[PATCH v2 09/10] tcp: Maintain host flowside for connections

[David Gibson <david@gibson.dropbear.id.au>]

We now maintain a struct flowside describing each TCP connection as it
appears to the guest.  We don't explicitly have the same information
for the connections as they appear to the host, however.  Rather, that
information is implicit in the state of the host side socket.  For
future generalisations of flow/connection tracking, we're going to
need to use this information more heavily, so properly populate the
other flowside in each flow table entry.

This does require an additional getsockname() call for each new connection.
We hope to optimise that away for at least some cases in future.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
 flow.c     | 26 ++++++++++++++++++++++++++
 flow.h     |  1 +
 tcp.c      | 46 ++++++++++++++++++++++++++++++++++++++++++----
 tcp_conn.h |  4 ----
 4 files changed, 69 insertions(+), 8 deletions(-)

diff --git a/flow.c b/flow.c
index 4521a43..f2a7377 100644
--- a/flow.c
+++ b/flow.c
@@ -7,6 +7,7 @@
 
 #include <unistd.h>
 #include <string.h>
+#include <errno.h>
 #include <arpa/inet.h>
 
 #include "util.h"
@@ -83,3 +84,28 @@ void flow_table_compact(struct ctx *c, union flow *hole)
 
 	memset(from, 0, sizeof(*from));
 }
+
+/** flowside_getsockname - Initialize flowside f{addr,port} from a bound socket
+ * @fs:		flowside to initialize
+ * @s:		bound socket
+ *
+ * #syscalls getsockname
+ */
+int flowside_getsockname(struct flowside *fs, int s)
+{
+	struct sockaddr_storage sa;
+	socklen_t sl = sizeof(sa);
+
+	/* FIXME: Workaround clang-tidy not realizing that getsockname() writes
+	 * the socket address.  See
+	 * https://github.com/llvm/llvm-project/issues/58992
+	 */
+	memset(&sa, 0, sizeof(struct sockaddr_in6));
+	if (getsockname(s, (struct sockaddr *)&sa, &sl) < 0)
+		return -errno;
+
+	inany_from_sockaddr(&fs->faddr, &fs->fport,
+			    (const struct sockaddr *)&sa);
+
+	return 0;
+}
diff --git a/flow.h b/flow.h
index b4f042b..4a27303 100644
--- a/flow.h
+++ b/flow.h
@@ -61,6 +61,7 @@ static inline bool flowside_complete(const struct flowside *fs)
 
 #define FLOWSIDE_STRLEN		(2*(INET6_ADDRSTRLEN+8) + 6)
 
+int flowside_getsockname(struct flowside *fs, int s);
 const char *flowside_fmt(const struct flowside *fs, char *buf, size_t size);
 
 /**
diff --git a/tcp.c b/tcp.c
index a9ddce6..297134f 100644
--- a/tcp.c
+++ b/tcp.c
@@ -397,6 +397,7 @@ struct tcp6_l2_head {	/* For MSS6 macro: keep in sync with tcp6_l2_buf_t */
 #define OPT_SACK	5
 #define OPT_TS		8
 
+#define SOCKSIDE(conn)		(&(conn)->f.side[0])
 #define TAPSIDE(conn)		(&(conn)->f.side[1])
 
 #define CONN_V4(conn)		(!!inany_v4(&TAPSIDE(conn)->faddr))
@@ -2020,6 +2021,19 @@ static void tcp_conn_from_tap(struct ctx *c,
 		conn_event(c, conn, TAP_SYN_ACK_SENT);
 	}
 
+	/* Initialise sock-side demiflow */
+	SOCKSIDE(conn)->eaddr = TAPSIDE(conn)->faddr;
+	SOCKSIDE(conn)->eport = TAPSIDE(conn)->fport;
+	if (flowside_getsockname(SOCKSIDE(conn), s) < 0) {
+		err("tcp: Failed to get local name for outgoing connection");
+		tcp_rst(c, conn);
+		return;
+	}
+
+	ASSERT(flowside_complete(SOCKSIDE(conn)));
+	debug("TCP: index %li, connection forwarded to socket, %s", FLOW_IDX(conn),
+	      flowside_fmt(SOCKSIDE(conn), fsstr, sizeof(fsstr)));
+
 	tcp_epoll_ctl(c, conn);
 }
 
@@ -2629,20 +2643,35 @@ static void tcp_snat_inbound(const struct ctx *c, union inany_addr *addr)
  * @s:		Accepted socket
  * @sa:		Peer socket address (from accept())
  * @now:	Current timestamp
+ *
+ * Return: true if able to create a tap connection, false otherwise
  */
-static void tcp_tap_conn_from_sock(struct ctx *c,
+static bool tcp_tap_conn_from_sock(struct ctx *c,
 				   union tcp_listen_epoll_ref ref,
 				   struct tcp_tap_conn *conn, int s,
 				   struct sockaddr *sa,
 				   const struct timespec *now)
 {
+	char fsstr[FLOWSIDE_STRLEN];
+
 	conn->f.type = FLOW_TCP;
 	conn->sock = s;
 	conn->timer = -1;
 	conn->ws_to_tap = conn->ws_from_tap = 0;
 	conn_event(c, conn, SOCK_ACCEPTED);
 
-	inany_from_sockaddr(&TAPSIDE(conn)->faddr, &TAPSIDE(conn)->fport, sa);
+	if (flowside_getsockname(SOCKSIDE(conn), s) < 0) {
+		err("tcp: Failed to get local name, connection dropped");
+		return false;
+	}
+	inany_from_sockaddr(&SOCKSIDE(conn)->eaddr, &SOCKSIDE(conn)->eport, sa);
+
+	ASSERT(flowside_complete(SOCKSIDE(conn)));
+	debug("TCP: index %li, new connection from socket, %s", FLOW_IDX(conn),
+	      flowside_fmt(SOCKSIDE(conn), fsstr, sizeof(fsstr)));
+
+	TAPSIDE(conn)->faddr = SOCKSIDE(conn)->eaddr;
+	TAPSIDE(conn)->fport = SOCKSIDE(conn)->eport;
 	tcp_snat_inbound(c, &TAPSIDE(conn)->faddr);
 
 	if (CONN_V4(conn)) {
@@ -2656,6 +2685,8 @@ static void tcp_tap_conn_from_sock(struct ctx *c,
 	TAPSIDE(conn)->eport = ref.port;
 
 	ASSERT(flowside_complete(TAPSIDE(conn)));
+	debug("TCP: index %li, connection forwarded to tap, %s", FLOW_IDX(conn),
+	      flowside_fmt(TAPSIDE(conn), fsstr, sizeof(fsstr)));
 
 	tcp_seq_init(c, conn, now);
 	tcp_hash_insert(c, conn);
@@ -2668,6 +2699,8 @@ static void tcp_tap_conn_from_sock(struct ctx *c,
 	conn_flag(c, conn, ACK_FROM_TAP_DUE);
 
 	tcp_get_sndbuf(conn);
+
+	return true;
 }
 
 /**
@@ -2704,8 +2737,13 @@ void tcp_listen_handler(struct ctx *c, union epoll_ref ref,
 				      s, (struct sockaddr *)&sa))
 		return;
 
-	tcp_tap_conn_from_sock(c, ref.tcp_listen, &flow->tcp, s,
-			       (struct sockaddr *)&sa, now);
+	if (tcp_tap_conn_from_sock(c, ref.tcp_listen, &flow->tcp, s,
+				   (struct sockaddr *)&sa, now))
+		return;
+
+	/* Failed to create the connection */
+	close(s);
+	c->flow_count--;
 }
 
 /**
diff --git a/tcp_conn.h b/tcp_conn.h
index 3482759..2ef0130 100644
--- a/tcp_conn.h
+++ b/tcp_conn.h
@@ -24,10 +24,6 @@
  * @ws_to_tap:		Window scaling factor advertised to tap/guest
  * @sndbuf:		Sending buffer in kernel, rounded to 2 ^ SNDBUF_BITS
  * @seq_dup_ack_approx:	Last duplicate ACK number sent to tap
- * @eaddr:		Guest side endpoint address (guest's local address)
- * @faddr:		Guest side forwarding address (guest's remote address)
- * @eport:		Guest side endpoint port (guest's local port)
- * @fport:		Guest side forwarding port (guest's remote port)
  * @wnd_from_tap:	Last window size from tap, unscaled (as received)
  * @wnd_to_tap:		Sending window advertised to tap, unscaled (as sent)
  * @seq_to_tap:		Next sequence for packets to tap
-- 
@@ -24,10 +24,6 @@
  * @ws_to_tap:		Window scaling factor advertised to tap/guest
  * @sndbuf:		Sending buffer in kernel, rounded to 2 ^ SNDBUF_BITS
  * @seq_dup_ack_approx:	Last duplicate ACK number sent to tap
- * @eaddr:		Guest side endpoint address (guest's local address)
- * @faddr:		Guest side forwarding address (guest's remote address)
- * @eport:		Guest side endpoint port (guest's local port)
- * @fport:		Guest side forwarding port (guest's remote port)
  * @wnd_from_tap:	Last window size from tap, unscaled (as received)
  * @wnd_to_tap:		Sending window advertised to tap, unscaled (as sent)
  * @seq_to_tap:		Next sequence for packets to tap
-- 
2.41.0