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.
37 #include <sys/types.h>
38 #include <sys/socket.h>
39 #include <netinet/in.h>
41 #include <rpc/pmap_prot.h>
42 #include <rpc/pmap_clnt.h>
47 #include "nfs_mount.h"
48 #include "mount_constants.h"
51 #ifdef HAVE_RPCSVC_NFS_PROT_H
52 #include <rpcsvc/nfs_prot.h>
54 #include <linux/nfs.h>
55 #define nfsstat nfs_stat
62 #define PMAP_TIMEOUT (10)
63 #define CONNECT_TIMEOUT (20)
64 #define MOUNT_TIMEOUT (30)
66 #if SIZEOF_SOCKLEN_T - 0 == 0
67 #define socklen_t unsigned int
70 extern int nfs_mount_data_version;
71 extern char *progname;
74 static const unsigned long nfs_to_mnt[] = {
81 static const unsigned long mnt_to_nfs[] = {
89 * Map an NFS version into the corresponding Mountd version
91 unsigned long nfsvers_to_mnt(const unsigned long vers)
94 return nfs_to_mnt[vers];
99 * Map a Mountd version into the corresponding NFS version
101 static unsigned long mntvers_to_nfs(const unsigned long vers)
104 return mnt_to_nfs[vers];
108 static const unsigned int probe_udp_only[] = {
113 static const unsigned int probe_udp_first[] = {
119 static const unsigned int probe_tcp_first[] = {
125 static const unsigned long probe_nfs2_only[] = {
130 static const unsigned long probe_nfs3_first[] = {
136 static const unsigned long probe_mnt1_first[] = {
142 static const unsigned long probe_mnt3_first[] = {
150 * nfs_name_to_address - resolve hostname to an IPv4 or IPv6 socket address
151 * @hostname: pointer to C string containing DNS hostname to resolve
152 * @sap: pointer to buffer to fill with socket address
153 * @len: IN: size of buffer to fill; OUT: size of socket address
155 * Returns 1 and places a socket address at @sap if successful;
158 int nfs_name_to_address(const char *hostname,
159 const sa_family_t af_hint,
160 struct sockaddr *sap, socklen_t *salen)
162 struct addrinfo *gai_results;
163 struct addrinfo gai_hint = {
164 .ai_family = af_hint,
165 .ai_flags = AI_ADDRCONFIG,
167 socklen_t len = *salen;
170 if (af_hint == AF_INET6)
171 gai_hint.ai_flags |= AI_V4MAPPED|AI_ALL;
175 error = getaddrinfo(hostname, NULL, &gai_hint, &gai_results);
177 nfs_error(_("%s: DNS resolution failed for %s: %s"),
178 progname, hostname, (error == EAI_SYSTEM ?
179 strerror(errno) : gai_strerror(error)));
183 switch (gai_results->ai_addr->sa_family) {
186 if (len >= gai_results->ai_addrlen) {
187 *salen = gai_results->ai_addrlen;
188 memcpy(sap, gai_results->ai_addr, *salen);
193 /* things are really broken if we get here, so warn */
194 nfs_error(_("%s: unrecognized DNS resolution results for %s"),
199 freeaddrinfo(gai_results);
204 * nfs_gethostbyname - resolve a hostname to an IPv4 address
205 * @hostname: pointer to a C string containing a DNS hostname
206 * @saddr: returns an IPv4 address
208 * Returns 1 if successful, otherwise zero.
210 int nfs_gethostbyname(const char *hostname, struct sockaddr_in *sin)
212 socklen_t len = sizeof(*sin);
214 return nfs_name_to_address(hostname, AF_INET,
215 (struct sockaddr *)sin, &len);
219 * Attempt to connect a socket, but time out after "timeout" seconds.
221 * On error return, caller closes the socket.
223 static int connect_to(int fd, struct sockaddr *addr,
224 socklen_t addrlen, int timeout)
228 struct timeval tv = {
232 saved = fcntl(fd, F_GETFL, 0);
233 fcntl(fd, F_SETFL, saved | O_NONBLOCK);
235 ret = connect(fd, addr, addrlen);
236 if (ret < 0 && errno != EINPROGRESS)
244 ret = select(fd + 1, &rset, &wset, NULL, &tv);
249 if (FD_ISSET(fd, &rset) || FD_ISSET(fd, &wset)) {
251 socklen_t len = sizeof(error);
252 if (getsockopt(fd, SOL_SOCKET, SO_ERROR, &error, &len) < 0)
262 fcntl(fd, F_SETFL, saved);
267 * Create a socket that is locally bound to a reserved or non-reserved port.
269 * The caller should check rpc_createerr to determine the cause of any error.
271 static int get_socket(struct sockaddr_in *saddr, unsigned int p_prot,
272 unsigned int timeout, int resvp, int conn)
275 struct sockaddr_in laddr;
276 socklen_t namelen = sizeof(laddr);
278 type = (p_prot == IPPROTO_UDP ? SOCK_DGRAM : SOCK_STREAM);
279 if ((so = socket (AF_INET, type, p_prot)) < 0)
282 laddr.sin_family = AF_INET;
284 laddr.sin_addr.s_addr = htonl(INADDR_ANY);
286 if (bindresvport(so, &laddr) < 0)
287 goto err_bindresvport;
289 cc = bind(so, (struct sockaddr *)&laddr, namelen);
293 if (type == SOCK_STREAM || (conn && type == SOCK_DGRAM)) {
294 cc = connect_to(so, (struct sockaddr *)saddr, namelen,
302 rpc_createerr.cf_stat = RPC_SYSTEMERROR;
303 rpc_createerr.cf_error.re_errno = errno;
305 nfs_error(_("%s: Unable to create %s socket: errno %d (%s)\n"),
306 progname, p_prot == IPPROTO_UDP ? _("UDP") : _("TCP"),
307 errno, strerror(errno));
312 rpc_createerr.cf_stat = RPC_SYSTEMERROR;
313 rpc_createerr.cf_error.re_errno = errno;
315 nfs_error(_("%s: Unable to bindresvport %s socket: errno %d"
317 progname, p_prot == IPPROTO_UDP ? _("UDP") : _("TCP"),
318 errno, strerror(errno));
324 rpc_createerr.cf_stat = RPC_SYSTEMERROR;
325 rpc_createerr.cf_error.re_errno = errno;
327 nfs_error(_("%s: Unable to bind to %s socket: errno %d (%s)\n"),
328 progname, p_prot == IPPROTO_UDP ? _("UDP") : _("TCP"),
329 errno, strerror(errno));
335 rpc_createerr.cf_stat = RPC_SYSTEMERROR;
336 rpc_createerr.cf_error.re_errno = errno;
338 nfs_error(_("%s: Unable to connect to %s:%d, errno %d (%s)\n"),
339 progname, inet_ntoa(saddr->sin_addr),
340 ntohs(saddr->sin_port), errno, strerror(errno));
347 * getport() is very similar to pmap_getport() with the exception that
348 * this version tries to use an ephemeral port, since reserved ports are
349 * not needed for GETPORT queries. This conserves the very limited
350 * reserved port space, which helps reduce failed socket binds
351 * during mount storms.
353 * A side effect of calling this function is that rpccreateerr is set.
355 static unsigned short getport(struct sockaddr_in *saddr,
356 unsigned long program,
357 unsigned long version,
360 struct sockaddr_in bind_saddr;
361 unsigned short port = 0;
367 bind_saddr.sin_port = htons(PMAPPORT);
369 socket = get_socket(&bind_saddr, proto, PMAP_TIMEOUT, FALSE, FALSE);
370 if (socket == RPC_ANYSOCK) {
371 if (proto == IPPROTO_TCP &&
372 rpc_createerr.cf_error.re_errno == ETIMEDOUT)
373 rpc_createerr.cf_stat = RPC_TIMEDOUT;
379 clnt = clntudp_bufcreate(&bind_saddr,
381 RETRY_TIMEOUT, &socket,
386 clnt = clnttcp_create(&bind_saddr,
389 RPCSMALLMSGSIZE, RPCSMALLMSGSIZE);
393 struct pmap parms = {
399 stat = clnt_call(clnt, PMAPPROC_GETPORT,
400 (xdrproc_t)xdr_pmap, (caddr_t)&parms,
401 (xdrproc_t)xdr_u_short, (caddr_t)&port,
404 clnt_geterr(clnt, &rpc_createerr.cf_error);
405 rpc_createerr.cf_stat = stat;
408 if (stat != RPC_SUCCESS)
411 rpc_createerr.cf_stat = RPC_PROGNOTREGISTERED;
419 * Use the portmapper to discover whether or not the service we want is
420 * available. The lists 'versions' and 'protos' define ordered sequences
421 * of service versions and udp/tcp protocols to probe for.
423 static int probe_port(clnt_addr_t *server, const unsigned long *versions,
424 const unsigned int *protos)
426 struct sockaddr_in *saddr = &server->saddr;
427 struct pmap *pmap = &server->pmap;
428 const unsigned long prog = pmap->pm_prog, *p_vers;
429 const unsigned int prot = (u_int)pmap->pm_prot, *p_prot;
430 const u_short port = (u_short) pmap->pm_port;
431 unsigned long vers = pmap->pm_vers;
432 unsigned short p_port;
434 p_prot = prot ? &prot : protos;
435 p_vers = vers ? &vers : versions;
436 rpc_createerr.cf_stat = 0;
438 p_port = getport(saddr, prog, *p_vers, *p_prot);
440 if (!port || port == p_port) {
441 saddr->sin_port = htons(p_port);
443 printf(_("%s: trying %s prog %ld vers "
444 "%ld prot %s port %d\n"),
446 inet_ntoa(saddr->sin_addr),
448 *p_prot == IPPROTO_UDP ?
452 if (clnt_ping(saddr, prog, *p_vers, *p_prot, NULL))
454 if (rpc_createerr.cf_stat == RPC_TIMEDOUT)
458 if (rpc_createerr.cf_stat != RPC_PROGNOTREGISTERED &&
459 rpc_createerr.cf_stat != RPC_PROGVERSMISMATCH)
467 if (vers || !*++p_vers)
476 pmap->pm_vers = *p_vers;
478 pmap->pm_prot = *p_prot;
480 pmap->pm_port = p_port;
481 rpc_createerr.cf_stat = 0;
485 static int probe_nfsport(clnt_addr_t *nfs_server)
487 struct pmap *pmap = &nfs_server->pmap;
489 if (pmap->pm_vers && pmap->pm_prot && pmap->pm_port)
492 if (nfs_mount_data_version >= 4)
493 return probe_port(nfs_server, probe_nfs3_first, probe_tcp_first);
495 return probe_port(nfs_server, probe_nfs2_only, probe_udp_only);
498 static int probe_mntport(clnt_addr_t *mnt_server)
500 struct pmap *pmap = &mnt_server->pmap;
502 if (pmap->pm_vers && pmap->pm_prot && pmap->pm_port)
505 if (nfs_mount_data_version >= 4)
506 return probe_port(mnt_server, probe_mnt3_first, probe_udp_first);
508 return probe_port(mnt_server, probe_mnt1_first, probe_udp_only);
512 * probe_bothports - discover the RPC endpoints of mountd and NFS server
513 * @mnt_server: pointer to address and pmap argument for mountd results
514 * @nfs_server: pointer to address and pmap argument for NFS server
516 * Returns 1 if successful, otherwise zero if some error occurred.
517 * Note that the arguments are both input and output arguments.
519 * A side effect of calling this function is that rpccreateerr is set.
521 int probe_bothports(clnt_addr_t *mnt_server, clnt_addr_t *nfs_server)
523 struct pmap *nfs_pmap = &nfs_server->pmap;
524 struct pmap *mnt_pmap = &mnt_server->pmap;
525 struct pmap save_nfs, save_mnt;
527 const unsigned long *probe_vers;
529 if (mnt_pmap->pm_vers && !nfs_pmap->pm_vers)
530 nfs_pmap->pm_vers = mntvers_to_nfs(mnt_pmap->pm_vers);
531 else if (nfs_pmap->pm_vers && !mnt_pmap->pm_vers)
532 mnt_pmap->pm_vers = nfsvers_to_mnt(nfs_pmap->pm_vers);
533 if (nfs_pmap->pm_vers)
536 memcpy(&save_nfs, nfs_pmap, sizeof(save_nfs));
537 memcpy(&save_mnt, mnt_pmap, sizeof(save_mnt));
538 probe_vers = (nfs_mount_data_version >= 4) ?
539 probe_mnt3_first : probe_mnt1_first;
541 for (; *probe_vers; probe_vers++) {
542 nfs_pmap->pm_vers = mntvers_to_nfs(*probe_vers);
543 if ((res = probe_nfsport(nfs_server) != 0)) {
544 mnt_pmap->pm_vers = *probe_vers;
545 if ((res = probe_mntport(mnt_server)) != 0)
547 memcpy(mnt_pmap, &save_mnt, sizeof(*mnt_pmap));
549 switch (rpc_createerr.cf_stat) {
550 case RPC_PROGVERSMISMATCH:
551 case RPC_PROGNOTREGISTERED:
556 memcpy(nfs_pmap, &save_nfs, sizeof(*nfs_pmap));
563 if (!probe_nfsport(nfs_server))
565 return probe_mntport(mnt_server);
568 static int probe_statd(void)
570 struct sockaddr_in addr;
573 memset(&addr, 0, sizeof(addr));
574 addr.sin_family = AF_INET;
575 addr.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
576 port = getport(&addr, 100024, 1, IPPROTO_UDP);
580 addr.sin_port = htons(port);
582 if (clnt_ping(&addr, 100024, 1, IPPROTO_UDP, NULL) <= 0)
589 * start_statd - attempt to start rpc.statd
591 * Returns 1 if statd is running; otherwise zero.
593 int start_statd(void)
603 if (stat(START_STATD, &stb) == 0) {
604 if (S_ISREG(stb.st_mode) && (stb.st_mode & S_IXUSR)) {
616 * nfs_call_umount - ask the server to remove a share from it's rmtab
617 * @mnt_server: address of RPC MNT program server
618 * @argp: directory path of share to "unmount"
620 * Returns one if the unmount call succeeded; zero if the unmount
621 * failed for any reason.
623 * Note that a side effect of calling this function is that rpccreateerr
626 int nfs_call_umount(clnt_addr_t *mnt_server, dirpath *argp)
629 enum clnt_stat res = 0;
632 if (!probe_mntport(mnt_server))
634 clnt = mnt_openclnt(mnt_server, &msock);
637 res = clnt_call(clnt, MOUNTPROC_UMNT,
638 (xdrproc_t)xdr_dirpath, (caddr_t)argp,
639 (xdrproc_t)xdr_void, NULL,
641 mnt_closeclnt(clnt, msock);
643 if (res == RPC_SUCCESS)
649 * mnt_openclnt - get a handle for a remote mountd service
650 * @mnt_server: address and pmap arguments of mountd service
651 * @msock: returns a file descriptor of the underlying transport socket
653 * Returns an active handle for the remote's mountd service
655 CLIENT *mnt_openclnt(clnt_addr_t *mnt_server, int *msock)
657 struct sockaddr_in *mnt_saddr = &mnt_server->saddr;
658 struct pmap *mnt_pmap = &mnt_server->pmap;
661 mnt_saddr->sin_port = htons((u_short)mnt_pmap->pm_port);
662 *msock = get_socket(mnt_saddr, mnt_pmap->pm_prot, MOUNT_TIMEOUT,
664 if (*msock == RPC_ANYSOCK) {
665 if (rpc_createerr.cf_error.re_errno == EADDRINUSE)
667 * Probably in-use by a TIME_WAIT connection,
668 * It is worth waiting a while and trying again.
670 rpc_createerr.cf_stat = RPC_TIMEDOUT;
674 switch (mnt_pmap->pm_prot) {
676 clnt = clntudp_bufcreate(mnt_saddr,
677 mnt_pmap->pm_prog, mnt_pmap->pm_vers,
678 RETRY_TIMEOUT, msock,
679 MNT_SENDBUFSIZE, MNT_RECVBUFSIZE);
682 clnt = clnttcp_create(mnt_saddr,
683 mnt_pmap->pm_prog, mnt_pmap->pm_vers,
685 MNT_SENDBUFSIZE, MNT_RECVBUFSIZE);
689 /* try to mount hostname:dirname */
690 clnt->cl_auth = authunix_create_default();
697 * mnt_closeclnt - terminate a handle for a remote mountd service
698 * @clnt: pointer to an active handle for a remote mountd service
699 * @msock: file descriptor of the underlying transport socket
702 void mnt_closeclnt(CLIENT *clnt, int msock)
704 auth_destroy(clnt->cl_auth);
710 * clnt_ping - send an RPC ping to the remote RPC service endpoint
711 * @saddr: server's address
712 * @prog: target RPC program number
713 * @vers: target RPC version number
714 * @prot: target RPC protocol
715 * @caddr: filled in with our network address
717 * Sigh... getport() doesn't actually check the version number.
718 * In order to make sure that the server actually supports the service
719 * we're requesting, we open and RPC client, and fire off a NULL
722 * caddr is the network address that the server will use to call us back.
723 * On multi-homed clients, this address depends on which NIC we use to
724 * route requests to the server.
726 * Returns one if successful, otherwise zero.
728 int clnt_ping(struct sockaddr_in *saddr, const unsigned long prog,
729 const unsigned long vers, const unsigned int prot,
730 struct sockaddr_in *caddr)
734 static char clnt_res;
735 struct sockaddr dissolve;
737 rpc_createerr.cf_stat = stat = 0;
738 sock = get_socket(saddr, prot, CONNECT_TIMEOUT, FALSE, TRUE);
739 if (sock == RPC_ANYSOCK) {
740 if (rpc_createerr.cf_error.re_errno == ETIMEDOUT) {
742 * TCP timeout. Bubble up the error to see
743 * how it should be handled.
745 rpc_createerr.cf_stat = RPC_TIMEDOUT;
751 /* Get the address of our end of this connection */
752 socklen_t len = sizeof(*caddr);
753 if (getsockname(sock, caddr, &len) != 0)
754 caddr->sin_family = 0;
759 /* The socket is connected (so we could getsockname successfully),
760 * but some servers on multi-homed hosts reply from
761 * the wrong address, so if we stay connected, we lose the reply.
763 dissolve.sa_family = AF_UNSPEC;
764 connect(sock, &dissolve, sizeof(dissolve));
766 clnt = clntudp_bufcreate(saddr, prog, vers,
767 RETRY_TIMEOUT, &sock,
768 RPCSMALLMSGSIZE, RPCSMALLMSGSIZE);
771 clnt = clnttcp_create(saddr, prog, vers, &sock,
772 RPCSMALLMSGSIZE, RPCSMALLMSGSIZE);
779 memset(&clnt_res, 0, sizeof(clnt_res));
780 stat = clnt_call(clnt, NULLPROC,
781 (xdrproc_t)xdr_void, (caddr_t)NULL,
782 (xdrproc_t)xdr_void, (caddr_t)&clnt_res,
785 clnt_geterr(clnt, &rpc_createerr.cf_error);
786 rpc_createerr.cf_stat = stat;
791 if (stat == RPC_SUCCESS)
798 * get_client_address - acquire our local network address
799 * @saddr: server's address
800 * @caddr: filled in with our network address
802 * Discover a network address that the server will use to call us back.
803 * On multi-homed clients, this address depends on which NIC we use to
804 * route requests to the server.
806 * Use a connected datagram socket so as not to leave a socket in TIME_WAIT.
808 * Returns one if successful, otherwise zero.
810 int get_client_address(struct sockaddr_in *saddr, struct sockaddr_in *caddr)
812 socklen_t len = sizeof(*caddr);
815 socket = get_socket(saddr, IPPROTO_UDP, CONNECT_TIMEOUT, FALSE, TRUE);
816 if (socket == RPC_ANYSOCK)
819 err = getsockname(socket, caddr, &len);
822 if (err && verbose) {
823 nfs_error(_("%s: getsockname failed: %s"),
824 progname, strerror(errno));