diff mbox series

[v4,1/2] gpiolib: move all includes to the top of gpio/consumer.h

Message ID 20250217103922.151047-1-brgl@bgdev.pl (mailing list archive)
State New
Headers show
Series [v4,1/2] gpiolib: move all includes to the top of gpio/consumer.h | expand

Commit Message

Bartosz Golaszewski Feb. 17, 2025, 10:39 a.m. UTC 
From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

We have several conditional includes depending on !CONFIG_GPIOLIB. This
is supposed to reduce compilation time with CONFIG_GPIOLIB=y but in
practice there's no difference on modern machines. It makes adding new
stubs that depend on more than just GPIOLIB harder so move them all to
the top, unduplicate them and replace asm/ with preferred linux/
alternatives.

Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
---
Changes in v4:
- rebased after fixing conflicts

 include/linux/gpio/consumer.h | 12 +++---------
 1 file changed, 3 insertions(+), 9 deletions(-)

Comments

Bartosz Golaszewski Feb. 18, 2025, 10:26 a.m. UTC | #1 
From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>


On Mon, 17 Feb 2025 11:39:21 +0100, Bartosz Golaszewski wrote:
> We have several conditional includes depending on !CONFIG_GPIOLIB. This
> is supposed to reduce compilation time with CONFIG_GPIOLIB=y but in
> practice there's no difference on modern machines. It makes adding new
> stubs that depend on more than just GPIOLIB harder so move them all to
> the top, unduplicate them and replace asm/ with preferred linux/
> alternatives.
> 
> [...]

Applied, thanks!

[1/2] gpiolib: move all includes to the top of gpio/consumer.h
      commit: dea69f2d1cc8d9ecdc72ba350d10a1e71940ef18
[2/2] gpiolib: don't build HTE code with CONFIG_HTE disabled
      commit: 63cdf6241ac7edd94cb4cd9a8f1ba2c3c61ed219

Best regards,
Andy Shevchenko Feb. 19, 2025, 12:32 p.m. UTC | #2 
On Mon, Feb 17, 2025 at 11:39:21AM +0100, Bartosz Golaszewski wrote:
> From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> 
> We have several conditional includes depending on !CONFIG_GPIOLIB. This
> is supposed to reduce compilation time with CONFIG_GPIOLIB=y but in
> practice there's no difference on modern machines.

It's not about modern machines. If every maintainer will think this way,
we end up in the complete and utter dead end with the headers.

I believe you at least had read the cover letter for the infamous Ingo's series
about headers and how it speeds up the build (in some cases up to 70% on as you
said "modern machines").

> It makes adding new stubs that depend on more than just GPIOLIB harder so
> move them all to the top, unduplicate them and replace asm/ with preferred
> linux/ alternatives.

NAK.

This makes dependency hell things much worse and this is a step back on the
untangling the current situation along with the slowing down the speed of the
build. Please. consider to revert or discard this patch.

...

>  #include <linux/bits.h>
> +#include <linux/bug.h>

Okay to replace, but not okay to move.

>  #include <linux/err.h>
> +#include <linux/errno.h>

Please, double check that it uses error codes from it, otherwise err.h includes
asm/errno.h with basic codes already.

> +#include <linux/kernel.h>

This is definitely no. Please, read what's written in the top of that file and
here is just a proxy for should come in the future a kind of might_sleep.h.
Do not move this one at all, please.

>  #include <linux/types.h>
Bartosz Golaszewski Feb. 19, 2025, 2:35 p.m. UTC | #3 
On Wed, 19 Feb 2025 13:32:49 +0100, Andy Shevchenko
<andriy.shevchenko@intel.com> said:
> On Mon, Feb 17, 2025 at 11:39:21AM +0100, Bartosz Golaszewski wrote:
>> From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
>>
>> We have several conditional includes depending on !CONFIG_GPIOLIB. This
>> is supposed to reduce compilation time with CONFIG_GPIOLIB=y but in
>> practice there's no difference on modern machines.
>
> It's not about modern machines. If every maintainer will think this way,
> we end up in the complete and utter dead end with the headers.
>
> I believe you at least had read the cover letter for the infamous Ingo's series
> about headers and how it speeds up the build (in some cases up to 70% on as you
> said "modern machines").
>
>> It makes adding new stubs that depend on more than just GPIOLIB harder so
>> move them all to the top, unduplicate them and replace asm/ with preferred
>> linux/ alternatives.
>
> NAK.
>
> This makes dependency hell things much worse and this is a step back on the
> untangling the current situation along with the slowing down the speed of the
> build. Please. consider to revert or discard this patch.
>
> ...
>
>>  #include <linux/bits.h>
>> +#include <linux/bug.h>
>
> Okay to replace, but not okay to move.
>
>>  #include <linux/err.h>
>> +#include <linux/errno.h>
>
> Please, double check that it uses error codes from it, otherwise err.h includes
> asm/errno.h with basic codes already.
>
>> +#include <linux/kernel.h>
>
> This is definitely no. Please, read what's written in the top of that file and
> here is just a proxy for should come in the future a kind of might_sleep.h.
> Do not move this one at all, please.
>
>>  #include <linux/types.h>
>

Fair enough. Does this look right to you?

diff --git a/include/linux/gpio/consumer.h b/include/linux/gpio/consumer.h
index 0b2b56199c36..38e313fd0e9a 100644
--- a/include/linux/gpio/consumer.h
+++ b/include/linux/gpio/consumer.h
@@ -3,10 +3,6 @@
 #define __LINUX_GPIO_CONSUMER_H

 #include <linux/bits.h>
-#include <linux/bug.h>
-#include <linux/err.h>
-#include <linux/errno.h>
-#include <linux/kernel.h>
 #include <linux/types.h>

 struct acpi_device;
@@ -185,6 +181,10 @@ struct gpio_desc
*devm_fwnode_gpiod_get_index(struct device *dev,

 #else /* CONFIG_GPIOLIB */

+#include <linux/bug.h>
+#include <linux/err.h>
+#include <linux/kernel.h>
+
 static inline int gpiod_count(struct device *dev, const char *con_id)
 {
	return 0;
@@ -549,6 +549,10 @@ struct gpio_desc
*devm_fwnode_gpiod_get_index(struct device *dev,
 int gpiod_enable_hw_timestamp_ns(struct gpio_desc *desc, unsigned long flags);
 int gpiod_disable_hw_timestamp_ns(struct gpio_desc *desc, unsigned long flags);
 #else
+
+#include <linux/bug.h>
+#include <linux/err.h>
+
 static inline int gpiod_enable_hw_timestamp_ns(struct gpio_desc *desc,
					       unsigned long flags)
 {
@@ -615,6 +619,8 @@ int devm_acpi_dev_add_driver_gpios(struct device *dev,

 #else  /* CONFIG_GPIOLIB && CONFIG_ACPI */

+#include <linux/err.h>
+
 static inline int acpi_dev_add_driver_gpios(struct acpi_device *adev,
			      const struct acpi_gpio_mapping *gpios)
 {
@@ -640,6 +646,8 @@ void gpiod_unexport(struct gpio_desc *desc);

 #else  /* CONFIG_GPIOLIB && CONFIG_GPIO_SYSFS */

+#include <linux/err.h>
+
 static inline int gpiod_export(struct gpio_desc *desc,
			       bool direction_may_change)
 {

Bart
Andy Shevchenko Feb. 19, 2025, 3:18 p.m. UTC | #4 
On Wed, Feb 19, 2025 at 06:35:59AM -0800, Bartosz Golaszewski wrote:
> On Wed, 19 Feb 2025 13:32:49 +0100, Andy Shevchenko
> <andriy.shevchenko@intel.com> said:
> > On Mon, Feb 17, 2025 at 11:39:21AM +0100, Bartosz Golaszewski wrote:
> >> From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> >>
> >> We have several conditional includes depending on !CONFIG_GPIOLIB. This
> >> is supposed to reduce compilation time with CONFIG_GPIOLIB=y but in
> >> practice there's no difference on modern machines.
> >
> > It's not about modern machines. If every maintainer will think this way,
> > we end up in the complete and utter dead end with the headers.
> >
> > I believe you at least had read the cover letter for the infamous Ingo's series
> > about headers and how it speeds up the build (in some cases up to 70% on as you
> > said "modern machines").
> >
> >> It makes adding new stubs that depend on more than just GPIOLIB harder so
> >> move them all to the top, unduplicate them and replace asm/ with preferred
> >> linux/ alternatives.
> >
> > NAK.
> >
> > This makes dependency hell things much worse and this is a step back on the
> > untangling the current situation along with the slowing down the speed of the
> > build. Please. consider to revert or discard this patch.
> >
> > ...
> >
> >>  #include <linux/bits.h>
> >> +#include <linux/bug.h>
> >
> > Okay to replace, but not okay to move.
> >
> >>  #include <linux/err.h>
> >> +#include <linux/errno.h>
> >
> > Please, double check that it uses error codes from it, otherwise err.h includes
> > asm/errno.h with basic codes already.
> >
> >> +#include <linux/kernel.h>
> >
> > This is definitely no. Please, read what's written in the top of that file and
> > here is just a proxy for should come in the future a kind of might_sleep.h.
> > Do not move this one at all, please.
> >
> >>  #include <linux/types.h>


Fair enough. Does this look right to you?

For kernel.h definitely, for err.h you proved your point (but which was unclear
to me as the repetitions are already being in a number).

For bug.h looks also good. But I prefer to use asm/bug.h as it's the one that
provides the APIs. Note, the reference to the recommended linux/* headers
rather than asm/* applies to the C code or custom headers which are not under
include/*. The latter should be optimized to what it uses exactly.

So, summarizing the above I would return kernel.h to where it belongs and
move back to asm/bug.h.
diff mbox series

Patch

diff --git a/include/linux/gpio/consumer.h b/include/linux/gpio/consumer.h
index 5cbd4afd7862..0dc49b5fca5c 100644
--- a/include/linux/gpio/consumer.h
+++ b/include/linux/gpio/consumer.h
@@ -3,7 +3,10 @@ 
 #define __LINUX_GPIO_CONSUMER_H
 
 #include <linux/bits.h>
+#include <linux/bug.h>
 #include <linux/err.h>
+#include <linux/errno.h>
+#include <linux/kernel.h>
 #include <linux/types.h>
 
 struct acpi_device;
@@ -184,11 +187,6 @@  struct gpio_desc *devm_fwnode_gpiod_get_index(struct device *dev,
 
 #else /* CONFIG_GPIOLIB */
 
-#include <linux/err.h>
-#include <linux/kernel.h>
-
-#include <asm/bug.h>
-
 static inline int gpiod_count(struct device *dev, const char *con_id)
 {
 	return 0;
@@ -609,8 +607,6 @@  int devm_acpi_dev_add_driver_gpios(struct device *dev,
 
 #else  /* CONFIG_GPIOLIB && CONFIG_ACPI */
 
-#include <linux/err.h>
-
 static inline int acpi_dev_add_driver_gpios(struct acpi_device *adev,
 			      const struct acpi_gpio_mapping *gpios)
 {
@@ -636,8 +632,6 @@  void gpiod_unexport(struct gpio_desc *desc);
 
 #else  /* CONFIG_GPIOLIB && CONFIG_GPIO_SYSFS */
 
-#include <asm/errno.h>
-
 static inline int gpiod_export(struct gpio_desc *desc,
 			       bool direction_may_change)
 {