[v2] media: venus: flush all buffers in output plane streamoff

[PATCH v2] media: venus: flush all buffers in output plane streamoff

Posted by Dikshita Agarwal 1 year, 11 months ago

For scenarios, when source change is followed by VIDIOC_STREAMOFF
on output plane, driver should discard any queued OUTPUT
buffers, which are not decoded or dequeued.
Flush with HFI_FLUSH_INPUT does not have any actual impact.
So, fix it, by invoking HFI_FLUSH_ALL, which will flush all
queued buffers.

Fixes: 85872f861d4c ("media: venus: Mark last capture buffer")
Signed-off-by: Dikshita Agarwal <quic_dikshita@quicinc.com>
---
Changes since v1:
- Added fixes tag (Bryan)

 drivers/media/platform/qcom/venus/vdec.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/media/platform/qcom/venus/vdec.c b/drivers/media/platform/qcom/venus/vdec.c
index 29130a9..0d2ab95 100644
--- a/drivers/media/platform/qcom/venus/vdec.c
+++ b/drivers/media/platform/qcom/venus/vdec.c
@@ -1255,7 +1255,7 @@ static int vdec_stop_output(struct venus_inst *inst)
 		break;
 	case VENUS_DEC_STATE_INIT:
 	case VENUS_DEC_STATE_CAPTURE_SETUP:
-		ret = hfi_session_flush(inst, HFI_FLUSH_INPUT, true);
+		ret = hfi_session_flush(inst, HFI_FLUSH_ALL, true);
 		break;
 	default:
 		break;
-- 
2.7.4

Re: [PATCH v2] media: venus: flush all buffers in output plane streamoff

Posted by Krzysztof Kozlowski 1 year, 7 months ago

On 10/01/2024 07:12, Dikshita Agarwal wrote:
> For scenarios, when source change is followed by VIDIOC_STREAMOFF
> on output plane, driver should discard any queued OUTPUT
> buffers, which are not decoded or dequeued.
> Flush with HFI_FLUSH_INPUT does not have any actual impact.
> So, fix it, by invoking HFI_FLUSH_ALL, which will flush all
> queued buffers.
> 
> Fixes: 85872f861d4c ("media: venus: Mark last capture buffer")
> Signed-off-by: Dikshita Agarwal <quic_dikshita@quicinc.com>

This is a friendly reminder during the review process.

It looks like you received a tag and forgot to add it.

If you do not know the process, here is a short explanation:
Please add Acked-by/Reviewed-by/Tested-by tags when posting new
versions, under or above your Signed-off-by tag. Tag is "received", when
provided in a message replied to you on the mailing list. Tools like b4
can help here. However, there's no need to repost patches *only* to add
the tags. The upstream maintainer will do that for tags received on the
version they apply.

https://elixir.bootlin.com/linux/v6.5-rc3/source/Documentation/process/submitting-patches.rst#L577

If a tag was not added on purpose, please state why and what changed.

Best regards,
Krzysztof

Re: [PATCH v2] media: venus: flush all buffers in output plane streamoff

Posted by Stanimir Varbanov 1 year, 8 months ago

Hello Dikshita,

On 10.01.24 г. 8:12 ч., Dikshita Agarwal wrote:
> For scenarios, when source change is followed by VIDIOC_STREAMOFF
> on output plane, driver should discard any queued OUTPUT
> buffers, which are not decoded or dequeued.
> Flush with HFI_FLUSH_INPUT does not have any actual impact.
> So, fix it, by invoking HFI_FLUSH_ALL, which will flush all
> queued buffers.
> 
> Fixes: 85872f861d4c ("media: venus: Mark last capture buffer")

Cc: stable is missing with a kernel version.

> Signed-off-by: Dikshita Agarwal <quic_dikshita@quicinc.com>
> ---
> Changes since v1:
> - Added fixes tag (Bryan)
> 
>   drivers/media/platform/qcom/venus/vdec.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/media/platform/qcom/venus/vdec.c b/drivers/media/platform/qcom/venus/vdec.c
> index 29130a9..0d2ab95 100644
> --- a/drivers/media/platform/qcom/venus/vdec.c
> +++ b/drivers/media/platform/qcom/venus/vdec.c
> @@ -1255,7 +1255,7 @@ static int vdec_stop_output(struct venus_inst *inst)
>   		break;
>   	case VENUS_DEC_STATE_INIT:
>   	case VENUS_DEC_STATE_CAPTURE_SETUP:
> -		ret = hfi_session_flush(inst, HFI_FLUSH_INPUT, true);
> +		ret = hfi_session_flush(inst, HFI_FLUSH_ALL, true);

Is it tested on msm8916? If I remember correctly FLUSH_ALL produces 
session error, could you confirm, please.

-- 
regards,
Stan

Re: [PATCH v2] media: venus: flush all buffers in output plane streamoff

Posted by Dikshita Agarwal 1 year, 11 months ago

Hi Bryan,

I realized later that I missed to add your reviewed-by tag to this v2
patch, Sorry for this silly mistake, will be more careful in future.

On 1/10/2024 11:42 AM, Dikshita Agarwal wrote:
> For scenarios, when source change is followed by VIDIOC_STREAMOFF
> on output plane, driver should discard any queued OUTPUT
> buffers, which are not decoded or dequeued.
> Flush with HFI_FLUSH_INPUT does not have any actual impact.
> So, fix it, by invoking HFI_FLUSH_ALL, which will flush all
> queued buffers.
> 
> Fixes: 85872f861d4c ("media: venus: Mark last capture buffer")
> Signed-off-by: Dikshita Agarwal <quic_dikshita@quicinc.com>
Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
> ---
> Changes since v1:
> - Added fixes tag (Bryan)
> 
>  drivers/media/platform/qcom/venus/vdec.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/media/platform/qcom/venus/vdec.c b/drivers/media/platform/qcom/venus/vdec.c
> index 29130a9..0d2ab95 100644
> --- a/drivers/media/platform/qcom/venus/vdec.c
> +++ b/drivers/media/platform/qcom/venus/vdec.c
> @@ -1255,7 +1255,7 @@ static int vdec_stop_output(struct venus_inst *inst)
>  		break;
>  	case VENUS_DEC_STATE_INIT:
>  	case VENUS_DEC_STATE_CAPTURE_SETUP:
> -		ret = hfi_session_flush(inst, HFI_FLUSH_INPUT, true);
> +		ret = hfi_session_flush(inst, HFI_FLUSH_ALL, true);
>  		break;
>  	default:
>  		break;

Re: [PATCH v2] media: venus: flush all buffers in output plane streamoff

Posted by Nathan Hebert 1 year, 11 months ago

On Tue, Jan 9, 2024 at 10:13 PM Dikshita Agarwal
<quic_dikshita@quicinc.com> wrote:
>
> For scenarios, when source change is followed by VIDIOC_STREAMOFF
> on output plane, driver should discard any queued OUTPUT
> buffers, which are not decoded or dequeued.
> Flush with HFI_FLUSH_INPUT does not have any actual impact.
> So, fix it, by invoking HFI_FLUSH_ALL, which will flush all
> queued buffers.
>
> Fixes: 85872f861d4c ("media: venus: Mark last capture buffer")
> Signed-off-by: Dikshita Agarwal <quic_dikshita@quicinc.com>
Tested on a SC7280 Chromebook with a custom client [0] to perform the
queueing and VIDIOC_STREAMOFF sequence as described in the commit
message. Before this patch, a buffer from before the VIDIOC_STREAMOFF
command would be dequeued and seen by the client. With this patch, it
is not seen by the client.

[0]: https://crrev.com/c/5191249

Tested-by: Nathan Hebert <nhebert@chromium.org>
>
> ---
> Changes since v1:
> - Added fixes tag (Bryan)
>
>  drivers/media/platform/qcom/venus/vdec.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/media/platform/qcom/venus/vdec.c b/drivers/media/platform/qcom/venus/vdec.c
> index 29130a9..0d2ab95 100644
> --- a/drivers/media/platform/qcom/venus/vdec.c
> +++ b/drivers/media/platform/qcom/venus/vdec.c
> @@ -1255,7 +1255,7 @@ static int vdec_stop_output(struct venus_inst *inst)
>                 break;
>         case VENUS_DEC_STATE_INIT:
>         case VENUS_DEC_STATE_CAPTURE_SETUP:
> -               ret = hfi_session_flush(inst, HFI_FLUSH_INPUT, true);
> +               ret = hfi_session_flush(inst, HFI_FLUSH_ALL, true);
>                 break;
>         default:
>                 break;
> --
> 2.7.4
>
>