xref: /titanic_50/usr/src/uts/common/fs/nfs/nfs_server.c (revision 03831d35f7499c87d51205817c93e9a8d42c4bae)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 /*
27  *	Copyright (c) 1983,1984,1985,1986,1987,1988,1989  AT&T.
28  *	All rights reserved.
29  *	Use is subject to license terms.
30  */
31 
32 #pragma ident	"%Z%%M%	%I%	%E% SMI"
33 
34 #include <sys/param.h>
35 #include <sys/types.h>
36 #include <sys/systm.h>
37 #include <sys/cred.h>
38 #include <sys/proc.h>
39 #include <sys/user.h>
40 #include <sys/buf.h>
41 #include <sys/vfs.h>
42 #include <sys/vnode.h>
43 #include <sys/pathname.h>
44 #include <sys/uio.h>
45 #include <sys/file.h>
46 #include <sys/stat.h>
47 #include <sys/errno.h>
48 #include <sys/socket.h>
49 #include <sys/sysmacros.h>
50 #include <sys/siginfo.h>
51 #include <sys/tiuser.h>
52 #include <sys/statvfs.h>
53 #include <sys/stream.h>
54 #include <sys/strsubr.h>
55 #include <sys/stropts.h>
56 #include <sys/timod.h>
57 #include <sys/t_kuser.h>
58 #include <sys/kmem.h>
59 #include <sys/kstat.h>
60 #include <sys/dirent.h>
61 #include <sys/cmn_err.h>
62 #include <sys/debug.h>
63 #include <sys/unistd.h>
64 #include <sys/vtrace.h>
65 #include <sys/mode.h>
66 #include <sys/acl.h>
67 #include <sys/sdt.h>
68 
69 #include <rpc/types.h>
70 #include <rpc/auth.h>
71 #include <rpc/auth_unix.h>
72 #include <rpc/auth_des.h>
73 #include <rpc/svc.h>
74 #include <rpc/xdr.h>
75 
76 #include <nfs/nfs.h>
77 #include <nfs/export.h>
78 #include <nfs/nfssys.h>
79 #include <nfs/nfs_clnt.h>
80 #include <nfs/nfs_acl.h>
81 #include <nfs/nfs_log.h>
82 #include <nfs/lm.h>
83 #include <nfs/nfs_dispatch.h>
84 #include <nfs/nfs4_drc.h>
85 
86 #include <rpcsvc/nfsauth_prot.h>
87 
88 #include <sys/modctl.h>
89 #include <sys/cladm.h>
90 #include <sys/clconf.h>
91 
92 #define	MAXHOST 32
93 const char *kinet_ntop6(uchar_t *, char *, size_t);
94 
95 /*
96  * Module linkage information.
97  */
98 
99 static struct modlmisc modlmisc = {
100 	&mod_miscops, "NFS server module"
101 };
102 
103 static struct modlinkage modlinkage = {
104 	MODREV_1, (void *)&modlmisc, NULL
105 };
106 
107 char _depends_on[] = "misc/klmmod";
108 
109 int
110 _init(void)
111 {
112 	int status;
113 
114 	if ((status = nfs_srvinit()) != 0) {
115 		cmn_err(CE_WARN, "_init: nfs_srvinit failed");
116 		return (status);
117 	}
118 
119 	status = mod_install((struct modlinkage *)&modlinkage);
120 	if (status != 0) {
121 		/*
122 		 * Could not load module, cleanup previous
123 		 * initialization work.
124 		 */
125 		nfs_srvfini();
126 	}
127 
128 	nfs_srv_quiesce_func = nfs_srv_quiesce_all;
129 
130 	return (status);
131 }
132 
133 int
134 _fini()
135 {
136 	return (EBUSY);
137 }
138 
139 int
140 _info(struct modinfo *modinfop)
141 {
142 	return (mod_info(&modlinkage, modinfop));
143 }
144 
145 /*
146  * PUBLICFH_CHECK() checks if the dispatch routine supports
147  * RPC_PUBLICFH_OK, if the filesystem is exported public, and if the
148  * incoming request is using the public filehandle. The check duplicates
149  * the exportmatch() call done in checkexport(), and we should consider
150  * modifying those routines to avoid the duplication. For now, we optimize
151  * by calling exportmatch() only after checking that the dispatch routine
152  * supports RPC_PUBLICFH_OK, and if the filesystem is explicitly exported
153  * public (i.e., not the placeholder).
154  */
155 #define	PUBLICFH_CHECK(disp, exi, fsid, xfid) \
156 		((disp->dis_flags & RPC_PUBLICFH_OK) && \
157 		((exi->exi_export.ex_flags & EX_PUBLIC) || \
158 		(exi == exi_public && exportmatch(exi_root, \
159 		fsid, xfid))))
160 
161 static void	nfs_srv_shutdown_all(int);
162 static void	rfs4_server_start(int);
163 static void	nullfree(void);
164 static void	rfs_dispatch(struct svc_req *, SVCXPRT *);
165 static void	acl_dispatch(struct svc_req *, SVCXPRT *);
166 static void	common_dispatch(struct svc_req *, SVCXPRT *,
167 		rpcvers_t, rpcvers_t, char *,
168 		struct rpc_disptable *);
169 static	int	checkauth(struct exportinfo *, struct svc_req *, cred_t *, int,
170 			bool_t);
171 static char	*client_name(struct svc_req *req);
172 static char	*client_addr(struct svc_req *req, char *buf);
173 extern	int	sec_svc_getcred(struct svc_req *, cred_t *cr, char **, int *);
174 extern	bool_t	sec_svc_inrootlist(int, caddr_t, int, caddr_t *);
175 
176 #define	NFSLOG_COPY_NETBUF(exi, xprt, nb)	{		\
177 	(nb)->maxlen = (xprt)->xp_rtaddr.maxlen;		\
178 	(nb)->len = (xprt)->xp_rtaddr.len;			\
179 	(nb)->buf = kmem_alloc((nb)->len, KM_SLEEP);		\
180 	bcopy((xprt)->xp_rtaddr.buf, (nb)->buf, (nb)->len);	\
181 	}
182 
183 /*
184  * Public Filehandle common nfs routines
185  */
186 static int	MCLpath(char **);
187 static void	URLparse(char *);
188 
189 /*
190  * NFS callout table.
191  * This table is used by svc_getreq() to dispatch a request with
192  * a given prog/vers pair to an appropriate service provider
193  * dispatch routine.
194  *
195  * NOTE: ordering is relied upon below when resetting the version min/max
196  * for NFS_PROGRAM.  Careful, if this is ever changed.
197  */
198 static SVC_CALLOUT __nfs_sc_clts[] = {
199 	{ NFS_PROGRAM,	   NFS_VERSMIN,	    NFS_VERSMAX,	rfs_dispatch },
200 	{ NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,	acl_dispatch }
201 };
202 
203 static SVC_CALLOUT_TABLE nfs_sct_clts = {
204 	sizeof (__nfs_sc_clts) / sizeof (__nfs_sc_clts[0]), FALSE,
205 	__nfs_sc_clts
206 };
207 
208 static SVC_CALLOUT __nfs_sc_cots[] = {
209 	{ NFS_PROGRAM,	   NFS_VERSMIN,	    NFS_VERSMAX,	rfs_dispatch },
210 	{ NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,	acl_dispatch }
211 };
212 
213 static SVC_CALLOUT_TABLE nfs_sct_cots = {
214 	sizeof (__nfs_sc_cots) / sizeof (__nfs_sc_cots[0]), FALSE, __nfs_sc_cots
215 };
216 
217 static SVC_CALLOUT __nfs_sc_rdma[] = {
218 	{ NFS_PROGRAM,	   NFS_VERSMIN,	    NFS_VERSMAX,	rfs_dispatch },
219 	{ NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,	acl_dispatch }
220 };
221 
222 static SVC_CALLOUT_TABLE nfs_sct_rdma = {
223 	sizeof (__nfs_sc_rdma) / sizeof (__nfs_sc_rdma[0]), FALSE, __nfs_sc_rdma
224 };
225 rpcvers_t nfs_versmin = NFS_VERSMIN_DEFAULT;
226 rpcvers_t nfs_versmax = NFS_VERSMAX_DEFAULT;
227 
228 /*
229  * Used to track the state of the server so that initialization
230  * can be done properly.
231  */
232 typedef enum {
233 	NFS_SERVER_STOPPED,	/* server state destroyed */
234 	NFS_SERVER_STOPPING,	/* server state being destroyed */
235 	NFS_SERVER_RUNNING,
236 	NFS_SERVER_QUIESCED,	/* server state preserved */
237 	NFS_SERVER_OFFLINE	/* server pool offline */
238 } nfs_server_running_t;
239 
240 static nfs_server_running_t nfs_server_upordown;
241 static kmutex_t nfs_server_upordown_lock;
242 static	kcondvar_t nfs_server_upordown_cv;
243 
244 int rfs4_dispatch(struct rpcdisp *, struct svc_req *, SVCXPRT *, char *);
245 
246 /*
247  * RDMA wait variables.
248  */
249 static kcondvar_t rdma_wait_cv;
250 static kmutex_t rdma_wait_mutex;
251 
252 /*
253  * Will be called at the point the server pool is being unregistered
254  * from the pool list. From that point onwards, the pool is waiting
255  * to be drained and as such the server state is stale and pertains
256  * to the old instantiation of the NFS server pool.
257  */
258 void
259 nfs_srv_offline(void)
260 {
261 	mutex_enter(&nfs_server_upordown_lock);
262 	if (nfs_server_upordown == NFS_SERVER_RUNNING) {
263 		nfs_server_upordown = NFS_SERVER_OFFLINE;
264 	}
265 	mutex_exit(&nfs_server_upordown_lock);
266 }
267 
268 /*
269  * Will be called at the point the server pool is being destroyed so
270  * all transports have been closed and no service threads are in
271  * existence.
272  *
273  * If we quiesce the server, we're shutting it down without destroying the
274  * server state. This allows it to warm start subsequently.
275  */
276 void
277 nfs_srv_stop_all(void)
278 {
279 	int quiesce = 0;
280 	nfs_srv_shutdown_all(quiesce);
281 }
282 
283 /*
284  * This alternative shutdown routine can be requested via nfssys()
285  */
286 void
287 nfs_srv_quiesce_all(void)
288 {
289 	int quiesce = 1;
290 	nfs_srv_shutdown_all(quiesce);
291 }
292 
293 static void
294 nfs_srv_shutdown_all(int quiesce) {
295 	mutex_enter(&nfs_server_upordown_lock);
296 	if (quiesce) {
297 		if (nfs_server_upordown == NFS_SERVER_RUNNING ||
298 			nfs_server_upordown == NFS_SERVER_OFFLINE) {
299 			nfs_server_upordown = NFS_SERVER_QUIESCED;
300 			cv_signal(&nfs_server_upordown_cv);
301 			cmn_err(CE_NOTE, "nfs_server: server is now quiesced; "
302 			    "NFSv4 state has been preserved");
303 		}
304 	} else {
305 		if (nfs_server_upordown == NFS_SERVER_OFFLINE) {
306 			nfs_server_upordown = NFS_SERVER_STOPPING;
307 			mutex_exit(&nfs_server_upordown_lock);
308 			rfs4_state_fini();
309 			rfs4_fini_drc(nfs4_drc);
310 			mutex_enter(&nfs_server_upordown_lock);
311 			nfs_server_upordown = NFS_SERVER_STOPPED;
312 			cv_signal(&nfs_server_upordown_cv);
313 		}
314 	}
315 	mutex_exit(&nfs_server_upordown_lock);
316 }
317 
318 static int
319 nfs_srv_set_sc_versions(struct file *fp, SVC_CALLOUT_TABLE **sctpp,
320 			rpcvers_t versmin, rpcvers_t versmax)
321 {
322 	struct strioctl strioc;
323 	struct T_info_ack tinfo;
324 	int		error, retval;
325 
326 	/*
327 	 * Find out what type of transport this is.
328 	 */
329 	strioc.ic_cmd = TI_GETINFO;
330 	strioc.ic_timout = -1;
331 	strioc.ic_len = sizeof (tinfo);
332 	strioc.ic_dp = (char *)&tinfo;
333 	tinfo.PRIM_type = T_INFO_REQ;
334 
335 	error = strioctl(fp->f_vnode, I_STR, (intptr_t)&strioc, 0, K_TO_K,
336 	    CRED(), &retval);
337 	if (error || retval)
338 		return (error);
339 
340 	/*
341 	 * Based on our query of the transport type...
342 	 *
343 	 * Reset the min/max versions based on the caller's request
344 	 * NOTE: This assumes that NFS_PROGRAM is first in the array!!
345 	 * And the second entry is the NFS_ACL_PROGRAM.
346 	 */
347 	switch (tinfo.SERV_type) {
348 	case T_CLTS:
349 		if (versmax == NFS_V4)
350 			return (EINVAL);
351 		__nfs_sc_clts[0].sc_versmin = versmin;
352 		__nfs_sc_clts[0].sc_versmax = versmax;
353 		__nfs_sc_clts[1].sc_versmin = versmin;
354 		__nfs_sc_clts[1].sc_versmax = versmax;
355 		*sctpp = &nfs_sct_clts;
356 		break;
357 	case T_COTS:
358 	case T_COTS_ORD:
359 		__nfs_sc_cots[0].sc_versmin = versmin;
360 		__nfs_sc_cots[0].sc_versmax = versmax;
361 		/* For the NFS_ACL program, check the max version */
362 		if (versmax > NFS_ACL_VERSMAX)
363 			versmax = NFS_ACL_VERSMAX;
364 		__nfs_sc_cots[1].sc_versmin = versmin;
365 		__nfs_sc_cots[1].sc_versmax = versmax;
366 		*sctpp = &nfs_sct_cots;
367 		break;
368 	default:
369 		error = EINVAL;
370 	}
371 
372 	return (error);
373 }
374 
375 /*
376  * NFS Server system call.
377  * Does all of the work of running a NFS server.
378  * uap->fd is the fd of an open transport provider
379  */
380 int
381 nfs_svc(struct nfs_svc_args *arg, model_t model)
382 {
383 	file_t *fp;
384 	SVCMASTERXPRT *xprt;
385 	int error;
386 	int readsize;
387 	char buf[KNC_STRSIZE];
388 	size_t len;
389 	STRUCT_HANDLE(nfs_svc_args, uap);
390 	struct netbuf addrmask;
391 	SVC_CALLOUT_TABLE *sctp = NULL;
392 
393 #ifdef lint
394 	model = model;		/* STRUCT macros don't always refer to it */
395 #endif
396 
397 	STRUCT_SET_HANDLE(uap, model, arg);
398 
399 	/* Check privileges in nfssys() */
400 
401 	if ((fp = getf(STRUCT_FGET(uap, fd))) == NULL)
402 		return (EBADF);
403 
404 	/*
405 	 * Set read buffer size to rsize
406 	 * and add room for RPC headers.
407 	 */
408 	readsize = nfs3tsize() + (RPC_MAXDATASIZE - NFS_MAXDATA);
409 	if (readsize < RPC_MAXDATASIZE)
410 		readsize = RPC_MAXDATASIZE;
411 
412 	error = copyinstr((const char *)STRUCT_FGETP(uap, netid), buf,
413 	    KNC_STRSIZE, &len);
414 	if (error) {
415 		releasef(STRUCT_FGET(uap, fd));
416 		return (error);
417 	}
418 
419 	addrmask.len = STRUCT_FGET(uap, addrmask.len);
420 	addrmask.maxlen = STRUCT_FGET(uap, addrmask.maxlen);
421 	addrmask.buf = kmem_alloc(addrmask.maxlen, KM_SLEEP);
422 	error = copyin(STRUCT_FGETP(uap, addrmask.buf), addrmask.buf,
423 	    addrmask.len);
424 	if (error) {
425 		releasef(STRUCT_FGET(uap, fd));
426 		kmem_free(addrmask.buf, addrmask.maxlen);
427 		return (error);
428 	}
429 
430 	nfs_versmin = STRUCT_FGET(uap, versmin);
431 	nfs_versmax = STRUCT_FGET(uap, versmax);
432 
433 	/* Double check the vers min/max ranges */
434 	if ((nfs_versmin > nfs_versmax) ||
435 		(nfs_versmin < NFS_VERSMIN) ||
436 		(nfs_versmax > NFS_VERSMAX)) {
437 		nfs_versmin = NFS_VERSMIN_DEFAULT;
438 		nfs_versmax = NFS_VERSMAX_DEFAULT;
439 	}
440 
441 	if (error =
442 	    nfs_srv_set_sc_versions(fp, &sctp, nfs_versmin, nfs_versmax)) {
443 		releasef(STRUCT_FGET(uap, fd));
444 		kmem_free(addrmask.buf, addrmask.maxlen);
445 		return (error);
446 	}
447 
448 	/* Initialize nfsv4 server */
449 	if (nfs_versmax == (rpcvers_t)NFS_V4)
450 		rfs4_server_start(STRUCT_FGET(uap, delegation));
451 
452 	/* Create a transport handle. */
453 	error = svc_tli_kcreate(fp, readsize, buf, &addrmask, &xprt,
454 				sctp, NULL, NFS_SVCPOOL_ID, TRUE);
455 
456 	if (error)
457 		kmem_free(addrmask.buf, addrmask.maxlen);
458 
459 	releasef(STRUCT_FGET(uap, fd));
460 
461 	/* save the cluster nodeid */
462 	if (cluster_bootflags & CLUSTER_BOOTED)
463 		lm_global_nlmid = clconf_get_nodeid();
464 
465 	return (error);
466 }
467 
468 static void
469 rfs4_server_start(int nfs4_srv_delegation)
470 {
471 	/*
472 	 * Determine if the server has previously been "started" and
473 	 * if not, do the per instance initialization
474 	 */
475 	mutex_enter(&nfs_server_upordown_lock);
476 
477 	if (nfs_server_upordown != NFS_SERVER_RUNNING) {
478 		/* Do we need to stop and wait on the previous server? */
479 		while (nfs_server_upordown == NFS_SERVER_STOPPING ||
480 			nfs_server_upordown == NFS_SERVER_OFFLINE)
481 			cv_wait(&nfs_server_upordown_cv,
482 			    &nfs_server_upordown_lock);
483 
484 		if (nfs_server_upordown != NFS_SERVER_RUNNING) {
485 			(void) svc_pool_control(NFS_SVCPOOL_ID,
486 			    SVCPSET_UNREGISTER_PROC, (void *)&nfs_srv_offline);
487 			(void) svc_pool_control(NFS_SVCPOOL_ID,
488 			    SVCPSET_SHUTDOWN_PROC, (void *)&nfs_srv_stop_all);
489 
490 			/* is this an nfsd warm start? */
491 			if (nfs_server_upordown == NFS_SERVER_QUIESCED) {
492 				int start_grace;
493 
494 				cmn_err(CE_NOTE, "nfs_server: "
495 				    "server was previously quiesced; "
496 				    "existing NFSv4 state will be re-used");
497 
498 				/*
499 				 * Cluster: this is also the signal that
500 				 * a failover has occurred, so create a new
501 				 * server instance, and start its grace period.
502 				 * We also need to reset all currently
503 				 * active grace periods in case of multiple
504 				 * failovers within the grace duration,
505 				 * to avoid partitioning clients of the same
506 				 * resource into different instances.
507 				 */
508 				if (cluster_bootflags & CLUSTER_BOOTED) {
509 					rfs4_grace_reset_all();
510 					start_grace = 1;
511 					rfs4_servinst_create(start_grace);
512 				}
513 			} else {
514 				rfs4_state_init();
515 				nfs4_drc = rfs4_init_drc(nfs4_drc_max,
516 							nfs4_drc_hash,
517 							nfs4_drc_lifetime);
518 			}
519 
520 			/*
521 			 * Check to see if delegation is to be
522 			 * enabled at the server
523 			 */
524 			if (nfs4_srv_delegation != FALSE)
525 				rfs4_set_deleg_policy(SRV_NORMAL_DELEGATE);
526 
527 			nfs_server_upordown = NFS_SERVER_RUNNING;
528 		}
529 		cv_signal(&nfs_server_upordown_cv);
530 	}
531 	mutex_exit(&nfs_server_upordown_lock);
532 }
533 
534 /*
535  * If RDMA device available,
536  * start RDMA listener.
537  */
538 int
539 rdma_start(struct rdma_svc_args *rsa)
540 {
541 	int error;
542 	rdma_xprt_group_t started_rdma_xprts;
543 
544 	/* Double check the vers min/max ranges */
545 	if ((rsa->nfs_versmin > rsa->nfs_versmax) ||
546 		(rsa->nfs_versmin < NFS_VERSMIN) ||
547 		(rsa->nfs_versmax > NFS_VERSMAX)) {
548 		rsa->nfs_versmin = NFS_VERSMIN_DEFAULT;
549 		rsa->nfs_versmax = NFS_VERSMAX_DEFAULT;
550 	}
551 	nfs_versmin = rsa->nfs_versmin;
552 	nfs_versmax = rsa->nfs_versmax;
553 
554 	/* Set the versions in the callout table */
555 	__nfs_sc_rdma[0].sc_versmin = rsa->nfs_versmin;
556 	__nfs_sc_rdma[0].sc_versmax = rsa->nfs_versmax;
557 	/* For the NFS_ACL program, check the max version */
558 	__nfs_sc_rdma[1].sc_versmin = rsa->nfs_versmin;
559 	if (rsa->nfs_versmax > NFS_ACL_VERSMAX)
560 		__nfs_sc_rdma[1].sc_versmax = NFS_ACL_VERSMAX;
561 	else
562 		__nfs_sc_rdma[1].sc_versmax = rsa->nfs_versmax;
563 
564 	/* Initialize nfsv4 server */
565 	if (rsa->nfs_versmax == (rpcvers_t)NFS_V4)
566 		rfs4_server_start(rsa->delegation);
567 
568 	started_rdma_xprts.rtg_count = 0;
569 	started_rdma_xprts.rtg_listhead = NULL;
570 	started_rdma_xprts.rtg_poolid = rsa->poolid;
571 	error = svc_rdma_kcreate(rsa->netid, &nfs_sct_rdma, rsa->poolid,
572 	    &started_rdma_xprts);
573 
574 	if (error == 0) {
575 		mutex_enter(&rdma_wait_mutex);
576 		if (!cv_wait_sig(&rdma_wait_cv, &rdma_wait_mutex)) {
577 			rdma_stop(started_rdma_xprts);
578 		}
579 		mutex_exit(&rdma_wait_mutex);
580 	}
581 
582 	return (error);
583 }
584 
585 /* ARGSUSED */
586 void
587 rpc_null(caddr_t *argp, caddr_t *resp)
588 {
589 }
590 
591 /* ARGSUSED */
592 static void
593 rfs_error(caddr_t *argp, caddr_t *resp)
594 {
595 	/* return (EOPNOTSUPP); */
596 }
597 
598 static void
599 nullfree(void)
600 {
601 }
602 
603 static char *rfscallnames_v2[] = {
604 	"RFS2_NULL",
605 	"RFS2_GETATTR",
606 	"RFS2_SETATTR",
607 	"RFS2_ROOT",
608 	"RFS2_LOOKUP",
609 	"RFS2_READLINK",
610 	"RFS2_READ",
611 	"RFS2_WRITECACHE",
612 	"RFS2_WRITE",
613 	"RFS2_CREATE",
614 	"RFS2_REMOVE",
615 	"RFS2_RENAME",
616 	"RFS2_LINK",
617 	"RFS2_SYMLINK",
618 	"RFS2_MKDIR",
619 	"RFS2_RMDIR",
620 	"RFS2_READDIR",
621 	"RFS2_STATFS"
622 };
623 
624 static struct rpcdisp rfsdisptab_v2[] = {
625 	/*
626 	 * NFS VERSION 2
627 	 */
628 
629 	/* RFS_NULL = 0 */
630 	{rpc_null,
631 	    xdr_void, NULL_xdrproc_t, 0,
632 	    xdr_void, NULL_xdrproc_t, 0,
633 	    nullfree, RPC_IDEMPOTENT,
634 	    0},
635 
636 	/* RFS_GETATTR = 1 */
637 	{rfs_getattr,
638 	    xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t),
639 	    xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat),
640 	    nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP,
641 	    rfs_getattr_getfh},
642 
643 	/* RFS_SETATTR = 2 */
644 	{rfs_setattr,
645 	    xdr_saargs, NULL_xdrproc_t, sizeof (struct nfssaargs),
646 	    xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat),
647 	    nullfree, RPC_MAPRESP,
648 	    rfs_setattr_getfh},
649 
650 	/* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */
651 	{rfs_error,
652 	    xdr_void, NULL_xdrproc_t, 0,
653 	    xdr_void, NULL_xdrproc_t, 0,
654 	    nullfree, RPC_IDEMPOTENT,
655 	    0},
656 
657 	/* RFS_LOOKUP = 4 */
658 	{rfs_lookup,
659 	    xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs),
660 	    xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres),
661 	    nullfree, RPC_IDEMPOTENT|RPC_MAPRESP|RPC_PUBLICFH_OK,
662 	    rfs_lookup_getfh},
663 
664 	/* RFS_READLINK = 5 */
665 	{rfs_readlink,
666 	    xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t),
667 	    xdr_rdlnres, NULL_xdrproc_t, sizeof (struct nfsrdlnres),
668 	    rfs_rlfree, RPC_IDEMPOTENT,
669 	    rfs_readlink_getfh},
670 
671 	/* RFS_READ = 6 */
672 	{rfs_read,
673 	    xdr_readargs, NULL_xdrproc_t, sizeof (struct nfsreadargs),
674 	    xdr_rdresult, NULL_xdrproc_t, sizeof (struct nfsrdresult),
675 	    rfs_rdfree, RPC_IDEMPOTENT,
676 	    rfs_read_getfh},
677 
678 	/* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */
679 	{rfs_error,
680 	    xdr_void, NULL_xdrproc_t, 0,
681 	    xdr_void, NULL_xdrproc_t, 0,
682 	    nullfree, RPC_IDEMPOTENT,
683 	    0},
684 
685 	/* RFS_WRITE = 8 */
686 	{rfs_write,
687 	    xdr_writeargs, NULL_xdrproc_t, sizeof (struct nfswriteargs),
688 	    xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat),
689 	    nullfree, RPC_MAPRESP,
690 	    rfs_write_getfh},
691 
692 	/* RFS_CREATE = 9 */
693 	{rfs_create,
694 	    xdr_creatargs, NULL_xdrproc_t, sizeof (struct nfscreatargs),
695 	    xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres),
696 	    nullfree, RPC_MAPRESP,
697 	    rfs_create_getfh},
698 
699 	/* RFS_REMOVE = 10 */
700 	{rfs_remove,
701 	    xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs),
702 #ifdef _LITTLE_ENDIAN
703 	    xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
704 #else
705 	    xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
706 #endif
707 	    nullfree, RPC_MAPRESP,
708 	    rfs_remove_getfh},
709 
710 	/* RFS_RENAME = 11 */
711 	{rfs_rename,
712 	    xdr_rnmargs, NULL_xdrproc_t, sizeof (struct nfsrnmargs),
713 #ifdef _LITTLE_ENDIAN
714 	    xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
715 #else
716 	    xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
717 #endif
718 	    nullfree, RPC_MAPRESP,
719 	    rfs_rename_getfh},
720 
721 	/* RFS_LINK = 12 */
722 	{rfs_link,
723 	    xdr_linkargs, NULL_xdrproc_t, sizeof (struct nfslinkargs),
724 #ifdef _LITTLE_ENDIAN
725 	    xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
726 #else
727 	    xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
728 #endif
729 	    nullfree, RPC_MAPRESP,
730 	    rfs_link_getfh},
731 
732 	/* RFS_SYMLINK = 13 */
733 	{rfs_symlink,
734 	    xdr_slargs, NULL_xdrproc_t, sizeof (struct nfsslargs),
735 #ifdef _LITTLE_ENDIAN
736 	    xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
737 #else
738 	    xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
739 #endif
740 	    nullfree, RPC_MAPRESP,
741 	    rfs_symlink_getfh},
742 
743 	/* RFS_MKDIR = 14 */
744 	{rfs_mkdir,
745 	    xdr_creatargs, NULL_xdrproc_t, sizeof (struct nfscreatargs),
746 	    xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres),
747 	    nullfree, RPC_MAPRESP,
748 	    rfs_mkdir_getfh},
749 
750 	/* RFS_RMDIR = 15 */
751 	{rfs_rmdir,
752 	    xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs),
753 #ifdef _LITTLE_ENDIAN
754 	    xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
755 #else
756 	    xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
757 #endif
758 	    nullfree, RPC_MAPRESP,
759 	    rfs_rmdir_getfh},
760 
761 	/* RFS_READDIR = 16 */
762 	{rfs_readdir,
763 	    xdr_rddirargs, NULL_xdrproc_t, sizeof (struct nfsrddirargs),
764 	    xdr_putrddirres, NULL_xdrproc_t, sizeof (struct nfsrddirres),
765 	    rfs_rddirfree, RPC_IDEMPOTENT,
766 	    rfs_readdir_getfh},
767 
768 	/* RFS_STATFS = 17 */
769 	{rfs_statfs,
770 	    xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t),
771 	    xdr_statfs, xdr_faststatfs, sizeof (struct nfsstatfs),
772 	    nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP,
773 	    rfs_statfs_getfh},
774 };
775 
776 static char *rfscallnames_v3[] = {
777 	"RFS3_NULL",
778 	"RFS3_GETATTR",
779 	"RFS3_SETATTR",
780 	"RFS3_LOOKUP",
781 	"RFS3_ACCESS",
782 	"RFS3_READLINK",
783 	"RFS3_READ",
784 	"RFS3_WRITE",
785 	"RFS3_CREATE",
786 	"RFS3_MKDIR",
787 	"RFS3_SYMLINK",
788 	"RFS3_MKNOD",
789 	"RFS3_REMOVE",
790 	"RFS3_RMDIR",
791 	"RFS3_RENAME",
792 	"RFS3_LINK",
793 	"RFS3_READDIR",
794 	"RFS3_READDIRPLUS",
795 	"RFS3_FSSTAT",
796 	"RFS3_FSINFO",
797 	"RFS3_PATHCONF",
798 	"RFS3_COMMIT"
799 };
800 
801 static struct rpcdisp rfsdisptab_v3[] = {
802 	/*
803 	 * NFS VERSION 3
804 	 */
805 
806 	/* RFS_NULL = 0 */
807 	{rpc_null,
808 	    xdr_void, NULL_xdrproc_t, 0,
809 	    xdr_void, NULL_xdrproc_t, 0,
810 	    nullfree, RPC_IDEMPOTENT,
811 	    0},
812 
813 	/* RFS3_GETATTR = 1 */
814 	{rfs3_getattr,
815 	    xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (GETATTR3args),
816 	    xdr_GETATTR3res, NULL_xdrproc_t, sizeof (GETATTR3res),
817 	    nullfree, (RPC_IDEMPOTENT | RPC_ALLOWANON),
818 	    rfs3_getattr_getfh},
819 
820 	/* RFS3_SETATTR = 2 */
821 	{rfs3_setattr,
822 	    xdr_SETATTR3args, NULL_xdrproc_t, sizeof (SETATTR3args),
823 	    xdr_SETATTR3res, NULL_xdrproc_t, sizeof (SETATTR3res),
824 	    nullfree, 0,
825 	    rfs3_setattr_getfh},
826 
827 	/* RFS3_LOOKUP = 3 */
828 	{rfs3_lookup,
829 	    xdr_diropargs3, NULL_xdrproc_t, sizeof (LOOKUP3args),
830 	    xdr_LOOKUP3res, NULL_xdrproc_t, sizeof (LOOKUP3res),
831 	    nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK),
832 	    rfs3_lookup_getfh},
833 
834 	/* RFS3_ACCESS = 4 */
835 	{rfs3_access,
836 	    xdr_ACCESS3args, NULL_xdrproc_t, sizeof (ACCESS3args),
837 	    xdr_ACCESS3res, NULL_xdrproc_t, sizeof (ACCESS3res),
838 	    nullfree, RPC_IDEMPOTENT,
839 	    rfs3_access_getfh},
840 
841 	/* RFS3_READLINK = 5 */
842 	{rfs3_readlink,
843 	    xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (READLINK3args),
844 	    xdr_READLINK3res, NULL_xdrproc_t, sizeof (READLINK3res),
845 	    rfs3_readlink_free, RPC_IDEMPOTENT,
846 	    rfs3_readlink_getfh},
847 
848 	/* RFS3_READ = 6 */
849 	{rfs3_read,
850 	    xdr_READ3args, NULL_xdrproc_t, sizeof (READ3args),
851 	    xdr_READ3res, NULL_xdrproc_t, sizeof (READ3res),
852 	    rfs3_read_free, RPC_IDEMPOTENT,
853 	    rfs3_read_getfh},
854 
855 	/* RFS3_WRITE = 7 */
856 	{rfs3_write,
857 	    xdr_WRITE3args, NULL_xdrproc_t, sizeof (WRITE3args),
858 	    xdr_WRITE3res, NULL_xdrproc_t, sizeof (WRITE3res),
859 	    nullfree, 0,
860 	    rfs3_write_getfh},
861 
862 	/* RFS3_CREATE = 8 */
863 	{rfs3_create,
864 	    xdr_CREATE3args, NULL_xdrproc_t, sizeof (CREATE3args),
865 	    xdr_CREATE3res, NULL_xdrproc_t, sizeof (CREATE3res),
866 	    nullfree, 0,
867 	    rfs3_create_getfh},
868 
869 	/* RFS3_MKDIR = 9 */
870 	{rfs3_mkdir,
871 	    xdr_MKDIR3args, NULL_xdrproc_t, sizeof (MKDIR3args),
872 	    xdr_MKDIR3res, NULL_xdrproc_t, sizeof (MKDIR3res),
873 	    nullfree, 0,
874 	    rfs3_mkdir_getfh},
875 
876 	/* RFS3_SYMLINK = 10 */
877 	{rfs3_symlink,
878 	    xdr_SYMLINK3args, NULL_xdrproc_t, sizeof (SYMLINK3args),
879 	    xdr_SYMLINK3res, NULL_xdrproc_t, sizeof (SYMLINK3res),
880 	    nullfree, 0,
881 	    rfs3_symlink_getfh},
882 
883 	/* RFS3_MKNOD = 11 */
884 	{rfs3_mknod,
885 	    xdr_MKNOD3args, NULL_xdrproc_t, sizeof (MKNOD3args),
886 	    xdr_MKNOD3res, NULL_xdrproc_t, sizeof (MKNOD3res),
887 	    nullfree, 0,
888 	    rfs3_mknod_getfh},
889 
890 	/* RFS3_REMOVE = 12 */
891 	{rfs3_remove,
892 	    xdr_diropargs3, NULL_xdrproc_t, sizeof (REMOVE3args),
893 	    xdr_REMOVE3res, NULL_xdrproc_t, sizeof (REMOVE3res),
894 	    nullfree, 0,
895 	    rfs3_remove_getfh},
896 
897 	/* RFS3_RMDIR = 13 */
898 	{rfs3_rmdir,
899 	    xdr_diropargs3, NULL_xdrproc_t, sizeof (RMDIR3args),
900 	    xdr_RMDIR3res, NULL_xdrproc_t, sizeof (RMDIR3res),
901 	    nullfree, 0,
902 	    rfs3_rmdir_getfh},
903 
904 	/* RFS3_RENAME = 14 */
905 	{rfs3_rename,
906 	    xdr_RENAME3args, NULL_xdrproc_t, sizeof (RENAME3args),
907 	    xdr_RENAME3res, NULL_xdrproc_t, sizeof (RENAME3res),
908 	    nullfree, 0,
909 	    rfs3_rename_getfh},
910 
911 	/* RFS3_LINK = 15 */
912 	{rfs3_link,
913 	    xdr_LINK3args, NULL_xdrproc_t, sizeof (LINK3args),
914 	    xdr_LINK3res, NULL_xdrproc_t, sizeof (LINK3res),
915 	    nullfree, 0,
916 	    rfs3_link_getfh},
917 
918 	/* RFS3_READDIR = 16 */
919 	{rfs3_readdir,
920 	    xdr_READDIR3args, NULL_xdrproc_t, sizeof (READDIR3args),
921 	    xdr_READDIR3res, NULL_xdrproc_t, sizeof (READDIR3res),
922 	    rfs3_readdir_free, RPC_IDEMPOTENT,
923 	    rfs3_readdir_getfh},
924 
925 	/* RFS3_READDIRPLUS = 17 */
926 	{rfs3_readdirplus,
927 	    xdr_READDIRPLUS3args, NULL_xdrproc_t, sizeof (READDIRPLUS3args),
928 	    xdr_READDIRPLUS3res, NULL_xdrproc_t, sizeof (READDIRPLUS3res),
929 	    rfs3_readdirplus_free, RPC_AVOIDWORK,
930 	    rfs3_readdirplus_getfh},
931 
932 	/* RFS3_FSSTAT = 18 */
933 	{rfs3_fsstat,
934 	    xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (FSSTAT3args),
935 	    xdr_FSSTAT3res, NULL_xdrproc_t, sizeof (FSSTAT3res),
936 	    nullfree, RPC_IDEMPOTENT,
937 	    rfs3_fsstat_getfh},
938 
939 	/* RFS3_FSINFO = 19 */
940 	{rfs3_fsinfo,
941 	    xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (FSINFO3args),
942 	    xdr_FSINFO3res, NULL_xdrproc_t, sizeof (FSINFO3res),
943 	    nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON,
944 	    rfs3_fsinfo_getfh},
945 
946 	/* RFS3_PATHCONF = 20 */
947 	{rfs3_pathconf,
948 	    xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (PATHCONF3args),
949 	    xdr_PATHCONF3res, NULL_xdrproc_t, sizeof (PATHCONF3res),
950 	    nullfree, RPC_IDEMPOTENT,
951 	    rfs3_pathconf_getfh},
952 
953 	/* RFS3_COMMIT = 21 */
954 	{rfs3_commit,
955 	    xdr_COMMIT3args, NULL_xdrproc_t, sizeof (COMMIT3args),
956 	    xdr_COMMIT3res, NULL_xdrproc_t, sizeof (COMMIT3res),
957 	    nullfree, RPC_IDEMPOTENT,
958 	    rfs3_commit_getfh},
959 };
960 
961 static char *rfscallnames_v4[] = {
962 	"RFS4_NULL",
963 	"RFS4_COMPOUND",
964 	"RFS4_NULL",
965 	"RFS4_NULL",
966 	"RFS4_NULL",
967 	"RFS4_NULL",
968 	"RFS4_NULL",
969 	"RFS4_NULL",
970 	"RFS4_CREATE"
971 };
972 
973 static struct rpcdisp rfsdisptab_v4[] = {
974 	/*
975 	 * NFS VERSION 4
976 	 */
977 
978 	/* RFS_NULL = 0 */
979 	{rpc_null,
980 	    xdr_void, NULL_xdrproc_t, 0,
981 	    xdr_void, NULL_xdrproc_t, 0,
982 	    nullfree, RPC_IDEMPOTENT, 0},
983 
984 	/* RFS4_compound = 1 */
985 	{rfs4_compound,
986 	    xdr_COMPOUND4args_srv, NULL_xdrproc_t, sizeof (COMPOUND4args),
987 	    xdr_COMPOUND4res_srv, NULL_xdrproc_t, sizeof (COMPOUND4res),
988 	    rfs4_compound_free, 0, 0},
989 };
990 
991 union rfs_args {
992 	/*
993 	 * NFS VERSION 2
994 	 */
995 
996 	/* RFS_NULL = 0 */
997 
998 	/* RFS_GETATTR = 1 */
999 	fhandle_t nfs2_getattr_args;
1000 
1001 	/* RFS_SETATTR = 2 */
1002 	struct nfssaargs nfs2_setattr_args;
1003 
1004 	/* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */
1005 
1006 	/* RFS_LOOKUP = 4 */
1007 	struct nfsdiropargs nfs2_lookup_args;
1008 
1009 	/* RFS_READLINK = 5 */
1010 	fhandle_t nfs2_readlink_args;
1011 
1012 	/* RFS_READ = 6 */
1013 	struct nfsreadargs nfs2_read_args;
1014 
1015 	/* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */
1016 
1017 	/* RFS_WRITE = 8 */
1018 	struct nfswriteargs nfs2_write_args;
1019 
1020 	/* RFS_CREATE = 9 */
1021 	struct nfscreatargs nfs2_create_args;
1022 
1023 	/* RFS_REMOVE = 10 */
1024 	struct nfsdiropargs nfs2_remove_args;
1025 
1026 	/* RFS_RENAME = 11 */
1027 	struct nfsrnmargs nfs2_rename_args;
1028 
1029 	/* RFS_LINK = 12 */
1030 	struct nfslinkargs nfs2_link_args;
1031 
1032 	/* RFS_SYMLINK = 13 */
1033 	struct nfsslargs nfs2_symlink_args;
1034 
1035 	/* RFS_MKDIR = 14 */
1036 	struct nfscreatargs nfs2_mkdir_args;
1037 
1038 	/* RFS_RMDIR = 15 */
1039 	struct nfsdiropargs nfs2_rmdir_args;
1040 
1041 	/* RFS_READDIR = 16 */
1042 	struct nfsrddirargs nfs2_readdir_args;
1043 
1044 	/* RFS_STATFS = 17 */
1045 	fhandle_t nfs2_statfs_args;
1046 
1047 	/*
1048 	 * NFS VERSION 3
1049 	 */
1050 
1051 	/* RFS_NULL = 0 */
1052 
1053 	/* RFS3_GETATTR = 1 */
1054 	GETATTR3args nfs3_getattr_args;
1055 
1056 	/* RFS3_SETATTR = 2 */
1057 	SETATTR3args nfs3_setattr_args;
1058 
1059 	/* RFS3_LOOKUP = 3 */
1060 	LOOKUP3args nfs3_lookup_args;
1061 
1062 	/* RFS3_ACCESS = 4 */
1063 	ACCESS3args nfs3_access_args;
1064 
1065 	/* RFS3_READLINK = 5 */
1066 	READLINK3args nfs3_readlink_args;
1067 
1068 	/* RFS3_READ = 6 */
1069 	READ3args nfs3_read_args;
1070 
1071 	/* RFS3_WRITE = 7 */
1072 	WRITE3args nfs3_write_args;
1073 
1074 	/* RFS3_CREATE = 8 */
1075 	CREATE3args nfs3_create_args;
1076 
1077 	/* RFS3_MKDIR = 9 */
1078 	MKDIR3args nfs3_mkdir_args;
1079 
1080 	/* RFS3_SYMLINK = 10 */
1081 	SYMLINK3args nfs3_symlink_args;
1082 
1083 	/* RFS3_MKNOD = 11 */
1084 	MKNOD3args nfs3_mknod_args;
1085 
1086 	/* RFS3_REMOVE = 12 */
1087 	REMOVE3args nfs3_remove_args;
1088 
1089 	/* RFS3_RMDIR = 13 */
1090 	RMDIR3args nfs3_rmdir_args;
1091 
1092 	/* RFS3_RENAME = 14 */
1093 	RENAME3args nfs3_rename_args;
1094 
1095 	/* RFS3_LINK = 15 */
1096 	LINK3args nfs3_link_args;
1097 
1098 	/* RFS3_READDIR = 16 */
1099 	READDIR3args nfs3_readdir_args;
1100 
1101 	/* RFS3_READDIRPLUS = 17 */
1102 	READDIRPLUS3args nfs3_readdirplus_args;
1103 
1104 	/* RFS3_FSSTAT = 18 */
1105 	FSSTAT3args nfs3_fsstat_args;
1106 
1107 	/* RFS3_FSINFO = 19 */
1108 	FSINFO3args nfs3_fsinfo_args;
1109 
1110 	/* RFS3_PATHCONF = 20 */
1111 	PATHCONF3args nfs3_pathconf_args;
1112 
1113 	/* RFS3_COMMIT = 21 */
1114 	COMMIT3args nfs3_commit_args;
1115 
1116 	/*
1117 	 * NFS VERSION 4
1118 	 */
1119 
1120 	/* RFS_NULL = 0 */
1121 
1122 	/* COMPUND = 1 */
1123 	COMPOUND4args nfs4_compound_args;
1124 };
1125 
1126 union rfs_res {
1127 	/*
1128 	 * NFS VERSION 2
1129 	 */
1130 
1131 	/* RFS_NULL = 0 */
1132 
1133 	/* RFS_GETATTR = 1 */
1134 	struct nfsattrstat nfs2_getattr_res;
1135 
1136 	/* RFS_SETATTR = 2 */
1137 	struct nfsattrstat nfs2_setattr_res;
1138 
1139 	/* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */
1140 
1141 	/* RFS_LOOKUP = 4 */
1142 	struct nfsdiropres nfs2_lookup_res;
1143 
1144 	/* RFS_READLINK = 5 */
1145 	struct nfsrdlnres nfs2_readlink_res;
1146 
1147 	/* RFS_READ = 6 */
1148 	struct nfsrdresult nfs2_read_res;
1149 
1150 	/* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */
1151 
1152 	/* RFS_WRITE = 8 */
1153 	struct nfsattrstat nfs2_write_res;
1154 
1155 	/* RFS_CREATE = 9 */
1156 	struct nfsdiropres nfs2_create_res;
1157 
1158 	/* RFS_REMOVE = 10 */
1159 	enum nfsstat nfs2_remove_res;
1160 
1161 	/* RFS_RENAME = 11 */
1162 	enum nfsstat nfs2_rename_res;
1163 
1164 	/* RFS_LINK = 12 */
1165 	enum nfsstat nfs2_link_res;
1166 
1167 	/* RFS_SYMLINK = 13 */
1168 	enum nfsstat nfs2_symlink_res;
1169 
1170 	/* RFS_MKDIR = 14 */
1171 	struct nfsdiropres nfs2_mkdir_res;
1172 
1173 	/* RFS_RMDIR = 15 */
1174 	enum nfsstat nfs2_rmdir_res;
1175 
1176 	/* RFS_READDIR = 16 */
1177 	struct nfsrddirres nfs2_readdir_res;
1178 
1179 	/* RFS_STATFS = 17 */
1180 	struct nfsstatfs nfs2_statfs_res;
1181 
1182 	/*
1183 	 * NFS VERSION 3
1184 	 */
1185 
1186 	/* RFS_NULL = 0 */
1187 
1188 	/* RFS3_GETATTR = 1 */
1189 	GETATTR3res nfs3_getattr_res;
1190 
1191 	/* RFS3_SETATTR = 2 */
1192 	SETATTR3res nfs3_setattr_res;
1193 
1194 	/* RFS3_LOOKUP = 3 */
1195 	LOOKUP3res nfs3_lookup_res;
1196 
1197 	/* RFS3_ACCESS = 4 */
1198 	ACCESS3res nfs3_access_res;
1199 
1200 	/* RFS3_READLINK = 5 */
1201 	READLINK3res nfs3_readlink_res;
1202 
1203 	/* RFS3_READ = 6 */
1204 	READ3res nfs3_read_res;
1205 
1206 	/* RFS3_WRITE = 7 */
1207 	WRITE3res nfs3_write_res;
1208 
1209 	/* RFS3_CREATE = 8 */
1210 	CREATE3res nfs3_create_res;
1211 
1212 	/* RFS3_MKDIR = 9 */
1213 	MKDIR3res nfs3_mkdir_res;
1214 
1215 	/* RFS3_SYMLINK = 10 */
1216 	SYMLINK3res nfs3_symlink_res;
1217 
1218 	/* RFS3_MKNOD = 11 */
1219 	MKNOD3res nfs3_mknod_res;
1220 
1221 	/* RFS3_REMOVE = 12 */
1222 	REMOVE3res nfs3_remove_res;
1223 
1224 	/* RFS3_RMDIR = 13 */
1225 	RMDIR3res nfs3_rmdir_res;
1226 
1227 	/* RFS3_RENAME = 14 */
1228 	RENAME3res nfs3_rename_res;
1229 
1230 	/* RFS3_LINK = 15 */
1231 	LINK3res nfs3_link_res;
1232 
1233 	/* RFS3_READDIR = 16 */
1234 	READDIR3res nfs3_readdir_res;
1235 
1236 	/* RFS3_READDIRPLUS = 17 */
1237 	READDIRPLUS3res nfs3_readdirplus_res;
1238 
1239 	/* RFS3_FSSTAT = 18 */
1240 	FSSTAT3res nfs3_fsstat_res;
1241 
1242 	/* RFS3_FSINFO = 19 */
1243 	FSINFO3res nfs3_fsinfo_res;
1244 
1245 	/* RFS3_PATHCONF = 20 */
1246 	PATHCONF3res nfs3_pathconf_res;
1247 
1248 	/* RFS3_COMMIT = 21 */
1249 	COMMIT3res nfs3_commit_res;
1250 
1251 	/*
1252 	 * NFS VERSION 4
1253 	 */
1254 
1255 	/* RFS_NULL = 0 */
1256 
1257 	/* RFS4_COMPOUND = 1 */
1258 	COMPOUND4res nfs4_compound_res;
1259 
1260 };
1261 
1262 static struct rpc_disptable rfs_disptable[] = {
1263 	{sizeof (rfsdisptab_v2) / sizeof (rfsdisptab_v2[0]),
1264 	    rfscallnames_v2,
1265 	    &rfsproccnt_v2_ptr, rfsdisptab_v2},
1266 	{sizeof (rfsdisptab_v3) / sizeof (rfsdisptab_v3[0]),
1267 	    rfscallnames_v3,
1268 	    &rfsproccnt_v3_ptr, rfsdisptab_v3},
1269 	{sizeof (rfsdisptab_v4) / sizeof (rfsdisptab_v4[0]),
1270 	    rfscallnames_v4,
1271 	    &rfsproccnt_v4_ptr, rfsdisptab_v4},
1272 };
1273 
1274 /*
1275  * If nfs_portmon is set, then clients are required to use privileged
1276  * ports (ports < IPPORT_RESERVED) in order to get NFS services.
1277  *
1278  * N.B.: this attempt to carry forward the already ill-conceived notion
1279  * of privileged ports for TCP/UDP is really quite ineffectual.  Not only
1280  * is it transport-dependent, it's laughably easy to spoof.  If you're
1281  * really interested in security, you must start with secure RPC instead.
1282  */
1283 static int nfs_portmon = 0;
1284 
1285 #ifdef DEBUG
1286 static int cred_hits = 0;
1287 static int cred_misses = 0;
1288 #endif
1289 
1290 
1291 #ifdef DEBUG
1292 /*
1293  * Debug code to allow disabling of rfs_dispatch() use of
1294  * fastxdrargs() and fastxdrres() calls for testing purposes.
1295  */
1296 static int rfs_no_fast_xdrargs = 0;
1297 static int rfs_no_fast_xdrres = 0;
1298 #endif
1299 
1300 union acl_args {
1301 	/*
1302 	 * ACL VERSION 2
1303 	 */
1304 
1305 	/* ACL2_NULL = 0 */
1306 
1307 	/* ACL2_GETACL = 1 */
1308 	GETACL2args acl2_getacl_args;
1309 
1310 	/* ACL2_SETACL = 2 */
1311 	SETACL2args acl2_setacl_args;
1312 
1313 	/* ACL2_GETATTR = 3 */
1314 	GETATTR2args acl2_getattr_args;
1315 
1316 	/* ACL2_ACCESS = 4 */
1317 	ACCESS2args acl2_access_args;
1318 
1319 	/* ACL2_GETXATTRDIR = 5 */
1320 	GETXATTRDIR2args acl2_getxattrdir_args;
1321 
1322 	/*
1323 	 * ACL VERSION 3
1324 	 */
1325 
1326 	/* ACL3_NULL = 0 */
1327 
1328 	/* ACL3_GETACL = 1 */
1329 	GETACL3args acl3_getacl_args;
1330 
1331 	/* ACL3_SETACL = 2 */
1332 	SETACL3args acl3_setacl;
1333 
1334 	/* ACL3_GETXATTRDIR = 3 */
1335 	GETXATTRDIR3args acl3_getxattrdir_args;
1336 
1337 };
1338 
1339 union acl_res {
1340 	/*
1341 	 * ACL VERSION 2
1342 	 */
1343 
1344 	/* ACL2_NULL = 0 */
1345 
1346 	/* ACL2_GETACL = 1 */
1347 	GETACL2res acl2_getacl_res;
1348 
1349 	/* ACL2_SETACL = 2 */
1350 	SETACL2res acl2_setacl_res;
1351 
1352 	/* ACL2_GETATTR = 3 */
1353 	GETATTR2res acl2_getattr_res;
1354 
1355 	/* ACL2_ACCESS = 4 */
1356 	ACCESS2res acl2_access_res;
1357 
1358 	/* ACL2_GETXATTRDIR = 5 */
1359 	GETXATTRDIR2args acl2_getxattrdir_res;
1360 
1361 	/*
1362 	 * ACL VERSION 3
1363 	 */
1364 
1365 	/* ACL3_NULL = 0 */
1366 
1367 	/* ACL3_GETACL = 1 */
1368 	GETACL3res acl3_getacl_res;
1369 
1370 	/* ACL3_SETACL = 2 */
1371 	SETACL3res acl3_setacl_res;
1372 
1373 	/* ACL3_GETXATTRDIR = 3 */
1374 	GETXATTRDIR3res acl3_getxattrdir_res;
1375 
1376 };
1377 
1378 static bool_t
1379 auth_tooweak(struct svc_req *req, char *res)
1380 {
1381 
1382 	if (req->rq_vers == NFS_VERSION && req->rq_proc == RFS_LOOKUP) {
1383 		struct nfsdiropres *dr = (struct nfsdiropres *)res;
1384 		if (dr->dr_status == WNFSERR_CLNT_FLAVOR)
1385 			return (TRUE);
1386 	} else if (req->rq_vers == NFS_V3 && req->rq_proc == NFSPROC3_LOOKUP) {
1387 		LOOKUP3res *resp = (LOOKUP3res *)res;
1388 		if (resp->status == WNFSERR_CLNT_FLAVOR)
1389 			return (TRUE);
1390 	}
1391 	return (FALSE);
1392 }
1393 
1394 
1395 static void
1396 common_dispatch(struct svc_req *req, SVCXPRT *xprt, rpcvers_t min_vers,
1397 		rpcvers_t max_vers, char *pgmname,
1398 		struct rpc_disptable *disptable)
1399 {
1400 	int which;
1401 	rpcvers_t vers;
1402 	char *args;
1403 	union {
1404 			union rfs_args ra;
1405 			union acl_args aa;
1406 		} args_buf;
1407 	char *res;
1408 	union {
1409 			union rfs_res rr;
1410 			union acl_res ar;
1411 		} res_buf;
1412 	struct rpcdisp *disp = NULL;
1413 	int dis_flags = 0;
1414 	cred_t *cr;
1415 	int error = 0;
1416 	int anon_ok;
1417 	struct exportinfo *exi = NULL;
1418 	unsigned int nfslog_rec_id;
1419 	int dupstat;
1420 	struct dupreq *dr;
1421 	int authres;
1422 	bool_t publicfh_ok = FALSE;
1423 	enum_t auth_flavor;
1424 	bool_t dupcached = FALSE;
1425 	struct netbuf	nb;
1426 	bool_t logging_enabled = FALSE;
1427 	struct exportinfo *nfslog_exi = NULL;
1428 	char **procnames;
1429 	char cbuf[INET6_ADDRSTRLEN];	/* to hold both IPv4 and IPv6 addr */
1430 
1431 	vers = req->rq_vers;
1432 
1433 	if (vers < min_vers || vers > max_vers) {
1434 		svcerr_progvers(req->rq_xprt, min_vers, max_vers);
1435 		error++;
1436 		cmn_err(CE_NOTE, "%s: bad version number %u", pgmname, vers);
1437 		goto done;
1438 	}
1439 	vers -= min_vers;
1440 
1441 	which = req->rq_proc;
1442 	if (which < 0 || which >= disptable[(int)vers].dis_nprocs) {
1443 		svcerr_noproc(req->rq_xprt);
1444 		error++;
1445 		goto done;
1446 	}
1447 
1448 	(*(disptable[(int)vers].dis_proccntp))[which].value.ui64++;
1449 
1450 	disp = &disptable[(int)vers].dis_table[which];
1451 	procnames = disptable[(int)vers].dis_procnames;
1452 
1453 	auth_flavor = req->rq_cred.oa_flavor;
1454 
1455 	/*
1456 	 * Deserialize into the args struct.
1457 	 */
1458 	args = (char *)&args_buf;
1459 
1460 #ifdef DEBUG
1461 	if (rfs_no_fast_xdrargs || (auth_flavor == RPCSEC_GSS) ||
1462 	    disp->dis_fastxdrargs == NULL_xdrproc_t ||
1463 	    !SVC_GETARGS(xprt, disp->dis_fastxdrargs, (char *)&args))
1464 #else
1465 	if ((auth_flavor == RPCSEC_GSS) ||
1466 	    disp->dis_fastxdrargs == NULL_xdrproc_t ||
1467 	    !SVC_GETARGS(xprt, disp->dis_fastxdrargs, (char *)&args))
1468 #endif
1469 	{
1470 		bzero(args, disp->dis_argsz);
1471 		if (!SVC_GETARGS(xprt, disp->dis_xdrargs, args)) {
1472 			svcerr_decode(xprt);
1473 			error++;
1474 			cmn_err(CE_NOTE,
1475 			    "Failed to decode arguments for %s version %u "
1476 			    "procedure %s client %s%s",
1477 			    pgmname, vers + min_vers, procnames[which],
1478 			    client_name(req), client_addr(req, cbuf));
1479 			goto done;
1480 		}
1481 	}
1482 
1483 	/*
1484 	 * If Version 4 use that specific dispatch function.
1485 	 */
1486 	if (req->rq_vers == 4) {
1487 		error += rfs4_dispatch(disp, req, xprt, args);
1488 		goto done;
1489 	}
1490 
1491 	dis_flags = disp->dis_flags;
1492 
1493 	/*
1494 	 * Find export information and check authentication,
1495 	 * setting the credential if everything is ok.
1496 	 */
1497 	if (disp->dis_getfh != NULL) {
1498 		void *fh;
1499 		fsid_t *fsid;
1500 		fid_t *fid, *xfid;
1501 		fhandle_t *fh2;
1502 		nfs_fh3 *fh3;
1503 
1504 		fh = (*disp->dis_getfh)(args);
1505 		switch (req->rq_vers) {
1506 		case NFS_VERSION:
1507 			fh2 = (fhandle_t *)fh;
1508 			fsid = &fh2->fh_fsid;
1509 			fid = (fid_t *)&fh2->fh_len;
1510 			xfid = (fid_t *)&fh2->fh_xlen;
1511 			break;
1512 		case NFS_V3:
1513 			fh3 = (nfs_fh3 *)fh;
1514 			fsid = &fh3->fh3_fsid;
1515 			fid = FH3TOFIDP(fh3);
1516 			xfid = FH3TOXFIDP(fh3);
1517 			break;
1518 		}
1519 
1520 		/*
1521 		 * Fix for bug 1038302 - corbin
1522 		 * There is a problem here if anonymous access is
1523 		 * disallowed.  If the current request is part of the
1524 		 * client's mount process for the requested filesystem,
1525 		 * then it will carry root (uid 0) credentials on it, and
1526 		 * will be denied by checkauth if that client does not
1527 		 * have explicit root=0 permission.  This will cause the
1528 		 * client's mount operation to fail.  As a work-around,
1529 		 * we check here to see if the request is a getattr or
1530 		 * statfs operation on the exported vnode itself, and
1531 		 * pass a flag to checkauth with the result of this test.
1532 		 *
1533 		 * The filehandle refers to the mountpoint itself if
1534 		 * the fh_data and fh_xdata portions of the filehandle
1535 		 * are equal.
1536 		 *
1537 		 * Added anon_ok argument to checkauth().
1538 		 */
1539 
1540 		if ((dis_flags & RPC_ALLOWANON) && EQFID(fid, xfid))
1541 			anon_ok = 1;
1542 		else
1543 			anon_ok = 0;
1544 
1545 		cr = xprt->xp_cred;
1546 		ASSERT(cr != NULL);
1547 #ifdef DEBUG
1548 		if (crgetref(cr) != 1) {
1549 			crfree(cr);
1550 			cr = crget();
1551 			xprt->xp_cred = cr;
1552 			cred_misses++;
1553 		} else
1554 			cred_hits++;
1555 #else
1556 		if (crgetref(cr) != 1) {
1557 			crfree(cr);
1558 			cr = crget();
1559 			xprt->xp_cred = cr;
1560 		}
1561 #endif
1562 
1563 		exi = checkexport(fsid, xfid);
1564 
1565 		if (exi != NULL) {
1566 			publicfh_ok = PUBLICFH_CHECK(disp, exi, fsid, xfid);
1567 
1568 			/*
1569 			 * Don't allow non-V4 clients access
1570 			 * to pseudo exports
1571 			 */
1572 			if (PSEUDO(exi)) {
1573 				svcerr_weakauth(xprt);
1574 				error++;
1575 				goto done;
1576 			}
1577 
1578 			authres = checkauth(exi, req, cr, anon_ok, publicfh_ok);
1579 			/*
1580 			 * authres >  0: authentication OK - proceed
1581 			 * authres == 0: authentication weak - return error
1582 			 * authres <  0: authentication timeout - drop
1583 			 */
1584 			if (authres <= 0) {
1585 				if (authres == 0) {
1586 					svcerr_weakauth(xprt);
1587 					error++;
1588 				}
1589 				goto done;
1590 			}
1591 		}
1592 	} else
1593 		cr = NULL;
1594 
1595 	if ((dis_flags & RPC_MAPRESP) && (auth_flavor != RPCSEC_GSS)) {
1596 		res = (char *)SVC_GETRES(xprt, disp->dis_ressz);
1597 		if (res == NULL)
1598 			res = (char *)&res_buf;
1599 	} else
1600 		res = (char *)&res_buf;
1601 
1602 	if (!(dis_flags & RPC_IDEMPOTENT)) {
1603 		dupstat = SVC_DUP_EXT(xprt, req, res, disp->dis_ressz, &dr,
1604 				&dupcached);
1605 
1606 		switch (dupstat) {
1607 		case DUP_ERROR:
1608 			svcerr_systemerr(xprt);
1609 			error++;
1610 			goto done;
1611 			/* NOTREACHED */
1612 		case DUP_INPROGRESS:
1613 			if (res != (char *)&res_buf)
1614 				SVC_FREERES(xprt);
1615 			error++;
1616 			goto done;
1617 			/* NOTREACHED */
1618 		case DUP_NEW:
1619 		case DUP_DROP:
1620 			curthread->t_flag |= T_DONTPEND;
1621 
1622 			(*disp->dis_proc)(args, res, exi, req, cr);
1623 
1624 			curthread->t_flag &= ~T_DONTPEND;
1625 			if (curthread->t_flag & T_WOULDBLOCK) {
1626 				curthread->t_flag &= ~T_WOULDBLOCK;
1627 				SVC_DUPDONE_EXT(xprt, dr, res, NULL,
1628 					disp->dis_ressz, DUP_DROP);
1629 				if (res != (char *)&res_buf)
1630 					SVC_FREERES(xprt);
1631 				error++;
1632 				goto done;
1633 			}
1634 			if (dis_flags & RPC_AVOIDWORK) {
1635 				SVC_DUPDONE_EXT(xprt, dr, res, NULL,
1636 					disp->dis_ressz, DUP_DROP);
1637 			} else {
1638 				SVC_DUPDONE_EXT(xprt, dr, res,
1639 					disp->dis_resfree == nullfree ? NULL :
1640 					disp->dis_resfree,
1641 					disp->dis_ressz, DUP_DONE);
1642 				dupcached = TRUE;
1643 			}
1644 			break;
1645 		case DUP_DONE:
1646 			break;
1647 		}
1648 
1649 	} else {
1650 		curthread->t_flag |= T_DONTPEND;
1651 
1652 		(*disp->dis_proc)(args, res, exi, req, cr);
1653 
1654 		curthread->t_flag &= ~T_DONTPEND;
1655 		if (curthread->t_flag & T_WOULDBLOCK) {
1656 			curthread->t_flag &= ~T_WOULDBLOCK;
1657 			if (res != (char *)&res_buf)
1658 				SVC_FREERES(xprt);
1659 			error++;
1660 			goto done;
1661 		}
1662 	}
1663 
1664 	if (auth_tooweak(req, res)) {
1665 		svcerr_weakauth(xprt);
1666 		error++;
1667 		goto done;
1668 	}
1669 
1670 	/*
1671 	 * Check to see if logging has been enabled on the server.
1672 	 * If so, then obtain the export info struct to be used for
1673 	 * the later writing of the log record.  This is done for
1674 	 * the case that a lookup is done across a non-logged public
1675 	 * file system.
1676 	 */
1677 	if (nfslog_buffer_list != NULL) {
1678 		nfslog_exi = nfslog_get_exi(exi, req, res, &nfslog_rec_id);
1679 		/*
1680 		 * Is logging enabled?
1681 		 */
1682 		logging_enabled = (nfslog_exi != NULL);
1683 
1684 		/*
1685 		 * Copy the netbuf for logging purposes, before it is
1686 		 * freed by svc_sendreply().
1687 		 */
1688 		if (logging_enabled) {
1689 			NFSLOG_COPY_NETBUF(nfslog_exi, xprt, &nb);
1690 			/*
1691 			 * If RPC_MAPRESP flag set (i.e. in V2 ops) the
1692 			 * res gets copied directly into the mbuf and
1693 			 * may be freed soon after the sendreply. So we
1694 			 * must copy it here to a safe place...
1695 			 */
1696 			if (res != (char *)&res_buf) {
1697 				bcopy(res, (char *)&res_buf, disp->dis_ressz);
1698 			}
1699 		}
1700 	}
1701 
1702 	/*
1703 	 * Serialize and send results struct
1704 	 */
1705 #ifdef DEBUG
1706 	if (rfs_no_fast_xdrres == 0 && res != (char *)&res_buf)
1707 #else
1708 	if (res != (char *)&res_buf)
1709 #endif
1710 	{
1711 		if (!svc_sendreply(xprt, disp->dis_fastxdrres, res)) {
1712 			cmn_err(CE_NOTE, "%s: bad sendreply", pgmname);
1713 			error++;
1714 		}
1715 	} else {
1716 		if (!svc_sendreply(xprt, disp->dis_xdrres, res)) {
1717 			cmn_err(CE_NOTE, "%s: bad sendreply", pgmname);
1718 			error++;
1719 		}
1720 	}
1721 
1722 	/*
1723 	 * Log if needed
1724 	 */
1725 	if (logging_enabled) {
1726 		nfslog_write_record(nfslog_exi, req, args, (char *)&res_buf,
1727 			cr, &nb, nfslog_rec_id, NFSLOG_ONE_BUFFER);
1728 		exi_rele(nfslog_exi);
1729 		kmem_free((&nb)->buf, (&nb)->len);
1730 	}
1731 
1732 	/*
1733 	 * Free results struct. With the addition of NFS V4 we can
1734 	 * have non-idempotent procedures with functions.
1735 	 */
1736 	if (disp->dis_resfree != nullfree && dupcached == FALSE) {
1737 		(*disp->dis_resfree)(res);
1738 	}
1739 
1740 done:
1741 	/*
1742 	 * Free arguments struct
1743 	 */
1744 	if (disp) {
1745 		if (!SVC_FREEARGS(xprt, disp->dis_xdrargs, args)) {
1746 			cmn_err(CE_NOTE, "%s: bad freeargs", pgmname);
1747 			error++;
1748 		}
1749 	} else {
1750 		if (!SVC_FREEARGS(xprt, (xdrproc_t)0, (caddr_t)0)) {
1751 			cmn_err(CE_NOTE, "%s: bad freeargs", pgmname);
1752 			error++;
1753 		}
1754 	}
1755 
1756 	if (exi != NULL)
1757 		exi_rele(exi);
1758 
1759 	global_svstat_ptr[req->rq_vers][NFS_BADCALLS].value.ui64 += error;
1760 
1761 	global_svstat_ptr[req->rq_vers][NFS_CALLS].value.ui64++;
1762 }
1763 
1764 static void
1765 rfs_dispatch(struct svc_req *req, SVCXPRT *xprt)
1766 {
1767 	common_dispatch(req, xprt, NFS_VERSMIN, NFS_VERSMAX,
1768 		"NFS", rfs_disptable);
1769 }
1770 
1771 static char *aclcallnames_v2[] = {
1772 	"ACL2_NULL",
1773 	"ACL2_GETACL",
1774 	"ACL2_SETACL",
1775 	"ACL2_GETATTR",
1776 	"ACL2_ACCESS",
1777 	"ACL2_GETXATTRDIR"
1778 };
1779 
1780 static struct rpcdisp acldisptab_v2[] = {
1781 	/*
1782 	 * ACL VERSION 2
1783 	 */
1784 
1785 	/* ACL2_NULL = 0 */
1786 	{rpc_null,
1787 	    xdr_void, NULL_xdrproc_t, 0,
1788 	    xdr_void, NULL_xdrproc_t, 0,
1789 	    nullfree, RPC_IDEMPOTENT,
1790 	    0},
1791 
1792 	/* ACL2_GETACL = 1 */
1793 	{acl2_getacl,
1794 	    xdr_GETACL2args, xdr_fastGETACL2args, sizeof (GETACL2args),
1795 	    xdr_GETACL2res, NULL_xdrproc_t, sizeof (GETACL2res),
1796 	    acl2_getacl_free, RPC_IDEMPOTENT,
1797 	    acl2_getacl_getfh},
1798 
1799 	/* ACL2_SETACL = 2 */
1800 	{acl2_setacl,
1801 	    xdr_SETACL2args, NULL_xdrproc_t, sizeof (SETACL2args),
1802 #ifdef _LITTLE_ENDIAN
1803 	    xdr_SETACL2res, xdr_fastSETACL2res, sizeof (SETACL2res),
1804 #else
1805 	    xdr_SETACL2res, NULL_xdrproc_t, sizeof (SETACL2res),
1806 #endif
1807 	    nullfree, RPC_MAPRESP,
1808 	    acl2_setacl_getfh},
1809 
1810 	/* ACL2_GETATTR = 3 */
1811 	{acl2_getattr,
1812 	    xdr_GETATTR2args, xdr_fastGETATTR2args, sizeof (GETATTR2args),
1813 #ifdef _LITTLE_ENDIAN
1814 	    xdr_GETATTR2res, xdr_fastGETATTR2res, sizeof (GETATTR2res),
1815 #else
1816 	    xdr_GETATTR2res, NULL_xdrproc_t, sizeof (GETATTR2res),
1817 #endif
1818 	    nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP,
1819 	    acl2_getattr_getfh},
1820 
1821 	/* ACL2_ACCESS = 4 */
1822 	{acl2_access,
1823 	    xdr_ACCESS2args, xdr_fastACCESS2args, sizeof (ACCESS2args),
1824 #ifdef _LITTLE_ENDIAN
1825 	    xdr_ACCESS2res, xdr_fastACCESS2res, sizeof (ACCESS2res),
1826 #else
1827 	    xdr_ACCESS2res, NULL_xdrproc_t, sizeof (ACCESS2res),
1828 #endif
1829 	    nullfree, RPC_IDEMPOTENT|RPC_MAPRESP,
1830 	    acl2_access_getfh},
1831 
1832 	/* ACL2_GETXATTRDIR = 5 */
1833 	{acl2_getxattrdir,
1834 	    xdr_GETXATTRDIR2args, NULL_xdrproc_t, sizeof (GETXATTRDIR2args),
1835 	    xdr_GETXATTRDIR2res, NULL_xdrproc_t, sizeof (GETXATTRDIR2res),
1836 	    nullfree, RPC_IDEMPOTENT,
1837 	    acl2_getxattrdir_getfh},
1838 };
1839 
1840 static char *aclcallnames_v3[] = {
1841 	"ACL3_NULL",
1842 	"ACL3_GETACL",
1843 	"ACL3_SETACL",
1844 	"ACL3_GETXATTRDIR"
1845 };
1846 
1847 static struct rpcdisp acldisptab_v3[] = {
1848 	/*
1849 	 * ACL VERSION 3
1850 	 */
1851 
1852 	/* ACL3_NULL = 0 */
1853 	{rpc_null,
1854 	    xdr_void, NULL_xdrproc_t, 0,
1855 	    xdr_void, NULL_xdrproc_t, 0,
1856 	    nullfree, RPC_IDEMPOTENT,
1857 	    0},
1858 
1859 	/* ACL3_GETACL = 1 */
1860 	{acl3_getacl,
1861 	    xdr_GETACL3args, NULL_xdrproc_t, sizeof (GETACL3args),
1862 	    xdr_GETACL3res, NULL_xdrproc_t, sizeof (GETACL3res),
1863 	    acl3_getacl_free, RPC_IDEMPOTENT,
1864 	    acl3_getacl_getfh},
1865 
1866 	/* ACL3_SETACL = 2 */
1867 	{acl3_setacl,
1868 	    xdr_SETACL3args, NULL_xdrproc_t, sizeof (SETACL3args),
1869 	    xdr_SETACL3res, NULL_xdrproc_t, sizeof (SETACL3res),
1870 	    nullfree, 0,
1871 	    acl3_setacl_getfh},
1872 
1873 	/* ACL3_GETXATTRDIR = 3 */
1874 	{acl3_getxattrdir,
1875 	    xdr_GETXATTRDIR3args, NULL_xdrproc_t, sizeof (GETXATTRDIR3args),
1876 	    xdr_GETXATTRDIR3res, NULL_xdrproc_t, sizeof (GETXATTRDIR3res),
1877 	    nullfree, RPC_IDEMPOTENT,
1878 	    acl3_getxattrdir_getfh},
1879 };
1880 
1881 static struct rpc_disptable acl_disptable[] = {
1882 	{sizeof (acldisptab_v2) / sizeof (acldisptab_v2[0]),
1883 		aclcallnames_v2,
1884 		&aclproccnt_v2_ptr, acldisptab_v2},
1885 	{sizeof (acldisptab_v3) / sizeof (acldisptab_v3[0]),
1886 		aclcallnames_v3,
1887 		&aclproccnt_v3_ptr, acldisptab_v3},
1888 };
1889 
1890 static void
1891 acl_dispatch(struct svc_req *req, SVCXPRT *xprt)
1892 {
1893 	common_dispatch(req, xprt, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,
1894 		"ACL", acl_disptable);
1895 }
1896 
1897 int
1898 checkwin(int flavor, int window, struct svc_req *req)
1899 {
1900 	struct authdes_cred *adc;
1901 
1902 	switch (flavor) {
1903 	case AUTH_DES:
1904 		adc = (struct authdes_cred *)req->rq_clntcred;
1905 		if (adc->adc_fullname.window > window)
1906 			return (0);
1907 		break;
1908 
1909 	default:
1910 		break;
1911 	}
1912 	return (1);
1913 }
1914 
1915 
1916 /*
1917  * checkauth() will check the access permission against the export
1918  * information.  Then map root uid/gid to appropriate uid/gid.
1919  *
1920  * This routine is used by NFS V3 and V2 code.
1921  */
1922 static int
1923 checkauth(struct exportinfo *exi, struct svc_req *req, cred_t *cr, int anon_ok,
1924     bool_t publicfh_ok)
1925 {
1926 	int i, nfsflavor, rpcflavor, stat, access;
1927 	struct secinfo *secp;
1928 	caddr_t principal;
1929 	char buf[INET6_ADDRSTRLEN]; /* to hold both IPv4 and IPv6 addr */
1930 	int anon_res = 0;
1931 
1932 	/*
1933 	 *	Check for privileged port number
1934 	 *	N.B.:  this assumes that we know the format of a netbuf.
1935 	 */
1936 	if (nfs_portmon) {
1937 		struct sockaddr *ca;
1938 		ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
1939 
1940 		if (ca == NULL)
1941 			return (0);
1942 
1943 		if ((ca->sa_family == AF_INET &&
1944 		    ntohs(((struct sockaddr_in *)ca)->sin_port) >=
1945 		    IPPORT_RESERVED) ||
1946 		    (ca->sa_family == AF_INET6 &&
1947 		    ntohs(((struct sockaddr_in6 *)ca)->sin6_port) >=
1948 		    IPPORT_RESERVED)) {
1949 			cmn_err(CE_NOTE,
1950 			    "nfs_server: client %s%ssent NFS request from "
1951 			    "unprivileged port",
1952 			    client_name(req), client_addr(req, buf));
1953 			return (0);
1954 		}
1955 	}
1956 
1957 	/*
1958 	 *  return 1 on success or 0 on failure
1959 	 */
1960 	stat = sec_svc_getcred(req, cr, &principal, &nfsflavor);
1961 
1962 	/*
1963 	 * A failed AUTH_UNIX svc_get_cred() implies we couldn't set
1964 	 * the credentials; below we map that to anonymous.
1965 	 */
1966 	if (!stat && nfsflavor != AUTH_UNIX) {
1967 		cmn_err(CE_NOTE,
1968 		    "nfs_server: couldn't get unix cred for %s",
1969 		    client_name(req));
1970 		return (0);
1971 	}
1972 
1973 	/*
1974 	 * Short circuit checkauth() on operations that support the
1975 	 * public filehandle, and if the request for that operation
1976 	 * is using the public filehandle. Note that we must call
1977 	 * sec_svc_getcred() first so that xp_cookie is set to the
1978 	 * right value. Normally xp_cookie is just the RPC flavor
1979 	 * of the the request, but in the case of RPCSEC_GSS it
1980 	 * could be a pseudo flavor.
1981 	 */
1982 	if (publicfh_ok)
1983 		return (1);
1984 
1985 	rpcflavor = req->rq_cred.oa_flavor;
1986 	/*
1987 	 * Check if the auth flavor is valid for this export
1988 	 */
1989 	access = nfsauth_access(exi, req);
1990 	if (access & NFSAUTH_DROP)
1991 		return (-1);	/* drop the request */
1992 
1993 	if (access & NFSAUTH_DENIED) {
1994 		/*
1995 		 * If anon_ok == 1 and we got NFSAUTH_DENIED, it was
1996 		 * probably due to the flavor not matching during the
1997 		 * the mount attempt. So map the flavor to AUTH_NONE
1998 		 * so that the credentials get mapped to the anonymous
1999 		 * user.
2000 		 */
2001 		if (anon_ok == 1)
2002 			rpcflavor = AUTH_NONE;
2003 		else
2004 			return (0);	/* deny access */
2005 
2006 	} else if (access & NFSAUTH_MAPNONE) {
2007 		/*
2008 		 * Access was granted even though the flavor mismatched
2009 		 * because AUTH_NONE was one of the exported flavors.
2010 		 */
2011 		rpcflavor = AUTH_NONE;
2012 
2013 	} else if (access & NFSAUTH_WRONGSEC) {
2014 		/*
2015 		 * NFSAUTH_WRONGSEC is used for NFSv4. Since V2/V3 already
2016 		 * negotiates the security flavor thru MOUNT protocol, the
2017 		 * only way it can get NFSAUTH_WRONGSEC here is from
2018 		 * NFS_ACL for V4. This could be for a limited view, so
2019 		 * map it to RO access. V4 lookup/readdir will take care
2020 		 * of the limited view portion.
2021 		 */
2022 		access |= NFSAUTH_RO;
2023 		access &= ~NFSAUTH_WRONGSEC;
2024 	}
2025 
2026 	switch (rpcflavor) {
2027 	case AUTH_NONE:
2028 		anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2029 				exi->exi_export.ex_anon);
2030 		(void) crsetgroups(cr, 0, NULL);
2031 		break;
2032 
2033 	case AUTH_UNIX:
2034 		if (!stat || crgetuid(cr) == 0 && !(access & NFSAUTH_ROOT)) {
2035 			anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2036 					exi->exi_export.ex_anon);
2037 			(void) crsetgroups(cr, 0, NULL);
2038 		}
2039 		break;
2040 
2041 	case AUTH_DES:
2042 	case RPCSEC_GSS:
2043 		/*
2044 		 *  Find the secinfo structure.  We should be able
2045 		 *  to find it by the time we reach here.
2046 		 *  nfsauth_access() has done the checking.
2047 		 */
2048 		secp = NULL;
2049 		for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
2050 			if (exi->exi_export.ex_secinfo[i].s_secinfo.sc_nfsnum ==
2051 			    nfsflavor) {
2052 				secp = &exi->exi_export.ex_secinfo[i];
2053 				break;
2054 			}
2055 		}
2056 
2057 		if (!secp) {
2058 			cmn_err(CE_NOTE, "nfs_server: client %s%shad "
2059 			    "no secinfo data for flavor %d",
2060 			    client_name(req), client_addr(req, buf),
2061 			    nfsflavor);
2062 			return (0);
2063 		}
2064 
2065 		if (!checkwin(rpcflavor, secp->s_window, req)) {
2066 			cmn_err(CE_NOTE,
2067 			    "nfs_server: client %s%sused invalid "
2068 			    "auth window value",
2069 			    client_name(req), client_addr(req, buf));
2070 			return (0);
2071 		}
2072 
2073 		/*
2074 		 * Map root principals listed in the share's root= list to root,
2075 		 * and map any others principals that were mapped to root by RPC
2076 		 * to anon.
2077 		 */
2078 		if (principal && sec_svc_inrootlist(rpcflavor, principal,
2079 			secp->s_rootcnt, secp->s_rootnames)) {
2080 			if (crgetuid(cr) == 0)
2081 				return (1);
2082 
2083 			(void) crsetugid(cr, 0, 0);
2084 
2085 			/*
2086 			 * NOTE: If and when kernel-land privilege tracing is
2087 			 * added this may have to be replaced with code that
2088 			 * retrieves root's supplementary groups (e.g., using
2089 			 * kgss_get_group_info().  In the meantime principals
2090 			 * mapped to uid 0 get all privileges, so setting cr's
2091 			 * supplementary groups for them does nothing.
2092 			 */
2093 			(void) crsetgroups(cr, 0, NULL);
2094 
2095 			return (1);
2096 		}
2097 
2098 		/*
2099 		 * Not a root princ, or not in root list, map UID 0/nobody to
2100 		 * the anon ID for the share.  (RPC sets cr's UIDs and GIDs to
2101 		 * UID_NOBODY and GID_NOBODY, respectively.)
2102 		 */
2103 		if (crgetuid(cr) != 0 &&
2104 		    (crgetuid(cr) != UID_NOBODY || crgetgid(cr) != GID_NOBODY))
2105 			return (1);
2106 
2107 		anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2108 			exi->exi_export.ex_anon);
2109 		(void) crsetgroups(cr, 0, NULL);
2110 		break;
2111 	default:
2112 		return (0);
2113 	} /* switch on rpcflavor */
2114 
2115 	/*
2116 	 * Even if anon access is disallowed via ex_anon == -1, we allow
2117 	 * this access if anon_ok is set.  So set creds to the default
2118 	 * "nobody" id.
2119 	 */
2120 	if (anon_res != 0) {
2121 		if (anon_ok == 0) {
2122 			cmn_err(CE_NOTE,
2123 			    "nfs_server: client %s%ssent wrong "
2124 			    "authentication for %s",
2125 			    client_name(req), client_addr(req, buf),
2126 			    exi->exi_export.ex_path ?
2127 			    exi->exi_export.ex_path : "?");
2128 			return (0);
2129 		}
2130 
2131 		if (crsetugid(cr, UID_NOBODY, GID_NOBODY) != 0)
2132 			return (0);
2133 	}
2134 
2135 	return (1);
2136 }
2137 
2138 /*
2139  * returns 0 on failure, -1 on a drop, -2 on wrong security flavor,
2140  * and 1 on success
2141  */
2142 int
2143 checkauth4(struct compound_state *cs, struct svc_req *req)
2144 {
2145 	int i, rpcflavor, access;
2146 	struct secinfo *secp;
2147 	char buf[MAXHOST + 1];
2148 	int anon_res = 0, nfsflavor;
2149 	struct exportinfo *exi;
2150 	cred_t	*cr;
2151 	caddr_t	principal;
2152 
2153 	exi = cs->exi;
2154 	cr = cs->cr;
2155 	principal = cs->principal;
2156 	nfsflavor = cs->nfsflavor;
2157 
2158 	ASSERT(cr != NULL);
2159 
2160 	rpcflavor = req->rq_cred.oa_flavor;
2161 	cs->access &= ~CS_ACCESS_LIMITED;
2162 
2163 	/*
2164 	 * Check the access right per auth flavor on the vnode of
2165 	 * this export for the given request.
2166 	 */
2167 	access = nfsauth4_access(cs->exi, cs->vp, req);
2168 
2169 	if (access & NFSAUTH_WRONGSEC)
2170 		return (-2);	/* no access for this security flavor */
2171 
2172 	if (access & NFSAUTH_DROP)
2173 		return (-1);	/* drop the request */
2174 
2175 	if (access & NFSAUTH_DENIED) {
2176 
2177 		if (exi->exi_export.ex_seccnt > 0)
2178 			return (0);	/* deny access */
2179 
2180 	} else if (access & NFSAUTH_LIMITED) {
2181 
2182 		cs->access |= CS_ACCESS_LIMITED;
2183 
2184 	} else if (access & NFSAUTH_MAPNONE) {
2185 		/*
2186 		 * Access was granted even though the flavor mismatched
2187 		 * because AUTH_NONE was one of the exported flavors.
2188 		 */
2189 		rpcflavor = AUTH_NONE;
2190 	}
2191 
2192 	/*
2193 	 * XXX probably need to redo some of it for nfsv4?
2194 	 * return 1 on success or 0 on failure
2195 	 */
2196 
2197 	switch (rpcflavor) {
2198 	case AUTH_NONE:
2199 		anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2200 				exi->exi_export.ex_anon);
2201 		(void) crsetgroups(cr, 0, NULL);
2202 		break;
2203 
2204 	case AUTH_UNIX:
2205 		if (crgetuid(cr) == 0 && !(access & NFSAUTH_ROOT)) {
2206 			anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2207 					exi->exi_export.ex_anon);
2208 			(void) crsetgroups(cr, 0, NULL);
2209 		}
2210 		break;
2211 
2212 	default:
2213 		/*
2214 		 *  Find the secinfo structure.  We should be able
2215 		 *  to find it by the time we reach here.
2216 		 *  nfsauth_access() has done the checking.
2217 		 */
2218 		secp = NULL;
2219 		for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
2220 			if (exi->exi_export.ex_secinfo[i].s_secinfo.sc_nfsnum ==
2221 			    nfsflavor) {
2222 				secp = &exi->exi_export.ex_secinfo[i];
2223 				break;
2224 			}
2225 		}
2226 
2227 		if (!secp) {
2228 			cmn_err(CE_NOTE, "nfs_server: client %s%shad "
2229 			    "no secinfo data for flavor %d",
2230 			    client_name(req), client_addr(req, buf),
2231 			    nfsflavor);
2232 			return (0);
2233 		}
2234 
2235 		if (!checkwin(rpcflavor, secp->s_window, req)) {
2236 			cmn_err(CE_NOTE,
2237 			    "nfs_server: client %s%sused invalid "
2238 			    "auth window value",
2239 			    client_name(req), client_addr(req, buf));
2240 			return (0);
2241 		}
2242 
2243 		/*
2244 		 * Map root principals listed in the share's root= list to root,
2245 		 * and map any others principals that were mapped to root by RPC
2246 		 * to anon.
2247 		 */
2248 		if (principal && sec_svc_inrootlist(rpcflavor, principal,
2249 			secp->s_rootcnt, secp->s_rootnames)) {
2250 			if (crgetuid(cr) == 0)
2251 				return (1);
2252 
2253 			(void) crsetugid(cr, 0, 0);
2254 
2255 			/*
2256 			 * NOTE: If and when kernel-land privilege tracing is
2257 			 * added this may have to be replaced with code that
2258 			 * retrieves root's supplementary groups (e.g., using
2259 			 * kgss_get_group_info().  In the meantime principals
2260 			 * mapped to uid 0 get all privileges, so setting cr's
2261 			 * supplementary groups for them does nothing.
2262 			 */
2263 			(void) crsetgroups(cr, 0, NULL);
2264 
2265 			return (1);
2266 		}
2267 
2268 		/*
2269 		 * Not a root princ, or not in root list, map UID 0/nobody to
2270 		 * the anon ID for the share.  (RPC sets cr's UIDs and GIDs to
2271 		 * UID_NOBODY and GID_NOBODY, respectively.)
2272 		 */
2273 		if (crgetuid(cr) != 0 &&
2274 		    (crgetuid(cr) != UID_NOBODY || crgetgid(cr) != GID_NOBODY))
2275 			return (1);
2276 
2277 		anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2278 			exi->exi_export.ex_anon);
2279 		(void) crsetgroups(cr, 0, NULL);
2280 		break;
2281 	} /* switch on rpcflavor */
2282 
2283 	/*
2284 	 * Even if anon access is disallowed via ex_anon == -1, we allow
2285 	 * this access if anon_ok is set.  So set creds to the default
2286 	 * "nobody" id.
2287 	 */
2288 
2289 	if (anon_res != 0) {
2290 		cmn_err(CE_NOTE,
2291 			"nfs_server: client %s%ssent wrong "
2292 			"authentication for %s",
2293 			client_name(req), client_addr(req, buf),
2294 			exi->exi_export.ex_path ?
2295 			exi->exi_export.ex_path : "?");
2296 		return (0);
2297 	}
2298 
2299 	return (1);
2300 }
2301 
2302 
2303 static char *
2304 client_name(struct svc_req *req)
2305 {
2306 	char *hostname = NULL;
2307 
2308 	/*
2309 	 * If it's a Unix cred then use the
2310 	 * hostname from the credential.
2311 	 */
2312 	if (req->rq_cred.oa_flavor == AUTH_UNIX) {
2313 		hostname = ((struct authunix_parms *)
2314 		    req->rq_clntcred)->aup_machname;
2315 	}
2316 	if (hostname == NULL)
2317 		hostname = "";
2318 
2319 	return (hostname);
2320 }
2321 
2322 static char *
2323 client_addr(struct svc_req *req, char *buf)
2324 {
2325 	struct sockaddr *ca;
2326 	uchar_t *b;
2327 	char *frontspace = "";
2328 
2329 	/*
2330 	 * We assume we are called in tandem with client_name and the
2331 	 * format string looks like "...client %s%sblah blah..."
2332 	 *
2333 	 * If it's a Unix cred then client_name returned
2334 	 * a host name, so we need insert a space between host name
2335 	 * and IP address.
2336 	 */
2337 	if (req->rq_cred.oa_flavor == AUTH_UNIX)
2338 		frontspace = " ";
2339 
2340 	/*
2341 	 * Convert the caller's IP address to a dotted string
2342 	 */
2343 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
2344 
2345 	if (ca->sa_family == AF_INET) {
2346 	    b = (uchar_t *)&((struct sockaddr_in *)ca)->sin_addr;
2347 	    (void) sprintf(buf, "%s(%d.%d.%d.%d) ", frontspace,
2348 		b[0] & 0xFF, b[1] & 0xFF, b[2] & 0xFF, b[3] & 0xFF);
2349 	} else if (ca->sa_family == AF_INET6) {
2350 		struct sockaddr_in6 *sin6;
2351 		sin6 = (struct sockaddr_in6 *)ca;
2352 		(void) kinet_ntop6((uchar_t *)&sin6->sin6_addr,
2353 				buf, INET6_ADDRSTRLEN);
2354 
2355 	} else {
2356 
2357 		/*
2358 		 * No IP address to print. If there was a host name
2359 		 * printed, then we print a space.
2360 		 */
2361 		(void) sprintf(buf, frontspace);
2362 	}
2363 
2364 	return (buf);
2365 }
2366 
2367 /*
2368  * NFS Server initialization routine.  This routine should only be called
2369  * once.  It performs the following tasks:
2370  *	- Call sub-initialization routines (localize access to variables)
2371  *	- Initialize all locks
2372  *	- initialize the version 3 write verifier
2373  */
2374 int
2375 nfs_srvinit(void)
2376 {
2377 	int error;
2378 
2379 	error = nfs_exportinit();
2380 	if (error != 0)
2381 		return (error);
2382 	error = rfs4_srvrinit();
2383 	if (error != 0) {
2384 		nfs_exportfini();
2385 		return (error);
2386 	}
2387 	rfs_srvrinit();
2388 	rfs3_srvrinit();
2389 	nfsauth_init();
2390 
2391 	/* Init the stuff to control start/stop */
2392 	nfs_server_upordown = NFS_SERVER_STOPPED;
2393 	mutex_init(&nfs_server_upordown_lock, NULL, MUTEX_DEFAULT, NULL);
2394 	cv_init(&nfs_server_upordown_cv, NULL, CV_DEFAULT, NULL);
2395 	mutex_init(&rdma_wait_mutex, NULL, MUTEX_DEFAULT, NULL);
2396 	cv_init(&rdma_wait_cv, NULL, CV_DEFAULT, NULL);
2397 
2398 	return (0);
2399 }
2400 
2401 /*
2402  * NFS Server finalization routine. This routine is called to cleanup the
2403  * initialization work previously performed if the NFS server module could
2404  * not be loaded correctly.
2405  */
2406 void
2407 nfs_srvfini(void)
2408 {
2409 	nfsauth_fini();
2410 	rfs3_srvrfini();
2411 	rfs_srvrfini();
2412 	nfs_exportfini();
2413 
2414 	mutex_destroy(&nfs_server_upordown_lock);
2415 	cv_destroy(&nfs_server_upordown_cv);
2416 	mutex_destroy(&rdma_wait_mutex);
2417 	cv_destroy(&rdma_wait_cv);
2418 }
2419 
2420 /*
2421  * Set up an iovec array of up to cnt pointers.
2422  */
2423 
2424 void
2425 mblk_to_iov(mblk_t *m, int cnt, struct iovec *iovp)
2426 {
2427 	while (m != NULL && cnt-- > 0) {
2428 		iovp->iov_base = (caddr_t)m->b_rptr;
2429 		iovp->iov_len = (m->b_wptr - m->b_rptr);
2430 		iovp++;
2431 		m = m->b_cont;
2432 	}
2433 }
2434 
2435 /*
2436  * Common code between NFS Version 2 and NFS Version 3 for the public
2437  * filehandle multicomponent lookups.
2438  */
2439 
2440 /*
2441  * Public filehandle evaluation of a multi-component lookup, following
2442  * symbolic links, if necessary. This may result in a vnode in another
2443  * filesystem, which is OK as long as the other filesystem is exported.
2444  *
2445  * Note that the exi will be set either to NULL or a new reference to the
2446  * exportinfo struct that corresponds to the vnode of the multi-component path.
2447  * It is the callers responsibility to release this reference.
2448  */
2449 int
2450 rfs_publicfh_mclookup(char *p, vnode_t *dvp, cred_t *cr, vnode_t **vpp,
2451     struct exportinfo **exi, struct sec_ol *sec)
2452 {
2453 	int pathflag;
2454 	vnode_t *mc_dvp = NULL;
2455 	vnode_t *realvp;
2456 	int error;
2457 
2458 	*exi = NULL;
2459 
2460 	/*
2461 	 * check if the given path is a url or native path. Since p is
2462 	 * modified by MCLpath(), it may be empty after returning from
2463 	 * there, and should be checked.
2464 	 */
2465 	if ((pathflag = MCLpath(&p)) == -1)
2466 		return (EIO);
2467 
2468 	/*
2469 	 * If pathflag is SECURITY_QUERY, turn the SEC_QUERY bit
2470 	 * on in sec->sec_flags. This bit will later serve as an
2471 	 * indication in makefh_ol() or makefh3_ol() to overload the
2472 	 * filehandle to contain the sec modes used by the server for
2473 	 * the path.
2474 	 */
2475 	if (pathflag == SECURITY_QUERY) {
2476 		if ((sec->sec_index = (uint_t)(*p)) > 0) {
2477 			sec->sec_flags |= SEC_QUERY;
2478 			p++;
2479 			if ((pathflag = MCLpath(&p)) == -1)
2480 				return (EIO);
2481 		} else {
2482 			cmn_err(CE_NOTE,
2483 			    "nfs_server: invalid security index %d, "
2484 			    "violating WebNFS SNEGO protocol.", sec->sec_index);
2485 			return (EIO);
2486 		}
2487 	}
2488 
2489 	if (p[0] == '\0') {
2490 		error = ENOENT;
2491 		goto publicfh_done;
2492 	}
2493 
2494 	error = rfs_pathname(p, &mc_dvp, vpp, dvp, cr, pathflag);
2495 
2496 	/*
2497 	 * If name resolves to "/" we get EINVAL since we asked for
2498 	 * the vnode of the directory that the file is in. Try again
2499 	 * with NULL directory vnode.
2500 	 */
2501 	if (error == EINVAL) {
2502 		error = rfs_pathname(p, NULL, vpp, dvp, cr, pathflag);
2503 		if (!error) {
2504 			ASSERT(*vpp != NULL);
2505 			if ((*vpp)->v_type == VDIR) {
2506 				VN_HOLD(*vpp);
2507 				mc_dvp = *vpp;
2508 			} else {
2509 				/*
2510 				 * This should not happen, the filesystem is
2511 				 * in an inconsistent state. Fail the lookup
2512 				 * at this point.
2513 				 */
2514 				VN_RELE(*vpp);
2515 				error = EINVAL;
2516 			}
2517 		}
2518 	}
2519 
2520 	if (error)
2521 		goto publicfh_done;
2522 
2523 	if (*vpp == NULL) {
2524 		error = ENOENT;
2525 		goto publicfh_done;
2526 	}
2527 
2528 	ASSERT(mc_dvp != NULL);
2529 	ASSERT(*vpp != NULL);
2530 
2531 	if ((*vpp)->v_type == VDIR) {
2532 		do {
2533 			/*
2534 			 * *vpp may be an AutoFS node, so we perform
2535 			 * a VOP_ACCESS() to trigger the mount of the intended
2536 			 * filesystem, so we can perform the lookup in the
2537 			 * intended filesystem.
2538 			 */
2539 			(void) VOP_ACCESS(*vpp, 0, 0, cr);
2540 
2541 			/*
2542 			 * If vnode is covered, get the
2543 			 * the topmost vnode.
2544 			 */
2545 			if (vn_mountedvfs(*vpp) != NULL) {
2546 				error = traverse(vpp);
2547 				if (error) {
2548 					VN_RELE(*vpp);
2549 					goto publicfh_done;
2550 				}
2551 			}
2552 
2553 			if (VOP_REALVP(*vpp, &realvp) == 0 && realvp != *vpp) {
2554 				/*
2555 				 * If realvp is different from *vpp
2556 				 * then release our reference on *vpp, so that
2557 				 * the export access check be performed on the
2558 				 * real filesystem instead.
2559 				 */
2560 				VN_HOLD(realvp);
2561 				VN_RELE(*vpp);
2562 				*vpp = realvp;
2563 			} else
2564 			    break;
2565 		/* LINTED */
2566 		} while (TRUE);
2567 
2568 		/*
2569 		 * Let nfs_vptexi() figure what the real parent is.
2570 		 */
2571 		VN_RELE(mc_dvp);
2572 		mc_dvp = NULL;
2573 
2574 	} else {
2575 		/*
2576 		 * If vnode is covered, get the
2577 		 * the topmost vnode.
2578 		 */
2579 		if (vn_mountedvfs(mc_dvp) != NULL) {
2580 			error = traverse(&mc_dvp);
2581 			if (error) {
2582 			    VN_RELE(*vpp);
2583 			    goto publicfh_done;
2584 			}
2585 		}
2586 
2587 		if (VOP_REALVP(mc_dvp, &realvp) == 0 && realvp != mc_dvp) {
2588 			/*
2589 			 * *vpp is a file, obtain realvp of the parent
2590 			 * directory vnode.
2591 			 */
2592 			VN_HOLD(realvp);
2593 			VN_RELE(mc_dvp);
2594 			mc_dvp = realvp;
2595 		}
2596 	}
2597 
2598 	/*
2599 	 * The pathname may take us from the public filesystem to another.
2600 	 * If that's the case then just set the exportinfo to the new export
2601 	 * and build filehandle for it. Thanks to per-access checking there's
2602 	 * no security issues with doing this. If the client is not allowed
2603 	 * access to this new export then it will get an access error when it
2604 	 * tries to use the filehandle
2605 	 */
2606 	if (error = nfs_check_vpexi(mc_dvp, *vpp, kcred, exi)) {
2607 		VN_RELE(*vpp);
2608 		goto publicfh_done;
2609 	}
2610 
2611 	/*
2612 	 * Not allowed access to pseudo exports.
2613 	 */
2614 	if (PSEUDO(*exi)) {
2615 		error = ENOENT;
2616 		VN_RELE(*vpp);
2617 		goto publicfh_done;
2618 	}
2619 
2620 	/*
2621 	 * Do a lookup for the index file. We know the index option doesn't
2622 	 * allow paths through handling in the share command, so mc_dvp will
2623 	 * be the parent for the index file vnode, if its present. Use
2624 	 * temporary pointers to preserve and reuse the vnode pointers of the
2625 	 * original directory in case there's no index file. Note that the
2626 	 * index file is a native path, and should not be interpreted by
2627 	 * the URL parser in rfs_pathname()
2628 	 */
2629 	if (((*exi)->exi_export.ex_flags & EX_INDEX) &&
2630 	    ((*vpp)->v_type == VDIR) && (pathflag == URLPATH)) {
2631 		vnode_t *tvp, *tmc_dvp;	/* temporary vnode pointers */
2632 
2633 		tmc_dvp = mc_dvp;
2634 		mc_dvp = tvp = *vpp;
2635 
2636 		error = rfs_pathname((*exi)->exi_export.ex_index, NULL, vpp,
2637 		    mc_dvp, cr, NATIVEPATH);
2638 
2639 		if (error == ENOENT) {
2640 			*vpp = tvp;
2641 			mc_dvp = tmc_dvp;
2642 			error = 0;
2643 		} else {	/* ok or error other than ENOENT */
2644 			if (tmc_dvp)
2645 				VN_RELE(tmc_dvp);
2646 			if (error)
2647 				goto publicfh_done;
2648 
2649 			/*
2650 			 * Found a valid vp for index "filename". Sanity check
2651 			 * for odd case where a directory is provided as index
2652 			 * option argument and leads us to another filesystem
2653 			 */
2654 
2655 			/* Release the reference on the old exi value */
2656 			ASSERT(*exi != NULL);
2657 			exi_rele(*exi);
2658 
2659 			if (error = nfs_check_vpexi(mc_dvp, *vpp, kcred, exi)) {
2660 				VN_RELE(*vpp);
2661 				goto publicfh_done;
2662 			}
2663 		}
2664 	}
2665 
2666 publicfh_done:
2667 	if (mc_dvp)
2668 		VN_RELE(mc_dvp);
2669 
2670 	return (error);
2671 }
2672 
2673 /*
2674  * Evaluate a multi-component path
2675  */
2676 int
2677 rfs_pathname(
2678 	char *path,			/* pathname to evaluate */
2679 	vnode_t **dirvpp,		/* ret for ptr to parent dir vnode */
2680 	vnode_t **compvpp,		/* ret for ptr to component vnode */
2681 	vnode_t *startdvp,		/* starting vnode */
2682 	cred_t *cr,			/* user's credential */
2683 	int pathflag)			/* flag to identify path, e.g. URL */
2684 {
2685 	char namebuf[TYPICALMAXPATHLEN];
2686 	struct pathname pn;
2687 	int error;
2688 
2689 	/*
2690 	 * If pathname starts with '/', then set startdvp to root.
2691 	 */
2692 	if (*path == '/') {
2693 		while (*path == '/')
2694 			path++;
2695 
2696 		startdvp = rootdir;
2697 	}
2698 
2699 	error = pn_get_buf(path, UIO_SYSSPACE, &pn, namebuf, sizeof (namebuf));
2700 	if (error == 0) {
2701 		/*
2702 		 * Call the URL parser for URL paths to modify the original
2703 		 * string to handle any '%' encoded characters that exist.
2704 		 * Done here to avoid an extra bcopy in the lookup.
2705 		 * We need to be careful about pathlen's. We know that
2706 		 * rfs_pathname() is called with a non-empty path. However,
2707 		 * it could be emptied due to the path simply being all /'s,
2708 		 * which is valid to proceed with the lookup, or due to the
2709 		 * URL parser finding an encoded null character at the
2710 		 * beginning of path which should not proceed with the lookup.
2711 		 */
2712 		if (pn.pn_pathlen != 0 && pathflag == URLPATH) {
2713 			URLparse(pn.pn_path);
2714 			if ((pn.pn_pathlen = strlen(pn.pn_path)) == 0)
2715 				return (ENOENT);
2716 		}
2717 		VN_HOLD(startdvp);
2718 		error = lookuppnvp(&pn, NULL, NO_FOLLOW, dirvpp, compvpp,
2719 		    rootdir, startdvp, cr);
2720 	}
2721 	if (error == ENAMETOOLONG) {
2722 		/*
2723 		 * This thread used a pathname > TYPICALMAXPATHLEN bytes long.
2724 		 */
2725 		if (error = pn_get(path, UIO_SYSSPACE, &pn))
2726 			return (error);
2727 		if (pn.pn_pathlen != 0 && pathflag == URLPATH) {
2728 			URLparse(pn.pn_path);
2729 			if ((pn.pn_pathlen = strlen(pn.pn_path)) == 0) {
2730 				pn_free(&pn);
2731 				return (ENOENT);
2732 			}
2733 		}
2734 		VN_HOLD(startdvp);
2735 		error = lookuppnvp(&pn, NULL, NO_FOLLOW, dirvpp, compvpp,
2736 		    rootdir, startdvp, cr);
2737 		pn_free(&pn);
2738 	}
2739 
2740 	return (error);
2741 }
2742 
2743 /*
2744  * Adapt the multicomponent lookup path depending on the pathtype
2745  */
2746 static int
2747 MCLpath(char **path)
2748 {
2749 	unsigned char c = (unsigned char)**path;
2750 
2751 	/*
2752 	 * If the MCL path is between 0x20 and 0x7E (graphic printable
2753 	 * character of the US-ASCII coded character set), its a URL path,
2754 	 * per RFC 1738.
2755 	 */
2756 	if (c >= 0x20 && c <= 0x7E)
2757 		return (URLPATH);
2758 
2759 	/*
2760 	 * If the first octet of the MCL path is not an ASCII character
2761 	 * then it must be interpreted as a tag value that describes the
2762 	 * format of the remaining octets of the MCL path.
2763 	 *
2764 	 * If the first octet of the MCL path is 0x81 it is a query
2765 	 * for the security info.
2766 	 */
2767 	switch (c) {
2768 	case 0x80:	/* native path, i.e. MCL via mount protocol */
2769 		(*path)++;
2770 		return (NATIVEPATH);
2771 	case 0x81:	/* security query */
2772 		(*path)++;
2773 		return (SECURITY_QUERY);
2774 	default:
2775 		return (-1);
2776 	}
2777 }
2778 
2779 #define	fromhex(c)  ((c >= '0' && c <= '9') ? (c - '0') : \
2780 			((c >= 'A' && c <= 'F') ? (c - 'A' + 10) :\
2781 			((c >= 'a' && c <= 'f') ? (c - 'a' + 10) : 0)))
2782 
2783 /*
2784  * The implementation of URLparse gaurantees that the final string will
2785  * fit in the original one. Replaces '%' occurrences followed by 2 characters
2786  * with its corresponding hexadecimal character.
2787  */
2788 static void
2789 URLparse(char *str)
2790 {
2791 	char *p, *q;
2792 
2793 	p = q = str;
2794 	while (*p) {
2795 		*q = *p;
2796 		if (*p++ == '%') {
2797 			if (*p) {
2798 				*q = fromhex(*p) * 16;
2799 				p++;
2800 				if (*p) {
2801 					*q += fromhex(*p);
2802 					p++;
2803 				}
2804 			}
2805 		}
2806 		q++;
2807 	}
2808 	*q = '\0';
2809 }
2810 
2811 
2812 /*
2813  * Get the export information for the lookup vnode, and verify its
2814  * useable.
2815  */
2816 int
2817 nfs_check_vpexi(vnode_t *mc_dvp, vnode_t *vp, cred_t *cr,
2818     struct exportinfo **exi)
2819 {
2820 	int walk;
2821 	int error = 0;
2822 
2823 	*exi = nfs_vptoexi(mc_dvp, vp, cr, &walk, NULL, FALSE);
2824 	if (*exi == NULL)
2825 		error = EACCES;
2826 	else {
2827 		/*
2828 		 * If nosub is set for this export then
2829 		 * a lookup relative to the public fh
2830 		 * must not terminate below the
2831 		 * exported directory.
2832 		 */
2833 		if ((*exi)->exi_export.ex_flags & EX_NOSUB && walk > 0)
2834 			error = EACCES;
2835 	}
2836 
2837 	return (error);
2838 }
2839