Commit my hack (yes, I still call it hack) - command line switch for zebra

daemon to change netlink receive buffer size.

doc/ChangeLog

@@ -1,3 +1,7 @@
+2004-08-31 Hasso Tepper <hasso at quagga.net>
+
+	* zebra.8: Document -s/--nl-bufsize command line switch.
+
 2004-08-27 Hasso Tepper <hasso at quagga.net>
 
 	* Update vtysh man page to reflect changes in shell.

doc/zebra.8

@@ -71,6 +71,18 @@ name at the moment. Default is \fIquagga\fR.
 \fB\-r\fR, \fB\-\-retain\fR 
 When the program terminates, retain routes added by \fBzebra\fR.
 .TP
+\fB\-s\fR, \fB\-\-nl-bufsize \fR\fInetlink-buffer-size\fR
+Set netlink receive buffer size. There are cases where zebra daemon can't
+handle flood of netlink messages from kernel. If you ever see "recvmsg overrun"
+messages in zebra log, you are in trouble.
+
+Solution is to increase receive buffer of netlink socket. Note that kernel
+doesn't allow to increase it over maximum value defined in
+\fI/proc/sys/net/core/rmem_max\fR. If you want to do it, you have to increase
+maximum before starting zebra.
+
+Note that this affects Linux only.
+.TP
 \fB\-v\fR, \fB\-\-version\fR
 Print the version and exit.
 .SH FILES

zebra/ChangeLog

@@ -1,3 +1,8 @@
+2004-08-31 Hasso Tepper <hasso at quagga.net>
+
+	* main.c, rt_netlink.c: Added -s command line switch for tuning
+	  netlink receive buffer size in Linux to avoid buffer overruns.
+
 2004-08-26  Miles Nordin  <carton@Ivy.NET>
 
 	* ipforward_sysctl.c (mib_ipv6): Use size_t for len, per

zebra/main.c

@@ -57,6 +57,11 @@ int retain_mode = 0;
 /* Don't delete kernel route. */
 int keep_kernel_mode = 0;
 
+#ifdef HAVE_NETLINK
+/* Receive buffer size for netlink socket */
+u_int32_t nl_rcvbufsize = 0;
+#endif /* HAVE_NETLINK */
+
 /* Command line options. */
 struct option longopts[] = 
 {
@@ -70,6 +75,9 @@ struct option longopts[] =
   { "vty_addr",    required_argument, NULL, 'A'},
   { "vty_port",    required_argument, NULL, 'P'},
   { "retain",      no_argument,       NULL, 'r'},
+#ifdef HAVE_NETLINK
+  { "nl-bufsize",  no_argument,       NULL, 's'},
+#endif /* HAVE_NETLINK */
   { "user",        required_argument, NULL, 'u'},
   { "version",     no_argument,       NULL, 'v'},
   { 0 }
@@ -111,23 +119,28 @@ usage (char *progname, int status)
     fprintf (stderr, "Try `%s --help' for more information.\n", progname);
   else
     {    
-      printf ("Usage : %s [OPTION...]\n\n\
-Daemon which manages kernel routing table management and \
-redistribution between different routing protocols.\n\n\
--b, --batch        Runs in batch mode\n\
--d, --daemon       Runs in daemon mode\n\
--f, --config_file  Set configuration file name\n\
--i, --pid_file     Set process identifier file name\n\
--k, --keep_kernel  Don't delete old routes which installed by zebra.\n\
--l, --log_mode     Set verbose log mode flag\n\
--A, --vty_addr     Set vty's bind address\n\
--P, --vty_port     Set vty's port number\n\
--r, --retain       When program terminates, retain added route by zebra.\n\
--u, --user         User and group to run as\n\
--v, --version      Print program version\n\
--h, --help         Display this help and exit\n\
-\n\
-Report bugs to %s\n", progname, ZEBRA_BUG_ADDRESS);
+      printf ("Usage : %s [OPTION...]\n\n"\
+	      "Daemon which manages kernel routing table management and "\
+	      "redistribution between different routing protocols.\n\n"\
+	      "-b, --batch        Runs in batch mode\n"\
+	      "-d, --daemon       Runs in daemon mode\n"\
+	      "-f, --config_file  Set configuration file name\n"\
+	      "-i, --pid_file     Set process identifier file name\n"\
+	      "-k, --keep_kernel  Don't delete old routes which installed by "\
+				  "zebra.\n"\
+	      "-l, --log_mode     Set verbose log mode flag\n"\
+	      "-A, --vty_addr     Set vty's bind address\n"\
+	      "-P, --vty_port     Set vty's port number\n"\
+	      "-r, --retain       When program terminates, retain added route "\
+				  "by zebra.\n"\
+	      "-u, --user         User and group to run as\n", progname);
+#ifdef HAVE_NETLINK
+      printf ("-s, --nl-bufsize   Set netlink receive buffer size\n");
+#endif /* HAVE_NETLINK */
+      printf ("-v, --version      Print program version\n"\
+	      "-h, --help         Display this help and exit\n"\
+	      "\n"\
+	      "Report bugs to %s\n", ZEBRA_BUG_ADDRESS);
     }
 
   exit (status);
@@ -216,7 +229,11 @@ main (int argc, char **argv)
     {
       int opt;
   
+#ifdef HAVE_NETLINK  
+      opt = getopt_long (argc, argv, "bdklf:i:hA:P:ru:vs:", longopts, 0);
+#else
       opt = getopt_long (argc, argv, "bdklf:i:hA:P:ru:v", longopts, 0);
+#endif /* HAVE_NETLINK */
 
       if (opt == EOF)
 	break;
@@ -259,6 +276,11 @@ main (int argc, char **argv)
 	case 'r':
 	  retain_mode = 1;
 	  break;
+#ifdef HAVE_NETLINK
+	case 's':
+	  nl_rcvbufsize = atoi (optarg);
+	  break;
+#endif /* HAVE_NETLINK */
   case 'u':
     zserv_privs.user = zserv_privs.group = optarg;
     break;

zebra/rt_netlink.c

@@ -84,6 +84,8 @@ extern struct zebra_t zebrad;
 
 extern struct zebra_privs_t zserv_privs;
 
+extern u_int32_t nl_rcvbufsize;
+
 /* Make socket for Linux netlink interface. */
 static int
 netlink_socket (struct nlsock *nl, unsigned long groups)
@@ -110,6 +112,48 @@ netlink_socket (struct nlsock *nl, unsigned long groups)
       return -1;
     }
 
+  /* Set receive buffer size if it's set from command line */
+  if (nl_rcvbufsize)
+    {
+      u_int32_t oldsize, oldlen;
+      u_int32_t newsize, newlen;
+
+      oldlen = sizeof(oldsize);
+      newlen = sizeof(newsize);
+
+      ret = getsockopt(sock, SOL_SOCKET, SO_RCVBUF, &oldsize, &oldlen);
+      if (ret < 0)
+	{
+	  zlog (NULL, LOG_ERR, "Can't get %s receive buffer size: %s", nl->name,
+		strerror (errno));
+	  close (sock);
+	  return -1;
+	}
+
+      ret = setsockopt(sock, SOL_SOCKET, SO_RCVBUF, &nl_rcvbufsize,
+		       sizeof(nl_rcvbufsize));
+      if (ret < 0)
+	{
+	  zlog (NULL, LOG_ERR, "Can't set %s receive buffer size: %s", nl->name,
+		strerror (errno));
+	  close (sock);
+	  return -1;
+	}
+
+      ret = getsockopt(sock, SOL_SOCKET, SO_RCVBUF, &newsize, &newlen);
+      if (ret < 0)
+	{
+	  zlog (NULL, LOG_ERR, "Can't get %s receive buffer size: %s", nl->name,
+		strerror (errno));
+	  close (sock);
+	  return -1;
+	}
+
+      zlog (NULL, LOG_INFO,
+	    "Setting netlink socket receive buffer size: %u -> %u",
+	    oldsize, newsize);
+    }
+
   memset (&snl, 0, sizeof snl);
   snl.nl_family = AF_NETLINK;
   snl.nl_groups = groups;