client->ops->close_channel(context);
+	pm_runtime_put(context->client->base.dev);
 	kfree(context);
 }
 
@@ -427,13 +433,20 @@ static int tegra_client_open(struct tegra_drm_file *fpriv,
 {
 	int err;
 
+	err = pm_runtime_resume_and_get(client->base.dev);
+	if (err)
+		return err;
+
 	err = client->ops->open_channel(client, context);
-	if (err < 0)
+	if (err < 0) {
+		pm_runtime_put(client->base.dev);
 		return err;
+	}
 
 	err = idr_alloc(&fpriv->legacy_contexts, context, 1, 0, GFP_KERNEL);
 	if (err < 0) {
 		client->ops->close_channel(context);
+		pm_runtime_put(client->base.dev);
 		return err;
 	}
 
@@ -936,6 +949,17 @@ int host1x_client_iommu_attach(struct host1x_client *client)
 	struct iommu_group *group = NULL;
 	int err;
 
+#if IS_ENABLED(CONFIG_ARM_DMA_USE_IOMMU)
+	if (client->dev->archdata.mapping) {
+		struct dma_iommu_mapping *mapping =
+				to_dma_iommu_mapping(client->dev);
+		arm_iommu_detach_device(client->dev);
+		arm_iommu_release_mapping(mapping);
+
+		domain = iommu_get_domain_for_dev(client->dev);
+	}
+#endif
+
 	/*
 	 * If the host1x client is already attached to an IOMMU domain that is
 	 * not the shared IOMMU domain, don't try to attach it to a different
@@ -1344,15 +1368,18 @@ static const struct of_device_id host1x_drm_subdevs[] = {
 	{ .compatible = "nvidia,tegra210-sor", },
 	{ .compatible = "nvidia,tegra210-sor1", },
 	{ .compatible = "nvidia,tegra210-vic", },
+	{ .compatible = "nvidia,tegra210-nvdec", },
 	{ .compatible = "nvidia,tegra186-display", },
 	{ .compatible = "nvidia,tegra186-dc", },
 	{ .compatible = "nvidia,tegra186-sor", },
 	{ .compatible = "nvidia,tegra186-sor1", },
 	{ .compatible = "nvidia,tegra186-vic", },
+	{ .compatible = "nvidia,tegra186-nvdec", },
 	{ .compatible = "nvidia,tegra194-display", },
 	{ .compatible = "nvidia,tegra194-dc", },
 	{ .compatible = "nvidia,tegra194-sor", },
 	{ .compatible = "nvidia,tegra194-vic", },
+	{ .compatible = "nvidia,tegra194-nvdec", },
 	{ /* sentinel */ }
 };
 
@@ -1376,6 +1403,7 @@ static struct platform_driver * const drivers[] = {
 	&tegra_gr2d_driver,
 	&tegra_gr3d_driver,
 	&tegra_vic_driver,
+	&tegra_nvdec_driver,
 };
 
 static int __init host1x_drm_init(void)

drivers/gpu/drm/tegra/drm.h

@@ -202,5 +202,6 @@ extern struct platform_driver tegra_sor_driver;
 extern struct platform_driver tegra_gr2d_driver;
 extern struct platform_driver tegra_gr3d_driver;
 extern struct platform_driver tegra_vic_driver;
+extern struct platform_driver tegra_nvdec_driver;
 
 #endif /* HOST1X_DRM_H */

drivers/gpu/drm/tegra/gem.c

@@ -23,86 +23,97 @@
 
 MODULE_IMPORT_NS(DMA_BUF);
 
-static void tegra_bo_put(struct host1x_bo *bo)
+static unsigned int sg_dma_count_chunks(struct scatterlist *sgl, unsigned int nents)
 {
-	struct tegra_bo *obj = host1x_to_tegra_bo(bo);
+	dma_addr_t next = ~(dma_addr_t)0;
+	unsigned int count = 0, i;
+	struct scatterlist *s;
 
-	drm_gem_object_put(&obj->gem);
-}
+	for_each_sg(sgl, s, nents, i) {
+		/* sg_dma_address(s) is only valid for entries that have sg_dma_len(s) != 0. */
+		if (!sg_dma_len(s))
+			continue;
 
-/* XXX move this into lib/scatterlist.c? */
-static int sg_alloc_table_from_sg(struct sg_table *sgt, struct scatterlist *sg,
-				  unsigned int nents, gfp_t gfp_mask)
-{
-	struct scatterlist *dst;
-	unsigned int i;
-	int err;
+		if (sg_dma_address(s) != next) {
+			next = sg_dma_address(s) + sg_dma_len(s);
+			count++;
+		}
+	}
 
-	err = sg_alloc_table(sgt, nents, gfp_mask);
-	if (err < 0)
-		return err;
+	return count;
+}
 
-	dst = sgt->sgl;
+static inline unsigned int sgt_dma_count_chunks(struct sg_table *sgt)
+{
+	return sg_dma_count_chunks(sgt->sgl, sgt->nents);
+}
 
-	for (i = 0; i < nents; i++) {
-		sg_set_page(dst, sg_page(sg), sg->length, 0);
-		dst = sg_next(dst);
-		sg = sg_next(sg);
-	}
+static void tegra_bo_put(struct host1x_bo *bo)
+{
+	struct tegra_bo *obj = host1x_to_tegra_bo(bo);
 
-	return 0;
+	drm_gem_object_put(&obj->gem);
 }
 
-static struct sg_table *tegra_bo_pin(struct device *dev, struct host1x_bo *bo,
-				     dma_addr_t *phys)
+static struct host1x_bo_mapping *tegra_bo_pin(struct device *dev, struct host1x_bo *bo,
+					      enum dma_data_direction direction)
 {
 	struct tegra_bo *obj = host1x_to_tegra_bo(bo);
-	struct sg_table *sgt;
+	struct drm_gem_object *gem = &obj->gem;
+	struct host1x_bo_mapping *map;
 	int err;
 
+	map = kzalloc(sizeof(*map), GFP_KERNEL);
+	if (!map)
+		return ERR_PTR(-ENOMEM);
+
+	kref_init(&map->ref);
+	map->bo = host1x_bo_get(bo);
+	map->direction = direction;
+	map->dev = dev;
+
 	/*
-	 * If we've manually mapped the buffer object through the IOMMU, make
-	 * sure to return the IOVA address of our mapping.
-	 *
-	 * Similarly, for buffers that have been allocated by the DMA API the
-	 * physical address can be used for devices that are not attached to
-	 * an IOMMU. For these devices, callers must pass a valid pointer via
-	 * the @phys argument.
-	 *
-	 * Imported buffers were also already mapped at import time, so the
-	 * existing mapping can be reused.
+	 * Imported buffers need special treatment to satisfy the semantics of DMA-BUF.
 	 */
-	if (phys) {
-		*phys = obj->iova;
-		return NULL;
+	if (gem->import_attach) {
+		struct dma_buf *buf = gem->import_attach->dmabuf;
+
+		map->attach = dma_buf_attach(buf, dev);
+		if (IS_ERR(map->attach)) {
+			err = PTR_ERR(map->attach);
+			goto free;
+		}
+
+		map->sgt = dma_buf_map_attachment(map->attach, direction);
+		if (IS_ERR(map->sgt)) {
+			dma_buf_detach(buf, map->attach);
+			err = PTR_ERR(map->sgt);
+			goto free;
+		}
+
+		err = sgt_dma_count_chunks(map->sgt);
+		map->size = gem->size;
+
+		goto out;
 	}
 
 	/*
 	 * If we don't have a mapping for this buffer yet, return an SG table
 	 * so that host1x can do the mapping for us via the DMA API.
 	 */
-	sgt = kzalloc(sizeof(*sgt), GFP_KERNEL);
-	if (!sgt)
-		return ERR_PTR(-ENOMEM);
+	map->sgt = kzalloc(sizeof(*map->sgt), GFP_KERNEL);
+	if (!map->sgt) {
+		err = -ENOMEM;
+		goto free;
+	}
 
 	if (obj->pages) {
 		/*
 		 * If the buffer object was allocated from the explicit IOMMU
 		 * API code paths, construct an SG table from the pages.
 		 */
-		err = sg_alloc_table_from_pages(sgt, obj->pages, obj->num_pages,
-						0, obj->gem.size, GFP_KERNEL);
-		if (err < 0)
-			goto free;
-	} else if (obj->sgt) {
-		/*
-		 * If the buffer object already has an SG table but no pages
-		 * were allocated for it, it means the buffer was imported and
-		 * the SG table needs to be copied to avoid overwriting any
-		 * other potential users of the original SG table.
-		 */
-		err = sg_alloc_table_from_sg(sgt, obj->sgt->sgl,
-					     obj->sgt->orig_nents, GFP_KERNEL);
+		err = sg_alloc_table_from_pages(map->sgt, obj->pages, obj->num_pages, 0, gem->size,
+						GFP_KERNEL);
 		if (err < 0)
 			goto free;
 	} else {
@@ -111,25 +122,53 @@ static struct sg_table *tegra_bo_pin(struct device *dev, struct host1x_bo *bo,
 		 * not imported, it had to be allocated with the DMA API, so
 		 * the DMA API helper can be used.
 		 */
-		err = dma_get_sgtable(dev, sgt, obj->vaddr, obj->iova,
-				      obj->gem.size);
+		err = dma_get_sgtable(dev, map->sgt, obj->vaddr, obj->iova, gem->size);
 		if (err < 0)
 			goto free;
 	}
 
-	return sgt;
+	err = dma_map_sgtable(dev, map->sgt, direction, 0);
+	if (err)
+		goto free_sgt;
+
+out:
+	/*
+	 * If we've manually mapped the buffer object through the IOMMU, make sure to return the
+	 * existing IOVA address of our mapping.
+	 */
+	if (!obj->mm) {
+		map->phys = sg_dma_address(map->sgt->sgl);
+		map->chunks = err;
+	} else {
+		map->phys = obj->iova;
+		map->chunks = 1;
+	}
+
+	map->size = gem->size;
 
+	return map;
+
+free_sgt:
+	sg_free_table(map->sgt);
 free:
-	kfree(sgt);
+	kfree(map->sgt);
+	kfree(map);
 	return ERR_PTR(err);
 }
 
-static void tegra_bo_unpin(struct device *dev, struct sg_table *sgt)
+static void tegra_bo_unpin(struct host1x_bo_mapping *map)
 {
-	if (sgt) {
-		sg_free_table(sgt);
-		kfree(sgt);
+	if (map->attach) {
+		dma_buf_unmap_attachment(map->attach, map->sgt, map->direction);
+		dma_buf_detach(map->attach->dmabuf, map->attach);
+	} else {
+		dma_unmap_sgtable(map->dev, map->sgt, map->direction, 0);
+		sg_free_table(map->sgt);
+		kfree(map->sgt);
 	}
+
+	host1x_bo_put(map->bo);
+	kfree(map);
 }
 
 static void *tegra_bo_mmap(struct host1x_bo *bo)
@@ -452,8 +491,18 @@ static struct tegra_bo *tegra_bo_import(struct drm_device *drm,
 void tegra_bo_free_object(struct drm_gem_object *gem)
 {
 	struct tegra_drm *tegra = gem->dev->dev_private;
+	struct host1x_bo_mapping *mapping, *tmp;
 	struct tegra_bo *bo = to_tegra_bo(gem);
 
+	/* remove all mappings of this buffer object from any caches */
+	list_for_each_entry_safe(mapping, tmp, &bo->base.mappings, list) {
+		if (mapping->cache)
+			host1x_bo_unpin(mapping);
+		else
+			dev_err(gem->dev->dev, "mapping %p stale for device %s\n", mapping,
+				dev_name(mapping->dev));
+	}
+
 	if (tegra->domain)
 		tegra_bo_iommu_unmap(tegra, bo);
 

drivers/gpu/drm/tegra/gr2d.c

@@ -4,14 +4,25 @@
  */
 
 #include <linux/clk.h>
+#include <linux/delay.h>
 #include <linux/iommu.h>
 #include <linux/module.h>
 #include <linux/of_device.h>
+#include <linux/pm_runtime.h>
+#include <linux/reset.h>
+
+#include <soc/tegra/common.h>
 
 #include "drm.h"
 #include "gem.h"
 #include "gr2d.h"
 
+enum {
+	RST_MC,
+	RST_GR2D,
+	RST_GR2D_MAX,
+};
+
 struct gr2d_soc {
 	unsigned int version;
 };
@@ -21,6 +32,9 @@ struct gr2d {
 	struct host1x_channel *channel;
 	struct clk *clk;
 
+	struct reset_control_bulk_data resets[RST_GR2D_MAX];
+	unsigned int nresets;
+
 	const struct gr2d_soc *soc;
 
 	DECLARE_BITMAP(addr_regs, GR2D_NUM_REGS);
@@ -56,15 +70,22 @@ static int gr2d_init(struct host1x_client *client)
 		goto free;
 	}
 
+	pm_runtime_enable(client->dev);
+	pm_runtime_use_autosuspend(client->dev);
+	pm_runtime_set_autosuspend_delay(client->dev, 200);
+
 	err = tegra_drm_register_client(dev->dev_private, drm);
 	if (err < 0) {
 		dev_err(client->dev, "failed to register client: %d\n", err);
-		goto detach;
+		goto disable_rpm;
 	}
 
 	return 0;
 
-detach:
+disable_rpm:
+	pm_runtime_dont_use_autosuspend(client->dev);
+	pm_runtime_force_suspend(client->dev);
+
 	host1x_client_iommu_detach(client);
 free:
 	host1x_syncpt_put(client->syncpts[0]);
@@ -85,10 +106,15 @@ static int gr2d_exit(struct host1x_client *client)
 	if (err < 0)
 		return err;
 
+	pm_runtime_dont_use_autosuspend(client->dev);
+	pm_runtime_force_suspend(client->dev);
+
 	host1x_client_iommu_detach(client);
 	host1x_syncpt_put(client->syncpts[0]);
 	host1x_channel_put(gr2d->channel);
 
+	gr2d->channel = NULL;
+
 	return 0;
 }
 
@@ -190,6 +216,27 @@ static const u32 gr2d_addr_regs[] = {
 	GR2D_VA_BASE_ADDR_SB,
 };
 
+static int gr2d_get_resets(struct device *dev, struct gr2d *gr2d)
+{
+	int err;
+
+	gr2d->resets[RST_MC].id = "mc";
+	gr2d->resets[RST_GR2D].id = "2d";
+	gr2d->nresets = RST_GR2D_MAX;
+
+	err = devm_reset_control_bulk_get_optional_exclusive_released(
+				dev, gr2d->nresets, gr2d->resets);
+	if (err) {
+		dev_err(dev, "failed to get reset: %d\n", err);
+		return err;
+	}
+
+	if (WARN_ON(!gr2d->resets[RST_GR2D].rstc))
+		return -ENOENT;
+
+	return 0;
+}
+
 static int gr2d_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
@@ -202,6 +249,8 @@ static int gr2d_probe(struct platform_device *pdev)
 	if (!gr2d)
 		return -ENOMEM;
 
+	platform_set_drvdata(pdev, gr2d);
+
 	gr2d->soc = of_device_get_match_data(dev);
 
 	syncpts = devm_kzalloc(dev, sizeof(*syncpts), GFP_KERNEL);
@@ -214,11 +263,9 @@ static int gr2d_probe(struct platform_device *pdev)
 		return PTR_ERR(gr2d->clk);
 	}
 
-	err = clk_prepare_enable(gr2d->clk);
-	if (err) {
-		dev_err(dev, "cannot turn on clock\n");
+	err = gr2d_get_resets(dev, gr2d);
+	if (err)
 		return err;
-	}
 
 	INIT_LIST_HEAD(&gr2d->client.base.list);
 	gr2d->client.base.ops = &gr2d_client_ops;
@@ -231,10 +278,13 @@ static int gr2d_probe(struct platform_device *pdev)
 	gr2d->client.version = gr2d->soc->version;
 	gr2d->client.ops = &gr2d_ops;
 
+	err = devm_tegra_core_dev_init_opp_table_common(dev);
+	if (err)
+		return err;
+
 	err = host1x_client_register(&gr2d->client.base);
 	if (err < 0) {
 		dev_err(dev, "failed to register host1x client: %d\n", err);
-		clk_disable_unprepare(gr2d->clk);
 		return err;
 	}
 
@@ -242,8 +292,6 @@ static int gr2d_probe(struct platform_device *pdev)
 	for (i = 0; i < ARRAY_SIZE(gr2d_addr_regs); i++)
 		set_bit(gr2d_addr_regs[i], gr2d->addr_regs);
 
-	platform_set_drvdata(pdev, gr2d);
-
 	return 0;
 }
 
@@ -259,15 +307,100 @@ static int gr2d_remove(struct platform_device *pdev)
 		return err;
 	}
 
+	return 0;
+}
+
+static int __maybe_unused gr2d_runtime_suspend(struct device *dev)
+{
+	struct gr2d *gr2d = dev_get_drvdata(dev);
+	int err;
+
+	host1x_channel_stop(gr2d->channel);
+	reset_control_bulk_release(gr2d->nresets, gr2d->resets);
+
+	/*
+	 * GR2D module shouldn't be reset while hardware is idling, otherwise
+	 * host1x's cmdproc will stuck on trying to access any G2 register
+	 * after reset. GR2D module could be either hot-reset or reset after
+	 * power-gating of the HEG partition. Hence we will put in reset only
+	 * the memory client part of the module, the HEG GENPD will take care
+	 * of resetting GR2D module across power-gating.
+	 *
+	 * On Tegra20 there is no HEG partition, but it's okay to have
+	 * undetermined h/w state since userspace is expected to reprogram
+	 * the state on each job submission anyways.
+	 */
+	err = reset_control_acquire(gr2d->resets[RST_MC].rstc);
+	if (err) {
+		dev_err(dev, "failed to acquire MC reset: %d\n", err);
+		goto acquire_reset;
+	}
+
+	err = reset_control_assert(gr2d->resets[RST_MC].rstc);
+	reset_control_release(gr2d->resets[RST_MC].rstc);
+	if (err) {
+		dev_err(dev, "failed to assert MC reset: %d\n", err);
+		goto acquire_reset;
+	}
+
 	clk_disable_unprepare(gr2d->clk);
 
 	return 0;
+
+acquire_reset:
+	reset_control_bulk_acquire(gr2d->nresets, gr2d->resets);
+	reset_control_bulk_deassert(gr2d->nresets, gr2d->resets);
+
+	return err;
 }
 
+static int __maybe_unused gr2d_runtime_resume(struct device *dev)
+{
+	struct gr2d *gr2d = dev_get_drvdata(dev);
+	int err;
+
+	err = reset_control_bulk_acquire(gr2d->nresets, gr2d->resets);
+	if (err) {
+		dev_err(dev, "failed to acquire reset: %d\n", err);
+		return err;
+	}
+
+	err = clk_prepare_enable(gr2d->clk);
+	if (err) {
+		dev_err(dev, "failed to enable clock: %d\n", err);
+		goto release_reset;
+	}
+
+	usleep_range(2000, 4000);
+
+	/* this is a reset array which deasserts both 2D MC and 2D itself */
+	err = reset_control_bulk_deassert(gr2d->nresets, gr2d->resets);
+	if (err) {
+		dev_err(dev, "failed to deassert reset: %d\n", err);
+		goto disable_clk;
+	}
+
+	return 0;
+
+disable_clk:
+	clk_disable_unprepare(gr2d->clk);
+release_reset:
+	reset_control_bulk_release(gr2d->nresets, gr2d->resets);
+
+	return err;
+}
+
+static const struct dev_pm_ops tegra_gr2d_pm = {
+	SET_RUNTIME_PM_OPS(gr2d_runtime_suspend, gr2d_runtime_resume, NULL)
+	SET_SYSTEM_SLEEP_PM_OPS(pm_runtime_force_suspend,
+				pm_runtime_force_resume)
+};
+
 struct platform_driver tegra_gr2d_driver = {
 	.driver = {
 		.name = "tegra-gr2d",
 		.of_match_table = gr2d_match,
+		.pm = &tegra_gr2d_pm,
 	},
 	.probe = gr2d_probe,
 	.remove = gr2d_remove,

drivers/gpu/drm/tegra/hdmi.c

@@ -11,10 +11,14 @@
 #include <linux/math64.h>
 #include <linux/module.h>
 #include <linux/of_device.h>
+#include <linux/pm_opp.h>
 #include <linux/pm_runtime.h>
 #include <linux/regulator/consumer.h>
 #include <linux/reset.h>
 
+#include <soc/tegra/common.h>
+#include <sound/hdmi-codec.h>
+
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_crtc.h>
 #include <drm/drm_debugfs.h>
@@ -78,6 +82,9 @@ struct tegra_hdmi {
 	bool dvi;
 
 	struct drm_info_list *debugfs_files;
+
+	struct platform_device *audio_pdev;
+	struct mutex audio_lock;
 };
 
 static inline struct tegra_hdmi *
@@ -360,6 +367,18 @@ static const struct tmds_config tegra124_tmds_config[] = {
 	},
 };
 
+static void tegra_hdmi_audio_lock(struct tegra_hdmi *hdmi)
+{
+	mutex_lock(&hdmi->audio_lock);
+	disable_irq(hdmi->irq);
+}
+
+static void tegra_hdmi_audio_unlock(struct tegra_hdmi *hdmi)
+{
+	enable_irq(hdmi->irq);
+	mutex_unlock(&hdmi->audio_lock);
+}
+
 static int
 tegra_hdmi_get_audio_config(unsigned int audio_freq, unsigned int pix_clock,
 			    struct tegra_hdmi_audio_config *config)
@@ -829,6 +848,23 @@ static void tegra_hdmi_setup_tmds(struct tegra_hdmi *hdmi,
 				  HDMI_NV_PDISP_SOR_IO_PEAK_CURRENT);
 }
 
+static int tegra_hdmi_reconfigure_audio(struct tegra_hdmi *hdmi)
+{
+	int err;
+
+	err = tegra_hdmi_setup_audio(hdmi);
+	if (err < 0) {
+		tegra_hdmi_disable_audio_infoframe(hdmi);
+		tegra_hdmi_disable_audio(hdmi);
+	} else {
+		tegra_hdmi_setup_audio_infoframe(hdmi);
+		tegra_hdmi_enable_audio_infoframe(hdmi);
+		tegra_hdmi_enable_audio(hdmi);
+	}
+
+	return err;
+}
+
 static bool tegra_output_is_hdmi(struct tegra_output *output)
 {
 	struct edid *edid;
@@ -1135,6 +1171,8 @@ static void tegra_hdmi_encoder_disable(struct drm_encoder *encoder)
 	u32 value;
 	int err;
 
+	tegra_hdmi_audio_lock(hdmi);
+
 	/*
 	 * The following accesses registers of the display controller, so make
 	 * sure it's only executed when the output is attached to one.
@@ -1159,6 +1197,10 @@ static void tegra_hdmi_encoder_disable(struct drm_encoder *encoder)
 	tegra_hdmi_writel(hdmi, 0, HDMI_NV_PDISP_INT_ENABLE);
 	tegra_hdmi_writel(hdmi, 0, HDMI_NV_PDISP_INT_MASK);
 
+	hdmi->pixel_clock = 0;
+
+	tegra_hdmi_audio_unlock(hdmi);
+
 	err = host1x_client_suspend(&hdmi->client);
 	if (err < 0)
 		dev_err(hdmi->dev, "failed to suspend: %d\n", err);
@@ -1182,6 +1224,8 @@ static void tegra_hdmi_encoder_enable(struct drm_encoder *encoder)
 		return;
 	}
 
+	tegra_hdmi_audio_lock(hdmi);
+
 	/*
 	 * Enable and unmask the HDA codec SCRATCH0 register interrupt. This
 	 * is used for interoperability between the HDA codec driver and the
@@ -1195,7 +1239,7 @@ static void tegra_hdmi_encoder_enable(struct drm_encoder *encoder)
 	h_back_porch = mode->htotal - mode->hsync_end;
 	h_front_porch = mode->hsync_start - mode->hdisplay;
 
-	err = clk_set_rate(hdmi->clk, hdmi->pixel_clock);
+	err = dev_pm_opp_set_rate(hdmi->dev, hdmi->pixel_clock);
 	if (err < 0) {
 		dev_err(hdmi->dev, "failed to set HDMI clock frequency: %d\n",
 			err);
@@ -1387,6 +1431,8 @@ static void tegra_hdmi_encoder_enable(struct drm_encoder *encoder)
 	}
 
 	/* TODO: add HDCP support */
+
+	tegra_hdmi_audio_unlock(hdmi);
 }
 
 static int
@@ -1416,6 +1462,91 @@ static const struct drm_encoder_helper_funcs tegra_hdmi_encoder_helper_funcs = {
 	.atomic_check = tegra_hdmi_encoder_atomic_check,
 };
 
+static int tegra_hdmi_hw_params(struct device *dev, void *data,
+				struct hdmi_codec_daifmt *fmt,
+				struct hdmi_codec_params *hparms)
+{
+	struct tegra_hdmi *hdmi = data;
+	int ret = 0;
+
+	tegra_hdmi_audio_lock(hdmi);
+
+	hdmi->format.sample_rate = hparms->sample_rate;
+	hdmi->format.channels = hparms->channels;
+
+	if (hdmi->pixel_clock && !hdmi->dvi)
+		ret = tegra_hdmi_reconfigure_audio(hdmi);
+
+	tegra_hdmi_audio_unlock(hdmi);
+
+	return ret;
+}
+
+static int tegra_hdmi_audio_startup(struct device *dev, void *data)
+{
+	struct tegra_hdmi *hdmi = data;
+	int ret;
+
+	ret = host1x_client_resume(&hdmi->client);
+	if (ret < 0)
+		dev_err(hdmi->dev, "failed to resume: %d\n", ret);
+
+	return ret;
+}
+
+static void tegra_hdmi_audio_shutdown(struct device *dev, void *data)
+{
+	struct tegra_hdmi *hdmi = data;
+	int ret;
+
+	tegra_hdmi_audio_lock(hdmi);
+
+	hdmi->format.sample_rate = 0;
+	hdmi->format.channels = 0;
+
+	tegra_hdmi_audio_unlock(hdmi);
+
+	ret = host1x_client_suspend(&hdmi->client);
+	if (ret < 0)
+		dev_err(hdmi->dev, "failed to suspend: %d\n", ret);
+}
+
+static const struct hdmi_codec_ops tegra_hdmi_codec_ops = {
+	.hw_params = tegra_hdmi_hw_params,
+	.audio_startup = tegra_hdmi_audio_startup,
+	.audio_shutdown = tegra_hdmi_audio_shutdown,
+};
+
+static int tegra_hdmi_codec_register(struct tegra_hdmi *hdmi)
+{
+	struct hdmi_codec_pdata codec_data = {};
+
+	if (hdmi->config->has_hda)
+		return 0;
+
+	codec_data.ops = &tegra_hdmi_codec_ops;
+	codec_data.data = hdmi;
+	codec_data.spdif = 1;
+
+	hdmi->audio_pdev = platform_device_register_data(hdmi->dev,
+							 HDMI_CODEC_DRV_NAME,
+							 PLATFORM_DEVID_AUTO,
+							 &codec_data,
+							 sizeof(codec_data));
+	if (IS_ERR(hdmi->audio_pdev))
+		return PTR_ERR(hdmi->audio_pdev);
+
+	hdmi->format.channels = 2;
+
+	return 0;
+}
+
+static void tegra_hdmi_codec_unregister(struct tegra_hdmi *hdmi)
+{
+	if (hdmi->audio_pdev)
+		platform_device_unregister(hdmi->audio_pdev);
+}
+
 static int tegra_hdmi_init(struct host1x_client *client)
 {
 	struct tegra_hdmi *hdmi = host1x_client_to_hdmi(client);
@@ -1453,28 +1584,47 @@ static int tegra_hdmi_init(struct host1x_client *client)
 	if (err < 0) {
 		dev_err(client->dev, "failed to enable HDMI regulator: %d\n",
 			err);
-		return err;
+		goto output_exit;
 	}
 
 	err = regulator_enable(hdmi->pll);
 	if (err < 0) {
 		dev_err(hdmi->dev, "failed to enable PLL regulator: %d\n", err);
-		return err;
+		goto disable_hdmi;
 	}
 
 	err = regulator_enable(hdmi->vdd);
 	if (err < 0) {
 		dev_err(hdmi->dev, "failed to enable VDD regulator: %d\n", err);
-		return err;
+		goto disable_pll;
+	}
+
+	err = tegra_hdmi_codec_register(hdmi);
+	if (err < 0) {
+		dev_err(hdmi->dev, "failed to register audio codec: %d\n", err);
+		goto disable_vdd;
 	}
 
 	return 0;
+
+disable_vdd:
+	regulator_disable(hdmi->vdd);
+disable_pll:
+	regulator_disable(hdmi->pll);
+disable_hdmi:
+	regulator_disable(hdmi->hdmi);
+output_exit:
+	tegra_output_exit(&hdmi->output);
+
+	return err;
 }
 
 static int tegra_hdmi_exit(struct host1x_client *client)
 {
 	struct tegra_hdmi *hdmi = host1x_client_to_hdmi(client);
 
+	tegra_hdmi_codec_unregister(hdmi);
+
 	tegra_output_exit(&hdmi->output);
 
 	regulator_disable(hdmi->vdd);
@@ -1599,7 +1749,6 @@ static irqreturn_t tegra_hdmi_irq(int irq, void *data)
 {
 	struct tegra_hdmi *hdmi = data;
 	u32 value;
-	int err;
 
 	value = tegra_hdmi_readl(hdmi, HDMI_NV_PDISP_INT_STATUS);
 	tegra_hdmi_writel(hdmi, value, HDMI_NV_PDISP_INT_STATUS);
@@ -1614,16 +1763,7 @@ static irqreturn_t tegra_hdmi_irq(int irq, void *data)
 			format = value & SOR_AUDIO_HDA_CODEC_SCRATCH0_FMT_MASK;
 
 			tegra_hda_parse_format(format, &hdmi->format);
-
-			err = tegra_hdmi_setup_audio(hdmi);
-			if (err < 0) {
-				tegra_hdmi_disable_audio_infoframe(hdmi);
-				tegra_hdmi_disable_audio(hdmi);
-			} else {
-				tegra_hdmi_setup_audio_infoframe(hdmi);
-				tegra_hdmi_enable_audio_infoframe(hdmi);
-				tegra_hdmi_enable_audio(hdmi);
-			}
+			tegra_hdmi_reconfigure_audio(hdmi);
 		} else {
 			tegra_hdmi_disable_audio_infoframe(hdmi);
 			tegra_hdmi_disable_audio(hdmi);
@@ -1651,6 +1791,8 @@ static int tegra_hdmi_probe(struct platform_device *pdev)
 	hdmi->stereo = false;
 	hdmi->dvi = false;
 
+	mutex_init(&hdmi->audio_lock);
+
 	hdmi->clk = devm_clk_get(&pdev->dev, NULL);
 	if (IS_ERR(hdmi->clk)) {
 		dev_err(&pdev->dev, "failed to get clock\n");
@@ -1732,7 +1874,14 @@ static int tegra_hdmi_probe(struct platform_device *pdev)
 	}
 
 	platform_set_drvdata(pdev, hdmi);
-	pm_runtime_enable(&pdev->dev);
+
+	err = devm_pm_runtime_enable(&pdev->dev);
+	if (err)
+		return err;
+
+	err = devm_tegra_core_dev_init_opp_table_common(&pdev->dev);
+	if (err)
+		return err;
 
 	INIT_LIST_HEAD(&hdmi->client.list);
 	hdmi->client.ops = &hdmi_client_ops;
@@ -1753,8 +1902,6 @@ static int tegra_hdmi_remove(struct platform_device *pdev)
 	struct tegra_hdmi *hdmi = platform_get_drvdata(pdev);
 	int err;
 
-	pm_runtime_disable(&pdev->dev);
-
 	err = host1x_client_unregister(&hdmi->client);
 	if (err < 0) {
 		dev_err(&pdev->dev, "failed to unregister host1x client: %d\n",

drivers/gpu/drm/tegra/hub.h

@@ -72,7 +72,6 @@ to_tegra_display_hub_state(struct drm_private_state *priv)
 	return container_of(priv, struct tegra_display_hub_state, base);
 }
 
-struct tegra_dc;
 struct tegra_plane;
 
 int tegra_display_hub_prepare(struct tegra_display_hub *hub);

drivers/gpu/drm/tegra/plane.c

@@ -74,7 +74,7 @@ tegra_plane_atomic_duplicate_state(struct drm_plane *plane)
 
 	for (i = 0; i < 3; i++) {
 		copy->iova[i] = DMA_MAPPING_ERROR;
-		copy->sgt[i] = NULL;
+		copy->map[i] = NULL;
 	}
 
 	return &copy->base;
@@ -138,55 +138,37 @@ const struct drm_plane_funcs tegra_plane_funcs = {
 
 static int tegra_dc_pin(struct tegra_dc *dc, struct tegra_plane_state *state)
 {
-	struct iommu_domain *domain = iommu_get_domain_for_dev(dc->dev);
 	unsigned int i;
 	int err;
 
 	for (i = 0; i < state->base.fb->format->num_planes; i++) {
 		struct tegra_bo *bo = tegra_fb_get_plane(state->base.fb, i);
-		dma_addr_t phys_addr, *phys;
-		struct sg_table *sgt;
+		struct host1x_bo_mapping *map;
 
-		/*
-		 * If we're not attached to a domain, we already stored the
-		 * physical address when the buffer was allocated. If we're
-		 * part of a group that's shared between all display
-		 * controllers, we've also already mapped the framebuffer
-		 * through the SMMU. In both cases we can short-circuit the
-		 * code below and retrieve the stored IOV address.
-		 */
-		if (!domain || dc->client.group)
-			phys = &phys_addr;
-		else
-			phys = NULL;
-
-		sgt = host1x_bo_pin(dc->dev, &bo->base, phys);
-		if (IS_ERR(sgt)) {
-			err = PTR_ERR(sgt);
+		map = host1x_bo_pin(dc->dev, &bo->base, DMA_TO_DEVICE, &dc->client.cache);
+		if (IS_ERR(map)) {
+			err = PTR_ERR(map);
 			goto unpin;
 		}
 
-		if (sgt) {
-			err = dma_map_sgtable(dc->dev, sgt, DMA_TO_DEVICE, 0);
-			if (err)
-				goto unpin;
-
+		if (!dc->client.group) {
 			/*
 			 * The display controller needs contiguous memory, so
 			 * fail if the buffer is discontiguous and we fail to
 			 * map its SG table to a single contiguous chunk of
 			 * I/O virtual memory.
 			 */
-			if (sgt->nents > 1) {
+			if (map->chunks > 1) {
 				err = -EINVAL;
 				goto unpin;
 			}
 
-			state->iova[i] = sg_dma_address(sgt->sgl);
-			state->sgt[i] = sgt;
+			state->iova[i] = map->phys;
 		} else {
-			state->iova[i] = phys_addr;
+			state->iova[i] = bo->iova;
 		}
+
+		state->map[i] = map;
 	}
 
 	return 0;
@@ -195,15 +177,9 @@ static int tegra_dc_pin(struct tegra_dc *dc, struct tegra_plane_state *state)
 	dev_err(dc->dev, "failed to map plane %u: %d\n", i, err);
 
 	while (i--) {
-		struct tegra_bo *bo = tegra_fb_get_plane(state->base.fb, i);
-		struct sg_table *sgt = state->sgt[i];
-
-		if (sgt)
-			dma_unmap_sgtable(dc->dev, sgt, DMA_TO_DEVICE, 0);
-
-		host1x_bo_unpin(dc->dev, &bo->base, sgt);
+		host1x_bo_unpin(state->map[i]);
 		state->iova[i] = DMA_MAPPING_ERROR;
-		state->sgt[i] = NULL;
+		state->map[i] = NULL;
 	}
 
 	return err;
@@ -214,15 +190,9 @@ static void tegra_dc_unpin(struct tegra_dc *dc, struct tegra_plane_state *state)
 	unsigned int i;
 
 	for (i = 0; i < state->base.fb->format->num_planes; i++) {
-		struct tegra_bo *bo = tegra_fb_get_plane(state->base.fb, i);
-		struct sg_table *sgt = state->sgt[i];
-
-		if (sgt)
-			dma_unmap_sgtable(dc->dev, sgt, DMA_TO_DEVICE, 0);
-
-		host1x_bo_unpin(dc->dev, &bo->base, sgt);
+		host1x_bo_unpin(state->map[i]);
 		state->iova[i] = DMA_MAPPING_ERROR;
-		state->sgt[i] = NULL;
+		state->map[i] = NULL;
 	}
 }
 
@@ -230,11 +200,14 @@ int tegra_plane_prepare_fb(struct drm_plane *plane,
 			   struct drm_plane_state *state)
 {
 	struct tegra_dc *dc = to_tegra_dc(state->crtc);
+	int err;
 
 	if (!state->fb)
 		return 0;
 
-	drm_gem_plane_helper_prepare_fb(plane, state);
+	err = drm_gem_plane_helper_prepare_fb(plane, state);
+	if (err < 0)
+		return err;
 
 	return tegra_dc_pin(dc, to_tegra_plane_state(state));
 }

drivers/gpu/drm/tegra/plane.h

@@ -43,7 +43,7 @@ struct tegra_plane_legacy_blending_state {
 struct tegra_plane_state {
 	struct drm_plane_state base;
 
-	struct sg_table *sgt[3];
+	struct host1x_bo_mapping *map[3];
 	dma_addr_t iova[3];
 
 	struct tegra_bo_tiling tiling;

drivers/gpu/drm/tegra/rgb.c

@@ -17,6 +17,8 @@ struct tegra_rgb {
 	struct tegra_output output;
 	struct tegra_dc *dc;
 
+	struct clk *pll_d_out0;
+	struct clk *pll_d2_out0;
 	struct clk *clk_parent;
 	struct clk *clk;
 };
@@ -116,13 +118,21 @@ static void tegra_rgb_encoder_enable(struct drm_encoder *encoder)
 		DISP_ORDER_RED_BLUE;
 	tegra_dc_writel(rgb->dc, value, DC_DISP_DISP_INTERFACE_CONTROL);
 
-	/* XXX: parameterize? */
-	value = SC0_H_QUALIFIER_NONE | SC1_H_QUALIFIER_NONE;
-	tegra_dc_writel(rgb->dc, value, DC_DISP_SHIFT_CLOCK_OPTIONS);
-
 	tegra_dc_commit(rgb->dc);
 }
 
+static bool tegra_rgb_pll_rate_change_allowed(struct tegra_rgb *rgb)
+{
+	if (!rgb->pll_d2_out0)
+		return false;
+
+	if (!clk_is_match(rgb->clk_parent, rgb->pll_d_out0) &&
+	    !clk_is_match(rgb->clk_parent, rgb->pll_d2_out0))
+		return false;
+
+	return true;
+}
+
 static int
 tegra_rgb_encoder_atomic_check(struct drm_encoder *encoder,
 			       struct drm_crtc_state *crtc_state,
@@ -151,8 +161,17 @@ tegra_rgb_encoder_atomic_check(struct drm_encoder *encoder,
 	 * and hope that the desired frequency can be matched (or at least
 	 * matched sufficiently close that the panel will still work).
 	 */
-	div = ((clk_get_rate(rgb->clk) * 2) / pclk) - 2;
-	pclk = 0;
+	if (tegra_rgb_pll_rate_change_allowed(rgb)) {
+		/*
+		 * Set display controller clock to x2 of PCLK in order to
+		 * produce higher resolution pulse positions.
+		 */
+		div = 2;
+		pclk *= 2;
+	} else {
+		div = ((clk_get_rate(rgb->clk) * 2) / pclk) - 2;
+		pclk = 0;
+	}
 
 	err = tegra_dc_state_setup_clock(dc, crtc_state, rgb->clk_parent,
 					 pclk, div);
@@ -210,6 +229,22 @@ int tegra_dc_rgb_probe(struct tegra_dc *dc)
 		return err;
 	}
 
+	rgb->pll_d_out0 = clk_get_sys(NULL, "pll_d_out0");
+	if (IS_ERR(rgb->pll_d_out0)) {
+		err = PTR_ERR(rgb->pll_d_out0);
+		dev_err(dc->dev, "failed to get pll_d_out0: %d\n", err);
+		return err;
+	}
+
+	if (dc->soc->has_pll_d2_out0) {
+		rgb->pll_d2_out0 = clk_get_sys(NULL, "pll_d2_out0");
+		if (IS_ERR(rgb->pll_d2_out0)) {
+			err = PTR_ERR(rgb->pll_d2_out0);
+			dev_err(dc->dev, "failed to get pll_d2_out0: %d\n", err);
+			return err;
+		}
+	}
+
 	dc->rgb = &rgb->output;
 
 	return 0;
@@ -217,9 +252,15 @@ int tegra_dc_rgb_probe(struct tegra_dc *dc)
 
 int tegra_dc_rgb_remove(struct tegra_dc *dc)
 {
+	struct tegra_rgb *rgb;
+
 	if (!dc->rgb)
 		return 0;
 
+	rgb = to_rgb(dc->rgb);
+	clk_put(rgb->pll_d2_out0);
+	clk_put(rgb->pll_d_out0);
+
 	tegra_output_remove(dc->rgb);
 	dc->rgb = NULL;
 

drivers/gpu/drm/tegra/submit.c

@@ -64,33 +64,62 @@ static void gather_bo_put(struct host1x_bo *host_bo)
 	kref_put(&bo->ref, gather_bo_release);
 }
 
-static struct sg_table *
-gather_bo_pin(struct device *dev, struct host1x_bo *host_bo, dma_addr_t *phys)
+static struct host1x_bo_mapping *
+gather_bo_pin(struct device *dev, struct host1x_bo *bo, enum dma_data_direction direction)
 {
-	struct gather_bo *bo = container_of(host_bo, struct gather_bo, base);
-	struct sg_table *sgt;
+	struct gather_bo *gather = container_of(bo, struct gather_bo, base);
+	struct host1x_bo_mapping *map;
 	int err;
 
-	sgt = kzalloc(sizeof(*sgt), GFP_KERNEL);
-	if (!sgt)
+	map = kzalloc(sizeof(*map), GFP_KERNEL);
+	if (!map)
 		return ERR_PTR(-ENOMEM);
 
-	err = dma_get_sgtable(bo->dev, sgt, bo->gather_data, bo->gather_data_dma,
-			      bo->gather_data_words * 4);
-	if (err) {
-		kfree(sgt);
-		return ERR_PTR(err);
+	kref_init(&map->ref);
+	map->bo = host1x_bo_get(bo);
+	map->direction = direction;
+	map->dev = dev;
+
+	map->sgt = kzalloc(sizeof(*map->sgt), GFP_KERNEL);
+	if (!map->sgt) {
+		err = -ENOMEM;
+		goto free;
 	}
 
-	return sgt;
+	err = dma_get_sgtable(gather->dev, map->sgt, gather->gather_data, gather->gather_data_dma,
+			      gather->gather_data_words * 4);
+	if (err)
+		goto free_sgt;
+
+	err = dma_map_sgtable(dev, map->sgt, direction, 0);
+	if (err)
+		goto free_sgt;
+
+	map->phys = sg_dma_address(map->sgt->sgl);
+	map->size = gather->gather_data_words * 4;
+	map->chunks = err;
+
+	return map;
+
+free_sgt:
+	sg_free_table(map->sgt);
+	kfree(map->sgt);
+free:
+	kfree(map);
+	return ERR_PTR(err);
 }
 
-static void gather_bo_unpin(struct device *dev, struct sg_table *sgt)
+static void gather_bo_unpin(struct host1x_bo_mapping *map)
 {
-	if (sgt) {
-		sg_free_table(sgt);
-		kfree(sgt);
-	}
+	if (!map)
+		return;
+
+	dma_unmap_sgtable(map->dev, map->sgt, map->direction, 0);
+	sg_free_table(map->sgt);
+	kfree(map->sgt);
+	host1x_bo_put(map->bo);
+
+	kfree(map);
 }
 
 static void *gather_bo_mmap(struct host1x_bo *host_bo)
@@ -475,8 +504,8 @@ static void release_job(struct host1x_job *job)
 	kfree(job_data->used_mappings);
 	kfree(job_data);
 
-	if (pm_runtime_enabled(client->base.dev))
-		pm_runtime_put_autosuspend(client->base.dev);
+	pm_runtime_mark_last_busy(client->base.dev);
+	pm_runtime_put_autosuspend(client->base.dev);
 }
 
 int tegra_drm_ioctl_channel_submit(struct drm_device *drm, void *data,
@@ -560,12 +589,10 @@ int tegra_drm_ioctl_channel_submit(struct drm_device *drm, void *data,
 	}
 
 	/* Boot engine. */
-	if (pm_runtime_enabled(context->client->base.dev)) {
-		err = pm_runtime_resume_and_get(context->client->base.dev);
-		if (err < 0) {
-			SUBMIT_ERR(context, "could not power up engine: %d", err);
-			goto unpin_job;
-		}
+	err = pm_runtime_resume_and_get(context->client->base.dev);
+	if (err < 0) {
+		SUBMIT_ERR(context, "could not power up engine: %d", err);
+		goto unpin_job;
 	}
 
 	job->user_data = job_data;

drivers/gpu/drm/tegra/uapi.c

@@ -17,11 +17,7 @@ static void tegra_drm_mapping_release(struct kref *ref)
 	struct tegra_drm_mapping *mapping =
 		container_of(ref, struct tegra_drm_mapping, ref);
 
-	if (mapping->sgt)
-		dma_unmap_sgtable(mapping->dev, mapping->sgt, mapping->direction,
-				  DMA_ATTR_SKIP_CPU_SYNC);
-
-	host1x_bo_unpin(mapping->dev, mapping->bo, mapping->sgt);
+	host1x_bo_unpin(mapping->map);
 	host1x_bo_put(mapping->bo);
 
 	kfree(mapping);
@@ -159,6 +155,7 @@ int tegra_drm_ioctl_channel_map(struct drm_device *drm, void *data, struct drm_f
 	struct drm_tegra_channel_map *args = data;
 	struct tegra_drm_mapping *mapping;
 	struct tegra_drm_context *context;
+	enum dma_data_direction direction;
 	int err = 0;
 
 	if (args->flags & ~DRM_TEGRA_CHANNEL_MAP_READ_WRITE)
@@ -180,68 +177,53 @@ int tegra_drm_ioctl_channel_map(struct drm_device *drm, void *data, struct drm_f
 
 	kref_init(&mapping->ref);
 
-	mapping->dev = context->client->base.dev;
 	mapping->bo = tegra_gem_lookup(file, args->handle);
 	if (!mapping->bo) {
 		err = -EINVAL;
-		goto unlock;
+		goto free;
 	}
 
-	if (context->client->base.group) {
-		/* IOMMU domain managed directly using IOMMU API */
-		host1x_bo_pin(mapping->dev, mapping->bo, &mapping->iova);
-	} else {
-		switch (args->flags & DRM_TEGRA_CHANNEL_MAP_READ_WRITE) {
-		case DRM_TEGRA_CHANNEL_MAP_READ_WRITE:
-			mapping->direction = DMA_BIDIRECTIONAL;
-			break;
-
-		case DRM_TEGRA_CHANNEL_MAP_WRITE:
-			mapping->direction = DMA_FROM_DEVICE;
-			break;
-
-		case DRM_TEGRA_CHANNEL_MAP_READ:
-			mapping->direction = DMA_TO_DEVICE;
-			break;
+	switch (args->flags & DRM_TEGRA_CHANNEL_MAP_READ_WRITE) {
+	case DRM_TEGRA_CHANNEL_MAP_READ_WRITE:
+		direction = DMA_BIDIRECTIONAL;
+		break;
 
-		default:
-			return -EINVAL;
-		}
+	case DRM_TEGRA_CHANNEL_MAP_WRITE:
+		direction = DMA_FROM_DEVICE;
+		break;
 
-		mapping->sgt = host1x_bo_pin(mapping->dev, mapping->bo, NULL);
-		if (IS_ERR(mapping->sgt)) {
-			err = PTR_ERR(mapping->sgt);
-			goto put_gem;
-		}
+	case DRM_TEGRA_CHANNEL_MAP_READ:
+		direction = DMA_TO_DEVICE;
+		break;
 
-		err = dma_map_sgtable(mapping->dev, mapping->sgt, mapping->direction,
-				      DMA_ATTR_SKIP_CPU_SYNC);
-		if (err)
-			goto unpin;
+	default:
+		err = -EINVAL;
+		goto put_gem;
+	}
 
-		mapping->iova = sg_dma_address(mapping->sgt->sgl);
+	mapping->map = host1x_bo_pin(context->client->base.dev, mapping->bo, direction, NULL);
+	if (IS_ERR(mapping->map)) {
+		err = PTR_ERR(mapping->map);
+		goto put_gem;
 	}
 
+	mapping->iova = mapping->map->phys;
 	mapping->iova_end = mapping->iova + host1x_to_tegra_bo(mapping->bo)->gem.size;
 
 	err = xa_alloc(&context->mappings, &args->mapping, mapping, XA_LIMIT(1, U32_MAX),
 		       GFP_KERNEL);
 	if (err < 0)
-		goto unmap;
+		goto unpin;
 
 	mutex_unlock(&fpriv->lock);
 
 	return 0;
 
-unmap:
-	if (mapping->sgt) {
-		dma_unmap_sgtable(mapping->dev, mapping->sgt, mapping->direction,
-				  DMA_ATTR_SKIP_CPU_SYNC);
-	}
 unpin:
-	host1x_bo_unpin(mapping->dev, mapping->bo, mapping->sgt);
+	host1x_bo_unpin(mapping->map);
 put_gem:
 	host1x_bo_put(mapping->bo);
+free:
 	kfree(mapping);
 unlock:
 	mutex_unlock(&fpriv->lock);

drivers/gpu/drm/tegra/uapi.h

@@ -27,10 +27,9 @@ struct tegra_drm_file {
 struct tegra_drm_mapping {
 	struct kref ref;
 
-	struct device *dev;
+	struct host1x_bo_mapping *map;
 	struct host1x_bo *bo;
-	struct sg_table *sgt;
-	enum dma_data_direction direction;
+
 	dma_addr_t iova;
 	dma_addr_t iova_end;
 };

drivers/gpu/drm/tegra/vic.c

@@ -5,6 +5,7 @@
 
 #include <linux/clk.h>
 #include <linux/delay.h>
+#include <linux/dma-mapping.h>
 #include <linux/host1x.h>
 #include <linux/iommu.h>
 #include <linux/module.h>
@@ -151,9 +152,13 @@ static int vic_init(struct host1x_client *client)
 		goto free_channel;
 	}
 
+	pm_runtime_enable(client->dev);
+	pm_runtime_use_autosuspend(client->dev);
+	pm_runtime_set_autosuspend_delay(client->dev, 500);
+
 	err = tegra_drm_register_client(tegra, drm);
 	if (err < 0)
-		goto free_syncpt;
+		goto disable_rpm;
 
 	/*
 	 * Inherit the DMA parameters (such as maximum segment size) from the
@@ -163,7 +168,10 @@ static int vic_init(struct host1x_client *client)
 
 	return 0;
 
-free_syncpt:
+disable_rpm:
+	pm_runtime_dont_use_autosuspend(client->dev);
+	pm_runtime_force_suspend(client->dev);
+
 	host1x_syncpt_put(client->syncpts[0]);
 free_channel:
 	host1x_channel_put(vic->channel);
@@ -188,10 +196,15 @@ static int vic_exit(struct host1x_client *client)
 	if (err < 0)
 		return err;
 
+	pm_runtime_dont_use_autosuspend(client->dev);
+	pm_runtime_force_suspend(client->dev);
+
 	host1x_syncpt_put(client->syncpts[0]);
 	host1x_channel_put(vic->channel);
 	host1x_client_iommu_detach(client);
 
+	vic->channel = NULL;
+
 	if (client->group) {
 		dma_unmap_single(vic->dev, vic->falcon.firmware.phys,
 				 vic->falcon.firmware.size, DMA_TO_DEVICE);
@@ -232,12 +245,12 @@ static int vic_load_firmware(struct vic *vic)
 
 	if (!client->group) {
 		virt = dma_alloc_coherent(vic->dev, size, &iova, GFP_KERNEL);
-
-		err = dma_mapping_error(vic->dev, iova);
-		if (err < 0)
-			return err;
+		if (!virt)
+			return -ENOMEM;
 	} else {
 		virt = tegra_drm_alloc(tegra, size, &iova);
+		if (IS_ERR(virt))
+			return PTR_ERR(virt);
 	}
 
 	vic->falcon.firmware.virt = virt;
@@ -315,6 +328,8 @@ static int vic_runtime_suspend(struct device *dev)
 	struct vic *vic = dev_get_drvdata(dev);
 	int err;
 
+	host1x_channel_stop(vic->channel);
+
 	err = reset_control_assert(vic->rst);
 	if (err < 0)
 		return err;
@@ -330,27 +345,17 @@ static int vic_open_channel(struct tegra_drm_client *client,
 			    struct tegra_drm_context *context)
 {
 	struct vic *vic = to_vic(client);
-	int err;
-
-	err = pm_runtime_resume_and_get(vic->dev);
-	if (err < 0)
-		return err;
 
 	context->channel = host1x_channel_get(vic->channel);
-	if (!context->channel) {
-		pm_runtime_put(vic->dev);
+	if (!context->channel)
 		return -ENOMEM;
-	}
 
 	return 0;
 }
 
 static void vic_close_channel(struct tegra_drm_context *context)
 {
-	struct vic *vic = to_vic(context->client);
-
 	host1x_channel_put(context->channel);
-	pm_runtime_put(vic->dev);
 }
 
 static const struct tegra_drm_client_ops vic_ops = {
@@ -441,6 +446,12 @@ static int vic_probe(struct platform_device *pdev)
 		return PTR_ERR(vic->clk);
 	}
 
+	err = clk_set_rate(vic->clk, ULONG_MAX);
+	if (err < 0) {
+		dev_err(&pdev->dev, "failed to set clock rate\n");
+		return err;
+	}
+
 	if (!dev->pm_domain) {
 		vic->rst = devm_reset_control_get(dev, "vic");
 		if (IS_ERR(vic->rst)) {
@@ -476,17 +487,8 @@ static int vic_probe(struct platform_device *pdev)
 		goto exit_falcon;
 	}
 
-	pm_runtime_enable(&pdev->dev);
-	if (!pm_runtime_enabled(&pdev->dev)) {
-		err = vic_runtime_resume(&pdev->dev);
-		if (err < 0)
-			goto unregister_client;
-	}
-
 	return 0;
 
-unregister_client:
-	host1x_client_unregister(&vic->client.base);
 exit_falcon:
 	falcon_exit(&vic->falcon);
 
@@ -505,11 +507,6 @@ static int vic_remove(struct platform_device *pdev)
 		return err;
 	}
 
-	if (pm_runtime_enabled(&pdev->dev))
-		pm_runtime_disable(&pdev->dev);
-	else
-		vic_runtime_suspend(&pdev->dev);
-
 	falcon_exit(&vic->falcon);
 
 	return 0;
@@ -517,6 +514,8 @@ static int vic_remove(struct platform_device *pdev)
 
 static const struct dev_pm_ops vic_pm_ops = {
 	SET_RUNTIME_PM_OPS(vic_runtime_suspend, vic_runtime_resume, NULL)
+	SET_SYSTEM_SLEEP_PM_OPS(pm_runtime_force_suspend,
+				pm_runtime_force_resume)
 };
 
 struct platform_driver tegra_vic_driver = {

drivers/gpu/host1x/Kconfig

@@ -2,6 +2,7 @@
 config TEGRA_HOST1X
 	tristate "NVIDIA Tegra host1x driver"
 	depends on ARCH_TEGRA || (ARM && COMPILE_TEST)
+	select DMA_SHARED_BUFFER
 	select IOMMU_IOVA
 	help
 	  Driver for the NVIDIA Tegra host1x hardware.

drivers/gpu/host1x/bus.c

@@ -5,6 +5,7 @@
  */
 
 #include <linux/debugfs.h>
+#include <linux/dma-mapping.h>
 #include <linux/host1x.h>
 #include <linux/of.h>
 #include <linux/seq_file.h>
@@ -742,6 +743,7 @@ EXPORT_SYMBOL(host1x_driver_unregister);
  */
 void __host1x_client_init(struct host1x_client *client, struct lock_class_key *key)
 {
+	host1x_bo_cache_init(&client->cache);
 	INIT_LIST_HEAD(&client->list);
 	__mutex_init(&client->lock, "host1x client lock", key);
 	client->usecount = 0;
@@ -761,7 +763,6 @@ EXPORT_SYMBOL(host1x_client_exit);
 /**
  * __host1x_client_register() - register a host1x client
  * @client: host1x client
- * @key: lock class key for the client-specific mutex
  *
  * Registers a host1x client with each host1x controller instance. Note that
  * each client will only match their parent host1x controller and will only be
@@ -830,6 +831,8 @@ int host1x_client_unregister(struct host1x_client *client)
 
 	mutex_unlock(&clients_lock);
 
+	host1x_bo_cache_destroy(&client->cache);
+
 	return 0;
 }
 EXPORT_SYMBOL(host1x_client_unregister);
@@ -904,3 +907,78 @@ int host1x_client_resume(struct host1x_client *client)
 	return err;
 }
 EXPORT_SYMBOL(host1x_client_resume);
+
+struct host1x_bo_mapping *host1x_bo_pin(struct device *dev, struct host1x_bo *bo,
+					enum dma_data_direction dir,
+					struct host1x_bo_cache *cache)
+{
+	struct host1x_bo_mapping *mapping;
+
+	if (cache) {
+		mutex_lock(&cache->lock);
+
+		list_for_each_entry(mapping, &cache->mappings, entry) {
+			if (mapping->bo == bo && mapping->direction == dir) {
+				kref_get(&mapping->ref);
+				goto unlock;
+			}
+		}
+	}
+
+	mapping = bo->ops->pin(dev, bo, dir);
+	if (IS_ERR(mapping))
+		goto unlock;
+
+	spin_lock(&mapping->bo->lock);
+	list_add_tail(&mapping->list, &bo->mappings);
+	spin_unlock(&mapping->bo->lock);
+
+	if (cache) {
+		INIT_LIST_HEAD(&mapping->entry);
+		mapping->cache = cache;
+
+		list_add_tail(&mapping->entry, &cache->mappings);
+
+		/* bump reference count to track the copy in the cache */
+		kref_get(&mapping->ref);
+	}
+
+unlock:
+	if (cache)
+		mutex_unlock(&cache->lock);
+
+	return mapping;
+}
+EXPORT_SYMBOL(host1x_bo_pin);
+
+static void __host1x_bo_unpin(struct kref *ref)
+{
+	struct host1x_bo_mapping *mapping = to_host1x_bo_mapping(ref);
+
+	/*
+	 * When the last reference of the mapping goes away, make sure to remove the mapping from
+	 * the cache.
+	 */
+	if (mapping->cache)
+		list_del(&mapping->entry);
+
+	spin_lock(&mapping->bo->lock);
+	list_del(&mapping->list);
+	spin_unlock(&mapping->bo->lock);
+
+	mapping->bo->ops->unpin(mapping);
+}
+
+void host1x_bo_unpin(struct host1x_bo_mapping *mapping)
+{
+	struct host1x_bo_cache *cache = mapping->cache;
+
+	if (cache)
+		mutex_lock(&cache->lock);
+
+	kref_put(&mapping->ref, __host1x_bo_unpin);
+
+	if (cache)
+		mutex_unlock(&cache->lock);
+}
+EXPORT_SYMBOL(host1x_bo_unpin);

drivers/gpu/host1x/channel.c

@@ -75,6 +75,14 @@ struct host1x_channel *host1x_channel_get_index(struct host1x *host,
 	return ch;
 }
 
+void host1x_channel_stop(struct host1x_channel *channel)
+{
+	struct host1x *host = dev_get_drvdata(channel->dev->parent);
+
+	host1x_hw_cdma_stop(host, &channel->cdma);
+}
+EXPORT_SYMBOL(host1x_channel_stop);
+
 static void release_channel(struct kref *kref)
 {
 	struct host1x_channel *channel =

drivers/gpu/host1x/debug.c

@@ -7,6 +7,7 @@
  */
 
 #include <linux/debugfs.h>
+#include <linux/pm_runtime.h>
 #include <linux/seq_file.h>
 #include <linux/uaccess.h>
 
@@ -52,6 +53,11 @@ static int show_channel(struct host1x_channel *ch, void *data, bool show_fifo)
 {
 	struct host1x *m = dev_get_drvdata(ch->dev->parent);
 	struct output *o = data;
+	int err;
+
+	err = pm_runtime_resume_and_get(m->dev);
+	if (err < 0)
+		return err;
 
 	mutex_lock(&ch->cdma.lock);
 	mutex_lock(&debug_lock);
@@ -64,6 +70,8 @@ static int show_channel(struct host1x_channel *ch, void *data, bool show_fifo)
 	mutex_unlock(&debug_lock);
 	mutex_unlock(&ch->cdma.lock);
 
+	pm_runtime_put(m->dev);
+
 	return 0;
 }
 
@@ -71,9 +79,14 @@ static void show_syncpts(struct host1x *m, struct output *o)
 {
 	struct list_head *pos;
 	unsigned int i;
+	int err;
 
 	host1x_debug_output(o, "---- syncpts ----\n");
 
+	err = pm_runtime_resume_and_get(m->dev);
+	if (err < 0)
+		return;
+
 	for (i = 0; i < host1x_syncpt_nb_pts(m); i++) {
 		u32 max = host1x_syncpt_read_max(m->syncpt + i);
 		u32 min = host1x_syncpt_load(m->syncpt + i);
@@ -101,6 +114,8 @@ static void show_syncpts(struct host1x *m, struct output *o)
 					    base_val);
 	}
 
+	pm_runtime_put(m->dev);
+
 	host1x_debug_output(o, "\n");
 }
 

drivers/gpu/host1x/dev.c

@@ -6,18 +6,26 @@
  */
 
 #include <linux/clk.h>
+#include <linux/delay.h>
 #include <linux/dma-mapping.h>
 #include <linux/io.h>
 #include <linux/list.h>
 #include <linux/module.h>
 #include <linux/of_device.h>
 #include <linux/of.h>
+#include <linux/pm_runtime.h>
 #include <linux/slab.h>
 
+#include <soc/tegra/common.h>
+
 #define CREATE_TRACE_POINTS
 #include <trace/events/host1x.h>
 #undef CREATE_TRACE_POINTS
 
+#if IS_ENABLED(CONFIG_ARM_DMA_USE_IOMMU)
+#include <asm/dma-iommu.h>
+#endif
+
 #include "bus.h"
 #include "channel.h"
 #include "debug.h"
@@ -132,6 +140,12 @@ static const struct host1x_sid_entry tegra186_sid_table[] = {
 		.offset = 0x30,
 		.limit = 0x34
 	},
+	{
+		/* NVDEC */
+		.base = 0x1b00,
+		.offset = 0x30,
+		.limit = 0x34
+	},
 };
 
 static const struct host1x_info host1x06_info = {
@@ -156,6 +170,18 @@ static const struct host1x_sid_entry tegra194_sid_table[] = {
 		.offset = 0x30,
 		.limit = 0x34
 	},
+	{
+		/* NVDEC */
+		.base = 0x1b00,
+		.offset = 0x30,
+		.limit = 0x34
+	},
+	{
+		/* NVDEC1 */
+		.base = 0x1bc0,
+		.offset = 0x30,
+		.limit = 0x34
+	},
 };
 
 static const struct host1x_info host1x07_info = {
@@ -190,6 +216,9 @@ static void host1x_setup_sid_table(struct host1x *host)
 	const struct host1x_info *info = host->info;
 	unsigned int i;
 
+	if (!info->has_hypervisor)
+		return;
+
 	for (i = 0; i < info->num_sid_entries; i++) {
 		const struct host1x_sid_entry *entry = &info->sid_table[i];
 
@@ -238,6 +267,17 @@ static struct iommu_domain *host1x_iommu_attach(struct host1x *host)
 	struct iommu_domain *domain = iommu_get_domain_for_dev(host->dev);
 	int err;
 
+#if IS_ENABLED(CONFIG_ARM_DMA_USE_IOMMU)
+	if (host->dev->archdata.mapping) {
+		struct dma_iommu_mapping *mapping =
+				to_dma_iommu_mapping(host->dev);
+		arm_iommu_detach_device(host->dev);
+		arm_iommu_release_mapping(mapping);
+
+		domain = iommu_get_domain_for_dev(host->dev);
+	}
+#endif
+
 	/*
 	 * We may not always want to enable IOMMU support (for example if the
 	 * host1x firewall is already enabled and we don't support addressing
@@ -347,6 +387,27 @@ static void host1x_iommu_exit(struct host1x *host)
 	}
 }
 
+static int host1x_get_resets(struct host1x *host)
+{
+	int err;
+
+	host->resets[0].id = "mc";
+	host->resets[1].id = "host1x";
+	host->nresets = ARRAY_SIZE(host->resets);
+
+	err = devm_reset_control_bulk_get_optional_exclusive_released(
+				host->dev, host->nresets, host->resets);
+	if (err) {
+		dev_err(host->dev, "failed to get reset: %d\n", err);
+		return err;
+	}
+
+	if (WARN_ON(!host->resets[1].rstc))
+		return -ENOENT;
+
+	return 0;
+}
+
 static int host1x_probe(struct platform_device *pdev)
 {
 	struct host1x *host;
@@ -386,6 +447,7 @@ static int host1x_probe(struct platform_device *pdev)
 	if (syncpt_irq < 0)
 		return syncpt_irq;
 
+	host1x_bo_cache_init(&host->cache);
 	mutex_init(&host->devices_lock);
 	INIT_LIST_HEAD(&host->devices);
 	INIT_LIST_HEAD(&host->list);
@@ -423,12 +485,9 @@ static int host1x_probe(struct platform_device *pdev)
 		return err;
 	}
 
-	host->rst = devm_reset_control_get(&pdev->dev, "host1x");
-	if (IS_ERR(host->rst)) {
-		err = PTR_ERR(host->rst);
-		dev_err(&pdev->dev, "failed to get reset: %d\n", err);
+	err = host1x_get_resets(host);
+	if (err)
 		return err;
-	}
 
 	err = host1x_iommu_init(host);
 	if (err < 0) {
@@ -443,22 +502,10 @@ static int host1x_probe(struct platform_device *pdev)
 		goto iommu_exit;
 	}
 
-	err = clk_prepare_enable(host->clk);
-	if (err < 0) {
-		dev_err(&pdev->dev, "failed to enable clock\n");
-		goto free_channels;
-	}
-
-	err = reset_control_deassert(host->rst);
-	if (err < 0) {
-		dev_err(&pdev->dev, "failed to deassert reset: %d\n", err);
-		goto unprepare_disable;
-	}
-
 	err = host1x_syncpt_init(host);
 	if (err) {
 		dev_err(&pdev->dev, "failed to initialize syncpts\n");
-		goto reset_assert;
+		goto free_channels;
 	}
 
 	err = host1x_intr_init(host, syncpt_irq);
@@ -467,10 +514,18 @@ static int host1x_probe(struct platform_device *pdev)
 		goto deinit_syncpt;
 	}
 
-	host1x_debug_init(host);
+	pm_runtime_enable(&pdev->dev);
+
+	err = devm_tegra_core_dev_init_opp_table_common(&pdev->dev);
+	if (err)
+		goto pm_disable;
 
-	if (host->info->has_hypervisor)
-		host1x_setup_sid_table(host);
+	/* the driver's code isn't ready yet for the dynamic RPM */
+	err = pm_runtime_resume_and_get(&pdev->dev);
+	if (err)
+		goto pm_disable;
+
+	host1x_debug_init(host);
 
 	err = host1x_register(host);
 	if (err < 0)
@@ -486,13 +541,14 @@ static int host1x_probe(struct platform_device *pdev)
 	host1x_unregister(host);
 deinit_debugfs:
 	host1x_debug_deinit(host);
+
+	pm_runtime_put_sync_suspend(&pdev->dev);
+pm_disable:
+	pm_runtime_disable(&pdev->dev);
+
 	host1x_intr_deinit(host);
 deinit_syncpt:
 	host1x_syncpt_deinit(host);
-reset_assert:
-	reset_control_assert(host->rst);
-unprepare_disable:
-	clk_disable_unprepare(host->clk);
 free_channels:
 	host1x_channel_list_free(&host->channel_list);
 iommu_exit:
@@ -507,19 +563,94 @@ static int host1x_remove(struct platform_device *pdev)
 
 	host1x_unregister(host);
 	host1x_debug_deinit(host);
+
+	pm_runtime_force_suspend(&pdev->dev);
+
 	host1x_intr_deinit(host);
 	host1x_syncpt_deinit(host);
-	reset_control_assert(host->rst);
-	clk_disable_unprepare(host->clk);
 	host1x_iommu_exit(host);
+	host1x_bo_cache_destroy(&host->cache);
 
 	return 0;
 }
 
+static int __maybe_unused host1x_runtime_suspend(struct device *dev)
+{
+	struct host1x *host = dev_get_drvdata(dev);
+	int err;
+
+	host1x_intr_stop(host);
+	host1x_syncpt_save(host);
+
+	err = reset_control_bulk_assert(host->nresets, host->resets);
+	if (err) {
+		dev_err(dev, "failed to assert reset: %d\n", err);
+		goto resume_host1x;
+	}
+
+	usleep_range(1000, 2000);
+
+	clk_disable_unprepare(host->clk);
+	reset_control_bulk_release(host->nresets, host->resets);
+
+	return 0;
+
+resume_host1x:
+	host1x_setup_sid_table(host);
+	host1x_syncpt_restore(host);
+	host1x_intr_start(host);
+
+	return err;
+}
+
+static int __maybe_unused host1x_runtime_resume(struct device *dev)
+{
+	struct host1x *host = dev_get_drvdata(dev);
+	int err;
+
+	err = reset_control_bulk_acquire(host->nresets, host->resets);
+	if (err) {
+		dev_err(dev, "failed to acquire reset: %d\n", err);
+		return err;
+	}
+
+	err = clk_prepare_enable(host->clk);
+	if (err) {
+		dev_err(dev, "failed to enable clock: %d\n", err);
+		goto release_reset;
+	}
+
+	err = reset_control_bulk_deassert(host->nresets, host->resets);
+	if (err < 0) {
+		dev_err(dev, "failed to deassert reset: %d\n", err);
+		goto disable_clk;
+	}
+
+	host1x_setup_sid_table(host);
+	host1x_syncpt_restore(host);
+	host1x_intr_start(host);
+
+	return 0;
+
+disable_clk:
+	clk_disable_unprepare(host->clk);
+release_reset:
+	reset_control_bulk_release(host->nresets, host->resets);
+
+	return err;
+}
+
+static const struct dev_pm_ops host1x_pm_ops = {
+	SET_RUNTIME_PM_OPS(host1x_runtime_suspend, host1x_runtime_resume,
+			   NULL)
+	/* TODO: add system suspend-resume once driver will be ready for that */
+};
+
 static struct platform_driver tegra_host1x_driver = {
 	.driver = {
 		.name = "tegra-host1x",
 		.of_match_table = host1x_of_match,
+		.pm = &host1x_pm_ops,
 	},
 	.probe = host1x_probe,
 	.remove = host1x_remove,

drivers/gpu/host1x/dev.h

@@ -118,7 +118,8 @@ struct host1x {
 	struct host1x_syncpt_base *bases;
 	struct device *dev;
 	struct clk *clk;
-	struct reset_control *rst;
+	struct reset_control_bulk_data resets[2];
+	unsigned int nresets;
 
 	struct iommu_group *group;
 	struct iommu_domain *domain;
@@ -149,6 +150,8 @@ struct host1x {
 	struct list_head list;
 
 	struct device_dma_parameters dma_parms;
+
+	struct host1x_bo_cache cache;
 };
 
 void host1x_hypervisor_writel(struct host1x *host1x, u32 r, u32 v);

drivers/gpu/host1x/hw/channel_hw.c

@@ -159,6 +159,27 @@ static void host1x_channel_set_streamid(struct host1x_channel *channel)
 #endif
 }
 
+static void host1x_enable_gather_filter(struct host1x_channel *ch)
+{
+#if HOST1X_HW >= 6
+	struct host1x *host = dev_get_drvdata(ch->dev->parent);
+	u32 val;
+
+	if (!host->hv_regs)
+		return;
+
+	val = host1x_hypervisor_readl(
+		host, HOST1X_HV_CH_KERNEL_FILTER_GBUFFER(ch->id / 32));
+	val |= BIT(ch->id % 32);
+	host1x_hypervisor_writel(
+		host, val, HOST1X_HV_CH_KERNEL_FILTER_GBUFFER(ch->id / 32));
+#elif HOST1X_HW >= 4
+	host1x_ch_writel(ch,
+			 HOST1X_CHANNEL_CHANNELCTRL_KERNEL_FILTER_GBUFFER(1),
+			 HOST1X_CHANNEL_CHANNELCTRL);
+#endif
+}
+
 static int channel_submit(struct host1x_job *job)
 {
 	struct host1x_channel *ch = job->channel;
@@ -190,6 +211,7 @@ static int channel_submit(struct host1x_job *job)
 	}
 
 	host1x_channel_set_streamid(ch);
+	host1x_enable_gather_filter(ch);
 
 	/* begin a CDMA submit */
 	err = host1x_cdma_begin(&ch->cdma, job);
@@ -249,27 +271,6 @@ static int channel_submit(struct host1x_job *job)
 	return err;
 }
 
-static void enable_gather_filter(struct host1x *host,
-				 struct host1x_channel *ch)
-{
-#if HOST1X_HW >= 6
-	u32 val;
-
-	if (!host->hv_regs)
-		return;
-
-	val = host1x_hypervisor_readl(
-		host, HOST1X_HV_CH_KERNEL_FILTER_GBUFFER(ch->id / 32));
-	val |= BIT(ch->id % 32);
-	host1x_hypervisor_writel(
-		host, val, HOST1X_HV_CH_KERNEL_FILTER_GBUFFER(ch->id / 32));
-#elif HOST1X_HW >= 4
-	host1x_ch_writel(ch,
-			 HOST1X_CHANNEL_CHANNELCTRL_KERNEL_FILTER_GBUFFER(1),
-			 HOST1X_CHANNEL_CHANNELCTRL);
-#endif
-}
-
 static int host1x_channel_init(struct host1x_channel *ch, struct host1x *dev,
 			       unsigned int index)
 {
@@ -278,7 +279,6 @@ static int host1x_channel_init(struct host1x_channel *ch, struct host1x *dev,
 #else
 	ch->regs = dev->regs + index * 0x100;
 #endif
-	enable_gather_filter(dev, ch);
 	return 0;
 }
 

drivers/gpu/host1x/intr.c

@@ -297,14 +297,11 @@ int host1x_intr_init(struct host1x *host, unsigned int irq_sync)
 			 "host1x_sp_%02u", id);
 	}
 
-	host1x_intr_start(host);
-
 	return 0;
 }
 
 void host1x_intr_deinit(struct host1x *host)
 {
-	host1x_intr_stop(host);
 }
 
 void host1x_intr_start(struct host1x *host)

drivers/gpu/host1x/job.c

@@ -134,20 +134,20 @@ EXPORT_SYMBOL(host1x_job_add_wait);
 
 static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 {
+	unsigned long mask = HOST1X_RELOC_READ | HOST1X_RELOC_WRITE;
 	struct host1x_client *client = job->client;
 	struct device *dev = client->dev;
 	struct host1x_job_gather *g;
-	struct iommu_domain *domain;
-	struct sg_table *sgt;
 	unsigned int i;
 	int err;
 
-	domain = iommu_get_domain_for_dev(dev);
 	job->num_unpins = 0;
 
 	for (i = 0; i < job->num_relocs; i++) {
 		struct host1x_reloc *reloc = &job->relocs[i];
-		dma_addr_t phys_addr, *phys;
+		enum dma_data_direction direction;
+		struct host1x_bo_mapping *map;
+		struct host1x_bo *bo;
 
 		reloc->target.bo = host1x_bo_get(reloc->target.bo);
 		if (!reloc->target.bo) {
@@ -155,64 +155,44 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 			goto unpin;
 		}
 
-		/*
-		 * If the client device is not attached to an IOMMU, the
-		 * physical address of the buffer object can be used.
-		 *
-		 * Similarly, when an IOMMU domain is shared between all
-		 * host1x clients, the IOVA is already available, so no
-		 * need to map the buffer object again.
-		 *
-		 * XXX Note that this isn't always safe to do because it
-		 * relies on an assumption that no cache maintenance is
-		 * needed on the buffer objects.
-		 */
-		if (!domain || client->group)
-			phys = &phys_addr;
-		else
-			phys = NULL;
-
-		sgt = host1x_bo_pin(dev, reloc->target.bo, phys);
-		if (IS_ERR(sgt)) {
-			err = PTR_ERR(sgt);
-			goto unpin;
-		}
+		bo = reloc->target.bo;
 
-		if (sgt) {
-			unsigned long mask = HOST1X_RELOC_READ |
-					     HOST1X_RELOC_WRITE;
-			enum dma_data_direction dir;
-
-			switch (reloc->flags & mask) {
-			case HOST1X_RELOC_READ:
-				dir = DMA_TO_DEVICE;
-				break;
+		switch (reloc->flags & mask) {
+		case HOST1X_RELOC_READ:
+			direction = DMA_TO_DEVICE;
+			break;
 
-			case HOST1X_RELOC_WRITE:
-				dir = DMA_FROM_DEVICE;
-				break;
+		case HOST1X_RELOC_WRITE:
+			direction = DMA_FROM_DEVICE;
+			break;
 
-			case HOST1X_RELOC_READ | HOST1X_RELOC_WRITE:
-				dir = DMA_BIDIRECTIONAL;
-				break;
+		case HOST1X_RELOC_READ | HOST1X_RELOC_WRITE:
+			direction = DMA_BIDIRECTIONAL;
+			break;
 
-			default:
-				err = -EINVAL;
-				goto unpin;
-			}
+		default:
+			err = -EINVAL;
+			goto unpin;
+		}
 
-			err = dma_map_sgtable(dev, sgt, dir, 0);
-			if (err)
-				goto unpin;
+		map = host1x_bo_pin(dev, bo, direction, &client->cache);
+		if (IS_ERR(map)) {
+			err = PTR_ERR(map);
+			goto unpin;
+		}
 
-			job->unpins[job->num_unpins].dev = dev;
-			job->unpins[job->num_unpins].dir = dir;
-			phys_addr = sg_dma_address(sgt->sgl);
+		/*
+		 * host1x clients are generally not able to do scatter-gather themselves, so fail
+		 * if the buffer is discontiguous and we fail to map its SG table to a single
+		 * contiguous chunk of I/O virtual memory.
+		 */
+		if (map->chunks > 1) {
+			err = -EINVAL;
+			goto unpin;
 		}
 
-		job->addr_phys[job->num_unpins] = phys_addr;
-		job->unpins[job->num_unpins].bo = reloc->target.bo;
-		job->unpins[job->num_unpins].sgt = sgt;
+		job->addr_phys[job->num_unpins] = map->phys;
+		job->unpins[job->num_unpins].map = map;
 		job->num_unpins++;
 	}
 
@@ -224,12 +204,11 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 		return 0;
 
 	for (i = 0; i < job->num_cmds; i++) {
+		struct host1x_bo_mapping *map;
 		size_t gather_size = 0;
 		struct scatterlist *sg;
-		dma_addr_t phys_addr;
 		unsigned long shift;
 		struct iova *alloc;
-		dma_addr_t *phys;
 		unsigned int j;
 
 		if (job->cmds[i].is_wait)
@@ -243,25 +222,16 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 			goto unpin;
 		}
 
-		/**
-		 * If the host1x is not attached to an IOMMU, there is no need
-		 * to map the buffer object for the host1x, since the physical
-		 * address can simply be used.
-		 */
-		if (!iommu_get_domain_for_dev(host->dev))
-			phys = &phys_addr;
-		else
-			phys = NULL;
-
-		sgt = host1x_bo_pin(host->dev, g->bo, phys);
-		if (IS_ERR(sgt)) {
-			err = PTR_ERR(sgt);
-			goto put;
+		map = host1x_bo_pin(host->dev, g->bo, DMA_TO_DEVICE, &host->cache);
+		if (IS_ERR(map)) {
+			err = PTR_ERR(map);
+			goto unpin;
 		}
 
 		if (host->domain) {
-			for_each_sgtable_sg(sgt, sg, j)
+			for_each_sgtable_sg(map->sgt, sg, j)
 				gather_size += sg->length;
+
 			gather_size = iova_align(&host->iova, gather_size);
 
 			shift = iova_shift(&host->iova);
@@ -272,33 +242,23 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 				goto put;
 			}
 
-			err = iommu_map_sgtable(host->domain,
-					iova_dma_addr(&host->iova, alloc),
-					sgt, IOMMU_READ);
+			err = iommu_map_sgtable(host->domain, iova_dma_addr(&host->iova, alloc),
+						map->sgt, IOMMU_READ);
 			if (err == 0) {
 				__free_iova(&host->iova, alloc);
 				err = -EINVAL;
 				goto put;
 			}
 
-			job->unpins[job->num_unpins].size = gather_size;
-			phys_addr = iova_dma_addr(&host->iova, alloc);
-		} else if (sgt) {
-			err = dma_map_sgtable(host->dev, sgt, DMA_TO_DEVICE, 0);
-			if (err)
-				goto put;
-
-			job->unpins[job->num_unpins].dir = DMA_TO_DEVICE;
-			job->unpins[job->num_unpins].dev = host->dev;
-			phys_addr = sg_dma_address(sgt->sgl);
+			map->phys = iova_dma_addr(&host->iova, alloc);
+			map->size = gather_size;
 		}
 
-		job->addr_phys[job->num_unpins] = phys_addr;
-		job->gather_addr_phys[i] = phys_addr;
-
-		job->unpins[job->num_unpins].bo = g->bo;
-		job->unpins[job->num_unpins].sgt = sgt;
+		job->addr_phys[job->num_unpins] = map->phys;
+		job->unpins[job->num_unpins].map = map;
 		job->num_unpins++;
+
+		job->gather_addr_phys[i] = map->phys;
 	}
 
 	return 0;
@@ -690,22 +650,16 @@ void host1x_job_unpin(struct host1x_job *job)
 	unsigned int i;
 
 	for (i = 0; i < job->num_unpins; i++) {
-		struct host1x_job_unpin_data *unpin = &job->unpins[i];
-		struct device *dev = unpin->dev ?: host->dev;
-		struct sg_table *sgt = unpin->sgt;
-
-		if (!job->enable_firewall && unpin->size && host->domain) {
-			iommu_unmap(host->domain, job->addr_phys[i],
-				    unpin->size);
-			free_iova(&host->iova,
-				iova_pfn(&host->iova, job->addr_phys[i]));
-		}
+		struct host1x_bo_mapping *map = job->unpins[i].map;
+		struct host1x_bo *bo = map->bo;
 
-		if (unpin->dev && sgt)
-			dma_unmap_sgtable(unpin->dev, sgt, unpin->dir, 0);
+		if (!job->enable_firewall && map->size && host->domain) {
+			iommu_unmap(host->domain, job->addr_phys[i], map->size);
+			free_iova(&host->iova, iova_pfn(&host->iova, job->addr_phys[i]));
+		}
 
-		host1x_bo_unpin(dev, unpin->bo, sgt);
-		host1x_bo_put(unpin->bo);
+		host1x_bo_unpin(map);
+		host1x_bo_put(bo);
 	}
 
 	job->num_unpins = 0;

drivers/gpu/host1x/job.h

@@ -35,11 +35,7 @@ struct host1x_job_cmd {
 };
 
 struct host1x_job_unpin_data {
-	struct host1x_bo *bo;
-	struct sg_table *sgt;
-	struct device *dev;
-	size_t size;
-	enum dma_data_direction dir;
+	struct host1x_bo_mapping *map;
 };
 
 /*

drivers/gpu/host1x/syncpt.c

@@ -143,6 +143,8 @@ void host1x_syncpt_restore(struct host1x *host)
 	for (i = 0; i < host1x_syncpt_nb_bases(host); i++)
 		host1x_hw_syncpt_restore_wait_base(host, sp_base + i);
 
+	host1x_hw_syncpt_enable_protection(host);
+
 	wmb();
 }
 
@@ -366,9 +368,6 @@ int host1x_syncpt_init(struct host1x *host)
 	host->syncpt = syncpt;
 	host->bases = bases;
 
-	host1x_syncpt_restore(host);
-	host1x_hw_syncpt_enable_protection(host);
-
 	/* Allocate sync point to use for clearing waits for expired fences */
 	host->nop_sp = host1x_syncpt_alloc(host, 0, "reserved-nop");
 	if (!host->nop_sp)

drivers/soc/tegra/common.c

@@ -10,6 +10,7 @@
 #include <linux/export.h>
 #include <linux/of.h>
 #include <linux/pm_opp.h>
+#include <linux/pm_runtime.h>
 
 #include <soc/tegra/common.h>
 #include <soc/tegra/fuse.h>
@@ -43,6 +44,7 @@ static int tegra_core_dev_init_opp_state(struct device *dev)
 {
 	unsigned long rate;
 	struct clk *clk;
+	bool rpm_enabled;
 	int err;
 
 	clk = devm_clk_get(dev, NULL);
@@ -57,8 +59,31 @@ static int tegra_core_dev_init_opp_state(struct device *dev)
 		return -EINVAL;
 	}
 
+	/*
+	 * Runtime PM of the device must be enabled in order to set up
+	 * GENPD's performance properly because GENPD core checks whether
+	 * device is suspended and this check doesn't work while RPM is
+	 * disabled. This makes sure the OPP vote below gets cached in
+	 * GENPD for the device. Instead, the vote is done the next time
+	 * the device gets runtime resumed.
+	 */
+	rpm_enabled = pm_runtime_enabled(dev);
+	if (!rpm_enabled)
+		pm_runtime_enable(dev);
+
+	/* should never happen in practice */
+	if (!pm_runtime_enabled(dev)) {
+		dev_WARN(dev, "failed to enable runtime PM\n");
+		pm_runtime_disable(dev);
+		return -EINVAL;
+	}
+
 	/* first dummy rate-setting initializes voltage vote */
 	err = dev_pm_opp_set_rate(dev, rate);
+
+	if (!rpm_enabled)
+		pm_runtime_disable(dev);
+
 	if (err) {
 		dev_err(dev, "failed to initialize OPP clock: %d\n", err);
 		return err;

include/linux/host1x.h

@@ -7,6 +7,8 @@
 #define __LINUX_HOST1X_H
 
 #include <linux/device.h>
+#include <linux/dma-direction.h>
+#include <linux/spinlock.h>
 #include <linux/types.h>
 
 enum host1x_class {
@@ -15,6 +17,8 @@ enum host1x_class {
 	HOST1X_CLASS_GR2D_SB = 0x52,
 	HOST1X_CLASS_VIC = 0x5D,
 	HOST1X_CLASS_GR3D = 0x60,
+	HOST1X_CLASS_NVDEC = 0xF0,
+	HOST1X_CLASS_NVDEC1 = 0xF5,
 };
 
 struct host1x;
@@ -23,6 +27,28 @@ struct iommu_group;
 
 u64 host1x_get_dma_mask(struct host1x *host1x);
 
+/**
+ * struct host1x_bo_cache - host1x buffer object cache
+ * @mappings: list of mappings
+ * @lock: synchronizes accesses to the list of mappings
+ */
+struct host1x_bo_cache {
+	struct list_head mappings;
+	struct mutex lock;
+};
+
+static inline void host1x_bo_cache_init(struct host1x_bo_cache *cache)
+{
+	INIT_LIST_HEAD(&cache->mappings);
+	mutex_init(&cache->lock);
+}
+
+static inline void host1x_bo_cache_destroy(struct host1x_bo_cache *cache)
+{
+	/* XXX warn if not empty? */
+	mutex_destroy(&cache->lock);
+}
+
 /**
  * struct host1x_client_ops - host1x client operations
  * @early_init: host1x client early initialization code
@@ -73,6 +99,8 @@ struct host1x_client {
 	struct host1x_client *parent;
 	unsigned int usecount;
 	struct mutex lock;
+
+	struct host1x_bo_cache cache;
 };
 
 /*
@@ -82,23 +110,48 @@ struct host1x_client {
 struct host1x_bo;
 struct sg_table;
 
+struct host1x_bo_mapping {
+	struct kref ref;
+	struct dma_buf_attachment *attach;
+	enum dma_data_direction direction;
+	struct list_head list;
+	struct host1x_bo *bo;
+	struct sg_table *sgt;
+	unsigned int chunks;
+	struct device *dev;
+	dma_addr_t phys;
+	size_t size;
+
+	struct host1x_bo_cache *cache;
+	struct list_head entry;
+};
+
+static inline struct host1x_bo_mapping *to_host1x_bo_mapping(struct kref *ref)
+{
+	return container_of(ref, struct host1x_bo_mapping, ref);
+}
+
 struct host1x_bo_ops {
 	struct host1x_bo *(*get)(struct host1x_bo *bo);
 	void (*put)(struct host1x_bo *bo);
-	struct sg_table *(*pin)(struct device *dev, struct host1x_bo *bo,
-				dma_addr_t *phys);
-	void (*unpin)(struct device *dev, struct sg_table *sgt);
+	struct host1x_bo_mapping *(*pin)(struct device *dev, struct host1x_bo *bo,
+					 enum dma_data_direction dir);
+	void (*unpin)(struct host1x_bo_mapping *map);
 	void *(*mmap)(struct host1x_bo *bo);
 	void (*munmap)(struct host1x_bo *bo, void *addr);
 };
 
 struct host1x_bo {
 	const struct host1x_bo_ops *ops;
+	struct list_head mappings;
+	spinlock_t lock;
 };
 
 static inline void host1x_bo_init(struct host1x_bo *bo,
 				  const struct host1x_bo_ops *ops)
 {
+	INIT_LIST_HEAD(&bo->mappings);
+	spin_lock_init(&bo->lock);
 	bo->ops = ops;
 }
 
@@ -112,18 +165,10 @@ static inline void host1x_bo_put(struct host1x_bo *bo)
 	bo->ops->put(bo);
 }
 
-static inline struct sg_table *host1x_bo_pin(struct device *dev,
-					     struct host1x_bo *bo,
-					     dma_addr_t *phys)
-{
-	return bo->ops->pin(dev, bo, phys);
-}
-
-static inline void host1x_bo_unpin(struct device *dev, struct host1x_bo *bo,
-				   struct sg_table *sgt)
-{
-	bo->ops->unpin(dev, sgt);
-}
+struct host1x_bo_mapping *host1x_bo_pin(struct device *dev, struct host1x_bo *bo,
+					enum dma_data_direction dir,
+					struct host1x_bo_cache *cache);
+void host1x_bo_unpin(struct host1x_bo_mapping *map);
 
 static inline void *host1x_bo_mmap(struct host1x_bo *bo)
 {
@@ -181,6 +226,7 @@ struct host1x_job;
 
 struct host1x_channel *host1x_channel_request(struct host1x_client *client);
 struct host1x_channel *host1x_channel_get(struct host1x_channel *channel);
+void host1x_channel_stop(struct host1x_channel *channel);
 void host1x_channel_put(struct host1x_channel *channel);
 int host1x_job_submit(struct host1x_job *job);
 

include/soc/tegra/common.h

@@ -39,4 +39,19 @@ devm_tegra_core_dev_init_opp_table(struct device *dev,
 }
 #endif
 
+static inline int
+devm_tegra_core_dev_init_opp_table_common(struct device *dev)
+{
+	struct tegra_core_opp_params opp_params = {};
+	int err;
+
+	opp_params.init_state = true;
+
+	err = devm_tegra_core_dev_init_opp_table(dev, &opp_params);
+	if (err != -ENODEV)
+		return err;
+
+	return 0;
+}
+
 #endif /* __SOC_TEGRA_COMMON_H__ */