[PATCH bpf v4 4/5] bpf, sockmap: Fix af_unix null-ptr-deref in proto update

[Michal Luczaj]

unix_stream_connect() sets sk_state (`WRITE_ONCE(sk->sk_state,
TCP_ESTABLISHED)`) _before_ it assigns a peer (`unix_peer(sk) = newsk`).
sk_state == TCP_ESTABLISHED makes sock_map_sk_state_allowed() believe that
socket is properly set up, which would include having a defined peer. IOW,
there's a window when unix_stream_bpf_update_proto() can be called on
socket which still has unix_peer(sk) == NULL.

         CPU0 bpf                            CPU1 connect
         --------                            ------------

                                WRITE_ONCE(sk->sk_state, TCP_ESTABLISHED)
sock_map_sk_state_allowed(sk)
...
sk_pair = unix_peer(sk)
sock_hold(sk_pair)
                                sock_hold(newsk)
                                smp_mb__after_atomic()
                                unix_peer(sk) = newsk

BUG: kernel NULL pointer dereference, address: 0000000000000080
RIP: 0010:unix_stream_bpf_update_proto+0xa0/0x1b0
Call Trace:
  sock_map_link+0x564/0x8b0
  sock_map_update_common+0x6e/0x340
  sock_map_update_elem_sys+0x17d/0x240
  __sys_bpf+0x26db/0x3250
  __x64_sys_bpf+0x21/0x30
  do_syscall_64+0x6b/0x3a0
  entry_SYSCALL_64_after_hwframe+0x76/0x7e

Initial idea was to move peer assignment _before_ the sk_state update[1],
but that involved an additional memory barrier, and changing the hot path
was rejected.
Then a NULL check during proto update in unix_stream_bpf_update_proto() was
considered[2], but the follow-up discussion[3] focused on the root cause,
i.e. sockmap update taking a wrong lock. Or, more specifically, missing
unix_state_lock()[4].
In the end it was concluded that teaching sockmap about the af_unix locking
would be unnecessarily complex[5].
Complexity aside, since BPF_PROG_TYPE_SCHED_CLS and BPF_PROG_TYPE_SCHED_ACT
are allowed to update sockmaps, sock_map_update_elem() taking the unix
lock, as it is currently implemented in unix_state_lock():
spin_lock(&unix_sk(s)->lock), would be problematic. unix_state_lock() taken
in a process context, followed by a softirq-context TC BPF program
attempting to take the same spinlock -- deadlock[6].
This way we circled back to the peer check idea[2].

[1]: https://lore.kernel.org/netdev/ba5c50aa-1df4-40c2-ab33-a72022c5a32e@xxxxxxx/
[2]: https://lore.kernel.org/netdev/20240610174906.32921-1-kuniyu@xxxxxxxxxx/
[3]: https://lore.kernel.org/netdev/7603c0e6-cd5b-452b-b710-73b64bd9de26@xxxxxxxxx/
[4]: https://lore.kernel.org/netdev/CAAVpQUA+8GL_j63CaKb8hbxoL21izD58yr1NvhOhU=j+35+3og@xxxxxxxxxxxxxx/
[5]: https://lore.kernel.org/bpf/CAAVpQUAHijOMext28Gi10dSLuMzGYh+jK61Ujn+fZ-wvcODR2A@xxxxxxxxxxxxxx/
[6]: https://lore.kernel.org/bpf/dd043c69-4d03-46fe-8325-8f97101435cf@xxxxxxxxx/

Summary of scenarios where af_unix/stream connect() may race a sockmap
update:

1. connect() vs. bpf(BPF_MAP_UPDATE_ELEM), i.e. sock_map_update_elem_sys()

   Implemented NULL check is sufficient. Once assigned, socket peer won't
   be released until socket fd is released. And that's not an issue because
   sock_map_update_elem_sys() bumps fd refcnf.

2. connect() vs BPF program doing update

   Update restricted per verifier.c:may_update_sockmap() to

      BPF_PROG_TYPE_TRACING/BPF_TRACE_ITER
      BPF_PROG_TYPE_SOCK_OPS (bpf_sock_map_update() only)
      BPF_PROG_TYPE_SOCKET_FILTER
      BPF_PROG_TYPE_SCHED_CLS
      BPF_PROG_TYPE_SCHED_ACT
      BPF_PROG_TYPE_XDP
      BPF_PROG_TYPE_SK_REUSEPORT
      BPF_PROG_TYPE_FLOW_DISSECTOR
      BPF_PROG_TYPE_SK_LOOKUP

   Plus one more race to consider:

            CPU0 bpf                            CPU1 connect
            --------                            ------------

                                   WRITE_ONCE(sk->sk_state, TCP_ESTABLISHED)
   sock_map_sk_state_allowed(sk)
                                   sock_hold(newsk)
                                   smp_mb__after_atomic()
                                   unix_peer(sk) = newsk
   sk_pair = unix_peer(sk)
   if (unlikely(!sk_pair))
      return -EINVAL;

                                                 CPU1 close
                                                 ----------

                                   skpair = unix_peer(sk);
                                   unix_peer(sk) = NULL;
                                   sock_put(skpair)
   // use after free?
   sock_hold(sk_pair)

   2.1 BPF program invoking helper function bpf_sock_map_update() ->
       BPF_CALL_4(bpf_sock_map_update(), ...)

       Helper limited to BPF_PROG_TYPE_SOCK_OPS. Nevertheless, a unix sock
       might be accessible via bpf_map_lookup_elem(). Which implies sk
       already having psock, which in turn implies sk already having
       sk_pair. Since sk_psock_destroy() is queued as RCU work, sk_pair
       won't go away while BPF executes the update.

   2.2 BPF program invoking helper function bpf_map_update_elem() ->
       sock_map_update_elem()

       2.2.1 Unix sock accessible to BPF prog only via sockmap lookup in
             BPF_PROG_TYPE_SOCKET_FILTER, BPF_PROG_TYPE_SCHED_CLS,
             BPF_PROG_TYPE_SCHED_ACT, BPF_PROG_TYPE_XDP,
             BPF_PROG_TYPE_SK_REUSEPORT, BPF_PROG_TYPE_FLOW_DISSECTOR,
             BPF_PROG_TYPE_SK_LOOKUP.

             Pretty much the same as case 2.1.

       2.2.2 Unix sock accessible to BPF program directly:
             BPF_PROG_TYPE_TRACING, narrowed down to BPF_TRACE_ITER.

             Sockmap iterator (sock_map_seq_ops) is safe: unix sock
             residing in a sockmap means that the sock already went through
             the proto update step.

             Unix sock iterator (bpf_iter_unix_seq_ops), on the other hand,
             gives access to socks that may still be unconnected. Which
             means iterator prog can race sockmap/proto update against
             connect().

             BUG: KASAN: null-ptr-deref in unix_stream_bpf_update_proto+0x253/0x4d0
             Write of size 4 at addr 0000000000000080 by task test_progs/3140
             Call Trace:
              dump_stack_lvl+0x5d/0x80
              kasan_report+0xe4/0x1c0
              kasan_check_range+0x125/0x200
              unix_stream_bpf_update_proto+0x253/0x4d0
              sock_map_link+0x71c/0xec0
              sock_map_update_common+0xbc/0x600
              sock_map_update_elem+0x19a/0x1f0
              bpf_prog_bbbf56096cdd4f01_selective_dump_unix+0x20c/0x217
              bpf_iter_run_prog+0x21e/0xae0
              bpf_iter_unix_seq_show+0x1e0/0x2a0
              bpf_seq_read+0x42c/0x10d0
              vfs_read+0x171/0xb20
              ksys_read+0xff/0x200
              do_syscall_64+0xf7/0x5e0
              entry_SYSCALL_64_after_hwframe+0x76/0x7e

             While the introduced NULL check prevents null-ptr-deref in the
             BPF program path as well, it is insufficient to guard against
             a poorly timed close() leading to a use-after-free. This will
             be addressed in a subsequent patch.

Reported-by: Michal Luczaj <mhal@xxxxxxx>
Closes: https://lore.kernel.org/netdev/ba5c50aa-1df4-40c2-ab33-a72022c5a32e@xxxxxxx/
Reported-by: 钱一铭 <yimingqian591@xxxxxxxxx>
Suggested-by: Kuniyuki Iwashima <kuniyu@xxxxxxxxxx>
Suggested-by: Martin KaFai Lau <martin.lau@xxxxxxxxx>
Fixes: c63829182c37 ("af_unix: Implement ->psock_update_sk_prot()")
Signed-off-by: Michal Luczaj <mhal@xxxxxxx>
---
 net/unix/unix_bpf.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/net/unix/unix_bpf.c b/net/unix/unix_bpf.c
index e0d30d6d22ac..57f3124c9d8d 100644
--- a/net/unix/unix_bpf.c
+++ b/net/unix/unix_bpf.c
@@ -185,6 +185,9 @@ int unix_stream_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool r
 	 */
 	if (!psock->sk_pair) {
 		sk_pair = unix_peer(sk);
+		if (unlikely(!sk_pair))
+			return -EINVAL;
+
 		sock_hold(sk_pair);
 		psock->sk_pair = sk_pair;
 	}

-- 
2.53.0