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 <rpc/pmap_prot.h>
38 #include <rpc/pmap_clnt.h>
39 #include <sys/socket.h>
44 #include "nfs_mount.h"
45 #include "mount_constants.h"
48 #ifdef HAVE_RPCSVC_NFS_PROT_H
49 #include <rpcsvc/nfs_prot.h>
51 #include <linux/nfs.h>
52 #define nfsstat nfs_stat
59 #define PMAP_TIMEOUT (10)
60 #define CONNECT_TIMEOUT (20)
61 #define MOUNT_TIMEOUT (30)
63 #if SIZEOF_SOCKLEN_T - 0 == 0
64 #define socklen_t unsigned int
67 extern int nfs_mount_data_version;
68 extern char *progname;
71 static const unsigned long nfs_to_mnt[] = {
78 static const unsigned long mnt_to_nfs[] = {
86 * Map an NFS version into the corresponding Mountd version
88 unsigned long nfsvers_to_mnt(const unsigned long vers)
91 return nfs_to_mnt[vers];
96 * Map a Mountd version into the corresponding NFS version
98 static unsigned long mntvers_to_nfs(const unsigned long vers)
101 return mnt_to_nfs[vers];
105 static const unsigned int probe_udp_only[] = {
110 static const unsigned int probe_udp_first[] = {
116 static const unsigned int probe_tcp_first[] = {
122 static const unsigned long probe_nfs2_only[] = {
127 static const unsigned long probe_nfs3_first[] = {
133 static const unsigned long probe_mnt1_first[] = {
139 static const unsigned long probe_mnt3_first[] = {
147 * nfs_name_to_address - resolve hostname to an IPv4 or IPv6 socket address
148 * @hostname: pointer to C string containing DNS hostname to resolve
149 * @sap: pointer to buffer to fill with socket address
150 * @len: IN: size of buffer to fill; OUT: size of socket address
152 * Returns 1 and places a socket address at @sap if successful;
155 int nfs_name_to_address(const char *hostname,
156 const sa_family_t af_hint,
157 struct sockaddr *sap, socklen_t *salen)
159 struct addrinfo *gai_results;
160 struct addrinfo gai_hint = {
161 .ai_family = af_hint,
162 .ai_flags = AI_ADDRCONFIG,
164 socklen_t len = *salen;
167 if (af_hint == AF_INET6)
168 gai_hint.ai_flags |= AI_V4MAPPED|AI_ALL;
172 error = getaddrinfo(hostname, NULL, &gai_hint, &gai_results);
174 nfs_error(_("%s: DNS resolution failed for %s: %s"),
175 progname, hostname, (error == EAI_SYSTEM ?
176 strerror(errno) : gai_strerror(error)));
180 switch (gai_results->ai_addr->sa_family) {
183 if (len >= gai_results->ai_addrlen) {
184 *salen = gai_results->ai_addrlen;
185 memcpy(sap, gai_results->ai_addr, *salen);
190 /* things are really broken if we get here, so warn */
191 nfs_error(_("%s: unrecognized DNS resolution results for %s"),
196 freeaddrinfo(gai_results);
201 * nfs_gethostbyname - resolve a hostname to an IPv4 address
202 * @hostname: pointer to a C string containing a DNS hostname
203 * @saddr: returns an IPv4 address
205 * Returns 1 if successful, otherwise zero.
207 int nfs_gethostbyname(const char *hostname, struct sockaddr_in *sin)
209 socklen_t len = sizeof(*sin);
211 return nfs_name_to_address(hostname, AF_INET,
212 (struct sockaddr *)sin, &len);
216 * Attempt to connect a socket, but time out after "timeout" seconds.
218 * On error return, caller closes the socket.
220 static int connect_to(int fd, struct sockaddr *addr,
221 socklen_t addrlen, int timeout)
225 struct timeval tv = {
229 saved = fcntl(fd, F_GETFL, 0);
230 fcntl(fd, F_SETFL, saved | O_NONBLOCK);
232 ret = connect(fd, addr, addrlen);
233 if (ret < 0 && errno != EINPROGRESS)
241 ret = select(fd + 1, &rset, &wset, NULL, &tv);
246 if (FD_ISSET(fd, &rset) || FD_ISSET(fd, &wset)) {
248 socklen_t len = sizeof(error);
249 if (getsockopt(fd, SOL_SOCKET, SO_ERROR, &error, &len) < 0)
259 fcntl(fd, F_SETFL, saved);
264 * Create a socket that is locally bound to a reserved or non-reserved port.
266 * The caller should check rpc_createerr to determine the cause of any error.
268 static int get_socket(struct sockaddr_in *saddr, unsigned int p_prot,
269 unsigned int timeout, int resvp, int conn)
272 struct sockaddr_in laddr;
273 socklen_t namelen = sizeof(laddr);
275 type = (p_prot == IPPROTO_UDP ? SOCK_DGRAM : SOCK_STREAM);
276 if ((so = socket (AF_INET, type, p_prot)) < 0)
279 laddr.sin_family = AF_INET;
281 laddr.sin_addr.s_addr = htonl(INADDR_ANY);
283 if (bindresvport(so, &laddr) < 0)
284 goto err_bindresvport;
286 cc = bind(so, (struct sockaddr *)&laddr, namelen);
290 if (type == SOCK_STREAM || (conn && type == SOCK_DGRAM)) {
291 cc = connect_to(so, (struct sockaddr *)saddr, namelen,
299 rpc_createerr.cf_stat = RPC_SYSTEMERROR;
300 rpc_createerr.cf_error.re_errno = errno;
302 nfs_error(_("%s: Unable to create %s socket: errno %d (%s)\n"),
303 progname, p_prot == IPPROTO_UDP ? _("UDP") : _("TCP"),
304 errno, strerror(errno));
309 rpc_createerr.cf_stat = RPC_SYSTEMERROR;
310 rpc_createerr.cf_error.re_errno = errno;
312 nfs_error(_("%s: Unable to bindresvport %s socket: errno %d"
314 progname, p_prot == IPPROTO_UDP ? _("UDP") : _("TCP"),
315 errno, strerror(errno));
321 rpc_createerr.cf_stat = RPC_SYSTEMERROR;
322 rpc_createerr.cf_error.re_errno = errno;
324 nfs_error(_("%s: Unable to bind to %s socket: errno %d (%s)\n"),
325 progname, p_prot == IPPROTO_UDP ? _("UDP") : _("TCP"),
326 errno, strerror(errno));
332 rpc_createerr.cf_stat = RPC_SYSTEMERROR;
333 rpc_createerr.cf_error.re_errno = errno;
335 nfs_error(_("%s: Unable to connect to %s:%d, errno %d (%s)\n"),
336 progname, inet_ntoa(saddr->sin_addr),
337 ntohs(saddr->sin_port), errno, strerror(errno));
344 * getport() is very similar to pmap_getport() with the exception that
345 * this version tries to use an ephemeral port, since reserved ports are
346 * not needed for GETPORT queries. This conserves the very limited
347 * reserved port space, which helps reduce failed socket binds
348 * during mount storms.
350 * A side effect of calling this function is that rpccreateerr is set.
352 static unsigned short getport(struct sockaddr_in *saddr,
353 unsigned long program,
354 unsigned long version,
357 struct sockaddr_in bind_saddr;
358 unsigned short port = 0;
364 bind_saddr.sin_port = htons(PMAPPORT);
366 socket = get_socket(&bind_saddr, proto, PMAP_TIMEOUT, FALSE, FALSE);
367 if (socket == RPC_ANYSOCK) {
368 if (proto == IPPROTO_TCP &&
369 rpc_createerr.cf_error.re_errno == ETIMEDOUT)
370 rpc_createerr.cf_stat = RPC_TIMEDOUT;
376 clnt = clntudp_bufcreate(&bind_saddr,
378 RETRY_TIMEOUT, &socket,
383 clnt = clnttcp_create(&bind_saddr,
386 RPCSMALLMSGSIZE, RPCSMALLMSGSIZE);
390 struct pmap parms = {
396 stat = clnt_call(clnt, PMAPPROC_GETPORT,
397 (xdrproc_t)xdr_pmap, (caddr_t)&parms,
398 (xdrproc_t)xdr_u_short, (caddr_t)&port,
401 clnt_geterr(clnt, &rpc_createerr.cf_error);
402 rpc_createerr.cf_stat = stat;
405 if (stat != RPC_SUCCESS)
408 rpc_createerr.cf_stat = RPC_PROGNOTREGISTERED;
416 * Use the portmapper to discover whether or not the service we want is
417 * available. The lists 'versions' and 'protos' define ordered sequences
418 * of service versions and udp/tcp protocols to probe for.
420 static int probe_port(clnt_addr_t *server, const unsigned long *versions,
421 const unsigned int *protos)
423 struct sockaddr_in *saddr = &server->saddr;
424 struct pmap *pmap = &server->pmap;
425 const unsigned long prog = pmap->pm_prog, *p_vers;
426 const unsigned int prot = (u_int)pmap->pm_prot, *p_prot;
427 const u_short port = (u_short) pmap->pm_port;
428 unsigned long vers = pmap->pm_vers;
429 unsigned short p_port;
431 p_prot = prot ? &prot : protos;
432 p_vers = vers ? &vers : versions;
433 rpc_createerr.cf_stat = 0;
435 p_port = getport(saddr, prog, *p_vers, *p_prot);
437 if (!port || port == p_port) {
438 saddr->sin_port = htons(p_port);
440 printf(_("%s: trying %s prog %ld vers "
441 "%ld prot %s port %d\n"),
443 inet_ntoa(saddr->sin_addr),
445 *p_prot == IPPROTO_UDP ?
449 if (clnt_ping(saddr, prog, *p_vers, *p_prot, NULL))
451 if (rpc_createerr.cf_stat == RPC_TIMEDOUT)
455 if (rpc_createerr.cf_stat != RPC_PROGNOTREGISTERED &&
456 rpc_createerr.cf_stat != RPC_PROGVERSMISMATCH)
464 if (vers || !*++p_vers)
473 pmap->pm_vers = *p_vers;
475 pmap->pm_prot = *p_prot;
477 pmap->pm_port = p_port;
478 rpc_createerr.cf_stat = 0;
482 static int probe_nfsport(clnt_addr_t *nfs_server)
484 struct pmap *pmap = &nfs_server->pmap;
486 if (pmap->pm_vers && pmap->pm_prot && pmap->pm_port)
489 if (nfs_mount_data_version >= 4)
490 return probe_port(nfs_server, probe_nfs3_first, probe_tcp_first);
492 return probe_port(nfs_server, probe_nfs2_only, probe_udp_only);
495 static int probe_mntport(clnt_addr_t *mnt_server)
497 struct pmap *pmap = &mnt_server->pmap;
499 if (pmap->pm_vers && pmap->pm_prot && pmap->pm_port)
502 if (nfs_mount_data_version >= 4)
503 return probe_port(mnt_server, probe_mnt3_first, probe_udp_first);
505 return probe_port(mnt_server, probe_mnt1_first, probe_udp_only);
509 * probe_bothports - discover the RPC endpoints of mountd and NFS server
510 * @mnt_server: pointer to address and pmap argument for mountd results
511 * @nfs_server: pointer to address and pmap argument for NFS server
513 * Returns 1 if successful, otherwise zero if some error occurred.
514 * Note that the arguments are both input and output arguments.
516 * A side effect of calling this function is that rpccreateerr is set.
518 int probe_bothports(clnt_addr_t *mnt_server, clnt_addr_t *nfs_server)
520 struct pmap *nfs_pmap = &nfs_server->pmap;
521 struct pmap *mnt_pmap = &mnt_server->pmap;
522 struct pmap save_nfs, save_mnt;
524 const unsigned long *probe_vers;
526 if (mnt_pmap->pm_vers && !nfs_pmap->pm_vers)
527 nfs_pmap->pm_vers = mntvers_to_nfs(mnt_pmap->pm_vers);
528 else if (nfs_pmap->pm_vers && !mnt_pmap->pm_vers)
529 mnt_pmap->pm_vers = nfsvers_to_mnt(nfs_pmap->pm_vers);
530 if (nfs_pmap->pm_vers)
533 memcpy(&save_nfs, nfs_pmap, sizeof(save_nfs));
534 memcpy(&save_mnt, mnt_pmap, sizeof(save_mnt));
535 probe_vers = (nfs_mount_data_version >= 4) ?
536 probe_mnt3_first : probe_mnt1_first;
538 for (; *probe_vers; probe_vers++) {
539 nfs_pmap->pm_vers = mntvers_to_nfs(*probe_vers);
540 if ((res = probe_nfsport(nfs_server) != 0)) {
541 mnt_pmap->pm_vers = *probe_vers;
542 if ((res = probe_mntport(mnt_server)) != 0)
544 memcpy(mnt_pmap, &save_mnt, sizeof(*mnt_pmap));
546 switch (rpc_createerr.cf_stat) {
547 case RPC_PROGVERSMISMATCH:
548 case RPC_PROGNOTREGISTERED:
553 memcpy(nfs_pmap, &save_nfs, sizeof(*nfs_pmap));
560 if (!probe_nfsport(nfs_server))
562 return probe_mntport(mnt_server);
565 static int probe_statd(void)
567 struct sockaddr_in addr;
570 memset(&addr, 0, sizeof(addr));
571 addr.sin_family = AF_INET;
572 addr.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
573 port = getport(&addr, 100024, 1, IPPROTO_UDP);
577 addr.sin_port = htons(port);
579 if (clnt_ping(&addr, 100024, 1, IPPROTO_UDP, NULL) <= 0)
586 * start_statd - attempt to start rpc.statd
588 * Returns 1 if statd is running; otherwise zero.
590 int start_statd(void)
600 if (stat(START_STATD, &stb) == 0) {
601 if (S_ISREG(stb.st_mode) && (stb.st_mode & S_IXUSR)) {
613 * nfs_call_umount - ask the server to remove a share from it's rmtab
614 * @mnt_server: address of RPC MNT program server
615 * @argp: directory path of share to "unmount"
617 * Returns one if the unmount call succeeded; zero if the unmount
618 * failed for any reason.
620 * Note that a side effect of calling this function is that rpccreateerr
623 int nfs_call_umount(clnt_addr_t *mnt_server, dirpath *argp)
626 enum clnt_stat res = 0;
629 if (!probe_mntport(mnt_server))
631 clnt = mnt_openclnt(mnt_server, &msock);
634 res = clnt_call(clnt, MOUNTPROC_UMNT,
635 (xdrproc_t)xdr_dirpath, (caddr_t)argp,
636 (xdrproc_t)xdr_void, NULL,
638 mnt_closeclnt(clnt, msock);
640 if (res == RPC_SUCCESS)
646 * mnt_openclnt - get a handle for a remote mountd service
647 * @mnt_server: address and pmap arguments of mountd service
648 * @msock: returns a file descriptor of the underlying transport socket
650 * Returns an active handle for the remote's mountd service
652 CLIENT *mnt_openclnt(clnt_addr_t *mnt_server, int *msock)
654 struct sockaddr_in *mnt_saddr = &mnt_server->saddr;
655 struct pmap *mnt_pmap = &mnt_server->pmap;
658 mnt_saddr->sin_port = htons((u_short)mnt_pmap->pm_port);
659 *msock = get_socket(mnt_saddr, mnt_pmap->pm_prot, MOUNT_TIMEOUT,
661 if (*msock == RPC_ANYSOCK) {
662 if (rpc_createerr.cf_error.re_errno == EADDRINUSE)
664 * Probably in-use by a TIME_WAIT connection,
665 * It is worth waiting a while and trying again.
667 rpc_createerr.cf_stat = RPC_TIMEDOUT;
671 switch (mnt_pmap->pm_prot) {
673 clnt = clntudp_bufcreate(mnt_saddr,
674 mnt_pmap->pm_prog, mnt_pmap->pm_vers,
675 RETRY_TIMEOUT, msock,
676 MNT_SENDBUFSIZE, MNT_RECVBUFSIZE);
679 clnt = clnttcp_create(mnt_saddr,
680 mnt_pmap->pm_prog, mnt_pmap->pm_vers,
682 MNT_SENDBUFSIZE, MNT_RECVBUFSIZE);
686 /* try to mount hostname:dirname */
687 clnt->cl_auth = authunix_create_default();
694 * mnt_closeclnt - terminate a handle for a remote mountd service
695 * @clnt: pointer to an active handle for a remote mountd service
696 * @msock: file descriptor of the underlying transport socket
699 void mnt_closeclnt(CLIENT *clnt, int msock)
701 auth_destroy(clnt->cl_auth);
707 * clnt_ping - send an RPC ping to the remote RPC service endpoint
708 * @saddr: server's address
709 * @prog: target RPC program number
710 * @vers: target RPC version number
711 * @prot: target RPC protocol
712 * @caddr: filled in with our network address
714 * Sigh... getport() doesn't actually check the version number.
715 * In order to make sure that the server actually supports the service
716 * we're requesting, we open and RPC client, and fire off a NULL
719 * caddr is the network address that the server will use to call us back.
720 * On multi-homed clients, this address depends on which NIC we use to
721 * route requests to the server.
723 * Returns one if successful, otherwise zero.
725 int clnt_ping(struct sockaddr_in *saddr, const unsigned long prog,
726 const unsigned long vers, const unsigned int prot,
727 struct sockaddr_in *caddr)
731 static char clnt_res;
732 struct sockaddr dissolve;
734 rpc_createerr.cf_stat = stat = 0;
735 sock = get_socket(saddr, prot, CONNECT_TIMEOUT, FALSE, TRUE);
736 if (sock == RPC_ANYSOCK) {
737 if (rpc_createerr.cf_error.re_errno == ETIMEDOUT) {
739 * TCP timeout. Bubble up the error to see
740 * how it should be handled.
742 rpc_createerr.cf_stat = RPC_TIMEDOUT;
748 /* Get the address of our end of this connection */
749 socklen_t len = sizeof(*caddr);
750 if (getsockname(sock, caddr, &len) != 0)
751 caddr->sin_family = 0;
756 /* The socket is connected (so we could getsockname successfully),
757 * but some servers on multi-homed hosts reply from
758 * the wrong address, so if we stay connected, we lose the reply.
760 dissolve.sa_family = AF_UNSPEC;
761 connect(sock, &dissolve, sizeof(dissolve));
763 clnt = clntudp_bufcreate(saddr, prog, vers,
764 RETRY_TIMEOUT, &sock,
765 RPCSMALLMSGSIZE, RPCSMALLMSGSIZE);
768 clnt = clnttcp_create(saddr, prog, vers, &sock,
769 RPCSMALLMSGSIZE, RPCSMALLMSGSIZE);
776 memset(&clnt_res, 0, sizeof(clnt_res));
777 stat = clnt_call(clnt, NULLPROC,
778 (xdrproc_t)xdr_void, (caddr_t)NULL,
779 (xdrproc_t)xdr_void, (caddr_t)&clnt_res,
782 clnt_geterr(clnt, &rpc_createerr.cf_error);
783 rpc_createerr.cf_stat = stat;
788 if (stat == RPC_SUCCESS)
795 * get_client_address - acquire our local network address
796 * @saddr: server's address
797 * @caddr: filled in with our network address
799 * Discover a network address that the server will use to call us back.
800 * On multi-homed clients, this address depends on which NIC we use to
801 * route requests to the server.
803 * Use a connected datagram socket so as not to leave a socket in TIME_WAIT.
805 * Returns one if successful, otherwise zero.
807 int get_client_address(struct sockaddr_in *saddr, struct sockaddr_in *caddr)
809 socklen_t len = sizeof(*caddr);
812 socket = get_socket(saddr, IPPROTO_UDP, CONNECT_TIMEOUT, FALSE, TRUE);
813 if (socket == RPC_ANYSOCK)
816 err = getsockname(socket, caddr, &len);
819 if (err && verbose) {
820 nfs_error(_("%s: getsockname failed: %s"),
821 progname, strerror(errno));