2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
7 * This code is derived from software contributed to Berkeley by
8 * Rick Macklem at The University of Guelph.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
39 #include "opt_inet6.h"
40 #include "opt_kgssapi.h"
41 #include "opt_kern_tls.h"
43 #include <fs/nfs/nfsport.h>
46 #include <rpc/rpcsec_gss.h>
47 #include <rpc/rpcsec_tls.h>
49 #include <fs/nfsserver/nfs_fha_new.h>
51 #include <security/mac/mac_framework.h>
55 char *nfsrv_zeropnfsdat = NULL;
58 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
60 int newnfs_nfsv3_procid[NFS_V3NPROCS] = {
85 SYSCTL_DECL(_vfs_nfsd);
87 NFSD_VNET_DEFINE_STATIC(int, nfs_privport) = 0;
88 SYSCTL_INT(_vfs_nfsd, OID_AUTO, nfs_privport, CTLFLAG_NFSD_VNET | CTLFLAG_RWTUN,
89 &NFSD_VNET_NAME(nfs_privport), 0,
90 "Only allow clients using a privileged port for NFSv2, 3 and 4");
92 NFSD_VNET_DEFINE_STATIC(int, nfs_minvers) = NFS_VER2;
93 SYSCTL_INT(_vfs_nfsd, OID_AUTO, server_min_nfsvers,
94 CTLFLAG_NFSD_VNET | CTLFLAG_RWTUN, &NFSD_VNET_NAME(nfs_minvers), 0,
95 "The lowest version of NFS handled by the server");
97 NFSD_VNET_DEFINE_STATIC(int, nfs_maxvers) = NFS_VER4;
98 SYSCTL_INT(_vfs_nfsd, OID_AUTO, server_max_nfsvers,
99 CTLFLAG_NFSD_VNET | CTLFLAG_RWTUN, &NFSD_VNET_NAME(nfs_maxvers), 0,
100 "The highest version of NFS handled by the server");
102 static int nfs_proc(struct nfsrv_descript *, u_int32_t, SVCXPRT *xprt,
103 struct nfsrvcache **);
105 extern u_long sb_max_adj;
106 extern int newnfs_numnfsd;
107 extern time_t nfsdev_time;
108 extern int nfsrv_writerpc[NFS_NPROCS];
109 extern volatile int nfsrv_devidcnt;
110 extern struct nfsv4_opflag nfsv4_opflag[NFSV42_NOPS];
112 NFSD_VNET_DECLARE(struct proc *, nfsd_master_proc);
114 NFSD_VNET_DEFINE(SVCPOOL *, nfsrvd_pool);
115 NFSD_VNET_DEFINE(int, nfsrv_numnfsd) = 0;
116 NFSD_VNET_DEFINE(struct nfsv4lock, nfsd_suspend_lock);
118 NFSD_VNET_DEFINE_STATIC(bool, nfsrvd_inited) = false;
121 * NFS server system calls
125 nfssvc_program(struct svc_req *rqst, SVCXPRT *xprt)
127 struct nfsrv_descript nd;
128 struct nfsrvcache *rp = NULL;
129 int cacherep, credflavor;
134 NFSD_CURVNET_SET_QUIET(NFSD_TD_TO_VNET(curthread));
135 memset(&nd, 0, sizeof(nd));
136 if (rqst->rq_vers == NFS_VER2) {
137 if (rqst->rq_proc > NFSV2PROC_STATFS ||
138 newnfs_nfsv3_procid[rqst->rq_proc] == NFSPROC_NOOP) {
143 nd.nd_procnum = newnfs_nfsv3_procid[rqst->rq_proc];
144 nd.nd_flag = ND_NFSV2;
145 } else if (rqst->rq_vers == NFS_VER3) {
146 if (rqst->rq_proc >= NFS_V3NPROCS) {
151 nd.nd_procnum = rqst->rq_proc;
152 nd.nd_flag = ND_NFSV3;
154 if (rqst->rq_proc != NFSPROC_NULL &&
155 rqst->rq_proc != NFSV4PROC_COMPOUND) {
160 nd.nd_procnum = rqst->rq_proc;
161 nd.nd_flag = ND_NFSV4;
165 * Note: we want rq_addr, not svc_getrpccaller for nd_nam2 -
166 * NFS_SRVMAXDATA uses a NULL value for nd_nam2 to detect TCP
169 nd.nd_mrep = rqst->rq_args;
170 rqst->rq_args = NULL;
171 newnfs_realign(&nd.nd_mrep, M_WAITOK);
172 nd.nd_md = nd.nd_mrep;
173 nd.nd_dpos = mtod(nd.nd_md, caddr_t);
174 nd.nd_nam = svc_getrpccaller(rqst);
175 nd.nd_nam2 = rqst->rq_addr;
179 if (NFSD_VNET(nfs_privport) != 0) {
180 /* Check if source port is privileged */
182 struct sockaddr *nam = nd.nd_nam;
183 struct sockaddr_in *sin;
185 sin = (struct sockaddr_in *)nam;
187 * INET/INET6 - same code:
188 * sin_port and sin6_port are at same offset
190 port = ntohs(sin->sin_port);
191 if (port >= IPPORT_RESERVED &&
192 nd.nd_procnum != NFSPROC_NULL) {
194 char buf[INET6_ADDRSTRLEN];
196 char buf[INET_ADDRSTRLEN];
199 #if defined(KLD_MODULE)
200 /* Do not use ip6_sprintf: the nfs module should work without INET6. */
201 #define ip6_sprintf(buf, a) \
202 (sprintf((buf), "%x:%x:%x:%x:%x:%x:%x:%x", \
203 (a)->s6_addr16[0], (a)->s6_addr16[1], \
204 (a)->s6_addr16[2], (a)->s6_addr16[3], \
205 (a)->s6_addr16[4], (a)->s6_addr16[5], \
206 (a)->s6_addr16[6], (a)->s6_addr16[7]), \
210 printf("NFS request from unprivileged port (%s:%d)\n",
212 sin->sin_family == AF_INET6 ?
213 ip6_sprintf(buf, &satosin6(sin)->sin6_addr) :
214 #if defined(KLD_MODULE)
218 inet_ntoa_r(sin->sin_addr, buf), port);
219 svcerr_weakauth(rqst);
226 if (nd.nd_procnum != NFSPROC_NULL) {
227 if (!svc_getcred(rqst, &nd.nd_cred, &credflavor)) {
228 svcerr_weakauth(rqst);
234 /* Set the flag based on credflavor */
235 if (credflavor == RPCSEC_GSS_KRB5) {
236 nd.nd_flag |= ND_GSS;
237 } else if (credflavor == RPCSEC_GSS_KRB5I) {
238 nd.nd_flag |= (ND_GSS | ND_GSSINTEGRITY);
239 } else if (credflavor == RPCSEC_GSS_KRB5P) {
240 nd.nd_flag |= (ND_GSS | ND_GSSPRIVACY);
241 } else if (credflavor != AUTH_SYS) {
242 svcerr_weakauth(rqst);
248 if ((xprt->xp_tls & RPCTLS_FLAGS_HANDSHAKE) != 0) {
249 nd.nd_flag |= ND_TLS;
250 if ((xprt->xp_tls & RPCTLS_FLAGS_VERIFIED) != 0)
251 nd.nd_flag |= ND_TLSCERT;
252 if ((xprt->xp_tls & RPCTLS_FLAGS_CERTUSER) != 0)
253 nd.nd_flag |= ND_TLSCERTUSER;
255 nd.nd_maxextsiz = 16384;
257 mac_cred_associate_nfsd(nd.nd_cred);
260 * Get a refcnt (shared lock) on nfsd_suspend_lock.
261 * NFSSVC_SUSPENDNFSD will take an exclusive lock on
262 * nfsd_suspend_lock to suspend these threads.
263 * The call to nfsv4_lock() that precedes nfsv4_getref()
264 * ensures that the acquisition of the exclusive lock
265 * takes priority over acquisition of the shared lock by
266 * waiting for any exclusive lock request to complete.
267 * This must be done here, before the check of
268 * nfsv4root exports by nfsvno_v4rootexport().
270 NFSLOCKV4ROOTMUTEX();
271 nfsv4_lock(&NFSD_VNET(nfsd_suspend_lock), 0, NULL,
272 NFSV4ROOTLOCKMUTEXPTR, NULL);
273 nfsv4_getref(&NFSD_VNET(nfsd_suspend_lock), NULL,
274 NFSV4ROOTLOCKMUTEXPTR, NULL);
275 NFSUNLOCKV4ROOTMUTEX();
277 if ((nd.nd_flag & ND_NFSV4) != 0) {
278 nd.nd_repstat = nfsvno_v4rootexport(&nd);
279 if (nd.nd_repstat != 0) {
280 NFSLOCKV4ROOTMUTEX();
281 nfsv4_relref(&NFSD_VNET(nfsd_suspend_lock));
282 NFSUNLOCKV4ROOTMUTEX();
283 svcerr_weakauth(rqst);
291 if ((xprt->xp_tls & RPCTLS_FLAGS_HANDSHAKE) != 0 &&
292 rpctls_getinfo(&maxlen, false, false))
293 nd.nd_maxextsiz = maxlen;
295 cacherep = nfs_proc(&nd, rqst->rq_xid, xprt, &rp);
296 NFSLOCKV4ROOTMUTEX();
297 nfsv4_relref(&NFSD_VNET(nfsd_suspend_lock));
298 NFSUNLOCKV4ROOTMUTEX();
301 nd.nd_mreq->m_len = 0;
304 if (nd.nd_mrep != NULL)
307 if (nd.nd_cred != NULL)
310 if (cacherep == RC_DROPIT) {
311 if (nd.nd_mreq != NULL)
317 if (nd.nd_mreq == NULL) {
323 if (nd.nd_repstat & NFSERR_AUTHERR) {
324 svcerr_auth(rqst, nd.nd_repstat & ~NFSERR_AUTHERR);
325 if (nd.nd_mreq != NULL)
327 } else if (!svc_sendreply_mbuf(rqst, nd.nd_mreq)) {
328 svcerr_systemerr(rqst);
331 nfsrvd_sentcache(rp, (rqst->rq_reply_seq != 0 ||
332 SVC_ACK(xprt, NULL)), rqst->rq_reply_seq);
337 NFSD_CURVNET_RESTORE();
338 ast_kclear(curthread);
343 * Check the cache and, optionally, do the RPC.
344 * Return the appropriate cache response.
347 nfs_proc(struct nfsrv_descript *nd, u_int32_t xid, SVCXPRT *xprt,
348 struct nfsrvcache **rpp)
350 int cacherep = RC_DOIT, isdgram, taglen = -1;
352 u_char tag[NFSV4_SMALLSTR + 1], *tagstr = NULL;
353 u_int32_t minorvers = 0;
357 if (nd->nd_nam2 == NULL) {
358 nd->nd_flag |= ND_STREAMSOCK;
366 * 1 - For NFSv2 over UDP, if we are near our malloc/mget
367 * limit, just drop the request. There is no
368 * NFSERR_RESOURCE or NFSERR_DELAY for NFSv2 and the
369 * client will timeout/retry over UDP in a little while.
370 * 2 - nd_repstat == 0 && nd_mreq == NULL, which
371 * means a normal nfs rpc, so check the cache
373 if ((nd->nd_flag & ND_NFSV2) && nd->nd_nam2 != NULL &&
374 nfsrv_mallocmget_limit()) {
375 cacherep = RC_DROPIT;
378 * For NFSv3, play it safe and assume that the client is
379 * doing retries on the same TCP connection.
381 if ((nd->nd_flag & (ND_NFSV4 | ND_STREAMSOCK)) ==
383 nd->nd_flag |= ND_SAMETCPCONN;
385 nd->nd_tcpconntime = NFSD_MONOSEC;
386 nd->nd_sockref = xprt->xp_sockref;
387 if ((nd->nd_flag & ND_NFSV4) != 0)
388 nfsd_getminorvers(nd, tag, &tagstr, &taglen,
390 if ((nd->nd_flag & ND_NFSV41) != 0)
391 /* NFSv4.1 caches replies in the session slots. */
394 cacherep = nfsrvd_getcache(nd);
397 nfsrc_trimcache(xprt->xp_sockref, ack, 0);
402 * Handle the request. There are three cases.
403 * RC_DOIT - do the RPC
404 * RC_REPLY - return the reply already created
405 * RC_DROPIT - just throw the request away
407 if (cacherep == RC_DOIT) {
408 if ((nd->nd_flag & ND_NFSV41) != 0)
410 nfsrvd_dorpc(nd, isdgram, tagstr, taglen, minorvers);
411 if ((nd->nd_flag & ND_NFSV41) != 0) {
412 if (nd->nd_repstat != NFSERR_REPLYFROMCACHE &&
413 (nd->nd_flag & ND_SAVEREPLY) != 0) {
414 /* Cache a copy of the reply. */
415 m = m_copym(nd->nd_mreq, 0, M_COPYALL,
419 if ((nd->nd_flag & ND_HASSEQUENCE) != 0)
420 nfsrv_cache_session(nd, &m);
421 if (nd->nd_repstat == NFSERR_REPLYFROMCACHE) {
424 m_freem(nd->nd_mreq);
430 if (nd->nd_repstat == NFSERR_DONTREPLY)
431 cacherep = RC_DROPIT;
434 *rpp = nfsrvd_updatecache(nd);
437 if (tagstr != NULL && taglen > NFSV4_SMALLSTR)
438 free(tagstr, M_TEMP);
445 nfssvc_loss(SVCXPRT *xprt)
451 NFSD_CURVNET_SET(NFSD_TD_TO_VNET(curthread));
452 nfsrc_trimcache(xprt->xp_sockref, ack, 1);
453 NFSD_CURVNET_RESTORE();
457 * Adds a socket to the list for servicing by nfsds.
460 nfsrvd_addsock(struct file *fp)
466 static u_int64_t sockref = 0;
471 error = soreserve(so, siz, siz);
476 * Steal the socket from userland so that it doesn't close
479 if (so->so_type == SOCK_DGRAM)
480 xprt = svc_dg_create(NFSD_VNET(nfsrvd_pool), so, 0, 0);
482 xprt = svc_vc_create(NFSD_VNET(nfsrvd_pool), so, 0, 0);
484 fp->f_ops = &badfileops;
486 xprt->xp_sockref = ++sockref;
487 if (NFSD_VNET(nfs_minvers) == NFS_VER2)
488 svc_reg(xprt, NFS_PROG, NFS_VER2, nfssvc_program,
490 if (NFSD_VNET(nfs_minvers) <= NFS_VER3 &&
491 NFSD_VNET(nfs_maxvers) >= NFS_VER3)
492 svc_reg(xprt, NFS_PROG, NFS_VER3, nfssvc_program,
494 if (NFSD_VNET(nfs_maxvers) >= NFS_VER4)
495 svc_reg(xprt, NFS_PROG, NFS_VER4, nfssvc_program,
497 if (so->so_type == SOCK_STREAM)
498 svc_loss_reg(xprt, nfssvc_loss);
509 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
510 * until it is killed by a signal.
513 nfsrvd_nfsd(struct thread *td, struct nfsd_nfsd_args *args)
515 char principal[MAXHOSTNAMELEN + 5];
518 bool_t ret2, ret3, ret4;
520 error = copyinstr(args->principal, principal, sizeof (principal),
526 * Only the first nfsd actually does any work. The RPC code
527 * adds threads to it as needed. Any extra processes offered
528 * by nfsd just exit. If nfsd is new enough, it will call us
529 * once with a structure that specifies how many threads to
533 if (NFSD_VNET(nfsrv_numnfsd) == 0) {
535 nfsdev_time = time_second;
538 p->p_flag2 |= P2_AST_SU;
540 newnfs_numnfsd++; /* Total num for all vnets. */
541 NFSD_VNET(nfsrv_numnfsd)++; /* Num for this vnet. */
544 error = nfsrv_createdevids(args, td);
546 /* An empty string implies AUTH_SYS only. */
547 if (principal[0] != '\0') {
548 ret2 = rpc_gss_set_svc_name_call(principal,
549 "kerberosv5", GSS_C_INDEFINITE, NFS_PROG,
551 ret3 = rpc_gss_set_svc_name_call(principal,
552 "kerberosv5", GSS_C_INDEFINITE, NFS_PROG,
554 ret4 = rpc_gss_set_svc_name_call(principal,
555 "kerberosv5", GSS_C_INDEFINITE, NFS_PROG,
558 if (!ret2 || !ret3 || !ret4)
559 printf("nfsd: can't register svc "
560 "name %s jid:%d\n", principal,
561 td->td_ucred->cr_prison->pr_id);
564 NFSD_VNET(nfsrvd_pool)->sp_minthreads =
566 NFSD_VNET(nfsrvd_pool)->sp_maxthreads =
570 * If this is a pNFS service, make Getattr do a
571 * vn_start_write(), so it can do a vn_set_extattr().
573 if (nfsrv_devidcnt > 0) {
574 nfsrv_writerpc[NFSPROC_GETATTR] = 1;
575 nfsv4_opflag[NFSV4OP_GETATTR].modifyfs = 1;
578 svc_run(NFSD_VNET(nfsrvd_pool));
580 /* Reset Getattr to not do a vn_start_write(). */
581 nfsrv_writerpc[NFSPROC_GETATTR] = 0;
582 nfsv4_opflag[NFSV4OP_GETATTR].modifyfs = 0;
584 if (principal[0] != '\0') {
585 rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER2);
586 rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER3);
587 rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER4);
592 NFSD_VNET(nfsrv_numnfsd)--;
595 p->p_flag2 &= ~P2_AST_SU;
606 * Initialize the data structures for the server.
607 * Handshake with any new nfsds starting up to avoid any chance of
611 nfsrvd_init(int terminating)
617 NFSD_VNET(nfsd_master_proc) = NULL;
619 nfsrv_freealllayoutsanddevids();
620 nfsrv_freeallbackchannel_xprts();
621 svcpool_close(NFSD_VNET(nfsrvd_pool));
622 free(nfsrv_zeropnfsdat, M_TEMP);
623 nfsrv_zeropnfsdat = NULL;
626 /* Initialize per-vnet globals once per vnet. */
627 if (NFSD_VNET(nfsrvd_inited))
629 NFSD_VNET(nfsrvd_inited) = true;
631 NFSD_VNET(nfsrvd_pool) = svcpool_create("nfsd",
632 SYSCTL_STATIC_CHILDREN(_vfs_nfsd));
633 NFSD_VNET(nfsrvd_pool)->sp_rcache = NULL;
634 NFSD_VNET(nfsrvd_pool)->sp_assign = fhanew_assign;
635 NFSD_VNET(nfsrvd_pool)->sp_done = fhanew_nd_complete;