[tcp] Send TCP keepalives on idle established connections

In some circumstances, intermediate devices may lose state in a way
that temporarily prevents the successful delivery of packets from a
TCP peer.  For example, a firewall may drop a NAT forwarding table
entry.

Since iPXE spends most of its time downloading files (and hence purely
receiving data, sending only TCP ACKs), this can easily happen in a
situation in which there is no reason for iPXE's TCP stack to generate
any retransmissions.  The temporary loss of connectivity can therefore
effectively become permanent.

Work around this problem by sending TCP keepalives after a period of
inactivity on an established connection.

TCP keepalives usually send a single garbage byte in sequence number
space that has already been ACKed by the peer.  Since we do not need
to elicit a response from the peer, we instead send pure ACKs (with no
garbage data) in order to keep the transmit code path simple.

Originally-implemented-by: Ladi Prosek <lprosek@redhat.com>
Debugged-by: Ladi Prosek <lprosek@redhat.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>

src/include/ipxe/tcp.h

  */
 #define TCP_MSL ( 2 * 60 * TICKS_PER_SEC )
 
+/**
+ * TCP keepalive period
+ *
+ * We send keepalive ACKs after this period of inactivity has elapsed
+ * on an established connection.
+ */
+#define TCP_KEEPALIVE_DELAY ( 15 * TICKS_PER_SEC )
+
 /**
  * TCP maximum header length
  *

src/net/tcp.c

 	struct process process;
 	/** Retransmission timer */
 	struct retry_timer timer;
+	/** Keepalive timer */
+	struct retry_timer keepalive;
 	/** Shutdown (TIME_WAIT) timer */
 	struct retry_timer wait;
 
 static struct process_descriptor tcp_process_desc;
 static struct interface_descriptor tcp_xfer_desc;
 static void tcp_expired ( struct retry_timer *timer, int over );
+static void tcp_keepalive_expired ( struct retry_timer *timer, int over );
 static void tcp_wait_expired ( struct retry_timer *timer, int over );
 static struct tcp_connection * tcp_demux ( unsigned int local_port );
 static int tcp_rx_ack ( struct tcp_connection *tcp, uint32_t ack,
 	intf_init ( &tcp->xfer, &tcp_xfer_desc, &tcp->refcnt );
 	process_init_stopped ( &tcp->process, &tcp_process_desc, &tcp->refcnt );
 	timer_init ( &tcp->timer, tcp_expired, &tcp->refcnt );
+	timer_init ( &tcp->keepalive, tcp_keepalive_expired, &tcp->refcnt );
 	timer_init ( &tcp->wait, tcp_wait_expired, &tcp->refcnt );
 	tcp->prev_tcp_state = TCP_CLOSED;
 	tcp->tcp_state = TCP_STATE_SENT ( TCP_SYN );
 		/* Remove from list and drop reference */
 		process_del ( &tcp->process );
 		stop_timer ( &tcp->timer );
+		stop_timer ( &tcp->keepalive );
 		stop_timer ( &tcp->wait );
 		list_del ( &tcp->list );
 		ref_put ( &tcp->refcnt );
 	if ( ! ( tcp->tcp_state & TCP_STATE_ACKED ( TCP_SYN ) ) )
 		tcp_rx_ack ( tcp, ( tcp->snd_seq + 1 ), 0 );
 
+	/* Stop keepalive timer */
+	stop_timer ( &tcp->keepalive );
+
 	/* If we have no data remaining to send, start sending FIN */
 	if ( list_empty ( &tcp->tx_queue ) &&
 	     ! ( tcp->tcp_state & TCP_STATE_SENT ( TCP_FIN ) ) ) {
 	}
 }
 
+/**
+ * Keepalive timer expired
+ *
+ * @v timer		Keepalive timer
+ * @v over		Failure indicator
+ */
+static void tcp_keepalive_expired ( struct retry_timer *timer,
+				    int over __unused ) {
+	struct tcp_connection *tcp =
+		container_of ( timer, struct tcp_connection, keepalive );
+
+	DBGC ( tcp, "TCP %p sending keepalive\n", tcp );
+
+	/* Reset keepalive timer */
+	start_timer_fixed ( &tcp->keepalive, TCP_KEEPALIVE_DELAY );
+
+	/* Send keepalive.  We do this only to preserve or restore
+	 * state in intermediate devices (e.g. firewall NAT tables);
+	 * we don't actually care about eliciting a response to verify
+	 * that the peer is still alive.  We therefore send just a
+	 * pure ACK, to keep our transmit path simple.
+	 */
+	tcp->flags |= TCP_ACK_PENDING;
+	tcp_xmit ( tcp );
+}
+
 /**
  * Shutdown timer expired
  *
 	/* Update window size */
 	tcp->snd_win = win;
 
+	/* Hold off (or start) the keepalive timer, if applicable */
+	if ( ! ( tcp->tcp_state & TCP_STATE_SENT ( TCP_FIN ) ) )
+		start_timer_fixed ( &tcp->keepalive, TCP_KEEPALIVE_DELAY );
+
 	/* Ignore ACKs that don't actually acknowledge any new data.
 	 * (In particular, do not stop the retransmission timer; this
 	 * avoids creating a sorceror's apprentice syndrome when a