freebsd-dev/lib/libc/net/getaddrinfo.c

2521 lines
57 KiB
C
Raw Normal View History

/* $KAME: getaddrinfo.c,v 1.15 2000/07/09 04:37:24 itojun Exp $ */
/*
* Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the project nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
/*
* "#ifdef FAITH" part is local hack for supporting IPv4-v6 translator.
*
* Issues to be discussed:
* - Thread safe-ness must be checked.
* - Return values. There are nonstandard return values defined and used
* in the source code. This is because RFC2553 is silent about which error
* code must be returned for which situation.
* - freeaddrinfo(NULL). RFC2553 is silent about it. XNET 5.2 says it is
* invalid. current code - SEGV on freeaddrinfo(NULL)
*
* Note:
* - The code filters out AFs that are not supported by the kernel,
* when globbing NULL hostname (to loopback, or wildcard). Is it the right
* thing to do? What is the relationship with post-RFC2553 AI_ADDRCONFIG
* in ai_flags?
* - (post-2553) semantics of AI_ADDRCONFIG itself is too vague.
* (1) what should we do against numeric hostname (2) what should we do
* against NULL hostname (3) what is AI_ADDRCONFIG itself. AF not ready?
* non-loopback address configured? global address configured?
*
* OS specific notes for netbsd/openbsd/freebsd4/bsdi4:
* - To avoid search order issue, we have a big amount of code duplicate
* from gethnamaddr.c and some other places. The issues that there's no
* lower layer function to lookup "IPv4 or IPv6" record. Calling
* gethostbyname2 from getaddrinfo will end up in wrong search order, as
* presented above.
*
* OS specific notes for freebsd4:
* - FreeBSD supported $GAI. The code does not.
* - FreeBSD allowed classful IPv4 numeric (127.1), the code does not.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "namespace.h"
#include <sys/types.h>
#include <sys/param.h>
#include <sys/socket.h>
#include <net/if.h>
#include <netinet/in.h>
#include <sys/queue.h>
#ifdef INET6
#include <net/if_var.h>
#include <sys/sysctl.h>
#include <netinet6/in6_var.h> /* XXX */
#endif
#include <arpa/inet.h>
#include <arpa/nameser.h>
#include <rpc/rpc.h>
#include <rpcsvc/yp_prot.h>
#include <rpcsvc/ypclnt.h>
#include <netdb.h>
#include <pthread.h>
#include <resolv.h>
#include <string.h>
#include <stdlib.h>
#include <stddef.h>
#include <ctype.h>
#include <unistd.h>
#include <stdio.h>
#include <errno.h>
#include "res_config.h"
#ifdef DEBUG
#include <syslog.h>
#endif
#include <stdarg.h>
#include <nsswitch.h>
#include "un-namespace.h"
#include "libc_private.h"
#if defined(__KAME__) && defined(INET6)
# define FAITH
#endif
#define SUCCESS 0
#define ANY 0
#define YES 1
#define NO 0
static const char in_addrany[] = { 0, 0, 0, 0 };
static const char in_loopback[] = { 127, 0, 0, 1 };
#ifdef INET6
static const char in6_addrany[] = {
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
};
static const char in6_loopback[] = {
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1
};
#endif
struct policyqueue {
TAILQ_ENTRY(policyqueue) pc_entry;
#ifdef INET6
struct in6_addrpolicy pc_policy;
#endif
};
TAILQ_HEAD(policyhead, policyqueue);
static const struct afd {
int a_af;
int a_addrlen;
int a_socklen;
int a_off;
const char *a_addrany;
2001-03-05 10:39:03 +00:00
const char *a_loopback;
int a_scoped;
} afdl [] = {
#ifdef INET6
#define N_INET6 0
{PF_INET6, sizeof(struct in6_addr),
sizeof(struct sockaddr_in6),
offsetof(struct sockaddr_in6, sin6_addr),
in6_addrany, in6_loopback, 1},
#define N_INET 1
#else
#define N_INET 0
#endif
{PF_INET, sizeof(struct in_addr),
sizeof(struct sockaddr_in),
offsetof(struct sockaddr_in, sin_addr),
in_addrany, in_loopback, 0},
{0, 0, 0, 0, NULL, NULL, 0},
};
struct explore {
int e_af;
int e_socktype;
int e_protocol;
const char *e_protostr;
int e_wild;
#define WILD_AF(ex) ((ex)->e_wild & 0x01)
#define WILD_SOCKTYPE(ex) ((ex)->e_wild & 0x02)
#define WILD_PROTOCOL(ex) ((ex)->e_wild & 0x04)
};
static const struct explore explore[] = {
#if 0
{ PF_LOCAL, 0, ANY, ANY, NULL, 0x01 },
#endif
#ifdef INET6
{ PF_INET6, SOCK_DGRAM, IPPROTO_UDP, "udp", 0x07 },
{ PF_INET6, SOCK_STREAM, IPPROTO_TCP, "tcp", 0x07 },
{ PF_INET6, SOCK_RAW, ANY, NULL, 0x05 },
#endif
{ PF_INET, SOCK_DGRAM, IPPROTO_UDP, "udp", 0x07 },
{ PF_INET, SOCK_STREAM, IPPROTO_TCP, "tcp", 0x07 },
{ PF_INET, SOCK_RAW, ANY, NULL, 0x05 },
{ PF_UNSPEC, SOCK_DGRAM, IPPROTO_UDP, "udp", 0x07 },
{ PF_UNSPEC, SOCK_STREAM, IPPROTO_TCP, "tcp", 0x07 },
{ PF_UNSPEC, SOCK_RAW, ANY, NULL, 0x05 },
{ -1, 0, 0, NULL, 0 },
};
#ifdef INET6
#define PTON_MAX 16
#else
#define PTON_MAX 4
#endif
#define AIO_SRCFLAG_DEPRECATED 0x1
struct ai_order {
union {
struct sockaddr_storage aiou_ss;
struct sockaddr aiou_sa;
} aio_src_un;
#define aio_srcsa aio_src_un.aiou_sa
u_int32_t aio_srcflag;
int aio_srcscope;
int aio_dstscope;
struct policyqueue *aio_srcpolicy;
struct policyqueue *aio_dstpolicy;
struct addrinfo *aio_ai;
int aio_matchlen;
};
static const ns_src default_dns_files[] = {
{ NSSRC_FILES, NS_SUCCESS },
{ NSSRC_DNS, NS_SUCCESS },
{ 0 }
};
struct res_target {
struct res_target *next;
const char *name; /* domain name */
int qclass, qtype; /* class and type of query */
u_char *answer; /* buffer to put answer */
int anslen; /* size of answer buffer */
int n; /* result length */
};
#define MAXPACKET (64*1024)
typedef union {
HEADER hdr;
u_char buf[MAXPACKET];
} querybuf;
2002-03-21 22:49:10 +00:00
static int str_isnumber(const char *);
2002-03-22 09:22:15 +00:00
static int explore_null(const struct addrinfo *,
const char *, struct addrinfo **);
static int explore_numeric(const struct addrinfo *, const char *,
const char *, struct addrinfo **);
2002-03-22 09:22:15 +00:00
static int explore_numeric_scope(const struct addrinfo *, const char *,
const char *, struct addrinfo **);
static int get_canonname(const struct addrinfo *,
struct addrinfo *, const char *);
static struct addrinfo *get_ai(const struct addrinfo *,
const struct afd *, const char *);
2002-03-21 22:49:10 +00:00
static int get_portmatch(const struct addrinfo *, const char *);
static int get_port(struct addrinfo *, const char *, int);
static const struct afd *find_afd(int);
static int addrconfig(struct addrinfo *);
static int comp_dst(const void *, const void *);
#ifdef INET6
static int ip6_str2scopeid(char *, struct sockaddr_in6 *, u_int32_t *);
#endif
static int gai_addr2scopetype(struct sockaddr *);
static int explore_fqdn(const struct addrinfo *, const char *,
const char *, struct addrinfo **);
static int reorder(struct addrinfo *);
static int get_addrselectpolicy(struct policyhead *);
static void free_addrselectpolicy(struct policyhead *);
static struct policyqueue *match_addrselectpolicy(struct sockaddr *,
struct policyhead *);
2002-03-22 09:22:15 +00:00
static struct addrinfo *getanswer(const querybuf *, int, const char *, int,
const struct addrinfo *);
#if defined(RESOLVSORT)
static int addr4sort(struct addrinfo *);
#endif
static int _dns_getaddrinfo(void *, void *, va_list);
2002-03-21 22:49:10 +00:00
static void _sethtent(void);
static void _endhtent(void);
static struct addrinfo *_gethtent(const char *, const struct addrinfo *);
static int _files_getaddrinfo(void *, void *, va_list);
#ifdef YP
2002-03-21 22:49:10 +00:00
static struct addrinfo *_yphostent(char *, const struct addrinfo *);
static int _yp_getaddrinfo(void *, void *, va_list);
#endif
2002-03-21 22:49:10 +00:00
static int res_queryN(const char *, struct res_target *);
static int res_searchN(const char *, struct res_target *);
2002-03-22 09:22:15 +00:00
static int res_querydomainN(const char *, const char *,
struct res_target *);
static struct ai_errlist {
const char *str;
int code;
} ai_errlist[] = {
{ "Success", 0, },
{ "Temporary failure in name resolution", EAI_AGAIN, },
{ "Invalid value for ai_flags", EAI_BADFLAGS, },
{ "Non-recoverable failure in name resolution", EAI_FAIL, },
{ "ai_family not supported", EAI_FAMILY, },
{ "Memory allocation failure", EAI_MEMORY, },
{ "hostname nor servname provided, or not known", EAI_NONAME, },
{ "servname not supported for ai_socktype", EAI_SERVICE, },
{ "ai_socktype not supported", EAI_SOCKTYPE, },
{ "System error returned in errno", EAI_SYSTEM, },
{ "Invalid value for hints", EAI_BADHINTS, },
{ "Resolved protocol is unknown", EAI_PROTOCOL, },
/* backward compatibility with userland code prior to 2553bis-02 */
{ "Address family for hostname not supported", 1, },
{ "No address associated with hostname", 7, },
{ NULL, -1, },
};
/*
Make the resolver(3) and many associated interfaces much more reentrant. The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now with what should be totally reentrant, and h_errno values will now be preserved correctly, but this does not affect interfaces such as gethostbyname(3) which are still mostly non-reentrant. In all of these relevant functions, the thread-safety has been pushed down as far as it seems possible right now. This means that operations that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected still under global locks that getaddrinfo(3) defines, but where possible the locking is greatly reduced. The most noticeable improvement is that multiple DNS lookups can now be run at the same time, and this shows major improvement in performance of DNS-lookup threaded programs, and solves the "Mozilla tab serialization" problem. No single-threaded applications need to be recompiled. Multi-threaded applications that reference "_res" to change resolver(3) options will need to be recompiled, and ones which reference "h_errno" will also if they desire the correct h_errno values. If the applications already understood that _res and h_errno were not thread-safe and had their own locking, they will see no performance improvement but will not actually break in any way. Please note that when NSS modules are used, or when nsdispatch(3) defaults to adding any lookups of its own to the individual libc _nsdispatch() calls, those MUST be reentrant as well.
2004-02-25 21:03:46 +00:00
* XXX: Many dependencies are not thread-safe. So, we share lock between
* getaddrinfo() and getipnodeby*(). Still, we cannot use
* getaddrinfo() and getipnodeby*() in conjunction with other
Make the resolver(3) and many associated interfaces much more reentrant. The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now with what should be totally reentrant, and h_errno values will now be preserved correctly, but this does not affect interfaces such as gethostbyname(3) which are still mostly non-reentrant. In all of these relevant functions, the thread-safety has been pushed down as far as it seems possible right now. This means that operations that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected still under global locks that getaddrinfo(3) defines, but where possible the locking is greatly reduced. The most noticeable improvement is that multiple DNS lookups can now be run at the same time, and this shows major improvement in performance of DNS-lookup threaded programs, and solves the "Mozilla tab serialization" problem. No single-threaded applications need to be recompiled. Multi-threaded applications that reference "_res" to change resolver(3) options will need to be recompiled, and ones which reference "h_errno" will also if they desire the correct h_errno values. If the applications already understood that _res and h_errno were not thread-safe and had their own locking, they will see no performance improvement but will not actually break in any way. Please note that when NSS modules are used, or when nsdispatch(3) defaults to adding any lookups of its own to the individual libc _nsdispatch() calls, those MUST be reentrant as well.
2004-02-25 21:03:46 +00:00
* functions which call them.
*/
pthread_mutex_t __getaddrinfo_thread_lock = PTHREAD_MUTEX_INITIALIZER;
#define THREAD_LOCK() \
if (__isthreaded) _pthread_mutex_lock(&__getaddrinfo_thread_lock);
#define THREAD_UNLOCK() \
if (__isthreaded) _pthread_mutex_unlock(&__getaddrinfo_thread_lock);
/* XXX macros that make external reference is BAD. */
#define GET_AI(ai, afd, addr) \
do { \
/* external reference: pai, error, and label free */ \
(ai) = get_ai(pai, (afd), (addr)); \
if ((ai) == NULL) { \
error = EAI_MEMORY; \
goto free; \
} \
} while (/*CONSTCOND*/0)
#define GET_PORT(ai, serv) \
do { \
/* external reference: error and label free */ \
error = get_port((ai), (serv), 0); \
if (error != 0) \
goto free; \
} while (/*CONSTCOND*/0)
#define GET_CANONNAME(ai, str) \
do { \
/* external reference: pai, error and label free */ \
error = get_canonname(pai, (ai), (str)); \
if (error != 0) \
goto free; \
} while (/*CONSTCOND*/0)
#define ERR(err) \
do { \
/* external reference: error, and label bad */ \
error = (err); \
goto bad; \
/*NOTREACHED*/ \
} while (/*CONSTCOND*/0)
#define MATCH_FAMILY(x, y, w) \
((x) == (y) || (/*CONSTCOND*/(w) && ((x) == PF_UNSPEC || (y) == PF_UNSPEC)))
#define MATCH(x, y, w) \
((x) == (y) || (/*CONSTCOND*/(w) && ((x) == ANY || (y) == ANY)))
char *
gai_strerror(ecode)
int ecode;
{
struct ai_errlist *p;
for (p = ai_errlist; p->str; p++) {
if (p->code == ecode)
return (char *)p->str;
}
return "Unknown error";
}
void
freeaddrinfo(ai)
struct addrinfo *ai;
{
struct addrinfo *next;
do {
next = ai->ai_next;
if (ai->ai_canonname)
free(ai->ai_canonname);
/* no need to free(ai->ai_addr) */
free(ai);
ai = next;
} while (ai);
}
static int
str_isnumber(p)
const char *p;
{
char *ep;
if (*p == '\0')
return NO;
ep = NULL;
errno = 0;
(void)strtoul(p, &ep, 10);
if (errno == 0 && ep && *ep == '\0')
return YES;
else
return NO;
}
int
getaddrinfo(hostname, servname, hints, res)
const char *hostname, *servname;
const struct addrinfo *hints;
struct addrinfo **res;
{
struct addrinfo sentinel;
struct addrinfo *cur;
int error = 0;
struct addrinfo ai;
struct addrinfo ai0;
struct addrinfo *pai;
const struct explore *ex;
int numeric = 0;
memset(&sentinel, 0, sizeof(sentinel));
cur = &sentinel;
pai = &ai;
pai->ai_flags = 0;
pai->ai_family = PF_UNSPEC;
pai->ai_socktype = ANY;
pai->ai_protocol = ANY;
pai->ai_addrlen = 0;
pai->ai_canonname = NULL;
pai->ai_addr = NULL;
pai->ai_next = NULL;
if (hostname == NULL && servname == NULL)
return EAI_NONAME;
if (hints) {
/* error check for hints */
if (hints->ai_addrlen || hints->ai_canonname ||
hints->ai_addr || hints->ai_next)
ERR(EAI_BADHINTS); /* xxx */
if (hints->ai_flags & ~AI_MASK)
ERR(EAI_BADFLAGS);
switch (hints->ai_family) {
case PF_UNSPEC:
case PF_INET:
#ifdef INET6
case PF_INET6:
#endif
break;
default:
ERR(EAI_FAMILY);
}
memcpy(pai, hints, sizeof(*pai));
/*
* if both socktype/protocol are specified, check if they
* are meaningful combination.
*/
if (pai->ai_socktype != ANY && pai->ai_protocol != ANY) {
for (ex = explore; ex->e_af >= 0; ex++) {
if (pai->ai_family != ex->e_af)
continue;
if (ex->e_socktype == ANY)
continue;
if (ex->e_protocol == ANY)
continue;
if (pai->ai_socktype == ex->e_socktype &&
pai->ai_protocol != ex->e_protocol) {
ERR(EAI_BADHINTS);
}
}
}
}
/*
* post-2553: AI_ALL and AI_V4MAPPED are effective only against
* AF_INET6 query. They need to be ignored if specified in other
* occassions.
*/
switch (pai->ai_flags & (AI_ALL | AI_V4MAPPED)) {
case AI_V4MAPPED:
case AI_ALL | AI_V4MAPPED:
if (pai->ai_family != AF_INET6)
pai->ai_flags &= ~(AI_ALL | AI_V4MAPPED);
break;
case AI_ALL:
#if 1
/* illegal */
ERR(EAI_BADFLAGS);
#else
pai->ai_flags &= ~(AI_ALL | AI_V4MAPPED);
#endif
break;
}
/*
* check for special cases. (1) numeric servname is disallowed if
* socktype/protocol are left unspecified. (2) servname is disallowed
* for raw and other inet{,6} sockets.
*/
if (MATCH_FAMILY(pai->ai_family, PF_INET, 1)
#ifdef PF_INET6
|| MATCH_FAMILY(pai->ai_family, PF_INET6, 1)
#endif
) {
ai0 = *pai; /* backup *pai */
if (pai->ai_family == PF_UNSPEC) {
#ifdef PF_INET6
pai->ai_family = PF_INET6;
#else
pai->ai_family = PF_INET;
#endif
}
error = get_portmatch(pai, servname);
if (error)
ERR(error);
*pai = ai0;
}
ai0 = *pai;
/* NULL hostname, or numeric hostname */
for (ex = explore; ex->e_af >= 0; ex++) {
*pai = ai0;
/* PF_UNSPEC entries are prepared for DNS queries only */
if (ex->e_af == PF_UNSPEC)
continue;
if (!MATCH_FAMILY(pai->ai_family, ex->e_af, WILD_AF(ex)))
continue;
if (!MATCH(pai->ai_socktype, ex->e_socktype, WILD_SOCKTYPE(ex)))
continue;
if (!MATCH(pai->ai_protocol, ex->e_protocol, WILD_PROTOCOL(ex)))
continue;
if (pai->ai_family == PF_UNSPEC)
pai->ai_family = ex->e_af;
if (pai->ai_socktype == ANY && ex->e_socktype != ANY)
pai->ai_socktype = ex->e_socktype;
if (pai->ai_protocol == ANY && ex->e_protocol != ANY)
pai->ai_protocol = ex->e_protocol;
if (hostname == NULL)
error = explore_null(pai, servname, &cur->ai_next);
else
error = explore_numeric_scope(pai, hostname, servname, &cur->ai_next);
if (error)
goto free;
while (cur && cur->ai_next)
cur = cur->ai_next;
}
/*
* XXX
* If numreic representation of AF1 can be interpreted as FQDN
* representation of AF2, we need to think again about the code below.
*/
if (sentinel.ai_next) {
numeric = 1;
goto good;
}
if (hostname == NULL)
ERR(EAI_NONAME); /* used to be EAI_NODATA */
if (pai->ai_flags & AI_NUMERICHOST)
ERR(EAI_NONAME);
if ((pai->ai_flags & AI_ADDRCONFIG) != 0 && !addrconfig(&ai0))
ERR(EAI_FAIL);
/*
* hostname as alphabetical name.
* we would like to prefer AF_INET6 than AF_INET, so we'll make a
* outer loop by AFs.
*/
for (ex = explore; ex->e_af >= 0; ex++) {
*pai = ai0;
/* require exact match for family field */
if (pai->ai_family != ex->e_af)
continue;
if (!MATCH(pai->ai_socktype, ex->e_socktype,
WILD_SOCKTYPE(ex))) {
continue;
}
if (!MATCH(pai->ai_protocol, ex->e_protocol,
WILD_PROTOCOL(ex))) {
continue;
}
if (pai->ai_socktype == ANY && ex->e_socktype != ANY)
pai->ai_socktype = ex->e_socktype;
if (pai->ai_protocol == ANY && ex->e_protocol != ANY)
pai->ai_protocol = ex->e_protocol;
error = explore_fqdn(pai, hostname, servname,
&cur->ai_next);
while (cur && cur->ai_next)
cur = cur->ai_next;
}
/* XXX inhibit errors if we have the result */
if (sentinel.ai_next)
error = 0;
good:
/*
* ensure we return either:
* - error == 0, non-NULL *res
* - error != 0, NULL *res
*/
if (error == 0) {
if (sentinel.ai_next) {
/*
* If the returned entry is for an active connection,
* and the given name is not numeric, reorder the
* list, so that the application would try the list
* in the most efficient order.
*/
if (hints == NULL || !(hints->ai_flags & AI_PASSIVE)) {
if (!numeric)
(void)reorder(&sentinel);
}
*res = sentinel.ai_next;
return SUCCESS;
} else
error = EAI_FAIL;
}
free:
bad:
if (sentinel.ai_next)
freeaddrinfo(sentinel.ai_next);
*res = NULL;
return error;
}
static int
reorder(sentinel)
struct addrinfo *sentinel;
{
struct addrinfo *ai, **aip;
struct ai_order *aio;
int i, n;
struct policyhead policyhead;
/* count the number of addrinfo elements for sorting. */
for (n = 0, ai = sentinel->ai_next; ai != NULL; ai = ai->ai_next, n++)
;
/*
* If the number is small enough, we can skip the reordering process.
*/
if (n <= 1)
return(n);
/* allocate a temporary array for sort and initialization of it. */
if ((aio = malloc(sizeof(*aio) * n)) == NULL)
return(n); /* give up reordering */
memset(aio, 0, sizeof(*aio) * n);
/* retrieve address selection policy from the kernel */
TAILQ_INIT(&policyhead);
if (!get_addrselectpolicy(&policyhead)) {
/* no policy is installed into kernel, we don't sort. */
free(aio);
return (n);
}
for (i = 0, ai = sentinel->ai_next; i < n; ai = ai->ai_next, i++) {
aio[i].aio_ai = ai;
aio[i].aio_dstscope = gai_addr2scopetype(ai->ai_addr);
aio[i].aio_dstpolicy = match_addrselectpolicy(ai->ai_addr,
&policyhead);
}
/* perform sorting. */
qsort(aio, n, sizeof(*aio), comp_dst);
/* reorder the addrinfo chain. */
for (i = 0, aip = &sentinel->ai_next; i < n; i++) {
*aip = aio[i].aio_ai;
aip = &aio[i].aio_ai->ai_next;
}
*aip = NULL;
/* cleanup and return */
free(aio);
free_addrselectpolicy(&policyhead);
return(n);
}
static int
get_addrselectpolicy(head)
struct policyhead *head;
{
#ifdef INET6
int mib[] = { CTL_NET, PF_INET6, IPPROTO_IPV6, IPV6CTL_ADDRCTLPOLICY };
size_t l;
char *buf;
struct in6_addrpolicy *pol, *ep;
if (sysctl(mib, sizeof(mib) / sizeof(mib[0]), NULL, &l, NULL, 0) < 0)
return (0);
if ((buf = malloc(l)) == NULL)
return (0);
if (sysctl(mib, sizeof(mib) / sizeof(mib[0]), buf, &l, NULL, 0) < 0) {
free(buf);
return (0);
}
ep = (struct in6_addrpolicy *)(buf + l);
for (pol = (struct in6_addrpolicy *)buf; pol + 1 <= ep; pol++) {
struct policyqueue *new;
if ((new = malloc(sizeof(*new))) == NULL) {
free_addrselectpolicy(head); /* make the list empty */
break;
}
new->pc_policy = *pol;
TAILQ_INSERT_TAIL(head, new, pc_entry);
}
free(buf);
return (1);
#else
return (0);
#endif
}
static void
free_addrselectpolicy(head)
struct policyhead *head;
{
struct policyqueue *ent, *nent;
for (ent = TAILQ_FIRST(head); ent; ent = nent) {
nent = TAILQ_NEXT(ent, pc_entry);
TAILQ_REMOVE(head, ent, pc_entry);
free(ent);
}
}
static struct policyqueue *
match_addrselectpolicy(addr, head)
struct sockaddr *addr;
struct policyhead *head;
{
#ifdef INET6
struct policyqueue *ent, *bestent = NULL;
struct in6_addrpolicy *pol;
int matchlen, bestmatchlen = -1;
u_char *mp, *ep, *k, *p, m;
struct sockaddr_in6 key;
switch(addr->sa_family) {
case AF_INET6:
key = *(struct sockaddr_in6 *)addr;
break;
case AF_INET:
/* convert the address into IPv4-mapped IPv6 address. */
memset(&key, 0, sizeof(key));
key.sin6_family = AF_INET6;
key.sin6_len = sizeof(key);
key.sin6_addr.s6_addr[10] = 0xff;
key.sin6_addr.s6_addr[11] = 0xff;
memcpy(&key.sin6_addr.s6_addr[12],
&((struct sockaddr_in *)addr)->sin_addr, 4);
break;
default:
return(NULL);
}
for (ent = TAILQ_FIRST(head); ent; ent = TAILQ_NEXT(ent, pc_entry)) {
pol = &ent->pc_policy;
matchlen = 0;
mp = (u_char *)&pol->addrmask.sin6_addr;
ep = mp + 16; /* XXX: scope field? */
k = (u_char *)&key.sin6_addr;
p = (u_char *)&pol->addr.sin6_addr;
for (; mp < ep && *mp; mp++, k++, p++) {
m = *mp;
if ((*k & m) != *p)
goto next; /* not match */
if (m == 0xff) /* short cut for a typical case */
matchlen += 8;
else {
while (m >= 0x80) {
matchlen++;
m <<= 1;
}
}
}
/* matched. check if this is better than the current best. */
if (matchlen > bestmatchlen) {
bestent = ent;
bestmatchlen = matchlen;
}
next:
continue;
}
return(bestent);
#else
return(NULL);
#endif
}
static int
comp_dst(arg1, arg2)
const void *arg1, *arg2;
{
const struct ai_order *dst1 = arg1, *dst2 = arg2;
/*
* Rule 1: Avoid unusable destinations.
* XXX: we currently do not consider if an appropriate route exists.
*/
if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
dst2->aio_srcsa.sa_family == AF_UNSPEC) {
return(-1);
}
if (dst1->aio_srcsa.sa_family == AF_UNSPEC &&
dst2->aio_srcsa.sa_family != AF_UNSPEC) {
return(1);
}
/* Rule 2: Prefer matching scope. */
if (dst1->aio_dstscope == dst1->aio_srcscope &&
dst2->aio_dstscope != dst2->aio_srcscope) {
return(-1);
}
if (dst1->aio_dstscope != dst1->aio_srcscope &&
dst2->aio_dstscope == dst2->aio_srcscope) {
return(1);
}
/* Rule 3: Avoid deprecated addresses. */
if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
dst2->aio_srcsa.sa_family != AF_UNSPEC) {
if (!(dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
(dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
return(-1);
}
if ((dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
!(dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
return(1);
}
}
/* Rule 4: Prefer home addresses. */
/* XXX: not implemented yet */
/* Rule 5: Prefer matching label. */
#ifdef INET6
if (dst1->aio_srcpolicy && dst1->aio_dstpolicy &&
dst1->aio_srcpolicy->pc_policy.label ==
dst1->aio_dstpolicy->pc_policy.label &&
(dst2->aio_srcpolicy == NULL || dst2->aio_dstpolicy == NULL ||
dst2->aio_srcpolicy->pc_policy.label !=
dst2->aio_dstpolicy->pc_policy.label)) {
return(-1);
}
if (dst2->aio_srcpolicy && dst2->aio_dstpolicy &&
dst2->aio_srcpolicy->pc_policy.label ==
dst2->aio_dstpolicy->pc_policy.label &&
(dst1->aio_srcpolicy == NULL || dst1->aio_dstpolicy == NULL ||
dst1->aio_srcpolicy->pc_policy.label !=
dst1->aio_dstpolicy->pc_policy.label)) {
return(1);
}
#endif
/* Rule 6: Prefer higher precedence. */
#ifdef INET6
if (dst1->aio_dstpolicy &&
(dst2->aio_dstpolicy == NULL ||
dst1->aio_dstpolicy->pc_policy.preced >
dst2->aio_dstpolicy->pc_policy.preced)) {
return(-1);
}
if (dst2->aio_dstpolicy &&
(dst1->aio_dstpolicy == NULL ||
dst2->aio_dstpolicy->pc_policy.preced >
dst1->aio_dstpolicy->pc_policy.preced)) {
return(1);
}
#endif
/* Rule 7: Prefer native transport. */
/* XXX: not implemented yet */
/* Rule 8: Prefer smaller scope. */
if (dst1->aio_dstscope >= 0 &&
dst1->aio_dstscope < dst2->aio_dstscope) {
return(-1);
}
if (dst2->aio_dstscope >= 0 &&
dst2->aio_dstscope < dst1->aio_dstscope) {
return(1);
}
/*
* Rule 9: Use longest matching prefix.
* We compare the match length in a same AF only.
*/
if (dst1->aio_ai->ai_addr->sa_family ==
dst2->aio_ai->ai_addr->sa_family) {
if (dst1->aio_matchlen > dst2->aio_matchlen) {
return(-1);
}
if (dst1->aio_matchlen < dst2->aio_matchlen) {
return(1);
}
}
/* Rule 10: Otherwise, leave the order unchanged. */
return(-1);
}
/*
* Copy from scope.c.
* XXX: we should standardize the functions and link them as standard
* library.
*/
static int
gai_addr2scopetype(sa)
struct sockaddr *sa;
{
#ifdef INET6
struct sockaddr_in6 *sa6;
#endif
struct sockaddr_in *sa4;
switch(sa->sa_family) {
#ifdef INET6
case AF_INET6:
sa6 = (struct sockaddr_in6 *)sa;
if (IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
/* just use the scope field of the multicast address */
return(sa6->sin6_addr.s6_addr[2] & 0x0f);
}
/*
* Unicast addresses: map scope type to corresponding scope
* value defined for multcast addresses.
* XXX: hardcoded scope type values are bad...
*/
if (IN6_IS_ADDR_LOOPBACK(&sa6->sin6_addr))
return(1); /* node local scope */
if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr))
return(2); /* link-local scope */
if (IN6_IS_ADDR_SITELOCAL(&sa6->sin6_addr))
return(5); /* site-local scope */
return(14); /* global scope */
break;
#endif
case AF_INET:
/*
* IPv4 pseudo scoping according to RFC 3484.
*/
sa4 = (struct sockaddr_in *)sa;
/* IPv4 autoconfiguration addresses have link-local scope. */
if (((u_char *)&sa4->sin_addr)[0] == 169 &&
((u_char *)&sa4->sin_addr)[1] == 254)
return(2);
/* Private addresses have site-local scope. */
if (((u_char *)&sa4->sin_addr)[0] == 10 ||
(((u_char *)&sa4->sin_addr)[0] == 172 &&
(((u_char *)&sa4->sin_addr)[1] & 0xf0) == 16) ||
(((u_char *)&sa4->sin_addr)[0] == 192 &&
((u_char *)&sa4->sin_addr)[1] == 168))
return(5);
/* Loopback addresses have link-local scope. */
if (((u_char *)&sa4->sin_addr)[0] == 127)
return(2);
return(14);
break;
default:
errno = EAFNOSUPPORT; /* is this a good error? */
return(-1);
}
}
/*
* hostname == NULL.
* passive socket -> anyaddr (0.0.0.0 or ::)
* non-passive socket -> localhost (127.0.0.1 or ::1)
*/
static int
explore_null(pai, servname, res)
const struct addrinfo *pai;
const char *servname;
struct addrinfo **res;
{
int s;
const struct afd *afd;
struct addrinfo *cur;
struct addrinfo sentinel;
int error;
*res = NULL;
sentinel.ai_next = NULL;
cur = &sentinel;
/*
* filter out AFs that are not supported by the kernel
* XXX errno?
*/
s = _socket(pai->ai_family, SOCK_DGRAM, 0);
if (s < 0) {
if (errno != EMFILE)
return 0;
} else
_close(s);
/*
* if the servname does not match socktype/protocol, ignore it.
*/
if (get_portmatch(pai, servname) != 0)
return 0;
afd = find_afd(pai->ai_family);
if (afd == NULL)
return 0;
if (pai->ai_flags & AI_PASSIVE) {
GET_AI(cur->ai_next, afd, afd->a_addrany);
/* xxx meaningless?
* GET_CANONNAME(cur->ai_next, "anyaddr");
*/
GET_PORT(cur->ai_next, servname);
} else {
GET_AI(cur->ai_next, afd, afd->a_loopback);
/* xxx meaningless?
* GET_CANONNAME(cur->ai_next, "localhost");
*/
GET_PORT(cur->ai_next, servname);
}
cur = cur->ai_next;
*res = sentinel.ai_next;
return 0;
free:
if (sentinel.ai_next)
freeaddrinfo(sentinel.ai_next);
return error;
}
/*
* numeric hostname
*/
static int
explore_numeric(pai, hostname, servname, res)
const struct addrinfo *pai;
const char *hostname;
const char *servname;
struct addrinfo **res;
{
const struct afd *afd;
struct addrinfo *cur;
struct addrinfo sentinel;
int error;
char pton[PTON_MAX];
*res = NULL;
sentinel.ai_next = NULL;
cur = &sentinel;
/*
* if the servname does not match socktype/protocol, ignore it.
*/
if (get_portmatch(pai, servname) != 0)
return 0;
afd = find_afd(pai->ai_family);
if (afd == NULL)
return 0;
switch (afd->a_af) {
#if 1 /*X/Open spec*/
case AF_INET:
if (inet_aton(hostname, (struct in_addr *)pton) == 1) {
if (pai->ai_family == afd->a_af ||
pai->ai_family == PF_UNSPEC /*?*/) {
GET_AI(cur->ai_next, afd, pton);
GET_PORT(cur->ai_next, servname);
while (cur && cur->ai_next)
cur = cur->ai_next;
} else
ERR(EAI_FAMILY); /*xxx*/
}
break;
#endif
default:
if (inet_pton(afd->a_af, hostname, pton) == 1) {
if (pai->ai_family == afd->a_af ||
pai->ai_family == PF_UNSPEC /*?*/) {
GET_AI(cur->ai_next, afd, pton);
GET_PORT(cur->ai_next, servname);
while (cur && cur->ai_next)
cur = cur->ai_next;
} else
ERR(EAI_FAMILY); /* XXX */
}
break;
}
*res = sentinel.ai_next;
return 0;
free:
bad:
if (sentinel.ai_next)
freeaddrinfo(sentinel.ai_next);
return error;
}
/*
* numeric hostname with scope
*/
static int
explore_numeric_scope(pai, hostname, servname, res)
const struct addrinfo *pai;
const char *hostname;
const char *servname;
struct addrinfo **res;
{
#if !defined(SCOPE_DELIMITER) || !defined(INET6)
return explore_numeric(pai, hostname, servname, res);
#else
const struct afd *afd;
struct addrinfo *cur;
int error;
char *cp, *hostname2 = NULL, *scope, *addr;
struct sockaddr_in6 *sin6;
/*
* if the servname does not match socktype/protocol, ignore it.
*/
if (get_portmatch(pai, servname) != 0)
return 0;
afd = find_afd(pai->ai_family);
if (afd == NULL)
return 0;
if (!afd->a_scoped)
return explore_numeric(pai, hostname, servname, res);
cp = strchr(hostname, SCOPE_DELIMITER);
if (cp == NULL)
return explore_numeric(pai, hostname, servname, res);
/*
* Handle special case of <scoped_address><delimiter><scope id>
*/
hostname2 = strdup(hostname);
if (hostname2 == NULL)
return EAI_MEMORY;
/* terminate at the delimiter */
hostname2[cp - hostname] = '\0';
addr = hostname2;
scope = cp + 1;
error = explore_numeric(pai, addr, servname, res);
if (error == 0) {
u_int32_t scopeid;
for (cur = *res; cur; cur = cur->ai_next) {
if (cur->ai_family != AF_INET6)
continue;
sin6 = (struct sockaddr_in6 *)(void *)cur->ai_addr;
if (ip6_str2scopeid(scope, sin6, &scopeid) == -1) {
free(hostname2);
return(EAI_NONAME); /* XXX: is return OK? */
}
sin6->sin6_scope_id = scopeid;
}
}
free(hostname2);
return error;
#endif
}
static int
get_canonname(pai, ai, str)
const struct addrinfo *pai;
struct addrinfo *ai;
const char *str;
{
if ((pai->ai_flags & AI_CANONNAME) != 0) {
ai->ai_canonname = (char *)malloc(strlen(str) + 1);
if (ai->ai_canonname == NULL)
return EAI_MEMORY;
strlcpy(ai->ai_canonname, str, strlen(str) + 1);
}
return 0;
}
static struct addrinfo *
get_ai(pai, afd, addr)
const struct addrinfo *pai;
const struct afd *afd;
const char *addr;
{
char *p;
struct addrinfo *ai;
#ifdef FAITH
struct in6_addr faith_prefix;
char *fp_str;
int translate = 0;
#endif
#ifdef FAITH
/*
* Transfrom an IPv4 addr into a special IPv6 addr format for
* IPv6->IPv4 translation gateway. (only TCP is supported now)
*
* +-----------------------------------+------------+
* | faith prefix part (12 bytes) | embedded |
* | | IPv4 addr part (4 bytes)
* +-----------------------------------+------------+
*
* faith prefix part is specified as ascii IPv6 addr format
* in environmental variable GAI.
* For FAITH to work correctly, routing to faith prefix must be
* setup toward a machine where a FAITH daemon operates.
* Also, the machine must enable some mechanizm
* (e.g. faith interface hack) to divert those packet with
* faith prefixed destination addr to user-land FAITH daemon.
*/
fp_str = getenv("GAI");
if (fp_str && inet_pton(AF_INET6, fp_str, &faith_prefix) == 1 &&
afd->a_af == AF_INET && pai->ai_socktype == SOCK_STREAM) {
u_int32_t v4a;
u_int8_t v4a_top;
memcpy(&v4a, addr, sizeof v4a);
v4a_top = v4a >> IN_CLASSA_NSHIFT;
if (!IN_MULTICAST(v4a) && !IN_EXPERIMENTAL(v4a) &&
v4a_top != 0 && v4a != IN_LOOPBACKNET) {
afd = &afdl[N_INET6];
memcpy(&faith_prefix.s6_addr[12], addr,
sizeof(struct in_addr));
translate = 1;
}
}
#endif
ai = (struct addrinfo *)malloc(sizeof(struct addrinfo)
+ (afd->a_socklen));
if (ai == NULL)
return NULL;
memcpy(ai, pai, sizeof(struct addrinfo));
ai->ai_addr = (struct sockaddr *)(void *)(ai + 1);
memset(ai->ai_addr, 0, (size_t)afd->a_socklen);
ai->ai_addr->sa_len = afd->a_socklen;
ai->ai_addrlen = afd->a_socklen;
ai->ai_addr->sa_family = ai->ai_family = afd->a_af;
p = (char *)(void *)(ai->ai_addr);
#ifdef FAITH
if (translate == 1)
memcpy(p + afd->a_off, &faith_prefix, (size_t)afd->a_addrlen);
else
#endif
memcpy(p + afd->a_off, addr, (size_t)afd->a_addrlen);
return ai;
}
static int
get_portmatch(ai, servname)
const struct addrinfo *ai;
const char *servname;
{
/* get_port does not touch first argument. when matchonly == 1. */
/* LINTED const cast */
return get_port((struct addrinfo *)ai, servname, 1);
}
static int
get_port(ai, servname, matchonly)
struct addrinfo *ai;
const char *servname;
int matchonly;
{
const char *proto;
struct servent *sp;
int port;
int allownumeric;
if (servname == NULL)
return 0;
switch (ai->ai_family) {
case AF_INET:
#ifdef AF_INET6
case AF_INET6:
#endif
break;
default:
return 0;
}
switch (ai->ai_socktype) {
case SOCK_RAW:
return EAI_SERVICE;
case SOCK_DGRAM:
case SOCK_STREAM:
allownumeric = 1;
break;
case ANY:
allownumeric = 0;
break;
default:
return EAI_SOCKTYPE;
}
if (str_isnumber(servname)) {
if (!allownumeric)
return EAI_SERVICE;
port = atoi(servname);
if (port < 0 || port > 65535)
return EAI_SERVICE;
port = htons(port);
} else {
switch (ai->ai_socktype) {
case SOCK_DGRAM:
proto = "udp";
break;
case SOCK_STREAM:
proto = "tcp";
break;
default:
proto = NULL;
break;
}
Make the resolver(3) and many associated interfaces much more reentrant. The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now with what should be totally reentrant, and h_errno values will now be preserved correctly, but this does not affect interfaces such as gethostbyname(3) which are still mostly non-reentrant. In all of these relevant functions, the thread-safety has been pushed down as far as it seems possible right now. This means that operations that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected still under global locks that getaddrinfo(3) defines, but where possible the locking is greatly reduced. The most noticeable improvement is that multiple DNS lookups can now be run at the same time, and this shows major improvement in performance of DNS-lookup threaded programs, and solves the "Mozilla tab serialization" problem. No single-threaded applications need to be recompiled. Multi-threaded applications that reference "_res" to change resolver(3) options will need to be recompiled, and ones which reference "h_errno" will also if they desire the correct h_errno values. If the applications already understood that _res and h_errno were not thread-safe and had their own locking, they will see no performance improvement but will not actually break in any way. Please note that when NSS modules are used, or when nsdispatch(3) defaults to adding any lookups of its own to the individual libc _nsdispatch() calls, those MUST be reentrant as well.
2004-02-25 21:03:46 +00:00
THREAD_LOCK();
if ((sp = getservbyname(servname, proto)) == NULL) {
THREAD_UNLOCK();
return EAI_SERVICE;
Make the resolver(3) and many associated interfaces much more reentrant. The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now with what should be totally reentrant, and h_errno values will now be preserved correctly, but this does not affect interfaces such as gethostbyname(3) which are still mostly non-reentrant. In all of these relevant functions, the thread-safety has been pushed down as far as it seems possible right now. This means that operations that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected still under global locks that getaddrinfo(3) defines, but where possible the locking is greatly reduced. The most noticeable improvement is that multiple DNS lookups can now be run at the same time, and this shows major improvement in performance of DNS-lookup threaded programs, and solves the "Mozilla tab serialization" problem. No single-threaded applications need to be recompiled. Multi-threaded applications that reference "_res" to change resolver(3) options will need to be recompiled, and ones which reference "h_errno" will also if they desire the correct h_errno values. If the applications already understood that _res and h_errno were not thread-safe and had their own locking, they will see no performance improvement but will not actually break in any way. Please note that when NSS modules are used, or when nsdispatch(3) defaults to adding any lookups of its own to the individual libc _nsdispatch() calls, those MUST be reentrant as well.
2004-02-25 21:03:46 +00:00
}
port = sp->s_port;
Make the resolver(3) and many associated interfaces much more reentrant. The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now with what should be totally reentrant, and h_errno values will now be preserved correctly, but this does not affect interfaces such as gethostbyname(3) which are still mostly non-reentrant. In all of these relevant functions, the thread-safety has been pushed down as far as it seems possible right now. This means that operations that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected still under global locks that getaddrinfo(3) defines, but where possible the locking is greatly reduced. The most noticeable improvement is that multiple DNS lookups can now be run at the same time, and this shows major improvement in performance of DNS-lookup threaded programs, and solves the "Mozilla tab serialization" problem. No single-threaded applications need to be recompiled. Multi-threaded applications that reference "_res" to change resolver(3) options will need to be recompiled, and ones which reference "h_errno" will also if they desire the correct h_errno values. If the applications already understood that _res and h_errno were not thread-safe and had their own locking, they will see no performance improvement but will not actually break in any way. Please note that when NSS modules are used, or when nsdispatch(3) defaults to adding any lookups of its own to the individual libc _nsdispatch() calls, those MUST be reentrant as well.
2004-02-25 21:03:46 +00:00
THREAD_UNLOCK();
}
if (!matchonly) {
switch (ai->ai_family) {
case AF_INET:
((struct sockaddr_in *)(void *)
ai->ai_addr)->sin_port = port;
break;
#ifdef INET6
case AF_INET6:
((struct sockaddr_in6 *)(void *)
ai->ai_addr)->sin6_port = port;
break;
#endif
}
}
return 0;
}
static const struct afd *
find_afd(af)
int af;
{
const struct afd *afd;
if (af == PF_UNSPEC)
return NULL;
for (afd = afdl; afd->a_af; afd++) {
if (afd->a_af == af)
return afd;
}
return NULL;
}
/*
* post-2553: AI_ADDRCONFIG check. if we use getipnodeby* as backend, backend
* will take care of it.
* the semantics of AI_ADDRCONFIG is not defined well. we are not sure
* if the code is right or not.
*
* XXX PF_UNSPEC -> PF_INET6 + PF_INET mapping needs to be in sync with
* _dns_getaddrinfo.
*/
static int
addrconfig(pai)
struct addrinfo *pai;
{
int s, af;
/*
* TODO:
* Note that implementation dependent test for address
* configuration should be done everytime called
* (or apropriate interval),
* because addresses will be dynamically assigned or deleted.
*/
af = pai->ai_family;
if (af == AF_UNSPEC) {
if ((s = _socket(AF_INET6, SOCK_DGRAM, 0)) < 0)
af = AF_INET;
else {
_close(s);
if ((s = _socket(AF_INET, SOCK_DGRAM, 0)) < 0)
af = AF_INET6;
else
_close(s);
}
}
if (af != AF_UNSPEC) {
if ((s = _socket(af, SOCK_DGRAM, 0)) < 0)
return 0;
_close(s);
}
pai->ai_family = af;
return 1;
}
#ifdef INET6
/* convert a string to a scope identifier. XXX: IPv6 specific */
static int
ip6_str2scopeid(scope, sin6, scopeid)
char *scope;
struct sockaddr_in6 *sin6;
u_int32_t *scopeid;
{
u_long lscopeid;
struct in6_addr *a6;
char *ep;
a6 = &sin6->sin6_addr;
/* empty scopeid portion is invalid */
if (*scope == '\0')
return -1;
if (IN6_IS_ADDR_LINKLOCAL(a6) || IN6_IS_ADDR_MC_LINKLOCAL(a6)) {
/*
* We currently assume a one-to-one mapping between links
* and interfaces, so we simply use interface indices for
* like-local scopes.
*/
*scopeid = if_nametoindex(scope);
if (*scopeid == 0)
goto trynumeric;
return 0;
}
/* still unclear about literal, allow numeric only - placeholder */
if (IN6_IS_ADDR_SITELOCAL(a6) || IN6_IS_ADDR_MC_SITELOCAL(a6))
goto trynumeric;
if (IN6_IS_ADDR_MC_ORGLOCAL(a6))
goto trynumeric;
else
goto trynumeric; /* global */
/* try to convert to a numeric id as a last resort */
trynumeric:
errno = 0;
lscopeid = strtoul(scope, &ep, 10);
*scopeid = (u_int32_t)(lscopeid & 0xffffffffUL);
if (errno == 0 && ep && *ep == '\0' && *scopeid == lscopeid)
return 0;
else
return -1;
}
#endif
/*
* FQDN hostname, DNS lookup
*/
static int
explore_fqdn(pai, hostname, servname, res)
const struct addrinfo *pai;
const char *hostname;
const char *servname;
struct addrinfo **res;
{
struct addrinfo *result;
struct addrinfo *cur;
int error = 0;
static const ns_dtab dtab[] = {
NS_FILES_CB(_files_getaddrinfo, NULL)
{ NSSRC_DNS, _dns_getaddrinfo, NULL }, /* force -DHESIOD */
NS_NIS_CB(_yp_getaddrinfo, NULL)
{ 0 }
};
result = NULL;
/*
* if the servname does not match socktype/protocol, ignore it.
*/
Make the resolver(3) and many associated interfaces much more reentrant. The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now with what should be totally reentrant, and h_errno values will now be preserved correctly, but this does not affect interfaces such as gethostbyname(3) which are still mostly non-reentrant. In all of these relevant functions, the thread-safety has been pushed down as far as it seems possible right now. This means that operations that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected still under global locks that getaddrinfo(3) defines, but where possible the locking is greatly reduced. The most noticeable improvement is that multiple DNS lookups can now be run at the same time, and this shows major improvement in performance of DNS-lookup threaded programs, and solves the "Mozilla tab serialization" problem. No single-threaded applications need to be recompiled. Multi-threaded applications that reference "_res" to change resolver(3) options will need to be recompiled, and ones which reference "h_errno" will also if they desire the correct h_errno values. If the applications already understood that _res and h_errno were not thread-safe and had their own locking, they will see no performance improvement but will not actually break in any way. Please note that when NSS modules are used, or when nsdispatch(3) defaults to adding any lookups of its own to the individual libc _nsdispatch() calls, those MUST be reentrant as well.
2004-02-25 21:03:46 +00:00
if (get_portmatch(pai, servname) != 0)
return 0;
switch (_nsdispatch(&result, dtab, NSDB_HOSTS, "getaddrinfo",
default_dns_files, hostname, pai)) {
case NS_TRYAGAIN:
error = EAI_AGAIN;
goto free;
case NS_UNAVAIL:
error = EAI_FAIL;
goto free;
case NS_NOTFOUND:
error = EAI_NONAME;
goto free;
case NS_SUCCESS:
error = 0;
for (cur = result; cur; cur = cur->ai_next) {
GET_PORT(cur, servname);
/* canonname should be filled already */
}
break;
}
*res = result;
return 0;
free:
if (result)
freeaddrinfo(result);
return error;
}
#ifdef DEBUG
static const char AskedForGot[] =
"gethostby*.getanswer: asked for \"%s\", got \"%s\"";
#endif
static FILE *hostf = NULL;
static struct addrinfo *
getanswer(answer, anslen, qname, qtype, pai)
const querybuf *answer;
int anslen;
const char *qname;
int qtype;
const struct addrinfo *pai;
{
struct addrinfo sentinel, *cur;
struct addrinfo ai;
const struct afd *afd;
char *canonname;
const HEADER *hp;
const u_char *cp;
int n;
const u_char *eom;
char *bp, *ep;
int type, class, ancount, qdcount;
int haveanswer, had_error;
char tbuf[MAXDNAME];
2002-03-21 22:49:10 +00:00
int (*name_ok)(const char *);
char hostbuf[8*1024];
memset(&sentinel, 0, sizeof(sentinel));
cur = &sentinel;
canonname = NULL;
eom = answer->buf + anslen;
switch (qtype) {
case T_A:
case T_AAAA:
case T_ANY: /*use T_ANY only for T_A/T_AAAA lookup*/
name_ok = res_hnok;
break;
default:
return (NULL); /* XXX should be abort(); */
}
/*
* find first satisfactory answer
*/
hp = &answer->hdr;
ancount = ntohs(hp->ancount);
qdcount = ntohs(hp->qdcount);
bp = hostbuf;
ep = hostbuf + sizeof hostbuf;
cp = answer->buf + HFIXEDSZ;
if (qdcount != 1) {
h_errno = NO_RECOVERY;
return (NULL);
}
n = dn_expand(answer->buf, eom, cp, bp, ep - bp);
if ((n < 0) || !(*name_ok)(bp)) {
h_errno = NO_RECOVERY;
return (NULL);
}
cp += n + QFIXEDSZ;
if (qtype == T_A || qtype == T_AAAA || qtype == T_ANY) {
/* res_send() has already verified that the query name is the
* same as the one we sent; this just gets the expanded name
* (i.e., with the succeeding search-domain tacked on).
*/
n = strlen(bp) + 1; /* for the \0 */
if (n >= MAXHOSTNAMELEN) {
h_errno = NO_RECOVERY;
return (NULL);
}
canonname = bp;
bp += n;
/* The qname can be abbreviated, but h_name is now absolute. */
qname = canonname;
}
haveanswer = 0;
had_error = 0;
while (ancount-- > 0 && cp < eom && !had_error) {
n = dn_expand(answer->buf, eom, cp, bp, ep - bp);
if ((n < 0) || !(*name_ok)(bp)) {
had_error++;
continue;
}
cp += n; /* name */
type = _getshort(cp);
cp += INT16SZ; /* type */
class = _getshort(cp);
cp += INT16SZ + INT32SZ; /* class, TTL */
n = _getshort(cp);
cp += INT16SZ; /* len */
if (class != C_IN) {
/* XXX - debug? syslog? */
cp += n;
continue; /* XXX - had_error++ ? */
}
if ((qtype == T_A || qtype == T_AAAA || qtype == T_ANY) &&
type == T_CNAME) {
n = dn_expand(answer->buf, eom, cp, tbuf, sizeof tbuf);
if ((n < 0) || !(*name_ok)(tbuf)) {
had_error++;
continue;
}
cp += n;
/* Get canonical name. */
n = strlen(tbuf) + 1; /* for the \0 */
if (n > ep - bp || n >= MAXHOSTNAMELEN) {
had_error++;
continue;
}
strlcpy(bp, tbuf, ep - bp);
canonname = bp;
bp += n;
continue;
}
if (qtype == T_ANY) {
if (!(type == T_A || type == T_AAAA)) {
cp += n;
continue;
}
} else if (type != qtype) {
#ifdef DEBUG
if (type != T_KEY && type != T_SIG)
syslog(LOG_NOTICE|LOG_AUTH,
"gethostby*.getanswer: asked for \"%s %s %s\", got type \"%s\"",
qname, p_class(C_IN), p_type(qtype),
p_type(type));
#endif
cp += n;
continue; /* XXX - had_error++ ? */
}
switch (type) {
case T_A:
case T_AAAA:
if (strcasecmp(canonname, bp) != 0) {
#ifdef DEBUG
syslog(LOG_NOTICE|LOG_AUTH,
AskedForGot, canonname, bp);
#endif
cp += n;
continue; /* XXX - had_error++ ? */
}
if (type == T_A && n != INADDRSZ) {
cp += n;
continue;
}
if (type == T_AAAA && n != IN6ADDRSZ) {
cp += n;
continue;
}
#ifdef FILTER_V4MAPPED
if (type == T_AAAA) {
struct in6_addr in6;
memcpy(&in6, cp, sizeof(in6));
if (IN6_IS_ADDR_V4MAPPED(&in6)) {
cp += n;
continue;
}
}
#endif
if (!haveanswer) {
int nn;
canonname = bp;
nn = strlen(bp) + 1; /* for the \0 */
bp += nn;
}
/* don't overwrite pai */
ai = *pai;
ai.ai_family = (type == T_A) ? AF_INET : AF_INET6;
afd = find_afd(ai.ai_family);
if (afd == NULL) {
cp += n;
continue;
}
cur->ai_next = get_ai(&ai, afd, (const char *)cp);
if (cur->ai_next == NULL)
had_error++;
while (cur && cur->ai_next)
cur = cur->ai_next;
cp += n;
break;
default:
abort();
}
if (!had_error)
haveanswer++;
}
if (haveanswer) {
#if defined(RESOLVSORT)
/*
* We support only IPv4 address for backward
* compatibility against gethostbyname(3).
*/
if (_res.nsort && qtype == T_A) {
if (addr4sort(&sentinel) < 0) {
freeaddrinfo(sentinel.ai_next);
h_errno = NO_RECOVERY;
return NULL;
}
}
#endif /*RESOLVSORT*/
if (!canonname)
(void)get_canonname(pai, sentinel.ai_next, qname);
else
(void)get_canonname(pai, sentinel.ai_next, canonname);
h_errno = NETDB_SUCCESS;
return sentinel.ai_next;
}
h_errno = NO_RECOVERY;
return NULL;
}
#ifdef RESOLVSORT
struct addr_ptr {
struct addrinfo *ai;
int aval;
};
static int
addr4sort(struct addrinfo *sentinel)
{
struct addrinfo *ai;
struct addr_ptr *addrs, addr;
struct sockaddr_in *sin;
int naddrs, i, j;
int needsort = 0;
if (!sentinel)
return -1;
naddrs = 0;
for (ai = sentinel->ai_next; ai; ai = ai->ai_next)
naddrs++;
if (naddrs < 2)
return 0; /* We don't need sorting. */
if ((addrs = malloc(sizeof(struct addr_ptr) * naddrs)) == NULL)
return -1;
i = 0;
for (ai = sentinel->ai_next; ai; ai = ai->ai_next) {
sin = (struct sockaddr_in *)ai->ai_addr;
for (j = 0; (unsigned)j < _res.nsort; j++) {
if (_res.sort_list[j].addr.s_addr ==
(sin->sin_addr.s_addr & _res.sort_list[j].mask))
break;
}
addrs[i].ai = ai;
addrs[i].aval = j;
if (needsort == 0 && i > 0 && j < addrs[i - 1].aval)
needsort = i;
i++;
}
if (!needsort) {
free(addrs);
return 0;
}
while (needsort < naddrs) {
for (j = needsort - 1; j >= 0; j--) {
if (addrs[j].aval > addrs[j+1].aval) {
addr = addrs[j];
addrs[j] = addrs[j + 1];
addrs[j + 1] = addr;
} else
break;
}
needsort++;
}
ai = sentinel;
for (i = 0; i < naddrs; ++i) {
ai->ai_next = addrs[i].ai;
ai = ai->ai_next;
}
ai->ai_next = NULL;
free(addrs);
return 0;
}
#endif /*RESOLVSORT*/
/*ARGSUSED*/
static int
_dns_getaddrinfo(rv, cb_data, ap)
void *rv;
void *cb_data;
va_list ap;
{
struct addrinfo *ai;
querybuf *buf, *buf2;
const char *name;
const struct addrinfo *pai;
struct addrinfo sentinel, *cur;
struct res_target q, q2;
name = va_arg(ap, char *);
pai = va_arg(ap, const struct addrinfo *);
memset(&q, 0, sizeof(q2));
memset(&q2, 0, sizeof(q2));
memset(&sentinel, 0, sizeof(sentinel));
cur = &sentinel;
buf = malloc(sizeof(*buf));
if (!buf) {
h_errno = NETDB_INTERNAL;
return NS_NOTFOUND;
}
buf2 = malloc(sizeof(*buf2));
if (!buf2) {
free(buf);
h_errno = NETDB_INTERNAL;
return NS_NOTFOUND;
}
switch (pai->ai_family) {
case AF_UNSPEC:
/* prefer IPv6 */
q.name = name;
q.qclass = C_IN;
q.qtype = T_AAAA;
q.answer = buf->buf;
q.anslen = sizeof(buf->buf);
q.next = &q2;
q2.name = name;
q2.qclass = C_IN;
q2.qtype = T_A;
q2.answer = buf2->buf;
q2.anslen = sizeof(buf2->buf);
break;
case AF_INET:
q.name = name;
q.qclass = C_IN;
q.qtype = T_A;
q.answer = buf->buf;
q.anslen = sizeof(buf->buf);
break;
case AF_INET6:
q.name = name;
q.qclass = C_IN;
q.qtype = T_AAAA;
q.answer = buf->buf;
q.anslen = sizeof(buf->buf);
break;
default:
free(buf);
free(buf2);
return NS_UNAVAIL;
}
if (res_searchN(name, &q) < 0) {
free(buf);
free(buf2);
return NS_NOTFOUND;
}
ai = getanswer(buf, q.n, q.name, q.qtype, pai);
if (ai) {
cur->ai_next = ai;
while (cur && cur->ai_next)
cur = cur->ai_next;
}
if (q.next) {
ai = getanswer(buf2, q2.n, q2.name, q2.qtype, pai);
if (ai)
cur->ai_next = ai;
}
free(buf);
free(buf2);
if (sentinel.ai_next == NULL)
switch (h_errno) {
case HOST_NOT_FOUND:
return NS_NOTFOUND;
case TRY_AGAIN:
return NS_TRYAGAIN;
default:
return NS_UNAVAIL;
}
*((struct addrinfo **)rv) = sentinel.ai_next;
return NS_SUCCESS;
}
static void
_sethtent()
{
if (!hostf)
hostf = fopen(_PATH_HOSTS, "r" );
else
rewind(hostf);
}
static void
_endhtent()
{
if (hostf) {
(void) fclose(hostf);
hostf = NULL;
}
}
static struct addrinfo *
_gethtent(name, pai)
const char *name;
const struct addrinfo *pai;
{
char *p;
char *cp, *tname, *cname;
struct addrinfo hints, *res0, *res;
int error;
const char *addr;
char hostbuf[8*1024];
if (!hostf && !(hostf = fopen(_PATH_HOSTS, "r" )))
return (NULL);
again:
if (!(p = fgets(hostbuf, sizeof hostbuf, hostf)))
return (NULL);
if (*p == '#')
goto again;
if (!(cp = strpbrk(p, "#\n")))
goto again;
*cp = '\0';
if (!(cp = strpbrk(p, " \t")))
goto again;
*cp++ = '\0';
addr = p;
cname = NULL;
/* if this is not something we're looking for, skip it. */
while (cp && *cp) {
if (*cp == ' ' || *cp == '\t') {
cp++;
continue;
}
tname = cp;
if (cname == NULL)
cname = cp;
if ((cp = strpbrk(cp, " \t")) != NULL)
*cp++ = '\0';
if (strcasecmp(name, tname) == 0)
goto found;
}
goto again;
found:
/* we should not glob socktype/protocol here */
memset(&hints, 0, sizeof(hints));
hints.ai_family = pai->ai_family;
hints.ai_socktype = SOCK_DGRAM;
hints.ai_protocol = 0;
hints.ai_flags = AI_NUMERICHOST;
error = getaddrinfo(addr, "0", &hints, &res0);
if (error)
goto again;
#ifdef FILTER_V4MAPPED
/* XXX should check all items in the chain */
if (res0->ai_family == AF_INET6 &&
IN6_IS_ADDR_V4MAPPED(&((struct sockaddr_in6 *)res0->ai_addr)->sin6_addr)) {
freeaddrinfo(res0);
goto again;
}
#endif
for (res = res0; res; res = res->ai_next) {
/* cover it up */
res->ai_flags = pai->ai_flags;
res->ai_socktype = pai->ai_socktype;
res->ai_protocol = pai->ai_protocol;
if (pai->ai_flags & AI_CANONNAME) {
if (get_canonname(pai, res, cname) != 0) {
freeaddrinfo(res0);
goto again;
}
}
}
return res0;
}
/*ARGSUSED*/
static int
_files_getaddrinfo(rv, cb_data, ap)
void *rv;
void *cb_data;
va_list ap;
{
const char *name;
const struct addrinfo *pai;
struct addrinfo sentinel, *cur;
struct addrinfo *p;
name = va_arg(ap, char *);
pai = va_arg(ap, struct addrinfo *);
memset(&sentinel, 0, sizeof(sentinel));
cur = &sentinel;
Make the resolver(3) and many associated interfaces much more reentrant. The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now with what should be totally reentrant, and h_errno values will now be preserved correctly, but this does not affect interfaces such as gethostbyname(3) which are still mostly non-reentrant. In all of these relevant functions, the thread-safety has been pushed down as far as it seems possible right now. This means that operations that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected still under global locks that getaddrinfo(3) defines, but where possible the locking is greatly reduced. The most noticeable improvement is that multiple DNS lookups can now be run at the same time, and this shows major improvement in performance of DNS-lookup threaded programs, and solves the "Mozilla tab serialization" problem. No single-threaded applications need to be recompiled. Multi-threaded applications that reference "_res" to change resolver(3) options will need to be recompiled, and ones which reference "h_errno" will also if they desire the correct h_errno values. If the applications already understood that _res and h_errno were not thread-safe and had their own locking, they will see no performance improvement but will not actually break in any way. Please note that when NSS modules are used, or when nsdispatch(3) defaults to adding any lookups of its own to the individual libc _nsdispatch() calls, those MUST be reentrant as well.
2004-02-25 21:03:46 +00:00
THREAD_LOCK();
_sethtent();
while ((p = _gethtent(name, pai)) != NULL) {
cur->ai_next = p;
while (cur && cur->ai_next)
cur = cur->ai_next;
}
_endhtent();
Make the resolver(3) and many associated interfaces much more reentrant. The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now with what should be totally reentrant, and h_errno values will now be preserved correctly, but this does not affect interfaces such as gethostbyname(3) which are still mostly non-reentrant. In all of these relevant functions, the thread-safety has been pushed down as far as it seems possible right now. This means that operations that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected still under global locks that getaddrinfo(3) defines, but where possible the locking is greatly reduced. The most noticeable improvement is that multiple DNS lookups can now be run at the same time, and this shows major improvement in performance of DNS-lookup threaded programs, and solves the "Mozilla tab serialization" problem. No single-threaded applications need to be recompiled. Multi-threaded applications that reference "_res" to change resolver(3) options will need to be recompiled, and ones which reference "h_errno" will also if they desire the correct h_errno values. If the applications already understood that _res and h_errno were not thread-safe and had their own locking, they will see no performance improvement but will not actually break in any way. Please note that when NSS modules are used, or when nsdispatch(3) defaults to adding any lookups of its own to the individual libc _nsdispatch() calls, those MUST be reentrant as well.
2004-02-25 21:03:46 +00:00
THREAD_UNLOCK();
*((struct addrinfo **)rv) = sentinel.ai_next;
if (sentinel.ai_next == NULL)
return NS_NOTFOUND;
return NS_SUCCESS;
}
#ifdef YP
static char *__ypdomain;
/*ARGSUSED*/
static struct addrinfo *
_yphostent(line, pai)
char *line;
const struct addrinfo *pai;
{
struct addrinfo sentinel, *cur;
struct addrinfo hints, *res, *res0;
int error;
char *p = line;
const char *addr, *canonname;
char *nextline;
char *cp;
addr = canonname = NULL;
memset(&sentinel, 0, sizeof(sentinel));
cur = &sentinel;
nextline:
/* terminate line */
cp = strchr(p, '\n');
if (cp) {
*cp++ = '\0';
nextline = cp;
} else
nextline = NULL;
cp = strpbrk(p, " \t");
if (cp == NULL) {
if (canonname == NULL)
return (NULL);
else
goto done;
}
*cp++ = '\0';
addr = p;
while (cp && *cp) {
if (*cp == ' ' || *cp == '\t') {
cp++;
continue;
}
if (!canonname)
canonname = cp;
if ((cp = strpbrk(cp, " \t")) != NULL)
*cp++ = '\0';
}
hints = *pai;
hints.ai_flags = AI_NUMERICHOST;
error = getaddrinfo(addr, NULL, &hints, &res0);
if (error == 0) {
for (res = res0; res; res = res->ai_next) {
/* cover it up */
res->ai_flags = pai->ai_flags;
if (pai->ai_flags & AI_CANONNAME)
(void)get_canonname(pai, res, canonname);
}
} else
res0 = NULL;
if (res0) {
cur->ai_next = res0;
while (cur && cur->ai_next)
cur = cur->ai_next;
}
if (nextline) {
p = nextline;
goto nextline;
}
done:
return sentinel.ai_next;
}
/*ARGSUSED*/
static int
_yp_getaddrinfo(rv, cb_data, ap)
void *rv;
void *cb_data;
va_list ap;
{
struct addrinfo sentinel, *cur;
struct addrinfo *ai = NULL;
static char *__ypcurrent;
int __ypcurrentlen, r;
const char *name;
const struct addrinfo *pai;
name = va_arg(ap, char *);
pai = va_arg(ap, const struct addrinfo *);
memset(&sentinel, 0, sizeof(sentinel));
cur = &sentinel;
Make the resolver(3) and many associated interfaces much more reentrant. The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now with what should be totally reentrant, and h_errno values will now be preserved correctly, but this does not affect interfaces such as gethostbyname(3) which are still mostly non-reentrant. In all of these relevant functions, the thread-safety has been pushed down as far as it seems possible right now. This means that operations that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected still under global locks that getaddrinfo(3) defines, but where possible the locking is greatly reduced. The most noticeable improvement is that multiple DNS lookups can now be run at the same time, and this shows major improvement in performance of DNS-lookup threaded programs, and solves the "Mozilla tab serialization" problem. No single-threaded applications need to be recompiled. Multi-threaded applications that reference "_res" to change resolver(3) options will need to be recompiled, and ones which reference "h_errno" will also if they desire the correct h_errno values. If the applications already understood that _res and h_errno were not thread-safe and had their own locking, they will see no performance improvement but will not actually break in any way. Please note that when NSS modules are used, or when nsdispatch(3) defaults to adding any lookups of its own to the individual libc _nsdispatch() calls, those MUST be reentrant as well.
2004-02-25 21:03:46 +00:00
THREAD_LOCK();
if (!__ypdomain) {
Make the resolver(3) and many associated interfaces much more reentrant. The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now with what should be totally reentrant, and h_errno values will now be preserved correctly, but this does not affect interfaces such as gethostbyname(3) which are still mostly non-reentrant. In all of these relevant functions, the thread-safety has been pushed down as far as it seems possible right now. This means that operations that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected still under global locks that getaddrinfo(3) defines, but where possible the locking is greatly reduced. The most noticeable improvement is that multiple DNS lookups can now be run at the same time, and this shows major improvement in performance of DNS-lookup threaded programs, and solves the "Mozilla tab serialization" problem. No single-threaded applications need to be recompiled. Multi-threaded applications that reference "_res" to change resolver(3) options will need to be recompiled, and ones which reference "h_errno" will also if they desire the correct h_errno values. If the applications already understood that _res and h_errno were not thread-safe and had their own locking, they will see no performance improvement but will not actually break in any way. Please note that when NSS modules are used, or when nsdispatch(3) defaults to adding any lookups of its own to the individual libc _nsdispatch() calls, those MUST be reentrant as well.
2004-02-25 21:03:46 +00:00
if (_yp_check(&__ypdomain) == 0) {
THREAD_UNLOCK();
return NS_UNAVAIL;
Make the resolver(3) and many associated interfaces much more reentrant. The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now with what should be totally reentrant, and h_errno values will now be preserved correctly, but this does not affect interfaces such as gethostbyname(3) which are still mostly non-reentrant. In all of these relevant functions, the thread-safety has been pushed down as far as it seems possible right now. This means that operations that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected still under global locks that getaddrinfo(3) defines, but where possible the locking is greatly reduced. The most noticeable improvement is that multiple DNS lookups can now be run at the same time, and this shows major improvement in performance of DNS-lookup threaded programs, and solves the "Mozilla tab serialization" problem. No single-threaded applications need to be recompiled. Multi-threaded applications that reference "_res" to change resolver(3) options will need to be recompiled, and ones which reference "h_errno" will also if they desire the correct h_errno values. If the applications already understood that _res and h_errno were not thread-safe and had their own locking, they will see no performance improvement but will not actually break in any way. Please note that when NSS modules are used, or when nsdispatch(3) defaults to adding any lookups of its own to the individual libc _nsdispatch() calls, those MUST be reentrant as well.
2004-02-25 21:03:46 +00:00
}
}
if (__ypcurrent)
free(__ypcurrent);
__ypcurrent = NULL;
/* hosts.byname is only for IPv4 (Solaris8) */
if (pai->ai_family == PF_UNSPEC || pai->ai_family == PF_INET) {
r = yp_match(__ypdomain, "hosts.byname", name,
(int)strlen(name), &__ypcurrent, &__ypcurrentlen);
if (r == 0) {
struct addrinfo ai4;
ai4 = *pai;
ai4.ai_family = AF_INET;
ai = _yphostent(__ypcurrent, &ai4);
if (ai) {
cur->ai_next = ai;
while (cur && cur->ai_next)
cur = cur->ai_next;
}
}
}
/* ipnodes.byname can hold both IPv4/v6 */
r = yp_match(__ypdomain, "ipnodes.byname", name,
(int)strlen(name), &__ypcurrent, &__ypcurrentlen);
if (r == 0) {
ai = _yphostent(__ypcurrent, pai);
if (ai) {
cur->ai_next = ai;
while (cur && cur->ai_next)
cur = cur->ai_next;
}
}
Make the resolver(3) and many associated interfaces much more reentrant. The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now with what should be totally reentrant, and h_errno values will now be preserved correctly, but this does not affect interfaces such as gethostbyname(3) which are still mostly non-reentrant. In all of these relevant functions, the thread-safety has been pushed down as far as it seems possible right now. This means that operations that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected still under global locks that getaddrinfo(3) defines, but where possible the locking is greatly reduced. The most noticeable improvement is that multiple DNS lookups can now be run at the same time, and this shows major improvement in performance of DNS-lookup threaded programs, and solves the "Mozilla tab serialization" problem. No single-threaded applications need to be recompiled. Multi-threaded applications that reference "_res" to change resolver(3) options will need to be recompiled, and ones which reference "h_errno" will also if they desire the correct h_errno values. If the applications already understood that _res and h_errno were not thread-safe and had their own locking, they will see no performance improvement but will not actually break in any way. Please note that when NSS modules are used, or when nsdispatch(3) defaults to adding any lookups of its own to the individual libc _nsdispatch() calls, those MUST be reentrant as well.
2004-02-25 21:03:46 +00:00
THREAD_UNLOCK();
if (sentinel.ai_next == NULL) {
h_errno = HOST_NOT_FOUND;
return NS_NOTFOUND;
}
*((struct addrinfo **)rv) = sentinel.ai_next;
return NS_SUCCESS;
}
#endif
/* resolver logic */
2002-03-21 22:49:10 +00:00
extern const char *__hostalias(const char *);
/*
* Formulate a normal query, send, and await answer.
* Returned answer is placed in supplied buffer "answer".
* Perform preliminary check of answer, returning success only
* if no error is indicated and the answer count is nonzero.
* Return the size of the response on success, -1 on error.
* Error number is left in h_errno.
*
* Caller must parse answer and determine whether it answers the question.
*/
static int
res_queryN(name, target)
const char *name; /* domain name */
struct res_target *target;
{
u_char *buf;
HEADER *hp;
int n;
struct res_target *t;
int rcode;
int ancount;
rcode = NOERROR;
ancount = 0;
if ((_res.options & RES_INIT) == 0 && res_init() == -1) {
h_errno = NETDB_INTERNAL;
return (-1);
}
buf = malloc(MAXPACKET);
if (!buf) {
h_errno = NETDB_INTERNAL;
return -1;
}
for (t = target; t; t = t->next) {
int class, type;
u_char *answer;
int anslen;
hp = (HEADER *)(void *)t->answer;
hp->rcode = NOERROR; /* default */
/* make it easier... */
class = t->qclass;
type = t->qtype;
answer = t->answer;
anslen = t->anslen;
#ifdef DEBUG
if (_res.options & RES_DEBUG)
printf(";; res_query(%s, %d, %d)\n", name, class, type);
#endif
n = res_mkquery(QUERY, name, class, type, NULL, 0, NULL,
buf, MAXPACKET);
if (n > 0 && (_res.options & RES_USE_EDNS0) != 0)
n = res_opt(n, buf, MAXPACKET, anslen);
if (n <= 0) {
#ifdef DEBUG
if (_res.options & RES_DEBUG)
printf(";; res_query: mkquery failed\n");
#endif
free(buf);
h_errno = NO_RECOVERY;
return (n);
}
n = res_send(buf, n, answer, anslen);
#if 0
if (n < 0) {
#ifdef DEBUG
if (_res.options & RES_DEBUG)
printf(";; res_query: send error\n");
#endif
free(buf);
h_errno = TRY_AGAIN;
return (n);
}
#endif
if (n < 0 || n > anslen)
hp->rcode = FORMERR; /* XXX not very informative */
if (hp->rcode != NOERROR || ntohs(hp->ancount) == 0) {
rcode = hp->rcode; /* record most recent error */
#ifdef DEBUG
if (_res.options & RES_DEBUG)
printf(";; rcode = %u, ancount=%u\n", hp->rcode,
ntohs(hp->ancount));
#endif
continue;
}
ancount += ntohs(hp->ancount);
t->n = n;
}
free(buf);
if (ancount == 0) {
switch (rcode) {
case NXDOMAIN:
h_errno = HOST_NOT_FOUND;
break;
case SERVFAIL:
h_errno = TRY_AGAIN;
break;
case NOERROR:
h_errno = NO_DATA;
break;
case FORMERR:
case NOTIMP:
case REFUSED:
default:
h_errno = NO_RECOVERY;
break;
}
return (-1);
}
return (ancount);
}
/*
* Formulate a normal query, send, and retrieve answer in supplied buffer.
* Return the size of the response on success, -1 on error.
* If enabled, implement search rules until answer or unrecoverable failure
* is detected. Error code, if any, is left in h_errno.
*/
static int
res_searchN(name, target)
const char *name; /* domain name */
struct res_target *target;
{
const char *cp, * const *domain;
HEADER *hp = (HEADER *)(void *)target->answer; /*XXX*/
u_int dots;
int trailing_dot, ret, saved_herrno;
int got_nodata = 0, got_servfail = 0, tried_as_is = 0;
if ((_res.options & RES_INIT) == 0 && res_init() == -1) {
h_errno = NETDB_INTERNAL;
return (-1);
}
errno = 0;
h_errno = HOST_NOT_FOUND; /* default, if we never query */
dots = 0;
for (cp = name; *cp; cp++)
dots += (*cp == '.');
trailing_dot = 0;
if (cp > name && *--cp == '.')
trailing_dot++;
/*
* if there aren't any dots, it could be a user-level alias
*/
if (!dots && (cp = __hostalias(name)) != NULL)
return (res_queryN(cp, target));
/*
* If there are dots in the name already, let's just give it a try
* 'as is'. The threshold can be set with the "ndots" option.
*/
saved_herrno = -1;
if (dots >= _res.ndots) {
ret = res_querydomainN(name, NULL, target);
if (ret > 0)
return (ret);
saved_herrno = h_errno;
tried_as_is++;
}
/*
* We do at least one level of search if
* - there is no dot and RES_DEFNAME is set, or
* - there is at least one dot, there is no trailing dot,
* and RES_DNSRCH is set.
*/
if ((!dots && (_res.options & RES_DEFNAMES)) ||
(dots && !trailing_dot && (_res.options & RES_DNSRCH))) {
int done = 0;
for (domain = (const char * const *)_res.dnsrch;
*domain && !done;
domain++) {
ret = res_querydomainN(name, *domain, target);
if (ret > 0)
return (ret);
/*
* If no server present, give up.
* If name isn't found in this domain,
* keep trying higher domains in the search list
* (if that's enabled).
* On a NO_DATA error, keep trying, otherwise
* a wildcard entry of another type could keep us
* from finding this entry higher in the domain.
* If we get some other error (negative answer or
* server failure), then stop searching up,
* but try the input name below in case it's
* fully-qualified.
*/
if (errno == ECONNREFUSED) {
h_errno = TRY_AGAIN;
return (-1);
}
switch (h_errno) {
case NO_DATA:
got_nodata++;
/* FALLTHROUGH */
case HOST_NOT_FOUND:
/* keep trying */
break;
case TRY_AGAIN:
if (hp->rcode == SERVFAIL) {
/* try next search element, if any */
got_servfail++;
break;
}
/* FALLTHROUGH */
default:
/* anything else implies that we're done */
done++;
}
/*
* if we got here for some reason other than DNSRCH,
* we only wanted one iteration of the loop, so stop.
*/
if (!(_res.options & RES_DNSRCH))
done++;
}
}
/*
* if we have not already tried the name "as is", do that now.
* note that we do this regardless of how many dots were in the
* name or whether it ends with a dot.
*/
if (!tried_as_is && (dots || !(_res.options & RES_NOTLDQUERY))) {
ret = res_querydomainN(name, NULL, target);
if (ret > 0)
return (ret);
}
/*
* if we got here, we didn't satisfy the search.
* if we did an initial full query, return that query's h_errno
* (note that we wouldn't be here if that query had succeeded).
* else if we ever got a nodata, send that back as the reason.
* else send back meaningless h_errno, that being the one from
* the last DNSRCH we did.
*/
if (saved_herrno != -1)
h_errno = saved_herrno;
else if (got_nodata)
h_errno = NO_DATA;
else if (got_servfail)
h_errno = TRY_AGAIN;
return (-1);
}
/*
* Perform a call on res_query on the concatenation of name and domain,
* removing a trailing dot from name if domain is NULL.
*/
static int
res_querydomainN(name, domain, target)
const char *name, *domain;
struct res_target *target;
{
char nbuf[MAXDNAME];
const char *longname = nbuf;
size_t n, d;
if ((_res.options & RES_INIT) == 0 && res_init() == -1) {
h_errno = NETDB_INTERNAL;
return (-1);
}
#ifdef DEBUG
if (_res.options & RES_DEBUG)
printf(";; res_querydomain(%s, %s)\n",
name, domain?domain:"<Nil>");
#endif
if (domain == NULL) {
/*
* Check for trailing '.';
* copy without '.' if present.
*/
n = strlen(name);
if (n >= MAXDNAME) {
h_errno = NO_RECOVERY;
return (-1);
}
if (n > 0 && name[--n] == '.') {
strncpy(nbuf, name, n);
nbuf[n] = '\0';
} else
longname = name;
} else {
n = strlen(name);
d = strlen(domain);
if (n + d + 1 >= MAXDNAME) {
h_errno = NO_RECOVERY;
return (-1);
}
snprintf(nbuf, sizeof(nbuf), "%s.%s", name, domain);
}
return (res_queryN(longname, target));
}