netdev - Re: [PATCH v2 15/24] net: qualcomm: MODULE_DEVICE

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

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <07871EF5-D1D3-4677-8100-B5E7F14DB302@holtmann.org>
Date:   Mon, 11 Jun 2018 17:52:31 +0200
From:   Marcel Holtmann <marcel@...tmann.org>
To:     Ricardo Ribalda Delgado <ricardo.ribalda@...il.com>
Cc:     LKML <linux-kernel@...r.kernel.org>,
        "open list:SERIAL DRIVERS" <linux-serial@...r.kernel.org>,
        Lino Sanfilippo <LinoSanfilippo@....de>,
        "David S. Miller" <davem@...emloft.net>,
        Stefan Wahren <stefan.wahren@...e.com>,
        Rob Herring <robh@...nel.org>, Johan Hovold <johan@...nel.org>,
        netdev@...r.kernel.org
Subject: Re: [PATCH v2 15/24] net: qualcomm: MODULE_DEVICE_TABLE(serdev)

Hi Ricardo,

>>>> the commit message is misleading me. If I build something with ACPI or DT support, then modinfo will show all modalias information for ACPI and DT compatible strings. What else does udev/modprobe actually need? Is something broken with the modalias export?
>>> 
>>> The main purpose is to autoload drivers for devices that have been
>>> created via sysfs or another module.
>>> 
>>> Eg1: We have a serial port on a standard computer that has connected a
>>> GPS module. Since it is something that is not in the ACPI nor the DT
>>> table the user will run
>>> 
>>> echo serdev_gps > /sys/bus/serial/devices/serial0/new_device
>>> 
>>> Eg2 module: https://github.com/ribalda/linux/blob/415bb3f0076c2b846ebe5409589b8e1e3004f55a/drivers/tty/serdev/test_platform.c
>>> 
>>> Modprobe does not know what module to load for that device unless
>>> there is a matching MODULE_DEVICE_TABLE
>>> Today, we have the same functionality for i2c devices
>>> https://www.kernel.org/doc/Documentation/i2c/instantiating-devices
>> 
>> but why does this have to be the driver name? I would rather say, create a generic string that describes the hardware and then use that. I think you also want the special handling, as from USB for example, that lets you reference an existing .compatible or ACPI ID as reference so that the driver_data is copied.
> 
> We can choose any name, but if there are no special variants (like the
> rave-sp driver) I would prefer using the module name. We can of course
> use more names, like the part number of the the device, but it is very
> convenient to also use the module name, and other subsystems also do
> that.
> 
> If you want to add specific names for this device please let me know
> and I will add them to the list. You know much more about this module
> than myself.

if we want to use the driver name, then why not build this into the new_device handling itself instead of duplicating that name in each serdev_device_id table. What is the reference here? For example platform device do matching on driver name if I recall this correctly.

However what is most important is that device_get_match_data() actually works. There should be no difference between DT, ACPI or a dynamic device.

Regards

Marcel