[PATCH] ipv4: ping: Fix potential use-after-free bug

[PATCH] ipv4: ping: Fix potential use-after-free bug

[Liang He]

In ping_unhash(), we should move sock_put(sk) after any possible
access point as the put function may free the object.

Fixes: c319b4d76b9e ("net: ipv4: add IPPROTO_ICMP socket kind")
Signed-off-by: Liang He <windhl@126.com>
---

 I have found other places containing similar code patterns.

 net/ipv4/ping.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/ipv4/ping.c b/net/ipv4/ping.c
index b83c2bd9d722..f90c86d37ffc 100644
--- a/net/ipv4/ping.c
+++ b/net/ipv4/ping.c
@@ -157,10 +157,10 @@ void ping_unhash(struct sock *sk)
 	spin_lock(&ping_table.lock);
 	if (sk_hashed(sk)) {
 		hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
-		sock_put(sk);
 		isk->inet_num = 0;
 		isk->inet_sport = 0;
 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
+		sock_put(sk);
 	}
 	spin_unlock(&ping_table.lock);
 }
-- 
2.25.1

Re: [PATCH] ipv4: ping: Fix potential use-after-free bug

[David Ahern]

On 9/16/22 4:07 AM, Liang He wrote:
> In ping_unhash(), we should move sock_put(sk) after any possible
> access point as the put function may free the object.

unhash handlers are called from sk_common_release which still has a
reference on the sock, so not really going to hit a UAF.

I do agree that it does not read correctly to 'put' a reference then
continue using the object. ie., the put should be moved to the end like
you have here. This is more of a tidiness exercise than a need to
backport to stable kernels.

> 
> Fixes: c319b4d76b9e ("net: ipv4: add IPPROTO_ICMP socket kind")
> Signed-off-by: Liang He <windhl@126.com>
> ---
> 
>  I have found other places containing similar code patterns.
> 
>  net/ipv4/ping.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/net/ipv4/ping.c b/net/ipv4/ping.c
> index b83c2bd9d722..f90c86d37ffc 100644
> --- a/net/ipv4/ping.c
> +++ b/net/ipv4/ping.c
> @@ -157,10 +157,10 @@ void ping_unhash(struct sock *sk)
>  	spin_lock(&ping_table.lock);
>  	if (sk_hashed(sk)) {
>  		hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
> -		sock_put(sk);
>  		isk->inet_num = 0;
>  		isk->inet_sport = 0;
>  		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
> +		sock_put(sk);
>  	}
>  	spin_unlock(&ping_table.lock);
>  }

Re:Re: [PATCH] ipv4: ping: Fix potential use-after-free bug

[Liang He]

At 2022-09-18 23:30:21, "David Ahern" <dsahern@kernel.org> wrote:
>On 9/16/22 4:07 AM, Liang He wrote:
>> In ping_unhash(), we should move sock_put(sk) after any possible
>> access point as the put function may free the object.
>
>unhash handlers are called from sk_common_release which still has a
>reference on the sock, so not really going to hit a UAF.
>

Thanks for this valuable lesson.

>I do agree that it does not read correctly to 'put' a reference then
>continue using the object. ie., the put should be moved to the end like
>you have here. This is more of a tidiness exercise than a need to
>backport to stable kernels.
>

OK, thanks.

>> 
>> Fixes: c319b4d76b9e ("net: ipv4: add IPPROTO_ICMP socket kind")
>> Signed-off-by: Liang He <windhl@126.com>
>> ---
>> 
>>  I have found other places containing similar code patterns.
>> 
>>  net/ipv4/ping.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>> 
>> diff --git a/net/ipv4/ping.c b/net/ipv4/ping.c
>> index b83c2bd9d722..f90c86d37ffc 100644
>> --- a/net/ipv4/ping.c
>> +++ b/net/ipv4/ping.c
>> @@ -157,10 +157,10 @@ void ping_unhash(struct sock *sk)
>>  	spin_lock(&ping_table.lock);
>>  	if (sk_hashed(sk)) {
>>  		hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
>> -		sock_put(sk);
>>  		isk->inet_num = 0;
>>  		isk->inet_sport = 0;
>>  		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
>> +		sock_put(sk);
>>  	}
>>  	spin_unlock(&ping_table.lock);
>>  }