Rewrote the kernel syncer. The old layering was horrible.

The new kernel syncer is cleanly split between generic UNIX module
and OS dependent submodules:

  -  krt.c (the generic part)
  -  krt-iface (low-level functions for interface handling)
  -  krt-scan (low-level functions for routing table scanning)
  -  krt-set (low-level functions for setting of kernel routes)

krt-set and krt-iface are common for all BSD-like Unices, krt-scan is heavily
system dependent (most Unices require /dev/kmem parsing, Linux uses /proc),
Netlink substitues all three modules.

We expect each UNIX port supports kernel routing table scanning, kernel
interface table scanning, kernel route manipulation and possibly also
asynchronous event notifications (new route, interface state change;
not implemented yet) and build the KRT protocol on the top of these
primitive operations.

1 files changed, 194 insertions, 0 deletions

diff --git a/sysdep/unix/krt-iface.c b/sysdep/unix/krt-iface.c
new file mode 100644
index 00000000..3b665a44
--- /dev/null
+++ b/sysdep/unix/krt-iface.c
@@ -0,0 +1,194 @@
+/*
+ *	BIRD -- Unix Interface Scanning and Syncing
+ *
+ *	(c) 1998--1999 Martin Mares <mj@ucw.cz>
+ *
+ *	Can be freely distributed and used under the terms of the GNU GPL.
+ */
+
+#include <string.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <net/if.h>
+#include <sys/ioctl.h>
+#include <errno.h>
+
+#define LOCAL_DEBUG
+
+#include "nest/bird.h"
+#include "nest/iface.h"
+#include "nest/route.h"
+#include "nest/protocol.h"
+#include "lib/timer.h"
+#include "lib/krt.h"
+
+#include "unix.h"
+
+int if_scan_sock = -1;
+
+static void
+scan_ifs(struct ifreq *r, int cnt)
+{
+  struct iface i;
+  char *err;
+  unsigned fl;
+  ip_addr netmask;
+  int l;
+
+  if_start_update();
+  for (cnt /= sizeof(struct ifreq); cnt; cnt--, r++)
+    {
+      bzero(&i, sizeof(i));
+      DBG("%s\n", r->ifr_ifrn.ifrn_name);
+      strncpy(i.name, r->ifr_ifrn.ifrn_name, sizeof(i.name) - 1);
+      get_sockaddr((struct sockaddr_in *) &r->ifr_addr, &i.ip, NULL);
+      if (ipa_nonzero(i.ip))
+	{
+	  l = ipa_classify(i.ip);
+	  if (l < 0 || !(l & IADDR_HOST))
+	    {
+	      log(L_ERR "%s: Invalid interface address", i.name);
+	      i.ip = IPA_NONE;
+	    }
+	  else if ((l & IADDR_SCOPE_MASK) == SCOPE_HOST)
+	    i.flags |= IF_LOOPBACK | IF_IGNORE;
+	}
+
+      if (ioctl(if_scan_sock, SIOCGIFFLAGS, r) < 0)
+	{
+	  err = "SIOCGIFFLAGS";
+	faulty:
+	  log(L_ERR "%s(%s): %m", err, i.name);
+	bad:
+	  i.flags = (i.flags & ~IF_LINK_UP) | IF_ADMIN_DOWN;
+	  continue;
+	}
+      fl = r->ifr_flags;
+      if (fl & IFF_UP)
+	i.flags |= IF_LINK_UP;
+
+      if (ioctl(if_scan_sock, SIOCGIFNETMASK, r) < 0)
+	{ err = "SIOCGIFNETMASK"; goto faulty; }
+      get_sockaddr((struct sockaddr_in *) &r->ifr_addr, &netmask, NULL);
+      l = ipa_mklen(netmask);
+      if (l < 0 || l == 31)
+	{
+	  log(L_ERR "%s: Invalid netmask", i.name);
+	  goto bad;
+	}
+      i.pxlen = l;
+
+      if (fl & IFF_POINTOPOINT)
+	{
+	  i.flags |= IF_UNNUMBERED;
+	  i.pxlen = BITS_PER_IP_ADDRESS;
+	  if (ioctl(if_scan_sock, SIOCGIFDSTADDR, r) < 0)
+	    { err = "SIOCGIFDSTADDR"; goto faulty; }
+	  get_sockaddr((struct sockaddr_in *) &r->ifr_addr, &i.opposite, NULL);
+	}
+      if (fl & IFF_LOOPBACK)
+	i.flags |= IF_LOOPBACK | IF_IGNORE;
+#ifndef CONFIG_ALL_MULTICAST
+      if (fl & IFF_MULTICAST)
+#endif
+	i.flags |= IF_MULTICAST;
+
+      i.prefix = ipa_and(i.ip, ipa_mkmask(i.pxlen));
+      if (i.pxlen < 32)
+	{
+	  i.brd = ipa_or(i.prefix, ipa_not(ipa_mkmask(i.pxlen)));
+	  if (ipa_equal(i.ip, i.prefix) || ipa_equal(i.ip, i.brd))
+	    {
+	      log(L_ERR "%s: Using network or broadcast address for interface", i.name);
+	      goto bad;
+	    }
+	  if (fl & IFF_BROADCAST)
+	    i.flags |= IF_BROADCAST;
+	  if (i.pxlen < 30)
+	    i.flags |= IF_MULTIACCESS;
+	  else
+	    i.opposite = ipa_opposite(i.ip);
+	}
+      else
+	i.brd = i.opposite;
+
+      if (ioctl(if_scan_sock, SIOCGIFMTU, r) < 0)
+	{ err = "SIOCGIFMTU"; goto faulty; }
+      i.mtu = r->ifr_mtu;
+
+#ifdef SIOCGIFINDEX
+      if (ioctl(if_scan_sock, SIOCGIFINDEX, r) < 0)
+	DBG("SIOCGIFINDEX failed: %m\n");
+      else
+	i.index = r->ifr_ifindex;
+#else
+      /* FIXME: What else? Guess ifindex (we need it at least for OSPF on unnumbered links)? */
+#endif
+
+      if_update(&i);
+    }
+  if_end_update();
+}
+
+void
+krt_if_scan(struct krt_proto *p)
+{
+  struct ifconf ic;
+  static int last_ifbuf_size = 4*sizeof(struct ifreq);
+  int res;
+
+  for(;;)
+    {
+      if (last_ifbuf_size)
+	{
+	  struct ifreq *r = alloca(last_ifbuf_size);
+	  ic.ifc_ifcu.ifcu_req = r;
+	  ic.ifc_len = last_ifbuf_size;
+	  res = ioctl(if_scan_sock, SIOCGIFCONF, &ic);
+	  if (res < 0 && errno != EFAULT)
+	    die("SIOCCGIFCONF: %m");
+	  if (res < last_ifbuf_size)
+	    {
+	      scan_ifs(r, ic.ifc_len);
+	      break;
+	    }
+	}
+#ifdef CLEAN_WAY_WORKING_ONLY_ON_LINUX_2_1	/* FIXME */
+      ic.ifc_req = NULL;
+      ic.ifc_len = 999999999;
+      if (ioctl(if_scan_sock, SIOCGIFCONF, &ic) < 0)
+	die("SIOCIFCONF: %m");
+      ic.ifc_len += sizeof(struct ifreq);
+      if (last_ifbuf_size < ic.ifc_len)
+	{
+	  last_ifbuf_size = ic.ifc_len;
+	  DBG("Increased ifconf buffer size to %d\n", last_ifbuf_size);
+	}
+#else
+      last_ifbuf_size *= 2;
+      DBG("Increased ifconf buffer size to %d\n", last_ifbuf_size);
+#endif
+    }
+}
+
+void
+krt_if_preconfig(struct krt_config *c)
+{
+}
+
+void
+krt_if_start(struct krt_proto *p)
+{
+  if (if_scan_sock < 0)
+    {
+      if_scan_sock = socket(PF_INET, SOCK_DGRAM, IPPROTO_IP);
+      DBG("Using socket %d for interface and route scanning\n", if_scan_sock);
+      if (if_scan_sock < 0)
+	die("Cannot create scanning socket: %m");
+    }
+}
+
+void
+krt_if_shutdown(struct krt_proto *p)
+{
+}