From: Joanne Koong <joannelkoong@xxxxxxxxx>
Date: Sun, 6 Nov 2022 11:18:44 -0800
> On Thu, Nov 3, 2022 at 10:24 AM Kuniyuki Iwashima <kuniyu@xxxxxxxxxx> wrote:
> >
> > When connect() is called on a socket bound to the wildcard address,
> > we change the socket's saddr to a local address. If the socket
> > fails to connect() to the destination, we have to reset the saddr.
> >
> > However, when an error occurs after inet_hash6?_connect() in
> > (dccp|tcp)_v[46]_conect(), we forget to reset saddr and leave
> > the socket bound to the address.
> >
> > From the user's point of view, whether saddr is reset or not varies
> > with errno. Let's fix this inconsistent behaviour.
> >
> > Note that with this patch, the repro [0] will trigger the WARN_ON()
> > in inet_csk_get_port() again, but this patch is not buggy and rather
> > fixes a bug papering over the bhash2's bug [1] for which we need
> > another fix.
> >
> > For the record, the repro causes -EADDRNOTAVAIL in inet_hash6_connect()
> > by this sequence:
> >
> > s1 = socket()
> > s1.setsockopt(SOL_SOCKET, SO_REUSEADDR, 1)
> > s1.bind(('127.0.0.1', 10000))
> > s1.sendto(b'hello', MSG_FASTOPEN, (('127.0.0.1', 10000)))
> > # or s1.connect(('127.0.0.1', 10000))
> >
> > s2 = socket()
> > s2.setsockopt(SOL_SOCKET, SO_REUSEADDR, 1)
> > s2.bind(('0.0.0.0', 10000))
> > s2.connect(('127.0.0.1', 10000)) # -EADDRNOTAVAIL
> >
> > s2.listen(32) # WARN_ON(inet_csk(sk)->icsk_bind2_hash != tb2);
> >
> > [0]: https://syzkaller.appspot.com/bug?extid=015d756bbd1f8b5c8f09
> > [1]: https://lore.kernel.org/netdev/20221029001249.86337-1-kuniyu@xxxxxxxxxx/
> >
> > Fixes: 3df80d9320bc ("[DCCP]: Introduce DCCPv6")
> > Fixes: 7c657876b63c ("[DCCP]: Initial implementation")
> > Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> > Signed-off-by: Kuniyuki Iwashima <kuniyu@xxxxxxxxxx>
> > ---
> > net/dccp/ipv4.c | 2 ++
> > net/dccp/ipv6.c | 2 ++
> > net/ipv4/tcp_ipv4.c | 2 ++
> > net/ipv6/tcp_ipv6.c | 2 ++
> > 4 files changed, 8 insertions(+)
> >
> > diff --git a/net/dccp/ipv4.c b/net/dccp/ipv4.c
> > index 713b7b8dad7e..40640c26680e 100644
> > --- a/net/dccp/ipv4.c
> > +++ b/net/dccp/ipv4.c
> > @@ -157,6 +157,8 @@ int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
> > * This unhashes the socket and releases the local port, if necessary.
> > */
> > dccp_set_state(sk, DCCP_CLOSED);
> > + if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
> > + inet_reset_saddr(sk);
> > ip_rt_put(rt);
> > sk->sk_route_caps = 0;
> > inet->inet_dport = 0;
> > diff --git a/net/dccp/ipv6.c b/net/dccp/ipv6.c
> > index e57b43006074..626166cb6d7e 100644
> > --- a/net/dccp/ipv6.c
> > +++ b/net/dccp/ipv6.c
> > @@ -985,6 +985,8 @@ static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
> >
> > late_failure:
> > dccp_set_state(sk, DCCP_CLOSED);
> > + if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
> > + inet_reset_saddr(sk);
> > __sk_dst_reset(sk);
> > failure:
> > inet->inet_dport = 0;
> > diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
> > index 87d440f47a70..6a3a732b584d 100644
> > --- a/net/ipv4/tcp_ipv4.c
> > +++ b/net/ipv4/tcp_ipv4.c
> > @@ -343,6 +343,8 @@ int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
> > * if necessary.
> > */
> > tcp_set_state(sk, TCP_CLOSE);
> > + if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
> > + inet_reset_saddr(sk);
> > ip_rt_put(rt);
> > sk->sk_route_caps = 0;
> > inet->inet_dport = 0;
> > diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
> > index 2a3f9296df1e..81b396e5cf79 100644
> > --- a/net/ipv6/tcp_ipv6.c
> > +++ b/net/ipv6/tcp_ipv6.c
> > @@ -359,6 +359,8 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
> >
> > late_failure:
> > tcp_set_state(sk, TCP_CLOSE);
> > + if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
> > + inet_reset_saddr(sk);
> > failure:
> > inet->inet_dport = 0;
> > sk->sk_route_caps = 0;
> > --
> > 2.30.2
> >
>
> inet_reset_saddr() sets both inet_saddr and inet_rcv_saddr to 0, but I
> think there are some edge cases where when dccp/tcp_v4/6_connect() is
> called, inet_saddr is 0 but inet_rcv_saddr is not, which means we'd
> need to reset inet_rcv_saddr to its original value. The example case
> I'm looking at is __inet_bind() where if the request is to bind to a
> multicast address,
>
> inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
> if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
> inet->inet_saddr = 0; /* Use device */
Thanks for reviewing.
We have to take care of these two error paths.
* (dccp|tcp)_v[46]_coonnect()
* __inet_stream_connect()
In __inet_stream_connect(), we call ->disconnect(), which already has the
same logic in this patch.
In your edge case, once (dccp|tcp)_v[46]_coonnect() succeeds, both of
inet->inet_saddr and sk_rcv_saddr are non-zero. If connect() fails after
that, in ->disconnect(), we cannot know if we should restore sk_rcv_saddr
only. Also, we don't have the previous address there.
For these reasons, we reset both addresses only if the sk was bound to
INADDR_ANY, which we can detect by the SOCK_BINDADDR_LOCK flag.
As you mentinoed, we can restore sk_rcv_saddr for the edge case in
(dccp|tcp)_v[46]_coonnect() but cannot in __inet_stream_connect().
If we do so, we need another flag for the case and another member to save
the old multicast/broadcast address. (+ where we need rehash for bhash2)
What do you think ?
