linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Vladimir Zapolskiy <vz@mleia.com>
To: Laurent Pinchart <laurent.pinchart@ideasonboard.com>,
	Vladimir Zapolskiy <vladimir_zapolskiy@mentor.com>
Cc: kieran.bingham@ideasonboard.com, Lee Jones <lee.jones@linaro.org>,
	Linus Walleij <linus.walleij@linaro.org>,
	Rob Herring <robh+dt@kernel.org>,
	Marek Vasut <marek.vasut@gmail.com>,
	Wolfram Sang <wsa@the-dreams.de>,
	devicetree@vger.kernel.org, linux-gpio@vger.kernel.org,
	linux-media@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 4/7] mfd: ds90ux9xx: add TI DS90Ux9xx de-/serializer MFD driver
Date: Fri, 12 Oct 2018 16:59:27 +0300	[thread overview]
Message-ID: <735126c6-9780-a9ef-4fd0-699d6876ed37@mleia.com> (raw)
In-Reply-To: <3599186.afdMBtdL0k@avalon>

Hello Laurent.

On 10/12/2018 04:01 PM, Laurent Pinchart wrote:
> Hello Vladimir,
> 
> On Friday, 12 October 2018 14:47:52 EEST Kieran Bingham wrote:
>> On 12/10/18 11:58, Vladimir Zapolskiy wrote:
>>> On 10/12/2018 12:20 PM, Kieran Bingham wrote:
>>>> On 12/10/18 09:39, Lee Jones wrote:
>>>>> On Fri, 12 Oct 2018, Vladimir Zapolskiy wrote:
>>>>>> On 10/12/2018 09:03 AM, Lee Jones wrote:
>>>>>>> On Tue, 09 Oct 2018, Vladimir Zapolskiy wrote:
>>>>>>>> From: Vladimir Zapolskiy <vladimir_zapolskiy@mentor.com>
>>>>>>>>
>>>>>>>> The change adds I2C device driver for TI DS90Ux9xx de-/serializers,
>>>>>>>> support of subdevice controllers is done in separate drivers, because
>>>>>>>> not all IC functionality may be needed in particular situations, and
>>>>>>>> this can be fine grained controlled in device tree.
>>>>>>>>
>>>>>>>> The development of the driver was a collaborative work, the
>>>>>>>> contribution done by Balasubramani Vivekanandan includes:
>>>>>>>> * original implementation of the driver based on a reference driver,
>>>>>>>> * regmap powered interrupt controller support on serializers,
>>>>>>>> * support of implicitly or improperly specified in device tree ICs,
>>>>>>>> * support of device properties and attributes: backward compatible
>>>>>>>>
>>>>>>>>   mode, low frequency operation mode, spread spectrum clock
>>>>>>>>   generator.
>>>>>>>>
>>>>>>>> Contribution by Steve Longerbeam:
>>>>>>>> * added ds90ux9xx_read_indirect() function,
>>>>>>>> * moved number of links property and added ds90ux9xx_num_fpd_links(),
>>>>>>>> * moved and updated ds90ux9xx_get_link_status() function to core
>>>>>>>> driver,
>>>>>>>> * added fpd_link_show device attribute.
>>>>>>>>
>>>>>>>> Sandeep Jain added support of pixel clock edge configuration.
>>>>>>>>
>>>>>>>> Signed-off-by: Vladimir Zapolskiy <vladimir_zapolskiy@mentor.com>
>>>>>>>> ---
>>>>>>>>
>>>>>>>>  drivers/mfd/Kconfig           |  14 +
>>>>>>>>  drivers/mfd/Makefile          |   1 +
>>>>>>>>  drivers/mfd/ds90ux9xx-core.c  | 879 ++++++++++++++++++++++++++++++++
>>>>>>>>  include/linux/mfd/ds90ux9xx.h |  42 ++
>>>>>>>>  4 files changed, 936 insertions(+)
>>>>>>>>  create mode 100644 drivers/mfd/ds90ux9xx-core.c
>>>>>>>>  create mode 100644 include/linux/mfd/ds90ux9xx.h
>>>>>>>>
>>>>>>>> diff --git a/drivers/mfd/Kconfig b/drivers/mfd/Kconfig
>>>>>>>> index 8c5dfdce4326..a969fa123f64 100644
>>>>>>>> --- a/drivers/mfd/Kconfig
>>>>>>>> +++ b/drivers/mfd/Kconfig
>>>>>>>> @@ -1280,6 +1280,20 @@ config MFD_DM355EVM_MSP
>>>>>>>>
>>>>>>>>  	  boards.  MSP430 firmware manages resets and power sequencing,
>>>>>>>>  	  inputs from buttons and the IR remote, LEDs, an RTC, and more.
>>>>>>>>
>>>>>>>> +config MFD_DS90UX9XX
>>>>>>>> +	tristate "TI DS90Ux9xx FPD-Link de-/serializer driver"
>>>>>>>> +	depends on I2C && OF
>>>>>>>> +	select MFD_CORE
>>>>>>>> +	select REGMAP_I2C
>>>>>>>> +	help
>>>>>>>> +	  Say yes here to enable support for TI DS90UX9XX de-/serializer
>>>>>>>> ICs.
>>>>>>>> +
>>>>>>>> +	  This driver provides basic support for setting up the
>>>>>>>> de-/serializer
>>>>>>>> +	  chips. Additional functionalities like connection handling to
>>>>>>>> +	  remote de-/serializers, I2C bridging, pin multiplexing, GPIO
>>>>>>>> +	  controller and so on are provided by separate drivers and should
>>>>>>>> +	  enabled individually.
>>>>>>>
>>>>>>> This is not an MFD driver.
>>>>>>
>>>>>> Why do you think so? The representation of the ICs into device tree
>>>>>> format of hardware description shows that this is a truly MFD driver
>>>>>> with multiple IP subcomponents naturally mapped into MFD cells.
>>>>>
>>>>> This driver does too much real work ('stuff') to be an MFD driver.
>>>>> MFD drivers should not need to care of; links, gates, modes, pixels,
>>>>> frequencies maps or properties.  Nor should they contain elaborate
>>>>> sysfs structures to control the aforementioned 'stuff'.
>>>>>
>>>>> Granted, there may be some code in there which could be appropriate
>>>>> for an MFD driver.  However most of it needs moving out into a
>>>>> function driver (or two).
>>>>>
>>>>>> Basically it is possible to replace explicit of_platform_populate() by
>>>>>> adding a "simple-mfd" compatible, if it is desired.
>>>>>>
>>>>>>> After a 30 second Google of what this device actually does, perhaps
>>>>>>> drivers/media might be a better fit?
>>>>>>
>>>>>> I assume it would be quite unusual to add a driver with NO media
>>>>>> functions and controls into drivers/media.
>>>>>
>>>>> drivers/media may very well not be the correct place for this.  In my
>>>>> 30 second Google, I saw that this device has a lot to do with cameras,
>>>>> hence my media association.
>>>>>
>>>>> If *all* else fails, there is always drivers/misc, but this should be
>>>>> avoided if at all possible.
>>>>
>>>> The device as a whole is FPD Link for camera devices I believe, but it
>>>
>>> I still don't understand (I could be biased though) why there is such
>>> a strong emphasis on cameras and media stuff in the discussion.
>>>
>>> No, "the device as a whole is FPD Link for camera devices" is a wrong
>>> statement. On hand I have a number of boards with
>>> serializers/deserializers from the TI DS90Ux9xx IC series and sensors are
>>> NOT connected to them.
> 
> I understand what is not connected to them, but could you tell us what is 
> connected then ? :-)

You got it right, the most two common ways of using the ICs:

1) SoC -> serializer ("local") -> deserializer ("remote") -> panel,
2) sensor -> serializer ("remote") -> deserializer ("local") -> SoC.

The point is that the published drivers naturally support both data chains
and even more of them, e.g. transferring audio data only or just setting
GPIO line signals on a "remote" PCB.

>> Yes - My apologies, this is a good point.
>>
>> Especially as the clue is in the name "Flat Panel Display".
>> I have been stuck with my GMSL hat on for too long.
>>
>> Even GMSL is in the same boat. It's just that 'we' are using GMSL for
>> cameras, but it can be used equally for displays and data.
>>
>> These devices are serialiser-deserialiser pairs with power and control
>> paths.
> 
> And data paths, that are designed to transport video (and audio in this case). 
> The devices are thus media-focussed, although in a broader sense than drivers/
> media/

The devices are media-focused only in the sense that media functionality
casts a shadow onto inalienable GPIO or I2C bridging functionality.

Anyway MFD driver representation of the ICs allows to keep pinmuxing, V4L2,
DRM, audio bridging, interrupt controller and all other subcontroller
functions separately, configure them separately, store under separate
driver frameworks etc. That's a perfect flexibility on drivers level
as I can see it.

>> Essentially they are multi purpose buses - which do not yet have a home.
>> We have used media as a home because of our use case.
>>
>> The use case whether they transfer frames from a camera or to a display
>> are of course closely related, but ultimately covered by two separate
>> subsystems at the pixel level (DRM vs V4L, or other for other data)
>>
>> Perhaps as they are buses - on a level with USB or I2C (except they can
>> of course carry I2C or Serial as well as 'bi-directional video' etc ),
>> they are looking for their own subsystem.
>>
>> Except I don't think we don't want to add a new subsystem for just one
>> (or two) devices...
> 
> I'm not sure a new subsystem is needed. As you've noted there's an overlap 
> between drivers/media/ and drivers/gpu/drm/ in terms of supported hardware. We 
> even have a devices supported by two drivers, one in drivers/media/ and one in 
> drivers/gpu/drm/ (I'm thinking about the adv7511 in particular). This is a 
> well known issue, and so far nothing has been done in mainline to try and 
> solve it.

Right, presumably this IC series would have *cell* drivers in both drivers/media
and drivers/gpu/drm/ and other folders like drivers/pinctrl/ or sound/.

The interesting thing is that the published drivers do NOT require any new
cell drivers (at least non-trivial ones with >200 LoC) in drivers/media/
or drivers/gpu/drm/ to get the expected operation of DS90Ux925/926/927/928
ICs (any ser-des connection combination), and we have a DS90Ux940 cell driver
targeted to drivers/media/

> Trying to find another home in drivers/mfd/ to escape from the problem isn't a 
> good solution in my opinion. The best option from a Linux point of view would 
> be to unify V4L2 and DRM/KMS when it comes to bridge support, but that's a 
> long way down the road (I won't complain if you want to give it a go though 
> :-)). As your use cases are display, focused, I would propose to start with 
> drivers/gpu/drm/bridge/, and leave the problem of camera support for first 
> person who will have such a use case.

Well, what should I do with pinctrl or audio bridging device drivers?

Stick all drivers together into an unmaintainable clod of tangled code?

Let's better exploit the excellent opportunity for code modularity given
by the MFD framework.

>>>> certainly has different functions which are broken out in this
>>>> implementation.
>>>
>>> No, there is absolutely nothing broken out from the presented MFD drivers,
>>> the drivers are completely integral and basically I don't expect any.
>>>
>>> If you are concerned about media functionality, the correspondent MFD
>>> *cell* drivers will be added into drivers/media, drivers/gpu/drm or
>>> whatever is to be a proper place.
>>>
>>>> I think it might be quite awkward having the i2c components in
>>>> drivers/i2c and the media components in drivers/media/i2c, so what about
>>>> creating drivers/media/i2c/fpd-link (or such) as a container?
>>>
>>> I open drivers/media/i2c/Kconfig and all entries with no exception are
>>> under from 'if VIDEO_V4L2'. The MFD drivers do NOT require on depend on
>>> VIDEO_V4L2 or any other multimedia frameworks, nor the MFD drivers export
>>> any multimedia controls.
>>>
>>>> Our GMSL implementation is also a complex camera(s) device - but does
>>>> not yet use the MFD framework, perhaps that's something to add to my
>>>> todo list.
> 
> Nope, please don't :-) The GMSL (de)serializers are no MFD devices either.
> 
>>> Okay, but the TI DS90Ux9xx is NOT a camera device, and it is NOT a
>>> multimedia device, but it is a pure MFD device so the argument is not
>>> applicable.
> 
> For the reasons pointed out in the review of other patches, and also pointed 
> out by Lee in his review of this patch, I disagree with that. This isn't an 
> MFD device.

Eh, it is an MFD device. Just probably drivers/mfd is really not the best
place to store this particular MFD device driver...

>>>> We currently keep all of the complexity within the max9286.c driver, but
>>>> I could foresee that being split further if more devices add to the
>>>> complexity of managing the bus. At which point we might want an
>>>> equivalent drivers/media/i2c/gmsl/ perhaps?
>>>>
>>>>>> Laurent, can you please share your opinion?
> 
> Done :-)
> 
> I'd like to mention that I would prefer focusing on the DT bindings first, and 

Sure, thank you for your comments :)

> then move to the driver side. In that area I would like to have a full example 
> of a system using these chips, as the "initial support" is too limited for a 
> proper review. This won't come as a surprise, but I will expect the OF graph 
> bindings to be used to model data connections.
> 

The leverage of "the initial support" to "the complete support" requires:
* audio bridge cell driver -- trivial, just one mute/unmute control,
* interrupt controller cell driver -- trivial, but for sake of perfection
  it requires some minimal changes in drivers/base/regmap/regmap-irq.c
* DS90Ux940 MIPI CSI-2 -- non-trivial one, but we have it ready, I just
  don't want to add it to the pile at the moment, otherwise we'll continue
  discussing cameras, and I'd like to postpone it :)

No more than that is needed to get absolutely complete support of 5 claimed
DS90UB9xx ICs, really. 5 other DS90UH9xx will require HDCP support in addition.

I'll try to roll out an example of DTS snippet soon.

--
Best wishes,
Vladimir

  reply	other threads:[~2018-10-12 13:59 UTC|newest]

Thread overview: 62+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-10-08 21:11 [PATCH 0/7] mfd/pinctrl: add initial support of TI DS90Ux9xx ICs Vladimir Zapolskiy
2018-10-08 21:11 ` [PATCH 1/7] dt-bindings: mfd: ds90ux9xx: add description " Vladimir Zapolskiy
2018-10-09  0:13   ` Marek Vasut
2018-10-09 11:11     ` Vladimir Zapolskiy
2018-10-09 20:55       ` Vladimir Zapolskiy
2018-10-09 21:03         ` Marek Vasut
2018-10-10  8:41   ` Linus Walleij
2018-10-12 11:44   ` Laurent Pinchart
2018-10-13 14:28     ` Vladimir Zapolskiy
2018-10-16 12:30       ` Laurent Pinchart
2018-10-30 16:43   ` Luca Ceresoli
2018-10-30 23:40     ` Vladimir Zapolskiy
2018-10-08 21:12 ` [PATCH 2/7] dt-bindings: mfd: ds90ux9xx: add description of TI DS90Ux9xx I2C bridge Vladimir Zapolskiy
2018-10-12 11:54   ` Laurent Pinchart
2018-10-30 16:43   ` Luca Ceresoli
2018-10-31 20:12     ` Vladimir Zapolskiy
2018-11-03 21:00       ` Luca Ceresoli
2018-11-03 22:07         ` Vladimir Zapolskiy
2018-10-08 21:12 ` [PATCH 3/7] dt-bindings: pinctrl: ds90ux9xx: add description of TI DS90Ux9xx pinmux Vladimir Zapolskiy
2018-10-10  8:45   ` Linus Walleij
2018-10-17 15:02     ` Rob Herring
2018-10-12 12:01   ` Laurent Pinchart
2018-10-13 13:47     ` Vladimir Zapolskiy
2018-10-16 12:48       ` Laurent Pinchart
2018-10-30 16:44         ` Luca Ceresoli
2018-10-31 20:31           ` Vladimir Zapolskiy
2018-10-08 21:12 ` [PATCH 4/7] mfd: ds90ux9xx: add TI DS90Ux9xx de-/serializer MFD driver Vladimir Zapolskiy
2018-10-09  4:08   ` kbuild test robot
2018-10-09 11:14     ` Vladimir Zapolskiy
2018-10-12  6:03   ` Lee Jones
2018-10-12  7:41     ` Vladimir Zapolskiy
2018-10-12  8:39       ` Lee Jones
2018-10-12  9:20         ` Kieran Bingham
2018-10-12 10:58           ` Vladimir Zapolskiy
2018-10-12 11:34             ` Lee Jones
2018-10-12 14:13               ` Vladimir Zapolskiy
2018-10-12 14:25                 ` Lee Jones
2018-10-12 11:47             ` Kieran Bingham
2018-10-12 13:01               ` Laurent Pinchart
2018-10-12 13:59                 ` Vladimir Zapolskiy [this message]
2018-10-16 13:08                   ` Laurent Pinchart
2018-10-23  8:16                   ` Vladimir Zapolskiy
2018-10-30 16:44                     ` Luca Ceresoli
2018-10-13 15:10                 ` Vladimir Zapolskiy
2018-10-16 13:12                   ` Laurent Pinchart
2018-10-16 18:32                     ` Vladimir Zapolskiy
2018-10-13 12:33               ` Vladimir Zapolskiy
2018-10-12 11:24         ` Vladimir Zapolskiy
2018-10-12 11:43           ` Lee Jones
2018-10-12 14:23             ` Vladimir Zapolskiy
2018-10-12 13:07           ` Laurent Pinchart
2018-10-08 21:12 ` [PATCH 5/7] mfd: ds90ux9xx: add I2C bridge/alias and link connection driver Vladimir Zapolskiy
2018-10-12  6:04   ` Lee Jones
2018-10-12  7:32     ` Vladimir Zapolskiy
2018-10-12  9:03       ` Lee Jones
2018-10-12 13:12       ` Laurent Pinchart
2018-10-12 14:36         ` Vladimir Zapolskiy
2018-10-16 14:06           ` Laurent Pinchart
2018-10-08 21:12 ` [PATCH 6/7] pinctrl: ds90ux9xx: add TI DS90Ux9xx pinmux and GPIO controller driver Vladimir Zapolskiy
2018-10-10  9:04   ` Linus Walleij
2018-10-08 21:12 ` [PATCH 7/7] MAINTAINERS: add entry for TI DS90Ux9xx FPD-Link III drivers Vladimir Zapolskiy
2018-10-12 11:34 ` [PATCH 0/7] mfd/pinctrl: add initial support of TI DS90Ux9xx ICs Laurent Pinchart

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=735126c6-9780-a9ef-4fd0-699d6876ed37@mleia.com \
    --to=vz@mleia.com \
    --cc=devicetree@vger.kernel.org \
    --cc=kieran.bingham@ideasonboard.com \
    --cc=laurent.pinchart@ideasonboard.com \
    --cc=lee.jones@linaro.org \
    --cc=linus.walleij@linaro.org \
    --cc=linux-gpio@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-media@vger.kernel.org \
    --cc=marek.vasut@gmail.com \
    --cc=robh+dt@kernel.org \
    --cc=vladimir_zapolskiy@mentor.com \
    --cc=wsa@the-dreams.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).