tcp_bpf: Add sk_rmem_alloc related logic for tcp_bpf ingress redirection

When we do sk_psock_verdict_apply->sk_psock_skb_ingress, an sk_msg will
be created out of the skb, and the rmem accounting of the sk_msg will be
handled by the skb.

For skmsgs in __SK_REDIRECT case of tcp_bpf_send_verdict, when redirecting
to the ingress of a socket, although we sk_rmem_schedule and add sk_msg to
the ingress_msg of sk_redir, we do not update sk_rmem_alloc. As a result,
except for the global memory limit, the rmem of sk_redir is nearly
unlimited. Thus, add sk_rmem_alloc related logic to limit the recv buffer.

Since the function sk_msg_recvmsg and __sk_psock_purge_ingress_msg are
used in these two paths. We use "msg->skb" to test whether the sk_msg is
skb backed up. If it's not, we shall do the memory accounting explicitly.

Fixes: 604326b ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: Zijian Zhang <zijianzhang@bytedance.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20241210012039.1669389-3-zijianzhang@bytedance.com

Author: Zijian Zhang

include/linux/skmsg.h

@@ -317,17 +317,22 @@ static inline void sock_drop(struct sock *sk, struct sk_buff *skb)
 	kfree_skb(skb);
 }
 
-static inline void sk_psock_queue_msg(struct sk_psock *psock,
+static inline bool sk_psock_queue_msg(struct sk_psock *psock,
 				      struct sk_msg *msg)
 {
+	bool ret;
+
 	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 {
+		ret = true;
+	} else {
 		sk_msg_free(psock->sk, msg);
 		kfree(msg);
+		ret = false;
 	}
 	spin_unlock_bh(&psock->ingress_lock);
+	return ret;
 }
 
 static inline struct sk_msg *sk_psock_dequeue_msg(struct sk_psock *psock)

net/core/skmsg.c

@@ -445,8 +445,10 @@ int sk_msg_recvmsg(struct sock *sk, struct sk_psock *psock, struct msghdr *msg,
 			if (likely(!peek)) {
 				sge->offset += copy;
 				sge->length -= copy;
-				if (!msg_rx->skb)
+				if (!msg_rx->skb) {
 					sk_mem_uncharge(sk, copy);
+					atomic_sub(copy, &sk->sk_rmem_alloc);
+				}
 				msg_rx->sg.size -= copy;
 
 				if (!sge->length) {
@@ -772,6 +774,8 @@ 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);
+		if (!msg->skb)
+			atomic_sub(msg->sg.size, &psock->sk->sk_rmem_alloc);
 		sk_msg_free(psock->sk, msg);
 		kfree(msg);
 	}

net/ipv4/tcp_bpf.c

@@ -56,6 +56,7 @@ static int bpf_tcp_ingress(struct sock *sk, struct sk_psock *psock,
 		}
 
 		sk_mem_charge(sk, size);
+		atomic_add(size, &sk->sk_rmem_alloc);
 		sk_msg_xfer(tmp, msg, i, size);
 		copied += size;
 		if (sge->length)
@@ -74,7 +75,8 @@ static int bpf_tcp_ingress(struct sock *sk, struct sk_psock *psock,
 
 	if (!ret) {
 		msg->sg.start = i;
-		sk_psock_queue_msg(psock, tmp);
+		if (!sk_psock_queue_msg(psock, tmp))
+			atomic_sub(copied, &sk->sk_rmem_alloc);
 		sk_psock_data_ready(sk, psock);
 	} else {
 		sk_msg_free(sk, tmp);