[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <87twm639kc.fsf@ketchup.mtl.sfl>
Date: Thu, 21 Jan 2016 16:29:55 -0500
From: Vivien Didelot <vivien.didelot@...oirfairelinux.com>
To: Florian Fainelli <f.fainelli@...il.com>,
Andrew Lunn <andrew@...n.ch>, narmstrong@...libre.com
Cc: netdev <netdev@...r.kernel.org>
Subject: Re: [PATCH RFC 26/28] dsa: Convert mv88e6xxx into a library allowing driver modules
Hi Andrew, Florian,
Florian Fainelli <f.fainelli@...il.com> writes:
> Le 23/12/2015 04:56, Andrew Lunn a écrit :
>> Turn mv88e6xxx into a library module, by exporting its symbols. Have
>> each driver register their own driver functions with the DSA core in
>> there init function.
>>
>> This results in each driver being a loadable module.
>>
>> Signed-off-by: Andrew Lunn <andrew@...n.ch>
>> ---
>
> [snip]
>
>> +obj-$(CONFIG_NET_DSA_MV88E6XXX) += mv88e6xxx.o
>> +obj-$(CONFIGNET_DSA_MV88E6131) += mv88e6123.o
>
> Missing underscore here between CONFIG and NET.
>
>> +obj-$(CONFIG_NET_DSA_MV88E6131) += mv88e6131.o
>> +obj-$(CONFIG_NET_DSA_MV88E6352) += mv88e6352.o
>> +obj-$(CONFIG_NET_DSA_MV88E6171) += mv88e6171.o
>> obj-$(CONFIG_NET_DSA_BCM_SF2) += bcm_sf2.o
>> diff --git a/drivers/net/dsa/mv88e6123.c b/drivers/net/dsa/mv88e6123.c
>> index bb39720f3e8b..9680e59fd2ae 100644
>> --- a/drivers/net/dsa/mv88e6123.c
>> +++ b/drivers/net/dsa/mv88e6123.c
>> @@ -167,7 +167,21 @@ static struct platform_driver mv88e6123_driver = {
>> .of_match_table = mv88e6123_of_match,
>> },
>> };
>> -module_platform_driver(mv88e6123_driver);
>>
>> +static int __init mv88e6123_init(void)
>> +{
>> + register_switch_driver(&mv88e6123_switch_driver);
>> +
>> + return platform_driver_register(&mv88e6123_driver);
>> +}
>> +
>> +static void __exit mv88e6123_exit(void)
>> +{
>> + platform_driver_unregister(&mv88e6123_driver);
>> + unregister_switch_driver(&mv88e6123_switch_driver);
>> +}
>
> I think Vivien had started something like this, but it could be nice to
> have a helper function/macro which reduces the boilerplate code, not
> critical for now though.
Indeed I did suggest this macro and exporting in the past, without
success though:
https://lkml.org/lkml/2015/5/2/152
Maybe the other approach with a reference driver (e.g. mv88e6352.c)
declaring a table of supported switch names is more natural?
Most Linux drivers factorize the support for similar devices this way
(e.g. lm75, ...).
Thanks,
-v
Powered by blists - more mailing lists