drm/amdgpu: optionally enable GART debugfs file

Keeping the pages array around can use a lot of system memory
when you want a large GART.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/Kconfig

@@ -15,3 +15,13 @@ config DRM_AMDGPU_USERPTR
 	help
 	  This option selects CONFIG_MMU_NOTIFIER if it isn't already
 	  selected to enabled full userptr support.
+
+config DRM_AMDGPU_GART_DEBUGFS
+	bool "Allow GART access through debugfs"
+	depends on DRM_AMDGPU
+	depends on DEBUG_FS
+	default n
+	help
+	  Selecting this option creates a debugfs file to inspect the mapped
+	  pages. Uses more memory for housekeeping, enable only for debugging.
+

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -611,7 +611,9 @@ struct amdgpu_gart {
 	unsigned			num_gpu_pages;
 	unsigned			num_cpu_pages;
 	unsigned			table_size;
+#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
 	struct page			**pages;
+#endif
 	bool				ready;
 	const struct amdgpu_gart_funcs *gart_funcs;
 };

drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c

@@ -238,17 +238,17 @@ void amdgpu_gart_unbind(struct amdgpu_device *adev, unsigned offset,
 	t = offset / AMDGPU_GPU_PAGE_SIZE;
 	p = t / (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE);
 	for (i = 0; i < pages; i++, p++) {
-		if (adev->gart.pages[p]) {
-			adev->gart.pages[p] = NULL;
-			page_base = adev->dummy_page.addr;
-			if (!adev->gart.ptr)
-				continue;
+#ifdef CONFIG_AMDGPU_GART_DEBUGFS
+		adev->gart.pages[p] = NULL;
+#endif
+		page_base = adev->dummy_page.addr;
+		if (!adev->gart.ptr)
+			continue;
 
-			for (j = 0; j < (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); j++, t++) {
-				amdgpu_gart_set_pte_pde(adev, adev->gart.ptr,
-							t, page_base, flags);
-				page_base += AMDGPU_GPU_PAGE_SIZE;
-			}
+		for (j = 0; j < (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); j++, t++) {
+			amdgpu_gart_set_pte_pde(adev, adev->gart.ptr,
+						t, page_base, flags);
+			page_base += AMDGPU_GPU_PAGE_SIZE;
 		}
 	}
 	mb();
@@ -286,7 +286,9 @@ int amdgpu_gart_bind(struct amdgpu_device *adev, unsigned offset,
 	p = t / (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE);
 
 	for (i = 0; i < pages; i++, p++) {
+#ifdef CONFIG_AMDGPU_GART_DEBUGFS
 		adev->gart.pages[p] = pagelist[i];
+#endif
 		if (adev->gart.ptr) {
 			page_base = dma_addr[i];
 			for (j = 0; j < (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); j++, t++) {
@@ -312,9 +314,9 @@ int amdgpu_gart_init(struct amdgpu_device *adev)
 {
 	int r;
 
-	if (adev->gart.pages) {
+	if (adev->dummy_page.page)
 		return 0;
-	}
+
 	/* We need PAGE_SIZE >= AMDGPU_GPU_PAGE_SIZE */
 	if (PAGE_SIZE < AMDGPU_GPU_PAGE_SIZE) {
 		DRM_ERROR("Page size is smaller than GPU page size!\n");
@@ -328,12 +330,16 @@ int amdgpu_gart_init(struct amdgpu_device *adev)
 	adev->gart.num_gpu_pages = adev->mc.gtt_size / AMDGPU_GPU_PAGE_SIZE;
 	DRM_INFO("GART: num cpu pages %u, num gpu pages %u\n",
 		 adev->gart.num_cpu_pages, adev->gart.num_gpu_pages);
+
+#ifdef CONFIG_AMDGPU_GART_DEBUGFS
 	/* Allocate pages table */
 	adev->gart.pages = vzalloc(sizeof(void *) * adev->gart.num_cpu_pages);
 	if (adev->gart.pages == NULL) {
 		amdgpu_gart_fini(adev);
 		return -ENOMEM;
 	}
+#endif
+
 	return 0;
 }
 
@@ -346,13 +352,14 @@ int amdgpu_gart_init(struct amdgpu_device *adev)
  */
 void amdgpu_gart_fini(struct amdgpu_device *adev)
 {
-	if (adev->gart.pages && adev->gart.ready) {
+	if (adev->gart.ready) {
 		/* unbind pages */
 		amdgpu_gart_unbind(adev, 0, adev->gart.num_cpu_pages);
 	}
 	adev->gart.ready = false;
+#ifdef CONFIG_AMDGPU_GART_DEBUGFS
 	vfree(adev->gart.pages);
 	adev->gart.pages = NULL;
-
+#endif
 	amdgpu_dummy_page_fini(adev);
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

@@ -1216,6 +1216,8 @@ static const struct file_operations amdgpu_ttm_vram_fops = {
 	.llseek = default_llseek
 };
 
+#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
+
 static ssize_t amdgpu_ttm_gtt_read(struct file *f, char __user *buf,
 				   size_t size, loff_t *pos)
 {
@@ -1263,6 +1265,8 @@ static const struct file_operations amdgpu_ttm_gtt_fops = {
 
 #endif
 
+#endif
+
 static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
 {
 #if defined(CONFIG_DEBUG_FS)
@@ -1278,6 +1282,7 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
 	i_size_write(ent->d_inode, adev->mc.mc_vram_size);
 	adev->mman.vram = ent;
 
+#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
 	ent = debugfs_create_file("amdgpu_gtt", S_IFREG | S_IRUGO, root,
 				  adev, &amdgpu_ttm_gtt_fops);
 	if (IS_ERR(ent))
@@ -1285,6 +1290,7 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
 	i_size_write(ent->d_inode, adev->mc.gtt_size);
 	adev->mman.gtt = ent;
 
+#endif
 	count = ARRAY_SIZE(amdgpu_ttm_debugfs_list);
 
 #ifdef CONFIG_SWIOTLB
@@ -1306,7 +1312,10 @@ static void amdgpu_ttm_debugfs_fini(struct amdgpu_device *adev)
 	debugfs_remove(adev->mman.vram);
 	adev->mman.vram = NULL;
 
+#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
 	debugfs_remove(adev->mman.gtt);
 	adev->mman.gtt = NULL;
 #endif
+
+#endif
 }