Instead of a heap allocation use a stack allocated sockaddr_storage to
support arbitrary length addr_len value (but bounds check it against the
maximum address length).

Signed-off-by: Kees Cook <kees@kernel.org>
---
 net/core/rtnetlink.c | 12 ++++--------
 1 file changed, 4 insertions(+), 8 deletions(-)

diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index f0a520987085..eddd10b74f06 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -2839,21 +2839,17 @@ static int do_setlink(const struct sk_buff *skb,
 	}
 
 	if (tb[IFLA_ADDRESS]) {
-		struct sockaddr *sa;
-		int len;
+		struct sockaddr_storage addr;
+		struct sockaddr *sa = (struct sockaddr *)&addr;
 
-		len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
-						  sizeof(*sa));
-		sa = kmalloc(len, GFP_KERNEL);
-		if (!sa) {
+		if (dev->addr_len > sizeof(addr.__data)) {
 			err = -ENOMEM;
 			goto errout;
 		}
 		sa->sa_family = dev->type;
-		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
+		memcpy(addr.__data, nla_data(tb[IFLA_ADDRESS]),
 		       dev->addr_len);
 		err = dev_set_mac_address_user(dev, sa, extack);
-		kfree(sa);
 		if (err)
 			goto errout;
 		status |= DO_SETLINK_MODIFIED;
-- 
2.34.1

On 11/4/24 11:25 PM, Kees Cook wrote:
> Instead of a heap allocation use a stack allocated sockaddr_storage to
> support arbitrary length addr_len value (but bounds check it against the
> maximum address length).
>
> Signed-off-by: Kees Cook <kees@kernel.org>
> ---
>   net/core/rtnetlink.c | 12 ++++--------
>   1 file changed, 4 insertions(+), 8 deletions(-)
>
> diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
> index f0a520987085..eddd10b74f06 100644
> --- a/net/core/rtnetlink.c
> +++ b/net/core/rtnetlink.c
> @@ -2839,21 +2839,17 @@ static int do_setlink(const struct sk_buff *skb,
>   	}
>   
>   	if (tb[IFLA_ADDRESS]) {
> -		struct sockaddr *sa;
> -		int len;
> +		struct sockaddr_storage addr;
> +		struct sockaddr *sa = (struct sockaddr *)&addr;


We already use too much stack space.


Please move addr into struct rtnl_newlink_tbs ?


>   
> -		len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
> -						  sizeof(*sa));
> -		sa = kmalloc(len, GFP_KERNEL);
> -		if (!sa) {
> +		if (dev->addr_len > sizeof(addr.__data)) {
>   			err = -ENOMEM;
>   			goto errout;
>   		}
>   		sa->sa_family = dev->type;
> -		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
> +		memcpy(addr.__data, nla_data(tb[IFLA_ADDRESS]),
>   		       dev->addr_len);
>   		err = dev_set_mac_address_user(dev, sa, extack);
> -		kfree(sa);
>   		if (err)
>   			goto errout;
>   		status |= DO_SETLINK_MODIFIED;

On Tue, Nov 05, 2024 at 11:59:49AM +0100, Eric Dumazet wrote:
> 
> On 11/4/24 11:25 PM, Kees Cook wrote:
> > Instead of a heap allocation use a stack allocated sockaddr_storage to
> > support arbitrary length addr_len value (but bounds check it against the
> > maximum address length).
> > 
> > Signed-off-by: Kees Cook <kees@kernel.org>
> > ---
> >   net/core/rtnetlink.c | 12 ++++--------
> >   1 file changed, 4 insertions(+), 8 deletions(-)
> > 
> > diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
> > index f0a520987085..eddd10b74f06 100644
> > --- a/net/core/rtnetlink.c
> > +++ b/net/core/rtnetlink.c
> > @@ -2839,21 +2839,17 @@ static int do_setlink(const struct sk_buff *skb,
> >   	}
> >   	if (tb[IFLA_ADDRESS]) {
> > -		struct sockaddr *sa;
> > -		int len;
> > +		struct sockaddr_storage addr;
> > +		struct sockaddr *sa = (struct sockaddr *)&addr;
> 
> We already use too much stack space.

I'm finally coming back to this. I was over-thinking: this only calls
dev_set_mac_address_user() so it really is a true struct sockaddr.

> Please move addr into struct rtnl_newlink_tbs ?

I couldn't figure out how that was meant to work -- I didn't see a
struct rtnl_newlink_tbs instance near this routine.

Regardless, I can just tweak the length and leave it heap allocated.

-Kees

-- 
Kees Cook