On 03/08/2019 09:17 AM, Eric Dumazet wrote:
>
>
> On 03/07/2019 11:25 PM, Xin Long wrote:
>> sk_setup_caps() is called to set sk->sk_dst_cache in pptp_connect,
>> so we have to dst_release(sk->sk_dst_cache) in pptp_sock_destruct,
>> otherwise, the dst refcnt will leak.
>>
>> It can be reproduced by this syz log:
>>
>> r1 = socket$pptp(0x18, 0x1, 0x2)
>> bind$pptp(r1, &(0x7f0000000100)={0x18, 0x2, {0x0, @local}}, 0x1e)
>> connect$pptp(r1, &(0x7f0000000000)={0x18, 0x2, {0x3, @remote}}, 0x1e)
>>
>> Consecutive dmesg warnings will occur:
>>
>> unregister_netdevice: waiting for lo to become free. Usage count = 1
>>
>> Fixes: 00959ade36ac ("PPTP: PPP over IPv4 (Point-to-Point Tunneling
>> Protocol)")
>> Reported-by: Xiumei Mu <x...@redhat.com>
>> Signed-off-by: Xin Long <lucien....@gmail.com>
>> ---
>> drivers/net/ppp/pptp.c | 1 +
>> 1 file changed, 1 insertion(+)
>>
>> diff --git a/drivers/net/ppp/pptp.c b/drivers/net/ppp/pptp.c
>> index 8f09edd..76172c2 100644
>> --- a/drivers/net/ppp/pptp.c
>> +++ b/drivers/net/ppp/pptp.c
>> @@ -532,6 +532,7 @@ static void pptp_sock_destruct(struct sock *sk)
>> pppox_unbind_sock(sk);
>> }
>> skb_queue_purge(&sk->sk_receive_queue);
>> + dst_release(rcu_dereference_check(sk->sk_dst_cache, 1));
>
> rcu_dereference_protected() is probably more appropriate. (No barrier is
> needed)
Before you ask, the first instance of this line probably came in commit
b6c6712a42ca3f9fa7f4a3d7c40e3a9dd1fd9e03 (net: sk_dst_cache RCUification)
but at that time rcu_dereference_protected() was not there yet.
We might cleanup this for all callers.
