gpu: host1x: Support DMA mapping of buffers

If host1x_bo_pin() returns an SG table, create a DMA mapping for the
buffer. For buffers that the host1x client has already mapped itself,
host1x_bo_pin() returns NULL and the existing DMA address is used.

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

drivers/gpu/drm/tegra/gem.c

···
 	struct sg_table *sgt;
 	int err;
 
-	if (phys)
+	/*
+	 * If we've manually mapped the buffer object through the IOMMU, make
+	 * sure to return the IOVA address of our mapping.
+	 */
+	if (phys && obj->mm) {
 		*phys = obj->iova;
+		return NULL;
+	}
 
+	/*
+	 * 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);
···
 
 static void tegra_bo_unpin(struct device *dev, struct sg_table *sgt)
 {
-	sg_free_table(sgt);
-	kfree(sgt);
+	if (sgt) {
+		sg_free_table(sgt);
+		kfree(sgt);
+	}
 }
 
 static void *tegra_bo_mmap(struct host1x_bo *bo)

drivers/gpu/host1x/dev.c

···
 #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"
···
 		dev_err(&pdev->dev, "failed to get reset: %d\n", err);
 		return 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);
-	}
-#endif
+
 	if (IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL))
+		goto skip_iommu;
+
+	if (iommu_get_domain_for_dev(&pdev->dev))
 		goto skip_iommu;
 
 	host->group = iommu_group_get(&pdev->dev);

drivers/gpu/host1x/job.c

···
 
 static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 {
-	struct device *dev = job->client->dev;
+	struct host1x_client *client = job->client;
+	struct device *dev = client->dev;
 	unsigned int i;
 	int err;
 
···
 
 	for (i = 0; i < job->num_relocs; i++) {
 		struct host1x_reloc *reloc = &job->relocs[i];
+		dma_addr_t phys_addr, *phys;
 		struct sg_table *sgt;
-		dma_addr_t phys_addr;
 
 		reloc->target.bo = host1x_bo_get(reloc->target.bo);
 		if (!reloc->target.bo) {
···
 			goto unpin;
 		}
 
-		sgt = host1x_bo_pin(dev, reloc->target.bo, &phys_addr);
+		if (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;
+		}
+
+		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_sg(dev, sgt->sgl, sgt->nents, dir);
+			if (!err) {
+				err = -ENOMEM;
+				goto unpin;
+			}
+
+			job->unpins[job->num_unpins].dev = dev;
+			job->unpins[job->num_unpins].dir = dir;
+			phys_addr = sg_dma_address(sgt->sgl);
 		}
 
 		job->addr_phys[job->num_unpins] = phys_addr;
···
 			goto unpin;
 		}
 
-		sgt = host1x_bo_pin(host->dev, g->bo, &phys_addr);
+		sgt = host1x_bo_pin(host->dev, g->bo, NULL);
 		if (IS_ERR(sgt)) {
 			err = PTR_ERR(sgt);
 			goto unpin;
···
 				goto unpin;
 			}
 
-			job->addr_phys[job->num_unpins] =
-				iova_dma_addr(&host->iova, alloc);
 			job->unpins[job->num_unpins].size = gather_size;
+			phys_addr = iova_dma_addr(&host->iova, alloc);
 		} else {
-			job->addr_phys[job->num_unpins] = phys_addr;
+			err = dma_map_sg(host->dev, sgt->sgl, sgt->nents,
+					 DMA_TO_DEVICE);
+			if (!err) {
+				err = -ENOMEM;
+				goto unpin;
+			}
+
+			job->unpins[job->num_unpins].dev = host->dev;
+			phys_addr = sg_dma_address(sgt->sgl);
 		}
 
-		job->gather_addr_phys[i] = job->addr_phys[job->num_unpins];
+		job->addr_phys[job->num_unpins] = phys_addr;
+		job->gather_addr_phys[i] = phys_addr;
 
+		job->unpins[job->num_unpins].dir = DMA_TO_DEVICE;
 		job->unpins[job->num_unpins].bo = g->bo;
 		job->unpins[job->num_unpins].sgt = sgt;
 		job->num_unpins++;
···
 
 	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 (!IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL) &&
 		    unpin->size && host->domain) {
···
 				iova_pfn(&host->iova, job->addr_phys[i]));
 		}
 
-		host1x_bo_unpin(host->dev, unpin->bo, unpin->sgt);
+		if (unpin->dev && sgt)
+			dma_unmap_sg(unpin->dev, sgt->sgl, sgt->nents,
+				     unpin->dir);
+
+		host1x_bo_unpin(dev, unpin->bo, sgt);
 		host1x_bo_put(unpin->bo);
 	}
 

drivers/gpu/host1x/job.h

···
 #ifndef __HOST1X_JOB_H
 #define __HOST1X_JOB_H
 
+#include <linux/dma-direction.h>
+
 struct host1x_job_gather {
 	unsigned int words;
 	dma_addr_t base;
···
 struct host1x_job_unpin_data {
 	struct host1x_bo *bo;
 	struct sg_table *sgt;
+	struct device *dev;
 	size_t size;
+	enum dma_data_direction dir;
 };
 
 /*