Received-SPF: pass (google.com: domain of linux-kernel+bounces-98696-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Precedence: bulk
Mime-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset=UTF-8
Date: Mon, 11 Mar 2024 11:26:16 +0100
Message-Id: <CZQUKBQF1GZ9.3RSNW5WQBU9L6@gimli.ms.mff.cuni.cz>
Cc: "Lee Jones" <lee@kernel.org>, "Rob Herring" <robh+dt@kernel.org>,
 "Krzysztof Kozlowski" <krzysztof.kozlowski+dt@linaro.org>, "Conor Dooley"
 <conor+dt@kernel.org>, "Liam Girdwood" <lgirdwood@gmail.com>, "Mark Brown"
 <broonie@kernel.org>, <devicetree@vger.kernel.org>,
 <linux-kernel@vger.kernel.org>, <linux-input@vger.kernel.org>,
 =?utf-8?q?Duje_Mihanovi=C4=87?= <duje.mihanovic@skole.hr>,
 <~postmarketos/upstreaming@lists.sr.ht>, <phone-devel@vger.kernel.org>
Subject: Re: [RFC PATCH v3 4/5] input: add onkey driver for Marvell 88PM886
 PMIC
To: "Krzysztof Kozlowski" <krzysztof.kozlowski@linaro.org>, "Dmitry
 Torokhov" <dmitry.torokhov@gmail.com>
From: "Karel Balej" <karelb@gimli.ms.mff.cuni.cz>
References: <20240303101506.4187-1-karelb@gimli.ms.mff.cuni.cz>
 <20240303101506.4187-5-karelb@gimli.ms.mff.cuni.cz>
 <ZeTgEmjJc_VhYpLm@google.com>
 <CZL8ZSZAVEBI.349BV2Y6AKIPN@gimli.ms.mff.cuni.cz>
 <ZeZxI_spu4vwxrs7@google.com>
 <CZQ1EP61IDOC.1PPYGMIOINGND@gimli.ms.mff.cuni.cz>
 <3601a374-4161-40e1-8a80-9bbfdae5bd8a@linaro.org>
In-Reply-To: <3601a374-4161-40e1-8a80-9bbfdae5bd8a@linaro.org>

Krzysztof Kozlowski, 2024-03-10T21:35:36+01:00:
> On 10/03/2024 12:35, Karel Balej wrote:
> > Dmitry Torokhov, 2024-03-04T17:10:59-08:00:
> >> On Mon, Mar 04, 2024 at 09:28:45PM +0100, Karel Balej wrote:
> >>> Dmitry,
> >>>
> >>> Dmitry Torokhov, 2024-03-03T12:39:46-08:00:
> >>>> On Sun, Mar 03, 2024 at 11:04:25AM +0100, Karel Balej wrote:
> >>>>> From: Karel Balej <balejk@matfyz.cz>
> >>>>>
> >>>>> Marvell 88PM886 PMIC provides onkey among other things. Add client
> >>>>> driver to handle it. The driver currently only provides a basic sup=
port
> >>>>> omitting additional functions found in the vendor version, such as =
long
> >>>>> onkey and GPIO integration.
> >>>>>
> >>>>> Signed-off-by: Karel Balej <balejk@matfyz.cz>
> >>>>> ---
> >>>>>
> >>>>> Notes:
> >>>>>     RFC v3:
> >>>>>     - Drop wakeup-source.
> >>>>>     RFC v2:
> >>>>>     - Address Dmitry's feedback:
> >>>>>       - Sort includes alphabetically.
> >>>>>       - Drop onkey->irq.
> >>>>>       - ret -> err in irq_handler and no initialization.
> >>>>>       - Break long lines and other formatting.
> >>>>>       - Do not clobber platform_get_irq error.
> >>>>>       - Do not set device parent manually.
> >>>>>       - Use input_set_capability.
> >>>>>       - Use the wakeup-source DT property.
> >>>>>       - Drop of_match_table.
> >>>>
> >>>> I only said that you should not be using of_match_ptr(), but you sti=
ll
> >>>> need to have of_match_table set and have MODULE_DEVICE_TABLE() for t=
he
> >>>> proper module loading support.
> >>>
> >>> I removed of_match_table because I no longer need compatible for this=
 --
> >>> there are no device tree properties and the driver is being instantia=
ted
> >>> by the MFD driver.
> >>>
> >>> Is the MODULE_DEVICE_TABLE() entry needed for the driver to probe whe=
n
> >>> compiled as module? If that is the case, given what I write above, am=
 I
> >>> correct that MODULE_DEVICE_TABLE(platform,...) would be the right thi=
ng
> >>> to use here?
> >>
> >> Yes, if uevent generated for the device is "platform:<name>" then
> >> MODULE_DEVICE_TABLE(platform,...) will suffice. I am not sure how MFD
> >> sets it up (OF modalias or platform), but you should be able to check
> >> the format looking at the "uevent" attribute for your device in sysfs
> >> (/sys/devices/bus/platform/...).=20
> >=20
> > The uevent is indeed platform.
> >=20
> > But since there is only one device, perhaps having a device table is
> > superfluous and using `MODULE_ALIAS("platform:88pm886-onkey")` is more
> > fitting?
>
> Adding aliases for standard IDs and standard cases is almost never
> correct. If you need module alias, it means your ID table is wrong (or
> missing, which is usually wrong).
>
> >=20
> > Although I don't understand why this is even necessary when the driver
> > name is such and the module is registered using
> > `module_platform_driver`...
>
> ID table and MODULE_DEVICE_TABLE() are necessary for modprobe to work.

I think I understand the practical reasons. My point was that I would
expect the alias to be added automatically even in the case that the
device table is absent based solely on the driver name and the
registration method (*module*_*platform*_driver). Why is that not the
case? Obviously the driver name matching the mfd_cell name is sufficient
for the driver to probe when it is built in so the name does seem to
serve as some identification for the device just as a device table entry
would.

Furthermore, drivers/input/serio/ioc3kbd.c does not seem to have an ID
table either, nor a MODULE_ALIAS -- is that a mistake? If not, what
mechanism causes the driver to probe when compiled as a module? It seems
to me to effectively be the same setup as with my driver and that does
not load automatically (because of the missing alias).

> Just run `modinfo`.

Thank you very much,
K. B.