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: stable/10/sys/fs/nfsserver/nfs_nfsdkrpc.c 314034 2017-02-21 09:29:46Z avg $");
36
37#include <sys/param.h>
38#include <sys/systm.h>
39
40#include "opt_inet6.h"
41#include "opt_kgssapi.h"
42
43#include <fs/nfs/nfsport.h>
44
45#include <rpc/rpc.h>
46#include <rpc/rpcsec_gss.h>
47
48#include <nfs/nfs_fha.h>
49#include <fs/nfsserver/nfs_fha_new.h>
50
51#include <security/mac/mac_framework.h>
52
53NFSDLOCKMUTEX;
54NFSV4ROOTLOCKMUTEX;
55struct nfsv4lock nfsd_suspend_lock;
56
57/*
58 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
59 */
60int newnfs_nfsv3_procid[NFS_V3NPROCS] = {
61	NFSPROC_NULL,
62	NFSPROC_GETATTR,
63	NFSPROC_SETATTR,
64	NFSPROC_NOOP,
65	NFSPROC_LOOKUP,
66	NFSPROC_READLINK,
67	NFSPROC_READ,
68	NFSPROC_NOOP,
69	NFSPROC_WRITE,
70	NFSPROC_CREATE,
71	NFSPROC_REMOVE,
72	NFSPROC_RENAME,
73	NFSPROC_LINK,
74	NFSPROC_SYMLINK,
75	NFSPROC_MKDIR,
76	NFSPROC_RMDIR,
77	NFSPROC_READDIR,
78	NFSPROC_FSSTAT,
79	NFSPROC_NOOP,
80	NFSPROC_NOOP,
81	NFSPROC_NOOP,
82	NFSPROC_NOOP,
83};
84
85
86SYSCTL_DECL(_vfs_nfsd);
87
88SVCPOOL		*nfsrvd_pool;
89
90static int	nfs_privport = 0;
91SYSCTL_INT(_vfs_nfsd, OID_AUTO, nfs_privport, CTLFLAG_RWTUN,
92    &nfs_privport, 0,
93    "Only allow clients using a privileged port for NFSv2 and 3");
94
95static int	nfs_minvers = NFS_VER2;
96SYSCTL_INT(_vfs_nfsd, OID_AUTO, server_min_nfsvers, CTLFLAG_RWTUN,
97    &nfs_minvers, 0, "The lowest version of NFS handled by the server");
98
99static int	nfs_maxvers = NFS_VER4;
100SYSCTL_INT(_vfs_nfsd, OID_AUTO, server_max_nfsvers, CTLFLAG_RWTUN,
101    &nfs_maxvers, 0, "The highest version of NFS handled by the server");
102
103static int nfs_proc(struct nfsrv_descript *, u_int32_t, SVCXPRT *xprt,
104    struct nfsrvcache **);
105
106extern u_long sb_max_adj;
107extern int newnfs_numnfsd;
108extern struct proc *nfsd_master_proc;
109
110/*
111 * NFS server system calls
112 */
113
114static void
115nfssvc_program(struct svc_req *rqst, SVCXPRT *xprt)
116{
117	struct nfsrv_descript nd;
118	struct nfsrvcache *rp = NULL;
119	int cacherep, credflavor;
120
121	memset(&nd, 0, sizeof(nd));
122	if (rqst->rq_vers == NFS_VER2) {
123		if (rqst->rq_proc > NFSV2PROC_STATFS ||
124		    newnfs_nfsv3_procid[rqst->rq_proc] == NFSPROC_NOOP) {
125			svcerr_noproc(rqst);
126			svc_freereq(rqst);
127			goto out;
128		}
129		nd.nd_procnum = newnfs_nfsv3_procid[rqst->rq_proc];
130		nd.nd_flag = ND_NFSV2;
131	} else if (rqst->rq_vers == NFS_VER3) {
132		if (rqst->rq_proc >= NFS_V3NPROCS) {
133			svcerr_noproc(rqst);
134			svc_freereq(rqst);
135			goto out;
136		}
137		nd.nd_procnum = rqst->rq_proc;
138		nd.nd_flag = ND_NFSV3;
139	} else {
140		if (rqst->rq_proc != NFSPROC_NULL &&
141		    rqst->rq_proc != NFSV4PROC_COMPOUND) {
142			svcerr_noproc(rqst);
143			svc_freereq(rqst);
144			goto out;
145		}
146		nd.nd_procnum = rqst->rq_proc;
147		nd.nd_flag = ND_NFSV4;
148	}
149
150	/*
151	 * Note: we want rq_addr, not svc_getrpccaller for nd_nam2 -
152	 * NFS_SRVMAXDATA uses a NULL value for nd_nam2 to detect TCP
153	 * mounts.
154	 */
155	nd.nd_mrep = rqst->rq_args;
156	rqst->rq_args = NULL;
157	newnfs_realign(&nd.nd_mrep, M_WAITOK);
158	nd.nd_md = nd.nd_mrep;
159	nd.nd_dpos = mtod(nd.nd_md, caddr_t);
160	nd.nd_nam = svc_getrpccaller(rqst);
161	nd.nd_nam2 = rqst->rq_addr;
162	nd.nd_mreq = NULL;
163	nd.nd_cred = NULL;
164
165	if (nfs_privport && (nd.nd_flag & ND_NFSV4) == 0) {
166		/* Check if source port is privileged */
167		u_short port;
168		struct sockaddr *nam = nd.nd_nam;
169		struct sockaddr_in *sin;
170
171		sin = (struct sockaddr_in *)nam;
172		/*
173		 * INET/INET6 - same code:
174		 *    sin_port and sin6_port are at same offset
175		 */
176		port = ntohs(sin->sin_port);
177		if (port >= IPPORT_RESERVED &&
178		    nd.nd_procnum != NFSPROC_NULL) {
179#ifdef INET6
180			char b6[INET6_ADDRSTRLEN];
181#if defined(KLD_MODULE)
182			/* Do not use ip6_sprintf: the nfs module should work without INET6. */
183#define	ip6_sprintf(buf, a)						\
184			(sprintf((buf), "%x:%x:%x:%x:%x:%x:%x:%x",	\
185			    (a)->s6_addr16[0], (a)->s6_addr16[1],	\
186			    (a)->s6_addr16[2], (a)->s6_addr16[3],	\
187			    (a)->s6_addr16[4], (a)->s6_addr16[5],	\
188			    (a)->s6_addr16[6], (a)->s6_addr16[7]),	\
189			    (buf))
190#endif
191#endif
192			printf("NFS request from unprivileged port (%s:%d)\n",
193#ifdef INET6
194			    sin->sin_family == AF_INET6 ?
195			    ip6_sprintf(b6, &satosin6(sin)->sin6_addr) :
196#if defined(KLD_MODULE)
197#undef ip6_sprintf
198#endif
199#endif
200			    inet_ntoa(sin->sin_addr), port);
201			svcerr_weakauth(rqst);
202			svc_freereq(rqst);
203			m_freem(nd.nd_mrep);
204			goto out;
205		}
206	}
207
208	if (nd.nd_procnum != NFSPROC_NULL) {
209		if (!svc_getcred(rqst, &nd.nd_cred, &credflavor)) {
210			svcerr_weakauth(rqst);
211			svc_freereq(rqst);
212			m_freem(nd.nd_mrep);
213			goto out;
214		}
215
216		/* Set the flag based on credflavor */
217		if (credflavor == RPCSEC_GSS_KRB5) {
218			nd.nd_flag |= ND_GSS;
219		} else if (credflavor == RPCSEC_GSS_KRB5I) {
220			nd.nd_flag |= (ND_GSS | ND_GSSINTEGRITY);
221		} else if (credflavor == RPCSEC_GSS_KRB5P) {
222			nd.nd_flag |= (ND_GSS | ND_GSSPRIVACY);
223		} else if (credflavor != AUTH_SYS) {
224			svcerr_weakauth(rqst);
225			svc_freereq(rqst);
226			m_freem(nd.nd_mrep);
227			goto out;
228		}
229
230#ifdef MAC
231		mac_cred_associate_nfsd(nd.nd_cred);
232#endif
233		/*
234		 * Get a refcnt (shared lock) on nfsd_suspend_lock.
235		 * NFSSVC_SUSPENDNFSD will take an exclusive lock on
236		 * nfsd_suspend_lock to suspend these threads.
237		 * The call to nfsv4_lock() that preceeds nfsv4_getref()
238		 * ensures that the acquisition of the exclusive lock
239		 * takes priority over acquisition of the shared lock by
240		 * waiting for any exclusive lock request to complete.
241		 * This must be done here, before the check of
242		 * nfsv4root exports by nfsvno_v4rootexport().
243		 */
244		NFSLOCKV4ROOTMUTEX();
245		nfsv4_lock(&nfsd_suspend_lock, 0, NULL, NFSV4ROOTLOCKMUTEXPTR,
246		    NULL);
247		nfsv4_getref(&nfsd_suspend_lock, NULL, NFSV4ROOTLOCKMUTEXPTR,
248		    NULL);
249		NFSUNLOCKV4ROOTMUTEX();
250
251		if ((nd.nd_flag & ND_NFSV4) != 0) {
252			nd.nd_repstat = nfsvno_v4rootexport(&nd);
253			if (nd.nd_repstat != 0) {
254				NFSLOCKV4ROOTMUTEX();
255				nfsv4_relref(&nfsd_suspend_lock);
256				NFSUNLOCKV4ROOTMUTEX();
257				svcerr_weakauth(rqst);
258				svc_freereq(rqst);
259				m_freem(nd.nd_mrep);
260				goto out;
261			}
262		}
263
264		cacherep = nfs_proc(&nd, rqst->rq_xid, xprt, &rp);
265		NFSLOCKV4ROOTMUTEX();
266		nfsv4_relref(&nfsd_suspend_lock);
267		NFSUNLOCKV4ROOTMUTEX();
268	} else {
269		NFSMGET(nd.nd_mreq);
270		nd.nd_mreq->m_len = 0;
271		cacherep = RC_REPLY;
272	}
273	if (nd.nd_mrep != NULL)
274		m_freem(nd.nd_mrep);
275
276	if (nd.nd_cred != NULL)
277		crfree(nd.nd_cred);
278
279	if (cacherep == RC_DROPIT) {
280		if (nd.nd_mreq != NULL)
281			m_freem(nd.nd_mreq);
282		svc_freereq(rqst);
283		goto out;
284	}
285
286	if (nd.nd_mreq == NULL) {
287		svcerr_decode(rqst);
288		svc_freereq(rqst);
289		goto out;
290	}
291
292	if (nd.nd_repstat & NFSERR_AUTHERR) {
293		svcerr_auth(rqst, nd.nd_repstat & ~NFSERR_AUTHERR);
294		if (nd.nd_mreq != NULL)
295			m_freem(nd.nd_mreq);
296	} else if (!svc_sendreply_mbuf(rqst, nd.nd_mreq)) {
297		svcerr_systemerr(rqst);
298	}
299	if (rp != NULL) {
300		nfsrvd_sentcache(rp, (rqst->rq_reply_seq != 0 ||
301		    SVC_ACK(xprt, NULL)), rqst->rq_reply_seq);
302	}
303	svc_freereq(rqst);
304
305out:
306	if (softdep_ast_cleanup != NULL)
307		softdep_ast_cleanup();
308	NFSEXITCODE(0);
309}
310
311/*
312 * Check the cache and, optionally, do the RPC.
313 * Return the appropriate cache response.
314 */
315static int
316nfs_proc(struct nfsrv_descript *nd, u_int32_t xid, SVCXPRT *xprt,
317    struct nfsrvcache **rpp)
318{
319	struct thread *td = curthread;
320	int cacherep = RC_DOIT, isdgram, taglen = -1;
321	struct mbuf *m;
322	u_char tag[NFSV4_SMALLSTR + 1], *tagstr = NULL;
323	u_int32_t minorvers = 0;
324	uint32_t ack;
325
326	*rpp = NULL;
327	if (nd->nd_nam2 == NULL) {
328		nd->nd_flag |= ND_STREAMSOCK;
329		isdgram = 0;
330	} else {
331		isdgram = 1;
332	}
333
334	/*
335	 * Two cases:
336	 * 1 - For NFSv2 over UDP, if we are near our malloc/mget
337	 *     limit, just drop the request. There is no
338	 *     NFSERR_RESOURCE or NFSERR_DELAY for NFSv2 and the
339	 *     client will timeout/retry over UDP in a little while.
340	 * 2 - nd_repstat == 0 && nd_mreq == NULL, which
341	 *     means a normal nfs rpc, so check the cache
342	 */
343	if ((nd->nd_flag & ND_NFSV2) && nd->nd_nam2 != NULL &&
344	    nfsrv_mallocmget_limit()) {
345		cacherep = RC_DROPIT;
346	} else {
347		/*
348		 * For NFSv3, play it safe and assume that the client is
349		 * doing retries on the same TCP connection.
350		 */
351		if ((nd->nd_flag & (ND_NFSV4 | ND_STREAMSOCK)) ==
352		    ND_STREAMSOCK)
353			nd->nd_flag |= ND_SAMETCPCONN;
354		nd->nd_retxid = xid;
355		nd->nd_tcpconntime = NFSD_MONOSEC;
356		nd->nd_sockref = xprt->xp_sockref;
357		if ((nd->nd_flag & ND_NFSV4) != 0)
358			nfsd_getminorvers(nd, tag, &tagstr, &taglen,
359			    &minorvers);
360		if ((nd->nd_flag & ND_NFSV41) != 0)
361			/* NFSv4.1 caches replies in the session slots. */
362			cacherep = RC_DOIT;
363		else {
364			cacherep = nfsrvd_getcache(nd);
365			ack = 0;
366			SVC_ACK(xprt, &ack);
367			nfsrc_trimcache(xprt->xp_sockref, ack, 0);
368		}
369	}
370
371	/*
372	 * Handle the request. There are three cases.
373	 * RC_DOIT - do the RPC
374	 * RC_REPLY - return the reply already created
375	 * RC_DROPIT - just throw the request away
376	 */
377	if (cacherep == RC_DOIT) {
378		if ((nd->nd_flag & ND_NFSV41) != 0)
379			nd->nd_xprt = xprt;
380		nfsrvd_dorpc(nd, isdgram, tagstr, taglen, minorvers, td);
381		if ((nd->nd_flag & ND_NFSV41) != 0) {
382			if (nd->nd_repstat != NFSERR_REPLYFROMCACHE &&
383			    (nd->nd_flag & ND_SAVEREPLY) != 0) {
384				/* Cache a copy of the reply. */
385				m = m_copym(nd->nd_mreq, 0, M_COPYALL,
386				    M_WAITOK);
387			} else
388				m = NULL;
389			if ((nd->nd_flag & ND_HASSEQUENCE) != 0)
390				nfsrv_cache_session(nd->nd_sessionid,
391				    nd->nd_slotid, nd->nd_repstat, &m);
392			if (nd->nd_repstat == NFSERR_REPLYFROMCACHE)
393				nd->nd_repstat = 0;
394			cacherep = RC_REPLY;
395		} else {
396			if (nd->nd_repstat == NFSERR_DONTREPLY)
397				cacherep = RC_DROPIT;
398			else
399				cacherep = RC_REPLY;
400			*rpp = nfsrvd_updatecache(nd);
401		}
402	}
403	if (tagstr != NULL && taglen > NFSV4_SMALLSTR)
404		free(tagstr, M_TEMP);
405
406	NFSEXITCODE2(0, nd);
407	return (cacherep);
408}
409
410static void
411nfssvc_loss(SVCXPRT *xprt)
412{
413	uint32_t ack;
414
415	ack = 0;
416	SVC_ACK(xprt, &ack);
417	nfsrc_trimcache(xprt->xp_sockref, ack, 1);
418}
419
420/*
421 * Adds a socket to the list for servicing by nfsds.
422 */
423int
424nfsrvd_addsock(struct file *fp)
425{
426	int siz;
427	struct socket *so;
428	int error = 0;
429	SVCXPRT *xprt;
430	static u_int64_t sockref = 0;
431
432	so = fp->f_data;
433
434	siz = sb_max_adj;
435	error = soreserve(so, siz, siz);
436	if (error)
437		goto out;
438
439	/*
440	 * Steal the socket from userland so that it doesn't close
441	 * unexpectedly.
442	 */
443	if (so->so_type == SOCK_DGRAM)
444		xprt = svc_dg_create(nfsrvd_pool, so, 0, 0);
445	else
446		xprt = svc_vc_create(nfsrvd_pool, so, 0, 0);
447	if (xprt) {
448		fp->f_ops = &badfileops;
449		fp->f_data = NULL;
450		xprt->xp_sockref = ++sockref;
451		if (nfs_minvers == NFS_VER2)
452			svc_reg(xprt, NFS_PROG, NFS_VER2, nfssvc_program,
453			    NULL);
454		if (nfs_minvers <= NFS_VER3 && nfs_maxvers >= NFS_VER3)
455			svc_reg(xprt, NFS_PROG, NFS_VER3, nfssvc_program,
456			    NULL);
457		if (nfs_maxvers >= NFS_VER4)
458			svc_reg(xprt, NFS_PROG, NFS_VER4, nfssvc_program,
459			    NULL);
460		if (so->so_type == SOCK_STREAM)
461			svc_loss_reg(xprt, nfssvc_loss);
462		SVC_RELEASE(xprt);
463	}
464
465out:
466	NFSEXITCODE(error);
467	return (error);
468}
469
470/*
471 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
472 * until it is killed by a signal.
473 */
474int
475nfsrvd_nfsd(struct thread *td, struct nfsd_nfsd_args *args)
476{
477	char principal[MAXHOSTNAMELEN + 5];
478	struct proc *p;
479	int error = 0;
480	bool_t ret2, ret3, ret4;
481
482	error = copyinstr(args->principal, principal, sizeof (principal),
483	    NULL);
484	if (error)
485		goto out;
486
487	/*
488	 * Only the first nfsd actually does any work. The RPC code
489	 * adds threads to it as needed. Any extra processes offered
490	 * by nfsd just exit. If nfsd is new enough, it will call us
491	 * once with a structure that specifies how many threads to
492	 * use.
493	 */
494	NFSD_LOCK();
495	if (newnfs_numnfsd == 0) {
496		p = td->td_proc;
497		PROC_LOCK(p);
498		p->p_flag2 |= P2_AST_SU;
499		PROC_UNLOCK(p);
500		newnfs_numnfsd++;
501
502		NFSD_UNLOCK();
503
504		/* An empty string implies AUTH_SYS only. */
505		if (principal[0] != '\0') {
506			ret2 = rpc_gss_set_svc_name_call(principal,
507			    "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER2);
508			ret3 = rpc_gss_set_svc_name_call(principal,
509			    "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER3);
510			ret4 = rpc_gss_set_svc_name_call(principal,
511			    "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER4);
512
513			if (!ret2 || !ret3 || !ret4)
514				printf("nfsd: can't register svc name\n");
515		}
516
517		nfsrvd_pool->sp_minthreads = args->minthreads;
518		nfsrvd_pool->sp_maxthreads = args->maxthreads;
519
520		svc_run(nfsrvd_pool);
521
522		if (principal[0] != '\0') {
523			rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER2);
524			rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER3);
525			rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER4);
526		}
527
528		NFSD_LOCK();
529		newnfs_numnfsd--;
530		nfsrvd_init(1);
531		PROC_LOCK(p);
532		p->p_flag2 &= ~P2_AST_SU;
533		PROC_UNLOCK(p);
534	}
535	NFSD_UNLOCK();
536
537out:
538	NFSEXITCODE(error);
539	return (error);
540}
541
542/*
543 * Initialize the data structures for the server.
544 * Handshake with any new nfsds starting up to avoid any chance of
545 * corruption.
546 */
547void
548nfsrvd_init(int terminating)
549{
550
551	NFSD_LOCK_ASSERT();
552
553	if (terminating) {
554		nfsd_master_proc = NULL;
555		NFSD_UNLOCK();
556		nfsrv_freeallbackchannel_xprts();
557		svcpool_close(nfsrvd_pool);
558		NFSD_LOCK();
559	} else {
560		NFSD_UNLOCK();
561		nfsrvd_pool = svcpool_create("nfsd",
562		    SYSCTL_STATIC_CHILDREN(_vfs_nfsd));
563		nfsrvd_pool->sp_rcache = NULL;
564		nfsrvd_pool->sp_assign = fhanew_assign;
565		nfsrvd_pool->sp_done = fha_nd_complete;
566		NFSD_LOCK();
567	}
568}
569
570