Re: [PATCH net] rxrpc: Fix a race between socket set up and I/O thread creation

[Paolo Abeni]

On 9/25/24 20:33, Simon Horman wrote:
> On Tue, Sep 24, 2024 at 11:08:01PM +0100, David Howells wrote:
> > In rxrpc_open_socket(), it sets up the socket and then sets up the I/O
> > thread that will handle it.  This is a problem, however, as there's a gap
> > between the two phases in which a packet may come into rxrpc_encap_rcv()
> > from the UDP packet but we oops when trying to wake the not-yet created I/O
> > thread.
> >
> > As a quick fix, just make rxrpc_encap_rcv() discard the packet if there's
> > no I/O thread yet.
> >
> > A better, but more intrusive fix would perhaps be to rearrange things such
> > that the socket creation is done by the I/O thread.
> >
> > Fixes: a275da62e8c1 ("rxrpc: Create a per-local endpoint receive queue and I/O thread")
> > Signed-off-by: David Howells <dhowells@xxxxxxxxxx>
> > Reviewed-by: Eric Dumazet <edumazet@xxxxxxxxxx>
>
> ...:wq
>
> > diff --git a/net/rxrpc/io_thread.c b/net/rxrpc/io_thread.c
> > index 0300baa9afcd..5c0a5374d51a 100644
> > --- a/net/rxrpc/io_thread.c
> > +++ b/net/rxrpc/io_thread.c
> > @@ -27,8 +27,9 @@ int rxrpc_encap_rcv(struct sock *udp_sk, struct sk_buff *skb)
> >   {
> >   	struct sk_buff_head *rx_queue;
> >   	struct rxrpc_local *local = rcu_dereference_sk_user_data(udp_sk);
> > +	struct task_struct *io_thread = READ_ONCE(local->io_thread);
>
> Hi David,
>
> The line above dereferences local.
> But the line below assumes that it may be NULL.
> This seems inconsistent.

sk->sk_user_data is cleared just before io_thread by rxrpc_io_thread(), 
I think accessing a NULL 'local' here should be possible.

@David, could you please respin addressing the above?

Thanks!

Paolo