1 files changed, 117 insertions, 44 deletions

diff --git a/drivers/accel/ivpu/ivpu_gem.c b/drivers/accel/ivpu/ivpu_gem.c
index 59cfcf3eaded..ece68f570b7e 100644
--- a/drivers/accel/ivpu/ivpu_gem.c
+++ b/drivers/accel/ivpu/ivpu_gem.c
@@ -15,6 +15,7 @@
 #include <drm/drm_utils.h>
 
 #include "ivpu_drv.h"
+#include "ivpu_fw.h"
 #include "ivpu_gem.h"
 #include "ivpu_hw.h"
 #include "ivpu_mmu.h"
@@ -27,8 +28,8 @@ static const struct drm_gem_object_funcs ivpu_gem_funcs;
 static inline void ivpu_dbg_bo(struct ivpu_device *vdev, struct ivpu_bo *bo, const char *action)
 {
 	ivpu_dbg(vdev, BO,
-		 "%6s: bo %8p vpu_addr %9llx size %8zu ctx %d has_pages %d dma_mapped %d mmu_mapped %d wc %d imported %d\n",
-		 action, bo, bo->vpu_addr, ivpu_bo_size(bo), bo->ctx_id,
+		 "%6s: bo %8p size %9zu ctx %d vpu_addr %9llx pages %d sgt %d mmu_mapped %d wc %d imported %d\n",
+		 action, bo, ivpu_bo_size(bo), bo->ctx_id, bo->vpu_addr,
 		 (bool)bo->base.pages, (bool)bo->base.sgt, bo->mmu_mapped, bo->base.map_wc,
 		 (bool)drm_gem_is_imported(&bo->base.base));
 }
@@ -43,22 +44,47 @@ static inline void ivpu_bo_unlock(struct ivpu_bo *bo)
 	dma_resv_unlock(bo->base.base.resv);
 }
 
+static struct sg_table *ivpu_bo_map_attachment(struct ivpu_device *vdev, struct ivpu_bo *bo)
+{
+	struct sg_table *sgt;
+
+	drm_WARN_ON(&vdev->drm, !bo->base.base.import_attach);
+
+	ivpu_bo_lock(bo);
+
+	sgt = bo->base.sgt;
+	if (!sgt) {
+		sgt = dma_buf_map_attachment(bo->base.base.import_attach, DMA_BIDIRECTIONAL);
+		if (IS_ERR(sgt))
+			ivpu_err(vdev, "Failed to map BO in IOMMU: %ld\n", PTR_ERR(sgt));
+		else
+			bo->base.sgt = sgt;
+	}
+
+	ivpu_bo_unlock(bo);
+
+	return sgt;
+}
+
 /*
- * ivpu_bo_pin() - pin the backing physical pages and map them to VPU.
+ * ivpu_bo_bind() - pin the backing physical pages and map them to VPU.
  *
  * This function pins physical memory pages, then maps the physical pages
  * to IOMMU address space and finally updates the VPU MMU page tables
  * to allow the VPU to translate VPU address to IOMMU address.
  */
-int __must_check ivpu_bo_pin(struct ivpu_bo *bo)
+int __must_check ivpu_bo_bind(struct ivpu_bo *bo)
 {
 	struct ivpu_device *vdev = ivpu_bo_to_vdev(bo);
 	struct sg_table *sgt;
 	int ret = 0;
 
-	ivpu_dbg_bo(vdev, bo, "pin");
+	ivpu_dbg_bo(vdev, bo, "bind");
 
-	sgt = drm_gem_shmem_get_pages_sgt(&bo->base);
+	if (bo->base.base.import_attach)
+		sgt = ivpu_bo_map_attachment(vdev, bo);
+	else
+		sgt = drm_gem_shmem_get_pages_sgt(&bo->base);
 	if (IS_ERR(sgt)) {
 		ret = PTR_ERR(sgt);
 		ivpu_err(vdev, "Failed to map BO in IOMMU: %d\n", ret);
@@ -70,7 +96,7 @@ int __must_check ivpu_bo_pin(struct ivpu_bo *bo)
 	if (!bo->mmu_mapped) {
 		drm_WARN_ON(&vdev->drm, !bo->ctx);
 		ret = ivpu_mmu_context_map_sgt(vdev, bo->ctx, bo->vpu_addr, sgt,
-					       ivpu_bo_is_snooped(bo));
+					       ivpu_bo_is_snooped(bo), ivpu_bo_is_read_only(bo));
 		if (ret) {
 			ivpu_err(vdev, "Failed to map BO in MMU: %d\n", ret);
 			goto unlock;
@@ -99,9 +125,9 @@ ivpu_bo_alloc_vpu_addr(struct ivpu_bo *bo, struct ivpu_mmu_context *ctx,
 	ret = ivpu_mmu_context_insert_node(ctx, range, ivpu_bo_size(bo), &bo->mm_node);
 	if (!ret) {
 		bo->ctx = ctx;
+		bo->ctx_id = ctx->id;
 		bo->vpu_addr = bo->mm_node.start;
-	} else {
-		ivpu_err(vdev, "Failed to add BO to context %u: %d\n", ctx->id, ret);
+		ivpu_dbg_bo(vdev, bo, "vaddr");
 	}
 
 	ivpu_bo_unlock(bo);
@@ -115,7 +141,7 @@ static void ivpu_bo_unbind_locked(struct ivpu_bo *bo)
 {
 	struct ivpu_device *vdev = ivpu_bo_to_vdev(bo);
 
-	lockdep_assert(dma_resv_held(bo->base.base.resv) || !kref_read(&bo->base.base.refcount));
+	dma_resv_assert_held(bo->base.base.resv);
 
 	if (bo->mmu_mapped) {
 		drm_WARN_ON(&vdev->drm, !bo->ctx);
@@ -130,13 +156,15 @@ static void ivpu_bo_unbind_locked(struct ivpu_bo *bo)
 		bo->ctx = NULL;
 	}
 
-	if (drm_gem_is_imported(&bo->base.base))
-		return;
-
 	if (bo->base.sgt) {
-		dma_unmap_sgtable(vdev->drm.dev, bo->base.sgt, DMA_BIDIRECTIONAL, 0);
-		sg_free_table(bo->base.sgt);
-		kfree(bo->base.sgt);
+		if (bo->base.base.import_attach) {
+			dma_buf_unmap_attachment(bo->base.base.import_attach,
+						 bo->base.sgt, DMA_BIDIRECTIONAL);
+		} else {
+			dma_unmap_sgtable(vdev->drm.dev, bo->base.sgt, DMA_BIDIRECTIONAL, 0);
+			sg_free_table(bo->base.sgt);
+			kfree(bo->base.sgt);
+		}
 		bo->base.sgt = NULL;
 	}
 }
@@ -182,10 +210,11 @@ struct drm_gem_object *ivpu_gem_create_object(struct drm_device *dev, size_t siz
 struct drm_gem_object *ivpu_gem_prime_import(struct drm_device *dev,
 					     struct dma_buf *dma_buf)
 {
+	struct ivpu_device *vdev = to_ivpu_device(dev);
 	struct device *attach_dev = dev->dev;
 	struct dma_buf_attachment *attach;
-	struct sg_table *sgt;
 	struct drm_gem_object *obj;
+	struct ivpu_bo *bo;
 	int ret;
 
 	attach = dma_buf_attach(dma_buf, attach_dev);
@@ -194,25 +223,25 @@ struct drm_gem_object *ivpu_gem_prime_import(struct drm_device *dev,
 
 	get_dma_buf(dma_buf);
 
-	sgt = dma_buf_map_attachment_unlocked(attach, DMA_BIDIRECTIONAL);
-	if (IS_ERR(sgt)) {
-		ret = PTR_ERR(sgt);
-		goto fail_detach;
-	}
-
-	obj = drm_gem_shmem_prime_import_sg_table(dev, attach, sgt);
+	obj = drm_gem_shmem_prime_import_sg_table(dev, attach, NULL);
 	if (IS_ERR(obj)) {
 		ret = PTR_ERR(obj);
-		goto fail_unmap;
+		goto fail_detach;
 	}
 
 	obj->import_attach = attach;
 	obj->resv = dma_buf->resv;
 
+	bo = to_ivpu_bo(obj);
+
+	mutex_lock(&vdev->bo_list_lock);
+	list_add_tail(&bo->bo_list_node, &vdev->bo_list);
+	mutex_unlock(&vdev->bo_list_lock);
+
+	ivpu_dbg(vdev, BO, "import: bo %8p size %9zu\n", bo, ivpu_bo_size(bo));
+
 	return obj;
 
-fail_unmap:
-	dma_buf_unmap_attachment_unlocked(attach, sgt, DMA_BIDIRECTIONAL);
 fail_detach:
 	dma_buf_detach(dma_buf, attach);
 	dma_buf_put(dma_buf);
@@ -220,7 +249,7 @@ fail_detach:
 	return ERR_PTR(ret);
 }
 
-static struct ivpu_bo *ivpu_bo_alloc(struct ivpu_device *vdev, u64 size, u32 flags, u32 ctx_id)
+static struct ivpu_bo *ivpu_bo_alloc(struct ivpu_device *vdev, u64 size, u32 flags)
 {
 	struct drm_gem_shmem_object *shmem;
 	struct ivpu_bo *bo;
@@ -238,7 +267,6 @@ static struct ivpu_bo *ivpu_bo_alloc(struct ivpu_device *vdev, u64 size, u32 fla
 		return ERR_CAST(shmem);
 
 	bo = to_ivpu_bo(&shmem->base);
-	bo->ctx_id = ctx_id;
 	bo->base.map_wc = flags & DRM_IVPU_BO_WC;
 	bo->flags = flags;
 
@@ -246,7 +274,7 @@ static struct ivpu_bo *ivpu_bo_alloc(struct ivpu_device *vdev, u64 size, u32 fla
 	list_add_tail(&bo->bo_list_node, &vdev->bo_list);
 	mutex_unlock(&vdev->bo_list_lock);
 
-	ivpu_dbg_bo(vdev, bo, "alloc");
+	ivpu_dbg(vdev, BO, " alloc: bo %8p size %9llu\n", bo, size);
 
 	return bo;
 }
@@ -259,8 +287,8 @@ static int ivpu_gem_bo_open(struct drm_gem_object *obj, struct drm_file *file)
 	struct ivpu_addr_range *range;
 
 	if (bo->ctx) {
-		ivpu_warn(vdev, "Can't add BO to ctx %u: already in ctx %u\n",
-			  file_priv->ctx.id, bo->ctx->id);
+		ivpu_dbg(vdev, IOCTL, "Can't add BO %pe to ctx %u: already in ctx %u\n",
+			 bo, file_priv->ctx.id, bo->ctx->id);
 		return -EALREADY;
 	}
 
@@ -281,23 +309,41 @@ static void ivpu_gem_bo_free(struct drm_gem_object *obj)
 
 	ivpu_dbg_bo(vdev, bo, "free");
 
+	drm_WARN_ON(&vdev->drm, list_empty(&bo->bo_list_node));
+
 	mutex_lock(&vdev->bo_list_lock);
 	list_del(&bo->bo_list_node);
-	mutex_unlock(&vdev->bo_list_lock);
 
 	drm_WARN_ON(&vdev->drm, !drm_gem_is_imported(&bo->base.base) &&
 		    !dma_resv_test_signaled(obj->resv, DMA_RESV_USAGE_READ));
 	drm_WARN_ON(&vdev->drm, ivpu_bo_size(bo) == 0);
 	drm_WARN_ON(&vdev->drm, bo->base.vaddr);
 
+	ivpu_bo_lock(bo);
 	ivpu_bo_unbind_locked(bo);
+	ivpu_bo_unlock(bo);
+
+	mutex_unlock(&vdev->bo_list_lock);
+
 	drm_WARN_ON(&vdev->drm, bo->mmu_mapped);
 	drm_WARN_ON(&vdev->drm, bo->ctx);
 
 	drm_WARN_ON(obj->dev, refcount_read(&bo->base.pages_use_count) > 1);
+	drm_WARN_ON(obj->dev, bo->base.base.vma_node.vm_files.rb_node);
 	drm_gem_shmem_free(&bo->base);
 }
 
+static enum drm_gem_object_status ivpu_gem_status(struct drm_gem_object *obj)
+{
+	struct ivpu_bo *bo = to_ivpu_bo(obj);
+	enum drm_gem_object_status status = 0;
+
+	if (ivpu_bo_is_resident(bo))
+		status |= DRM_GEM_OBJECT_RESIDENT;
+
+	return status;
+}
+
 static const struct drm_gem_object_funcs ivpu_gem_funcs = {
 	.free = ivpu_gem_bo_free,
 	.open = ivpu_gem_bo_open,
@@ -308,6 +354,7 @@ static const struct drm_gem_object_funcs ivpu_gem_funcs = {
 	.vmap = drm_gem_shmem_object_vmap,
 	.vunmap = drm_gem_shmem_object_vunmap,
 	.mmap = drm_gem_shmem_object_mmap,
+	.status = ivpu_gem_status,
 	.vm_ops = &drm_gem_shmem_vm_ops,
 };
 
@@ -320,25 +367,33 @@ int ivpu_bo_create_ioctl(struct drm_device *dev, void *data, struct drm_file *fi
 	struct ivpu_bo *bo;
 	int ret;
 
-	if (args->flags & ~DRM_IVPU_BO_FLAGS)
+	if (args->flags & ~DRM_IVPU_BO_FLAGS) {
+		ivpu_dbg(vdev, IOCTL, "Invalid BO flags 0x%x\n", args->flags);
 		return -EINVAL;
+	}
 
-	if (size == 0)
+	if (size == 0) {
+		ivpu_dbg(vdev, IOCTL, "Invalid BO size %llu\n", args->size);
 		return -EINVAL;
+	}
 
-	bo = ivpu_bo_alloc(vdev, size, args->flags, file_priv->ctx.id);
+	bo = ivpu_bo_alloc(vdev, size, args->flags);
 	if (IS_ERR(bo)) {
-		ivpu_err(vdev, "Failed to allocate BO: %pe (ctx %u size %llu flags 0x%x)",
+		ivpu_dbg(vdev, IOCTL, "Failed to allocate BO: %pe ctx %u size %llu flags 0x%x\n",
 			 bo, file_priv->ctx.id, args->size, args->flags);
 		return PTR_ERR(bo);
 	}
 
+	drm_WARN_ON(&vdev->drm, bo->base.base.handle_count != 0);
+
 	ret = drm_gem_handle_create(file, &bo->base.base, &args->handle);
-	if (ret)
-		ivpu_err(vdev, "Failed to create handle for BO: %pe (ctx %u size %llu flags 0x%x)",
+	if (ret) {
+		ivpu_dbg(vdev, IOCTL, "Failed to create handle for BO: %pe ctx %u size %llu flags 0x%x\n",
 			 bo, file_priv->ctx.id, args->size, args->flags);
-	else
+	} else {
 		args->vpu_addr = bo->vpu_addr;
+		drm_WARN_ON(&vdev->drm, bo->base.base.handle_count != 1);
+	}
 
 	drm_gem_object_put(&bo->base.base);
 
@@ -360,18 +415,21 @@ ivpu_bo_create(struct ivpu_device *vdev, struct ivpu_mmu_context *ctx,
 	drm_WARN_ON(&vdev->drm, !PAGE_ALIGNED(range->end));
 	drm_WARN_ON(&vdev->drm, !PAGE_ALIGNED(size));
 
-	bo = ivpu_bo_alloc(vdev, size, flags, IVPU_GLOBAL_CONTEXT_MMU_SSID);
+	bo = ivpu_bo_alloc(vdev, size, flags);
 	if (IS_ERR(bo)) {
-		ivpu_err(vdev, "Failed to allocate BO: %pe (vpu_addr 0x%llx size %llu flags 0x%x)",
+		ivpu_err(vdev, "Failed to allocate BO: %pe vpu_addr 0x%llx size %llu flags 0x%x\n",
 			 bo, range->start, size, flags);
 		return NULL;
 	}
 
 	ret = ivpu_bo_alloc_vpu_addr(bo, ctx, range);
-	if (ret)
+	if (ret) {
+		ivpu_err(vdev, "Failed to allocate NPU address for BO: %pe ctx %u size %llu: %d\n",
+			 bo, ctx->id, size, ret);
 		goto err_put;
+	}
 
-	ret = ivpu_bo_pin(bo);
+	ret = ivpu_bo_bind(bo);
 	if (ret)
 		goto err_put;
 
@@ -391,6 +449,21 @@ err_put:
 	return NULL;
 }
 
+struct ivpu_bo *ivpu_bo_create_runtime(struct ivpu_device *vdev, u64 addr, u64 size, u32 flags)
+{
+	struct ivpu_addr_range range;
+
+	if (!ivpu_is_within_range(addr, size, &vdev->hw->ranges.runtime)) {
+		ivpu_err(vdev, "Invalid runtime BO address 0x%llx size %llu\n", addr, size);
+		return NULL;
+	}
+
+	if (ivpu_hw_range_init(vdev, &range, addr, size))
+		return NULL;
+
+	return ivpu_bo_create(vdev, &vdev->gctx, &range, size, flags);
+}
+
 struct ivpu_bo *ivpu_bo_create_global(struct ivpu_device *vdev, u64 size, u32 flags)
 {
 	return ivpu_bo_create(vdev, &vdev->gctx, &vdev->hw->ranges.global, size, flags);