nfs_nfsdkrpc.c revision 192780 
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 192780 2009-05-26 01:09:33Z 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
45NFSDLOCKMUTEX;
46
47/*
48 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
49 */
50static int newnfs_nfsv3_procid[NFS_V3NPROCS] = {
51	NFSPROC_NULL,
52	NFSPROC_GETATTR,
53	NFSPROC_SETATTR,
54	NFSPROC_NOOP,
55	NFSPROC_LOOKUP,
56	NFSPROC_READLINK,
57	NFSPROC_READ,
58	NFSPROC_NOOP,
59	NFSPROC_WRITE,
60	NFSPROC_CREATE,
61	NFSPROC_REMOVE,
62	NFSPROC_RENAME,
63	NFSPROC_LINK,
64	NFSPROC_SYMLINK,
65	NFSPROC_MKDIR,
66	NFSPROC_RMDIR,
67	NFSPROC_READDIR,
68	NFSPROC_FSSTAT,
69	NFSPROC_NOOP,
70	NFSPROC_NOOP,
71	NFSPROC_NOOP,
72	NFSPROC_NOOP,
73};
74
75
76SYSCTL_DECL(_vfs_newnfs);
77
78SVCPOOL		*nfsrvd_pool;
79
80static int	nfs_privport = 0;
81SYSCTL_INT(_vfs_newnfs, OID_AUTO, nfs_privport, CTLFLAG_RW,
82    &nfs_privport, 0,
83    "Only allow clients using a privileged port for NFSv2 and 3");
84
85static int nfs_proc(struct nfsrv_descript *, u_int32_t, struct socket *,
86    u_int64_t, struct nfsrvcache **);
87
88extern u_long sb_max_adj;
89extern int newnfs_numnfsd;
90
91/*
92 * NFS server system calls
93 */
94
95static void
96nfssvc_program(struct svc_req *rqst, SVCXPRT *xprt)
97{
98	struct nfsrv_descript nd;
99	struct nfsrvcache *rp = NULL;
100	int cacherep, credflavor;
101
102	memset(&nd, 0, sizeof(nd));
103	if (rqst->rq_vers == NFS_VER2) {
104		if (rqst->rq_proc > NFSV2PROC_STATFS) {
105			svcerr_noproc(rqst);
106			svc_freereq(rqst);
107			return;
108		}
109		nd.nd_procnum = newnfs_nfsv3_procid[rqst->rq_proc];
110		nd.nd_flag = ND_NFSV2;
111	} else if (rqst->rq_vers == NFS_VER3) {
112		if (rqst->rq_proc >= NFS_V3NPROCS) {
113			svcerr_noproc(rqst);
114			svc_freereq(rqst);
115			return;
116		}
117		nd.nd_procnum = rqst->rq_proc;
118		nd.nd_flag = ND_NFSV3;
119	} else {
120		if (rqst->rq_proc != NFSPROC_NULL &&
121		    rqst->rq_proc != NFSV4PROC_COMPOUND) {
122			svcerr_noproc(rqst);
123			svc_freereq(rqst);
124			return;
125		}
126		nd.nd_procnum = rqst->rq_proc;
127		nd.nd_flag = ND_NFSV4;
128	}
129
130	/*
131	 * Note: we want rq_addr, not svc_getrpccaller for nd_nam2 -
132	 * NFS_SRVMAXDATA uses a NULL value for nd_nam2 to detect TCP
133	 * mounts.
134	 */
135	nd.nd_mrep = rqst->rq_args;
136	rqst->rq_args = NULL;
137	newnfs_realign(&nd.nd_mrep);
138	nd.nd_md = nd.nd_mrep;
139	nd.nd_dpos = mtod(nd.nd_md, caddr_t);
140	nd.nd_nam = svc_getrpccaller(rqst);
141	nd.nd_nam2 = rqst->rq_addr;
142	nd.nd_mreq = NULL;
143	nd.nd_cred = NULL;
144
145	if (nfs_privport && (nd.nd_flag & ND_NFSV4) == 0) {
146		/* Check if source port is privileged */
147		u_short port;
148		struct sockaddr *nam = nd.nd_nam;
149		struct sockaddr_in *sin;
150
151		sin = (struct sockaddr_in *)nam;
152		/*
153		 * INET/INET6 - same code:
154		 *    sin_port and sin6_port are at same offset
155		 */
156		port = ntohs(sin->sin_port);
157		if (port >= IPPORT_RESERVED &&
158		    nd.nd_procnum != NFSPROC_NULL) {
159#ifdef INET6
160			char b6[INET6_ADDRSTRLEN];
161#if defined(KLD_MODULE)
162			/* Do not use ip6_sprintf: the nfs module should work without INET6. */
163#define	ip6_sprintf(buf, a)						\
164			(sprintf((buf), "%x:%x:%x:%x:%x:%x:%x:%x",	\
165			    (a)->s6_addr16[0], (a)->s6_addr16[1],	\
166			    (a)->s6_addr16[2], (a)->s6_addr16[3],	\
167			    (a)->s6_addr16[4], (a)->s6_addr16[5],	\
168			    (a)->s6_addr16[6], (a)->s6_addr16[7]),	\
169			    (buf))
170#endif
171#endif
172			printf("NFS request from unprivileged port (%s:%d)\n",
173#ifdef INET6
174			    sin->sin_family == AF_INET6 ?
175			    ip6_sprintf(b6, &satosin6(sin)->sin6_addr) :
176#if defined(KLD_MODULE)
177#undef ip6_sprintf
178#endif
179#endif
180			    inet_ntoa(sin->sin_addr), port);
181			svcerr_weakauth(rqst);
182			svc_freereq(rqst);
183			m_freem(nd.nd_mrep);
184			return;
185		}
186	}
187
188	if (nd.nd_procnum != NFSPROC_NULL) {
189		if (!svc_getcred(rqst, &nd.nd_cred, &credflavor)) {
190			svcerr_weakauth(rqst);
191			svc_freereq(rqst);
192			m_freem(nd.nd_mrep);
193			return;
194		}
195
196		/* Set the flag based on credflavor */
197		if (credflavor == RPCSEC_GSS_KRB5) {
198			nd.nd_flag |= ND_GSS;
199		} else if (credflavor == RPCSEC_GSS_KRB5I) {
200			nd.nd_flag |= (ND_GSS | ND_GSSINTEGRITY);
201		} else if (credflavor == RPCSEC_GSS_KRB5P) {
202			nd.nd_flag |= (ND_GSS | ND_GSSPRIVACY);
203		} else if (credflavor != AUTH_SYS) {
204			svcerr_weakauth(rqst);
205			svc_freereq(rqst);
206			m_freem(nd.nd_mrep);
207			return;
208		}
209
210#ifdef MAC
211		mac_cred_associate_nfsd(nd.nd_cred);
212#endif
213		if ((nd.nd_flag & ND_NFSV4) != 0) {
214			nd.nd_repstat = nfsvno_v4rootexport(&nd);
215			if (nd.nd_repstat != 0) {
216				svcerr_weakauth(rqst);
217				svc_freereq(rqst);
218				m_freem(nd.nd_mrep);
219				return;
220			}
221		}
222
223		cacherep = nfs_proc(&nd, rqst->rq_xid, xprt->xp_socket,
224		    xprt->xp_sockref, &rp);
225	} else {
226		NFSMGET(nd.nd_mreq);
227		nd.nd_mreq->m_len = 0;
228		cacherep = RC_REPLY;
229	}
230	if (nd.nd_mrep != NULL)
231		m_freem(nd.nd_mrep);
232
233	if (nd.nd_cred != NULL)
234		crfree(nd.nd_cred);
235
236	if (cacherep == RC_DROPIT) {
237		if (nd.nd_mreq != NULL)
238			m_freem(nd.nd_mreq);
239		svc_freereq(rqst);
240		return;
241	}
242
243	if (nd.nd_mreq == NULL) {
244		svcerr_decode(rqst);
245		svc_freereq(rqst);
246		return;
247	}
248
249	if (nd.nd_repstat & NFSERR_AUTHERR) {
250		svcerr_auth(rqst, nd.nd_repstat & ~NFSERR_AUTHERR);
251		if (nd.nd_mreq != NULL)
252			m_freem(nd.nd_mreq);
253	} else if (!svc_sendreply_mbuf(rqst, nd.nd_mreq)) {
254		svcerr_systemerr(rqst);
255	}
256	if (rp != NULL)
257		nfsrvd_sentcache(rp, xprt->xp_socket, 0);
258	svc_freereq(rqst);
259}
260
261/*
262 * Check the cache and, optionally, do the RPC.
263 * Return the appropriate cache response.
264 */
265static int
266nfs_proc(struct nfsrv_descript *nd, u_int32_t xid, struct socket *so,
267    u_int64_t sockref, struct nfsrvcache **rpp)
268{
269	struct thread *td = curthread;
270	int cacherep = RC_DOIT, isdgram;
271
272	*rpp = NULL;
273	if (nd->nd_nam2 == NULL) {
274		nd->nd_flag |= ND_STREAMSOCK;
275		isdgram = 0;
276	} else {
277		isdgram = 1;
278	}
279	NFSGETTIME(&nd->nd_starttime);
280
281	/*
282	 * Two cases:
283	 * 1 - For NFSv2 over UDP, if we are near our malloc/mget
284	 *     limit, just drop the request. There is no
285	 *     NFSERR_RESOURCE or NFSERR_DELAY for NFSv2 and the
286	 *     client will timeout/retry over UDP in a little while.
287	 * 2 - nd_repstat == 0 && nd_mreq == NULL, which
288	 *     means a normal nfs rpc, so check the cache
289	 */
290	if ((nd->nd_flag & ND_NFSV2) && nd->nd_nam2 != NULL &&
291	    nfsrv_mallocmget_limit()) {
292		cacherep = RC_DROPIT;
293	} else {
294		/*
295		 * For NFSv3, play it safe and assume that the client is
296		 * doing retries on the same TCP connection.
297		 */
298		if ((nd->nd_flag & (ND_NFSV4 | ND_STREAMSOCK)) ==
299		    ND_STREAMSOCK)
300			nd->nd_flag |= ND_SAMETCPCONN;
301		nd->nd_retxid = xid;
302		nd->nd_tcpconntime = NFSD_MONOSEC;
303		nd->nd_sockref = sockref;
304		cacherep = nfsrvd_getcache(nd, so);
305	}
306
307	/*
308	 * Handle the request. There are three cases.
309	 * RC_DOIT - do the RPC
310	 * RC_REPLY - return the reply already created
311	 * RC_DROPIT - just throw the request away
312	 */
313	if (cacherep == RC_DOIT) {
314		nfsrvd_dorpc(nd, isdgram, td);
315		if (nd->nd_repstat == NFSERR_DONTREPLY)
316			cacherep = RC_DROPIT;
317		else
318			cacherep = RC_REPLY;
319		*rpp = nfsrvd_updatecache(nd, so);
320	}
321	return (cacherep);
322}
323
324/*
325 * Adds a socket to the list for servicing by nfsds.
326 */
327int
328nfsrvd_addsock(struct file *fp)
329{
330	int siz;
331	struct socket *so;
332	int error;
333	SVCXPRT *xprt;
334	static u_int64_t sockref = 0;
335
336	so = fp->f_data;
337
338	siz = sb_max_adj;
339	error = soreserve(so, siz, siz);
340	if (error) {
341		return (error);
342	}
343
344	/*
345	 * Steal the socket from userland so that it doesn't close
346	 * unexpectedly.
347	 */
348	if (so->so_type == SOCK_DGRAM)
349		xprt = svc_dg_create(nfsrvd_pool, so, 0, 0);
350	else
351		xprt = svc_vc_create(nfsrvd_pool, so, 0, 0);
352	if (xprt) {
353		fp->f_ops = &badfileops;
354		fp->f_data = NULL;
355		xprt->xp_sockref = ++sockref;
356		svc_reg(xprt, NFS_PROG, NFS_VER2, nfssvc_program, NULL);
357		svc_reg(xprt, NFS_PROG, NFS_VER3, nfssvc_program, NULL);
358		svc_reg(xprt, NFS_PROG, NFS_VER4, nfssvc_program, NULL);
359	}
360
361	return (0);
362}
363
364/*
365 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
366 * until it is killed by a signal.
367 */
368int
369nfsrvd_nfsd(struct thread *td, struct nfsd_nfsd_args *args)
370{
371#ifdef KGSSAPI
372	char principal[MAXHOSTNAMELEN + 5];
373	int error;
374	bool_t ret2, ret3, ret4;
375#endif
376
377#ifdef KGSSAPI
378	error = copyinstr(args->principal, principal, sizeof (principal),
379	    NULL);
380	if (error)
381		return (error);
382#endif
383
384	/*
385	 * Only the first nfsd actually does any work. The RPC code
386	 * adds threads to it as needed. Any extra processes offered
387	 * by nfsd just exit. If nfsd is new enough, it will call us
388	 * once with a structure that specifies how many threads to
389	 * use.
390	 */
391	NFSD_LOCK();
392	if (newnfs_numnfsd == 0) {
393		newnfs_numnfsd++;
394
395		NFSD_UNLOCK();
396
397#ifdef KGSSAPI
398		/* An empty string implies AUTH_SYS only. */
399		if (principal[0] != '\0') {
400			ret2 = rpc_gss_set_svc_name(principal, "kerberosv5",
401			    GSS_C_INDEFINITE, NFS_PROG, NFS_VER2);
402			ret3 = rpc_gss_set_svc_name(principal, "kerberosv5",
403			    GSS_C_INDEFINITE, NFS_PROG, NFS_VER3);
404			ret4 = rpc_gss_set_svc_name(principal, "kerberosv5",
405			    GSS_C_INDEFINITE, NFS_PROG, NFS_VER4);
406
407			if (!ret2 || !ret3 || !ret4) {
408				NFSD_LOCK();
409				newnfs_numnfsd--;
410				nfsrvd_init(1);
411				NFSD_UNLOCK();
412				return (EAUTH);
413			}
414		}
415#endif
416
417		nfsrvd_pool->sp_minthreads = args->minthreads;
418		nfsrvd_pool->sp_maxthreads = args->maxthreads;
419
420		svc_run(nfsrvd_pool);
421
422#ifdef KGSSAPI
423		if (principal[0] != '\0') {
424			rpc_gss_clear_svc_name(NFS_PROG, NFS_VER2);
425			rpc_gss_clear_svc_name(NFS_PROG, NFS_VER3);
426			rpc_gss_clear_svc_name(NFS_PROG, NFS_VER4);
427		}
428#endif
429
430		NFSD_LOCK();
431		newnfs_numnfsd--;
432		nfsrvd_init(1);
433	}
434	NFSD_UNLOCK();
435
436	return (0);
437}
438
439/*
440 * Initialize the data structures for the server.
441 * Handshake with any new nfsds starting up to avoid any chance of
442 * corruption.
443 */
444void
445nfsrvd_init(int terminating)
446{
447
448	NFSD_LOCK_ASSERT();
449
450	if (terminating) {
451		NFSD_UNLOCK();
452		svcpool_destroy(nfsrvd_pool);
453		nfsrvd_pool = NULL;
454		NFSD_LOCK();
455	}
456
457	NFSD_UNLOCK();
458
459	nfsrvd_pool = svcpool_create("nfsd", SYSCTL_STATIC_CHILDREN(_vfs_newnfs));
460	nfsrvd_pool->sp_rcache = NULL;
461	nfsrvd_pool->sp_assign = NULL;
462	nfsrvd_pool->sp_done = NULL;
463
464	NFSD_LOCK();
465}
466
467