NetBSD/usr.sbin/lpr/lpd/lpd.c

773 lines
18 KiB
C

/* $NetBSD: lpd.c,v 1.28 2001/03/28 03:17:42 simonb Exp $ */
/*
* Copyright (c) 1983, 1993, 1994
* 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.
*/
#include <sys/cdefs.h>
#ifndef lint
__COPYRIGHT("@(#) Copyright (c) 1983, 1993, 1994\n\
The Regents of the University of California. All rights reserved.\n");
#endif /* not lint */
#ifndef lint
#if 0
static char sccsid[] = "@(#)lpd.c 8.7 (Berkeley) 5/10/95";
#else
__RCSID("$NetBSD: lpd.c,v 1.28 2001/03/28 03:17:42 simonb Exp $");
#endif
#endif /* not lint */
/*
* lpd -- line printer daemon.
*
* Listen for a connection and perform the requested operation.
* Operations are:
* \1printer\n
* check the queue for jobs and print any found.
* \2printer\n
* receive a job from another machine and queue it.
* \3printer [users ...] [jobs ...]\n
* return the current state of the queue (short form).
* \4printer [users ...] [jobs ...]\n
* return the current state of the queue (long form).
* \5printer person [users ...] [jobs ...]\n
* remove jobs from the queue.
*
* Strategy to maintain protected spooling area:
* 1. Spooling area is writable only by daemon and spooling group
* 2. lpr runs setuid root and setgrp spooling group; it uses
* root to access any file it wants (verifying things before
* with an access call) and group id to know how it should
* set up ownership of files in the spooling area.
* 3. Files in spooling area are owned by root, group spooling
* group, with mode 660.
* 4. lpd, lpq and lprm run setuid daemon and setgrp spooling group to
* access files and printer. Users can't get to anything
* w/o help of lpq and lprm programs.
*/
#include <sys/param.h>
#include <sys/wait.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <sys/un.h>
#include <sys/stat.h>
#include <sys/file.h>
#include <netinet/in.h>
#include <err.h>
#include <netdb.h>
#include <unistd.h>
#include <syslog.h>
#include <signal.h>
#include <errno.h>
#include <fcntl.h>
#include <dirent.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <ctype.h>
#include <arpa/inet.h>
#include "lp.h"
#include "lp.local.h"
#include "pathnames.h"
#include "extern.h"
/* XXX from libc/net/rcmd.c */
extern int __ivaliduser_sa __P((FILE *, struct sockaddr *, socklen_t,
const char *, const char *));
int lflag; /* log requests flag */
int rflag; /* allow of for remote printers */
int sflag; /* secure (no inet) flag */
int from_remote; /* from remote socket */
char **blist; /* list of addresses to bind(2) to */
int blist_size;
int blist_addrs;
int main __P((int, char **));
static void reapchild __P((int));
static void mcleanup __P((int));
static void doit __P((void));
static void startup __P((void));
static void chkhost __P((struct sockaddr *));
static int ckqueue __P((char *));
static void usage __P((void));
static int *socksetup __P((int, int, const char *));
uid_t uid, euid;
int child_count;
int
main(argc, argv)
int argc;
char **argv;
{
fd_set defreadfds;
struct sockaddr_un un, fromunix;
struct sockaddr_storage frominet;
sigset_t nmask, omask;
int lfd, errs, i, f, funix, *finet;
int child_max = 32; /* more then enough to hose the system */
int options = 0;
struct servent *sp;
const char *port = "printer";
euid = geteuid(); /* these shouldn't be different */
uid = getuid();
gethostname(host, sizeof(host));
host[sizeof(host) - 1] = '\0';
name = argv[0];
errs = 0;
while ((i = getopt(argc, argv, "b:dln:srw:")) != -1)
switch (i) {
case 'b':
if (blist_addrs >= blist_size) {
blist_size += sizeof(char *) * 4;
if (blist == NULL)
blist = malloc(blist_size);
else
blist = realloc(blist, blist_size);
if (blist == NULL)
err(1, "cant allocate bind addr list");
}
blist[blist_addrs++] = strdup(optarg);
break;
case 'd':
options |= SO_DEBUG;
break;
case 'l':
lflag++;
break;
case 'n':
child_max = atoi(optarg);
if (child_max < 0 || child_max > 1024)
errx(1, "invalid number of children: %s",
optarg);
break;
case 'r':
rflag++;
break;
case 's':
sflag++;
break;
case 'w':
wait_time = atoi(optarg);
if (wait_time < 0)
errx(1, "wait time must be postive: %s",
optarg);
if (wait_time < 30)
warnx("warning: wait time less than 30 seconds");
break;
default:
errs++;
}
argc -= optind;
argv += optind;
if (errs)
usage();
switch (argc) {
case 1:
if ((i = atoi(argv[0])) == 0)
usage();
if (i < 0 || i > USHRT_MAX)
errx(1, "port # %d is invalid", i);
port = argv[0];
break;
case 0:
sp = getservbyname(port, "tcp");
if (sp == NULL)
errx(1, "%s/tcp: unknown service", port);
break;
default:
usage();
}
#ifndef DEBUG
/*
* Set up standard environment by detaching from the parent.
*/
daemon(0, 0);
#endif
openlog("lpd", LOG_PID, LOG_LPR);
syslog(LOG_INFO, "restarted");
(void)umask(0);
lfd = open(_PATH_MASTERLOCK, O_WRONLY|O_CREAT, 0644);
if (lfd < 0) {
syslog(LOG_ERR, "%s: %m", _PATH_MASTERLOCK);
exit(1);
}
if (flock(lfd, LOCK_EX|LOCK_NB) < 0) {
if (errno == EWOULDBLOCK) /* active deamon present */
exit(0);
syslog(LOG_ERR, "%s: %m", _PATH_MASTERLOCK);
exit(1);
}
ftruncate(lfd, 0);
/*
* write process id for others to know
*/
(void)snprintf(line, sizeof(line), "%u\n", getpid());
f = strlen(line);
if (write(lfd, line, f) != f) {
syslog(LOG_ERR, "%s: %m", _PATH_MASTERLOCK);
exit(1);
}
signal(SIGCHLD, reapchild);
/*
* Restart all the printers.
*/
startup();
(void)unlink(_PATH_SOCKETNAME);
funix = socket(AF_LOCAL, SOCK_STREAM, 0);
if (funix < 0) {
syslog(LOG_ERR, "socket: %m");
exit(1);
}
sigemptyset(&nmask);
sigaddset(&nmask, SIGHUP);
sigaddset(&nmask, SIGINT);
sigaddset(&nmask, SIGQUIT);
sigaddset(&nmask, SIGTERM);
sigprocmask(SIG_BLOCK, &nmask, &omask);
(void) umask(07);
signal(SIGHUP, mcleanup);
signal(SIGINT, mcleanup);
signal(SIGQUIT, mcleanup);
signal(SIGTERM, mcleanup);
memset(&un, 0, sizeof(un));
un.sun_family = AF_LOCAL;
strncpy(un.sun_path, _PATH_SOCKETNAME, sizeof(un.sun_path) - 1);
#ifndef SUN_LEN
#define SUN_LEN(unp) (strlen((unp)->sun_path) + 2)
#endif
if (bind(funix, (struct sockaddr *)&un, SUN_LEN(&un)) < 0) {
syslog(LOG_ERR, "ubind: %m");
exit(1);
}
(void) umask(0);
sigprocmask(SIG_SETMASK, &omask, (sigset_t *)0);
FD_ZERO(&defreadfds);
FD_SET(funix, &defreadfds);
listen(funix, 5);
if (!sflag || blist_addrs)
finet = socksetup(PF_UNSPEC, options, port);
else
finet = NULL; /* pretend we couldn't open TCP socket. */
if (blist != NULL) {
for (i = 0; i < blist_addrs; i++)
free(blist[i]);
free(blist);
}
if (finet) {
for (i = 1; i <= *finet; i++) {
FD_SET(finet[i], &defreadfds);
listen(finet[i], 5);
}
}
/*
* Main loop: accept, do a request, continue.
*/
memset(&frominet, 0, sizeof(frominet));
memset(&fromunix, 0, sizeof(fromunix));
for (;;) {
int domain, nfds, s, fromlen;
fd_set readfds;
/* "short" so it overflows in about 2 hours */
short sleeptime = 10;
while (child_max < child_count) {
syslog(LOG_WARNING,
"too many children, sleeping for %d seconds",
sleeptime);
sleep(sleeptime);
sleeptime <<= 1;
if (sleeptime < 0) {
syslog(LOG_CRIT, "sleeptime overflowed! help!");
sleeptime = 10;
}
}
FD_COPY(&defreadfds, &readfds);
nfds = select(20, &readfds, 0, 0, 0);
if (nfds <= 0) {
if (nfds < 0 && errno != EINTR)
syslog(LOG_WARNING, "select: %m");
continue;
}
if (FD_ISSET(funix, &readfds)) {
domain = AF_LOCAL;
fromlen = sizeof(fromunix);
s = accept(funix,
(struct sockaddr *)&fromunix, &fromlen);
} else {
for (i = 1; i <= *finet; i++)
if (FD_ISSET(finet[i], &readfds)) {
domain = AF_INET, fromlen = sizeof(frominet);
s = accept(finet[i], (struct sockaddr *)&frominet, &fromlen);
}
}
if (s < 0) {
if (errno != EINTR)
syslog(LOG_WARNING, "accept: %m");
continue;
}
switch (fork()) {
case 0:
signal(SIGCHLD, SIG_IGN);
signal(SIGHUP, SIG_IGN);
signal(SIGINT, SIG_IGN);
signal(SIGQUIT, SIG_IGN);
signal(SIGTERM, SIG_IGN);
(void)close(funix);
if (!sflag && finet)
for (i = 1; i <= *finet; i++)
(void)close(finet[i]);
dup2(s, 1);
(void)close(s);
if (domain == AF_INET) {
/* for both AF_INET and AF_INET6 */
from_remote = 1;
chkhost((struct sockaddr *)&frominet);
} else
from_remote = 0;
doit();
exit(0);
case -1:
syslog(LOG_WARNING, "fork: %m, sleeping for 10 seconds...");
sleep(10);
continue;
default:
child_count++;
}
(void)close(s);
}
}
static void
reapchild(signo)
int signo;
{
union wait status;
while (wait3((int *)&status, WNOHANG, 0) > 0)
child_count--;
}
static void
mcleanup(signo)
int signo;
{
if (lflag)
syslog(LOG_INFO, "exiting");
unlink(_PATH_SOCKETNAME);
exit(0);
}
/*
* Stuff for handling job specifications
*/
char *user[MAXUSERS]; /* users to process */
int users; /* # of users in user array */
int requ[MAXREQUESTS]; /* job number of spool entries */
int requests; /* # of spool requests */
char *person; /* name of person doing lprm */
char fromb[NI_MAXHOST]; /* buffer for client's machine name */
char cbuf[BUFSIZ]; /* command line buffer */
char *cmdnames[] = {
"null",
"printjob",
"recvjob",
"displayq short",
"displayq long",
"rmjob"
};
static void
doit()
{
char *cp;
int n;
for (;;) {
cp = cbuf;
do {
if (cp >= &cbuf[sizeof(cbuf) - 1])
fatal("Command line too long");
if ((n = read(1, cp, 1)) != 1) {
if (n < 0)
fatal("Lost connection");
return;
}
} while (*cp++ != '\n');
*--cp = '\0';
cp = cbuf;
if (lflag) {
if (*cp >= '\1' && *cp <= '\5') {
syslog(LOG_INFO, "%s requests %s %s",
from, cmdnames[(int)*cp], cp+1);
setproctitle("serving %s: %s %s", from,
cmdnames[(int)*cp], cp+1);
}
else
syslog(LOG_INFO, "bad request (%d) from %s",
*cp, from);
}
switch (*cp++) {
case '\1': /* check the queue and print any jobs there */
printer = cp;
printjob();
break;
case '\2': /* receive files to be queued */
if (!from_remote) {
syslog(LOG_INFO, "illegal request (%d)", *cp);
exit(1);
}
printer = cp;
recvjob();
break;
case '\3': /* display the queue (short form) */
case '\4': /* display the queue (long form) */
printer = cp;
while (*cp) {
if (*cp != ' ') {
cp++;
continue;
}
*cp++ = '\0';
while (isspace(*cp))
cp++;
if (*cp == '\0')
break;
if (isdigit(*cp)) {
if (requests >= MAXREQUESTS)
fatal("Too many requests");
requ[requests++] = atoi(cp);
} else {
if (users >= MAXUSERS)
fatal("Too many users");
user[users++] = cp;
}
}
displayq(cbuf[0] - '\3');
exit(0);
case '\5': /* remove a job from the queue */
if (!from_remote) {
syslog(LOG_INFO, "illegal request (%d)", *cp);
exit(1);
}
printer = cp;
while (*cp && *cp != ' ')
cp++;
if (!*cp)
break;
*cp++ = '\0';
person = cp;
while (*cp) {
if (*cp != ' ') {
cp++;
continue;
}
*cp++ = '\0';
while (isspace(*cp))
cp++;
if (*cp == '\0')
break;
if (isdigit(*cp)) {
if (requests >= MAXREQUESTS)
fatal("Too many requests");
requ[requests++] = atoi(cp);
} else {
if (users >= MAXUSERS)
fatal("Too many users");
user[users++] = cp;
}
}
rmjob();
break;
}
fatal("Illegal service request");
}
}
/*
* Make a pass through the printcap database and start printing any
* files left from the last time the machine went down.
*/
static void
startup()
{
char *buf;
char *cp;
/*
* Restart the daemons.
*/
while (cgetnext(&buf, printcapdb) > 0) {
if (ckqueue(buf) <= 0) {
free(buf);
continue; /* no work to do for this printer */
}
for (cp = buf; *cp; cp++)
if (*cp == '|' || *cp == ':') {
*cp = '\0';
break;
}
if (lflag)
syslog(LOG_INFO, "work for %s", buf);
switch (fork()) {
case -1:
syslog(LOG_WARNING, "startup: cannot fork");
mcleanup(0);
case 0:
printer = buf;
setproctitle("working on printer %s", printer);
cgetclose();
printjob();
/* NOTREACHED */
default:
child_count++;
free(buf);
}
}
}
/*
* Make sure there's some work to do before forking off a child
*/
static int
ckqueue(cap)
char *cap;
{
struct dirent *d;
DIR *dirp;
char *spooldir;
if (cgetstr(cap, "sd", &spooldir) == -1)
spooldir = _PATH_DEFSPOOL;
if ((dirp = opendir(spooldir)) == NULL)
return (-1);
while ((d = readdir(dirp)) != NULL) {
if (d->d_name[0] != 'c' || d->d_name[1] != 'f')
continue; /* daemon control files only */
closedir(dirp);
return (1); /* found something */
}
closedir(dirp);
return (0);
}
#define DUMMY ":nobody::"
/*
* Check to see if the from host has access to the line printer.
*/
static void
chkhost(f)
struct sockaddr *f;
{
struct addrinfo hints, *res, *r;
FILE *hostf;
int first = 1, good = 0;
char host[NI_MAXHOST], ip[NI_MAXHOST];
char serv[NI_MAXSERV];
int error;
error = getnameinfo(f, f->sa_len, NULL, 0, serv, sizeof(serv),
NI_NUMERICSERV);
if (error || atoi(serv) >= IPPORT_RESERVED)
fatal("Malformed from address");
/* Need real hostname for temporary filenames */
error = getnameinfo(f, f->sa_len, host, sizeof(host), NULL, 0,
NI_NAMEREQD);
if (error) {
error = getnameinfo(f, f->sa_len, host, sizeof(host), NULL, 0,
NI_NUMERICHOST);
if (error)
fatal("Host name for your address unknown");
else
fatal("Host name for your address (%s) unknown", host);
}
(void)strncpy(fromb, host, sizeof(fromb) - 1);
fromb[sizeof(fromb) - 1] = '\0';
from = fromb;
/* need address in stringform for comparison (no DNS lookup here) */
error = getnameinfo(f, f->sa_len, host, sizeof(host), NULL, 0,
NI_NUMERICHOST);
if (error)
fatal("Cannot print address");
/* Check for spoof, ala rlogind */
memset(&hints, 0, sizeof(hints));
hints.ai_family = PF_UNSPEC;
hints.ai_socktype = SOCK_DGRAM; /*dummy*/
error = getaddrinfo(fromb, NULL, &hints, &res);
if (error) {
fatal("hostname for your address (%s) unknown: %s", host,
gai_strerror(error));
}
good = 0;
for (r = res; good == 0 && r; r = r->ai_next) {
error = getnameinfo(r->ai_addr, r->ai_addrlen, ip, sizeof(ip),
NULL, 0, NI_NUMERICHOST);
if (!error && !strcmp(host, ip))
good = 1;
}
if (res)
freeaddrinfo(res);
if (good == 0)
fatal("address for your hostname (%s) not matched", host);
setproctitle("serving %s", from);
hostf = fopen(_PATH_HOSTSEQUIV, "r");
again:
if (hostf) {
if (__ivaliduser_sa(hostf, f, f->sa_len, DUMMY, DUMMY) == 0) {
(void)fclose(hostf);
return;
}
(void)fclose(hostf);
}
if (first == 1) {
first = 0;
hostf = fopen(_PATH_HOSTSLPD, "r");
goto again;
}
fatal("Your host does not have line printer access");
/*NOTREACHED*/
}
static void
usage()
{
fprintf(stderr, "usage: %s [-dlrs] [-b bind-address] [-n maxchild] "
"[-w maxwait] [port]\n", getprogname());
exit(1);
}
/* setup server socket for specified address family */
/* if af is PF_UNSPEC more than one socket may be returned */
/* the returned list is dynamically allocated, so caller needs to free it */
int *
socksetup(af, options, port)
int af, options;
const char *port;
{
struct addrinfo hints, *res, *r;
int error, maxs = 0, *s, *socks = NULL, blidx = 0;
const int on = 1;
do {
memset(&hints, 0, sizeof(hints));
hints.ai_flags = AI_PASSIVE;
hints.ai_family = af;
hints.ai_socktype = SOCK_STREAM;
error = getaddrinfo((blist_addrs == 0) ? NULL : blist[blidx],
port ? port : "printer", &hints, &res);
if (error) {
if (blist_addrs)
syslog(LOG_ERR, "%s: %s", blist[blidx],
gai_strerror(error));
else
syslog(LOG_ERR, "%s", gai_strerror(error));
mcleanup(0);
}
/* Count max number of sockets we may open */
for (r = res; r; r = r->ai_next, maxs++)
;
if (socks == NULL) {
socks = malloc((maxs + 1) * sizeof(int));
if (socks)
*socks = 0; /* num of sockets ctr at start */
} else
socks = realloc(socks, (maxs + 1) * sizeof(int));
if (!socks) {
syslog(LOG_ERR, "couldn't allocate memory for sockets");
mcleanup(0);
}
s = socks + *socks + 1;
for (r = res; r; r = r->ai_next) {
*s = socket(r->ai_family, r->ai_socktype,
r->ai_protocol);
if (*s < 0) {
syslog(LOG_DEBUG, "socket(): %m");
continue;
}
if (options & SO_DEBUG)
if (setsockopt(*s, SOL_SOCKET, SO_DEBUG,
&on, sizeof(on)) < 0) {
syslog(LOG_ERR,
"setsockopt (SO_DEBUG): %m");
close (*s);
continue;
}
if (bind(*s, r->ai_addr, r->ai_addrlen) < 0) {
syslog(LOG_DEBUG, "bind(): %m");
close (*s);
continue;
}
*socks = *socks + 1;
s++;
}
if (res)
freeaddrinfo(res);
} while (++blidx < blist_addrs);
if (socks == NULL || *socks == 0) {
syslog(LOG_ERR, "Couldn't bind to any socket");
if (socks != NULL)
free(socks);
mcleanup(0);
}
return(socks);
}