container_of: add type safety

[PATCH] container_of: add type safety

Posted by Michael S. Tsirkin 2 years, 6 months ago

Using a wrong member in container_of will result in an error.
No so for container_of_const - it is just a cast so will
happily give you a wrong pointer.

Use logic from container_of to add safety.

Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Sakari Ailus <sakari.ailus@linux.intel.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Sakari Ailus <sakari.ailus@linux.intel.com>
Cc: Rafael J. Wysocki <rafael@kernel.org>
Link: https://lore.kernel.org/r/20221205121206.166576-1-gregkh@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
 include/linux/container_of.h | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/include/linux/container_of.h b/include/linux/container_of.h
index 1d898f9158b4..5d87faf72e0a 100644
--- a/include/linux/container_of.h
+++ b/include/linux/container_of.h
@@ -29,10 +29,13 @@
  * @type:		the type of the container struct this is embedded in.
  * @member:		the name of the member within the struct.
  */
-#define container_of_const(ptr, type, member)				\
+#define container_of_const(ptr, type, member) ({			\
+	static_assert(__same_type(*(ptr), ((type *)0)->member) ||	\
+		      __same_type(*(ptr), void),			\
+		      "pointer type mismatch in container_of()");	\
 	_Generic(ptr,							\
 		const typeof(*(ptr)) *: ((const type *)container_of(ptr, type, member)),\
 		default: ((type *)container_of(ptr, type, member))	\
-	)
+	); })
 
 #endif	/* _LINUX_CONTAINER_OF_H */
-- 
MST

Re: [PATCH] container_of: add type safety

Posted by Greg Kroah-Hartman 2 years, 6 months ago

On Wed, Mar 08, 2023 at 12:35:03PM -0500, Michael S. Tsirkin wrote:
> Using a wrong member in container_of will result in an error.
> No so for container_of_const - it is just a cast so will
> happily give you a wrong pointer.
> 
> Use logic from container_of to add safety.

Also your subject line is wrong :(

Re: [PATCH] container_of: add type safety

Posted by Greg Kroah-Hartman 2 years, 6 months ago

On Wed, Mar 08, 2023 at 12:35:03PM -0500, Michael S. Tsirkin wrote:
> Using a wrong member in container_of will result in an error.
> No so for container_of_const - it is just a cast so will
> happily give you a wrong pointer.
> 
> Use logic from container_of to add safety.
> 
> Cc: Jason Gunthorpe <jgg@ziepe.ca>
> Cc: Sakari Ailus <sakari.ailus@linux.intel.com>
> Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
> Cc: Jason Gunthorpe <jgg@nvidia.com>
> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> Cc: Sakari Ailus <sakari.ailus@linux.intel.com>
> Cc: Rafael J. Wysocki <rafael@kernel.org>
> Link: https://lore.kernel.org/r/20221205121206.166576-1-gregkh@linuxfoundation.org

That's the wrong link, that's not this patch, that was an old patch.

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

I did not sign off on this.  NEVER do that, you just made a legal
statement in my name, why?

Why did you not sign off on it?

totally confused...

> ---
>  include/linux/container_of.h | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/include/linux/container_of.h b/include/linux/container_of.h
> index 1d898f9158b4..5d87faf72e0a 100644
> --- a/include/linux/container_of.h
> +++ b/include/linux/container_of.h
> @@ -29,10 +29,13 @@
>   * @type:		the type of the container struct this is embedded in.
>   * @member:		the name of the member within the struct.
>   */
> -#define container_of_const(ptr, type, member)				\
> +#define container_of_const(ptr, type, member) ({			\
> +	static_assert(__same_type(*(ptr), ((type *)0)->member) ||	\
> +		      __same_type(*(ptr), void),			\
> +		      "pointer type mismatch in container_of()");	\

Why is this needed because:


>  	_Generic(ptr,							\
>  		const typeof(*(ptr)) *: ((const type *)container_of(ptr, type, member)),\
>  		default: ((type *)container_of(ptr, type, member))	\

container_of() is used here, so shouldn't the assert trigger there if
you get things wrong?

So why is this change needed at all?

thanks,

greg k-h

Re: [PATCH] container_of: add type safety

Posted by Michael S. Tsirkin 2 years, 6 months ago

On Wed, Mar 08, 2023 at 06:57:13PM +0100, Greg Kroah-Hartman wrote:
> On Wed, Mar 08, 2023 at 12:35:03PM -0500, Michael S. Tsirkin wrote:
> > Using a wrong member in container_of will result in an error.
> > No so for container_of_const - it is just a cast so will
> > happily give you a wrong pointer.
> > 
> > Use logic from container_of to add safety.
> > 
> > Cc: Jason Gunthorpe <jgg@ziepe.ca>
> > Cc: Sakari Ailus <sakari.ailus@linux.intel.com>
> > Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
> > Cc: Jason Gunthorpe <jgg@nvidia.com>
> > Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> > Cc: Sakari Ailus <sakari.ailus@linux.intel.com>
> > Cc: Rafael J. Wysocki <rafael@kernel.org>
> > Link: https://lore.kernel.org/r/20221205121206.166576-1-gregkh@linuxfoundation.org
> 
> That's the wrong link, that's not this patch, that was an old patch.
> 
> > Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> 
> I did not sign off on this.  NEVER do that, you just made a legal
> statement in my name, why?
> 
> Why did you not sign off on it?
> 
> totally confused...

Ooops. I started with people who signed off on the original one
and replaced S.O.B with a CC. Forgot to replace it for yours
and then the script checking patches saw a signoff and was happy
so I didn't notice I forgot to sign it myself ...

However ...


> > ---
> >  include/linux/container_of.h | 7 +++++--
> >  1 file changed, 5 insertions(+), 2 deletions(-)
> > 
> > diff --git a/include/linux/container_of.h b/include/linux/container_of.h
> > index 1d898f9158b4..5d87faf72e0a 100644
> > --- a/include/linux/container_of.h
> > +++ b/include/linux/container_of.h
> > @@ -29,10 +29,13 @@
> >   * @type:		the type of the container struct this is embedded in.
> >   * @member:		the name of the member within the struct.
> >   */
> > -#define container_of_const(ptr, type, member)				\
> > +#define container_of_const(ptr, type, member) ({			\
> > +	static_assert(__same_type(*(ptr), ((type *)0)->member) ||	\
> > +		      __same_type(*(ptr), void),			\
> > +		      "pointer type mismatch in container_of()");	\
> 
> Why is this needed because:
> 
> 
> >  	_Generic(ptr,							\
> >  		const typeof(*(ptr)) *: ((const type *)container_of(ptr, type, member)),\
> >  		default: ((type *)container_of(ptr, type, member))	\
> 
> container_of() is used here, so shouldn't the assert trigger there if
> you get things wrong?
> 
> So why is this change needed at all?
> 
> thanks,
> 
> greg k-h


Hmm. I think I was confused. Error did not seem to trigger but
I tried again and it does. Ignore this please.

Sorry about the noise.

-- 
MST