[PATCH bpf] bpf, sockmap: Reject sk_msg egress redirects to non-TCP sockets

From: Jakub Sitnicki <jakub@cloudflare.com>
To: bpf@vger.kernel.org
Cc: netdev@vger.kernel.org, kernel-team@cloudflare.com,
	Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Andrii Nakryiko <andrii@kernel.org>,
	John Fastabend <john.fastabend@gmail.com>,
	Cong Wang <cong.wang@bytedance.com>
Subject: [PATCH bpf] bpf, sockmap: Reject sk_msg egress redirects to non-TCP sockets
Date: Wed, 20 Sep 2023 12:20:55 +0200	[thread overview]
Message-ID: <20230920102055.42662-1-jakub@cloudflare.com> (raw)

With a SOCKMAP/SOCKHASH map and an sk_msg program user can steer messages
sent from one TCP socket (s1) to actually egress from another TCP
socket (s2):

tcp_bpf_sendmsg(s1)		// = sk_prot->sendmsg
  tcp_bpf_send_verdict(s1)	// __SK_REDIRECT case
    tcp_bpf_sendmsg_redir(s2)
      tcp_bpf_push_locked(s2)
	tcp_bpf_push(s2)
	  tcp_rate_check_app_limited(s2) // expects tcp_sock
	  tcp_sendmsg_locked(s2)	 // ditto

There is a hard-coded assumption in the call-chain, that the egress
socket (s2) is a TCP socket.

However in commit 122e6c79efe1 ("sock_map: Update sock type checks for
UDP") we have enabled redirects to non-TCP sockets. This was done for the
sake of BPF sk_skb programs. There was no indention to support sk_msg
send-to-egress use case.

As a result, attempts to send-to-egress through a non-TCP socket lead to a
crash due to invalid downcast from sock to tcp_sock:

 BUG: kernel NULL pointer dereference, address: 000000000000002f
 ...
 Call Trace:
  <TASK>
  ? show_regs+0x60/0x70
  ? __die+0x1f/0x70
  ? page_fault_oops+0x80/0x160
  ? do_user_addr_fault+0x2d7/0x800
  ? rcu_is_watching+0x11/0x50
  ? exc_page_fault+0x70/0x1c0
  ? asm_exc_page_fault+0x27/0x30
  ? tcp_tso_segs+0x14/0xa0
  tcp_write_xmit+0x67/0xce0
  __tcp_push_pending_frames+0x32/0xf0
  tcp_push+0x107/0x140
  tcp_sendmsg_locked+0x99f/0xbb0
  tcp_bpf_push+0x19d/0x3a0
  tcp_bpf_sendmsg_redir+0x55/0xd0
  tcp_bpf_send_verdict+0x407/0x550
  tcp_bpf_sendmsg+0x1a1/0x390
  inet_sendmsg+0x6a/0x70
  sock_sendmsg+0x9d/0xc0
  ? sockfd_lookup_light+0x12/0x80
  __sys_sendto+0x10e/0x160
  ? syscall_enter_from_user_mode+0x20/0x60
  ? __this_cpu_preempt_check+0x13/0x20
  ? lockdep_hardirqs_on+0x82/0x110
  __x64_sys_sendto+0x1f/0x30
  do_syscall_64+0x38/0x90
  entry_SYSCALL_64_after_hwframe+0x63/0xcd

Reject selecting a non-TCP sockets as redirect target from a BPF sk_msg
program to prevent the crash. When attempted, user will receive an EACCES
error from send/sendto/sendmsg() syscall.

Fixes: 122e6c79efe1 ("sock_map: Update sock type checks for UDP")
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
---
FYI, I'm working on revamping the sockmap_listen selftest, which exercises
some of redirect combinations, to cover the whole combination matrix so
that we can catch these kinds of problems early on.

 net/core/sock_map.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/net/core/sock_map.c b/net/core/sock_map.c
index cb11750b1df5..4292c2ed1828 100644
--- a/net/core/sock_map.c
+++ b/net/core/sock_map.c
@@ -668,6 +668,8 @@ BPF_CALL_4(bpf_msg_redirect_map, struct sk_msg *, msg,
 	sk = __sock_map_lookup_elem(map, key);
 	if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
 		return SK_DROP;
+	if (!(flags & BPF_F_INGRESS) && !sk_is_tcp(sk))
+		return SK_DROP;
 
 	msg->flags = flags;
 	msg->sk_redir = sk;
@@ -1267,6 +1269,8 @@ BPF_CALL_4(bpf_msg_redirect_hash, struct sk_msg *, msg,
 	sk = __sock_hash_lookup_elem(map, key);
 	if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
 		return SK_DROP;
+	if (!(flags & BPF_F_INGRESS) && !sk_is_tcp(sk))
+		return SK_DROP;
 
 	msg->flags = flags;
 	msg->sk_redir = sk;
-- 
2.41.0

