Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [RFC PATCH v11 6/9] media: tegra: Add Tegra210 Video input driver
To:     Sowjanya Komatineni <skomatineni@nvidia.com>,
        thierry.reding@gmail.com, jonathanh@nvidia.com, frankc@nvidia.com,
        hverkuil@xs4all.nl, sakari.ailus@iki.fi, helen.koike@collabora.com
Cc:     sboyd@kernel.org, linux-media@vger.kernel.org,
        devicetree@vger.kernel.org, linux-clk@vger.kernel.org,
        linux-tegra@vger.kernel.org, linux-kernel@vger.kernel.org
References: <1588197606-32124-1-git-send-email-skomatineni@nvidia.com>
 <1588197606-32124-7-git-send-email-skomatineni@nvidia.com>
 <bacc4308-4b95-f566-b80e-096ff96407b5@gmail.com>
 <4da289e6-036f-853b-beb4-379d6462adb0@gmail.com>
 <c6d54885-6f23-f60c-a17b-3481fc4d6adf@gmail.com>
 <b14b9dc5-7ac9-7735-d98d-eebc7e151cba@nvidia.com>
 <7d31d24f-f353-7e82-3ff9-cdba8b773d1e@nvidia.com>
 <06a4a067-8d54-4322-b2a6-14e82eaeda29@nvidia.com>
 <47873bbd-cf90-4595-5a99-7e9113327ecc@nvidia.com>
 <f6088e0f-4ac7-a6be-3ede-0233dc88ef2c@nvidia.com>
From:   Dmitry Osipenko <digetx@gmail.com>
Message-ID: <71532440-f455-cc24-74f7-9ccad5947099@gmail.com>
Date:   Thu, 30 Apr 2020 22:33:44 +0300
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.7.0
MIME-Version: 1.0
In-Reply-To: <f6088e0f-4ac7-a6be-3ede-0233dc88ef2c@nvidia.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

30.04.2020 22:09, Sowjanya Komatineni пишет:
> 
> On 4/30/20 11:18 AM, Sowjanya Komatineni wrote:
>>
>> On 4/30/20 10:06 AM, Sowjanya Komatineni wrote:
>>>
>>> On 4/30/20 9:29 AM, Sowjanya Komatineni wrote:
>>>>
>>>> On 4/30/20 9:04 AM, Sowjanya Komatineni wrote:
>>>>>
>>>>> On 4/30/20 7:13 AM, Dmitry Osipenko wrote:
>>>>>> 30.04.2020 17:02, Dmitry Osipenko пишет:
>>>>>>> 30.04.2020 16:56, Dmitry Osipenko пишет:
>>>>>>>> 30.04.2020 01:00, Sowjanya Komatineni пишет:
>>>>>>>>> +static int chan_capture_kthread_finish(void *data)
>>>>>>>>> +{
>>>>>>>>> +    struct tegra_vi_channel *chan = data;
>>>>>>>>> +    struct tegra_channel_buffer *buf;
>>>>>>>>> +
>>>>>>>>> +    set_freezable();
>>>>>>>>> +
>>>>>>>>> +    while (1) {
>>>>>>>>> +        try_to_freeze();
>>>>>>>> I guess it won't be great to freeze in the middle of a capture
>>>>>>>> process, so:
>>>>>>>>         if (list_empty(&chan->done))
>>>>>>>>             try_to_freeze();
>>>>>>> And here should be some locking protection in order not race with
>>>>>>> the
>>>>>>> chan_capture_kthread_start because kthread_finish could freeze
>>>>>>> before
>>>>>>> kthread_start.
>>>>>> Or maybe both start / finish threads should simply be allowed to
>>>>>> freeze
>>>>>> only when both capture and done lists are empty.
>>>>>>
>>>>>> if (list_empty(&chan->capture) &&
>>>>>>      list_empty(&chan->done))
>>>>>>     try_to_freeze();
>>>>>
>>>>> good to freeze when not in middle of the frame capture but why
>>>>> should we not allow freeze in between captures?
>>>>>
>>>>> Other drivers do allow freeze in between frame captures.
>>>>>
>>>>> I guess we can freeze before dequeue for capture and in finish
>>>>> thread we can freeze after capture done. This also don't need to
>>>>> check for list_empty with freeze to allow between frame captures.
>>>>>
>>>> Also if we add check for both lists empty, freeze is not allowed as
>>>> long as streaming is going on and in case of continuous streaming
>>>> freeze will never happen.
>>>
>> To allow freeze b/w frames (but not in middle of a frame),
>>
>> for capture_start thread, probably we can do unconditional
>> try_to_freeze()

Is it possible to use wait_event_freezable()?

https://www.kernel.org/doc/Documentation/power/freezing-of-tasks.txt

Will the wait_event_interruptible() be woken up when system freezes?

>> for capture_finish thread, at end of capture done we can do
>> try_to_freeze() only when done list is empty

This doesn't prevent situation where the done-list is empty and the
"finish" thread freezes, in the same time the "start" thread issues new
capture and then freezes too.

1. "start" thread issues capture

2. "finish" thread wakes and waits for the capture to complete

3. "start" thread begins another capture, waits for FRAME_START

4. system freezing activates

5. "finish" thread completes the capture and freezes because done-list
is empty

6. "start" thread gets FRAME_START, issues another capture and freezes

> My understanding is buffer updates/release should not happen after
> frozen state. So we should let frame capture of outstanding buffer to
> finish before freezing in capture_finish thread.
> 
> But for capture_start thread we can unconditionally freeze before
> dequeuing next buffer for capture.
> 
> With this when both threads are in frozen state and no buffer
> updates/captures will happen after frozen state.
> 
> I think its not required to finish streaming of all frames completely to
> let threads to enter frozen state as streaming can be continuous as well.

Yes, only freezing in the middle of IO should be avoided.

https://lwn.net/Articles/705269/

>>> Hi Dmitry,
>>>
>>> Will update in v12 to not allow freeze in middle of a frame capture.
>>>
>>> Can you please confirm on above if you agree to allow freeze to
>>> happen in b/w frame captures?
>>>
>>> Also as most feedback has been received from you by now, appreciate
>>> if you can provide all in this v11 if you have anything else so we
>>> will not have any new changes after v12.

I'll take another look tomorrow / during weekend and let you know.