[git commit master 1/1] dhcprelay: code shrink, and explain its workings a bit more

Denys Vlasenko vda.linux at googlemail.com
Tue Oct 5 23:45:24 UTC 2010


commit: http://git.busybox.net/busybox/commit/?id=c0f39b0fb288206036d2c442588cf597390c3622
branch: http://git.busybox.net/busybox/commit/?id=refs/heads/master

function                                             old     new   delta
sendto_ip4                                             -      55     +55
dhcprelay_main                                      1059     942    -117
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 0/1 up/down: 55/-117)           Total: -62 bytes

Signed-off-by: Denys Vlasenko <vda.linux at googlemail.com>
---
 networking/udhcp/dhcprelay.c |  189 ++++++++++++++++++++++++++----------------
 1 files changed, 118 insertions(+), 71 deletions(-)

diff --git a/networking/udhcp/dhcprelay.c b/networking/udhcp/dhcprelay.c
index a2c7f35..759a4ba 100644
--- a/networking/udhcp/dhcprelay.c
+++ b/networking/udhcp/dhcprelay.c
@@ -11,9 +11,12 @@
  */
 #include "common.h"
 
-#define SERVER_PORT      67
-#define SELECT_TIMEOUT    5 /* select timeout in sec. */
-#define MAX_LIFETIME   2*60 /* lifetime of an xid entry in sec. */
+#define SERVER_PORT    67
+
+/* lifetime of an xid entry in sec. */
+#define MAX_LIFETIME   2*60
+/* select timeout in sec. */
+#define SELECT_TIMEOUT (MAX_LIFETIME / 8)
 
 /* This list holds information about clients. The xid_* functions manipulate this list. */
 struct xid_item {
@@ -67,11 +70,11 @@ static struct xid_item *xid_find(uint32_t xid)
 	struct xid_item *item = dhcprelay_xid_list.next;
 	while (item != NULL) {
 		if (item->xid == xid) {
-			return item;
+			break;
 		}
 		item = item->next;
 	}
-	return NULL;
+	return item;
 }
 
 static void xid_del(uint32_t xid)
@@ -110,62 +113,72 @@ static int get_dhcp_packet_type(struct dhcp_packet *p)
 }
 
 /**
- * get_client_devices - parses the devices list
- * dev_list - comma separated list of devices
+ * make_iface_list - parses client/server interface names
  * returns array
  */
-static char **get_client_devices(char *dev_list, int *client_number)
+static char **make_iface_list(char **client_and_server_ifaces, int *client_number)
 {
-	char *s, **client_dev;
+	char *s, **iface_list;
 	int i, cn;
 
-	/* copy list */
-	dev_list = xstrdup(dev_list);
-
-	/* get number of items, replace ',' with NULs */
-	s = dev_list;
-	cn = 1;
+	/* get number of items */
+	cn = 2; /* 1 server iface + at least 1 client one */
+	s = client_and_server_ifaces[0]; /* list of client ifaces */
 	while (*s) {
-		if (*s == ',') {
-			*s = '\0';
+		if (*s == ',')
 			cn++;
-		}
 		s++;
 	}
 	*client_number = cn;
 
 	/* create vector of pointers */
-	client_dev = xzalloc(cn * sizeof(*client_dev));
-	client_dev[0] = dev_list;
+	iface_list = xzalloc(cn * sizeof(iface_list[0]));
+
+	iface_list[0] = client_and_server_ifaces[1]; /* server iface */
+
 	i = 1;
-	while (i != cn) {
-		client_dev[i] = client_dev[i - 1] + strlen(client_dev[i - 1]) + 1;
-		i++;
+	s = xstrdup(client_and_server_ifaces[0]); /* list of client ifaces */
+	goto store_client_iface_name;
+
+	while (i < cn) {
+		if (*s++ == ',') {
+			s[-1] = '\0';
+ store_client_iface_name:
+			iface_list[i++] = s;
+		}
 	}
-	return client_dev;
+
+	return iface_list;
 }
 
 /* Creates listen sockets (in fds) bound to client and server ifaces,
  * and returns numerically max fd.
  */
-static int init_sockets(char **client_ifaces, int num_clients,
-			char *server_iface, int *fds)
+static int init_sockets(char **iface_list, int num_clients, int *fds)
 {
 	int i, n;
 
-	/* talk to real server on bootps */
-	fds[0] = udhcp_listen_socket(/*INADDR_ANY,*/ SERVER_PORT, server_iface);
-	n = fds[0];
-
-	for (i = 1; i < num_clients; i++) {
-		/* listen for clients on bootps */
-		fds[i] = udhcp_listen_socket(/*INADDR_ANY,*/ SERVER_PORT, client_ifaces[i-1]);
-		if (fds[i] > n)
+	n = 0;
+	for (i = 0; i < num_clients; i++) {
+		fds[i] = udhcp_listen_socket(/*INADDR_ANY,*/ SERVER_PORT, iface_list[i]);
+		if (n < fds[i])
 			n = fds[i];
 	}
 	return n;
 }
 
+static int sendto_ip4(int sock, const void *msg, int msg_len, struct sockaddr_in *to)
+{
+	int err;
+
+	errno = 0;
+	err = sendto(sock, msg, msg_len, 0, (struct sockaddr*) to, sizeof(*to));
+	err -= msg_len;
+	if (err)
+		bb_perror_msg("sendto");
+	return err;
+}
+
 /**
  * pass_to_server() - forwards dhcp packets from client to server
  * p - packet to send
@@ -174,7 +187,7 @@ static int init_sockets(char **client_ifaces, int num_clients,
 static void pass_to_server(struct dhcp_packet *p, int packet_len, int client, int *fds,
 			struct sockaddr_in *client_addr, struct sockaddr_in *server_addr)
 {
-	int res, type;
+	int type;
 
 	/* check packet_type */
 	type = get_dhcp_packet_type(p);
@@ -188,13 +201,12 @@ static void pass_to_server(struct dhcp_packet *p, int packet_len, int client, in
 	/* create new xid entry */
 	xid_add(p->xid, client_addr, client);
 
-	/* forward request to LAN (server) */
-	errno = 0;
-	res = sendto(fds[0], p, packet_len, 0, (struct sockaddr*)server_addr,
-			sizeof(struct sockaddr_in));
-	if (res != packet_len) {
-		bb_perror_msg("sendto");
-	}
+	/* forward request to server */
+	/* note that we send from fds[0] which is bound to SERVER_PORT (67).
+	 * IOW: we send _from_ SERVER_PORT! Although this may look strange,
+	 * RFC 1542 not only allows, but prescribes this for BOOTP relays.
+	 */
+	sendto_ip4(fds[0], p, packet_len, server_addr);
 }
 
 /**
@@ -203,7 +215,7 @@ static void pass_to_server(struct dhcp_packet *p, int packet_len, int client, in
  */
 static void pass_to_client(struct dhcp_packet *p, int packet_len, int *fds)
 {
-	int res, type;
+	int type;
 	struct xid_item *item;
 
 	/* check xid */
@@ -218,14 +230,12 @@ static void pass_to_client(struct dhcp_packet *p, int packet_len, int *fds)
 		return;
 	}
 
+//TODO: also do it if (p->flags & htons(BROADCAST_FLAG)) is set!
 	if (item->ip.sin_addr.s_addr == htonl(INADDR_ANY))
 		item->ip.sin_addr.s_addr = htonl(INADDR_BROADCAST);
-	errno = 0;
-	res = sendto(fds[item->client], p, packet_len, 0, (struct sockaddr*) &(item->ip),
-			sizeof(item->ip));
-	if (res != packet_len) {
-		bb_perror_msg("sendto");
-		return;
+
+	if (sendto_ip4(fds[item->client], p, packet_len, &item->ip) != 0) {
+		return; /* send error occurred */
 	}
 
 	/* remove xid entry */
@@ -235,36 +245,30 @@ static void pass_to_client(struct dhcp_packet *p, int packet_len, int *fds)
 int dhcprelay_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
 int dhcprelay_main(int argc, char **argv)
 {
-	struct dhcp_packet dhcp_msg;
 	struct sockaddr_in server_addr;
-	struct sockaddr_in client_addr;
-	fd_set rfds;
-	char **client_ifaces;
+	char **iface_list;
 	int *fds;
 	int num_sockets, max_socket;
 	uint32_t our_nip;
 
 	server_addr.sin_family = AF_INET;
+	server_addr.sin_addr.s_addr = htonl(INADDR_BROADCAST);
 	server_addr.sin_port = htons(SERVER_PORT);
 
-	/* dhcprelay client_iface1,client_iface2,... server_iface [server_IP] */
+	/* dhcprelay CLIENT_IFACE1[,CLIENT_IFACE2...] SERVER_IFACE [SERVER_IP] */
 	if (argc == 4) {
 		if (!inet_aton(argv[3], &server_addr.sin_addr))
 			bb_perror_msg_and_die("bad server IP");
-	} else if (argc == 3) {
-		server_addr.sin_addr.s_addr = htonl(INADDR_BROADCAST);
-	} else {
+	} else if (argc != 3) {
 		bb_show_usage();
 	}
 
-	/* Produce list of client ifaces */
-	client_ifaces = get_client_devices(argv[1], &num_sockets);
+	iface_list = make_iface_list(argv + 1, &num_sockets);
 
-	num_sockets++; /* for server socket at fds[0] */
 	fds = xmalloc(num_sockets * sizeof(fds[0]));
 
 	/* Create sockets and bind one to every iface */
-	max_socket = init_sockets(client_ifaces, num_sockets, argv[2], fds);
+	max_socket = init_sockets(iface_list, num_sockets, fds);
 
 	/* Get our IP on server_iface */
 	if (udhcp_read_interface(argv[2], NULL, &our_nip, NULL))
@@ -272,11 +276,10 @@ int dhcprelay_main(int argc, char **argv)
 
 	/* Main loop */
 	while (1) {
-//reinit stuff from time to time? go back to get_client_devices
-//every N minutes?
+// reinit stuff from time to time? go back to make_iface_list
+// every N minutes?
+		fd_set rfds;
 		struct timeval tv;
-		size_t packlen;
-		socklen_t addr_size;
 		int i;
 
 		FD_ZERO(&rfds);
@@ -285,6 +288,9 @@ int dhcprelay_main(int argc, char **argv)
 		tv.tv_sec = SELECT_TIMEOUT;
 		tv.tv_usec = 0;
 		if (select(max_socket + 1, &rfds, NULL, NULL, &tv) > 0) {
+			int packlen;
+			struct dhcp_packet dhcp_msg;
+
 			/* server */
 			if (FD_ISSET(fds[0], &rfds)) {
 				packlen = udhcp_recv_kernel_packet(&dhcp_msg, fds[0]);
@@ -292,24 +298,65 @@ int dhcprelay_main(int argc, char **argv)
 					pass_to_client(&dhcp_msg, packlen, fds);
 				}
 			}
+
 			/* clients */
 			for (i = 1; i < num_sockets; i++) {
+				struct sockaddr_in client_addr;
+				socklen_t addr_size;
+
 				if (!FD_ISSET(fds[i], &rfds))
 					continue;
-				addr_size = sizeof(struct sockaddr_in);
+
+				addr_size = sizeof(client_addr);
 				packlen = recvfrom(fds[i], &dhcp_msg, sizeof(dhcp_msg), 0,
 						(struct sockaddr *)(&client_addr), &addr_size);
 				if (packlen <= 0)
 					continue;
 
 				/* Get our IP on corresponding client_iface */
-//why? what if server can't route such IP?
-				if (udhcp_read_interface(client_ifaces[i-1], NULL, &dhcp_msg.gateway_nip, NULL)) {
-					/* Fall back to our server_iface's IP */
-//this makes more sense!
+// RFC 1542
+// 4.1 General BOOTP Processing for Relay Agents
+// 4.1.1 BOOTREQUEST Messages
+//   If the relay agent does decide to relay the request, it MUST examine
+//   the 'giaddr' ("gateway" IP address) field.  If this field is zero,
+//   the relay agent MUST fill this field with the IP address of the
+//   interface on which the request was received.  If the interface has
+//   more than one IP address logically associated with it, the relay
+//   agent SHOULD choose one IP address associated with that interface and
+//   use it consistently for all BOOTP messages it relays.  If the
+//   'giaddr' field contains some non-zero value, the 'giaddr' field MUST
+//   NOT be modified.  The relay agent MUST NOT, under any circumstances,
+//   fill the 'giaddr' field with a broadcast address as is suggested in
+//   [1] (Section 8, sixth paragraph).
+
+// but why? what if server can't route such IP? Client ifaces may be, say, NATed!
+
+// 4.1.2 BOOTREPLY Messages
+//   BOOTP relay agents relay BOOTREPLY messages only to BOOTP clients.
+//   It is the responsibility of BOOTP servers to send BOOTREPLY messages
+//   directly to the relay agent identified in the 'giaddr' field.
+// (yeah right, unless it is impossible... see comment above)
+//   Therefore, a relay agent may assume that all BOOTREPLY messages it
+//   receives are intended for BOOTP clients on its directly-connected
+//   networks.
+//
+//   When a relay agent receives a BOOTREPLY message, it should examine
+//   the BOOTP 'giaddr', 'yiaddr', 'chaddr', 'htype', and 'hlen' fields.
+//   These fields should provide adequate information for the relay agent
+//   to deliver the BOOTREPLY message to the client.
+//
+//   The 'giaddr' field can be used to identify the logical interface from
+//   which the reply must be sent (i.e., the host or router interface
+//   connected to the same network as the BOOTP client).  If the content
+//   of the 'giaddr' field does not match one of the relay agent's
+//   directly-connected logical interfaces, the BOOTREPLY messsage MUST be
+//   silently discarded.
+				if (udhcp_read_interface(iface_list[i], NULL, &dhcp_msg.gateway_nip, NULL)) {
+					/* Fall back to our IP on server iface */
+// this makes more sense!
 					dhcp_msg.gateway_nip = our_nip;
 				}
-//maybe set dhcp_msg.flags |= BROADCAST_FLAG too?
+// maybe dhcp_msg.hops++? drop packets with too many hops (RFC 1542 says 4 or 16)?
 				pass_to_server(&dhcp_msg, packlen, i, fds, &client_addr, &server_addr);
 			}
 		}
-- 
1.7.1



More information about the busybox-cvs mailing list