linux-kernel - Re: [PATCH 10/28] venus: vdec: call session

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 04 May 2018 16:39:59 +0530
From:   Vikash Garodia <vgarodia@...eaurora.org>
To:     Stanimir Varbanov <stanimir.varbanov@...aro.org>
Cc:     Mauro Carvalho Chehab <mchehab@...nel.org>,
        Hans Verkuil <hverkuil@...all.nl>, linux-media@...r.kernel.org,
        linux-kernel@...r.kernel.org, linux-arm-msm@...r.kernel.org,
        linux-media-owner@...r.kernel.org
Subject: Re: [PATCH 10/28] venus: vdec: call session_continue in insufficient
 event

Hi Stanimir,

On 2018-05-03 17:06, Stanimir Varbanov wrote:
> Hi Vikash,
> 
> Thanks for the comments!
> 
> On  2.05.2018 09:26, Vikash Garodia wrote:
>> Hello Stanimir,
>> 
>> On 2018-04-24 18:14, Stanimir Varbanov wrote:
>>> Call session_continue for Venus 4xx version even when the event
>>> says that the buffer resources are not sufficient. Leaving a
>>> comment with more information about the workaround.
>>> 
>>> Signed-off-by: Stanimir Varbanov <stanimir.varbanov@...aro.org>
>>> ---
>>>  drivers/media/platform/qcom/venus/vdec.c | 8 ++++++++
>>>  1 file changed, 8 insertions(+)
>>> 
>>> diff --git a/drivers/media/platform/qcom/venus/vdec.c
>>> b/drivers/media/platform/qcom/venus/vdec.c
>>> index c45452634e7e..91c7384ff9c8 100644
>>> --- a/drivers/media/platform/qcom/venus/vdec.c
>>> +++ b/drivers/media/platform/qcom/venus/vdec.c
>>> @@ -873,6 +873,14 @@ static void vdec_event_notify(struct venus_inst
>>> *inst, u32 event,
>>> 
>>>              dev_dbg(dev, "event not sufficient resources (%ux%u)\n",
>>>                  data->width, data->height);
>>> +            /*
>>> +             * Workaround: Even that the firmware send and event for
>>> +             * insufficient buffer resources it is safe to call
>>> +             * session_continue because actually the event says that
>>> +             * the number of capture buffers is lower.
>>> +             */
>>> +            if (IS_V4(core))
>>> +                hfi_session_continue(inst);
>>>              break;
>>>          case HFI_EVENT_RELEASE_BUFFER_REFERENCE:
>>>              venus_helper_release_buf_ref(inst, data->tag);
>> 
>> Insufficient event from video firmware could be sent either,
>> 1. due to insufficient buffer resources
>> 2. due to lower capture buffers
>> 
>> It cannot be assumed that the event received by the host is due to 
>> lower capture
>> buffers. Incase the buffer resource is insufficient, let say there is 
>> a bitstream
>> resolution switch from 720p to 1080p, capture buffers needs to be 
>> reallocated.
> 
> I agree with you. I will rework this part and call session_continue
> only for case #2.

Even if the capture buffers are lower, driver should consider 
reallocation of capture
buffers with required higher count. Without this, it may happen that for 
a given video
frame, the decoded output will not be generated. The fact that the DPB 
buffer count is
same as capture buffers, will be lower than required. Hence the frame 
which needs YUV
reference beyond the DPB count, will get stuck as it cannot be decoded 
due to unavailability
of sufficient DPB buffers.
Say for ex. 10 DPB and capture buffers are allocated. For a given 
bitstream, firmware requested
the count to be 15. Frame 1 to 10 gets decoded and stored in DPB as 
references for future frame
decoding. Now when the 11th frame is queued to firmware, it can be 
decode but cannot be stored
as reference to decode future (12th) frame. Hence 11 frame will get 
stuck and will not be given
back to host driver.

>> 
>> The driver should be sending the V4L2_EVENT_SOURCE_CHANGE to client 
>> instead of ignoring
>> the event from firmware.
> 
> The v4l2 event is sent always to v4l clients.
> 
> regards,
> Stan