[PATCH 1/1] checkpatch: Deprecate container_of() in favour of container_of_const()

[PATCH 1/1] checkpatch: Deprecate container_of() in favour of container_of_const()

[Sakari Ailus]

container_of() discards constness of its first argument and thus its use
may result in inadvertly unconstifying an object that is const. While
container_of_const() addresses this problem, the vast majority of new uses
of the two use the container_of() variant lacking the check.

Stem the flow of new container_of() use by adding it to the list of
deprecated functions in scripts/checkpatch.pl. Once all existing
container_of() users have been fixed to respect the constness attribute,
the check can be added to container_of() and we can switch back to
container_of() again.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
---
 scripts/checkpatch.pl | 1 +
 1 file changed, 1 insertion(+)

diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
index 664f7b7a622c..32edebfc1935 100755
--- a/scripts/checkpatch.pl
+++ b/scripts/checkpatch.pl
@@ -859,6 +859,7 @@ our %deprecated_apis = (
 	"kunmap_atomic"				=> "kunmap_local",
 	"srcu_read_lock_lite"			=> "srcu_read_lock_fast",
 	"srcu_read_unlock_lite"			=> "srcu_read_unlock_fast",
+	"container_of"				=> "container_of_const",
 );
 
 #Create a search pattern for all these strings to speed up a loop below
-- 
2.39.5

Re: [PATCH 1/1] checkpatch: Deprecate container_of() in favour of container_of_const()

[Greg Kroah-Hartman]

On Mon, Aug 11, 2025 at 11:57:31AM +0300, Sakari Ailus wrote:
> container_of() discards constness of its first argument and thus its use
> may result in inadvertly unconstifying an object that is const. While
> container_of_const() addresses this problem, the vast majority of new uses
> of the two use the container_of() variant lacking the check.
> 
> Stem the flow of new container_of() use by adding it to the list of
> deprecated functions in scripts/checkpatch.pl. Once all existing
> container_of() users have been fixed to respect the constness attribute,
> the check can be added to container_of() and we can switch back to
> container_of() again.
> 
> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
> ---
>  scripts/checkpatch.pl | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
> index 664f7b7a622c..32edebfc1935 100755
> --- a/scripts/checkpatch.pl
> +++ b/scripts/checkpatch.pl
> @@ -859,6 +859,7 @@ our %deprecated_apis = (
>  	"kunmap_atomic"				=> "kunmap_local",
>  	"srcu_read_lock_lite"			=> "srcu_read_lock_fast",
>  	"srcu_read_unlock_lite"			=> "srcu_read_unlock_fast",
> +	"container_of"				=> "container_of_const",
>  );
>  
>  #Create a search pattern for all these strings to speed up a loop below
> -- 
> 2.39.5
> 

Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Re: [PATCH 1/1] checkpatch: Deprecate container_of() in favour of container_of_const()

[Sakari Ailus]

Ping.

On Mon, Aug 11, 2025 at 11:57:31AM +0300, Sakari Ailus wrote:
> container_of() discards constness of its first argument and thus its use
> may result in inadvertly unconstifying an object that is const. While
> container_of_const() addresses this problem, the vast majority of new uses
> of the two use the container_of() variant lacking the check.
> 
> Stem the flow of new container_of() use by adding it to the list of
> deprecated functions in scripts/checkpatch.pl. Once all existing
> container_of() users have been fixed to respect the constness attribute,
> the check can be added to container_of() and we can switch back to
> container_of() again.
> 
> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
> ---
>  scripts/checkpatch.pl | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
> index 664f7b7a622c..32edebfc1935 100755
> --- a/scripts/checkpatch.pl
> +++ b/scripts/checkpatch.pl
> @@ -859,6 +859,7 @@ our %deprecated_apis = (
>  	"kunmap_atomic"				=> "kunmap_local",
>  	"srcu_read_lock_lite"			=> "srcu_read_lock_fast",
>  	"srcu_read_unlock_lite"			=> "srcu_read_unlock_fast",
> +	"container_of"				=> "container_of_const",
>  );
>  
>  #Create a search pattern for all these strings to speed up a loop below
> -- 
> 2.39.5
> 
> 

-- 
Sakari Ailus

Re: [PATCH 1/1] checkpatch: Deprecate container_of() in favour of container_of_const()

[Joe Perches]

On Wed, 2025-11-05 at 10:08 +0200, Sakari Ailus wrote:
> Ping.
> 
> On Mon, Aug 11, 2025 at 11:57:31AM +0300, Sakari Ailus wrote:
> > container_of() discards constness of its first argument and thus its use
> > may result in inadvertly unconstifying an object that is const. While
> > container_of_const() addresses this problem, the vast majority of new uses
> > of the two use the container_of() variant lacking the check.
> > 
> > Stem the flow of new container_of() use by adding it to the list of
> > deprecated functions in scripts/checkpatch.pl. Once all existing
> > container_of() users have been fixed to respect the constness attribute,
> > the check can be added to container_of() and we can switch back to
> > container_of() again.
> > 
> > Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
> > ---
> >  scripts/checkpatch.pl | 1 +
> >  1 file changed, 1 insertion(+)
> > 
> > diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
> > index 664f7b7a622c..32edebfc1935 100755
> > --- a/scripts/checkpatch.pl
> > +++ b/scripts/checkpatch.pl
> > @@ -859,6 +859,7 @@ our %deprecated_apis = (
> >  	"kunmap_atomic"				=> "kunmap_local",
> >  	"srcu_read_lock_lite"			=> "srcu_read_lock_fast",
> >  	"srcu_read_unlock_lite"			=> "srcu_read_unlock_fast",
> > +	"container_of"				=> "container_of_const",
> >  );
> >  
> >  #Create a search pattern for all these strings to speed up a loop below
> > -- 
> > 2.39.5
> > 
> > 

Re: [PATCH 1/1] checkpatch: Deprecate container_of() in favour of container_of_const()

[Christoph Hellwig]

Well, if container_of_const is a drop-in replacement for all uses of
container_of, just fix the implementation?  If it is an almost but not
complete replacement, find a different name for the few uses.  And if
it's not anywhere near drop in, just using container_of make a lot more
sense.

Just in case it wasn't clear:  We really should keep the shorter nice
name as the default if at all possible.  Both because it reads better
and because it avoids churn.

Re: [PATCH 1/1] checkpatch: Deprecate container_of() in favour of container_of_const()

[Sakari Ailus]

Hi Christoph,

On Wed, Nov 05, 2025 at 05:01:46AM -0800, Christoph Hellwig wrote:
> Well, if container_of_const is a drop-in replacement for all uses of
> container_of, just fix the implementation?  If it is an almost but not
> complete replacement, find a different name for the few uses.  And if
> it's not anywhere near drop in, just using container_of make a lot more
> sense.
> 
> Just in case it wasn't clear:  We really should keep the shorter nice
> name as the default if at all possible.  Both because it reads better
> and because it avoids churn.

I'm in favour of the container_of() name instead of container_of_const(),
but there are a vast number of uses of container_of() in the kernel and
fixing these to respect const qualifier is going to take time. The intent
here is to prevent adding new uses of container_of() that do not respect
constness.

I tested this again on v6.18-rc2 and with my .config (far from
allyesconfig) there were 840 warnings from discarding const qualifier.
Presumably a lot of these are duplicates, but then again it'll be a
thankless (and possibly endless?) task to fix these if we don't make effort
to avoid new ones from being added at the same time.

-- 
Kind regards,

Sakari Ailus