Re: [PATCH] net: can: Enable xilinx driver for all ARCHs

[Jean Delvare <jdelvare@suse.de>]

Le Monday 09 March 2015 à 10:56 +0100, Marc Kleine-Budde a écrit :
> On 03/09/2015 10:52 AM, Michal Simek wrote:
> > On 03/09/2015 10:13 AM, Marc Kleine-Budde wrote:
> >> On 03/09/2015 09:58 AM, Michal Simek wrote:
> >>> On 03/09/2015 09:50 AM, Marc Kleine-Budde wrote:
> >>>> On 03/09/2015 09:48 AM, Michal Simek wrote:
> >>>>> Remove Kconfig dependency and enable driver for
> >>>>> all ARCHs.
> >>>>>
> >>>>> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
> >>>>> Acked-by: Sören Brinkmann <soren.brinkmann@xilinx.com>
> >>>>> ---
> >>>>>
> >>>>> Test for all archs done by Kbuild test robot without any problem.
> >>>>> ---
> >>>>>  drivers/net/can/Kconfig | 1 -
> >>>>>  1 file changed, 1 deletion(-)
> >>>>>
> >>>>> diff --git a/drivers/net/can/Kconfig b/drivers/net/can/Kconfig
> >>>>> index 98d73aab52fe..f690c3fb3088 100644
> >>>>> --- a/drivers/net/can/Kconfig
> >>>>> +++ b/drivers/net/can/Kconfig
> >>>>> @@ -131,7 +131,6 @@ config CAN_RCAR
> >>>>>  
> >>>>>  config CAN_XILINXCAN
> >>>>>  	tristate "Xilinx CAN"
> >>>>> -	depends on ARCH_ZYNQ || MICROBLAZE || COMPILE_TEST
> >>>>
> >>>> I think that's what COMPILE_TEST ist for?
> >>>
> >>> For compilation yes but not for enabling. Currently this driver
> >>> can be also used on ARM64 that's why people suggesting directly
> >>> to remove dependency on arch and then COMPILE_TEST can be removed
> >>> too.
> >>
> >> In the past (May 2014) I had people complaining that certain ARM SoC
> >> specific drivers are enabled on ARM in general, not just on that SoC. As
> >> I'm not following arm64 in detail, has the notion of using depends on
> >> ARCH changes since then?
> >>
> >>> Or do you want me to change description to mention that this is for
> >>> ARM64 enabling?
> >>
> >> Given this is consensus, a remark to ARM64 would be appreciated. :)
> > 
> > David has applied this patch which is in general just the same as this one.
> > 
> > (linux-next) Remove architecture dependency
> > 28811a8c00fe0d899b8a544421f3b4947425d5e8
> > 
> > Mark Brown has suggested to do so for spi and don't check architecture at
> > all.
> > I have also sent similar patch for watchdog driver too.
> > 
> > I think it is up to you if you want to add ARM64 to Kconfig or just remove
> > that arch dependencies.
> > To be honest my goal is to enable this driver for ARM64.
> > Please tell me what way you prefer.
> 
> I like the idea of removing the depends on ARCH completely. Jean, what
> do you think?

Removing the dependency completely will let the option be displayed on
systems where the driver is useless. I am in favor of having hardware
dependencies on as many drivers as possible to avoid bothering the user
with irrelevant questions. The list of Kconfig entries has grown a lot
over time!

If the current dependency is too strict then I would suggest to extend
it or to make it broader (depends on ARM || ARM64 || COMPILE_TEST would
be acceptable IMHO.) Dropping it completely only makes sense if the part
is used on so many systems that the dependency becomes too long or is a
pain to maintain.

Thanks,
-- 
Jean Delvare
SUSE L3 Support