1 /* Host name resolution and matching.
2 Copyright (C) 1995, 1996, 1997, 2000, 2001 Free Software Foundation, Inc.
4 This file is part of GNU Wget.
6 GNU Wget is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or
9 (at your option) any later version.
11 GNU Wget is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with Wget; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
20 In addition, as a special exception, the Free Software Foundation
21 gives permission to link the code of its release of Wget with the
22 OpenSSL project's "OpenSSL" library (or with modified versions of it
23 that use the same license as the "OpenSSL" library), and distribute
24 the linked executables. You must obey the GNU General Public License
25 in all respects for all of the code used other than "OpenSSL". If you
26 modify this file, you may extend this exception to your version of the
27 file, but you are not obligated to do so. If you do not wish to do
28 so, delete this exception statement from your version. */
44 #include <sys/types.h>
48 # define SET_H_ERRNO(err) WSASetLastError (err)
50 # include <sys/socket.h>
51 # include <netinet/in.h>
53 # include <arpa/inet.h>
56 # define SET_H_ERRNO(err) ((void)(h_errno = (err)))
60 #define NO_ADDRESS NO_DATA
63 #ifdef HAVE_SYS_UTSNAME_H
64 # include <sys/utsname.h>
85 int ip_default_family = AF_UNSPEC;
87 int ip_default_family = AF_INET;
90 /* Mapping between known hosts and to lists of their addresses. */
92 static struct hash_table *host_name_addresses_map;
94 /* Lists of addresses. This should eventually be extended to handle
98 int count; /* number of adrresses */
99 ip_address *addresses; /* pointer to the string of addresses */
101 int faulty; /* number of addresses known not to work. */
102 int refcount; /* so we know whether to free it or not. */
105 /* Get the bounds of the address list. */
108 address_list_get_bounds (const struct address_list *al, int *start, int *end)
114 /* Copy address number INDEX to IP_STORE. */
117 address_list_copy_one (const struct address_list *al, int index,
118 ip_address *ip_store)
120 assert (index >= al->faulty && index < al->count);
121 memcpy (ip_store, al->addresses + index, sizeof (ip_address));
124 /* Check whether two address lists have all their IPs in common. */
127 address_list_match_all (const struct address_list *al1,
128 const struct address_list *al2)
135 if (al1->count != al2->count)
138 /* For the comparison to be complete, we'd need to sort the IP
139 addresses first. But that's not necessary because this is only
140 used as an optimization. */
143 /* In the non-IPv6 case, there is only one address type, so we can
144 compare the whole array with memcmp. */
145 return 0 == memcmp (al1->addresses, al2->addresses,
146 al1->count * sizeof (ip_address));
147 #else /* ENABLE_IPV6 */
148 for (i = 0; i < al1->count; ++i)
150 const ip_address *ip1 = &al1->addresses[i];
151 const ip_address *ip2 = &al2->addresses[i];
153 if (ip1->type != ip2->type)
159 if (ADDRESS_IPV4_IN_ADDR (ip1).s_addr
160 != ADDRESS_IPV4_IN_ADDR (ip2).s_addr)
164 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
165 if (ADDRESS_IPV6_SCOPE (ip1) != ADDRESS_IPV6_SCOPE (ip2))
167 #endif /* HAVE_SOCKADDR_IN6_SCOPE_ID */
168 if (!IN6_ARE_ADDR_EQUAL (&ADDRESS_IPV6_IN6_ADDR (ip1),
169 &ADDRESS_IPV6_IN6_ADDR (ip2)))
177 #endif /* ENABLE_IPV6 */
180 /* Mark the INDEXth element of AL as faulty, so that the next time
181 this address list is used, the faulty element will be skipped. */
184 address_list_set_faulty (struct address_list *al, int index)
186 /* We assume that the address list is traversed in order, so that a
187 "faulty" attempt is always preceded with all-faulty addresses,
188 and this is how Wget uses it. */
189 assert (index == al->faulty);
192 if (al->faulty >= al->count)
193 /* All addresses have been proven faulty. Since there's not much
194 sense in returning the user an empty address list the next
195 time, we'll rather make them all clean, so that they can be
202 * address_list_from_addrinfo
204 * This function transform an addrinfo links list in and address_list.
207 * addrinfo* Linkt list of addrinfo
210 * address_list* New allocated address_list
212 static struct address_list *
213 address_list_from_addrinfo (const struct addrinfo *ai)
215 struct address_list *al;
216 const struct addrinfo *ptr;
221 for (ptr = ai; ptr != NULL ; ptr = ptr->ai_next)
222 if (ptr->ai_family == AF_INET || ptr->ai_family == AF_INET6)
227 al = xmalloc (sizeof (struct address_list));
228 al->addresses = xmalloc (cnt * sizeof (ip_address));
234 for (ptr = ai; ptr != NULL; ptr = ptr->ai_next)
235 if (ptr->ai_family == AF_INET6)
237 const struct sockaddr_in6 *sin6 =
238 (const struct sockaddr_in6 *)ptr->ai_addr;
239 ip->type = IPV6_ADDRESS;
240 ADDRESS_IPV6_IN6_ADDR (ip) = sin6->sin6_addr;
241 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
242 ADDRESS_IPV6_SCOPE (ip) = sin6->sin6_scope_id;
246 else if (ptr->ai_family == AF_INET)
248 const struct sockaddr_in *sin =
249 (const struct sockaddr_in *)ptr->ai_addr;
250 ip->type = IPV4_ADDRESS;
251 ADDRESS_IPV4_IN_ADDR (ip) = sin->sin_addr;
254 assert (ip - al->addresses == cnt);
258 /* Create an address_list out of a NULL-terminated vector of
259 addresses, as returned by gethostbyname. */
260 static struct address_list *
261 address_list_from_vector (char **h_addr_list)
264 struct address_list *al = xmalloc (sizeof (struct address_list));
267 while (h_addr_list[count])
273 al->addresses = xmalloc (count * sizeof (ip_address));
276 for (i = 0; i < count; i++)
278 ip_address *ip = &al->addresses[i];
279 ip->type = IPV4_ADDRESS;
280 memcpy (ADDRESS_IPV4_DATA (ip), h_addr_list[i], 4);
286 /* Like address_list_from_vector, but initialized with a single
289 static struct address_list *
290 address_list_from_single (const ip_address *addr)
292 struct address_list *al = xmalloc (sizeof (struct address_list));
295 al->addresses = xmalloc (sizeof (ip_address));
297 memcpy (al->addresses, addr, sizeof (ip_address));
304 address_list_delete (struct address_list *al)
306 xfree (al->addresses);
311 address_list_release (struct address_list *al)
314 DEBUGP (("Releasing %p (new refcount %d).\n", al, al->refcount));
315 if (al->refcount <= 0)
317 DEBUGP (("Deleting unused %p.\n", al));
318 address_list_delete (al);
323 * sockaddr_set_address
325 * This function takes a sockaddr struct and fills in the protocol type,
326 * the port number and the address. If ENABLE_IPV6 is defined, the sa
327 * parameter should point to a sockaddr_storage structure; if not, it
328 * should point to a sockaddr_in structure.
329 * If the address parameter is NULL, the function will use the unspecified
330 * address (0.0.0.0 for IPv4 and :: for IPv6).
331 * Unsupported address family will abort the whole programm.
334 * struct sockaddr* The space to be filled
335 * unsigned short The port
336 * const ip_address The IP address
339 * - Only modifies 1st parameter.
342 sockaddr_set_address (struct sockaddr *sa, unsigned short port,
343 const ip_address *addr)
345 if (addr->type == IPV4_ADDRESS)
347 struct sockaddr_in *sin = (struct sockaddr_in *)sa;
349 sin->sin_family = AF_INET;
350 sin->sin_port = htons (port);
352 sin->sin_addr.s_addr = INADDR_ANY;
354 sin->sin_addr = ADDRESS_IPV4_IN_ADDR (addr);
357 else if (addr->type == IPV6_ADDRESS)
359 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa;
361 sin6->sin6_family = AF_INET6;
362 sin6->sin6_port = htons (port);
363 /* #### How can ADDR be NULL? We have dereferenced it above by
364 accessing addr->type! */
367 sin6->sin6_addr = in6addr_any;
368 /* #### Should we set the scope_id here? */
372 sin6->sin6_addr = ADDRESS_IPV6_IN6_ADDR (addr);
373 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
374 sin6->sin6_scope_id = ADDRESS_IPV6_SCOPE (addr);
378 #endif /* ENABLE_IPV6 */
384 sockaddr_get_address (const struct sockaddr *sa, unsigned short *port,
387 if (sa->sa_family == AF_INET)
389 struct sockaddr_in *sin = (struct sockaddr_in *)sa;
391 addr->type = IPV4_ADDRESS;
392 ADDRESS_IPV4_IN_ADDR (addr) = sin->sin_addr;
394 *port = ntohs (sin->sin_port);
397 else if (sa->sa_family == AF_INET6)
399 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa;
401 addr->type = IPV6_ADDRESS;
402 ADDRESS_IPV6_IN6_ADDR (addr) = sin6->sin6_addr;
403 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
404 ADDRESS_IPV6_SCOPE (addr) = sin6->sin6_scope_id;
407 *port = ntohs (sin6->sin6_port);
414 #if 0 /* currently unused */
418 * This funtion only fill the port of the socket information.
419 * If the protocol is not supported nothing is done.
420 * Unsuported adress family will abort the whole programm.
423 * that the IP-Protocol already is set.
426 * wget_sockaddr* The space there port should be entered
427 * unsigned int The port that should be entered in host order
430 * - Only modify 1. param
433 sockaddr_set_port (struct sockaddr *sa, unsigned short port)
435 if (sa->sa_family == AF_INET)
437 struct sockaddr_in *sin = (struct sockaddr_in *)sa;
438 sin->sin_port = htons (port);
441 else if (sa->sa_family == AF_INET6)
443 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa;
444 sin6->sin6_port = htons (port);
455 * This function only return the port from the input structure
456 * Unsuported adress family will abort the whole programm.
459 * that the IP-Protocol already is set.
462 * wget_sockaddr* Information where to get the port
465 * unsigned short Port Number in host order.
468 sockaddr_get_port (const struct sockaddr *sa)
470 if (sa->sa_family == AF_INET) {
471 const struct sockaddr_in *sin = (const struct sockaddr_in *)sa;
472 return htons (sin->sin_port);
474 } else if (sa->sa_family == AF_INET6) {
475 const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)sa;
476 return htons (sin6->sin6_port);
480 /* do not complain about return nothing */
487 * This function return the length of the sockaddr corresponding to
488 * the acutall prefered protocol for (bind, connect etc...)
489 * Unsuported adress family will abort the whole programm.
492 * that the IP-Protocol already is set.
495 * - Public IP-Family Information
498 * int structure length for socket options
501 sockaddr_len (const struct sockaddr *sa)
503 if (sa->sa_family == AF_INET)
505 return sizeof (struct sockaddr_in);
508 else if (sa->sa_family == AF_INET6)
510 return sizeof (struct sockaddr_in6);
515 /* do not complain about return nothing */
519 /* Versions of gethostbyname and getaddrinfo that support timeout. */
523 struct ghbnwt_context {
524 const char *host_name;
525 struct hostent *hptr;
529 gethostbyname_with_timeout_callback (void *arg)
531 struct ghbnwt_context *ctx = (struct ghbnwt_context *)arg;
532 ctx->hptr = gethostbyname (ctx->host_name);
535 /* Just like gethostbyname, except it times out after TIMEOUT seconds.
536 In case of timeout, NULL is returned and errno is set to ETIMEDOUT.
537 The function makes sure that when NULL is returned for reasons
538 other than timeout, errno is reset. */
540 static struct hostent *
541 gethostbyname_with_timeout (const char *host_name, double timeout)
543 struct ghbnwt_context ctx;
544 ctx.host_name = host_name;
545 if (run_with_timeout (timeout, gethostbyname_with_timeout_callback, &ctx))
547 SET_H_ERRNO (HOST_NOT_FOUND);
556 #else /* ENABLE_IPV6 */
558 struct gaiwt_context {
561 const struct addrinfo *hints;
562 struct addrinfo **res;
567 getaddrinfo_with_timeout_callback (void *arg)
569 struct gaiwt_context *ctx = (struct gaiwt_context *)arg;
570 ctx->exit_code = getaddrinfo (ctx->node, ctx->service, ctx->hints, ctx->res);
573 /* Just like getaddrinfo, except it times out after TIMEOUT seconds.
574 In case of timeout, the EAI_SYSTEM error code is returned and errno
575 is set to ETIMEDOUT. */
578 getaddrinfo_with_timeout (const char *node, const char *service,
579 const struct addrinfo *hints, struct addrinfo **res,
582 struct gaiwt_context ctx;
584 ctx.service = service;
588 if (run_with_timeout (timeout, getaddrinfo_with_timeout_callback, &ctx))
593 return ctx.exit_code;
596 #endif /* ENABLE_IPV6 */
598 /* Pretty-print ADDR. When compiled without IPv6, this is the same as
599 inet_ntoa. With IPv6, it either prints an IPv6 address or an IPv4
603 pretty_print_address (const ip_address *addr)
608 return inet_ntoa (ADDRESS_IPV4_IN_ADDR (addr));
612 static char buf[128];
613 inet_ntop (AF_INET6, &ADDRESS_IPV6_IN6_ADDR (addr), buf, sizeof (buf));
615 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
617 /* append "%SCOPE_ID" for all ?non-global? addresses */
618 char *p = buf + strlen (buf);
620 number_to_string (p, ADDRESS_IPV6_SCOPE (addr));
624 buf[sizeof (buf) - 1] = '\0';
633 /* Add host name HOST with the address ADDR_TEXT to the cache.
634 ADDR_LIST is a NULL-terminated list of addresses, as in struct
638 cache_host_lookup (const char *host, struct address_list *al)
640 if (!host_name_addresses_map)
641 host_name_addresses_map = make_nocase_string_hash_table (0);
644 hash_table_put (host_name_addresses_map, xstrdup_lower (host), al);
650 debug_logprintf ("Caching %s =>", host);
651 for (i = 0; i < al->count; i++)
652 debug_logprintf (" %s", pretty_print_address (al->addresses + i));
653 debug_logprintf ("\n");
658 struct address_list *
659 lookup_host (const char *host, int flags)
661 struct address_list *al = NULL;
665 struct addrinfo hints, *res;
667 /* Is this necessary? Should this function be changed to accept a
669 if (flags & LH_IPV4_ONLY)
671 else if (flags & LH_IPV6_ONLY)
674 family = ip_default_family;
677 /* First, try to check whether the address is already a numeric
678 address, in which case we don't need to cache it or bother with
679 setting up timeouts. Plus, if memory serves me right, Ultrix's
680 gethostbyname can't handle numeric addresses (!).
682 Where getaddrinfo is available, we do it using the AI_NUMERICHOST
683 flag. Without IPv6, we use inet_addr succeeds. */
686 memset (&hints, 0, sizeof (hints));
687 hints.ai_family = family;
688 hints.ai_socktype = SOCK_STREAM;
689 hints.ai_flags = AI_NUMERICHOST;
690 if (flags & LH_PASSIVE)
691 hints.ai_flags = AI_PASSIVE;
693 /* no need to call getaddrinfo_with_timeout here, as we're not
694 * relying on the DNS, but we're only doing an address translation
695 * from presentation (ASCII) to network format */
696 err = getaddrinfo (host, NULL, &hints, &res);
697 if (err == 0 && res != NULL)
699 al = address_list_from_addrinfo (res);
705 uint32_t addr_ipv4 = (uint32_t)inet_addr (host);
706 if (addr_ipv4 != (uint32_t) -1)
708 /* The return value of inet_addr is in network byte order, so
709 we can just copy it to IP. */
711 ip.type = IPV4_ADDRESS;
712 memcpy (ADDRESS_IPV4_DATA (&ip), &addr_ipv4, 4);
713 return address_list_from_single (&ip);
718 /* Then, try to find the host in the cache. */
720 if (host_name_addresses_map)
722 al = hash_table_get (host_name_addresses_map, host);
725 DEBUGP (("Found %s in host_name_addresses_map (%p)\n", host, al));
731 if (!(flags & LH_SILENT))
732 logprintf (LOG_VERBOSE, _("Resolving %s... "), host);
734 /* Host name lookup goes on below. */
738 memset (&hints, 0, sizeof (hints));
739 hints.ai_family = family;
740 hints.ai_socktype = SOCK_STREAM;
741 if (flags & LH_PASSIVE)
742 hints.ai_flags = AI_PASSIVE;
744 err = getaddrinfo_with_timeout (host, NULL, &hints, &res, opt.dns_timeout);
746 if (err != 0 || res == NULL)
748 if (!(flags & LH_SILENT))
749 logprintf (LOG_VERBOSE, _("failed: %s.\n"),
750 err != EAI_SYSTEM ? gai_strerror (err) : strerror (errno));
753 al = address_list_from_addrinfo (res);
758 struct hostent *hptr = gethostbyname_with_timeout (host, opt.dns_timeout);
761 if (!(flags & LH_SILENT))
763 if (errno != ETIMEDOUT)
764 logprintf (LOG_VERBOSE, _("failed: %s.\n"), herrmsg (h_errno));
766 logputs (LOG_VERBOSE, _("failed: timed out.\n"));
770 assert (hptr->h_length == 4);
771 /* Do all systems have h_addr_list, or is it a newer thing? If
772 the latter, use address_list_from_single. */
773 al = address_list_from_vector (hptr->h_addr_list);
777 /* Print the addresses determined by DNS lookup, but no more than
779 if (!(flags & LH_SILENT))
782 int printmax = al->count <= 3 ? al->count : 3;
783 for (i = 0; i < printmax; i++)
785 logprintf (LOG_VERBOSE, "%s",
786 pretty_print_address (al->addresses + i));
787 if (i < printmax - 1)
788 logputs (LOG_VERBOSE, ", ");
790 if (printmax != al->count)
791 logputs (LOG_VERBOSE, ", ...");
792 logputs (LOG_VERBOSE, "\n");
795 /* Cache the lookup information. */
797 cache_host_lookup (host, al);
802 /* Determine whether a URL is acceptable to be followed, according to
803 a list of domains to accept. */
805 accept_domain (struct url *u)
807 assert (u->host != NULL);
810 if (!sufmatch ((const char **)opt.domains, u->host))
813 if (opt.exclude_domains)
815 if (sufmatch ((const char **)opt.exclude_domains, u->host))
821 /* Check whether WHAT is matched in LIST, each element of LIST being a
822 pattern to match WHAT against, using backward matching (see
823 match_backwards() in utils.c).
825 If an element of LIST matched, 1 is returned, 0 otherwise. */
827 sufmatch (const char **list, const char *what)
832 for (i = 0; list[i]; i++)
834 for (j = strlen (list[i]), k = lw; j >= 0 && k >= 0; j--, k--)
835 if (TOLOWER (list[i][j]) != TOLOWER (what[k]))
837 /* The domain must be first to reach to beginning. */
844 /* Print error messages for host errors. */
848 /* Can't use switch since some constants are equal (at least on my
849 system), and the compiler signals "duplicate case value". */
850 if (error == HOST_NOT_FOUND
851 || error == NO_RECOVERY
853 || error == NO_ADDRESS
854 || error == TRY_AGAIN)
855 return _("Host not found");
857 return _("Unknown error");
861 host_cleanup_mapper (void *key, void *value, void *arg_ignored)
863 struct address_list *al;
865 xfree (key); /* host */
867 al = (struct address_list *)value;
868 assert (al->refcount == 1);
869 address_list_delete (al);
877 if (host_name_addresses_map)
879 hash_table_map (host_name_addresses_map, host_cleanup_mapper, NULL);
880 hash_table_destroy (host_name_addresses_map);
881 host_name_addresses_map = NULL;