| /* vi: set sw=4 ts=4: */ |
| /* |
| * Mini ping implementation for busybox |
| * |
| * Copyright (C) 1999 by Randolph Chung <tausq@debian.org> |
| * |
| * Adapted from the ping in netkit-base 0.10: |
| * Copyright (c) 1989 The Regents of the University of California. |
| * All rights reserved. |
| * |
| * This code is derived from software contributed to Berkeley by |
| * Mike Muuss. |
| * |
| * Licensed under GPLv2 or later, see file LICENSE in this source tree. |
| */ |
| /* from ping6.c: |
| * Copyright (C) 1999 by Randolph Chung <tausq@debian.org> |
| * |
| * This version of ping is adapted from the ping in netkit-base 0.10, |
| * which is: |
| * |
| * Original copyright notice is retained at the end of this file. |
| * |
| * This version is an adaptation of ping.c from busybox. |
| * The code was modified by Bart Visscher <magick@linux-fan.com> |
| */ |
| //config:config PING |
| //config: bool "ping (10 kb)" |
| //config: default y |
| //config: help |
| //config: ping uses the ICMP protocol's mandatory ECHO_REQUEST datagram to |
| //config: elicit an ICMP ECHO_RESPONSE from a host or gateway. |
| //config: |
| //config:config PING6 |
| //config: bool "ping6 (11 kb)" |
| //config: default y |
| //config: depends on FEATURE_IPV6 |
| //config: help |
| //config: Alias to "ping -6". |
| //config: |
| //config:config FEATURE_FANCY_PING |
| //config: bool "Enable fancy ping output" |
| //config: default y |
| //config: depends on PING || PING6 |
| //config: help |
| //config: With this option off, ping will say "HOST is alive!" |
| //config: or terminate with SIGALRM in 5 seconds otherwise. |
| //config: No command-line options will be recognized. |
| |
| /* Needs socket(AF_INET, SOCK_RAW, IPPROTO_ICMP), therefore BB_SUID_MAYBE: */ |
| //applet:IF_PING(APPLET(ping, BB_DIR_BIN, BB_SUID_MAYBE)) |
| //applet:IF_PING6(APPLET(ping6, BB_DIR_BIN, BB_SUID_MAYBE)) |
| |
| //kbuild:lib-$(CONFIG_PING) += ping.o |
| //kbuild:lib-$(CONFIG_PING6) += ping.o |
| |
| //usage:#if !ENABLE_FEATURE_FANCY_PING |
| //usage:# define ping_trivial_usage |
| //usage: "HOST" |
| //usage:# define ping_full_usage "\n\n" |
| //usage: "Send ICMP ECHO_REQUESTs to HOST" |
| //usage:# define ping6_trivial_usage |
| //usage: "HOST" |
| //usage:# define ping6_full_usage "\n\n" |
| //usage: "Send ICMP ECHO_REQUESTs to HOST" |
| //usage:#else |
| //usage:# define ping_trivial_usage |
| //usage: "[OPTIONS] HOST" |
| //usage:# define ping_full_usage "\n\n" |
| //usage: "Send ICMP ECHO_REQUESTs to HOST\n" |
| //usage: IF_PING6( |
| //usage: "\n -4,-6 Force IP or IPv6 name resolution" |
| //usage: ) |
| //usage: "\n -c CNT Send only CNT pings" |
| //usage: "\n -s SIZE Send SIZE data bytes in packets (default 56)" |
| //usage: "\n -i SECS Interval" |
| //usage: "\n -A Ping as soon as reply is recevied" |
| //usage: "\n -t TTL Set TTL" |
| //usage: "\n -I IFACE/IP Source interface or IP address" |
| //usage: "\n -W SEC Seconds to wait for the first response (default 10)" |
| //usage: "\n (after all -c CNT packets are sent)" |
| //usage: "\n -w SEC Seconds until ping exits (default:infinite)" |
| //usage: "\n (can exit earlier with -c CNT)" |
| //usage: "\n -q Quiet, only display output at start/finish" |
| //usage: "\n -p HEXBYTE Payload pattern" |
| //usage: |
| //usage:# define ping6_trivial_usage |
| //usage: "[OPTIONS] HOST" |
| //usage:# define ping6_full_usage "\n\n" |
| //usage: "Send ICMP ECHO_REQUESTs to HOST\n" |
| //usage: "\n -c CNT Send only CNT pings" |
| //usage: "\n -s SIZE Send SIZE data bytes in packets (default 56)" |
| //usage: "\n -i SECS Interval" |
| //usage: "\n -A Ping as soon as reply is recevied" |
| ///////: "\n -t TTL Set TTL" |
| ///////^^^^^ -t not tested for IPv6, might be not working |
| //usage: "\n -I IFACE/IP Source interface or IP address" |
| //usage: "\n -W SEC Seconds to wait for the first response (default 10)" |
| //usage: "\n (after all -c CNT packets are sent)" |
| //usage: "\n -w SEC Seconds until ping exits (default:infinite)" |
| //usage: "\n (can exit earlier with -c CNT)" |
| //usage: "\n -q Quiet, only display output at start/finish" |
| //usage: "\n -p HEXBYTE Payload pattern" |
| //usage: |
| //usage:#endif |
| //usage: |
| //usage:#define ping_example_usage |
| //usage: "$ ping localhost\n" |
| //usage: "PING slag (127.0.0.1): 56 data bytes\n" |
| //usage: "64 bytes from 127.0.0.1: icmp_seq=0 ttl=255 time=20.1 ms\n" |
| //usage: "\n" |
| //usage: "--- debian ping statistics ---\n" |
| //usage: "1 packets transmitted, 1 packets received, 0% packet loss\n" |
| //usage: "round-trip min/avg/max = 20.1/20.1/20.1 ms\n" |
| //usage:#define ping6_example_usage |
| //usage: "$ ping6 ip6-localhost\n" |
| //usage: "PING ip6-localhost (::1): 56 data bytes\n" |
| //usage: "64 bytes from ::1: icmp6_seq=0 ttl=64 time=20.1 ms\n" |
| //usage: "\n" |
| //usage: "--- ip6-localhost ping statistics ---\n" |
| //usage: "1 packets transmitted, 1 packets received, 0% packet loss\n" |
| //usage: "round-trip min/avg/max = 20.1/20.1/20.1 ms\n" |
| |
| #include <net/if.h> |
| #if defined(__FreeBSD__) |
| # include <netinet/in.h> /* struct ip and friends */ |
| # include <netinet/ip.h> |
| #endif |
| #include <netinet/ip_icmp.h> |
| #include "libbb.h" |
| #include "common_bufsiz.h" |
| |
| #ifdef __BIONIC__ |
| /* should be in netinet/ip_icmp.h */ |
| # define ICMP_DEST_UNREACH 3 /* Destination Unreachable */ |
| # define ICMP_SOURCE_QUENCH 4 /* Source Quench */ |
| # define ICMP_REDIRECT 5 /* Redirect (change route) */ |
| # define ICMP_ECHO 8 /* Echo Request */ |
| # define ICMP_TIME_EXCEEDED 11 /* Time Exceeded */ |
| # define ICMP_PARAMETERPROB 12 /* Parameter Problem */ |
| # define ICMP_TIMESTAMP 13 /* Timestamp Request */ |
| # define ICMP_TIMESTAMPREPLY 14 /* Timestamp Reply */ |
| # define ICMP_INFO_REQUEST 15 /* Information Request */ |
| # define ICMP_INFO_REPLY 16 /* Information Reply */ |
| # define ICMP_ADDRESS 17 /* Address Mask Request */ |
| # define ICMP_ADDRESSREPLY 18 /* Address Mask Reply */ |
| #endif |
| |
| /* Some operating systems, like GNU/Hurd, don't define SOL_RAW, but do have |
| * IPPROTO_RAW. Since the IPPROTO definitions are also valid to use for |
| * setsockopt (and take the same value as their corresponding SOL definitions, |
| * if they exist), we can just fall back on IPPROTO_RAW. */ |
| #ifndef SOL_RAW |
| # define SOL_RAW IPPROTO_RAW |
| #endif |
| |
| #if ENABLE_PING6 |
| # include <netinet/icmp6.h> |
| /* I see RENUMBERED constants in bits/in.h - !!? |
| * What a fuck is going on with libc? Is it a glibc joke? */ |
| # ifdef IPV6_2292HOPLIMIT |
| # undef IPV6_HOPLIMIT |
| # define IPV6_HOPLIMIT IPV6_2292HOPLIMIT |
| # endif |
| #endif |
| |
| #if defined(__FreeBSD__) |
| /** |
| * On BSD the IPv4 struct is called struct ip and instead of iXX |
| * the members are called ip_XX. One could change this code to use |
| * struct ip but that would require to define _BSD_SOURCE and that |
| * might have other complications. Instead make sure struct iphdr |
| * is present on FreeBSD. The below is taken from GLIBC. |
| * |
| * The GNU C Library is free software; you can redistribute it and/or |
| * modify it under the terms of the GNU Lesser General Public |
| * License as published by the Free Software Foundation; either |
| * version 2.1 of the License, or (at your option) any later version. |
| */ |
| struct iphdr { |
| # if BYTE_ORDER == LITTLE_ENDIAN |
| unsigned int ihl:4; |
| unsigned int version:4; |
| # elif BYTE_ORDER == BIG_ENDIAN |
| unsigned int version:4; |
| unsigned int ihl:4; |
| # endif |
| uint8_t tos; |
| uint16_t tot_len; |
| uint16_t id; |
| uint16_t frag_off; |
| uint8_t ttl; |
| uint8_t protocol; |
| uint16_t check; |
| uint32_t saddr; |
| uint32_t daddr; |
| /*The options start here. */ |
| }; |
| #endif |
| |
| enum { |
| DEFDATALEN = 56, |
| MAXIPLEN = 60, |
| MAXICMPLEN = 76, |
| MAX_DUP_CHK = (8 * 128), |
| MAXWAIT = 10, |
| PINGINTERVAL = 1, /* 1 second */ |
| pingsock = 0, |
| }; |
| |
| static void |
| #if ENABLE_PING6 |
| create_icmp_socket(len_and_sockaddr *lsa) |
| #else |
| create_icmp_socket(void) |
| #define create_icmp_socket(lsa) create_icmp_socket() |
| #endif |
| { |
| int sock; |
| #if ENABLE_PING6 |
| if (lsa->u.sa.sa_family == AF_INET6) |
| sock = socket(AF_INET6, SOCK_RAW, IPPROTO_ICMPV6); |
| else |
| #endif |
| sock = socket(AF_INET, SOCK_RAW, 1); /* 1 == ICMP */ |
| if (sock < 0) { |
| if (errno == EPERM) |
| bb_simple_error_msg_and_die(bb_msg_perm_denied_are_you_root); |
| bb_simple_perror_msg_and_die(bb_msg_can_not_create_raw_socket); |
| } |
| |
| xmove_fd(sock, pingsock); |
| } |
| |
| #if !ENABLE_FEATURE_FANCY_PING |
| |
| /* Simple version */ |
| |
| struct globals { |
| char *hostname; |
| char packet[DEFDATALEN + MAXIPLEN + MAXICMPLEN]; |
| uint16_t myid; |
| } FIX_ALIASING; |
| #define G (*(struct globals*)bb_common_bufsiz1) |
| #define INIT_G() do { setup_common_bufsiz(); } while (0) |
| |
| static void noresp(int ign UNUSED_PARAM) |
| { |
| printf("No response from %s\n", G.hostname); |
| exit(EXIT_FAILURE); |
| } |
| |
| static void ping4(len_and_sockaddr *lsa) |
| { |
| struct icmp *pkt; |
| int c; |
| |
| pkt = (struct icmp *) G.packet; |
| /*memset(pkt, 0, sizeof(G.packet)); already is */ |
| pkt->icmp_type = ICMP_ECHO; |
| pkt->icmp_id = G.myid; |
| pkt->icmp_cksum = inet_cksum(pkt, sizeof(G.packet)); |
| |
| xsendto(pingsock, G.packet, DEFDATALEN + ICMP_MINLEN, &lsa->u.sa, lsa->len); |
| |
| /* listen for replies */ |
| while (1) { |
| #if 0 |
| struct sockaddr_in from; |
| socklen_t fromlen = sizeof(from); |
| |
| c = recvfrom(pingsock, G.packet, sizeof(G.packet), 0, |
| (struct sockaddr *) &from, &fromlen); |
| #else |
| c = recv(pingsock, G.packet, sizeof(G.packet), 0); |
| #endif |
| if (c < 0) { |
| if (errno != EINTR) |
| bb_simple_perror_msg("recvfrom"); |
| continue; |
| } |
| if (c >= 76) { /* ip + icmp */ |
| struct iphdr *iphdr = (struct iphdr *) G.packet; |
| |
| pkt = (struct icmp *) (G.packet + (iphdr->ihl << 2)); /* skip ip hdr */ |
| if (pkt->icmp_id != G.myid) |
| continue; /* not our ping */ |
| if (pkt->icmp_type == ICMP_ECHOREPLY) |
| break; |
| } |
| } |
| } |
| |
| #if ENABLE_PING6 |
| static void ping6(len_and_sockaddr *lsa) |
| { |
| struct icmp6_hdr *pkt; |
| int c; |
| int sockopt; |
| |
| pkt = (struct icmp6_hdr *) G.packet; |
| /*memset(pkt, 0, sizeof(G.packet)); already is */ |
| pkt->icmp6_type = ICMP6_ECHO_REQUEST; |
| pkt->icmp6_id = G.myid; |
| |
| sockopt = offsetof(struct icmp6_hdr, icmp6_cksum); |
| setsockopt_int(pingsock, SOL_RAW, IPV6_CHECKSUM, sockopt); |
| |
| xsendto(pingsock, G.packet, DEFDATALEN + sizeof(struct icmp6_hdr), &lsa->u.sa, lsa->len); |
| |
| /* listen for replies */ |
| while (1) { |
| #if 0 |
| struct sockaddr_in6 from; |
| socklen_t fromlen = sizeof(from); |
| |
| c = recvfrom(pingsock, G.packet, sizeof(G.packet), 0, |
| (struct sockaddr *) &from, &fromlen); |
| #else |
| c = recv(pingsock, G.packet, sizeof(G.packet), 0); |
| #endif |
| if (c < 0) { |
| if (errno != EINTR) |
| bb_simple_perror_msg("recvfrom"); |
| continue; |
| } |
| if (c >= ICMP_MINLEN) { /* icmp6_hdr */ |
| if (pkt->icmp6_id != G.myid) |
| continue; /* not our ping */ |
| if (pkt->icmp6_type == ICMP6_ECHO_REPLY) |
| break; |
| } |
| } |
| } |
| #endif |
| |
| #if !ENABLE_PING6 |
| # define common_ping_main(af, argv) common_ping_main(argv) |
| #endif |
| static int common_ping_main(sa_family_t af, char **argv) |
| { |
| len_and_sockaddr *lsa; |
| |
| INIT_G(); |
| |
| #if ENABLE_PING6 |
| while ((++argv)[0] && argv[0][0] == '-') { |
| if (argv[0][1] == '4') { |
| af = AF_INET; |
| continue; |
| } |
| if (argv[0][1] == '6') { |
| af = AF_INET6; |
| continue; |
| } |
| bb_show_usage(); |
| } |
| #else |
| argv++; |
| #endif |
| |
| G.hostname = *argv; |
| if (!G.hostname) |
| bb_show_usage(); |
| |
| #if ENABLE_PING6 |
| lsa = xhost_and_af2sockaddr(G.hostname, 0, af); |
| #else |
| lsa = xhost_and_af2sockaddr(G.hostname, 0, AF_INET); |
| #endif |
| /* Set timer _after_ DNS resolution */ |
| signal(SIGALRM, noresp); |
| alarm(5); /* give the host 5000ms to respond */ |
| |
| create_icmp_socket(lsa); |
| G.myid = (uint16_t) getpid(); |
| /* we can use native-endian ident, but other Unix ping/traceroute |
| * utils use *big-endian pid*, and e.g. traceroute on our machine may be |
| * *not* from busybox, idents may collide. Follow the convention: |
| */ |
| G.myid = htons(G.myid); |
| #if ENABLE_PING6 |
| if (lsa->u.sa.sa_family == AF_INET6) |
| ping6(lsa); |
| else |
| #endif |
| ping4(lsa); |
| if (ENABLE_FEATURE_CLEAN_UP) |
| close(pingsock); |
| printf("%s is alive!\n", G.hostname); |
| return EXIT_SUCCESS; |
| } |
| |
| |
| #else /* FEATURE_FANCY_PING */ |
| |
| |
| /* Full(er) version */ |
| |
| /* -c NUM, -t NUM, -w NUM, -W NUM */ |
| #define OPT_STRING "qvAc:+s:t:+w:+W:+I:np:i:4"IF_PING6("6") |
| enum { |
| OPT_QUIET = 1 << 0, |
| OPT_VERBOSE = 1 << 1, |
| OPT_A = 1 << 2, |
| OPT_c = 1 << 3, |
| OPT_s = 1 << 4, |
| OPT_t = 1 << 5, |
| OPT_w = 1 << 6, |
| OPT_W = 1 << 7, |
| OPT_I = 1 << 8, |
| /*OPT_n = 1 << 9, - ignored */ |
| OPT_p = 1 << 10, |
| OPT_i = 1 << 11, |
| OPT_IPV4 = 1 << 12, |
| OPT_IPV6 = (1 << 13) * ENABLE_PING6, |
| }; |
| |
| |
| struct globals { |
| int if_index; |
| char *str_I; |
| len_and_sockaddr *source_lsa; |
| unsigned datalen; |
| unsigned pingcount; /* must be int-sized */ |
| unsigned opt_ttl; |
| unsigned long ntransmitted, nreceived, nrepeats; |
| uint16_t myid; |
| uint8_t pattern; |
| unsigned tmin, tmax; /* in us */ |
| unsigned long long tsum; /* in us, sum of all times */ |
| unsigned cur_us; /* low word only, we don't need more */ |
| unsigned deadline_us; |
| unsigned interval_us; |
| unsigned timeout; |
| unsigned sizeof_rcv_packet; |
| char *rcv_packet; /* [datalen + MAXIPLEN + MAXICMPLEN] */ |
| void *snd_packet; /* [datalen + ipv4/ipv6_const] */ |
| const char *hostname; |
| const char *dotted; |
| union { |
| struct sockaddr sa; |
| struct sockaddr_in sin; |
| #if ENABLE_PING6 |
| struct sockaddr_in6 sin6; |
| #endif |
| } pingaddr; |
| unsigned char rcvd_tbl[MAX_DUP_CHK / 8]; |
| } FIX_ALIASING; |
| #define G (*(struct globals*)bb_common_bufsiz1) |
| #define if_index (G.if_index ) |
| #define source_lsa (G.source_lsa ) |
| #define str_I (G.str_I ) |
| #define datalen (G.datalen ) |
| #define pingcount (G.pingcount ) |
| #define opt_ttl (G.opt_ttl ) |
| #define myid (G.myid ) |
| #define tmin (G.tmin ) |
| #define tmax (G.tmax ) |
| #define tsum (G.tsum ) |
| #define timeout (G.timeout ) |
| #define hostname (G.hostname ) |
| #define dotted (G.dotted ) |
| #define pingaddr (G.pingaddr ) |
| #define rcvd_tbl (G.rcvd_tbl ) |
| #define INIT_G() do { \ |
| setup_common_bufsiz(); \ |
| BUILD_BUG_ON(sizeof(G) > COMMON_BUFSIZE); \ |
| datalen = DEFDATALEN; \ |
| timeout = MAXWAIT; \ |
| tmin = UINT_MAX; \ |
| } while (0) |
| |
| |
| #define BYTE(bit) rcvd_tbl[(bit)>>3] |
| #define MASK(bit) (1 << ((bit) & 7)) |
| #define SET(bit) (BYTE(bit) |= MASK(bit)) |
| #define CLR(bit) (BYTE(bit) &= (~MASK(bit))) |
| #define TST(bit) (BYTE(bit) & MASK(bit)) |
| |
| static void print_stats_and_exit(int junk) NORETURN; |
| static void print_stats_and_exit(int junk UNUSED_PARAM) |
| { |
| unsigned long ul; |
| unsigned long nrecv; |
| |
| signal(SIGINT, SIG_IGN); |
| |
| nrecv = G.nreceived; |
| printf("\n--- %s ping statistics ---\n" |
| "%lu packets transmitted, " |
| "%lu packets received, ", |
| hostname, G.ntransmitted, nrecv |
| ); |
| if (G.nrepeats) |
| printf("%lu duplicates, ", G.nrepeats); |
| ul = G.ntransmitted; |
| if (ul != 0) |
| ul = (ul - nrecv) * 100 / ul; |
| printf("%lu%% packet loss\n", ul); |
| if (tmin != UINT_MAX) { |
| unsigned tavg = tsum / (nrecv + G.nrepeats); |
| printf("round-trip min/avg/max = %u.%03u/%u.%03u/%u.%03u ms\n", |
| tmin / 1000, tmin % 1000, |
| tavg / 1000, tavg % 1000, |
| tmax / 1000, tmax % 1000); |
| } |
| /* if condition is true, exit with 1 -- 'failure' */ |
| exit(nrecv == 0 || (G.deadline_us && nrecv < pingcount)); |
| } |
| |
| static void sendping_tail(void (*sp)(int), int size_pkt) |
| { |
| int sz; |
| |
| if (G.deadline_us) { |
| unsigned n = G.cur_us - G.deadline_us; |
| if ((int)n >= 0) |
| print_stats_and_exit(0); |
| } |
| |
| CLR((uint16_t)G.ntransmitted % MAX_DUP_CHK); |
| G.ntransmitted++; |
| size_pkt += datalen; |
| |
| /* sizeof(pingaddr) can be larger than real sa size, but I think |
| * it doesn't matter */ |
| sz = xsendto(pingsock, G.snd_packet, size_pkt, &pingaddr.sa, sizeof(pingaddr)); |
| if (sz != size_pkt) |
| bb_simple_error_msg_and_die(bb_msg_write_error); |
| |
| if (pingcount == 0 || G.ntransmitted < pingcount) { |
| /* Didn't send all pings yet - schedule next in -i SEC interval */ |
| struct itimerval i; |
| signal(SIGALRM, sp); |
| /*ualarm(G.interval_us, 0); - does not work for >=1sec on some libc */ |
| i.it_interval.tv_sec = 0; |
| i.it_interval.tv_usec = 0; |
| i.it_value.tv_sec = G.interval_us / 1000000; |
| i.it_value.tv_usec = G.interval_us % 1000000; |
| setitimer(ITIMER_REAL, &i, NULL); |
| } else { /* -c NN, and all NN are sent */ |
| /* Wait for the last ping to come back. |
| * -W timeout: wait for a response in seconds. |
| * Affects only timeout in absence of any responses, |
| * otherwise ping waits for two RTTs. */ |
| unsigned expire = timeout; |
| |
| if (G.nreceived) { |
| /* approx. 2*tmax, in seconds (2 RTT) */ |
| expire = tmax / (512*1024); |
| if (expire == 0) |
| expire = 1; |
| } |
| signal(SIGALRM, print_stats_and_exit); |
| alarm(expire); |
| } |
| } |
| |
| static void sendping4(int junk UNUSED_PARAM) |
| { |
| struct icmp *pkt = G.snd_packet; |
| |
| memset(pkt, G.pattern, datalen + ICMP_MINLEN + 4); |
| pkt->icmp_type = ICMP_ECHO; |
| /*pkt->icmp_code = 0;*/ |
| pkt->icmp_cksum = 0; /* cksum is calculated with this field set to 0 */ |
| pkt->icmp_seq = htons(G.ntransmitted); /* don't ++ here, it can be a macro */ |
| pkt->icmp_id = myid; |
| |
| /* If datalen < 4, we store timestamp _past_ the packet, |
| * but it's ok - we allocated 4 extra bytes in xzalloc() just in case. |
| */ |
| /*if (datalen >= 4)*/ |
| /* No hton: we'll read it back on the same machine */ |
| *(uint32_t*)&pkt->icmp_dun = G.cur_us = monotonic_us(); |
| |
| pkt->icmp_cksum = inet_cksum(pkt, datalen + ICMP_MINLEN); |
| |
| sendping_tail(sendping4, ICMP_MINLEN); |
| } |
| #if ENABLE_PING6 |
| static void sendping6(int junk UNUSED_PARAM) |
| { |
| struct icmp6_hdr *pkt = G.snd_packet; |
| |
| memset(pkt, G.pattern, datalen + sizeof(struct icmp6_hdr) + 4); |
| pkt->icmp6_type = ICMP6_ECHO_REQUEST; |
| /*pkt->icmp6_code = 0;*/ |
| /*pkt->icmp6_cksum = 0;*/ |
| pkt->icmp6_seq = htons(G.ntransmitted); /* don't ++ here, it can be a macro */ |
| pkt->icmp6_id = myid; |
| |
| /*if (datalen >= 4)*/ |
| *(bb__aliased_uint32_t*)(&pkt->icmp6_data8[4]) = G.cur_us = monotonic_us(); |
| |
| //TODO? pkt->icmp_cksum = inet_cksum(...); |
| |
| sendping_tail(sendping6, sizeof(struct icmp6_hdr)); |
| } |
| #endif |
| |
| static const char *icmp_type_name(int id) |
| { |
| switch (id) { |
| case ICMP_ECHOREPLY: return "Echo Reply"; |
| case ICMP_DEST_UNREACH: return "Destination Unreachable"; |
| case ICMP_SOURCE_QUENCH: return "Source Quench"; |
| case ICMP_REDIRECT: return "Redirect (change route)"; |
| case ICMP_ECHO: return "Echo Request"; |
| case ICMP_TIME_EXCEEDED: return "Time Exceeded"; |
| case ICMP_PARAMETERPROB: return "Parameter Problem"; |
| case ICMP_TIMESTAMP: return "Timestamp Request"; |
| case ICMP_TIMESTAMPREPLY: return "Timestamp Reply"; |
| case ICMP_INFO_REQUEST: return "Information Request"; |
| case ICMP_INFO_REPLY: return "Information Reply"; |
| case ICMP_ADDRESS: return "Address Mask Request"; |
| case ICMP_ADDRESSREPLY: return "Address Mask Reply"; |
| default: return "unknown ICMP type"; |
| } |
| } |
| #if ENABLE_PING6 |
| /* RFC3542 changed some definitions from RFC2292 for no good reason, whee! |
| * the newer 3542 uses a MLD_ prefix where as 2292 uses ICMP6_ prefix */ |
| #ifndef MLD_LISTENER_QUERY |
| # define MLD_LISTENER_QUERY ICMP6_MEMBERSHIP_QUERY |
| #endif |
| #ifndef MLD_LISTENER_REPORT |
| # define MLD_LISTENER_REPORT ICMP6_MEMBERSHIP_REPORT |
| #endif |
| #ifndef MLD_LISTENER_REDUCTION |
| # define MLD_LISTENER_REDUCTION ICMP6_MEMBERSHIP_REDUCTION |
| #endif |
| static const char *icmp6_type_name(int id) |
| { |
| switch (id) { |
| case ICMP6_DST_UNREACH: return "Destination Unreachable"; |
| case ICMP6_PACKET_TOO_BIG: return "Packet too big"; |
| case ICMP6_TIME_EXCEEDED: return "Time Exceeded"; |
| case ICMP6_PARAM_PROB: return "Parameter Problem"; |
| case ICMP6_ECHO_REPLY: return "Echo Reply"; |
| case ICMP6_ECHO_REQUEST: return "Echo Request"; |
| case MLD_LISTENER_QUERY: return "Listener Query"; |
| case MLD_LISTENER_REPORT: return "Listener Report"; |
| case MLD_LISTENER_REDUCTION: return "Listener Reduction"; |
| default: return "unknown ICMP type"; |
| } |
| } |
| #endif |
| |
| static void unpack_tail(int sz, uint32_t *tp, |
| const char *from_str, |
| uint16_t recv_seq, int ttl) |
| { |
| unsigned char *b, m; |
| const char *dupmsg = " (DUP!)"; |
| unsigned triptime = triptime; /* for gcc */ |
| |
| if (tp) { |
| /* (int32_t) cast is for hypothetical 64-bit unsigned */ |
| /* (doesn't hurt 32-bit real-world anyway) */ |
| triptime = (int32_t) ((uint32_t)monotonic_us() - *tp); |
| tsum += triptime; |
| if (triptime < tmin) |
| tmin = triptime; |
| if (triptime > tmax) |
| tmax = triptime; |
| } |
| |
| b = &BYTE(recv_seq % MAX_DUP_CHK); |
| m = MASK(recv_seq % MAX_DUP_CHK); |
| /*if TST(recv_seq % MAX_DUP_CHK):*/ |
| if (*b & m) { |
| ++G.nrepeats; |
| } else { |
| /*SET(recv_seq % MAX_DUP_CHK):*/ |
| *b |= m; |
| ++G.nreceived; |
| dupmsg += 7; |
| } |
| |
| if (option_mask32 & OPT_QUIET) |
| return; |
| |
| printf("%d bytes from %s: seq=%u ttl=%d", sz, |
| from_str, recv_seq, ttl); |
| if (tp) |
| printf(" time=%u.%03u ms", triptime / 1000, triptime % 1000); |
| puts(dupmsg); |
| fflush_all(); |
| } |
| static int unpack4(char *buf, int sz, struct sockaddr_in *from) |
| { |
| struct icmp *icmppkt; |
| struct iphdr *iphdr; |
| int hlen; |
| |
| /* discard if too short */ |
| if (sz < (datalen + ICMP_MINLEN)) |
| return 0; |
| |
| /* check IP header */ |
| iphdr = (struct iphdr *) buf; |
| hlen = iphdr->ihl << 2; |
| sz -= hlen; |
| icmppkt = (struct icmp *) (buf + hlen); |
| if (icmppkt->icmp_id != myid) |
| return 0; /* not our ping */ |
| |
| if (icmppkt->icmp_type == ICMP_ECHOREPLY) { |
| uint16_t recv_seq = ntohs(icmppkt->icmp_seq); |
| uint32_t *tp = NULL; |
| |
| if (sz >= ICMP_MINLEN + sizeof(uint32_t)) |
| tp = (uint32_t *) icmppkt->icmp_data; |
| unpack_tail(sz, tp, |
| inet_ntoa(*(struct in_addr *) &from->sin_addr.s_addr), |
| recv_seq, iphdr->ttl); |
| return 1; |
| } |
| if (icmppkt->icmp_type != ICMP_ECHO) { |
| bb_error_msg("warning: got ICMP %d (%s)", |
| icmppkt->icmp_type, |
| icmp_type_name(icmppkt->icmp_type)); |
| } |
| return 0; |
| } |
| #if ENABLE_PING6 |
| static int unpack6(char *packet, int sz, struct sockaddr_in6 *from, int hoplimit) |
| { |
| struct icmp6_hdr *icmppkt; |
| char buf[INET6_ADDRSTRLEN]; |
| |
| /* discard if too short */ |
| if (sz < (datalen + sizeof(struct icmp6_hdr))) |
| return 0; |
| |
| icmppkt = (struct icmp6_hdr *) packet; |
| if (icmppkt->icmp6_id != myid) |
| return 0; /* not our ping */ |
| |
| if (icmppkt->icmp6_type == ICMP6_ECHO_REPLY) { |
| uint16_t recv_seq = ntohs(icmppkt->icmp6_seq); |
| uint32_t *tp = NULL; |
| |
| if (sz >= sizeof(struct icmp6_hdr) + sizeof(uint32_t)) |
| tp = (uint32_t *) &icmppkt->icmp6_data8[4]; |
| unpack_tail(sz, tp, |
| inet_ntop(AF_INET6, &from->sin6_addr, |
| buf, sizeof(buf)), |
| recv_seq, hoplimit); |
| return 1; |
| } |
| if (icmppkt->icmp6_type != ICMP6_ECHO_REQUEST) { |
| bb_error_msg("warning: got ICMP %d (%s)", |
| icmppkt->icmp6_type, |
| icmp6_type_name(icmppkt->icmp6_type)); |
| } |
| return 0; |
| } |
| #endif |
| |
| static void ping4(len_and_sockaddr *lsa) |
| { |
| int sockopt; |
| |
| pingaddr.sin = lsa->u.sin; |
| if (source_lsa) { |
| if (setsockopt(pingsock, IPPROTO_IP, IP_MULTICAST_IF, |
| &source_lsa->u.sa, source_lsa->len)) |
| bb_simple_error_msg_and_die("can't set multicast source interface"); |
| xbind(pingsock, &source_lsa->u.sa, source_lsa->len); |
| } |
| |
| /* enable broadcast pings */ |
| setsockopt_broadcast(pingsock); |
| |
| /* set recv buf (needed if we can get lots of responses: flood ping, |
| * broadcast ping etc) */ |
| sockopt = (datalen * 2) + 7 * 1024; /* giving it a bit of extra room */ |
| setsockopt_SOL_SOCKET_int(pingsock, SO_RCVBUF, sockopt); |
| |
| if (opt_ttl != 0) { |
| setsockopt_int(pingsock, IPPROTO_IP, IP_TTL, opt_ttl); |
| /* above doesn't affect packets sent to bcast IP, so... */ |
| setsockopt_int(pingsock, IPPROTO_IP, IP_MULTICAST_TTL, opt_ttl); |
| } |
| |
| signal(SIGINT, print_stats_and_exit); |
| |
| /* start the ping's going ... */ |
| send_ping: |
| sendping4(0); |
| |
| /* listen for replies */ |
| while (1) { |
| struct sockaddr_in from; |
| socklen_t fromlen = (socklen_t) sizeof(from); |
| int c; |
| |
| c = recvfrom(pingsock, G.rcv_packet, G.sizeof_rcv_packet, 0, |
| (struct sockaddr *) &from, &fromlen); |
| if (c < 0) { |
| if (errno != EINTR) |
| bb_simple_perror_msg("recvfrom"); |
| continue; |
| } |
| c = unpack4(G.rcv_packet, c, &from); |
| if (pingcount && G.nreceived >= pingcount) |
| break; |
| if (c && (option_mask32 & OPT_A)) { |
| goto send_ping; |
| } |
| } |
| } |
| #if ENABLE_PING6 |
| static void ping6(len_and_sockaddr *lsa) |
| { |
| int sockopt; |
| struct msghdr msg; |
| struct sockaddr_in6 from; |
| struct iovec iov; |
| char control_buf[CMSG_SPACE(36)]; |
| |
| pingaddr.sin6 = lsa->u.sin6; |
| if (source_lsa) |
| xbind(pingsock, &source_lsa->u.sa, source_lsa->len); |
| |
| #ifdef ICMP6_FILTER |
| { |
| struct icmp6_filter filt; |
| if (!(option_mask32 & OPT_VERBOSE)) { |
| ICMP6_FILTER_SETBLOCKALL(&filt); |
| ICMP6_FILTER_SETPASS(ICMP6_ECHO_REPLY, &filt); |
| } else { |
| ICMP6_FILTER_SETPASSALL(&filt); |
| } |
| if (setsockopt(pingsock, IPPROTO_ICMPV6, ICMP6_FILTER, &filt, |
| sizeof(filt)) < 0) |
| bb_error_msg_and_die("setsockopt(%s)", "ICMP6_FILTER"); |
| } |
| #endif /*ICMP6_FILTER*/ |
| |
| /* enable broadcast pings */ |
| setsockopt_broadcast(pingsock); |
| |
| /* set recv buf (needed if we can get lots of responses: flood ping, |
| * broadcast ping etc) */ |
| sockopt = (datalen * 2) + 7 * 1024; /* giving it a bit of extra room */ |
| setsockopt_SOL_SOCKET_int(pingsock, SO_RCVBUF, sockopt); |
| |
| sockopt = offsetof(struct icmp6_hdr, icmp6_cksum); |
| BUILD_BUG_ON(offsetof(struct icmp6_hdr, icmp6_cksum) != 2); |
| setsockopt_int(pingsock, SOL_RAW, IPV6_CHECKSUM, sockopt); |
| |
| /* request ttl info to be returned in ancillary data */ |
| setsockopt_1(pingsock, SOL_IPV6, IPV6_HOPLIMIT); |
| |
| if (if_index) |
| pingaddr.sin6.sin6_scope_id = if_index; |
| |
| signal(SIGINT, print_stats_and_exit); |
| |
| msg.msg_name = &from; |
| msg.msg_namelen = sizeof(from); |
| msg.msg_iov = &iov; |
| msg.msg_iovlen = 1; |
| msg.msg_control = control_buf; |
| iov.iov_base = G.rcv_packet; |
| iov.iov_len = G.sizeof_rcv_packet; |
| |
| /* start the ping's going ... */ |
| send_ping: |
| sendping6(0); |
| |
| /* listen for replies */ |
| while (1) { |
| int c; |
| struct cmsghdr *mp; |
| int hoplimit = -1; |
| |
| msg.msg_controllen = sizeof(control_buf); |
| c = recvmsg(pingsock, &msg, 0); |
| if (c < 0) { |
| if (errno != EINTR) |
| bb_simple_perror_msg("recvfrom"); |
| continue; |
| } |
| for (mp = CMSG_FIRSTHDR(&msg); mp; mp = CMSG_NXTHDR(&msg, mp)) { |
| if (mp->cmsg_level == SOL_IPV6 |
| && mp->cmsg_type == IPV6_HOPLIMIT |
| /* don't check len - we trust the kernel: */ |
| /* && mp->cmsg_len >= CMSG_LEN(sizeof(int)) */ |
| ) { |
| /*hoplimit = *(int*)CMSG_DATA(mp); - unaligned access */ |
| move_from_unaligned_int(hoplimit, CMSG_DATA(mp)); |
| } |
| } |
| c = unpack6(G.rcv_packet, c, &from, hoplimit); |
| if (pingcount && G.nreceived >= pingcount) |
| break; |
| if (c && (option_mask32 & OPT_A)) { |
| goto send_ping; |
| } |
| } |
| } |
| #endif |
| |
| static void ping(len_and_sockaddr *lsa) |
| { |
| printf("PING %s (%s)", hostname, dotted); |
| if (source_lsa) { |
| printf(" from %s", |
| xmalloc_sockaddr2dotted_noport(&source_lsa->u.sa)); |
| } |
| printf(": %d data bytes\n", datalen); |
| |
| create_icmp_socket(lsa); |
| /* untested whether "-I addr" really works for IPv6: */ |
| if (str_I) |
| setsockopt_bindtodevice(pingsock, str_I); |
| |
| G.sizeof_rcv_packet = datalen + MAXIPLEN + MAXICMPLEN; |
| G.rcv_packet = xzalloc(G.sizeof_rcv_packet); |
| #if ENABLE_PING6 |
| if (lsa->u.sa.sa_family == AF_INET6) { |
| /* +4 reserves a place for timestamp, which may end up sitting |
| * _after_ packet. Saves one if() - see sendping4/6() */ |
| G.snd_packet = xzalloc(datalen + sizeof(struct icmp6_hdr) + 4); |
| ping6(lsa); |
| } else |
| #endif |
| { |
| G.snd_packet = xzalloc(datalen + ICMP_MINLEN + 4); |
| ping4(lsa); |
| } |
| } |
| |
| static int common_ping_main(int opt, char **argv) |
| { |
| len_and_sockaddr *lsa; |
| char *str_s, *str_p; |
| char *str_i = (char*)"1"; |
| duration_t interval; |
| |
| INIT_G(); |
| |
| opt |= getopt32(argv, "^" |
| OPT_STRING |
| /* exactly one arg; -v and -q don't mix */ |
| "\0" "=1:q--v:v--q", |
| &pingcount, &str_s, &opt_ttl, &G.deadline_us, &timeout, &str_I, &str_p, &str_i |
| ); |
| if (opt & OPT_s) |
| datalen = xatou16(str_s); // -s |
| if (opt & OPT_I) { // -I |
| if_index = if_nametoindex(str_I); |
| if (!if_index) { |
| /* TODO: I'm not sure it takes IPv6 unless in [XX:XX..] format */ |
| source_lsa = xdotted2sockaddr(str_I, 0); |
| str_I = NULL; /* don't try to bind to device later */ |
| } |
| } |
| if (opt & OPT_p) |
| G.pattern = xstrtou_range(str_p, 16, 0, 255); |
| if (G.deadline_us) { |
| unsigned d = G.deadline_us < INT_MAX/1000000 ? G.deadline_us : INT_MAX/1000000; |
| G.deadline_us = 1 | ((d * 1000000) + monotonic_us()); |
| } |
| interval = parse_duration_str(str_i); |
| if (interval > INT_MAX/1000000) |
| interval = INT_MAX/1000000; |
| G.interval_us = interval * 1000000; |
| |
| myid = (uint16_t) getpid(); |
| /* we can use native-endian ident, but other Unix ping/traceroute |
| * utils use *big-endian pid*, and e.g. traceroute on our machine may be |
| * *not* from busybox, idents may collide. Follow the convention: |
| */ |
| myid = htons(myid); |
| hostname = argv[optind]; |
| #if ENABLE_PING6 |
| { |
| sa_family_t af = AF_UNSPEC; |
| if (opt & OPT_IPV4) |
| af = AF_INET; |
| if (opt & OPT_IPV6) |
| af = AF_INET6; |
| lsa = xhost_and_af2sockaddr(hostname, 0, af); |
| } |
| #else |
| lsa = xhost_and_af2sockaddr(hostname, 0, AF_INET); |
| #endif |
| |
| if (source_lsa && source_lsa->u.sa.sa_family != lsa->u.sa.sa_family) |
| /* leaking it here... */ |
| source_lsa = NULL; |
| |
| dotted = xmalloc_sockaddr2dotted_noport(&lsa->u.sa); |
| ping(lsa); |
| print_stats_and_exit(0); |
| /*return EXIT_SUCCESS;*/ |
| } |
| #endif /* FEATURE_FANCY_PING */ |
| |
| |
| #if ENABLE_PING |
| int ping_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE; |
| int ping_main(int argc UNUSED_PARAM, char **argv) |
| { |
| # if !ENABLE_FEATURE_FANCY_PING |
| return common_ping_main(AF_UNSPEC, argv); |
| # else |
| return common_ping_main(0, argv); |
| # endif |
| } |
| #endif |
| |
| #if ENABLE_PING6 |
| int ping6_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE; |
| int ping6_main(int argc UNUSED_PARAM, char **argv) |
| { |
| # if !ENABLE_FEATURE_FANCY_PING |
| return common_ping_main(AF_INET6, argv); |
| # else |
| return common_ping_main(OPT_IPV6, argv); |
| # endif |
| } |
| #endif |
| |
| /* from ping6.c: |
| * Copyright (c) 1989 The Regents of the University of California. |
| * All rights reserved. |
| * |
| * This code is derived from software contributed to Berkeley by |
| * Mike Muuss. |
| * |
| * Redistribution and use in source and binary forms, with or without |
| * modification, are permitted provided that the following conditions |
| * are met: |
| * 1. Redistributions of source code must retain the above copyright |
| * notice, this list of conditions and the following disclaimer. |
| * 2. Redistributions in binary form must reproduce the above copyright |
| * notice, this list of conditions and the following disclaimer in the |
| * documentation and/or other materials provided with the distribution. |
| * |
| * 3. <BSD Advertising Clause omitted per the July 22, 1999 licensing change |
| * ftp://ftp.cs.berkeley.edu/pub/4bsd/README.Impt.License.Change> |
| * |
| * 4. Neither the name of the University nor the names of its contributors |
| * may be used to endorse or promote products derived from this software |
| * without specific prior written permission. |
| * |
| * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ''AS IS'' AND |
| * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
| * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| * SUCH DAMAGE. |
| */ |