kern.c revision 143723
1/*-
2 * Copyright (c) 1997 Berkeley Software Design, Inc. All rights reserved.
3 *
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
6 * are met:
7 * 1. Redistributions of source code must retain the above copyright
8 *    notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 *    notice, this list of conditions and the following disclaimer in the
11 *    documentation and/or other materials provided with the distribution.
12 * 3. Berkeley Software Design Inc's name may not be used to endorse or
13 *    promote products derived from this software without specific prior
14 *    written permission.
15 *
16 * THIS SOFTWARE IS PROVIDED BY BERKELEY SOFTWARE DESIGN INC ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED.  IN NO EVENT SHALL BERKELEY SOFTWARE DESIGN INC BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 *
28 *      from BSDI kern.c,v 1.2 1998/11/25 22:38:27 don Exp
29 * $FreeBSD: head/usr.sbin/rpc.lockd/kern.c 143723 2005-03-16 23:55:48Z cognet $
30 */
31
32#include <sys/param.h>
33#include <sys/mount.h>
34#include <sys/queue.h>
35#include <sys/socket.h>
36#include <sys/stat.h>
37
38#include <netinet/in.h>
39#include <arpa/inet.h>
40
41#include <err.h>
42#include <errno.h>
43#include <fcntl.h>
44#include <paths.h>
45#include <pwd.h>
46#include <stdio.h>
47#include <stdlib.h>
48#include <string.h>
49#include <syslog.h>
50#include <unistd.h>
51#include <netdb.h>
52
53#include "nlm_prot.h"
54#include <nfs/rpcv2.h>
55#include <nfs/nfsproto.h>
56#include <nfsclient/nfs_lock.h>
57
58#include "lockd.h"
59#include "lockd_lock.h"
60#include <nfsclient/nfs.h>
61
62#define DAEMON_USERNAME	"daemon"
63
64/* Lock request owner. */
65typedef struct __owner {
66	pid_t	 pid;				/* Process ID. */
67	time_t	 tod;				/* Time-of-day. */
68} OWNER;
69static OWNER owner;
70
71static char hostname[MAXHOSTNAMELEN + 1];	/* Hostname. */
72
73static void	client_cleanup(void);
74static void	set_auth(CLIENT *cl, struct xucred *ucred);
75int	lock_request(LOCKD_MSG *);
76int	test_request(LOCKD_MSG *);
77void	show(LOCKD_MSG *);
78int	unlock_request(LOCKD_MSG *);
79static	int	devfd;
80
81static int
82nfslockdans(int vers, struct lockd_ans *ansp)
83{
84
85	ansp->la_vers = vers;
86	return (write(devfd, ansp, sizeof *ansp) <= 0);
87}
88
89/*
90 * will break because fifo needs to be repopened when EOF'd
91 */
92#define lockd_seteuid(uid)	seteuid(uid)
93
94#define d_calls (debug_level > 1)
95#define d_args (debug_level > 2)
96
97static const char *
98from_addr(saddr)
99	struct sockaddr *saddr;
100{
101	static char inet_buf[INET6_ADDRSTRLEN];
102
103	if (getnameinfo(saddr, saddr->sa_len, inet_buf, sizeof(inet_buf),
104			NULL, 0, NI_NUMERICHOST) == 0)
105		return inet_buf;
106	return "???";
107}
108
109void
110client_cleanup(void)
111{
112	(void)lockd_seteuid(0);
113	exit(-1);
114}
115
116/*
117 * client_request --
118 *	Loop around messages from the kernel, forwarding them off to
119 *	NLM servers.
120 */
121pid_t
122client_request(void)
123{
124	LOCKD_MSG msg;
125	fd_set rdset;
126	int nr, ret;
127	pid_t child;
128	uid_t daemon_uid;
129	mode_t old_umask;
130	struct passwd *pw;
131
132	/* Open the dev . */
133	devfd = open(_PATH_DEV _PATH_NFSLCKDEV, O_RDWR | O_NONBLOCK);
134	if (devfd < 0) {
135		syslog(LOG_ERR, "open: %s: %m", _PATH_NFSLCKDEV);
136		goto err;
137	}
138	/*
139	 * Create a separate process, the client code is really a separate
140	 * daemon that shares a lot of code.
141	 */
142	switch (child = fork()) {
143	case -1:
144		err(1, "fork");
145	case 0:
146		break;
147	default:
148		return (child);
149	}
150
151	signal(SIGHUP, (sig_t)client_cleanup);
152	signal(SIGTERM, (sig_t)client_cleanup);
153
154	/* Setup. */
155	(void)time(&owner.tod);
156	owner.pid = getpid();
157	(void)gethostname(hostname, sizeof(hostname) - 1);
158
159	pw = getpwnam(DAEMON_USERNAME);
160	if (pw == NULL) {
161		syslog(LOG_ERR, "getpwnam: %s: %m", DAEMON_USERNAME);
162		goto err;
163	}
164	daemon_uid = pw->pw_uid;
165	/* drop our root priviledges */
166	(void)lockd_seteuid(daemon_uid);
167
168	for (;;) {
169		/* Read the fixed length message. */
170		if ((nr = read(devfd, &msg, sizeof(msg))) == sizeof(msg)) {
171			if (d_args)
172				show(&msg);
173
174			if (msg.lm_version != LOCKD_MSG_VERSION) {
175				syslog(LOG_ERR,
176				    "unknown msg type: %d", msg.lm_version);
177			}
178			/*
179			 * Send it to the NLM server and don't grant the lock
180			 * if we fail for any reason.
181			 */
182			switch (msg.lm_fl.l_type) {
183			case F_RDLCK:
184			case F_WRLCK:
185				if (msg.lm_getlk)
186					ret = test_request(&msg);
187				else
188					ret = lock_request(&msg);
189				break;
190			case F_UNLCK:
191				ret = unlock_request(&msg);
192				break;
193			default:
194				ret = 1;
195				syslog(LOG_ERR,
196				    "unknown lock type: %d", msg.lm_fl.l_type);
197				break;
198			}
199			if (ret) {
200				struct lockd_ans ans;
201
202				ans.la_msg_ident = msg.lm_msg_ident;
203				ans.la_errno = EHOSTUNREACH;
204
205				if (nfslockdans(LOCKD_ANS_VERSION, &ans)) {
206					syslog((errno == EPIPE ? LOG_INFO :
207						LOG_ERR), "process %lu: %m",
208						(u_long)msg.lm_msg_ident.pid);
209				}
210			}
211		} else if (nr == -1) {
212			if (errno != EAGAIN) {
213				syslog(LOG_ERR, "read: %s: %m", _PATH_NFSLCKDEV);
214				goto err;
215			}
216		} else if (nr != 0) {
217			syslog(LOG_ERR,
218			    "%s: discard %d bytes", _PATH_NFSLCKDEV, nr);
219		}
220	}
221
222	/* Reached only on error. */
223err:
224	(void)lockd_seteuid(0);
225	_exit (1);
226}
227
228void
229set_auth(cl, xucred)
230	CLIENT *cl;
231	struct xucred *xucred;
232{
233        if (cl->cl_auth != NULL)
234                cl->cl_auth->ah_ops->ah_destroy(cl->cl_auth);
235        cl->cl_auth = authunix_create(hostname,
236                        xucred->cr_uid,
237                        xucred->cr_groups[0],
238                        xucred->cr_ngroups - 1,
239                        &xucred->cr_groups[1]);
240}
241
242
243/*
244 * test_request --
245 *	Convert a lock LOCKD_MSG into an NLM request, and send it off.
246 */
247int
248test_request(LOCKD_MSG *msg)
249{
250	CLIENT *cli;
251	struct timeval timeout = {0, 0};	/* No timeout, no response. */
252	char dummy;
253
254	if (d_calls)
255		syslog(LOG_DEBUG, "test request: %s: %s to %s",
256		    msg->lm_nfsv3 ? "V4" : "V1/3",
257		    msg->lm_fl.l_type == F_WRLCK ? "write" : "read",
258		    from_addr((struct sockaddr *)&msg->lm_addr));
259
260	if (msg->lm_nfsv3) {
261		struct nlm4_testargs arg4;
262
263		arg4.cookie.n_bytes = (char *)&msg->lm_msg_ident;
264		arg4.cookie.n_len = sizeof(msg->lm_msg_ident);
265		arg4.exclusive = msg->lm_fl.l_type == F_WRLCK ? 1 : 0;
266		arg4.alock.caller_name = hostname;
267		arg4.alock.fh.n_bytes = (char *)&msg->lm_fh;
268		arg4.alock.fh.n_len = msg->lm_fh_len;
269		arg4.alock.oh.n_bytes = (char *)&owner;
270		arg4.alock.oh.n_len = sizeof(owner);
271		arg4.alock.svid = msg->lm_msg_ident.pid;
272		arg4.alock.l_offset = msg->lm_fl.l_start;
273		arg4.alock.l_len = msg->lm_fl.l_len;
274
275		if ((cli = get_client(
276		    (struct sockaddr *)&msg->lm_addr,
277		    NLM_VERS4)) == NULL)
278			return (1);
279
280		set_auth(cli, &msg->lm_cred);
281		(void)clnt_call(cli, NLM_TEST_MSG,
282		    (xdrproc_t)xdr_nlm4_testargs, &arg4,
283		    (xdrproc_t)xdr_void, &dummy, timeout);
284	} else {
285		struct nlm_testargs arg;
286
287		arg.cookie.n_bytes = (char *)&msg->lm_msg_ident;
288		arg.cookie.n_len = sizeof(msg->lm_msg_ident);
289		arg.exclusive = msg->lm_fl.l_type == F_WRLCK ? 1 : 0;
290		arg.alock.caller_name = hostname;
291		arg.alock.fh.n_bytes = (char *)&msg->lm_fh;
292		arg.alock.fh.n_len = msg->lm_fh_len;
293		arg.alock.oh.n_bytes = (char *)&owner;
294		arg.alock.oh.n_len = sizeof(owner);
295		arg.alock.svid = msg->lm_msg_ident.pid;
296		arg.alock.l_offset = msg->lm_fl.l_start;
297		arg.alock.l_len = msg->lm_fl.l_len;
298
299		if ((cli = get_client(
300		    (struct sockaddr *)&msg->lm_addr,
301		    NLM_VERS)) == NULL)
302			return (1);
303
304		set_auth(cli, &msg->lm_cred);
305		(void)clnt_call(cli, NLM_TEST_MSG,
306		    (xdrproc_t)xdr_nlm_testargs, &arg,
307		    (xdrproc_t)xdr_void, &dummy, timeout);
308	}
309	return (0);
310}
311
312/*
313 * lock_request --
314 *	Convert a lock LOCKD_MSG into an NLM request, and send it off.
315 */
316int
317lock_request(LOCKD_MSG *msg)
318{
319	CLIENT *cli;
320	struct nlm4_lockargs arg4;
321	struct nlm_lockargs arg;
322	struct timeval timeout = {0, 0};	/* No timeout, no response. */
323	char dummy;
324
325	if (d_calls)
326		syslog(LOG_DEBUG, "lock request: %s: %s to %s",
327		    msg->lm_nfsv3 ? "V4" : "V1/3",
328		    msg->lm_fl.l_type == F_WRLCK ? "write" : "read",
329		    from_addr((struct sockaddr *)&msg->lm_addr));
330
331	if (msg->lm_nfsv3) {
332		arg4.cookie.n_bytes = (char *)&msg->lm_msg_ident;
333		arg4.cookie.n_len = sizeof(msg->lm_msg_ident);
334		arg4.block = msg->lm_wait ? 1 : 0;
335		arg4.exclusive = msg->lm_fl.l_type == F_WRLCK ? 1 : 0;
336		arg4.alock.caller_name = hostname;
337		arg4.alock.fh.n_bytes = (char *)&msg->lm_fh;
338		arg4.alock.fh.n_len = msg->lm_fh_len;
339		arg4.alock.oh.n_bytes = (char *)&owner;
340		arg4.alock.oh.n_len = sizeof(owner);
341		arg4.alock.svid = msg->lm_msg_ident.pid;
342		arg4.alock.l_offset = msg->lm_fl.l_start;
343		arg4.alock.l_len = msg->lm_fl.l_len;
344		arg4.reclaim = 0;
345		arg4.state = nsm_state;
346
347		if ((cli = get_client(
348		    (struct sockaddr *)&msg->lm_addr,
349		    NLM_VERS4)) == NULL)
350			return (1);
351
352		set_auth(cli, &msg->lm_cred);
353		(void)clnt_call(cli, NLM_LOCK_MSG,
354		    (xdrproc_t)xdr_nlm4_lockargs, &arg4,
355		    (xdrproc_t)xdr_void, &dummy, timeout);
356	} else {
357		arg.cookie.n_bytes = (char *)&msg->lm_msg_ident;
358		arg.cookie.n_len = sizeof(msg->lm_msg_ident);
359		arg.block = msg->lm_wait ? 1 : 0;
360		arg.exclusive = msg->lm_fl.l_type == F_WRLCK ? 1 : 0;
361		arg.alock.caller_name = hostname;
362		arg.alock.fh.n_bytes = (char *)&msg->lm_fh;
363		arg.alock.fh.n_len = msg->lm_fh_len;
364		arg.alock.oh.n_bytes = (char *)&owner;
365		arg.alock.oh.n_len = sizeof(owner);
366		arg.alock.svid = msg->lm_msg_ident.pid;
367		arg.alock.l_offset = msg->lm_fl.l_start;
368		arg.alock.l_len = msg->lm_fl.l_len;
369		arg.reclaim = 0;
370		arg.state = nsm_state;
371
372		if ((cli = get_client(
373		    (struct sockaddr *)&msg->lm_addr,
374		    NLM_VERS)) == NULL)
375			return (1);
376
377		set_auth(cli, &msg->lm_cred);
378		(void)clnt_call(cli, NLM_LOCK_MSG,
379		    (xdrproc_t)xdr_nlm_lockargs, &arg,
380		    (xdrproc_t)xdr_void, &dummy, timeout);
381	}
382	return (0);
383}
384
385/*
386 * unlock_request --
387 *	Convert an unlock LOCKD_MSG into an NLM request, and send it off.
388 */
389int
390unlock_request(LOCKD_MSG *msg)
391{
392	CLIENT *cli;
393	struct nlm4_unlockargs arg4;
394	struct nlm_unlockargs arg;
395	struct timeval timeout = {0, 0};	/* No timeout, no response. */
396	char dummy;
397
398	if (d_calls)
399		syslog(LOG_DEBUG, "unlock request: %s: to %s",
400		    msg->lm_nfsv3 ? "V4" : "V1/3",
401		    from_addr((struct sockaddr *)&msg->lm_addr));
402
403	if (msg->lm_nfsv3) {
404		arg4.cookie.n_bytes = (char *)&msg->lm_msg_ident;
405		arg4.cookie.n_len = sizeof(msg->lm_msg_ident);
406		arg4.alock.caller_name = hostname;
407		arg4.alock.fh.n_bytes = (char *)&msg->lm_fh;
408		arg4.alock.fh.n_len = msg->lm_fh_len;
409		arg4.alock.oh.n_bytes = (char *)&owner;
410		arg4.alock.oh.n_len = sizeof(owner);
411		arg4.alock.svid = msg->lm_msg_ident.pid;
412		arg4.alock.l_offset = msg->lm_fl.l_start;
413		arg4.alock.l_len = msg->lm_fl.l_len;
414
415		if ((cli = get_client(
416		    (struct sockaddr *)&msg->lm_addr,
417		    NLM_VERS4)) == NULL)
418			return (1);
419
420		set_auth(cli, &msg->lm_cred);
421		(void)clnt_call(cli, NLM_UNLOCK_MSG,
422		    (xdrproc_t)xdr_nlm4_unlockargs, &arg4,
423		    (xdrproc_t)xdr_void, &dummy, timeout);
424	} else {
425		arg.cookie.n_bytes = (char *)&msg->lm_msg_ident;
426		arg.cookie.n_len = sizeof(msg->lm_msg_ident);
427		arg.alock.caller_name = hostname;
428		arg.alock.fh.n_bytes = (char *)&msg->lm_fh;
429		arg.alock.fh.n_len = msg->lm_fh_len;
430		arg.alock.oh.n_bytes = (char *)&owner;
431		arg.alock.oh.n_len = sizeof(owner);
432		arg.alock.svid = msg->lm_msg_ident.pid;
433		arg.alock.l_offset = msg->lm_fl.l_start;
434		arg.alock.l_len = msg->lm_fl.l_len;
435
436		if ((cli = get_client(
437		    (struct sockaddr *)&msg->lm_addr,
438		    NLM_VERS)) == NULL)
439			return (1);
440
441		set_auth(cli, &msg->lm_cred);
442		(void)clnt_call(cli, NLM_UNLOCK_MSG,
443		    (xdrproc_t)xdr_nlm_unlockargs, &arg,
444		    (xdrproc_t)xdr_void, &dummy, timeout);
445	}
446
447	return (0);
448}
449
450int
451lock_answer(int pid, netobj *netcookie, int result, int *pid_p, int version)
452{
453	struct lockd_ans ans;
454
455	if (netcookie->n_len != sizeof(ans.la_msg_ident)) {
456		if (pid == -1) {	/* we're screwed */
457			syslog(LOG_ERR, "inedible nlm cookie");
458			return -1;
459		}
460		ans.la_msg_ident.pid = pid;
461		ans.la_msg_ident.msg_seq = -1;
462	} else {
463		memcpy(&ans.la_msg_ident, netcookie->n_bytes,
464		    sizeof(ans.la_msg_ident));
465	}
466
467	if (d_calls)
468		syslog(LOG_DEBUG, "lock answer: pid %lu: %s %d",
469		    (unsigned long)ans.la_msg_ident.pid,
470		    version == NLM_VERS4 ? "nlmv4" : "nlmv3",
471		    result);
472
473	ans.la_set_getlk_pid = 0;
474	if (version == NLM_VERS4)
475		switch (result) {
476		case nlm4_granted:
477			ans.la_errno = 0;
478			break;
479		default:
480			ans.la_errno = EACCES;
481			break;
482		case nlm4_denied:
483			if (pid_p == NULL)
484				ans.la_errno = EAGAIN;
485			else {
486				/* this is an answer to a nlm_test msg */
487				ans.la_set_getlk_pid = 1;
488				ans.la_getlk_pid = *pid_p;
489				ans.la_errno = 0;
490			}
491			break;
492		case nlm4_denied_nolocks:
493			ans.la_errno = EAGAIN;
494			break;
495		case nlm4_blocked:
496			return -1;
497			/* NOTREACHED */
498		case nlm4_denied_grace_period:
499			ans.la_errno = EAGAIN;
500			break;
501		case nlm4_deadlck:
502			ans.la_errno = EDEADLK;
503			break;
504		case nlm4_rofs:
505			ans.la_errno = EROFS;
506			break;
507		case nlm4_stale_fh:
508			ans.la_errno = ESTALE;
509			break;
510		case nlm4_fbig:
511			ans.la_errno = EFBIG;
512			break;
513		case nlm4_failed:
514			ans.la_errno = EACCES;
515			break;
516		}
517	else
518		switch (result) {
519		case nlm_granted:
520			ans.la_errno = 0;
521			break;
522		default:
523			ans.la_errno = EACCES;
524			break;
525		case nlm_denied:
526			if (pid_p == NULL)
527				ans.la_errno = EAGAIN;
528			else {
529				/* this is an answer to a nlm_test msg */
530				ans.la_set_getlk_pid = 1;
531				ans.la_getlk_pid = *pid_p;
532				ans.la_errno = 0;
533			}
534			break;
535		case nlm_denied_nolocks:
536			ans.la_errno = EAGAIN;
537			break;
538		case nlm_blocked:
539			return -1;
540			/* NOTREACHED */
541		case nlm_denied_grace_period:
542			ans.la_errno = EAGAIN;
543			break;
544		case nlm_deadlck:
545			ans.la_errno = EDEADLK;
546			break;
547		}
548
549	if (nfslockdans(LOCKD_ANS_VERSION, &ans)) {
550		syslog(((errno == EPIPE || errno == ESRCH) ?
551			LOG_INFO : LOG_ERR),
552			"process %lu: %m", (u_long)ans.la_msg_ident.pid);
553		return -1;
554	}
555	return 0;
556}
557
558/*
559 * show --
560 *	Display the contents of a kernel LOCKD_MSG structure.
561 */
562void
563show(LOCKD_MSG *mp)
564{
565	static char hex[] = "0123456789abcdef";
566	struct fid *fidp;
567	fsid_t *fsidp;
568	size_t len;
569	u_int8_t *p, *t, buf[NFS_SMALLFH*3+1];
570
571	syslog(LOG_DEBUG, "process ID: %lu\n", (long)mp->lm_msg_ident.pid);
572
573	fsidp = (fsid_t *)&mp->lm_fh;
574	fidp = (struct fid *)((u_int8_t *)&mp->lm_fh + sizeof(fsid_t));
575
576	for (t = buf, p = (u_int8_t *)mp->lm_fh,
577	    len = mp->lm_fh_len;
578	    len > 0; ++p, --len) {
579		*t++ = '\\';
580		*t++ = hex[(*p & 0xf0) >> 4];
581		*t++ = hex[*p & 0x0f];
582	}
583	*t = '\0';
584
585	syslog(LOG_DEBUG, "fh_len %d, fh %s\n", (int)mp->lm_fh_len, buf);
586
587	/* Show flock structure. */
588	syslog(LOG_DEBUG, "start %qu; len %qu; pid %lu; type %d; whence %d\n",
589	    (unsigned long long)mp->lm_fl.l_start,
590	    (unsigned long long)mp->lm_fl.l_len, (u_long)mp->lm_fl.l_pid,
591	    mp->lm_fl.l_type, mp->lm_fl.l_whence);
592
593	/* Show wait flag. */
594	syslog(LOG_DEBUG, "wait was %s\n", mp->lm_wait ? "set" : "not set");
595}
596