nfs_nfsdkrpc.c revision 223309 
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 */
33
34#include <sys/cdefs.h>
35__FBSDID("$FreeBSD: head/sys/fs/nfsserver/nfs_nfsdkrpc.c 223309 2011-06-19 22:08:55Z rmacklem $");
36
37#include "opt_inet6.h"
38#include "opt_kgssapi.h"
39
40#include <fs/nfs/nfsport.h>
41
42#include <rpc/rpc.h>
43#include <rpc/rpcsec_gss.h>
44
45#include <security/mac/mac_framework.h>
46
47NFSDLOCKMUTEX;
48
49/*
50 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
51 */
52static int newnfs_nfsv3_procid[NFS_V3NPROCS] = {
53	NFSPROC_NULL,
54	NFSPROC_GETATTR,
55	NFSPROC_SETATTR,
56	NFSPROC_NOOP,
57	NFSPROC_LOOKUP,
58	NFSPROC_READLINK,
59	NFSPROC_READ,
60	NFSPROC_NOOP,
61	NFSPROC_WRITE,
62	NFSPROC_CREATE,
63	NFSPROC_REMOVE,
64	NFSPROC_RENAME,
65	NFSPROC_LINK,
66	NFSPROC_SYMLINK,
67	NFSPROC_MKDIR,
68	NFSPROC_RMDIR,
69	NFSPROC_READDIR,
70	NFSPROC_FSSTAT,
71	NFSPROC_NOOP,
72	NFSPROC_NOOP,
73	NFSPROC_NOOP,
74	NFSPROC_NOOP,
75};
76
77
78SYSCTL_DECL(_vfs_nfsd);
79
80SVCPOOL		*nfsrvd_pool;
81
82static int	nfs_privport = 0;
83SYSCTL_INT(_vfs_nfsd, OID_AUTO, nfs_privport, CTLFLAG_RW,
84    &nfs_privport, 0,
85    "Only allow clients using a privileged port for NFSv2 and 3");
86
87static int	nfs_minvers = NFS_VER2;
88SYSCTL_INT(_vfs_nfsd, OID_AUTO, server_min_nfsvers, CTLFLAG_RW,
89    &nfs_minvers, 0, "The lowest version of NFS handled by the server");
90
91static int	nfs_maxvers = NFS_VER4;
92SYSCTL_INT(_vfs_nfsd, OID_AUTO, server_max_nfsvers, CTLFLAG_RW,
93    &nfs_maxvers, 0, "The highest version of NFS handled by the server");
94
95static int nfs_proc(struct nfsrv_descript *, u_int32_t, struct socket *,
96    u_int64_t, struct nfsrvcache **);
97
98extern u_long sb_max_adj;
99extern int newnfs_numnfsd;
100extern struct proc *nfsd_master_proc;
101
102/*
103 * NFS server system calls
104 */
105
106static void
107nfssvc_program(struct svc_req *rqst, SVCXPRT *xprt)
108{
109	struct nfsrv_descript nd;
110	struct nfsrvcache *rp = NULL;
111	int cacherep, credflavor;
112
113	memset(&nd, 0, sizeof(nd));
114	if (rqst->rq_vers == NFS_VER2) {
115		if (rqst->rq_proc > NFSV2PROC_STATFS) {
116			svcerr_noproc(rqst);
117			svc_freereq(rqst);
118			return;
119		}
120		nd.nd_procnum = newnfs_nfsv3_procid[rqst->rq_proc];
121		nd.nd_flag = ND_NFSV2;
122	} else if (rqst->rq_vers == NFS_VER3) {
123		if (rqst->rq_proc >= NFS_V3NPROCS) {
124			svcerr_noproc(rqst);
125			svc_freereq(rqst);
126			return;
127		}
128		nd.nd_procnum = rqst->rq_proc;
129		nd.nd_flag = ND_NFSV3;
130	} else {
131		if (rqst->rq_proc != NFSPROC_NULL &&
132		    rqst->rq_proc != NFSV4PROC_COMPOUND) {
133			svcerr_noproc(rqst);
134			svc_freereq(rqst);
135			return;
136		}
137		nd.nd_procnum = rqst->rq_proc;
138		nd.nd_flag = ND_NFSV4;
139	}
140
141	/*
142	 * Note: we want rq_addr, not svc_getrpccaller for nd_nam2 -
143	 * NFS_SRVMAXDATA uses a NULL value for nd_nam2 to detect TCP
144	 * mounts.
145	 */
146	nd.nd_mrep = rqst->rq_args;
147	rqst->rq_args = NULL;
148	newnfs_realign(&nd.nd_mrep);
149	nd.nd_md = nd.nd_mrep;
150	nd.nd_dpos = mtod(nd.nd_md, caddr_t);
151	nd.nd_nam = svc_getrpccaller(rqst);
152	nd.nd_nam2 = rqst->rq_addr;
153	nd.nd_mreq = NULL;
154	nd.nd_cred = NULL;
155
156	if (nfs_privport && (nd.nd_flag & ND_NFSV4) == 0) {
157		/* Check if source port is privileged */
158		u_short port;
159		struct sockaddr *nam = nd.nd_nam;
160		struct sockaddr_in *sin;
161
162		sin = (struct sockaddr_in *)nam;
163		/*
164		 * INET/INET6 - same code:
165		 *    sin_port and sin6_port are at same offset
166		 */
167		port = ntohs(sin->sin_port);
168		if (port >= IPPORT_RESERVED &&
169		    nd.nd_procnum != NFSPROC_NULL) {
170#ifdef INET6
171			char b6[INET6_ADDRSTRLEN];
172#if defined(KLD_MODULE)
173			/* Do not use ip6_sprintf: the nfs module should work without INET6. */
174#define	ip6_sprintf(buf, a)						\
175			(sprintf((buf), "%x:%x:%x:%x:%x:%x:%x:%x",	\
176			    (a)->s6_addr16[0], (a)->s6_addr16[1],	\
177			    (a)->s6_addr16[2], (a)->s6_addr16[3],	\
178			    (a)->s6_addr16[4], (a)->s6_addr16[5],	\
179			    (a)->s6_addr16[6], (a)->s6_addr16[7]),	\
180			    (buf))
181#endif
182#endif
183			printf("NFS request from unprivileged port (%s:%d)\n",
184#ifdef INET6
185			    sin->sin_family == AF_INET6 ?
186			    ip6_sprintf(b6, &satosin6(sin)->sin6_addr) :
187#if defined(KLD_MODULE)
188#undef ip6_sprintf
189#endif
190#endif
191			    inet_ntoa(sin->sin_addr), port);
192			svcerr_weakauth(rqst);
193			svc_freereq(rqst);
194			m_freem(nd.nd_mrep);
195			return;
196		}
197	}
198
199	if (nd.nd_procnum != NFSPROC_NULL) {
200		if (!svc_getcred(rqst, &nd.nd_cred, &credflavor)) {
201			svcerr_weakauth(rqst);
202			svc_freereq(rqst);
203			m_freem(nd.nd_mrep);
204			return;
205		}
206
207		/* Set the flag based on credflavor */
208		if (credflavor == RPCSEC_GSS_KRB5) {
209			nd.nd_flag |= ND_GSS;
210		} else if (credflavor == RPCSEC_GSS_KRB5I) {
211			nd.nd_flag |= (ND_GSS | ND_GSSINTEGRITY);
212		} else if (credflavor == RPCSEC_GSS_KRB5P) {
213			nd.nd_flag |= (ND_GSS | ND_GSSPRIVACY);
214		} else if (credflavor != AUTH_SYS) {
215			svcerr_weakauth(rqst);
216			svc_freereq(rqst);
217			m_freem(nd.nd_mrep);
218			return;
219		}
220
221#ifdef MAC
222		mac_cred_associate_nfsd(nd.nd_cred);
223#endif
224		if ((nd.nd_flag & ND_NFSV4) != 0) {
225			nd.nd_repstat = nfsvno_v4rootexport(&nd);
226			if (nd.nd_repstat != 0) {
227				svcerr_weakauth(rqst);
228				svc_freereq(rqst);
229				m_freem(nd.nd_mrep);
230				return;
231			}
232		}
233
234		cacherep = nfs_proc(&nd, rqst->rq_xid, xprt->xp_socket,
235		    xprt->xp_sockref, &rp);
236	} else {
237		NFSMGET(nd.nd_mreq);
238		nd.nd_mreq->m_len = 0;
239		cacherep = RC_REPLY;
240	}
241	if (nd.nd_mrep != NULL)
242		m_freem(nd.nd_mrep);
243
244	if (nd.nd_cred != NULL)
245		crfree(nd.nd_cred);
246
247	if (cacherep == RC_DROPIT) {
248		if (nd.nd_mreq != NULL)
249			m_freem(nd.nd_mreq);
250		svc_freereq(rqst);
251		return;
252	}
253
254	if (nd.nd_mreq == NULL) {
255		svcerr_decode(rqst);
256		svc_freereq(rqst);
257		return;
258	}
259
260	if (nd.nd_repstat & NFSERR_AUTHERR) {
261		svcerr_auth(rqst, nd.nd_repstat & ~NFSERR_AUTHERR);
262		if (nd.nd_mreq != NULL)
263			m_freem(nd.nd_mreq);
264	} else if (!svc_sendreply_mbuf(rqst, nd.nd_mreq)) {
265		svcerr_systemerr(rqst);
266	}
267	if (rp != NULL)
268		nfsrvd_sentcache(rp, xprt->xp_socket, 0);
269	svc_freereq(rqst);
270}
271
272/*
273 * Check the cache and, optionally, do the RPC.
274 * Return the appropriate cache response.
275 */
276static int
277nfs_proc(struct nfsrv_descript *nd, u_int32_t xid, struct socket *so,
278    u_int64_t sockref, struct nfsrvcache **rpp)
279{
280	struct thread *td = curthread;
281	int cacherep = RC_DOIT, isdgram;
282
283	*rpp = NULL;
284	if (nd->nd_nam2 == NULL) {
285		nd->nd_flag |= ND_STREAMSOCK;
286		isdgram = 0;
287	} else {
288		isdgram = 1;
289	}
290	NFSGETTIME(&nd->nd_starttime);
291
292	/*
293	 * Two cases:
294	 * 1 - For NFSv2 over UDP, if we are near our malloc/mget
295	 *     limit, just drop the request. There is no
296	 *     NFSERR_RESOURCE or NFSERR_DELAY for NFSv2 and the
297	 *     client will timeout/retry over UDP in a little while.
298	 * 2 - nd_repstat == 0 && nd_mreq == NULL, which
299	 *     means a normal nfs rpc, so check the cache
300	 */
301	if ((nd->nd_flag & ND_NFSV2) && nd->nd_nam2 != NULL &&
302	    nfsrv_mallocmget_limit()) {
303		cacherep = RC_DROPIT;
304	} else {
305		/*
306		 * For NFSv3, play it safe and assume that the client is
307		 * doing retries on the same TCP connection.
308		 */
309		if ((nd->nd_flag & (ND_NFSV4 | ND_STREAMSOCK)) ==
310		    ND_STREAMSOCK)
311			nd->nd_flag |= ND_SAMETCPCONN;
312		nd->nd_retxid = xid;
313		nd->nd_tcpconntime = NFSD_MONOSEC;
314		nd->nd_sockref = sockref;
315		cacherep = nfsrvd_getcache(nd, so);
316	}
317
318	/*
319	 * Handle the request. There are three cases.
320	 * RC_DOIT - do the RPC
321	 * RC_REPLY - return the reply already created
322	 * RC_DROPIT - just throw the request away
323	 */
324	if (cacherep == RC_DOIT) {
325		nfsrvd_dorpc(nd, isdgram, td);
326		if (nd->nd_repstat == NFSERR_DONTREPLY)
327			cacherep = RC_DROPIT;
328		else
329			cacherep = RC_REPLY;
330		*rpp = nfsrvd_updatecache(nd, so);
331	}
332	return (cacherep);
333}
334
335/*
336 * Adds a socket to the list for servicing by nfsds.
337 */
338int
339nfsrvd_addsock(struct file *fp)
340{
341	int siz;
342	struct socket *so;
343	int error;
344	SVCXPRT *xprt;
345	static u_int64_t sockref = 0;
346
347	so = fp->f_data;
348
349	siz = sb_max_adj;
350	error = soreserve(so, siz, siz);
351	if (error) {
352		return (error);
353	}
354
355	/*
356	 * Steal the socket from userland so that it doesn't close
357	 * unexpectedly.
358	 */
359	if (so->so_type == SOCK_DGRAM)
360		xprt = svc_dg_create(nfsrvd_pool, so, 0, 0);
361	else
362		xprt = svc_vc_create(nfsrvd_pool, so, 0, 0);
363	if (xprt) {
364		fp->f_ops = &badfileops;
365		fp->f_data = NULL;
366		xprt->xp_sockref = ++sockref;
367		if (nfs_minvers == NFS_VER2)
368			svc_reg(xprt, NFS_PROG, NFS_VER2, nfssvc_program,
369			    NULL);
370		if (nfs_minvers <= NFS_VER3 && nfs_maxvers >= NFS_VER3)
371			svc_reg(xprt, NFS_PROG, NFS_VER3, nfssvc_program,
372			    NULL);
373		if (nfs_maxvers >= NFS_VER4)
374			svc_reg(xprt, NFS_PROG, NFS_VER4, nfssvc_program,
375			    NULL);
376		SVC_RELEASE(xprt);
377	}
378
379	return (0);
380}
381
382/*
383 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
384 * until it is killed by a signal.
385 */
386int
387nfsrvd_nfsd(struct thread *td, struct nfsd_nfsd_args *args)
388{
389	char principal[MAXHOSTNAMELEN + 5];
390	int error;
391	bool_t ret2, ret3, ret4;
392
393	error = copyinstr(args->principal, principal, sizeof (principal),
394	    NULL);
395	if (error)
396		return (error);
397
398	/*
399	 * Only the first nfsd actually does any work. The RPC code
400	 * adds threads to it as needed. Any extra processes offered
401	 * by nfsd just exit. If nfsd is new enough, it will call us
402	 * once with a structure that specifies how many threads to
403	 * use.
404	 */
405	NFSD_LOCK();
406	if (newnfs_numnfsd == 0) {
407		newnfs_numnfsd++;
408
409		NFSD_UNLOCK();
410
411		/* An empty string implies AUTH_SYS only. */
412		if (principal[0] != '\0') {
413			ret2 = rpc_gss_set_svc_name_call(principal,
414			    "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER2);
415			ret3 = rpc_gss_set_svc_name_call(principal,
416			    "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER3);
417			ret4 = rpc_gss_set_svc_name_call(principal,
418			    "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER4);
419
420			if (!ret2 || !ret3 || !ret4)
421				printf("nfsd: can't register svc name\n");
422		}
423
424		nfsrvd_pool->sp_minthreads = args->minthreads;
425		nfsrvd_pool->sp_maxthreads = args->maxthreads;
426
427		svc_run(nfsrvd_pool);
428
429		if (principal[0] != '\0') {
430			rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER2);
431			rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER3);
432			rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER4);
433		}
434
435		NFSD_LOCK();
436		newnfs_numnfsd--;
437		nfsrvd_init(1);
438	}
439	NFSD_UNLOCK();
440
441	return (0);
442}
443
444/*
445 * Initialize the data structures for the server.
446 * Handshake with any new nfsds starting up to avoid any chance of
447 * corruption.
448 */
449void
450nfsrvd_init(int terminating)
451{
452
453	NFSD_LOCK_ASSERT();
454
455	if (terminating) {
456		nfsd_master_proc = NULL;
457		NFSD_UNLOCK();
458		svcpool_destroy(nfsrvd_pool);
459		nfsrvd_pool = NULL;
460		NFSD_LOCK();
461	}
462
463	NFSD_UNLOCK();
464
465	nfsrvd_pool = svcpool_create("nfsd", SYSCTL_STATIC_CHILDREN(_vfs_nfsd));
466	nfsrvd_pool->sp_rcache = NULL;
467	nfsrvd_pool->sp_assign = NULL;
468	nfsrvd_pool->sp_done = NULL;
469
470	NFSD_LOCK();
471}
472
473