[v2] netlink: Return unsigned value for nla_len()

  The return value from nla_len() is never expected to be negative, and can
never be more than struct nlattr::nla_len (a u16). Adjust the prototype
on the function. This will let GCC's value range optimization passes
know that the return can never be negative, and can never be larger than
u16. As recently discussed[1], this silences the following warning in
GCC 12+:

net/wireless/nl80211.c: In function 'nl80211_set_cqm_rssi.isra':
net/wireless/nl80211.c:12892:17: warning: 'memcpy' specified bound 18446744073709551615 exceeds maximum object size 9223372036854775807 [-Wstringop-overflow=]
12892 |                 memcpy(cqm_config->rssi_thresholds, thresholds,
      |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
12893 |                        flex_array_size(cqm_config, rssi_thresholds,
      |                        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
12894 |                                        n_thresholds));
      |                                        ~~~~~~~~~~~~~~

A future change would be to clamp the subtraction to make sure it never
wraps around if nla_len is somehow less than NLA_HDRLEN, which would
have the additional benefit of being defensive in the face of nlattr
corruption or logic errors.

Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202311090752.hWcJWAHL-lkp@intel.com/ [1]
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Johannes Berg <johannes@sipsolutions.net>
Cc: Jeff Johnson <quic_jjohnson@quicinc.com>
Cc: Michael Walle <mwalle@kernel.org>
Cc: Max Schulze <max.schulze@online.de>
Cc: netdev@vger.kernel.org
Cc: linux-wireless@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
---
 v2:
 - do not clamp return value (kuba)
 - adjust NLA_HDRLEN to be u16 also
 v1: https://lore.kernel.org/all/20231130200058.work.520-kees@kernel.org/
---
 include/net/netlink.h        | 2 +-
 include/uapi/linux/netlink.h | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
  

Le 02/12/2023 à 21:25, Kees Cook a écrit :
> The return value from nla_len() is never expected to be negative, and can
> never be more than struct nlattr::nla_len (a u16). Adjust the prototype
> on the function. This will let GCC's value range optimization passes
> know that the return can never be negative, and can never be larger than
> u16. As recently discussed[1], this silences the following warning in
> GCC 12+:
> 
> net/wireless/nl80211.c: In function 'nl80211_set_cqm_rssi.isra':
> net/wireless/nl80211.c:12892:17: warning: 'memcpy' specified bound 18446744073709551615 exceeds maximum object size 9223372036854775807 [-Wstringop-overflow=]
> 12892 |                 memcpy(cqm_config->rssi_thresholds, thresholds,
>       |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> 12893 |                        flex_array_size(cqm_config, rssi_thresholds,
>       |                        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> 12894 |                                        n_thresholds));
>       |                                        ~~~~~~~~~~~~~~
> 
> A future change would be to clamp the subtraction to make sure it never
> wraps around if nla_len is somehow less than NLA_HDRLEN, which would
> have the additional benefit of being defensive in the face of nlattr
> corruption or logic errors.
> 
> Reported-by: kernel test robot <lkp@intel.com>
> Closes: https://lore.kernel.org/oe-kbuild-all/202311090752.hWcJWAHL-lkp@intel.com/ [1]
> Cc: Jakub Kicinski <kuba@kernel.org>
> Cc: "David S. Miller" <davem@davemloft.net>
> Cc: Eric Dumazet <edumazet@google.com>
> Cc: Paolo Abeni <pabeni@redhat.com>
> Cc: Johannes Berg <johannes@sipsolutions.net>
> Cc: Jeff Johnson <quic_jjohnson@quicinc.com>
> Cc: Michael Walle <mwalle@kernel.org>
> Cc: Max Schulze <max.schulze@online.de>
> Cc: netdev@vger.kernel.org
> Cc: linux-wireless@vger.kernel.org
> Signed-off-by: Kees Cook <keescook@chromium.org>
> ---
>  v2:
>  - do not clamp return value (kuba)
>  - adjust NLA_HDRLEN to be u16 also
>  v1: https://lore.kernel.org/all/20231130200058.work.520-kees@kernel.org/
> ---
>  include/net/netlink.h        | 2 +-
>  include/uapi/linux/netlink.h | 2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/include/net/netlink.h b/include/net/netlink.h
> index 83bdf787aeee..7678a596a86b 100644
> --- a/include/net/netlink.h
> +++ b/include/net/netlink.h
> @@ -1200,7 +1200,7 @@ static inline void *nla_data(const struct nlattr *nla)
>   * nla_len - length of payload
>   * @nla: netlink attribute
>   */
> -static inline int nla_len(const struct nlattr *nla)
> +static inline u16 nla_len(const struct nlattr *nla)
>  {
>  	return nla->nla_len - NLA_HDRLEN;
>  }
> diff --git a/include/uapi/linux/netlink.h b/include/uapi/linux/netlink.h
> index f87aaf28a649..270feed9fd63 100644
> --- a/include/uapi/linux/netlink.h
> +++ b/include/uapi/linux/netlink.h
> @@ -247,7 +247,7 @@ struct nlattr {
>  
>  #define NLA_ALIGNTO		4
>  #define NLA_ALIGN(len)		(((len) + NLA_ALIGNTO - 1) & ~(NLA_ALIGNTO - 1))
> -#define NLA_HDRLEN		((int) NLA_ALIGN(sizeof(struct nlattr)))
> +#define NLA_HDRLEN		((__u16) NLA_ALIGN(sizeof(struct nlattr)))
I wonder if this may break the compilation of some userspace tools with errors
like comparing signed and unsigned values.


Regards,
Nicolas
  

On Mon, Dec 04, 2023 at 10:22:25AM +0100, Nicolas Dichtel wrote:
> Le 02/12/2023 à 21:25, Kees Cook a écrit :
> > The return value from nla_len() is never expected to be negative, and can
> > never be more than struct nlattr::nla_len (a u16). Adjust the prototype
> > on the function. This will let GCC's value range optimization passes
> > know that the return can never be negative, and can never be larger than
> > u16. As recently discussed[1], this silences the following warning in
> > GCC 12+:
> > 
> > net/wireless/nl80211.c: In function 'nl80211_set_cqm_rssi.isra':
> > net/wireless/nl80211.c:12892:17: warning: 'memcpy' specified bound 18446744073709551615 exceeds maximum object size 9223372036854775807 [-Wstringop-overflow=]
> > 12892 |                 memcpy(cqm_config->rssi_thresholds, thresholds,
> >       |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > 12893 |                        flex_array_size(cqm_config, rssi_thresholds,
> >       |                        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > 12894 |                                        n_thresholds));
> >       |                                        ~~~~~~~~~~~~~~
> > 
> > A future change would be to clamp the subtraction to make sure it never
> > wraps around if nla_len is somehow less than NLA_HDRLEN, which would
> > have the additional benefit of being defensive in the face of nlattr
> > corruption or logic errors.
> > 
> > Reported-by: kernel test robot <lkp@intel.com>
> > Closes: https://lore.kernel.org/oe-kbuild-all/202311090752.hWcJWAHL-lkp@intel.com/ [1]
> > Cc: Jakub Kicinski <kuba@kernel.org>
> > Cc: "David S. Miller" <davem@davemloft.net>
> > Cc: Eric Dumazet <edumazet@google.com>
> > Cc: Paolo Abeni <pabeni@redhat.com>
> > Cc: Johannes Berg <johannes@sipsolutions.net>
> > Cc: Jeff Johnson <quic_jjohnson@quicinc.com>
> > Cc: Michael Walle <mwalle@kernel.org>
> > Cc: Max Schulze <max.schulze@online.de>
> > Cc: netdev@vger.kernel.org
> > Cc: linux-wireless@vger.kernel.org
> > Signed-off-by: Kees Cook <keescook@chromium.org>
> > ---
> >  v2:
> >  - do not clamp return value (kuba)
> >  - adjust NLA_HDRLEN to be u16 also
> >  v1: https://lore.kernel.org/all/20231130200058.work.520-kees@kernel.org/
> > ---
> >  include/net/netlink.h        | 2 +-
> >  include/uapi/linux/netlink.h | 2 +-
> >  2 files changed, 2 insertions(+), 2 deletions(-)
> > 
> > diff --git a/include/net/netlink.h b/include/net/netlink.h
> > index 83bdf787aeee..7678a596a86b 100644
> > --- a/include/net/netlink.h
> > +++ b/include/net/netlink.h
> > @@ -1200,7 +1200,7 @@ static inline void *nla_data(const struct nlattr *nla)
> >   * nla_len - length of payload
> >   * @nla: netlink attribute
> >   */
> > -static inline int nla_len(const struct nlattr *nla)
> > +static inline u16 nla_len(const struct nlattr *nla)
> >  {
> >  	return nla->nla_len - NLA_HDRLEN;
> >  }
> > diff --git a/include/uapi/linux/netlink.h b/include/uapi/linux/netlink.h
> > index f87aaf28a649..270feed9fd63 100644
> > --- a/include/uapi/linux/netlink.h
> > +++ b/include/uapi/linux/netlink.h
> > @@ -247,7 +247,7 @@ struct nlattr {
> >  
> >  #define NLA_ALIGNTO		4
> >  #define NLA_ALIGN(len)		(((len) + NLA_ALIGNTO - 1) & ~(NLA_ALIGNTO - 1))
> > -#define NLA_HDRLEN		((int) NLA_ALIGN(sizeof(struct nlattr)))
> > +#define NLA_HDRLEN		((__u16) NLA_ALIGN(sizeof(struct nlattr)))
> I wonder if this may break the compilation of some userspace tools with errors
> like comparing signed and unsigned values.

Should I drop this part, then?
  

Le 04/12/2023 à 23:21, Kees Cook a écrit :
[snip]
>>> diff --git a/include/uapi/linux/netlink.h b/include/uapi/linux/netlink.h
>>> index f87aaf28a649..270feed9fd63 100644
>>> --- a/include/uapi/linux/netlink.h
>>> +++ b/include/uapi/linux/netlink.h
>>> @@ -247,7 +247,7 @@ struct nlattr {
>>>  
>>>  #define NLA_ALIGNTO		4
>>>  #define NLA_ALIGN(len)		(((len) + NLA_ALIGNTO - 1) & ~(NLA_ALIGNTO - 1))
>>> -#define NLA_HDRLEN		((int) NLA_ALIGN(sizeof(struct nlattr)))
>>> +#define NLA_HDRLEN		((__u16) NLA_ALIGN(sizeof(struct nlattr)))
>> I wonder if this may break the compilation of some userspace tools with errors
>> like comparing signed and unsigned values.
> 
> Should I drop this part, then?
> 
Yes please.


Thank you,
Nicolas