Change default time period to 30s and relax default loss interval to 5x send interval to address nagging low level loss reports

Add usage note about loss percentage reporting and calculation

README.md

@@ -1,9 +1,26 @@
 # dpinger
 
-dpinger is a daemon for continous monitoring of latency and loss on a network connection. It is
+dpinger is a daemon for continuous monitoring of latency and loss on a network connection. It is
 intended for use by firewalls to monitor link health, as well as for providing information to
 various monitoring systems such as Cacti, Nagios, Zabbix, etc. 
 
+The output of dpinger can either be file or socket based, and consists of three numbers:
+  
+    <Average Latency in μs> <Standard Deviation in μs> <Percentage of Loss>
+    
+dpinger also provides for invocation of a command based upon threshold values
+for Average Latency or Percentage of Loss. Arguments to the command are:
+
+    <Target IP> <Alarm on/off> <Average Latency> <Standard Deviation> <Percentage of Loss>
+
+In addition to command invocation, dpinger can also log alerts via syslog. 
+
+If several instances of dpinger are being used to monitor different targets, or the same target
+with different source addresses, etc., an Identifier can be added to the output to identify
+which instance of dpinger is the source. This is particularly useful with syslog.
+
+<br>
+
 Usage examples:
 
     dpinger -t 300s -r 60s 192.168.0.1 >> /tmp/dpinger.out
@@ -38,3 +55,9 @@ Produce a report every 60 seconds and append it to /tmp/dpinger.out.
   Monitor IP address fe80::1 for latency and loss. Send echo requests every 200 milliseconds.
   Make current status available on demand via a Unix domain socket /tmp/igb1.status. Record
   process id in /run/dpinger.
+
+    dpinger -S -i Comcast -s 5s -t 600s -r 0 -L 10% -p /run/dpinger 8.8.8.8
+
+  Monitor IP address 8.8.8.8 for latency and loss. Send echo requests every five seconds and
+  average results over 10 minutes. Log alerts via syslog including identifier string "Comcast"
+  if average loss exceeds 10 percent. Record process id in /run/dpinger.

dpinger.c

@@ -38,6 +38,8 @@
 #include <fcntl.h>
 #include <signal.h>
 
+#include <netdb.h>
+#include <net/if.h>
 #include <sys/socket.h>
 #include <sys/un.h>
 #include <sys/stat.h>
@@ -50,6 +52,18 @@
 #include <pthread.h>
 #include <syslog.h>
 
+// TODO:
+//
+// After December 31st, 2016, review use of fcntl() for setting non blocking
+// and close on exec. It would be preferable to use accept4(), SOCK_CLOEXEC
+// and SOCK_NONBLOCK. These are currently avoided to allow use on older
+// systems such as FreeBSD 9.3, Linux 2.6.26.
+// For Linux accept4() currently requires defining _GNU_SOURCE which we would
+// like to avoid.
+// For FreeBSD, these definitions were introduced with FreeBSD 10.0 and are
+// not present in 9.3 which is supported through 2016.
+
+
 // Who we are
 static const char *             progname;
 
@@ -62,10 +76,11 @@ static unsigned int             flag_rewind = 0;
 static unsigned int             flag_syslog = 0;
 
 // String representation of target
-static char                     dest_str[INET6_ADDRSTRLEN];
+#define ADDR_STR_MAX            (INET6_ADDRSTRLEN + IF_NAMESIZE + 1)
+static char                     dest_str[ADDR_STR_MAX];
 
 // Time period over which we are averaging results in ms
-static unsigned long            time_period_msec = 25000;
+static unsigned long            time_period_msec = 30000;
 
 // Interval between sends in ms
 static unsigned long            send_interval_msec = 250;
@@ -662,6 +677,7 @@ usocket_thread(
     while (1)
     {
         sock_fd = accept(usocket_fd, NULL, NULL);
+        (void) fcntl(sock_fd, F_SETFL, FD_CLOEXEC);
         (void) fcntl(sock_fd, F_SETFL, fcntl(sock_fd, F_GETFL, 0) | O_NONBLOCK);
 
         report(&average_latency_usec, &latency_deviation, &average_loss_percent);
@@ -772,8 +788,8 @@ usage(void)
     fprintf(stderr, "    -S log warnings via syslog\n");
     fprintf(stderr, "    -B bind (source) address\n");
     fprintf(stderr, "    -s time interval between echo requests (default 250ms)\n");
-    fprintf(stderr, "    -l time interval before packets are treated as lost (default 2x send interval)\n");
-    fprintf(stderr, "    -t time period over which results are averaged (default 25s)\n");
+    fprintf(stderr, "    -l time interval before packets are treated as lost (default 5x send interval)\n");
+    fprintf(stderr, "    -t time period over which results are averaged (default 30s)\n");
     fprintf(stderr, "    -r time interval between reports (default 1s)\n");
     fprintf(stderr, "    -o output file for reports (default stdout)\n");
     fprintf(stderr, "    -A time interval between alerts (default 1s)\n");
@@ -784,11 +800,13 @@ usage(void)
     fprintf(stderr, "    -u unix socket name for polling\n");
     fprintf(stderr, "    -p process id file name\n\n");
     fprintf(stderr, "  notes:\n");
+    fprintf(stderr, "    IP addresses can be in either IPv4 or IPv6 format\n\n");
     fprintf(stderr, "    time values can be expressed with a suffix of 'm' (milliseconds) or 's' (seconds)\n");
     fprintf(stderr, "    if no suffix is specified, milliseconds is the default\n\n");
-    fprintf(stderr, "    IP addresses can be in either IPv4 or IPv6 format\n\n");
     fprintf(stderr, "    the output format is \"latency_avg latency_stddev loss_pct\"\n");
-    fprintf(stderr, "    latency values are output in microseconds\n\n");
+    fprintf(stderr, "    latency values are output in microseconds\n");
+    fprintf(stderr, "    loss percentage is reported in whole numbers of 0-100\n");
+    fprintf(stderr, "    resolution of loss calculation is: 100 * send_interval / (time_period - loss_interval)\n\n");
     fprintf(stderr, "    the alert_cmd is invoked as \"alert_cmd dest_addr alarm_flag latency_avg loss_avg\"\n");
     fprintf(stderr, "    alarm_flag is set to 1 if either latency or loss is in alarm state\n");
     fprintf(stderr, "    alarm_flag will return to 0 when both have have cleared alarm state\n\n");
@@ -824,8 +842,8 @@ parse_args(
     int                         argc,
     char * const                argv[])
 {
-    struct in_addr              addr;
-    struct in6_addr             addr6;
+    struct addrinfo             hint;
+    struct addrinfo *           addr_info;
     const char *                dest_arg;
     const char *                bind_arg = NULL;
     size_t                      len;
@@ -904,6 +922,7 @@ parse_args(
             {
                 fatal("invalid latency alarm threshold %s\n", optarg);
             }
+            latency_alarm_threshold_usec = latency_alarm_threshold_msec * 1000;
             break;
 
         case 'L':
@@ -956,6 +975,7 @@ parse_args(
         usage();
         fatal(NULL);
     }
+    dest_arg = argv[optind];
 
     // Ensure we have something to do: at least one of alarm, report, socket
     if (report_interval_msec == 0 && latency_alarm_threshold_msec == 0 && loss_alarm_threshold_percent == 0 && usocket_name == NULL)
@@ -963,9 +983,6 @@ parse_args(
         fatal("no activity enabled\n");
     }
 
-    // Destination address
-    dest_arg = argv[optind];
-
     // Ensure we have something to average over
     if (time_period_msec < send_interval_msec)
     {
@@ -979,61 +996,50 @@ parse_args(
         fatal("ratio of time period to send interval cannot exceed 65536:1\n");
     }
 
-    // Check for an IPv4 address
-    r = inet_pton(AF_INET, dest_arg, &addr);
-    if (r)
+    // Check destination address
+    memset(&hint, 0, sizeof(struct addrinfo));
+    hint.ai_flags = AI_NUMERICHOST;
+    hint.ai_family = AF_UNSPEC;
+    hint.ai_socktype = SOCK_RAW;
+
+    r = getaddrinfo(dest_arg, NULL, &hint, &addr_info);
+    if (r != 0)
     {
-        struct sockaddr_in * dest = (struct sockaddr_in *) &dest_addr;
-        dest->sin_family = AF_INET;
-        dest->sin_addr = addr;
-        dest_addr_len = sizeof(struct sockaddr_in);
-
-        if (bind_arg)
-        {
-            r = inet_pton(AF_INET, bind_arg, &addr);
-            if (r == 0)
-            {
-                fatal("Invalid bind IP address %s\n", bind_arg);
-            }
-
-            struct sockaddr_in * bind4 = (struct sockaddr_in *) &bind_addr;
-            bind4->sin_family = AF_INET;
-            bind4->sin_addr = addr;
-            bind_addr_len = sizeof(struct sockaddr_in);
-        }
+        fatal("invalid destination IP address %s\n", dest_arg);
     }
-    else
+
+    if (addr_info->ai_family == AF_INET6)
     {
-        // Perhaps it's an IPv6 address?
-        r = inet_pton(AF_INET6, dest_arg, &addr6);
-        if (r == 0)
-        {
-            fatal("Invalid destination IP address %s\n", dest_arg);
-        }
-
-        struct sockaddr_in6 * dest6 = (struct sockaddr_in6 *) &dest_addr;
-        dest6->sin6_family = AF_INET6;
-        dest6->sin6_addr = addr6;
-        dest_addr_len = sizeof(struct sockaddr_in6);
-
         af_family = AF_INET6;
         ip_proto = IPPROTO_ICMPV6;
         echo_request_type = ICMP6_ECHO_REQUEST;
         echo_reply_type = ICMP6_ECHO_REPLY;
+    }
+    else if (addr_info->ai_family != AF_INET)
+    {
+        fatal("invalid destination IP address %s\n", dest_arg);
+    }
 
-        if (bind_arg)
+
+    dest_addr_len = addr_info->ai_addrlen;
+    memcpy(&dest_addr, addr_info->ai_addr, dest_addr_len);
+    freeaddrinfo(addr_info);
+
+    // Check bind address
+    if (bind_arg)
+    {
+        // Address family must match
+        hint.ai_family = af_family;
+
+        r = getaddrinfo(bind_arg, NULL, &hint, &addr_info);
+        if (r != 0)
         {
-            r = inet_pton(AF_INET6, bind_arg, &addr6);
-            if (r == 0)
-            {
-                fatal("Invalid source IP address %s\n", bind_arg);
-            }
-
-            struct sockaddr_in6 * bind6 = (struct sockaddr_in6 *) &bind_addr;
-            bind6->sin6_family = AF_INET6;
-            bind6->sin6_addr = addr6;
-            bind_addr_len = sizeof(struct sockaddr_in6);
+           fatal("invalid bind IP address %s\n", bind_arg);
         }
+
+        bind_addr_len = addr_info->ai_addrlen;
+        memcpy(&bind_addr, addr_info->ai_addr, bind_addr_len);
+        freeaddrinfo(addr_info);
     }
 }
 
@@ -1046,10 +1052,8 @@ main(
     int                         argc,
     char                        *argv[])
 {
-    char                        bind_str[INET6_ADDRSTRLEN] = "(none)";
-    const void *                addr;
-    const char *                p;
-    int                         pidfile_fd;
+    char                        bind_str[ADDR_STR_MAX] = "(none)";
+    int                         pidfile_fd = -1;
     pthread_t                   thread;
     struct                      sigaction act;
     int                         r;
@@ -1064,12 +1068,15 @@ main(
         perror("socket");
         fatal("cannot create send socket\n");
     }
+    (void) fcntl(send_sock, F_SETFL, FD_CLOEXEC);
+
     recv_sock = socket(af_family, SOCK_RAW, ip_proto);
     if (recv_sock == -1)
     {
         perror("socket");
         fatal("cannot create recv socket\n");
     }
+    (void) fcntl(recv_sock, F_SETFL, FD_CLOEXEC);
 
     // Bind our sockets to an address if requested
     if (bind_addr_len)
@@ -1095,7 +1102,7 @@ main(
     // Create report file
     if (report_name)
     {
-        report_fd = open(report_name, O_WRONLY | O_CREAT | O_TRUNC, 0644);
+        report_fd = open(report_name, O_WRONLY | O_CREAT | O_TRUNC | O_CLOEXEC, 0644);
         if (report_fd == -1)
         {
             perror("open");
@@ -1117,12 +1124,13 @@ main(
             fatal("socket name too large\n");
         }
 
-        usocket_fd = socket(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC, 0);
+        usocket_fd = socket(AF_UNIX, SOCK_STREAM, 0);
         if (usocket_fd == -1)
         {
             perror("socket");
             fatal("cannot create unix domain socket\n");
         }
+        (void) fcntl(usocket_fd, F_SETFL, FD_CLOEXEC);
 
         (void) unlink(usocket_name);
 
@@ -1154,7 +1162,7 @@ main(
     // Create pid file
     if (pidfile_name)
     {
-        pidfile_fd = open(pidfile_name, O_WRONLY | O_CREAT | O_TRUNC, 0644);
+        pidfile_fd = open(pidfile_name, O_WRONLY | O_CREAT | O_TRUNC | O_CLOEXEC, 0644);
         if (pidfile_fd == -1)
         {
             perror("open");
@@ -1190,7 +1198,7 @@ main(
     sigaction(SIGINT, &act, NULL);
 
     // Write pid file
-    if (pidfile_name)
+    if (pidfile_fd != -1)
     {
         char                    buf[64];
         int                     len;
@@ -1208,7 +1216,7 @@ main(
             fatal("error writing pidfile\n");
         }
 
-        r= close(pidfile_fd);
+        r = close(pidfile_fd);
         if (r == -1)
         {
             perror("close");
@@ -1227,51 +1235,31 @@ main(
     // Set the default loss interval
     if (loss_interval_msec == 0)
     {
-        loss_interval_msec = send_interval_msec * 2;
+        loss_interval_msec = send_interval_msec * 5;
     }
+    loss_interval_usec = loss_interval_msec * 1000;
 
-    // Log our parameters
-    if (af_family == AF_INET)
+    // Log our general parameters
+    r = getnameinfo((struct sockaddr *) &dest_addr, dest_addr_len, dest_str, sizeof(dest_str), NULL, 0, NI_NUMERICHOST);
+    if (r != 0)
     {
-        addr = (const void *) &((struct sockaddr_in *) &dest_addr)->sin_addr;
-    }
-    else
-    {
-        addr = (const void *) &((struct sockaddr_in6 *) &dest_addr)->sin6_addr;
-    }
-    p = inet_ntop(af_family, addr, dest_str, sizeof(dest_str));
-    if (p == NULL)
-    {
-        fatal("inet_ntop of destination address failed\n");
+        fatal("getnameinfo of destination address failed\n");
     }
 
     if (bind_addr_len)
     {
-        if (af_family == AF_INET)
+        r = getnameinfo((struct sockaddr *) &bind_addr, bind_addr_len, bind_str, sizeof(bind_str), NULL, 0, NI_NUMERICHOST);
+        if (r != 0)
         {
-            addr = (const void *) &((struct sockaddr_in *) &bind_addr)->sin_addr;
-        }
-        else
-        {
-            addr = (const void *) &((struct sockaddr_in6 *) &bind_addr)->sin6_addr;
-        }
-        p = inet_ntop(af_family, addr, bind_str, sizeof(bind_str));
-        if (p == NULL)
-        {
-            fatal("inet_ntop of bind address failed\n");
+            fatal("getnameinfo of bind address failed\n");
         }
     }
 
-    // Log our general parameters
     logger("send_interval %lums  loss_interval %lums  time_period %lums  report_interval %lums  alert_interval %lums  latency_alarm %lums  loss_alarm %lu%%  dest_addr %s  bind_addr %s  identifier \"%s\"\n",
            send_interval_msec, loss_interval_msec, time_period_msec, report_interval_msec,
            alert_interval_msec, latency_alarm_threshold_msec, loss_alarm_threshold_percent,
            dest_str, bind_str, identifier);
 
-    // Convert loss interval and alarm threshold to microseconds
-    loss_interval_usec = loss_interval_msec * 1000;
-    latency_alarm_threshold_usec = latency_alarm_threshold_msec * 1000;
-
     // Set my echo id
     echo_id = htons(getpid());