tls: rx: periodically flush socket backlog

We continuously hold the socket lock during large reads and writes.
This may inflate RTT and negatively impact TCP performance.
Flush the backlog periodically. I tried to pick a flush period (128kB)
which gives significant benefit but the max Bps rate is not yet visibly
impacted.
Signed-off-by: NJakub Kicinski <kuba@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

net/core/sock.c

@@ -2870,6 +2870,7 @@ void __sk_flush_backlog(struct sock *sk)
 	__release_sock(sk);
 	spin_unlock_bh(&sk->sk_lock.slock);
 }
+EXPORT_SYMBOL_GPL(__sk_flush_backlog);
 
 /**
  * sk_wait_data - wait for data to arrive at sk_receive_queue

net/tls/tls_sw.c

@@ -1738,6 +1738,24 @@ static int process_rx_list(struct tls_sw_context_rx *ctx,
 	return copied ? : err;
 }
 
+static void
+tls_read_flush_backlog(struct sock *sk, struct tls_prot_info *prot,
+		       size_t len_left, size_t decrypted, ssize_t done,
+		       size_t *flushed_at)
+{
+	size_t max_rec;
+
+	if (len_left <= decrypted)
+		return;
+
+	max_rec = prot->overhead_size - prot->tail_size + TLS_MAX_PAYLOAD_SIZE;
+	if (done - *flushed_at < SZ_128K && tcp_inq(sk) > max_rec)
+		return;
+
+	*flushed_at = done;
+	sk_flush_backlog(sk);
+}
+
 int tls_sw_recvmsg(struct sock *sk,
 		   struct msghdr *msg,
 		   size_t len,
@@ -1750,6 +1768,7 @@ int tls_sw_recvmsg(struct sock *sk,
 	struct sk_psock *psock;
 	unsigned char control = 0;
 	ssize_t decrypted = 0;
+	size_t flushed_at = 0;
 	struct strp_msg *rxm;
 	struct tls_msg *tlm;
 	struct sk_buff *skb;
@@ -1839,6 +1858,10 @@ int tls_sw_recvmsg(struct sock *sk,
 		if (err <= 0)
 			goto recv_end;
 
+		/* periodically flush backlog, and feed strparser */
+		tls_read_flush_backlog(sk, prot, len, to_decrypt,
+				       decrypted + copied, &flushed_at);
+
 		ctx->recv_pkt = NULL;
 		__strp_unpause(&ctx->strp);
 		__skb_queue_tail(&ctx->rx_list, skb);