Re: [PATCH bpf-next v3 2/3] bpf, sockmap: avoid using sk_socket after free when reading

[Jiayuan Chen]

March 20, 2025 at 08:34, "Cong Wang" <xiyou.wangcong@xxxxxxxxx> wrote:

> 
> On Mon, Mar 17, 2025 at 05:22:55PM +0800, Jiayuan Chen wrote:
> 
> > 
> > There are potential concurrency issues, as shown below.
> > 
> >  '''
> > 
> >  CPU0 CPU1
> > 
> >  sk_psock_verdict_data_ready:
> > 
> >  socket *sock = sk->sk_socket
> > 
> >  if (!sock) return
> > 
> >  close(fd):
> > 
> >  ...
> > 
> >  ops->release()
> > 
> >  if (!sock->ops) return
> > 
> >  sock->ops = NULL
> > 
> >  rcu_call(sock)
> > 
> >  free(sock)
> > 
> >  READ_ONCE(sock->ops)
> > 
> >  ^
> > 
> >  use 'sock' after free
> > 
> >  '''
> > 
> >  
> > 
> >  RCU is not applicable to Unix sockets read path, because the Unix socket
> > 
> >  implementation itself assumes it's always in process context and heavily
> > 
> >  uses mutex_lock, so, we can't call read_skb within rcu lock.
> > 
> 
> Hm, I guess the RCU work in sk_psock_drop() does not work for Unix
> 
> domain sockets either?
> 
> Thanks.
>

Although the Unix domain socket framework does not use RCU locks, the
entire sockmap process protects access to psock via RCU:
'''
rcu_read_lock();
psock = sk_psock(sk_other);
if (psock) {
 ...
}
rcu_read_unlock(); // `sk_psock_drop` will not execute until the unlock
'''

Therefore, I believe there are no issues with the psock operations here.

Thanks~