1 /*-
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
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.
19 *
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
30 * SUCH DAMAGE.
31 *
32 * @(#)nfs_syscalls.c 8.5 (Berkeley) 3/30/95
33 */
34
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37
38 #include "opt_inet6.h"
39 #include "opt_kgssapi.h"
40
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/sysproto.h>
44 #include <sys/kernel.h>
45 #include <sys/sysctl.h>
46 #include <sys/file.h>
47 #include <sys/filedesc.h>
48 #include <sys/jail.h>
49 #include <sys/vnode.h>
50 #include <sys/malloc.h>
51 #include <sys/mount.h>
52 #include <sys/priv.h>
53 #include <sys/proc.h>
54 #include <sys/bio.h>
55 #include <sys/buf.h>
56 #include <sys/mbuf.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/domain.h>
60 #include <sys/protosw.h>
61 #include <sys/namei.h>
62 #include <sys/fcntl.h>
63 #include <sys/lockf.h>
64 #include <sys/eventhandler.h>
65
66 #include <netinet/in.h>
67 #include <netinet/tcp.h>
68 #ifdef INET6
69 #include <net/if.h>
70 #include <netinet6/in6_var.h>
71 #endif
72
73 #include <rpc/rpc.h>
74 #include <rpc/rpcsec_gss.h>
75 #include <rpc/replay.h>
76
77 #include <nfs/xdr_subs.h>
78 #include <nfs/nfsproto.h>
79 #include <nfsserver/nfs.h>
80 #include <nfsserver/nfsm_subs.h>
81 #include <nfsserver/nfsrvcache.h>
82 #include <nfsserver/nfs_fha.h>
83
84 #include <security/mac/mac_framework.h>
85
86 static MALLOC_DEFINE(M_NFSSVC, "nfss_srvsock", "Nfs server structure");
87
88 MALLOC_DEFINE(M_NFSRVDESC, "nfss_srvdesc", "NFS server socket descriptor");
89 MALLOC_DEFINE(M_NFSD, "nfss_daemon", "Nfs server daemon structure");
90
91 #define TRUE 1
92 #define FALSE 0
93
94 SYSCTL_DECL(_vfs_nfsrv);
95
96 SVCPOOL *nfsrv_pool;
97 int nfsd_waiting = 0;
98 int nfsrv_numnfsd = 0;
99 struct callout nfsrv_callout;
100 static eventhandler_tag nfsrv_nmbclusters_tag;
101
102 static int nfs_privport = 0;
103 SYSCTL_INT(_vfs_nfsrv, NFS_NFSPRIVPORT, nfs_privport, CTLFLAG_RW,
104 &nfs_privport, 0,
105 "Only allow clients using a privileged port");
106 SYSCTL_INT(_vfs_nfsrv, OID_AUTO, gatherdelay, CTLFLAG_RW,
107 &nfsrvw_procrastinate, 0,
108 "Delay value for write gathering");
109 SYSCTL_INT(_vfs_nfsrv, OID_AUTO, gatherdelay_v3, CTLFLAG_RW,
110 &nfsrvw_procrastinate_v3, 0,
111 "Delay in seconds for NFSv3 write gathering");
112
113 static int nfssvc_addsock(struct file *, struct thread *);
114 static int nfssvc_nfsd(struct thread *, struct nfsd_nfsd_args *);
115
116 extern u_long sb_max_adj;
117
118 int32_t (*nfsrv3_procs[NFS_NPROCS])(struct nfsrv_descript *nd,
119 struct nfssvc_sock *slp, struct mbuf **mreqp) = {
120 nfsrv_null,
121 nfsrv_getattr,
122 nfsrv_setattr,
123 nfsrv_lookup,
124 nfsrv3_access,
125 nfsrv_readlink,
126 nfsrv_read,
127 nfsrv_write,
128 nfsrv_create,
129 nfsrv_mkdir,
130 nfsrv_symlink,
131 nfsrv_mknod,
132 nfsrv_remove,
133 nfsrv_rmdir,
134 nfsrv_rename,
135 nfsrv_link,
136 nfsrv_readdir,
137 nfsrv_readdirplus,
138 nfsrv_statfs,
139 nfsrv_fsinfo,
140 nfsrv_pathconf,
141 nfsrv_commit,
142 nfsrv_noop
143 };
144
145 /*
146 * NFS server system calls
147 */
148 /*
149 * This is now called from nfssvc() in nfs/nfs_nfssvc.c.
150 */
151
152 /*
153 * Nfs server psuedo system call for the nfsd's
154 * Based on the flag value it either:
155 * - adds a socket to the selection list
156 * - remains in the kernel as an nfsd
157 * - remains in the kernel as an nfsiod
158 * For INET6 we suppose that nfsd provides only IN6P_IPV6_V6ONLY sockets
159 * and that mountd provides
160 * - sockaddr with no IPv4-mapped addresses
161 * - mask for both INET and INET6 families if there is IPv4-mapped overlap
162 */
163 int
164 nfssvc_nfsserver(struct thread *td, struct nfssvc_args *uap)
165 {
166 struct file *fp;
167 struct nfsd_addsock_args addsockarg;
168 struct nfsd_nfsd_args nfsdarg;
169 int error;
170
171 if (uap->flag & NFSSVC_ADDSOCK) {
172 error = copyin(uap->argp, (caddr_t)&addsockarg,
173 sizeof(addsockarg));
174 if (error)
175 return (error);
176 if ((error = fget(td, addsockarg.sock, &fp)) != 0)
177 return (error);
178 if (fp->f_type != DTYPE_SOCKET) {
179 fdrop(fp, td);
180 return (error); /* XXXRW: Should be EINVAL? */
181 }
182 error = nfssvc_addsock(fp, td);
183 fdrop(fp, td);
184 } else if (uap->flag & NFSSVC_OLDNFSD)
185 error = nfssvc_nfsd(td, NULL);
186 else if (uap->flag & NFSSVC_NFSD) {
187 if (!uap->argp)
188 return (EINVAL);
189 error = copyin(uap->argp, (caddr_t)&nfsdarg,
190 sizeof(nfsdarg));
191 if (error)
192 return (error);
193 error = nfssvc_nfsd(td, &nfsdarg);
194 } else
195 error = ENXIO;
196 return (error);
197 }
198
199 /*
200 * Generate the rpc reply header
201 * siz arg. is used to decide if adding a cluster is worthwhile
202 */
203 struct mbuf *
204 nfs_rephead(int siz, struct nfsrv_descript *nd, int err,
205 struct mbuf **mbp, caddr_t *bposp)
206 {
207 u_int32_t *tl;
208 struct mbuf *mreq;
209 caddr_t bpos;
210 struct mbuf *mb;
211
212 if (err == EBADRPC)
213 return (NULL);
214
215 nd->nd_repstat = err;
216 if (err && (nd->nd_flag & ND_NFSV3) == 0) /* XXX recheck */
217 siz = 0;
218
219 MGET(mreq, M_WAIT, MT_DATA);
220
221 /*
222 * If this is a big reply, use a cluster
223 */
224 mreq->m_len = 0;
225 if (siz >= MINCLSIZE) {
226 MCLGET(mreq, M_WAIT);
227 }
228 mb = mreq;
229 bpos = mtod(mb, caddr_t);
230
231 if (err != NFSERR_RETVOID) {
232 tl = nfsm_build(u_int32_t *, NFSX_UNSIGNED);
233 if (err)
234 *tl = txdr_unsigned(nfsrv_errmap(nd, err));
235 else
236 *tl = 0;
237 }
238
239 *mbp = mb;
240 *bposp = bpos;
241 if (err != 0 && err != NFSERR_RETVOID)
242 nfsrvstats.srvrpc_errs++;
243
244 return (mreq);
245 }
246
247 static void
248 nfssvc_program(struct svc_req *rqst, SVCXPRT *xprt)
249 {
250 rpcproc_t procnum;
251 int32_t (*proc)(struct nfsrv_descript *nd, struct nfssvc_sock *slp,
252 struct mbuf **mreqp);
253 int flag;
254 struct nfsrv_descript nd;
255 struct mbuf *mreq, *mrep;
256 int error;
257
258 if (rqst->rq_vers == NFS_VER2) {
259 if (rqst->rq_proc > NFSV2PROC_STATFS) {
260 svcerr_noproc(rqst);
261 svc_freereq(rqst);
262 return;
263 }
264 procnum = nfsrv_nfsv3_procid[rqst->rq_proc];
265 flag = 0;
266 } else {
267 if (rqst->rq_proc >= NFS_NPROCS) {
268 svcerr_noproc(rqst);
269 svc_freereq(rqst);
270 return;
271 }
272 procnum = rqst->rq_proc;
273 flag = ND_NFSV3;
274 }
275 proc = nfsrv3_procs[procnum];
276
277 mreq = mrep = NULL;
278 mreq = rqst->rq_args;
279 rqst->rq_args = NULL;
280 (void)nfs_realign(&mreq, M_WAIT);
281
282 /*
283 * Note: we want rq_addr, not svc_getrpccaller for nd_nam2 -
284 * NFS_SRVMAXDATA uses a NULL value for nd_nam2 to detect TCP
285 * mounts.
286 */
287 memset(&nd, 0, sizeof(nd));
288 nd.nd_md = nd.nd_mrep = mreq;
289 nd.nd_dpos = mtod(mreq, caddr_t);
290 nd.nd_nam = svc_getrpccaller(rqst);
291 nd.nd_nam2 = rqst->rq_addr;
292 nd.nd_procnum = procnum;
293 nd.nd_cr = NULL;
294 nd.nd_flag = flag;
295
296 if (nfs_privport) {
297 /* Check if source port is privileged */
298 u_short port;
299 struct sockaddr *nam = nd.nd_nam;
300 struct sockaddr_in *sin;
301
302 sin = (struct sockaddr_in *)nam;
303 /*
304 * INET/INET6 - same code:
305 * sin_port and sin6_port are at same offset
306 */
307 port = ntohs(sin->sin_port);
308 if (port >= IPPORT_RESERVED &&
309 nd.nd_procnum != NFSPROC_NULL) {
310 #ifdef INET6
311 char b6[INET6_ADDRSTRLEN];
312 #if defined(KLD_MODULE)
313 /* Do not use ip6_sprintf: the nfs module should work without INET6. */
314 #define ip6_sprintf(buf, a) \
315 (sprintf((buf), "%x:%x:%x:%x:%x:%x:%x:%x", \
316 (a)->s6_addr16[0], (a)->s6_addr16[1], \
317 (a)->s6_addr16[2], (a)->s6_addr16[3], \
318 (a)->s6_addr16[4], (a)->s6_addr16[5], \
319 (a)->s6_addr16[6], (a)->s6_addr16[7]), \
320 (buf))
321 #endif
322 #endif
323 printf("NFS request from unprivileged port (%s:%d)\n",
324 #ifdef INET6
325 sin->sin_family == AF_INET6 ?
326 ip6_sprintf(b6, &satosin6(sin)->sin6_addr) :
327 #if defined(KLD_MODULE)
328 #undef ip6_sprintf
329 #endif
330 #endif
331 inet_ntoa(sin->sin_addr), port);
332 m_freem(mreq);
333 svcerr_weakauth(rqst);
334 svc_freereq(rqst);
335 return;
336 }
337 }
338
339 if (proc != nfsrv_null) {
340 if (!svc_getcred(rqst, &nd.nd_cr, &nd.nd_credflavor)) {
341 m_freem(mreq);
342 svcerr_weakauth(rqst);
343 svc_freereq(rqst);
344 return;
345 }
346 #ifdef MAC
347 mac_cred_associate_nfsd(nd.nd_cr);
348 #endif
349 }
350 nfsrvstats.srvrpccnt[nd.nd_procnum]++;
351
352 error = proc(&nd, NULL, &mrep);
353
354 if (nd.nd_cr)
355 crfree(nd.nd_cr);
356
357 if (mrep == NULL) {
358 svcerr_decode(rqst);
359 svc_freereq(rqst);
360 return;
361 }
362 if (error && error != NFSERR_RETVOID) {
363 svcerr_systemerr(rqst);
364 svc_freereq(rqst);
365 return;
366 }
367 if (nd.nd_repstat & NFSERR_AUTHERR) {
368 svcerr_auth(rqst, nd.nd_repstat & ~NFSERR_AUTHERR);
369 m_freem(mrep);
370 } else {
371 if (!svc_sendreply_mbuf(rqst, mrep))
372 svcerr_systemerr(rqst);
373 }
374 svc_freereq(rqst);
375 }
376
377 /*
378 * Adds a socket to the list for servicing by nfsds.
379 */
380 static int
381 nfssvc_addsock(struct file *fp, struct thread *td)
382 {
383 int siz;
384 struct socket *so;
385 int error;
386 SVCXPRT *xprt;
387
388 so = fp->f_data;
389
390 siz = sb_max_adj;
391 error = soreserve(so, siz, siz);
392 if (error)
393 return (error);
394
395 /*
396 * Steal the socket from userland so that it doesn't close
397 * unexpectedly.
398 */
399 if (so->so_type == SOCK_DGRAM)
400 xprt = svc_dg_create(nfsrv_pool, so, 0, 0);
401 else
402 xprt = svc_vc_create(nfsrv_pool, so, 0, 0);
403 if (xprt) {
404 fp->f_ops = &badfileops;
405 fp->f_data = NULL;
406 svc_reg(xprt, NFS_PROG, NFS_VER2, nfssvc_program, NULL);
407 svc_reg(xprt, NFS_PROG, NFS_VER3, nfssvc_program, NULL);
408 SVC_RELEASE(xprt);
409 }
410
411 return (0);
412 }
413
414 /*
415 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
416 * until it is killed by a signal.
417 */
418 static int
419 nfssvc_nfsd(struct thread *td, struct nfsd_nfsd_args *args)
420 {
421 char principal[128];
422 int error;
423
424 if (args) {
425 error = copyinstr(args->principal, principal,
426 sizeof(principal), NULL);
427 if (error)
428 return (error);
429 } else {
430 memcpy(principal, "nfs@", 4);
431 getcredhostname(td->td_ucred, principal + 4,
432 sizeof(principal) - 4);
433 }
434
435 /*
436 * Only the first nfsd actually does any work. The RPC code
437 * adds threads to it as needed. Any extra processes offered
438 * by nfsd just exit. If nfsd is new enough, it will call us
439 * once with a structure that specifies how many threads to
440 * use.
441 */
442 NFSD_LOCK();
443 if (nfsrv_numnfsd == 0) {
444 nfsrv_numnfsd++;
445
446 NFSD_UNLOCK();
447
448 rpc_gss_set_svc_name_call(principal, "kerberosv5",
449 GSS_C_INDEFINITE, NFS_PROG, NFS_VER2);
450 rpc_gss_set_svc_name_call(principal, "kerberosv5",
451 GSS_C_INDEFINITE, NFS_PROG, NFS_VER3);
452
453 if (args) {
454 nfsrv_pool->sp_minthreads = args->minthreads;
455 nfsrv_pool->sp_maxthreads = args->maxthreads;
456 } else {
457 nfsrv_pool->sp_minthreads = 4;
458 nfsrv_pool->sp_maxthreads = 4;
459 }
460
461 svc_run(nfsrv_pool);
462
463 rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER2);
464 rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER3);
465
466 NFSD_LOCK();
467 nfsrv_numnfsd--;
468 nfsrv_init(TRUE);
469 }
470 NFSD_UNLOCK();
471
472 return (0);
473 }
474
475 /*
476 * Size the NFS server's duplicate request cache at 1/2 the
477 * nmbclusters, floating within a (64, 2048) range. This is to
478 * prevent all mbuf clusters being tied up in the NFS dupreq
479 * cache for small values of nmbclusters.
480 */
481 static size_t
482 nfsrv_replay_size(void)
483 {
484 size_t replaysiz;
485
486 replaysiz = nmbclusters / 2;
487 if (replaysiz > NFSRVCACHE_MAX_SIZE)
488 replaysiz = NFSRVCACHE_MAX_SIZE;
489 if (replaysiz < NFSRVCACHE_MIN_SIZE)
490 replaysiz = NFSRVCACHE_MIN_SIZE;
491 replaysiz *= MCLBYTES;
492
493 return (replaysiz);
494 }
495
496 /*
497 * Called when nmbclusters changes - we resize the replay cache
498 * accordingly.
499 */
500 static void
501 nfsrv_nmbclusters_change(void *tag)
502 {
503
504 if (nfsrv_pool)
505 replay_setsize(nfsrv_pool->sp_rcache, nfsrv_replay_size());
506 }
507
508 /*
509 * Initialize the data structures for the server.
510 * Handshake with any new nfsds starting up to avoid any chance of
511 * corruption.
512 */
513 void
514 nfsrv_init(int terminating)
515 {
516
517 NFSD_LOCK_ASSERT();
518
519 if (terminating) {
520 NFSD_UNLOCK();
521 EVENTHANDLER_DEREGISTER(nmbclusters_change,
522 nfsrv_nmbclusters_tag);
523 svcpool_destroy(nfsrv_pool);
524 nfsrv_pool = NULL;
525 NFSD_LOCK();
526 } else
527 nfs_pub.np_valid = 0;
528
529 NFSD_UNLOCK();
530
531 nfsrv_pool = svcpool_create("nfsd", SYSCTL_STATIC_CHILDREN(_vfs_nfsrv));
532 nfsrv_pool->sp_rcache = replay_newcache(nfsrv_replay_size());
533 nfsrv_pool->sp_assign = fha_assign;
534 nfsrv_pool->sp_done = fha_nd_complete;
535 nfsrv_nmbclusters_tag = EVENTHANDLER_REGISTER(nmbclusters_change,
536 nfsrv_nmbclusters_change, NULL, EVENTHANDLER_PRI_FIRST);
537
538 NFSD_LOCK();
539 }
Cache object: 8b19bdee441929124220e1cfd28e57e8
|