NetBSD/usr.sbin/ypbind/ypbind.c

1272 lines
30 KiB
C
Raw Normal View History

/* $NetBSD: ypbind.c,v 1.85 2011/05/24 07:01:53 dholland Exp $ */
/*
1994-05-25 13:55:39 +04:00
* Copyright (c) 1992, 1993 Theo de Raadt <deraadt@fsa.ca>
* 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.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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 AUTHOR 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.
*/
#include <sys/cdefs.h>
#ifndef LINT
__RCSID("$NetBSD: ypbind.c,v 1.85 2011/05/24 07:01:53 dholland Exp $");
#endif
1993-04-26 12:09:02 +04:00
#include <sys/types.h>
2011-05-23 06:54:53 +04:00
#include <sys/param.h>
#include <sys/file.h>
1993-04-26 12:09:02 +04:00
#include <sys/ioctl.h>
#include <sys/signal.h>
#include <sys/socket.h>
#include <sys/stat.h>
2011-05-23 06:54:53 +04:00
#include <sys/syslog.h>
#include <sys/uio.h>
#include <arpa/inet.h>
#include <net/if.h>
#include <ctype.h>
#include <dirent.h>
#include <err.h>
#include <errno.h>
1999-02-12 18:04:00 +03:00
#include <fcntl.h>
2011-05-23 06:54:53 +04:00
#include <ifaddrs.h>
#include <limits.h>
2011-05-23 06:54:53 +04:00
#include <netdb.h>
#include <stdarg.h>
1993-04-26 12:09:02 +04:00
#include <stdio.h>
#include <stdlib.h>
1993-04-26 12:09:02 +04:00
#include <string.h>
2011-05-23 06:54:53 +04:00
#include <syslog.h>
#include <unistd.h>
#include <util.h>
1993-04-26 12:09:02 +04:00
#include <rpc/rpc.h>
#include <rpc/xdr.h>
#include <rpc/pmap_clnt.h>
#include <rpc/pmap_prot.h>
#include <rpc/pmap_rmt.h>
#include <rpcsvc/yp_prot.h>
#include <rpcsvc/ypclnt.h>
#include "pathnames.h"
2011-05-24 10:56:16 +04:00
#define YPSERVERSSUFF ".ypservers"
#define BINDINGDIR (_PATH_VAR_YP "binding")
#ifndef O_SHLOCK
#define O_SHLOCK 0
#endif
2011-05-24 10:56:16 +04:00
int _yp_invalid_domain(const char *); /* XXX libc internal */
2011-05-24 10:56:16 +04:00
////////////////////////////////////////////////////////////
// types and globals
typedef enum {
YPBIND_DIRECT, YPBIND_BROADCAST,
2011-05-24 10:56:16 +04:00
} ypbind_mode_t;
1993-04-26 12:09:02 +04:00
struct domain {
struct domain *dom_next;
char dom_name[YPMAXDOMAIN + 1];
1993-04-26 12:09:02 +04:00
struct sockaddr_in dom_server_addr;
int dom_socket;
CLIENT *dom_client;
long dom_vers;
time_t dom_checktime;
time_t dom_asktime;
1993-04-26 12:09:02 +04:00
int dom_lockfd;
int dom_alive;
2011-05-23 06:06:41 +04:00
uint32_t dom_xid;
FILE *dom_serversfile; /* /var/yp/binding/foo.ypservers */
int dom_been_ypset; /* ypset been done on this domain? */
ypbind_mode_t dom_ypbindmode; /* broadcast or direct */
1993-04-26 12:09:02 +04:00
};
2011-05-24 10:56:16 +04:00
#define BUFSIZE 1400
static char *domainname;
1993-04-26 12:09:02 +04:00
2011-05-24 10:59:07 +04:00
static struct domain *domains;
static int check;
1993-04-26 12:09:02 +04:00
static ypbind_mode_t default_ypbindmode;
static int allow_local_ypset = 0, allow_any_ypset = 0;
static int insecure;
static int rpcsock, pingsock;
static struct rmtcallargs rmtca;
static struct rmtcallres rmtcr;
static bool_t rmtcr_outval;
2011-05-23 06:06:41 +04:00
static unsigned long rmtcr_port;
static SVCXPRT *udptransp, *tcptransp;
////////////////////////////////////////////////////////////
// utilities
static int
open_locked(const char *path, int flags, mode_t mode)
{
int fd;
fd = open(path, flags|O_SHLOCK, mode);
if (fd < 0) {
return -1;
}
#if O_SHLOCK == 0
/* dholland 20110522 wouldn't it be better to check this for error? */
(void)flock(fd, LOCK_SH);
#endif
return fd;
}
2011-05-24 10:56:16 +04:00
////////////////////////////////////////////////////////////
// logging
#ifdef DEBUG
2011-05-24 10:56:16 +04:00
#define DPRINTF(...) (debug ? (void)printf(__VA_ARGS__) : (void)0)
static int debug;
#else
#define DPRINTF(...)
#endif
1998-07-13 16:01:03 +04:00
static void yp_log(int, const char *, ...) __printflike(2, 3);
static void
yp_log(int pri, const char *fmt, ...)
{
va_list ap;
va_start(ap, fmt);
#if defined(DEBUG)
if (debug) {
(void)vprintf(fmt, ap);
(void)printf("\n");
} else
#endif
vsyslog(pri, fmt, ap);
va_end(ap);
}
////////////////////////////////////////////////////////////
// ypservers file
/*
* Get pathname for the ypservers file for a given domain
* (/var/yp/binding/DOMAIN.ypservers)
*/
static const char *
ypservers_filename(const char *domain)
{
static char ret[PATH_MAX];
(void)snprintf(ret, sizeof(ret), "%s/%s%s",
BINDINGDIR, domain, YPSERVERSSUFF);
return ret;
}
2011-05-24 10:56:16 +04:00
////////////////////////////////////////////////////////////
// struct domain
2011-05-24 10:56:16 +04:00
static struct domain *
2011-05-24 10:58:54 +04:00
domain_find(uint32_t xid)
2011-05-24 10:56:16 +04:00
{
struct domain *dom;
2011-05-24 10:56:16 +04:00
for (dom = domains; dom != NULL; dom = dom->dom_next)
if (dom->dom_xid == xid)
2011-05-24 10:56:16 +04:00
break;
return dom;
2011-05-24 10:56:16 +04:00
}
static uint32_t
unique_xid(struct domain *dom)
2011-05-24 10:56:16 +04:00
{
uint32_t tmp_xid;
tmp_xid = ((uint32_t)(unsigned long)dom) & 0xffffffff;
2011-05-24 10:58:54 +04:00
while (domain_find(tmp_xid) != NULL)
2011-05-24 10:56:16 +04:00
tmp_xid++;
return tmp_xid;
}
static struct domain *
domain_create(const char *name)
{
struct domain *dom;
const char *pathname;
struct stat st;
dom = malloc(sizeof *dom);
if (dom == NULL) {
yp_log(LOG_ERR, "domain_create: Out of memory");
exit(1);
}
dom->dom_next = NULL;
(void)strlcpy(dom->dom_name, name, sizeof(dom->dom_name));
(void)memset(&dom->dom_server_addr, 0, sizeof(dom->dom_server_addr));
dom->dom_socket = -1;
dom->dom_client = NULL;
dom->dom_vers = YPVERS;
dom->dom_checktime = 0;
dom->dom_asktime = 0;
dom->dom_lockfd = -1;
dom->dom_alive = 0;
dom->dom_xid = unique_xid(dom);
dom->dom_been_ypset = 0;
dom->dom_serversfile = NULL;
/*
* Per traditional ypbind(8) semantics, if a ypservers
* file does not exist, we revert to broadcast mode.
*
* The sysadmin can force broadcast mode by passing the
* -broadcast flag. There is currently no way to fail and
* reject domains for which there is no ypservers file.
*/
dom->dom_ypbindmode = default_ypbindmode;
if (dom->dom_ypbindmode == YPBIND_DIRECT) {
pathname = ypservers_filename(domainname);
if (stat(pathname, &st) < 0) {
/* XXX syslog a warning here? */
DPRINTF("%s does not exist, defaulting to broadcast\n",
pathname);
dom->dom_ypbindmode = YPBIND_BROADCAST;
}
}
/* add to global list */
dom->dom_next = domains;
domains = dom;
return dom;
}
2011-05-24 10:56:16 +04:00
////////////////////////////////////////////////////////////
// locks
static int
makelock(struct domain *dom)
{
int fd;
char path[MAXPATHLEN];
1998-07-13 16:01:03 +04:00
(void)snprintf(path, sizeof(path), "%s/%s.%ld", BINDINGDIR,
dom->dom_name, dom->dom_vers);
fd = open_locked(path, O_CREAT|O_RDWR|O_TRUNC, 0644);
if (fd == -1) {
1998-07-13 16:01:03 +04:00
(void)mkdir(BINDINGDIR, 0755);
fd = open_locked(path, O_CREAT|O_RDWR|O_TRUNC, 0644);
if (fd == -1) {
return -1;
}
}
return fd;
}
1993-04-26 12:09:02 +04:00
static void
removelock(struct domain *dom)
{
char path[MAXPATHLEN];
1993-04-26 12:09:02 +04:00
1998-07-13 16:01:03 +04:00
(void)snprintf(path, sizeof(path), "%s/%s.%ld",
BINDINGDIR, dom->dom_name, dom->dom_vers);
1998-07-13 16:01:03 +04:00
(void)unlink(path);
}
/*
2009-11-06 18:36:55 +03:00
* purge_bindingdir: remove old binding files (i.e. "rm BINDINGDIR\/\*.[0-9]")
*
* local YP functions [e.g. yp_master()] will fail without even talking
* to ypbind if there is a stale (non-flock'd) binding file present.
* we have to scan the entire BINDINGDIR for binding files, because
* ypbind may bind more than just the yp_get_default_domain() domain.
*/
static int
2009-11-06 18:36:55 +03:00
purge_bindingdir(const char *dirpath)
2009-11-06 18:41:58 +03:00
{
DIR *dirp;
int unlinkedfiles, l;
struct dirent *dp;
char pathname[MAXPATHLEN];
if ((dirp = opendir(dirpath)) == NULL)
return(-1); /* at this point, shouldn't ever happen */
do {
unlinkedfiles = 0;
while ((dp = readdir(dirp)) != NULL) {
l = dp->d_namlen;
/* 'rm *.[0-9]' */
if (l > 2 && dp->d_name[l-2] == '.' &&
dp->d_name[l-1] >= '0' && dp->d_name[l-1] <= '9') {
(void)snprintf(pathname, sizeof(pathname),
"%s/%s", dirpath, dp->d_name);
if (unlink(pathname) < 0 && errno != ENOENT)
return(-1);
unlinkedfiles++;
}
}
/* rescan dir if we removed it */
if (unlinkedfiles)
rewinddir(dirp);
} while (unlinkedfiles);
closedir(dirp);
return(0);
}
2011-05-24 10:56:16 +04:00
////////////////////////////////////////////////////////////
// sunrpc twaddle
/*
* LOOPBACK IS MORE IMPORTANT: PUT IN HACK
*/
static void
rpc_received(char *dom_name, struct sockaddr_in *raddrp, int force,
int is_ypset)
2011-05-24 10:56:16 +04:00
{
struct domain *dom;
2011-05-24 10:56:16 +04:00
struct iovec iov[2];
struct ypbind_resp ybr;
ssize_t result;
2011-05-24 10:56:16 +04:00
int fd;
DPRINTF("returned from %s about %s\n",
inet_ntoa(raddrp->sin_addr), dom_name);
2011-05-24 10:56:16 +04:00
if (dom_name == NULL)
2011-05-24 10:56:16 +04:00
return;
if (_yp_invalid_domain(dom_name))
2011-05-24 10:56:16 +04:00
return;
/* don't support insecure servers by default */
if (!insecure && ntohs(raddrp->sin_port) >= IPPORT_RESERVED)
return;
for (dom = domains; dom != NULL; dom = dom->dom_next)
if (!strcmp(dom->dom_name, dom_name))
2011-05-24 10:56:16 +04:00
break;
if (dom == NULL) {
2011-05-24 10:56:16 +04:00
if (force == 0)
return;
dom = domain_create(dom_name);
2011-05-24 10:56:16 +04:00
}
if (is_ypset) {
dom->dom_been_ypset = 1;
}
2011-05-24 10:56:16 +04:00
/* soft update, alive */
if (dom->dom_alive == 1 && force == 0) {
if (!memcmp(&dom->dom_server_addr, raddrp,
sizeof(dom->dom_server_addr))) {
dom->dom_alive = 1;
2011-05-24 10:56:16 +04:00
/* recheck binding in 60 sec */
dom->dom_checktime = time(NULL) + 60;
2011-05-24 10:56:16 +04:00
}
return;
}
(void)memcpy(&dom->dom_server_addr, raddrp,
sizeof(dom->dom_server_addr));
2011-05-24 10:56:16 +04:00
/* recheck binding in 60 seconds */
dom->dom_checktime = time(NULL) + 60;
dom->dom_alive = 1;
2011-05-24 10:56:16 +04:00
if (dom->dom_lockfd != -1)
(void)close(dom->dom_lockfd);
2011-05-24 10:56:16 +04:00
if ((fd = makelock(dom)) == -1)
2011-05-24 10:56:16 +04:00
return;
/*
* ok, if BINDINGDIR exists, and we can create the binding file,
* then write to it..
*/
dom->dom_lockfd = fd;
2011-05-24 10:56:16 +04:00
iov[0].iov_base = &(udptransp->xp_port);
iov[0].iov_len = sizeof udptransp->xp_port;
iov[1].iov_base = &ybr;
iov[1].iov_len = sizeof ybr;
(void)memset(&ybr, 0, sizeof ybr);
ybr.ypbind_status = YPBIND_SUCC_VAL;
ybr.ypbind_respbody.ypbind_bindinfo.ypbind_binding_addr =
raddrp->sin_addr;
ybr.ypbind_respbody.ypbind_bindinfo.ypbind_binding_port =
raddrp->sin_port;
result = writev(dom->dom_lockfd, iov, 2);
if (result < 0 || (size_t)result != iov[0].iov_len + iov[1].iov_len) {
if (result < 0)
yp_log(LOG_WARNING, "writev: %s", strerror(errno));
else
yp_log(LOG_WARNING, "writev: short count");
(void)close(dom->dom_lockfd);
removelock(dom);
dom->dom_lockfd = -1;
2011-05-24 10:56:16 +04:00
}
}
static void *
/*ARGSUSED*/
ypbindproc_null_2(SVCXPRT *transp, void *argp)
1993-04-26 12:09:02 +04:00
{
static char res;
DPRINTF("ypbindproc_null_2\n");
1998-07-13 16:01:03 +04:00
(void)memset(&res, 0, sizeof(res));
1993-04-26 12:09:02 +04:00
return (void *)&res;
}
static void *
/*ARGSUSED*/
ypbindproc_domain_2(SVCXPRT *transp, void *argp)
1993-04-26 12:09:02 +04:00
{
static struct ypbind_resp res;
struct domain *dom;
char *arg = *(char **) argp;
time_t now;
int count;
1993-04-26 12:09:02 +04:00
DPRINTF("ypbindproc_domain_2 %s\n", arg);
if (_yp_invalid_domain(arg))
return NULL;
1998-07-13 16:01:03 +04:00
(void)memset(&res, 0, sizeof res);
1993-04-26 12:09:02 +04:00
res.ypbind_status = YPBIND_FAIL_VAL;
for (count = 0, dom = domains;
dom != NULL;
dom = dom->dom_next, count++) {
if (count > 100)
return NULL; /* prevent denial of service */
if (!strcmp(dom->dom_name, arg))
1993-04-26 12:09:02 +04:00
break;
}
1993-04-26 12:09:02 +04:00
if (dom == NULL) {
dom = domain_create(arg);
removelock(dom);
1993-04-26 12:09:02 +04:00
check++;
DPRINTF("unknown domain %s\n", arg);
1993-04-26 12:09:02 +04:00
return NULL;
}
if (dom->dom_alive == 0) {
DPRINTF("dead domain %s\n", arg);
1993-04-26 12:09:02 +04:00
return NULL;
}
1993-04-26 12:09:02 +04:00
#ifdef HEURISTIC
(void)time(&now);
if (now < dom->dom_asktime + 5) {
1993-04-26 12:09:02 +04:00
/*
* Hmm. More than 2 requests in 5 seconds have indicated
* that my binding is possibly incorrect.
* Ok, do an immediate poll of the server.
1993-04-26 12:09:02 +04:00
*/
if (dom->dom_checktime >= now) {
/* don't flood it */
dom->dom_checktime = 0;
check++;
}
1993-04-26 12:09:02 +04:00
}
dom->dom_asktime = now;
1993-04-26 12:09:02 +04:00
#endif
res.ypbind_status = YPBIND_SUCC_VAL;
res.ypbind_respbody.ypbind_bindinfo.ypbind_binding_addr.s_addr =
dom->dom_server_addr.sin_addr.s_addr;
1993-04-26 12:09:02 +04:00
res.ypbind_respbody.ypbind_bindinfo.ypbind_binding_port =
dom->dom_server_addr.sin_port;
DPRINTF("domain %s at %s/%d\n", dom->dom_name,
inet_ntoa(dom->dom_server_addr.sin_addr),
ntohs(dom->dom_server_addr.sin_port));
1993-04-26 12:09:02 +04:00
return &res;
}
static void *
ypbindproc_setdom_2(SVCXPRT *transp, void *argp)
1993-04-26 12:09:02 +04:00
{
struct ypbind_setdom *sd = argp;
1993-04-26 12:09:02 +04:00
struct sockaddr_in *fromsin, bindsin;
1994-09-28 04:39:54 +03:00
static bool_t res;
1993-04-26 12:09:02 +04:00
DPRINTF("ypbindproc_setdom_2 %s\n", inet_ntoa(bindsin.sin_addr));
1998-07-13 16:01:03 +04:00
(void)memset(&res, 0, sizeof(res));
1993-04-26 12:09:02 +04:00
fromsin = svc_getcaller(transp);
if (allow_any_ypset) {
/* nothing */
} else if (allow_local_ypset) {
if (fromsin->sin_addr.s_addr != htonl(INADDR_LOOPBACK)) {
DPRINTF("ypset denied from %s\n",
inet_ntoa(fromsin->sin_addr));
return NULL;
}
} else {
DPRINTF("ypset denied\n");
return NULL;
1993-04-26 12:09:02 +04:00
}
if (ntohs(fromsin->sin_port) >= IPPORT_RESERVED) {
DPRINTF("ypset from unprivileged port denied\n");
1994-09-28 04:39:54 +03:00
return &res;
}
if (sd->ypsetdom_vers != YPVERS) {
DPRINTF("ypset with wrong version denied\n");
1994-09-28 04:39:54 +03:00
return &res;
}
1993-04-26 12:09:02 +04:00
1998-07-13 16:01:03 +04:00
(void)memset(&bindsin, 0, sizeof bindsin);
1993-04-26 12:09:02 +04:00
bindsin.sin_family = AF_INET;
1995-01-15 11:55:59 +03:00
bindsin.sin_len = sizeof(bindsin);
bindsin.sin_addr = sd->ypsetdom_addr;
bindsin.sin_port = sd->ypsetdom_port;
rpc_received(sd->ypsetdom_domain, &bindsin, 1, 1);
1993-04-26 12:09:02 +04:00
DPRINTF("ypset to %s for domain %s succeeded\n",
inet_ntoa(bindsin.sin_addr), sd->ypsetdom_domain);
1993-04-26 12:09:02 +04:00
res = 1;
1994-09-28 04:39:54 +03:00
return &res;
1993-04-26 12:09:02 +04:00
}
static void
ypbindprog_2(struct svc_req *rqstp, register SVCXPRT *transp)
1993-04-26 12:09:02 +04:00
{
union {
char ypbindproc_domain_2_arg[YPMAXDOMAIN + 1];
1993-04-26 12:09:02 +04:00
struct ypbind_setdom ypbindproc_setdom_2_arg;
void *alignment;
1993-04-26 12:09:02 +04:00
} argument;
struct authunix_parms *creds;
char *result;
xdrproc_t xdr_argument, xdr_result;
void *(*local)(SVCXPRT *, void *);
1993-04-26 12:09:02 +04:00
switch (rqstp->rq_proc) {
case YPBINDPROC_NULL:
xdr_argument = xdr_void;
xdr_result = xdr_void;
local = ypbindproc_null_2;
1993-04-26 12:09:02 +04:00
break;
case YPBINDPROC_DOMAIN:
xdr_argument = xdr_ypdomain_wrap_string;
1993-04-26 12:09:02 +04:00
xdr_result = xdr_ypbind_resp;
local = ypbindproc_domain_2;
1993-04-26 12:09:02 +04:00
break;
case YPBINDPROC_SETDOM:
1995-01-15 11:55:59 +03:00
switch (rqstp->rq_cred.oa_flavor) {
1993-04-26 12:09:02 +04:00
case AUTH_UNIX:
creds = (struct authunix_parms *)rqstp->rq_clntcred;
1995-01-15 11:55:59 +03:00
if (creds->aup_uid != 0) {
1993-04-26 12:09:02 +04:00
svcerr_auth(transp, AUTH_BADCRED);
return;
}
break;
default:
svcerr_auth(transp, AUTH_TOOWEAK);
return;
}
xdr_argument = xdr_ypbind_setdom;
xdr_result = xdr_void;
local = ypbindproc_setdom_2;
1993-04-26 12:09:02 +04:00
break;
default:
svcerr_noproc(transp);
return;
}
1998-07-13 16:01:03 +04:00
(void)memset(&argument, 0, sizeof(argument));
if (!svc_getargs(transp, xdr_argument, (caddr_t)(void *)&argument)) {
1993-04-26 12:09:02 +04:00
svcerr_decode(transp);
return;
}
result = (*local)(transp, &argument);
1993-04-26 12:09:02 +04:00
if (result != NULL && !svc_sendreply(transp, xdr_result, result)) {
svcerr_systemerr(transp);
}
return;
}
static void
sunrpc_setup(void)
{
int one;
(void)pmap_unset(YPBINDPROG, YPBINDVERS);
udptransp = svcudp_create(RPC_ANYSOCK);
if (udptransp == NULL)
errx(1, "Cannot create udp service.");
if (!svc_register(udptransp, YPBINDPROG, YPBINDVERS, ypbindprog_2,
IPPROTO_UDP))
errx(1, "Unable to register (YPBINDPROG, YPBINDVERS, udp).");
tcptransp = svctcp_create(RPC_ANYSOCK, 0, 0);
if (tcptransp == NULL)
errx(1, "Cannot create tcp service.");
if (!svc_register(tcptransp, YPBINDPROG, YPBINDVERS, ypbindprog_2,
IPPROTO_TCP))
errx(1, "Unable to register (YPBINDPROG, YPBINDVERS, tcp).");
/* XXX use SOCK_STREAM for direct queries? */
if ((rpcsock = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP)) == -1)
err(1, "rpc socket");
if ((pingsock = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP)) == -1)
err(1, "ping socket");
(void)fcntl(rpcsock, F_SETFL, fcntl(rpcsock, F_GETFL, 0) | FNDELAY);
(void)fcntl(pingsock, F_SETFL, fcntl(pingsock, F_GETFL, 0) | FNDELAY);
one = 1;
(void)setsockopt(rpcsock, SOL_SOCKET, SO_BROADCAST, &one,
(socklen_t)sizeof(one));
rmtca.prog = YPPROG;
rmtca.vers = YPVERS;
rmtca.proc = YPPROC_DOMAIN_NONACK;
rmtca.xdr_args = NULL; /* set at call time */
rmtca.args_ptr = NULL; /* set at call time */
rmtcr.port_ptr = &rmtcr_port;
rmtcr.xdr_results = xdr_bool;
rmtcr.results_ptr = (caddr_t)(void *)&rmtcr_outval;
}
2011-05-24 10:56:16 +04:00
////////////////////////////////////////////////////////////
// operational logic
1993-04-26 12:09:02 +04:00
2011-05-24 10:56:16 +04:00
static int
broadcast(char *buf, int outlen)
{
struct ifaddrs *ifap, *ifa;
struct sockaddr_in bindsin;
struct in_addr in;
1993-04-26 12:09:02 +04:00
2011-05-24 10:56:16 +04:00
(void)memset(&bindsin, 0, sizeof bindsin);
bindsin.sin_family = AF_INET;
bindsin.sin_len = sizeof(bindsin);
bindsin.sin_port = htons(PMAPPORT);
2011-05-24 10:56:16 +04:00
if (getifaddrs(&ifap) != 0) {
yp_log(LOG_WARNING, "broadcast: getifaddrs: %s",
strerror(errno));
2011-05-24 10:56:16 +04:00
return (-1);
1993-04-26 12:09:02 +04:00
}
2011-05-24 10:56:16 +04:00
for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
if (ifa->ifa_addr->sa_family != AF_INET)
continue;
if ((ifa->ifa_flags & IFF_UP) == 0)
continue;
1993-04-26 12:09:02 +04:00
2011-05-24 10:56:16 +04:00
switch (ifa->ifa_flags & (IFF_LOOPBACK | IFF_BROADCAST)) {
case IFF_BROADCAST:
if (!ifa->ifa_broadaddr)
continue;
if (ifa->ifa_broadaddr->sa_family != AF_INET)
continue;
in = ((struct sockaddr_in *)(void *)ifa->ifa_broadaddr)->sin_addr;
break;
case IFF_LOOPBACK:
in = ((struct sockaddr_in *)(void *)ifa->ifa_addr)->sin_addr;
break;
default:
continue;
}
1993-04-26 12:09:02 +04:00
2011-05-24 10:56:16 +04:00
bindsin.sin_addr = in;
DPRINTF("broadcast %x\n", bindsin.sin_addr.s_addr);
if (sendto(rpcsock, buf, outlen, 0,
(struct sockaddr *)(void *)&bindsin,
(socklen_t)bindsin.sin_len) == -1)
yp_log(LOG_WARNING, "broadcast: sendto: %s",
strerror(errno));
2011-05-24 10:56:16 +04:00
}
freeifaddrs(ifap);
return (0);
}
1993-04-26 12:09:02 +04:00
static int
direct(char *buf, int outlen, struct domain *dom)
{
const char *path;
char line[_POSIX2_LINE_MAX];
char *p;
struct hostent *hp;
struct sockaddr_in bindsin;
int i, count = 0;
/*
* XXX what happens if someone's editor unlinks and replaces
* the servers file?
*/
if (dom->dom_serversfile != NULL) {
rewind(dom->dom_serversfile);
} else {
path = ypservers_filename(dom->dom_name);
dom->dom_serversfile = fopen(path, "r");
if (dom->dom_serversfile == NULL) {
/*
* XXX there should be a time restriction on
* this (and/or on trying the open) so we
* don't flood the log. Or should we fall back
* to broadcast mode?
*/
yp_log(LOG_ERR, "%s: %s", path,
strerror(errno));
return -1;
}
}
(void)memset(&bindsin, 0, sizeof bindsin);
bindsin.sin_family = AF_INET;
bindsin.sin_len = sizeof(bindsin);
bindsin.sin_port = htons(PMAPPORT);
while (fgets(line, (int)sizeof(line), dom->dom_serversfile) != NULL) {
/* skip lines that are too big */
p = strchr(line, '\n');
if (p == NULL) {
int c;
while ((c = getc(dom->dom_serversfile)) != '\n' && c != EOF)
;
continue;
}
*p = '\0';
p = line;
while (isspace((unsigned char)*p))
p++;
if (*p == '#')
continue;
hp = gethostbyname(p);
if (!hp) {
yp_log(LOG_WARNING, "%s: %s", p, hstrerror(h_errno));
continue;
}
/* step through all addresses in case first is unavailable */
for (i = 0; hp->h_addr_list[i]; i++) {
(void)memcpy(&bindsin.sin_addr, hp->h_addr_list[0],
hp->h_length);
if (sendto(rpcsock, buf, outlen, 0,
(struct sockaddr *)(void *)&bindsin,
(socklen_t)sizeof(bindsin)) < 0) {
yp_log(LOG_WARNING, "direct: sendto: %s",
strerror(errno));
continue;
} else
count++;
}
}
if (!count) {
yp_log(LOG_WARNING, "No contactable servers found in %s",
ypservers_filename(dom->dom_name));
return -1;
}
return 0;
}
static int
direct_set(char *buf, int outlen, struct domain *dom)
{
struct sockaddr_in bindsin;
char path[MAXPATHLEN];
struct iovec iov[2];
struct ypbind_resp ybr;
SVCXPRT dummy_svc;
int fd;
ssize_t bytes;
/*
* Gack, we lose if binding file went away. We reset
* "been_set" if this happens, otherwise we'll never
* bind again.
*/
(void)snprintf(path, sizeof(path), "%s/%s.%ld", BINDINGDIR,
dom->dom_name, dom->dom_vers);
fd = open_locked(path, O_RDONLY, 0644);
if (fd == -1) {
yp_log(LOG_WARNING, "%s: %s", path, strerror(errno));
dom->dom_been_ypset = 0;
return -1;
}
/* Read the binding file... */
iov[0].iov_base = &(dummy_svc.xp_port);
iov[0].iov_len = sizeof(dummy_svc.xp_port);
iov[1].iov_base = &ybr;
iov[1].iov_len = sizeof(ybr);
bytes = readv(fd, iov, 2);
(void)close(fd);
if (bytes <0 || (size_t)bytes != (iov[0].iov_len + iov[1].iov_len)) {
/* Binding file corrupt? */
if (bytes < 0)
yp_log(LOG_WARNING, "%s: %s", path, strerror(errno));
else
yp_log(LOG_WARNING, "%s: short read", path);
dom->dom_been_ypset = 0;
return -1;
}
bindsin.sin_addr =
ybr.ypbind_respbody.ypbind_bindinfo.ypbind_binding_addr;
if (sendto(rpcsock, buf, outlen, 0,
(struct sockaddr *)(void *)&bindsin,
(socklen_t)sizeof(bindsin)) < 0) {
yp_log(LOG_WARNING, "direct_set: sendto: %s", strerror(errno));
return -1;
}
return 0;
}
static enum clnt_stat
handle_replies(void)
1993-04-26 12:09:02 +04:00
{
char buf[BUFSIZE];
socklen_t fromlen;
ssize_t inlen;
struct domain *dom;
1993-04-26 12:09:02 +04:00
struct sockaddr_in raddr;
struct rpc_msg msg;
XDR xdr;
recv_again:
DPRINTF("handle_replies receiving\n");
1998-07-13 16:01:03 +04:00
(void)memset(&xdr, 0, sizeof(xdr));
(void)memset(&msg, 0, sizeof(msg));
1993-04-26 12:09:02 +04:00
msg.acpted_rply.ar_verf = _null_auth;
msg.acpted_rply.ar_results.where = (caddr_t)(void *)&rmtcr;
1993-04-26 12:09:02 +04:00
msg.acpted_rply.ar_results.proc = xdr_rmtcallres;
try_again:
1998-07-13 16:01:03 +04:00
fromlen = sizeof(struct sockaddr);
1993-04-26 12:09:02 +04:00
inlen = recvfrom(rpcsock, buf, sizeof buf, 0,
(struct sockaddr *)(void *)&raddr, &fromlen);
1995-01-15 11:55:59 +03:00
if (inlen < 0) {
if (errno == EINTR)
1993-04-26 12:09:02 +04:00
goto try_again;
DPRINTF("handle_replies: recvfrom failed (%s)\n",
strerror(errno));
1993-04-26 12:09:02 +04:00
return RPC_CANTRECV;
}
2011-05-23 06:06:41 +04:00
if ((size_t)inlen < sizeof(uint32_t))
1993-04-26 12:09:02 +04:00
goto recv_again;
/*
* see if reply transaction id matches sent id.
* If so, decode the results.
*/
2011-05-23 06:06:41 +04:00
xdrmem_create(&xdr, buf, (unsigned)inlen, XDR_DECODE);
1995-01-15 11:55:59 +03:00
if (xdr_replymsg(&xdr, &msg)) {
if ((msg.rm_reply.rp_stat == MSG_ACCEPTED) &&
(msg.acpted_rply.ar_stat == SUCCESS)) {
2011-05-23 06:06:41 +04:00
raddr.sin_port = htons((uint16_t)rmtcr_port);
dom = domain_find(msg.rm_xid);
if (dom != NULL)
rpc_received(dom->dom_name, &raddr, 0, 0);
1993-04-26 12:09:02 +04:00
}
}
xdr.x_op = XDR_FREE;
msg.acpted_rply.ar_results.proc = xdr_void;
xdr_destroy(&xdr);
return RPC_SUCCESS;
}
static enum clnt_stat
handle_ping(void)
{
char buf[BUFSIZE];
socklen_t fromlen;
ssize_t inlen;
struct domain *dom;
struct sockaddr_in raddr;
struct rpc_msg msg;
XDR xdr;
bool_t res;
recv_again:
DPRINTF("handle_ping receiving\n");
1998-07-13 16:01:03 +04:00
(void)memset(&xdr, 0, sizeof(xdr));
(void)memset(&msg, 0, sizeof(msg));
msg.acpted_rply.ar_verf = _null_auth;
msg.acpted_rply.ar_results.where = (caddr_t)(void *)&res;
msg.acpted_rply.ar_results.proc = xdr_bool;
try_again:
fromlen = sizeof (struct sockaddr);
inlen = recvfrom(pingsock, buf, sizeof buf, 0,
(struct sockaddr *)(void *)&raddr, &fromlen);
1995-01-15 11:55:59 +03:00
if (inlen < 0) {
if (errno == EINTR)
goto try_again;
DPRINTF("handle_ping: recvfrom failed (%s)\n",
strerror(errno));
return RPC_CANTRECV;
}
2011-05-23 06:06:41 +04:00
if ((size_t)inlen < sizeof(uint32_t))
goto recv_again;
/*
* see if reply transaction id matches sent id.
* If so, decode the results.
*/
2011-05-23 06:06:41 +04:00
xdrmem_create(&xdr, buf, (unsigned)inlen, XDR_DECODE);
1995-01-15 11:55:59 +03:00
if (xdr_replymsg(&xdr, &msg)) {
if ((msg.rm_reply.rp_stat == MSG_ACCEPTED) &&
(msg.acpted_rply.ar_stat == SUCCESS)) {
dom = domain_find(msg.rm_xid);
if (dom != NULL)
rpc_received(dom->dom_name, &raddr, 0, 0);
}
}
2011-05-24 10:56:16 +04:00
xdr.x_op = XDR_FREE;
msg.acpted_rply.ar_results.proc = xdr_void;
xdr_destroy(&xdr);
return RPC_SUCCESS;
}
static int
nag_servers(struct domain *dom)
2011-05-24 10:56:16 +04:00
{
char *dom_name = dom->dom_name;
2011-05-24 10:56:16 +04:00
struct rpc_msg msg;
char buf[BUFSIZE];
enum clnt_stat st;
int outlen;
AUTH *rpcua;
XDR xdr;
DPRINTF("nag_servers\n");
rmtca.xdr_args = xdr_ypdomain_wrap_string;
rmtca.args_ptr = (caddr_t)(void *)&dom_name;
2011-05-24 10:56:16 +04:00
(void)memset(&xdr, 0, sizeof xdr);
(void)memset(&msg, 0, sizeof msg);
rpcua = authunix_create_default();
if (rpcua == NULL) {
DPRINTF("cannot get unix auth\n");
return RPC_SYSTEMERROR;
}
msg.rm_direction = CALL;
msg.rm_call.cb_rpcvers = RPC_MSG_VERSION;
msg.rm_call.cb_prog = PMAPPROG;
msg.rm_call.cb_vers = PMAPVERS;
msg.rm_call.cb_proc = PMAPPROC_CALLIT;
msg.rm_call.cb_cred = rpcua->ah_cred;
msg.rm_call.cb_verf = rpcua->ah_verf;
msg.rm_xid = dom->dom_xid;
2011-05-24 10:56:16 +04:00
xdrmem_create(&xdr, buf, (unsigned)sizeof(buf), XDR_ENCODE);
if (!xdr_callmsg(&xdr, &msg)) {
st = RPC_CANTENCODEARGS;
AUTH_DESTROY(rpcua);
return st;
}
if (!xdr_rmtcall_args(&xdr, &rmtca)) {
st = RPC_CANTENCODEARGS;
AUTH_DESTROY(rpcua);
return st;
}
outlen = (int)xdr_getpos(&xdr);
xdr_destroy(&xdr);
if (outlen < 1) {
st = RPC_CANTENCODEARGS;
AUTH_DESTROY(rpcua);
return st;
}
AUTH_DESTROY(rpcua);
if (dom->dom_lockfd != -1) {
(void)close(dom->dom_lockfd);
dom->dom_lockfd = -1;
removelock(dom);
2011-05-24 10:56:16 +04:00
}
if (dom->dom_alive == 2) {
2011-05-24 10:56:16 +04:00
/*
* This resolves the following situation:
* ypserver on other subnet was once bound,
* but rebooted and is now using a different port
*/
struct sockaddr_in bindsin;
(void)memset(&bindsin, 0, sizeof bindsin);
bindsin.sin_family = AF_INET;
bindsin.sin_len = sizeof(bindsin);
bindsin.sin_port = htons(PMAPPORT);
bindsin.sin_addr = dom->dom_server_addr.sin_addr;
2011-05-24 10:56:16 +04:00
if (sendto(rpcsock, buf, outlen, 0,
(struct sockaddr *)(void *)&bindsin,
(socklen_t)sizeof bindsin) == -1)
yp_log(LOG_WARNING, "nag_servers: sendto: %s",
strerror(errno));
2011-05-24 10:56:16 +04:00
}
switch (dom->dom_ypbindmode) {
2011-05-24 10:56:16 +04:00
case YPBIND_BROADCAST:
if (dom->dom_been_ypset) {
return direct_set(buf, outlen, dom);
}
2011-05-24 10:56:16 +04:00
return broadcast(buf, outlen);
case YPBIND_DIRECT:
return direct(buf, outlen, dom);
2011-05-24 10:56:16 +04:00
}
/*NOTREACHED*/
return -1;
}
static int
ping(struct domain *dom)
2011-05-24 10:56:16 +04:00
{
char *dom_name = dom->dom_name;
2011-05-24 10:56:16 +04:00
struct rpc_msg msg;
char buf[BUFSIZE];
enum clnt_stat st;
int outlen;
AUTH *rpcua;
XDR xdr;
(void)memset(&xdr, 0, sizeof xdr);
(void)memset(&msg, 0, sizeof msg);
rpcua = authunix_create_default();
if (rpcua == NULL) {
DPRINTF("cannot get unix auth\n");
return RPC_SYSTEMERROR;
}
msg.rm_direction = CALL;
msg.rm_call.cb_rpcvers = RPC_MSG_VERSION;
msg.rm_call.cb_prog = YPPROG;
msg.rm_call.cb_vers = YPVERS;
msg.rm_call.cb_proc = YPPROC_DOMAIN_NONACK;
msg.rm_call.cb_cred = rpcua->ah_cred;
msg.rm_call.cb_verf = rpcua->ah_verf;
msg.rm_xid = dom->dom_xid;
2011-05-24 10:56:16 +04:00
xdrmem_create(&xdr, buf, (unsigned)sizeof(buf), XDR_ENCODE);
if (!xdr_callmsg(&xdr, &msg)) {
st = RPC_CANTENCODEARGS;
AUTH_DESTROY(rpcua);
return st;
}
if (!xdr_ypdomain_wrap_string(&xdr, &dom_name)) {
2011-05-24 10:56:16 +04:00
st = RPC_CANTENCODEARGS;
AUTH_DESTROY(rpcua);
return st;
}
outlen = (int)xdr_getpos(&xdr);
xdr_destroy(&xdr);
if (outlen < 1) {
st = RPC_CANTENCODEARGS;
AUTH_DESTROY(rpcua);
return st;
}
AUTH_DESTROY(rpcua);
dom->dom_alive = 2;
DPRINTF("ping %x\n", dom->dom_server_addr.sin_addr.s_addr);
2011-05-24 10:56:16 +04:00
if (sendto(pingsock, buf, outlen, 0,
(struct sockaddr *)(void *)&dom->dom_server_addr,
(socklen_t)(sizeof dom->dom_server_addr)) == -1)
yp_log(LOG_WARNING, "ping: sendto: %s", strerror(errno));
2011-05-24 10:56:16 +04:00
return 0;
}
/*
* State transition is done like this:
*
* STATE EVENT ACTION NEWSTATE TIMEOUT
* no binding timeout broadcast no binding 5 sec
* no binding answer -- binding 60 sec
* binding timeout ping server checking 5 sec
* checking timeout ping server + broadcast checking 5 sec
* checking answer -- binding 60 sec
*/
static void
2011-05-24 10:56:16 +04:00
checkwork(void)
{
struct domain *dom;
2011-05-24 10:56:16 +04:00
time_t t;
check = 0;
(void)time(&t);
for (dom = domains; dom != NULL; dom = dom->dom_next) {
if (dom->dom_checktime < t) {
if (dom->dom_alive == 1)
(void)ping(dom);
2011-05-24 10:56:16 +04:00
else
(void)nag_servers(dom);
2011-05-24 10:56:16 +04:00
(void)time(&t);
dom->dom_checktime = t + 5;
2011-05-24 10:56:16 +04:00
}
}
}
2011-05-24 10:56:16 +04:00
////////////////////////////////////////////////////////////
// main
static void
usage(void)
{
const char *opt = "";
#ifdef DEBUG
opt = " [-d]";
#endif
(void)fprintf(stderr,
"Usage: %s [-broadcast] [-insecure] [-ypset] [-ypsetme]%s\n",
getprogname(), opt);
exit(1);
}
2011-05-24 10:56:16 +04:00
int
main(int argc, char *argv[])
1993-04-26 12:09:02 +04:00
{
2011-05-24 10:56:16 +04:00
struct timeval tv;
fd_set fdsr;
int width, lockfd;
int evil = 0;
1993-04-26 12:09:02 +04:00
2011-05-24 10:56:16 +04:00
setprogname(argv[0]);
(void)yp_get_default_domain(&domainname);
if (domainname[0] == '\0')
errx(1, "Domainname not set. Aborting.");
if (_yp_invalid_domain(domainname))
errx(1, "Invalid domainname: %s", domainname);
1993-04-26 12:09:02 +04:00
default_ypbindmode = YPBIND_DIRECT;
1993-04-26 12:09:02 +04:00
2011-05-24 10:56:16 +04:00
while (--argc) {
++argv;
if (!strcmp("-insecure", *argv)) {
2011-05-24 10:56:16 +04:00
insecure = 1;
} else if (!strcmp("-ypset", *argv)) {
allow_any_ypset = 1;
allow_local_ypset = 1;
} else if (!strcmp("-ypsetme", *argv)) {
allow_any_ypset = 0;
allow_local_ypset = 1;
} else if (!strcmp("-broadcast", *argv)) {
default_ypbindmode = YPBIND_BROADCAST;
2011-05-24 10:56:16 +04:00
#ifdef DEBUG
} else if (!strcmp("-d", *argv)) {
debug = 1;
2011-05-24 10:56:16 +04:00
#endif
} else {
2011-05-24 10:56:16 +04:00
usage();
}
2011-05-24 10:56:16 +04:00
}
2011-05-24 10:56:16 +04:00
/* initialise syslog */
openlog("ypbind", LOG_PERROR | LOG_PID, LOG_DAEMON);
/* acquire ypbind.lock */
lockfd = open_locked(_PATH_YPBIND_LOCK, O_CREAT|O_RDWR|O_TRUNC, 0644);
2011-05-24 10:56:16 +04:00
if (lockfd == -1)
err(1, "Cannot create %s", _PATH_YPBIND_LOCK);
1993-04-26 12:09:02 +04:00
/* initialize sunrpc stuff */
sunrpc_setup();
2011-05-24 10:56:16 +04:00
/* blow away old bindings in BINDINGDIR */
if (purge_bindingdir(BINDINGDIR) < 0)
errx(1, "unable to purge old bindings from %s", BINDINGDIR);
2011-05-24 10:56:16 +04:00
/* build initial domain binding, make it "unsuccessful" */
2011-05-24 10:59:07 +04:00
domains = domain_create(domainname);
removelock(domains);
2011-05-24 10:56:16 +04:00
checkwork();
2011-05-24 10:56:16 +04:00
for (;;) {
width = svc_maxfd;
if (rpcsock > width)
width = rpcsock;
if (pingsock > width)
width = pingsock;
width++;
fdsr = svc_fdset;
FD_SET(rpcsock, &fdsr);
FD_SET(pingsock, &fdsr);
tv.tv_sec = 1;
tv.tv_usec = 0;
switch (select(width, &fdsr, NULL, NULL, &tv)) {
case 0:
checkwork();
break;
case -1:
yp_log(LOG_WARNING, "select: %s", strerror(errno));
2011-05-24 10:56:16 +04:00
break;
default:
if (FD_ISSET(rpcsock, &fdsr))
(void)handle_replies();
if (FD_ISSET(pingsock, &fdsr))
(void)handle_ping();
svc_getreqset(&fdsr);
if (check)
checkwork();
break;
}
2011-05-24 10:59:07 +04:00
if (!evil && domains->dom_alive) {
2011-05-24 10:56:16 +04:00
evil = 1;
#ifdef DEBUG
if (!debug)
#endif
(void)daemon(0, 0);
(void)pidfile(NULL);
}
}
}