drm/vc4: Add an API for creating GPU shaders in GEM BOs.

Since we have no MMU, the kernel needs to validate that the submitted
shader code won't make any accesses to memory that the user doesn't
control, which involves banning some operations (general purpose DMA
writes), and tracking where we need to write out pointers for other
operations (texture sampling).  Once it's validated, we return a GEM
BO containing the shader, which doesn't allow mapping for write or
exporting to other subsystems.

v2: Use __u32-style types.

Signed-off-by: Eric Anholt <eric@anholt.net>

drivers/gpu/drm/vc4/Makefile

@@ -10,7 +10,8 @@ vc4-y := \
 	vc4_kms.o \
 	vc4_hdmi.o \
 	vc4_hvs.o \
-	vc4_plane.o
+	vc4_plane.o \
+	vc4_validate_shaders.o
 
 vc4-$(CONFIG_DEBUG_FS) += vc4_debugfs.o
 

drivers/gpu/drm/vc4/vc4_bo.c

@@ -79,6 +79,12 @@ static void vc4_bo_destroy(struct vc4_bo *bo)
 	struct drm_gem_object *obj = &bo->base.base;
 	struct vc4_dev *vc4 = to_vc4_dev(obj->dev);
 
+	if (bo->validated_shader) {
+		kfree(bo->validated_shader->texture_samples);
+		kfree(bo->validated_shader);
+		bo->validated_shader = NULL;
+	}
+
 	vc4->bo_stats.num_allocated--;
 	vc4->bo_stats.size_allocated -= obj->size;
 	drm_gem_cma_free_object(obj);
@@ -315,6 +321,12 @@ void vc4_free_object(struct drm_gem_object *gem_bo)
 		goto out;
 	}
 
+	if (bo->validated_shader) {
+		kfree(bo->validated_shader->texture_samples);
+		kfree(bo->validated_shader);
+		bo->validated_shader = NULL;
+	}
+
 	bo->free_time = jiffies;
 	list_add(&bo->size_head, cache_list);
 	list_add(&bo->unref_head, &vc4->bo_cache.time_list);
@@ -347,6 +359,78 @@ static void vc4_bo_cache_time_timer(unsigned long data)
 	schedule_work(&vc4->bo_cache.time_work);
 }
 
+struct dma_buf *
+vc4_prime_export(struct drm_device *dev, struct drm_gem_object *obj, int flags)
+{
+	struct vc4_bo *bo = to_vc4_bo(obj);
+
+	if (bo->validated_shader) {
+		DRM_ERROR("Attempting to export shader BO\n");
+		return ERR_PTR(-EINVAL);
+	}
+
+	return drm_gem_prime_export(dev, obj, flags);
+}
+
+int vc4_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+	struct drm_gem_object *gem_obj;
+	struct vc4_bo *bo;
+	int ret;
+
+	ret = drm_gem_mmap(filp, vma);
+	if (ret)
+		return ret;
+
+	gem_obj = vma->vm_private_data;
+	bo = to_vc4_bo(gem_obj);
+
+	if (bo->validated_shader && (vma->vm_flags & VM_WRITE)) {
+		DRM_ERROR("mmaping of shader BOs for writing not allowed.\n");
+		return -EINVAL;
+	}
+
+	/*
+	 * Clear the VM_PFNMAP flag that was set by drm_gem_mmap(), and set the
+	 * vm_pgoff (used as a fake buffer offset by DRM) to 0 as we want to map
+	 * the whole buffer.
+	 */
+	vma->vm_flags &= ~VM_PFNMAP;
+	vma->vm_pgoff = 0;
+
+	ret = dma_mmap_writecombine(bo->base.base.dev->dev, vma,
+				    bo->base.vaddr, bo->base.paddr,
+				    vma->vm_end - vma->vm_start);
+	if (ret)
+		drm_gem_vm_close(vma);
+
+	return ret;
+}
+
+int vc4_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)
+{
+	struct vc4_bo *bo = to_vc4_bo(obj);
+
+	if (bo->validated_shader && (vma->vm_flags & VM_WRITE)) {
+		DRM_ERROR("mmaping of shader BOs for writing not allowed.\n");
+		return -EINVAL;
+	}
+
+	return drm_gem_cma_prime_mmap(obj, vma);
+}
+
+void *vc4_prime_vmap(struct drm_gem_object *obj)
+{
+	struct vc4_bo *bo = to_vc4_bo(obj);
+
+	if (bo->validated_shader) {
+		DRM_ERROR("mmaping of shader BOs not allowed.\n");
+		return ERR_PTR(-EINVAL);
+	}
+
+	return drm_gem_cma_prime_vmap(obj);
+}
+
 int vc4_create_bo_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *file_priv)
 {
@@ -387,6 +471,62 @@ int vc4_mmap_bo_ioctl(struct drm_device *dev, void *data,
 	return 0;
 }
 
+int
+vc4_create_shader_bo_ioctl(struct drm_device *dev, void *data,
+			   struct drm_file *file_priv)
+{
+	struct drm_vc4_create_shader_bo *args = data;
+	struct vc4_bo *bo = NULL;
+	int ret;
+
+	if (args->size == 0)
+		return -EINVAL;
+
+	if (args->size % sizeof(u64) != 0)
+		return -EINVAL;
+
+	if (args->flags != 0) {
+		DRM_INFO("Unknown flags set: 0x%08x\n", args->flags);
+		return -EINVAL;
+	}
+
+	if (args->pad != 0) {
+		DRM_INFO("Pad set: 0x%08x\n", args->pad);
+		return -EINVAL;
+	}
+
+	bo = vc4_bo_create(dev, args->size, true);
+	if (!bo)
+		return -ENOMEM;
+
+	ret = copy_from_user(bo->base.vaddr,
+			     (void __user *)(uintptr_t)args->data,
+			     args->size);
+	if (ret != 0)
+		goto fail;
+	/* Clear the rest of the memory from allocating from the BO
+	 * cache.
+	 */
+	memset(bo->base.vaddr + args->size, 0,
+	       bo->base.base.size - args->size);
+
+	bo->validated_shader = vc4_validate_shader(&bo->base);
+	if (!bo->validated_shader) {
+		ret = -EINVAL;
+		goto fail;
+	}
+
+	/* We have to create the handle after validation, to avoid
+	 * races for users to do doing things like mmap the shader BO.
+	 */
+	ret = drm_gem_handle_create(file_priv, &bo->base.base, &args->handle);
+
+ fail:
+	drm_gem_object_unreference_unlocked(&bo->base.base);
+
+	return ret;
+}
+
 void vc4_bo_cache_init(struct drm_device *dev)
 {
 	struct vc4_dev *vc4 = to_vc4_dev(dev);

drivers/gpu/drm/vc4/vc4_drv.c

@@ -64,7 +64,7 @@ static const struct file_operations vc4_drm_fops = {
 	.open = drm_open,
 	.release = drm_release,
 	.unlocked_ioctl = drm_ioctl,
-	.mmap = drm_gem_cma_mmap,
+	.mmap = vc4_mmap,
 	.poll = drm_poll,
 	.read = drm_read,
 #ifdef CONFIG_COMPAT
@@ -76,6 +76,7 @@ static const struct file_operations vc4_drm_fops = {
 static const struct drm_ioctl_desc vc4_drm_ioctls[] = {
 	DRM_IOCTL_DEF_DRV(VC4_CREATE_BO, vc4_create_bo_ioctl, 0),
 	DRM_IOCTL_DEF_DRV(VC4_MMAP_BO, vc4_mmap_bo_ioctl, 0),
+	DRM_IOCTL_DEF_DRV(VC4_CREATE_SHADER_BO, vc4_create_shader_bo_ioctl, 0),
 };
 
 static struct drm_driver vc4_drm_driver = {
@@ -102,12 +103,12 @@ static struct drm_driver vc4_drm_driver = {
 	.prime_handle_to_fd = drm_gem_prime_handle_to_fd,
 	.prime_fd_to_handle = drm_gem_prime_fd_to_handle,
 	.gem_prime_import = drm_gem_prime_import,
-	.gem_prime_export = drm_gem_prime_export,
+	.gem_prime_export = vc4_prime_export,
 	.gem_prime_get_sg_table	= drm_gem_cma_prime_get_sg_table,
 	.gem_prime_import_sg_table = drm_gem_cma_prime_import_sg_table,
-	.gem_prime_vmap = drm_gem_cma_prime_vmap,
+	.gem_prime_vmap = vc4_prime_vmap,
 	.gem_prime_vunmap = drm_gem_cma_prime_vunmap,
-	.gem_prime_mmap = drm_gem_cma_prime_mmap,
+	.gem_prime_mmap = vc4_prime_mmap,
 
 	.dumb_create = vc4_dumb_create,
 	.dumb_map_offset = drm_gem_cma_dumb_map_offset,

drivers/gpu/drm/vc4/vc4_drv.h

@@ -69,6 +69,11 @@ struct vc4_bo {
 
 	/* List entry for the BO's position in vc4_dev->bo_cache.size_list */
 	struct list_head size_head;
+
+	/* Struct for shader validation state, if created by
+	 * DRM_IOCTL_VC4_CREATE_SHADER_BO.
+	 */
+	struct vc4_validated_shader_info *validated_shader;
 };
 
 static inline struct vc4_bo *
@@ -117,6 +122,42 @@ to_vc4_encoder(struct drm_encoder *encoder)
 #define HVS_READ(offset) readl(vc4->hvs->regs + offset)
 #define HVS_WRITE(offset, val) writel(val, vc4->hvs->regs + offset)
 
+/**
+ * struct vc4_texture_sample_info - saves the offsets into the UBO for texture
+ * setup parameters.
+ *
+ * This will be used at draw time to relocate the reference to the texture
+ * contents in p0, and validate that the offset combined with
+ * width/height/stride/etc. from p1 and p2/p3 doesn't sample outside the BO.
+ * Note that the hardware treats unprovided config parameters as 0, so not all
+ * of them need to be set up for every texure sample, and we'll store ~0 as
+ * the offset to mark the unused ones.
+ *
+ * See the VC4 3D architecture guide page 41 ("Texture and Memory Lookup Unit
+ * Setup") for definitions of the texture parameters.
+ */
+struct vc4_texture_sample_info {
+	bool is_direct;
+	uint32_t p_offset[4];
+};
+
+/**
+ * struct vc4_validated_shader_info - information about validated shaders that
+ * needs to be used from command list validation.
+ *
+ * For a given shader, each time a shader state record references it, we need
+ * to verify that the shader doesn't read more uniforms than the shader state
+ * record's uniform BO pointer can provide, and we need to apply relocations
+ * and validate the shader state record's uniforms that define the texture
+ * samples.
+ */
+struct vc4_validated_shader_info {
+	uint32_t uniforms_size;
+	uint32_t uniforms_src_size;
+	uint32_t num_texture_samples;
+	struct vc4_texture_sample_info *texture_samples;
+};
+
 /**
  * _wait_for - magic (register) wait macro
  *
@@ -157,8 +198,13 @@ struct dma_buf *vc4_prime_export(struct drm_device *dev,
 				 struct drm_gem_object *obj, int flags);
 int vc4_create_bo_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *file_priv);
+int vc4_create_shader_bo_ioctl(struct drm_device *dev, void *data,
+			       struct drm_file *file_priv);
 int vc4_mmap_bo_ioctl(struct drm_device *dev, void *data,
 		      struct drm_file *file_priv);
+int vc4_mmap(struct file *filp, struct vm_area_struct *vma);
+int vc4_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma);
+void *vc4_prime_vmap(struct drm_gem_object *obj);
 void vc4_bo_cache_init(struct drm_device *dev);
 void vc4_bo_cache_destroy(struct drm_device *dev);
 int vc4_bo_stats_debugfs(struct seq_file *m, void *arg);
@@ -194,3 +240,7 @@ struct drm_plane *vc4_plane_init(struct drm_device *dev,
 				 enum drm_plane_type type);
 u32 vc4_plane_write_dlist(struct drm_plane *plane, u32 __iomem *dlist);
 u32 vc4_plane_dlist_size(struct drm_plane_state *state);
+
+/* vc4_validate_shader.c */
+struct vc4_validated_shader_info *
+vc4_validate_shader(struct drm_gem_cma_object *shader_obj);