To be able to constify instances of struct attribute it has to be
possible to add them to struct attribute_group.
The current type of the attrs member however is not compatible with that.
Introduce a union that allows registration of both const and non-const
attributes to enable a piecewise transition.
As both union member types are compatible no logic needs to be adapted.

Technically it is now possible register a const struct
attribute and receive it as mutable pointer in the callbacks.
This is a soundness issue.
But this same soundness issue already exists today in
sysfs_create_file().
Also the struct definition and callback implementation are always
closely linked and are meant to be moved to const in lockstep.

Similar to commit 906c508afdca ("sysfs: attribute_group: allow registration of const bin_attribute")

Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
---
 include/linux/sysfs.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/include/linux/sysfs.h b/include/linux/sysfs.h
index 0f2fcd244523f050c5286f19d4fe1846506f9214..f5e25bed777a6a6e717f10973f1abcd12111f5c5 100644
--- a/include/linux/sysfs.h
+++ b/include/linux/sysfs.h
@@ -105,7 +105,10 @@ struct attribute_group {
 	size_t			(*bin_size)(struct kobject *,
 					    const struct bin_attribute *,
 					    int);
-	struct attribute	**attrs;
+	union {
+		struct attribute	**attrs;
+		const struct attribute	*const *attrs_new;
+	};
 	union {
 		struct bin_attribute		**bin_attrs;
 		const struct bin_attribute	*const *bin_attrs_new;

-- 
2.48.1

On Thu, Jan 16, 2025 at 06:32:27PM +0100, Thomas Weißschuh wrote:
> To be able to constify instances of struct attribute it has to be
> possible to add them to struct attribute_group.
> The current type of the attrs member however is not compatible with that.
> Introduce a union that allows registration of both const and non-const
> attributes to enable a piecewise transition.
> As both union member types are compatible no logic needs to be adapted.
> 
> Technically it is now possible register a const struct
> attribute and receive it as mutable pointer in the callbacks.
> This is a soundness issue.
> But this same soundness issue already exists today in
> sysfs_create_file().
> Also the struct definition and callback implementation are always
> closely linked and are meant to be moved to const in lockstep.
> 
> Similar to commit 906c508afdca ("sysfs: attribute_group: allow registration of const bin_attribute")
> 
> Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
> ---
>  include/linux/sysfs.h | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/include/linux/sysfs.h b/include/linux/sysfs.h
> index 0f2fcd244523f050c5286f19d4fe1846506f9214..f5e25bed777a6a6e717f10973f1abcd12111f5c5 100644
> --- a/include/linux/sysfs.h
> +++ b/include/linux/sysfs.h
> @@ -105,7 +105,10 @@ struct attribute_group {
>  	size_t			(*bin_size)(struct kobject *,
>  					    const struct bin_attribute *,
>  					    int);
> -	struct attribute	**attrs;
> +	union {
> +		struct attribute	**attrs;
> +		const struct attribute	*const *attrs_new;
> +	};

I'm all for the idea, BUT, let's finish up doing this one:

>  	union {
>  		struct bin_attribute		**bin_attrs;
>  		const struct bin_attribute	*const *bin_attrs_new;

first please.

That way we can see just how "easy" the switch from _new to not-new goes :)

thanks,

greg k-h

Hi Greg,

On 2025-01-17 08:01:00+0100, Greg Kroah-Hartman wrote:
> On Thu, Jan 16, 2025 at 06:32:27PM +0100, Thomas Weißschuh wrote:
> > To be able to constify instances of struct attribute it has to be
> > possible to add them to struct attribute_group.
> > The current type of the attrs member however is not compatible with that.
> > Introduce a union that allows registration of both const and non-const
> > attributes to enable a piecewise transition.
> > As both union member types are compatible no logic needs to be adapted.
> > 
> > Technically it is now possible register a const struct
> > attribute and receive it as mutable pointer in the callbacks.
> > This is a soundness issue.
> > But this same soundness issue already exists today in
> > sysfs_create_file().
> > Also the struct definition and callback implementation are always
> > closely linked and are meant to be moved to const in lockstep.
> > 
> > Similar to commit 906c508afdca ("sysfs: attribute_group: allow registration of const bin_attribute")
> > 
> > Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
> > ---
> >  include/linux/sysfs.h | 5 ++++-
> >  1 file changed, 4 insertions(+), 1 deletion(-)
> > 
> > diff --git a/include/linux/sysfs.h b/include/linux/sysfs.h
> > index 0f2fcd244523f050c5286f19d4fe1846506f9214..f5e25bed777a6a6e717f10973f1abcd12111f5c5 100644
> > --- a/include/linux/sysfs.h
> > +++ b/include/linux/sysfs.h
> > @@ -105,7 +105,10 @@ struct attribute_group {
> >  	size_t			(*bin_size)(struct kobject *,
> >  					    const struct bin_attribute *,
> >  					    int);
> > -	struct attribute	**attrs;
> > +	union {
> > +		struct attribute	**attrs;
> > +		const struct attribute	*const *attrs_new;
> > +	};
> 
> I'm all for the idea, BUT, let's finish up doing this one:
> 
> >  	union {
> >  		struct bin_attribute		**bin_attrs;
> >  		const struct bin_attribute	*const *bin_attrs_new;
> 
> first please.
> 
> That way we can see just how "easy" the switch from _new to not-new goes :)

I'd like to resend these preparatory patches so they go into v6.17-rc1
and I can work on the follow-up changes.
In my opinion the switch from _new will work nicely. There have been no
new users of _new in -next at all.

Any objections?


Thomas

On Sat, Jun 28, 2025 at 10:19:07AM +0200, Thomas Weißschuh wrote:
> Hi Greg,
> 
> On 2025-01-17 08:01:00+0100, Greg Kroah-Hartman wrote:
> > On Thu, Jan 16, 2025 at 06:32:27PM +0100, Thomas Weißschuh wrote:
> > > To be able to constify instances of struct attribute it has to be
> > > possible to add them to struct attribute_group.
> > > The current type of the attrs member however is not compatible with that.
> > > Introduce a union that allows registration of both const and non-const
> > > attributes to enable a piecewise transition.
> > > As both union member types are compatible no logic needs to be adapted.
> > > 
> > > Technically it is now possible register a const struct
> > > attribute and receive it as mutable pointer in the callbacks.
> > > This is a soundness issue.
> > > But this same soundness issue already exists today in
> > > sysfs_create_file().
> > > Also the struct definition and callback implementation are always
> > > closely linked and are meant to be moved to const in lockstep.
> > > 
> > > Similar to commit 906c508afdca ("sysfs: attribute_group: allow registration of const bin_attribute")
> > > 
> > > Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
> > > ---
> > >  include/linux/sysfs.h | 5 ++++-
> > >  1 file changed, 4 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/include/linux/sysfs.h b/include/linux/sysfs.h
> > > index 0f2fcd244523f050c5286f19d4fe1846506f9214..f5e25bed777a6a6e717f10973f1abcd12111f5c5 100644
> > > --- a/include/linux/sysfs.h
> > > +++ b/include/linux/sysfs.h
> > > @@ -105,7 +105,10 @@ struct attribute_group {
> > >  	size_t			(*bin_size)(struct kobject *,
> > >  					    const struct bin_attribute *,
> > >  					    int);
> > > -	struct attribute	**attrs;
> > > +	union {
> > > +		struct attribute	**attrs;
> > > +		const struct attribute	*const *attrs_new;
> > > +	};
> > 
> > I'm all for the idea, BUT, let's finish up doing this one:
> > 
> > >  	union {
> > >  		struct bin_attribute		**bin_attrs;
> > >  		const struct bin_attribute	*const *bin_attrs_new;
> > 
> > first please.
> > 
> > That way we can see just how "easy" the switch from _new to not-new goes :)
> 
> I'd like to resend these preparatory patches so they go into v6.17-rc1
> and I can work on the follow-up changes.
> In my opinion the switch from _new will work nicely. There have been no
> new users of _new in -next at all.
> 
> Any objections?

Sure, please do.

greg k-h