Skip to contentSkip to conversation

iio: accel: fxls8962af: fix i2c dependency

PATCH In review Arnd Bergmann 3 participants 7 messages Jul 21, 2021 – Jul 24, 2021 IIO

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 iio: accel: fxls8962af: fix i2c dependency

From: Arnd Bergmann <[email protected]>

With CONFIG_SPI=y and CONFIG_I2C=m, building fxls8962af into vmlinux causes a link error against the I2C module:

aarch64-linux-ld: drivers/iio/accel/fxls8962af-core.o: in function `fxls8962af_fifo_flush':
fxls8962af-core.c:(.text+0x3a0): undefined reference to `i2c_verify_client'

Work around it by adding a Kconfig dependency that forces the SPI driver to be a loadable module whenever I2C is a module.

Fixes af959b7b96b8 ("iio: accel: fxls8962af: fix errata bug E3 - I2C burst reads")

---
I'm not overly happy with the fix either, but couldn't think of a better idea. If someone provide a different fix, please ignore mine.

---
 drivers/iio/accel/Kconfig | 2 ++
 1 file changed, 2 insertions(+)
drivers/iio/accel/Kconfig +2 -0
index 0e56ace61103..8d8b1ba42ff8 100644
@@ -231,6 +231,7 @@ config DMARD10
 
 config FXLS8962AF
 	tristate
+	depends on I2C || !I2C # cannot be built-in for modular I2C
 
 config FXLS8962AF_I2C
 	tristate "NXP FXLS8962AF/FXLS8964AF Accelerometer I2C Driver"
@@ -247,6 +248,7 @@ config FXLS8962AF_I2C
 config FXLS8962AF_SPI
 	tristate "NXP FXLS8962AF/FXLS8964AF Accelerometer SPI Driver"
 	depends on SPI
+	depends on I2C || !I2C
 	select FXLS8962AF
 	select REGMAP_SPI
 	help
Signature · 1 line
2.29.2
#1 lore raw
Quoted 11 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

With CONFIG_SPI=y and CONFIG_I2C=m, building fxls8962af into vmlinux causes a link error against the I2C module:

aarch64-linux-ld: drivers/iio/accel/fxls8962af-core.o: in function `fxls8962af_fifo_flush':
fxls8962af-core.c:(.text+0x3a0): undefined reference to `i2c_verify_client'

Work around it by adding a Kconfig dependency that forces the SPI driver to be a loadable module whenever I2C is a module.

...

Quoted 3 lines

config FXLS8962AF
tristate
+ depends on I2C || !I2C # cannot be built-in for modular I2C

Can you enlighten me how this will not be a no-op?

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#2 lore raw
Quoted 14 lines · Andy Shevchenko
Quoted 11 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

With CONFIG_SPI=y and CONFIG_I2C=m, building fxls8962af into vmlinux causes a link error against the I2C module:

aarch64-linux-ld: drivers/iio/accel/fxls8962af-core.o: in function `fxls8962af_fifo_flush':
fxls8962af-core.c:(.text+0x3a0): undefined reference to `i2c_verify_client'

Work around it by adding a Kconfig dependency that forces the SPI driver to be a loadable module whenever I2C is a module.

Quoted 7 lines

...

Quoted 3 lines

config FXLS8962AF
tristate
+ depends on I2C || !I2C # cannot be built-in for modular I2C

Can you enlighten me how this will not be a no-op?

This part does nothing, it only causes a warning when FXLS8962AF gets selected =y when I2C=m.

The important bit is the other hunk that adds the same dependency to the FXLS8962AF_SPI symbol, which enforces that either I2C is completely disabled, or treated as a dependency that prevents the user from setting FXLS8962AF_SPI=y when that would cause a link failure.

The effect is similar to a 'depends on SND_SOC_I2C_AND_SPI', except we only need it on the SPI symbol here because the SPI core cannot be in a module itself.

Arnd

#3 lore raw
Quoted 3 lines · Arnd Bergmann
Quoted 1 line · Andy Shevchenko

On Wed, Jul 21, 2021 at 6:13 PM Arnd Bergmann <[email protected]> wrote:

...

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

config FXLS8962AF
tristate
+ depends on I2C || !I2C # cannot be built-in for modular I2C

Can you enlighten me how this will not be a no-op?

This part does nothing, it only causes a warning when FXLS8962AF gets selected =y when I2C=m.

This is something new to me. But shouldn't the other chunk guarantee that warning won't happen?

Quoted 5 lines

The important bit is the other hunk that adds the same dependency to the FXLS8962AF_SPI symbol, which enforces that either I2C is completely disabled, or treated as a dependency that prevents the user from setting FXLS8962AF_SPI=y when that would cause a link failure.

This part I understand and neither object to nor comment on.

Quoted 3 lines

The effect is similar to a 'depends on SND_SOC_I2C_AND_SPI', except we only need it on the SPI symbol here because the SPI core cannot be in a module itself.

I see. Thanks for elaboration.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#4 lore raw
Quoted 11 lines · Andy Shevchenko
Quoted 3 lines · Arnd Bergmann
Quoted 1 line · Andy Shevchenko

On Wed, Jul 21, 2021 at 6:13 PM Arnd Bergmann <[email protected]> wrote:

...

Quoted 4 lines
Quoted 4 lines
Quoted 3 lines

config FXLS8962AF
tristate
+ depends on I2C || !I2C # cannot be built-in for modular I2C

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

Can you enlighten me how this will not be a no-op?

This part does nothing, it only causes a warning when FXLS8962AF gets selected =y when I2C=m.

This is something new to me. But shouldn't the other chunk guarantee that warning won't happen?

Correct, it works without that, but if that fails after something changes, this version would provide better diagnostics than the FXLS8962AF core driver causing a link failure, and I found it documents better why the other driver needs the dependency.

Let me know if you prefer me to resend the patch without this hunk.

Arnd

maintainer
#5 lore raw
Quoted 21 lines · On Wed, 21 Jul 2021 20:40:30 +0200 Arnd Bergmann <[email protected]> wrote:
Quoted 11 lines · Andy Shevchenko
Quoted 3 lines · Arnd Bergmann
Quoted 1 line · Andy Shevchenko

On Wed, Jul 21, 2021 at 6:13 PM Arnd Bergmann <[email protected]> wrote:

...

Quoted 4 lines
Quoted 4 lines
Quoted 3 lines

config FXLS8962AF
tristate
+ depends on I2C || !I2C # cannot be built-in for modular I2C

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

Can you enlighten me how this will not be a no-op?

This part does nothing, it only causes a warning when FXLS8962AF gets selected =y when I2C=m.

This is something new to me. But shouldn't the other chunk guarantee that warning won't happen?

Quoted 8 lines

Correct, it works without that, but if that fails after something changes, this version would provide better diagnostics than the FXLS8962AF core driver causing a link failure, and I found it documents better why the other driver needs the dependency.

Let me know if you prefer me to resend the patch without this hunk.

Arnd

Hi Arnd,

I didn't think of this particularly combination when we dealt with last build issue the workaround brought in. I've applied this to the fixes-togreg branch of iio.git as an immediately solution, but longer term we should think about just using a function pointer to allow us to move this into the i2c specific module. If we do that we can drop this complex build logic later.

Thanks,

Jonathan

#6 lore raw
Quoted 8 lines · Jonathan Cameron

On Wed, 21 Jul 2021 20:40:30 +0200 Arnd Bergmann <[email protected]> wrote:

I didn't think of this particularly combination when we dealt with last build issue the workaround brought in. I've applied this to the fixes-togreg branch of iio.git as an immediately solution, but longer term we should think about just using a function pointer to allow us to move this into the i2c specific module. If we do that we can drop this complex build logic later.

Ok, that sounds good to me, thanks!

Arnd