Message-ID: <5f5c1b41-0c03-c906-0756-b5338a0be88c@microchip.com>
Date:   Wed, 22 Jun 2022 14:09:48 +0000
From:   <Eugen.Hristev@...rochip.com>
To:     <hverkuil@...all.nl>, <linux-arm-kernel@...ts.infradead.org>,
        <linux-media@...r.kernel.org>
CC:     <devicetree@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <Claudiu.Beznea@...rochip.com>, <Nicolas.Ferre@...rochip.com>,
        <jacopo@...ndi.org>
Subject: Re: [PATCH v10 0/5] media: atmel: atmel-isc: implement media
 controller

On 6/22/22 4:47 PM, Hans Verkuil wrote:
> On 22/06/2022 14:42, Eugen.Hristev@...rochip.com wrote:
>> On 6/22/22 3:25 PM, Eugen Hristev - M18282 wrote:
>>> On 6/22/22 2:53 PM, Hans Verkuil wrote:
>>>> Hi Eugen,
>>>>
>>>> On 03/05/2022 11:51, Eugen Hristev wrote:
>>>>> This series is a split from the series :
>>>>> [PATCH v9 00/13] media: atmel: atmel-isc: implement media controller
>>>>> and it includes the media controller part.
>>>>> previous fixes were sent on a different patch series.
>>>>>
>>>>> As discussed on the ML, moving forward with having the media link validate at
>>>>> start/stop streaming call.
>>>>> I will test the patch :
>>>>> [RFC PATCHv2] vb2: add support for (un)prepare_streaming queue ops
>>>>> afterwards, but that patch requires moving my logic to the new vb2 callbacks.
>>>>
>>>> I'm looking at merging this series, but I would like to have the output of
>>>> 'v4l2-compliance -m /dev/mediaX' to verify that the MC links etc. is all
>>>> correct.
>>>
>>> Hello Hans,
>>>
>>> Please have a look at attached file . Unless you want me to add the
>>> whole output to the e-mail ?
>>>
>>> I also added output of media-ctl -p for your convenience.
>>> the subdev2 is a device and driver that is not upstream and has some
>>> compliance issues, they are reported by the v4l2-compliance tool, but
>>> they should not affect this series, it's a synopsys driver that was
>>> rejected on mainline a few years ago, I took it for internal usage, but
>>> it's not cleaned up nor worked a lot upon.
>>>
>>>>
>>>> And one more question which may have been answered already in the past:
>>>>
>>>> Changing to the MC will break existing applications, doesn't it? Or did I
>>>> miss something?
>>>>
>>>
>>> The existing applications will have to configure the pipeline now. It
>>> will no longer work by configuring just the top video node /dev/video0 .
>>> They would have to use media-ctl for it, something similar with this set
>>> of commands:
>>
>> To add on top of that, actually, the reality is that without the MC
>> support in atmel-isc , some of our platforms do not work at all, because
>> the csi2dc driver which is in the middle of the pipeline, is a MC
>> driver. So it will not work without configuring it with MC anyway. It
>> used to work in a very preliminary version of the csi2dc driver which I
>> sent a few years ago, but that way of handling things was rejected.
>> Hence I changed the csi2dc to being full-MC driver (requested for new
>> drivers) and now I am completing the conversion for the whole pipeline.
>> We are using this MC-centric approach in production for our products to
>> be as close as possible to mainline, and backported it to our 5.15
>> internal releases, which people are using right now.
> 
> I'm not all that keen on breaking userspace for those who do NOT use the
> Atmel BSP. Basically some platforms are currently broken, and with this patch
> series some other platforms are broken, but at least can be fixed by changing
> userspace.
> 
> How feasible is it to do something similar that TI did for the cal driver?
> (drivers/media/platform/ti/cal)
> 
> I.e., based on a module option the MC is enabled or disabled. And if a
> csi2dc is present, then the MC API is always enabled.

Some platforms that are now broken, never worked, because they would 
need all the modules to be configured with MC anyway.
The other platforms, that work now by just configuring the top video 
node, work, but in a quite limited way, as sensor drivers can be 
MC-capable and configurable from userspace, but the top video driver 
will overwrite their configuration based on it's own decision algorithm.
With the MC approach, all the platforms work, the drawback is, as you 
said, that the userspace has to configure the pipeline from head to toe; 
but , we knew that: moving to a MC approach, makes the old way of 
configuring the image capture simply 'not enough'.

It would be difficult to maintain a driver that would use the MC API and 
handle the things for itself and just check the pipeline; and in the 
same time if a module parameter is different, pass configuration down 
the pipeline and have an algorithm implemented that would interact with 
the subdev, ask for it's capabilities, and then decide on its own what 
the subdev would use. The driver would be a bit big and it would have a 
lot of code. That is one of the advantages of these patches, to simplify 
the driver.

I would prefer to not have to keep that code, and move to MC approach, 
but in the end you have arguments and you are in charge.

Eugen


> 
> Regards,
> 
>          Hans
> 
>>
>>>
>>> media-ctl -d /dev/media0 --set-v4l2 '"imx219
>>> 1-0010":0[fmt:SRGGB10_1X10/1920x1080]'
>>> media-ctl -d /dev/media0 --set-v4l2
>>> '"dw-csi.0":0[fmt:SRGGB10_1X10/1920x1080]'
>>> media-ctl -d /dev/media0 --set-v4l2 '"csi2dc":0[fmt:SRGGB10_1X10/1920x1080]'
>>> media-ctl -d /dev/media0 --set-v4l2
>>> '"atmel_isc_scaler":0[fmt:SRGGB10_1X10/1920x1080]'
>>>
>>> Thank you for taking care of this !
>>>
>>> Eugen
>>>
>>>> Regards,
>>>>
>>>>            Hans
>>>>
>>>>>
>>>>> Full series history:
>>>>>
>>>>> Changes in v10:
>>>>> -> split the series into this first fixes part.
>>>>> -> moved IO_MC addition from first patch to the second patch on the driver changes
>>>>> -> edited commit messages
>>>>> -> DT nodes now disabled by default.
>>>>>
>>>>> Changes in v9:
>>>>> -> kernel robot reported isc_link_validate is not static, changed to static.
>>>>>
>>>>> Changes in v8:
>>>>> -> scaler: modified crop bounds to have the exact source size
>>>>>
>>>>> Changes in v7:
>>>>> -> scaler: modified crop bounds to have maximum isc size
>>>>> -> format propagation: did small changes as per Jacopo review
>>>>>
>>>>>
>>>>> Changes in v6:
>>>>> -> worked a bit on scaler, added try crop and other changes as per Jacopo review
>>>>> -> worked on isc-base enum_fmt , reworked as per Jacopo review
>>>>>
>>>>> Changes in v5:
>>>>> -> removed patch that removed the 'stop' variable as it was still required
>>>>> -> added two new trivial patches
>>>>> -> reworked some parts of the scaler and format propagation after discussions with Jacopo
>>>>>
>>>>>
>>>>> Changes in v4:
>>>>> -> as reviewed by Hans, added new patch to remove the 'stop' variable and reworked
>>>>> one patch that was using it
>>>>> -> as reviewed by Jacopo, reworked some parts of the media controller implementation
>>>>>
>>>>>
>>>>> Changes in v3:
>>>>> - change in bindings, small fixes in csi2dc driver and conversion to mc
>>>>> for the isc-base.
>>>>> - removed some MAINTAINERS patches and used patterns in MAINTAINERS
>>>>>
>>>>> Changes in v2:
>>>>> - integrated many changes suggested by Jacopo in the review of the v1 series.
>>>>> - add a few new patches
>>>>>
>>>>> Eugen Hristev (5):
>>>>>      media: atmel: atmel-isc: prepare for media controller support
>>>>>      media: atmel: atmel-isc: implement media controller
>>>>>      ARM: dts: at91: sama7g5: add nodes for video capture
>>>>>      ARM: configs: at91: sama7: add xisc and csi2dc
>>>>>      ARM: multi_v7_defconfig: add atmel video pipeline modules
>>>>>
>>>>>     arch/arm/boot/dts/sama7g5.dtsi                |  51 ++
>>>>>     arch/arm/configs/multi_v7_defconfig           |   3 +
>>>>>     arch/arm/configs/sama7_defconfig              |   2 +
>>>>>     drivers/media/platform/atmel/Makefile         |   2 +-
>>>>>     drivers/media/platform/atmel/atmel-isc-base.c | 485 +++++++++---------
>>>>>     .../media/platform/atmel/atmel-isc-scaler.c   | 267 ++++++++++
>>>>>     drivers/media/platform/atmel/atmel-isc.h      |  50 +-
>>>>>     .../media/platform/atmel/atmel-sama5d2-isc.c  |  34 +-
>>>>>     .../media/platform/atmel/atmel-sama7g5-isc.c  |  32 +-
>>>>>     9 files changed, 685 insertions(+), 241 deletions(-)
>>>>>     create mode 100644 drivers/media/platform/atmel/atmel-isc-scaler.c
>>>>>
>>>>
>>>
>>
> 

Powered by blists - more mailing lists