bpf: sockmap/tls, close can race with map free

When a map free is called and in parallel a socket is closed we
have two paths that can potentially reset the socket prot ops, the
bpf close() path and the map free path. This creates a problem
with which prot ops should be used from the socket closed side.

If the map_free side completes first then we want to call the
original lowest level ops. However, if the tls path runs first
we want to call the sockmap ops. Additionally there was no locking
around prot updates in TLS code paths so the prot ops could
be changed multiple times once from TLS path and again from sockmap
side potentially leaving ops pointed at either TLS or sockmap
when psock and/or tls context have already been destroyed.

To fix this race first only update ops inside callback lock
so that TLS, sockmap and lowest level all agree on prot state.
Second and a ULP callback update() so that lower layers can
inform the upper layer when they are being removed allowing the
upper layer to reset prot ops.

This gets us close to allowing sockmap and tls to be stacked
in arbitrary order but will save that patch for *next trees.

v4:
 - make sure we don't free things for device;
 - remove the checks which swap the callbacks back
   only if TLS is at the top.

Reported-by: syzbot+06537213db7ba2745c4a@syzkaller.appspotmail.com
Fixes: 02c558b2d5d6 ("bpf: sockmap, support for msg_peek in sk_msg with redirect ingress")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

3 files changed, 43 insertions, 7 deletions

diff --git a/net/core/skmsg.c b/net/core/skmsg.c
index 93bffaad21354..6832eeb4b7854 100644
--- a/net/core/skmsg.c
+++ b/net/core/skmsg.c
@@ -585,12 +585,12 @@ EXPORT_SYMBOL_GPL(sk_psock_destroy);
 
 void sk_psock_drop(struct sock *sk, struct sk_psock *psock)
 {
-	rcu_assign_sk_user_data(sk, NULL);
 	sk_psock_cork_free(psock);
 	sk_psock_zap_ingress(psock);
-	sk_psock_restore_proto(sk, psock);
 
 	write_lock_bh(&sk->sk_callback_lock);
+	sk_psock_restore_proto(sk, psock);
+	rcu_assign_sk_user_data(sk, NULL);
 	if (psock->progs.skb_parser)
 		sk_psock_stop_strp(sk, psock);
 	write_unlock_bh(&sk->sk_callback_lock);
diff --git a/net/ipv4/tcp_ulp.c b/net/ipv4/tcp_ulp.c
index 3d8a1d8354719..4849edb62d529 100644
--- a/net/ipv4/tcp_ulp.c
+++ b/net/ipv4/tcp_ulp.c
@@ -96,6 +96,19 @@ void tcp_get_available_ulp(char *buf, size_t maxlen)
 	rcu_read_unlock();
 }
 
+void tcp_update_ulp(struct sock *sk, struct proto *proto)
+{
+	struct inet_connection_sock *icsk = inet_csk(sk);
+
+	if (!icsk->icsk_ulp_ops) {
+		sk->sk_prot = proto;
+		return;
+	}
+
+	if (icsk->icsk_ulp_ops->update)
+		icsk->icsk_ulp_ops->update(sk, proto);
+}
+
 void tcp_cleanup_ulp(struct sock *sk)
 {
 	struct inet_connection_sock *icsk = inet_csk(sk);
diff --git a/net/tls/tls_main.c b/net/tls/tls_main.c
index 48f1c26459d04..f208f8455ef2e 100644
--- a/net/tls/tls_main.c
+++ b/net/tls/tls_main.c
@@ -328,7 +328,10 @@ static void tls_sk_proto_unhash(struct sock *sk)
 
 	ctx = tls_get_ctx(sk);
 	tls_sk_proto_cleanup(sk, ctx, timeo);
+	write_lock_bh(&sk->sk_callback_lock);
 	icsk->icsk_ulp_data = NULL;
+	sk->sk_prot = ctx->sk_proto;
+	write_unlock_bh(&sk->sk_callback_lock);
 
 	if (ctx->sk_proto->unhash)
 		ctx->sk_proto->unhash(sk);
@@ -337,7 +340,7 @@ static void tls_sk_proto_unhash(struct sock *sk)
 
 static void tls_sk_proto_close(struct sock *sk, long timeout)
 {
-	void (*sk_proto_close)(struct sock *sk, long timeout);
+	struct inet_connection_sock *icsk = inet_csk(sk);
 	struct tls_context *ctx = tls_get_ctx(sk);
 	long timeo = sock_sndtimeo(sk, 0);
 	bool free_ctx;
@@ -347,12 +350,15 @@ static void tls_sk_proto_close(struct sock *sk, long timeout)
 
 	lock_sock(sk);
 	free_ctx = ctx->tx_conf != TLS_HW && ctx->rx_conf != TLS_HW;
-	sk_proto_close = ctx->sk_proto_close;
 
 	if (ctx->tx_conf != TLS_BASE || ctx->rx_conf != TLS_BASE)
 		tls_sk_proto_cleanup(sk, ctx, timeo);
 
+	write_lock_bh(&sk->sk_callback_lock);
+	if (free_ctx)
+		icsk->icsk_ulp_data = NULL;
 	sk->sk_prot = ctx->sk_proto;
+	write_unlock_bh(&sk->sk_callback_lock);
 	release_sock(sk);
 	if (ctx->tx_conf == TLS_SW)
 		tls_sw_free_ctx_tx(ctx);
@@ -360,7 +366,7 @@ static void tls_sk_proto_close(struct sock *sk, long timeout)
 		tls_sw_strparser_done(ctx);
 	if (ctx->rx_conf == TLS_SW)
 		tls_sw_free_ctx_rx(ctx);
-	sk_proto_close(sk, timeout);
+	ctx->sk_proto_close(sk, timeout);
 
 	if (free_ctx)
 		tls_ctx_free(ctx);
@@ -827,7 +833,7 @@ static int tls_init(struct sock *sk)
 	int rc = 0;
 
 	if (tls_hw_prot(sk))
-		goto out;
+		return 0;
 
 	/* The TLS ulp is currently supported only for TCP sockets
 	 * in ESTABLISHED state.
@@ -838,22 +844,38 @@ static int tls_init(struct sock *sk)
 	if (sk->sk_state != TCP_ESTABLISHED)
 		return -ENOTSUPP;
 
+	tls_build_proto(sk);
+
 	/* allocate tls context */
+	write_lock_bh(&sk->sk_callback_lock);
 	ctx = create_ctx(sk);
 	if (!ctx) {
 		rc = -ENOMEM;
 		goto out;
 	}
 
-	tls_build_proto(sk);
 	ctx->tx_conf = TLS_BASE;
 	ctx->rx_conf = TLS_BASE;
 	ctx->sk_proto = sk->sk_prot;
 	update_sk_prot(sk, ctx);
 out:
+	write_unlock_bh(&sk->sk_callback_lock);
 	return rc;
 }
 
+static void tls_update(struct sock *sk, struct proto *p)
+{
+	struct tls_context *ctx;
+
+	ctx = tls_get_ctx(sk);
+	if (likely(ctx)) {
+		ctx->sk_proto_close = p->close;
+		ctx->sk_proto = p;
+	} else {
+		sk->sk_prot = p;
+	}
+}
+
 void tls_register_device(struct tls_device *device)
 {
 	spin_lock_bh(&device_spinlock);
@@ -874,6 +896,7 @@ static struct tcp_ulp_ops tcp_tls_ulp_ops __read_mostly = {
 	.name			= "tls",
 	.owner			= THIS_MODULE,
 	.init			= tls_init,
+	.update			= tls_update,
 };
 
 static int __init tls_register(void)