| 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
| ||
|
Message-ID: <CAGTfZH1ZAD_R1q8COUE2hdrq24BS1muJ=7-_1VeHriBqH+aAZQ@mail.gmail.com>
Date: Tue, 31 May 2016 23:34:48 +0900
From: Chanwoo Choi <cwchoi00@...il.com>
To: Rob Herring <robh@...nel.org>
Cc: Venkat Reddy Talla <vreddytalla@...dia.com>,
MyungJoo Ham <myungjoo.ham@...sung.com>,
Pawel Moll <pawel.moll@....com>,
Mark Rutland <mark.rutland@....com>,
Ian Campbell <ijc+devicetree@...lion.org.uk>,
"devicetree@...r.kernel.org" <devicetree@...r.kernel.org>,
Kumar Gala <galak@...eaurora.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
Laxman Dewangan <ldewangan@...dia.com>
Subject: Re: [PATCH v4] extcon: gpio: Add the support for Device tree bindings
Hi Rob,
On Tue, May 31, 2016 at 10:35 PM, Rob Herring <robh@...nel.org> wrote:
> On Tue, May 31, 2016 at 2:35 AM, Chanwoo Choi <cw00.choi@...sung.com> wrote:
>> Hi Rob,
>>
>> On 2016년 05월 31일 15:44, Chanwoo Choi wrote:
>>> On 2016년 05월 28일 00:29, Rob Herring wrote:
>>>> On Thu, May 26, 2016 at 05:17:45PM +0530, Venkat Reddy Talla wrote:
>>>>> Add the support for Device tree bindings of extcon-gpio driver.
>>>>> The extcon-gpio device tree node must include the both 'extcon-id' and
>>>>> 'gpios' property.
>>>>
>>>> I think extcon bindings are a mess in general...
>>>>
>>>>> For example:
>>>>> usb_cable: extcon-gpio-0 {
>>>>> compatible = "extcon-gpio";
>>>>> extcon-id = <EXTCON_USB>;
>>>>> gpios = <&gpio6 1 GPIO_ACTIVE_HIGH>;
>>>>> }
>>>>> ta_cable: extcon-gpio-1 {
>>>>> compatible = "extcon-gpio";
>>>>> extcon-id = <EXTCON_CHG_USB_DCP>;
>>>>> gpios = <&gpio3 2 GPIO_ACTIVE_LOW>;
>>>>> debounce-ms = <50>; /* 50 millisecond */
>>>>> wakeup-source;
>>>>> }
>>>>
>>>> This is all 1 logical connector, the USB connector. Why are you
>>>> describing cables? Those are not part of the h/w and are dynamic.
>>>> Describe this as a connector which is one thing (i.e. node). Use a
>>>> compatible string that reflects the type of connector
>>>> (usb-microab-connector), not the driver you want to use. Then define
>>>> GPIO lines needed to provide state information like VBus, ID, charger
>>>> modes and control lines like soft connect (D+ pullup enable), VBus
>>>> enable, etc.
>>>
>>> You're right. The extcon-gpio driver will not use the "extcon-gpio" raw compatible.
>>> As you commented[1], the each connector will have the unique name to use the extcon-gpio.c driver.
>>> [1] https://lkml.org/lkml/2015/10/21/906
>>>
>>>
>>> For example,
>>> The extcon-gpio.c driver may have the different name including the h/w information
>>> according to the kind of external connector.
>>>
>>> static const struct of_device_id gpio_extcon_of_match[] = {
>>> {
>>> .compatible = "extcon-chg-sdp", /* SDP charger connector */
>>> .data = EXTCON_CHG_SDP_DATA,
>>> }, {
>>> .compatible = "extcon-chg-dcp", /* DCP charger connector */
>>> .data = EXTCON_CHG_DCP_DATA,
>>> }, {
>>> .compatible = "extcon-jack-microphone", /* Microphone jack connector */
>>> .data = EXTCON_JACK_MICROPHONE_DATA,
>>> }, {
>>> .compatible = "extcon-disp-hdmi", /* HDMI connector*/
>>> .data = EXTCON_DISP_HDMI_DATA,
>>> },
>>> ......
>>> };
>>
>> I reply it again.
>>
>> The extcon-gpio.c is very similar with existing gpio_keys.c driver[1]
>> [1] drivers/input/keyboard/gpio_keys.c
>
> There is a big difference in that each gpio-key is independent. The
> only state is pressed or not. A USB connector has multiple pieces of
> state information. You may be treating them independently, but I don't
> think they should be.
I think that is it not different because the EXTCON can only notify the whether
external connector is attached or detached such as gpio-key (pressed or not).
EXTCON don't handle the any additional state except for attached or detached.
>
>> The gpio_keys.c driver use the following style to support the device-tree.
>> It use the "gpio-keys" compatible and this dt node include the specific
>> 'key code' such as 'extcon-id = <EXTCON_CHG_USB_DCP>;'
>
> This is state information about what is currently attached. The
> analogy with gpio-keys would be multiple key codes on one gpio which
> would be broken...
I compared between 'gpis-keys' and ' extcon-gpio' driver as following:
name | gpio-keys | extcon-gpio
---------------------------------------------------------------------------------------------
gpio | gpios = <> | extcon-gpio = <>
---------------------------------------------------------------------------------------------
type | linux,code = <> | extcon-id = <>
---------------------------------------------------------------------------------------------
key & | KEY_POWER | EXTCON_USB
extcon id | KEY_VOLUME_UP | EXTCON_CHG_USB_SDP
| KEY_VOLUME_DOWN | EXTCON_JACK_MICROPHONE
| etc | etc
---------------------------------------------------------------------------------------------
state | pressed or not | attached or detached
---------------------------------------------------------------------------------------------
>
>>
>> gpio_keys {
>> compatible = "gpio-keys";
>>
>> power_key {
>> gpios = <&gpx2 7 GPIO_ACTIVE_LOW>;
>> linux,code = <KEY_POWER>;
>> label = "power key";
>> debounce-interval = <10>;
>> wakeup-source;
>> };
>> };
>>
>> If the extcon-gpio.c driver should have the separate compatible according to
>> the kind of external connector, the list of compatible name of extcon-gpio.c driver
>> will be increased when new external connector is attached.
>
> So? Different h/w needs different compatible strings.
>
> But again, you are mixing describing the connector (only what is
> soldered on a board) and state information (what is attached). Do not
> put state information into DT (describe the gpio signals or chip that
> provides the state information).
>
>> The extcon-gpio.c driver can separate the kind of external connector
>> by using the 'extcon-id' property.
>
> This use of DT is just broken. Come up with another way.
>
> Rob
Thanks,
Chanwoo Choi
Powered by blists - more mailing lists