This patch adds ZeroConf IP address allocation support to BusyBox, adding less than 3KB ARM code. Run by zcip eth0 /etc/zcip.script A sample script is included. It's a pretty close match to draft-ietf-zeroconf-ipv4-linklocal-17.txt (the July 2004 version) but this version has only had light testing so far: no busy networks, in particular. For now, avoid large deployments relying on this implementation without additional testing. --- busybox-1.00/docs/busybox_header.pod 2004-08-16 03:29:29.000000000 -0700 +++ busybox/docs/busybox_header.pod 2004-10-18 15:33:42.000000000 -0700 @@ -102,7 +102,7 @@ touch, tr, traceroute, true, tty, udhcpc, udhcpd, umount, uname, uncompress, uniq, unix2dos, unzip, uptime, usleep, uudecode, uuencode, vconfig, vi, vlock, watch, watchdog, wc, wget, which, who, whoami, xargs, - yes, zcat + yes, zcat, zcip =head1 COMMAND DESCRIPTIONS --- busybox-1.00/examples/zcip.script 1969-12-31 16:00:00.000000000 -0800 +++ busybox/examples/zcip.script 2004-10-18 15:33:42.000000000 -0700 @@ -0,0 +1,34 @@ +#!/bin/sh + +# only for use as a "zcip" callback script +if [ "x$interface" = x ] +then + exit 1 +fi + +# zcip should start on boot/resume and various media changes +case "$1" in +init) + # for now, zcip requires the link to be already up, + # and it drops links when they go down. that may + # not be the most robust model. + exit 0 + ;; +config) + if [ x$ip = x ] + then + exit 1 + fi + # FIXME remember $ip for $interface, to use on restart + exec ip address add dev $interface \ + scope link local $ip/16 broadcast + + ;; +deconfig) + if [ x$ip = x ] + then + exit 1 + fi + exec ip address del dev $interface local $ip + ;; +esac +exit 1 --- busybox-1.00/include/applets.h 2004-08-26 16:01:34.000000000 -0700 +++ busybox/include/applets.h 2004-10-18 15:33:42.000000000 -0700 @@ -670,6 +670,9 @@ #ifdef CONFIG_GUNZIP APPLET(zcat, gunzip_main, _BB_DIR_BIN, _BB_SUID_NEVER) #endif +#ifdef CONFIG_ZCIP + APPLET(zcip, zcip_main, _BB_DIR_SBIN, _BB_SUID_NEVER) +#endif #if !defined(PROTOTYPES) && !defined(MAKE_USAGE) { 0,NULL,0 } --- busybox-1.00/include/usage.h 2004-09-14 09:23:56.000000000 -0700 +++ busybox/include/usage.h 2004-10-18 15:33:42.000000000 -0700 @@ -2882,4 +2882,14 @@ #define zcat_full_usage \ "Uncompress to stdout." +#define zcip_trivial_usage \ + "[OPTIONS] ifname script" +#define zcip_full_usage \ + "zcip manages a ZeroConf IPv4 link-local address.\n" \ + "Options:\n" \ + "\t-f foreground mode\n" \ + "\t-q quit after address (no daemon)\n" \ + "\t-r 169.254.x.x request this address first\n" \ + "\t-v verbose; show version\n" + #endif /* __BB_USAGE_H__ */ --- busybox-1.00/networking/Config.in 2004-09-23 13:08:46.000000000 -0700 +++ busybox/networking/Config.in 2004-10-18 15:33:42.000000000 -0700 @@ -630,5 +630,14 @@ source networking/udhcp/Config.in +config CONFIG_ZCIP + bool "zcip" + default n + help + zcip supports ZeroConf IPv4 address selection: + it manages an address on the 169.254/16 network. + + See http://www.zeroconf.org for further details. + endmenu --- busybox-1.00/networking/Makefile.in 2004-10-08 00:45:43.000000000 -0700 +++ busybox/networking/Makefile.in 2004-10-20 01:48:43.504468768 -0700 @@ -50,6 +50,7 @@ NETWORKING-$(CONFIG_TRACEROUTE) += traceroute.o NETWORKING-$(CONFIG_VCONFIG) += vconfig.o NETWORKING-$(CONFIG_WGET) += wget.o +NETWORKING-$(CONFIG_ZCIP) += zcip.o libraries-y+=$(NETWORKING_DIR)$(NETWORKING_AR) --- busybox-1.00/networking/zcip.c 1969-12-31 16:00:00.000000000 -0800 +++ busybox/networking/zcip.c 2004-10-20 02:18:13.390405208 -0700 @@ -0,0 +1,558 @@ +/* + * ZeroConf IPv4 Link-Local addressing (see ) + * + * Copyright (C) 2003 by Arthur van Hoff (avh@strangeberry.com) + * Copyright (C) 2004 by David Brownell + * + * 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 + * (at your option) 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., 59 Temple Place, Suite 330, Boston, MA + * 02111-1307 USA + */ + +/* + * This is a user-mode implementation that's a pretty close match to + * draft-ietf-zeroconf-ipv4-linklocal-17.txt (the July 2004 version). + * It can build as part of BusyBox or by itself: + * + * cc -Os -DNO_BUSYBOX -DDEBUG -o zcip zcip.c + * + * ZCIP just manages the 169.254.*.* addresses. That network is not + * routed at the IP level, though various proxies or bridges can + * certainly be used. Its naming is built over multicast DNS. + */ + +// #define DEBUG + + // TODO: + // - more real-world usage/testing, especially daemon mode + // - kernel packet filters to reduce scheduling noise + // - avoid silent scripts failures, especially under load... + // - implement some syslog option '-s' + // - link status monitoring + +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include + +#include +#include + + +struct arp_packet { + struct ether_header hdr; + struct arphdr arp; + struct ether_addr source_addr; + struct in_addr source_ip; + struct ether_addr target_addr; + struct in_addr target_ip; + unsigned char pad[18]; +} __attribute__ ((__packed__)); + +/* 169.254.0.0 */ +static uint32_t LINKLOCAL_ADDR = 0xa9fe0000; +static uint32_t LINKLOCAL_MASK = 0xFFFF0000; + +/* protocol timeout parameters, specified in seconds */ +static const unsigned PROBE_WAIT = 1; +static const unsigned PROBE_MIN = 1; +static const unsigned PROBE_MAX = 2; +static const unsigned PROBE_NUM = 3; +static const unsigned MAX_COLLISIONS = 10; +static const unsigned RATE_LIMIT_INTERVAL = 60; +static const unsigned ANNOUNCE_WAIT = 2; +static const unsigned ANNOUNCE_NUM = 2; +static const unsigned ANNOUNCE_INTERVAL = 2; +static const unsigned DEFEND_INTERVAL = 10; + +static const unsigned char ZCIP_VERSION[] = "0.7 (20 October 2004)"; +static char *prog; + +static struct in_addr null_ip = { 0 }; +static struct ether_addr null_addr = { {0, 0, 0, 0, 0, 0} }; + +static int verbose = 0; + +#ifdef DEBUG + +#define DBG(fmt,args...) \ + fprintf(stderr, "%s: " fmt , prog , ## args) +#define VDBG(fmt,args...) do { \ + if (verbose) fprintf(stderr, "%s: " fmt , prog ,## args); \ + } while (0) + +/** + * Convert an ethernet address to a printable string. + */ +static char * +ether2str(const struct ether_addr *addr) +{ + static char str[32]; + + snprintf(str, sizeof (str), "%02x:%02x:%02x:%02x:%02x:%02x", + addr->ether_addr_octet[0], addr->ether_addr_octet[1], + addr->ether_addr_octet[2], addr->ether_addr_octet[3], + addr->ether_addr_octet[4], addr->ether_addr_octet[5]); + return str; +} + +#else + +#define DBG(fmt,args...) \ + do { } while (0) +#define VDBG DBG +#endif /* DEBUG */ + +/** + * Pick a random link local IP address on 169.254/16, except that + * the first and last 256 addresses are reserved. + */ +static void +pick(struct in_addr *ip) +{ + ip->s_addr = htonl(LINKLOCAL_ADDR | + ((abs(lrand48()) % 0xFD00) + 0x0100)); +} + +/** + * Broadcast an ARP packet. + */ +static int +arp(int fd, struct sockaddr *saddr, int op, + struct ether_addr *source_addr, struct in_addr source_ip, + struct ether_addr *target_addr, struct in_addr target_ip) +{ + struct arp_packet p; + + memset(&p, 0, sizeof (p)); + + // ether header + p.hdr.ether_type = htons(ETHERTYPE_ARP); + memcpy(p.hdr.ether_shost, source_addr, ETH_ALEN); + memset(p.hdr.ether_dhost, 0xff, ETH_ALEN); + + // arp request + p.arp.ar_hrd = htons(ARPHRD_ETHER); + p.arp.ar_pro = htons(ETHERTYPE_IP); + p.arp.ar_hln = ETH_ALEN; + p.arp.ar_pln = 4; + p.arp.ar_op = htons(op); + memcpy(&p.source_addr, source_addr, ETH_ALEN); + memcpy(&p.source_ip, &source_ip, sizeof (p.source_ip)); + memcpy(&p.target_addr, target_addr, ETH_ALEN); + memcpy(&p.target_ip, &target_ip, sizeof (p.target_ip)); + + // send it + if (sendto(fd, &p, sizeof (p), 0, saddr, sizeof (*saddr)) < 0) { + perror("sendto"); + return -errno; + } + return 0; +} + +/** + * Run a script. + */ +int +run(char *script, char *arg, char *intf, struct in_addr *ip) +{ + int pid, status; + char *why; + + if (script != NULL) { + VDBG("%s run %s %s\n", intf, script, arg); +#ifdef __uClinux__ +#error need to use vfork/execle +#else + pid = fork(); + if (pid < 0) { + why = "fork"; + goto bad; + } + if (pid == 0) { + // child process + setenv("interface", intf, 1); + if (ip != NULL) { + setenv("ip", inet_ntoa(*ip), 1); +// FIXME syslog these calls + } + execl(script, script, arg, NULL); + why = "execl"; + goto bad; + } +#endif + if (waitpid(pid, &status, 0) <= 0) { + why = "waitpid"; + goto bad; + } + if (WEXITSTATUS(status) != 0) { + fprintf(stderr, "%s: script %s failed, exit=%d\n", + prog, script, WEXITSTATUS(status)); + return -errno; + } + } + return 0; +bad: +// FIXME syslog errors here + perror(why); + return -errno; +} + +#ifndef NO_BUSYBOX +#include "busybox.h" +#define main zcip_main +#endif + +/** + * Print usage information. + */ +static void __attribute__ ((noreturn)) +usage(const char *msg) +{ + fprintf(stderr, "%s: %s\n", prog, msg); +#ifdef NO_BUSYBOX + fprintf(stderr, "Usage: %s [OPTIONS] ifname script\n" + "\t-f foreground mode\n" + "\t-q quit after address (no daemon)\n" + "\t-r 169.254.x.x request this address first\n" + "\t-v verbose; show version\n", prog); + exit(0); +#else + bb_show_usage(); +#endif +} + +/** + * Return milliseconds of random delay, up to "secs" seconds. + */ +static inline unsigned +ms_rdelay(unsigned secs) +{ + return lrand48() % (secs * 1000); +} + +/** + * main program + */ +int +main(int argc, char *argv[]) +{ + char *intf = NULL; + char *script = NULL; + char *why; + struct sockaddr saddr; + struct arp_packet p; + struct ether_addr addr; + struct in_addr ip = { 0 }; + int fd; + int quit = 0; + int ready = 0; + int foreground = 0; + suseconds_t timeout = 0; // milliseconds + time_t defend = 0; + int collisions = 0; + int nprobes = 0; + int nclaims = 0; + int t; + + // parse commandline: prog [options] ifname script + prog = argv[0]; + while ((t = getopt(argc, argv, "fqr:v")) != EOF) { + switch (t) { + case 'f': + foreground = 1; + continue; + case 'q': + quit = 1; + continue; + case 'r': + if (inet_aton(optarg, &ip) == 0 + || (ntohl(ip.s_addr) & LINKLOCAL_MASK) + != LINKLOCAL_ADDR) { + usage("invalid link address"); + } + continue; + case 'v': + if (!verbose) + printf("%s: version %s\n", prog, ZCIP_VERSION); + verbose++; + continue; + default: + usage("bad option"); + } + } + if (optind < argc - 1) { + intf = argv[optind++]; + script = argv[optind++]; + } + if (optind != argc || !intf) + usage("wrong number of arguments"); + + // initialize the interface (modprobe, ifup, etc) + if (run(script, "init", intf, NULL) < 0) + return EXIT_FAILURE; + + // initialize saddr + memset(&saddr, 0, sizeof (saddr)); + strncpy(saddr.sa_data, intf, sizeof (saddr.sa_data)); + + // open an ARP socket + if ((fd = socket(PF_PACKET, SOCK_PACKET, htons(ETH_P_ARP))) < 0) { + why = "open"; + goto bad; + } + // bind to the interface's ARP socket + if (bind(fd, &saddr, sizeof (saddr)) < 0) { + why = "bind"; + goto bad; + } else { + struct ifreq ifr; + short seed[3]; + + // get the interface's ethernet address + memset(&ifr, 0, sizeof (ifr)); + strncpy(ifr.ifr_name, intf, sizeof (ifr.ifr_name)); + if (ioctl(fd, SIOCGIFHWADDR, &ifr) < 0) { + why = "get ethernet address"; + goto bad; + } + memcpy(&addr, &ifr.ifr_hwaddr.sa_data, ETHER_ADDR_LEN); + + // start with some stable ip address, either a function of + // the hardware address or else the last address we used. + // retry timers and new addresses shouldn't be predictable. + memcpy(seed, &ifr.ifr_hwaddr.sa_data, ETHER_ADDR_LEN); + if (ip.s_addr == 0) { + seed48(seed); + pick(&ip); + } + t = time(0); + seed[0] ^= t; + seed[1] ^= t >> 16; + seed48(seed); + } + + // FIXME cases to handle: + // - zcip already running! + // - link already has local address... just defend/update + + // daemonize now; don't delay system startup + if (!foreground) { + if (daemon(0, verbose) < 0) { + why = "daemon"; + goto bad; + } + } + + // run the dynamic address negotiation protocol, + // restarting after address collisions: + // - start with some address we want to try + // - short random delay + // - arp probes to see if another host else uses it + // - arp announcements that we're claiming it + // - use it + // - defend it, within limits + while (1) { + struct pollfd fds[1]; + struct timeval tv1; + + fds[0].fd = fd; + fds[0].events = POLLIN | POLLERR; + fds[0].revents = 0; + + // poll, being ready to adjust current timeout + if (timeout > 0) { + gettimeofday(&tv1, NULL); + tv1.tv_usec += (timeout % 1000) * 1000; + if (tv1.tv_usec > 1000000) { + tv1.tv_usec -= 1000000; + tv1.tv_sec++; + } + tv1.tv_sec += timeout / 1000; + } else if (timeout == 0) { + timeout = ms_rdelay(PROBE_WAIT); + // FIXME setsockopt(fd, SO_ATTACH_FILTER, ...) to + // make the kernel filter out all packets except + // ones we'd care about. + } + VDBG("...wait %ld %s nprobes=%d, nclaims=%d\n", + timeout, intf, nprobes, nclaims); + switch (poll(fds, 1, timeout)) { + + // timeouts trigger protocol transitions + case 0: + // probes + if (nprobes < PROBE_NUM) { + nprobes++; + VDBG("probe/%d %s@%s\n", + nprobes, intf, inet_ntoa(ip)); + (void)arp(fd, &saddr, ARPOP_REQUEST, + &addr, null_ip, + &null_addr, ip); + if (nprobes < PROBE_NUM) { + timeout = PROBE_MIN * 1000; + timeout += ms_rdelay(PROBE_MAX + - PROBE_MIN); + } else + timeout = ANNOUNCE_WAIT * 1000; + } + // then announcements + else if (nclaims < ANNOUNCE_NUM) { + nclaims++; + VDBG("announce/%d %s@%s\n", + nclaims, intf, inet_ntoa(ip)); + (void)arp(fd, &saddr, ARPOP_REQUEST, + &addr, ip, &addr, ip); + if (nclaims < ANNOUNCE_NUM) { + timeout = ANNOUNCE_INTERVAL * 1000; + } else { + // link is ok to use earlier + run(script, "config", intf, &ip); + ready = 1; + collisions = 0; + timeout = -1; + + // NOTE: all other exit paths + // should deconfig ... + if (quit) + return EXIT_SUCCESS; + // FIXME update filters + } + } + break; + + // packets arriving + case 1: + // maybe adjust timeout + if (timeout > 0) { + struct timeval tv2; + + gettimeofday(&tv2, NULL); + if (timercmp(&tv1, &tv2, <)) { + timeout = -1; + } else { + timersub(&tv1, &tv2, &tv1); + timeout = 1000 * tv1.tv_sec + + tv1.tv_usec / 1000; + } + } + if ((fds[0].revents & POLLIN) == 0) { + if (fds[0].revents & POLLERR) { + // FIXME: links routinely go down; + // this shouldn't necessarily exit. + fprintf(stderr, "%s %s: poll error\n", + prog, intf); + if (ready) { + run(script, "deconfig", + intf, &ip); + } + return EXIT_FAILURE; + } + continue; + } + // read ARP packet + if (recv(fd, &p, sizeof (p), 0) < 0) { +// FIXME stderr may be bad here (syslog) + why = "recv"; + goto bad; + } + if (p.hdr.ether_type != htons(ETHERTYPE_ARP)) + continue; + + VDBG("%s recv arp type=%d, op=%d,\n", + intf, ntohs(p.hdr.ether_type), + ntohs(p.arp.ar_op)); + VDBG("\tsource=%s %s\n", + ether2str(&p.source_addr), + inet_ntoa(p.source_ip)); + VDBG("\ttarget=%s %s\n", + ether2str(&p.target_addr), + inet_ntoa(p.target_ip)); + if (p.arp.ar_op != htons(ARPOP_REQUEST) + && p.arp.ar_op != htons(ARPOP_REPLY)) + continue; + + // some cases are always collisions + if ((p.source_ip.s_addr == ip.s_addr) + && (memcmp(&addr, &p.source_addr, + ETH_ALEN) != 0)) { +collision: + VDBG("%s ARP conflict from %s\n", intf, + ether2str(&p.source_addr)); + if (ready) { + time_t now = time(0); + + if ((defend + DEFEND_INTERVAL) + < now) { + defend = now; + (void)arp(fd, &saddr, + ARPOP_REQUEST, + &addr, ip, + &addr, ip); + VDBG("%s defend\n", intf); + timeout = -1; + continue; + } + defend = now; + ready = 0; + run(script, "deconfig", intf, &ip); + // FIXME rm filters: setsockopt(fd, + // SO_DETACH_FILTER, ...) + } + collisions++; + if (collisions >= MAX_COLLISIONS) { + VDBG("%s ratelimit\n", intf); + sleep(RATE_LIMIT_INTERVAL); + } + // restart the whole protocol + pick(&ip); + timeout = 0; + nprobes = 0; + nclaims = 0; + } + // two hosts probing one address is a collision too + else if (p.target_ip.s_addr == ip.s_addr + && nclaims == 0 + && p.arp.ar_op == htons(ARPOP_REQUEST) + && memcmp(&addr, &p.target_addr, + ETH_ALEN) != 0) { + goto collision; + } + break; + + default: +// FIXME stderr may be bad here (syslog) + why = "poll"; + goto bad; + } + } +bad: + perror(why); + return EXIT_FAILURE; +}