[neighbour] Generalise concept of neighbour discovery

Split the protocol-independent portions of arp.c into a separate file
neighbour.c, to allow for sharing of functionality between IPv4+ARP
and IPv6+NDP.

Signed-off-by: Michael Brown <mcb30@ipxe.org>

5 files changed, 579 insertions, 357 deletions

diff --git a/src/include/ipxe/arp.h b/src/include/ipxe/arp.h
index 00396d82..e30ae6b7 100644
--- a/src/include/ipxe/arp.h
+++ b/src/include/ipxe/arp.h
@@ -11,6 +11,7 @@ FILE_LICENCE ( GPL2_OR_LATER );
 
 #include <ipxe/tables.h>
 #include <ipxe/netdevice.h>
+#include <ipxe/neighbour.h>
 
 /** A network-layer protocol that relies upon ARP */
 struct arp_net_protocol {
@@ -34,9 +35,26 @@ struct arp_net_protocol {
 #define __arp_net_protocol __table_entry ( ARP_NET_PROTOCOLS, 01 )
 
 extern struct net_protocol arp_protocol __net_protocol;
+extern struct neighbour_discovery arp_discovery;
 
-extern int arp_tx ( struct io_buffer *iobuf, struct net_device *netdev,
-		    struct net_protocol *net_protocol, const void *net_dest,
-		    const void *net_source, const void *ll_source );
+/**
+ * Transmit packet, determining link-layer address via ARP
+ *
+ * @v iobuf		I/O buffer
+ * @v netdev		Network device
+ * @v net_protocol	Network-layer protocol
+ * @v net_dest		Destination network-layer address
+ * @v net_source	Source network-layer address
+ * @v ll_source		Source link-layer address
+ * @ret rc		Return status code
+ */
+static inline int arp_tx ( struct io_buffer *iobuf, struct net_device *netdev,
+			   struct net_protocol *net_protocol,
+			   const void *net_dest, const void *net_source,
+			   const void *ll_source ) {
+
+	return neighbour_tx ( iobuf, netdev, net_protocol, net_dest,
+			      &arp_discovery, net_source, ll_source );
+}
 
 #endif /* _IPXE_ARP_H */
diff --git a/src/include/ipxe/errfile.h b/src/include/ipxe/errfile.h
index 801d579b..83675796 100644
--- a/src/include/ipxe/errfile.h
+++ b/src/include/ipxe/errfile.h
@@ -214,6 +214,7 @@ FILE_LICENCE ( GPL2_OR_LATER );
 #define ERRFILE_nfs_open		( ERRFILE_NET | 0x00340000 )
 #define ERRFILE_mount			( ERRFILE_NET | 0x00350000 )
 #define ERRFILE_oncrpc_iob		( ERRFILE_NET | 0x00360000 )
+#define ERRFILE_neighbour		( ERRFILE_NET | 0x00370000 )
 
 #define ERRFILE_image		      ( ERRFILE_IMAGE | 0x00000000 )
 #define ERRFILE_elf		      ( ERRFILE_IMAGE | 0x00010000 )
diff --git a/src/include/ipxe/neighbour.h b/src/include/ipxe/neighbour.h
new file mode 100644
index 00000000..5720e8b0
--- /dev/null
+++ b/src/include/ipxe/neighbour.h
@@ -0,0 +1,44 @@
+#ifndef _IPXE_NEIGHBOUR_H
+#define _IPXE_NEIGHBOUR_H
+
+/** @file
+ *
+ * Neighbour discovery
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER );
+
+#include <ipxe/netdevice.h>
+
+/** A neighbour discovery protocol */
+struct neighbour_discovery {
+	/** Name */
+	const char *name;
+	/**
+	 * Transmit neighbour discovery request
+	 *
+	 * @v netdev		Network device
+	 * @v net_protocol	Network-layer protocol
+	 * @v net_dest		Destination network-layer address
+	 * @v net_source	Source network-layer address
+	 * @ret rc		Return status code
+	 */
+	int ( * tx_request ) ( struct net_device *netdev,
+			       struct net_protocol *net_protocol,
+			       const void *net_dest, const void *net_source );
+};
+
+extern int neighbour_tx ( struct io_buffer *iobuf, struct net_device *netdev,
+			  struct net_protocol *net_protocol,
+			  const void *net_dest,
+			  struct neighbour_discovery *discovery,
+			  const void *net_source, const void *ll_source );
+extern int neighbour_update ( struct net_device *netdev,
+			      struct net_protocol *net_protocol,
+			      const void *net_dest, const void *ll_dest );
+extern int neighbour_define ( struct net_device *netdev,
+			      struct net_protocol *net_protocol,
+			      const void *net_dest, const void *ll_dest );
+
+#endif /* _IPXE_NEIGHBOUR_H */
diff --git a/src/net/arp.c b/src/net/arp.c
index b94eb906..261e681e 100644
--- a/src/net/arp.c
+++ b/src/net/arp.c
@@ -28,11 +28,7 @@ FILE_LICENCE ( GPL2_OR_LATER );
 #include <ipxe/if_arp.h>
 #include <ipxe/iobuf.h>
 #include <ipxe/netdevice.h>
-#include <ipxe/list.h>
-#include <ipxe/retry.h>
-#include <ipxe/timer.h>
-#include <ipxe/malloc.h>
-#include <ipxe/refcnt.h>
+#include <ipxe/neighbour.h>
 #include <ipxe/arp.h>
 
 /** @file
@@ -45,291 +41,30 @@ FILE_LICENCE ( GPL2_OR_LATER );
  *
  */
 
-/** ARP minimum timeout */
-#define ARP_MIN_TIMEOUT ( TICKS_PER_SEC / 8 )
-
-/** ARP maximum timeout */
-#define ARP_MAX_TIMEOUT ( TICKS_PER_SEC * 3 )
-
-/** An ARP cache entry */
-struct arp_entry {
-	/** Reference count */
-	struct refcnt refcnt;
-	/** List of ARP cache entries */
-	struct list_head list;
-	/** Network device */
-	struct net_device *netdev;
-	/** Network-layer protocol */
-	struct net_protocol *net_protocol;
-	/** Network-layer destination address */
-	uint8_t net_dest[MAX_NET_ADDR_LEN];
-	/** Network-layer source address */
-	uint8_t net_source[MAX_NET_ADDR_LEN];
-	/** Link-layer destination address */
-	uint8_t ll_dest[MAX_LL_ADDR_LEN];
-	/** Retransmission timer */
-	struct retry_timer timer;
-	/** Pending I/O buffers */
-	struct list_head tx_queue;
-};
-
-/** The ARP cache */
-static LIST_HEAD ( arp_entries );
-
 struct net_protocol arp_protocol __net_protocol;
 
-static void arp_expired ( struct retry_timer *timer, int over );
-
-/**
- * Free ARP cache entry
- *
- * @v refcnt		Reference count
- */
-static void arp_free ( struct refcnt *refcnt ) {
-	struct arp_entry *arp =
-		container_of ( refcnt, struct arp_entry, refcnt );
-
-	/* Sanity check */
-	assert ( list_empty ( &arp->tx_queue ) );
-
-	/* Drop reference to network device */
-	netdev_put ( arp->netdev );
-
-	/* Free entry */
-	free ( arp );
-}
-
-/**
- * Create ARP cache entry
- *
- * @v netdev		Network device
- * @v net_protocol	Network-layer protocol
- * @v net_dest		Destination network-layer address
- * @v net_source	Source network-layer address
- * @ret arp		ARP cache entry, or NULL if allocation failed
- */
-static struct arp_entry * arp_create ( struct net_device *netdev,
-				       struct net_protocol *net_protocol,
-				       const void *net_dest,
-				       const void *net_source ) {
-	struct arp_entry *arp;
-
-	/* Allocate and initialise entry */
-	arp = zalloc ( sizeof ( *arp ) );
-	if ( ! arp )
-		return NULL;
-	ref_init ( &arp->refcnt, arp_free );
-	arp->netdev = netdev_get ( netdev );
-	arp->net_protocol = net_protocol;
-	memcpy ( arp->net_dest, net_dest,
-		 net_protocol->net_addr_len );
-	memcpy ( arp->net_source, net_source,
-		 net_protocol->net_addr_len );
-	timer_init ( &arp->timer, arp_expired, &arp->refcnt );
-	arp->timer.min_timeout = ARP_MIN_TIMEOUT;
-	arp->timer.max_timeout = ARP_MAX_TIMEOUT;
-	INIT_LIST_HEAD ( &arp->tx_queue );
-
-	/* Start timer running to trigger initial transmission */
-	start_timer_nodelay ( &arp->timer );
-
-	/* Transfer ownership to cache */
-	list_add ( &arp->list, &arp_entries );
-
-	DBGC ( arp, "ARP %p %s %s %s created\n", arp, netdev->name,
-	       net_protocol->name, net_protocol->ntoa ( net_dest ) );
-	return arp;
-}
-
-/**
- * Find entry in the ARP cache
- *
- * @v netdev		Network device
- * @v net_protocol	Network-layer protocol
- * @v net_dest		Destination network-layer address
- * @ret arp		ARP cache entry, or NULL if not found
- */
-static struct arp_entry * arp_find ( struct net_device *netdev,
-				     struct net_protocol *net_protocol,
-				     const void *net_dest ) {
-	struct arp_entry *arp;
-
-	list_for_each_entry ( arp, &arp_entries, list ) {
-		if ( ( arp->netdev == netdev ) &&
-		     ( arp->net_protocol == net_protocol ) &&
-		     ( memcmp ( arp->net_dest, net_dest,
-				net_protocol->net_addr_len ) == 0 ) ) {
-
-			/* Move to start of cache */
-			list_del ( &arp->list );
-			list_add ( &arp->list, &arp_entries );
-
-			return arp;
-		}
-	}
-	return NULL;
-}
-
-/**
- * Destroy ARP cache entry
- *
- * @v arp		ARP cache entry
- * @v rc		Reason for destruction
- */
-static void arp_destroy ( struct arp_entry *arp, int rc ) {
-	struct net_device *netdev = arp->netdev;
-	struct net_protocol *net_protocol = arp->net_protocol;
-	struct io_buffer *iobuf;
-
-	/* Take ownership from cache */
-	list_del ( &arp->list );
-
-	/* Stop timer */
-	stop_timer ( &arp->timer );
-
-	/* Discard any outstanding I/O buffers */
-	while ( ( iobuf = list_first_entry ( &arp->tx_queue, struct io_buffer,
-					     list ) ) != NULL ) {
-		DBGC2 ( arp, "ARP %p %s %s %s discarding deferred packet: "
-			"%s\n", arp, netdev->name, net_protocol->name,
-			net_protocol->ntoa ( arp->net_dest ), strerror ( rc ) );
-		list_del ( &iobuf->list );
-		netdev_tx_err ( arp->netdev, iobuf, rc );
-	}
-
-	DBGC ( arp, "ARP %p %s %s %s destroyed: %s\n", arp, netdev->name,
-	       net_protocol->name, net_protocol->ntoa ( arp->net_dest ),
-	       strerror ( rc ) );
-
-	/* Drop remaining reference */
-	ref_put ( &arp->refcnt );
-}
-
 /**
- * Test if ARP cache entry has a valid link-layer address
+ * Transmit ARP request
  *
- * @v arp		ARP cache entry
- * @ret resolved	ARP cache entry is resolved
- */
-static inline int arp_resolved ( struct arp_entry *arp ) {
-	return ( ! timer_running ( &arp->timer ) );
-}
-
-/**
- * Transmit packet, determining link-layer address via ARP
- *
- * @v iobuf		I/O buffer
  * @v netdev		Network device
  * @v net_protocol	Network-layer protocol
  * @v net_dest		Destination network-layer address
  * @v net_source	Source network-layer address
- * @v ll_source		Source link-layer address
  * @ret rc		Return status code
  */
-int arp_tx ( struct io_buffer *iobuf, struct net_device *netdev,
-	     struct net_protocol *net_protocol, const void *net_dest,
-	     const void *net_source, const void *ll_source ) {
-	struct arp_entry *arp;
-
-	/* Find or create ARP cache entry */
-	arp = arp_find ( netdev, net_protocol, net_dest );
-	if ( ! arp ) {
-		arp = arp_create ( netdev, net_protocol, net_dest,
-				   net_source );
-		if ( ! arp )
-			return -ENOMEM;
-	}
-
-	/* If a link-layer address is available then transmit
-	 * immediately, otherwise queue for later transmission.
-	 */
-	if ( arp_resolved ( arp ) ) {
-		return net_tx ( iobuf, netdev, net_protocol, arp->ll_dest,
-				ll_source );
-	} else {
-		DBGC2 ( arp, "ARP %p %s %s %s deferring packet\n",
-			arp, netdev->name, net_protocol->name,
-			net_protocol->ntoa ( net_dest ) );
-		list_add_tail ( &iobuf->list, &arp->tx_queue );
-		return -EAGAIN;
-	}
-}
-
-/**
- * Update ARP cache entry
- *
- * @v arp		ARP cache entry
- * @v ll_dest		Destination link-layer address
- */
-static void arp_update ( struct arp_entry *arp, const void *ll_dest ) {
-	struct net_device *netdev = arp->netdev;
-	struct ll_protocol *ll_protocol = netdev->ll_protocol;
-	struct net_protocol *net_protocol = arp->net_protocol;
-	struct io_buffer *iobuf;
-	int rc;
-
-	DBGC ( arp, "ARP %p %s %s %s updated => %s\n", arp, netdev->name,
-	       net_protocol->name, net_protocol->ntoa ( arp->net_dest ),
-	       ll_protocol->ntoa ( ll_dest ) );
-
-	/* Fill in link-layer address */
-	memcpy ( arp->ll_dest, ll_dest, ll_protocol->ll_addr_len );
-
-	/* Stop retransmission timer */
-	stop_timer ( &arp->timer );
-
-	/* Transmit any packets in queue.  Take out a temporary
-	 * reference on the entry to prevent it from going out of
-	 * scope during the call to net_tx().
-	 */
-	ref_get ( &arp->refcnt );
-	while ( ( iobuf = list_first_entry ( &arp->tx_queue, struct io_buffer,
-					     list ) ) != NULL ) {
-		DBGC2 ( arp, "ARP %p %s %s %s transmitting deferred packet\n",
-			arp, netdev->name, net_protocol->name,
-			net_protocol->ntoa ( arp->net_dest ) );
-		list_del ( &iobuf->list );
-		if ( ( rc = net_tx ( iobuf, netdev, net_protocol, ll_dest,
-				     netdev->ll_addr ) ) != 0 ) {
-			DBGC ( arp, "ARP %p could not transmit deferred "
-			       "packet: %s\n", arp, strerror ( rc ) );
-			/* Ignore error and continue */
-		}
-	}
-	ref_put ( &arp->refcnt );
-}
-
-/**
- * Handle ARP timer expiry
- *
- * @v timer		Retry timer
- * @v fail		Failure indicator
- */
-static void arp_expired ( struct retry_timer *timer, int fail ) {
-	struct arp_entry *arp = container_of ( timer, struct arp_entry, timer );
-	struct net_device *netdev = arp->netdev;
+static int arp_tx_request ( struct net_device *netdev,
+			    struct net_protocol *net_protocol,
+			    const void *net_dest, const void *net_source ) {
 	struct ll_protocol *ll_protocol = netdev->ll_protocol;
-	struct net_protocol *net_protocol = arp->net_protocol;
 	struct io_buffer *iobuf;
 	struct arphdr *arphdr;
 	int rc;
 
-	/* If we have failed, destroy the cache entry */
-	if ( fail ) {
-		arp_destroy ( arp, -ETIMEDOUT );
-		return;
-	}
-
-	/* Restart the timer */
-	start_timer ( &arp->timer );
-
 	/* Allocate ARP packet */
 	iobuf = alloc_iob ( MAX_LL_HEADER_LEN + sizeof ( *arphdr ) +
 			    ( 2 * ( MAX_LL_ADDR_LEN + MAX_NET_ADDR_LEN ) ) );
-	if ( ! iobuf ) {
-		/* Leave timer running and try again later */
-		return;
-	}
+	if ( ! iobuf )
+		return -ENOMEM;
 	iob_reserve ( iobuf, MAX_LL_HEADER_LEN );
 
 	/* Build up ARP request */
@@ -342,21 +77,30 @@ static void arp_expired ( struct retry_timer *timer, int fail ) {
 	memcpy ( iob_put ( iobuf, ll_protocol->ll_addr_len ),
 		 netdev->ll_addr, ll_protocol->ll_addr_len );
 	memcpy ( iob_put ( iobuf, net_protocol->net_addr_len ),
-		 arp->net_source, net_protocol->net_addr_len );
+		 net_source, net_protocol->net_addr_len );
 	memset ( iob_put ( iobuf, ll_protocol->ll_addr_len ),
 		 0, ll_protocol->ll_addr_len );
 	memcpy ( iob_put ( iobuf, net_protocol->net_addr_len ),
-		 arp->net_dest, net_protocol->net_addr_len );
+		 net_dest, net_protocol->net_addr_len );
 
 	/* Transmit ARP request */
 	if ( ( rc = net_tx ( iobuf, netdev, &arp_protocol,
 			     netdev->ll_broadcast, netdev->ll_addr ) ) != 0 ) {
-		DBGC ( arp, "ARP %p could not transmit request: %s\n",
-		       arp, strerror ( rc ) );
-		return;
+		DBGC ( netdev, "ARP %s %s %s could not transmit request: %s\n",
+		       netdev->name, net_protocol->name,
+		       net_protocol->ntoa ( net_dest ), strerror ( rc ) );
+		return rc;
 	}
+
+	return 0;
 }
 
+/** ARP neighbour discovery protocol */
+struct neighbour_discovery arp_discovery = {
+	.name = "ARP",
+	.tx_request = arp_tx_request,
+};
+
 /**
  * Identify ARP protocol
  *
@@ -368,9 +112,8 @@ static struct arp_net_protocol * arp_find_protocol ( uint16_t net_proto ) {
 	struct arp_net_protocol *arp_net_protocol;
 
 	for_each_table_entry ( arp_net_protocol, ARP_NET_PROTOCOLS ) {
-		if ( arp_net_protocol->net_protocol->net_proto == net_proto ) {
+		if ( arp_net_protocol->net_protocol->net_proto == net_proto )
 			return arp_net_protocol;
-		}
 	}
 	return NULL;
 }
@@ -392,7 +135,6 @@ static int arp_rx ( struct io_buffer *iobuf, struct net_device *netdev,
 	struct arp_net_protocol *arp_net_protocol;
 	struct net_protocol *net_protocol;
 	struct ll_protocol *ll_protocol;
-	struct arp_entry *arp;
 	int rc;
 
 	/* Identify network-layer and link-layer protocols */
@@ -412,11 +154,9 @@ static int arp_rx ( struct io_buffer *iobuf, struct net_device *netdev,
 		goto done;
 	}
 
-	/* See if we have an entry for this sender, and update it if so */
-	arp = arp_find ( netdev, net_protocol, arp_sender_pa ( arphdr ) );
-	if ( arp ) {
-		arp_update ( arp, arp_sender_ha ( arphdr ) );
-	}
+	/* Update neighbour cache entry for this sender, if any */
+	neighbour_update ( netdev, net_protocol, arp_sender_pa ( arphdr ),
+			   arp_sender_ha ( arphdr ) );
 
 	/* If it's not a request, there's nothing more to do */
 	if ( arphdr->ar_op != htons ( ARPOP_REQUEST ) ) {
@@ -431,10 +171,10 @@ static int arp_rx ( struct io_buffer *iobuf, struct net_device *netdev,
 	}
 
 	/* Change request to a reply */
-	DBGC ( netdev, "ARP reply %s %s %s => %s %s\n",
-	       netdev->name, net_protocol->name,
-	       net_protocol->ntoa ( arp_target_pa ( arphdr ) ),
-	       ll_protocol->name, ll_protocol->ntoa ( netdev->ll_addr ) );
+	DBGC2 ( netdev, "ARP %s %s %s reply => %s %s\n",
+		netdev->name, net_protocol->name,
+		net_protocol->ntoa ( arp_target_pa ( arphdr ) ),
+		ll_protocol->name, ll_protocol->ntoa ( netdev->ll_addr ) );
 	arphdr->ar_op = htons ( ARPOP_REPLY );
 	memswap ( arp_sender_ha ( arphdr ), arp_target_ha ( arphdr ),
 		 arphdr->ar_hln + arphdr->ar_pln );
@@ -444,8 +184,10 @@ static int arp_rx ( struct io_buffer *iobuf, struct net_device *netdev,
 	if ( ( rc = net_tx ( iob_disown ( iobuf ), netdev, &arp_protocol,
 			     arp_target_ha ( arphdr ),
 			     netdev->ll_addr ) ) != 0 ) {
-		DBGC ( netdev, "ARP could not transmit reply via %s: %s\n",
-		       netdev->name, strerror ( rc ) );
+		DBGC ( netdev, "ARP %s %s %s could not transmit reply: %s\n",
+		       netdev->name, net_protocol->name,
+		       net_protocol->ntoa ( arp_target_pa ( arphdr ) ),
+		       strerror ( rc ) );
 		goto done;
 	}
 
@@ -469,72 +211,10 @@ static const char * arp_ntoa ( const void *net_addr __unused ) {
 	return "<ARP>";
 }
 
-/** ARP protocol */
+/** ARP network protocol */
 struct net_protocol arp_protocol __net_protocol = {
 	.name = "ARP",
 	.net_proto = htons ( ETH_P_ARP ),
 	.rx = arp_rx,
 	.ntoa = arp_ntoa,
 };
-
-/**
- * Update ARP cache on network device creation
- *
- * @v netdev		Network device
- */
-static int arp_probe ( struct net_device *netdev __unused ) {
-	/* Nothing to do */
-	return 0;
-}
-
-/**
- * Update ARP cache on network device state change or removal
- *
- * @v netdev		Network device
- */
-static void arp_flush ( struct net_device *netdev ) {
-	struct arp_entry *arp;
-	struct arp_entry *tmp;
-
-	/* Remove all ARP cache entries when a network device is closed */
-	if ( ! netdev_is_open ( netdev ) ) {
-		list_for_each_entry_safe ( arp, tmp, &arp_entries, list )
-			arp_destroy ( arp, -ENODEV );
-	}
-}
-
-/** ARP driver (for net device notifications) */
-struct net_driver arp_net_driver __net_driver = {
-	.name = "ARP",
-	.probe = arp_probe,
-	.notify = arp_flush,
-	.remove = arp_flush,
-};
-
-/**
- * Discard some cached ARP entries
- *
- * @ret discarded	Number of cached items discarded
- */
-static unsigned int arp_discard ( void ) {
-	struct arp_entry *arp;
-
-	/* Drop oldest cache entry, if any */
-	arp = list_last_entry ( &arp_entries, struct arp_entry, list );
-	if ( arp ) {
-		arp_destroy ( arp, -ENOBUFS );
-		return 1;
-	} else {
-		return 0;
-	}
-}
-
-/** ARP cache discarder
- *
- * ARP cache entries are deemed to have a high replacement cost, since
- * flushing an active ARP cache entry midway through a TCP transfer
- * will cause substantial disruption.
- */
-struct cache_discarder arp_discarder __cache_discarder ( CACHE_EXPENSIVE ) = {
-	.discard = arp_discard,
-};
diff --git a/src/net/neighbour.c b/src/net/neighbour.c
new file mode 100644
index 00000000..210b7138
--- /dev/null
+++ b/src/net/neighbour.c
@@ -0,0 +1,479 @@
+/*
+ * Copyright (C) 2013 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER );
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include <ipxe/refcnt.h>
+#include <ipxe/list.h>
+#include <ipxe/iobuf.h>
+#include <ipxe/retry.h>
+#include <ipxe/timer.h>
+#include <ipxe/malloc.h>
+#include <ipxe/neighbour.h>
+
+/** @file
+ *
+ * Neighbour discovery
+ *
+ * This file implements the abstract functions of neighbour discovery,
+ * independent of the underlying network protocol (e.g. ARP or NDP).
+ *
+ */
+
+/** A neighbour cache entry */
+struct neighbour {
+	/** Reference count */
+	struct refcnt refcnt;
+	/** List of neighbour cache entries */
+	struct list_head list;
+
+	/** Network device */
+	struct net_device *netdev;
+	/** Network-layer protocol */
+	struct net_protocol *net_protocol;
+	/** Network-layer destination address */
+	uint8_t net_dest[MAX_NET_ADDR_LEN];
+	/** Link-layer destination address */
+	uint8_t ll_dest[MAX_LL_ADDR_LEN];
+
+	/** Neighbour discovery protocol (if any) */
+	struct neighbour_discovery *discovery;
+	/** Network-layer source address (if any) */
+	uint8_t net_source[MAX_NET_ADDR_LEN];
+	/** Retransmission timer */
+	struct retry_timer timer;
+
+	/** Pending I/O buffers */
+	struct list_head tx_queue;
+};
+
+/** Neighbour discovery minimum timeout */
+#define NEIGHBOUR_MIN_TIMEOUT ( TICKS_PER_SEC / 8 )
+
+/** Neighbour discovery maximum timeout */
+#define NEIGHBOUR_MAX_TIMEOUT ( TICKS_PER_SEC * 3 )
+
+/** The neighbour cache */
+static LIST_HEAD ( neighbours );
+
+static void neighbour_expired ( struct retry_timer *timer, int over );
+
+/**
+ * Free neighbour cache entry
+ *
+ * @v refcnt		Reference count
+ */
+static void neighbour_free ( struct refcnt *refcnt ) {
+	struct neighbour *neighbour =
+		container_of ( refcnt, struct neighbour, refcnt );
+
+	/* Sanity check */
+	assert ( list_empty ( &neighbour->tx_queue ) );
+
+	/* Drop reference to network device */
+	netdev_put ( neighbour->netdev );
+
+	/* Free neighbour */
+	free ( neighbour );
+}
+
+/**
+ * Test if neighbour cache entry has a valid link-layer address
+ *
+ * @v neighbour		Neighbour cache entry
+ * @ret has_ll_dest	Neighbour cache entry has a valid link-layer address
+ */
+static inline __attribute__ (( always_inline )) int
+neighbour_has_ll_dest ( struct neighbour *neighbour ) {
+	return ( ! timer_running ( &neighbour->timer ) );
+}
+
+/**
+ * Create neighbour cache entry
+ *
+ * @v netdev		Network device
+ * @v net_protocol	Network-layer protocol
+ * @v net_dest		Destination network-layer address
+ * @ret neighbour	Neighbour cache entry, or NULL if allocation failed
+ */
+static struct neighbour * neighbour_create ( struct net_device *netdev,
+					     struct net_protocol *net_protocol,
+					     const void *net_dest ) {
+	struct neighbour *neighbour;
+
+	/* Allocate and initialise entry */
+	neighbour = zalloc ( sizeof ( *neighbour ) );
+	if ( ! neighbour )
+		return NULL;
+	ref_init ( &neighbour->refcnt, neighbour_free );
+	neighbour->netdev = netdev_get ( netdev );
+	neighbour->net_protocol = net_protocol;
+	memcpy ( neighbour->net_dest, net_dest,
+		 net_protocol->net_addr_len );
+	timer_init ( &neighbour->timer, neighbour_expired, &neighbour->refcnt );
+	neighbour->timer.min_timeout = NEIGHBOUR_MIN_TIMEOUT;
+	neighbour->timer.max_timeout = NEIGHBOUR_MAX_TIMEOUT;
+	INIT_LIST_HEAD ( &neighbour->tx_queue );
+
+	/* Transfer ownership to cache */
+	list_add ( &neighbour->list, &neighbours );
+
+	DBGC ( neighbour, "NEIGHBOUR %s %s %s created\n", netdev->name,
+	       net_protocol->name, net_protocol->ntoa ( net_dest ) );
+	return neighbour;
+}
+
+/**
+ * Find neighbour cache entry
+ *
+ * @v netdev		Network device
+ * @v net_protocol	Network-layer protocol
+ * @v net_dest		Destination network-layer address
+ * @ret neighbour	Neighbour cache entry, or NULL if not found
+ */
+static struct neighbour * neighbour_find ( struct net_device *netdev,
+					   struct net_protocol *net_protocol,
+					   const void *net_dest ) {
+	struct neighbour *neighbour;
+
+	list_for_each_entry ( neighbour, &neighbours, list ) {
+		if ( ( neighbour->netdev == netdev ) &&
+		     ( neighbour->net_protocol == net_protocol ) &&
+		     ( memcmp ( neighbour->net_dest, net_dest,
+				net_protocol->net_addr_len ) == 0 ) ) {
+
+			/* Move to start of cache */
+			list_del ( &neighbour->list );
+			list_add ( &neighbour->list, &neighbours );
+
+			return neighbour;
+		}
+	}
+	return NULL;
+}
+
+/**
+ * Start neighbour discovery
+ *
+ * @v neighbour		Neighbour cache entry
+ * @v discovery		Neighbour discovery protocol
+ * @v net_source	Source network-layer address
+ */
+static void neighbour_discover ( struct neighbour *neighbour,
+				 struct neighbour_discovery *discovery,
+				 const void *net_source ) {
+	struct net_device *netdev = neighbour->netdev;
+	struct net_protocol *net_protocol = neighbour->net_protocol;
+
+	/* Record discovery protocol and source network-layer address */
+	neighbour->discovery = discovery;
+	memcpy ( neighbour->net_source, net_source,
+		 net_protocol->net_addr_len );
+
+	/* Start timer to trigger neighbour discovery */
+	start_timer_nodelay ( &neighbour->timer );
+
+	DBGC ( neighbour, "NEIGHBOUR %s %s %s discovering via %s\n",
+	       netdev->name, net_protocol->name,
+	       net_protocol->ntoa ( neighbour->net_dest ),
+	       neighbour->discovery->name );
+}
+
+/**
+ * Complete neighbour discovery
+ *
+ * @v neighbour		Neighbour cache entry
+ * @v ll_dest		Destination link-layer address
+ */
+static void neighbour_discovered ( struct neighbour *neighbour,
+				   const void *ll_dest ) {
+	struct net_device *netdev = neighbour->netdev;
+	struct ll_protocol *ll_protocol = netdev->ll_protocol;
+	struct net_protocol *net_protocol = neighbour->net_protocol;
+	struct io_buffer *iobuf;
+	int rc;
+
+	/* Fill in link-layer address */
+	memcpy ( neighbour->ll_dest, ll_dest, ll_protocol->ll_addr_len );
+	DBGC ( neighbour, "NEIGHBOUR %s %s %s is %s %s\n", netdev->name,
+	       net_protocol->name, net_protocol->ntoa ( neighbour->net_dest ),
+	       ll_protocol->name, ll_protocol->ntoa ( neighbour->ll_dest ) );
+
+	/* Stop retransmission timer */
+	stop_timer ( &neighbour->timer );
+
+	/* Transmit any packets in queue.  Take out a temporary
+	 * reference on the entry to prevent it from going out of
+	 * scope during the call to net_tx().
+	 */
+	ref_get ( &neighbour->refcnt );
+	while ( ( iobuf = list_first_entry ( &neighbour->tx_queue,
+					     struct io_buffer, list )) != NULL){
+		DBGC2 ( neighbour, "NEIGHBOUR %s %s %s transmitting deferred "
+			"packet\n", netdev->name, net_protocol->name,
+			net_protocol->ntoa ( neighbour->net_dest ) );
+		list_del ( &iobuf->list );
+		if ( ( rc = net_tx ( iobuf, netdev, net_protocol, ll_dest,
+				     netdev->ll_addr ) ) != 0 ) {
+			DBGC ( neighbour, "NEIGHBOUR %s %s %s could not "
+			       "transmit deferred packet: %s\n",
+			       netdev->name, net_protocol->name,
+			       net_protocol->ntoa ( neighbour->net_dest ),
+			       strerror ( rc ) );
+			/* Ignore error and continue */
+		}
+	}
+	ref_put ( &neighbour->refcnt );
+}
+
+/**
+ * Destroy neighbour cache entry
+ *
+ * @v neighbour		Neighbour cache entry
+ * @v rc		Reason for destruction
+ */
+static void neighbour_destroy ( struct neighbour *neighbour, int rc ) {
+	struct net_device *netdev = neighbour->netdev;
+	struct net_protocol *net_protocol = neighbour->net_protocol;
+	struct io_buffer *iobuf;
+
+	/* Take ownership from cache */
+	list_del ( &neighbour->list );
+
+	/* Stop timer */
+	stop_timer ( &neighbour->timer );
+
+	/* Discard any outstanding I/O buffers */
+	while ( ( iobuf = list_first_entry ( &neighbour->tx_queue,
+					     struct io_buffer, list )) != NULL){
+		DBGC2 ( neighbour, "NEIGHBOUR %s %s %s discarding deferred "
+			"packet: %s\n", netdev->name, net_protocol->name,
+			net_protocol->ntoa ( neighbour->net_dest ),
+			strerror ( rc ) );
+		list_del ( &iobuf->list );
+		netdev_tx_err ( neighbour->netdev, iobuf, rc );
+	}
+
+	DBGC ( neighbour, "NEIGHBOUR %s %s %s destroyed: %s\n", netdev->name,
+	       net_protocol->name, net_protocol->ntoa ( neighbour->net_dest ),
+	       strerror ( rc ) );
+
+	/* Drop remaining reference */
+	ref_put ( &neighbour->refcnt );
+}
+
+/**
+ * Handle neighbour timer expiry
+ *
+ * @v timer		Retry timer
+ * @v fail		Failure indicator
+ */
+static void neighbour_expired ( struct retry_timer *timer, int fail ) {
+	struct neighbour *neighbour =
+		container_of ( timer, struct neighbour, timer );
+	struct net_device *netdev = neighbour->netdev;
+	struct net_protocol *net_protocol = neighbour->net_protocol;
+	struct neighbour_discovery *discovery =
+		neighbour->discovery;
+	const void *net_dest = neighbour->net_dest;
+	const void *net_source = neighbour->net_source;
+	int rc;
+
+	/* If we have failed, destroy the cache entry */
+	if ( fail ) {
+		neighbour_destroy ( neighbour, -ETIMEDOUT );
+		return;
+	}
+
+	/* Restart the timer */
+	start_timer ( &neighbour->timer );
+
+	/* Transmit neighbour request */
+	if ( ( rc = discovery->tx_request ( netdev, net_protocol, net_dest,
+					    net_source ) ) != 0 ) {
+		DBGC ( neighbour, "NEIGHBOUR %s %s %s could not transmit %s "
+		       "request: %s\n", netdev->name, net_protocol->name,
+		       net_protocol->ntoa ( neighbour->net_dest ),
+		       neighbour->discovery->name, strerror ( rc ) );
+		/* Retransmit when timer expires */
+		return;
+	}
+}
+
+/**
+ * Transmit packet, determining link-layer address via neighbour discovery
+ *
+ * @v iobuf		I/O buffer
+ * @v netdev		Network device
+ * @v discovery		Neighbour discovery protocol
+ * @v net_protocol	Network-layer protocol
+ * @v net_dest		Destination network-layer address
+ * @v net_source	Source network-layer address
+ * @v ll_source		Source link-layer address
+ * @ret rc		Return status code
+ */
+int neighbour_tx ( struct io_buffer *iobuf, struct net_device *netdev,
+		   struct net_protocol *net_protocol, const void *net_dest,
+		   struct neighbour_discovery *discovery,
+		   const void *net_source, const void *ll_source ) {
+	struct neighbour *neighbour;
+
+	/* Find or create neighbour cache entry */
+	neighbour = neighbour_find ( netdev, net_protocol, net_dest );
+	if ( ! neighbour ) {
+		neighbour = neighbour_create ( netdev, net_protocol, net_dest );
+		if ( ! neighbour )
+			return -ENOMEM;
+		neighbour_discover ( neighbour, discovery, net_source );
+	}
+
+	/* If a link-layer address is available then transmit
+	 * immediately, otherwise queue for later transmission.
+	 */
+	if ( neighbour_has_ll_dest ( neighbour ) ) {
+		return net_tx ( iobuf, netdev, net_protocol, neighbour->ll_dest,
+				ll_source );
+	} else {
+		DBGC2 ( neighbour, "NEIGHBOUR %s %s %s deferring packet\n",
+			netdev->name, net_protocol->name,
+			net_protocol->ntoa ( net_dest ) );
+		list_add_tail ( &iobuf->list, &neighbour->tx_queue );
+		return -EAGAIN;
+	}
+}
+
+/**
+ * Update existing neighbour cache entry
+ *
+ * @v netdev		Network device
+ * @v net_protocol	Network-layer protocol
+ * @v net_dest		Destination network-layer address
+ * @v ll_dest		Destination link-layer address
+ * @ret rc		Return status code
+ */
+int neighbour_update ( struct net_device *netdev,
+		       struct net_protocol *net_protocol,
+		       const void *net_dest, const void *ll_dest ) {
+	struct neighbour *neighbour;
+
+	/* Find neighbour cache entry */
+	neighbour = neighbour_find ( netdev, net_protocol, net_dest );
+	if ( ! neighbour )
+		return -ENOENT;
+
+	/* Set destination address */
+	neighbour_discovered ( neighbour, ll_dest );
+
+	return 0;
+}
+
+/**
+ * Define neighbour cache entry
+ *
+ * @v netdev		Network device
+ * @v net_protocol	Network-layer protocol
+ * @v net_dest		Destination network-layer address
+ * @v ll_dest		Destination link-layer address, if known
+ * @ret rc		Return status code
+ */
+int neighbour_define ( struct net_device *netdev,
+		       struct net_protocol *net_protocol,
+		       const void *net_dest, const void *ll_dest ) {
+	struct neighbour *neighbour;
+
+	/* Find or create neighbour cache entry */
+	neighbour = neighbour_find ( netdev, net_protocol, net_dest );
+	if ( ! neighbour ) {
+		neighbour = neighbour_create ( netdev, net_protocol, net_dest );
+		if ( ! neighbour )
+			return -ENOMEM;
+	}
+
+	/* Set destination address */
+	neighbour_discovered ( neighbour, ll_dest );
+
+	return 0;
+}
+
+/**
+ * Update neighbour cache on network device creation
+ *
+ * @v netdev		Network device
+ */
+static int neighbour_probe ( struct net_device *netdev __unused ) {
+	/* Nothing to do */
+	return 0;
+}
+
+/**
+ * Update neighbour cache on network device state change or removal
+ *
+ * @v netdev		Network device
+ */
+static void neighbour_flush ( struct net_device *netdev ) {
+	struct neighbour *neighbour;
+	struct neighbour *tmp;
+
+	/* Remove all neighbour cache entries when a network device is closed */
+	if ( ! netdev_is_open ( netdev ) ) {
+		list_for_each_entry_safe ( neighbour, tmp, &neighbours, list )
+			neighbour_destroy ( neighbour, -ENODEV );
+	}
+}
+
+/** Neighbour driver (for net device notifications) */
+struct net_driver neighbour_net_driver __net_driver = {
+	.name = "Neighbour",
+	.probe = neighbour_probe,
+	.notify = neighbour_flush,
+	.remove = neighbour_flush,
+};
+
+/**
+ * Discard some cached neighbour entries
+ *
+ * @ret discarded	Number of cached items discarded
+ */
+static unsigned int neighbour_discard ( void ) {
+	struct neighbour *neighbour;
+
+	/* Drop oldest cache entry, if any */
+	neighbour = list_last_entry ( &neighbours, struct neighbour, list );
+	if ( neighbour ) {
+		neighbour_destroy ( neighbour, -ENOBUFS );
+		return 1;
+	} else {
+		return 0;
+	}
+}
+
+/**
+ * Neighbour cache discarder
+ *
+ * Neighbour cache entries are deemed to have a high replacement cost,
+ * since flushing an active neighbour cache entry midway through a TCP
+ * transfer will cause substantial disruption.
+ */
+struct cache_discarder neighbour_discarder __cache_discarder (CACHE_EXPENSIVE)={
+	.discard = neighbour_discard,
+};