Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [RFC][PATCH 13/15] videobuf2: do not sync buffers for DMABUF
 queues
To:     Sergey Senozhatsky <senozhatsky@chromium.org>
Cc:     Hans Verkuil <hans.verkuil@cisco.com>,
        Tomasz Figa <tfiga@chromium.org>,
        Mauro Carvalho Chehab <mchehab@kernel.org>,
        Kyungmin Park <kyungmin.park@samsung.com>,
        Marek Szyprowski <m.szyprowski@samsung.com>,
        Sakari Ailus <sakari.ailus@iki.fi>,
        Laurent Pinchart <laurent.pinchart@ideasonboard.com>,
        Pawel Osciak <posciak@chromium.org>,
        linux-media@vger.kernel.org, linux-kernel@vger.kernel.org
References: <20191217032034.54897-1-senozhatsky@chromium.org>
 <20191217032034.54897-14-senozhatsky@chromium.org>
 <2d0e1a9b-6c5e-ff70-9862-32c8b8aaf65f@xs4all.nl>
 <20200122050515.GB49953@google.com>
From:   Hans Verkuil <hverkuil@xs4all.nl>
Message-ID: <57f711a0-6183-74f6-ab24-ebe414cb6881@xs4all.nl>
Date:   Thu, 23 Jan 2020 12:35:33 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.9.0
MIME-Version: 1.0
In-Reply-To: <20200122050515.GB49953@google.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 1/22/20 6:05 AM, Sergey Senozhatsky wrote:
> On (20/01/10 11:30), Hans Verkuil wrote:
> [..]
>>> diff --git a/drivers/media/common/videobuf2/videobuf2-v4l2.c b/drivers/media/common/videobuf2/videobuf2-v4l2.c
>>> index 1762849288ae..2b9d3318e6fb 100644
>>> --- a/drivers/media/common/videobuf2/videobuf2-v4l2.c
>>> +++ b/drivers/media/common/videobuf2/videobuf2-v4l2.c
>>> @@ -341,8 +341,22 @@ static void set_buffer_cache_hints(struct vb2_queue *q,
>>>  				   struct vb2_buffer *vb,
>>>  				   struct v4l2_buffer *b)
>>>  {
>>> -	vb->need_cache_sync_on_prepare = 1;
>>> +	/*
>>> +	 * DMA exporter should take care of cache syncs, so we can avoid
>>> +	 * explicit ->prepare()/->finish() syncs.
>>> +	 */
>>> +	if (q->memory == VB2_MEMORY_DMABUF) {
>>> +		vb->need_cache_sync_on_finish = 0;
>>> +		vb->need_cache_sync_on_prepare = 0;
>>> +		return;
>>> +	}
>>>  
>>> +	/*
>>> +	 * For other ->memory types we always need ->prepare() cache
>>> +	 * sync. ->finish() cache sync, however, can be avoided when queue
>>> +	 * direction is TO_DEVICE.
>>> +	 */
>>> +	vb->need_cache_sync_on_prepare = 1;
>>
>> I'm trying to remember: what needs to be done in prepare()
>> for a capture buffer? I thought that for capture you only
>> needed to invalidate the cache in finish(), but nothing needs
>> to be done in the prepare().
> 
> Hmm. Not sure. A precaution in case if user-space wrote to that buffer?

But whatever was written in the buffer is going to be overwritten anyway.

Unless I am mistaken the current situation is that the cache syncs are done
in both prepare and finish, regardless of the DMA direction.

I would keep that behavior to avoid introducing any unexpected regressions.

Then, if q->allow_cache_hint is set, then default to a cache sync (cache clean)
in the prepare for OUTPUT buffers and a cache sync (cache invalidate) in the
finish for CAPTURE buffers.

This also means that any drivers that want to access a buffer in between the
prepare...finish calls will need to do a begin/end_cpu_access. But that's a
separate matter.

Regards,

	Hans

> 
> +	if (q->dma_dir == DMA_FROM_DEVICE)
> +		q->need_cache_sync_on_prepare = 0;
> 
> ?
> 
> 	-ss
>