freebsd-dev/contrib/bind/bin/named-xfer/named-xfer.c

2178 lines
52 KiB
C
Raw Normal View History

/*
* The original version of xfer by Kevin Dunlap.
* Completed and integrated with named by David Waitzman
* (dwaitzman@bbn.com) 3/14/88.
* Modified by M. Karels and O. Kure 10-88.
* Modified extensively since then by just about everybody.
*/
/*
* Copyright (c) 1988, 1990
* The Regents of the University of California. 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. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University 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 REGENTS 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 REGENTS 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.
*/
/*
* Portions Copyright (c) 1993 by Digital Equipment Corporation.
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies, and that
* the name of Digital Equipment Corporation not be used in advertising or
* publicity pertaining to distribution of the document or software without
* specific, written prior permission.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND DIGITAL EQUIPMENT CORP. DISCLAIMS ALL
* WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL DIGITAL EQUIPMENT
* CORPORATION BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
* DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
* PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
* ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
* SOFTWARE.
*/
/*
* Portions Copyright (c) 1995 by International Business Machines, Inc.
*
* International Business Machines, Inc. (hereinafter called IBM) grants
* permission under its copyrights to use, copy, modify, and distribute this
* Software with or without fee, provided that the above copyright notice and
* all paragraphs of this notice appear in all copies, and that the name of IBM
* not be used in connection with the marketing of any product incorporating
* the Software or modifications thereof, without specific, written prior
* permission.
*
* To the extent it has a right to do so, IBM grants an immunity from suit
* under its patents, if any, for the use, sale or manufacture of products to
* the extent that such products are used for performing Domain Name System
* dynamic updates in TCP/IP networks by means of the Software. No immunity is
* granted for any product per se or for any other function of any product.
*
* THE SOFTWARE IS PROVIDED "AS IS", AND IBM DISCLAIMS ALL WARRANTIES,
* INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
* PARTICULAR PURPOSE. IN NO EVENT SHALL IBM BE LIABLE FOR ANY SPECIAL,
* DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER ARISING
* OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE, EVEN
* IF IBM IS APPRISED OF THE POSSIBILITY OF SUCH DAMAGES.
*/
/* Portions Copyright (c) 1996, 1997 by Internet Software Consortium.
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND INTERNET SOFTWARE CONSORTIUM DISCLAIMS
* ALL WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL INTERNET SOFTWARE
* CONSORTIUM BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
* DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
* PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
* ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
* SOFTWARE.
*/
#if !defined(lint) && !defined(SABER)
char copyright[] =
"@(#) Copyright (c) 1988, 1990 The Regents of the University of California.\n\
portions Copyright (c) 1993 Digital Equipment Corporation\n\
portions Copyright (c) 1995, 1996 Internet Software Consorium\n\
All rights reserved.\n";
#endif /* not lint */
#if !defined(lint) && !defined(SABER)
static char sccsid[] = "@(#)named-xfer.c 4.18 (Berkeley) 3/7/91";
static char rcsid[] = "$Id: named-xfer.c,v 8.38 1998/03/27 00:19:28 halley Exp $";
#endif /* not lint */
#include "port_before.h"
#include "fd_setsize.h"
#include <sys/types.h>
#include <sys/param.h>
#include <sys/file.h>
#include <sys/stat.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <netdb.h>
#include <arpa/inet.h>
#include <arpa/nameser.h>
#include <limits.h>
#include <ctype.h>
#include <errno.h>
#include <math.h>
#include <resolv.h>
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <syslog.h>
#include <unistd.h>
#include <stdarg.h>
#include <isc/eventlib.h>
#include <isc/logging.h>
#include "port_after.h"
#define MAIN_PROGRAM
#include "../named/named.h"
#undef MAIN_PROGRAM
#define MAX_XFER_RESTARTS 2
# ifdef SHORT_FNAMES
extern long pathconf __P((const char *path, int name)); /* XXX */
# endif
static struct zoneinfo zone; /* zone information */
static char ddtfilename[] = _PATH_TMPXFER,
*ddtfile = ddtfilename,
*tmpname,
*domain; /* domain being xfered */
static int quiet = 0,
read_interrupted = 0,
curclass,
domain_len; /* strlen(domain) */
static FILE *fp = NULL,
*dbfp = NULL;
static char *ProgName;
static void usage(const char *);
static int getzone(struct zoneinfo *, u_int32_t, int),
print_output(struct zoneinfo *, u_int32_t,
u_char *, int, u_char *),
netread(int, char *, int, int),
writemsg(int, const u_char *, int);
static SIG_FN read_alarm(void);
static SIG_FN term_handler(void);
static const char *soa_zinfo(struct zoneinfo *, u_char *, u_char*);
struct zoneinfo zp_start, zp_finish;
static int restarts = 0;
FILE *ddt = NULL;
/*
* Debugging printf.
*/
#ifdef DEBUG
void
dprintf(int level, const char *format, ...) {
va_list ap;
va_start(ap, format);
if (ddt != NULL && debug >= level)
(void) vfprintf(ddt, format, ap);
va_end(ap);
}
#endif /*DEBUG*/
static
int init_xfer_logging() {
log_channel chan;
if (log_new_context(ns_log_max_category, NULL, &log_ctx) < 0) {
perror("log_new_context");
return (0);
}
log_option(log_ctx, LOG_OPTION_DEBUG, debug);
log_option(log_ctx, LOG_OPTION_LEVEL, debug);
log_ctx_valid = 1;
chan = log_new_syslog_channel(0, 0, LOG_DAEMON);
if (chan == NULL)
return (0);
if (log_add_channel(log_ctx, ns_log_default, chan) < 0) {
perror("log_add_channel syslog");
return (0);
}
if (debug) {
unsigned int flags = LOG_USE_CONTEXT_LEVEL|LOG_REQUIRE_DEBUG;
chan = log_new_file_channel(flags, 0, NULL, ddt, 0, ULONG_MAX);
if (chan == NULL)
return (0);
if (log_add_channel(log_ctx, ns_log_default, chan) < 0) {
perror("log_add_channel debug");
return (0);
}
}
return (1);
}
void cleanup_for_exit(void) {
#ifdef DEBUG
if (!debug)
#endif
(void) unlink(tmpname);
}
int
main(int argc, char *argv[]) {
struct zoneinfo *zp;
struct hostent *hp;
struct in_addr axfr_src;
char *dbfile = NULL, *tracefile = NULL, *tm = NULL;
int dbfd, ddtd, result, c, fd, closed = 0;
u_int32_t serial_no = 0;
u_int port = htons(NAMESERVER_PORT);
struct stat statbuf;
#ifdef STUBS
int stub_only = 0;
#endif
int class = C_IN;
int n;
long num_files;
ProgName = strrchr(argv[0], '/');
if (ProgName != NULL)
ProgName++;
else
ProgName = argv[0];
(void) umask(022);
/* this is a hack; closing everything in the parent is hard. */
num_files = MIN(sysconf(_SC_OPEN_MAX), FD_SETSIZE);
for (fd = num_files - 1; fd > STDERR_FILENO; fd--)
closed += (close(fd) == 0);
#ifdef RENICE
nice(-40); /* this is the recommended procedure to */
nice(20); /* reset the priority of the current process */
nice(0); /* to "normal" (== 0) - see nice(3) */
#endif
#ifdef LOG_PERROR
n = LOG_PERROR;
#else
n = 0;
#endif
#ifdef SYSLOG_42BSD
openlog(ProgName, LOG_PID);
#else
openlog(ProgName, LOG_PID|LOG_CONS|n, LOG_DAEMON);
#endif
axfr_src.s_addr = 0;
#ifdef STUBS
while ((c = getopt(argc, argv, "C:d:l:s:t:z:f:p:P:qx:S")) != EOF)
#else
while ((c = getopt(argc, argv, "C:d:l:s:t:z:f:p:P:qx:")) != EOF)
#endif
switch (c) {
case 'C':
class = get_class(optarg);
break;
case 'd':
#ifdef DEBUG
debug = atoi(optarg);
#endif
break;
case 'l':
ddtfile = (char *)malloc(strlen(optarg) +
sizeof(".XXXXXX") + 1);
if (!ddtfile)
panic("malloc(ddtfile)", NULL);
#ifdef SHORT_FNAMES
filenamecpy(ddtfile, optarg);
#else
(void) strcpy(ddtfile, optarg);
#endif /* SHORT_FNAMES */
(void) strcat(ddtfile, ".XXXXXX");
break;
case 's':
serial_no = strtoul(optarg, (char **)NULL, 10);
break;
case 't':
tracefile = optarg;
break;
case 'z': /* zone == domain */
domain = optarg;
domain_len = strlen(domain);
while ((domain_len > 0) &&
(domain[domain_len-1] == '.'))
domain[--domain_len] = '\0';
break;
case 'f':
dbfile = optarg;
tmpname = (char *)malloc((unsigned)strlen(optarg) +
sizeof(".XXXXXX") + 1);
if (!tmpname)
panic("malloc(tmpname)", NULL);
#ifdef SHORT_FNAMES
filenamecpy(tmpname, optarg);
#else
(void) strcpy(tmpname, optarg);
#endif /* SHORT_FNAMES */
break;
case 'p':
port = htons((u_int16_t)atoi(optarg));
break;
case 'P':
port = (u_int16_t)atoi(optarg);
break;
#ifdef STUBS
case 'S':
stub_only = 1;
break;
#endif
case 'q':
quiet++;
break;
case 'x':
if (!inet_aton(optarg, &axfr_src))
panic("bad -x addr: %s", optarg);
break;
case '?':
default:
usage("unrecognized argument");
/* NOTREACHED */
}
if (!domain || !dbfile || optind >= argc) {
if (!domain)
usage("no domain");
if (!dbfile)
usage("no dbfile");
if (optind >= argc)
usage("not enough arguments");
/* NOTREACHED */
}
if (stat(dbfile, &statbuf) != -1 &&
!S_ISREG(statbuf.st_mode) &&
!S_ISFIFO(statbuf.st_mode))
usage("dbfile must be a regular file or FIFO");
if (tracefile && (fp = fopen(tracefile, "w")) == NULL)
perror(tracefile);
(void) strcat(tmpname, ".XXXXXX");
/* tmpname is now something like "/etc/named/named.bu.db.XXXXXX" */
if ((dbfd = mkstemp(tmpname)) == -1) {
perror(tmpname);
if (!quiet)
syslog(LOG_ERR, "can't make tmpfile (%s): %m\n",
tmpname);
exit(XFER_FAIL);
}
#ifdef HAVE_FCHMOD /* XXX */
if (fchmod(dbfd, 0644) == -1)
#else
if (chmod(tmpname, 0644) == -1)
#endif
{
perror(tmpname);
if (!quiet)
syslog(LOG_ERR, "can't [f]chmod tmpfile (%s): %m\n",
tmpname);
exit(XFER_FAIL);
}
if ((dbfp = fdopen(dbfd, "r+")) == NULL) {
perror(tmpname);
if (!quiet)
syslog(LOG_ERR, "can't fdopen tmpfile (%s)", tmpname);
exit(XFER_FAIL);
}
#ifdef DEBUG
if (debug) {
/* ddtfile is now something like "/usr/tmp/xfer.ddt.XXXXXX" */
if ((ddtd = mkstemp(ddtfile)) == -1) {
perror(ddtfile);
debug = 0;
}
#ifdef HAVE_FCHMOD
else if (fchmod(ddtd, 0644) == -1)
#else
else if (chmod(ddtfile, 0644) == -1)
#endif
{
perror(ddtfile);
debug = 0;
} else if ((ddt = fdopen(ddtd, "w")) == NULL) {
perror(ddtfile);
debug = 0;
} else
setvbuf(ddt, NULL, _IOLBF, 0);
}
#endif
if (!init_xfer_logging()) {
perror("init_xfer_logging");
}
/*
* Ignore many types of signals that named (assumed to be our parent)
* considers important- if not, the user controlling named with
* signals usually kills us.
*/
(void) signal(SIGHUP, SIG_IGN);
#ifdef SIGSYS
(void) signal(SIGSYS, SIG_IGN);
#endif
#ifdef DEBUG
if (debug == 0)
#endif
{
(void) signal(SIGINT, SIG_IGN);
(void) signal(SIGQUIT, SIG_IGN);
}
(void) signal(SIGILL, SIG_IGN);
#if defined(SIGUSR1) && defined(SIGUSR2)
(void) signal(SIGUSR1, SIG_IGN);
(void) signal(SIGUSR2, SIG_IGN);
#else /* SIGUSR1&&SIGUSR2 */
(void) signal(SIGEMT, SIG_IGN);
(void) signal(SIGFPE, SIG_IGN);
#endif /* SIGUSR1&&SIGUSR2 */
dprintf(1, "domain `%s'; file `%s'; serial %u; closed %d\n",
domain, dbfile, serial_no, closed);
buildservicelist();
buildprotolist();
/* init zone data */
zp = &zone;
#ifdef STUBS
if (stub_only)
zp->z_type = Z_STUB;
else
#endif
zp->z_type = Z_SECONDARY;
zp->z_class = class;
zp->z_origin = domain;
zp->z_source = dbfile;
zp->z_axfr_src = axfr_src;
zp->z_addrcnt = 0;
dprintf(1, "zone found (%d): \"%s\", source = %s\n",
zp->z_type,
(zp->z_origin[0] == '\0') ? "." : zp->z_origin,
zp->z_source);
for (; optind != argc; optind++) {
tm = argv[optind];
if (!inet_aton(tm, &zp->z_addr[zp->z_addrcnt])) {
hp = gethostbyname(tm);
if (hp == NULL) {
syslog(LOG_NOTICE,
"uninterpretable server (%s) for %s\n",
tm, zp->z_origin);
continue;
}
memcpy(&zp->z_addr[zp->z_addrcnt],
hp->h_addr,
INADDRSZ);
dprintf(1, "Arg: \"%s\"\n", tm);
}
if (++zp->z_addrcnt >= NSMAX) {
zp->z_addrcnt = NSMAX;
dprintf(1, "NSMAX reached\n");
break;
}
}
dprintf(1, "addrcnt = %d\n", zp->z_addrcnt);
res_init();
_res.options &= ~(RES_DEFNAMES | RES_DNSRCH | RES_RECURSE);
result = getzone(zp, serial_no, port);
(void) my_fclose(dbfp);
switch (result) {
case XFER_SUCCESS: /* ok exit */
if (rename(tmpname, dbfile) == -1) {
perror("rename");
if (!quiet)
syslog(LOG_ERR, "rename %s to %s: %m",
tmpname, dbfile);
exit(XFER_FAIL);
}
exit(XFER_SUCCESS);
case XFER_UPTODATE: /* the zone was already uptodate */
(void) unlink(tmpname);
exit(XFER_UPTODATE);
default:
result = XFER_FAIL;
/* fall through */
case XFER_TIMEOUT:
case XFER_FAIL:
cleanup_for_exit();
exit(result); /* error or timeout */
}
/*NOTREACHED*/
return (0); /* Make gcc happy. */
}
static char *UsageText[] = {
"\t-z zone_to_transfer\n",
"\t-f db_file\n",
"\t-s serial_no\n",
"\t[-d debug_level]\n",
"\t[-l debug_log_file]\n",
"\t[-t trace_file]\n",
"\t[-p port]\n",
#ifdef STUBS
"\t[-S]\n",
#endif
"\t[-C class]\n",
"\t[-x axfr-src]\n",
"\tservers...\n",
NULL
};
static void
usage(const char *msg) {
char * const *line;
fprintf(stderr, "Usage error: %s\n", msg);
fprintf(stderr, "Usage: %s\n", ProgName);
for (line = UsageText; *line; line++)
fputs(*line, stderr);
exit(XFER_FAIL);
}
#define DEF_DNAME '\001' /* '\0' means the root domain */
/* XXX: The following variables should probably all be "static" */
u_int32_t minimum_ttl = 0;
int soa_cnt = 0;
#ifdef STUBS
int ns_cnt = 0;
#endif
int query_type = 0;
int prev_comment = 0; /* was previous record a comment? */
char zone_top[MAXDNAME]; /* the top of the zone */
char prev_origin[MAXDNAME]; /* from most recent $ORIGIN line */
char prev_dname[MAXDNAME] = { DEF_DNAME }; /* from previous record */
char prev_ns_dname[MAXDNAME] = { DEF_DNAME }; /* from most recent NS record */
static int
getzone(struct zoneinfo *zp, u_int32_t serial_no, int port) {
HEADER *hp;
u_int len;
u_int32_t serial;
int s, n, l, error = 0;
u_int cnt;
u_char *cp, *nmp, *eom, *tmp ;
u_char *buf = NULL;
u_int bufsize = 0;
char name[MAXDNAME], name2[MAXDNAME];
struct sockaddr_in sin;
#ifdef POSIX_SIGNALS
struct sigaction sv, osv;
#else
struct sigvec sv, osv;
#endif
int qdcount, ancount, aucount, class, type;
const char *badsoa_msg = "Nil";
#ifdef DEBUG
if (debug) {
(void)fprintf(ddt,"getzone() %s ", zp->z_origin);
switch (zp->z_type) {
case Z_STUB:
fprintf(ddt,"stub\n");
break;
case Z_SECONDARY:
fprintf(ddt,"secondary\n");
break;
default:
fprintf(ddt,"unknown type\n");
}
}
#endif
#ifdef POSIX_SIGNALS
memset(&sv, 0, sizeof sv);
sv.sa_handler = (SIG_FN (*)()) read_alarm;
/* SA_ONSTACK isn't recommended for strict POSIX code */
/* is it absolutely necessary? */
/* sv.sa_flags = SA_ONSTACK; */
sigfillset(&sv.sa_mask);
(void) sigaction(SIGALRM, &sv, &osv);
memset(&sv, 0, sizeof sv);
sv.sa_handler = (SIG_FN (*)()) term_handler;
sigfillset(&sv.sa_mask);
(void) sigaction(SIGTERM, &sv, &osv);
#else
memset(&sv, 0, sizeof sv);
sv.sv_handler = read_alarm;
sv.sv_mask = ~0;
(void) sigvec(SIGALRM, &sv, &osv);
memset(&sv, 0, sizeof sv);
sv.sv_handler = term_handler;
sv.sv_mask = ~0;
(void) sigvec(SIGTERM, &sv, &osv);
#endif
strcpy(zone_top, zp->z_origin);
if ((l = strlen(zone_top)) != 0 && zone_top[l - 1] == '.')
zone_top[l - 1] = '\0';
strcpy(prev_origin, zone_top);
for (cnt = 0; cnt < zp->z_addrcnt; cnt++) {
curclass = zp->z_class;
error = 0;
if (buf == NULL) {
if ((buf = (u_char *)malloc(2 * PACKETSZ)) == NULL) {
syslog(LOG_INFO, "malloc(%u) failed",
2 * PACKETSZ);
error++;
break;
}
bufsize = 2 * PACKETSZ;
}
if ((s = socket(AF_INET, SOCK_STREAM, PF_UNSPEC)) < 0) {
syslog(LOG_INFO, "socket: %m");
error++;
break;
}
if (zp->z_axfr_src.s_addr != 0) {
memset(&sin, 0, sizeof sin);
sin.sin_family = AF_INET;
sin.sin_port = 0; /* "ANY" */
sin.sin_addr = zp->z_axfr_src;
dprintf(2, "binding to address [%s]\n",
inet_ntoa(sin.sin_addr));
if (bind(s, (struct sockaddr *)&sin, sizeof sin) < 0)
syslog(LOG_INFO, "warning: bind(%s) failed",
inet_ntoa(zp->z_axfr_src));
}
memset(&sin, 0, sizeof sin);
sin.sin_family = AF_INET;
sin.sin_port = port;
sin.sin_addr = zp->z_addr[cnt];
dprintf(2, "connecting to server #%d [%s].%d\n",
cnt+1, inet_ntoa(sin.sin_addr), ntohs(sin.sin_port));
if (connect(s, (struct sockaddr *)&sin, sizeof(sin)) < 0) {
if (!quiet)
syslog(LOG_INFO,
"connect(%s) for zone %s failed: %m",
inet_ntoa(sin.sin_addr), zp->z_origin);
error++;
(void) my_close(s);
continue;
}
n = res_mkquery(QUERY, zp->z_origin, curclass,
T_SOA, NULL, 0, NULL, buf, bufsize);
if (n < 0) {
if (!quiet)
syslog(LOG_INFO,
"zone %s: res_mkquery T_SOA failed",
zp->z_origin);
(void) my_close(s);
#ifdef POSIX_SIGNALS
(void) sigaction(SIGALRM, &osv, (struct sigaction *)0);
#else
(void) sigvec(SIGALRM, &osv, (struct sigvec *)0);
#endif
return (XFER_FAIL);
}
/*
* Send length & message for SOA query
*/
if (writemsg(s, buf, n) < 0) {
syslog(LOG_INFO, "writemsg: %m");
error++;
(void) my_close(s);
continue;
}
/*
* Get out your butterfly net and catch the SOA
*/
if (netread(s, (char *)buf, INT16SZ, XFER_TIMER) < 0) {
error++;
(void) my_close(s);
continue;
}
if ((len = ns_get16(buf)) == 0) {
(void) my_close(s);
continue;
}
if (len > bufsize) {
if ((buf = (u_char *)realloc(buf, len)) == NULL) {
syslog(LOG_INFO,
"malloc(%u) failed for SOA from server [%s], zone %s\n",
len,
inet_ntoa(sin.sin_addr),
zp->z_origin);
(void) my_close(s);
continue;
}
bufsize = len;
}
if (netread(s, (char *)buf, len, XFER_TIMER) < 0) {
error++;
(void) my_close(s);
continue;
}
#ifdef DEBUG
if (debug >= 3) {
(void)fprintf(ddt,"len = %d\n", len);
fp_nquery(buf, len, ddt);
}
#endif
hp = (HEADER *) buf;
qdcount = ntohs(hp->qdcount);
ancount = ntohs(hp->ancount);
aucount = ntohs(hp->nscount);
/*
* close socket if any of these apply:
* 1) rcode != NOERROR
* 2) not an authority response
* 3) not an answer to our question
* 4) both the number of answers and authority count < 1)
*/
if (hp->rcode != NOERROR || !hp->aa || qdcount != 1 ||
(ancount < 1 && aucount < 1)) {
#ifndef ultrix /*XXX*/
syslog(LOG_NOTICE,
"[%s] %s for %s, SOA query got rcode %d, aa %d, ancount %d, aucount %d",
inet_ntoa(sin.sin_addr),
(hp->aa
? (qdcount==1 ?"no SOA found" :"bad response")
: "not authoritative"),
zp->z_origin[0] != '\0' ? zp->z_origin : ".",
hp->rcode, hp->aa, ancount, aucount);
#endif
error++;
(void) my_close(s);
continue;
}
zp_start = *zp;
if ((int)len < HFIXEDSZ + QFIXEDSZ) {
badsoa_msg = "too short";
badsoa:
syslog(LOG_INFO,
"malformed SOA from [%s], zone %s: %s",
inet_ntoa(sin.sin_addr), zp->z_origin,
badsoa_msg);
error++;
(void) my_close(s);
continue;
}
/*
* Step through response.
*/
tmp = buf + HFIXEDSZ;
eom = buf + len;
/* Query Section. */
n = dn_expand(buf, eom, tmp, name2, sizeof name2);
if (n < 0) {
badsoa_msg = "qname error";
goto badsoa;
}
tmp += n;
if (tmp + 2 * INT16SZ > eom) {
badsoa_msg = "query error";
goto badsoa;
}
NS_GET16(type, tmp);
NS_GET16(class, tmp);
if (class != curclass || type != T_SOA ||
strcasecmp(zp->z_origin, name2) != 0) {
syslog(LOG_INFO,
"wrong query in resp from [%s], zone %s: [%s %s %s]\n",
inet_ntoa(sin.sin_addr), zp->z_origin,
name2, p_class(class), p_type(type));
error++;
(void) my_close(s);
continue;
}
/* ... Answer Section.
* We may have to loop a little, to bypass SIG SOA's in
* the response.
*/
do {
u_char *cp4;
u_short type, class, dlen;
u_int32_t ttl;
n = dn_expand(buf, eom, tmp, name2, sizeof name2);
if (n < 0) {
badsoa_msg = "aname error";
goto badsoa;
}
tmp += n;
/* Are type, class, and ttl OK? */
cp4 = tmp; /* Leave tmp pointing to type field */
if (eom - cp4 < 3 * INT16SZ + INT32SZ) {
badsoa_msg = "zinfo too short";
goto badsoa;
}
NS_GET16(type, cp4);
NS_GET16(class, cp4);
NS_GET32(ttl, cp4);
NS_GET16(dlen, cp4);
if (cp4 + dlen > eom) {
badsoa_msg = "zinfo dlen too big";
goto badsoa;
}
if (type == T_SOA)
break;
/* Skip to next record, if any. */
dprintf(1, "skipping %s %s RR in response\n",
name2, p_type(type));
tmp = cp4 + dlen;
} while (1);
if (strcasecmp(zp->z_origin, name2) != 0) {
syslog(LOG_INFO,
"wrong answer in resp from [%s], zone %s: [%s %s %s]\n",
inet_ntoa(sin.sin_addr), zp->z_origin,
name2, p_class(class), p_type(type));
error++;
(void) my_close(s);
continue;
}
badsoa_msg = soa_zinfo(&zp_start, tmp, eom);
if (badsoa_msg)
goto badsoa;
if (SEQ_GT(zp_start.z_serial, serial_no) || !serial_no) {
const char *l, *nl;
dprintf(1, "need update, serial %u\n",
zp_start.z_serial);
hp = (HEADER *) buf;
soa_cnt = 0;
#ifdef STUBS
ns_cnt = 0;
#endif
gettime(&tt);
for (l = Version; l; l = nl) {
size_t len;
if ((nl = strchr(l, '\n')) != NULL) {
len = nl - l;
nl = nl + 1;
} else {
len = strlen(l);
nl = NULL;
}
while (isspace((unsigned char) *l))
l++;
if (*l)
fprintf(dbfp, "; BIND version %.*s\n",
(int)len, l);
}
fprintf(dbfp, "; zone '%s' last serial %u\n",
domain, serial_no);
fprintf(dbfp, "; from %s at %s",
inet_ntoa(sin.sin_addr),
ctimel(tt.tv_sec));
for (;;) {
if ((soa_cnt == 0) || (zp->z_type == Z_STUB)) {
#ifdef STUBS
if (zp->z_type == Z_STUB) {
if (soa_cnt == 1 &&
ns_cnt == 0)
query_type = T_NS;
else
query_type = T_SOA;
} else
#endif
query_type = T_AXFR;
n = res_mkquery(QUERY, zp->z_origin,
curclass, query_type,
NULL, 0,
NULL, buf, bufsize);
if (n < 0) {
if (!quiet) {
#ifdef STUBS
if (zp->z_type == Z_STUB)
syslog(LOG_INFO,
(query_type == T_SOA)
? "zone %s: res_mkquery T_SOA failed"
: "zone %s: res_mkquery T_NS failed",
zp->z_origin);
else
#endif
syslog(LOG_INFO,
"zone %s: res_mkquery T_AXFR failed",
zp->z_origin);
}
(void) my_close(s);
#ifdef POSIX_SIGNALS
sigaction(SIGALRM, &osv,
(struct sigaction *)0);
#else
sigvec(SIGALRM, &osv,
(struct sigvec *)0);
#endif
return (XFER_FAIL);
}
/*
* Send length & msg for zone transfer
*/
if (writemsg(s, buf, n) < 0) {
syslog(LOG_INFO,
"writemsg: %m");
error++;
(void) my_close(s);
break;
}
}
/*
* Receive length & response
*/
if (netread(s, (char *)buf, INT16SZ,
(soa_cnt == 0) ?300 :XFER_TIMER)
< 0) {
error++;
break;
}
if ((len = ns_get16(buf)) == 0)
break;
if (len > bufsize) {
buf = (u_char *)realloc(buf, len);
if (buf == NULL) {
syslog(LOG_INFO,
"malloc(%u) failed for packet from server [%s], zone %s\n",
len,
inet_ntoa(sin.sin_addr),
zp->z_origin);
error++;
break;
}
bufsize = len;
}
hp = (HEADER *)buf;
eom = buf + len;
if (netread(s, (char *)buf, len, XFER_TIMER)
< 0) {
error++;
break;
}
#ifdef DEBUG
if (debug >= 3) {
(void)fprintf(ddt,"len = %d\n", len);
fp_nquery(buf, len, ddt);
}
if (fp)
fp_nquery(buf, len, fp);
#endif
if (len < HFIXEDSZ) {
struct sockaddr_in my_addr;
char my_addr_text[30];
int alen;
badrec:
error++;
alen = sizeof my_addr;
if (getsockname(s, (struct sockaddr *)
&my_addr, &alen) < 0)
sprintf(my_addr_text,
"[errno %d]", errno);
else
sprintf(my_addr_text,
"[%s].%u",
inet_ntoa(my_addr.
sin_addr),
ntohs(my_addr.sin_port)
);
syslog(LOG_INFO,
"[%s] record too short from [%s], zone %s\n",
my_addr_text,
inet_ntoa(sin.sin_addr),
zp->z_origin);
break;
}
cp = buf + HFIXEDSZ;
if (hp->qdcount) {
if ((n = dn_skipname(cp, eom)) == -1
|| n + QFIXEDSZ >= eom - cp)
goto badrec;
cp += n + QFIXEDSZ;
}
nmp = cp;
if ((n = dn_skipname(cp, eom)) == -1)
goto badrec;
tmp = cp + n;
#ifdef STUBS
if (zp->z_type == Z_STUB) {
ancount = ntohs(hp->ancount);
n = 0;
for (cnt = 0; cnt < (u_int)ancount; cnt++) {
n = print_output(zp, serial_no, buf,
len, cp);
if (n < 0)
break;
cp += n;
}
/*
* If we've processed the answer section and
* didn't get any useful answers, bail out.
*/
if (query_type == T_SOA && soa_cnt == 0) {
syslog(LOG_ERR,
"stubs: no SOA in answer");
error++;
break;
}
if (query_type == T_NS && ns_cnt == 0) {
syslog(LOG_ERR,
"stubs: no NS in answer");
error++;
break;
}
if (n >= 0 && hp->nscount) {
ancount = ntohs(hp->nscount);
for (cnt = 0;
cnt < (u_int)ancount;
cnt++) {
n = print_output(zp, serial_no,
buf, len,
cp);
if (n < 0)
break;
cp += n;
}
}
ancount = ntohs(hp->arcount);
for (cnt = 0;
n > 0 && cnt < (u_int)ancount;
cnt++) {
n = print_output(zp, serial_no, buf,
len, cp);
cp += n;
}
if (n < 0) {
syslog(LOG_INFO,
"print_output: unparseable answer (%d), zone %s",
hp->rcode, zp->z_origin);
error++;
break;
}
if (cp != eom) {
syslog(LOG_INFO,
"print_output: short answer (%d, %d), zone %s",
cp - buf, eom - buf,
zp->z_origin);
error++;
break;
}
} else {
#endif /*STUBS*/
ancount = ntohs(hp->ancount);
for (n = cnt = 0;
cnt < (u_int)ancount;
cnt++) {
n = print_output(zp, serial_no, buf,
len, cp);
if (n < 0)
break;
cp += n;
}
if (n < 0) {
syslog(LOG_INFO,
"print_output: unparseable answer (%d), zone %s",
hp->rcode, zp->z_origin);
error++;
break;
}
if (cp != eom) {
syslog(LOG_INFO,
"print_output: short answer (%d, %d), zone %s",
cp - buf, eom - buf,
zp->z_origin);
error++;
break;
}
#ifdef STUBS
}
#endif
if (soa_cnt >= 2)
break;
}
(void) my_close(s);
if (error == 0) {
#ifdef POSIX_SIGNALS
(void) sigaction(SIGALRM, &osv,
(struct sigaction *)0);
#else
(void) sigvec(SIGALRM, &osv, (struct sigvec *)0);
#endif
return (XFER_SUCCESS);
}
dprintf(2, "error receiving zone transfer\n");
} else if (zp_start.z_serial == serial_no) {
(void) my_close(s);
dprintf(1, "zone up-to-date, serial %u\n", zp_start.z_serial);
return (XFER_UPTODATE);
} else {
(void) my_close(s);
if (!quiet)
syslog(LOG_NOTICE,
"serial from [%s], zone %s: %u lower than current: %u\n",
inet_ntoa(sin.sin_addr), zp->z_origin,
zp_start.z_serial, serial_no);
return (XFER_FAIL);
}
}
#ifdef POSIX_SIGNALS
(void) sigaction(SIGALRM, &osv, (struct sigaction *)0);
#else
(void) sigvec(SIGALRM, &osv, (struct sigvec *)0);
#endif
if (!error)
return (XFER_TIMEOUT);
return (XFER_FAIL);
}
static SIG_FN
term_handler() {
cleanup_for_exit();
_exit(XFER_FAIL); /* not safe to call exit() from a signal handler */
}
/*
* Set flag saying to read was interrupted
* used for a read timer
*/
static SIG_FN
read_alarm() {
read_interrupted = 1;
}
static int
netread(int fd, char *buf, int len, int timeout) {
static const char setitimerStr[] = "setitimer: %m";
struct itimerval ival, zeroival;
struct sockaddr_in sa;
int n, salen;
#if defined(NETREAD_BROKEN)
int retries = 0;
#endif
memset(&zeroival, 0, sizeof zeroival);
ival = zeroival;
ival.it_value.tv_sec = timeout;
while (len > 0) {
if (setitimer(ITIMER_REAL, &ival, NULL) < 0) {
syslog(LOG_INFO, setitimerStr);
return (-1);
}
errno = 0;
salen = sizeof sa;
n = recvfrom(fd, buf, len, 0, (struct sockaddr *)&sa, &salen);
if (n == 0 && errno == 0) {
#if defined(NETREAD_BROKEN)
if (++retries < 42) /* doug adams */
continue;
#endif
syslog(LOG_INFO, "premature EOF, fetching \"%s\"",
domain);
return (-1);
}
if (n < 0) {
if (errno == 0) {
#if defined(NETREAD_BROKEN)
if (++retries < 42) /* doug adams */
continue;
#endif
syslog(LOG_INFO,
"recv(len=%d): n=%d && !errno",
len, n);
return (-1);
}
if (errno == EINTR) {
if (!read_interrupted) {
/* It wasn't a timeout; ignore it. */
continue;
}
errno = ETIMEDOUT;
}
syslog(LOG_INFO, "recv(len=%d): %m", len);
return (-1);
}
buf += n;
len -= n;
}
if (setitimer(ITIMER_REAL, &zeroival, NULL) < 0) {
syslog(LOG_INFO, setitimerStr);
return (-1);
}
return (0);
}
/*
* Write a counted buffer to a file descriptor preceded by a length word.
*/
static int
writemsg(int rfd, const u_char *msg, int msglen) {
struct iovec iov[2];
u_char len[INT16SZ];
int ret;
__putshort(msglen, len);
iov[0].iov_base = (char *)len;
iov[0].iov_len = INT16SZ;
iov[1].iov_base = (char *)msg;
iov[1].iov_len = msglen;
ret = writev(rfd, iov, 2);
if (ret != INT16SZ + msglen) {
syslog(LOG_DEBUG, "writemsg(%d,%#x,%d) failed: %s",
rfd, msg, msglen, strerror(errno));
return (-1);
}
return (ret);
}
static const char *
soa_zinfo(struct zoneinfo *zp, u_char *cp, u_char *eom) {
int n, type, class;
u_int32_t ttl;
u_int16_t dlen;
u_char *rdatap;
/* Are type, class, and ttl OK? */
if (eom - cp < 3 * INT16SZ + INT32SZ)
return ("zinfo too short");
NS_GET16(type, cp);
NS_GET16(class, cp);
NS_GET32(ttl, cp);
NS_GET16(dlen, cp);
rdatap = cp;
if (type != T_SOA || class != curclass)
return ("zinfo wrong typ/cla/ttl");
/* Skip master name and contact name, we can't validate them. */
if ((n = dn_skipname(cp, eom)) == -1)
return ("zinfo mname");
cp += n;
if ((n = dn_skipname(cp, eom)) == -1)
return ("zinfo hname");
cp += n;
/* Grab the data fields. */
if (eom - cp < 5 * INT32SZ)
return ("zinfo dlen");
NS_GET32(zp->z_serial, cp);
NS_GET32(zp->z_refresh, cp);
NS_GET32(zp->z_retry, cp);
NS_GET32(zp->z_expire, cp);
NS_GET32(zp->z_minimum, cp);
if (cp != rdatap + dlen)
return ("bad soa dlen");
return (NULL);
}
#define BOUNDS_CHECK(ptr, count) \
do { \
if ((ptr) + (count) > eom) { \
hp->rcode = FORMERR; \
return (-1); \
} \
} while (0)
/*
* Parse the message, determine if it should be printed, and if so, print it
* in .db file form. Does minimal error checking on the message content.
*
* XXX why aren't we using ns_sprintrr() ?
*/
static int
print_output(struct zoneinfo *zp, u_int32_t serial_no, u_char *msg,
int msglen, u_char *rrp) {
u_char *cp;
HEADER *hp = (HEADER *) msg;
u_int32_t addr, ttl, tmpnum;
int i, j, tab, result, n1, n;
u_int class, type, dlen;
char data[MAXDATA];
u_char *cp1, *cp2, *temp_ptr, *eom, *rr_type_ptr;
u_char *cdata, *rdatap;
char *origin, dname[MAXDNAME];
const char *proto;
const char *ignore = "";
const char *badsoa_msg;
int escaped = 0;
eom = msg + msglen;
cp = rrp;
n = dn_expand(msg, msg + msglen, cp, dname, sizeof dname);
if (n < 0) {
hp->rcode = FORMERR;
return (-1);
}
cp += n;
BOUNDS_CHECK(cp, 3 * INT16SZ + INT32SZ);
rr_type_ptr = cp;
NS_GET16(type, cp);
NS_GET16(class, cp);
NS_GET32(ttl, cp);
/*
* Following the Clarification draft's direction, we treat TTLs with
* the MSB set as if they were 0.
*/
if (ttl > MAXIMUM_TTL) {
syslog(LOG_INFO, "%s: TTL > %u, converted to 0", dname,
MAXIMUM_TTL);
ttl = 0;
}
NS_GET16(dlen, cp);
BOUNDS_CHECK(cp, dlen);
rdatap = cp;
origin = dname;
while (*origin) {
if (!escaped && *origin == '.') {
origin++; /* skip over '.' */
break;
}
escaped = (*origin++ == '\\') && !escaped;
}
dprintf(3, "print_output: dname %s type %d class %d ttl %u\n",
dname, type, class, ttl);
/*
* Convert the resource record data into the internal database format.
* CP points to the raw resource record.
* After this switch:
* CP has been updated to point past the RR.
* CP1 points to the internal database version.
* N is the length of the internal database version.
*/
switch (type) {
case T_A:
case T_WKS:
case T_HINFO:
case T_TXT:
case T_X25:
case T_ISDN:
case T_LOC:
case T_NSAP:
case T_AAAA:
case T_KEY:
cp1 = cp;
n = dlen;
cp += n;
break;
case T_CNAME:
case T_MB:
case T_MG:
case T_MR:
case T_NS:
case T_PTR:
n = dn_expand(msg, msg + msglen, cp, data, sizeof data);
if (n < 0) {
hp->rcode = FORMERR;
return (-1);
}
cp += n;
cp1 = (u_char *)data;
n = strlen(data) + 1;
break;
case T_MINFO:
case T_SOA:
case T_RP:
n = dn_expand(msg, msg + msglen, cp, data, sizeof data);
if (n < 0) {
hp->rcode = FORMERR;
return (-1);
}
cp += n;
n = strlen(data) + 1;
cp1 = (u_char *)data + n;
n1 = sizeof data - n;
if (type == T_SOA)
n1 -= 5 * INT32SZ;
n = dn_expand(msg, msg + msglen, cp, (char *)cp1, n1);
if (n < 0) {
hp->rcode = FORMERR;
return (-1);
}
cp += n;
cp1 += strlen((char *) cp1) + 1;
if (type == T_SOA) {
BOUNDS_CHECK(cp, 5 * INT32SZ);
temp_ptr = cp + 4 * INT32SZ;
NS_GET32(minimum_ttl, temp_ptr);
/*
* Following the Clarification draft's direction,
* we treat TTLs with the MSB set as if they were 0.
*/
if (minimum_ttl > MAXIMUM_TTL) {
syslog(LOG_INFO,
"%s: SOA minimum TTL > %u, converted to 0",
dname, MAXIMUM_TTL);
minimum_ttl = 0;
}
n = 5 * INT32SZ;
memcpy(cp1, cp, n);
cp += n;
cp1 += n;
}
n = cp1 - (u_char *)data;
cp1 = (u_char *)data;
break;
case T_NAPTR:
/* Grab weight and port. */
BOUNDS_CHECK(cp, INT16SZ*2);
memcpy(data, cp, INT16SZ*2);
cp1 = (u_char *)data + INT16SZ*2;
cp += INT16SZ*2;
/* Flags */
BOUNDS_CHECK(cp, 1);
n = *cp++;
BOUNDS_CHECK(cp, n);
*cp1++ = n;
memcpy(cp1, cp, n);
cp += n; cp1 += n;
/* Service */
BOUNDS_CHECK(cp, 1);
n = *cp++;
BOUNDS_CHECK(cp, n);
*cp1++ = n;
memcpy(cp1, cp, n);
cp += n; cp1 += n;
/* Regexp */
BOUNDS_CHECK(cp, 1);
n = *cp++;
BOUNDS_CHECK(cp, n);
*cp1++ = n;
memcpy(cp1, cp, n);
cp += n; cp1 += n;
/* Replacement */
n = dn_expand(msg, msg + msglen, cp, (char *)cp1,
sizeof data - ((char *)cp1 - data));
if (n < 0)
return (-1);
cp += n;
/* compute end of data */
cp1 += strlen((char *)cp1) + 1;
/* compute size of data */
n = cp1 - (u_char *)data;
cp1 = (u_char *)data;
break;
case T_MX:
case T_AFSDB:
case T_RT:
case T_SRV:
/* grab preference */
BOUNDS_CHECK(cp, INT16SZ);
memcpy(data, cp, INT16SZ);
cp1 = (u_char *)data + INT16SZ;
cp += INT16SZ;
if (type == T_SRV) {
BOUNDS_CHECK(cp, INT16SZ*2);
memcpy(cp1, cp, INT16SZ*2);
cp1 += INT16SZ*2;
cp += INT16SZ*2;
}
/* get name */
n = dn_expand(msg, msg + msglen, cp,
(char *)cp1,
sizeof data - (cp1 - (u_char *)data));
if (n < 0)
return (-1);
cp += n;
/* compute end of data */
cp1 += strlen((char *) cp1) + 1;
/* compute size of data */
n = cp1 - (u_char *)data;
cp1 = (u_char *)data;
break;
case T_PX:
/* grab preference */
BOUNDS_CHECK(cp, INT16SZ);
memcpy(data, cp, INT16SZ);
cp1 = (u_char *)data + INT16SZ;
cp += INT16SZ;
/* get MAP822 name */
n = dn_expand(msg, msg + msglen, cp,
(char *)cp1, sizeof data - INT16SZ);
if (n < 0)
return (-1);
cp += n;
cp1 += (n = (strlen((char *) cp1) + 1));
n1 = sizeof data - n;
/* get MAPX400 name */
n = dn_expand(msg, msg + msglen, cp, (char *)cp1, n1);
if (n < 0)
return (-1);
cp += n;
cp1 += strlen((char *) cp1) + 1;
n = cp1 - (u_char *)data;
cp1 = (u_char *)data;
break;
case T_SIG:
/* CP is the raw resource record as it arrived.
* CP1, after this switch, points to the internal database version. */
cp1 = (u_char *)data;
/* first just copy over the type_covered, algorithm, */
/* labels, orig ttl, two timestamps, and the footprint */
BOUNDS_CHECK(cp, NS_SIG_SIGNER);
memcpy(cp1, cp, NS_SIG_SIGNER);
cp += NS_SIG_SIGNER;
cp1 += NS_SIG_SIGNER;
/* then the signer's name */
n = dn_expand(msg, msg + msglen, cp,
(char *)cp1, (sizeof data) - 18);
if (n < 0)
return (-1);
cp += n;
cp1 += strlen((char*)cp1)+1;
/* finally, we copy over the variable-length signature.
Its size is the total data length, minus what we copied. */
n = dlen - (NS_SIG_SIGNER + n);
if (n > ((int)(sizeof data) - (int)(cp1 - (u_char *)data))) {
hp->rcode = FORMERR;
return (-1); /* out of room! */
}
memcpy(cp1, cp, n);
cp += n;
cp1 += n;
/* compute size of data */
n = cp1 - (u_char *)data;
cp1 = (u_char *)data;
break;
case T_NXT:
n = dn_expand(msg, msg + msglen, cp,
(char *)data, sizeof data);
if (n < 0) {
hp->rcode = FORMERR;
return (-1);
}
cp += n;
cp1 = (u_char *)data + strlen(data) + 1;
n = dlen - n;
if (n > ((int)(sizeof data) - (int)(cp1 - (u_char *)data))) {
hp->rcode = FORMERR;
return (-1); /* out of room! */
}
if (n > 0) { /* Actually, n should never be less than 4 */
memcpy(cp1, cp, n);
cp += n;
} else {
hp->rcode = FORMERR;
return (-1);
}
n += cp1 - (u_char *)data;
cp1 = (u_char *)data;
break;
default:
syslog(LOG_INFO, "\"%s %s %s\" - unknown type (%d)",
dname, p_class(class), p_type(type), type);
hp->rcode = NOTIMP;
return (-1);
}
if (n > MAXDATA) {
dprintf(1, "update type %d: %d bytes is too much data\n",
type, n);
hp->rcode = FORMERR;
return (-1);
}
if (cp != rdatap + dlen) {
dprintf(1,
"encoded rdata length is %u, but actual length was %u\n",
dlen, (u_int)(cp - rdatap));
hp->rcode = FORMERR;
return (-1);
}
cdata = cp1;
result = cp - rrp;
/*
* Special handling for SOA records.
*/
if (type == T_SOA) {
if (strcasecmp(dname, zp->z_origin) != 0) {
syslog(LOG_INFO,
"wrong zone name in AXFR (wanted \"%s\", got \"%s\")",
zp->z_origin, dname);
hp->rcode = FORMERR;
return (-1);
}
if (!soa_cnt) {
badsoa_msg = soa_zinfo(&zp_start, rr_type_ptr, eom);
if (badsoa_msg) {
syslog(LOG_INFO,
"malformed SOA for zone %s: %s",
zp->z_origin, badsoa_msg);
hp->rcode = FORMERR;
return (-1);
}
if (SEQ_GT(zp_start.z_serial, serial_no) ||
!serial_no)
soa_cnt++;
else {
syslog(LOG_INFO,
"serial went backwards after transfer started");
return (-1);
}
} else {
badsoa_msg = soa_zinfo(&zp_finish, rr_type_ptr, eom);
if (badsoa_msg) {
syslog(LOG_INFO,
"malformed SOA for zone %s: %s",
zp->z_origin, badsoa_msg);
hp->rcode = FORMERR;
return (-1);
}
dprintf(2, "SOA, serial %u\n", zp_finish.z_serial);
if (zp_start.z_serial != zp_finish.z_serial) {
dprintf(1, "serial changed, restart\n");
restarts++;
if (restarts > MAX_XFER_RESTARTS) {
syslog(LOG_INFO,
"too many transfer restarts for zone %s",
zp->z_origin);
hp->rcode = FORMERR;
return (-1);
}
soa_cnt = 0;
#ifdef STUBS
ns_cnt = 0;
#endif
minimum_ttl = 0;
strcpy(prev_origin, zp->z_origin);
prev_dname[0] = DEF_DNAME;
/*
* Flush buffer, truncate file
* and seek to beginning to restart.
*/
fflush(dbfp);
if (ftruncate(fileno(dbfp), 0) != 0) {
if (!quiet)
syslog(LOG_INFO,
"ftruncate %s: %m\n",
tmpname);
return (-1);
}
fseek(dbfp, 0L, 0);
return (result);
}
soa_cnt++;
return (result);
}
}
#ifdef STUBS
if (zp->z_type == Z_STUB) {
if (query_type == T_NS && type == T_NS)
ns_cnt++;
/*
* If we're processing a response to an SOA query, we don't
* want to print anything from the response except for the SOA.
* We do want to check everything in the packet, which is
* why we do this check now instead of earlier.
*/
if (query_type == T_SOA && type != T_SOA)
return (result);
}
#endif
if (!soa_cnt || soa_cnt >= 2) {
char *gripe;
if (!soa_cnt)
gripe = "got RR before first SOA";
else
gripe = "got RR after second SOA";
syslog(LOG_INFO, "%s in zone %s", gripe, zp->z_origin);
hp->rcode = FORMERR;
return (-1);
}
/*
* If they are trying to tell us info about something that is
* not in the zone that we are transfering, then ignore it!
* They don't have the authority to tell us this info.
*
* We have to do a bit of checking here - the name that we are
* checking is is fully qualified & may be in a subdomain of the
* zone in question. We also need to ignore any final dots.
*
* If a domain has both NS records and non-NS records, (for
* example, NS and MX records), then we should ignore the non-NS
* records (except that we should not ignore glue A records).
* XXX: It is difficult to do this properly, so we just compare
* the current dname with that in the most recent NS record.
* This defends against the most common error case,
* where the remote server sends MX records soon after the
* NS records for a particular domain. If sent earlier, we lose. XXX
*/
if (!samedomain(dname, domain)) {
(void) fprintf(dbfp, "; Ignoring info about %s, not in zone %s.\n",
dname, domain);
ignore = "; ";
} else if (type != T_NS && type != T_A &&
strcasecmp(zone_top, dname) != 0 &&
strcasecmp(prev_ns_dname, dname) == 0)
{
(void) fprintf(dbfp, "; Ignoring extra info about %s, invalid after NS delegation.\n",
dname);
ignore = "; ";
}
/*
* If the current record is not being ignored, but the
* previous record was ignored, then we invalidate information
* that might have been altered by ignored records.
* (This means that we sometimes output unnecessary $ORIGIN
* lines, but that is harmless.)
*
* Also update prev_comment now.
*/
if (prev_comment && ignore[0] == '\0') {
prev_dname[0] = DEF_DNAME;
prev_origin[0] = DEF_DNAME;
}
prev_comment = (ignore[0] != '\0');
/*
* set prev_ns_dname if necessary
*/
if (type == T_NS) {
(void) strcpy(prev_ns_dname, dname);
}
/*
* If the origin has changed, print the new origin
*/
if (strcasecmp(prev_origin, origin)) {
(void) strcpy(prev_origin, origin);
(void) fprintf(dbfp, "%s$ORIGIN %s.\n", ignore, origin);
}
tab = 0;
if (strcasecmp(prev_dname, dname)) {
/*
* set the prev_dname to be the current dname, then cut off all
* characters of dname after (and including) the first '.'
*/
char *cutp;
(void) strcpy(prev_dname, dname);
escaped = 0;
cutp = dname;
while (*cutp) {
if (!escaped && *cutp == '.')
break;
escaped = (*cutp++ == '\\') && !escaped;
}
*cutp = '\0';
if (dname[0] == 0) {
if (origin[0] == 0)
(void) fprintf(dbfp, "%s.\t", ignore);
else
(void) fprintf(dbfp, "%s.%s.\t",
ignore, origin); /* ??? */
} else {
char *backslash;
backslash = (*dname == '@' || *dname == '$') ?
"\\" : "";
(void) fprintf(dbfp, "%s%s%s\t", ignore,
backslash, dname);
}
if (strlen(dname) < (size_t)8)
tab = 1;
} else {
(void) fprintf(dbfp, "%s\t", ignore);
tab = 1;
}
if (ttl != minimum_ttl)
(void) fprintf(dbfp, "%d\t", (int) ttl);
else if (tab)
(void) putc('\t', dbfp);
(void) fprintf(dbfp, "%s\t%s\t", p_class(class), p_type(type));
cp = cdata;
/*
* Print type specific data
*/
switch (type) {
case T_A:
switch (class) {
case C_IN:
case C_HS:
fputs(inet_ntoa(ina_get(cp)), dbfp);
break;
}
(void) fprintf(dbfp, "\n");
break;
case T_CNAME:
case T_MB:
case T_MG:
case T_MR:
case T_PTR:
if (cp[0] == '\0')
(void) fprintf(dbfp, ".\n");
else
(void) fprintf(dbfp, "%s.\n", cp);
break;
case T_NS:
cp = cdata;
if (cp[0] == '\0')
(void) fprintf(dbfp, ".\t");
else
(void) fprintf(dbfp, "%s.", cp);
(void) fprintf(dbfp, "\n");
break;
case T_HINFO:
case T_ISDN:
cp2 = cp + n;
for (i = 0; i < 2; i++) {
if (i != 0)
(void) putc(' ', dbfp);
n = *cp++;
cp1 = cp + n;
if (cp1 > cp2)
cp1 = cp2;
(void) putc('"', dbfp);
j = 0;
while (cp < cp1) {
if (*cp == '\0') {
cp = cp1;
break;
}
if (strchr("\n\"\\", *cp))
(void) putc('\\', dbfp);
(void) putc(*cp++, dbfp);
j++;
}
if (j == 0 && (type != T_ISDN || i == 0))
(void) putc('?', dbfp);
(void) putc('"', dbfp);
}
(void) putc('\n', dbfp);
break;
case T_SOA:
(void) fprintf(dbfp, "%s.", cp);
cp += strlen((char *) cp) + 1;
(void) fprintf(dbfp, " %s. (\n", cp);
cp += strlen((char *) cp) + 1;
NS_GET32(tmpnum, cp);
(void) fprintf(dbfp, "%s\t\t%u", ignore, tmpnum);
NS_GET32(tmpnum, cp);
(void) fprintf(dbfp, " %u", tmpnum);
NS_GET32(tmpnum, cp);
(void) fprintf(dbfp, " %u", tmpnum);
NS_GET32(tmpnum, cp);
(void) fprintf(dbfp, " %u", tmpnum);
NS_GET32(tmpnum, cp);
(void) fprintf(dbfp, " %u )\n", tmpnum);
break;
case T_MX:
case T_AFSDB:
case T_RT:
NS_GET16(tmpnum, cp);
(void) fprintf(dbfp, "%u", tmpnum);
(void) fprintf(dbfp, " %s.\n", cp);
break;
case T_PX:
NS_GET16(tmpnum, cp);
(void) fprintf(dbfp, "%u", tmpnum);
(void) fprintf(dbfp, " %s.", cp);
cp += strlen((char *) cp) + 1;
(void) fprintf(dbfp, " %s.\n", cp);
break;
case T_TXT:
case T_X25:
cp1 = cp + n;
while (cp < cp1) {
(void) putc('"', dbfp);
if ((i = *cp++) != 0) {
for (j = i; j > 0 && cp < cp1; j--) {
if (strchr("\n\"\\", *cp))
(void) putc('\\', dbfp);
(void) putc(*cp++, dbfp);
}
}
(void) putc('"', dbfp);
if (cp < cp1)
(void) putc(' ', dbfp);
}
(void) putc('\n', dbfp);
break;
case T_NSAP:
fprintf(dbfp, "%s\n", inet_nsap_ntoa(n, cp, NULL));
break;
case T_AAAA: {
char t[sizeof "ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255"];
fprintf(dbfp, "%s\n", inet_ntop(AF_INET6, cp, t, sizeof t));
break;
}
case T_LOC: {
char t[255];
(void) fprintf(dbfp, "%s\n", loc_ntoa(cp, t));
break;
}
case T_NAPTR: {
u_int32_t order, preference;
/* Order */
NS_GET16(order, cp);
fprintf(dbfp, "%u", order);
/* Preference */
NS_GET16(preference, cp);
fprintf(dbfp, " %u", preference);
/* Flags */
if ((n = *cp++) != 0) {
fprintf(dbfp, " \"%.*s\"", (int)n, cp);
cp += n;
}
/* Service */
if ((n = *cp++) != 0) {
fprintf(dbfp, " \"%.*s\"", (int)n, cp);
cp += n;
}
/* Regexp */
if ((n = *cp++) != 0) {
fprintf(dbfp, " \"%.*s\"", (int)n, cp);
cp += n;
}
/* Replacement */
fprintf(dbfp, " %s.\n", cp);
break;
}
case T_SRV: {
u_int priority, weight, port;
NS_GET16(priority, cp);
NS_GET16(weight, cp);
NS_GET16(port, cp);
fprintf(dbfp, "\t%u %u %u %s.\n",
priority, weight, port, cp);
break;
}
case T_WKS:
fputs(inet_ntoa(ina_get(cp)), dbfp);
cp += INADDRSZ;
fputc(' ', dbfp);
proto = protocolname(*cp);
cp += sizeof(char);
(void) fprintf(dbfp, "%s ", proto);
i = 0;
while (cp < cdata + n) {
j = *cp++;
do {
if (j & 0200)
(void) fprintf(dbfp, " %s",
servicename(i, proto));
j <<= 1;
} while (++i & 07);
}
(void) fprintf(dbfp, "\n");
break;
case T_MINFO:
case T_RP:
(void) fprintf(dbfp, "%s.", cp);
cp += strlen((char *) cp) + 1;
(void) fprintf(dbfp, " %s.\n", cp);
break;
case T_KEY: {
char databuf[16+NS_MD5RSA_MAX_BASE64]; /* 16 for slop */
u_int keyflags;
/* get & format key flags */
keyflags = ns_get16(cp);
(void) fprintf(dbfp, "0x%04x ", keyflags);
cp += INT16SZ;
/* protocol id */
(void) fprintf(dbfp, " %u", *cp++);
/* algorithm id */
(void) fprintf(dbfp, " %u ", *cp++);
/* key itself (which may have zero length) */
n = b64_ntop(cp, (cp1 + n) - cp, databuf, sizeof databuf);
if (n < 0)
fprintf(dbfp, "; BAD BASE64\n");
else
fprintf(dbfp, "%s\n", databuf);
break;
}
case T_SIG: {
char databuf[16+NS_MD5RSA_MAX_BASE64]; /* 16 for slop */
/* get & format rr type which signature covers */
(void) fprintf(dbfp,"%s", p_type(ns_get16((u_char*)cp)));
cp += INT16SZ;
/* algorithm id */
(void) fprintf(dbfp," %d",*cp++);
/* labels (# of labels in name) - skip in textual record */
cp++;
/* orig time to live (TTL)) */
(void) fprintf(dbfp," %u", (u_int32_t)ns_get32((u_char*)cp));
cp += INT32SZ;
/* expiration time */
(void) fprintf(dbfp," %s", p_secstodate(ns_get32((u_char*)cp)));
cp += INT32SZ;
/* time signed */
(void) fprintf(dbfp," %s", p_secstodate(ns_get32((u_char*)cp)));
cp += INT32SZ;
/* Key footprint */
(void) fprintf(dbfp," %d", ns_get16((u_char*)cp));
cp += INT16SZ;
/* signer's name */
(void) fprintf(dbfp, " %s. ", cp);
cp += strlen((char *) cp) + 1;
/* signature itself */
n = b64_ntop(cp, (cdata + n) - cp, databuf, sizeof databuf);
if (n < 0)
fprintf (dbfp, "; BAD BASE64\n");
else
fprintf (dbfp, "%s\n", databuf);
break;
}
case T_NXT:
fprintf(dbfp, "%s.", (char *)cp);
i = strlen((char *)cp)+1;
cp += i;
n -= i;
for (i=0; i < n*NS_NXT_BITS; i++)
if (NS_NXT_BIT_ISSET (i, cp))
fprintf(dbfp, " %s", p_type(i));
fprintf(dbfp,"\n");
break;
default:
cp1 = cp + n;
while (cp < cp1)
fprintf(dbfp, "0x%02.2X ", *cp++ & 0xFF);
(void) fprintf(dbfp, "???\n");
}
if (ferror(dbfp)) {
syslog(LOG_ERR, "%s: %m", tmpname);
cleanup_for_exit();
exit(XFER_FAIL);
}
return (result);
}
#ifdef SHORT_FNAMES
/*
** This routine handles creating temporary files with mkstemp
** in the presence of a 14 char filename system. Pathconf()
** does not work over NFS.
*/
filenamecpy(char *ddtfile, char *optarg) {
int namelen, extra, len;
char *dirname, *filename;
/* determine the length of filename allowed */
if((dirname = strrchr(optarg, '/')) == NULL){
filename = optarg;
} else {
*dirname++ = '\0';
filename = dirname;
}
namelen = pathconf(dirname == NULL? "." : optarg, _PC_NAME_MAX);
if(namelen <= 0)
namelen = 255; /* length could not be determined */
if(dirname != NULL)
*--dirname = '/';
/* copy a shorter name if it will be longer than allowed */
extra = (strlen(filename)+strlen(".XXXXXX")) - namelen;
if(extra > 0){
len = strlen(optarg) - extra;
(void) strncpy(ddtfile, optarg, len);
ddtfile[len] = '\0';
} else
(void) strcpy(ddtfile, optarg);
}
#endif /* SHORT_FNAMES */