-/* Dealing with host names.
- Copyright (C) 1995, 1996, 1997, 2000 Free Software Foundation, Inc.
+/* Host name resolution and matching.
+ Copyright (C) 1995, 1996, 1997, 2000, 2001 Free Software Foundation, Inc.
This file is part of GNU Wget.
#include <config.h>
+#ifndef WINDOWS
+#include <netdb.h>
+#endif
+
#include <stdio.h>
#include <stdlib.h>
#ifdef HAVE_STRING_H
#else
# include <sys/socket.h>
# include <netinet/in.h>
-#ifndef __BEOS__
-# include <arpa/inet.h>
-#endif
+# ifndef __BEOS__
+# include <arpa/inet.h>
+# endif
# include <netdb.h>
#endif /* WINDOWS */
extern int errno;
#endif
-/* Mapping between all known hosts to their addresses (n.n.n.n). */
+#ifndef h_errno
+# ifndef __CYGWIN__
+extern int h_errno;
+# endif
+#endif
+
+#ifdef INET6
+int ip_default_family = AF_INET6;
+#else
+int ip_default_family = AF_INET;
+#endif
-/* #### We should map to *lists* of IP addresses. */
+/* Mapping between known hosts and to lists of their addresses. */
-struct hash_table *host_name_address_map;
+static struct hash_table *host_name_addresses_map;
+\f
+/* Lists of addresses. This should eventually be extended to handle
+ IPv6. */
-/* The following two tables are obsolete, since we no longer do host
- canonicalization. */
+struct address_list {
+ int count; /* number of adrresses */
+ ip_address *addresses; /* pointer to the string of addresses */
-/* Mapping between all known addresses (n.n.n.n) to their hosts. This
- is the inverse of host_name_address_map. These two tables share
- the strdup'ed strings. */
-struct hash_table *host_address_name_map;
+ int faulty; /* number of addresses known not to work. */
+ int refcount; /* so we know whether to free it or not. */
+};
-/* Mapping between auxilliary (slave) and master host names. */
-struct hash_table *host_slave_master_map;
+/* Get the bounds of the address list. */
-/* The same as gethostbyname, but supports internet addresses of the
- form `N.N.N.N'. On some systems gethostbyname() knows how to do
- this automatically. */
-struct hostent *
-ngethostbyname (const char *name)
+void
+address_list_get_bounds (struct address_list *al, int *start, int *end)
{
- struct hostent *hp;
- unsigned long addr;
+ *start = al->faulty;
+ *end = al->count;
+}
- addr = (unsigned long)inet_addr (name);
- if ((int)addr != -1)
- hp = gethostbyaddr ((char *)&addr, sizeof (addr), AF_INET);
- else
- hp = gethostbyname (name);
- return hp;
+/* Copy address number INDEX to IP_STORE. */
+
+void
+address_list_copy_one (struct address_list *al, int index, ip_address *ip_store)
+{
+ assert (index >= al->faulty && index < al->count);
+ memcpy (ip_store, al->addresses + index, sizeof (ip_address));
}
-/* Add host name HOST with the address ADDR_TEXT to the cache.
- Normally this means that the (HOST, ADDR_TEXT) pair will be to
- host_name_address_map and to host_address_name_map. (It is the
- caller's responsibility to make sure that HOST is not already in
- host_name_address_map.)
+/* Check whether two address lists have all their IPs in common. */
+
+int
+address_list_match_all (struct address_list *al1, struct address_list *al2)
+{
+ if (al1 == al2)
+ return 1;
+ if (al1->count != al2->count)
+ return 0;
+ return 0 == memcmp (al1->addresses, al2->addresses,
+ al1->count * sizeof (ip_address));
+}
+
+/* Mark the INDEXth element of AL as faulty, so that the next time
+ this address list is used, the faulty element will be skipped. */
+
+void
+address_list_set_faulty (struct address_list *al, int index)
+{
+ /* We assume that the address list is traversed in order, so that a
+ "faulty" attempt is always preceded with all-faulty addresses,
+ and this is how Wget uses it. */
+ assert (index == al->faulty);
+
+ ++al->faulty;
+ if (al->faulty >= al->count)
+ /* All addresses have been proven faulty. Since there's not much
+ sense in returning the user an empty address list the next
+ time, we'll rather make them all clean, so that they can be
+ retried anew. */
+ al->faulty = 0;
+}
+
+#ifdef INET6
+/**
+ * address_list_from_addrinfo
+ *
+ * This function transform an addrinfo links list in and address_list.
+ *
+ * Input:
+ * addrinfo* Linkt list of addrinfo
+ *
+ * Output:
+ * address_list* New allocated address_list
+ */
+static struct address_list *
+address_list_from_addrinfo (struct addrinfo *ai)
+{
+ struct address_list *al;
+ struct addrinfo *ai_head = ai;
+ int cnt = 0;
+ int i;
+
+ for (ai = ai_head; ai; ai = ai->ai_next)
+ if (ai->ai_family == AF_INET || ai->ai_family == AF_INET6)
+ ++cnt;
+ if (cnt == 0)
+ return NULL;
+
+ al = xmalloc (sizeof (struct address_list));
+ al->addresses = xmalloc (cnt * sizeof (ip_address));
+ al->count = cnt;
+ al->faulty = 0;
+ al->refcount = 1;
+
+ for (i = 0, ai = ai_head; ai; ai = ai->ai_next)
+ if (ai->ai_family == AF_INET6)
+ {
+ struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)ai->ai_addr;
+ memcpy (al->addresses + i, &sin6->sin6_addr, 16);
+ ++i;
+ }
+ else if (ai->ai_family == AF_INET)
+ {
+ struct sockaddr_in *sin = (struct sockaddr_in *)ai->ai_addr;
+ map_ipv4_to_ip ((ip4_address *)&sin->sin_addr, al->addresses + i);
+ ++i;
+ }
+ assert (i == cnt);
+ return al;
+}
+#else
+/* Create an address_list out of a NULL-terminated list of addresses,
+ as returned by gethostbyname. */
+static struct address_list *
+address_list_new (char **h_addr_list)
+{
+ int count = 0, i;
+
+ struct address_list *al = xmalloc (sizeof (struct address_list));
- If the ADDR_TEXT has already been seen and belongs to another host,
- HOST will be added to host_slave_master_map instead. */
+ while (h_addr_list[count])
+ ++count;
+ assert (count > 0);
+ al->count = count;
+ al->faulty = 0;
+ al->addresses = xmalloc (count * sizeof (ip_address));
+ al->refcount = 1;
+
+ for (i = 0; i < count; i++)
+ map_ipv4_to_ip ((ip4_address *)h_addr_list[i], al->addresses + i);
+
+ return al;
+}
+#endif
+
+/* Like address_list_new, but initialized with only one address. */
+
+static struct address_list *
+address_list_new_one (ip_address *addr)
+{
+ struct address_list *al = xmalloc (sizeof (struct address_list));
+ al->count = 1;
+ al->faulty = 0;
+ al->addresses = xmalloc (sizeof (ip_address));
+ al->refcount = 1;
+ memcpy (al->addresses, addr, sizeof (ip_address));
+
+ return al;
+}
static void
-add_host_to_cache (const char *host, const char *addr_text)
+address_list_delete (struct address_list *al)
{
- char *canonical_name = hash_table_get (host_address_name_map, addr_text);
- if (canonical_name)
+ xfree (al->addresses);
+ xfree (al);
+}
+
+void
+address_list_release (struct address_list *al)
+{
+ --al->refcount;
+ DEBUGP (("Releasing %p (new refcount %d).\n", al, al->refcount));
+ if (al->refcount <= 0)
{
- DEBUGP (("Mapping %s to %s in host_slave_master_map.\n",
- host, canonical_name));
- /* We've already dealt with that host under another name. */
- hash_table_put (host_slave_master_map,
- xstrdup_lower (host),
- xstrdup_lower (canonical_name));
+ DEBUGP (("Deleting unused %p.\n", al));
+ address_list_delete (al);
}
- else
+}
+\f
+/**
+ * wget_sockaddr_set_address
+ *
+ * This function takes an wget_sockaddr and fill in the protocol type,
+ * the port number and the address, there NULL in address means wildcard.
+ * Unsuported adress family will abort the whole programm.
+ *
+ * Input:
+ * wget_sockaddr* The space to be filled
+ * int The wished protocol
+ * unsigned short The port
+ * const ip_address The Binary IP adress
+ *
+ * Return:
+ * - Only modify 1. param
+ */
+void
+wget_sockaddr_set_address (wget_sockaddr *sa,
+ int ip_family, unsigned short port, ip_address *addr)
+{
+ if (ip_family == AF_INET)
+ {
+ sa->sin.sin_family = ip_family;
+ sa->sin.sin_port = htons (port);
+ if (addr == NULL)
+ memset (&sa->sin.sin_addr, 0, sizeof(ip4_address));
+ else
+ {
+ ip4_address addr4;
+ if (!map_ip_to_ipv4 (addr, &addr4))
+ /* should the callers have prevented this? */
+ abort ();
+ memcpy (&sa->sin.sin_addr, &addr4, sizeof(ip4_address));
+ }
+ return;
+ }
+#ifdef INET6
+ if (ip_family == AF_INET6)
{
- /* This is really the first time we're dealing with that host. */
- char *h_copy = xstrdup_lower (host);
- char *a_copy = xstrdup (addr_text);
- DEBUGP (("Caching %s <-> %s\n", h_copy, a_copy));
- hash_table_put (host_name_address_map, h_copy, a_copy);
- hash_table_put (host_address_name_map, a_copy, h_copy);
+ sa->sin6.sin6_family = ip_family;
+ sa->sin6.sin6_port = htons (port);
+ if (addr == NULL)
+ memset (&sa->sin6.sin6_addr, 0 , 16);
+ else
+ memcpy (&sa->sin6.sin6_addr, addr, 16);
+ return;
}
+#endif
+ abort();
}
-/* Store the address of HOSTNAME, internet-style (four octets in
- network order), to WHERE. First try to get the address from the
- cache; if it is not available, call the DNS functions and update
- the cache.
-
- Return 1 on successful finding of the hostname, 0 otherwise. */
-int
-store_hostaddress (unsigned char *where, const char *hostname)
+/**
+ * wget_sockaddr_set_port
+ *
+ * This funtion only fill the port of the socket information.
+ * If the protocol is not supported nothing is done.
+ * Unsuported adress family will abort the whole programm.
+ *
+ * Require:
+ * that the IP-Protocol already is set.
+ *
+ * Input:
+ * wget_sockaddr* The space there port should be entered
+ * unsigned int The port that should be entered in host order
+ *
+ * Return:
+ * - Only modify 1. param
+ */
+void
+wget_sockaddr_set_port (wget_sockaddr *sa, unsigned short port)
{
- unsigned long addr;
- char *addr_text;
- char *canonical_name;
- struct hostent *hptr;
- struct in_addr in;
- char *inet_s;
-
- /* If the address is of the form d.d.d.d, there will be no trouble
- with it. */
- addr = (unsigned long)inet_addr (hostname);
- /* If we have the numeric address, just store it. */
- if ((int)addr != -1)
+ if (sa->sa.sa_family == AF_INET)
{
- /* ADDR is defined to be in network byte order, meaning the code
- works on little and big endian 32-bit architectures without
- change. On big endian 64-bit architectures we need to be
- careful to copy the correct four bytes. */
- int offset;
- have_addr:
-#ifdef WORDS_BIGENDIAN
- offset = sizeof (unsigned long) - 4;
+ sa->sin.sin_port = htons (port);
+ return;
+ }
+#ifdef INET6
+ if (sa->sa.sa_family == AF_INET6)
+ {
+ sa->sin6.sin6_port = htons (port);
+ return;
+ }
+#endif
+ abort();
+}
+
+/**
+ * wget_sockaddr_get_addr
+ *
+ * This function return the adress from an sockaddr as byte string.
+ * Unsuported adress family will abort the whole programm.
+ *
+ * Require:
+ * that the IP-Protocol already is set.
+ *
+ * Input:
+ * wget_sockaddr* Socket Information
+ *
+ * Output:
+ * unsigned char * IP address as byte string.
+ */
+void *
+wget_sockaddr_get_addr (wget_sockaddr *sa)
+{
+ if (sa->sa.sa_family == AF_INET)
+ return &sa->sin.sin_addr;
+#ifdef INET6
+ if (sa->sa.sa_family == AF_INET6)
+ return &sa->sin6.sin6_addr;
+#endif
+ abort();
+ /* unreached */
+ return NULL;
+}
+
+/**
+ * wget_sockaddr_get_port
+ *
+ * This function only return the port from the input structure
+ * Unsuported adress family will abort the whole programm.
+ *
+ * Require:
+ * that the IP-Protocol already is set.
+ *
+ * Input:
+ * wget_sockaddr* Information where to get the port
+ *
+ * Output:
+ * unsigned short Port Number in host order.
+ */
+unsigned short
+wget_sockaddr_get_port (const wget_sockaddr *sa)
+{
+ if (sa->sa.sa_family == AF_INET)
+ return htons (sa->sin.sin_port);
+#ifdef INET6
+ if (sa->sa.sa_family == AF_INET6)
+ return htons (sa->sin6.sin6_port);
+#endif
+ abort();
+ /* do not complain about return nothing */
+ return -1;
+}
+
+/**
+ * sockaddr_len
+ *
+ * This function return the length of the sockaddr corresponding to
+ * the acutall prefered protocol for (bind, connect etc...)
+ * Unsuported adress family will abort the whole programm.
+ *
+ * Require:
+ * that the IP-Protocol already is set.
+ *
+ * Input:
+ * - Public IP-Family Information
+ *
+ * Output:
+ * int structure length for socket options
+ */
+int
+sockaddr_len ()
+{
+ if (ip_default_family == AF_INET)
+ return sizeof (struct sockaddr_in);
+#ifdef INET6
+ if (ip_default_family == AF_INET6)
+ return sizeof (struct sockaddr_in6);
+#endif
+ abort();
+ /* do not complain about return nothing */
+ return 0;
+}
+
+/**
+ * Map an IPv4 adress to the internal adress format.
+ */
+void
+map_ipv4_to_ip (ip4_address *ipv4, ip_address *ip)
+{
+#ifdef INET6
+ static unsigned char ipv64[12] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0xff, 0xff};
+ memcpy ((char *)ip + 12, ipv4 , 4);
+ memcpy ((char *)ip + 0, ipv64, 12);
+#else
+ if ((char *)ip != (char *)ipv4)
+ memcpy (ip, ipv4, 4);
+#endif
+}
+
+/* Detect whether an IP adress represents an IPv4 address and, if so,
+ copy it to IPV4. 0 is returned on failure.
+ This operation always succeeds when Wget is compiled without IPv6.
+ If IPV4 is NULL, don't copy, just detect. */
+
+int
+map_ip_to_ipv4 (ip_address *ip, ip4_address *ipv4)
+{
+#ifdef INET6
+ static unsigned char ipv64[12] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0xff, 0xff};
+ if (0 != memcmp (ip, ipv64, 12))
+ return 0;
+ if (ipv4)
+ memcpy (ipv4, (char *)ip + 12, 4);
#else
- offset = 0;
+ if (ipv4)
+ memcpy (ipv4, (char *)ip, 4);
#endif
- memcpy (where, (char *)&addr + offset, 4);
- return 1;
+ return 1;
+}
+\f
+/* Pretty-print ADDR. When compiled without IPv6, this is the same as
+ inet_ntoa. With IPv6, it either prints an IPv6 address or an IPv4
+ address. */
+
+char *
+pretty_print_address (ip_address *addr)
+{
+#ifdef INET6
+ ip4_address addr4;
+ static char buf[128];
+
+ if (map_ip_to_ipv4 (addr, &addr4))
+ return inet_ntoa (*(struct in_addr *)&addr4);
+
+ if (!inet_ntop (AF_INET6, addr, buf, sizeof (buf)))
+ return "<unknown>";
+ return buf;
+#endif
+ return inet_ntoa (*(struct in_addr *)addr);
+}
+
+/* Add host name HOST with the address ADDR_TEXT to the cache.
+ ADDR_LIST is a NULL-terminated list of addresses, as in struct
+ hostent. */
+
+static void
+cache_host_lookup (const char *host, struct address_list *al)
+{
+ if (!host_name_addresses_map)
+ host_name_addresses_map = make_nocase_string_hash_table (0);
+
+ ++al->refcount;
+ hash_table_put (host_name_addresses_map, xstrdup_lower (host), al);
+
+#ifdef DEBUG
+ if (opt.debug)
+ {
+ int i;
+ debug_logprintf ("Caching %s =>", host);
+ for (i = 0; i < al->count; i++)
+ debug_logprintf (" %s", pretty_print_address (al->addresses + i));
+ debug_logprintf ("\n");
}
+#endif
+}
+
+struct address_list *
+lookup_host (const char *host, int silent)
+{
+ struct address_list *al = NULL;
+ unsigned long addr_ipv4; /* #### use a 32-bit type here. */
+ ip_address addr;
+
+ /* First, try to check whether the address is already a numeric
+ address. */
- /* By now we know that the address is not of the form d.d.d.d. Try
- to find it in our cache of host addresses. */
- addr_text = hash_table_get (host_name_address_map, hostname);
- if (addr_text)
+#ifdef INET6
+ if (inet_pton (AF_INET6, host, &addr) > 0)
+ return address_list_new_one (&addr);
+#endif
+
+ addr_ipv4 = (unsigned long)inet_addr (host);
+ if ((int)addr_ipv4 != -1)
{
- DEBUGP (("Found %s in host_name_address_map: %s\n",
- hostname, addr_text));
- addr = (unsigned long)inet_addr (addr_text);
- goto have_addr;
+ /* ADDR is defined to be in network byte order, which is what
+ this returns, so we can just copy it to STORE_IP. However,
+ on big endian 64-bit architectures the value will be stored
+ in the *last*, not first four bytes. OFFSET makes sure that
+ we copy the correct four bytes. */
+ int offset = 0;
+#ifdef WORDS_BIGENDIAN
+ offset = sizeof (unsigned long) - sizeof (ip4_address);
+#endif
+ map_ipv4_to_ip ((ip4_address *)((char *)&addr_ipv4 + offset), &addr);
+ return address_list_new_one (&addr);
}
- /* Maybe this host is known to us under another name. If so, we'll
- find it in host_slave_master_map, and use the master name to find
- its address in host_name_address_map. */
- canonical_name = hash_table_get (host_slave_master_map, hostname);
- if (canonical_name)
+ if (host_name_addresses_map)
{
- addr_text = hash_table_get (host_name_address_map, canonical_name);
- assert (addr_text != NULL);
- DEBUGP (("Found %s as slave of %s -> %s\n",
- hostname, canonical_name, addr_text));
- addr = (unsigned long)inet_addr (addr_text);
- goto have_addr;
+ al = hash_table_get (host_name_addresses_map, host);
+
+ if (al)
+ {
+ DEBUGP (("Found %s in host_name_addresses_map (%p)\n", host, al));
+ ++al->refcount;
+ return al;
+ }
}
- /* Since all else has failed, let's try gethostbyname(). Note that
- we use gethostbyname() rather than ngethostbyname(), because we
- already know that the address is not numerical. */
- hptr = gethostbyname (hostname);
- if (!hptr)
- return 0;
- /* Copy the address of the host to socket description. */
- memcpy (where, hptr->h_addr_list[0], hptr->h_length);
- assert (hptr->h_length == 4);
-
- /* Now that we've gone through the truoble of calling
- gethostbyname(), we can store this valuable information to the
- cache. First, we have to look for it by address to know if it's
- already in the cache by another name. */
- /* Originally, we copied to in.s_addr, but it appears to be missing
- on some systems. */
- memcpy (&in, *hptr->h_addr_list, sizeof (in));
- inet_s = inet_ntoa (in);
- add_host_to_cache (hostname, inet_s);
- return 1;
-}
+ if (!silent)
+ logprintf (LOG_VERBOSE, _("Resolving %s... "), host);
+
+ /* Host name lookup goes on below. #### We should implement
+ getaddrinfo_with_timeout and gethostbyname_with_timeout the same
+ way connect.c implements connect_with_timeout. */
+
+#ifdef INET6
+ {
+ struct addrinfo hints, *ai;
+ int err;
+
+ memset (&hints, 0, sizeof (hints));
+ if (ip_default_family == AF_INET)
+ hints.ai_family = AF_INET;
+ else
+ hints.ai_family = PF_UNSPEC;
+ hints.ai_socktype = SOCK_STREAM;
+ err = getaddrinfo (host, NULL, &hints, &ai);
+
+ if (err != 0 || ai == NULL)
+ {
+ if (!silent)
+ logprintf (LOG_VERBOSE, _("failed: %s.\n"), gai_strerror (err));
+ return NULL;
+ }
+ al = address_list_from_addrinfo (ai);
+ freeaddrinfo (ai);
+ }
+#else
+ {
+ struct hostent *hptr = gethostbyname (host);
+ if (!hptr)
+ {
+ if (!silent)
+ logprintf (LOG_VERBOSE, _("failed: %s.\n"), herrmsg (h_errno));
+ return NULL;
+ }
+ /* Do all systems have h_addr_list, or is it a newer thing? If
+ the latter, use address_list_new_one. */
+ al = address_list_new (hptr->h_addr_list);
+ }
+#endif
+
+ if (!silent)
+ logprintf (LOG_VERBOSE, _("done.\n"));
+
+ /* Cache the lookup information. */
+ cache_host_lookup (host, al);
+ return al;
+}
+\f
/* Determine whether a URL is acceptable to be followed, according to
a list of domains to accept. */
int
return _("Unknown error");
}
-void
-host_cleanup (void)
+static int
+host_cleanup_mapper (void *key, void *value, void *arg_ignored)
{
- /* host_name_address_map and host_address_name_map share the
- strings. Because of that, calling free_keys_and_values once
- suffices for both. */
- free_keys_and_values (host_name_address_map);
- hash_table_destroy (host_name_address_map);
- hash_table_destroy (host_address_name_map);
- free_keys_and_values (host_slave_master_map);
- hash_table_destroy (host_slave_master_map);
+ struct address_list *al;
+
+ xfree (key); /* host */
+
+ al = (struct address_list *)value;
+ assert (al->refcount == 1);
+ address_list_delete (al);
+
+ return 0;
}
void
-host_init (void)
+host_cleanup (void)
{
- host_name_address_map = make_string_hash_table (0);
- host_address_name_map = make_string_hash_table (0);
- host_slave_master_map = make_string_hash_table (0);
+ if (host_name_addresses_map)
+ {
+ hash_table_map (host_name_addresses_map, host_cleanup_mapper, NULL);
+ hash_table_destroy (host_name_addresses_map);
+ host_name_addresses_map = NULL;
+ }
}