xref: /titanic_50/usr/src/uts/common/fs/nfs/nfs_server.c (revision 9d3d2ed09c8e9ba0b2ba44fdd1dd300b2c3f9e8e)
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, Version 1.0 only
6  * (the "License").  You may not use this file except in compliance
7  * with the License.
8  *
9  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10  * or http://www.opensolaris.org/os/licensing.
11  * See the License for the specific language governing permissions
12  * and limitations under the License.
13  *
14  * When distributing Covered Code, include this CDDL HEADER in each
15  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16  * If applicable, add the following below this CDDL HEADER, with the
17  * fields enclosed by brackets "[]" replaced with your own identifying
18  * information: Portions Copyright [yyyy] [name of copyright owner]
19  *
20  * CDDL HEADER END
21  */
22 /*
23  * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 /*
28  *	Copyright (c) 1983,1984,1985,1986,1987,1988,1989  AT&T.
29  *	All rights reserved.
30  *	Use is subject to license terms.
31  */
32 
33 #pragma ident	"%Z%%M%	%I%	%E% SMI"
34 
35 #include <sys/param.h>
36 #include <sys/types.h>
37 #include <sys/systm.h>
38 #include <sys/cred.h>
39 #include <sys/proc.h>
40 #include <sys/user.h>
41 #include <sys/buf.h>
42 #include <sys/vfs.h>
43 #include <sys/vnode.h>
44 #include <sys/pathname.h>
45 #include <sys/uio.h>
46 #include <sys/file.h>
47 #include <sys/stat.h>
48 #include <sys/errno.h>
49 #include <sys/socket.h>
50 #include <sys/sysmacros.h>
51 #include <sys/siginfo.h>
52 #include <sys/tiuser.h>
53 #include <sys/statvfs.h>
54 #include <sys/stream.h>
55 #include <sys/strsubr.h>
56 #include <sys/stropts.h>
57 #include <sys/timod.h>
58 #include <sys/t_kuser.h>
59 #include <sys/kmem.h>
60 #include <sys/kstat.h>
61 #include <sys/dirent.h>
62 #include <sys/cmn_err.h>
63 #include <sys/debug.h>
64 #include <sys/unistd.h>
65 #include <sys/vtrace.h>
66 #include <sys/mode.h>
67 #include <sys/acl.h>
68 #include <sys/sdt.h>
69 
70 #include <rpc/types.h>
71 #include <rpc/auth.h>
72 #include <rpc/auth_unix.h>
73 #include <rpc/auth_des.h>
74 #include <rpc/svc.h>
75 #include <rpc/xdr.h>
76 
77 #include <nfs/nfs.h>
78 #include <nfs/export.h>
79 #include <nfs/nfssys.h>
80 #include <nfs/nfs_clnt.h>
81 #include <nfs/nfs_acl.h>
82 #include <nfs/nfs_log.h>
83 #include <nfs/lm.h>
84 #include <nfs/nfs_dispatch.h>
85 #include <nfs/nfs4_drc.h>
86 
87 #include <rpcsvc/nfsauth_prot.h>
88 
89 #include <sys/modctl.h>
90 #include <sys/cladm.h>
91 #include <sys/clconf.h>
92 
93 #define	MAXHOST 32
94 const char *kinet_ntop6(uchar_t *, char *, size_t);
95 
96 /*
97  * Module linkage information.
98  */
99 
100 static struct modlmisc modlmisc = {
101 	&mod_miscops, "NFS server module"
102 };
103 
104 static struct modlinkage modlinkage = {
105 	MODREV_1, (void *)&modlmisc, NULL
106 };
107 
108 char _depends_on[] = "misc/klmmod";
109 
110 int
111 _init(void)
112 {
113 	int status;
114 
115 	if ((status = nfs_srvinit()) != 0) {
116 		cmn_err(CE_WARN, "_init: nfs_srvinit failed");
117 		return (status);
118 	}
119 
120 	status = mod_install((struct modlinkage *)&modlinkage);
121 	if (status != 0) {
122 		/*
123 		 * Could not load module, cleanup previous
124 		 * initialization work.
125 		 */
126 		nfs_srvfini();
127 	}
128 
129 	nfs_srv_quiesce_func = nfs_srv_quiesce_all;
130 
131 	return (status);
132 }
133 
134 int
135 _fini()
136 {
137 	return (EBUSY);
138 }
139 
140 int
141 _info(struct modinfo *modinfop)
142 {
143 	return (mod_info(&modlinkage, modinfop));
144 }
145 
146 /*
147  * PUBLICFH_CHECK() checks if the dispatch routine supports
148  * RPC_PUBLICFH_OK, if the filesystem is exported public, and if the
149  * incoming request is using the public filehandle. The check duplicates
150  * the exportmatch() call done in checkexport(), and we should consider
151  * modifying those routines to avoid the duplication. For now, we optimize
152  * by calling exportmatch() only after checking that the dispatch routine
153  * supports RPC_PUBLICFH_OK, and if the filesystem is explicitly exported
154  * public (i.e., not the placeholder).
155  */
156 #define	PUBLICFH_CHECK(disp, exi, fh) \
157 		((disp->dis_flags & RPC_PUBLICFH_OK) && \
158 		((exi->exi_export.ex_flags & EX_PUBLIC) || \
159 		(exi == exi_public && exportmatch(exi_root, \
160 		&fh->fh_fsid, (fid_t *)&fh->fh_xlen))))
161 
162 static void	nfs_srv_shutdown_all(int);
163 static void	rfs4_server_start(int);
164 static void	nullfree(void);
165 static void	rfs_dispatch(struct svc_req *, SVCXPRT *);
166 static void	acl_dispatch(struct svc_req *, SVCXPRT *);
167 static void	common_dispatch(struct svc_req *, SVCXPRT *,
168 		rpcvers_t, rpcvers_t, char *,
169 		struct rpc_disptable *);
170 static	int	checkauth(struct exportinfo *, struct svc_req *, cred_t *, int,
171 			bool_t);
172 static char	*client_name(struct svc_req *req);
173 static char	*client_addr(struct svc_req *req, char *buf);
174 extern	int	sec_svc_getcred(struct svc_req *, cred_t *cr, char **, int *);
175 extern	bool_t	sec_svc_inrootlist(int, caddr_t, int, caddr_t *);
176 
177 #define	NFSLOG_COPY_NETBUF(exi, xprt, nb)	{		\
178 	(nb)->maxlen = (xprt)->xp_rtaddr.maxlen;		\
179 	(nb)->len = (xprt)->xp_rtaddr.len;			\
180 	(nb)->buf = kmem_alloc((nb)->len, KM_SLEEP);		\
181 	bcopy((xprt)->xp_rtaddr.buf, (nb)->buf, (nb)->len);	\
182 	}
183 
184 /*
185  * Public Filehandle common nfs routines
186  */
187 static int	MCLpath(char **);
188 static void	URLparse(char *);
189 
190 /*
191  * NFS callout table.
192  * This table is used by svc_getreq() to dispatch a request with
193  * a given prog/vers pair to an appropriate service provider
194  * dispatch routine.
195  *
196  * NOTE: ordering is relied upon below when resetting the version min/max
197  * for NFS_PROGRAM.  Careful, if this is ever changed.
198  */
199 static SVC_CALLOUT __nfs_sc_clts[] = {
200 	{ NFS_PROGRAM,	   NFS_VERSMIN,	    NFS_VERSMAX,	rfs_dispatch },
201 	{ NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,	acl_dispatch }
202 };
203 
204 static SVC_CALLOUT_TABLE nfs_sct_clts = {
205 	sizeof (__nfs_sc_clts) / sizeof (__nfs_sc_clts[0]), FALSE,
206 	__nfs_sc_clts
207 };
208 
209 static SVC_CALLOUT __nfs_sc_cots[] = {
210 	{ NFS_PROGRAM,	   NFS_VERSMIN,	    NFS_VERSMAX,	rfs_dispatch },
211 	{ NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,	acl_dispatch }
212 };
213 
214 static SVC_CALLOUT_TABLE nfs_sct_cots = {
215 	sizeof (__nfs_sc_cots) / sizeof (__nfs_sc_cots[0]), FALSE, __nfs_sc_cots
216 };
217 
218 static SVC_CALLOUT __nfs_sc_rdma[] = {
219 	{ NFS_PROGRAM,	   NFS_VERSMIN,	    NFS_VERSMAX,	rfs_dispatch },
220 	{ NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,	acl_dispatch }
221 };
222 
223 static SVC_CALLOUT_TABLE nfs_sct_rdma = {
224 	sizeof (__nfs_sc_rdma) / sizeof (__nfs_sc_rdma[0]), FALSE, __nfs_sc_rdma
225 };
226 rpcvers_t nfs_versmin = NFS_VERSMIN_DEFAULT;
227 rpcvers_t nfs_versmax = NFS_VERSMAX_DEFAULT;
228 
229 /*
230  * Used to track the state of the server so that initialization
231  * can be done properly.
232  */
233 typedef enum {
234 	NFS_SERVER_STOPPED,	/* server state destroyed */
235 	NFS_SERVER_STOPPING,	/* server state being destroyed */
236 	NFS_SERVER_RUNNING,
237 	NFS_SERVER_QUIESCED,	/* server state preserved */
238 	NFS_SERVER_OFFLINE	/* server pool offline */
239 } nfs_server_running_t;
240 
241 static nfs_server_running_t nfs_server_upordown;
242 static kmutex_t nfs_server_upordown_lock;
243 static	kcondvar_t nfs_server_upordown_cv;
244 
245 int rfs4_dispatch(struct rpcdisp *, struct svc_req *, SVCXPRT *, char *);
246 
247 /*
248  * RDMA wait variables.
249  */
250 static kcondvar_t rdma_wait_cv;
251 static kmutex_t rdma_wait_mutex;
252 
253 /*
254  * Will be called at the point the server pool is being unregistered
255  * from the pool list. From that point onwards, the pool is waiting
256  * to be drained and as such the server state is stale and pertains
257  * to the old instantiation of the NFS server pool.
258  */
259 void
260 nfs_srv_offline(void)
261 {
262 	mutex_enter(&nfs_server_upordown_lock);
263 	if (nfs_server_upordown == NFS_SERVER_RUNNING) {
264 		nfs_server_upordown = NFS_SERVER_OFFLINE;
265 	}
266 	mutex_exit(&nfs_server_upordown_lock);
267 }
268 
269 /*
270  * Will be called at the point the server pool is being destroyed so
271  * all transports have been closed and no service threads are in
272  * existence.
273  *
274  * If we quiesce the server, we're shutting it down without destroying the
275  * server state. This allows it to warm start subsequently.
276  */
277 void
278 nfs_srv_stop_all(void)
279 {
280 	int quiesce = 0;
281 	nfs_srv_shutdown_all(quiesce);
282 }
283 
284 /*
285  * This alternative shutdown routine can be requested via nfssys()
286  */
287 void
288 nfs_srv_quiesce_all(void)
289 {
290 	int quiesce = 1;
291 	nfs_srv_shutdown_all(quiesce);
292 }
293 
294 static void
295 nfs_srv_shutdown_all(int quiesce) {
296 	mutex_enter(&nfs_server_upordown_lock);
297 	if (quiesce) {
298 		if (nfs_server_upordown == NFS_SERVER_RUNNING ||
299 			nfs_server_upordown == NFS_SERVER_OFFLINE) {
300 			nfs_server_upordown = NFS_SERVER_QUIESCED;
301 			cv_signal(&nfs_server_upordown_cv);
302 			cmn_err(CE_NOTE, "nfs_server: server is now quiesced; "
303 			    "NFSv4 state has been preserved");
304 		}
305 	} else {
306 		if (nfs_server_upordown == NFS_SERVER_OFFLINE) {
307 			nfs_server_upordown = NFS_SERVER_STOPPING;
308 			mutex_exit(&nfs_server_upordown_lock);
309 			rfs4_state_fini();
310 			rfs4_fini_drc(nfs4_drc);
311 			mutex_enter(&nfs_server_upordown_lock);
312 			nfs_server_upordown = NFS_SERVER_STOPPED;
313 			cv_signal(&nfs_server_upordown_cv);
314 		}
315 	}
316 	mutex_exit(&nfs_server_upordown_lock);
317 }
318 
319 static int
320 nfs_srv_set_sc_versions(struct file *fp, SVC_CALLOUT_TABLE **sctpp,
321 			rpcvers_t versmin, rpcvers_t versmax)
322 {
323 	struct strioctl strioc;
324 	struct T_info_ack tinfo;
325 	int		error, retval;
326 
327 	/*
328 	 * Find out what type of transport this is.
329 	 */
330 	strioc.ic_cmd = TI_GETINFO;
331 	strioc.ic_timout = -1;
332 	strioc.ic_len = sizeof (tinfo);
333 	strioc.ic_dp = (char *)&tinfo;
334 	tinfo.PRIM_type = T_INFO_REQ;
335 
336 	error = strioctl(fp->f_vnode, I_STR, (intptr_t)&strioc, 0, K_TO_K,
337 	    CRED(), &retval);
338 	if (error || retval)
339 		return (error);
340 
341 	/*
342 	 * Based on our query of the transport type...
343 	 *
344 	 * Reset the min/max versions based on the caller's request
345 	 * NOTE: This assumes that NFS_PROGRAM is first in the array!!
346 	 * And the second entry is the NFS_ACL_PROGRAM.
347 	 */
348 	switch (tinfo.SERV_type) {
349 	case T_CLTS:
350 		if (versmax == NFS_V4)
351 			return (EINVAL);
352 		__nfs_sc_clts[0].sc_versmin = versmin;
353 		__nfs_sc_clts[0].sc_versmax = versmax;
354 		__nfs_sc_clts[1].sc_versmin = versmin;
355 		__nfs_sc_clts[1].sc_versmax = versmax;
356 		*sctpp = &nfs_sct_clts;
357 		break;
358 	case T_COTS:
359 	case T_COTS_ORD:
360 		__nfs_sc_cots[0].sc_versmin = versmin;
361 		__nfs_sc_cots[0].sc_versmax = versmax;
362 		/* For the NFS_ACL program, check the max version */
363 		if (versmax > NFS_ACL_VERSMAX)
364 			versmax = NFS_ACL_VERSMAX;
365 		__nfs_sc_cots[1].sc_versmin = versmin;
366 		__nfs_sc_cots[1].sc_versmax = versmax;
367 		*sctpp = &nfs_sct_cots;
368 		break;
369 	default:
370 		error = EINVAL;
371 	}
372 
373 	return (error);
374 }
375 
376 /*
377  * NFS Server system call.
378  * Does all of the work of running a NFS server.
379  * uap->fd is the fd of an open transport provider
380  */
381 int
382 nfs_svc(struct nfs_svc_args *arg, model_t model)
383 {
384 	file_t *fp;
385 	SVCMASTERXPRT *xprt;
386 	int error;
387 	int readsize;
388 	char buf[KNC_STRSIZE];
389 	size_t len;
390 	STRUCT_HANDLE(nfs_svc_args, uap);
391 	struct netbuf addrmask;
392 	SVC_CALLOUT_TABLE *sctp = NULL;
393 
394 #ifdef lint
395 	model = model;		/* STRUCT macros don't always refer to it */
396 #endif
397 
398 	STRUCT_SET_HANDLE(uap, model, arg);
399 
400 	/* Check privileges in nfssys() */
401 
402 	if ((fp = getf(STRUCT_FGET(uap, fd))) == NULL)
403 		return (EBADF);
404 
405 	/*
406 	 * Set read buffer size to rsize
407 	 * and add room for RPC headers.
408 	 */
409 	readsize = nfs3tsize() + (RPC_MAXDATASIZE - NFS_MAXDATA);
410 	if (readsize < RPC_MAXDATASIZE)
411 		readsize = RPC_MAXDATASIZE;
412 
413 	error = copyinstr((const char *)STRUCT_FGETP(uap, netid), buf,
414 	    KNC_STRSIZE, &len);
415 	if (error) {
416 		releasef(STRUCT_FGET(uap, fd));
417 		return (error);
418 	}
419 
420 	addrmask.len = STRUCT_FGET(uap, addrmask.len);
421 	addrmask.maxlen = STRUCT_FGET(uap, addrmask.maxlen);
422 	addrmask.buf = kmem_alloc(addrmask.maxlen, KM_SLEEP);
423 	error = copyin(STRUCT_FGETP(uap, addrmask.buf), addrmask.buf,
424 	    addrmask.len);
425 	if (error) {
426 		releasef(STRUCT_FGET(uap, fd));
427 		kmem_free(addrmask.buf, addrmask.maxlen);
428 		return (error);
429 	}
430 
431 	nfs_versmin = STRUCT_FGET(uap, versmin);
432 	nfs_versmax = STRUCT_FGET(uap, versmax);
433 
434 	/* Double check the vers min/max ranges */
435 	if ((nfs_versmin > nfs_versmax) ||
436 		(nfs_versmin < NFS_VERSMIN) ||
437 		(nfs_versmax > NFS_VERSMAX)) {
438 		nfs_versmin = NFS_VERSMIN_DEFAULT;
439 		nfs_versmax = NFS_VERSMAX_DEFAULT;
440 	}
441 
442 	if (error =
443 	    nfs_srv_set_sc_versions(fp, &sctp, nfs_versmin, nfs_versmax)) {
444 		releasef(STRUCT_FGET(uap, fd));
445 		kmem_free(addrmask.buf, addrmask.maxlen);
446 		return (error);
447 	}
448 
449 	/* Initialize nfsv4 server */
450 	if (nfs_versmax == (rpcvers_t)NFS_V4)
451 		rfs4_server_start(STRUCT_FGET(uap, delegation));
452 
453 	/* Create a transport handle. */
454 	error = svc_tli_kcreate(fp, readsize, buf, &addrmask, &xprt,
455 				sctp, NULL, NFS_SVCPOOL_ID, TRUE);
456 
457 	if (error)
458 		kmem_free(addrmask.buf, addrmask.maxlen);
459 
460 	releasef(STRUCT_FGET(uap, fd));
461 
462 	/* save the cluster nodeid */
463 	if (cluster_bootflags & CLUSTER_BOOTED)
464 		lm_global_nlmid = clconf_get_nodeid();
465 
466 	return (error);
467 }
468 
469 static void
470 rfs4_server_start(int nfs4_srv_delegation)
471 {
472 	/*
473 	 * Determine if the server has previously been "started" and
474 	 * if not, do the per instance initialization
475 	 */
476 	mutex_enter(&nfs_server_upordown_lock);
477 
478 	if (nfs_server_upordown != NFS_SERVER_RUNNING) {
479 		/* Do we need to stop and wait on the previous server? */
480 		while (nfs_server_upordown == NFS_SERVER_STOPPING ||
481 			nfs_server_upordown == NFS_SERVER_OFFLINE)
482 			cv_wait(&nfs_server_upordown_cv,
483 			    &nfs_server_upordown_lock);
484 
485 		if (nfs_server_upordown != NFS_SERVER_RUNNING) {
486 			(void) svc_pool_control(NFS_SVCPOOL_ID,
487 			    SVCPSET_UNREGISTER_PROC, (void *)&nfs_srv_offline);
488 			(void) svc_pool_control(NFS_SVCPOOL_ID,
489 			    SVCPSET_SHUTDOWN_PROC, (void *)&nfs_srv_stop_all);
490 
491 			/* is this an nfsd warm start? */
492 			if (nfs_server_upordown == NFS_SERVER_QUIESCED) {
493 				int start_grace;
494 
495 				cmn_err(CE_NOTE, "nfs_server: "
496 				    "server was previously quiesced; "
497 				    "existing NFSv4 state will be re-used");
498 
499 				/*
500 				 * Cluster: this is also the signal that
501 				 * a failover has occurred, so create a new
502 				 * server instance, and start its grace period.
503 				 * We also need to reset all currently
504 				 * active grace periods in case of multiple
505 				 * failovers within the grace duration,
506 				 * to avoid partitioning clients of the same
507 				 * resource into different instances.
508 				 */
509 				if (cluster_bootflags & CLUSTER_BOOTED) {
510 					rfs4_grace_reset_all();
511 					start_grace = 1;
512 					rfs4_servinst_create(start_grace);
513 				}
514 			} else {
515 				rfs4_state_init();
516 				nfs4_drc = rfs4_init_drc(nfs4_drc_max,
517 							nfs4_drc_hash,
518 							nfs4_drc_lifetime);
519 			}
520 
521 			/*
522 			 * Check to see if delegation is to be
523 			 * enabled at the server
524 			 */
525 			if (nfs4_srv_delegation != FALSE)
526 				rfs4_set_deleg_policy(SRV_NORMAL_DELEGATE);
527 
528 			nfs_server_upordown = NFS_SERVER_RUNNING;
529 		}
530 		cv_signal(&nfs_server_upordown_cv);
531 	}
532 	mutex_exit(&nfs_server_upordown_lock);
533 }
534 
535 /*
536  * If RDMA device available,
537  * start RDMA listener.
538  */
539 int
540 rdma_start(struct rdma_svc_args *rsa)
541 {
542 	int error;
543 	rdma_xprt_group_t started_rdma_xprts;
544 
545 	/* Double check the vers min/max ranges */
546 	if ((rsa->nfs_versmin > rsa->nfs_versmax) ||
547 		(rsa->nfs_versmin < NFS_VERSMIN) ||
548 		(rsa->nfs_versmax > NFS_VERSMAX)) {
549 		rsa->nfs_versmin = NFS_VERSMIN_DEFAULT;
550 		rsa->nfs_versmax = NFS_VERSMAX_DEFAULT;
551 	}
552 	nfs_versmin = rsa->nfs_versmin;
553 	nfs_versmax = rsa->nfs_versmax;
554 
555 	/* Set the versions in the callout table */
556 	__nfs_sc_rdma[0].sc_versmin = rsa->nfs_versmin;
557 	__nfs_sc_rdma[0].sc_versmax = rsa->nfs_versmax;
558 	/* For the NFS_ACL program, check the max version */
559 	__nfs_sc_rdma[1].sc_versmin = rsa->nfs_versmin;
560 	if (rsa->nfs_versmax > NFS_ACL_VERSMAX)
561 		__nfs_sc_rdma[1].sc_versmax = NFS_ACL_VERSMAX;
562 	else
563 		__nfs_sc_rdma[1].sc_versmax = rsa->nfs_versmax;
564 
565 	/* Initialize nfsv4 server */
566 	if (rsa->nfs_versmax == (rpcvers_t)NFS_V4)
567 		rfs4_server_start(rsa->delegation);
568 
569 	started_rdma_xprts.rtg_count = 0;
570 	started_rdma_xprts.rtg_listhead = NULL;
571 	started_rdma_xprts.rtg_poolid = rsa->poolid;
572 	error = svc_rdma_kcreate(rsa->netid, &nfs_sct_rdma, rsa->poolid,
573 	    &started_rdma_xprts);
574 
575 	if (error == 0) {
576 		mutex_enter(&rdma_wait_mutex);
577 		if (!cv_wait_sig(&rdma_wait_cv, &rdma_wait_mutex)) {
578 			rdma_stop(started_rdma_xprts);
579 		}
580 		mutex_exit(&rdma_wait_mutex);
581 	}
582 
583 	return (error);
584 }
585 
586 /* ARGSUSED */
587 void
588 rpc_null(caddr_t *argp, caddr_t *resp)
589 {
590 }
591 
592 /* ARGSUSED */
593 static void
594 rfs_error(caddr_t *argp, caddr_t *resp)
595 {
596 	/* return (EOPNOTSUPP); */
597 }
598 
599 static void
600 nullfree(void)
601 {
602 }
603 
604 static char *rfscallnames_v2[] = {
605 	"RFS2_NULL",
606 	"RFS2_GETATTR",
607 	"RFS2_SETATTR",
608 	"RFS2_ROOT",
609 	"RFS2_LOOKUP",
610 	"RFS2_READLINK",
611 	"RFS2_READ",
612 	"RFS2_WRITECACHE",
613 	"RFS2_WRITE",
614 	"RFS2_CREATE",
615 	"RFS2_REMOVE",
616 	"RFS2_RENAME",
617 	"RFS2_LINK",
618 	"RFS2_SYMLINK",
619 	"RFS2_MKDIR",
620 	"RFS2_RMDIR",
621 	"RFS2_READDIR",
622 	"RFS2_STATFS"
623 };
624 
625 static struct rpcdisp rfsdisptab_v2[] = {
626 	/*
627 	 * NFS VERSION 2
628 	 */
629 
630 	/* RFS_NULL = 0 */
631 	{rpc_null,
632 	    xdr_void, NULL_xdrproc_t, 0,
633 	    xdr_void, NULL_xdrproc_t, 0,
634 	    nullfree, RPC_IDEMPOTENT,
635 	    0},
636 
637 	/* RFS_GETATTR = 1 */
638 	{rfs_getattr,
639 	    xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t),
640 	    xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat),
641 	    nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP,
642 	    rfs_getattr_getfh},
643 
644 	/* RFS_SETATTR = 2 */
645 	{rfs_setattr,
646 	    xdr_saargs, NULL_xdrproc_t, sizeof (struct nfssaargs),
647 	    xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat),
648 	    nullfree, RPC_MAPRESP,
649 	    rfs_setattr_getfh},
650 
651 	/* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */
652 	{rfs_error,
653 	    xdr_void, NULL_xdrproc_t, 0,
654 	    xdr_void, NULL_xdrproc_t, 0,
655 	    nullfree, RPC_IDEMPOTENT,
656 	    0},
657 
658 	/* RFS_LOOKUP = 4 */
659 	{rfs_lookup,
660 	    xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs),
661 	    xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres),
662 	    nullfree, RPC_IDEMPOTENT|RPC_MAPRESP|RPC_PUBLICFH_OK,
663 	    rfs_lookup_getfh},
664 
665 	/* RFS_READLINK = 5 */
666 	{rfs_readlink,
667 	    xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t),
668 	    xdr_rdlnres, NULL_xdrproc_t, sizeof (struct nfsrdlnres),
669 	    rfs_rlfree, RPC_IDEMPOTENT,
670 	    rfs_readlink_getfh},
671 
672 	/* RFS_READ = 6 */
673 	{rfs_read,
674 	    xdr_readargs, NULL_xdrproc_t, sizeof (struct nfsreadargs),
675 	    xdr_rdresult, NULL_xdrproc_t, sizeof (struct nfsrdresult),
676 	    rfs_rdfree, RPC_IDEMPOTENT,
677 	    rfs_read_getfh},
678 
679 	/* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */
680 	{rfs_error,
681 	    xdr_void, NULL_xdrproc_t, 0,
682 	    xdr_void, NULL_xdrproc_t, 0,
683 	    nullfree, RPC_IDEMPOTENT,
684 	    0},
685 
686 	/* RFS_WRITE = 8 */
687 	{rfs_write,
688 	    xdr_writeargs, NULL_xdrproc_t, sizeof (struct nfswriteargs),
689 	    xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat),
690 	    nullfree, RPC_MAPRESP,
691 	    rfs_write_getfh},
692 
693 	/* RFS_CREATE = 9 */
694 	{rfs_create,
695 	    xdr_creatargs, NULL_xdrproc_t, sizeof (struct nfscreatargs),
696 	    xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres),
697 	    nullfree, RPC_MAPRESP,
698 	    rfs_create_getfh},
699 
700 	/* RFS_REMOVE = 10 */
701 	{rfs_remove,
702 	    xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs),
703 #ifdef _LITTLE_ENDIAN
704 	    xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
705 #else
706 	    xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
707 #endif
708 	    nullfree, RPC_MAPRESP,
709 	    rfs_remove_getfh},
710 
711 	/* RFS_RENAME = 11 */
712 	{rfs_rename,
713 	    xdr_rnmargs, NULL_xdrproc_t, sizeof (struct nfsrnmargs),
714 #ifdef _LITTLE_ENDIAN
715 	    xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
716 #else
717 	    xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
718 #endif
719 	    nullfree, RPC_MAPRESP,
720 	    rfs_rename_getfh},
721 
722 	/* RFS_LINK = 12 */
723 	{rfs_link,
724 	    xdr_linkargs, NULL_xdrproc_t, sizeof (struct nfslinkargs),
725 #ifdef _LITTLE_ENDIAN
726 	    xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
727 #else
728 	    xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
729 #endif
730 	    nullfree, RPC_MAPRESP,
731 	    rfs_link_getfh},
732 
733 	/* RFS_SYMLINK = 13 */
734 	{rfs_symlink,
735 	    xdr_slargs, NULL_xdrproc_t, sizeof (struct nfsslargs),
736 #ifdef _LITTLE_ENDIAN
737 	    xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
738 #else
739 	    xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
740 #endif
741 	    nullfree, RPC_MAPRESP,
742 	    rfs_symlink_getfh},
743 
744 	/* RFS_MKDIR = 14 */
745 	{rfs_mkdir,
746 	    xdr_creatargs, NULL_xdrproc_t, sizeof (struct nfscreatargs),
747 	    xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres),
748 	    nullfree, RPC_MAPRESP,
749 	    rfs_mkdir_getfh},
750 
751 	/* RFS_RMDIR = 15 */
752 	{rfs_rmdir,
753 	    xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs),
754 #ifdef _LITTLE_ENDIAN
755 	    xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
756 #else
757 	    xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
758 #endif
759 	    nullfree, RPC_MAPRESP,
760 	    rfs_rmdir_getfh},
761 
762 	/* RFS_READDIR = 16 */
763 	{rfs_readdir,
764 	    xdr_rddirargs, NULL_xdrproc_t, sizeof (struct nfsrddirargs),
765 	    xdr_putrddirres, NULL_xdrproc_t, sizeof (struct nfsrddirres),
766 	    rfs_rddirfree, RPC_IDEMPOTENT,
767 	    rfs_readdir_getfh},
768 
769 	/* RFS_STATFS = 17 */
770 	{rfs_statfs,
771 	    xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t),
772 	    xdr_statfs, xdr_faststatfs, sizeof (struct nfsstatfs),
773 	    nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP,
774 	    rfs_statfs_getfh},
775 };
776 
777 static char *rfscallnames_v3[] = {
778 	"RFS3_NULL",
779 	"RFS3_GETATTR",
780 	"RFS3_SETATTR",
781 	"RFS3_LOOKUP",
782 	"RFS3_ACCESS",
783 	"RFS3_READLINK",
784 	"RFS3_READ",
785 	"RFS3_WRITE",
786 	"RFS3_CREATE",
787 	"RFS3_MKDIR",
788 	"RFS3_SYMLINK",
789 	"RFS3_MKNOD",
790 	"RFS3_REMOVE",
791 	"RFS3_RMDIR",
792 	"RFS3_RENAME",
793 	"RFS3_LINK",
794 	"RFS3_READDIR",
795 	"RFS3_READDIRPLUS",
796 	"RFS3_FSSTAT",
797 	"RFS3_FSINFO",
798 	"RFS3_PATHCONF",
799 	"RFS3_COMMIT"
800 };
801 
802 static struct rpcdisp rfsdisptab_v3[] = {
803 	/*
804 	 * NFS VERSION 3
805 	 */
806 
807 	/* RFS_NULL = 0 */
808 	{rpc_null,
809 	    xdr_void, NULL_xdrproc_t, 0,
810 	    xdr_void, NULL_xdrproc_t, 0,
811 	    nullfree, RPC_IDEMPOTENT,
812 	    0},
813 
814 	/* RFS3_GETATTR = 1 */
815 	{rfs3_getattr,
816 	    xdr_nfs_fh3, xdr_fastnfs_fh3, sizeof (GETATTR3args),
817 	    xdr_GETATTR3res, NULL_xdrproc_t, sizeof (GETATTR3res),
818 	    nullfree, (RPC_IDEMPOTENT | RPC_ALLOWANON),
819 	    rfs3_getattr_getfh},
820 
821 	/* RFS3_SETATTR = 2 */
822 	{rfs3_setattr,
823 	    xdr_SETATTR3args, NULL_xdrproc_t, sizeof (SETATTR3args),
824 	    xdr_SETATTR3res, NULL_xdrproc_t, sizeof (SETATTR3res),
825 	    nullfree, 0,
826 	    rfs3_setattr_getfh},
827 
828 	/* RFS3_LOOKUP = 3 */
829 	{rfs3_lookup,
830 	    xdr_diropargs3, NULL_xdrproc_t, sizeof (LOOKUP3args),
831 	    xdr_LOOKUP3res, NULL_xdrproc_t, sizeof (LOOKUP3res),
832 	    nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK),
833 	    rfs3_lookup_getfh},
834 
835 	/* RFS3_ACCESS = 4 */
836 	{rfs3_access,
837 	    xdr_ACCESS3args, NULL_xdrproc_t, sizeof (ACCESS3args),
838 	    xdr_ACCESS3res, NULL_xdrproc_t, sizeof (ACCESS3res),
839 	    nullfree, RPC_IDEMPOTENT,
840 	    rfs3_access_getfh},
841 
842 	/* RFS3_READLINK = 5 */
843 	{rfs3_readlink,
844 	    xdr_nfs_fh3, xdr_fastnfs_fh3, sizeof (READLINK3args),
845 	    xdr_READLINK3res, NULL_xdrproc_t, sizeof (READLINK3res),
846 	    rfs3_readlink_free, RPC_IDEMPOTENT,
847 	    rfs3_readlink_getfh},
848 
849 	/* RFS3_READ = 6 */
850 	{rfs3_read,
851 	    xdr_READ3args, NULL_xdrproc_t, sizeof (READ3args),
852 	    xdr_READ3res, NULL_xdrproc_t, sizeof (READ3res),
853 	    rfs3_read_free, RPC_IDEMPOTENT,
854 	    rfs3_read_getfh},
855 
856 	/* RFS3_WRITE = 7 */
857 	{rfs3_write,
858 	    xdr_WRITE3args, NULL_xdrproc_t, sizeof (WRITE3args),
859 	    xdr_WRITE3res, NULL_xdrproc_t, sizeof (WRITE3res),
860 	    nullfree, 0,
861 	    rfs3_write_getfh},
862 
863 	/* RFS3_CREATE = 8 */
864 	{rfs3_create,
865 	    xdr_CREATE3args, NULL_xdrproc_t, sizeof (CREATE3args),
866 	    xdr_CREATE3res, NULL_xdrproc_t, sizeof (CREATE3res),
867 	    nullfree, 0,
868 	    rfs3_create_getfh},
869 
870 	/* RFS3_MKDIR = 9 */
871 	{rfs3_mkdir,
872 	    xdr_MKDIR3args, NULL_xdrproc_t, sizeof (MKDIR3args),
873 	    xdr_MKDIR3res, NULL_xdrproc_t, sizeof (MKDIR3res),
874 	    nullfree, 0,
875 	    rfs3_mkdir_getfh},
876 
877 	/* RFS3_SYMLINK = 10 */
878 	{rfs3_symlink,
879 	    xdr_SYMLINK3args, NULL_xdrproc_t, sizeof (SYMLINK3args),
880 	    xdr_SYMLINK3res, NULL_xdrproc_t, sizeof (SYMLINK3res),
881 	    nullfree, 0,
882 	    rfs3_symlink_getfh},
883 
884 	/* RFS3_MKNOD = 11 */
885 	{rfs3_mknod,
886 	    xdr_MKNOD3args, NULL_xdrproc_t, sizeof (MKNOD3args),
887 	    xdr_MKNOD3res, NULL_xdrproc_t, sizeof (MKNOD3res),
888 	    nullfree, 0,
889 	    rfs3_mknod_getfh},
890 
891 	/* RFS3_REMOVE = 12 */
892 	{rfs3_remove,
893 	    xdr_diropargs3, NULL_xdrproc_t, sizeof (REMOVE3args),
894 	    xdr_REMOVE3res, NULL_xdrproc_t, sizeof (REMOVE3res),
895 	    nullfree, 0,
896 	    rfs3_remove_getfh},
897 
898 	/* RFS3_RMDIR = 13 */
899 	{rfs3_rmdir,
900 	    xdr_diropargs3, NULL_xdrproc_t, sizeof (RMDIR3args),
901 	    xdr_RMDIR3res, NULL_xdrproc_t, sizeof (RMDIR3res),
902 	    nullfree, 0,
903 	    rfs3_rmdir_getfh},
904 
905 	/* RFS3_RENAME = 14 */
906 	{rfs3_rename,
907 	    xdr_RENAME3args, NULL_xdrproc_t, sizeof (RENAME3args),
908 	    xdr_RENAME3res, NULL_xdrproc_t, sizeof (RENAME3res),
909 	    nullfree, 0,
910 	    rfs3_rename_getfh},
911 
912 	/* RFS3_LINK = 15 */
913 	{rfs3_link,
914 	    xdr_LINK3args, NULL_xdrproc_t, sizeof (LINK3args),
915 	    xdr_LINK3res, NULL_xdrproc_t, sizeof (LINK3res),
916 	    nullfree, 0,
917 	    rfs3_link_getfh},
918 
919 	/* RFS3_READDIR = 16 */
920 	{rfs3_readdir,
921 	    xdr_READDIR3args, NULL_xdrproc_t, sizeof (READDIR3args),
922 	    xdr_READDIR3res, NULL_xdrproc_t, sizeof (READDIR3res),
923 	    rfs3_readdir_free, RPC_IDEMPOTENT,
924 	    rfs3_readdir_getfh},
925 
926 	/* RFS3_READDIRPLUS = 17 */
927 	{rfs3_readdirplus,
928 	    xdr_READDIRPLUS3args, NULL_xdrproc_t, sizeof (READDIRPLUS3args),
929 	    xdr_READDIRPLUS3res, NULL_xdrproc_t, sizeof (READDIRPLUS3res),
930 	    rfs3_readdirplus_free, RPC_AVOIDWORK,
931 	    rfs3_readdirplus_getfh},
932 
933 	/* RFS3_FSSTAT = 18 */
934 	{rfs3_fsstat,
935 	    xdr_nfs_fh3, xdr_fastnfs_fh3, sizeof (FSSTAT3args),
936 	    xdr_FSSTAT3res, NULL_xdrproc_t, sizeof (FSSTAT3res),
937 	    nullfree, RPC_IDEMPOTENT,
938 	    rfs3_fsstat_getfh},
939 
940 	/* RFS3_FSINFO = 19 */
941 	{rfs3_fsinfo,
942 	    xdr_nfs_fh3, xdr_fastnfs_fh3, sizeof (FSINFO3args),
943 	    xdr_FSINFO3res, NULL_xdrproc_t, sizeof (FSINFO3res),
944 	    nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON,
945 	    rfs3_fsinfo_getfh},
946 
947 	/* RFS3_PATHCONF = 20 */
948 	{rfs3_pathconf,
949 	    xdr_nfs_fh3, xdr_fastnfs_fh3, sizeof (PATHCONF3args),
950 	    xdr_PATHCONF3res, NULL_xdrproc_t, sizeof (PATHCONF3res),
951 	    nullfree, RPC_IDEMPOTENT,
952 	    rfs3_pathconf_getfh},
953 
954 	/* RFS3_COMMIT = 21 */
955 	{rfs3_commit,
956 	    xdr_COMMIT3args, NULL_xdrproc_t, sizeof (COMMIT3args),
957 	    xdr_COMMIT3res, NULL_xdrproc_t, sizeof (COMMIT3res),
958 	    nullfree, RPC_IDEMPOTENT,
959 	    rfs3_commit_getfh},
960 };
961 
962 static char *rfscallnames_v4[] = {
963 	"RFS4_NULL",
964 	"RFS4_COMPOUND",
965 	"RFS4_NULL",
966 	"RFS4_NULL",
967 	"RFS4_NULL",
968 	"RFS4_NULL",
969 	"RFS4_NULL",
970 	"RFS4_NULL",
971 	"RFS4_CREATE"
972 };
973 
974 static struct rpcdisp rfsdisptab_v4[] = {
975 	/*
976 	 * NFS VERSION 4
977 	 */
978 
979 	/* RFS_NULL = 0 */
980 	{rpc_null,
981 	    xdr_void, NULL_xdrproc_t, 0,
982 	    xdr_void, NULL_xdrproc_t, 0,
983 	    nullfree, RPC_IDEMPOTENT, 0},
984 
985 	/* RFS4_compound = 1 */
986 	{rfs4_compound,
987 	    xdr_COMPOUND4args_srv, NULL_xdrproc_t, sizeof (COMPOUND4args),
988 	    xdr_COMPOUND4res_srv, NULL_xdrproc_t, sizeof (COMPOUND4res),
989 	    rfs4_compound_free, 0, 0},
990 };
991 
992 union rfs_args {
993 	/*
994 	 * NFS VERSION 2
995 	 */
996 
997 	/* RFS_NULL = 0 */
998 
999 	/* RFS_GETATTR = 1 */
1000 	fhandle_t nfs2_getattr_args;
1001 
1002 	/* RFS_SETATTR = 2 */
1003 	struct nfssaargs nfs2_setattr_args;
1004 
1005 	/* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */
1006 
1007 	/* RFS_LOOKUP = 4 */
1008 	struct nfsdiropargs nfs2_lookup_args;
1009 
1010 	/* RFS_READLINK = 5 */
1011 	fhandle_t nfs2_readlink_args;
1012 
1013 	/* RFS_READ = 6 */
1014 	struct nfsreadargs nfs2_read_args;
1015 
1016 	/* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */
1017 
1018 	/* RFS_WRITE = 8 */
1019 	struct nfswriteargs nfs2_write_args;
1020 
1021 	/* RFS_CREATE = 9 */
1022 	struct nfscreatargs nfs2_create_args;
1023 
1024 	/* RFS_REMOVE = 10 */
1025 	struct nfsdiropargs nfs2_remove_args;
1026 
1027 	/* RFS_RENAME = 11 */
1028 	struct nfsrnmargs nfs2_rename_args;
1029 
1030 	/* RFS_LINK = 12 */
1031 	struct nfslinkargs nfs2_link_args;
1032 
1033 	/* RFS_SYMLINK = 13 */
1034 	struct nfsslargs nfs2_symlink_args;
1035 
1036 	/* RFS_MKDIR = 14 */
1037 	struct nfscreatargs nfs2_mkdir_args;
1038 
1039 	/* RFS_RMDIR = 15 */
1040 	struct nfsdiropargs nfs2_rmdir_args;
1041 
1042 	/* RFS_READDIR = 16 */
1043 	struct nfsrddirargs nfs2_readdir_args;
1044 
1045 	/* RFS_STATFS = 17 */
1046 	fhandle_t nfs2_statfs_args;
1047 
1048 	/*
1049 	 * NFS VERSION 3
1050 	 */
1051 
1052 	/* RFS_NULL = 0 */
1053 
1054 	/* RFS3_GETATTR = 1 */
1055 	GETATTR3args nfs3_getattr_args;
1056 
1057 	/* RFS3_SETATTR = 2 */
1058 	SETATTR3args nfs3_setattr_args;
1059 
1060 	/* RFS3_LOOKUP = 3 */
1061 	LOOKUP3args nfs3_lookup_args;
1062 
1063 	/* RFS3_ACCESS = 4 */
1064 	ACCESS3args nfs3_access_args;
1065 
1066 	/* RFS3_READLINK = 5 */
1067 	READLINK3args nfs3_readlink_args;
1068 
1069 	/* RFS3_READ = 6 */
1070 	READ3args nfs3_read_args;
1071 
1072 	/* RFS3_WRITE = 7 */
1073 	WRITE3args nfs3_write_args;
1074 
1075 	/* RFS3_CREATE = 8 */
1076 	CREATE3args nfs3_create_args;
1077 
1078 	/* RFS3_MKDIR = 9 */
1079 	MKDIR3args nfs3_mkdir_args;
1080 
1081 	/* RFS3_SYMLINK = 10 */
1082 	SYMLINK3args nfs3_symlink_args;
1083 
1084 	/* RFS3_MKNOD = 11 */
1085 	MKNOD3args nfs3_mknod_args;
1086 
1087 	/* RFS3_REMOVE = 12 */
1088 	REMOVE3args nfs3_remove_args;
1089 
1090 	/* RFS3_RMDIR = 13 */
1091 	RMDIR3args nfs3_rmdir_args;
1092 
1093 	/* RFS3_RENAME = 14 */
1094 	RENAME3args nfs3_rename_args;
1095 
1096 	/* RFS3_LINK = 15 */
1097 	LINK3args nfs3_link_args;
1098 
1099 	/* RFS3_READDIR = 16 */
1100 	READDIR3args nfs3_readdir_args;
1101 
1102 	/* RFS3_READDIRPLUS = 17 */
1103 	READDIRPLUS3args nfs3_readdirplus_args;
1104 
1105 	/* RFS3_FSSTAT = 18 */
1106 	FSSTAT3args nfs3_fsstat_args;
1107 
1108 	/* RFS3_FSINFO = 19 */
1109 	FSINFO3args nfs3_fsinfo_args;
1110 
1111 	/* RFS3_PATHCONF = 20 */
1112 	PATHCONF3args nfs3_pathconf_args;
1113 
1114 	/* RFS3_COMMIT = 21 */
1115 	COMMIT3args nfs3_commit_args;
1116 
1117 	/*
1118 	 * NFS VERSION 4
1119 	 */
1120 
1121 	/* RFS_NULL = 0 */
1122 
1123 	/* COMPUND = 1 */
1124 	COMPOUND4args nfs4_compound_args;
1125 };
1126 
1127 union rfs_res {
1128 	/*
1129 	 * NFS VERSION 2
1130 	 */
1131 
1132 	/* RFS_NULL = 0 */
1133 
1134 	/* RFS_GETATTR = 1 */
1135 	struct nfsattrstat nfs2_getattr_res;
1136 
1137 	/* RFS_SETATTR = 2 */
1138 	struct nfsattrstat nfs2_setattr_res;
1139 
1140 	/* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */
1141 
1142 	/* RFS_LOOKUP = 4 */
1143 	struct nfsdiropres nfs2_lookup_res;
1144 
1145 	/* RFS_READLINK = 5 */
1146 	struct nfsrdlnres nfs2_readlink_res;
1147 
1148 	/* RFS_READ = 6 */
1149 	struct nfsrdresult nfs2_read_res;
1150 
1151 	/* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */
1152 
1153 	/* RFS_WRITE = 8 */
1154 	struct nfsattrstat nfs2_write_res;
1155 
1156 	/* RFS_CREATE = 9 */
1157 	struct nfsdiropres nfs2_create_res;
1158 
1159 	/* RFS_REMOVE = 10 */
1160 	enum nfsstat nfs2_remove_res;
1161 
1162 	/* RFS_RENAME = 11 */
1163 	enum nfsstat nfs2_rename_res;
1164 
1165 	/* RFS_LINK = 12 */
1166 	enum nfsstat nfs2_link_res;
1167 
1168 	/* RFS_SYMLINK = 13 */
1169 	enum nfsstat nfs2_symlink_res;
1170 
1171 	/* RFS_MKDIR = 14 */
1172 	struct nfsdiropres nfs2_mkdir_res;
1173 
1174 	/* RFS_RMDIR = 15 */
1175 	enum nfsstat nfs2_rmdir_res;
1176 
1177 	/* RFS_READDIR = 16 */
1178 	struct nfsrddirres nfs2_readdir_res;
1179 
1180 	/* RFS_STATFS = 17 */
1181 	struct nfsstatfs nfs2_statfs_res;
1182 
1183 	/*
1184 	 * NFS VERSION 3
1185 	 */
1186 
1187 	/* RFS_NULL = 0 */
1188 
1189 	/* RFS3_GETATTR = 1 */
1190 	GETATTR3res nfs3_getattr_res;
1191 
1192 	/* RFS3_SETATTR = 2 */
1193 	SETATTR3res nfs3_setattr_res;
1194 
1195 	/* RFS3_LOOKUP = 3 */
1196 	LOOKUP3res nfs3_lookup_res;
1197 
1198 	/* RFS3_ACCESS = 4 */
1199 	ACCESS3res nfs3_access_res;
1200 
1201 	/* RFS3_READLINK = 5 */
1202 	READLINK3res nfs3_readlink_res;
1203 
1204 	/* RFS3_READ = 6 */
1205 	READ3res nfs3_read_res;
1206 
1207 	/* RFS3_WRITE = 7 */
1208 	WRITE3res nfs3_write_res;
1209 
1210 	/* RFS3_CREATE = 8 */
1211 	CREATE3res nfs3_create_res;
1212 
1213 	/* RFS3_MKDIR = 9 */
1214 	MKDIR3res nfs3_mkdir_res;
1215 
1216 	/* RFS3_SYMLINK = 10 */
1217 	SYMLINK3res nfs3_symlink_res;
1218 
1219 	/* RFS3_MKNOD = 11 */
1220 	MKNOD3res nfs3_mknod_res;
1221 
1222 	/* RFS3_REMOVE = 12 */
1223 	REMOVE3res nfs3_remove_res;
1224 
1225 	/* RFS3_RMDIR = 13 */
1226 	RMDIR3res nfs3_rmdir_res;
1227 
1228 	/* RFS3_RENAME = 14 */
1229 	RENAME3res nfs3_rename_res;
1230 
1231 	/* RFS3_LINK = 15 */
1232 	LINK3res nfs3_link_res;
1233 
1234 	/* RFS3_READDIR = 16 */
1235 	READDIR3res nfs3_readdir_res;
1236 
1237 	/* RFS3_READDIRPLUS = 17 */
1238 	READDIRPLUS3res nfs3_readdirplus_res;
1239 
1240 	/* RFS3_FSSTAT = 18 */
1241 	FSSTAT3res nfs3_fsstat_res;
1242 
1243 	/* RFS3_FSINFO = 19 */
1244 	FSINFO3res nfs3_fsinfo_res;
1245 
1246 	/* RFS3_PATHCONF = 20 */
1247 	PATHCONF3res nfs3_pathconf_res;
1248 
1249 	/* RFS3_COMMIT = 21 */
1250 	COMMIT3res nfs3_commit_res;
1251 
1252 	/*
1253 	 * NFS VERSION 4
1254 	 */
1255 
1256 	/* RFS_NULL = 0 */
1257 
1258 	/* RFS4_COMPOUND = 1 */
1259 	COMPOUND4res nfs4_compound_res;
1260 
1261 };
1262 
1263 static struct rpc_disptable rfs_disptable[] = {
1264 	{sizeof (rfsdisptab_v2) / sizeof (rfsdisptab_v2[0]),
1265 	    rfscallnames_v2,
1266 	    &rfsproccnt_v2_ptr, rfsdisptab_v2},
1267 	{sizeof (rfsdisptab_v3) / sizeof (rfsdisptab_v3[0]),
1268 	    rfscallnames_v3,
1269 	    &rfsproccnt_v3_ptr, rfsdisptab_v3},
1270 	{sizeof (rfsdisptab_v4) / sizeof (rfsdisptab_v4[0]),
1271 	    rfscallnames_v4,
1272 	    &rfsproccnt_v4_ptr, rfsdisptab_v4},
1273 };
1274 
1275 /*
1276  * If nfs_portmon is set, then clients are required to use privileged
1277  * ports (ports < IPPORT_RESERVED) in order to get NFS services.
1278  *
1279  * N.B.: this attempt to carry forward the already ill-conceived notion
1280  * of privileged ports for TCP/UDP is really quite ineffectual.  Not only
1281  * is it transport-dependent, it's laughably easy to spoof.  If you're
1282  * really interested in security, you must start with secure RPC instead.
1283  */
1284 static int nfs_portmon = 0;
1285 
1286 #ifdef DEBUG
1287 static int cred_hits = 0;
1288 static int cred_misses = 0;
1289 #endif
1290 
1291 
1292 #ifdef DEBUG
1293 /*
1294  * Debug code to allow disabling of rfs_dispatch() use of
1295  * fastxdrargs() and fastxdrres() calls for testing purposes.
1296  */
1297 static int rfs_no_fast_xdrargs = 0;
1298 static int rfs_no_fast_xdrres = 0;
1299 #endif
1300 
1301 union acl_args {
1302 	/*
1303 	 * ACL VERSION 2
1304 	 */
1305 
1306 	/* ACL2_NULL = 0 */
1307 
1308 	/* ACL2_GETACL = 1 */
1309 	GETACL2args acl2_getacl_args;
1310 
1311 	/* ACL2_SETACL = 2 */
1312 	SETACL2args acl2_setacl_args;
1313 
1314 	/* ACL2_GETATTR = 3 */
1315 	GETATTR2args acl2_getattr_args;
1316 
1317 	/* ACL2_ACCESS = 4 */
1318 	ACCESS2args acl2_access_args;
1319 
1320 	/* ACL2_GETXATTRDIR = 5 */
1321 	GETXATTRDIR2args acl2_getxattrdir_args;
1322 
1323 	/*
1324 	 * ACL VERSION 3
1325 	 */
1326 
1327 	/* ACL3_NULL = 0 */
1328 
1329 	/* ACL3_GETACL = 1 */
1330 	GETACL3args acl3_getacl_args;
1331 
1332 	/* ACL3_SETACL = 2 */
1333 	SETACL3args acl3_setacl;
1334 
1335 	/* ACL3_GETXATTRDIR = 3 */
1336 	GETXATTRDIR3args acl3_getxattrdir_args;
1337 
1338 };
1339 
1340 union acl_res {
1341 	/*
1342 	 * ACL VERSION 2
1343 	 */
1344 
1345 	/* ACL2_NULL = 0 */
1346 
1347 	/* ACL2_GETACL = 1 */
1348 	GETACL2res acl2_getacl_res;
1349 
1350 	/* ACL2_SETACL = 2 */
1351 	SETACL2res acl2_setacl_res;
1352 
1353 	/* ACL2_GETATTR = 3 */
1354 	GETATTR2res acl2_getattr_res;
1355 
1356 	/* ACL2_ACCESS = 4 */
1357 	ACCESS2res acl2_access_res;
1358 
1359 	/* ACL2_GETXATTRDIR = 5 */
1360 	GETXATTRDIR2args acl2_getxattrdir_res;
1361 
1362 	/*
1363 	 * ACL VERSION 3
1364 	 */
1365 
1366 	/* ACL3_NULL = 0 */
1367 
1368 	/* ACL3_GETACL = 1 */
1369 	GETACL3res acl3_getacl_res;
1370 
1371 	/* ACL3_SETACL = 2 */
1372 	SETACL3res acl3_setacl_res;
1373 
1374 	/* ACL3_GETXATTRDIR = 3 */
1375 	GETXATTRDIR3res acl3_getxattrdir_res;
1376 
1377 };
1378 
1379 static bool_t
1380 auth_tooweak(struct svc_req *req, char *res)
1381 {
1382 
1383 	if (req->rq_vers == NFS_VERSION && req->rq_proc == RFS_LOOKUP) {
1384 		struct nfsdiropres *dr = (struct nfsdiropres *)res;
1385 		if (dr->dr_status == WNFSERR_CLNT_FLAVOR)
1386 			return (TRUE);
1387 	} else if (req->rq_vers == NFS_V3 && req->rq_proc == NFSPROC3_LOOKUP) {
1388 		LOOKUP3res *resp = (LOOKUP3res *)res;
1389 		if (resp->status == WNFSERR_CLNT_FLAVOR)
1390 			return (TRUE);
1391 	}
1392 	return (FALSE);
1393 }
1394 
1395 
1396 static void
1397 common_dispatch(struct svc_req *req, SVCXPRT *xprt, rpcvers_t min_vers,
1398 		rpcvers_t max_vers, char *pgmname,
1399 		struct rpc_disptable *disptable)
1400 {
1401 	int which;
1402 	rpcvers_t vers;
1403 	char *args;
1404 	union {
1405 			union rfs_args ra;
1406 			union acl_args aa;
1407 		} args_buf;
1408 	char *res;
1409 	union {
1410 			union rfs_res rr;
1411 			union acl_res ar;
1412 		} res_buf;
1413 	struct rpcdisp *disp = NULL;
1414 	int dis_flags = 0;
1415 	cred_t *cr;
1416 	int error = 0;
1417 	int anon_ok;
1418 	struct exportinfo *exi = NULL;
1419 	unsigned int nfslog_rec_id;
1420 	int dupstat;
1421 	struct dupreq *dr;
1422 	int authres;
1423 	bool_t publicfh_ok = FALSE;
1424 	enum_t auth_flavor;
1425 	bool_t dupcached = FALSE;
1426 	struct netbuf	nb;
1427 	bool_t logging_enabled = FALSE;
1428 	struct exportinfo *nfslog_exi = NULL;
1429 
1430 	vers = req->rq_vers;
1431 
1432 	if (vers < min_vers || vers > max_vers) {
1433 		svcerr_progvers(req->rq_xprt, min_vers, max_vers);
1434 		error++;
1435 		cmn_err(CE_NOTE, "%s: bad version number %u", pgmname, vers);
1436 		goto done;
1437 	}
1438 	vers -= min_vers;
1439 
1440 	which = req->rq_proc;
1441 	if (which < 0 || which >= disptable[(int)vers].dis_nprocs) {
1442 		svcerr_noproc(req->rq_xprt);
1443 		error++;
1444 		goto done;
1445 	}
1446 
1447 	(*(disptable[(int)vers].dis_proccntp))[which].value.ui64++;
1448 
1449 	disp = &disptable[(int)vers].dis_table[which];
1450 
1451 	auth_flavor = req->rq_cred.oa_flavor;
1452 	/*
1453 	 * Deserialize into the args struct.
1454 	 */
1455 
1456 	args = (char *)&args_buf;
1457 
1458 #ifdef DEBUG
1459 	if (rfs_no_fast_xdrargs || (auth_flavor == RPCSEC_GSS) ||
1460 	    disp->dis_fastxdrargs == NULL_xdrproc_t ||
1461 	    !SVC_GETARGS(xprt, disp->dis_fastxdrargs, (char *)&args)) {
1462 #else
1463 	if ((auth_flavor == RPCSEC_GSS) ||
1464 	    disp->dis_fastxdrargs == NULL_xdrproc_t ||
1465 	    !SVC_GETARGS(xprt, disp->dis_fastxdrargs, (char *)&args)) {
1466 #endif
1467 		bzero(args, disp->dis_argsz);
1468 		if (!SVC_GETARGS(xprt, disp->dis_xdrargs, args)) {
1469 			svcerr_decode(xprt);
1470 			error++;
1471 			cmn_err(CE_NOTE, "%s: bad getargs for %u/%d",
1472 			    pgmname, vers + min_vers, which);
1473 			goto done;
1474 		}
1475 	}
1476 
1477 	/*
1478 	 * If Version 4 use that specific dispatch function.
1479 	 */
1480 	if (req->rq_vers == 4) {
1481 		error += rfs4_dispatch(disp, req, xprt, args);
1482 		goto done;
1483 	}
1484 
1485 	dis_flags = disp->dis_flags;
1486 
1487 	/*
1488 	 * Find export information and check authentication,
1489 	 * setting the credential if everything is ok.
1490 	 */
1491 	if (disp->dis_getfh != NULL) {
1492 		fhandle_t *fh;
1493 
1494 		fh = (*disp->dis_getfh)(args);
1495 
1496 		/*
1497 		 * Fix for bug 1038302 - corbin
1498 		 * There is a problem here if anonymous access is
1499 		 * disallowed.  If the current request is part of the
1500 		 * client's mount process for the requested filesystem,
1501 		 * then it will carry root (uid 0) credentials on it, and
1502 		 * will be denied by checkauth if that client does not
1503 		 * have explicit root=0 permission.  This will cause the
1504 		 * client's mount operation to fail.  As a work-around,
1505 		 * we check here to see if the request is a getattr or
1506 		 * statfs operation on the exported vnode itself, and
1507 		 * pass a flag to checkauth with the result of this test.
1508 		 *
1509 		 * The filehandle refers to the mountpoint itself if
1510 		 * the fh_data and fh_xdata portions of the filehandle
1511 		 * are equal.
1512 		 *
1513 		 * Added anon_ok argument to checkauth().
1514 		 */
1515 
1516 		if ((dis_flags & RPC_ALLOWANON) &&
1517 		    EQFID((fid_t *)&fh->fh_len, (fid_t *)&fh->fh_xlen))
1518 			anon_ok = 1;
1519 		else
1520 			anon_ok = 0;
1521 
1522 		cr = xprt->xp_cred;
1523 		ASSERT(cr != NULL);
1524 #ifdef DEBUG
1525 		if (crgetref(cr) != 1) {
1526 			crfree(cr);
1527 			cr = crget();
1528 			xprt->xp_cred = cr;
1529 			cred_misses++;
1530 		} else
1531 			cred_hits++;
1532 #else
1533 		if (crgetref(cr) != 1) {
1534 			crfree(cr);
1535 			cr = crget();
1536 			xprt->xp_cred = cr;
1537 		}
1538 #endif
1539 
1540 		exi = checkexport(&fh->fh_fsid, (fid_t *)&fh->fh_xlen);
1541 
1542 		if (exi != NULL) {
1543 			publicfh_ok = PUBLICFH_CHECK(disp, exi, fh);
1544 
1545 			/*
1546 			 * Don't allow non-V4 clients access
1547 			 * to pseudo exports
1548 			 */
1549 			if (PSEUDO(exi)) {
1550 				svcerr_weakauth(xprt);
1551 				error++;
1552 				goto done;
1553 			}
1554 
1555 			authres = checkauth(exi, req, cr, anon_ok, publicfh_ok);
1556 			/*
1557 			 * authres >  0: authentication OK - proceed
1558 			 * authres == 0: authentication weak - return error
1559 			 * authres <  0: authentication timeout - drop
1560 			 */
1561 			if (authres <= 0) {
1562 				if (authres == 0) {
1563 					svcerr_weakauth(xprt);
1564 					error++;
1565 				}
1566 				goto done;
1567 			}
1568 		}
1569 	} else
1570 		cr = NULL;
1571 
1572 	if ((dis_flags & RPC_MAPRESP) && (auth_flavor != RPCSEC_GSS)) {
1573 		res = (char *)SVC_GETRES(xprt, disp->dis_ressz);
1574 		if (res == NULL)
1575 			res = (char *)&res_buf;
1576 	} else
1577 		res = (char *)&res_buf;
1578 
1579 	if (!(dis_flags & RPC_IDEMPOTENT)) {
1580 		dupstat = SVC_DUP_EXT(xprt, req, res, disp->dis_ressz, &dr,
1581 				&dupcached);
1582 
1583 		switch (dupstat) {
1584 		case DUP_ERROR:
1585 			svcerr_systemerr(xprt);
1586 			error++;
1587 			goto done;
1588 			/* NOTREACHED */
1589 		case DUP_INPROGRESS:
1590 			if (res != (char *)&res_buf)
1591 				SVC_FREERES(xprt);
1592 			error++;
1593 			goto done;
1594 			/* NOTREACHED */
1595 		case DUP_NEW:
1596 		case DUP_DROP:
1597 			curthread->t_flag |= T_DONTPEND;
1598 
1599 			(*disp->dis_proc)(args, res, exi, req, cr);
1600 
1601 			curthread->t_flag &= ~T_DONTPEND;
1602 			if (curthread->t_flag & T_WOULDBLOCK) {
1603 				curthread->t_flag &= ~T_WOULDBLOCK;
1604 				SVC_DUPDONE_EXT(xprt, dr, res, NULL,
1605 					disp->dis_ressz, DUP_DROP);
1606 				if (res != (char *)&res_buf)
1607 					SVC_FREERES(xprt);
1608 				error++;
1609 				goto done;
1610 			}
1611 			if (dis_flags & RPC_AVOIDWORK) {
1612 				SVC_DUPDONE_EXT(xprt, dr, res, NULL,
1613 					disp->dis_ressz, DUP_DROP);
1614 			} else {
1615 				SVC_DUPDONE_EXT(xprt, dr, res,
1616 					disp->dis_resfree == nullfree ? NULL :
1617 					disp->dis_resfree,
1618 					disp->dis_ressz, DUP_DONE);
1619 				dupcached = TRUE;
1620 			}
1621 			break;
1622 		case DUP_DONE:
1623 			break;
1624 		}
1625 
1626 	} else {
1627 		curthread->t_flag |= T_DONTPEND;
1628 
1629 		(*disp->dis_proc)(args, res, exi, req, cr);
1630 
1631 		curthread->t_flag &= ~T_DONTPEND;
1632 		if (curthread->t_flag & T_WOULDBLOCK) {
1633 			curthread->t_flag &= ~T_WOULDBLOCK;
1634 			if (res != (char *)&res_buf)
1635 				SVC_FREERES(xprt);
1636 			error++;
1637 			goto done;
1638 		}
1639 	}
1640 
1641 	if (auth_tooweak(req, res)) {
1642 		svcerr_weakauth(xprt);
1643 		error++;
1644 		goto done;
1645 	}
1646 
1647 	/*
1648 	 * Check to see if logging has been enabled on the server.
1649 	 * If so, then obtain the export info struct to be used for
1650 	 * the later writing of the log record.  This is done for
1651 	 * the case that a lookup is done across a non-logged public
1652 	 * file system.
1653 	 */
1654 	if (nfslog_buffer_list != NULL) {
1655 		nfslog_exi = nfslog_get_exi(exi, req, res, &nfslog_rec_id);
1656 		/*
1657 		 * Is logging enabled?
1658 		 */
1659 		logging_enabled = (nfslog_exi != NULL);
1660 
1661 		/*
1662 		 * Copy the netbuf for logging purposes, before it is
1663 		 * freed by svc_sendreply().
1664 		 */
1665 		if (logging_enabled) {
1666 			NFSLOG_COPY_NETBUF(nfslog_exi, xprt, &nb);
1667 			/*
1668 			 * If RPC_MAPRESP flag set (i.e. in V2 ops) the
1669 			 * res gets copied directly into the mbuf and
1670 			 * may be freed soon after the sendreply. So we
1671 			 * must copy it here to a safe place...
1672 			 */
1673 			if (res != (char *)&res_buf) {
1674 				bcopy(res, (char *)&res_buf, disp->dis_ressz);
1675 			}
1676 		}
1677 	}
1678 
1679 	/*
1680 	 * Serialize and send results struct
1681 	 */
1682 #ifdef DEBUG
1683 	if (rfs_no_fast_xdrres == 0 && res != (char *)&res_buf) {
1684 #else
1685 	if (res != (char *)&res_buf) {
1686 #endif
1687 		if (!svc_sendreply(xprt, disp->dis_fastxdrres, res)) {
1688 			cmn_err(CE_NOTE, "%s: bad sendreply", pgmname);
1689 			error++;
1690 		}
1691 	} else {
1692 		if (!svc_sendreply(xprt, disp->dis_xdrres, res)) {
1693 			cmn_err(CE_NOTE, "%s: bad sendreply", pgmname);
1694 			error++;
1695 		}
1696 	}
1697 
1698 	/*
1699 	 * Log if needed
1700 	 */
1701 	if (logging_enabled) {
1702 		nfslog_write_record(nfslog_exi, req, args, (char *)&res_buf,
1703 			cr, &nb, nfslog_rec_id, NFSLOG_ONE_BUFFER);
1704 		exi_rele(nfslog_exi);
1705 		kmem_free((&nb)->buf, (&nb)->len);
1706 	}
1707 
1708 	/*
1709 	 * Free results struct. With the addition of NFS V4 we can
1710 	 * have non-idempotent procedures with functions.
1711 	 */
1712 	if (disp->dis_resfree != nullfree && dupcached == FALSE) {
1713 		(*disp->dis_resfree)(res);
1714 	}
1715 
1716 done:
1717 	/*
1718 	 * Free arguments struct
1719 	 */
1720 	if (disp) {
1721 		if (!SVC_FREEARGS(xprt, disp->dis_xdrargs, args)) {
1722 			cmn_err(CE_NOTE, "%s: bad freeargs", pgmname);
1723 			error++;
1724 		}
1725 	} else {
1726 		if (!SVC_FREEARGS(xprt, (xdrproc_t)0, (caddr_t)0)) {
1727 			cmn_err(CE_NOTE, "%s: bad freeargs", pgmname);
1728 			error++;
1729 		}
1730 	}
1731 
1732 	if (exi != NULL)
1733 		exi_rele(exi);
1734 
1735 	global_svstat_ptr[req->rq_vers][NFS_BADCALLS].value.ui64 += error;
1736 
1737 	global_svstat_ptr[req->rq_vers][NFS_CALLS].value.ui64++;
1738 }
1739 
1740 static void
1741 rfs_dispatch(struct svc_req *req, SVCXPRT *xprt)
1742 {
1743 	common_dispatch(req, xprt, NFS_VERSMIN, NFS_VERSMAX,
1744 		"nfs_server", rfs_disptable);
1745 }
1746 
1747 static char *aclcallnames_v2[] = {
1748 	"ACL2_NULL",
1749 	"ACL2_GETACL",
1750 	"ACL2_SETACL",
1751 	"ACL2_GETATTR",
1752 	"ACL2_ACCESS",
1753 	"ACL2_GETXATTRDIR"
1754 };
1755 
1756 static struct rpcdisp acldisptab_v2[] = {
1757 	/*
1758 	 * ACL VERSION 2
1759 	 */
1760 
1761 	/* ACL2_NULL = 0 */
1762 	{rpc_null,
1763 	    xdr_void, NULL_xdrproc_t, 0,
1764 	    xdr_void, NULL_xdrproc_t, 0,
1765 	    nullfree, RPC_IDEMPOTENT,
1766 	    0},
1767 
1768 	/* ACL2_GETACL = 1 */
1769 	{acl2_getacl,
1770 	    xdr_GETACL2args, xdr_fastGETACL2args, sizeof (GETACL2args),
1771 	    xdr_GETACL2res, NULL_xdrproc_t, sizeof (GETACL2res),
1772 	    acl2_getacl_free, RPC_IDEMPOTENT,
1773 	    acl2_getacl_getfh},
1774 
1775 	/* ACL2_SETACL = 2 */
1776 	{acl2_setacl,
1777 	    xdr_SETACL2args, NULL_xdrproc_t, sizeof (SETACL2args),
1778 #ifdef _LITTLE_ENDIAN
1779 	    xdr_SETACL2res, xdr_fastSETACL2res, sizeof (SETACL2res),
1780 #else
1781 	    xdr_SETACL2res, NULL_xdrproc_t, sizeof (SETACL2res),
1782 #endif
1783 	    nullfree, RPC_MAPRESP,
1784 	    acl2_setacl_getfh},
1785 
1786 	/* ACL2_GETATTR = 3 */
1787 	{acl2_getattr,
1788 	    xdr_GETATTR2args, xdr_fastGETATTR2args, sizeof (GETATTR2args),
1789 #ifdef _LITTLE_ENDIAN
1790 	    xdr_GETATTR2res, xdr_fastGETATTR2res, sizeof (GETATTR2res),
1791 #else
1792 	    xdr_GETATTR2res, NULL_xdrproc_t, sizeof (GETATTR2res),
1793 #endif
1794 	    nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP,
1795 	    acl2_getattr_getfh},
1796 
1797 	/* ACL2_ACCESS = 4 */
1798 	{acl2_access,
1799 	    xdr_ACCESS2args, xdr_fastACCESS2args, sizeof (ACCESS2args),
1800 #ifdef _LITTLE_ENDIAN
1801 	    xdr_ACCESS2res, xdr_fastACCESS2res, sizeof (ACCESS2res),
1802 #else
1803 	    xdr_ACCESS2res, NULL_xdrproc_t, sizeof (ACCESS2res),
1804 #endif
1805 	    nullfree, RPC_IDEMPOTENT|RPC_MAPRESP,
1806 	    acl2_access_getfh},
1807 
1808 	/* ACL2_GETXATTRDIR = 5 */
1809 	{acl2_getxattrdir,
1810 	    xdr_GETXATTRDIR2args, NULL_xdrproc_t, sizeof (GETXATTRDIR2args),
1811 	    xdr_GETXATTRDIR2res, NULL_xdrproc_t, sizeof (GETXATTRDIR2res),
1812 	    nullfree, RPC_IDEMPOTENT,
1813 	    acl2_getxattrdir_getfh},
1814 };
1815 
1816 static char *aclcallnames_v3[] = {
1817 	"ACL3_NULL",
1818 	"ACL3_GETACL",
1819 	"ACL3_SETACL",
1820 	"ACL3_GETXATTRDIR"
1821 };
1822 
1823 static struct rpcdisp acldisptab_v3[] = {
1824 	/*
1825 	 * ACL VERSION 3
1826 	 */
1827 
1828 	/* ACL3_NULL = 0 */
1829 	{rpc_null,
1830 	    xdr_void, NULL_xdrproc_t, 0,
1831 	    xdr_void, NULL_xdrproc_t, 0,
1832 	    nullfree, RPC_IDEMPOTENT,
1833 	    0},
1834 
1835 	/* ACL3_GETACL = 1 */
1836 	{acl3_getacl,
1837 	    xdr_GETACL3args, NULL_xdrproc_t, sizeof (GETACL3args),
1838 	    xdr_GETACL3res, NULL_xdrproc_t, sizeof (GETACL3res),
1839 	    acl3_getacl_free, RPC_IDEMPOTENT,
1840 	    acl3_getacl_getfh},
1841 
1842 	/* ACL3_SETACL = 2 */
1843 	{acl3_setacl,
1844 	    xdr_SETACL3args, NULL_xdrproc_t, sizeof (SETACL3args),
1845 	    xdr_SETACL3res, NULL_xdrproc_t, sizeof (SETACL3res),
1846 	    nullfree, 0,
1847 	    acl3_setacl_getfh},
1848 
1849 	/* ACL3_GETXATTRDIR = 3 */
1850 	{acl3_getxattrdir,
1851 	    xdr_GETXATTRDIR3args, NULL_xdrproc_t, sizeof (GETXATTRDIR3args),
1852 	    xdr_GETXATTRDIR3res, NULL_xdrproc_t, sizeof (GETXATTRDIR3res),
1853 	    nullfree, RPC_IDEMPOTENT,
1854 	    acl3_getxattrdir_getfh},
1855 };
1856 
1857 static struct rpc_disptable acl_disptable[] = {
1858 	{sizeof (acldisptab_v2) / sizeof (acldisptab_v2[0]),
1859 		aclcallnames_v2,
1860 		&aclproccnt_v2_ptr, acldisptab_v2},
1861 	{sizeof (acldisptab_v3) / sizeof (acldisptab_v3[0]),
1862 		aclcallnames_v3,
1863 		&aclproccnt_v3_ptr, acldisptab_v3},
1864 };
1865 
1866 static void
1867 acl_dispatch(struct svc_req *req, SVCXPRT *xprt)
1868 {
1869 	common_dispatch(req, xprt, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,
1870 		"acl_server", acl_disptable);
1871 }
1872 
1873 int
1874 checkwin(int flavor, int window, struct svc_req *req)
1875 {
1876 	struct authdes_cred *adc;
1877 
1878 	switch (flavor) {
1879 	case AUTH_DES:
1880 		adc = (struct authdes_cred *)req->rq_clntcred;
1881 		if (adc->adc_fullname.window > window)
1882 			return (0);
1883 		break;
1884 
1885 	default:
1886 		break;
1887 	}
1888 	return (1);
1889 }
1890 
1891 
1892 /*
1893  * checkauth() will check the access permission against the export
1894  * information.  Then map root uid/gid to appropriate uid/gid.
1895  *
1896  * This routine is used by NFS V3 and V2 code.
1897  */
1898 static int
1899 checkauth(struct exportinfo *exi, struct svc_req *req, cred_t *cr, int anon_ok,
1900     bool_t publicfh_ok)
1901 {
1902 	int i, nfsflavor, rpcflavor, stat, access;
1903 	struct secinfo *secp;
1904 	caddr_t principal;
1905 	char buf[INET6_ADDRSTRLEN]; /* to hold both IPv4 and IPv6 addr */
1906 	int anon_res = 0;
1907 
1908 	/*
1909 	 *	Check for privileged port number
1910 	 *	N.B.:  this assumes that we know the format of a netbuf.
1911 	 */
1912 	if (nfs_portmon) {
1913 		struct sockaddr *ca;
1914 		ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
1915 
1916 		if (ca == NULL)
1917 			return (0);
1918 
1919 		if ((ca->sa_family == AF_INET &&
1920 		    ntohs(((struct sockaddr_in *)ca)->sin_port) >=
1921 		    IPPORT_RESERVED) ||
1922 		    (ca->sa_family == AF_INET6 &&
1923 		    ntohs(((struct sockaddr_in6 *)ca)->sin6_port) >=
1924 		    IPPORT_RESERVED)) {
1925 			cmn_err(CE_NOTE,
1926 			    "nfs_server: client %s%ssent NFS request from "
1927 			    "unprivileged port",
1928 			    client_name(req), client_addr(req, buf));
1929 			return (0);
1930 		}
1931 	}
1932 
1933 	/*
1934 	 *  return 1 on success or 0 on failure
1935 	 */
1936 	stat = sec_svc_getcred(req, cr, &principal, &nfsflavor);
1937 
1938 	/*
1939 	 * A failed AUTH_UNIX svc_get_cred() implies we couldn't set
1940 	 * the credentials; below we map that to anonymous.
1941 	 */
1942 	if (!stat && nfsflavor != AUTH_UNIX) {
1943 		cmn_err(CE_NOTE,
1944 		    "nfs_server: couldn't get unix cred for %s",
1945 		    client_name(req));
1946 		return (0);
1947 	}
1948 
1949 	/*
1950 	 * Short circuit checkauth() on operations that support the
1951 	 * public filehandle, and if the request for that operation
1952 	 * is using the public filehandle. Note that we must call
1953 	 * sec_svc_getcred() first so that xp_cookie is set to the
1954 	 * right value. Normally xp_cookie is just the RPC flavor
1955 	 * of the the request, but in the case of RPCSEC_GSS it
1956 	 * could be a pseudo flavor.
1957 	 */
1958 	if (publicfh_ok)
1959 		return (1);
1960 
1961 	rpcflavor = req->rq_cred.oa_flavor;
1962 	/*
1963 	 * Check if the auth flavor is valid for this export
1964 	 */
1965 	access = nfsauth_access(exi, req);
1966 	if (access & NFSAUTH_DROP)
1967 		return (-1);	/* drop the request */
1968 
1969 	if (access & NFSAUTH_DENIED) {
1970 		/*
1971 		 * If anon_ok == 1 and we got NFSAUTH_DENIED, it was
1972 		 * probably due to the flavor not matching during the
1973 		 * the mount attempt. So map the flavor to AUTH_NONE
1974 		 * so that the credentials get mapped to the anonymous
1975 		 * user.
1976 		 */
1977 		if (anon_ok == 1)
1978 			rpcflavor = AUTH_NONE;
1979 		else
1980 			return (0);	/* deny access */
1981 
1982 	} else if (access & NFSAUTH_MAPNONE) {
1983 		/*
1984 		 * Access was granted even though the flavor mismatched
1985 		 * because AUTH_NONE was one of the exported flavors.
1986 		 */
1987 		rpcflavor = AUTH_NONE;
1988 
1989 	} else if (access & NFSAUTH_WRONGSEC) {
1990 		/*
1991 		 * NFSAUTH_WRONGSEC is used for NFSv4. Since V2/V3 already
1992 		 * negotiates the security flavor thru MOUNT protocol, the
1993 		 * only way it can get NFSAUTH_WRONGSEC here is from
1994 		 * NFS_ACL for V4. This could be for a limited view, so
1995 		 * map it to RO access. V4 lookup/readdir will take care
1996 		 * of the limited view portion.
1997 		 */
1998 		access |= NFSAUTH_RO;
1999 		access &= ~NFSAUTH_WRONGSEC;
2000 	}
2001 
2002 	switch (rpcflavor) {
2003 	case AUTH_NONE:
2004 		anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2005 				exi->exi_export.ex_anon);
2006 		(void) crsetgroups(cr, 0, NULL);
2007 		break;
2008 
2009 	case AUTH_UNIX:
2010 		if (!stat || crgetuid(cr) == 0 && !(access & NFSAUTH_ROOT)) {
2011 			anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2012 					exi->exi_export.ex_anon);
2013 			(void) crsetgroups(cr, 0, NULL);
2014 		}
2015 		break;
2016 
2017 	case AUTH_DES:
2018 	case RPCSEC_GSS:
2019 		/*
2020 		 *  Find the secinfo structure.  We should be able
2021 		 *  to find it by the time we reach here.
2022 		 *  nfsauth_access() has done the checking.
2023 		 */
2024 		secp = NULL;
2025 		for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
2026 			if (exi->exi_export.ex_secinfo[i].s_secinfo.sc_nfsnum ==
2027 			    nfsflavor) {
2028 				secp = &exi->exi_export.ex_secinfo[i];
2029 				break;
2030 			}
2031 		}
2032 
2033 		if (!secp) {
2034 			cmn_err(CE_NOTE, "nfs_server: client %s%shad "
2035 			    "no secinfo data for flavor %d",
2036 			    client_name(req), client_addr(req, buf),
2037 			    nfsflavor);
2038 			return (0);
2039 		}
2040 
2041 		if (!checkwin(rpcflavor, secp->s_window, req)) {
2042 			cmn_err(CE_NOTE,
2043 			    "nfs_server: client %s%sused invalid "
2044 			    "auth window value",
2045 			    client_name(req), client_addr(req, buf));
2046 			return (0);
2047 		}
2048 
2049 		/*
2050 		 * Map root principals listed in the share's root= list to root,
2051 		 * and map any others principals that were mapped to root by RPC
2052 		 * to anon.
2053 		 */
2054 		if (principal && sec_svc_inrootlist(rpcflavor, principal,
2055 			secp->s_rootcnt, secp->s_rootnames)) {
2056 			if (crgetuid(cr) == 0)
2057 				return (1);
2058 
2059 			(void) crsetugid(cr, 0, 0);
2060 
2061 			/*
2062 			 * NOTE: If and when kernel-land privilege tracing is
2063 			 * added this may have to be replaced with code that
2064 			 * retrieves root's supplementary groups (e.g., using
2065 			 * kgss_get_group_info().  In the meantime principals
2066 			 * mapped to uid 0 get all privileges, so setting cr's
2067 			 * supplementary groups for them does nothing.
2068 			 */
2069 			(void) crsetgroups(cr, 0, NULL);
2070 
2071 			return (1);
2072 		}
2073 
2074 		/*
2075 		 * Not a root princ, or not in root list, map UID 0/nobody to
2076 		 * the anon ID for the share.  (RPC sets cr's UIDs and GIDs to
2077 		 * UID_NOBODY and GID_NOBODY, respectively.)
2078 		 */
2079 		if (crgetuid(cr) != 0 &&
2080 		    (crgetuid(cr) != UID_NOBODY || crgetgid(cr) != GID_NOBODY))
2081 			return (1);
2082 
2083 		anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2084 			exi->exi_export.ex_anon);
2085 		(void) crsetgroups(cr, 0, NULL);
2086 		break;
2087 	default:
2088 		return (0);
2089 	} /* switch on rpcflavor */
2090 
2091 	/*
2092 	 * Even if anon access is disallowed via ex_anon == -1, we allow
2093 	 * this access if anon_ok is set.  So set creds to the default
2094 	 * "nobody" id.
2095 	 */
2096 	if (anon_res != 0) {
2097 		if (anon_ok == 0) {
2098 			cmn_err(CE_NOTE,
2099 			    "nfs_server: client %s%ssent wrong "
2100 			    "authentication for %s",
2101 			    client_name(req), client_addr(req, buf),
2102 			    exi->exi_export.ex_path ?
2103 			    exi->exi_export.ex_path : "?");
2104 			return (0);
2105 		}
2106 
2107 		if (crsetugid(cr, UID_NOBODY, GID_NOBODY) != 0)
2108 			return (0);
2109 	}
2110 
2111 	return (1);
2112 }
2113 
2114 /*
2115  * returns 0 on failure, -1 on a drop, -2 on wrong security flavor,
2116  * and 1 on success
2117  */
2118 int
2119 checkauth4(struct compound_state *cs, struct svc_req *req)
2120 {
2121 	int i, rpcflavor, access;
2122 	struct secinfo *secp;
2123 	char buf[MAXHOST + 1];
2124 	int anon_res = 0, nfsflavor;
2125 	struct exportinfo *exi;
2126 	cred_t	*cr;
2127 	caddr_t	principal;
2128 
2129 	exi = cs->exi;
2130 	cr = cs->cr;
2131 	principal = cs->principal;
2132 	nfsflavor = cs->nfsflavor;
2133 
2134 	ASSERT(cr != NULL);
2135 
2136 	rpcflavor = req->rq_cred.oa_flavor;
2137 	cs->access &= ~CS_ACCESS_LIMITED;
2138 
2139 	/*
2140 	 * Check the access right per auth flavor on the vnode of
2141 	 * this export for the given request.
2142 	 */
2143 	access = nfsauth4_access(cs->exi, cs->vp, req);
2144 
2145 	if (access & NFSAUTH_WRONGSEC)
2146 		return (-2);	/* no access for this security flavor */
2147 
2148 	if (access & NFSAUTH_DROP)
2149 		return (-1);	/* drop the request */
2150 
2151 	if (access & NFSAUTH_DENIED) {
2152 
2153 		if (exi->exi_export.ex_seccnt > 0)
2154 			return (0);	/* deny access */
2155 
2156 	} else if (access & NFSAUTH_LIMITED) {
2157 
2158 		cs->access |= CS_ACCESS_LIMITED;
2159 
2160 	} else if (access & NFSAUTH_MAPNONE) {
2161 		/*
2162 		 * Access was granted even though the flavor mismatched
2163 		 * because AUTH_NONE was one of the exported flavors.
2164 		 */
2165 		rpcflavor = AUTH_NONE;
2166 	}
2167 
2168 	/*
2169 	 * XXX probably need to redo some of it for nfsv4?
2170 	 * return 1 on success or 0 on failure
2171 	 */
2172 
2173 	switch (rpcflavor) {
2174 	case AUTH_NONE:
2175 		anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2176 				exi->exi_export.ex_anon);
2177 		(void) crsetgroups(cr, 0, NULL);
2178 		break;
2179 
2180 	case AUTH_UNIX:
2181 		if (crgetuid(cr) == 0 && !(access & NFSAUTH_ROOT)) {
2182 			anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2183 					exi->exi_export.ex_anon);
2184 			(void) crsetgroups(cr, 0, NULL);
2185 		}
2186 		break;
2187 
2188 	default:
2189 		/*
2190 		 *  Find the secinfo structure.  We should be able
2191 		 *  to find it by the time we reach here.
2192 		 *  nfsauth_access() has done the checking.
2193 		 */
2194 		secp = NULL;
2195 		for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
2196 			if (exi->exi_export.ex_secinfo[i].s_secinfo.sc_nfsnum ==
2197 			    nfsflavor) {
2198 				secp = &exi->exi_export.ex_secinfo[i];
2199 				break;
2200 			}
2201 		}
2202 
2203 		if (!secp) {
2204 			cmn_err(CE_NOTE, "nfs_server: client %s%shad "
2205 			    "no secinfo data for flavor %d",
2206 			    client_name(req), client_addr(req, buf),
2207 			    nfsflavor);
2208 			return (0);
2209 		}
2210 
2211 		if (!checkwin(rpcflavor, secp->s_window, req)) {
2212 			cmn_err(CE_NOTE,
2213 			    "nfs_server: client %s%sused invalid "
2214 			    "auth window value",
2215 			    client_name(req), client_addr(req, buf));
2216 			return (0);
2217 		}
2218 
2219 		/*
2220 		 * Map root principals listed in the share's root= list to root,
2221 		 * and map any others principals that were mapped to root by RPC
2222 		 * to anon.
2223 		 */
2224 		if (principal && sec_svc_inrootlist(rpcflavor, principal,
2225 			secp->s_rootcnt, secp->s_rootnames)) {
2226 			if (crgetuid(cr) == 0)
2227 				return (1);
2228 
2229 			(void) crsetugid(cr, 0, 0);
2230 
2231 			/*
2232 			 * NOTE: If and when kernel-land privilege tracing is
2233 			 * added this may have to be replaced with code that
2234 			 * retrieves root's supplementary groups (e.g., using
2235 			 * kgss_get_group_info().  In the meantime principals
2236 			 * mapped to uid 0 get all privileges, so setting cr's
2237 			 * supplementary groups for them does nothing.
2238 			 */
2239 			(void) crsetgroups(cr, 0, NULL);
2240 
2241 			return (1);
2242 		}
2243 
2244 		/*
2245 		 * Not a root princ, or not in root list, map UID 0/nobody to
2246 		 * the anon ID for the share.  (RPC sets cr's UIDs and GIDs to
2247 		 * UID_NOBODY and GID_NOBODY, respectively.)
2248 		 */
2249 		if (crgetuid(cr) != 0 &&
2250 		    (crgetuid(cr) != UID_NOBODY || crgetgid(cr) != GID_NOBODY))
2251 			return (1);
2252 
2253 		anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2254 			exi->exi_export.ex_anon);
2255 		(void) crsetgroups(cr, 0, NULL);
2256 		break;
2257 	} /* switch on rpcflavor */
2258 
2259 	/*
2260 	 * Even if anon access is disallowed via ex_anon == -1, we allow
2261 	 * this access if anon_ok is set.  So set creds to the default
2262 	 * "nobody" id.
2263 	 */
2264 
2265 	if (anon_res != 0) {
2266 		cmn_err(CE_NOTE,
2267 			"nfs_server: client %s%ssent wrong "
2268 			"authentication for %s",
2269 			client_name(req), client_addr(req, buf),
2270 			exi->exi_export.ex_path ?
2271 			exi->exi_export.ex_path : "?");
2272 		return (0);
2273 	}
2274 
2275 	return (1);
2276 }
2277 
2278 
2279 static char *
2280 client_name(struct svc_req *req)
2281 {
2282 	char *hostname = NULL;
2283 
2284 	/*
2285 	 * If it's a Unix cred then use the
2286 	 * hostname from the credential.
2287 	 */
2288 	if (req->rq_cred.oa_flavor == AUTH_UNIX) {
2289 		hostname = ((struct authunix_parms *)
2290 		    req->rq_clntcred)->aup_machname;
2291 	}
2292 	if (hostname == NULL)
2293 		hostname = "";
2294 
2295 	return (hostname);
2296 }
2297 
2298 static char *
2299 client_addr(struct svc_req *req, char *buf)
2300 {
2301 	struct sockaddr *ca;
2302 	uchar_t *b;
2303 	char *frontspace = "";
2304 
2305 	/*
2306 	 * We assume we are called in tandem with client_name and the
2307 	 * format string looks like "...client %s%sblah blah..."
2308 	 *
2309 	 * If it's a Unix cred then client_name returned
2310 	 * a host name, so we need insert a space between host name
2311 	 * and IP address.
2312 	 */
2313 	if (req->rq_cred.oa_flavor == AUTH_UNIX)
2314 		frontspace = " ";
2315 
2316 	/*
2317 	 * Convert the caller's IP address to a dotted string
2318 	 */
2319 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
2320 
2321 	if (ca->sa_family == AF_INET) {
2322 	    b = (uchar_t *)&((struct sockaddr_in *)ca)->sin_addr;
2323 	    (void) sprintf(buf, "%s(%d.%d.%d.%d) ", frontspace,
2324 		b[0] & 0xFF, b[1] & 0xFF, b[2] & 0xFF, b[3] & 0xFF);
2325 	} else if (ca->sa_family == AF_INET6) {
2326 		struct sockaddr_in6 *sin6;
2327 		sin6 = (struct sockaddr_in6 *)ca;
2328 		(void) kinet_ntop6((uchar_t *)&sin6->sin6_addr,
2329 				buf, INET6_ADDRSTRLEN);
2330 
2331 	} else {
2332 
2333 		/*
2334 		 * No IP address to print. If there was a host name
2335 		 * printed, then we print a space.
2336 		 */
2337 		(void) sprintf(buf, frontspace);
2338 	}
2339 
2340 	return (buf);
2341 }
2342 
2343 /*
2344  * NFS Server initialization routine.  This routine should only be called
2345  * once.  It performs the following tasks:
2346  *	- Call sub-initialization routines (localize access to variables)
2347  *	- Initialize all locks
2348  *	- initialize the version 3 write verifier
2349  */
2350 int
2351 nfs_srvinit(void)
2352 {
2353 	int error;
2354 
2355 	error = nfs_exportinit();
2356 	if (error != 0)
2357 		return (error);
2358 	error = rfs4_srvrinit();
2359 	if (error != 0) {
2360 		nfs_exportfini();
2361 		return (error);
2362 	}
2363 	rfs_srvrinit();
2364 	rfs3_srvrinit();
2365 	nfsauth_init();
2366 
2367 	/* Init the stuff to control start/stop */
2368 	nfs_server_upordown = NFS_SERVER_STOPPED;
2369 	mutex_init(&nfs_server_upordown_lock, NULL, MUTEX_DEFAULT, NULL);
2370 	cv_init(&nfs_server_upordown_cv, NULL, CV_DEFAULT, NULL);
2371 	mutex_init(&rdma_wait_mutex, NULL, MUTEX_DEFAULT, NULL);
2372 	cv_init(&rdma_wait_cv, NULL, CV_DEFAULT, NULL);
2373 
2374 	return (0);
2375 }
2376 
2377 /*
2378  * NFS Server finalization routine. This routine is called to cleanup the
2379  * initialization work previously performed if the NFS server module could
2380  * not be loaded correctly.
2381  */
2382 void
2383 nfs_srvfini(void)
2384 {
2385 	nfsauth_fini();
2386 	rfs3_srvrfini();
2387 	rfs_srvrfini();
2388 	nfs_exportfini();
2389 
2390 	mutex_destroy(&nfs_server_upordown_lock);
2391 	cv_destroy(&nfs_server_upordown_cv);
2392 	mutex_destroy(&rdma_wait_mutex);
2393 	cv_destroy(&rdma_wait_cv);
2394 }
2395 
2396 /*
2397  * Set up an iovec array of up to cnt pointers.
2398  */
2399 
2400 void
2401 mblk_to_iov(mblk_t *m, int cnt, struct iovec *iovp)
2402 {
2403 	while (m != NULL && cnt-- > 0) {
2404 		iovp->iov_base = (caddr_t)m->b_rptr;
2405 		iovp->iov_len = (m->b_wptr - m->b_rptr);
2406 		iovp++;
2407 		m = m->b_cont;
2408 	}
2409 }
2410 
2411 /*
2412  * Common code between NFS Version 2 and NFS Version 3 for the public
2413  * filehandle multicomponent lookups.
2414  */
2415 
2416 /*
2417  * Public filehandle evaluation of a multi-component lookup, following
2418  * symbolic links, if necessary. This may result in a vnode in another
2419  * filesystem, which is OK as long as the other filesystem is exported.
2420  *
2421  * Note that the exi will be set either to NULL or a new reference to the
2422  * exportinfo struct that corresponds to the vnode of the multi-component path.
2423  * It is the callers responsibility to release this reference.
2424  */
2425 int
2426 rfs_publicfh_mclookup(char *p, vnode_t *dvp, cred_t *cr, vnode_t **vpp,
2427     struct exportinfo **exi, struct sec_ol *sec)
2428 {
2429 	int pathflag;
2430 	vnode_t *mc_dvp = NULL;
2431 	vnode_t *realvp;
2432 	int error;
2433 
2434 	*exi = NULL;
2435 
2436 	/*
2437 	 * check if the given path is a url or native path. Since p is
2438 	 * modified by MCLpath(), it may be empty after returning from
2439 	 * there, and should be checked.
2440 	 */
2441 	if ((pathflag = MCLpath(&p)) == -1)
2442 		return (EIO);
2443 
2444 	/*
2445 	 * If pathflag is SECURITY_QUERY, turn the SEC_QUERY bit
2446 	 * on in sec->sec_flags. This bit will later serve as an
2447 	 * indication in makefh_ol() or makefh3_ol() to overload the
2448 	 * filehandle to contain the sec modes used by the server for
2449 	 * the path.
2450 	 */
2451 	if (pathflag == SECURITY_QUERY) {
2452 		if ((sec->sec_index = (uint_t)(*p)) > 0) {
2453 			sec->sec_flags |= SEC_QUERY;
2454 			p++;
2455 			if ((pathflag = MCLpath(&p)) == -1)
2456 				return (EIO);
2457 		} else {
2458 			cmn_err(CE_NOTE,
2459 			    "nfs_server: invalid security index %d, "
2460 			    "violating WebNFS SNEGO protocol.", sec->sec_index);
2461 			return (EIO);
2462 		}
2463 	}
2464 
2465 	if (p[0] == '\0') {
2466 		error = ENOENT;
2467 		goto publicfh_done;
2468 	}
2469 
2470 	error = rfs_pathname(p, &mc_dvp, vpp, dvp, cr, pathflag);
2471 
2472 	/*
2473 	 * If name resolves to "/" we get EINVAL since we asked for
2474 	 * the vnode of the directory that the file is in. Try again
2475 	 * with NULL directory vnode.
2476 	 */
2477 	if (error == EINVAL) {
2478 		error = rfs_pathname(p, NULL, vpp, dvp, cr, pathflag);
2479 		if (!error) {
2480 			ASSERT(*vpp != NULL);
2481 			if ((*vpp)->v_type == VDIR) {
2482 				VN_HOLD(*vpp);
2483 				mc_dvp = *vpp;
2484 			} else {
2485 				/*
2486 				 * This should not happen, the filesystem is
2487 				 * in an inconsistent state. Fail the lookup
2488 				 * at this point.
2489 				 */
2490 				VN_RELE(*vpp);
2491 				error = EINVAL;
2492 			}
2493 		}
2494 	}
2495 
2496 	if (error)
2497 		goto publicfh_done;
2498 
2499 	if (*vpp == NULL) {
2500 		error = ENOENT;
2501 		goto publicfh_done;
2502 	}
2503 
2504 	ASSERT(mc_dvp != NULL);
2505 	ASSERT(*vpp != NULL);
2506 
2507 	if ((*vpp)->v_type == VDIR) {
2508 		do {
2509 			/*
2510 			 * *vpp may be an AutoFS node, so we perform
2511 			 * a VOP_ACCESS() to trigger the mount of the intended
2512 			 * filesystem, so we can perform the lookup in the
2513 			 * intended filesystem.
2514 			 */
2515 			(void) VOP_ACCESS(*vpp, 0, 0, cr);
2516 
2517 			/*
2518 			 * If vnode is covered, get the
2519 			 * the topmost vnode.
2520 			 */
2521 			if (vn_mountedvfs(*vpp) != NULL) {
2522 				error = traverse(vpp);
2523 				if (error) {
2524 					VN_RELE(*vpp);
2525 					goto publicfh_done;
2526 				}
2527 			}
2528 
2529 			if (VOP_REALVP(*vpp, &realvp) == 0 && realvp != *vpp) {
2530 				/*
2531 				 * If realvp is different from *vpp
2532 				 * then release our reference on *vpp, so that
2533 				 * the export access check be performed on the
2534 				 * real filesystem instead.
2535 				 */
2536 				VN_HOLD(realvp);
2537 				VN_RELE(*vpp);
2538 				*vpp = realvp;
2539 			} else
2540 			    break;
2541 		/* LINTED */
2542 		} while (TRUE);
2543 
2544 		/*
2545 		 * Let nfs_vptexi() figure what the real parent is.
2546 		 */
2547 		VN_RELE(mc_dvp);
2548 		mc_dvp = NULL;
2549 
2550 	} else {
2551 		/*
2552 		 * If vnode is covered, get the
2553 		 * the topmost vnode.
2554 		 */
2555 		if (vn_mountedvfs(mc_dvp) != NULL) {
2556 			error = traverse(&mc_dvp);
2557 			if (error) {
2558 			    VN_RELE(*vpp);
2559 			    goto publicfh_done;
2560 			}
2561 		}
2562 
2563 		if (VOP_REALVP(mc_dvp, &realvp) == 0 && realvp != mc_dvp) {
2564 			/*
2565 			 * *vpp is a file, obtain realvp of the parent
2566 			 * directory vnode.
2567 			 */
2568 			VN_HOLD(realvp);
2569 			VN_RELE(mc_dvp);
2570 			mc_dvp = realvp;
2571 		}
2572 	}
2573 
2574 	/*
2575 	 * The pathname may take us from the public filesystem to another.
2576 	 * If that's the case then just set the exportinfo to the new export
2577 	 * and build filehandle for it. Thanks to per-access checking there's
2578 	 * no security issues with doing this. If the client is not allowed
2579 	 * access to this new export then it will get an access error when it
2580 	 * tries to use the filehandle
2581 	 */
2582 	if (error = nfs_check_vpexi(mc_dvp, *vpp, kcred, exi)) {
2583 		VN_RELE(*vpp);
2584 		goto publicfh_done;
2585 	}
2586 
2587 	/*
2588 	 * Not allowed access to pseudo exports.
2589 	 */
2590 	if (PSEUDO(*exi)) {
2591 		error = ENOENT;
2592 		VN_RELE(*vpp);
2593 		goto publicfh_done;
2594 	}
2595 
2596 	/*
2597 	 * Do a lookup for the index file. We know the index option doesn't
2598 	 * allow paths through handling in the share command, so mc_dvp will
2599 	 * be the parent for the index file vnode, if its present. Use
2600 	 * temporary pointers to preserve and reuse the vnode pointers of the
2601 	 * original directory in case there's no index file. Note that the
2602 	 * index file is a native path, and should not be interpreted by
2603 	 * the URL parser in rfs_pathname()
2604 	 */
2605 	if (((*exi)->exi_export.ex_flags & EX_INDEX) &&
2606 	    ((*vpp)->v_type == VDIR) && (pathflag == URLPATH)) {
2607 		vnode_t *tvp, *tmc_dvp;	/* temporary vnode pointers */
2608 
2609 		tmc_dvp = mc_dvp;
2610 		mc_dvp = tvp = *vpp;
2611 
2612 		error = rfs_pathname((*exi)->exi_export.ex_index, NULL, vpp,
2613 		    mc_dvp, cr, NATIVEPATH);
2614 
2615 		if (error == ENOENT) {
2616 			*vpp = tvp;
2617 			mc_dvp = tmc_dvp;
2618 			error = 0;
2619 		} else {	/* ok or error other than ENOENT */
2620 			if (tmc_dvp)
2621 				VN_RELE(tmc_dvp);
2622 			if (error)
2623 				goto publicfh_done;
2624 
2625 			/*
2626 			 * Found a valid vp for index "filename". Sanity check
2627 			 * for odd case where a directory is provided as index
2628 			 * option argument and leads us to another filesystem
2629 			 */
2630 
2631 			/* Release the reference on the old exi value */
2632 			ASSERT(*exi != NULL);
2633 			exi_rele(*exi);
2634 
2635 			if (error = nfs_check_vpexi(mc_dvp, *vpp, kcred, exi)) {
2636 				VN_RELE(*vpp);
2637 				goto publicfh_done;
2638 			}
2639 		}
2640 	}
2641 
2642 publicfh_done:
2643 	if (mc_dvp)
2644 		VN_RELE(mc_dvp);
2645 
2646 	return (error);
2647 }
2648 
2649 /*
2650  * Evaluate a multi-component path
2651  */
2652 int
2653 rfs_pathname(
2654 	char *path,			/* pathname to evaluate */
2655 	vnode_t **dirvpp,		/* ret for ptr to parent dir vnode */
2656 	vnode_t **compvpp,		/* ret for ptr to component vnode */
2657 	vnode_t *startdvp,		/* starting vnode */
2658 	cred_t *cr,			/* user's credential */
2659 	int pathflag)			/* flag to identify path, e.g. URL */
2660 {
2661 	char namebuf[TYPICALMAXPATHLEN];
2662 	struct pathname pn;
2663 	int error;
2664 
2665 	/*
2666 	 * If pathname starts with '/', then set startdvp to root.
2667 	 */
2668 	if (*path == '/') {
2669 		while (*path == '/')
2670 			path++;
2671 
2672 		startdvp = rootdir;
2673 	}
2674 
2675 	error = pn_get_buf(path, UIO_SYSSPACE, &pn, namebuf, sizeof (namebuf));
2676 	if (error == 0) {
2677 		/*
2678 		 * Call the URL parser for URL paths to modify the original
2679 		 * string to handle any '%' encoded characters that exist.
2680 		 * Done here to avoid an extra bcopy in the lookup.
2681 		 * We need to be careful about pathlen's. We know that
2682 		 * rfs_pathname() is called with a non-empty path. However,
2683 		 * it could be emptied due to the path simply being all /'s,
2684 		 * which is valid to proceed with the lookup, or due to the
2685 		 * URL parser finding an encoded null character at the
2686 		 * beginning of path which should not proceed with the lookup.
2687 		 */
2688 		if (pn.pn_pathlen != 0 && pathflag == URLPATH) {
2689 			URLparse(pn.pn_path);
2690 			if ((pn.pn_pathlen = strlen(pn.pn_path)) == 0)
2691 				return (ENOENT);
2692 		}
2693 		VN_HOLD(startdvp);
2694 		error = lookuppnvp(&pn, NULL, NO_FOLLOW, dirvpp, compvpp,
2695 		    rootdir, startdvp, cr);
2696 	}
2697 	if (error == ENAMETOOLONG) {
2698 		/*
2699 		 * This thread used a pathname > TYPICALMAXPATHLEN bytes long.
2700 		 */
2701 		if (error = pn_get(path, UIO_SYSSPACE, &pn))
2702 			return (error);
2703 		if (pn.pn_pathlen != 0 && pathflag == URLPATH) {
2704 			URLparse(pn.pn_path);
2705 			if ((pn.pn_pathlen = strlen(pn.pn_path)) == 0) {
2706 				pn_free(&pn);
2707 				return (ENOENT);
2708 			}
2709 		}
2710 		VN_HOLD(startdvp);
2711 		error = lookuppnvp(&pn, NULL, NO_FOLLOW, dirvpp, compvpp,
2712 		    rootdir, startdvp, cr);
2713 		pn_free(&pn);
2714 	}
2715 
2716 	return (error);
2717 }
2718 
2719 /*
2720  * Adapt the multicomponent lookup path depending on the pathtype
2721  */
2722 static int
2723 MCLpath(char **path)
2724 {
2725 	unsigned char c = (unsigned char)**path;
2726 
2727 	/*
2728 	 * If the MCL path is between 0x20 and 0x7E (graphic printable
2729 	 * character of the US-ASCII coded character set), its a URL path,
2730 	 * per RFC 1738.
2731 	 */
2732 	if (c >= 0x20 && c <= 0x7E)
2733 		return (URLPATH);
2734 
2735 	/*
2736 	 * If the first octet of the MCL path is not an ASCII character
2737 	 * then it must be interpreted as a tag value that describes the
2738 	 * format of the remaining octets of the MCL path.
2739 	 *
2740 	 * If the first octet of the MCL path is 0x81 it is a query
2741 	 * for the security info.
2742 	 */
2743 	switch (c) {
2744 	case 0x80:	/* native path, i.e. MCL via mount protocol */
2745 		(*path)++;
2746 		return (NATIVEPATH);
2747 	case 0x81:	/* security query */
2748 		(*path)++;
2749 		return (SECURITY_QUERY);
2750 	default:
2751 		return (-1);
2752 	}
2753 }
2754 
2755 #define	fromhex(c)  ((c >= '0' && c <= '9') ? (c - '0') : \
2756 			((c >= 'A' && c <= 'F') ? (c - 'A' + 10) :\
2757 			((c >= 'a' && c <= 'f') ? (c - 'a' + 10) : 0)))
2758 
2759 /*
2760  * The implementation of URLparse gaurantees that the final string will
2761  * fit in the original one. Replaces '%' occurrences followed by 2 characters
2762  * with its corresponding hexadecimal character.
2763  */
2764 static void
2765 URLparse(char *str)
2766 {
2767 	char *p, *q;
2768 
2769 	p = q = str;
2770 	while (*p) {
2771 		*q = *p;
2772 		if (*p++ == '%') {
2773 			if (*p) {
2774 				*q = fromhex(*p) * 16;
2775 				p++;
2776 				if (*p) {
2777 					*q += fromhex(*p);
2778 					p++;
2779 				}
2780 			}
2781 		}
2782 		q++;
2783 	}
2784 	*q = '\0';
2785 }
2786 
2787 
2788 /*
2789  * Get the export information for the lookup vnode, and verify its
2790  * useable.
2791  */
2792 int
2793 nfs_check_vpexi(vnode_t *mc_dvp, vnode_t *vp, cred_t *cr,
2794     struct exportinfo **exi)
2795 {
2796 	int walk;
2797 	int error = 0;
2798 
2799 	*exi = nfs_vptoexi(mc_dvp, vp, cr, &walk, NULL, FALSE);
2800 	if (*exi == NULL)
2801 		error = EACCES;
2802 	else {
2803 		/*
2804 		 * If nosub is set for this export then
2805 		 * a lookup relative to the public fh
2806 		 * must not terminate below the
2807 		 * exported directory.
2808 		 */
2809 		if ((*exi)->exi_export.ex_flags & EX_NOSUB && walk > 0)
2810 			error = EACCES;
2811 	}
2812 
2813 	return (error);
2814 }
2815