IPv4/IPv6 integrated socket code.

2 files changed, 154 insertions, 208 deletions

diff --git a/sysdep/linux/netlink.c b/sysdep/linux/netlink.c
index 7063e2ca..a0f85186 100644
--- a/sysdep/linux/netlink.c
+++ b/sysdep/linux/netlink.c
@@ -104,9 +104,9 @@ nl_request_dump(int cmd)
   req.nh.nlmsg_type = cmd;
   req.nh.nlmsg_len = sizeof(req);
   req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP;
-  /* Is it important which PF_* is used for link-level interface scan?
-     It seems that some information is available only when PF_INET is used. */
-  req.g.rtgen_family = (cmd == RTM_GETLINK) ? PF_INET : BIRD_PF;
+  /* Is it important which AF_* is used for link-level interface scan?
+     It seems that some information is available only when AF_INET is used. */
+  req.g.rtgen_family = (cmd == RTM_GETLINK) ? AF_INET : BIRD_AF;
   nl_send(&nl_scan, &req.nh);
 }
 
@@ -1069,7 +1069,7 @@ nl_open_async(void)
   sk->type = SK_MAGIC;
   sk->rx_hook = nl_async_hook;
   sk->fd = fd;
-  if (sk_open(sk))
+  if (sk_open(sk) < 0)
     bug("Netlink: sk_open failed");
 }
 
diff --git a/sysdep/linux/sysio.h b/sysdep/linux/sysio.h
index dc807392..5fd75c90 100644
--- a/sysdep/linux/sysio.h
+++ b/sysdep/linux/sysio.h
@@ -6,232 +6,151 @@
  *	Can be freely distributed and used under the terms of the GNU GPL.
  */
 
-#include <net/if.h>
-
-#ifdef IPV6
-
-#ifndef IPV6_UNICAST_HOPS
-/* Needed on glibc 2.0 systems */
-#include <linux/in6.h>
-#define CONFIG_IPV6_GLIBC_20
-#endif
-
-static inline void
-set_inaddr(struct in6_addr *ia, ip_addr a)
-{
-  ipa_hton(a);
-  memcpy(ia, &a, sizeof(a));
-}
-
-static inline void
-get_inaddr(ip_addr *a, struct in6_addr *ia)
-{
-  memcpy(a, ia, sizeof(*a));
-  ipa_ntoh(*a);
-}
-
-#else
-
-static inline void
-set_inaddr(struct in_addr *ia, ip_addr a)
-{
-  ipa_hton(a);
-  memcpy(&ia->s_addr, &a, sizeof(a));
-}
-
-static inline void
-get_inaddr(ip_addr *a, struct in_addr *ia)
-{
-  memcpy(a, &ia->s_addr, sizeof(*a));
-  ipa_ntoh(*a);
-}
-
 
 #ifndef HAVE_STRUCT_IP_MREQN
 /* Several versions of glibc don't define this structure, so we have to do it ourselves */
 struct ip_mreqn
 {
-	struct in_addr	imr_multiaddr;		/* IP multicast address of group */
-	struct in_addr	imr_address;		/* local IP address of interface */
-	int		imr_ifindex;		/* Interface index */
+  struct in_addr imr_multiaddr;			/* IP multicast address of group */
+  struct in_addr imr_address;			/* local IP address of interface */
+  int		 imr_ifindex;			/* Interface index */
 };
 #endif
 
+#ifndef IP_MINTTL
+#define IP_MINTTL 21
+#endif
 
-static inline void fill_mreqn(struct ip_mreqn *m, ip_addr maddr, struct iface *ifa)
-{
-  bzero(m, sizeof(*m));
-  m->imr_ifindex = ifa->index;
-  set_inaddr(&m->imr_multiaddr, maddr);
-}
+#ifndef IPV6_TCLASS
+#define IPV6_TCLASS 67
+#endif
 
-static inline char *
-sysio_setup_multicast(sock *s)
-{
-  struct ip_mreqn m;
-  int zero = 0;
+#ifndef IPV6_MINHOPCOUNT
+#define IPV6_MINHOPCOUNT 73
+#endif
 
-  if (setsockopt(s->fd, SOL_IP, IP_MULTICAST_LOOP, &zero, sizeof(zero)) < 0)
-    return "IP_MULTICAST_LOOP";
 
-  if (setsockopt(s->fd, SOL_IP, IP_MULTICAST_TTL, &s->ttl, sizeof(s->ttl)) < 0)
-    return "IP_MULTICAST_TTL";
+#ifndef TCP_MD5SIG
 
-  /* This defines where should we send _outgoing_ multicasts */
-  fill_mreqn(&m, IPA_NONE, s->iface);
-  if (setsockopt(s->fd, SOL_IP, IP_MULTICAST_IF, &m, sizeof(m)) < 0)
-    return "IP_MULTICAST_IF";
+#define TCP_MD5SIG  14
+#define TCP_MD5SIG_MAXKEYLEN 80
 
-  return NULL;
-}
+struct tcp_md5sig {
+  struct  sockaddr_storage tcpm_addr;             /* address associated */
+  u16   __tcpm_pad1;                              /* zero */
+  u16   tcpm_keylen;                              /* key length */
+  u32   __tcpm_pad2;                              /* zero */
+  u8    tcpm_key[TCP_MD5SIG_MAXKEYLEN];           /* key (binary) */
+};
 
-static inline char *
-sysio_join_group(sock *s, ip_addr maddr)
-{
-  struct ip_mreqn m;
+#endif
 
-  /* And this one sets interface for _receiving_ multicasts from */
-  fill_mreqn(&m, maddr, s->iface);
-  if (setsockopt(s->fd, SOL_IP, IP_ADD_MEMBERSHIP, &m, sizeof(m)) < 0)
-    return "IP_ADD_MEMBERSHIP";
 
-  return NULL;
-}
+/* Linux does not care if sa_len is larger than needed */
+#define SA_LEN(x) sizeof(sockaddr)
 
-static inline char *
-sysio_leave_group(sock *s, ip_addr maddr)
-{
-  struct ip_mreqn m;
 
-  /* And this one sets interface for _receiving_ multicasts from */
-  fill_mreqn(&m, maddr, s->iface);
-  if (setsockopt(s->fd, SOL_IP, IP_DROP_MEMBERSHIP, &m, sizeof(m)) < 0)
-    return "IP_DROP_MEMBERSHIP";
+/*
+ *	Linux IPv4 multicast syscalls
+ */
 
-  return NULL;
-}
+#define INIT_MREQ4(maddr,ifa) \
+  { .imr_multiaddr = ipa_to_in4(maddr), .imr_ifindex = ifa->index }
 
-#endif
+static inline int
+sk_setup_multicast4(sock *s)
+{
+  struct ip_mreqn mr = { .imr_ifindex = s->iface->index };
+  int ttl = s->ttl;
+  int n = 0;
 
+  /* This defines where should we send _outgoing_ multicasts */
+  if (setsockopt(s->fd, SOL_IP, IP_MULTICAST_IF, &mr, sizeof(mr)) < 0)
+    ERR("IP_MULTICAST_IF");
 
-/* For the case that we have older libc headers */
-/* Copied from Linux kernel file include/linux/tcp.h */
+  if (setsockopt(s->fd, SOL_IP, IP_MULTICAST_TTL, &ttl, sizeof(ttl)) < 0)
+    ERR("IP_MULTICAST_TTL");
 
-#ifndef TCP_MD5SIG
+  if (setsockopt(s->fd, SOL_IP, IP_MULTICAST_LOOP, &n, sizeof(n)) < 0)
+    ERR("IP_MULTICAST_LOOP");
 
-#define TCP_MD5SIG  14
-#define TCP_MD5SIG_MAXKEYLEN 80
+  return 0;
+}
 
-#include <linux/types.h>
+static inline int
+sk_join_group4(sock *s, ip_addr maddr)
+{
+  struct ip_mreqn mr = INIT_MREQ4(maddr, s->iface);
 
-struct tcp_md5sig {
-  struct  sockaddr_storage tcpm_addr;             /* address associated */
-  __u16   __tcpm_pad1;                            /* zero */
-  __u16   tcpm_keylen;                            /* key length */
-  __u32   __tcpm_pad2;                            /* zero */
-  __u8    tcpm_key[TCP_MD5SIG_MAXKEYLEN];         /* key (binary) */
-};
+  if (setsockopt(s->fd, SOL_IP, IP_ADD_MEMBERSHIP, &mr, sizeof(mr)) < 0)
+    ERR("IP_ADD_MEMBERSHIP");
 
-#endif
+  return 0;
+}
 
-static int
-sk_set_md5_auth_int(sock *s, sockaddr *sa, char *passwd)
+static inline int
+sk_leave_group4(sock *s, ip_addr maddr)
 {
-  struct tcp_md5sig md5;
+  struct ip_mreqn mr = INIT_MREQ4(maddr, s->iface);
 
-  memset(&md5, 0, sizeof(md5));
-  memcpy(&md5.tcpm_addr, (struct sockaddr *) sa, sizeof(*sa));
+  if (setsockopt(s->fd, SOL_IP, IP_DROP_MEMBERSHIP, &mr, sizeof(mr)) < 0)
+    ERR("IP_DROP_MEMBERSHIP");
 
-  if (passwd)
-    {
-      int len = strlen(passwd);
-
-      if (len > TCP_MD5SIG_MAXKEYLEN)
-	{
-	  log(L_ERR "MD5 password too long");
-	  return -1;
-	}
-
-      md5.tcpm_keylen = len;
-      memcpy(&md5.tcpm_key, passwd, len);
-    }
-
-  int rv = setsockopt(s->fd, SOL_TCP, TCP_MD5SIG, &md5, sizeof(md5));
-
-  if (rv < 0) 
-    {
-      if (errno == ENOPROTOOPT)
-	log(L_ERR "Kernel does not support TCP MD5 signatures");
-      else
-	log(L_ERR "sk_set_md5_auth_int: setsockopt: %m");
-    }
-
-  return rv;
+  return 0;
 }
 
 
-#ifndef IPV6
+/*
+ *	Linux IPv4 packet control messages
+ */
 
-/* RX/TX packet info handling for IPv4 */
 /* Mostly similar to standardized IPv6 code */
 
-#define CMSG_RX_SPACE (CMSG_SPACE(sizeof(struct in_pktinfo)) + CMSG_SPACE(sizeof(int)))
-#define CMSG_TX_SPACE CMSG_SPACE(sizeof(struct in_pktinfo))
+#define CMSG4_SPACE_PKTINFO CMSG_SPACE(sizeof(struct in_pktinfo))
+#define CMSG4_SPACE_TTL CMSG_SPACE(sizeof(int))
 
-static char *
-sysio_register_cmsgs(sock *s)
+static inline int
+sk_request_cmsg4_pktinfo(sock *s)
 {
-  int ok = 1;
+  int y = 1;
 
-  if ((s->flags & SKF_LADDR_RX) &&
-      (setsockopt(s->fd, SOL_IP, IP_PKTINFO, &ok, sizeof(ok)) < 0))
-    return "IP_PKTINFO";
+  if (setsockopt(s->fd, SOL_IP, IP_PKTINFO, &y, sizeof(y)) < 0)
+    ERR("IP_PKTINFO");
 
-  if ((s->flags & SKF_TTL_RX) &&
-      (setsockopt(s->fd, SOL_IP, IP_RECVTTL, &ok, sizeof(ok)) < 0))
-    return "IP_RECVTTL";
-
-  return NULL;
+  return 0;
 }
 
-static void
-sysio_process_rx_cmsgs(sock *s, struct msghdr *msg)
+static inline int
+sk_request_cmsg4_ttl(sock *s)
 {
-  struct cmsghdr *cm;
-  struct in_pktinfo *pi = NULL;
-  int *ttl = NULL;
+  int y = 1;
 
-  for (cm = CMSG_FIRSTHDR(msg); cm != NULL; cm = CMSG_NXTHDR(msg, cm))
-  {
-    if (cm->cmsg_level == SOL_IP && cm->cmsg_type == IP_PKTINFO)
-      pi = (struct in_pktinfo *) CMSG_DATA(cm);
+  if (setsockopt(s->fd, SOL_IP, IP_RECVTTL, &y, sizeof(y)) < 0)
+    ERR("IP_RECVTTL");
 
-    if (cm->cmsg_level == SOL_IP && cm->cmsg_type == IP_TTL)
-      ttl = (int *) CMSG_DATA(cm);
-  }
+  return 0;
+}
 
-  if (s->flags & SKF_LADDR_RX)
+static inline void
+sk_process_cmsg4_pktinfo(sock *s, struct cmsghdr *cm)
+{
+  if (cm->cmsg_type == IP_PKTINFO)
   {
-    if (pi)
-    {
-      get_inaddr(&s->laddr, &pi->ipi_addr);
-      s->lifindex = pi->ipi_ifindex;
-    }
-    else
-    {
-      s->laddr = IPA_NONE;
-      s->lifindex = 0;
-    }
+    struct in_pktinfo *pi = (struct in_pktinfo *) CMSG_DATA(cm);
+    s->laddr = ipa_from_in4(pi->ipi_addr);
+    s->lifindex = pi->ipi_ifindex;
   }
+}
 
-  if (s->flags & SKF_TTL_RX)
-    s->ttl = ttl ? *ttl : -1;
-
-  return;
+static inline void
+sk_process_cmsg4_ttl(sock *s, struct cmsghdr *cm)
+{
+  if (cm->cmsg_type == IP_TTL)
+    s->rcv_ttl = * (int *) CMSG_DATA(cm);
 }
 
-static void
-sysio_prepare_tx_cmsgs(sock *s, struct msghdr *msg, void *cbuf, size_t cbuflen)
+static inline void
+sk_prepare_cmsgs4(sock *s, struct msghdr *msg, void *cbuf, size_t cbuflen)
 {
   struct cmsghdr *cm;
   struct in_pktinfo *pi;
@@ -246,78 +165,105 @@ sysio_prepare_tx_cmsgs(sock *s, struct msghdr *msg, void *cbuf, size_t cbuflen)
 
   pi = (struct in_pktinfo *) CMSG_DATA(cm);
   pi->ipi_ifindex = s->iface ? s->iface->index : 0;
-  set_inaddr(&pi->ipi_spec_dst, s->saddr);
-  set_inaddr(&pi->ipi_addr, IPA_NONE);
+  pi->ipi_spec_dst = ipa_to_in4(s->saddr);
+  pi->ipi_addr = ipa_to_in4(IPA_NONE);
 
   msg->msg_controllen = cm->cmsg_len;
 }
 
 
-#endif
+/*
+ *	Miscellaneous Linux socket syscalls
+ */
 
+int
+sk_set_md5_auth(sock *s, ip_addr a, struct iface *ifa, char *passwd)
+{
+  struct tcp_md5sig md5;
 
-#ifndef IP_MINTTL
-#define IP_MINTTL 21
-#endif
+  memset(&md5, 0, sizeof(md5));
+  sockaddr_fill((sockaddr *) &md5.tcpm_addr, s->af, a, ifa, 0);
 
-#ifndef IPV6_MINHOPCOUNT
-#define IPV6_MINHOPCOUNT 73
-#endif
+  if (passwd)
+  {
+    int len = strlen(passwd);
+
+    if (len > TCP_MD5SIG_MAXKEYLEN)
+      ERR_MSG("MD5 password too long");
 
+    md5.tcpm_keylen = len;
+    memcpy(&md5.tcpm_key, passwd, len);
+  }
+
+  if (setsockopt(s->fd, SOL_TCP, TCP_MD5SIG, &md5, sizeof(md5)) < 0)
+  {
+    if (errno == ENOPROTOOPT)
+      ERR_MSG("Kernel does not support TCP MD5 signatures");
+    else
+      ERR("TCP_MD5SIG");
+  }
 
-#ifndef IPV6
+  return 0;
+}
 
-static int
+static inline int
 sk_set_min_ttl4(sock *s, int ttl)
 {
   if (setsockopt(s->fd, SOL_IP, IP_MINTTL, &ttl, sizeof(ttl)) < 0)
   {
     if (errno == ENOPROTOOPT)
-      log(L_ERR "Kernel does not support IPv4 TTL security");
+      ERR_MSG("Kernel does not support IPv4 TTL security");
     else
-      log(L_ERR "sk_set_min_ttl4: setsockopt: %m");
-
-    return -1;
+      ERR("IP_MINTTL");
   }
 
   return 0;
 }
 
-#else
-
-static int
+static inline int
 sk_set_min_ttl6(sock *s, int ttl)
 {
   if (setsockopt(s->fd, SOL_IPV6, IPV6_MINHOPCOUNT, &ttl, sizeof(ttl)) < 0)
   {
     if (errno == ENOPROTOOPT)
-      log(L_ERR "Kernel does not support IPv6 TTL security");
+      ERR_MSG("Kernel does not support IPv6 TTL security");
     else
-      log(L_ERR "sk_set_min_ttl6: setsockopt: %m");
-
-    return -1;
+      ERR("IPV6_MINHOPCOUNT");
   }
 
   return 0;
 }
 
-#endif
+static inline int
+sk_disable_mtu_disc4(sock *s)
+{
+  int dont = IP_PMTUDISC_DONT;
 
+  if (setsockopt(s->fd, SOL_IP, IP_MTU_DISCOVER, &dont, sizeof(dont)) < 0)
+    ERR("IP_MTU_DISCOVER");
 
-#ifndef IPV6_TCLASS
-#define IPV6_TCLASS 67
-#endif
+  return 0;
+}
+
+static inline int
+sk_disable_mtu_disc6(sock *s)
+{
+  int dont = IPV6_PMTUDISC_DONT;
+
+  if (setsockopt(s->fd, SOL_IPV6, IPV6_MTU_DISCOVER, &dont, sizeof(dont)) < 0)
+    ERR("IPV6_MTU_DISCOVER");
+
+  return 0;
+}
 
 int sk_priority_control = 7;
 
-static int
+static inline int
 sk_set_priority(sock *s, int prio)
 {
   if (setsockopt(s->fd, SOL_SOCKET, SO_PRIORITY, &prio, sizeof(prio)) < 0)
-  {
-    log(L_WARN "sk_set_priority: setsockopt: %m");
-    return -1;
-  }
+    ERR("SO_PRIORITY");
 
   return 0;
 }
+