[PATCH 0/4] drm: Ignore non-existing color planes in helpers

[PATCH 0/4] drm: Ignore non-existing color planes in helpers

[Thomas Zimmermann]

Some DRM helpers assume that all potential color planes of a framebuffer
are available; even if the color format didn't specify them. Non-existing
planes are silently ignored. This behavior is inconsistent with other
helpers and apparently leads to subtle bugs with uninitialized GEM buffer
mappings. [1]

Change all affected code to look at the framebuffer format's number of
color planes and only process these planes. The update has been discussed
in [2] before.

Tested with GEM SHMEM helpers on simpledrm.

[1] https://lore.kernel.org/dri-devel/20210730183511.20080-1-tzimmermann@suse.de/T/#md0172b10bb588d8f20f4f456e304f08d2a4505f7
[2] https://lore.kernel.org/dri-devel/877dc0d9-c6c6-022c-20d8-14b33e863934@suse.de/

Thomas Zimmermann (4):
  drm/gem: Share code between drm_gem_fb_{begin,end}_cpu_access()
  drm/gem: Ignore color planes that are unused by framebuffer format
  drm/gem-vram: Ignore planes that are unused by framebuffer format
  drm/gem: Warn on trying to use a non-existing framebuffer plane

 drivers/gpu/drm/drm_gem_atomic_helper.c      |   6 +-
 drivers/gpu/drm/drm_gem_framebuffer_helper.c | 103 +++++++++----------
 drivers/gpu/drm/drm_gem_vram_helper.c        |  36 ++++---
 include/drm/drm_gem_framebuffer_helper.h     |  10 +-
 4 files changed, 81 insertions(+), 74 deletions(-)


base-commit: b0b93865a24c910fcbfa6e6fa0955fae930a56d3
-- 
2.36.0

[PATCH 1/4] drm/gem: Share code between drm_gem_fb_{begin, end}_cpu_access()

[Thomas Zimmermann]

The error-recovery code in drm_gem_fb_begin() is the same pattern
as drm_gem_fb_end(). Implement both this an internal helper. No
functional changes.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
---
 drivers/gpu/drm/drm_gem_framebuffer_helper.c | 62 ++++++++------------
 1 file changed, 26 insertions(+), 36 deletions(-)

diff --git a/drivers/gpu/drm/drm_gem_framebuffer_helper.c b/drivers/gpu/drm/drm_gem_framebuffer_helper.c
index f4619803acd0..2fcacab9f812 100644
--- a/drivers/gpu/drm/drm_gem_framebuffer_helper.c
+++ b/drivers/gpu/drm/drm_gem_framebuffer_helper.c
@@ -403,6 +403,27 @@ void drm_gem_fb_vunmap(struct drm_framebuffer *fb,
 }
 EXPORT_SYMBOL(drm_gem_fb_vunmap);
 
+static void __drm_gem_fb_end_cpu_access(struct drm_framebuffer *fb, enum dma_data_direction dir,
+					unsigned int num_planes)
+{
+	struct dma_buf_attachment *import_attach;
+	struct drm_gem_object *obj;
+	int ret;
+
+	while (num_planes) {
+		--num_planes;
+		obj = drm_gem_fb_get_obj(fb, num_planes);
+		if (!obj)
+			continue;
+		import_attach = obj->import_attach;
+		if (!import_attach)
+			continue;
+		ret = dma_buf_end_cpu_access(import_attach->dmabuf, dir);
+		if (ret)
+			drm_err(fb->dev, "dma_buf_end_cpu_access() failed: %d\n", ret);
+	}
+}
+
 /**
  * drm_gem_fb_begin_cpu_access - prepares GEM buffer objects for CPU access
  * @fb: the framebuffer
@@ -422,7 +443,7 @@ int drm_gem_fb_begin_cpu_access(struct drm_framebuffer *fb, enum dma_data_direct
 	struct dma_buf_attachment *import_attach;
 	struct drm_gem_object *obj;
 	size_t i;
-	int ret, ret2;
+	int ret;
 
 	for (i = 0; i < ARRAY_SIZE(fb->obj); ++i) {
 		obj = drm_gem_fb_get_obj(fb, i);
@@ -433,28 +454,13 @@ int drm_gem_fb_begin_cpu_access(struct drm_framebuffer *fb, enum dma_data_direct
 			continue;
 		ret = dma_buf_begin_cpu_access(import_attach->dmabuf, dir);
 		if (ret)
-			goto err_dma_buf_end_cpu_access;
+			goto err___drm_gem_fb_end_cpu_access;
 	}
 
 	return 0;
 
-err_dma_buf_end_cpu_access:
-	while (i) {
-		--i;
-		obj = drm_gem_fb_get_obj(fb, i);
-		if (!obj)
-			continue;
-		import_attach = obj->import_attach;
-		if (!import_attach)
-			continue;
-		ret2 = dma_buf_end_cpu_access(import_attach->dmabuf, dir);
-		if (ret2) {
-			drm_err(fb->dev,
-				"dma_buf_end_cpu_access() failed during error handling: %d\n",
-				ret2);
-		}
-	}
-
+err___drm_gem_fb_end_cpu_access:
+	__drm_gem_fb_end_cpu_access(fb, dir, i);
 	return ret;
 }
 EXPORT_SYMBOL(drm_gem_fb_begin_cpu_access);
@@ -472,23 +478,7 @@ EXPORT_SYMBOL(drm_gem_fb_begin_cpu_access);
  */
 void drm_gem_fb_end_cpu_access(struct drm_framebuffer *fb, enum dma_data_direction dir)
 {
-	size_t i = ARRAY_SIZE(fb->obj);
-	struct dma_buf_attachment *import_attach;
-	struct drm_gem_object *obj;
-	int ret;
-
-	while (i) {
-		--i;
-		obj = drm_gem_fb_get_obj(fb, i);
-		if (!obj)
-			continue;
-		import_attach = obj->import_attach;
-		if (!import_attach)
-			continue;
-		ret = dma_buf_end_cpu_access(import_attach->dmabuf, dir);
-		if (ret)
-			drm_err(fb->dev, "dma_buf_end_cpu_access() failed: %d\n", ret);
-	}
+	__drm_gem_fb_end_cpu_access(fb, dir, ARRAY_SIZE(fb->obj));
 }
 EXPORT_SYMBOL(drm_gem_fb_end_cpu_access);
 
-- 
2.36.0

[PATCH 2/4] drm/gem: Ignore color planes that are unused by framebuffer format

[Thomas Zimmermann]

Only handle color planes that exist in a framebuffer's color format.
Ignore non-existing planes.

So far, several helpers assumed that all 4 planes are available and
silently ignored non-existing planes. This lead to subtil bugs with
uninitialized data in instances of struct iosys_map. [1]

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://lore.kernel.org/dri-devel/20210730183511.20080-1-tzimmermann@suse.de/T/#md0172b10bb588d8f20f4f456e304f08d2a4505f7 # 1
---
 drivers/gpu/drm/drm_gem_atomic_helper.c      |  6 ++--
 drivers/gpu/drm/drm_gem_framebuffer_helper.c | 37 +++++++++++---------
 include/drm/drm_gem_framebuffer_helper.h     | 10 ++----
 3 files changed, 27 insertions(+), 26 deletions(-)

diff --git a/drivers/gpu/drm/drm_gem_atomic_helper.c b/drivers/gpu/drm/drm_gem_atomic_helper.c
index a5026f617739..f16d60217c6c 100644
--- a/drivers/gpu/drm/drm_gem_atomic_helper.c
+++ b/drivers/gpu/drm/drm_gem_atomic_helper.c
@@ -169,8 +169,10 @@ int drm_gem_plane_helper_prepare_fb(struct drm_plane *plane,
 		struct drm_gem_object *obj = drm_gem_fb_get_obj(state->fb, i);
 		struct dma_fence *new;
 
-		if (WARN_ON_ONCE(!obj))
-			continue;
+		if (!obj) {
+			ret = -EINVAL;
+			goto error;
+		}
 
 		ret = dma_resv_get_singleton(obj->resv, usage, &new);
 		if (ret)
diff --git a/drivers/gpu/drm/drm_gem_framebuffer_helper.c b/drivers/gpu/drm/drm_gem_framebuffer_helper.c
index 2fcacab9f812..09e90e19cd93 100644
--- a/drivers/gpu/drm/drm_gem_framebuffer_helper.c
+++ b/drivers/gpu/drm/drm_gem_framebuffer_helper.c
@@ -92,9 +92,9 @@ drm_gem_fb_init(struct drm_device *dev,
  */
 void drm_gem_fb_destroy(struct drm_framebuffer *fb)
 {
-	size_t i;
+	unsigned int i;
 
-	for (i = 0; i < ARRAY_SIZE(fb->obj); i++)
+	for (i = 0; i < fb->format->num_planes; i++)
 		drm_gem_object_put(fb->obj[i]);
 
 	drm_framebuffer_cleanup(fb);
@@ -329,24 +329,26 @@ EXPORT_SYMBOL_GPL(drm_gem_fb_create_with_dirty);
  * The argument returns the addresses of the data stored in each BO. This
  * is different from @map if the framebuffer's offsets field is non-zero.
  *
+ * Both, @map and @data, must each refer to arrays with at least
+ * fb->format->num_planes elements.
+ *
  * See drm_gem_fb_vunmap() for unmapping.
  *
  * Returns:
  * 0 on success, or a negative errno code otherwise.
  */
-int drm_gem_fb_vmap(struct drm_framebuffer *fb,
-		    struct iosys_map map[static DRM_FORMAT_MAX_PLANES],
-		    struct iosys_map data[DRM_FORMAT_MAX_PLANES])
+int drm_gem_fb_vmap(struct drm_framebuffer *fb, struct iosys_map *map,
+		    struct iosys_map *data)
 {
 	struct drm_gem_object *obj;
 	unsigned int i;
 	int ret;
 
-	for (i = 0; i < DRM_FORMAT_MAX_PLANES; ++i) {
+	for (i = 0; i < fb->format->num_planes; ++i) {
 		obj = drm_gem_fb_get_obj(fb, i);
 		if (!obj) {
-			iosys_map_clear(&map[i]);
-			continue;
+			ret = -EINVAL;
+			goto err_drm_gem_vunmap;
 		}
 		ret = drm_gem_vmap(obj, &map[i]);
 		if (ret)
@@ -354,7 +356,7 @@ int drm_gem_fb_vmap(struct drm_framebuffer *fb,
 	}
 
 	if (data) {
-		for (i = 0; i < DRM_FORMAT_MAX_PLANES; ++i) {
+		for (i = 0; i < fb->format->num_planes; ++i) {
 			memcpy(&data[i], &map[i], sizeof(data[i]));
 			if (iosys_map_is_null(&data[i]))
 				continue;
@@ -385,10 +387,9 @@ EXPORT_SYMBOL(drm_gem_fb_vmap);
  *
  * See drm_gem_fb_vmap() for more information.
  */
-void drm_gem_fb_vunmap(struct drm_framebuffer *fb,
-		       struct iosys_map map[static DRM_FORMAT_MAX_PLANES])
+void drm_gem_fb_vunmap(struct drm_framebuffer *fb, struct iosys_map *map)
 {
-	unsigned int i = DRM_FORMAT_MAX_PLANES;
+	unsigned int i = fb->format->num_planes;
 	struct drm_gem_object *obj;
 
 	while (i) {
@@ -442,13 +443,15 @@ int drm_gem_fb_begin_cpu_access(struct drm_framebuffer *fb, enum dma_data_direct
 {
 	struct dma_buf_attachment *import_attach;
 	struct drm_gem_object *obj;
-	size_t i;
+	unsigned int i;
 	int ret;
 
-	for (i = 0; i < ARRAY_SIZE(fb->obj); ++i) {
+	for (i = 0; i < fb->format->num_planes; ++i) {
 		obj = drm_gem_fb_get_obj(fb, i);
-		if (!obj)
-			continue;
+		if (!obj) {
+			ret = -EINVAL;
+			goto err___drm_gem_fb_end_cpu_access;
+		}
 		import_attach = obj->import_attach;
 		if (!import_attach)
 			continue;
@@ -478,7 +481,7 @@ EXPORT_SYMBOL(drm_gem_fb_begin_cpu_access);
  */
 void drm_gem_fb_end_cpu_access(struct drm_framebuffer *fb, enum dma_data_direction dir)
 {
-	__drm_gem_fb_end_cpu_access(fb, dir, ARRAY_SIZE(fb->obj));
+	__drm_gem_fb_end_cpu_access(fb, dir, fb->format->num_planes);
 }
 EXPORT_SYMBOL(drm_gem_fb_end_cpu_access);
 
diff --git a/include/drm/drm_gem_framebuffer_helper.h b/include/drm/drm_gem_framebuffer_helper.h
index 1091e4fa08cb..d302521f3dd4 100644
--- a/include/drm/drm_gem_framebuffer_helper.h
+++ b/include/drm/drm_gem_framebuffer_helper.h
@@ -4,8 +4,6 @@
 #include <linux/dma-buf.h>
 #include <linux/iosys-map.h>
 
-#include <drm/drm_fourcc.h>
-
 struct drm_afbc_framebuffer;
 struct drm_device;
 struct drm_fb_helper_surface_size;
@@ -39,11 +37,9 @@ struct drm_framebuffer *
 drm_gem_fb_create_with_dirty(struct drm_device *dev, struct drm_file *file,
 			     const struct drm_mode_fb_cmd2 *mode_cmd);
 
-int drm_gem_fb_vmap(struct drm_framebuffer *fb,
-		    struct iosys_map map[static DRM_FORMAT_MAX_PLANES],
-		    struct iosys_map data[DRM_FORMAT_MAX_PLANES]);
-void drm_gem_fb_vunmap(struct drm_framebuffer *fb,
-		       struct iosys_map map[static DRM_FORMAT_MAX_PLANES]);
+int drm_gem_fb_vmap(struct drm_framebuffer *fb, struct iosys_map *map,
+		    struct iosys_map *data);
+void drm_gem_fb_vunmap(struct drm_framebuffer *fb, struct iosys_map *map);
 int drm_gem_fb_begin_cpu_access(struct drm_framebuffer *fb, enum dma_data_direction dir);
 void drm_gem_fb_end_cpu_access(struct drm_framebuffer *fb, enum dma_data_direction dir);
 
-- 
2.36.0

[PATCH 3/4] drm/gem-vram: Ignore planes that are unused by framebuffer format

[Thomas Zimmermann]

Only handle color planes that exist in a framebuffer's color format.
Ignore non-existing planes.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
---
 drivers/gpu/drm/drm_gem_vram_helper.c | 36 ++++++++++++++++++---------
 1 file changed, 24 insertions(+), 12 deletions(-)

diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c b/drivers/gpu/drm/drm_gem_vram_helper.c
index 123045b58fec..3305541a10ab 100644
--- a/drivers/gpu/drm/drm_gem_vram_helper.c
+++ b/drivers/gpu/drm/drm_gem_vram_helper.c
@@ -9,6 +9,7 @@
 #include <drm/drm_file.h>
 #include <drm/drm_framebuffer.h>
 #include <drm/drm_gem_atomic_helper.h>
+#include <drm/drm_gem_framebuffer_helper.h>
 #include <drm/drm_gem_ttm_helper.h>
 #include <drm/drm_gem_vram_helper.h>
 #include <drm/drm_managed.h>
@@ -648,17 +649,22 @@ int
 drm_gem_vram_plane_helper_prepare_fb(struct drm_plane *plane,
 				     struct drm_plane_state *new_state)
 {
-	size_t i;
+	struct drm_framebuffer *fb = new_state->fb;
 	struct drm_gem_vram_object *gbo;
+	struct drm_gem_object *obj;
+	unsigned int i;
 	int ret;
 
-	if (!new_state->fb)
+	if (!fb)
 		return 0;
 
-	for (i = 0; i < ARRAY_SIZE(new_state->fb->obj); ++i) {
-		if (!new_state->fb->obj[i])
-			continue;
-		gbo = drm_gem_vram_of_gem(new_state->fb->obj[i]);
+	for (i = 0; i < fb->format->num_planes; ++i) {
+		obj = drm_gem_fb_get_obj(fb, i);
+		if (!obj) {
+			ret = -EINVAL;
+			goto err_drm_gem_vram_unpin;
+		}
+		gbo = drm_gem_vram_of_gem(obj);
 		ret = drm_gem_vram_pin(gbo, DRM_GEM_VRAM_PL_FLAG_VRAM);
 		if (ret)
 			goto err_drm_gem_vram_unpin;
@@ -673,7 +679,10 @@ drm_gem_vram_plane_helper_prepare_fb(struct drm_plane *plane,
 err_drm_gem_vram_unpin:
 	while (i) {
 		--i;
-		gbo = drm_gem_vram_of_gem(new_state->fb->obj[i]);
+		obj = drm_gem_fb_get_obj(fb, i);
+		if (!obj)
+			continue;
+		gbo = drm_gem_vram_of_gem(obj);
 		drm_gem_vram_unpin(gbo);
 	}
 	return ret;
@@ -694,16 +703,19 @@ void
 drm_gem_vram_plane_helper_cleanup_fb(struct drm_plane *plane,
 				     struct drm_plane_state *old_state)
 {
-	size_t i;
+	struct drm_framebuffer *fb = old_state->fb;
 	struct drm_gem_vram_object *gbo;
+	struct drm_gem_object *obj;
+	unsigned int i;
 
-	if (!old_state->fb)
+	if (!fb)
 		return;
 
-	for (i = 0; i < ARRAY_SIZE(old_state->fb->obj); ++i) {
-		if (!old_state->fb->obj[i])
+	for (i = 0; i < fb->format->num_planes; ++i) {
+		obj = drm_gem_fb_get_obj(fb, i);
+		if (!obj)
 			continue;
-		gbo = drm_gem_vram_of_gem(old_state->fb->obj[i]);
+		gbo = drm_gem_vram_of_gem(obj);
 		drm_gem_vram_unpin(gbo);
 	}
 }
-- 
2.36.0

[PATCH 4/4] drm/gem: Warn on trying to use a non-existing framebuffer plane

[Thomas Zimmermann]

Warn if callers of drm_gem_fb_get_obj() try to use a GEM buffer for
a non-existing or unset plane.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
---
 drivers/gpu/drm/drm_gem_framebuffer_helper.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/drm_gem_framebuffer_helper.c b/drivers/gpu/drm/drm_gem_framebuffer_helper.c
index 09e90e19cd93..291f9ae4359d 100644
--- a/drivers/gpu/drm/drm_gem_framebuffer_helper.c
+++ b/drivers/gpu/drm/drm_gem_framebuffer_helper.c
@@ -53,7 +53,11 @@ MODULE_IMPORT_NS(DMA_BUF);
 struct drm_gem_object *drm_gem_fb_get_obj(struct drm_framebuffer *fb,
 					  unsigned int plane)
 {
-	if (plane >= ARRAY_SIZE(fb->obj))
+	struct drm_device *dev = fb->dev;
+
+	if (drm_WARN_ON_ONCE(dev, plane >= ARRAY_SIZE(fb->obj)))
+		return NULL;
+	else if (drm_WARN_ON_ONCE(dev, !fb->obj[plane]))
 		return NULL;
 
 	return fb->obj[plane];
-- 
2.36.0

Re: [PATCH 0/4] drm: Ignore non-existing color planes in helpers

[Thomas Zimmermann]

[-- Attachment #1.1: Type: text/plain, Size: 1722 bytes --]

pinging for review

Am 09.05.22 um 10:15 schrieb Thomas Zimmermann:
> Some DRM helpers assume that all potential color planes of a framebuffer
> are available; even if the color format didn't specify them. Non-existing
> planes are silently ignored. This behavior is inconsistent with other
> helpers and apparently leads to subtle bugs with uninitialized GEM buffer
> mappings. [1]
> 
> Change all affected code to look at the framebuffer format's number of
> color planes and only process these planes. The update has been discussed
> in [2] before.
> 
> Tested with GEM SHMEM helpers on simpledrm.
> 
> [1] https://lore.kernel.org/dri-devel/20210730183511.20080-1-tzimmermann@suse.de/T/#md0172b10bb588d8f20f4f456e304f08d2a4505f7
> [2] https://lore.kernel.org/dri-devel/877dc0d9-c6c6-022c-20d8-14b33e863934@suse.de/
> 
> Thomas Zimmermann (4):
>    drm/gem: Share code between drm_gem_fb_{begin,end}_cpu_access()
>    drm/gem: Ignore color planes that are unused by framebuffer format
>    drm/gem-vram: Ignore planes that are unused by framebuffer format
>    drm/gem: Warn on trying to use a non-existing framebuffer plane
> 
>   drivers/gpu/drm/drm_gem_atomic_helper.c      |   6 +-
>   drivers/gpu/drm/drm_gem_framebuffer_helper.c | 103 +++++++++----------
>   drivers/gpu/drm/drm_gem_vram_helper.c        |  36 ++++---
>   include/drm/drm_gem_framebuffer_helper.h     |  10 +-
>   4 files changed, 81 insertions(+), 74 deletions(-)
> 
> 
> base-commit: b0b93865a24c910fcbfa6e6fa0955fae930a56d3

-- 
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 840 bytes --]

Re: [PATCH 1/4] drm/gem: Share code between drm_gem_fb_{begin, end}_cpu_access()

[Javier Martinez Canillas]

Hello Thomas,

On 5/9/22 10:15, Thomas Zimmermann wrote:
> The error-recovery code in drm_gem_fb_begin() is the same pattern
> as drm_gem_fb_end(). Implement both this an internal helper. No

Maybe "both of these using using an" or something like that ?

> functional changes.
> 
> Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
> ---
>  drivers/gpu/drm/drm_gem_framebuffer_helper.c | 62 ++++++++------------
>  1 file changed, 26 insertions(+), 36 deletions(-)
>

Nice cleanup. For the patch:

Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>

I just have a few minor comments below.

> diff --git a/drivers/gpu/drm/drm_gem_framebuffer_helper.c b/drivers/gpu/drm/drm_gem_framebuffer_helper.c
> index f4619803acd0..2fcacab9f812 100644
> --- a/drivers/gpu/drm/drm_gem_framebuffer_helper.c
> +++ b/drivers/gpu/drm/drm_gem_framebuffer_helper.c
> @@ -403,6 +403,27 @@ void drm_gem_fb_vunmap(struct drm_framebuffer *fb,
>  }
>  EXPORT_SYMBOL(drm_gem_fb_vunmap);
>  
> +static void __drm_gem_fb_end_cpu_access(struct drm_framebuffer *fb, enum dma_data_direction dir,
> +					unsigned int num_planes)
> +{
> +	struct dma_buf_attachment *import_attach;
> +	struct drm_gem_object *obj;
> +	int ret;
> +
> +	while (num_planes) {
> +		--num_planes;
> +		obj = drm_gem_fb_get_obj(fb, num_planes);
> +		if (!obj)
> +			continue;
> +		import_attach = obj->import_attach;
> +		if (!import_attach)
> +			continue;
> +		ret = dma_buf_end_cpu_access(import_attach->dmabuf, dir);
> +		if (ret)
> +			drm_err(fb->dev, "dma_buf_end_cpu_access() failed: %d\n", ret);

I wonder if would be useful to also print the plane index and access mode here.

> +	}
> +}
> +
>  /**
>   * drm_gem_fb_begin_cpu_access - prepares GEM buffer objects for CPU access
>   * @fb: the framebuffer
> @@ -422,7 +443,7 @@ int drm_gem_fb_begin_cpu_access(struct drm_framebuffer *fb, enum dma_data_direct
>  	struct dma_buf_attachment *import_attach;
>  	struct drm_gem_object *obj;
>  	size_t i;
> -	int ret, ret2;
> +	int ret;
>  
>  	for (i = 0; i < ARRAY_SIZE(fb->obj); ++i) {
>  		obj = drm_gem_fb_get_obj(fb, i);
> @@ -433,28 +454,13 @@ int drm_gem_fb_begin_cpu_access(struct drm_framebuffer *fb, enum dma_data_direct
>  			continue;
>  		ret = dma_buf_begin_cpu_access(import_attach->dmabuf, dir);
>  		if (ret)
> -			goto err_dma_buf_end_cpu_access;
> +			goto err___drm_gem_fb_end_cpu_access;

I wouldn't rename this. As I read it, the goto label doesn't denote what
function is called but rather what action is being taken in an error path.

Since finally what's being done is a dma_buf_end_cpu_access in the helper,
I think that just keeping err_dma_buf_end_cpu_access is enough. Also, the
additional underscores added make it harder to read IMO.

-- 
Best regards,

Javier Martinez Canillas
Linux Engineering
Red Hat

Re: [PATCH 2/4] drm/gem: Ignore color planes that are unused by framebuffer format

[Javier Martinez Canillas]

On 5/9/22 10:16, Thomas Zimmermann wrote:
> Only handle color planes that exist in a framebuffer's color format.
> Ignore non-existing planes.
> 
> So far, several helpers assumed that all 4 planes are available and
> silently ignored non-existing planes. This lead to subtil bugs with
> uninitialized data in instances of struct iosys_map. [1]
> 
> Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
> Link: https://lore.kernel.org/dri-devel/20210730183511.20080-1-tzimmermann@suse.de/T/#md0172b10bb588d8f20f4f456e304f08d2a4505f7 # 1
> ---

Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>

-- 
Best regards,

Javier Martinez Canillas
Linux Engineering
Red Hat

Re: [PATCH 0/4] drm: Ignore non-existing color planes in helpers

[Noralf Trønnes]

Den 09.05.2022 10.15, skrev Thomas Zimmermann:
> Some DRM helpers assume that all potential color planes of a framebuffer
> are available; even if the color format didn't specify them. Non-existing
> planes are silently ignored. This behavior is inconsistent with other
> helpers and apparently leads to subtle bugs with uninitialized GEM buffer
> mappings. [1]
> 
> Change all affected code to look at the framebuffer format's number of
> color planes and only process these planes. The update has been discussed
> in [2] before.
> 
> Tested with GEM SHMEM helpers on simpledrm.
> 

Tested with drm/gud, now there's only one UBSAN warning:

Tested-by: Noralf Trønnes <noralf@tronnes.org>

I was hoping to get some time to give a review, but it doesn't look like
I'm able to do that.

Noralf.


> [1] https://lore.kernel.org/dri-devel/20210730183511.20080-1-tzimmermann@suse.de/T/#md0172b10bb588d8f20f4f456e304f08d2a4505f7
> [2] https://lore.kernel.org/dri-devel/877dc0d9-c6c6-022c-20d8-14b33e863934@suse.de/
> 
> Thomas Zimmermann (4):
>   drm/gem: Share code between drm_gem_fb_{begin,end}_cpu_access()
>   drm/gem: Ignore color planes that are unused by framebuffer format
>   drm/gem-vram: Ignore planes that are unused by framebuffer format
>   drm/gem: Warn on trying to use a non-existing framebuffer plane
> 
>  drivers/gpu/drm/drm_gem_atomic_helper.c      |   6 +-
>  drivers/gpu/drm/drm_gem_framebuffer_helper.c | 103 +++++++++----------
>  drivers/gpu/drm/drm_gem_vram_helper.c        |  36 ++++---
>  include/drm/drm_gem_framebuffer_helper.h     |  10 +-
>  4 files changed, 81 insertions(+), 74 deletions(-)
> 
> 
> base-commit: b0b93865a24c910fcbfa6e6fa0955fae930a56d3

Re: [PATCH 3/4] drm/gem-vram: Ignore planes that are unused by framebuffer format

[Javier Martinez Canillas]

On 5/9/22 10:16, Thomas Zimmermann wrote:
> Only handle color planes that exist in a framebuffer's color format.
> Ignore non-existing planes.
> 
> Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
> ---

Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>

> @@ -673,7 +679,10 @@ drm_gem_vram_plane_helper_prepare_fb(struct drm_plane *plane,
>  err_drm_gem_vram_unpin:
>  	while (i) {
>  		--i;
> -		gbo = drm_gem_vram_of_gem(new_state->fb->obj[i]);
> +		obj = drm_gem_fb_get_obj(fb, i);
> +		if (!obj)
> +			continue;
> +		gbo = drm_gem_vram_of_gem(obj);
>  		drm_gem_vram_unpin(gbo);

The code in this error path to unpin the GEM vram objects...

>  	}
>  	return ret;
> @@ -694,16 +703,19 @@ void
>  drm_gem_vram_plane_helper_cleanup_fb(struct drm_plane *plane,
>  				     struct drm_plane_state *old_state)
>  {
> -	size_t i;
> +	struct drm_framebuffer *fb = old_state->fb;
>  	struct drm_gem_vram_object *gbo;
> +	struct drm_gem_object *obj;
> +	unsigned int i;
>  
> -	if (!old_state->fb)
> +	if (!fb)
>  		return;
>  
> -	for (i = 0; i < ARRAY_SIZE(old_state->fb->obj); ++i) {
> -		if (!old_state->fb->obj[i])
> +	for (i = 0; i < fb->format->num_planes; ++i) {
> +		obj = drm_gem_fb_get_obj(fb, i);
> +		if (!obj)
>  			continue;
> -		gbo = drm_gem_vram_of_gem(old_state->fb->obj[i]);
> +		gbo = drm_gem_vram_of_gem(obj);
>  		drm_gem_vram_unpin(gbo);

... and this, seems to be basically the same.

So maybe as a follow-up you can do a similar cleanup like you did in patch
1/4 and use a helper function to handle both.

-- 
Best regards,

Javier Martinez Canillas
Linux Engineering
Red Hat

Re: [PATCH 4/4] drm/gem: Warn on trying to use a non-existing framebuffer plane

[Javier Martinez Canillas]

On 5/9/22 10:16, Thomas Zimmermann wrote:
> Warn if callers of drm_gem_fb_get_obj() try to use a GEM buffer for
> a non-existing or unset plane.
> 
> Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
> ---

Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>

-- 
Best regards,

Javier Martinez Canillas
Linux Engineering
Red Hat

Re: [PATCH 1/4] drm/gem: Share code between drm_gem_fb_{begin, end}_cpu_access()

[Thomas Zimmermann]

[-- Attachment #1.1: Type: text/plain, Size: 3416 bytes --]

Hi Javier

Am 16.05.22 um 15:00 schrieb Javier Martinez Canillas:
> Hello Thomas,
> 
> On 5/9/22 10:15, Thomas Zimmermann wrote:
>> The error-recovery code in drm_gem_fb_begin() is the same pattern
>> as drm_gem_fb_end(). Implement both this an internal helper. No
> 
> Maybe "both of these using using an" or something like that ?

Of course.

> 
>> functional changes.
>>
>> Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
>> ---
>>   drivers/gpu/drm/drm_gem_framebuffer_helper.c | 62 ++++++++------------
>>   1 file changed, 26 insertions(+), 36 deletions(-)
>>
> 
> Nice cleanup. For the patch:
> 
> Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
> 
> I just have a few minor comments below.
> 
>> diff --git a/drivers/gpu/drm/drm_gem_framebuffer_helper.c b/drivers/gpu/drm/drm_gem_framebuffer_helper.c
>> index f4619803acd0..2fcacab9f812 100644
>> --- a/drivers/gpu/drm/drm_gem_framebuffer_helper.c
>> +++ b/drivers/gpu/drm/drm_gem_framebuffer_helper.c
>> @@ -403,6 +403,27 @@ void drm_gem_fb_vunmap(struct drm_framebuffer *fb,
>>   }
>>   EXPORT_SYMBOL(drm_gem_fb_vunmap);
>>   
>> +static void __drm_gem_fb_end_cpu_access(struct drm_framebuffer *fb, enum dma_data_direction dir,
>> +					unsigned int num_planes)
>> +{
>> +	struct dma_buf_attachment *import_attach;
>> +	struct drm_gem_object *obj;
>> +	int ret;
>> +
>> +	while (num_planes) {
>> +		--num_planes;
>> +		obj = drm_gem_fb_get_obj(fb, num_planes);
>> +		if (!obj)
>> +			continue;
>> +		import_attach = obj->import_attach;
>> +		if (!import_attach)
>> +			continue;
>> +		ret = dma_buf_end_cpu_access(import_attach->dmabuf, dir);
>> +		if (ret)
>> +			drm_err(fb->dev, "dma_buf_end_cpu_access() failed: %d\n", ret);
> 
> I wonder if would be useful to also print the plane index and access mode here.

Ok.

> 
>> +	}
>> +}
>> +
>>   /**
>>    * drm_gem_fb_begin_cpu_access - prepares GEM buffer objects for CPU access
>>    * @fb: the framebuffer
>> @@ -422,7 +443,7 @@ int drm_gem_fb_begin_cpu_access(struct drm_framebuffer *fb, enum dma_data_direct
>>   	struct dma_buf_attachment *import_attach;
>>   	struct drm_gem_object *obj;
>>   	size_t i;
>> -	int ret, ret2;
>> +	int ret;
>>   
>>   	for (i = 0; i < ARRAY_SIZE(fb->obj); ++i) {
>>   		obj = drm_gem_fb_get_obj(fb, i);
>> @@ -433,28 +454,13 @@ int drm_gem_fb_begin_cpu_access(struct drm_framebuffer *fb, enum dma_data_direct
>>   			continue;
>>   		ret = dma_buf_begin_cpu_access(import_attach->dmabuf, dir);
>>   		if (ret)
>> -			goto err_dma_buf_end_cpu_access;
>> +			goto err___drm_gem_fb_end_cpu_access;
> 
> I wouldn't rename this. As I read it, the goto label doesn't denote what
> function is called but rather what action is being taken in an error path.
> 
> Since finally what's being done is a dma_buf_end_cpu_access in the helper,
> I think that just keeping err_dma_buf_end_cpu_access is enough. Also, the
> additional underscores added make it harder to read IMO.

I usually name the goto labels after the function that comes next. It's 
not really pretty here, but being inconsistent would probably annoy me 
more than the underscores.

Best regards
Thomas

> 

-- 
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 840 bytes --]

Re: [PATCH 3/4] drm/gem-vram: Ignore planes that are unused by framebuffer format

[Thomas Zimmermann]

[-- Attachment #1.1: Type: text/plain, Size: 2118 bytes --]

Hi

Am 16.05.22 um 15:34 schrieb Javier Martinez Canillas:
> On 5/9/22 10:16, Thomas Zimmermann wrote:
>> Only handle color planes that exist in a framebuffer's color format.
>> Ignore non-existing planes.
>>
>> Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
>> ---
> 
> Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
> 
>> @@ -673,7 +679,10 @@ drm_gem_vram_plane_helper_prepare_fb(struct drm_plane *plane,
>>   err_drm_gem_vram_unpin:
>>   	while (i) {
>>   		--i;
>> -		gbo = drm_gem_vram_of_gem(new_state->fb->obj[i]);
>> +		obj = drm_gem_fb_get_obj(fb, i);
>> +		if (!obj)
>> +			continue;
>> +		gbo = drm_gem_vram_of_gem(obj);
>>   		drm_gem_vram_unpin(gbo);
> 
> The code in this error path to unpin the GEM vram objects...
> 
>>   	}
>>   	return ret;
>> @@ -694,16 +703,19 @@ void
>>   drm_gem_vram_plane_helper_cleanup_fb(struct drm_plane *plane,
>>   				     struct drm_plane_state *old_state)
>>   {
>> -	size_t i;
>> +	struct drm_framebuffer *fb = old_state->fb;
>>   	struct drm_gem_vram_object *gbo;
>> +	struct drm_gem_object *obj;
>> +	unsigned int i;
>>   
>> -	if (!old_state->fb)
>> +	if (!fb)
>>   		return;
>>   
>> -	for (i = 0; i < ARRAY_SIZE(old_state->fb->obj); ++i) {
>> -		if (!old_state->fb->obj[i])
>> +	for (i = 0; i < fb->format->num_planes; ++i) {
>> +		obj = drm_gem_fb_get_obj(fb, i);
>> +		if (!obj)
>>   			continue;
>> -		gbo = drm_gem_vram_of_gem(old_state->fb->obj[i]);
>> +		gbo = drm_gem_vram_of_gem(obj);
>>   		drm_gem_vram_unpin(gbo);
> 
> ... and this, seems to be basically the same.
> 
> So maybe as a follow-up you can do a similar cleanup like you did in patch
> 1/4 and use a helper function to handle both.

I was thinking the same when I wrote this patch and somehow decided 
against it.  But since you also mention it, I'm going to add this change 
to the patchset.

Best regards
Thomas

> 

-- 
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 840 bytes --]