i2c: cadence: fix Kconfig dependency

On Mon, 2014-04-07 at 11:02AM +0400, Alexander Shiyan wrote:
> Mon, 7 Apr 2014 08:56:57 +0200 от Uwe Kleine-König <u.kleine-koenig@pengutronix.de>:
> > Hello,
> ...
> > > > On Sun, Apr 06, 2014 at 09:07:07PM +0200, Wolfram Sang wrote:
> > > > > During development, the driver first really needed to depend on
> > > > > COMMON_CLK only. Later, it was switched to writel_relaxed, but it was
> > > > > forgotten to update the dependencies, so build errors occured:
> > > > > 
> > > > > config: make ARCH=i386 allyesconfig
> > > > > 
> > > > > All error/warnings:
> > > > > 
> > > > >    drivers/i2c/busses/i2c-cadence.c: In function 'cdns_i2c_clear_bus_hold':
> > > > > >> drivers/i2c/busses/i2c-cadence.c:168:3: error: implicit declaration
> > > > > >> of function 'writel_relaxed' [-Werror=implicit-function-declaration]
> > > > > 
> > > > > Use a very safe dependency for now.
> > > > > 
> > > > > Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
> > > > > ---
> > > > >  drivers/i2c/busses/Kconfig | 2 +-
> > > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > > 
> > > > > diff --git a/drivers/i2c/busses/Kconfig b/drivers/i2c/busses/Kconfig
> > > > > index 93165ff453ab..3d3b9b3577c5 100644
> > > > > --- a/drivers/i2c/busses/Kconfig
> > > > > +++ b/drivers/i2c/busses/Kconfig
> > > > > @@ -378,7 +378,7 @@ config I2C_BLACKFIN_TWI_CLK_KHZ
> > > > >  
> > > > >  config I2C_CADENCE
> > > > >  	tristate "Cadence I2C Controller"
> > > > > -	depends on COMMON_CLK
> > > > > +	depends on ARCH_ZYNQ
> > > > I'd suggest:
> > > > 
> > > > 	depends on ARM && (ARCH_ZYNC || COMPILE_TEST)
> > > 
> > > ARCH_ZYNC || (ARM && COMPILE_TEST)
> > > 
> > > Same, but more clear.
> > "more clear" might be subjective, at least for me I don't see a benefit
> > from one over the other.
> > 
> > hmm, considering a (hypothetical) driver that depends on I2C and is
> > available on Zync but compiles on ARM, it would need either: 
> 
> No. Look at drivers/i2c/Kconfig:

I'd propose the below fix. The relaxed IO helpers are currently
available on ARM only and the clock notifier used, depends on
COMMON_CLK. I2C-related dependencies should be okay due to the location in
Kconfig of this option.

	Thanks,
	Sören

---------------8<---------------8<----------------8<-----------------8<-----------8<-------
Date: Sun, 6 Apr 2014 15:19:34 -0700
Subject: [PATCH] i2c: cadence: Add dependency on ARM

Due to switching to the (read|write)l_relaxed IO helpers, building this
driver fails on non-ARM architectures, which don't provide these helpers
with the following error:
  All error/warnings:

     drivers/i2c/busses/i2c-cadence.c: In function 'cdns_i2c_clear_bus_hold':
  >> drivers/i2c/busses/i2c-cadence.c:168:3: error: implicit declaration of function 'writel_relaxed' [-Werror=implicit-function-declaration]
        cdns_i2c_writereg(reg & ~CDNS_I2C_CR_HOLD, CDNS_I2C_CR_OFFSET);
        ^
     cc1: some warnings being treated as errors

Make the driver additionally depend on CONFIG_ARM to fix this.

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Soren Brinkmann <soren.brinkmann@xilinx.com>
---
 drivers/i2c/busses/Kconfig | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

> I'd propose the below fix. The relaxed IO helpers are currently
> available on ARM only and the clock notifier used, depends on
> COMMON_CLK. I2C-related dependencies should be okay due to the location in
> Kconfig of this option.

If somebody thinks an update is still needed, please resend against
latest i2c-next. Thanks.

Hi Wolfram,

On Wed, 2014-05-21 at 10:29AM +0200, Wolfram Sang wrote:
> > I'd propose the below fix. The relaxed IO helpers are currently
> > available on ARM only and the clock notifier used, depends on
> > COMMON_CLK. I2C-related dependencies should be okay due to the location in
> > Kconfig of this option.
> 
> If somebody thinks an update is still needed, please resend against
> latest i2c-next. Thanks.
> 

I think we're fine for now. Once the *_relaxed IO helpers are available
on all archs it makes sense to revisit the COMPILE_TEST option. Until
then we are restricted to ARM anyway and the additional Zynq restriction
doesn't make a big difference, IMHO.

@Michal: Do you have an update on the IO helpers?

	Thanks,
	Sören


--
To unsubscribe from this list: send the line "unsubscribe linux-i2c" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

On 05/21/2014 06:00 PM, Sören Brinkmann wrote:
> Hi Wolfram,
> 
> On Wed, 2014-05-21 at 10:29AM +0200, Wolfram Sang wrote:
>>> I'd propose the below fix. The relaxed IO helpers are currently
>>> available on ARM only and the clock notifier used, depends on
>>> COMMON_CLK. I2C-related dependencies should be okay due to the location in
>>> Kconfig of this option.
>>
>> If somebody thinks an update is still needed, please resend against
>> latest i2c-next. Thanks.
>>
> 
> I think we're fine for now. Once the *_relaxed IO helpers are available
> on all archs it makes sense to revisit the COMPILE_TEST option. Until
> then we are restricted to ARM anyway and the additional Zynq restriction
> doesn't make a big difference, IMHO.
> 
> @Michal: Do you have an update on the IO helpers?

My patch was replaced by Will's one but I can't see it in linux-next
that's why we should ask directly Will.
Will: What's the status on it?

Thanks,
Michal