2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
37 #include <sys/param.h>
38 #include <sys/systm.h>
40 #include "opt_inet6.h"
41 #include "opt_kgssapi.h"
43 #include <fs/nfs/nfsport.h>
46 #include <rpc/rpcsec_gss.h>
48 #include <nfs/nfs_fha.h>
49 #include <fs/nfsserver/nfs_fha_new.h>
51 #include <security/mac/mac_framework.h>
55 struct nfsv4lock nfsd_suspend_lock;
58 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
60 int newnfs_nfsv3_procid[NFS_V3NPROCS] = {
86 SYSCTL_DECL(_vfs_nfsd);
90 static int nfs_privport = 0;
91 SYSCTL_INT(_vfs_nfsd, OID_AUTO, nfs_privport, CTLFLAG_RW,
93 "Only allow clients using a privileged port for NFSv2 and 3");
95 static int nfs_minvers = NFS_VER2;
96 SYSCTL_INT(_vfs_nfsd, OID_AUTO, server_min_nfsvers, CTLFLAG_RW,
97 &nfs_minvers, 0, "The lowest version of NFS handled by the server");
99 static int nfs_maxvers = NFS_VER4;
100 SYSCTL_INT(_vfs_nfsd, OID_AUTO, server_max_nfsvers, CTLFLAG_RW,
101 &nfs_maxvers, 0, "The highest version of NFS handled by the server");
103 static int nfs_proc(struct nfsrv_descript *, u_int32_t, SVCXPRT *xprt,
104 struct nfsrvcache **);
106 extern u_long sb_max_adj;
107 extern int newnfs_numnfsd;
108 extern struct proc *nfsd_master_proc;
111 * NFS server system calls
115 nfssvc_program(struct svc_req *rqst, SVCXPRT *xprt)
117 struct nfsrv_descript nd;
118 struct nfsrvcache *rp = NULL;
119 int cacherep, credflavor;
121 memset(&nd, 0, sizeof(nd));
122 if (rqst->rq_vers == NFS_VER2) {
123 if (rqst->rq_proc > NFSV2PROC_STATFS) {
128 nd.nd_procnum = newnfs_nfsv3_procid[rqst->rq_proc];
129 nd.nd_flag = ND_NFSV2;
130 } else if (rqst->rq_vers == NFS_VER3) {
131 if (rqst->rq_proc >= NFS_V3NPROCS) {
136 nd.nd_procnum = rqst->rq_proc;
137 nd.nd_flag = ND_NFSV3;
139 if (rqst->rq_proc != NFSPROC_NULL &&
140 rqst->rq_proc != NFSV4PROC_COMPOUND) {
145 nd.nd_procnum = rqst->rq_proc;
146 nd.nd_flag = ND_NFSV4;
150 * Note: we want rq_addr, not svc_getrpccaller for nd_nam2 -
151 * NFS_SRVMAXDATA uses a NULL value for nd_nam2 to detect TCP
154 nd.nd_mrep = rqst->rq_args;
155 rqst->rq_args = NULL;
156 newnfs_realign(&nd.nd_mrep, M_WAITOK);
157 nd.nd_md = nd.nd_mrep;
158 nd.nd_dpos = mtod(nd.nd_md, caddr_t);
159 nd.nd_nam = svc_getrpccaller(rqst);
160 nd.nd_nam2 = rqst->rq_addr;
164 if (nfs_privport && (nd.nd_flag & ND_NFSV4) == 0) {
165 /* Check if source port is privileged */
167 struct sockaddr *nam = nd.nd_nam;
168 struct sockaddr_in *sin;
170 sin = (struct sockaddr_in *)nam;
172 * INET/INET6 - same code:
173 * sin_port and sin6_port are at same offset
175 port = ntohs(sin->sin_port);
176 if (port >= IPPORT_RESERVED &&
177 nd.nd_procnum != NFSPROC_NULL) {
179 char b6[INET6_ADDRSTRLEN];
180 #if defined(KLD_MODULE)
181 /* Do not use ip6_sprintf: the nfs module should work without INET6. */
182 #define ip6_sprintf(buf, a) \
183 (sprintf((buf), "%x:%x:%x:%x:%x:%x:%x:%x", \
184 (a)->s6_addr16[0], (a)->s6_addr16[1], \
185 (a)->s6_addr16[2], (a)->s6_addr16[3], \
186 (a)->s6_addr16[4], (a)->s6_addr16[5], \
187 (a)->s6_addr16[6], (a)->s6_addr16[7]), \
191 printf("NFS request from unprivileged port (%s:%d)\n",
193 sin->sin_family == AF_INET6 ?
194 ip6_sprintf(b6, &satosin6(sin)->sin6_addr) :
195 #if defined(KLD_MODULE)
199 inet_ntoa(sin->sin_addr), port);
200 svcerr_weakauth(rqst);
207 if (nd.nd_procnum != NFSPROC_NULL) {
208 if (!svc_getcred(rqst, &nd.nd_cred, &credflavor)) {
209 svcerr_weakauth(rqst);
215 /* Set the flag based on credflavor */
216 if (credflavor == RPCSEC_GSS_KRB5) {
217 nd.nd_flag |= ND_GSS;
218 } else if (credflavor == RPCSEC_GSS_KRB5I) {
219 nd.nd_flag |= (ND_GSS | ND_GSSINTEGRITY);
220 } else if (credflavor == RPCSEC_GSS_KRB5P) {
221 nd.nd_flag |= (ND_GSS | ND_GSSPRIVACY);
222 } else if (credflavor != AUTH_SYS) {
223 svcerr_weakauth(rqst);
230 mac_cred_associate_nfsd(nd.nd_cred);
233 * Get a refcnt (shared lock) on nfsd_suspend_lock.
234 * NFSSVC_SUSPENDNFSD will take an exclusive lock on
235 * nfsd_suspend_lock to suspend these threads.
236 * This must be done here, before the check of
237 * nfsv4root exports by nfsvno_v4rootexport().
239 NFSLOCKV4ROOTMUTEX();
240 nfsv4_getref(&nfsd_suspend_lock, NULL, NFSV4ROOTLOCKMUTEXPTR,
242 NFSUNLOCKV4ROOTMUTEX();
244 if ((nd.nd_flag & ND_NFSV4) != 0) {
245 nd.nd_repstat = nfsvno_v4rootexport(&nd);
246 if (nd.nd_repstat != 0) {
247 NFSLOCKV4ROOTMUTEX();
248 nfsv4_relref(&nfsd_suspend_lock);
249 NFSUNLOCKV4ROOTMUTEX();
250 svcerr_weakauth(rqst);
257 cacherep = nfs_proc(&nd, rqst->rq_xid, xprt, &rp);
258 NFSLOCKV4ROOTMUTEX();
259 nfsv4_relref(&nfsd_suspend_lock);
260 NFSUNLOCKV4ROOTMUTEX();
263 nd.nd_mreq->m_len = 0;
266 if (nd.nd_mrep != NULL)
269 if (nd.nd_cred != NULL)
272 if (cacherep == RC_DROPIT) {
273 if (nd.nd_mreq != NULL)
279 if (nd.nd_mreq == NULL) {
285 if (nd.nd_repstat & NFSERR_AUTHERR) {
286 svcerr_auth(rqst, nd.nd_repstat & ~NFSERR_AUTHERR);
287 if (nd.nd_mreq != NULL)
289 } else if (!svc_sendreply_mbuf(rqst, nd.nd_mreq)) {
290 svcerr_systemerr(rqst);
293 nfsrvd_sentcache(rp, (rqst->rq_reply_seq != 0 ||
294 SVC_ACK(xprt, NULL)), rqst->rq_reply_seq);
303 * Check the cache and, optionally, do the RPC.
304 * Return the appropriate cache response.
307 nfs_proc(struct nfsrv_descript *nd, u_int32_t xid, SVCXPRT *xprt,
308 struct nfsrvcache **rpp)
310 struct thread *td = curthread;
311 int cacherep = RC_DOIT, isdgram;
315 if (nd->nd_nam2 == NULL) {
316 nd->nd_flag |= ND_STREAMSOCK;
324 * 1 - For NFSv2 over UDP, if we are near our malloc/mget
325 * limit, just drop the request. There is no
326 * NFSERR_RESOURCE or NFSERR_DELAY for NFSv2 and the
327 * client will timeout/retry over UDP in a little while.
328 * 2 - nd_repstat == 0 && nd_mreq == NULL, which
329 * means a normal nfs rpc, so check the cache
331 if ((nd->nd_flag & ND_NFSV2) && nd->nd_nam2 != NULL &&
332 nfsrv_mallocmget_limit()) {
333 cacherep = RC_DROPIT;
336 * For NFSv3, play it safe and assume that the client is
337 * doing retries on the same TCP connection.
339 if ((nd->nd_flag & (ND_NFSV4 | ND_STREAMSOCK)) ==
341 nd->nd_flag |= ND_SAMETCPCONN;
343 nd->nd_tcpconntime = NFSD_MONOSEC;
344 nd->nd_sockref = xprt->xp_sockref;
345 cacherep = nfsrvd_getcache(nd);
348 nfsrc_trimcache(xprt->xp_sockref, ack, 0);
352 * Handle the request. There are three cases.
353 * RC_DOIT - do the RPC
354 * RC_REPLY - return the reply already created
355 * RC_DROPIT - just throw the request away
357 if (cacherep == RC_DOIT) {
358 nfsrvd_dorpc(nd, isdgram, td);
359 if (nd->nd_repstat == NFSERR_DONTREPLY)
360 cacherep = RC_DROPIT;
363 *rpp = nfsrvd_updatecache(nd);
371 nfssvc_loss(SVCXPRT *xprt)
377 nfsrc_trimcache(xprt->xp_sockref, ack, 1);
381 * Adds a socket to the list for servicing by nfsds.
384 nfsrvd_addsock(struct file *fp)
390 static u_int64_t sockref = 0;
395 error = soreserve(so, siz, siz);
400 * Steal the socket from userland so that it doesn't close
403 if (so->so_type == SOCK_DGRAM)
404 xprt = svc_dg_create(nfsrvd_pool, so, 0, 0);
406 xprt = svc_vc_create(nfsrvd_pool, so, 0, 0);
408 fp->f_ops = &badfileops;
410 xprt->xp_sockref = ++sockref;
411 if (nfs_minvers == NFS_VER2)
412 svc_reg(xprt, NFS_PROG, NFS_VER2, nfssvc_program,
414 if (nfs_minvers <= NFS_VER3 && nfs_maxvers >= NFS_VER3)
415 svc_reg(xprt, NFS_PROG, NFS_VER3, nfssvc_program,
417 if (nfs_maxvers >= NFS_VER4)
418 svc_reg(xprt, NFS_PROG, NFS_VER4, nfssvc_program,
420 if (so->so_type == SOCK_STREAM)
421 svc_loss_reg(xprt, nfssvc_loss);
431 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
432 * until it is killed by a signal.
435 nfsrvd_nfsd(struct thread *td, struct nfsd_nfsd_args *args)
437 char principal[MAXHOSTNAMELEN + 5];
439 bool_t ret2, ret3, ret4;
441 error = copyinstr(args->principal, principal, sizeof (principal),
447 * Only the first nfsd actually does any work. The RPC code
448 * adds threads to it as needed. Any extra processes offered
449 * by nfsd just exit. If nfsd is new enough, it will call us
450 * once with a structure that specifies how many threads to
454 if (newnfs_numnfsd == 0) {
459 /* An empty string implies AUTH_SYS only. */
460 if (principal[0] != '\0') {
461 ret2 = rpc_gss_set_svc_name_call(principal,
462 "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER2);
463 ret3 = rpc_gss_set_svc_name_call(principal,
464 "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER3);
465 ret4 = rpc_gss_set_svc_name_call(principal,
466 "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER4);
468 if (!ret2 || !ret3 || !ret4)
469 printf("nfsd: can't register svc name\n");
472 nfsrvd_pool->sp_minthreads = args->minthreads;
473 nfsrvd_pool->sp_maxthreads = args->maxthreads;
475 svc_run(nfsrvd_pool);
477 if (principal[0] != '\0') {
478 rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER2);
479 rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER3);
480 rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER4);
495 * Initialize the data structures for the server.
496 * Handshake with any new nfsds starting up to avoid any chance of
500 nfsrvd_init(int terminating)
506 nfsd_master_proc = NULL;
508 svcpool_destroy(nfsrvd_pool);
515 nfsrvd_pool = svcpool_create("nfsd", SYSCTL_STATIC_CHILDREN(_vfs_nfsd));
516 nfsrvd_pool->sp_rcache = NULL;
517 nfsrvd_pool->sp_assign = fhanew_assign;
518 nfsrvd_pool->sp_done = fha_nd_complete;