Cherry pick of minmax.h commits from 5.15.y

[PATCH 0/7 5.10.y] Cherry pick of minmax.h commits from 5.15.y

Posted by Eliav Farber 2 weeks, 1 day ago

This series backports seven commits from v5.15.y that update minmax.h
and related code:

 - ed6e37e30826 ("tracing: Define the is_signed_type() macro once")
 - 998f03984e25 ("minmax: sanity check constant bounds when clamping")
 - d470787b25e6 ("minmax: clamp more efficiently by avoiding extra
   comparison")
 - 1c2ee5bc9f11 ("minmax: fix header inclusions")
 - d53b5d862acd ("minmax: allow min()/max()/clamp() if the arguments
   have the same signedness.")
 - 7ed91c5560df ("minmax: allow comparisons of 'int' against 'unsigned
   char/short'")
 - 22f7794ef5a3 ("minmax: relax check to allow comparison between
   unsigned arguments and signed constants")

The main motivation is commit d53b5d862acd, which removes the strict
type check in min()/max() when both arguments have the same signedness.
Without this, kernel 5.10 builds can emit warnings that become build
failures when -Werror is used.

Additionally, commit ed6e37e30826 from tracing is required as a
dependency; without it, compilation fails.

Andy Shevchenko (1):
  minmax: fix header inclusions

Bart Van Assche (1):
  tracing: Define the is_signed_type() macro once

David Laight (3):
  minmax: allow min()/max()/clamp() if the arguments have the same
    signedness.
  minmax: allow comparisons of 'int' against 'unsigned char/short'
  minmax: relax check to allow comparison between unsigned arguments and
    signed constants

Jason A. Donenfeld (2):
  minmax: sanity check constant bounds when clamping
  minmax: clamp more efficiently by avoiding extra comparison

 include/linux/compiler.h     |  6 +++
 include/linux/minmax.h       | 89 ++++++++++++++++++++++++++----------
 include/linux/overflow.h     |  1 -
 include/linux/trace_events.h |  2 -
 4 files changed, 70 insertions(+), 28 deletions(-)

-- 
2.47.3

Re: [PATCH 0/7 5.10.y] Cherry pick of minmax.h commits from 5.15.y

Posted by David Laight 1 week, 6 days ago

On Tue, 16 Sep 2025 21:22:52 +0000
Eliav Farber <farbere@amazon.com> wrote:

> This series backports seven commits from v5.15.y that update minmax.h
> and related code:
> 
>  - ed6e37e30826 ("tracing: Define the is_signed_type() macro once")
>  - 998f03984e25 ("minmax: sanity check constant bounds when clamping")
>  - d470787b25e6 ("minmax: clamp more efficiently by avoiding extra
>    comparison")
>  - 1c2ee5bc9f11 ("minmax: fix header inclusions")
>  - d53b5d862acd ("minmax: allow min()/max()/clamp() if the arguments
>    have the same signedness.")
>  - 7ed91c5560df ("minmax: allow comparisons of 'int' against 'unsigned
>    char/short'")
>  - 22f7794ef5a3 ("minmax: relax check to allow comparison between
>    unsigned arguments and signed constants")

I think you need to pick up the later changes (from Linus) as well.
Without them nested min() and max() can generate very long lines from
the pre-processor (tens of megabytes) that cause very slow and/or
failing compilations on 32bit and other memory-limited systems.

There are a few other changes needed at the same time.
The current min() and max() can't be used in a few places because
they aren't 'constant enough' with constant arguments.

	David


> 
> The main motivation is commit d53b5d862acd, which removes the strict
> type check in min()/max() when both arguments have the same signedness.
> Without this, kernel 5.10 builds can emit warnings that become build
> failures when -Werror is used.
> 
> Additionally, commit ed6e37e30826 from tracing is required as a
> dependency; without it, compilation fails.
> 
> Andy Shevchenko (1):
>   minmax: fix header inclusions
> 
> Bart Van Assche (1):
>   tracing: Define the is_signed_type() macro once
> 
> David Laight (3):
>   minmax: allow min()/max()/clamp() if the arguments have the same
>     signedness.
>   minmax: allow comparisons of 'int' against 'unsigned char/short'
>   minmax: relax check to allow comparison between unsigned arguments and
>     signed constants
> 
> Jason A. Donenfeld (2):
>   minmax: sanity check constant bounds when clamping
>   minmax: clamp more efficiently by avoiding extra comparison
> 
>  include/linux/compiler.h     |  6 +++
>  include/linux/minmax.h       | 89 ++++++++++++++++++++++++++----------
>  include/linux/overflow.h     |  1 -
>  include/linux/trace_events.h |  2 -
>  4 files changed, 70 insertions(+), 28 deletions(-)
>

RE: [PATCH 0/7 5.10.y] Cherry pick of minmax.h commits from 5.15.y

Posted by Farber, Eliav 1 week, 5 days ago

> On Tue, 16 Sep 2025 21:22:52 +0000
> Eliav Farber <farbere@amazon.com> wrote:
>
> > This series backports seven commits from v5.15.y that update minmax.h 
> > and related code:
> >
> >  - ed6e37e30826 ("tracing: Define the is_signed_type() macro once")
> >  - 998f03984e25 ("minmax: sanity check constant bounds when clamping")
> >  - d470787b25e6 ("minmax: clamp more efficiently by avoiding extra
> >    comparison")
> >  - 1c2ee5bc9f11 ("minmax: fix header inclusions")
> >  - d53b5d862acd ("minmax: allow min()/max()/clamp() if the arguments
> >    have the same signedness.")
> >  - 7ed91c5560df ("minmax: allow comparisons of 'int' against 'unsigned
> >    char/short'")
> >  - 22f7794ef5a3 ("minmax: relax check to allow comparison between
> >    unsigned arguments and signed constants")
>
> I think you need to pick up the later changes (from Linus) as well.
> Without them nested min() and max() can generate very long lines from the pre-processor (tens of megabytes) that cause very slow and/or failing compilations on 32bit and other memory-limited systems.
>
> There are a few other changes needed at the same time.
> The current min() and max() can't be used in a few places because they aren't 'constant enough' with constant arguments.

I aligned minmax.h to include all changes in v6.17-rc6.
https://lore.kernel.org/stable/20250919101727.16152-1-farbere@amazon.com/T/#t

---
Regards, Eliav