2 * network.c -- Provide common network functions for NFS mount/umount
4 * Copyright (C) 2007 Oracle. All rights reserved.
5 * Copyright (C) 2007 Chuck Lever <chuck.lever@oracle.com>
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public
9 * License as published by the Free Software Foundation; either
10 * version 2 of the License, or (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License for more details.
17 * You should have received a copy of the GNU General Public
18 * License along with this program; if not, write to the
19 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
20 * Boston, MA 021110-1307, USA.
33 #include <rpc/pmap_prot.h>
34 #include <rpc/pmap_clnt.h>
35 #include <sys/socket.h>
40 #include "nfs_mount.h"
41 #include "mount_constants.h"
44 #ifdef HAVE_RPCSVC_NFS_PROT_H
45 #include <rpcsvc/nfs_prot.h>
47 #include <linux/nfs.h>
48 #define nfsstat nfs_stat
55 #define PMAP_TIMEOUT (10)
56 #define CONNECT_TIMEOUT (20)
57 #define MOUNT_TIMEOUT (30)
59 #if SIZEOF_SOCKLEN_T - 0 == 0
60 #define socklen_t unsigned int
63 extern int nfs_mount_data_version;
64 extern char *progname;
67 static const unsigned long nfs_to_mnt[] = {
74 static const unsigned long mnt_to_nfs[] = {
82 * Map an NFS version into the corresponding Mountd version
84 unsigned long nfsvers_to_mnt(const unsigned long vers)
87 return nfs_to_mnt[vers];
92 * Map a Mountd version into the corresponding NFS version
94 static unsigned long mntvers_to_nfs(const unsigned long vers)
97 return mnt_to_nfs[vers];
101 static const unsigned int probe_udp_only[] = {
106 static const unsigned int probe_udp_first[] = {
112 static const unsigned int probe_tcp_first[] = {
118 static const unsigned long probe_nfs2_only[] = {
123 static const unsigned long probe_nfs3_first[] = {
129 static const unsigned long probe_mnt1_first[] = {
135 static const unsigned long probe_mnt3_first[] = {
142 int nfs_gethostbyname(const char *hostname, struct sockaddr_in *saddr)
146 saddr->sin_family = AF_INET;
147 if (!inet_aton(hostname, &saddr->sin_addr)) {
148 if ((hp = gethostbyname(hostname)) == NULL) {
149 nfs_error(_("%s: can't get address for %s\n"),
153 if (hp->h_length > sizeof(*saddr)) {
154 nfs_error(_("%s: got bad hp->h_length\n"),
156 hp->h_length = sizeof(*saddr);
158 memcpy(&saddr->sin_addr, hp->h_addr, hp->h_length);
165 * Attempt to connect a socket, but time out after "timeout" seconds.
167 * On error return, caller closes the socket.
169 static int connect_to(int fd, struct sockaddr *addr,
170 socklen_t addrlen, int timeout)
174 struct timeval tv = {
178 saved = fcntl(fd, F_GETFL, 0);
179 fcntl(fd, F_SETFL, saved | O_NONBLOCK);
181 ret = connect(fd, addr, addrlen);
182 if (ret < 0 && errno != EINPROGRESS)
190 ret = select(fd + 1, &rset, &wset, NULL, &tv);
195 if (FD_ISSET(fd, &rset) || FD_ISSET(fd, &wset)) {
197 socklen_t len = sizeof(error);
198 if (getsockopt(fd, SOL_SOCKET, SO_ERROR, &error, &len) < 0)
208 fcntl(fd, F_SETFL, saved);
213 * Create a socket that is locally bound to a reserved or non-reserved port.
215 * The caller should check rpc_createerr to determine the cause of any error.
217 static int get_socket(struct sockaddr_in *saddr, unsigned int p_prot,
218 unsigned int timeout, int resvp, int conn)
221 struct sockaddr_in laddr;
222 socklen_t namelen = sizeof(laddr);
224 type = (p_prot == IPPROTO_UDP ? SOCK_DGRAM : SOCK_STREAM);
225 if ((so = socket (AF_INET, type, p_prot)) < 0)
228 laddr.sin_family = AF_INET;
230 laddr.sin_addr.s_addr = htonl(INADDR_ANY);
232 if (bindresvport(so, &laddr) < 0)
233 goto err_bindresvport;
235 cc = bind(so, (struct sockaddr *)&laddr, namelen);
239 if (type == SOCK_STREAM || (conn && type == SOCK_DGRAM)) {
240 cc = connect_to(so, (struct sockaddr *)saddr, namelen,
248 rpc_createerr.cf_stat = RPC_SYSTEMERROR;
249 rpc_createerr.cf_error.re_errno = errno;
251 nfs_error(_("%s: Unable to create %s socket: errno %d (%s)\n"),
252 progname, p_prot == IPPROTO_UDP ? _("UDP") : _("TCP"),
253 errno, strerror(errno));
258 rpc_createerr.cf_stat = RPC_SYSTEMERROR;
259 rpc_createerr.cf_error.re_errno = errno;
261 nfs_error(_("%s: Unable to bindresvport %s socket: errno %d"
263 progname, p_prot == IPPROTO_UDP ? _("UDP") : _("TCP"),
264 errno, strerror(errno));
270 rpc_createerr.cf_stat = RPC_SYSTEMERROR;
271 rpc_createerr.cf_error.re_errno = errno;
273 nfs_error(_("%s: Unable to bind to %s socket: errno %d (%s)\n"),
274 progname, p_prot == IPPROTO_UDP ? _("UDP") : _("TCP"),
275 errno, strerror(errno));
281 rpc_createerr.cf_stat = RPC_SYSTEMERROR;
282 rpc_createerr.cf_error.re_errno = errno;
284 nfs_error(_("%s: Unable to connect to %s:%d, errno %d (%s)\n"),
285 progname, inet_ntoa(saddr->sin_addr),
286 ntohs(saddr->sin_port), errno, strerror(errno));
293 * getport() is very similar to pmap_getport() with the exception that
294 * this version tries to use an ephemeral port, since reserved ports are
295 * not needed for GETPORT queries. This conserves the very limited
296 * reserved port space, which helps reduce failed socket binds
297 * during mount storms.
299 * A side effect of calling this function is that rpccreateerr is set.
301 static unsigned short getport(struct sockaddr_in *saddr,
302 unsigned long program,
303 unsigned long version,
306 unsigned short port = 0;
311 saddr->sin_port = htons(PMAPPORT);
313 socket = get_socket(saddr, proto, PMAP_TIMEOUT, FALSE, FALSE);
314 if (socket == RPC_ANYSOCK) {
315 if (proto == IPPROTO_TCP &&
316 rpc_createerr.cf_error.re_errno == ETIMEDOUT)
317 rpc_createerr.cf_stat = RPC_TIMEDOUT;
323 clnt = clntudp_bufcreate(saddr,
325 RETRY_TIMEOUT, &socket,
330 clnt = clnttcp_create(saddr, PMAPPROG, PMAPVERS, &socket,
331 RPCSMALLMSGSIZE, RPCSMALLMSGSIZE);
335 struct pmap parms = {
341 stat = clnt_call(clnt, PMAPPROC_GETPORT,
342 (xdrproc_t)xdr_pmap, (caddr_t)&parms,
343 (xdrproc_t)xdr_u_short, (caddr_t)&port,
346 clnt_geterr(clnt, &rpc_createerr.cf_error);
347 rpc_createerr.cf_stat = stat;
350 if (stat != RPC_SUCCESS)
353 rpc_createerr.cf_stat = RPC_PROGNOTREGISTERED;
361 * Use the portmapper to discover whether or not the service we want is
362 * available. The lists 'versions' and 'protos' define ordered sequences
363 * of service versions and udp/tcp protocols to probe for.
365 static int probe_port(clnt_addr_t *server, const unsigned long *versions,
366 const unsigned int *protos)
368 struct sockaddr_in *saddr = &server->saddr;
369 struct pmap *pmap = &server->pmap;
370 const unsigned long prog = pmap->pm_prog, *p_vers;
371 const unsigned int prot = (u_int)pmap->pm_prot, *p_prot;
372 const u_short port = (u_short) pmap->pm_port;
373 unsigned long vers = pmap->pm_vers;
374 unsigned short p_port;
376 p_prot = prot ? &prot : protos;
377 p_vers = vers ? &vers : versions;
378 rpc_createerr.cf_stat = 0;
380 p_port = getport(saddr, prog, *p_vers, *p_prot);
382 if (!port || port == p_port) {
383 saddr->sin_port = htons(p_port);
385 printf(_("%s: trying %s prog %ld vers "
386 "%ld prot %s port %d\n"),
388 inet_ntoa(saddr->sin_addr),
390 *p_prot == IPPROTO_UDP ?
394 if (clnt_ping(saddr, prog, *p_vers, *p_prot, NULL))
396 if (rpc_createerr.cf_stat == RPC_TIMEDOUT)
400 if (rpc_createerr.cf_stat != RPC_PROGNOTREGISTERED)
408 if (vers == pmap->pm_vers) {
412 if (vers || !*++p_vers)
421 pmap->pm_vers = *p_vers;
423 pmap->pm_prot = *p_prot;
425 pmap->pm_port = p_port;
426 rpc_createerr.cf_stat = 0;
430 static int probe_nfsport(clnt_addr_t *nfs_server)
432 struct pmap *pmap = &nfs_server->pmap;
434 if (pmap->pm_vers && pmap->pm_prot && pmap->pm_port)
437 if (nfs_mount_data_version >= 4)
438 return probe_port(nfs_server, probe_nfs3_first, probe_tcp_first);
440 return probe_port(nfs_server, probe_nfs2_only, probe_udp_only);
443 static int probe_mntport(clnt_addr_t *mnt_server)
445 struct pmap *pmap = &mnt_server->pmap;
447 if (pmap->pm_vers && pmap->pm_prot && pmap->pm_port)
450 if (nfs_mount_data_version >= 4)
451 return probe_port(mnt_server, probe_mnt3_first, probe_udp_first);
453 return probe_port(mnt_server, probe_mnt1_first, probe_udp_only);
456 int probe_bothports(clnt_addr_t *mnt_server, clnt_addr_t *nfs_server)
458 struct pmap *nfs_pmap = &nfs_server->pmap;
459 struct pmap *mnt_pmap = &mnt_server->pmap;
460 struct pmap save_nfs, save_mnt;
462 const unsigned long *probe_vers;
464 if (mnt_pmap->pm_vers && !nfs_pmap->pm_vers)
465 nfs_pmap->pm_vers = mntvers_to_nfs(mnt_pmap->pm_vers);
466 else if (nfs_pmap->pm_vers && !mnt_pmap->pm_vers)
467 mnt_pmap->pm_vers = nfsvers_to_mnt(nfs_pmap->pm_vers);
468 if (nfs_pmap->pm_vers)
471 memcpy(&save_nfs, nfs_pmap, sizeof(save_nfs));
472 memcpy(&save_mnt, mnt_pmap, sizeof(save_mnt));
473 probe_vers = (nfs_mount_data_version >= 4) ?
474 probe_mnt3_first : probe_mnt1_first;
476 for (; *probe_vers; probe_vers++) {
477 nfs_pmap->pm_vers = mntvers_to_nfs(*probe_vers);
478 if ((res = probe_nfsport(nfs_server) != 0)) {
479 mnt_pmap->pm_vers = nfsvers_to_mnt(nfs_pmap->pm_vers);
480 if ((res = probe_mntport(mnt_server)) != 0)
482 memcpy(mnt_pmap, &save_mnt, sizeof(*mnt_pmap));
484 switch (rpc_createerr.cf_stat) {
485 case RPC_PROGVERSMISMATCH:
486 case RPC_PROGNOTREGISTERED:
491 memcpy(nfs_pmap, &save_nfs, sizeof(*nfs_pmap));
498 if (!probe_nfsport(nfs_server))
500 return probe_mntport(mnt_server);
503 static int probe_statd(void)
505 struct sockaddr_in addr;
508 memset(&addr, 0, sizeof(addr));
509 addr.sin_family = AF_INET;
510 addr.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
511 port = getport(&addr, 100024, 1, IPPROTO_UDP);
515 addr.sin_port = htons(port);
517 if (clnt_ping(&addr, 100024, 1, IPPROTO_UDP, NULL) <= 0)
524 * Attempt to start rpc.statd
526 int start_statd(void)
536 if (stat(START_STATD, &stb) == 0) {
537 if (S_ISREG(stb.st_mode) && (stb.st_mode & S_IXUSR)) {
549 * nfs_call_umount - ask the server to remove a share from it's rmtab
550 * @mnt_server: address of RPC MNT program server
551 * @argp: directory path of share to "unmount"
553 * Returns one if the unmount call succeeded; zero if the unmount
554 * failed for any reason.
556 * Note that a side effect of calling this function is that rpccreateerr
559 int nfs_call_umount(clnt_addr_t *mnt_server, dirpath *argp)
562 enum clnt_stat res = 0;
565 switch (mnt_server->pmap.pm_vers) {
569 if (!probe_mntport(mnt_server))
571 clnt = mnt_openclnt(mnt_server, &msock);
574 res = clnt_call(clnt, MOUNTPROC_UMNT,
575 (xdrproc_t)xdr_dirpath, (caddr_t)argp,
576 (xdrproc_t)xdr_void, NULL,
578 mnt_closeclnt(clnt, msock);
579 if (res == RPC_SUCCESS)
587 if (res == RPC_SUCCESS)
592 CLIENT *mnt_openclnt(clnt_addr_t *mnt_server, int *msock)
594 struct sockaddr_in *mnt_saddr = &mnt_server->saddr;
595 struct pmap *mnt_pmap = &mnt_server->pmap;
598 mnt_saddr->sin_port = htons((u_short)mnt_pmap->pm_port);
599 *msock = get_socket(mnt_saddr, mnt_pmap->pm_prot, MOUNT_TIMEOUT,
601 if (*msock == RPC_ANYSOCK) {
602 if (rpc_createerr.cf_error.re_errno == EADDRINUSE)
604 * Probably in-use by a TIME_WAIT connection,
605 * It is worth waiting a while and trying again.
607 rpc_createerr.cf_stat = RPC_TIMEDOUT;
611 switch (mnt_pmap->pm_prot) {
613 clnt = clntudp_bufcreate(mnt_saddr,
614 mnt_pmap->pm_prog, mnt_pmap->pm_vers,
615 RETRY_TIMEOUT, msock,
616 MNT_SENDBUFSIZE, MNT_RECVBUFSIZE);
619 clnt = clnttcp_create(mnt_saddr,
620 mnt_pmap->pm_prog, mnt_pmap->pm_vers,
622 MNT_SENDBUFSIZE, MNT_RECVBUFSIZE);
626 /* try to mount hostname:dirname */
627 clnt->cl_auth = authunix_create_default();
633 void mnt_closeclnt(CLIENT *clnt, int msock)
635 auth_destroy(clnt->cl_auth);
641 * Sigh... getport() doesn't actually check the version number.
642 * In order to make sure that the server actually supports the service
643 * we're requesting, we open and RPC client, and fire off a NULL
646 int clnt_ping(struct sockaddr_in *saddr, const unsigned long prog,
647 const unsigned long vers, const unsigned int prot,
648 struct sockaddr_in *caddr)
652 static char clnt_res;
653 struct sockaddr dissolve;
655 rpc_createerr.cf_stat = stat = 0;
656 sock = get_socket(saddr, prot, CONNECT_TIMEOUT, FALSE, TRUE);
657 if (sock == RPC_ANYSOCK) {
658 if (rpc_createerr.cf_error.re_errno == ETIMEDOUT) {
660 * TCP timeout. Bubble up the error to see
661 * how it should be handled.
663 rpc_createerr.cf_stat = RPC_TIMEDOUT;
669 /* Get the address of our end of this connection */
670 socklen_t len = sizeof(*caddr);
671 if (getsockname(sock, caddr, &len) != 0)
672 caddr->sin_family = 0;
677 /* The socket is connected (so we could getsockname successfully),
678 * but some servers on multi-homed hosts reply from
679 * the wrong address, so if we stay connected, we lose the reply.
681 dissolve.sa_family = AF_UNSPEC;
682 connect(sock, &dissolve, sizeof(dissolve));
684 clnt = clntudp_bufcreate(saddr, prog, vers,
685 RETRY_TIMEOUT, &sock,
686 RPCSMALLMSGSIZE, RPCSMALLMSGSIZE);
689 clnt = clnttcp_create(saddr, prog, vers, &sock,
690 RPCSMALLMSGSIZE, RPCSMALLMSGSIZE);
697 memset(&clnt_res, 0, sizeof(clnt_res));
698 stat = clnt_call(clnt, NULLPROC,
699 (xdrproc_t)xdr_void, (caddr_t)NULL,
700 (xdrproc_t)xdr_void, (caddr_t)&clnt_res,
703 clnt_geterr(clnt, &rpc_createerr.cf_error);
704 rpc_createerr.cf_stat = stat;
709 if (stat == RPC_SUCCESS)