[rft, PATCH v1 1/1] cpumask: Don't use "proxy" headers

[rft, PATCH v1 1/1] cpumask: Don't use "proxy" headers

[Andy Shevchenko]

Update header inclusions to follow IWYU (Include What You Use)
principle.

Note that kernel.h is discouraged to be included as it's written
at the top of that file.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---

Not fully compile tested, might give some compilation errors

 arch/x86/include/asm/cpumask.h |  2 ++
 include/linux/cpumask.h        | 12 +++++++-----
 2 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/arch/x86/include/asm/cpumask.h b/arch/x86/include/asm/cpumask.h
index 70f6b60ad67b..9df9e9cde670 100644
--- a/arch/x86/include/asm/cpumask.h
+++ b/arch/x86/include/asm/cpumask.h
@@ -2,6 +2,8 @@
 #ifndef _ASM_X86_CPUMASK_H
 #define _ASM_X86_CPUMASK_H
 #ifndef __ASSEMBLER__
+
+#include <linux/compiler.h>
 #include <linux/cpumask.h>
 
 extern void setup_cpu_local_masks(void);
diff --git a/include/linux/cpumask.h b/include/linux/cpumask.h
index 66694ee8d86e..6abf3d7e9435 100644
--- a/include/linux/cpumask.h
+++ b/include/linux/cpumask.h
@@ -7,14 +7,16 @@
  * set of CPUs in a system, one bit position per CPU number.  In general,
  * only nr_cpu_ids (<= NR_CPUS) bits are valid.
  */
-#include <linux/cleanup.h>
-#include <linux/kernel.h>
-#include <linux/bitmap.h>
-#include <linux/cpumask_types.h>
 #include <linux/atomic.h>
-#include <linux/bug.h>
+#include <linux/bitmap.h>
+#include <linux/cleanup.h>
+#include <linux/cpumask_types.h>
 #include <linux/gfp_types.h>
 #include <linux/numa.h>
+#include <linux/threads.h>
+#include <linux/types.h>
+
+#include <asm/bug.h>
 
 /**
  * cpumask_pr_args - printf args to output a cpumask
-- 
2.50.1

Re: [rft, PATCH v1 1/1] cpumask: Don't use "proxy" headers

[Yury Norov]

On Mon, Nov 24, 2025 at 09:39:59PM +0100, Andy Shevchenko wrote:
> Update header inclusions to follow IWYU (Include What You Use)
> principle.
> 
> Note that kernel.h is discouraged to be included as it's written
> at the top of that file.
> 
> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> ---
> 
> Not fully compile tested, might give some compilation errors
> 
>  arch/x86/include/asm/cpumask.h |  2 ++
>  include/linux/cpumask.h        | 12 +++++++-----
>  2 files changed, 9 insertions(+), 5 deletions(-)
> 
> diff --git a/arch/x86/include/asm/cpumask.h b/arch/x86/include/asm/cpumask.h
> index 70f6b60ad67b..9df9e9cde670 100644
> --- a/arch/x86/include/asm/cpumask.h
> +++ b/arch/x86/include/asm/cpumask.h
> @@ -2,6 +2,8 @@
>  #ifndef _ASM_X86_CPUMASK_H
>  #define _ASM_X86_CPUMASK_H
>  #ifndef __ASSEMBLER__
> +
> +#include <linux/compiler.h>
>  #include <linux/cpumask.h>
>  
>  extern void setup_cpu_local_masks(void);
> diff --git a/include/linux/cpumask.h b/include/linux/cpumask.h
> index 66694ee8d86e..6abf3d7e9435 100644
> --- a/include/linux/cpumask.h
> +++ b/include/linux/cpumask.h
> @@ -7,14 +7,16 @@
>   * set of CPUs in a system, one bit position per CPU number.  In general,
>   * only nr_cpu_ids (<= NR_CPUS) bits are valid.
>   */
> -#include <linux/cleanup.h>
> -#include <linux/kernel.h>

I recall, removing kernel.h from cpumasks was a troublesome exercise
back then. But it may be my false memory.

Added this for testing in bitmap-for-next.

Thanks,
Yury

> -#include <linux/bitmap.h>
> -#include <linux/cpumask_types.h>
>  #include <linux/atomic.h>
> -#include <linux/bug.h>
> +#include <linux/bitmap.h>
> +#include <linux/cleanup.h>
> +#include <linux/cpumask_types.h>
>  #include <linux/gfp_types.h>
>  #include <linux/numa.h>
> +#include <linux/threads.h>
> +#include <linux/types.h>
> +
> +#include <asm/bug.h>
>  
>  /**
>   * cpumask_pr_args - printf args to output a cpumask
> -- 
> 2.50.1

Re: [rft, PATCH v1 1/1] cpumask: Don't use "proxy" headers

[Andy Shevchenko]

On Mon, Nov 24, 2025 at 04:17:51PM -0500, Yury Norov wrote:
> On Mon, Nov 24, 2025 at 09:39:59PM +0100, Andy Shevchenko wrote:

> > Not fully compile tested, might give some compilation errors

...

> > -#include <linux/cleanup.h>
> > -#include <linux/kernel.h>
> 
> I recall, removing kernel.h from cpumasks was a troublesome exercise
> back then. But it may be my false memory.

Yeah, there are several headers that are too tangled together.
At least it builds on my configurations on x86_64.

> Added this for testing in bitmap-for-next.

Thanks, let's see how it will go. But note that it might fall the builds in
some cases.

-- 
With Best Regards,
Andy Shevchenko

Re: [rft, PATCH v1 1/1] cpumask: Don't use "proxy" headers

[Andy Shevchenko]

On Tue, Nov 25, 2025 at 12:42:23AM +0200, Andy Shevchenko wrote:
> On Mon, Nov 24, 2025 at 04:17:51PM -0500, Yury Norov wrote:
> > On Mon, Nov 24, 2025 at 09:39:59PM +0100, Andy Shevchenko wrote:
> 
> > > Not fully compile tested, might give some compilation errors

...

> > > -#include <linux/cleanup.h>
> > > -#include <linux/kernel.h>
> > 
> > I recall, removing kernel.h from cpumasks was a troublesome exercise
> > back then. But it may be my false memory.
> 
> Yeah, there are several headers that are too tangled together.
> At least it builds on my configurations on x86_64.
> 
> > Added this for testing in bitmap-for-next.
> 
> Thanks, let's see how it will go. But note that it might fall the builds in
> some cases.

So, we survived already a couple of Linux Next integration builds, I believe
we are quite okay with this change and if anything appears, it can be fixed
in a followups.

Thanks!

-- 
With Best Regards,
Andy Shevchenko

Re: [rft, PATCH v1 1/1] cpumask: Don't use "proxy" headers

[Yury Norov]

On Wed, Nov 26, 2025 at 09:59:16AM +0200, Andy Shevchenko wrote:
> On Tue, Nov 25, 2025 at 12:42:23AM +0200, Andy Shevchenko wrote:
> > On Mon, Nov 24, 2025 at 04:17:51PM -0500, Yury Norov wrote:
> > > On Mon, Nov 24, 2025 at 09:39:59PM +0100, Andy Shevchenko wrote:
> > 
> > > > Not fully compile tested, might give some compilation errors
> 
> ...
> 
> > > > -#include <linux/cleanup.h>
> > > > -#include <linux/kernel.h>
> > > 
> > > I recall, removing kernel.h from cpumasks was a troublesome exercise
> > > back then. But it may be my false memory.
> > 
> > Yeah, there are several headers that are too tangled together.
> > At least it builds on my configurations on x86_64.
> > 
> > > Added this for testing in bitmap-for-next.
> > 
> > Thanks, let's see how it will go. But note that it might fall the builds in
> > some cases.
> 
> So, we survived already a couple of Linux Next integration builds, I believe
> we are quite okay with this change and if anything appears, it can be fixed
> in a followups.

Yeah. I'll just keep the patch in my branch, so hopefully it will land
in mainline in the next merge cycle.

Thanks,
Yury