/* ping.c - check network connectivity * * Copyright 2014 Rob Landley * * Not in SUSv4. * * Note: ping_group_range should never have existed. To disable it, do: * echo 0 999999999 > /proc/sys/net/ipv4/ping_group_range * (Android does this by default in its init script.) * * Yes, I wimped out and capped -s at sizeof(toybuf), waiting for a complaint... // -s > 4064 = sizeof(toybuf)-sizeof(struct icmphdr)-CMSG_SPACE(sizeof(uint8_t)), then kernel adds 20 bytes USE_PING(NEWTOY(ping, "<1>1m#t#<0>255=64c#<0=3s#<0>4064=56i%W#<0=3w#<0qf46I:[-46]", TOYFLAG_USR|TOYFLAG_BIN)) USE_PING(OLDTOY(ping6, ping, TOYFLAG_USR|TOYFLAG_BIN)) config PING bool "ping" default y help usage: ping [OPTIONS] HOST Check network connectivity by sending packets to a host and reporting its response. Send ICMP ECHO_REQUEST packets to ipv4 or ipv6 addresses and prints each echo it receives back, with round trip time. Returns true if host alive. Options: -4, -6 Force IPv4 or IPv6 -c CNT Send CNT many packets (default 3, 0 = infinite) -f Flood (print . and \b to show drops, default -c 15 -i 0.2) -i TIME Interval between packets (default 1, need root for < .2) -I IFACE/IP Source interface or address -m MARK Tag outgoing packets using SO_MARK -q Quiet (stops after one returns true if host is alive) -s SIZE Data SIZE in bytes (default 56) -t TTL Set Time To Live (number of hops) -W SEC Seconds to wait for response after last -c packet (default 3) -w SEC Exit after this many seconds */ #define FOR_ping #include "toys.h" #include #include GLOBALS( char *I; long w, W, i, s, c, t, m; struct sockaddr *sa; int sock; unsigned long sent, recv, fugit, min, max; ) // Print a summary. Called as a single handler or at exit. static void summary(int sig) { if (!FLAG(q) && TT.sent && TT.sa) { printf("\n--- %s ping statistics ---\n", ntop(TT.sa)); printf("%lu packets transmitted, %lu received, %ld%% packet loss\n", TT.sent, TT.recv, ((TT.sent-TT.recv)*100)/(TT.sent?TT.sent:1)); if (TT.recv) printf("round-trip min/avg/max = %lu/%lu/%lu ms\n", TT.min, TT.fugit/TT.recv, TT.max); } TT.sa = 0; } // assumes aligned and can read even number of bytes static unsigned short pingchksum(unsigned short *data, int len) { unsigned short u = 0, d; // circular carry is endian independent: bits from high byte go to low byte while (len>0) { d = *data++; if (len == 1) d &= 255<= (u += d)) u++; len -= 2; } return u; } static int xrecvmsgwait(int fd, struct msghdr *msg, int flag, union socksaddr *sa, int timeout) { socklen_t sl = sizeof(*sa); int len; if (timeout >= 0) { struct pollfd pfd; pfd.fd = fd; pfd.events = POLLIN; if (!xpoll(&pfd, 1, timeout)) return 0; } msg->msg_name = (void *)sa; msg->msg_namelen = sl; len = recvmsg(fd, msg, flag); if (len<0) perror_exit("recvmsg"); return len; } void ping_main(void) { struct addrinfo *ai, *ai2; struct ifaddrs *ifa, *ifa2 = 0; struct icmphdr *ih = (void *)toybuf; struct msghdr msg; struct cmsghdr *cmsg; struct iovec iov; union socksaddr srcaddr, srcaddr2; struct sockaddr *sa = (void *)&srcaddr; int family = 0, ttl = 0, len; long long tnext, tW, tnow, tw; unsigned short seq = 0, pkttime; // Set nonstatic default values if (!FLAG(i)) TT.i = FLAG(f) ? 200 : 1000; else if (TT.i<200 && getuid()) error_exit("need root for -i <200"); if (!FLAG(s)) TT.s = 56; // 64-PHDR_LEN if (FLAG(f) && !FLAG(c)) TT.c = 15; // ipv4 or ipv6? (0 = autodetect if -I or arg have only one address type.) if (FLAG(6) || strchr(toys.which->name, '6')) family = AF_INET6; else if (FLAG(4)) family = AF_INET; else family = 0; // If -I srcaddr look it up. Allow numeric address of correct type. memset(&srcaddr, 0, sizeof(srcaddr)); if (TT.I) { if (!FLAG(6) && inet_pton(AF_INET, TT.I, (void *)&srcaddr.in.sin_addr)) family = AF_INET; else if (!FLAG(4) && inet_pton(AF_INET6, TT.I, (void *)&srcaddr.in6.sin6_addr)) family = AF_INET6; else if (getifaddrs(&ifa2)) perror_exit("getifaddrs"); } // Look up HOST address, filtering for correct type and interface. // If -I but no -46 then find compatible type between -I and HOST ai2 = xgetaddrinfo(*toys.optargs, 0, family, 0, 0, 0); for (ai = ai2; ai; ai = ai->ai_next) { // correct type? if (family && family!=ai->ai_family) continue; if (ai->ai_family!=AF_INET && ai->ai_family!=AF_INET6) continue; // correct interface? if (!TT.I || !ifa2) break; for (ifa = ifa2; ifa; ifa = ifa->ifa_next) { if (!ifa->ifa_addr || ifa->ifa_addr->sa_family!=ai->ai_family || strcmp(ifa->ifa_name, TT.I)) continue; sa = (void *)ifa->ifa_addr; break; } if (ifa) break; } if (!ai) error_exit("no v%d addr for -I %s", 4+2*(family==AF_INET6), TT.I); TT.sa = ai->ai_addr; // Open DGRAM socket sa->sa_family = ai->ai_family; TT.sock = socket(ai->ai_family, SOCK_DGRAM, len = (ai->ai_family == AF_INET) ? IPPROTO_ICMP : IPPROTO_ICMPV6); if (TT.sock == -1) { perror_msg("socket SOCK_DGRAM %x", len); if (errno == EACCES) { fprintf(stderr, "Kernel bug workaround:\n" "echo 0 99999999 | sudo tee /proc/sys/net/ipv4/ping_group_range\n"); } xexit(); } if (TT.I) xbind(TT.sock, sa, sizeof(srcaddr)); len = 1; xsetsockopt(TT.sock, SOL_IP, IP_RECVTTL, &len, sizeof(len)); if (FLAG(m)) { len = TT.m; xsetsockopt(TT.sock, SOL_SOCKET, SO_MARK, &len, sizeof(len)); } if (TT.t) { len = TT.t; if (ai->ai_family == AF_INET) xsetsockopt(TT.sock, IPPROTO_IP, IP_TTL, &len, 4); else xsetsockopt(TT.sock, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &len, sizeof(len)); } if (!FLAG(q)) { printf("Ping %s (%s)", *toys.optargs, ntop(TT.sa)); if (TT.I) { *toybuf = 0; printf(" from %s (%s)", TT.I, ntop(sa)); } // 20 byte TCP header, 8 byte ICMP header, plus data payload printf(": %ld(%ld) bytes.\n", TT.s, TT.s+28); } TT.min = ULONG_MAX; toys.exitval = 1; tW = tw = 0; tnext = millitime(); if (TT.w) tw = TT.w*1000+tnext; memset(&msg, 0, sizeof(msg)); // left enought space to store ttl value len = CMSG_SPACE(sizeof(uint8_t)); iov.iov_base = (void *)toybuf; iov.iov_len = sizeof(toybuf) - len; msg.msg_iov = &iov; msg.msg_iovlen = 1; msg.msg_control = &toybuf[iov.iov_len]; msg.msg_controllen = len; sigatexit(summary); // Send/receive packets for (;;) { int waitms = INT_MAX; // Exit due to timeout? (TODO: timeout is after last packet, waiting if // any packets ever dropped. Not timeout since packet was dropped.) tnow = millitime(); if (tW) { if (0>=(waitms = tW-tnow) || !(TT.sent-TT.recv)) break; waitms = tW-tnow; } if (tw) { if (tnow>tw) break; else if (waitms>tw-tnow) waitms = tw-tnow; } // Time to send the next packet? if (!tW && tnext-tnow <= 0) { tnext += TT.i; memset(ih, 0, sizeof(*ih)); ih->type = (ai->ai_family == AF_INET) ? 8 : 128; ih->un.echo.id = getpid(); ih->un.echo.sequence = ++seq; if (TT.s >= 4) *(unsigned *)(ih+1) = tnow; ih->checksum = pingchksum((void *)toybuf, TT.s+sizeof(*ih)); xsendto(TT.sock, toybuf, TT.s+sizeof(*ih), TT.sa); TT.sent++; if (FLAG(f) && !FLAG(q)) xputc('.'); // last packet? if (TT.c) if (!--TT.c) { tW = tnow + TT.W*1000; waitms = 1; // check for immediate return even when W=0 } } // This is down here so it's against new period if we just sent a packet if (!tW && waitms>tnext-tnow) waitms = tnext-tnow; // wait for next packet or timeout if (waitms<0) waitms = 0; if (!(len = xrecvmsgwait(TT.sock, &msg, 0, &srcaddr2, waitms))) continue; TT.recv++; TT.fugit += (pkttime = millitime()-*(unsigned *)(ih+1)); if (pkttime < TT.min) TT.min = pkttime; if (pkttime > TT.max) TT.max = pkttime; // reply id == 0 for ipv4, 129 for ipv6 cmsg = CMSG_FIRSTHDR(&msg); for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) { if (cmsg->cmsg_level == IPPROTO_IP && cmsg->cmsg_type == IP_TTL) { ttl = *(uint8_t *)CMSG_DATA(cmsg); break; } }; if (!FLAG(q)) { if (FLAG(f)) xputc('\b'); else { printf("%d bytes from %s: icmp_seq=%d ttl=%d", len, ntop(&srcaddr2.s), ih->un.echo.sequence, ttl); if (len >= sizeof(*ih)+4) printf(" time=%u ms", pkttime); xputc('\n'); } } toys.exitval = 0; } sigatexit(0); summary(0); if (CFG_TOYBOX_FREE) { freeaddrinfo(ai2); if (ifa2) freeifaddrs(ifa2); } }