Linux Samsung SOC development
 help / color / mirror / Atom feed
* [PATCH v2 1/4] drm/exynos: fb: use drm_format_num_planes to get buffer count
@ 2015-04-27 21:10 Tobias Jakobi
  2015-04-27 21:10 ` [PATCH v2 2/4] drm/exynos: plane: honor buffer offset for dma_addr Tobias Jakobi
                   ` (4 more replies)
  0 siblings, 5 replies; 6+ messages in thread
From: Tobias Jakobi @ 2015-04-27 21:10 UTC (permalink / raw)
  To: linux-samsung-soc
  Cc: dri-devel, gustavo.padovan, jy0922.shim, inki.dae, Tobias Jakobi

The previous code had some special case handling for the buffer
count in exynos_drm_format_num_buffers().

This code was incorrect though, since this special case doesn't
exist for DRM. It stemmed from the existence of the special NV12M
V4L2 format. NV12 is a bi-planar format (separate planes for luma
and chroma) and V4L2 differentiates between a NV12 buffer where
luma and chroma is contiguous in memory (so no data between
luma/chroma), and a NV12 buffer where luma and chroma have two
explicit memory locations (which is then called NV12M).

This distinction doesn't exist for DRM. A bi-planar format always
explicitly comes with the information about its two planes (even
if these planes should be contiguous).

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
---
 drivers/gpu/drm/exynos/exynos_drm_fb.c | 39 +---------------------------------
 1 file changed, 1 insertion(+), 38 deletions(-)

diff --git a/drivers/gpu/drm/exynos/exynos_drm_fb.c b/drivers/gpu/drm/exynos/exynos_drm_fb.c
index 929cb03..142eb4e 100644
--- a/drivers/gpu/drm/exynos/exynos_drm_fb.c
+++ b/drivers/gpu/drm/exynos/exynos_drm_fb.c
@@ -171,43 +171,6 @@ exynos_drm_framebuffer_init(struct drm_device *dev,
 	return &exynos_fb->fb;
 }
 
-static u32 exynos_drm_format_num_buffers(struct drm_mode_fb_cmd2 *mode_cmd)
-{
-	unsigned int cnt = 0;
-
-	if (mode_cmd->pixel_format != DRM_FORMAT_NV12)
-		return drm_format_num_planes(mode_cmd->pixel_format);
-
-	while (cnt != MAX_FB_BUFFER) {
-		if (!mode_cmd->handles[cnt])
-			break;
-		cnt++;
-	}
-
-	/*
-	 * check if NV12 or NV12M.
-	 *
-	 * NV12
-	 * handles[0] = base1, offsets[0] = 0
-	 * handles[1] = base1, offsets[1] = Y_size
-	 *
-	 * NV12M
-	 * handles[0] = base1, offsets[0] = 0
-	 * handles[1] = base2, offsets[1] = 0
-	 */
-	if (cnt == 2) {
-		/*
-		 * in case of NV12 format, offsets[1] is not 0 and
-		 * handles[0] is same as handles[1].
-		 */
-		if (mode_cmd->offsets[1] &&
-			mode_cmd->handles[0] == mode_cmd->handles[1])
-			cnt = 1;
-	}
-
-	return cnt;
-}
-
 static struct drm_framebuffer *
 exynos_user_fb_create(struct drm_device *dev, struct drm_file *file_priv,
 		      struct drm_mode_fb_cmd2 *mode_cmd)
@@ -230,7 +193,7 @@ exynos_user_fb_create(struct drm_device *dev, struct drm_file *file_priv,
 
 	drm_helper_mode_fill_fb_struct(&exynos_fb->fb, mode_cmd);
 	exynos_fb->exynos_gem_obj[0] = to_exynos_gem_obj(obj);
-	exynos_fb->buf_cnt = exynos_drm_format_num_buffers(mode_cmd);
+	exynos_fb->buf_cnt = drm_format_num_planes(mode_cmd->pixel_format);
 
 	DRM_DEBUG_KMS("buf_cnt = %d\n", exynos_fb->buf_cnt);
 
-- 
2.0.5

^ permalink raw reply related	[flat|nested] 6+ messages in thread

* [PATCH v2 2/4] drm/exynos: plane: honor buffer offset for dma_addr
  2015-04-27 21:10 [PATCH v2 1/4] drm/exynos: fb: use drm_format_num_planes to get buffer count Tobias Jakobi
@ 2015-04-27 21:10 ` Tobias Jakobi
  2015-04-27 21:10 ` [PATCH v2 3/4] drm/exynos: mixer: remove buffer count handling in vp_video_buffer() Tobias Jakobi
                   ` (3 subsequent siblings)
  4 siblings, 0 replies; 6+ messages in thread
From: Tobias Jakobi @ 2015-04-27 21:10 UTC (permalink / raw)
  To: linux-samsung-soc; +Cc: Tobias Jakobi, gustavo.padovan, dri-devel

Previously we were ignoring the buffer offsets that are
passed through the addfb2 ioctl. This didn't cause any
major issues, since for uni-planar formats (like XRGB8888)
userspace would most of the time just use offsets[0]=0.

However with NV12 offsets[1] is very likely non-zero.
So properly apply the offsets to our dma addresses.

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
---
 drivers/gpu/drm/exynos/exynos_drm_plane.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/exynos/exynos_drm_plane.c b/drivers/gpu/drm/exynos/exynos_drm_plane.c
index 13ea334..b1180fb 100644
--- a/drivers/gpu/drm/exynos/exynos_drm_plane.c
+++ b/drivers/gpu/drm/exynos/exynos_drm_plane.c
@@ -76,7 +76,7 @@ int exynos_check_plane(struct drm_plane *plane, struct drm_framebuffer *fb)
 			return -EFAULT;
 		}
 
-		exynos_plane->dma_addr[i] = buffer->dma_addr;
+		exynos_plane->dma_addr[i] = buffer->dma_addr + fb->offsets[i];
 
 		DRM_DEBUG_KMS("buffer: %d, dma_addr = 0x%lx\n",
 				i, (unsigned long)exynos_plane->dma_addr[i]);
-- 
2.0.5

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

^ permalink raw reply related	[flat|nested] 6+ messages in thread

* [PATCH v2 3/4] drm/exynos: mixer: remove buffer count handling in vp_video_buffer()
  2015-04-27 21:10 [PATCH v2 1/4] drm/exynos: fb: use drm_format_num_planes to get buffer count Tobias Jakobi
  2015-04-27 21:10 ` [PATCH v2 2/4] drm/exynos: plane: honor buffer offset for dma_addr Tobias Jakobi
@ 2015-04-27 21:10 ` Tobias Jakobi
  2015-04-27 21:10 ` [PATCH v2 4/4] drm/exynos: mixer: also allow NV21 for the video processor Tobias Jakobi
                   ` (2 subsequent siblings)
  4 siblings, 0 replies; 6+ messages in thread
From: Tobias Jakobi @ 2015-04-27 21:10 UTC (permalink / raw)
  To: linux-samsung-soc
  Cc: dri-devel, gustavo.padovan, jy0922.shim, inki.dae, Tobias Jakobi

The video processor (VP) supports four formats: NV12, NV21 and its
tiled variants. All these formats are bi-planar, so the buffer
count in vp_video_buffer() is always 2.

Also properly exit if we're called with an invalid (non-VP) pixelformat.

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
---
 drivers/gpu/drm/exynos/exynos_mixer.c | 17 ++---------------
 1 file changed, 2 insertions(+), 15 deletions(-)

diff --git a/drivers/gpu/drm/exynos/exynos_mixer.c b/drivers/gpu/drm/exynos/exynos_mixer.c
index 0474fd3..970b293 100644
--- a/drivers/gpu/drm/exynos/exynos_mixer.c
+++ b/drivers/gpu/drm/exynos/exynos_mixer.c
@@ -388,7 +388,6 @@ static void vp_video_buffer(struct mixer_context *ctx, int win)
 	struct mixer_resources *res = &ctx->mixer_res;
 	unsigned long flags;
 	struct exynos_drm_plane *plane;
-	unsigned int buf_num = 1;
 	dma_addr_t luma_addr[2], chroma_addr[2];
 	bool tiled_mode = false;
 	bool crcb_mode = false;
@@ -399,27 +398,15 @@ static void vp_video_buffer(struct mixer_context *ctx, int win)
 	switch (plane->pixel_format) {
 	case DRM_FORMAT_NV12:
 		crcb_mode = false;
-		buf_num = 2;
 		break;
-	/* TODO: single buffer format NV12, NV21 */
 	default:
-		/* ignore pixel format at disable time */
-		if (!plane->dma_addr[0])
-			break;
-
 		DRM_ERROR("pixel format for vp is wrong [%d].\n",
 				plane->pixel_format);
 		return;
 	}
 
-	if (buf_num == 2) {
-		luma_addr[0] = plane->dma_addr[0];
-		chroma_addr[0] = plane->dma_addr[1];
-	} else {
-		luma_addr[0] = plane->dma_addr[0];
-		chroma_addr[0] = plane->dma_addr[0]
-			+ (plane->pitch * plane->fb_height);
-	}
+	luma_addr[0] = plane->dma_addr[0];
+	chroma_addr[0] = plane->dma_addr[1];
 
 	if (plane->scan_flag & DRM_MODE_FLAG_INTERLACE) {
 		ctx->interlace = true;
-- 
2.0.5

^ permalink raw reply related	[flat|nested] 6+ messages in thread

* [PATCH v2 4/4] drm/exynos: mixer: also allow NV21 for the video processor
  2015-04-27 21:10 [PATCH v2 1/4] drm/exynos: fb: use drm_format_num_planes to get buffer count Tobias Jakobi
  2015-04-27 21:10 ` [PATCH v2 2/4] drm/exynos: plane: honor buffer offset for dma_addr Tobias Jakobi
  2015-04-27 21:10 ` [PATCH v2 3/4] drm/exynos: mixer: remove buffer count handling in vp_video_buffer() Tobias Jakobi
@ 2015-04-27 21:10 ` Tobias Jakobi
  2015-04-28  6:38 ` [PATCH v2 1/4] drm/exynos: fb: use drm_format_num_planes to get buffer count Joonyoung Shim
  2015-05-04  7:21 ` Inki Dae
  4 siblings, 0 replies; 6+ messages in thread
From: Tobias Jakobi @ 2015-04-27 21:10 UTC (permalink / raw)
  To: linux-samsung-soc; +Cc: Tobias Jakobi, gustavo.padovan, dri-devel

All the necessary code is already there, just need to
handle the format in the switch statement.

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
---
 drivers/gpu/drm/exynos/exynos_mixer.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/gpu/drm/exynos/exynos_mixer.c b/drivers/gpu/drm/exynos/exynos_mixer.c
index 970b293..1bcbcfd 100644
--- a/drivers/gpu/drm/exynos/exynos_mixer.c
+++ b/drivers/gpu/drm/exynos/exynos_mixer.c
@@ -399,6 +399,9 @@ static void vp_video_buffer(struct mixer_context *ctx, int win)
 	case DRM_FORMAT_NV12:
 		crcb_mode = false;
 		break;
+	case DRM_FORMAT_NV21:
+		crcb_mode = true;
+		break;
 	default:
 		DRM_ERROR("pixel format for vp is wrong [%d].\n",
 				plane->pixel_format);
-- 
2.0.5

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

^ permalink raw reply related	[flat|nested] 6+ messages in thread

* Re: [PATCH v2 1/4] drm/exynos: fb: use drm_format_num_planes to get buffer count
  2015-04-27 21:10 [PATCH v2 1/4] drm/exynos: fb: use drm_format_num_planes to get buffer count Tobias Jakobi
                   ` (2 preceding siblings ...)
  2015-04-27 21:10 ` [PATCH v2 4/4] drm/exynos: mixer: also allow NV21 for the video processor Tobias Jakobi
@ 2015-04-28  6:38 ` Joonyoung Shim
  2015-05-04  7:21 ` Inki Dae
  4 siblings, 0 replies; 6+ messages in thread
From: Joonyoung Shim @ 2015-04-28  6:38 UTC (permalink / raw)
  To: Tobias Jakobi, linux-samsung-soc; +Cc: dri-devel, gustavo.padovan, inki.dae

Hi Tobias,

On 04/28/2015 06:10 AM, Tobias Jakobi wrote:
> The previous code had some special case handling for the buffer
> count in exynos_drm_format_num_buffers().
> 
> This code was incorrect though, since this special case doesn't
> exist for DRM. It stemmed from the existence of the special NV12M
> V4L2 format. NV12 is a bi-planar format (separate planes for luma
> and chroma) and V4L2 differentiates between a NV12 buffer where
> luma and chroma is contiguous in memory (so no data between
> luma/chroma), and a NV12 buffer where luma and chroma have two
> explicit memory locations (which is then called NV12M).
> 
> This distinction doesn't exist for DRM. A bi-planar format always
> explicitly comes with the information about its two planes (even
> if these planes should be contiguous).
> 
> Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
> ---
>  drivers/gpu/drm/exynos/exynos_drm_fb.c | 39 +---------------------------------
>  1 file changed, 1 insertion(+), 38 deletions(-)
> 

Looks good to me about this series,

Acked-by: Joonyoung Shim <jy0922.shim@samsung.com>

Thanks.

> diff --git a/drivers/gpu/drm/exynos/exynos_drm_fb.c b/drivers/gpu/drm/exynos/exynos_drm_fb.c
> index 929cb03..142eb4e 100644
> --- a/drivers/gpu/drm/exynos/exynos_drm_fb.c
> +++ b/drivers/gpu/drm/exynos/exynos_drm_fb.c
> @@ -171,43 +171,6 @@ exynos_drm_framebuffer_init(struct drm_device *dev,
>  	return &exynos_fb->fb;
>  }
>  
> -static u32 exynos_drm_format_num_buffers(struct drm_mode_fb_cmd2 *mode_cmd)
> -{
> -	unsigned int cnt = 0;
> -
> -	if (mode_cmd->pixel_format != DRM_FORMAT_NV12)
> -		return drm_format_num_planes(mode_cmd->pixel_format);
> -
> -	while (cnt != MAX_FB_BUFFER) {
> -		if (!mode_cmd->handles[cnt])
> -			break;
> -		cnt++;
> -	}
> -
> -	/*
> -	 * check if NV12 or NV12M.
> -	 *
> -	 * NV12
> -	 * handles[0] = base1, offsets[0] = 0
> -	 * handles[1] = base1, offsets[1] = Y_size
> -	 *
> -	 * NV12M
> -	 * handles[0] = base1, offsets[0] = 0
> -	 * handles[1] = base2, offsets[1] = 0
> -	 */
> -	if (cnt == 2) {
> -		/*
> -		 * in case of NV12 format, offsets[1] is not 0 and
> -		 * handles[0] is same as handles[1].
> -		 */
> -		if (mode_cmd->offsets[1] &&
> -			mode_cmd->handles[0] == mode_cmd->handles[1])
> -			cnt = 1;
> -	}
> -
> -	return cnt;
> -}
> -
>  static struct drm_framebuffer *
>  exynos_user_fb_create(struct drm_device *dev, struct drm_file *file_priv,
>  		      struct drm_mode_fb_cmd2 *mode_cmd)
> @@ -230,7 +193,7 @@ exynos_user_fb_create(struct drm_device *dev, struct drm_file *file_priv,
>  
>  	drm_helper_mode_fill_fb_struct(&exynos_fb->fb, mode_cmd);
>  	exynos_fb->exynos_gem_obj[0] = to_exynos_gem_obj(obj);
> -	exynos_fb->buf_cnt = exynos_drm_format_num_buffers(mode_cmd);
> +	exynos_fb->buf_cnt = drm_format_num_planes(mode_cmd->pixel_format);
>  
>  	DRM_DEBUG_KMS("buf_cnt = %d\n", exynos_fb->buf_cnt);
>  
> 

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH v2 1/4] drm/exynos: fb: use drm_format_num_planes to get buffer count
  2015-04-27 21:10 [PATCH v2 1/4] drm/exynos: fb: use drm_format_num_planes to get buffer count Tobias Jakobi
                   ` (3 preceding siblings ...)
  2015-04-28  6:38 ` [PATCH v2 1/4] drm/exynos: fb: use drm_format_num_planes to get buffer count Joonyoung Shim
@ 2015-05-04  7:21 ` Inki Dae
  4 siblings, 0 replies; 6+ messages in thread
From: Inki Dae @ 2015-05-04  7:21 UTC (permalink / raw)
  To: Tobias Jakobi; +Cc: linux-samsung-soc, gustavo.padovan, dri-devel

On 2015년 04월 28일 06:10, Tobias Jakobi wrote:
> The previous code had some special case handling for the buffer
> count in exynos_drm_format_num_buffers().
> 
> This code was incorrect though, since this special case doesn't
> exist for DRM. It stemmed from the existence of the special NV12M
> V4L2 format. NV12 is a bi-planar format (separate planes for luma
> and chroma) and V4L2 differentiates between a NV12 buffer where
> luma and chroma is contiguous in memory (so no data between
> luma/chroma), and a NV12 buffer where luma and chroma have two
> explicit memory locations (which is then called NV12M).
> 
> This distinction doesn't exist for DRM. A bi-planar format always
> explicitly comes with the information about its two planes (even
> if these planes should be contiguous).
> 

For all patches, Applied.

Thanks,
Inki Dae

> Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
> ---
>  drivers/gpu/drm/exynos/exynos_drm_fb.c | 39 +---------------------------------
>  1 file changed, 1 insertion(+), 38 deletions(-)
> 
> diff --git a/drivers/gpu/drm/exynos/exynos_drm_fb.c b/drivers/gpu/drm/exynos/exynos_drm_fb.c
> index 929cb03..142eb4e 100644
> --- a/drivers/gpu/drm/exynos/exynos_drm_fb.c
> +++ b/drivers/gpu/drm/exynos/exynos_drm_fb.c
> @@ -171,43 +171,6 @@ exynos_drm_framebuffer_init(struct drm_device *dev,
>  	return &exynos_fb->fb;
>  }
>  
> -static u32 exynos_drm_format_num_buffers(struct drm_mode_fb_cmd2 *mode_cmd)
> -{
> -	unsigned int cnt = 0;
> -
> -	if (mode_cmd->pixel_format != DRM_FORMAT_NV12)
> -		return drm_format_num_planes(mode_cmd->pixel_format);
> -
> -	while (cnt != MAX_FB_BUFFER) {
> -		if (!mode_cmd->handles[cnt])
> -			break;
> -		cnt++;
> -	}
> -
> -	/*
> -	 * check if NV12 or NV12M.
> -	 *
> -	 * NV12
> -	 * handles[0] = base1, offsets[0] = 0
> -	 * handles[1] = base1, offsets[1] = Y_size
> -	 *
> -	 * NV12M
> -	 * handles[0] = base1, offsets[0] = 0
> -	 * handles[1] = base2, offsets[1] = 0
> -	 */
> -	if (cnt == 2) {
> -		/*
> -		 * in case of NV12 format, offsets[1] is not 0 and
> -		 * handles[0] is same as handles[1].
> -		 */
> -		if (mode_cmd->offsets[1] &&
> -			mode_cmd->handles[0] == mode_cmd->handles[1])
> -			cnt = 1;
> -	}
> -
> -	return cnt;
> -}
> -
>  static struct drm_framebuffer *
>  exynos_user_fb_create(struct drm_device *dev, struct drm_file *file_priv,
>  		      struct drm_mode_fb_cmd2 *mode_cmd)
> @@ -230,7 +193,7 @@ exynos_user_fb_create(struct drm_device *dev, struct drm_file *file_priv,
>  
>  	drm_helper_mode_fill_fb_struct(&exynos_fb->fb, mode_cmd);
>  	exynos_fb->exynos_gem_obj[0] = to_exynos_gem_obj(obj);
> -	exynos_fb->buf_cnt = exynos_drm_format_num_buffers(mode_cmd);
> +	exynos_fb->buf_cnt = drm_format_num_planes(mode_cmd->pixel_format);
>  
>  	DRM_DEBUG_KMS("buf_cnt = %d\n", exynos_fb->buf_cnt);
>  
> 

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2015-05-04  7:21 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2015-04-27 21:10 [PATCH v2 1/4] drm/exynos: fb: use drm_format_num_planes to get buffer count Tobias Jakobi
2015-04-27 21:10 ` [PATCH v2 2/4] drm/exynos: plane: honor buffer offset for dma_addr Tobias Jakobi
2015-04-27 21:10 ` [PATCH v2 3/4] drm/exynos: mixer: remove buffer count handling in vp_video_buffer() Tobias Jakobi
2015-04-27 21:10 ` [PATCH v2 4/4] drm/exynos: mixer: also allow NV21 for the video processor Tobias Jakobi
2015-04-28  6:38 ` [PATCH v2 1/4] drm/exynos: fb: use drm_format_num_planes to get buffer count Joonyoung Shim
2015-05-04  7:21 ` Inki Dae

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox