* [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling
@ 2023-02-03 9:16 Benjamin Gaignard
2023-02-03 9:16 ` [PATCH v8 1/6] media: verisilicon: Do not set context src/dst formats in reset functions Benjamin Gaignard
` (6 more replies)
0 siblings, 7 replies; 10+ messages in thread
From: Benjamin Gaignard @ 2023-02-03 9:16 UTC (permalink / raw)
To: ezequiel, p.zabel, mchehab, shawnguo, s.hauer, kernel, festevam,
linux-imx, hverkuil-cisco, nicolas.dufresne
Cc: linux-media, linux-rockchip, linux-kernel, linux-arm-kernel,
kernel, Benjamin Gaignard
When decoding a 10bits bitstreams HEVC driver should only expose 10bits pixel formats.
To fulfill this requirement it is needed to call hantro_reset_raw_fmt()
and to only change driver internal state in case of success.
Fluster score for HEVC (140/147) doesn't change after this series.
Fluster score for VP9 is 146/303.
version 8:
- Correct patch 4.
- Add a patch for VP9.
version 7:
- Remove unused ctx variable in hantro_try_ctrl().
- Change HANTRO_DEFAULT_BIT_DEPTH value to 8.
- Simplify hantro_check_depth_match logic.
- Keep ctx->bit_depth as integer value because it is use
to compute buffers size for hevc.
version 6:
- Split the patches in multiple sub-patches.
- Rework hantro_reset_encoded_fmt() usage.
version 5:
- Add Nicolas's review tags
- Add Fixes tags
version 4:
- Split the change in 2 patches.
- Change hantro_check_depth_match() prototype to avoid using
ctx->bit_depth
- Return the result of hantro_reset_raw_fmt() to the caller.
- Only set ctx->bit_depth when hantro_reset_raw_fmt() returns is ok.
Benjamin Gaignard (6):
media: verisilicon: Do not set context src/dst formats in reset
functions
media: verisilicon: Do not use ctx fields as format storage when
resetting
media: verisilicon: Do not set ctx->bit_depth in hantro_try_ctrl()
media: verisilicon: Do not change context bit depth before validating
the format
media: verisilicon: HEVC: Only propose 10 bits compatible pixels
formats
media: verisilicon: VP9: Only propose 10 bits compatible pixels
formats
.../media/platform/verisilicon/hantro_drv.c | 49 +++++++---
.../platform/verisilicon/hantro_postproc.c | 2 +-
.../media/platform/verisilicon/hantro_v4l2.c | 90 +++++++++----------
.../media/platform/verisilicon/hantro_v4l2.h | 3 +-
.../media/platform/verisilicon/imx8m_vpu_hw.c | 2 +
5 files changed, 86 insertions(+), 60 deletions(-)
--
2.34.1
^ permalink raw reply [flat|nested] 10+ messages in thread
* [PATCH v8 1/6] media: verisilicon: Do not set context src/dst formats in reset functions
2023-02-03 9:16 [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling Benjamin Gaignard
@ 2023-02-03 9:16 ` Benjamin Gaignard
2023-02-03 9:16 ` [PATCH v8 2/6] media: verisilicon: Do not use ctx fields as format storage when resetting Benjamin Gaignard
` (5 subsequent siblings)
6 siblings, 0 replies; 10+ messages in thread
From: Benjamin Gaignard @ 2023-02-03 9:16 UTC (permalink / raw)
To: ezequiel, p.zabel, mchehab, shawnguo, s.hauer, kernel, festevam,
linux-imx, hverkuil-cisco, nicolas.dufresne
Cc: linux-media, linux-rockchip, linux-kernel, linux-arm-kernel,
kernel, Benjamin Gaignard
Setting context source and destination formats should only be done
in hantro_set_fmt_out() and hantro_set_fmt_cap() after check that
the targeted queue is not busy.
Remove these calls from hantro_reset_encoded_fmt() and
hantro_reset_raw_fmt() to clean the driver.
Fixes: dc39473d0340 ("media: hantro: imx8m: Enable 10bit decoding")
Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
---
drivers/media/platform/verisilicon/hantro_v4l2.c | 9 ++-------
1 file changed, 2 insertions(+), 7 deletions(-)
diff --git a/drivers/media/platform/verisilicon/hantro_v4l2.c b/drivers/media/platform/verisilicon/hantro_v4l2.c
index 2c7a805289e7..33cb865238de 100644
--- a/drivers/media/platform/verisilicon/hantro_v4l2.c
+++ b/drivers/media/platform/verisilicon/hantro_v4l2.c
@@ -381,13 +381,10 @@ hantro_reset_encoded_fmt(struct hantro_ctx *ctx)
vpu_fmt = hantro_get_default_fmt(ctx, true);
- if (ctx->is_encoder) {
- ctx->vpu_dst_fmt = vpu_fmt;
+ if (ctx->is_encoder)
fmt = &ctx->dst_fmt;
- } else {
- ctx->vpu_src_fmt = vpu_fmt;
+ else
fmt = &ctx->src_fmt;
- }
hantro_reset_fmt(fmt, vpu_fmt);
fmt->width = vpu_fmt->frmsize.min_width;
@@ -407,11 +404,9 @@ hantro_reset_raw_fmt(struct hantro_ctx *ctx)
raw_vpu_fmt = hantro_get_default_fmt(ctx, false);
if (ctx->is_encoder) {
- ctx->vpu_src_fmt = raw_vpu_fmt;
raw_fmt = &ctx->src_fmt;
encoded_fmt = &ctx->dst_fmt;
} else {
- ctx->vpu_dst_fmt = raw_vpu_fmt;
raw_fmt = &ctx->dst_fmt;
encoded_fmt = &ctx->src_fmt;
}
--
2.34.1
^ permalink raw reply [flat|nested] 10+ messages in thread
* [PATCH v8 2/6] media: verisilicon: Do not use ctx fields as format storage when resetting
2023-02-03 9:16 [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling Benjamin Gaignard
2023-02-03 9:16 ` [PATCH v8 1/6] media: verisilicon: Do not set context src/dst formats in reset functions Benjamin Gaignard
@ 2023-02-03 9:16 ` Benjamin Gaignard
2023-02-18 12:56 ` Ezequiel Garcia
2023-02-03 9:16 ` [PATCH v8 3/6] media: verisilicon: Do not set ctx->bit_depth in hantro_try_ctrl() Benjamin Gaignard
` (4 subsequent siblings)
6 siblings, 1 reply; 10+ messages in thread
From: Benjamin Gaignard @ 2023-02-03 9:16 UTC (permalink / raw)
To: ezequiel, p.zabel, mchehab, shawnguo, s.hauer, kernel, festevam,
linux-imx, hverkuil-cisco, nicolas.dufresne
Cc: linux-media, linux-rockchip, linux-kernel, linux-arm-kernel,
kernel, Benjamin Gaignard
Source and destination pixel formats fields of context structure should
not be used as storage when resetting the format.
Use local variables instead and let hantro_set_fmt_out() and
hantro_set_fmt_cap() set them correctly later.
Fixes: dc39473d0340 ("media: hantro: imx8m: Enable 10bit decoding")
Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
---
.../media/platform/verisilicon/hantro_v4l2.c | 40 +++++++++----------
1 file changed, 18 insertions(+), 22 deletions(-)
diff --git a/drivers/media/platform/verisilicon/hantro_v4l2.c b/drivers/media/platform/verisilicon/hantro_v4l2.c
index 33cb865238de..e60151a8a401 100644
--- a/drivers/media/platform/verisilicon/hantro_v4l2.c
+++ b/drivers/media/platform/verisilicon/hantro_v4l2.c
@@ -377,47 +377,43 @@ static void
hantro_reset_encoded_fmt(struct hantro_ctx *ctx)
{
const struct hantro_fmt *vpu_fmt;
- struct v4l2_pix_format_mplane *fmt;
+ struct v4l2_pix_format_mplane fmt;
vpu_fmt = hantro_get_default_fmt(ctx, true);
+ if (!vpu_fmt)
+ return;
+ hantro_reset_fmt(&fmt, vpu_fmt);
+ fmt.width = vpu_fmt->frmsize.min_width;
+ fmt.height = vpu_fmt->frmsize.min_height;
if (ctx->is_encoder)
- fmt = &ctx->dst_fmt;
- else
- fmt = &ctx->src_fmt;
-
- hantro_reset_fmt(fmt, vpu_fmt);
- fmt->width = vpu_fmt->frmsize.min_width;
- fmt->height = vpu_fmt->frmsize.min_height;
- if (ctx->is_encoder)
- hantro_set_fmt_cap(ctx, fmt);
+ hantro_set_fmt_cap(ctx, &fmt);
else
- hantro_set_fmt_out(ctx, fmt);
+ hantro_set_fmt_out(ctx, &fmt);
}
static void
hantro_reset_raw_fmt(struct hantro_ctx *ctx)
{
const struct hantro_fmt *raw_vpu_fmt;
- struct v4l2_pix_format_mplane *raw_fmt, *encoded_fmt;
+ struct v4l2_pix_format_mplane raw_fmt, *encoded_fmt;
raw_vpu_fmt = hantro_get_default_fmt(ctx, false);
+ if (!raw_vpu_fmt)
+ return;
- if (ctx->is_encoder) {
- raw_fmt = &ctx->src_fmt;
+ if (ctx->is_encoder)
encoded_fmt = &ctx->dst_fmt;
- } else {
- raw_fmt = &ctx->dst_fmt;
+ else
encoded_fmt = &ctx->src_fmt;
- }
- hantro_reset_fmt(raw_fmt, raw_vpu_fmt);
- raw_fmt->width = encoded_fmt->width;
- raw_fmt->height = encoded_fmt->height;
+ hantro_reset_fmt(&raw_fmt, raw_vpu_fmt);
+ raw_fmt.width = encoded_fmt->width;
+ raw_fmt.height = encoded_fmt->height;
if (ctx->is_encoder)
- hantro_set_fmt_out(ctx, raw_fmt);
+ hantro_set_fmt_out(ctx, &raw_fmt);
else
- hantro_set_fmt_cap(ctx, raw_fmt);
+ hantro_set_fmt_cap(ctx, &raw_fmt);
}
void hantro_reset_fmts(struct hantro_ctx *ctx)
--
2.34.1
^ permalink raw reply [flat|nested] 10+ messages in thread
* [PATCH v8 3/6] media: verisilicon: Do not set ctx->bit_depth in hantro_try_ctrl()
2023-02-03 9:16 [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling Benjamin Gaignard
2023-02-03 9:16 ` [PATCH v8 1/6] media: verisilicon: Do not set context src/dst formats in reset functions Benjamin Gaignard
2023-02-03 9:16 ` [PATCH v8 2/6] media: verisilicon: Do not use ctx fields as format storage when resetting Benjamin Gaignard
@ 2023-02-03 9:16 ` Benjamin Gaignard
2023-02-03 9:16 ` [PATCH v8 4/6] media: verisilicon: Do not change context bit depth before validating the format Benjamin Gaignard
` (3 subsequent siblings)
6 siblings, 0 replies; 10+ messages in thread
From: Benjamin Gaignard @ 2023-02-03 9:16 UTC (permalink / raw)
To: ezequiel, p.zabel, mchehab, shawnguo, s.hauer, kernel, festevam,
linux-imx, hverkuil-cisco, nicolas.dufresne
Cc: linux-media, linux-rockchip, linux-kernel, linux-arm-kernel,
kernel, Benjamin Gaignard
In hantro_try_ctrl() we should only check the values inside
control parameters and not set ctx->bit_depth. That must
be done in controls set function.
Create a set control function for hevc where ctx->bit_depth is
set at the right time.
Fixes: dc39473d0340 ("media: hantro: imx8m: Enable 10bit decoding")
Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
---
.../media/platform/verisilicon/hantro_drv.c | 32 ++++++++++++++-----
1 file changed, 24 insertions(+), 8 deletions(-)
diff --git a/drivers/media/platform/verisilicon/hantro_drv.c b/drivers/media/platform/verisilicon/hantro_drv.c
index 8cb4a68c9119..6d8bc55ea627 100644
--- a/drivers/media/platform/verisilicon/hantro_drv.c
+++ b/drivers/media/platform/verisilicon/hantro_drv.c
@@ -251,11 +251,6 @@ queue_init(void *priv, struct vb2_queue *src_vq, struct vb2_queue *dst_vq)
static int hantro_try_ctrl(struct v4l2_ctrl *ctrl)
{
- struct hantro_ctx *ctx;
-
- ctx = container_of(ctrl->handler,
- struct hantro_ctx, ctrl_handler);
-
if (ctrl->id == V4L2_CID_STATELESS_H264_SPS) {
const struct v4l2_ctrl_h264_sps *sps = ctrl->p_new.p_h264_sps;
@@ -274,8 +269,6 @@ static int hantro_try_ctrl(struct v4l2_ctrl *ctrl)
if (sps->bit_depth_luma_minus8 != 0 && sps->bit_depth_luma_minus8 != 2)
/* Only 8-bit and 10-bit are supported */
return -EINVAL;
-
- ctx->bit_depth = sps->bit_depth_luma_minus8 + 8;
} else if (ctrl->id == V4L2_CID_STATELESS_VP9_FRAME) {
const struct v4l2_ctrl_vp9_frame *dec_params = ctrl->p_new.p_vp9_frame;
@@ -324,6 +317,24 @@ static int hantro_vp9_s_ctrl(struct v4l2_ctrl *ctrl)
return 0;
}
+static int hantro_hevc_s_ctrl(struct v4l2_ctrl *ctrl)
+{
+ struct hantro_ctx *ctx;
+
+ ctx = container_of(ctrl->handler,
+ struct hantro_ctx, ctrl_handler);
+
+ switch (ctrl->id) {
+ case V4L2_CID_STATELESS_HEVC_SPS:
+ ctx->bit_depth = ctrl->p_new.p_hevc_sps->bit_depth_luma_minus8 + 8;
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
static const struct v4l2_ctrl_ops hantro_ctrl_ops = {
.try_ctrl = hantro_try_ctrl,
};
@@ -336,6 +347,11 @@ static const struct v4l2_ctrl_ops hantro_vp9_ctrl_ops = {
.s_ctrl = hantro_vp9_s_ctrl,
};
+static const struct v4l2_ctrl_ops hantro_hevc_ctrl_ops = {
+ .try_ctrl = hantro_try_ctrl,
+ .s_ctrl = hantro_hevc_s_ctrl,
+};
+
#define HANTRO_JPEG_ACTIVE_MARKERS (V4L2_JPEG_ACTIVE_MARKER_APP0 | \
V4L2_JPEG_ACTIVE_MARKER_COM | \
V4L2_JPEG_ACTIVE_MARKER_DQT | \
@@ -470,7 +486,7 @@ static const struct hantro_ctrl controls[] = {
.codec = HANTRO_HEVC_DECODER,
.cfg = {
.id = V4L2_CID_STATELESS_HEVC_SPS,
- .ops = &hantro_ctrl_ops,
+ .ops = &hantro_hevc_ctrl_ops,
},
}, {
.codec = HANTRO_HEVC_DECODER,
--
2.34.1
^ permalink raw reply [flat|nested] 10+ messages in thread
* [PATCH v8 4/6] media: verisilicon: Do not change context bit depth before validating the format
2023-02-03 9:16 [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling Benjamin Gaignard
` (2 preceding siblings ...)
2023-02-03 9:16 ` [PATCH v8 3/6] media: verisilicon: Do not set ctx->bit_depth in hantro_try_ctrl() Benjamin Gaignard
@ 2023-02-03 9:16 ` Benjamin Gaignard
2023-02-03 9:16 ` [PATCH v8 5/6] media: verisilicon: HEVC: Only propose 10 bits compatible pixels formats Benjamin Gaignard
` (2 subsequent siblings)
6 siblings, 0 replies; 10+ messages in thread
From: Benjamin Gaignard @ 2023-02-03 9:16 UTC (permalink / raw)
To: ezequiel, p.zabel, mchehab, shawnguo, s.hauer, kernel, festevam,
linux-imx, hverkuil-cisco, nicolas.dufresne
Cc: linux-media, linux-rockchip, linux-kernel, linux-arm-kernel,
kernel, Benjamin Gaignard, Nicolas Dufresne
It is needed to check if the proposed pixels format is valid before
updating context bit depth and other internal states.
Stop using ctx->bit_depth to check format depth match and return
result to the caller.
Fixes: dc39473d0340 ("media: hantro: imx8m: Enable 10bit decoding")
Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
---
.../platform/verisilicon/hantro_postproc.c | 2 +-
.../media/platform/verisilicon/hantro_v4l2.c | 51 ++++++++++---------
.../media/platform/verisilicon/hantro_v4l2.h | 3 +-
3 files changed, 30 insertions(+), 26 deletions(-)
version 8:
- Set HANTRO_DEFAULT_BIT_DEPTH to 8
diff --git a/drivers/media/platform/verisilicon/hantro_postproc.c b/drivers/media/platform/verisilicon/hantro_postproc.c
index 09d8cf942689..6437423ccf3a 100644
--- a/drivers/media/platform/verisilicon/hantro_postproc.c
+++ b/drivers/media/platform/verisilicon/hantro_postproc.c
@@ -197,7 +197,7 @@ int hantro_postproc_alloc(struct hantro_ctx *ctx)
unsigned int i, buf_size;
/* this should always pick native format */
- fmt = hantro_get_default_fmt(ctx, false);
+ fmt = hantro_get_default_fmt(ctx, false, ctx->bit_depth);
if (!fmt)
return -EINVAL;
v4l2_fill_pixfmt_mp(&pix_mp, fmt->fourcc, ctx->src_fmt.width,
diff --git a/drivers/media/platform/verisilicon/hantro_v4l2.c b/drivers/media/platform/verisilicon/hantro_v4l2.c
index e60151a8a401..7c9a04171b45 100644
--- a/drivers/media/platform/verisilicon/hantro_v4l2.c
+++ b/drivers/media/platform/verisilicon/hantro_v4l2.c
@@ -28,6 +28,8 @@
#include "hantro_hw.h"
#include "hantro_v4l2.h"
+#define HANTRO_DEFAULT_BIT_DEPTH 8
+
static int hantro_set_fmt_out(struct hantro_ctx *ctx,
struct v4l2_pix_format_mplane *pix_mp);
static int hantro_set_fmt_cap(struct hantro_ctx *ctx,
@@ -76,18 +78,13 @@ int hantro_get_format_depth(u32 fourcc)
}
static bool
-hantro_check_depth_match(const struct hantro_ctx *ctx,
- const struct hantro_fmt *fmt)
+hantro_check_depth_match(const struct hantro_fmt *fmt, int bit_depth)
{
- int fmt_depth, ctx_depth = 8;
+ int fmt_depth;
if (!fmt->match_depth && !fmt->postprocessed)
return true;
- /* 0 means default depth, which is 8 */
- if (ctx->bit_depth)
- ctx_depth = ctx->bit_depth;
-
fmt_depth = hantro_get_format_depth(fmt->fourcc);
/*
@@ -95,9 +92,9 @@ hantro_check_depth_match(const struct hantro_ctx *ctx,
* It may be possible to relax that on some HW.
*/
if (!fmt->match_depth)
- return fmt_depth <= ctx_depth;
+ return fmt_depth <= bit_depth;
- return fmt_depth == ctx_depth;
+ return fmt_depth == bit_depth;
}
static const struct hantro_fmt *
@@ -119,7 +116,7 @@ hantro_find_format(const struct hantro_ctx *ctx, u32 fourcc)
}
const struct hantro_fmt *
-hantro_get_default_fmt(const struct hantro_ctx *ctx, bool bitstream)
+hantro_get_default_fmt(const struct hantro_ctx *ctx, bool bitstream, int bit_depth)
{
const struct hantro_fmt *formats;
unsigned int i, num_fmts;
@@ -128,7 +125,7 @@ hantro_get_default_fmt(const struct hantro_ctx *ctx, bool bitstream)
for (i = 0; i < num_fmts; i++) {
if (bitstream == (formats[i].codec_mode !=
HANTRO_MODE_NONE) &&
- hantro_check_depth_match(ctx, &formats[i]))
+ hantro_check_depth_match(&formats[i], bit_depth))
return &formats[i];
}
return NULL;
@@ -203,7 +200,7 @@ static int vidioc_enum_fmt(struct file *file, void *priv,
if (skip_mode_none == mode_none)
continue;
- if (!hantro_check_depth_match(ctx, fmt))
+ if (!hantro_check_depth_match(fmt, ctx->bit_depth))
continue;
if (j == f->index) {
f->pixelformat = fmt->fourcc;
@@ -223,7 +220,7 @@ static int vidioc_enum_fmt(struct file *file, void *priv,
for (i = 0; i < num_fmts; i++) {
fmt = &formats[i];
- if (!hantro_check_depth_match(ctx, fmt))
+ if (!hantro_check_depth_match(fmt, ctx->bit_depth))
continue;
if (j == f->index) {
f->pixelformat = fmt->fourcc;
@@ -291,7 +288,7 @@ static int hantro_try_fmt(const struct hantro_ctx *ctx,
fmt = hantro_find_format(ctx, pix_mp->pixelformat);
if (!fmt) {
- fmt = hantro_get_default_fmt(ctx, coded);
+ fmt = hantro_get_default_fmt(ctx, coded, HANTRO_DEFAULT_BIT_DEPTH);
pix_mp->pixelformat = fmt->fourcc;
}
@@ -379,7 +376,7 @@ hantro_reset_encoded_fmt(struct hantro_ctx *ctx)
const struct hantro_fmt *vpu_fmt;
struct v4l2_pix_format_mplane fmt;
- vpu_fmt = hantro_get_default_fmt(ctx, true);
+ vpu_fmt = hantro_get_default_fmt(ctx, true, HANTRO_DEFAULT_BIT_DEPTH);
if (!vpu_fmt)
return;
@@ -392,15 +389,16 @@ hantro_reset_encoded_fmt(struct hantro_ctx *ctx)
hantro_set_fmt_out(ctx, &fmt);
}
-static void
-hantro_reset_raw_fmt(struct hantro_ctx *ctx)
+int
+hantro_reset_raw_fmt(struct hantro_ctx *ctx, int bit_depth)
{
const struct hantro_fmt *raw_vpu_fmt;
struct v4l2_pix_format_mplane raw_fmt, *encoded_fmt;
+ int ret;
- raw_vpu_fmt = hantro_get_default_fmt(ctx, false);
+ raw_vpu_fmt = hantro_get_default_fmt(ctx, false, bit_depth);
if (!raw_vpu_fmt)
- return;
+ return -EINVAL;
if (ctx->is_encoder)
encoded_fmt = &ctx->dst_fmt;
@@ -411,15 +409,20 @@ hantro_reset_raw_fmt(struct hantro_ctx *ctx)
raw_fmt.width = encoded_fmt->width;
raw_fmt.height = encoded_fmt->height;
if (ctx->is_encoder)
- hantro_set_fmt_out(ctx, &raw_fmt);
+ ret = hantro_set_fmt_out(ctx, &raw_fmt);
else
- hantro_set_fmt_cap(ctx, &raw_fmt);
+ ret = hantro_set_fmt_cap(ctx, &raw_fmt);
+
+ if (!ret)
+ ctx->bit_depth = bit_depth;
+
+ return ret;
}
void hantro_reset_fmts(struct hantro_ctx *ctx)
{
hantro_reset_encoded_fmt(ctx);
- hantro_reset_raw_fmt(ctx);
+ hantro_reset_raw_fmt(ctx, HANTRO_DEFAULT_BIT_DEPTH);
}
static void
@@ -519,7 +522,7 @@ static int hantro_set_fmt_out(struct hantro_ctx *ctx,
* changes to the raw format.
*/
if (!ctx->is_encoder)
- hantro_reset_raw_fmt(ctx);
+ hantro_reset_raw_fmt(ctx, hantro_get_format_depth(pix_mp->pixelformat));
/* Colorimetry information are always propagated. */
ctx->dst_fmt.colorspace = pix_mp->colorspace;
@@ -582,7 +585,7 @@ static int hantro_set_fmt_cap(struct hantro_ctx *ctx,
* changes to the raw format.
*/
if (ctx->is_encoder)
- hantro_reset_raw_fmt(ctx);
+ hantro_reset_raw_fmt(ctx, HANTRO_DEFAULT_BIT_DEPTH);
/* Colorimetry information are always propagated. */
ctx->src_fmt.colorspace = pix_mp->colorspace;
diff --git a/drivers/media/platform/verisilicon/hantro_v4l2.h b/drivers/media/platform/verisilicon/hantro_v4l2.h
index 64f6f57e9d7a..9ea2fef57dcd 100644
--- a/drivers/media/platform/verisilicon/hantro_v4l2.h
+++ b/drivers/media/platform/verisilicon/hantro_v4l2.h
@@ -21,9 +21,10 @@
extern const struct v4l2_ioctl_ops hantro_ioctl_ops;
extern const struct vb2_ops hantro_queue_ops;
+int hantro_reset_raw_fmt(struct hantro_ctx *ctx, int bit_depth);
void hantro_reset_fmts(struct hantro_ctx *ctx);
int hantro_get_format_depth(u32 fourcc);
const struct hantro_fmt *
-hantro_get_default_fmt(const struct hantro_ctx *ctx, bool bitstream);
+hantro_get_default_fmt(const struct hantro_ctx *ctx, bool bitstream, int bit_depth);
#endif /* HANTRO_V4L2_H_ */
--
2.34.1
^ permalink raw reply [flat|nested] 10+ messages in thread
* [PATCH v8 5/6] media: verisilicon: HEVC: Only propose 10 bits compatible pixels formats
2023-02-03 9:16 [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling Benjamin Gaignard
` (3 preceding siblings ...)
2023-02-03 9:16 ` [PATCH v8 4/6] media: verisilicon: Do not change context bit depth before validating the format Benjamin Gaignard
@ 2023-02-03 9:16 ` Benjamin Gaignard
2023-02-18 12:57 ` Ezequiel Garcia
2023-02-03 9:16 ` [PATCH v8 6/6] media: verisilicon: VP9: " Benjamin Gaignard
2023-02-18 12:54 ` [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling Ezequiel Garcia
6 siblings, 1 reply; 10+ messages in thread
From: Benjamin Gaignard @ 2023-02-03 9:16 UTC (permalink / raw)
To: ezequiel, p.zabel, mchehab, shawnguo, s.hauer, kernel, festevam,
linux-imx, hverkuil-cisco, nicolas.dufresne
Cc: linux-media, linux-rockchip, linux-kernel, linux-arm-kernel,
kernel, Benjamin Gaignard, Nicolas Dufresne
When decoding a 10bits bitstreams HEVC driver should only expose
10bits pixel formats.
To fulfill this requirement it is needed to call hantro_reset_raw_fmt()
when bit depth change and to correctly set match_depth in pixel formats
enumeration.
Fixes: dc39473d0340 ("media: hantro: imx8m: Enable 10bit decoding")
Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
---
drivers/media/platform/verisilicon/hantro_drv.c | 11 +++++++++--
drivers/media/platform/verisilicon/imx8m_vpu_hw.c | 2 ++
2 files changed, 11 insertions(+), 2 deletions(-)
diff --git a/drivers/media/platform/verisilicon/hantro_drv.c b/drivers/media/platform/verisilicon/hantro_drv.c
index 6d8bc55ea627..fa31b200b097 100644
--- a/drivers/media/platform/verisilicon/hantro_drv.c
+++ b/drivers/media/platform/verisilicon/hantro_drv.c
@@ -326,8 +326,15 @@ static int hantro_hevc_s_ctrl(struct v4l2_ctrl *ctrl)
switch (ctrl->id) {
case V4L2_CID_STATELESS_HEVC_SPS:
- ctx->bit_depth = ctrl->p_new.p_hevc_sps->bit_depth_luma_minus8 + 8;
- break;
+ {
+ const struct v4l2_ctrl_hevc_sps *sps = ctrl->p_new.p_hevc_sps;
+ int bit_depth = sps->bit_depth_luma_minus8 + 8;
+
+ if (ctx->bit_depth == bit_depth)
+ return 0;
+
+ return hantro_reset_raw_fmt(ctx, bit_depth);
+ }
default:
return -EINVAL;
}
diff --git a/drivers/media/platform/verisilicon/imx8m_vpu_hw.c b/drivers/media/platform/verisilicon/imx8m_vpu_hw.c
index b390228fd3b4..f850d8bddef6 100644
--- a/drivers/media/platform/verisilicon/imx8m_vpu_hw.c
+++ b/drivers/media/platform/verisilicon/imx8m_vpu_hw.c
@@ -152,6 +152,7 @@ static const struct hantro_fmt imx8m_vpu_g2_postproc_fmts[] = {
{
.fourcc = V4L2_PIX_FMT_NV12,
.codec_mode = HANTRO_MODE_NONE,
+ .match_depth = true,
.postprocessed = true,
.frmsize = {
.min_width = FMT_MIN_WIDTH,
@@ -165,6 +166,7 @@ static const struct hantro_fmt imx8m_vpu_g2_postproc_fmts[] = {
{
.fourcc = V4L2_PIX_FMT_P010,
.codec_mode = HANTRO_MODE_NONE,
+ .match_depth = true,
.postprocessed = true,
.frmsize = {
.min_width = FMT_MIN_WIDTH,
--
2.34.1
^ permalink raw reply [flat|nested] 10+ messages in thread
* [PATCH v8 6/6] media: verisilicon: VP9: Only propose 10 bits compatible pixels formats
2023-02-03 9:16 [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling Benjamin Gaignard
` (4 preceding siblings ...)
2023-02-03 9:16 ` [PATCH v8 5/6] media: verisilicon: HEVC: Only propose 10 bits compatible pixels formats Benjamin Gaignard
@ 2023-02-03 9:16 ` Benjamin Gaignard
2023-02-18 12:54 ` [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling Ezequiel Garcia
6 siblings, 0 replies; 10+ messages in thread
From: Benjamin Gaignard @ 2023-02-03 9:16 UTC (permalink / raw)
To: ezequiel, p.zabel, mchehab, shawnguo, s.hauer, kernel, festevam,
linux-imx, hverkuil-cisco, nicolas.dufresne
Cc: linux-media, linux-rockchip, linux-kernel, linux-arm-kernel,
kernel, Benjamin Gaignard
When decoding a 10bits bitstreams VP9 driver should only expose
10bits pixel formats.
To fulfill this requirement it is needed to call hantro_reset_raw_fmt()
when bit depth change and to correctly set match_depth in pixel formats
enumeration.
Fixes: dc39473d0340 ("media: hantro: imx8m: Enable 10bit decoding")
Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
---
drivers/media/platform/verisilicon/hantro_drv.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)
diff --git a/drivers/media/platform/verisilicon/hantro_drv.c b/drivers/media/platform/verisilicon/hantro_drv.c
index fa31b200b097..1c4b90c905ea 100644
--- a/drivers/media/platform/verisilicon/hantro_drv.c
+++ b/drivers/media/platform/verisilicon/hantro_drv.c
@@ -308,8 +308,14 @@ static int hantro_vp9_s_ctrl(struct v4l2_ctrl *ctrl)
switch (ctrl->id) {
case V4L2_CID_STATELESS_VP9_FRAME:
- ctx->bit_depth = ctrl->p_new.p_vp9_frame->bit_depth;
- break;
+ {
+ int bit_depth = ctrl->p_new.p_vp9_frame->bit_depth;
+
+ if (ctx->bit_depth == bit_depth)
+ return 0;
+
+ return hantro_reset_raw_fmt(ctx, bit_depth);
+ }
default:
return -EINVAL;
}
--
2.34.1
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling
2023-02-03 9:16 [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling Benjamin Gaignard
` (5 preceding siblings ...)
2023-02-03 9:16 ` [PATCH v8 6/6] media: verisilicon: VP9: " Benjamin Gaignard
@ 2023-02-18 12:54 ` Ezequiel Garcia
6 siblings, 0 replies; 10+ messages in thread
From: Ezequiel Garcia @ 2023-02-18 12:54 UTC (permalink / raw)
To: Benjamin Gaignard
Cc: robert.mader, p.zabel, mchehab, shawnguo, s.hauer, kernel,
festevam, linux-imx, hverkuil-cisco, nicolas.dufresne,
linux-media, linux-rockchip, linux-kernel, linux-arm-kernel,
kernel
Hi Benjamin,
We are almost there, just some minor comments.
On Fri, Feb 3 2023 at 10:16:16 AM +0100, Benjamin Gaignard
<benjamin.gaignard@collabora.com> wrote:
> When decoding a 10bits bitstreams HEVC driver should only expose
> 10bits pixel formats.
> To fulfill this requirement it is needed to call
> hantro_reset_raw_fmt()
> and to only change driver internal state in case of success.
>
> Fluster score for HEVC (140/147) doesn't change after this series.
> Fluster score for VP9 is 146/303.
>
Given the series is changing the format negotation which affects
all codecs, can you test MPEG-2, VP8, H.264 and JPEG encoding?
Can you also run v4l2-compliance on all the /dev/videoX devices?
(Adding Robert Mader, who recently helped test the JPEG encoder).
Thanks,
Ezequiel
> version 8:
> - Correct patch 4.
> - Add a patch for VP9.
>
> version 7:
> - Remove unused ctx variable in hantro_try_ctrl().
> - Change HANTRO_DEFAULT_BIT_DEPTH value to 8.
> - Simplify hantro_check_depth_match logic.
> - Keep ctx->bit_depth as integer value because it is use
> to compute buffers size for hevc.
>
> version 6:
> - Split the patches in multiple sub-patches.
> - Rework hantro_reset_encoded_fmt() usage.
>
> version 5:
> - Add Nicolas's review tags
> - Add Fixes tags
>
> version 4:
> - Split the change in 2 patches.
> - Change hantro_check_depth_match() prototype to avoid using
> ctx->bit_depth
> - Return the result of hantro_reset_raw_fmt() to the caller.
> - Only set ctx->bit_depth when hantro_reset_raw_fmt() returns is ok.
>
> Benjamin Gaignard (6):
> media: verisilicon: Do not set context src/dst formats in reset
> functions
> media: verisilicon: Do not use ctx fields as format storage when
> resetting
> media: verisilicon: Do not set ctx->bit_depth in hantro_try_ctrl()
> media: verisilicon: Do not change context bit depth before
> validating
> the format
> media: verisilicon: HEVC: Only propose 10 bits compatible pixels
> formats
> media: verisilicon: VP9: Only propose 10 bits compatible pixels
> formats
>
> .../media/platform/verisilicon/hantro_drv.c | 49 +++++++---
> .../platform/verisilicon/hantro_postproc.c | 2 +-
> .../media/platform/verisilicon/hantro_v4l2.c | 90
> +++++++++----------
> .../media/platform/verisilicon/hantro_v4l2.h | 3 +-
> .../media/platform/verisilicon/imx8m_vpu_hw.c | 2 +
> 5 files changed, 86 insertions(+), 60 deletions(-)
>
> --
> 2.34.1
>
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v8 2/6] media: verisilicon: Do not use ctx fields as format storage when resetting
2023-02-03 9:16 ` [PATCH v8 2/6] media: verisilicon: Do not use ctx fields as format storage when resetting Benjamin Gaignard
@ 2023-02-18 12:56 ` Ezequiel Garcia
0 siblings, 0 replies; 10+ messages in thread
From: Ezequiel Garcia @ 2023-02-18 12:56 UTC (permalink / raw)
To: Benjamin Gaignard
Cc: p.zabel, mchehab, shawnguo, s.hauer, kernel, festevam, linux-imx,
hverkuil-cisco, nicolas.dufresne, linux-media, linux-rockchip,
linux-kernel, linux-arm-kernel, kernel
Hi Benjamin,
On Fri, Feb 3 2023 at 10:16:18 AM +0100, Benjamin Gaignard
<benjamin.gaignard@collabora.com> wrote:
> Source and destination pixel formats fields of context structure
> should
> not be used as storage when resetting the format.
> Use local variables instead and let hantro_set_fmt_out() and
> hantro_set_fmt_cap() set them correctly later.
>
> Fixes: dc39473d0340 ("media: hantro: imx8m: Enable 10bit decoding")
>
The above Fixes tag looks incorrect. I am unsure what would be the
right Fixes, perhaps we can avoid putting any?
Same for all the other patches, the Fixes tag look wrong.
Thanks,
Ezequiel
> Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
> ---
> .../media/platform/verisilicon/hantro_v4l2.c | 40
> +++++++++----------
> 1 file changed, 18 insertions(+), 22 deletions(-)
>
> diff --git a/drivers/media/platform/verisilicon/hantro_v4l2.c
> b/drivers/media/platform/verisilicon/hantro_v4l2.c
> index 33cb865238de..e60151a8a401 100644
> --- a/drivers/media/platform/verisilicon/hantro_v4l2.c
> +++ b/drivers/media/platform/verisilicon/hantro_v4l2.c
> @@ -377,47 +377,43 @@ static void
> hantro_reset_encoded_fmt(struct hantro_ctx *ctx)
> {
> const struct hantro_fmt *vpu_fmt;
> - struct v4l2_pix_format_mplane *fmt;
> + struct v4l2_pix_format_mplane fmt;
>
> vpu_fmt = hantro_get_default_fmt(ctx, true);
> + if (!vpu_fmt)
> + return;
>
> + hantro_reset_fmt(&fmt, vpu_fmt);
> + fmt.width = vpu_fmt->frmsize.min_width;
> + fmt.height = vpu_fmt->frmsize.min_height;
> if (ctx->is_encoder)
> - fmt = &ctx->dst_fmt;
> - else
> - fmt = &ctx->src_fmt;
> -
> - hantro_reset_fmt(fmt, vpu_fmt);
> - fmt->width = vpu_fmt->frmsize.min_width;
> - fmt->height = vpu_fmt->frmsize.min_height;
> - if (ctx->is_encoder)
> - hantro_set_fmt_cap(ctx, fmt);
> + hantro_set_fmt_cap(ctx, &fmt);
> else
> - hantro_set_fmt_out(ctx, fmt);
> + hantro_set_fmt_out(ctx, &fmt);
> }
>
> static void
> hantro_reset_raw_fmt(struct hantro_ctx *ctx)
> {
> const struct hantro_fmt *raw_vpu_fmt;
> - struct v4l2_pix_format_mplane *raw_fmt, *encoded_fmt;
> + struct v4l2_pix_format_mplane raw_fmt, *encoded_fmt;
>
> raw_vpu_fmt = hantro_get_default_fmt(ctx, false);
> + if (!raw_vpu_fmt)
> + return;
>
> - if (ctx->is_encoder) {
> - raw_fmt = &ctx->src_fmt;
> + if (ctx->is_encoder)
> encoded_fmt = &ctx->dst_fmt;
> - } else {
> - raw_fmt = &ctx->dst_fmt;
> + else
> encoded_fmt = &ctx->src_fmt;
> - }
>
> - hantro_reset_fmt(raw_fmt, raw_vpu_fmt);
> - raw_fmt->width = encoded_fmt->width;
> - raw_fmt->height = encoded_fmt->height;
> + hantro_reset_fmt(&raw_fmt, raw_vpu_fmt);
> + raw_fmt.width = encoded_fmt->width;
> + raw_fmt.height = encoded_fmt->height;
> if (ctx->is_encoder)
> - hantro_set_fmt_out(ctx, raw_fmt);
> + hantro_set_fmt_out(ctx, &raw_fmt);
> else
> - hantro_set_fmt_cap(ctx, raw_fmt);
> + hantro_set_fmt_cap(ctx, &raw_fmt);
> }
>
> void hantro_reset_fmts(struct hantro_ctx *ctx)
> --
> 2.34.1
>
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v8 5/6] media: verisilicon: HEVC: Only propose 10 bits compatible pixels formats
2023-02-03 9:16 ` [PATCH v8 5/6] media: verisilicon: HEVC: Only propose 10 bits compatible pixels formats Benjamin Gaignard
@ 2023-02-18 12:57 ` Ezequiel Garcia
0 siblings, 0 replies; 10+ messages in thread
From: Ezequiel Garcia @ 2023-02-18 12:57 UTC (permalink / raw)
To: Benjamin Gaignard
Cc: p.zabel, mchehab, shawnguo, s.hauer, kernel, festevam, linux-imx,
hverkuil-cisco, nicolas.dufresne, linux-media, linux-rockchip,
linux-kernel, linux-arm-kernel, kernel, Nicolas Dufresne
Hi Benjamin,
On Fri, Feb 3 2023 at 10:16:21 AM +0100, Benjamin Gaignard
<benjamin.gaignard@collabora.com> wrote:
> When decoding a 10bits bitstreams HEVC driver should only expose
> 10bits pixel formats.
> To fulfill this requirement it is needed to call
> hantro_reset_raw_fmt()
> when bit depth change and to correctly set match_depth in pixel
> formats
> enumeration.
>
> Fixes: dc39473d0340 ("media: hantro: imx8m: Enable 10bit decoding")
>
> Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
> Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
> ---
> drivers/media/platform/verisilicon/hantro_drv.c | 11 +++++++++--
> drivers/media/platform/verisilicon/imx8m_vpu_hw.c | 2 ++
> 2 files changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/media/platform/verisilicon/hantro_drv.c
> b/drivers/media/platform/verisilicon/hantro_drv.c
> index 6d8bc55ea627..fa31b200b097 100644
> --- a/drivers/media/platform/verisilicon/hantro_drv.c
> +++ b/drivers/media/platform/verisilicon/hantro_drv.c
> @@ -326,8 +326,15 @@ static int hantro_hevc_s_ctrl(struct v4l2_ctrl
> *ctrl)
>
> switch (ctrl->id) {
> case V4L2_CID_STATELESS_HEVC_SPS:
> - ctx->bit_depth = ctrl->p_new.p_hevc_sps->bit_depth_luma_minus8 + 8;
> - break;
> + {
Should be:
case V4L2_CID_STATELESS_HEVC_SPS: {
...
Same for VP9.
Thanks,
Ezequiel
> + const struct v4l2_ctrl_hevc_sps *sps = ctrl->p_new.p_hevc_sps;
> + int bit_depth = sps->bit_depth_luma_minus8 + 8;
> +
> + if (ctx->bit_depth == bit_depth)
> + return 0;
> +
> + return hantro_reset_raw_fmt(ctx, bit_depth);
> + }
> default:
> return -EINVAL;
> }
> diff --git a/drivers/media/platform/verisilicon/imx8m_vpu_hw.c
> b/drivers/media/platform/verisilicon/imx8m_vpu_hw.c
> index b390228fd3b4..f850d8bddef6 100644
> --- a/drivers/media/platform/verisilicon/imx8m_vpu_hw.c
> +++ b/drivers/media/platform/verisilicon/imx8m_vpu_hw.c
> @@ -152,6 +152,7 @@ static const struct hantro_fmt
> imx8m_vpu_g2_postproc_fmts[] = {
> {
> .fourcc = V4L2_PIX_FMT_NV12,
> .codec_mode = HANTRO_MODE_NONE,
> + .match_depth = true,
> .postprocessed = true,
> .frmsize = {
> .min_width = FMT_MIN_WIDTH,
> @@ -165,6 +166,7 @@ static const struct hantro_fmt
> imx8m_vpu_g2_postproc_fmts[] = {
> {
> .fourcc = V4L2_PIX_FMT_P010,
> .codec_mode = HANTRO_MODE_NONE,
> + .match_depth = true,
> .postprocessed = true,
> .frmsize = {
> .min_width = FMT_MIN_WIDTH,
> --
> 2.34.1
>
^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2023-02-18 12:58 UTC | newest]
Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-02-03 9:16 [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling Benjamin Gaignard
2023-02-03 9:16 ` [PATCH v8 1/6] media: verisilicon: Do not set context src/dst formats in reset functions Benjamin Gaignard
2023-02-03 9:16 ` [PATCH v8 2/6] media: verisilicon: Do not use ctx fields as format storage when resetting Benjamin Gaignard
2023-02-18 12:56 ` Ezequiel Garcia
2023-02-03 9:16 ` [PATCH v8 3/6] media: verisilicon: Do not set ctx->bit_depth in hantro_try_ctrl() Benjamin Gaignard
2023-02-03 9:16 ` [PATCH v8 4/6] media: verisilicon: Do not change context bit depth before validating the format Benjamin Gaignard
2023-02-03 9:16 ` [PATCH v8 5/6] media: verisilicon: HEVC: Only propose 10 bits compatible pixels formats Benjamin Gaignard
2023-02-18 12:57 ` Ezequiel Garcia
2023-02-03 9:16 ` [PATCH v8 6/6] media: verisilicon: VP9: " Benjamin Gaignard
2023-02-18 12:54 ` [PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling Ezequiel Garcia
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®