From mboxrd@z Thu Jan 1 00:00:00 1970 From: arnd@arndb.de (Arnd Bergmann) Date: Thu, 05 Jun 2014 17:29:01 +0200 Subject: [PATCH] ASoC: sigmadsp: remove I2C support In-Reply-To: <53908B5B.10807@metafoo.de> References: <6181626.txoA4bpImf@wuerfel> <4610731.9TBGyCy3TV@wuerfel> <53908B5B.10807@metafoo.de> Message-ID: <6805774.tHfga8A7Qh@wuerfel> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Thursday 05 June 2014 17:23:07 Lars-Peter Clausen wrote: > On 06/05/2014 05:10 PM, Arnd Bergmann wrote: > > On Thursday 05 June 2014 16:55:47 Lars-Peter Clausen wrote: > >> On 06/05/2014 04:47 PM, Arnd Bergmann wrote: > >>> The newly added ADAU1781 codec comes with separate drivers for > >>> I2C and SPI modes, which gives us a new variation of the dependency > >>> problems: If SPI is enabled and the I2C core is a loadable module, > >>> selecting SND_SOC_ADAU1781_SPI as built-in leads to a link error because > >>> the common base code from sigmadsp.c now also gets built-in: > >>> > >>> sound/built-in.o: In function `sigma_action_write_i2c': > >>> :(.text+0x5d8d4): undefined reference to `i2c_master_send' > >>> > >>> Rather than adding yet more complex Kconfig logic, this converts the last > >>> user of process_sigma_firmware() to process_sigma_firmware_regmap(), > >>> which seems trivial after the driver itself now uses regmap as well. > >>> With that done, there is no more linker dependency from > >>> SND_SOC_ADAU1781_SPI to i2c_master_send. > >>> > >>> Signed-off-by: Arnd Bergmann > >>> Cc: Lars-Peter Clausen > >> > >> No this patch won't work. Regmap support for the adau1701 is rather limited > >> which is why we need native i2c support in the lib. > > > > Ah, I see. > > > >> But yea, I just noticed the same problem an hour ago and was thinking > >> hopefully I'll get to fixing this before it shows up in Arnd's randconfig > >> builds > > > > It's actually my second attempt after I finished writing the first patch > > for it. Is that better? > > That will fix the symptoms of the issue for SND_SOC_ALL_CODECS, but the > problem is still there. I'll send a patch which factors the regmap and I2C > sigmadsp code out into their own modules. > Ok, that sounds best, thanks! Arnd