drm/tegra: Implement correct DMA-BUF semantics

DMA-BUF requires that each device that accesses a DMA-BUF attaches to it
separately. To do so the host1x_bo_pin() and host1x_bo_unpin() functions
need to be reimplemented so that they can return a mapping, which either
represents an attachment or a map of the driver's own GEM object.

Signed-off-by: Thierry Reding <[e-mail]>

drivers/gpu/drm/tegra/gem.c

···
 
 MODULE_IMPORT_NS(DMA_BUF);
 
+static unsigned int sg_dma_count_chunks(struct scatterlist *sgl, unsigned int nents)
+{
+	dma_addr_t next = ~(dma_addr_t)0;
+	unsigned int count = 0, i;
+	struct scatterlist *s;
+
+	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;
+
+		if (sg_dma_address(s) != next) {
+			next = sg_dma_address(s) + sg_dma_len(s);
+			count++;
+		}
+	}
+
+	return count;
+}
+
+static inline unsigned int sgt_dma_count_chunks(struct sg_table *sgt)
+{
+	return sg_dma_count_chunks(sgt->sgl, sgt->nents);
+}
+
 static void tegra_bo_put(struct host1x_bo *bo)
 {
 	struct tegra_bo *obj = host1x_to_tegra_bo(bo);
···
 	drm_gem_object_put(&obj->gem);
 }
 
-/* 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;
-
-	err = sg_alloc_table(sgt, nents, gfp_mask);
-	if (err < 0)
-		return err;
-
-	dst = sgt->sgl;
-
-	for (i = 0; i < nents; i++) {
-		sg_set_page(dst, sg_page(sg), sg->length, 0);
-		dst = sg_next(dst);
-		sg = sg_next(sg);
-	}
-
-	return 0;
-}
-
-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);
+
+	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 {
···
 		 * 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)
+		return;
+
+	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)

drivers/gpu/drm/tegra/plane.c

···
 
 	for (i = 0; i < 3; i++) {
 		copy->iova[i] = DMA_MAPPING_ERROR;
-		copy->sgt[i] = NULL;
+		copy->map[i] = NULL;
 	}
 
 	return &copy->base;
···
 
 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);
+		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;
···
 	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;
···
 	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;
 	}
 }
 

drivers/gpu/drm/tegra/plane.h

···
 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/submit.c

···
 	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);
+	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)

drivers/gpu/drm/tegra/uapi.c

···
 	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);
···
 	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)
···
 
 	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;
+	switch (args->flags & DRM_TEGRA_CHANNEL_MAP_READ_WRITE) {
+	case DRM_TEGRA_CHANNEL_MAP_READ_WRITE:
+		direction = DMA_BIDIRECTIONAL;
+		break;
 
-		case DRM_TEGRA_CHANNEL_MAP_WRITE:
-			mapping->direction = DMA_FROM_DEVICE;
-			break;
+	case DRM_TEGRA_CHANNEL_MAP_WRITE:
+		direction = DMA_FROM_DEVICE;
+		break;
 
-		case DRM_TEGRA_CHANNEL_MAP_READ:
-			mapping->direction = DMA_TO_DEVICE;
-			break;
+	case DRM_TEGRA_CHANNEL_MAP_READ:
+		direction = DMA_TO_DEVICE;
+		break;
 
-		default:
-			return -EINVAL;
-		}
-
-		mapping->sgt = host1x_bo_pin(mapping->dev, mapping->bo, NULL);
-		if (IS_ERR(mapping->sgt)) {
-			err = PTR_ERR(mapping->sgt);
-			goto put_gem;
-		}
-
-		err = dma_map_sgtable(mapping->dev, mapping->sgt, mapping->direction,
-				      DMA_ATTR_SKIP_CPU_SYNC);
-		if (err)
-			goto unpin;
-
-		mapping->iova = sg_dma_address(mapping->sgt->sgl);
+	default:
+		err = -EINVAL;
+		goto put_gem;
 	}
 
+	mapping->map = host1x_bo_pin(context->client->base.dev, mapping->bo, direction);
+	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

···
 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/host1x/job.c

···
 
 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) {
···
 			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;
+		bo = reloc->target.bo;
 
-		sgt = host1x_bo_pin(dev, reloc->target.bo, phys);
-		if (IS_ERR(sgt)) {
-			err = PTR_ERR(sgt);
+		switch (reloc->flags & mask) {
+		case HOST1X_RELOC_READ:
+			direction = DMA_TO_DEVICE;
+			break;
+
+		case HOST1X_RELOC_WRITE:
+			direction = DMA_FROM_DEVICE;
+			break;
+
+		case HOST1X_RELOC_READ | HOST1X_RELOC_WRITE:
+			direction = DMA_BIDIRECTIONAL;
+			break;
+
+		default:
+			err = -EINVAL;
 			goto unpin;
 		}
 
-		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;
-
-			case HOST1X_RELOC_WRITE:
-				dir = DMA_FROM_DEVICE;
-				break;
-
-			case HOST1X_RELOC_READ | HOST1X_RELOC_WRITE:
-				dir = DMA_BIDIRECTIONAL;
-				break;
-
-			default:
-				err = -EINVAL;
-				goto unpin;
-			}
-
-			err = dma_map_sgtable(dev, sgt, dir, 0);
-			if (err)
-				goto unpin;
-
-			job->unpins[job->num_unpins].dev = dev;
-			job->unpins[job->num_unpins].dir = dir;
-			phys_addr = sg_dma_address(sgt->sgl);
+		map = host1x_bo_pin(dev, bo, direction);
+		if (IS_ERR(map)) {
+			err = PTR_ERR(map);
+			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;
+		/*
+		 * 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] = map->phys;
+		job->unpins[job->num_unpins].map = map;
 		job->num_unpins++;
 	}
 
···
 		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)
···
 			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);
+		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);
···
 				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;
···
 	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;
+		struct host1x_bo_mapping *map = job->unpins[i].map;
+		struct host1x_bo *bo = map->bo;
 
-		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]));
+		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]));
 		}
 
-		if (unpin->dev && sgt)
-			dma_unmap_sgtable(unpin->dev, sgt, unpin->dir, 0);
-
-		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

···
 };
 
 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;
 };
 
 /*

include/linux/host1x.h

···
 #define __LINUX_HOST1X_H
 
 #include <linux/device.h>
+#include <linux/dma-direction.h>
 #include <linux/types.h>
 
 enum host1x_class {
···
 struct host1x_bo;
 struct sg_table;
 
+struct host1x_bo_mapping {
+	struct dma_buf_attachment *attach;
+	enum dma_data_direction direction;
+	struct host1x_bo *bo;
+	struct sg_table *sgt;
+	unsigned int chunks;
+	struct device *dev;
+	dma_addr_t phys;
+	size_t size;
+};
+
 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);
 };
···
 	bo->ops->put(bo);
 }
 
-static inline struct sg_table *host1x_bo_pin(struct device *dev,
-					     struct host1x_bo *bo,
-					     dma_addr_t *phys)
+static inline struct host1x_bo_mapping *host1x_bo_pin(struct device *dev, struct host1x_bo *bo,
+						      enum dma_data_direction dir)
 {
-	return bo->ops->pin(dev, bo, phys);
+	return bo->ops->pin(dev, bo, dir);
 }
 
-static inline void host1x_bo_unpin(struct device *dev, struct host1x_bo *bo,
-				   struct sg_table *sgt)
+static inline void host1x_bo_unpin(struct host1x_bo_mapping *map)
 {
-	bo->ops->unpin(dev, sgt);
+	map->bo->ops->unpin(map);
 }
 
 static inline void *host1x_bo_mmap(struct host1x_bo *bo)