bpf, sockmap: Add sk_rmem_alloc check for tcp_bpf_ingress()

We use sk_msg to redirect with sock hash, like this:

  skA   redirect    skB
  Tx <----------->  Rx

And construct a scenario where the packet sending speed is high, the
packet receiving speed is slow, so the packets are stacked in the ingress
queue on the receiving side. After a period of time, the memory is
exhausted and the system ooms.

To fix, we add sk_rmem_alloc while sk_msg queued in the ingress queue
and subtract sk_rmem_alloc while sk_msg dequeued from the ingress queue
and check sk_rmem_alloc at the beginning of bpf_tcp_ingress().

Signed-off-by: Wang Yufen <[email protected]>

Author: wangyufen316

include/linux/skmsg.h

@@ -308,9 +308,10 @@ static inline void sk_psock_queue_msg(struct sk_psock *psock,
 				      struct sk_msg *msg)
 {
 	spin_lock_bh(&psock->ingress_lock);
-	if (sk_psock_test_state(psock, SK_PSOCK_TX_ENABLED))
+	if (sk_psock_test_state(psock, SK_PSOCK_TX_ENABLED)) {
 		list_add_tail(&msg->list, &psock->ingress_msg);
-	else {
+		atomic_add(msg->sg.size, &psock->sk->sk_rmem_alloc);
+	} else {
 		sk_msg_free(psock->sk, msg);
 		kfree(msg);
 	}
@@ -323,8 +324,10 @@ static inline struct sk_msg *sk_psock_dequeue_msg(struct sk_psock *psock)
 
 	spin_lock_bh(&psock->ingress_lock);
 	msg = list_first_entry_or_null(&psock->ingress_msg, struct sk_msg, list);
-	if (msg)
+	if (msg) {
 		list_del(&msg->list);
+		atomic_sub(msg->sg.size, &psock->sk->sk_rmem_alloc);
+	}
 	spin_unlock_bh(&psock->ingress_lock);
 	return msg;
 }

net/core/skmsg.c

@@ -445,6 +445,7 @@ int sk_msg_recvmsg(struct sock *sk, struct sk_psock *psock, struct msghdr *msg,
 				if (!msg_rx->skb)
 					sk_mem_uncharge(sk, copy);
 				msg_rx->sg.size -= copy;
+				atomic_sub(copy, &sk->sk_rmem_alloc);
 
 				if (!sge->length) {
 					sk_msg_iter_var_next(i);
@@ -754,6 +755,7 @@ static void __sk_psock_purge_ingress_msg(struct sk_psock *psock)
 
 	list_for_each_entry_safe(msg, tmp, &psock->ingress_msg, list) {
 		list_del(&msg->list);
+		atomic_sub(msg->sg.size, &psock->sk->sk_rmem_alloc);
 		sk_msg_free(psock->sk, msg);
 		kfree(msg);
 	}

net/ipv4/tcp_bpf.c

@@ -24,6 +24,12 @@ static int bpf_tcp_ingress(struct sock *sk, struct sk_psock *psock,
 		return -ENOMEM;
 
 	lock_sock(sk);
+	if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf) {
+		release_sock(sk);
+		kfree(tmp);
+		return -EAGAIN;
+	}
+
 	tmp->sg.start = msg->sg.start;
 	i = msg->sg.start;
 	do {