xref: /titanic_51/usr/src/uts/common/fs/nfs/nfs4_srv.c (revision d7b41ca400354f0823f3c776bf7b2b43d2182821)
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  */
30 
31 #pragma ident	"%Z%%M%	%I%	%E% SMI"
32 
33 #include <sys/param.h>
34 #include <sys/types.h>
35 #include <sys/systm.h>
36 #include <sys/cred.h>
37 #include <sys/buf.h>
38 #include <sys/vfs.h>
39 #include <sys/vnode.h>
40 #include <sys/uio.h>
41 #include <sys/errno.h>
42 #include <sys/sysmacros.h>
43 #include <sys/statvfs.h>
44 #include <sys/kmem.h>
45 #include <sys/dirent.h>
46 #include <sys/cmn_err.h>
47 #include <sys/debug.h>
48 #include <sys/systeminfo.h>
49 #include <sys/flock.h>
50 #include <sys/pathname.h>
51 #include <sys/nbmlock.h>
52 #include <sys/share.h>
53 #include <sys/atomic.h>
54 #include <sys/policy.h>
55 #include <sys/fem.h>
56 #include <sys/sdt.h>
57 
58 #include <rpc/types.h>
59 #include <rpc/auth.h>
60 #include <rpc/rpcsec_gss.h>
61 #include <rpc/svc.h>
62 
63 #include <nfs/nfs.h>
64 #include <nfs/export.h>
65 #include <nfs/lm.h>
66 #include <nfs/nfs4.h>
67 
68 #include <sys/strsubr.h>
69 #include <sys/strsun.h>
70 
71 #include <inet/common.h>
72 #include <inet/ip.h>
73 #include <inet/ip6.h>
74 
75 #include <sys/tsol/label.h>
76 #include <sys/tsol/tndb.h>
77 
78 #define	RFS4_MAXLOCK_TRIES 4	/* Try to get the lock this many times */
79 static int rfs4_maxlock_tries = RFS4_MAXLOCK_TRIES;
80 #define	RFS4_LOCK_DELAY 10	/* Milliseconds */
81 static clock_t rfs4_lock_delay = RFS4_LOCK_DELAY;
82 
83 /* End of Tunables */
84 
85 /*
86  * Used to bump the stateid4.seqid value and show changes in the stateid
87  */
88 #define	next_stateid(sp) (++(sp)->bits.chgseq)
89 
90 /*
91  * RFS4_MINLEN_ENTRY4: XDR-encoded size of smallest possible dirent.
92  *	This is used to return NFS4ERR_TOOSMALL when clients specify
93  *	maxcount that isn't large enough to hold the smallest possible
94  *	XDR encoded dirent.
95  *
96  *	    sizeof cookie (8 bytes) +
97  *	    sizeof name_len (4 bytes) +
98  *	    sizeof smallest (padded) name (4 bytes) +
99  *	    sizeof bitmap4_len (12 bytes) +   NOTE: we always encode len=2 bm4
100  *	    sizeof attrlist4_len (4 bytes) +
101  *	    sizeof next boolean (4 bytes)
102  *
103  * RFS4_MINLEN_RDDIR4: XDR-encoded size of READDIR op reply containing
104  * the smallest possible entry4 (assumes no attrs requested).
105  *	sizeof nfsstat4 (4 bytes) +
106  *	sizeof verifier4 (8 bytes) +
107  *	sizeof entry4list bool (4 bytes) +
108  *	sizeof entry4 	(36 bytes) +
109  *	sizeof eof bool  (4 bytes)
110  *
111  * RFS4_MINLEN_RDDIR_BUF: minimum length of buffer server will provide to
112  *	VOP_READDIR.  Its value is the size of the maximum possible dirent
113  *	for solaris.  The DIRENT64_RECLEN macro returns	the size of dirent
114  *	required for a given name length.  MAXNAMELEN is the maximum
115  *	filename length allowed in Solaris.  The first two DIRENT64_RECLEN()
116  *	macros are to allow for . and .. entries -- just a minor tweak to try
117  *	and guarantee that buffer we give to VOP_READDIR will be large enough
118  *	to hold ., .., and the largest possible solaris dirent64.
119  */
120 #define	RFS4_MINLEN_ENTRY4 36
121 #define	RFS4_MINLEN_RDDIR4 (4 + NFS4_VERIFIER_SIZE + 4 + RFS4_MINLEN_ENTRY4 + 4)
122 #define	RFS4_MINLEN_RDDIR_BUF \
123 	(DIRENT64_RECLEN(1) + DIRENT64_RECLEN(2) + DIRENT64_RECLEN(MAXNAMELEN))
124 
125 /*
126  * It would be better to pad to 4 bytes since that's what XDR would do,
127  * but the dirents UFS gives us are already padded to 8, so just take
128  * what we're given.  Dircount is only a hint anyway.  Currently the
129  * solaris kernel is ASCII only, so there's no point in calling the
130  * UTF8 functions.
131  *
132  * dirent64: named padded to provide 8 byte struct alignment
133  *	d_ino(8) + d_off(8) + d_reclen(2) + d_name(namelen + null(1) + pad)
134  *
135  * cookie: uint64_t   +  utf8namelen: uint_t  +   utf8name padded to 8 bytes
136  *
137  */
138 #define	DIRENT64_TO_DIRCOUNT(dp) \
139 	(3 * BYTES_PER_XDR_UNIT + DIRENT64_NAMELEN((dp)->d_reclen))
140 
141 /*
142  * types of label comparison
143  */
144 #define	EQUALITY_CHECK	0
145 #define	DOMINANCE_CHECK	1
146 
147 time_t rfs4_start_time;			/* Initialized in rfs4_srvrinit */
148 
149 static sysid_t lockt_sysid;		/* dummy sysid for all LOCKT calls */
150 
151 u_longlong_t nfs4_srv_caller_id;
152 
153 verifier4	Write4verf;
154 verifier4	Readdir4verf;
155 
156 void		rfs4_init_compound_state(struct compound_state *);
157 
158 static void	nullfree(caddr_t);
159 static void	rfs4_op_inval(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
160 			struct compound_state *);
161 static void	rfs4_op_access(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
162 			struct compound_state *);
163 static void	rfs4_op_close(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
164 			struct compound_state *);
165 static void	rfs4_op_commit(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
166 			struct compound_state *);
167 static void	rfs4_op_create(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
168 			struct compound_state *);
169 static void	rfs4_op_create_free(nfs_resop4 *resop);
170 static void	rfs4_op_delegreturn(nfs_argop4 *, nfs_resop4 *,
171 				    struct svc_req *, struct compound_state *);
172 static void	rfs4_op_getattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
173 			struct compound_state *);
174 static void	rfs4_op_getattr_free(nfs_resop4 *);
175 static void	rfs4_op_getfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
176 			struct compound_state *);
177 static void	rfs4_op_getfh_free(nfs_resop4 *);
178 static void	rfs4_op_illegal(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
179 			struct compound_state *);
180 static void	rfs4_op_link(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
181 			struct compound_state *);
182 static void	rfs4_op_lock(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
183 			struct compound_state *);
184 static void	lock_denied_free(nfs_resop4 *);
185 static void	rfs4_op_locku(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
186 			struct compound_state *);
187 static void	rfs4_op_lockt(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
188 			struct compound_state *);
189 static void	rfs4_op_lookup(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
190 			struct compound_state *);
191 static void	rfs4_op_lookupp(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
192 			struct compound_state *);
193 static void	rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop,
194 				struct svc_req *req, struct compound_state *cs);
195 static void	rfs4_op_nverify(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
196 			struct compound_state *);
197 static void	rfs4_op_open(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
198 			struct compound_state *);
199 static void	rfs4_op_open_confirm(nfs_argop4 *, nfs_resop4 *,
200 			struct svc_req *, struct compound_state *);
201 static void	rfs4_op_open_downgrade(nfs_argop4 *, nfs_resop4 *,
202 			struct svc_req *, struct compound_state *);
203 static void	rfs4_op_putfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
204 			struct compound_state *);
205 static void	rfs4_op_putpubfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
206 			struct compound_state *);
207 static void	rfs4_op_putrootfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
208 			struct compound_state *);
209 static void	rfs4_op_read(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
210 			struct compound_state *);
211 static void	rfs4_op_read_free(nfs_resop4 *);
212 static void	rfs4_op_readdir_free(nfs_resop4 *resop);
213 static void	rfs4_op_readlink(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
214 			struct compound_state *);
215 static void	rfs4_op_readlink_free(nfs_resop4 *);
216 static void	rfs4_op_release_lockowner(nfs_argop4 *, nfs_resop4 *,
217 			struct svc_req *, struct compound_state *);
218 static void	rfs4_op_remove(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
219 			struct compound_state *);
220 static void	rfs4_op_rename(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
221 			struct compound_state *);
222 static void	rfs4_op_renew(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
223 			struct compound_state *);
224 static void	rfs4_op_restorefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
225 			struct compound_state *);
226 static void	rfs4_op_savefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
227 			struct compound_state *);
228 static void	rfs4_op_setattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
229 			struct compound_state *);
230 static void	rfs4_op_verify(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
231 			struct compound_state *);
232 static void	rfs4_op_write(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
233 			struct compound_state *);
234 static void	rfs4_op_setclientid(nfs_argop4 *, nfs_resop4 *,
235 			struct svc_req *, struct compound_state *);
236 static void	rfs4_op_setclientid_confirm(nfs_argop4 *, nfs_resop4 *,
237 			struct svc_req *req, struct compound_state *);
238 static void	rfs4_op_secinfo(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
239 			struct compound_state *);
240 static void	rfs4_op_secinfo_free(nfs_resop4 *);
241 
242 static nfsstat4 check_open_access(uint32_t,
243 				struct compound_state *, struct svc_req *);
244 nfsstat4 rfs4_client_sysid(rfs4_client_t *, sysid_t *);
245 static int	vop_shrlock(vnode_t *, int, struct shrlock *, int);
246 static int 	rfs4_shrlock(rfs4_state_t *, int);
247 static int	rfs4_share(rfs4_state_t *);
248 void rfs4_ss_clid(rfs4_client_t *, struct svc_req *);
249 
250 /*
251  * translation table for attrs
252  */
253 struct nfs4_ntov_table {
254 	union nfs4_attr_u *na;
255 	uint8_t amap[NFS4_MAXNUM_ATTRS];
256 	int attrcnt;
257 	bool_t vfsstat;
258 };
259 
260 static void	nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp);
261 static void	nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp,
262 				    struct nfs4_svgetit_arg *sargp);
263 
264 static nfsstat4	do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp,
265 		    struct compound_state *cs, struct nfs4_svgetit_arg *sargp,
266 		    struct nfs4_ntov_table *ntovp, nfs4_attr_cmd_t cmd);
267 
268 fem_t	*deleg_rdops;
269 fem_t	*deleg_wrops;
270 
271 rfs4_servinst_t	*rfs4_cur_servinst = NULL;	/* current server instance */
272 kmutex_t	rfs4_servinst_lock;		/* protects linked list */
273 int		rfs4_seen_first_compound;	/* set first time we see one */
274 
275 #ifdef DEBUG
276 int	rfs4_servinst_debug = 0;
277 #endif
278 
279 /*
280  * NFS4 op dispatch table
281  */
282 
283 struct rfsv4disp {
284 	void	(*dis_proc)();		/* proc to call */
285 	void	(*dis_resfree)();	/* frees space allocated by proc */
286 	int	dis_flags;		/* RPC_IDEMPOTENT, etc... */
287 };
288 
289 static struct rfsv4disp rfsv4disptab[] = {
290 	/*
291 	 * NFS VERSION 4
292 	 */
293 
294 	/* RFS_NULL = 0 */
295 	{rfs4_op_illegal, nullfree, 0},
296 
297 	/* UNUSED = 1 */
298 	{rfs4_op_illegal, nullfree, 0},
299 
300 	/* UNUSED = 2 */
301 	{rfs4_op_illegal, nullfree, 0},
302 
303 	/* OP_ACCESS = 3 */
304 	{rfs4_op_access, nullfree, RPC_IDEMPOTENT},
305 
306 	/* OP_CLOSE = 4 */
307 	{rfs4_op_close, nullfree, 0},
308 
309 	/* OP_COMMIT = 5 */
310 	{rfs4_op_commit, nullfree, RPC_IDEMPOTENT},
311 
312 	/* OP_CREATE = 6 */
313 	{rfs4_op_create, nullfree, 0},
314 
315 	/* OP_DELEGPURGE = 7 */
316 	{rfs4_op_inval, nullfree, 0},
317 
318 	/* OP_DELEGRETURN = 8 */
319 	{rfs4_op_delegreturn, nullfree, 0},
320 
321 	/* OP_GETATTR = 9 */
322 	{rfs4_op_getattr, rfs4_op_getattr_free, RPC_IDEMPOTENT},
323 
324 	/* OP_GETFH = 10 */
325 	{rfs4_op_getfh, rfs4_op_getfh_free, RPC_ALL},
326 
327 	/* OP_LINK = 11 */
328 	{rfs4_op_link, nullfree, 0},
329 
330 	/* OP_LOCK = 12 */
331 	{rfs4_op_lock, lock_denied_free, 0},
332 
333 	/* OP_LOCKT = 13 */
334 	{rfs4_op_lockt, lock_denied_free, 0},
335 
336 	/* OP_LOCKU = 14 */
337 	{rfs4_op_locku, nullfree, 0},
338 
339 	/* OP_LOOKUP = 15 */
340 	{rfs4_op_lookup, nullfree, (RPC_IDEMPOTENT|RPC_PUBLICFH_OK)},
341 
342 	/* OP_LOOKUPP = 16 */
343 	{rfs4_op_lookupp, nullfree, (RPC_IDEMPOTENT|RPC_PUBLICFH_OK)},
344 
345 	/* OP_NVERIFY = 17 */
346 	{rfs4_op_nverify, nullfree, RPC_IDEMPOTENT},
347 
348 	/* OP_OPEN = 18 */
349 	{rfs4_op_open, rfs4_free_reply, 0},
350 
351 	/* OP_OPENATTR = 19 */
352 	{rfs4_op_openattr, nullfree, 0},
353 
354 	/* OP_OPEN_CONFIRM = 20 */
355 	{rfs4_op_open_confirm, nullfree, 0},
356 
357 	/* OP_OPEN_DOWNGRADE = 21 */
358 	{rfs4_op_open_downgrade, nullfree, 0},
359 
360 	/* OP_OPEN_PUTFH = 22 */
361 	{rfs4_op_putfh, nullfree, RPC_ALL},
362 
363 	/* OP_PUTPUBFH = 23 */
364 	{rfs4_op_putpubfh, nullfree, RPC_ALL},
365 
366 	/* OP_PUTROOTFH = 24 */
367 	{rfs4_op_putrootfh, nullfree, RPC_ALL},
368 
369 	/* OP_READ = 25 */
370 	{rfs4_op_read, rfs4_op_read_free, RPC_IDEMPOTENT},
371 
372 	/* OP_READDIR = 26 */
373 	{rfs4_op_readdir, rfs4_op_readdir_free, RPC_IDEMPOTENT},
374 
375 	/* OP_READLINK = 27 */
376 	{rfs4_op_readlink, rfs4_op_readlink_free, RPC_IDEMPOTENT},
377 
378 	/* OP_REMOVE = 28 */
379 	{rfs4_op_remove, nullfree, 0},
380 
381 	/* OP_RENAME = 29 */
382 	{rfs4_op_rename, nullfree, 0},
383 
384 	/* OP_RENEW = 30 */
385 	{rfs4_op_renew, nullfree, 0},
386 
387 	/* OP_RESTOREFH = 31 */
388 	{rfs4_op_restorefh, nullfree, RPC_ALL},
389 
390 	/* OP_SAVEFH = 32 */
391 	{rfs4_op_savefh, nullfree, RPC_ALL},
392 
393 	/* OP_SECINFO = 33 */
394 	{rfs4_op_secinfo, rfs4_op_secinfo_free, 0},
395 
396 	/* OP_SETATTR = 34 */
397 	{rfs4_op_setattr, nullfree, 0},
398 
399 	/* OP_SETCLIENTID = 35 */
400 	{rfs4_op_setclientid, nullfree, 0},
401 
402 	/* OP_SETCLIENTID_CONFIRM = 36 */
403 	{rfs4_op_setclientid_confirm, nullfree, 0},
404 
405 	/* OP_VERIFY = 37 */
406 	{rfs4_op_verify, nullfree, RPC_IDEMPOTENT},
407 
408 	/* OP_WRITE = 38 */
409 	{rfs4_op_write, nullfree, 0},
410 
411 	/* OP_RELEASE_LOCKOWNER = 39 */
412 	{rfs4_op_release_lockowner, nullfree, 0},
413 };
414 
415 static uint_t rfsv4disp_cnt = sizeof (rfsv4disptab) / sizeof (rfsv4disptab[0]);
416 
417 #define	OP_ILLEGAL_IDX (rfsv4disp_cnt)
418 
419 #ifdef DEBUG
420 
421 int rfs4_fillone_debug = 0;
422 int rfs4_shrlock_debug = 0;
423 int rfs4_no_stub_access = 1;
424 int rfs4_rddir_debug = 0;
425 
426 static char *rfs4_op_string[] = {
427 	"rfs4_op_null",
428 	"rfs4_op_1 unused",
429 	"rfs4_op_2 unused",
430 	"rfs4_op_access",
431 	"rfs4_op_close",
432 	"rfs4_op_commit",
433 	"rfs4_op_create",
434 	"rfs4_op_delegpurge",
435 	"rfs4_op_delegreturn",
436 	"rfs4_op_getattr",
437 	"rfs4_op_getfh",
438 	"rfs4_op_link",
439 	"rfs4_op_lock",
440 	"rfs4_op_lockt",
441 	"rfs4_op_locku",
442 	"rfs4_op_lookup",
443 	"rfs4_op_lookupp",
444 	"rfs4_op_nverify",
445 	"rfs4_op_open",
446 	"rfs4_op_openattr",
447 	"rfs4_op_open_confirm",
448 	"rfs4_op_open_downgrade",
449 	"rfs4_op_putfh",
450 	"rfs4_op_putpubfh",
451 	"rfs4_op_putrootfh",
452 	"rfs4_op_read",
453 	"rfs4_op_readdir",
454 	"rfs4_op_readlink",
455 	"rfs4_op_remove",
456 	"rfs4_op_rename",
457 	"rfs4_op_renew",
458 	"rfs4_op_restorefh",
459 	"rfs4_op_savefh",
460 	"rfs4_op_secinfo",
461 	"rfs4_op_setattr",
462 	"rfs4_op_setclientid",
463 	"rfs4_op_setclient_confirm",
464 	"rfs4_op_verify",
465 	"rfs4_op_write",
466 	"rfs4_op_release_lockowner",
467 	"rfs4_op_illegal"
468 };
469 #endif
470 
471 void rfs4_ss_chkclid(rfs4_client_t *);
472 
473 #ifdef	nextdp
474 #undef nextdp
475 #endif
476 #define	nextdp(dp)	((struct dirent64 *)((char *)(dp) + (dp)->d_reclen))
477 
478 static const fs_operation_def_t nfs4_rd_deleg_tmpl[] = {
479 	VOPNAME_OPEN, deleg_rdopen,
480 	VOPNAME_WRITE, deleg_write,
481 	VOPNAME_SETATTR, deleg_setattr,
482 	VOPNAME_RWLOCK, deleg_rd_rwlock,
483 	VOPNAME_SPACE, deleg_space,
484 	VOPNAME_SETSECATTR, deleg_setsecattr,
485 	VOPNAME_VNEVENT, deleg_vnevent,
486 	NULL, NULL
487 };
488 static const fs_operation_def_t nfs4_wr_deleg_tmpl[] = {
489 	VOPNAME_OPEN, deleg_wropen,
490 	VOPNAME_READ, deleg_read,
491 	VOPNAME_WRITE, deleg_write,
492 	VOPNAME_SETATTR, deleg_setattr,
493 	VOPNAME_RWLOCK, deleg_wr_rwlock,
494 	VOPNAME_SPACE, deleg_space,
495 	VOPNAME_SETSECATTR, deleg_setsecattr,
496 	VOPNAME_VNEVENT, deleg_vnevent,
497 	NULL, NULL
498 };
499 
500 int
501 rfs4_srvrinit(void)
502 {
503 	timespec32_t verf;
504 	int error;
505 	extern void rfs4_attr_init();
506 	extern krwlock_t rfs4_deleg_policy_lock;
507 
508 	/*
509 	 * The following algorithm attempts to find a unique verifier
510 	 * to be used as the write verifier returned from the server
511 	 * to the client.  It is important that this verifier change
512 	 * whenever the server reboots.  Of secondary importance, it
513 	 * is important for the verifier to be unique between two
514 	 * different servers.
515 	 *
516 	 * Thus, an attempt is made to use the system hostid and the
517 	 * current time in seconds when the nfssrv kernel module is
518 	 * loaded.  It is assumed that an NFS server will not be able
519 	 * to boot and then to reboot in less than a second.  If the
520 	 * hostid has not been set, then the current high resolution
521 	 * time is used.  This will ensure different verifiers each
522 	 * time the server reboots and minimize the chances that two
523 	 * different servers will have the same verifier.
524 	 * XXX - this is broken on LP64 kernels.
525 	 */
526 	verf.tv_sec = (time_t)nfs_atoi(hw_serial);
527 	if (verf.tv_sec != 0) {
528 		verf.tv_nsec = gethrestime_sec();
529 	} else {
530 		timespec_t tverf;
531 
532 		gethrestime(&tverf);
533 		verf.tv_sec = (time_t)tverf.tv_sec;
534 		verf.tv_nsec = tverf.tv_nsec;
535 	}
536 
537 	Write4verf = *(uint64_t *)&verf;
538 
539 	rfs4_attr_init();
540 	mutex_init(&rfs4_deleg_lock, NULL, MUTEX_DEFAULT, NULL);
541 
542 	/* Used to manage create/destroy of server state */
543 	mutex_init(&rfs4_state_lock, NULL, MUTEX_DEFAULT, NULL);
544 
545 	/* Used to manage access to server instance linked list */
546 	mutex_init(&rfs4_servinst_lock, NULL, MUTEX_DEFAULT, NULL);
547 
548 	/* Used to manage access to rfs4_deleg_policy */
549 	rw_init(&rfs4_deleg_policy_lock, NULL, RW_DEFAULT, NULL);
550 
551 	error = fem_create("deleg_rdops", nfs4_rd_deleg_tmpl, &deleg_rdops);
552 	if (error != 0) {
553 		rfs4_disable_delegation();
554 	} else {
555 		error = fem_create("deleg_wrops", nfs4_wr_deleg_tmpl,
556 				&deleg_wrops);
557 		if (error != 0) {
558 			rfs4_disable_delegation();
559 			fem_free(deleg_rdops);
560 		}
561 	}
562 
563 	nfs4_srv_caller_id = fs_new_caller_id();
564 
565 	lockt_sysid = lm_alloc_sysidt();
566 
567 	return (0);
568 }
569 
570 void
571 rfs4_srvrfini(void)
572 {
573 	extern krwlock_t rfs4_deleg_policy_lock;
574 
575 	if (lockt_sysid != LM_NOSYSID) {
576 		lm_free_sysidt(lockt_sysid);
577 		lockt_sysid = LM_NOSYSID;
578 	}
579 
580 	mutex_destroy(&rfs4_deleg_lock);
581 	mutex_destroy(&rfs4_state_lock);
582 	rw_destroy(&rfs4_deleg_policy_lock);
583 
584 	fem_free(deleg_rdops);
585 	fem_free(deleg_wrops);
586 }
587 
588 void
589 rfs4_init_compound_state(struct compound_state *cs)
590 {
591 	bzero(cs, sizeof (*cs));
592 	cs->cont = TRUE;
593 	cs->access = CS_ACCESS_DENIED;
594 	cs->deleg = FALSE;
595 	cs->mandlock = FALSE;
596 	cs->fh.nfs_fh4_val = cs->fhbuf;
597 }
598 
599 void
600 rfs4_grace_start(rfs4_servinst_t *sip)
601 {
602 	time_t now = gethrestime_sec();
603 
604 	NFS4_DEBUG(rfs4_servinst_debug, (CE_NOTE,
605 	    "rfs4_grace_start: inst %p: 0x%lx", (void *)sip, now));
606 
607 	rw_enter(&sip->rwlock, RW_WRITER);
608 	sip->start_time = now;
609 	sip->grace_period = rfs4_grace_period;
610 	rw_exit(&sip->rwlock);
611 }
612 
613 /*
614  * returns true if the instance's grace period has never been started
615  */
616 int
617 rfs4_servinst_grace_new(rfs4_servinst_t *sip)
618 {
619 	time_t start_time;
620 
621 	rw_enter(&sip->rwlock, RW_READER);
622 	start_time = sip->start_time;
623 	rw_exit(&sip->rwlock);
624 
625 	return (start_time == 0);
626 }
627 
628 /*
629  * Indicates if server instance is within the
630  * grace period.
631  */
632 int
633 rfs4_servinst_in_grace(rfs4_servinst_t *sip)
634 {
635 	time_t grace_expiry;
636 
637 	rw_enter(&sip->rwlock, RW_READER);
638 	grace_expiry = sip->start_time + sip->grace_period;
639 	rw_exit(&sip->rwlock);
640 
641 	return (gethrestime_sec() < grace_expiry);
642 }
643 
644 int
645 rfs4_clnt_in_grace(rfs4_client_t *cp)
646 {
647 	ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0);
648 
649 	return (rfs4_servinst_in_grace(cp->server_instance));
650 }
651 
652 /*
653  * reset all currently active grace periods
654  */
655 void
656 rfs4_grace_reset_all(void)
657 {
658 #ifdef DEBUG
659 	int n = 0;
660 #endif
661 	rfs4_servinst_t *sip;
662 
663 	mutex_enter(&rfs4_servinst_lock);
664 	for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev) {
665 		if (rfs4_servinst_in_grace(sip)) {
666 			rfs4_grace_start(sip);
667 #ifdef DEBUG
668 			n++;
669 #endif
670 		}
671 	}
672 	mutex_exit(&rfs4_servinst_lock);
673 
674 	NFS4_DEBUG(rfs4_servinst_debug, (CE_NOTE,
675 	    "rfs4_grace_reset_all: reset %d instances", n));
676 }
677 
678 /*
679  * start any new instances' grace periods
680  */
681 void
682 rfs4_grace_start_new(void)
683 {
684 #ifdef DEBUG
685 	int n = 0;
686 #endif
687 	rfs4_servinst_t *sip;
688 
689 	mutex_enter(&rfs4_servinst_lock);
690 	for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev) {
691 		if (rfs4_servinst_grace_new(sip))
692 			rfs4_grace_start(sip);
693 #ifdef DEBUG
694 		n++;
695 #endif
696 	}
697 	mutex_exit(&rfs4_servinst_lock);
698 
699 	NFS4_DEBUG(rfs4_servinst_debug, (CE_NOTE,
700 	    "rfs4_grace_start_new: started %d new instances", n));
701 }
702 
703 /*
704  * Create a new server instance, and make it the currently active instance.
705  * Note that starting the grace period too early will reduce the clients'
706  * recovery window.
707  */
708 void
709 rfs4_servinst_create(int start_grace)
710 {
711 	rfs4_servinst_t *sip;
712 
713 	sip = kmem_alloc(sizeof (rfs4_servinst_t), KM_SLEEP);
714 	rw_init(&sip->rwlock, NULL, RW_DEFAULT, NULL);
715 
716 	sip->start_time = (time_t)0;
717 	sip->grace_period = (time_t)0;
718 	sip->next = NULL;
719 	sip->prev = NULL;
720 
721 	mutex_enter(&rfs4_servinst_lock);
722 	if (rfs4_cur_servinst == NULL) {
723 		NFS4_DEBUG(rfs4_servinst_debug, (CE_NOTE,
724 		    "rfs4_servinst_create: creating first instance"));
725 	} else {
726 		/* add to linked list */
727 		sip->prev = rfs4_cur_servinst;
728 		rfs4_cur_servinst->next = sip;
729 	}
730 	if (start_grace)
731 		rfs4_grace_start(sip);
732 	/* make the new instance "current" */
733 	rfs4_cur_servinst = sip;
734 	mutex_exit(&rfs4_servinst_lock);
735 
736 	NFS4_DEBUG(rfs4_servinst_debug, (CE_NOTE,
737 	    "rfs4_servinst_create: new current instance: %p; start_grace: %d",
738 	    (void *)sip, start_grace));
739 }
740 
741 /*
742  * In future, we might add a rfs4_servinst_destroy(sip) but, for now, destroy
743  * all instances directly.
744  */
745 void
746 rfs4_servinst_destroy_all(void)
747 {
748 	rfs4_servinst_t *sip, *prev, *current;
749 #ifdef DEBUG
750 	int n = 0;
751 #endif
752 
753 	mutex_enter(&rfs4_servinst_lock);
754 	ASSERT(rfs4_cur_servinst != NULL);
755 	current = rfs4_cur_servinst;
756 	rfs4_cur_servinst = NULL;
757 	for (sip = current; sip != NULL; sip = prev) {
758 		prev = sip->prev;
759 		rw_destroy(&sip->rwlock);
760 		kmem_free(sip, sizeof (rfs4_servinst_t));
761 #ifdef DEBUG
762 		n++;
763 #endif
764 	}
765 	mutex_exit(&rfs4_servinst_lock);
766 
767 	NFS4_DEBUG(rfs4_servinst_debug, (CE_NOTE,
768 	    "rfs4_servinst_destroy_all: destroyed %d instances", n));
769 }
770 
771 /*
772  * Assign the current server instance to a client_t.
773  * Should be called with cp->dbe held.
774  */
775 void
776 rfs4_servinst_assign(rfs4_client_t *cp, rfs4_servinst_t *sip)
777 {
778 	ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0);
779 
780 	NFS4_DEBUG(rfs4_servinst_debug, (CE_NOTE,
781 	    "rfs4_servinst_assign: client: %p, old: %p, new: %p", (void *)cp,
782 	    (void *)cp->server_instance, (void *)sip));
783 
784 	/*
785 	 * The lock ensures that if the current instance is in the process
786 	 * of changing, we will see the new one.
787 	 */
788 	mutex_enter(&rfs4_servinst_lock);
789 	cp->server_instance = sip;
790 	mutex_exit(&rfs4_servinst_lock);
791 }
792 
793 rfs4_servinst_t *
794 rfs4_servinst(rfs4_client_t *cp)
795 {
796 	ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0);
797 
798 	return (cp->server_instance);
799 }
800 
801 /* ARGSUSED */
802 static void
803 nullfree(caddr_t resop)
804 {
805 }
806 
807 /*
808  * This is a fall-through for invalid or not implemented (yet) ops
809  */
810 /* ARGSUSED */
811 static void
812 rfs4_op_inval(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
813 	struct compound_state *cs)
814 {
815 	*cs->statusp = *((nfsstat4 *)&(resop)->nfs_resop4_u) = NFS4ERR_INVAL;
816 }
817 
818 /*
819  * Check if the security flavor, nfsnum, is in the flavor_list.
820  */
821 bool_t
822 in_flavor_list(int nfsnum, int *flavor_list, int count)
823 {
824 	int i;
825 
826 	for (i = 0; i < count; i++) {
827 		if (nfsnum == flavor_list[i])
828 			return (TRUE);
829 	}
830 	return (FALSE);
831 }
832 
833 /*
834  * Used by rfs4_op_secinfo to get the security information from the
835  * export structure associated with the component.
836  */
837 /* ARGSUSED */
838 static nfsstat4
839 do_rfs4_op_secinfo(struct compound_state *cs, char *nm, SECINFO4res *resp)
840 {
841 	int error, different_export = 0;
842 	vnode_t *dvp, *vp, *tvp;
843 	struct exportinfo *exi = NULL;
844 	fid_t fid;
845 	uint_t count, i;
846 	secinfo4 *resok_val;
847 	struct secinfo *secp;
848 	bool_t did_traverse;
849 	int dotdot, walk;
850 
851 	dvp = cs->vp;
852 	dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0');
853 
854 	/*
855 	 * If dotdotting, then need to check whether it's above the
856 	 * root of a filesystem, or above an export point.
857 	 */
858 	if (dotdot) {
859 
860 		/*
861 		 * If dotdotting at the root of a filesystem, then
862 		 * need to traverse back to the mounted-on filesystem
863 		 * and do the dotdot lookup there.
864 		 */
865 		if (cs->vp->v_flag & VROOT) {
866 
867 			/*
868 			 * If at the system root, then can
869 			 * go up no further.
870 			 */
871 			if (VN_CMP(dvp, rootdir))
872 				return (puterrno4(ENOENT));
873 
874 			/*
875 			 * Traverse back to the mounted-on filesystem
876 			 */
877 			dvp = untraverse(cs->vp);
878 
879 			/*
880 			 * Set the different_export flag so we remember
881 			 * to pick up a new exportinfo entry for
882 			 * this new filesystem.
883 			 */
884 			different_export = 1;
885 		} else {
886 
887 			/*
888 			 * If dotdotting above an export point then set
889 			 * the different_export to get new export info.
890 			 */
891 			different_export = nfs_exported(cs->exi, cs->vp);
892 		}
893 	}
894 
895 	/*
896 	 * Get the vnode for the component "nm".
897 	 */
898 	error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cs->cr);
899 	if (error)
900 		return (puterrno4(error));
901 
902 	/*
903 	 * If the vnode is in a pseudo filesystem, or if the security flavor
904 	 * used in the request is valid but not an explicitly shared flavor,
905 	 * or the access bit indicates that this is a limited access,
906 	 * check whether this vnode is visible.
907 	 */
908 	if (!different_export &&
909 	    (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) ||
910 	    cs->access & CS_ACCESS_LIMITED)) {
911 		if (! nfs_visible(cs->exi, vp, &different_export)) {
912 			VN_RELE(vp);
913 			return (puterrno4(ENOENT));
914 		}
915 	}
916 
917 	/*
918 	 * If it's a mountpoint, then traverse it.
919 	 */
920 	if (vn_ismntpt(vp)) {
921 		tvp = vp;
922 		if ((error = traverse(&tvp)) != 0) {
923 			VN_RELE(vp);
924 			return (puterrno4(error));
925 		}
926 		/* remember that we had to traverse mountpoint */
927 		did_traverse = TRUE;
928 		vp = tvp;
929 		different_export = 1;
930 	} else if (vp->v_vfsp != dvp->v_vfsp) {
931 		/*
932 		 * If vp isn't a mountpoint and the vfs ptrs aren't the same,
933 		 * then vp is probably an LOFS object.  We don't need the
934 		 * realvp, we just need to know that we might have crossed
935 		 * a server fs boundary and need to call checkexport4.
936 		 * (LOFS lookup hides server fs mountpoints, and actually calls
937 		 * traverse)
938 		 */
939 		different_export = 1;
940 		did_traverse = FALSE;
941 	}
942 
943 	/*
944 	 * Get the export information for it.
945 	 */
946 	if (different_export) {
947 
948 		bzero(&fid, sizeof (fid));
949 		fid.fid_len = MAXFIDSZ;
950 		error = vop_fid_pseudo(vp, &fid);
951 		if (error) {
952 			VN_RELE(vp);
953 			return (puterrno4(error));
954 		}
955 
956 		if (dotdot)
957 			exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE);
958 		else
959 			exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
960 
961 		if (exi == NULL) {
962 			if (did_traverse == TRUE) {
963 				/*
964 				 * If this vnode is a mounted-on vnode,
965 				 * but the mounted-on file system is not
966 				 * exported, send back the secinfo for
967 				 * the exported node that the mounted-on
968 				 * vnode lives in.
969 				 */
970 				exi = cs->exi;
971 			} else {
972 				VN_RELE(vp);
973 				return (puterrno4(EACCES));
974 			}
975 		}
976 	} else {
977 		exi = cs->exi;
978 	}
979 	ASSERT(exi != NULL);
980 
981 
982 	/*
983 	 * Create the secinfo result based on the security information
984 	 * from the exportinfo structure (exi).
985 	 *
986 	 * Return all flavors for a pseudo node.
987 	 * For a real export node, return the flavor that the client
988 	 * has access with.
989 	 */
990 	ASSERT(RW_LOCK_HELD(&exported_lock));
991 	if (PSEUDO(exi)) {
992 		count = exi->exi_export.ex_seccnt; /* total sec count */
993 		resok_val = kmem_alloc(count * sizeof (secinfo4), KM_SLEEP);
994 		secp = exi->exi_export.ex_secinfo;
995 
996 		for (i = 0; i < count; i++) {
997 		    resok_val[i].flavor = secp[i].s_secinfo.sc_rpcnum;
998 		    if (resok_val[i].flavor == RPCSEC_GSS) {
999 			rpcsec_gss_info *info;
1000 
1001 			info = &resok_val[i].flavor_info;
1002 			info->qop = secp[i].s_secinfo.sc_qop;
1003 			info->service =
1004 				(rpc_gss_svc_t)secp[i].s_secinfo.sc_service;
1005 
1006 			/* get oid opaque data */
1007 			info->oid.sec_oid4_len =
1008 				secp[i].s_secinfo.sc_gss_mech_type->length;
1009 			info->oid.sec_oid4_val =
1010 				kmem_alloc(
1011 				    secp[i].s_secinfo.sc_gss_mech_type->length,
1012 				    KM_SLEEP);
1013 			bcopy(secp[i].s_secinfo.sc_gss_mech_type->elements,
1014 				info->oid.sec_oid4_val, info->oid.sec_oid4_len);
1015 		    }
1016 		}
1017 		resp->SECINFO4resok_len = count;
1018 		resp->SECINFO4resok_val = resok_val;
1019 	} else {
1020 		int ret_cnt = 0, k = 0;
1021 		int *flavor_list;
1022 
1023 		count = exi->exi_export.ex_seccnt; /* total sec count */
1024 		secp = exi->exi_export.ex_secinfo;
1025 
1026 		flavor_list = kmem_alloc(count * sizeof (int), KM_SLEEP);
1027 		/* find out which flavors to return */
1028 		for (i = 0; i < count; i ++) {
1029 			int access, flavor, perm;
1030 
1031 			flavor = secp[i].s_secinfo.sc_nfsnum;
1032 			perm = secp[i].s_flags;
1033 
1034 			access = nfsauth4_secinfo_access(exi, cs->req,
1035 						flavor, perm);
1036 
1037 			if (! (access & NFSAUTH_DENIED) &&
1038 			    ! (access & NFSAUTH_WRONGSEC)) {
1039 				flavor_list[ret_cnt] = flavor;
1040 				ret_cnt++;
1041 			}
1042 		}
1043 
1044 		/* Create the returning SECINFO value */
1045 		resok_val = kmem_alloc(ret_cnt * sizeof (secinfo4), KM_SLEEP);
1046 
1047 		for (i = 0; i < count; i++) {
1048 		/* If the flavor is in the flavor list, fill in resok_val. */
1049 		    if (in_flavor_list(secp[i].s_secinfo.sc_nfsnum,
1050 						flavor_list, ret_cnt)) {
1051 			resok_val[k].flavor = secp[i].s_secinfo.sc_rpcnum;
1052 			if (resok_val[k].flavor == RPCSEC_GSS) {
1053 			    rpcsec_gss_info *info;
1054 
1055 			    info = &resok_val[k].flavor_info;
1056 			    info->qop = secp[i].s_secinfo.sc_qop;
1057 			    info->service =
1058 				(rpc_gss_svc_t)secp[i].s_secinfo.sc_service;
1059 
1060 			    /* get oid opaque data */
1061 			    info->oid.sec_oid4_len =
1062 				secp[i].s_secinfo.sc_gss_mech_type->length;
1063 			    info->oid.sec_oid4_val =
1064 				kmem_alloc(
1065 				    secp[i].s_secinfo.sc_gss_mech_type->length,
1066 				    KM_SLEEP);
1067 			    bcopy(secp[i].s_secinfo.sc_gss_mech_type->elements,
1068 				info->oid.sec_oid4_val, info->oid.sec_oid4_len);
1069 			}
1070 			k++;
1071 		    }
1072 		    if (k >= ret_cnt)
1073 			break;
1074 		}
1075 		resp->SECINFO4resok_len = ret_cnt;
1076 		resp->SECINFO4resok_val = resok_val;
1077 		kmem_free(flavor_list, count * sizeof (int));
1078 	}
1079 
1080 	VN_RELE(vp);
1081 	return (NFS4_OK);
1082 }
1083 
1084 /*
1085  * SECINFO (Operation 33): Obtain required security information on
1086  * the component name in the format of (security-mechanism-oid, qop, service)
1087  * triplets.
1088  */
1089 /* ARGSUSED */
1090 static void
1091 rfs4_op_secinfo(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1092 	struct compound_state *cs)
1093 {
1094 	SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo;
1095 	utf8string *utfnm = &argop->nfs_argop4_u.opsecinfo.name;
1096 	uint_t len;
1097 	char *nm;
1098 
1099 	/*
1100 	 * Current file handle (cfh) should have been set before getting
1101 	 * into this function. If not, return error.
1102 	 */
1103 	if (cs->vp == NULL) {
1104 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1105 		return;
1106 	}
1107 
1108 	if (cs->vp->v_type != VDIR) {
1109 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
1110 		return;
1111 	}
1112 
1113 	/*
1114 	 * Verify the component name. If failed, error out, but
1115 	 * do not error out if the component name is a "..".
1116 	 * SECINFO will return its parents secinfo data for SECINFO "..".
1117 	 */
1118 	if (!utf8_dir_verify(utfnm)) {
1119 		if (utfnm->utf8string_len != 2 ||
1120 				utfnm->utf8string_val[0] != '.' ||
1121 				utfnm->utf8string_val[1] != '.') {
1122 			*cs->statusp = resp->status = NFS4ERR_INVAL;
1123 			return;
1124 		}
1125 	}
1126 
1127 	nm = utf8_to_str(utfnm, &len, NULL);
1128 	if (nm == NULL) {
1129 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1130 		return;
1131 	}
1132 
1133 	if (len > MAXNAMELEN) {
1134 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1135 		kmem_free(nm, len);
1136 		return;
1137 	}
1138 
1139 	*cs->statusp = resp->status = do_rfs4_op_secinfo(cs, nm, resp);
1140 
1141 	kmem_free(nm, len);
1142 }
1143 
1144 /*
1145  * Free SECINFO result.
1146  */
1147 /* ARGSUSED */
1148 static void
1149 rfs4_op_secinfo_free(nfs_resop4 *resop)
1150 {
1151 	SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo;
1152 	int count, i;
1153 	secinfo4 *resok_val;
1154 
1155 	/* If this is not an Ok result, nothing to free. */
1156 	if (resp->status != NFS4_OK) {
1157 		return;
1158 	}
1159 
1160 	count = resp->SECINFO4resok_len;
1161 	resok_val = resp->SECINFO4resok_val;
1162 
1163 	for (i = 0; i < count; i++) {
1164 	    if (resok_val[i].flavor == RPCSEC_GSS) {
1165 		rpcsec_gss_info *info;
1166 
1167 		info = &resok_val[i].flavor_info;
1168 		kmem_free(info->oid.sec_oid4_val, info->oid.sec_oid4_len);
1169 	    }
1170 	}
1171 	kmem_free(resok_val, count * sizeof (secinfo4));
1172 	resp->SECINFO4resok_len = 0;
1173 	resp->SECINFO4resok_val = NULL;
1174 }
1175 
1176 /*
1177  * do label check on client label and server's file lable.
1178  */
1179 static boolean_t
1180 do_rfs4_label_check(bslabel_t *clabel, vnode_t *vp, int flag)
1181 {
1182 	bslabel_t *slabel;
1183 	ts_label_t *tslabel;
1184 	boolean_t result;
1185 
1186 	if ((tslabel = nfs4_getflabel(vp)) == NULL) {
1187 		return (B_FALSE);
1188 	}
1189 	slabel = label2bslabel(tslabel);
1190 	DTRACE_PROBE4(tx__rfs4__log__info__labelcheck, char *,
1191 	    "comparing server's file label(1) with client label(2) (vp(3))",
1192 	    bslabel_t *, slabel, bslabel_t *, clabel, vnode_t *, vp);
1193 
1194 	if (flag == EQUALITY_CHECK)
1195 		result = blequal(clabel, slabel);
1196 	else
1197 		result = bldominates(clabel, slabel);
1198 	label_rele(tslabel);
1199 	return (result);
1200 }
1201 
1202 /* ARGSUSED */
1203 static void
1204 rfs4_op_access(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1205 	struct compound_state *cs)
1206 {
1207 	ACCESS4args *args = &argop->nfs_argop4_u.opaccess;
1208 	ACCESS4res *resp = &resop->nfs_resop4_u.opaccess;
1209 	int error;
1210 	vnode_t *vp;
1211 	struct vattr va;
1212 	int checkwriteperm;
1213 	cred_t *cr = cs->cr;
1214 	bslabel_t *clabel, *slabel;
1215 	ts_label_t *tslabel;
1216 	boolean_t admin_low_client;
1217 
1218 #if 0	/* XXX allow access even if !cs->access. Eventually only pseudo fs */
1219 	if (cs->access == CS_ACCESS_DENIED) {
1220 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1221 		return;
1222 	}
1223 #endif
1224 	if (cs->vp == NULL) {
1225 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1226 		return;
1227 	}
1228 
1229 	ASSERT(cr != NULL);
1230 
1231 	vp = cs->vp;
1232 
1233 	/*
1234 	 * If the file system is exported read only, it is not appropriate
1235 	 * to check write permissions for regular files and directories.
1236 	 * Special files are interpreted by the client, so the underlying
1237 	 * permissions are sent back to the client for interpretation.
1238 	 */
1239 	if (rdonly4(cs->exi, cs->vp, req) &&
1240 		(vp->v_type == VREG || vp->v_type == VDIR))
1241 		checkwriteperm = 0;
1242 	else
1243 		checkwriteperm = 1;
1244 
1245 	/*
1246 	 * XXX
1247 	 * We need the mode so that we can correctly determine access
1248 	 * permissions relative to a mandatory lock file.  Access to
1249 	 * mandatory lock files is denied on the server, so it might
1250 	 * as well be reflected to the server during the open.
1251 	 */
1252 	va.va_mask = AT_MODE;
1253 	error = VOP_GETATTR(vp, &va, 0, cr);
1254 	if (error) {
1255 		*cs->statusp = resp->status = puterrno4(error);
1256 		return;
1257 	}
1258 	resp->access = 0;
1259 	resp->supported = 0;
1260 
1261 	if (is_system_labeled()) {
1262 		ASSERT(req->rq_label != NULL);
1263 		clabel = req->rq_label;
1264 		DTRACE_PROBE2(tx__rfs4__log__info__opaccess__clabel, char *,
1265 		    "got client label from request(1)",
1266 		    struct svc_req *, req);
1267 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
1268 			if ((tslabel = nfs4_getflabel(vp)) == NULL) {
1269 				*cs->statusp = resp->status = puterrno4(EACCES);
1270 				return;
1271 			}
1272 			slabel = label2bslabel(tslabel);
1273 			DTRACE_PROBE3(tx__rfs4__log__info__opaccess__slabel,
1274 			    char *, "got server label(1) for vp(2)",
1275 			    bslabel_t *, slabel, vnode_t *, vp);
1276 
1277 			admin_low_client = B_FALSE;
1278 		} else
1279 			admin_low_client = B_TRUE;
1280 	}
1281 
1282 	if (args->access & ACCESS4_READ) {
1283 		error = VOP_ACCESS(vp, VREAD, 0, cr);
1284 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1285 		    (!is_system_labeled() || admin_low_client ||
1286 		    bldominates(clabel, slabel)))
1287 			resp->access |= ACCESS4_READ;
1288 		resp->supported |= ACCESS4_READ;
1289 	}
1290 	if ((args->access & ACCESS4_LOOKUP) && vp->v_type == VDIR) {
1291 		error = VOP_ACCESS(vp, VEXEC, 0, cr);
1292 		if (!error && (!is_system_labeled() || admin_low_client ||
1293 		    bldominates(clabel, slabel)))
1294 			resp->access |= ACCESS4_LOOKUP;
1295 		resp->supported |= ACCESS4_LOOKUP;
1296 	}
1297 	if (checkwriteperm &&
1298 	    (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND))) {
1299 		error = VOP_ACCESS(vp, VWRITE, 0, cr);
1300 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1301 		    (!is_system_labeled() || admin_low_client ||
1302 		    blequal(clabel, slabel)))
1303 			resp->access |=
1304 			    (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND));
1305 		resp->supported |= (ACCESS4_MODIFY|ACCESS4_EXTEND);
1306 	}
1307 
1308 	if (checkwriteperm &&
1309 	    (args->access & ACCESS4_DELETE) && vp->v_type == VDIR) {
1310 		error = VOP_ACCESS(vp, VWRITE, 0, cr);
1311 		if (!error && (!is_system_labeled() || admin_low_client ||
1312 		    blequal(clabel, slabel)))
1313 			resp->access |= ACCESS4_DELETE;
1314 		resp->supported |= ACCESS4_DELETE;
1315 	}
1316 	if (args->access & ACCESS4_EXECUTE && vp->v_type != VDIR) {
1317 		error = VOP_ACCESS(vp, VEXEC, 0, cr);
1318 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1319 		    (!is_system_labeled() || admin_low_client ||
1320 		    bldominates(clabel, slabel)))
1321 			resp->access |= ACCESS4_EXECUTE;
1322 		resp->supported |= ACCESS4_EXECUTE;
1323 	}
1324 
1325 	if (is_system_labeled() && !admin_low_client)
1326 		label_rele(tslabel);
1327 
1328 	*cs->statusp = resp->status = NFS4_OK;
1329 }
1330 
1331 /* ARGSUSED */
1332 static void
1333 rfs4_op_commit(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1334 	struct compound_state *cs)
1335 {
1336 	COMMIT4args *args = &argop->nfs_argop4_u.opcommit;
1337 	COMMIT4res *resp = &resop->nfs_resop4_u.opcommit;
1338 	int error;
1339 	vnode_t *vp = cs->vp;
1340 	cred_t *cr = cs->cr;
1341 	vattr_t va;
1342 
1343 	if (vp == NULL) {
1344 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1345 		return;
1346 	}
1347 	if (cs->access == CS_ACCESS_DENIED) {
1348 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1349 		return;
1350 	}
1351 
1352 	if (args->offset + args->count < args->offset) {
1353 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1354 		return;
1355 	}
1356 
1357 	va.va_mask = AT_UID;
1358 	error = VOP_GETATTR(vp, &va, 0, cr);
1359 
1360 	/*
1361 	 * If we can't get the attributes, then we can't do the
1362 	 * right access checking.  So, we'll fail the request.
1363 	 */
1364 	if (error) {
1365 		*cs->statusp = resp->status = puterrno4(error);
1366 		return;
1367 	}
1368 	if (rdonly4(cs->exi, cs->vp, req)) {
1369 		*cs->statusp = resp->status = NFS4ERR_ROFS;
1370 		return;
1371 	}
1372 
1373 	if (vp->v_type != VREG) {
1374 		if (vp->v_type == VDIR)
1375 			resp->status = NFS4ERR_ISDIR;
1376 		else
1377 			resp->status = NFS4ERR_INVAL;
1378 		*cs->statusp = resp->status;
1379 		return;
1380 	}
1381 
1382 	if (crgetuid(cr) != va.va_uid &&
1383 	    (error = VOP_ACCESS(vp, VWRITE, 0, cs->cr))) {
1384 		*cs->statusp = resp->status = puterrno4(error);
1385 		return;
1386 	}
1387 
1388 	error = VOP_PUTPAGE(vp, args->offset, args->count, 0, cr);
1389 	if (!error)
1390 		error = VOP_FSYNC(vp, FNODSYNC, cr);
1391 
1392 	if (error) {
1393 		*cs->statusp = resp->status = puterrno4(error);
1394 		return;
1395 	}
1396 
1397 	*cs->statusp = resp->status = NFS4_OK;
1398 	resp->writeverf = Write4verf;
1399 }
1400 
1401 /*
1402  * rfs4_op_mknod is called from rfs4_op_create after all initial verification
1403  * was completed. It does the nfsv4 create for special files.
1404  */
1405 /* ARGSUSED */
1406 static vnode_t *
1407 do_rfs4_op_mknod(CREATE4args *args, CREATE4res *resp, struct svc_req *req,
1408 	struct compound_state *cs, vattr_t *vap, char *nm)
1409 {
1410 	int error;
1411 	cred_t *cr = cs->cr;
1412 	vnode_t *dvp = cs->vp;
1413 	vnode_t *vp = NULL;
1414 	int mode;
1415 	enum vcexcl excl;
1416 
1417 	switch (args->type) {
1418 	case NF4CHR:
1419 	case NF4BLK:
1420 		if (secpolicy_sys_devices(cr) != 0) {
1421 			*cs->statusp = resp->status = NFS4ERR_PERM;
1422 			return (NULL);
1423 		}
1424 		if (args->type == NF4CHR)
1425 			vap->va_type = VCHR;
1426 		else
1427 			vap->va_type = VBLK;
1428 		vap->va_rdev = makedevice(args->ftype4_u.devdata.specdata1,
1429 					args->ftype4_u.devdata.specdata2);
1430 		vap->va_mask |= AT_RDEV;
1431 		break;
1432 	case NF4SOCK:
1433 		vap->va_type = VSOCK;
1434 		break;
1435 	case NF4FIFO:
1436 		vap->va_type = VFIFO;
1437 		break;
1438 	default:
1439 		*cs->statusp = resp->status = NFS4ERR_BADTYPE;
1440 		return (NULL);
1441 	}
1442 
1443 	/*
1444 	 * Must specify the mode.
1445 	 */
1446 	if (!(vap->va_mask & AT_MODE)) {
1447 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1448 		return (NULL);
1449 	}
1450 
1451 	excl = EXCL;
1452 
1453 	mode = 0;
1454 
1455 	error = VOP_CREATE(dvp, nm, vap, excl, mode, &vp, cr, 0);
1456 	if (error) {
1457 		*cs->statusp = resp->status = puterrno4(error);
1458 		return (NULL);
1459 	}
1460 	return (vp);
1461 }
1462 
1463 /*
1464  * nfsv4 create is used to create non-regular files. For regular files,
1465  * use nfsv4 open.
1466  */
1467 /* ARGSUSED */
1468 static void
1469 rfs4_op_create(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1470 	struct compound_state *cs)
1471 {
1472 	CREATE4args *args = &argop->nfs_argop4_u.opcreate;
1473 	CREATE4res *resp = &resop->nfs_resop4_u.opcreate;
1474 	int error;
1475 	struct vattr bva, iva, iva2, ava, *vap;
1476 	cred_t *cr = cs->cr;
1477 	vnode_t *dvp = cs->vp;
1478 	vnode_t *vp = NULL;
1479 	char *nm, *lnm;
1480 	uint_t len, llen;
1481 	int syncval = 0;
1482 	struct nfs4_svgetit_arg sarg;
1483 	struct nfs4_ntov_table ntov;
1484 	struct statvfs64 sb;
1485 	nfsstat4 status;
1486 
1487 	resp->attrset = 0;
1488 
1489 	if (dvp == NULL) {
1490 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1491 		return;
1492 	}
1493 
1494 	/*
1495 	 * If there is an unshared filesystem mounted on this vnode,
1496 	 * do not allow to create an object in this directory.
1497 	 */
1498 	if (vn_ismntpt(dvp)) {
1499 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1500 		return;
1501 	}
1502 
1503 	/* Verify that type is correct */
1504 	switch (args->type) {
1505 	case NF4LNK:
1506 	case NF4BLK:
1507 	case NF4CHR:
1508 	case NF4SOCK:
1509 	case NF4FIFO:
1510 	case NF4DIR:
1511 		break;
1512 	default:
1513 		*cs->statusp = resp->status = NFS4ERR_BADTYPE;
1514 		return;
1515 	};
1516 
1517 	if (cs->access == CS_ACCESS_DENIED) {
1518 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1519 		return;
1520 	}
1521 	if (dvp->v_type != VDIR) {
1522 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
1523 		return;
1524 	}
1525 	if (!utf8_dir_verify(&args->objname)) {
1526 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1527 		return;
1528 	}
1529 
1530 	if (rdonly4(cs->exi, cs->vp, req)) {
1531 		*cs->statusp = resp->status = NFS4ERR_ROFS;
1532 		return;
1533 	}
1534 
1535 	/*
1536 	 * Name of newly created object
1537 	 */
1538 	nm = utf8_to_fn(&args->objname, &len, NULL);
1539 	if (nm == NULL) {
1540 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1541 		return;
1542 	}
1543 
1544 	if (len > MAXNAMELEN) {
1545 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1546 		kmem_free(nm, len);
1547 		return;
1548 	}
1549 
1550 	resp->attrset = 0;
1551 
1552 	sarg.sbp = &sb;
1553 	nfs4_ntov_table_init(&ntov);
1554 
1555 	status = do_rfs4_set_attrs(&resp->attrset,
1556 					&args->createattrs, cs, &sarg,
1557 					&ntov, NFS4ATTR_SETIT);
1558 
1559 	if (sarg.vap->va_mask == 0 && status == NFS4_OK)
1560 		status = NFS4ERR_INVAL;
1561 
1562 	if (status != NFS4_OK) {
1563 		*cs->statusp = resp->status = status;
1564 		kmem_free(nm, len);
1565 		nfs4_ntov_table_free(&ntov, &sarg);
1566 		resp->attrset = 0;
1567 		return;
1568 	}
1569 
1570 	/* Get "before" change value */
1571 	bva.va_mask = AT_CTIME|AT_SEQ;
1572 	error = VOP_GETATTR(dvp, &bva, 0, cr);
1573 	if (error) {
1574 		*cs->statusp = resp->status = puterrno4(error);
1575 		kmem_free(nm, len);
1576 		nfs4_ntov_table_free(&ntov, &sarg);
1577 		resp->attrset = 0;
1578 		return;
1579 	}
1580 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bva.va_ctime)
1581 
1582 	vap = sarg.vap;
1583 
1584 	/*
1585 	 * Set default initial values for attributes when not specified
1586 	 * in createattrs.
1587 	 */
1588 	if ((vap->va_mask & AT_UID) == 0) {
1589 		vap->va_uid = crgetuid(cr);
1590 		vap->va_mask |= AT_UID;
1591 	}
1592 	if ((vap->va_mask & AT_GID) == 0) {
1593 		vap->va_gid = crgetgid(cr);
1594 		vap->va_mask |= AT_GID;
1595 	}
1596 
1597 	vap->va_mask |= AT_TYPE;
1598 	switch (args->type) {
1599 	case NF4DIR:
1600 		vap->va_type = VDIR;
1601 		if ((vap->va_mask & AT_MODE) == 0) {
1602 			vap->va_mode = 0700;	/* default: owner rwx only */
1603 			vap->va_mask |= AT_MODE;
1604 		}
1605 		error = VOP_MKDIR(dvp, nm, vap, &vp, cr);
1606 		if (error)
1607 			break;
1608 
1609 		/*
1610 		 * Get the initial "after" sequence number, if it fails,
1611 		 * set to zero
1612 		 */
1613 		iva.va_mask = AT_SEQ;
1614 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr))
1615 			iva.va_seq = 0;
1616 		break;
1617 	case NF4LNK:
1618 		vap->va_type = VLNK;
1619 		if ((vap->va_mask & AT_MODE) == 0) {
1620 			vap->va_mode = 0700;	/* default: owner rwx only */
1621 			vap->va_mask |= AT_MODE;
1622 		}
1623 
1624 		/*
1625 		 * symlink names must be treated as data
1626 		 */
1627 		lnm = utf8_to_str(&args->ftype4_u.linkdata, &llen, NULL);
1628 
1629 		if (lnm == NULL) {
1630 			*cs->statusp = resp->status = NFS4ERR_INVAL;
1631 			kmem_free(nm, len);
1632 			nfs4_ntov_table_free(&ntov, &sarg);
1633 			resp->attrset = 0;
1634 			return;
1635 		}
1636 
1637 		if (llen > MAXPATHLEN) {
1638 			*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1639 			kmem_free(nm, len);
1640 			kmem_free(lnm, llen);
1641 			nfs4_ntov_table_free(&ntov, &sarg);
1642 			resp->attrset = 0;
1643 			return;
1644 		}
1645 
1646 		error = VOP_SYMLINK(dvp, nm, vap, lnm, cr);
1647 		if (lnm != NULL)
1648 			kmem_free(lnm, llen);
1649 		if (error)
1650 			break;
1651 
1652 		/*
1653 		 * Get the initial "after" sequence number, if it fails,
1654 		 * set to zero
1655 		 */
1656 		iva.va_mask = AT_SEQ;
1657 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr))
1658 			iva.va_seq = 0;
1659 
1660 		error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr);
1661 		if (error)
1662 			break;
1663 
1664 		/*
1665 		 * va_seq is not safe over VOP calls, check it again
1666 		 * if it has changed zero out iva to force atomic = FALSE.
1667 		 */
1668 		iva2.va_mask = AT_SEQ;
1669 		if (VOP_GETATTR(dvp, &iva2, 0, cs->cr) ||
1670 						iva2.va_seq != iva.va_seq)
1671 			iva.va_seq = 0;
1672 		break;
1673 	default:
1674 		/*
1675 		 * probably a special file.
1676 		 */
1677 		if ((vap->va_mask & AT_MODE) == 0) {
1678 			vap->va_mode = 0600;	/* default: owner rw only */
1679 			vap->va_mask |= AT_MODE;
1680 		}
1681 		syncval = FNODSYNC;
1682 		/*
1683 		 * We know this will only generate one VOP call
1684 		 */
1685 		vp = do_rfs4_op_mknod(args, resp, req, cs, vap, nm);
1686 
1687 		if (vp == NULL) {
1688 			kmem_free(nm, len);
1689 			nfs4_ntov_table_free(&ntov, &sarg);
1690 			resp->attrset = 0;
1691 			return;
1692 		}
1693 
1694 		/*
1695 		 * Get the initial "after" sequence number, if it fails,
1696 		 * set to zero
1697 		 */
1698 		iva.va_mask = AT_SEQ;
1699 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr))
1700 			iva.va_seq = 0;
1701 
1702 		break;
1703 	}
1704 	kmem_free(nm, len);
1705 
1706 	if (error) {
1707 		*cs->statusp = resp->status = puterrno4(error);
1708 	}
1709 
1710 	/*
1711 	 * Force modified data and metadata out to stable storage.
1712 	 */
1713 	(void) VOP_FSYNC(dvp, 0, cr);
1714 
1715 	if (resp->status != NFS4_OK) {
1716 		if (vp != NULL)
1717 			VN_RELE(vp);
1718 		nfs4_ntov_table_free(&ntov, &sarg);
1719 		resp->attrset = 0;
1720 		return;
1721 	}
1722 
1723 	/*
1724 	 * Finish setup of cinfo response, "before" value already set.
1725 	 * Get "after" change value, if it fails, simply return the
1726 	 * before value.
1727 	 */
1728 	ava.va_mask = AT_CTIME|AT_SEQ;
1729 	if (VOP_GETATTR(dvp, &ava, 0, cr)) {
1730 		ava.va_ctime = bva.va_ctime;
1731 		ava.va_seq = 0;
1732 	}
1733 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, ava.va_ctime);
1734 
1735 	/*
1736 	 * True verification that object was created with correct
1737 	 * attrs is impossible.  The attrs could have been changed
1738 	 * immediately after object creation.  If attributes did
1739 	 * not verify, the only recourse for the server is to
1740 	 * destroy the object.  Maybe if some attrs (like gid)
1741 	 * are set incorrectly, the object should be destroyed;
1742 	 * however, seems bad as a default policy.  Do we really
1743 	 * want to destroy an object over one of the times not
1744 	 * verifying correctly?  For these reasons, the server
1745 	 * currently sets bits in attrset for createattrs
1746 	 * that were set; however, no verification is done.
1747 	 *
1748 	 * vmask_to_nmask accounts for vattr bits set on create
1749 	 *	[do_rfs4_set_attrs() only sets resp bits for
1750 	 *	 non-vattr/vfs bits.]
1751 	 * Mask off any bits set by default so as not to return
1752 	 * more attrset bits than were requested in createattrs
1753 	 */
1754 	nfs4_vmask_to_nmask(sarg.vap->va_mask, &resp->attrset);
1755 	resp->attrset &= args->createattrs.attrmask;
1756 	nfs4_ntov_table_free(&ntov, &sarg);
1757 
1758 	error = makefh4(&cs->fh, vp, cs->exi);
1759 	if (error) {
1760 		*cs->statusp = resp->status = puterrno4(error);
1761 	}
1762 
1763 	/*
1764 	 * The cinfo.atomic = TRUE only if we got no errors, we have
1765 	 * non-zero va_seq's, and it has incremented by exactly one
1766 	 * during the creation and it didn't change during the VOP_LOOKUP
1767 	 * or VOP_FSYNC.
1768 	 */
1769 	if (!error && bva.va_seq && iva.va_seq && ava.va_seq &&
1770 			iva.va_seq == (bva.va_seq + 1) &&
1771 			iva.va_seq == ava.va_seq)
1772 		resp->cinfo.atomic = TRUE;
1773 	else
1774 		resp->cinfo.atomic = FALSE;
1775 
1776 	(void) VOP_FSYNC(vp, syncval, cr);
1777 
1778 	if (resp->status != NFS4_OK) {
1779 		VN_RELE(vp);
1780 		return;
1781 	}
1782 	if (cs->vp)
1783 		VN_RELE(cs->vp);
1784 
1785 	cs->vp = vp;
1786 	*cs->statusp = resp->status = NFS4_OK;
1787 }
1788 
1789 
1790 /*ARGSUSED*/
1791 static void
1792 rfs4_op_delegreturn(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1793 	struct compound_state *cs)
1794 {
1795 	DELEGRETURN4args *args = &argop->nfs_argop4_u.opdelegreturn;
1796 	DELEGRETURN4res *resp = &resop->nfs_resop4_u.opdelegreturn;
1797 	rfs4_deleg_state_t *dsp;
1798 	nfsstat4 status;
1799 
1800 	status = rfs4_get_deleg_state(&args->deleg_stateid, &dsp);
1801 	resp->status = *cs->statusp = status;
1802 	if (status != NFS4_OK)
1803 		return;
1804 
1805 	/* Ensure specified filehandle matches */
1806 	if (cs->vp != dsp->finfo->vp) {
1807 		resp->status = *cs->statusp = NFS4ERR_BAD_STATEID;
1808 	} else
1809 		rfs4_return_deleg(dsp, FALSE);
1810 
1811 	rfs4_update_lease(dsp->client);
1812 
1813 	rfs4_deleg_state_rele(dsp);
1814 }
1815 
1816 /*
1817  * Check to see if a given "flavor" is an explicitly shared flavor.
1818  * The assumption of this routine is the "flavor" is already a valid
1819  * flavor in the secinfo list of "exi".
1820  *
1821  *	e.g.
1822  *		# share -o sec=flavor1 /export
1823  *		# share -o sec=flavor2 /export/home
1824  *
1825  *		flavor2 is not an explicitly shared flavor for /export,
1826  *		however it is in the secinfo list for /export thru the
1827  *		server namespace setup.
1828  */
1829 int
1830 is_exported_sec(int flavor, struct exportinfo *exi)
1831 {
1832 	int	i;
1833 	struct secinfo *sp;
1834 
1835 	sp = exi->exi_export.ex_secinfo;
1836 	for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
1837 		if (flavor == sp[i].s_secinfo.sc_nfsnum ||
1838 		    sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) {
1839 			return (SEC_REF_EXPORTED(&sp[i]));
1840 		}
1841 	}
1842 
1843 	/* Should not reach this point based on the assumption */
1844 	return (0);
1845 }
1846 
1847 /*
1848  * Check if the security flavor used in the request matches what is
1849  * required at the export point or at the root pseudo node (exi_root).
1850  *
1851  * returns 1 if there's a match or if exported with AUTH_NONE; 0 otherwise.
1852  *
1853  */
1854 static int
1855 secinfo_match_or_authnone(struct compound_state *cs)
1856 {
1857 	int	i;
1858 	struct secinfo *sp;
1859 
1860 	/*
1861 	 * Check cs->nfsflavor (from the request) against
1862 	 * the current export data in cs->exi.
1863 	 */
1864 	sp = cs->exi->exi_export.ex_secinfo;
1865 	for (i = 0; i < cs->exi->exi_export.ex_seccnt; i++) {
1866 		if (cs->nfsflavor == sp[i].s_secinfo.sc_nfsnum ||
1867 		    sp[i].s_secinfo.sc_nfsnum == AUTH_NONE)
1868 			return (1);
1869 	}
1870 
1871 	return (0);
1872 }
1873 
1874 /*
1875  * Check the access authority for the client and return the correct error.
1876  */
1877 nfsstat4
1878 call_checkauth4(struct compound_state *cs, struct svc_req *req)
1879 {
1880 	int	authres;
1881 
1882 	/*
1883 	 * First, check if the security flavor used in the request
1884 	 * are among the flavors set in the server namespace.
1885 	 */
1886 	if (!secinfo_match_or_authnone(cs)) {
1887 		*cs->statusp = NFS4ERR_WRONGSEC;
1888 		return (*cs->statusp);
1889 	}
1890 
1891 	authres = checkauth4(cs, req);
1892 
1893 	if (authres > 0) {
1894 		*cs->statusp = NFS4_OK;
1895 		if (! (cs->access & CS_ACCESS_LIMITED))
1896 			cs->access = CS_ACCESS_OK;
1897 	} else if (authres == 0) {
1898 		*cs->statusp = NFS4ERR_ACCESS;
1899 	} else if (authres == -2) {
1900 		*cs->statusp = NFS4ERR_WRONGSEC;
1901 	} else {
1902 		*cs->statusp = NFS4ERR_DELAY;
1903 	}
1904 	return (*cs->statusp);
1905 }
1906 
1907 /*
1908  * bitmap4_to_attrmask is called by getattr and readdir.
1909  * It sets up the vattr mask and determines whether vfsstat call is needed
1910  * based on the input bitmap.
1911  * Returns nfsv4 status.
1912  */
1913 static nfsstat4
1914 bitmap4_to_attrmask(bitmap4 breq, struct nfs4_svgetit_arg *sargp)
1915 {
1916 	int i;
1917 	uint_t	va_mask;
1918 	struct statvfs64 *sbp = sargp->sbp;
1919 
1920 	sargp->sbp = NULL;
1921 	sargp->flag = 0;
1922 	sargp->rdattr_error = NFS4_OK;
1923 	sargp->mntdfid_set = FALSE;
1924 	if (sargp->cs->vp)
1925 		sargp->xattr = get_fh4_flag(&sargp->cs->fh,
1926 					    FH4_ATTRDIR | FH4_NAMEDATTR);
1927 	else
1928 		sargp->xattr = 0;
1929 
1930 	/*
1931 	 * Set rdattr_error_req to true if return error per
1932 	 * failed entry rather than fail the readdir.
1933 	 */
1934 	if (breq & FATTR4_RDATTR_ERROR_MASK)
1935 		sargp->rdattr_error_req = 1;
1936 	else
1937 		sargp->rdattr_error_req = 0;
1938 
1939 	/*
1940 	 * generate the va_mask
1941 	 * Handle the easy cases first
1942 	 */
1943 	switch (breq) {
1944 	case NFS4_NTOV_ATTR_MASK:
1945 		sargp->vap->va_mask = NFS4_NTOV_ATTR_AT_MASK;
1946 		return (NFS4_OK);
1947 
1948 	case NFS4_FS_ATTR_MASK:
1949 		sargp->vap->va_mask = NFS4_FS_ATTR_AT_MASK;
1950 		sargp->sbp = sbp;
1951 		return (NFS4_OK);
1952 
1953 	case NFS4_NTOV_ATTR_CACHE_MASK:
1954 		sargp->vap->va_mask = NFS4_NTOV_ATTR_CACHE_AT_MASK;
1955 		return (NFS4_OK);
1956 
1957 	case FATTR4_LEASE_TIME_MASK:
1958 		sargp->vap->va_mask = 0;
1959 		return (NFS4_OK);
1960 
1961 	default:
1962 		va_mask = 0;
1963 		for (i = 0; i < nfs4_ntov_map_size; i++) {
1964 			if ((breq & nfs4_ntov_map[i].fbit) &&
1965 							nfs4_ntov_map[i].vbit)
1966 				va_mask |= nfs4_ntov_map[i].vbit;
1967 		}
1968 
1969 		/*
1970 		 * Check is vfsstat is needed
1971 		 */
1972 		if (breq & NFS4_FS_ATTR_MASK)
1973 			sargp->sbp = sbp;
1974 
1975 		sargp->vap->va_mask = va_mask;
1976 		return (NFS4_OK);
1977 	}
1978 	/* NOTREACHED */
1979 }
1980 
1981 /*
1982  * bitmap4_get_sysattrs is called by getattr and readdir.
1983  * It calls both VOP_GETATTR and VFS_STATVFS calls to get the attrs.
1984  * Returns nfsv4 status.
1985  */
1986 static nfsstat4
1987 bitmap4_get_sysattrs(struct nfs4_svgetit_arg *sargp)
1988 {
1989 	int error;
1990 	struct compound_state *cs = sargp->cs;
1991 	vnode_t *vp = cs->vp;
1992 
1993 	if (sargp->sbp != NULL) {
1994 		if (error = VFS_STATVFS(vp->v_vfsp, sargp->sbp)) {
1995 			sargp->sbp = NULL;	/* to identify error */
1996 			return (puterrno4(error));
1997 		}
1998 	}
1999 
2000 	return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr));
2001 }
2002 
2003 static void
2004 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp)
2005 {
2006 	ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size,
2007 			KM_SLEEP);
2008 	ntovp->attrcnt = 0;
2009 	ntovp->vfsstat = FALSE;
2010 }
2011 
2012 static void
2013 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp,
2014 	struct nfs4_svgetit_arg *sargp)
2015 {
2016 	int i;
2017 	union nfs4_attr_u *na;
2018 	uint8_t *amap;
2019 
2020 	/*
2021 	 * XXX Should do the same checks for whether the bit is set
2022 	 */
2023 	for (i = 0, na = ntovp->na, amap = ntovp->amap;
2024 		i < ntovp->attrcnt; i++, na++, amap++) {
2025 		(void) (*nfs4_ntov_map[*amap].sv_getit)(
2026 			NFS4ATTR_FREEIT, sargp, na);
2027 	}
2028 	if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) {
2029 		/*
2030 		 * xdr_free for getattr will be done later
2031 		 */
2032 		for (i = 0, na = ntovp->na, amap = ntovp->amap;
2033 			i < ntovp->attrcnt; i++, na++, amap++) {
2034 			xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na);
2035 		}
2036 	}
2037 	kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size);
2038 }
2039 
2040 /*
2041  * do_rfs4_op_getattr gets the system attrs and converts into fattr4.
2042  */
2043 static nfsstat4
2044 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp,
2045 	struct nfs4_svgetit_arg *sargp)
2046 {
2047 	int error = 0;
2048 	int i, k;
2049 	struct nfs4_ntov_table ntov;
2050 	XDR xdr;
2051 	ulong_t xdr_size;
2052 	char *xdr_attrs;
2053 	nfsstat4 status = NFS4_OK;
2054 	nfsstat4 prev_rdattr_error = sargp->rdattr_error;
2055 	union nfs4_attr_u *na;
2056 	uint8_t *amap;
2057 
2058 	sargp->op = NFS4ATTR_GETIT;
2059 	sargp->flag = 0;
2060 
2061 	fattrp->attrmask = 0;
2062 	/* if no bits requested, then return empty fattr4 */
2063 	if (breq == 0) {
2064 		fattrp->attrlist4_len = 0;
2065 		fattrp->attrlist4 = NULL;
2066 		return (NFS4_OK);
2067 	}
2068 
2069 	/*
2070 	 * return NFS4ERR_INVAL when client requests write-only attrs
2071 	 */
2072 	if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK))
2073 		return (NFS4ERR_INVAL);
2074 
2075 	nfs4_ntov_table_init(&ntov);
2076 	na = ntov.na;
2077 	amap = ntov.amap;
2078 
2079 	/*
2080 	 * Now loop to get or verify the attrs
2081 	 */
2082 	for (i = 0; i < nfs4_ntov_map_size; i++) {
2083 		if (breq & nfs4_ntov_map[i].fbit) {
2084 			if ((*nfs4_ntov_map[i].sv_getit)(
2085 				    NFS4ATTR_SUPPORTED, sargp, NULL) == 0) {
2086 
2087 				error = (*nfs4_ntov_map[i].sv_getit)(
2088 						NFS4ATTR_GETIT, sargp, na);
2089 
2090 				/*
2091 				 * Possible error values:
2092 				 * >0 if sv_getit failed to
2093 				 * get the attr; 0 if succeeded;
2094 				 * <0 if rdattr_error and the
2095 				 * attribute cannot be returned.
2096 				 */
2097 				if (error && !(sargp->rdattr_error_req))
2098 					goto done;
2099 				/*
2100 				 * If error then just for entry
2101 				 */
2102 				if (error == 0) {
2103 					fattrp->attrmask |=
2104 						nfs4_ntov_map[i].fbit;
2105 					*amap++ =
2106 						(uint8_t)nfs4_ntov_map[i].nval;
2107 					na++;
2108 					(ntov.attrcnt)++;
2109 				} else if ((error > 0) &&
2110 					(sargp->rdattr_error == NFS4_OK)) {
2111 					sargp->rdattr_error = puterrno4(error);
2112 				}
2113 				error = 0;
2114 			}
2115 		}
2116 	}
2117 
2118 	/*
2119 	 * If rdattr_error was set after the return value for it was assigned,
2120 	 * update it.
2121 	 */
2122 	if (prev_rdattr_error != sargp->rdattr_error) {
2123 		na = ntov.na;
2124 		amap = ntov.amap;
2125 		for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2126 			k = *amap;
2127 			if (k < FATTR4_RDATTR_ERROR) {
2128 				continue;
2129 			}
2130 			if ((k == FATTR4_RDATTR_ERROR) &&
2131 			    ((*nfs4_ntov_map[k].sv_getit)(
2132 				NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) {
2133 
2134 				(void) (*nfs4_ntov_map[k].sv_getit)(
2135 						NFS4ATTR_GETIT, sargp, na);
2136 			}
2137 			break;
2138 		}
2139 	}
2140 
2141 	xdr_size = 0;
2142 	na = ntov.na;
2143 	amap = ntov.amap;
2144 	for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2145 		xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na);
2146 	}
2147 
2148 	fattrp->attrlist4_len = xdr_size;
2149 	if (xdr_size) {
2150 		/* freed by rfs4_op_getattr_free() */
2151 		fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP);
2152 
2153 		xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE);
2154 
2155 		na = ntov.na;
2156 		amap = ntov.amap;
2157 		for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2158 			if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) {
2159 				cmn_err(CE_WARN, "do_rfs4_op_getattr: xdr "
2160 					"encode of attribute %d failed\n",
2161 					*amap);
2162 				status = NFS4ERR_SERVERFAULT;
2163 				break;
2164 			}
2165 		}
2166 		/* xdrmem_destroy(&xdrs); */	/* NO-OP */
2167 	} else {
2168 		fattrp->attrlist4 = NULL;
2169 	}
2170 done:
2171 
2172 	nfs4_ntov_table_free(&ntov, sargp);
2173 
2174 	if (error != 0)
2175 		status = puterrno4(error);
2176 
2177 	return (status);
2178 }
2179 
2180 /* ARGSUSED */
2181 static void
2182 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2183 	struct compound_state *cs)
2184 {
2185 	GETATTR4args *args = &argop->nfs_argop4_u.opgetattr;
2186 	GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr;
2187 	struct nfs4_svgetit_arg sarg;
2188 	struct statvfs64 sb;
2189 	nfsstat4 status;
2190 
2191 	if (cs->vp == NULL) {
2192 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2193 		return;
2194 	}
2195 
2196 	if (cs->access == CS_ACCESS_DENIED) {
2197 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2198 		return;
2199 	}
2200 
2201 	sarg.sbp = &sb;
2202 	sarg.cs = cs;
2203 
2204 	status = bitmap4_to_attrmask(args->attr_request, &sarg);
2205 	if (status == NFS4_OK) {
2206 		status = bitmap4_get_sysattrs(&sarg);
2207 		if (status == NFS4_OK)
2208 			status = do_rfs4_op_getattr(args->attr_request,
2209 				&resp->obj_attributes, &sarg);
2210 	}
2211 	*cs->statusp = resp->status = status;
2212 }
2213 
2214 static void
2215 rfs4_op_getattr_free(nfs_resop4 *resop)
2216 {
2217 	GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr;
2218 
2219 	nfs4_fattr4_free(&resp->obj_attributes);
2220 }
2221 
2222 /* ARGSUSED */
2223 static void
2224 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2225 	struct compound_state *cs)
2226 {
2227 	GETFH4res *resp = &resop->nfs_resop4_u.opgetfh;
2228 
2229 	if (cs->vp == NULL) {
2230 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2231 		return;
2232 	}
2233 	if (cs->access == CS_ACCESS_DENIED) {
2234 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2235 		return;
2236 	}
2237 
2238 	resp->object.nfs_fh4_val =
2239 		kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP);
2240 	nfs_fh4_copy(&cs->fh, &resp->object);
2241 	*cs->statusp = resp->status = NFS4_OK;
2242 }
2243 
2244 static void
2245 rfs4_op_getfh_free(nfs_resop4 *resop)
2246 {
2247 	GETFH4res *resp = &resop->nfs_resop4_u.opgetfh;
2248 
2249 	if (resp->status == NFS4_OK &&
2250 	    resp->object.nfs_fh4_val != NULL) {
2251 		kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len);
2252 		resp->object.nfs_fh4_val = NULL;
2253 		resp->object.nfs_fh4_len = 0;
2254 	}
2255 }
2256 
2257 /*
2258  * illegal: args: void
2259  *	    res : status (NFS4ERR_OP_ILLEGAL)
2260  */
2261 /* ARGSUSED */
2262 static void
2263 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop,
2264 	struct svc_req *req, struct compound_state *cs)
2265 {
2266 	ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal;
2267 
2268 	resop->resop = OP_ILLEGAL;
2269 	*cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL;
2270 }
2271 
2272 /*
2273  * link: args: SAVED_FH: file, CURRENT_FH: target directory
2274  *	 res: status. If success - CURRENT_FH unchanged, return change_info
2275  */
2276 /* ARGSUSED */
2277 static void
2278 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2279 	struct compound_state *cs)
2280 {
2281 	LINK4args *args = &argop->nfs_argop4_u.oplink;
2282 	LINK4res *resp = &resop->nfs_resop4_u.oplink;
2283 	int error;
2284 	vnode_t *vp;
2285 	vnode_t *dvp;
2286 	struct vattr bdva, idva, adva;
2287 	char *nm;
2288 	uint_t  len;
2289 
2290 	/* SAVED_FH: source object */
2291 	vp = cs->saved_vp;
2292 	if (vp == NULL) {
2293 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2294 		return;
2295 	}
2296 
2297 	/* CURRENT_FH: target directory */
2298 	dvp = cs->vp;
2299 	if (dvp == NULL) {
2300 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2301 		return;
2302 	}
2303 
2304 	/*
2305 	 * If there is a non-shared filesystem mounted on this vnode,
2306 	 * do not allow to link any file in this directory.
2307 	 */
2308 	if (vn_ismntpt(dvp)) {
2309 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2310 		return;
2311 	}
2312 
2313 	if (cs->access == CS_ACCESS_DENIED) {
2314 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2315 		return;
2316 	}
2317 
2318 	/* Check source object's type validity */
2319 	if (vp->v_type == VDIR) {
2320 		*cs->statusp = resp->status = NFS4ERR_ISDIR;
2321 		return;
2322 	}
2323 
2324 	/* Check target directory's type */
2325 	if (dvp->v_type != VDIR) {
2326 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2327 		return;
2328 	}
2329 
2330 	if (cs->saved_exi != cs->exi) {
2331 		*cs->statusp = resp->status = NFS4ERR_XDEV;
2332 		return;
2333 	}
2334 
2335 	if (!utf8_dir_verify(&args->newname)) {
2336 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2337 		return;
2338 	}
2339 
2340 	nm = utf8_to_fn(&args->newname, &len, NULL);
2341 	if (nm == NULL) {
2342 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2343 		return;
2344 	}
2345 
2346 	if (len > MAXNAMELEN) {
2347 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
2348 		kmem_free(nm, len);
2349 		return;
2350 	}
2351 
2352 	if (rdonly4(cs->exi, cs->vp, req)) {
2353 		*cs->statusp = resp->status = NFS4ERR_ROFS;
2354 		kmem_free(nm, len);
2355 		return;
2356 	}
2357 
2358 	/* Get "before" change value */
2359 	bdva.va_mask = AT_CTIME|AT_SEQ;
2360 	error = VOP_GETATTR(dvp, &bdva, 0, cs->cr);
2361 	if (error) {
2362 		*cs->statusp = resp->status = puterrno4(error);
2363 		kmem_free(nm, len);
2364 		return;
2365 	}
2366 
2367 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime)
2368 
2369 	error = VOP_LINK(dvp, vp, nm, cs->cr);
2370 
2371 	kmem_free(nm, len);
2372 
2373 	/*
2374 	 * Get the initial "after" sequence number, if it fails, set to zero
2375 	 */
2376 	idva.va_mask = AT_SEQ;
2377 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr))
2378 		idva.va_seq = 0;
2379 
2380 	/*
2381 	 * Force modified data and metadata out to stable storage.
2382 	 */
2383 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr);
2384 	(void) VOP_FSYNC(dvp, 0, cs->cr);
2385 
2386 	if (error) {
2387 		*cs->statusp = resp->status = puterrno4(error);
2388 		return;
2389 	}
2390 
2391 	/*
2392 	 * Get "after" change value, if it fails, simply return the
2393 	 * before value.
2394 	 */
2395 	adva.va_mask = AT_CTIME|AT_SEQ;
2396 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr)) {
2397 		adva.va_ctime = bdva.va_ctime;
2398 		adva.va_seq = 0;
2399 	}
2400 
2401 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
2402 
2403 	/*
2404 	 * The cinfo.atomic = TRUE only if we have
2405 	 * non-zero va_seq's, and it has incremented by exactly one
2406 	 * during the VOP_LINK and it didn't change during the VOP_FSYNC.
2407 	 */
2408 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
2409 			idva.va_seq == (bdva.va_seq + 1) &&
2410 			idva.va_seq == adva.va_seq)
2411 		resp->cinfo.atomic = TRUE;
2412 	else
2413 		resp->cinfo.atomic = FALSE;
2414 
2415 	*cs->statusp = resp->status = NFS4_OK;
2416 }
2417 
2418 /*
2419  * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work.
2420  */
2421 
2422 /* ARGSUSED */
2423 static nfsstat4
2424 do_rfs4_op_lookup(char *nm, uint_t buflen, struct svc_req *req,
2425 	struct compound_state *cs)
2426 {
2427 	int error;
2428 	int different_export = 0;
2429 	vnode_t *vp, *tvp, *pre_tvp = NULL, *oldvp = NULL;
2430 	struct exportinfo *exi = NULL, *pre_exi = NULL;
2431 	nfsstat4 stat;
2432 	fid_t fid;
2433 	int attrdir, dotdot, walk;
2434 	bool_t is_newvp = FALSE;
2435 
2436 	if (cs->vp->v_flag & V_XATTRDIR) {
2437 		attrdir = 1;
2438 		ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR));
2439 	} else {
2440 		attrdir = 0;
2441 		ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR));
2442 	}
2443 
2444 	dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0');
2445 
2446 	/*
2447 	 * If dotdotting, then need to check whether it's
2448 	 * above the root of a filesystem, or above an
2449 	 * export point.
2450 	 */
2451 	if (dotdot) {
2452 
2453 		/*
2454 		 * If dotdotting at the root of a filesystem, then
2455 		 * need to traverse back to the mounted-on filesystem
2456 		 * and do the dotdot lookup there.
2457 		 */
2458 		if (cs->vp->v_flag & VROOT) {
2459 
2460 			/*
2461 			 * If at the system root, then can
2462 			 * go up no further.
2463 			 */
2464 			if (VN_CMP(cs->vp, rootdir))
2465 				return (puterrno4(ENOENT));
2466 
2467 			/*
2468 			 * Traverse back to the mounted-on filesystem
2469 			 */
2470 			cs->vp = untraverse(cs->vp);
2471 
2472 			/*
2473 			 * Set the different_export flag so we remember
2474 			 * to pick up a new exportinfo entry for
2475 			 * this new filesystem.
2476 			 */
2477 			different_export = 1;
2478 		} else {
2479 
2480 			/*
2481 			 * If dotdotting above an export point then set
2482 			 * the different_export to get new export info.
2483 			 */
2484 			different_export = nfs_exported(cs->exi, cs->vp);
2485 		}
2486 	}
2487 
2488 	error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr);
2489 	if (error)
2490 		return (puterrno4(error));
2491 
2492 	/*
2493 	 * If the vnode is in a pseudo filesystem, check whether it is visible.
2494 	 *
2495 	 * XXX if the vnode is a symlink and it is not visible in
2496 	 * a pseudo filesystem, return ENOENT (not following symlink).
2497 	 * V4 client can not mount such symlink. This is a regression
2498 	 * from V2/V3.
2499 	 *
2500 	 * In the same exported filesystem, if the security flavor used
2501 	 * is not an explicitly shared flavor, limit the view to the visible
2502 	 * list entries only. This is not a WRONGSEC case because it's already
2503 	 * checked via PUTROOTFH/PUTPUBFH or PUTFH.
2504 	 */
2505 	if (!different_export &&
2506 	    (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) ||
2507 	    cs->access & CS_ACCESS_LIMITED)) {
2508 		if (! nfs_visible(cs->exi, vp, &different_export)) {
2509 			VN_RELE(vp);
2510 			return (puterrno4(ENOENT));
2511 		}
2512 	}
2513 
2514 	/*
2515 	 * If it's a mountpoint, then traverse it.
2516 	 */
2517 	if (vn_ismntpt(vp)) {
2518 		pre_exi = cs->exi;	/* save pre-traversed exportinfo */
2519 		pre_tvp = vp;		/* save pre-traversed vnode	*/
2520 
2521 		/*
2522 		 * hold pre_tvp to counteract rele by traverse.  We will
2523 		 * need pre_tvp below if checkexport4 fails
2524 		 */
2525 		VN_HOLD(pre_tvp);
2526 		tvp = vp;
2527 		if ((error = traverse(&tvp)) != 0) {
2528 			VN_RELE(vp);
2529 			VN_RELE(pre_tvp);
2530 			return (puterrno4(error));
2531 		}
2532 		vp = tvp;
2533 		different_export = 1;
2534 	} else if (vp->v_vfsp != cs->vp->v_vfsp) {
2535 		/*
2536 		 * The vfsp comparison is to handle the case where
2537 		 * a LOFS mount is shared.  lo_lookup traverses mount points,
2538 		 * and NFS is unaware of local fs transistions because
2539 		 * v_vfsmountedhere isn't set.  For this special LOFS case,
2540 		 * the dir and the obj returned by lookup will have different
2541 		 * vfs ptrs.
2542 		 */
2543 		different_export = 1;
2544 	}
2545 
2546 	if (different_export) {
2547 
2548 		bzero(&fid, sizeof (fid));
2549 		fid.fid_len = MAXFIDSZ;
2550 		error = vop_fid_pseudo(vp, &fid);
2551 		if (error) {
2552 			VN_RELE(vp);
2553 			if (pre_tvp)
2554 				VN_RELE(pre_tvp);
2555 			return (puterrno4(error));
2556 		}
2557 
2558 		if (dotdot)
2559 			exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE);
2560 		else
2561 			exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
2562 
2563 		if (exi == NULL) {
2564 			if (pre_tvp) {
2565 				/*
2566 				 * If this vnode is a mounted-on vnode,
2567 				 * but the mounted-on file system is not
2568 				 * exported, send back the filehandle for
2569 				 * the mounted-on vnode, not the root of
2570 				 * the mounted-on file system.
2571 				 */
2572 				VN_RELE(vp);
2573 				vp = pre_tvp;
2574 				exi = pre_exi;
2575 			} else {
2576 				VN_RELE(vp);
2577 				return (puterrno4(EACCES));
2578 			}
2579 		} else if (pre_tvp) {
2580 			/* we're done with pre_tvp now. release extra hold */
2581 			VN_RELE(pre_tvp);
2582 		}
2583 
2584 		cs->exi = exi;
2585 
2586 		/*
2587 		 * Now we do a checkauth4. The reason is that
2588 		 * this client/user may not have access to the new
2589 		 * exported file system, and if he does,
2590 		 * the client/user may be mapped to a different uid.
2591 		 *
2592 		 * We start with a new cr, because the checkauth4 done
2593 		 * in the PUT*FH operation over wrote the cred's uid,
2594 		 * gid, etc, and we want the real thing before calling
2595 		 * checkauth4()
2596 		 */
2597 		crfree(cs->cr);
2598 		cs->cr = crdup(cs->basecr);
2599 
2600 		if (cs->vp)
2601 			oldvp = cs->vp;
2602 		cs->vp = vp;
2603 		is_newvp = TRUE;
2604 
2605 		stat = call_checkauth4(cs, req);
2606 		if (stat != NFS4_OK) {
2607 			VN_RELE(cs->vp);
2608 			cs->vp = oldvp;
2609 			return (stat);
2610 		}
2611 	}
2612 
2613 	/*
2614 	 * After various NFS checks, do a label check on the path
2615 	 * component. The label on this path should either be the
2616 	 * global zone's label or a zone's label. We are only
2617 	 * interested in the zone's label because exported files
2618 	 * in global zone is accessible (though read-only) to
2619 	 * clients. The exportability/visibility check is already
2620 	 * done before reaching this code.
2621 	 */
2622 	if (is_system_labeled()) {
2623 		bslabel_t *clabel;
2624 
2625 		ASSERT(req->rq_label != NULL);
2626 		clabel = req->rq_label;
2627 		DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *,
2628 		    "got client label from request(1)", struct svc_req *, req);
2629 
2630 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
2631 			if (!do_rfs4_label_check(clabel, vp, DOMINANCE_CHECK)) {
2632 				error = EACCES;
2633 				goto err_out;
2634 			}
2635 		} else {
2636 			/*
2637 			 * We grant access to admin_low label clients
2638 			 * only if the client is trusted, i.e. also
2639 			 * running Solaris Trusted Extension.
2640 			 */
2641 			struct sockaddr	*ca;
2642 			int		addr_type;
2643 			void		*ipaddr;
2644 			tsol_tpc_t	*tp;
2645 
2646 			ca = (struct sockaddr *)svc_getrpccaller(
2647 			    req->rq_xprt)->buf;
2648 			if (ca->sa_family == AF_INET) {
2649 				addr_type = IPV4_VERSION;
2650 				ipaddr = &((struct sockaddr_in *)ca)->sin_addr;
2651 			} else if (ca->sa_family == AF_INET6) {
2652 				addr_type = IPV6_VERSION;
2653 				ipaddr = &((struct sockaddr_in6 *)
2654 				    ca)->sin6_addr;
2655 			}
2656 			tp = find_tpc(ipaddr, addr_type, B_FALSE);
2657 			if (tp == NULL || tp->tpc_tp.tp_doi !=
2658 			    l_admin_low->tsl_doi || tp->tpc_tp.host_type !=
2659 			    SUN_CIPSO) {
2660 				error = EACCES;
2661 				goto err_out;
2662 			}
2663 		}
2664 	}
2665 
2666 	error = makefh4(&cs->fh, vp, cs->exi);
2667 
2668 err_out:
2669 	if (error) {
2670 		if (is_newvp) {
2671 			VN_RELE(cs->vp);
2672 			cs->vp = oldvp;
2673 		} else
2674 			VN_RELE(vp);
2675 		return (puterrno4(error));
2676 	}
2677 
2678 	if (!is_newvp) {
2679 		if (cs->vp)
2680 			VN_RELE(cs->vp);
2681 		cs->vp = vp;
2682 	} else if (oldvp)
2683 		VN_RELE(oldvp);
2684 
2685 	/*
2686 	 * if did lookup on attrdir and didn't lookup .., set named
2687 	 * attr fh flag
2688 	 */
2689 	if (attrdir && ! dotdot)
2690 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
2691 
2692 	/* Assume false for now, open proc will set this */
2693 	cs->mandlock = FALSE;
2694 
2695 	return (NFS4_OK);
2696 }
2697 
2698 /* ARGSUSED */
2699 static void
2700 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2701 	struct compound_state *cs)
2702 {
2703 	LOOKUP4args *args = &argop->nfs_argop4_u.oplookup;
2704 	LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup;
2705 	char *nm;
2706 	uint_t len;
2707 
2708 	if (cs->vp == NULL) {
2709 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2710 		return;
2711 	}
2712 
2713 	if (cs->vp->v_type == VLNK) {
2714 		*cs->statusp = resp->status = NFS4ERR_SYMLINK;
2715 		return;
2716 	}
2717 
2718 	if (cs->vp->v_type != VDIR) {
2719 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2720 		return;
2721 	}
2722 
2723 	if (!utf8_dir_verify(&args->objname)) {
2724 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2725 		return;
2726 	}
2727 
2728 	nm = utf8_to_str(&args->objname, &len, NULL);
2729 	if (nm == NULL) {
2730 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2731 		return;
2732 	}
2733 
2734 	if (len > MAXNAMELEN) {
2735 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
2736 		kmem_free(nm, len);
2737 		return;
2738 	}
2739 
2740 	*cs->statusp = resp->status = do_rfs4_op_lookup(nm, len, req, cs);
2741 
2742 	kmem_free(nm, len);
2743 }
2744 
2745 /* ARGSUSED */
2746 static void
2747 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
2748 	struct compound_state *cs)
2749 {
2750 	LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp;
2751 
2752 	if (cs->vp == NULL) {
2753 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2754 		return;
2755 	}
2756 
2757 	if (cs->vp->v_type != VDIR) {
2758 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2759 		return;
2760 	}
2761 
2762 	*cs->statusp = resp->status = do_rfs4_op_lookup("..", 3, req, cs);
2763 
2764 	/*
2765 	 * From NFSV4 Specification, LOOKUPP should not check for
2766 	 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead.
2767 	 */
2768 	if (resp->status == NFS4ERR_WRONGSEC) {
2769 		*cs->statusp = resp->status = NFS4_OK;
2770 	}
2771 }
2772 
2773 
2774 /*ARGSUSED2*/
2775 static void
2776 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2777 	struct compound_state *cs)
2778 {
2779 	OPENATTR4args	*args = &argop->nfs_argop4_u.opopenattr;
2780 	OPENATTR4res	*resp = &resop->nfs_resop4_u.opopenattr;
2781 	vnode_t		*avp = NULL;
2782 	int		lookup_flags = LOOKUP_XATTR, error;
2783 	int		exp_ro = 0;
2784 
2785 	if (cs->vp == NULL) {
2786 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2787 		return;
2788 	}
2789 
2790 	/*
2791 	 * Make a couple of checks made by copen()
2792 	 *
2793 	 * Check to make sure underlying fs supports xattrs.  This
2794 	 * is required because solaris filesystem implementations
2795 	 * (UFS/TMPFS) don't enforce the noxattr mount option
2796 	 * in VOP_LOOKUP(LOOKUP_XATTR).  If fs doesn't support this
2797 	 * pathconf cmd or if fs supports cmd but doesn't claim
2798 	 * support for xattr, return NOTSUPP.  It would be better
2799 	 * to use VOP_PATHCONF( _PC_XATTR_ENABLED) for this; however,
2800 	 * that cmd is not available to VOP_PATHCONF interface
2801 	 * (it's only implemented inside pathconf syscall)...
2802 	 *
2803 	 * Verify permission to put attributes on files (access
2804 	 * checks from copen).
2805 	 */
2806 
2807 	if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0) {
2808 		error = ENOTSUP;
2809 		goto error_out;
2810 	}
2811 
2812 	if ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr) != 0) &&
2813 	    (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr) != 0) &&
2814 	    (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr) != 0)) {
2815 		error = EACCES;
2816 		goto error_out;
2817 	}
2818 
2819 	/*
2820 	 * The CREATE_XATTR_DIR VOP flag cannot be specified if
2821 	 * the file system is exported read-only -- regardless of
2822 	 * createdir flag.  Otherwise the attrdir would be created
2823 	 * (assuming server fs isn't mounted readonly locally).  If
2824 	 * VOP_LOOKUP returns ENOENT in this case, the error will
2825 	 * be translated into EROFS.  ENOSYS is mapped to ENOTSUP
2826 	 * because specfs has no VOP_LOOKUP op, so the macro would
2827 	 * return ENOSYS.  EINVAL is returned by all (current)
2828 	 * Solaris file system implementations when any of their
2829 	 * restrictions are violated (xattr(dir) can't have xattrdir).
2830 	 * Returning NOTSUPP is more appropriate in this case
2831 	 * because the object will never be able to have an attrdir.
2832 	 */
2833 	if (args->createdir && ! (exp_ro = rdonly4(cs->exi, cs->vp, req)))
2834 		lookup_flags |= CREATE_XATTR_DIR;
2835 
2836 	error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr);
2837 
2838 	if (error) {
2839 		if (error == ENOENT && args->createdir && exp_ro)
2840 			error = EROFS;
2841 		else if (error == EINVAL || error == ENOSYS)
2842 			error = ENOTSUP;
2843 		goto error_out;
2844 	}
2845 
2846 	ASSERT(avp->v_flag & V_XATTRDIR);
2847 
2848 	error = makefh4(&cs->fh, avp, cs->exi);
2849 
2850 	if (error) {
2851 		VN_RELE(avp);
2852 		goto error_out;
2853 	}
2854 
2855 	VN_RELE(cs->vp);
2856 	cs->vp = avp;
2857 
2858 	/*
2859 	 * There is no requirement for an attrdir fh flag
2860 	 * because the attrdir has a vnode flag to distinguish
2861 	 * it from regular (non-xattr) directories.  The
2862 	 * FH4_ATTRDIR flag is set for future sanity checks.
2863 	 */
2864 	set_fh4_flag(&cs->fh, FH4_ATTRDIR);
2865 	*cs->statusp = resp->status = NFS4_OK;
2866 	return;
2867 
2868 error_out:
2869 
2870 	*cs->statusp = resp->status = puterrno4(error);
2871 }
2872 
2873 static int
2874 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred)
2875 {
2876 	int error;
2877 	int i;
2878 	clock_t delaytime;
2879 	caller_context_t ct;
2880 
2881 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
2882 
2883 	/*
2884 	 * Don't block on mandatory locks. If this routine returns
2885 	 * EAGAIN, the caller should return NFS4ERR_LOCKED.
2886 	 */
2887 	uio->uio_fmode = FNONBLOCK;
2888 
2889 	ct.cc_sysid = 0;
2890 	ct.cc_pid = 0;
2891 	ct.cc_caller_id = nfs4_srv_caller_id;
2892 
2893 	for (i = 0; i < rfs4_maxlock_tries; i++) {
2894 
2895 
2896 		if (direction == FREAD) {
2897 			(void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, &ct);
2898 			error = VOP_READ(vp, uio, ioflag, cred, &ct);
2899 			VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, &ct);
2900 		} else {
2901 			(void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, &ct);
2902 			error = VOP_WRITE(vp, uio, ioflag, cred, &ct);
2903 			VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, &ct);
2904 		}
2905 
2906 		if (error != EAGAIN)
2907 			break;
2908 
2909 		if (i < rfs4_maxlock_tries - 1) {
2910 			delay(delaytime);
2911 			delaytime *= 2;
2912 		}
2913 	}
2914 
2915 	return (error);
2916 }
2917 
2918 /* ARGSUSED */
2919 static void
2920 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2921 	struct compound_state *cs)
2922 {
2923 	READ4args *args = &argop->nfs_argop4_u.opread;
2924 	READ4res *resp = &resop->nfs_resop4_u.opread;
2925 	int error;
2926 	int verror;
2927 	vnode_t *vp;
2928 	struct vattr va;
2929 	struct iovec iov;
2930 	struct uio uio;
2931 	u_offset_t offset;
2932 	bool_t *deleg = &cs->deleg;
2933 	nfsstat4 stat;
2934 	int in_crit = 0;
2935 	mblk_t *mp;
2936 	int alloc_err = 0;
2937 
2938 	vp = cs->vp;
2939 	if (vp == NULL) {
2940 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2941 		return;
2942 	}
2943 	if (cs->access == CS_ACCESS_DENIED) {
2944 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2945 		return;
2946 	}
2947 
2948 	/*
2949 	 * Enter the critical region before calling VOP_RWLOCK
2950 	 * to avoid a deadlock with write requests.
2951 	 */
2952 	if (nbl_need_check(vp)) {
2953 		nbl_start_crit(vp, RW_READER);
2954 		in_crit = 1;
2955 		if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0)) {
2956 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
2957 			goto out;
2958 		}
2959 	}
2960 
2961 	if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE,
2962 					deleg, TRUE)) != NFS4_OK) {
2963 		*cs->statusp = resp->status = stat;
2964 		goto out;
2965 	}
2966 
2967 	va.va_mask = AT_MODE|AT_SIZE|AT_UID;
2968 	verror = VOP_GETATTR(vp, &va, 0, cs->cr);
2969 
2970 	/*
2971 	 * If we can't get the attributes, then we can't do the
2972 	 * right access checking.  So, we'll fail the request.
2973 	 */
2974 	if (verror) {
2975 		*cs->statusp = resp->status = puterrno4(verror);
2976 		goto out;
2977 	}
2978 
2979 	if (vp->v_type != VREG) {
2980 		*cs->statusp = resp->status =
2981 			((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
2982 		goto out;
2983 	}
2984 
2985 	if (crgetuid(cs->cr) != va.va_uid &&
2986 	    (error = VOP_ACCESS(vp, VREAD, 0, cs->cr)) &&
2987 	    (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr))) {
2988 		*cs->statusp = resp->status = puterrno4(error);
2989 		goto out;
2990 	}
2991 
2992 	if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */
2993 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2994 		goto out;
2995 	}
2996 
2997 	offset = args->offset;
2998 	if (offset >= va.va_size) {
2999 		*cs->statusp = resp->status = NFS4_OK;
3000 		resp->eof = TRUE;
3001 		resp->data_len = 0;
3002 		resp->data_val = NULL;
3003 		resp->mblk = NULL;
3004 		*cs->statusp = resp->status = NFS4_OK;
3005 		goto out;
3006 	}
3007 
3008 	if (args->count == 0) {
3009 		*cs->statusp = resp->status = NFS4_OK;
3010 		resp->eof = FALSE;
3011 		resp->data_len = 0;
3012 		resp->data_val = NULL;
3013 		resp->mblk = NULL;
3014 		goto out;
3015 	}
3016 
3017 	/*
3018 	 * Do not allocate memory more than maximum allowed
3019 	 * transfer size
3020 	 */
3021 	if (args->count > rfs4_tsize(req))
3022 		args->count = rfs4_tsize(req);
3023 
3024 	/*
3025 	 * mp will contain the data to be sent out in the read reply.
3026 	 * It will be freed after the reply has been sent.
3027 	 * Let's roundup the data to a BYTES_PER_XDR_UNIT multiple,
3028 	 * so that the call to xdrmblk_putmblk() never fails.
3029 	 * If the first alloc of the requested size fails, then
3030 	 * decrease the size to something more reasonable and wait
3031 	 * for the allocation to occur.
3032 	 */
3033 	mp = allocb(RNDUP(args->count), BPRI_MED);
3034 	if (mp == NULL) {
3035 		if (args->count > MAXBSIZE)
3036 			args->count = MAXBSIZE;
3037 		mp = allocb_wait(RNDUP(args->count), BPRI_MED,
3038 				STR_NOSIG, &alloc_err);
3039 	}
3040 	ASSERT(mp != NULL);
3041 	ASSERT(alloc_err == 0);
3042 
3043 	iov.iov_base = (caddr_t)mp->b_datap->db_base;
3044 	iov.iov_len = args->count;
3045 	uio.uio_iov = &iov;
3046 	uio.uio_iovcnt = 1;
3047 	uio.uio_segflg = UIO_SYSSPACE;
3048 	uio.uio_extflg = UIO_COPY_CACHED;
3049 	uio.uio_loffset = args->offset;
3050 	uio.uio_resid = args->count;
3051 
3052 	error = do_io(FREAD, vp, &uio, 0, cs->cr);
3053 
3054 	va.va_mask = AT_SIZE;
3055 	verror = VOP_GETATTR(vp, &va, 0, cs->cr);
3056 
3057 	if (error) {
3058 		freeb(mp);
3059 		*cs->statusp = resp->status = puterrno4(error);
3060 		goto out;
3061 	}
3062 
3063 	*cs->statusp = resp->status = NFS4_OK;
3064 
3065 	ASSERT(uio.uio_resid >= 0);
3066 	resp->data_len = args->count - uio.uio_resid;
3067 	resp->data_val = (char *)mp->b_datap->db_base;
3068 	resp->mblk = mp;
3069 
3070 	if (!verror && offset + resp->data_len == va.va_size)
3071 		resp->eof = TRUE;
3072 	else
3073 		resp->eof = FALSE;
3074 
3075 out:
3076 	if (in_crit)
3077 		nbl_end_crit(vp);
3078 }
3079 
3080 static void
3081 rfs4_op_read_free(nfs_resop4 *resop)
3082 {
3083 	READ4res *resp = &resop->nfs_resop4_u.opread;
3084 
3085 	if (resp->status == NFS4_OK && resp->mblk != NULL) {
3086 		freeb(resp->mblk);
3087 		resp->mblk = NULL;
3088 		resp->data_val = NULL;
3089 		resp->data_len = 0;
3090 	}
3091 }
3092 
3093 static void
3094 rfs4_op_readdir_free(nfs_resop4 *resop)
3095 {
3096 	READDIR4res *resp = &resop->nfs_resop4_u.opreaddir;
3097 
3098 	if (resp->status == NFS4_OK && resp->mblk != NULL) {
3099 		freeb(resp->mblk);
3100 		resp->mblk = NULL;
3101 		resp->data_len = 0;
3102 	}
3103 }
3104 
3105 
3106 /* ARGSUSED */
3107 static void
3108 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
3109 	struct compound_state *cs)
3110 {
3111 	PUTPUBFH4res *resp = &resop->nfs_resop4_u.opputpubfh;
3112 	int error;
3113 	vnode_t *vp;
3114 	struct exportinfo *exi, *sav_exi;
3115 	nfs_fh4_fmt_t *fh_fmtp;
3116 
3117 	if (cs->vp) {
3118 		VN_RELE(cs->vp);
3119 		cs->vp = NULL;
3120 	}
3121 
3122 	if (cs->cr)
3123 		crfree(cs->cr);
3124 
3125 	cs->cr = crdup(cs->basecr);
3126 
3127 	vp = exi_public->exi_vp;
3128 	if (vp == NULL) {
3129 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
3130 		return;
3131 	}
3132 
3133 	error = makefh4(&cs->fh, vp, exi_public);
3134 	if (error != 0) {
3135 		*cs->statusp = resp->status = puterrno4(error);
3136 		return;
3137 	}
3138 	sav_exi = cs->exi;
3139 	if (exi_public == exi_root) {
3140 		/*
3141 		 * No filesystem is actually shared public, so we default
3142 		 * to exi_root. In this case, we must check whether root
3143 		 * is exported.
3144 		 */
3145 		fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val;
3146 
3147 		/*
3148 		 * if root filesystem is exported, the exportinfo struct that we
3149 		 * should use is what checkexport4 returns, because root_exi is
3150 		 * actually a mostly empty struct.
3151 		 */
3152 		exi = checkexport4(&fh_fmtp->fh4_fsid,
3153 			(fid_t *)&fh_fmtp->fh4_xlen, NULL);
3154 		cs->exi = ((exi != NULL) ? exi : exi_public);
3155 	} else {
3156 		/*
3157 		 * it's a properly shared filesystem
3158 		 */
3159 		cs->exi = exi_public;
3160 	}
3161 
3162 	VN_HOLD(vp);
3163 	cs->vp = vp;
3164 
3165 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3166 		VN_RELE(cs->vp);
3167 		cs->vp = NULL;
3168 		cs->exi = sav_exi;
3169 		return;
3170 	}
3171 
3172 	*cs->statusp = resp->status = NFS4_OK;
3173 }
3174 
3175 /*
3176  * XXX - issue with put*fh operations. Suppose /export/home is exported.
3177  * Suppose an NFS client goes to mount /export/home/joe. If /export, home,
3178  * or joe have restrictive search permissions, then we shouldn't let
3179  * the client get a file handle. This is easy to enforce. However, we
3180  * don't know what security flavor should be used until we resolve the
3181  * path name. Another complication is uid mapping. If root is
3182  * the user, then it will be mapped to the anonymous user by default,
3183  * but we won't know that till we've resolved the path name. And we won't
3184  * know what the anonymous user is.
3185  * Luckily, SECINFO is specified to take a full filename.
3186  * So what we will have to in rfs4_op_lookup is check that flavor of
3187  * the target object matches that of the request, and if root was the
3188  * caller, check for the root= and anon= options, and if necessary,
3189  * repeat the lookup using the right cred_t. But that's not done yet.
3190  */
3191 /* ARGSUSED */
3192 static void
3193 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3194 	struct compound_state *cs)
3195 {
3196 	PUTFH4args *args = &argop->nfs_argop4_u.opputfh;
3197 	PUTFH4res *resp = &resop->nfs_resop4_u.opputfh;
3198 	nfs_fh4_fmt_t *fh_fmtp;
3199 
3200 	if (cs->vp) {
3201 		VN_RELE(cs->vp);
3202 		cs->vp = NULL;
3203 	}
3204 
3205 	if (cs->cr) {
3206 		crfree(cs->cr);
3207 		cs->cr = NULL;
3208 	}
3209 
3210 
3211 	if (args->object.nfs_fh4_len < NFS_FH4_LEN) {
3212 		*cs->statusp = resp->status = NFS4ERR_BADHANDLE;
3213 		return;
3214 	}
3215 
3216 	fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val;
3217 	cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen,
3218 				NULL);
3219 
3220 	if (cs->exi == NULL) {
3221 		*cs->statusp = resp->status = NFS4ERR_STALE;
3222 		return;
3223 	}
3224 
3225 	cs->cr = crdup(cs->basecr);
3226 
3227 	ASSERT(cs->cr != NULL);
3228 
3229 	if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) {
3230 		*cs->statusp = resp->status;
3231 		return;
3232 	}
3233 
3234 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3235 		VN_RELE(cs->vp);
3236 		cs->vp = NULL;
3237 		return;
3238 	}
3239 
3240 	nfs_fh4_copy(&args->object, &cs->fh);
3241 	*cs->statusp = resp->status = NFS4_OK;
3242 	cs->deleg = FALSE;
3243 }
3244 
3245 /* ARGSUSED */
3246 static void
3247 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3248 	struct compound_state *cs)
3249 
3250 {
3251 	PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh;
3252 	int error;
3253 	fid_t fid;
3254 	struct exportinfo *exi, *sav_exi;
3255 
3256 	if (cs->vp) {
3257 		VN_RELE(cs->vp);
3258 		cs->vp = NULL;
3259 	}
3260 
3261 	if (cs->cr)
3262 		crfree(cs->cr);
3263 
3264 	cs->cr = crdup(cs->basecr);
3265 
3266 	/*
3267 	 * Using rootdir, the system root vnode,
3268 	 * get its fid.
3269 	 */
3270 	bzero(&fid, sizeof (fid));
3271 	fid.fid_len = MAXFIDSZ;
3272 	error = vop_fid_pseudo(rootdir, &fid);
3273 	if (error != 0) {
3274 		*cs->statusp = resp->status = puterrno4(error);
3275 		return;
3276 	}
3277 
3278 	/*
3279 	 * Then use the root fsid & fid it to find out if it's exported
3280 	 *
3281 	 * If the server root isn't exported directly, then
3282 	 * it should at least be a pseudo export based on
3283 	 * one or more exports further down in the server's
3284 	 * file tree.
3285 	 */
3286 	exi = checkexport4(&rootdir->v_vfsp->vfs_fsid, &fid, NULL);
3287 	if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) {
3288 		NFS4_DEBUG(rfs4_debug,
3289 			(CE_WARN, "rfs4_op_putrootfh: export check failure"));
3290 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
3291 		return;
3292 	}
3293 
3294 	/*
3295 	 * Now make a filehandle based on the root
3296 	 * export and root vnode.
3297 	 */
3298 	error = makefh4(&cs->fh, rootdir, exi);
3299 	if (error != 0) {
3300 		*cs->statusp = resp->status = puterrno4(error);
3301 		return;
3302 	}
3303 
3304 	sav_exi = cs->exi;
3305 	cs->exi = exi;
3306 
3307 	VN_HOLD(rootdir);
3308 	cs->vp = rootdir;
3309 
3310 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3311 		VN_RELE(rootdir);
3312 		cs->vp = NULL;
3313 		cs->exi = sav_exi;
3314 		return;
3315 	}
3316 
3317 	*cs->statusp = resp->status = NFS4_OK;
3318 	cs->deleg = FALSE;
3319 }
3320 
3321 /*
3322  * A directory entry is a valid nfsv4 entry if
3323  * - it has a non-zero ino
3324  * - it is not a dot or dotdot name
3325  * - it is visible in a pseudo export or in a real export that can
3326  *   only have a limited view.
3327  */
3328 static bool_t
3329 valid_nfs4_entry(struct exportinfo *exi, struct dirent64 *dp,
3330 		int *expseudo, int check_visible)
3331 {
3332 	if (dp->d_ino == 0 || NFS_IS_DOTNAME(dp->d_name)) {
3333 		*expseudo = 0;
3334 		return (FALSE);
3335 	}
3336 
3337 	if (! check_visible) {
3338 		*expseudo = 0;
3339 		return (TRUE);
3340 	}
3341 
3342 	return (nfs_visible_inode(exi, dp->d_ino, expseudo));
3343 }
3344 
3345 /*
3346  * set_rdattr_params sets up the variables used to manage what information
3347  * to get for each directory entry.
3348  */
3349 static nfsstat4
3350 set_rdattr_params(struct nfs4_svgetit_arg *sargp,
3351 		bitmap4 attrs, bool_t *need_to_lookup)
3352 {
3353 	uint_t	va_mask;
3354 	nfsstat4 status;
3355 	bitmap4 objbits;
3356 
3357 	status = bitmap4_to_attrmask(attrs, sargp);
3358 	if (status != NFS4_OK) {
3359 		/*
3360 		 * could not even figure attr mask
3361 		 */
3362 		return (status);
3363 	}
3364 	va_mask = sargp->vap->va_mask;
3365 
3366 	/*
3367 	 * dirent's d_ino is always correct value for mounted_on_fileid.
3368 	 * mntdfid_set is set once here, but mounted_on_fileid is
3369 	 * set in main dirent processing loop for each dirent.
3370 	 * The mntdfid_set is a simple optimization that lets the
3371 	 * server attr code avoid work when caller is readdir.
3372 	 */
3373 	sargp->mntdfid_set = TRUE;
3374 
3375 	/*
3376 	 * Lookup entry only if client asked for any of the following:
3377 	 * a) vattr attrs
3378 	 * b) vfs attrs
3379 	 * c) attrs w/per-object scope requested (change, filehandle, etc)
3380 	 *    other than mounted_on_fileid (which we can take from dirent)
3381 	 */
3382 	objbits = attrs ? attrs & NFS4_VP_ATTR_MASK : 0;
3383 
3384 	if (va_mask || sargp->sbp || (objbits & ~FATTR4_MOUNTED_ON_FILEID_MASK))
3385 		*need_to_lookup = TRUE;
3386 	else
3387 		*need_to_lookup = FALSE;
3388 
3389 	if (sargp->sbp == NULL)
3390 		return (NFS4_OK);
3391 
3392 	/*
3393 	 * If filesystem attrs are requested, get them now from the
3394 	 * directory vp, as most entries will have same filesystem. The only
3395 	 * exception are mounted over entries but we handle
3396 	 * those as we go (XXX mounted over detection not yet implemented).
3397 	 */
3398 	sargp->vap->va_mask = 0;	/* to avoid VOP_GETATTR */
3399 	status = bitmap4_get_sysattrs(sargp);
3400 	sargp->vap->va_mask = va_mask;
3401 
3402 	if ((status != NFS4_OK) && sargp->rdattr_error_req) {
3403 		/*
3404 		 * Failed to get filesystem attributes.
3405 		 * Return a rdattr_error for each entry, but don't fail.
3406 		 * However, don't get any obj-dependent attrs.
3407 		 */
3408 		sargp->rdattr_error = status;	/* for rdattr_error */
3409 		*need_to_lookup = FALSE;
3410 		/*
3411 		 * At least get fileid for regular readdir output
3412 		 */
3413 		sargp->vap->va_mask &= AT_NODEID;
3414 		status = NFS4_OK;
3415 	}
3416 
3417 	return (status);
3418 }
3419 
3420 /*
3421  * readlink: args: CURRENT_FH.
3422  *	res: status. If success - CURRENT_FH unchanged, return linktext.
3423  */
3424 
3425 /* ARGSUSED */
3426 static void
3427 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3428 	struct compound_state *cs)
3429 {
3430 	READLINK4res *resp = &resop->nfs_resop4_u.opreadlink;
3431 	int error;
3432 	vnode_t *vp;
3433 	struct iovec iov;
3434 	struct vattr va;
3435 	struct uio uio;
3436 	char *data;
3437 
3438 	/* CURRENT_FH: directory */
3439 	vp = cs->vp;
3440 	if (vp == NULL) {
3441 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3442 		return;
3443 	}
3444 
3445 	if (cs->access == CS_ACCESS_DENIED) {
3446 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3447 		return;
3448 	}
3449 
3450 	if (vp->v_type == VDIR) {
3451 		*cs->statusp = resp->status = NFS4ERR_ISDIR;
3452 		return;
3453 	}
3454 
3455 	if (vp->v_type != VLNK) {
3456 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3457 		return;
3458 	}
3459 
3460 	va.va_mask = AT_MODE;
3461 	error = VOP_GETATTR(vp, &va, 0, cs->cr);
3462 	if (error) {
3463 		*cs->statusp = resp->status = puterrno4(error);
3464 		return;
3465 	}
3466 
3467 	if (MANDLOCK(vp, va.va_mode)) {
3468 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3469 		return;
3470 	}
3471 
3472 	data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP);
3473 
3474 	iov.iov_base = data;
3475 	iov.iov_len = MAXPATHLEN;
3476 	uio.uio_iov = &iov;
3477 	uio.uio_iovcnt = 1;
3478 	uio.uio_segflg = UIO_SYSSPACE;
3479 	uio.uio_extflg = UIO_COPY_CACHED;
3480 	uio.uio_loffset = 0;
3481 	uio.uio_resid = MAXPATHLEN;
3482 
3483 	error = VOP_READLINK(vp, &uio, cs->cr);
3484 
3485 	if (error) {
3486 		kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1);
3487 		*cs->statusp = resp->status = puterrno4(error);
3488 		return;
3489 	}
3490 
3491 	*(data + MAXPATHLEN - uio.uio_resid) = '\0';
3492 
3493 	/*
3494 	 * treat link name as data
3495 	 */
3496 	(void) str_to_utf8(data, &resp->link);
3497 
3498 	kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1);
3499 	*cs->statusp = resp->status = NFS4_OK;
3500 }
3501 
3502 static void
3503 rfs4_op_readlink_free(nfs_resop4 *resop)
3504 {
3505 	READLINK4res *resp = &resop->nfs_resop4_u.opreadlink;
3506 	utf8string *symlink = &resp->link;
3507 
3508 	if (symlink->utf8string_val) {
3509 		UTF8STRING_FREE(*symlink)
3510 	}
3511 }
3512 
3513 /*
3514  * release_lockowner:
3515  *	Release any state associated with the supplied
3516  *	lockowner. Note if any lo_state is holding locks we will not
3517  *	rele that lo_state and thus the lockowner will not be destroyed.
3518  *	A client using lock after the lock owner stateid has been released
3519  *	will suffer the consequence of NFS4ERR_BAD_STATEID and would have
3520  *	to reissue the lock with new_lock_owner set to TRUE.
3521  *	args: lock_owner
3522  *	res:  status
3523  */
3524 /* ARGSUSED */
3525 static void
3526 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop,
3527 	struct svc_req *req, struct compound_state *cs)
3528 {
3529 	RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner;
3530 	RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner;
3531 	rfs4_lockowner_t *lo;
3532 	rfs4_openowner_t *oop;
3533 	rfs4_state_t *sp;
3534 	rfs4_lo_state_t *lsp;
3535 	rfs4_client_t *cp;
3536 	bool_t create = FALSE;
3537 	locklist_t *llist;
3538 	sysid_t sysid;
3539 
3540 	/* Make sure there is a clientid around for this request */
3541 	cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE);
3542 
3543 	if (cp == NULL) {
3544 		*cs->statusp = resp->status =
3545 			rfs4_check_clientid(&ap->lock_owner.clientid, 0);
3546 		return;
3547 	}
3548 	rfs4_client_rele(cp);
3549 
3550 	lo = rfs4_findlockowner(&ap->lock_owner, &create);
3551 	if (lo == NULL) {
3552 		*cs->statusp = resp->status = NFS4_OK;
3553 		return;
3554 	}
3555 	ASSERT(lo->client != NULL);
3556 
3557 	/*
3558 	 * Check for EXPIRED client. If so will reap state with in a lease
3559 	 * period or on next set_clientid_confirm step
3560 	 */
3561 	if (rfs4_lease_expired(lo->client)) {
3562 		rfs4_lockowner_rele(lo);
3563 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
3564 		return;
3565 	}
3566 
3567 	/*
3568 	 * If no sysid has been assigned, then no locks exist; just return.
3569 	 */
3570 	rfs4_dbe_lock(lo->client->dbe);
3571 	if (lo->client->sysidt == LM_NOSYSID) {
3572 		rfs4_lockowner_rele(lo);
3573 		rfs4_dbe_unlock(lo->client->dbe);
3574 		return;
3575 	}
3576 
3577 	sysid = lo->client->sysidt;
3578 	rfs4_dbe_unlock(lo->client->dbe);
3579 
3580 	/*
3581 	 * Mark the lockowner invalid.
3582 	 */
3583 	rfs4_dbe_hide(lo->dbe);
3584 
3585 	/*
3586 	 * sysid-pid pair should now not be used since the lockowner is
3587 	 * invalid. If the client were to instantiate the lockowner again
3588 	 * it would be assigned a new pid. Thus we can get the list of
3589 	 * current locks.
3590 	 */
3591 
3592 	llist = flk_get_active_locks(sysid, lo->pid);
3593 	/* If we are still holding locks fail */
3594 	if (llist != NULL) {
3595 
3596 		*cs->statusp = resp->status = NFS4ERR_LOCKS_HELD;
3597 
3598 		flk_free_locklist(llist);
3599 		/*
3600 		 * We need to unhide the lockowner so the client can
3601 		 * try it again. The bad thing here is if the client
3602 		 * has a logic error that took it here in the first place
3603 		 * he probably has lost accounting of the locks that it
3604 		 * is holding. So we may have dangling state until the
3605 		 * open owner state is reaped via close. One scenario
3606 		 * that could possibly occur is that the client has
3607 		 * sent the unlock request(s) in separate threads
3608 		 * and has not waited for the replies before sending the
3609 		 * RELEASE_LOCKOWNER request. Presumably, it would expect
3610 		 * and deal appropriately with NFS4ERR_LOCKS_HELD, by
3611 		 * reissuing the request.
3612 		 */
3613 		rfs4_dbe_unhide(lo->dbe);
3614 		rfs4_lockowner_rele(lo);
3615 		return;
3616 	}
3617 
3618 	/*
3619 	 * For the corresponding client we need to check each open
3620 	 * owner for any opens that have lockowner state associated
3621 	 * with this lockowner.
3622 	 */
3623 
3624 	rfs4_dbe_lock(lo->client->dbe);
3625 	for (oop = lo->client->openownerlist.next->oop; oop != NULL;
3626 	    oop = oop->openownerlist.next->oop) {
3627 
3628 		rfs4_dbe_lock(oop->dbe);
3629 		for (sp = oop->ownerstateids.next->sp; sp != NULL;
3630 		    sp = sp->ownerstateids.next->sp) {
3631 
3632 			rfs4_dbe_lock(sp->dbe);
3633 			for (lsp = sp->lockownerlist.next->lsp;
3634 			    lsp != NULL; lsp = lsp->lockownerlist.next->lsp) {
3635 				if (lsp->locker == lo) {
3636 					rfs4_dbe_lock(lsp->dbe);
3637 					rfs4_dbe_invalidate(lsp->dbe);
3638 					rfs4_dbe_unlock(lsp->dbe);
3639 				}
3640 			}
3641 			rfs4_dbe_unlock(sp->dbe);
3642 		}
3643 		rfs4_dbe_unlock(oop->dbe);
3644 	}
3645 	rfs4_dbe_unlock(lo->client->dbe);
3646 
3647 	rfs4_lockowner_rele(lo);
3648 
3649 	*cs->statusp = resp->status = NFS4_OK;
3650 }
3651 
3652 /*
3653  * short utility function to lookup a file and recall the delegation
3654  */
3655 static rfs4_file_t *
3656 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp,
3657 	int *lkup_error, cred_t *cr)
3658 {
3659 	vnode_t *vp;
3660 	rfs4_file_t *fp = NULL;
3661 	bool_t fcreate = FALSE;
3662 	int error;
3663 
3664 	if (vpp)
3665 		*vpp = NULL;
3666 
3667 	if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr)) == 0) {
3668 		if (vp->v_type == VREG)
3669 			fp = rfs4_findfile(vp, NULL, &fcreate);
3670 		if (vpp)
3671 			*vpp = vp;
3672 		else
3673 			VN_RELE(vp);
3674 	}
3675 
3676 	if (lkup_error)
3677 		*lkup_error = error;
3678 
3679 	return (fp);
3680 }
3681 
3682 /*
3683  * remove: args: CURRENT_FH: directory; name.
3684  *	res: status. If success - CURRENT_FH unchanged, return change_info
3685  *		for directory.
3686  */
3687 /* ARGSUSED */
3688 static void
3689 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3690 	struct compound_state *cs)
3691 {
3692 	REMOVE4args *args = &argop->nfs_argop4_u.opremove;
3693 	REMOVE4res *resp = &resop->nfs_resop4_u.opremove;
3694 	int error;
3695 	vnode_t *dvp, *vp;
3696 	struct vattr bdva, idva, adva;
3697 	char *nm;
3698 	uint_t len;
3699 	rfs4_file_t *fp;
3700 	int in_crit = 0;
3701 	bslabel_t *clabel;
3702 
3703 	/* CURRENT_FH: directory */
3704 	dvp = cs->vp;
3705 	if (dvp == NULL) {
3706 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3707 		return;
3708 	}
3709 
3710 	if (cs->access == CS_ACCESS_DENIED) {
3711 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3712 		return;
3713 	}
3714 
3715 	/*
3716 	 * If there is an unshared filesystem mounted on this vnode,
3717 	 * Do not allow to remove anything in this directory.
3718 	 */
3719 	if (vn_ismntpt(dvp)) {
3720 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3721 		return;
3722 	}
3723 
3724 	if (dvp->v_type != VDIR) {
3725 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
3726 		return;
3727 	}
3728 
3729 	if (!utf8_dir_verify(&args->target)) {
3730 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3731 		return;
3732 	}
3733 
3734 	/*
3735 	 * Lookup the file so that we can check if it's a directory
3736 	 */
3737 	nm = utf8_to_fn(&args->target, &len, NULL);
3738 	if (nm == NULL) {
3739 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3740 		return;
3741 	}
3742 
3743 	if (len > MAXNAMELEN) {
3744 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
3745 		kmem_free(nm, len);
3746 		return;
3747 	}
3748 
3749 	if (rdonly4(cs->exi, cs->vp, req)) {
3750 		*cs->statusp = resp->status = NFS4ERR_ROFS;
3751 		kmem_free(nm, len);
3752 		return;
3753 	}
3754 
3755 	/*
3756 	 * Lookup the file to determine type and while we are see if
3757 	 * there is a file struct around and check for delegation.
3758 	 * We don't need to acquire va_seq before this lookup, if
3759 	 * it causes an update, cinfo.before will not match, which will
3760 	 * trigger a cache flush even if atomic is TRUE.
3761 	 */
3762 	if (fp = rfs4_lookup_and_findfile(dvp, nm, &vp, &error, cs->cr)) {
3763 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
3764 						NULL)) {
3765 			VN_RELE(vp);
3766 			rfs4_file_rele(fp);
3767 			*cs->statusp = resp->status = NFS4ERR_DELAY;
3768 			kmem_free(nm, len);
3769 			return;
3770 		}
3771 	}
3772 
3773 	/* Didn't find anything to remove */
3774 	if (vp == NULL) {
3775 		*cs->statusp = resp->status = error;
3776 		kmem_free(nm, len);
3777 		return;
3778 	}
3779 
3780 	if (nbl_need_check(vp)) {
3781 		nbl_start_crit(vp, RW_READER);
3782 		in_crit = 1;
3783 		if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0)) {
3784 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
3785 			kmem_free(nm, len);
3786 			nbl_end_crit(vp);
3787 			VN_RELE(vp);
3788 			if (fp) {
3789 				rfs4_clear_dont_grant(fp);
3790 				rfs4_file_rele(fp);
3791 			}
3792 			return;
3793 		}
3794 	}
3795 
3796 	/* check label before allowing removal */
3797 	if (is_system_labeled()) {
3798 		ASSERT(req->rq_label != NULL);
3799 		clabel = req->rq_label;
3800 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
3801 		    "got client label from request(1)",
3802 		    struct svc_req *, req);
3803 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
3804 			if (!do_rfs4_label_check(clabel, vp, EQUALITY_CHECK)) {
3805 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
3806 				kmem_free(nm, len);
3807 				if (in_crit)
3808 					nbl_end_crit(vp);
3809 				VN_RELE(vp);
3810 				if (fp) {
3811 					rfs4_clear_dont_grant(fp);
3812 					rfs4_file_rele(fp);
3813 				}
3814 				return;
3815 			}
3816 		}
3817 	}
3818 
3819 	/* Get dir "before" change value */
3820 	bdva.va_mask = AT_CTIME|AT_SEQ;
3821 	error = VOP_GETATTR(dvp, &bdva, 0, cs->cr);
3822 	if (error) {
3823 		*cs->statusp = resp->status = puterrno4(error);
3824 		kmem_free(nm, len);
3825 		return;
3826 	}
3827 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime)
3828 
3829 	/* Actually do the REMOVE operation */
3830 	if (vp->v_type == VDIR) {
3831 		/*
3832 		 * Can't remove a directory that has a mounted-on filesystem.
3833 		 */
3834 		if (vn_ismntpt(vp)) {
3835 			error = EACCES;
3836 		} else {
3837 			/*
3838 			 * System V defines rmdir to return EEXIST,
3839 			 * not * ENOTEMPTY, if the directory is not
3840 			 * empty.  A System V NFS server needs to map
3841 			 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to
3842 			 * transmit over the wire.
3843 			 */
3844 			if ((error = VOP_RMDIR(dvp, nm, rootdir, cs->cr))
3845 				== EEXIST)
3846 				error = ENOTEMPTY;
3847 		}
3848 	} else {
3849 		if ((error = VOP_REMOVE(dvp, nm, cs->cr)) == 0 &&
3850 			fp != NULL) {
3851 			struct vattr va;
3852 			vnode_t *tvp;
3853 
3854 			rfs4_dbe_lock(fp->dbe);
3855 			tvp = fp->vp;
3856 			if (tvp)
3857 				VN_HOLD(tvp);
3858 			rfs4_dbe_unlock(fp->dbe);
3859 
3860 			if (tvp) {
3861 				/*
3862 				 * This is va_seq safe because we are not
3863 				 * manipulating dvp.
3864 				 */
3865 				va.va_mask = AT_NLINK;
3866 				if (!VOP_GETATTR(tvp, &va, 0, cs->cr) &&
3867 					va.va_nlink == 0) {
3868 					/* Remove state on file remove */
3869 					if (in_crit) {
3870 						nbl_end_crit(vp);
3871 						in_crit = 0;
3872 					}
3873 					rfs4_close_all_state(fp);
3874 				}
3875 				VN_RELE(tvp);
3876 			}
3877 		}
3878 	}
3879 
3880 	if (in_crit)
3881 		nbl_end_crit(vp);
3882 	VN_RELE(vp);
3883 
3884 	if (fp) {
3885 		rfs4_clear_dont_grant(fp);
3886 		rfs4_file_rele(fp);
3887 	}
3888 	kmem_free(nm, len);
3889 
3890 	if (error) {
3891 		*cs->statusp = resp->status = puterrno4(error);
3892 		return;
3893 	}
3894 
3895 	/*
3896 	 * Get the initial "after" sequence number, if it fails, set to zero
3897 	 */
3898 	idva.va_mask = AT_SEQ;
3899 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr))
3900 		idva.va_seq = 0;
3901 
3902 	/*
3903 	 * Force modified data and metadata out to stable storage.
3904 	 */
3905 	(void) VOP_FSYNC(dvp, 0, cs->cr);
3906 
3907 	/*
3908 	 * Get "after" change value, if it fails, simply return the
3909 	 * before value.
3910 	 */
3911 	adva.va_mask = AT_CTIME|AT_SEQ;
3912 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr)) {
3913 		adva.va_ctime = bdva.va_ctime;
3914 		adva.va_seq = 0;
3915 	}
3916 
3917 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
3918 
3919 	/*
3920 	 * The cinfo.atomic = TRUE only if we have
3921 	 * non-zero va_seq's, and it has incremented by exactly one
3922 	 * during the VOP_REMOVE/RMDIR and it didn't change during
3923 	 * the VOP_FSYNC.
3924 	 */
3925 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
3926 			idva.va_seq == (bdva.va_seq + 1) &&
3927 			idva.va_seq == adva.va_seq)
3928 		resp->cinfo.atomic = TRUE;
3929 	else
3930 		resp->cinfo.atomic = FALSE;
3931 
3932 	*cs->statusp = resp->status = NFS4_OK;
3933 }
3934 
3935 /*
3936  * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory,
3937  *		oldname and newname.
3938  *	res: status. If success - CURRENT_FH unchanged, return change_info
3939  *		for both from and target directories.
3940  */
3941 /* ARGSUSED */
3942 static void
3943 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3944 	struct compound_state *cs)
3945 {
3946 	RENAME4args *args = &argop->nfs_argop4_u.oprename;
3947 	RENAME4res *resp = &resop->nfs_resop4_u.oprename;
3948 	int error;
3949 	vnode_t *odvp;
3950 	vnode_t *ndvp;
3951 	vnode_t *srcvp, *targvp;
3952 	struct vattr obdva, oidva, oadva;
3953 	struct vattr nbdva, nidva, nadva;
3954 	char *onm, *nnm;
3955 	uint_t olen, nlen;
3956 	rfs4_file_t *fp, *sfp;
3957 	int in_crit_src, in_crit_targ;
3958 	int fp_rele_grant_hold, sfp_rele_grant_hold;
3959 	bslabel_t *clabel;
3960 
3961 	fp = sfp = NULL;
3962 	srcvp = targvp = NULL;
3963 	in_crit_src = in_crit_targ = 0;
3964 	fp_rele_grant_hold = sfp_rele_grant_hold = 0;
3965 
3966 	/* CURRENT_FH: target directory */
3967 	ndvp = cs->vp;
3968 	if (ndvp == NULL) {
3969 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3970 		return;
3971 	}
3972 
3973 	/* SAVED_FH: from directory */
3974 	odvp = cs->saved_vp;
3975 	if (odvp == NULL) {
3976 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3977 		return;
3978 	}
3979 
3980 	if (cs->access == CS_ACCESS_DENIED) {
3981 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3982 		return;
3983 	}
3984 
3985 	/*
3986 	 * If there is an unshared filesystem mounted on this vnode,
3987 	 * do not allow to rename objects in this directory.
3988 	 */
3989 	if (vn_ismntpt(odvp)) {
3990 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3991 		return;
3992 	}
3993 
3994 	/*
3995 	 * If there is an unshared filesystem mounted on this vnode,
3996 	 * do not allow to rename to this directory.
3997 	 */
3998 	if (vn_ismntpt(ndvp)) {
3999 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4000 		return;
4001 	}
4002 
4003 	if (odvp->v_type != VDIR || ndvp->v_type != VDIR) {
4004 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
4005 		return;
4006 	}
4007 
4008 	if (cs->saved_exi != cs->exi) {
4009 		*cs->statusp = resp->status = NFS4ERR_XDEV;
4010 		return;
4011 	}
4012 
4013 	if (!utf8_dir_verify(&args->oldname)) {
4014 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4015 		return;
4016 	}
4017 
4018 	if (!utf8_dir_verify(&args->newname)) {
4019 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4020 		return;
4021 	}
4022 
4023 	onm = utf8_to_fn(&args->oldname, &olen, NULL);
4024 	if (onm == NULL) {
4025 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4026 		return;
4027 	}
4028 
4029 	nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4030 	if (nnm == NULL) {
4031 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4032 		kmem_free(onm, olen);
4033 		return;
4034 	}
4035 
4036 	if (olen > MAXNAMELEN || nlen > MAXNAMELEN) {
4037 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
4038 		kmem_free(onm, olen);
4039 		kmem_free(nnm, nlen);
4040 		return;
4041 	}
4042 
4043 
4044 	if (rdonly4(cs->exi, cs->vp, req)) {
4045 		*cs->statusp = resp->status = NFS4ERR_ROFS;
4046 		kmem_free(onm, olen);
4047 		kmem_free(nnm, nlen);
4048 		return;
4049 	}
4050 
4051 	/* check label of the target dir */
4052 	if (is_system_labeled()) {
4053 		ASSERT(req->rq_label != NULL);
4054 		clabel = req->rq_label;
4055 		DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *,
4056 		    "got client label from request(1)",
4057 		    struct svc_req *, req);
4058 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
4059 			if (!do_rfs4_label_check(clabel, ndvp,
4060 			    EQUALITY_CHECK)) {
4061 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
4062 				return;
4063 			}
4064 		}
4065 	}
4066 
4067 	/*
4068 	 * Is the source a file and have a delegation?
4069 	 * We don't need to acquire va_seq before these lookups, if
4070 	 * it causes an update, cinfo.before will not match, which will
4071 	 * trigger a cache flush even if atomic is TRUE.
4072 	 */
4073 	if (sfp = rfs4_lookup_and_findfile(odvp, onm, &srcvp, &error, cs->cr)) {
4074 		if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE,
4075 						NULL)) {
4076 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4077 			goto err_out;
4078 		}
4079 	}
4080 
4081 	if (srcvp == NULL) {
4082 		*cs->statusp = resp->status = puterrno4(error);
4083 		kmem_free(onm, olen);
4084 		kmem_free(nnm, nlen);
4085 		return;
4086 	}
4087 
4088 	sfp_rele_grant_hold = 1;
4089 
4090 	/* Does the destination exist and a file and have a delegation? */
4091 	if (fp = rfs4_lookup_and_findfile(ndvp, nnm, &targvp, NULL, cs->cr)) {
4092 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
4093 						NULL)) {
4094 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4095 			goto err_out;
4096 		}
4097 	}
4098 	fp_rele_grant_hold = 1;
4099 
4100 
4101 	/* Check for NBMAND lock on both source and target */
4102 	if (nbl_need_check(srcvp)) {
4103 		nbl_start_crit(srcvp, RW_READER);
4104 		in_crit_src = 1;
4105 		if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0)) {
4106 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4107 			goto err_out;
4108 		}
4109 	}
4110 
4111 	if (targvp && nbl_need_check(targvp)) {
4112 		nbl_start_crit(targvp, RW_READER);
4113 		in_crit_targ = 1;
4114 		if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0)) {
4115 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4116 			goto err_out;
4117 		}
4118 	}
4119 
4120 	/* Get source "before" change value */
4121 	obdva.va_mask = AT_CTIME|AT_SEQ;
4122 	error = VOP_GETATTR(odvp, &obdva, 0, cs->cr);
4123 	if (!error) {
4124 		nbdva.va_mask = AT_CTIME|AT_SEQ;
4125 		error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr);
4126 	}
4127 	if (error) {
4128 		*cs->statusp = resp->status = puterrno4(error);
4129 		goto err_out;
4130 	}
4131 
4132 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime)
4133 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime)
4134 
4135 	if ((error = VOP_RENAME(odvp, onm, ndvp, nnm, cs->cr)) == 0 &&
4136 		fp != NULL) {
4137 		struct vattr va;
4138 		vnode_t *tvp;
4139 
4140 		rfs4_dbe_lock(fp->dbe);
4141 		tvp = fp->vp;
4142 		if (tvp)
4143 			VN_HOLD(tvp);
4144 		rfs4_dbe_unlock(fp->dbe);
4145 
4146 		if (tvp) {
4147 			va.va_mask = AT_NLINK;
4148 			if (!VOP_GETATTR(tvp, &va, 0, cs->cr) &&
4149 				va.va_nlink == 0) {
4150 				/* The file is gone and so should the state */
4151 				if (in_crit_targ) {
4152 					nbl_end_crit(targvp);
4153 					in_crit_targ = 0;
4154 				}
4155 				rfs4_close_all_state(fp);
4156 			}
4157 			VN_RELE(tvp);
4158 		}
4159 	}
4160 	if (error == 0) {
4161 		char *tmp;
4162 
4163 		/* fix the path name for the renamed file */
4164 		mutex_enter(&srcvp->v_lock);
4165 		tmp = srcvp->v_path;
4166 		srcvp->v_path = NULL;
4167 		mutex_exit(&srcvp->v_lock);
4168 		vn_setpath(rootdir, ndvp, srcvp, nnm, nlen - 1);
4169 		if (tmp != NULL)
4170 			kmem_free(tmp, strlen(tmp) + 1);
4171 	}
4172 
4173 	if (in_crit_src)
4174 		nbl_end_crit(srcvp);
4175 	if (srcvp)
4176 		VN_RELE(srcvp);
4177 	if (in_crit_targ)
4178 		nbl_end_crit(targvp);
4179 	if (targvp)
4180 		VN_RELE(targvp);
4181 
4182 	if (sfp) {
4183 		rfs4_clear_dont_grant(sfp);
4184 		rfs4_file_rele(sfp);
4185 	}
4186 	if (fp) {
4187 		rfs4_clear_dont_grant(fp);
4188 		rfs4_file_rele(fp);
4189 	}
4190 
4191 	kmem_free(onm, olen);
4192 	kmem_free(nnm, nlen);
4193 
4194 	/*
4195 	 * Get the initial "after" sequence number, if it fails, set to zero
4196 	 */
4197 	oidva.va_mask = AT_SEQ;
4198 	if (VOP_GETATTR(odvp, &oidva, 0, cs->cr))
4199 		oidva.va_seq = 0;
4200 
4201 	nidva.va_mask = AT_SEQ;
4202 	if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr))
4203 		nidva.va_seq = 0;
4204 
4205 	/*
4206 	 * Force modified data and metadata out to stable storage.
4207 	 */
4208 	(void) VOP_FSYNC(odvp, 0, cs->cr);
4209 	(void) VOP_FSYNC(ndvp, 0, cs->cr);
4210 
4211 	if (error) {
4212 		*cs->statusp = resp->status = puterrno4(error);
4213 		return;
4214 	}
4215 
4216 	/*
4217 	 * Get "after" change values, if it fails, simply return the
4218 	 * before value.
4219 	 */
4220 	oadva.va_mask = AT_CTIME|AT_SEQ;
4221 	if (VOP_GETATTR(odvp, &oadva, 0, cs->cr)) {
4222 		oadva.va_ctime = obdva.va_ctime;
4223 		oadva.va_seq = 0;
4224 	}
4225 
4226 	nadva.va_mask = AT_CTIME|AT_SEQ;
4227 	if (VOP_GETATTR(odvp, &nadva, 0, cs->cr)) {
4228 		nadva.va_ctime = nbdva.va_ctime;
4229 		nadva.va_seq = 0;
4230 	}
4231 
4232 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime)
4233 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime)
4234 
4235 	/*
4236 	 * The cinfo.atomic = TRUE only if we have
4237 	 * non-zero va_seq's, and it has incremented by exactly one
4238 	 * during the VOP_RENAME and it didn't change during the VOP_FSYNC.
4239 	 */
4240 	if (obdva.va_seq && oidva.va_seq && oadva.va_seq &&
4241 			oidva.va_seq == (obdva.va_seq + 1) &&
4242 			oidva.va_seq == oadva.va_seq)
4243 		resp->source_cinfo.atomic = TRUE;
4244 	else
4245 		resp->source_cinfo.atomic = FALSE;
4246 
4247 	if (nbdva.va_seq && nidva.va_seq && nadva.va_seq &&
4248 			nidva.va_seq == (nbdva.va_seq + 1) &&
4249 			nidva.va_seq == nadva.va_seq)
4250 		resp->target_cinfo.atomic = TRUE;
4251 	else
4252 		resp->target_cinfo.atomic = FALSE;
4253 
4254 #ifdef	VOLATILE_FH_TEST
4255 	{
4256 	extern void add_volrnm_fh(struct exportinfo *, vnode_t *);
4257 
4258 	/*
4259 	 * Add the renamed file handle to the volatile rename list
4260 	 */
4261 	if (cs->exi->exi_export.ex_flags & EX_VOLRNM) {
4262 		/* file handles may expire on rename */
4263 		vnode_t *vp;
4264 
4265 		nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4266 		/*
4267 		 * Already know that nnm will be a valid string
4268 		 */
4269 		error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr);
4270 		kmem_free(nnm, nlen);
4271 		if (!error) {
4272 			add_volrnm_fh(cs->exi, vp);
4273 			VN_RELE(vp);
4274 		}
4275 	}
4276 	}
4277 #endif	/* VOLATILE_FH_TEST */
4278 
4279 	*cs->statusp = resp->status = NFS4_OK;
4280 	return;
4281 
4282 err_out:
4283 	kmem_free(onm, olen);
4284 	kmem_free(nnm, nlen);
4285 
4286 	if (in_crit_src) nbl_end_crit(srcvp);
4287 	if (in_crit_targ) nbl_end_crit(targvp);
4288 	if (targvp) VN_RELE(targvp);
4289 	if (srcvp) VN_RELE(srcvp);
4290 	if (sfp) {
4291 		if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp);
4292 		rfs4_file_rele(sfp);
4293 	}
4294 	if (fp) {
4295 		if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp);
4296 		rfs4_file_rele(fp);
4297 	}
4298 }
4299 
4300 /* ARGSUSED */
4301 static void
4302 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4303 	struct compound_state *cs)
4304 {
4305 	RENEW4args *args = &argop->nfs_argop4_u.oprenew;
4306 	RENEW4res *resp = &resop->nfs_resop4_u.oprenew;
4307 	rfs4_client_t *cp;
4308 
4309 	if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) {
4310 		*cs->statusp = resp->status =
4311 			rfs4_check_clientid(&args->clientid, 0);
4312 		return;
4313 	}
4314 
4315 	if (rfs4_lease_expired(cp)) {
4316 		rfs4_client_rele(cp);
4317 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
4318 		return;
4319 	}
4320 
4321 	rfs4_update_lease(cp);
4322 
4323 	mutex_enter(cp->cbinfo.cb_lock);
4324 	if (cp->cbinfo.cb_notified_of_cb_path_down == FALSE) {
4325 		cp->cbinfo.cb_notified_of_cb_path_down = TRUE;
4326 		*cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN;
4327 	} else {
4328 		*cs->statusp = resp->status = NFS4_OK;
4329 	}
4330 	mutex_exit(cp->cbinfo.cb_lock);
4331 
4332 	rfs4_client_rele(cp);
4333 
4334 }
4335 
4336 /* ARGSUSED */
4337 static void
4338 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
4339 	struct compound_state *cs)
4340 {
4341 	RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh;
4342 
4343 	/* No need to check cs->access - we are not accessing any object */
4344 	if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) {
4345 		*cs->statusp = resp->status = NFS4ERR_RESTOREFH;
4346 		return;
4347 	}
4348 	if (cs->vp != NULL) {
4349 		VN_RELE(cs->vp);
4350 	}
4351 	cs->vp = cs->saved_vp;
4352 	cs->saved_vp = NULL;
4353 	cs->exi = cs->saved_exi;
4354 	nfs_fh4_copy(&cs->saved_fh, &cs->fh);
4355 	*cs->statusp = resp->status = NFS4_OK;
4356 	cs->deleg = FALSE;
4357 }
4358 
4359 /* ARGSUSED */
4360 static void
4361 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4362 	struct compound_state *cs)
4363 {
4364 	SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh;
4365 
4366 	/* No need to check cs->access - we are not accessing any object */
4367 	if (cs->vp == NULL) {
4368 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4369 		return;
4370 	}
4371 	if (cs->saved_vp != NULL) {
4372 		VN_RELE(cs->saved_vp);
4373 	}
4374 	cs->saved_vp = cs->vp;
4375 	VN_HOLD(cs->saved_vp);
4376 	cs->saved_exi = cs->exi;
4377 	/*
4378 	 * since SAVEFH is fairly rare, don't alloc space for its fh
4379 	 * unless necessary.
4380 	 */
4381 	if (cs->saved_fh.nfs_fh4_val == NULL) {
4382 		cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP);
4383 	}
4384 	nfs_fh4_copy(&cs->fh, &cs->saved_fh);
4385 	*cs->statusp = resp->status = NFS4_OK;
4386 }
4387 
4388 /*
4389  * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to
4390  * return the bitmap of attrs that were set successfully. It is also
4391  * called by Verify/Nverify to test the vattr/vfsstat attrs. It should
4392  * always be called only after rfs4_do_set_attrs().
4393  *
4394  * Verify that the attributes are same as the expected ones. sargp->vap
4395  * and sargp->sbp contain the input attributes as translated from fattr4.
4396  *
4397  * This function verifies only the attrs that correspond to a vattr or
4398  * vfsstat struct. That is because of the extra step needed to get the
4399  * corresponding system structs. Other attributes have already been set or
4400  * verified by do_rfs4_set_attrs.
4401  *
4402  * Return 0 if all attrs match, -1 if some don't, error if error processing.
4403  */
4404 static int
4405 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp,
4406 	bitmap4 *resp, struct nfs4_ntov_table *ntovp)
4407 {
4408 	int error, ret_error = 0;
4409 	int i, k;
4410 	uint_t sva_mask = sargp->vap->va_mask;
4411 	uint_t vbit;
4412 	union nfs4_attr_u *na;
4413 	uint8_t *amap;
4414 	bool_t getsb = ntovp->vfsstat;
4415 
4416 	if (sva_mask != 0) {
4417 		/*
4418 		 * Okay to overwrite sargp->vap because we verify based
4419 		 * on the incoming values.
4420 		 */
4421 		ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0,
4422 				sargp->cs->cr);
4423 		if (ret_error) {
4424 			if (resp == NULL)
4425 				return (ret_error);
4426 			/*
4427 			 * Must return bitmap of successful attrs
4428 			 */
4429 			sva_mask = 0;	/* to prevent checking vap later */
4430 		} else {
4431 			/*
4432 			 * Some file systems clobber va_mask. it is probably
4433 			 * wrong of them to do so, nonethless we practice
4434 			 * defensive coding.
4435 			 * See bug id 4276830.
4436 			 */
4437 			sargp->vap->va_mask = sva_mask;
4438 		}
4439 	}
4440 
4441 	if (getsb) {
4442 		/*
4443 		 * Now get the superblock and loop on the bitmap, as there is
4444 		 * no simple way of translating from superblock to bitmap4.
4445 		 */
4446 		ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp);
4447 		if (ret_error) {
4448 			if (resp == NULL)
4449 				goto errout;
4450 			getsb = FALSE;
4451 		}
4452 	}
4453 
4454 	/*
4455 	 * Now loop and verify each attribute which getattr returned
4456 	 * whether it's the same as the input.
4457 	 */
4458 	if (resp == NULL && !getsb && (sva_mask == 0))
4459 		goto errout;
4460 
4461 	na = ntovp->na;
4462 	amap = ntovp->amap;
4463 	k = 0;
4464 	for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) {
4465 		k = *amap;
4466 		ASSERT(nfs4_ntov_map[k].nval == k);
4467 		vbit = nfs4_ntov_map[k].vbit;
4468 
4469 		/*
4470 		 * If vattr attribute but VOP_GETATTR failed, or it's
4471 		 * superblock attribute but VFS_STATVFS failed, skip
4472 		 */
4473 		if (vbit) {
4474 			if ((vbit & sva_mask) == 0)
4475 				continue;
4476 		} else if (!(getsb && nfs4_ntov_map[k].vfsstat)) {
4477 			continue;
4478 		}
4479 		error = (*nfs4_ntov_map[k].sv_getit)(
4480 				NFS4ATTR_VERIT, sargp, na);
4481 		if (resp != NULL) {
4482 			if (error)
4483 				ret_error = -1;	/* not all match */
4484 			else	/* update response bitmap */
4485 				*resp |= nfs4_ntov_map[k].fbit;
4486 			continue;
4487 		}
4488 		if (error) {
4489 			ret_error = -1;	/* not all match */
4490 			break;
4491 		}
4492 	}
4493 errout:
4494 	return (ret_error);
4495 }
4496 
4497 /*
4498  * Decode the attribute to be set/verified. If the attr requires a sys op
4499  * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't
4500  * call the sv_getit function for it, because the sys op hasn't yet been done.
4501  * Return 0 for success, error code if failed.
4502  *
4503  * Note: the decoded arg is not freed here but in nfs4_ntov_table_free.
4504  */
4505 static int
4506 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp,
4507 	int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap)
4508 {
4509 	int error = 0;
4510 	bool_t set_later;
4511 
4512 	sargp->vap->va_mask |= nfs4_ntov_map[k].vbit;
4513 
4514 	if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) {
4515 		set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat;
4516 		/*
4517 		 * don't verify yet if a vattr or sb dependent attr,
4518 		 * because we don't have their sys values yet.
4519 		 * Will be done later.
4520 		 */
4521 		if (! (set_later && (cmd == NFS4ATTR_VERIT))) {
4522 			/*
4523 			 * ACLs are a special case, since setting the MODE
4524 			 * conflicts with setting the ACL.  We delay setting
4525 			 * the ACL until all other attributes have been set.
4526 			 * The ACL gets set in do_rfs4_op_setattr().
4527 			 */
4528 			if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) {
4529 				error = (*nfs4_ntov_map[k].sv_getit)(cmd,
4530 				    sargp, nap);
4531 				if (error) {
4532 					xdr_free(nfs4_ntov_map[k].xfunc,
4533 					    (caddr_t)nap);
4534 				}
4535 			}
4536 		}
4537 	} else {
4538 #ifdef  DEBUG
4539 		cmn_err(CE_NOTE, "decode_fattr4_attr: error "
4540 			"decoding attribute %d\n", k);
4541 #endif
4542 		error = EINVAL;
4543 	}
4544 	if (!error && resp_bval && !set_later) {
4545 		*resp_bval |= nfs4_ntov_map[k].fbit;
4546 	}
4547 
4548 	return (error);
4549 }
4550 
4551 /*
4552  * Set vattr based on incoming fattr4 attrs - used by setattr.
4553  * Set response mask. Ignore any values that are not writable vattr attrs.
4554  */
4555 static nfsstat4
4556 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
4557 		struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp,
4558 		nfs4_attr_cmd_t cmd)
4559 {
4560 	int error = 0;
4561 	int i;
4562 	char *attrs = fattrp->attrlist4;
4563 	uint32_t attrslen = fattrp->attrlist4_len;
4564 	XDR xdr;
4565 	nfsstat4 status = NFS4_OK;
4566 	vnode_t *vp = cs->vp;
4567 	union nfs4_attr_u *na;
4568 	uint8_t *amap;
4569 
4570 #ifndef lint
4571 	/*
4572 	 * Make sure that maximum attribute number can be expressed as an
4573 	 * 8 bit quantity.
4574 	 */
4575 	ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1));
4576 #endif
4577 
4578 	if (vp == NULL) {
4579 		if (resp)
4580 			*resp = 0;
4581 		return (NFS4ERR_NOFILEHANDLE);
4582 	}
4583 	if (cs->access == CS_ACCESS_DENIED) {
4584 		if (resp)
4585 			*resp = 0;
4586 		return (NFS4ERR_ACCESS);
4587 	}
4588 
4589 	sargp->op = cmd;
4590 	sargp->cs = cs;
4591 	sargp->flag = 0;	/* may be set later */
4592 	sargp->vap->va_mask = 0;
4593 	sargp->rdattr_error = NFS4_OK;
4594 	sargp->rdattr_error_req = FALSE;
4595 	/* sargp->sbp is set by the caller */
4596 
4597 	xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE);
4598 
4599 	na = ntovp->na;
4600 	amap = ntovp->amap;
4601 
4602 	/*
4603 	 * The following loop iterates on the nfs4_ntov_map checking
4604 	 * if the fbit is set in the requested bitmap.
4605 	 * If set then we process the arguments using the
4606 	 * rfs4_fattr4 conversion functions to populate the setattr
4607 	 * vattr and va_mask. Any settable attrs that are not using vattr
4608 	 * will be set in this loop.
4609 	 */
4610 	for (i = 0; i < nfs4_ntov_map_size; i++) {
4611 		if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) {
4612 			continue;
4613 		}
4614 		/*
4615 		 * If setattr, must be a writable attr.
4616 		 * If verify/nverify, must be a readable attr.
4617 		 */
4618 		if ((error = (*nfs4_ntov_map[i].sv_getit)(
4619 				    NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) {
4620 			/*
4621 			 * Client tries to set/verify an
4622 			 * unsupported attribute, tries to set
4623 			 * a read only attr or verify a write
4624 			 * only one - error!
4625 			 */
4626 			break;
4627 		}
4628 		/*
4629 		 * Decode the attribute to set/verify
4630 		 */
4631 		error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval,
4632 					&xdr, resp ? resp : NULL, na);
4633 		if (error)
4634 			break;
4635 		*amap++ = (uint8_t)nfs4_ntov_map[i].nval;
4636 		na++;
4637 		(ntovp->attrcnt)++;
4638 		if (nfs4_ntov_map[i].vfsstat)
4639 			ntovp->vfsstat = TRUE;
4640 	}
4641 
4642 	if (error != 0)
4643 		status = (error == ENOTSUP ?	NFS4ERR_ATTRNOTSUPP :
4644 						puterrno4(error));
4645 	/* xdrmem_destroy(&xdrs); */	/* NO-OP */
4646 	return (status);
4647 }
4648 
4649 static nfsstat4
4650 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
4651 		stateid4 *stateid)
4652 {
4653 	int error = 0;
4654 	struct nfs4_svgetit_arg sarg;
4655 	bool_t trunc;
4656 
4657 	nfsstat4 status = NFS4_OK;
4658 	cred_t *cr = cs->cr;
4659 	vnode_t *vp = cs->vp;
4660 	struct nfs4_ntov_table ntov;
4661 	struct statvfs64 sb;
4662 	struct vattr bva;
4663 	struct flock64 bf;
4664 	int in_crit = 0;
4665 	uint_t saved_mask = 0;
4666 	caller_context_t ct;
4667 
4668 	*resp = 0;
4669 	sarg.sbp = &sb;
4670 	nfs4_ntov_table_init(&ntov);
4671 	status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov,
4672 			NFS4ATTR_SETIT);
4673 	if (status != NFS4_OK) {
4674 		/*
4675 		 * failed set attrs
4676 		 */
4677 		goto done;
4678 	}
4679 	if ((sarg.vap->va_mask == 0) &&
4680 	    (! (fattrp->attrmask & FATTR4_ACL_MASK))) {
4681 		/*
4682 		 * no further work to be done
4683 		 */
4684 		goto done;
4685 	}
4686 
4687 	/*
4688 	 * If we got a request to set the ACL and the MODE, only
4689 	 * allow changing VSUID, VSGID, and VSVTX.  Attempting
4690 	 * to change any other bits, along with setting an ACL,
4691 	 * gives NFS4ERR_INVAL.
4692 	 */
4693 	if ((fattrp->attrmask & FATTR4_ACL_MASK) &&
4694 	    (fattrp->attrmask & FATTR4_MODE_MASK)) {
4695 		vattr_t va;
4696 
4697 		va.va_mask = AT_MODE;
4698 		error = VOP_GETATTR(vp, &va, 0, cs->cr);
4699 		if (error) {
4700 			status = puterrno4(error);
4701 			goto done;
4702 		}
4703 		if ((sarg.vap->va_mode ^ va.va_mode) &
4704 		    ~(VSUID | VSGID | VSVTX)) {
4705 			status = NFS4ERR_INVAL;
4706 			goto done;
4707 		}
4708 	}
4709 
4710 	/* Check stateid only if size has been set */
4711 	if (sarg.vap->va_mask & AT_SIZE) {
4712 		trunc = (sarg.vap->va_size == 0);
4713 		status = rfs4_check_stateid(FWRITE, cs->vp, stateid,
4714 			trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE);
4715 		if (status != NFS4_OK)
4716 			goto done;
4717 	}
4718 
4719 	ct.cc_sysid = 0;
4720 	ct.cc_pid = 0;
4721 	ct.cc_caller_id = nfs4_srv_caller_id;
4722 
4723 	/* XXX start of possible race with delegations */
4724 
4725 	/*
4726 	 * We need to specially handle size changes because it is
4727 	 * possible for the client to create a file with read-only
4728 	 * modes, but with the file opened for writing. If the client
4729 	 * then tries to set the file size, e.g. ftruncate(3C),
4730 	 * fcntl(F_FREESP), the normal access checking done in
4731 	 * VOP_SETATTR would prevent the client from doing it even though
4732 	 * it should be allowed to do so.  To get around this, we do the
4733 	 * access checking for ourselves and use VOP_SPACE which doesn't
4734 	 * do the access checking.
4735 	 * Also the client should not be allowed to change the file
4736 	 * size if there is a conflicting non-blocking mandatory lock in
4737 	 * the region of the change.
4738 	 */
4739 	if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) {
4740 		u_offset_t offset;
4741 		ssize_t length;
4742 
4743 		/*
4744 		 * ufs_setattr clears AT_SIZE from vap->va_mask, but
4745 		 * before returning, sarg.vap->va_mask is used to
4746 		 * generate the setattr reply bitmap.  We also clear
4747 		 * AT_SIZE below before calling VOP_SPACE.  For both
4748 		 * of these cases, the va_mask needs to be saved here
4749 		 * and restored after calling VOP_SETATTR.
4750 		 */
4751 		saved_mask = sarg.vap->va_mask;
4752 
4753 		/*
4754 		 * Check any possible conflict due to NBMAND locks.
4755 		 * Get into critical region before VOP_GETATTR, so the
4756 		 * size attribute is valid when checking conflicts.
4757 		 */
4758 		if (nbl_need_check(vp)) {
4759 			nbl_start_crit(vp, RW_READER);
4760 			in_crit = 1;
4761 		}
4762 
4763 		bva.va_mask = AT_UID|AT_SIZE;
4764 		if (error = VOP_GETATTR(vp, &bva, 0, cr)) {
4765 			status = puterrno4(error);
4766 			goto done;
4767 		}
4768 
4769 		if (in_crit) {
4770 			if (sarg.vap->va_size < bva.va_size) {
4771 				offset = sarg.vap->va_size;
4772 				length = bva.va_size - sarg.vap->va_size;
4773 			} else {
4774 				offset = bva.va_size;
4775 				length = sarg.vap->va_size - bva.va_size;
4776 			}
4777 			if (nbl_conflict(vp, NBL_WRITE, offset, length, 0)) {
4778 				status = NFS4ERR_LOCKED;
4779 				goto done;
4780 			}
4781 		}
4782 
4783 		if (crgetuid(cr) == bva.va_uid) {
4784 			sarg.vap->va_mask &= ~AT_SIZE;
4785 			bf.l_type = F_WRLCK;
4786 			bf.l_whence = 0;
4787 			bf.l_start = (off64_t)sarg.vap->va_size;
4788 			bf.l_len = 0;
4789 			bf.l_sysid = 0;
4790 			bf.l_pid = 0;
4791 			error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE,
4792 					(offset_t)sarg.vap->va_size, cr, &ct);
4793 		}
4794 	}
4795 
4796 	if (!error && sarg.vap->va_mask != 0)
4797 		error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct);
4798 
4799 	/* restore va_mask -- ufs_setattr clears AT_SIZE */
4800 	if (saved_mask & AT_SIZE)
4801 		sarg.vap->va_mask |= AT_SIZE;
4802 
4803 	/*
4804 	 * If an ACL was being set, it has been delayed until now,
4805 	 * in order to set the mode (via the VOP_SETATTR() above) first.
4806 	 */
4807 	if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) {
4808 		int i;
4809 
4810 		for (i = 0; i < NFS4_MAXNUM_ATTRS; i++)
4811 			if (ntov.amap[i] == FATTR4_ACL)
4812 				break;
4813 		if (i < NFS4_MAXNUM_ATTRS) {
4814 			error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)(
4815 			    NFS4ATTR_SETIT, &sarg, &ntov.na[i]);
4816 			if (error == 0) {
4817 				*resp |= FATTR4_ACL_MASK;
4818 			} else if (error == ENOTSUP) {
4819 				(void) rfs4_verify_attr(&sarg, resp, &ntov);
4820 				status = NFS4ERR_ATTRNOTSUPP;
4821 				goto done;
4822 			}
4823 		} else {
4824 			NFS4_DEBUG(rfs4_debug,
4825 			    (CE_NOTE, "do_rfs4_op_setattr: "
4826 			    "unable to find ACL in fattr4"));
4827 			error = EINVAL;
4828 		}
4829 	}
4830 
4831 	if (error) {
4832 		status = puterrno4(error);
4833 
4834 		/*
4835 		 * Set the response bitmap when setattr failed.
4836 		 * If VOP_SETATTR partially succeeded, test by doing a
4837 		 * VOP_GETATTR on the object and comparing the data
4838 		 * to the setattr arguments.
4839 		 */
4840 		(void) rfs4_verify_attr(&sarg, resp, &ntov);
4841 	} else {
4842 		/*
4843 		 * Force modified metadata out to stable storage.
4844 		 */
4845 		(void) VOP_FSYNC(vp, FNODSYNC, cr);
4846 		/*
4847 		 * Set response bitmap
4848 		 */
4849 		nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp);
4850 	}
4851 
4852 /* Return early and already have a NFSv4 error */
4853 done:
4854 	/*
4855 	 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr
4856 	 * conversion sets both readable and writeable NFS4 attrs
4857 	 * for AT_MTIME and AT_ATIME.  The line below masks out
4858 	 * unrequested attrs from the setattr result bitmap.  This
4859 	 * is placed after the done: label to catch the ATTRNOTSUP
4860 	 * case.
4861 	 */
4862 	*resp &= fattrp->attrmask;
4863 
4864 	if (in_crit)
4865 		nbl_end_crit(vp);
4866 
4867 	nfs4_ntov_table_free(&ntov, &sarg);
4868 
4869 	return (status);
4870 }
4871 
4872 /* ARGSUSED */
4873 static void
4874 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4875 	struct compound_state *cs)
4876 {
4877 	SETATTR4args *args = &argop->nfs_argop4_u.opsetattr;
4878 	SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr;
4879 	bslabel_t *clabel;
4880 
4881 	if (cs->vp == NULL) {
4882 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4883 		return;
4884 	}
4885 
4886 	/*
4887 	 * If there is an unshared filesystem mounted on this vnode,
4888 	 * do not allow to setattr on this vnode.
4889 	 */
4890 	if (vn_ismntpt(cs->vp)) {
4891 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4892 		return;
4893 	}
4894 
4895 	resp->attrsset = 0;
4896 
4897 	if (rdonly4(cs->exi, cs->vp, req)) {
4898 		*cs->statusp = resp->status = NFS4ERR_ROFS;
4899 		return;
4900 	}
4901 
4902 	/* check label before setting attributes */
4903 	if (is_system_labeled()) {
4904 		ASSERT(req->rq_label != NULL);
4905 		clabel = req->rq_label;
4906 		DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *,
4907 		    "got client label from request(1)",
4908 		    struct svc_req *, req);
4909 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
4910 			if (!do_rfs4_label_check(clabel, cs->vp,
4911 			    EQUALITY_CHECK)) {
4912 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
4913 				return;
4914 			}
4915 		}
4916 	}
4917 
4918 	*cs->statusp = resp->status =
4919 		do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs,
4920 			&args->stateid);
4921 }
4922 
4923 /* ARGSUSED */
4924 static void
4925 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4926 	struct compound_state *cs)
4927 {
4928 	/*
4929 	 * verify and nverify are exactly the same, except that nverify
4930 	 * succeeds when some argument changed, and verify succeeds when
4931 	 * when none changed.
4932 	 */
4933 
4934 	VERIFY4args  *args = &argop->nfs_argop4_u.opverify;
4935 	VERIFY4res *resp = &resop->nfs_resop4_u.opverify;
4936 
4937 	int error;
4938 	struct nfs4_svgetit_arg sarg;
4939 	struct statvfs64 sb;
4940 	struct nfs4_ntov_table ntov;
4941 
4942 	if (cs->vp == NULL) {
4943 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4944 		return;
4945 	}
4946 
4947 	sarg.sbp = &sb;
4948 	nfs4_ntov_table_init(&ntov);
4949 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
4950 				&sarg, &ntov, NFS4ATTR_VERIT);
4951 	if (resp->status != NFS4_OK) {
4952 		/*
4953 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
4954 		 * so could return -1 for "no match".
4955 		 */
4956 		if (resp->status == -1)
4957 			resp->status = NFS4ERR_NOT_SAME;
4958 		goto done;
4959 	}
4960 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
4961 	switch (error) {
4962 	case 0:
4963 		resp->status = NFS4_OK;
4964 		break;
4965 	case -1:
4966 		resp->status = NFS4ERR_NOT_SAME;
4967 		break;
4968 	default:
4969 		resp->status = puterrno4(error);
4970 		break;
4971 	}
4972 done:
4973 	*cs->statusp = resp->status;
4974 	nfs4_ntov_table_free(&ntov, &sarg);
4975 }
4976 
4977 /* ARGSUSED */
4978 static void
4979 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4980 	struct compound_state *cs)
4981 {
4982 	/*
4983 	 * verify and nverify are exactly the same, except that nverify
4984 	 * succeeds when some argument changed, and verify succeeds when
4985 	 * when none changed.
4986 	 */
4987 
4988 	NVERIFY4args  *args = &argop->nfs_argop4_u.opnverify;
4989 	NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify;
4990 
4991 	int error;
4992 	struct nfs4_svgetit_arg sarg;
4993 	struct statvfs64 sb;
4994 	struct nfs4_ntov_table ntov;
4995 
4996 	if (cs->vp == NULL) {
4997 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4998 		return;
4999 	}
5000 	sarg.sbp = &sb;
5001 	nfs4_ntov_table_init(&ntov);
5002 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
5003 				&sarg, &ntov, NFS4ATTR_VERIT);
5004 	if (resp->status != NFS4_OK) {
5005 		/*
5006 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
5007 		 * so could return -1 for "no match".
5008 		 */
5009 		if (resp->status == -1)
5010 			resp->status = NFS4_OK;
5011 		goto done;
5012 	}
5013 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
5014 	switch (error) {
5015 	case 0:
5016 		resp->status = NFS4ERR_SAME;
5017 		break;
5018 	case -1:
5019 		resp->status = NFS4_OK;
5020 		break;
5021 	default:
5022 		resp->status = puterrno4(error);
5023 		break;
5024 	}
5025 done:
5026 	*cs->statusp = resp->status;
5027 	nfs4_ntov_table_free(&ntov, &sarg);
5028 }
5029 
5030 /*
5031  * XXX - This should live in an NFS header file.
5032  */
5033 #define	MAX_IOVECS	12
5034 
5035 /* ARGSUSED */
5036 static void
5037 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5038 	struct compound_state *cs)
5039 {
5040 	WRITE4args  *args = &argop->nfs_argop4_u.opwrite;
5041 	WRITE4res *resp = &resop->nfs_resop4_u.opwrite;
5042 	int error;
5043 	vnode_t *vp;
5044 	struct vattr bva;
5045 	u_offset_t rlimit;
5046 	struct uio uio;
5047 	struct iovec iov[MAX_IOVECS];
5048 	struct iovec *iovp;
5049 	int iovcnt;
5050 	int ioflag;
5051 	cred_t *savecred, *cr;
5052 	bool_t *deleg = &cs->deleg;
5053 	nfsstat4 stat;
5054 	int in_crit = 0;
5055 
5056 	vp = cs->vp;
5057 	if (vp == NULL) {
5058 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5059 		return;
5060 	}
5061 	if (cs->access == CS_ACCESS_DENIED) {
5062 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5063 		return;
5064 	}
5065 
5066 	cr = cs->cr;
5067 
5068 	/*
5069 	 * We have to enter the critical region before calling VOP_RWLOCK
5070 	 * to avoid a deadlock with ufs.
5071 	 */
5072 	if (nbl_need_check(vp)) {
5073 		nbl_start_crit(vp, RW_READER);
5074 		in_crit = 1;
5075 		if (nbl_conflict(vp, NBL_WRITE,
5076 				args->offset, args->data_len, 0)) {
5077 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
5078 			goto out;
5079 		}
5080 	}
5081 
5082 	if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE,
5083 					deleg, TRUE)) != NFS4_OK) {
5084 		*cs->statusp = resp->status = stat;
5085 		goto out;
5086 	}
5087 
5088 	bva.va_mask = AT_MODE | AT_UID;
5089 	error = VOP_GETATTR(vp, &bva, 0, cr);
5090 
5091 	/*
5092 	 * If we can't get the attributes, then we can't do the
5093 	 * right access checking.  So, we'll fail the request.
5094 	 */
5095 	if (error) {
5096 		*cs->statusp = resp->status = puterrno4(error);
5097 		goto out;
5098 	}
5099 
5100 	if (rdonly4(cs->exi, cs->vp, req)) {
5101 		*cs->statusp = resp->status = NFS4ERR_ROFS;
5102 		goto out;
5103 	}
5104 
5105 	if (vp->v_type != VREG) {
5106 		*cs->statusp = resp->status =
5107 			((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
5108 		goto out;
5109 	}
5110 
5111 	if (crgetuid(cr) != bva.va_uid &&
5112 	    (error = VOP_ACCESS(vp, VWRITE, 0, cr))) {
5113 		*cs->statusp = resp->status = puterrno4(error);
5114 		goto out;
5115 	}
5116 
5117 	if (MANDLOCK(vp, bva.va_mode)) {
5118 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5119 		goto out;
5120 	}
5121 
5122 	if (args->data_len == 0) {
5123 		*cs->statusp = resp->status = NFS4_OK;
5124 		resp->count = 0;
5125 		resp->committed = args->stable;
5126 		resp->writeverf = Write4verf;
5127 		goto out;
5128 	}
5129 
5130 	if (args->mblk != NULL) {
5131 		mblk_t *m;
5132 		uint_t bytes, round_len;
5133 
5134 		iovcnt = 0;
5135 		bytes = 0;
5136 		round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT);
5137 		for (m = args->mblk;
5138 		    m != NULL && bytes < round_len;
5139 		    m = m->b_cont) {
5140 			iovcnt++;
5141 			bytes += MBLKL(m);
5142 		}
5143 #ifdef DEBUG
5144 		/* should have ended on an mblk boundary */
5145 		if (bytes != round_len) {
5146 			printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n",
5147 			    bytes, round_len, args->data_len);
5148 			printf("args=%p, args->mblk=%p, m=%p", (void *)args,
5149 			    (void *)args->mblk, (void *)m);
5150 			ASSERT(bytes == round_len);
5151 		}
5152 #endif
5153 		if (iovcnt <= MAX_IOVECS) {
5154 			iovp = iov;
5155 		} else {
5156 			iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP);
5157 		}
5158 		mblk_to_iov(args->mblk, iovcnt, iovp);
5159 	} else {
5160 		iovcnt = 1;
5161 		iovp = iov;
5162 		iovp->iov_base = args->data_val;
5163 		iovp->iov_len = args->data_len;
5164 	}
5165 
5166 	uio.uio_iov = iovp;
5167 	uio.uio_iovcnt = iovcnt;
5168 
5169 	uio.uio_segflg = UIO_SYSSPACE;
5170 	uio.uio_extflg = UIO_COPY_DEFAULT;
5171 	uio.uio_loffset = args->offset;
5172 	uio.uio_resid = args->data_len;
5173 	uio.uio_llimit = curproc->p_fsz_ctl;
5174 	rlimit = uio.uio_llimit - args->offset;
5175 	if (rlimit < (u_offset_t)uio.uio_resid)
5176 		uio.uio_resid = (int)rlimit;
5177 
5178 	if (args->stable == UNSTABLE4)
5179 		ioflag = 0;
5180 	else if (args->stable == FILE_SYNC4)
5181 		ioflag = FSYNC;
5182 	else if (args->stable == DATA_SYNC4)
5183 		ioflag = FDSYNC;
5184 	else {
5185 		if (iovp != iov)
5186 			kmem_free(iovp, sizeof (*iovp) * iovcnt);
5187 		*cs->statusp = resp->status = NFS4ERR_INVAL;
5188 		goto out;
5189 	}
5190 
5191 	/*
5192 	 * We're changing creds because VM may fault and we need
5193 	 * the cred of the current thread to be used if quota
5194 	 * checking is enabled.
5195 	 */
5196 	savecred = curthread->t_cred;
5197 	curthread->t_cred = cr;
5198 	error = do_io(FWRITE, vp, &uio, ioflag, cr);
5199 	curthread->t_cred = savecred;
5200 
5201 	if (iovp != iov)
5202 		kmem_free(iovp, sizeof (*iovp) * iovcnt);
5203 
5204 	if (error) {
5205 		*cs->statusp = resp->status = puterrno4(error);
5206 		goto out;
5207 	}
5208 
5209 	*cs->statusp = resp->status = NFS4_OK;
5210 	resp->count = args->data_len - uio.uio_resid;
5211 
5212 	if (ioflag == 0)
5213 		resp->committed = UNSTABLE4;
5214 	else
5215 		resp->committed = FILE_SYNC4;
5216 
5217 	resp->writeverf = Write4verf;
5218 
5219 out:
5220 	if (in_crit)
5221 		nbl_end_crit(vp);
5222 }
5223 
5224 
5225 /* XXX put in a header file */
5226 extern int	sec_svc_getcred(struct svc_req *, cred_t *,  caddr_t *, int *);
5227 
5228 void
5229 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi,
5230 	struct svc_req *req, cred_t *cr)
5231 {
5232 	uint_t i;
5233 	struct compound_state cs;
5234 
5235 	rfs4_init_compound_state(&cs);
5236 	/*
5237 	 * Form a reply tag by copying over the reqeuest tag.
5238 	 */
5239 	resp->tag.utf8string_val =
5240 				kmem_alloc(args->tag.utf8string_len, KM_SLEEP);
5241 	resp->tag.utf8string_len = args->tag.utf8string_len;
5242 	bcopy(args->tag.utf8string_val, resp->tag.utf8string_val,
5243 					resp->tag.utf8string_len);
5244 
5245 	cs.statusp = &resp->status;
5246 
5247 	/*
5248 	 * XXX for now, minorversion should be zero
5249 	 */
5250 	if (args->minorversion != NFS4_MINORVERSION) {
5251 		resp->array_len = 0;
5252 		resp->array = NULL;
5253 		resp->status = NFS4ERR_MINOR_VERS_MISMATCH;
5254 		return;
5255 	}
5256 
5257 	resp->array_len = args->array_len;
5258 	resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4),
5259 		KM_SLEEP);
5260 
5261 	ASSERT(exi == NULL);
5262 	ASSERT(cr == NULL);
5263 
5264 	cr = crget();
5265 	ASSERT(cr != NULL);
5266 
5267 	if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) {
5268 		crfree(cr);
5269 		return;
5270 	}
5271 
5272 	cs.basecr = cr;
5273 
5274 	cs.req = req;
5275 
5276 	/*
5277 	 * For now, NFS4 compound processing must be protected by
5278 	 * exported_lock because it can access more than one exportinfo
5279 	 * per compound and share/unshare can now change multiple
5280 	 * exinfo structs.  The NFS2/3 code only refs 1 exportinfo
5281 	 * per proc (excluding public exinfo), and exi_count design
5282 	 * is sufficient to protect concurrent execution of NFS2/3
5283 	 * ops along with unexport.  This lock will be removed as
5284 	 * part of the NFSv4 phase 2 namespace redesign work.
5285 	 */
5286 	rw_enter(&exported_lock, RW_READER);
5287 
5288 	/*
5289 	 * If this is the first compound we've seen, we need to start all
5290 	 * new instances' grace periods.
5291 	 */
5292 	if (rfs4_seen_first_compound == 0) {
5293 		rfs4_grace_start_new();
5294 		/*
5295 		 * This must be set after rfs4_grace_start_new(), otherwise
5296 		 * another thread could proceed past here before the former
5297 		 * is finished.
5298 		 */
5299 		rfs4_seen_first_compound = 1;
5300 	}
5301 
5302 	for (i = 0; i < args->array_len && cs.cont; i++) {
5303 		nfs_argop4 *argop;
5304 		nfs_resop4 *resop;
5305 		uint_t op;
5306 
5307 		argop = &args->array[i];
5308 		resop = &resp->array[i];
5309 		resop->resop = argop->argop;
5310 		op = (uint_t)resop->resop;
5311 
5312 		if (op < rfsv4disp_cnt) {
5313 			/*
5314 			 * Count the individual ops here; NULL and COMPOUND
5315 			 * are counted in common_dispatch()
5316 			 */
5317 			rfsproccnt_v4_ptr[op].value.ui64++;
5318 
5319 			NFS4_DEBUG(rfs4_debug > 1,
5320 				(CE_NOTE, "Executing %s", rfs4_op_string[op]));
5321 			(*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs);
5322 			NFS4_DEBUG(rfs4_debug > 1,
5323 				(CE_NOTE, "%s returned %d",
5324 				rfs4_op_string[op], *cs.statusp));
5325 			if (*cs.statusp != NFS4_OK)
5326 				cs.cont = FALSE;
5327 		} else {
5328 			/*
5329 			 * This is effectively dead code since XDR code
5330 			 * will have already returned BADXDR if op doesn't
5331 			 * decode to legal value.  This only done for a
5332 			 * day when XDR code doesn't verify v4 opcodes.
5333 			 */
5334 			op = OP_ILLEGAL;
5335 			rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++;
5336 
5337 			rfs4_op_illegal(argop, resop, req, &cs);
5338 			cs.cont = FALSE;
5339 		}
5340 
5341 		/*
5342 		 * If not at last op, and if we are to stop, then
5343 		 * compact the results array.
5344 		 */
5345 		if ((i + 1) < args->array_len && !cs.cont) {
5346 			nfs_resop4 *new_res = kmem_alloc(
5347 				(i+1) * sizeof (nfs_resop4), KM_SLEEP);
5348 			bcopy(resp->array,
5349 				new_res, (i+1) * sizeof (nfs_resop4));
5350 			kmem_free(resp->array,
5351 				args->array_len * sizeof (nfs_resop4));
5352 
5353 			resp->array_len =  i + 1;
5354 			resp->array = new_res;
5355 		}
5356 	}
5357 
5358 	rw_exit(&exported_lock);
5359 
5360 	if (cs.vp)
5361 		VN_RELE(cs.vp);
5362 	if (cs.saved_vp)
5363 		VN_RELE(cs.saved_vp);
5364 	if (cs.saved_fh.nfs_fh4_val)
5365 		kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE);
5366 
5367 	if (cs.basecr)
5368 		crfree(cs.basecr);
5369 	if (cs.cr)
5370 		crfree(cs.cr);
5371 	/*
5372 	 * done with this compound request, free the label
5373 	 */
5374 
5375 	if (req->rq_label != NULL) {
5376 		kmem_free(req->rq_label, sizeof (bslabel_t));
5377 		req->rq_label = NULL;
5378 	}
5379 }
5380 
5381 /*
5382  * XXX because of what appears to be duplicate calls to rfs4_compound_free
5383  * XXX zero out the tag and array values. Need to investigate why the
5384  * XXX calls occur, but at least prevent the panic for now.
5385  */
5386 void
5387 rfs4_compound_free(COMPOUND4res *resp)
5388 {
5389 	uint_t i;
5390 
5391 	if (resp->tag.utf8string_val) {
5392 		UTF8STRING_FREE(resp->tag)
5393 	}
5394 
5395 	for (i = 0; i < resp->array_len; i++) {
5396 		nfs_resop4 *resop;
5397 		uint_t op;
5398 
5399 		resop = &resp->array[i];
5400 		op = (uint_t)resop->resop;
5401 		if (op < rfsv4disp_cnt) {
5402 			(*rfsv4disptab[op].dis_resfree)(resop);
5403 		}
5404 	}
5405 	if (resp->array != NULL) {
5406 		kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4));
5407 	}
5408 }
5409 
5410 /*
5411  * Process the value of the compound request rpc flags, as a bit-AND
5412  * of the individual per-op flags (idempotent, allowork, publicfh_ok)
5413  */
5414 void
5415 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp)
5416 {
5417 	int i;
5418 	int flag = RPC_ALL;
5419 
5420 	for (i = 0; flag && i < args->array_len; i++) {
5421 		uint_t op;
5422 
5423 		op = (uint_t)args->array[i].argop;
5424 
5425 		if (op < rfsv4disp_cnt)
5426 			flag &= rfsv4disptab[op].dis_flags;
5427 		else
5428 			flag = 0;
5429 	}
5430 	*flagp = flag;
5431 }
5432 
5433 nfsstat4
5434 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp)
5435 {
5436 	nfsstat4 e;
5437 
5438 	rfs4_dbe_lock(cp->dbe);
5439 
5440 	if (cp->sysidt != LM_NOSYSID) {
5441 		*sp = cp->sysidt;
5442 		e = NFS4_OK;
5443 
5444 	} else if ((cp->sysidt = lm_alloc_sysidt()) != LM_NOSYSID) {
5445 		*sp = cp->sysidt;
5446 		e = NFS4_OK;
5447 
5448 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
5449 			"rfs4_client_sysid: allocated 0x%x\n", *sp));
5450 	} else
5451 		e = NFS4ERR_DELAY;
5452 
5453 	rfs4_dbe_unlock(cp->dbe);
5454 	return (e);
5455 }
5456 
5457 #if defined(DEBUG) && ! defined(lint)
5458 static void lock_print(char *str, int operation, struct flock64 *flk)
5459 {
5460 	char *op, *type;
5461 
5462 	switch (operation) {
5463 	case F_GETLK: op = "F_GETLK";
5464 		break;
5465 	case F_SETLK: op = "F_SETLK";
5466 		break;
5467 	default: op = "F_UNKNOWN";
5468 		break;
5469 	}
5470 	switch (flk->l_type) {
5471 	case F_UNLCK: type = "F_UNLCK";
5472 		break;
5473 	case F_RDLCK: type = "F_RDLCK";
5474 		break;
5475 	case F_WRLCK: type = "F_WRLCK";
5476 		break;
5477 	default: type = "F_UNKNOWN";
5478 		break;
5479 	}
5480 
5481 	ASSERT(flk->l_whence == 0);
5482 	cmn_err(CE_NOTE, "%s:  %s, type = %s, off = %llx len = %llx pid = %d",
5483 		str, op, type,
5484 		(longlong_t)flk->l_start,
5485 		flk->l_len ? (longlong_t)flk->l_len : ~0LL,
5486 		flk->l_pid);
5487 }
5488 
5489 #define	LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f)
5490 #else
5491 #define	LOCK_PRINT(d, s, t, f)
5492 #endif
5493 
5494 /*ARGSUSED*/
5495 static bool_t
5496 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs)
5497 {
5498 	return (TRUE);
5499 }
5500 
5501 /*
5502  * Look up the pathname using the vp in cs as the directory vnode.
5503  * cs->vp will be the vnode for the file on success
5504  */
5505 
5506 static nfsstat4
5507 rfs4_lookup(component4 *component, struct svc_req *req,
5508 	    struct compound_state *cs)
5509 {
5510 	char *nm;
5511 	uint32_t len;
5512 	nfsstat4 status;
5513 
5514 	if (cs->vp == NULL) {
5515 		return (NFS4ERR_NOFILEHANDLE);
5516 	}
5517 	if (cs->vp->v_type != VDIR) {
5518 		return (NFS4ERR_NOTDIR);
5519 	}
5520 
5521 	if (!utf8_dir_verify(component))
5522 		return (NFS4ERR_INVAL);
5523 
5524 	nm = utf8_to_fn(component, &len, NULL);
5525 	if (nm == NULL) {
5526 		return (NFS4ERR_INVAL);
5527 	}
5528 
5529 	if (len > MAXNAMELEN) {
5530 		kmem_free(nm, len);
5531 		return (NFS4ERR_NAMETOOLONG);
5532 	}
5533 
5534 	status = do_rfs4_op_lookup(nm, len, req, cs);
5535 
5536 	kmem_free(nm, len);
5537 
5538 	return (status);
5539 }
5540 
5541 static nfsstat4
5542 rfs4_lookupfile(component4 *component, struct svc_req *req,
5543 		struct compound_state *cs, uint32_t access,
5544 		change_info4 *cinfo)
5545 {
5546 	nfsstat4 status;
5547 	vnode_t *dvp = cs->vp;
5548 	vattr_t bva, ava, fva;
5549 	int error;
5550 
5551 	/* Get "before" change value */
5552 	bva.va_mask = AT_CTIME|AT_SEQ;
5553 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr);
5554 	if (error)
5555 		return (puterrno4(error));
5556 
5557 	/* rfs4_lookup may VN_RELE directory */
5558 	VN_HOLD(dvp);
5559 
5560 	status = rfs4_lookup(component, req, cs);
5561 	if (status != NFS4_OK) {
5562 		VN_RELE(dvp);
5563 		return (status);
5564 	}
5565 
5566 	/*
5567 	 * Get "after" change value, if it fails, simply return the
5568 	 * before value.
5569 	 */
5570 	ava.va_mask = AT_CTIME|AT_SEQ;
5571 	if (VOP_GETATTR(dvp, &ava, 0, cs->cr)) {
5572 		ava.va_ctime = bva.va_ctime;
5573 		ava.va_seq = 0;
5574 	}
5575 	VN_RELE(dvp);
5576 
5577 	/*
5578 	 * Validate the file is a file
5579 	 */
5580 	fva.va_mask = AT_TYPE|AT_MODE;
5581 	error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr);
5582 	if (error)
5583 		return (puterrno4(error));
5584 
5585 	if (fva.va_type != VREG) {
5586 		if (fva.va_type == VDIR)
5587 			return (NFS4ERR_ISDIR);
5588 		if (fva.va_type == VLNK)
5589 			return (NFS4ERR_SYMLINK);
5590 		return (NFS4ERR_INVAL);
5591 	}
5592 
5593 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime);
5594 	NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
5595 
5596 	/*
5597 	 * It is undefined if VOP_LOOKUP will change va_seq, so
5598 	 * cinfo.atomic = TRUE only if we have
5599 	 * non-zero va_seq's, and they have not changed.
5600 	 */
5601 	if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq)
5602 		cinfo->atomic = TRUE;
5603 	else
5604 		cinfo->atomic = FALSE;
5605 
5606 	/* Check for mandatory locking */
5607 	cs->mandlock = MANDLOCK(cs->vp, fva.va_mode);
5608 	return (check_open_access(access, cs, req));
5609 }
5610 
5611 static nfsstat4
5612 create_vnode(vnode_t *dvp, char *nm,  vattr_t *vap, createmode4 mode,
5613 	    timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created)
5614 {
5615 	int error;
5616 	nfsstat4 status = NFS4_OK;
5617 	vattr_t va;
5618 
5619 tryagain:
5620 
5621 	/*
5622 	 * The file open mode used is VWRITE.  If the client needs
5623 	 * some other semantic, then it should do the access checking
5624 	 * itself.  It would have been nice to have the file open mode
5625 	 * passed as part of the arguments.
5626 	 */
5627 
5628 	*created = TRUE;
5629 	error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0);
5630 
5631 	if (error) {
5632 		*created = FALSE;
5633 
5634 		/*
5635 		 * If we got something other than file already exists
5636 		 * then just return this error.  Otherwise, we got
5637 		 * EEXIST.  If we were doing a GUARDED create, then
5638 		 * just return this error.  Otherwise, we need to
5639 		 * make sure that this wasn't a duplicate of an
5640 		 * exclusive create request.
5641 		 *
5642 		 * The assumption is made that a non-exclusive create
5643 		 * request will never return EEXIST.
5644 		 */
5645 
5646 		if (error != EEXIST || mode == GUARDED4) {
5647 			status = puterrno4(error);
5648 			return (status);
5649 		}
5650 		error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr);
5651 
5652 		if (error) {
5653 			/*
5654 			 * We couldn't find the file that we thought that
5655 			 * we just created.  So, we'll just try creating
5656 			 * it again.
5657 			 */
5658 			if (error == ENOENT)
5659 				goto tryagain;
5660 
5661 			status = puterrno4(error);
5662 			return (status);
5663 		}
5664 
5665 		if (mode == UNCHECKED4) {
5666 			/* existing object must be regular file */
5667 			if ((*vpp)->v_type != VREG) {
5668 				if ((*vpp)->v_type == VDIR)
5669 					status = NFS4ERR_ISDIR;
5670 				else if ((*vpp)->v_type == VLNK)
5671 					status = NFS4ERR_SYMLINK;
5672 				else
5673 					status = NFS4ERR_INVAL;
5674 				VN_RELE(*vpp);
5675 				return (status);
5676 			}
5677 
5678 			return (NFS4_OK);
5679 		}
5680 
5681 		/* Check for duplicate request */
5682 		ASSERT(mtime != 0);
5683 		va.va_mask = AT_MTIME;
5684 		error = VOP_GETATTR(*vpp, &va, 0, cr);
5685 		if (!error) {
5686 			/* We found the file */
5687 			if (va.va_mtime.tv_sec != mtime->tv_sec ||
5688 			    va.va_mtime.tv_nsec != mtime->tv_nsec) {
5689 				/* but its not our creation */
5690 				VN_RELE(*vpp);
5691 				return (NFS4ERR_EXIST);
5692 			}
5693 			*created = TRUE; /* retrans of create == created */
5694 			return (NFS4_OK);
5695 		}
5696 		VN_RELE(*vpp);
5697 		return (NFS4ERR_EXIST);
5698 	}
5699 
5700 	return (NFS4_OK);
5701 }
5702 
5703 static nfsstat4
5704 check_open_access(uint32_t access,
5705 		struct compound_state *cs, struct svc_req *req)
5706 {
5707 	int error;
5708 	vnode_t *vp;
5709 	bool_t readonly;
5710 	cred_t *cr = cs->cr;
5711 
5712 	/* For now we don't allow mandatory locking as per V2/V3 */
5713 	if (cs->access == CS_ACCESS_DENIED || cs->mandlock) {
5714 		return (NFS4ERR_ACCESS);
5715 	}
5716 
5717 	vp = cs->vp;
5718 	ASSERT(cr != NULL && vp->v_type == VREG);
5719 
5720 	/*
5721 	 * If the file system is exported read only and we are trying
5722 	 * to open for write, then return NFS4ERR_ROFS
5723 	 */
5724 
5725 	readonly = rdonly4(cs->exi, cs->vp, req);
5726 
5727 	if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly)
5728 		return (NFS4ERR_ROFS);
5729 
5730 	if (access & OPEN4_SHARE_ACCESS_READ) {
5731 		if ((VOP_ACCESS(vp, VREAD, 0, cr) != 0) &&
5732 		    (VOP_ACCESS(vp, VEXEC, 0, cr) != 0)) {
5733 			return (NFS4ERR_ACCESS);
5734 		}
5735 	}
5736 
5737 	if (access & OPEN4_SHARE_ACCESS_WRITE) {
5738 		error = VOP_ACCESS(vp, VWRITE, 0, cr);
5739 		if (error)
5740 			return (NFS4ERR_ACCESS);
5741 	}
5742 
5743 	return (NFS4_OK);
5744 }
5745 
5746 static nfsstat4
5747 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs,
5748 		change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid)
5749 {
5750 	struct nfs4_svgetit_arg sarg;
5751 	struct nfs4_ntov_table ntov;
5752 
5753 	bool_t ntov_table_init = FALSE;
5754 	struct statvfs64 sb;
5755 	nfsstat4 status;
5756 	vnode_t *vp;
5757 	vattr_t bva, ava, iva, cva, *vap;
5758 	vnode_t *dvp;
5759 	timespec32_t *mtime;
5760 	char *nm = NULL;
5761 	uint_t buflen;
5762 	bool_t created;
5763 	bool_t setsize = FALSE;
5764 	len_t reqsize;
5765 	int error;
5766 	bool_t trunc;
5767 	caller_context_t ct;
5768 	component4 *component;
5769 	bslabel_t *clabel;
5770 
5771 	sarg.sbp = &sb;
5772 
5773 	dvp = cs->vp;
5774 
5775 	/* Check if the file system is read only */
5776 	if (rdonly4(cs->exi, dvp, req))
5777 		return (NFS4ERR_ROFS);
5778 
5779 	/* check the label of including directory */
5780 	if (is_system_labeled()) {
5781 		ASSERT(req->rq_label != NULL);
5782 		clabel = req->rq_label;
5783 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
5784 		    "got client label from request(1)",
5785 		    struct svc_req *, req);
5786 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
5787 			if (!do_rfs4_label_check(clabel, dvp, EQUALITY_CHECK)) {
5788 				return (NFS4ERR_ACCESS);
5789 			}
5790 		}
5791 	}
5792 
5793 	/*
5794 	 * Get the last component of path name in nm. cs will reference
5795 	 * the including directory on success.
5796 	 */
5797 	component = &args->open_claim4_u.file;
5798 	if (!utf8_dir_verify(component))
5799 		return (NFS4ERR_INVAL);
5800 
5801 	nm = utf8_to_fn(component, &buflen, NULL);
5802 
5803 	if (nm == NULL)
5804 		return (NFS4ERR_RESOURCE);
5805 
5806 	if (buflen > MAXNAMELEN) {
5807 		kmem_free(nm, buflen);
5808 		return (NFS4ERR_NAMETOOLONG);
5809 	}
5810 
5811 	bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ;
5812 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr);
5813 	if (error) {
5814 		kmem_free(nm, buflen);
5815 		return (puterrno4(error));
5816 	}
5817 
5818 	if (bva.va_type != VDIR) {
5819 		kmem_free(nm, buflen);
5820 		return (NFS4ERR_NOTDIR);
5821 	}
5822 
5823 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime)
5824 
5825 	switch (args->mode) {
5826 	case GUARDED4:
5827 		/*FALLTHROUGH*/
5828 	case UNCHECKED4:
5829 		nfs4_ntov_table_init(&ntov);
5830 		ntov_table_init = TRUE;
5831 
5832 		*attrset = 0;
5833 		status = do_rfs4_set_attrs(attrset,
5834 					&args->createhow4_u.createattrs,
5835 					cs, &sarg, &ntov, NFS4ATTR_SETIT);
5836 
5837 		if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) &&
5838 		    sarg.vap->va_type != VREG) {
5839 			if (sarg.vap->va_type == VDIR)
5840 				status = NFS4ERR_ISDIR;
5841 			else if (sarg.vap->va_type == VLNK)
5842 				status = NFS4ERR_SYMLINK;
5843 			else
5844 				status = NFS4ERR_INVAL;
5845 		}
5846 
5847 		if (status != NFS4_OK) {
5848 			kmem_free(nm, buflen);
5849 			nfs4_ntov_table_free(&ntov, &sarg);
5850 			*attrset = 0;
5851 			return (status);
5852 		}
5853 
5854 		vap = sarg.vap;
5855 		vap->va_type = VREG;
5856 		vap->va_mask |= AT_TYPE;
5857 
5858 		if ((vap->va_mask & AT_MODE) == 0) {
5859 			vap->va_mask |= AT_MODE;
5860 			vap->va_mode = (mode_t)0600;
5861 		}
5862 
5863 		if (vap->va_mask & AT_SIZE) {
5864 
5865 			/* Disallow create with a non-zero size */
5866 
5867 			if ((reqsize = sarg.vap->va_size) != 0) {
5868 				kmem_free(nm, buflen);
5869 				nfs4_ntov_table_free(&ntov, &sarg);
5870 				*attrset = 0;
5871 				return (NFS4ERR_INVAL);
5872 			}
5873 			setsize = TRUE;
5874 		}
5875 		break;
5876 
5877 	case EXCLUSIVE4:
5878 		/* prohibit EXCL create of named attributes */
5879 		if (dvp->v_flag & V_XATTRDIR) {
5880 			kmem_free(nm, buflen);
5881 			*attrset = 0;
5882 			return (NFS4ERR_INVAL);
5883 		}
5884 
5885 		cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE;
5886 		cva.va_type = VREG;
5887 		/*
5888 		 * Ensure no time overflows. Assumes underlying
5889 		 * filesystem supports at least 32 bits.
5890 		 * Truncate nsec to usec resolution to allow valid
5891 		 * compares even if the underlying filesystem truncates.
5892 		 */
5893 		mtime = (timespec32_t *)&args->createhow4_u.createverf;
5894 		cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX;
5895 		cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000;
5896 		cva.va_mode = (mode_t)0;
5897 		vap = &cva;
5898 		break;
5899 	}
5900 
5901 	status = create_vnode(dvp, nm, vap, args->mode, mtime,
5902 						cs->cr, &vp, &created);
5903 	kmem_free(nm, buflen);
5904 
5905 	if (status != NFS4_OK) {
5906 		if (ntov_table_init)
5907 			nfs4_ntov_table_free(&ntov, &sarg);
5908 		*attrset = 0;
5909 		return (status);
5910 	}
5911 
5912 	trunc = (setsize && !created);
5913 
5914 	if (args->mode != EXCLUSIVE4) {
5915 		bitmap4 createmask = args->createhow4_u.createattrs.attrmask;
5916 
5917 		/*
5918 		 * True verification that object was created with correct
5919 		 * attrs is impossible.  The attrs could have been changed
5920 		 * immediately after object creation.  If attributes did
5921 		 * not verify, the only recourse for the server is to
5922 		 * destroy the object.  Maybe if some attrs (like gid)
5923 		 * are set incorrectly, the object should be destroyed;
5924 		 * however, seems bad as a default policy.  Do we really
5925 		 * want to destroy an object over one of the times not
5926 		 * verifying correctly?  For these reasons, the server
5927 		 * currently sets bits in attrset for createattrs
5928 		 * that were set; however, no verification is done.
5929 		 *
5930 		 * vmask_to_nmask accounts for vattr bits set on create
5931 		 *	[do_rfs4_set_attrs() only sets resp bits for
5932 		 *	 non-vattr/vfs bits.]
5933 		 * Mask off any bits we set by default so as not to return
5934 		 * more attrset bits than were requested in createattrs
5935 		 */
5936 		if (created) {
5937 			nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset);
5938 			*attrset &= createmask;
5939 		} else {
5940 			/*
5941 			 * We did not create the vnode (we tried but it
5942 			 * already existed).  In this case, the only createattr
5943 			 * that the spec allows the server to set is size,
5944 			 * and even then, it can only be set if it is 0.
5945 			 */
5946 			*attrset = 0;
5947 			if (trunc)
5948 				*attrset = FATTR4_SIZE_MASK;
5949 		}
5950 	}
5951 	if (ntov_table_init)
5952 		nfs4_ntov_table_free(&ntov, &sarg);
5953 
5954 	/*
5955 	 * Get the initial "after" sequence number, if it fails,
5956 	 * set to zero, time to before.
5957 	 */
5958 	iva.va_mask = AT_CTIME|AT_SEQ;
5959 	if (VOP_GETATTR(dvp, &iva, 0, cs->cr)) {
5960 		iva.va_seq = 0;
5961 		iva.va_ctime = bva.va_ctime;
5962 	}
5963 
5964 	/*
5965 	 * create_vnode attempts to create the file exclusive,
5966 	 * if it already exists the VOP_CREATE will fail and
5967 	 * may not increase va_seq. It is atomic if
5968 	 * we haven't changed the directory, but if it has changed
5969 	 * we don't know what changed it.
5970 	 */
5971 	if (!created) {
5972 		if (bva.va_seq && iva.va_seq &&
5973 			bva.va_seq == iva.va_seq)
5974 			cinfo->atomic = TRUE;
5975 		else
5976 			cinfo->atomic = FALSE;
5977 		NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime);
5978 	} else {
5979 		/*
5980 		 * The entry was created, we need to sync the
5981 		 * directory metadata.
5982 		 */
5983 		(void) VOP_FSYNC(dvp, 0, cs->cr);
5984 
5985 		/*
5986 		 * Get "after" change value, if it fails, simply return the
5987 		 * before value.
5988 		 */
5989 		ava.va_mask = AT_CTIME|AT_SEQ;
5990 		if (VOP_GETATTR(dvp, &ava, 0, cs->cr)) {
5991 			ava.va_ctime = bva.va_ctime;
5992 			ava.va_seq = 0;
5993 		}
5994 
5995 		NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
5996 
5997 		/*
5998 		 * The cinfo->atomic = TRUE only if we have
5999 		 * non-zero va_seq's, and it has incremented by exactly one
6000 		 * during the create_vnode and it didn't
6001 		 * change during the VOP_FSYNC.
6002 		 */
6003 		if (bva.va_seq && iva.va_seq && ava.va_seq &&
6004 				iva.va_seq == (bva.va_seq + 1) &&
6005 				iva.va_seq == ava.va_seq)
6006 			cinfo->atomic = TRUE;
6007 		else
6008 			cinfo->atomic = FALSE;
6009 	}
6010 
6011 	/* Check for mandatory locking and that the size gets set. */
6012 	cva.va_mask = AT_MODE;
6013 	if (setsize)
6014 		cva.va_mask |= AT_SIZE;
6015 
6016 	/* Assume the worst */
6017 	cs->mandlock = TRUE;
6018 
6019 	if (VOP_GETATTR(vp, &cva, 0, cs->cr) == 0) {
6020 		cs->mandlock = MANDLOCK(cs->vp, cva.va_mode);
6021 
6022 		/*
6023 		 * Truncate the file if necessary; this would be
6024 		 * the case for create over an existing file.
6025 		 */
6026 
6027 		if (trunc) {
6028 			int in_crit = 0;
6029 			rfs4_file_t *fp;
6030 			bool_t create = FALSE;
6031 
6032 			/*
6033 			 * We are writing over an existing file.
6034 			 * Check to see if we need to recall a delegation.
6035 			 */
6036 			rfs4_hold_deleg_policy();
6037 			if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) {
6038 				if (rfs4_check_delegated_byfp(FWRITE, fp,
6039 					(reqsize == 0), FALSE, FALSE,
6040 							&clientid)) {
6041 
6042 					rfs4_file_rele(fp);
6043 					rfs4_rele_deleg_policy();
6044 					VN_RELE(vp);
6045 					*attrset = 0;
6046 					return (NFS4ERR_DELAY);
6047 				}
6048 				rfs4_file_rele(fp);
6049 			}
6050 			rfs4_rele_deleg_policy();
6051 
6052 			if (nbl_need_check(vp)) {
6053 				in_crit = 1;
6054 
6055 				ASSERT(reqsize == 0);
6056 
6057 				nbl_start_crit(vp, RW_READER);
6058 				if (nbl_conflict(vp, NBL_WRITE, 0,
6059 						cva.va_size, 0)) {
6060 					in_crit = 0;
6061 					nbl_end_crit(vp);
6062 					VN_RELE(vp);
6063 					*attrset = 0;
6064 					return (NFS4ERR_ACCESS);
6065 				}
6066 			}
6067 			ct.cc_sysid = 0;
6068 			ct.cc_pid = 0;
6069 			ct.cc_caller_id = nfs4_srv_caller_id;
6070 
6071 			cva.va_mask = AT_SIZE;
6072 			cva.va_size = reqsize;
6073 			(void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct);
6074 			if (in_crit)
6075 				nbl_end_crit(vp);
6076 		}
6077 	}
6078 
6079 	error = makefh4(&cs->fh, vp, cs->exi);
6080 
6081 	/*
6082 	 * Force modified data and metadata out to stable storage.
6083 	 */
6084 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr);
6085 
6086 	if (error) {
6087 		VN_RELE(vp);
6088 		*attrset = 0;
6089 		return (puterrno4(error));
6090 	}
6091 
6092 	/* if parent dir is attrdir, set namedattr fh flag */
6093 	if (dvp->v_flag & V_XATTRDIR)
6094 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
6095 
6096 	if (cs->vp)
6097 		VN_RELE(cs->vp);
6098 
6099 	cs->vp = vp;
6100 
6101 	/*
6102 	 * if we did not create the file, we will need to check
6103 	 * the access bits on the file
6104 	 */
6105 
6106 	if (!created) {
6107 		if (setsize)
6108 			args->share_access |= OPEN4_SHARE_ACCESS_WRITE;
6109 		status = check_open_access(args->share_access, cs, req);
6110 		if (status != NFS4_OK)
6111 			*attrset = 0;
6112 	}
6113 	return (status);
6114 }
6115 
6116 /*ARGSUSED*/
6117 static void
6118 rfs4_do_open(struct compound_state *cs, struct svc_req *req,
6119 		rfs4_openowner_t *oo, delegreq_t deleg,
6120 		uint32_t access, uint32_t deny,
6121 		OPEN4res *resp)
6122 {
6123 	/* XXX Currently not using req  */
6124 	rfs4_state_t *state;
6125 	rfs4_file_t *file;
6126 	bool_t screate = TRUE;
6127 	bool_t fcreate = TRUE;
6128 	uint32_t amodes;
6129 	uint32_t dmodes;
6130 	rfs4_deleg_state_t *dsp;
6131 	struct shrlock shr;
6132 	struct shr_locowner shr_loco;
6133 	sysid_t sysid;
6134 	nfsstat4 status;
6135 	int fflags = 0;
6136 	int recall = 0;
6137 	int err;
6138 
6139 	/* get the file struct and hold a lock on it during initial open */
6140 	file = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate);
6141 	if (file == NULL) {
6142 		NFS4_DEBUG(rfs4_debug,
6143 			(CE_NOTE, "rfs4_do_open: can't find file"));
6144 		resp->status = NFS4ERR_SERVERFAULT;
6145 		return;
6146 	}
6147 
6148 	state = rfs4_findstate_by_owner_file(oo, file, &screate);
6149 	if (state == NULL) {
6150 		NFS4_DEBUG(rfs4_debug,
6151 			(CE_NOTE, "rfs4_do_open: can't find state"));
6152 		resp->status = NFS4ERR_RESOURCE;
6153 		/* No need to keep any reference */
6154 		rfs4_file_rele_withunlock(file);
6155 		return;
6156 	}
6157 
6158 	/*
6159 	 * Check for conflicts in deny and access before checking for
6160 	 * conflicts in delegation.  We don't want to recall a
6161 	 * delegation based on an open that will eventually fail based
6162 	 * on shares modes.
6163 	 */
6164 
6165 	shr.s_access = (short)access;
6166 	shr.s_deny = (short)deny;
6167 	shr.s_pid = rfs4_dbe_getid(oo->dbe);
6168 
6169 	if ((status = rfs4_client_sysid(oo->client, &sysid)) != NFS4_OK) {
6170 		resp->status = status;
6171 		rfs4_file_rele(file);
6172 		/* Not a fully formed open; "close" it */
6173 		if (screate == TRUE)
6174 			rfs4_state_close(state, FALSE, FALSE, cs->cr);
6175 		rfs4_state_rele(state);
6176 		return;
6177 	}
6178 	shr.s_sysid = sysid;
6179 	shr_loco.sl_pid = shr.s_pid;
6180 	shr_loco.sl_id = shr.s_sysid;
6181 	shr.s_owner = (caddr_t)&shr_loco;
6182 	shr.s_own_len = sizeof (shr_loco);
6183 
6184 	fflags = 0;
6185 	if (access & OPEN4_SHARE_ACCESS_READ)
6186 		fflags |= FREAD;
6187 	if (access & OPEN4_SHARE_ACCESS_WRITE)
6188 		fflags |= FWRITE;
6189 
6190 	if ((err = vop_shrlock(cs->vp, F_SHARE, &shr, fflags)) != 0) {
6191 
6192 		resp->status = err == EAGAIN ?
6193 			NFS4ERR_SHARE_DENIED : puterrno4(err);
6194 
6195 		rfs4_file_rele(file);
6196 		/* Not a fully formed open; "close" it */
6197 		if (screate == TRUE)
6198 			rfs4_state_close(state, FALSE, FALSE, cs->cr);
6199 		rfs4_state_rele(state);
6200 		return;
6201 	}
6202 
6203 	rfs4_dbe_lock(state->dbe);
6204 	rfs4_dbe_lock(file->dbe);
6205 
6206 	/*
6207 	 * Calculate the new deny and access mode that this open is adding to
6208 	 * the file for this open owner;
6209 	 */
6210 	dmodes = (deny & ~state->share_deny);
6211 	amodes = (access & ~state->share_access);
6212 
6213 	/*
6214 	 * Check to see if this file is delegated and if so, if a
6215 	 * recall needs to be done.
6216 	 */
6217 	if (rfs4_check_recall(state, access)) {
6218 		rfs4_dbe_unlock(file->dbe);
6219 		rfs4_dbe_unlock(state->dbe);
6220 		rfs4_recall_deleg(file, FALSE, state->owner->client);
6221 		delay(NFS4_DELEGATION_CONFLICT_DELAY);
6222 		rfs4_dbe_lock(state->dbe);
6223 		rfs4_dbe_lock(file->dbe);
6224 		/* Let's see if the delegation was returned */
6225 		if (rfs4_check_recall(state, access)) {
6226 			rfs4_dbe_unlock(file->dbe);
6227 			rfs4_dbe_unlock(state->dbe);
6228 			rfs4_file_rele(file);
6229 			rfs4_update_lease(state->owner->client);
6230 			/* recalculate flags to match what was added */
6231 			fflags = 0;
6232 			if (amodes & OPEN4_SHARE_ACCESS_READ)
6233 				fflags |= FREAD;
6234 			if (amodes & OPEN4_SHARE_ACCESS_WRITE)
6235 				fflags |= FWRITE;
6236 			(void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags);
6237 			/* Not a fully formed open; "close" it */
6238 			if (screate == TRUE)
6239 				rfs4_state_close(state, FALSE, FALSE, cs->cr);
6240 			rfs4_state_rele(state);
6241 			resp->status = NFS4ERR_DELAY;
6242 			return;
6243 		}
6244 	}
6245 
6246 	if (dmodes & OPEN4_SHARE_DENY_READ)
6247 		file->deny_read++;
6248 	if (dmodes & OPEN4_SHARE_DENY_WRITE)
6249 		file->deny_write++;
6250 	file->share_deny |= deny;
6251 	state->share_deny |= deny;
6252 
6253 	if (amodes & OPEN4_SHARE_ACCESS_READ)
6254 		file->access_read++;
6255 	if (amodes & OPEN4_SHARE_ACCESS_WRITE)
6256 		file->access_write++;
6257 	file->share_access |= access;
6258 	state->share_access |= access;
6259 
6260 	/*
6261 	 * Check for delegation here. if the deleg argument is not
6262 	 * DELEG_ANY, then this is a reclaim from a client and
6263 	 * we must honor the delegation requested. If necessary we can
6264 	 * set the recall flag.
6265 	 */
6266 
6267 	dsp = rfs4_grant_delegation(deleg, state, &recall);
6268 
6269 	cs->deleg = (file->dinfo->dtype == OPEN_DELEGATE_WRITE);
6270 
6271 	next_stateid(&state->stateid);
6272 
6273 	resp->stateid = state->stateid.stateid;
6274 
6275 	rfs4_dbe_unlock(file->dbe);
6276 	rfs4_dbe_unlock(state->dbe);
6277 
6278 	if (dsp) {
6279 		rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall);
6280 		rfs4_deleg_state_rele(dsp);
6281 	}
6282 
6283 	rfs4_file_rele(file);
6284 	rfs4_state_rele(state);
6285 
6286 	resp->status = NFS4_OK;
6287 }
6288 
6289 /*ARGSUSED*/
6290 static void
6291 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req,
6292 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6293 {
6294 	change_info4 *cinfo = &resp->cinfo;
6295 	bitmap4 *attrset = &resp->attrset;
6296 
6297 	if (args->opentype == OPEN4_NOCREATE)
6298 		resp->status = rfs4_lookupfile(&args->open_claim4_u.file,
6299 					req, cs, args->share_access, cinfo);
6300 	else {
6301 		/* inhibit delegation grants during exclusive create */
6302 
6303 		if (args->mode == EXCLUSIVE4)
6304 			rfs4_disable_delegation();
6305 
6306 		resp->status = rfs4_createfile(args, req, cs, cinfo, attrset,
6307 					oo->client->clientid);
6308 	}
6309 
6310 	if (resp->status == NFS4_OK) {
6311 
6312 		/* cs->vp cs->fh now reference the desired file */
6313 
6314 		rfs4_do_open(cs, req, oo, DELEG_ANY, args->share_access,
6315 						args->share_deny, resp);
6316 
6317 		/*
6318 		 * If rfs4_createfile set attrset, we must
6319 		 * clear this attrset before the response is copied.
6320 		 */
6321 		if (resp->status != NFS4_OK && resp->attrset) {
6322 			resp->attrset = 0;
6323 		}
6324 	}
6325 	else
6326 		*cs->statusp = resp->status;
6327 
6328 	if (args->mode == EXCLUSIVE4)
6329 		rfs4_enable_delegation();
6330 }
6331 
6332 /*ARGSUSED*/
6333 static void
6334 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req,
6335 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6336 {
6337 	change_info4 *cinfo = &resp->cinfo;
6338 	vattr_t va;
6339 	vtype_t v_type = cs->vp->v_type;
6340 	int error = 0;
6341 
6342 	/* Verify that we have a regular file */
6343 	if (v_type != VREG) {
6344 		if (v_type == VDIR)
6345 			resp->status = NFS4ERR_ISDIR;
6346 		else if (v_type == VLNK)
6347 			resp->status = NFS4ERR_SYMLINK;
6348 		else
6349 			resp->status = NFS4ERR_INVAL;
6350 		return;
6351 	}
6352 
6353 	va.va_mask = AT_MODE|AT_UID;
6354 	error = VOP_GETATTR(cs->vp, &va, 0, cs->cr);
6355 	if (error) {
6356 		resp->status = puterrno4(error);
6357 		return;
6358 	}
6359 
6360 	cs->mandlock = MANDLOCK(cs->vp, va.va_mode);
6361 
6362 	/*
6363 	 * Check if we have access to the file, Note the the file
6364 	 * could have originally been open UNCHECKED or GUARDED
6365 	 * with mode bits that will now fail, but there is nothing
6366 	 * we can really do about that except in the case that the
6367 	 * owner of the file is the one requesting the open.
6368 	 */
6369 	if (crgetuid(cs->cr) != va.va_uid) {
6370 		resp->status = check_open_access(args->share_access, cs, req);
6371 		if (resp->status != NFS4_OK) {
6372 			return;
6373 		}
6374 	}
6375 
6376 	/*
6377 	 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero
6378 	 */
6379 	cinfo->before = 0;
6380 	cinfo->after = 0;
6381 	cinfo->atomic = FALSE;
6382 
6383 	rfs4_do_open(cs, req, oo,
6384 		NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type),
6385 		args->share_access, args->share_deny, resp);
6386 }
6387 
6388 static void
6389 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req,
6390 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6391 {
6392 	int error;
6393 	nfsstat4 status;
6394 	stateid4 stateid =
6395 			args->open_claim4_u.delegate_cur_info.delegate_stateid;
6396 	rfs4_deleg_state_t *dsp;
6397 
6398 	/*
6399 	 * Find the state info from the stateid and confirm that the
6400 	 * file is delegated.  If the state openowner is the same as
6401 	 * the supplied openowner we're done. If not, get the file
6402 	 * info from the found state info. Use that file info to
6403 	 * create the state for this lock owner. Note solaris doen't
6404 	 * really need the pathname to find the file. We may want to
6405 	 * lookup the pathname and make sure that the vp exist and
6406 	 * matches the vp in the file structure. However it is
6407 	 * possible that the pathname nolonger exists (local process
6408 	 * unlinks the file), so this may not be that useful.
6409 	 */
6410 
6411 	status = rfs4_get_deleg_state(&stateid, &dsp);
6412 	if (status != NFS4_OK) {
6413 		resp->status = status;
6414 		return;
6415 	}
6416 
6417 	ASSERT(dsp->finfo->dinfo->dtype != OPEN_DELEGATE_NONE);
6418 
6419 	/*
6420 	 * New lock owner, create state. Since this was probably called
6421 	 * in response to a CB_RECALL we set deleg to DELEG_NONE
6422 	 */
6423 
6424 	ASSERT(cs->vp != NULL);
6425 	VN_RELE(cs->vp);
6426 	VN_HOLD(dsp->finfo->vp);
6427 	cs->vp = dsp->finfo->vp;
6428 
6429 	if (error = makefh4(&cs->fh, cs->vp, cs->exi)) {
6430 		rfs4_deleg_state_rele(dsp);
6431 		*cs->statusp = resp->status = puterrno4(error);
6432 		return;
6433 	}
6434 
6435 	/* Mark progress for delegation returns */
6436 	dsp->finfo->dinfo->time_lastwrite = gethrestime_sec();
6437 	rfs4_deleg_state_rele(dsp);
6438 	rfs4_do_open(cs, req, oo, DELEG_NONE,
6439 				args->share_access, args->share_deny, resp);
6440 }
6441 
6442 /*ARGSUSED*/
6443 static void
6444 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req,
6445 			OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6446 {
6447 	/*
6448 	 * Lookup the pathname, it must already exist since this file
6449 	 * was delegated.
6450 	 *
6451 	 * Find the file and state info for this vp and open owner pair.
6452 	 *	check that they are in fact delegated.
6453 	 *	check that the state access and deny modes are the same.
6454 	 *
6455 	 * Return the delgation possibly seting the recall flag.
6456 	 */
6457 	rfs4_file_t *file;
6458 	rfs4_state_t *state;
6459 	bool_t create = FALSE;
6460 	bool_t dcreate = FALSE;
6461 	rfs4_deleg_state_t *dsp;
6462 	nfsace4 *ace;
6463 
6464 
6465 	/* Note we ignore oflags */
6466 	resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev,
6467 				req, cs, args->share_access, &resp->cinfo);
6468 
6469 	if (resp->status != NFS4_OK) {
6470 		return;
6471 	}
6472 
6473 	/* get the file struct and hold a lock on it during initial open */
6474 	file = rfs4_findfile_withlock(cs->vp, NULL, &create);
6475 	if (file == NULL) {
6476 		NFS4_DEBUG(rfs4_debug,
6477 			(CE_NOTE, "rfs4_do_opendelprev: can't find file"));
6478 		resp->status = NFS4ERR_SERVERFAULT;
6479 		return;
6480 	}
6481 
6482 	state = rfs4_findstate_by_owner_file(oo, file, &create);
6483 	if (state == NULL) {
6484 		NFS4_DEBUG(rfs4_debug,
6485 			(CE_NOTE, "rfs4_do_opendelprev: can't find state"));
6486 		resp->status = NFS4ERR_SERVERFAULT;
6487 		rfs4_file_rele_withunlock(file);
6488 		return;
6489 	}
6490 
6491 	rfs4_dbe_lock(state->dbe);
6492 	rfs4_dbe_lock(file->dbe);
6493 	if (args->share_access != state->share_access ||
6494 			args->share_deny != state->share_deny ||
6495 			state->finfo->dinfo->dtype == OPEN_DELEGATE_NONE) {
6496 		NFS4_DEBUG(rfs4_debug,
6497 			(CE_NOTE, "rfs4_do_opendelprev: state mixup"));
6498 		rfs4_dbe_unlock(file->dbe);
6499 		rfs4_dbe_unlock(state->dbe);
6500 		rfs4_file_rele(file);
6501 		rfs4_state_rele(state);
6502 		resp->status = NFS4ERR_SERVERFAULT;
6503 		return;
6504 	}
6505 	rfs4_dbe_unlock(file->dbe);
6506 	rfs4_dbe_unlock(state->dbe);
6507 
6508 	dsp = rfs4_finddeleg(state, &dcreate);
6509 	if (dsp == NULL) {
6510 		rfs4_state_rele(state);
6511 		rfs4_file_rele(file);
6512 		resp->status = NFS4ERR_SERVERFAULT;
6513 		return;
6514 	}
6515 
6516 	next_stateid(&state->stateid);
6517 
6518 	resp->stateid = state->stateid.stateid;
6519 
6520 	resp->delegation.delegation_type = dsp->dtype;
6521 
6522 	if (dsp->dtype == OPEN_DELEGATE_READ) {
6523 		open_read_delegation4 *rv =
6524 			&resp->delegation.open_delegation4_u.read;
6525 
6526 		rv->stateid = dsp->delegid.stateid;
6527 		rv->recall = FALSE; /* no policy in place to set to TRUE */
6528 		ace = &rv->permissions;
6529 	} else {
6530 		open_write_delegation4 *rv =
6531 			&resp->delegation.open_delegation4_u.write;
6532 
6533 		rv->stateid = dsp->delegid.stateid;
6534 		rv->recall = FALSE;  /* no policy in place to set to TRUE */
6535 		ace = &rv->permissions;
6536 		rv->space_limit.limitby = NFS_LIMIT_SIZE;
6537 		rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX;
6538 	}
6539 
6540 	/* XXX For now */
6541 	ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE;
6542 	ace->flag = 0;
6543 	ace->access_mask = 0;
6544 	ace->who.utf8string_len = 0;
6545 	ace->who.utf8string_val = 0;
6546 
6547 	rfs4_deleg_state_rele(dsp);
6548 	rfs4_state_rele(state);
6549 	rfs4_file_rele(file);
6550 }
6551 
6552 typedef enum {
6553 	NFS4_CHKSEQ_OKAY = 0,
6554 	NFS4_CHKSEQ_REPLAY = 1,
6555 	NFS4_CHKSEQ_BAD = 2
6556 } rfs4_chkseq_t;
6557 
6558 /*
6559  * Generic function for sequence number checks.
6560  */
6561 static rfs4_chkseq_t
6562 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop,
6563 		seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres)
6564 {
6565 	/* Same sequence ids and matching operations? */
6566 	if (seqid == rqst_seq && resop->resop == lastop->resop) {
6567 		if (copyres == TRUE) {
6568 			rfs4_free_reply(resop);
6569 			rfs4_copy_reply(resop, lastop);
6570 		}
6571 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
6572 			"Replayed SEQID %d\n", seqid));
6573 		return (NFS4_CHKSEQ_REPLAY);
6574 	}
6575 
6576 	/* If the incoming sequence is not the next expected then it is bad */
6577 	if (rqst_seq != seqid + 1) {
6578 		if (rqst_seq == seqid) {
6579 			NFS4_DEBUG(rfs4_debug,
6580 				(CE_NOTE, "BAD SEQID: Replayed sequence id "
6581 				"but last op was %d current op is %d\n",
6582 				lastop->resop, resop->resop));
6583 			return (NFS4_CHKSEQ_BAD);
6584 		}
6585 		NFS4_DEBUG(rfs4_debug,
6586 			(CE_NOTE, "BAD SEQID: got %u expecting %u\n",
6587 				rqst_seq, seqid));
6588 		return (NFS4_CHKSEQ_BAD);
6589 	}
6590 
6591 	/* Everything okay -- next expected */
6592 	return (NFS4_CHKSEQ_OKAY);
6593 }
6594 
6595 
6596 static rfs4_chkseq_t
6597 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop)
6598 {
6599 	rfs4_chkseq_t rc;
6600 
6601 	rfs4_dbe_lock(op->dbe);
6602 	rc = rfs4_check_seqid(op->open_seqid, op->reply, seqid, resop, TRUE);
6603 	rfs4_dbe_unlock(op->dbe);
6604 
6605 	if (rc == NFS4_CHKSEQ_OKAY)
6606 		rfs4_update_lease(op->client);
6607 
6608 	return (rc);
6609 }
6610 
6611 static rfs4_chkseq_t
6612 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op,
6613 	nfs_resop4 *resop)
6614 {
6615 	rfs4_chkseq_t rc;
6616 
6617 	rfs4_dbe_lock(op->dbe);
6618 	rc = rfs4_check_seqid(op->open_seqid, op->reply,
6619 		olo_seqid, resop, FALSE);
6620 	rfs4_dbe_unlock(op->dbe);
6621 
6622 	return (rc);
6623 }
6624 
6625 static rfs4_chkseq_t
6626 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lp, nfs_resop4 *resop)
6627 {
6628 	rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY;
6629 
6630 	rfs4_dbe_lock(lp->dbe);
6631 	if (!lp->skip_seqid_check)
6632 		rc = rfs4_check_seqid(lp->seqid, lp->reply,
6633 			seqid, resop, TRUE);
6634 	rfs4_dbe_unlock(lp->dbe);
6635 
6636 	return (rc);
6637 }
6638 
6639 static void
6640 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop,
6641 	    struct svc_req *req, struct compound_state *cs)
6642 {
6643 	OPEN4args *args = &argop->nfs_argop4_u.opopen;
6644 	OPEN4res *resp = &resop->nfs_resop4_u.opopen;
6645 	open_owner4 *owner = &args->owner;
6646 	open_claim_type4 claim = args->claim;
6647 	rfs4_client_t *cp;
6648 	rfs4_openowner_t *oo;
6649 	bool_t create;
6650 	bool_t replay = FALSE;
6651 	int can_reclaim;
6652 
6653 
6654 	if (cs->vp == NULL) {
6655 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
6656 		return;
6657 	}
6658 
6659 	/*
6660 	 * Need to check clientid and lease expiration first based on
6661 	 * error ordering and incrementing sequence id.
6662 	 */
6663 	cp = rfs4_findclient_by_id(owner->clientid, FALSE);
6664 	if (cp == NULL) {
6665 		*cs->statusp = resp->status =
6666 			rfs4_check_clientid(&owner->clientid, 0);
6667 		return;
6668 	}
6669 
6670 	if (rfs4_lease_expired(cp)) {
6671 		rfs4_client_close(cp);
6672 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
6673 		return;
6674 	}
6675 	can_reclaim = cp->can_reclaim;
6676 
6677 	/*
6678 	 * Find the open_owner for use from this point forward.  Take
6679 	 * care in updating the sequence id based on the type of error
6680 	 * being returned.
6681 	 */
6682 retry:
6683 	create = TRUE;
6684 	oo = rfs4_findopenowner(owner, &create, args->seqid);
6685 	if (oo == NULL) {
6686 		*cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
6687 		rfs4_client_rele(cp);
6688 		return;
6689 	}
6690 
6691 	/* Hold off access to the sequence space while the open is done */
6692 	rfs4_sw_enter(&oo->oo_sw);
6693 
6694 	/*
6695 	 * If the open_owner existed before at the server, then check
6696 	 * the sequence id.
6697 	 */
6698 	if (!create && !oo->postpone_confirm) {
6699 		switch (rfs4_check_open_seqid(args->seqid, oo, resop)) {
6700 		case NFS4_CHKSEQ_BAD:
6701 			if ((args->seqid > oo->open_seqid) &&
6702 				oo->need_confirm) {
6703 				rfs4_free_opens(oo, TRUE, FALSE);
6704 				rfs4_sw_exit(&oo->oo_sw);
6705 				rfs4_openowner_rele(oo);
6706 				goto retry;
6707 			}
6708 			resp->status = NFS4ERR_BAD_SEQID;
6709 			goto out;
6710 		case NFS4_CHKSEQ_REPLAY: /* replay of previous request */
6711 			replay = TRUE;
6712 			goto out;
6713 		default:
6714 			break;
6715 		}
6716 
6717 		/*
6718 		 * Sequence was ok and open owner exists
6719 		 * check to see if we have yet to see an
6720 		 * open_confirm.
6721 		 */
6722 		if (oo->need_confirm) {
6723 			rfs4_free_opens(oo, TRUE, FALSE);
6724 			rfs4_sw_exit(&oo->oo_sw);
6725 			rfs4_openowner_rele(oo);
6726 			goto retry;
6727 		}
6728 	}
6729 	/* Grace only applies to regular-type OPENs */
6730 	if (rfs4_clnt_in_grace(cp) &&
6731 	    (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) {
6732 		*cs->statusp = resp->status = NFS4ERR_GRACE;
6733 		goto out;
6734 	}
6735 
6736 	/*
6737 	 * If previous state at the server existed then can_reclaim
6738 	 * will be set. If not reply NFS4ERR_NO_GRACE to the
6739 	 * client.
6740 	 */
6741 	if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) {
6742 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
6743 		goto out;
6744 	}
6745 
6746 
6747 	/*
6748 	 * Reject the open if the client has missed the grace period
6749 	 */
6750 	if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) {
6751 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
6752 		goto out;
6753 	}
6754 
6755 	/* Couple of up-front bookkeeping items */
6756 	if (oo->need_confirm) {
6757 		/*
6758 		 * If this is a reclaim OPEN then we should not ask
6759 		 * for a confirmation of the open_owner per the
6760 		 * protocol specification.
6761 		 */
6762 		if (claim == CLAIM_PREVIOUS)
6763 			oo->need_confirm = FALSE;
6764 		else
6765 			resp->rflags |= OPEN4_RESULT_CONFIRM;
6766 	}
6767 	resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX;
6768 
6769 	/*
6770 	 * If there is an unshared filesystem mounted on this vnode,
6771 	 * do not allow to open/create in this directory.
6772 	 */
6773 	if (vn_ismntpt(cs->vp)) {
6774 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
6775 		goto out;
6776 	}
6777 
6778 	/*
6779 	 * access must READ, WRITE, or BOTH.  No access is invalid.
6780 	 * deny can be READ, WRITE, BOTH, or NONE.
6781 	 * bits not defined for access/deny are invalid.
6782 	 */
6783 	if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) ||
6784 	    (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) ||
6785 	    (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) {
6786 		*cs->statusp = resp->status = NFS4ERR_INVAL;
6787 		goto out;
6788 	}
6789 
6790 
6791 	/*
6792 	 * make sure attrset is zero before response is built.
6793 	 */
6794 	resp->attrset = 0;
6795 
6796 	switch (claim) {
6797 	case CLAIM_NULL:
6798 		rfs4_do_opennull(cs, req, args, oo, resp);
6799 	    break;
6800 	case CLAIM_PREVIOUS:
6801 		rfs4_do_openprev(cs, req, args, oo, resp);
6802 	    break;
6803 	case CLAIM_DELEGATE_CUR:
6804 		rfs4_do_opendelcur(cs, req, args, oo, resp);
6805 	    break;
6806 	case CLAIM_DELEGATE_PREV:
6807 		rfs4_do_opendelprev(cs, req, args, oo, resp);
6808 	    break;
6809 	default:
6810 		resp->status = NFS4ERR_INVAL;
6811 		break;
6812 	}
6813 
6814 out:
6815 	rfs4_client_rele(cp);
6816 
6817 	/* Catch sequence id handling here to make it a little easier */
6818 	switch (resp->status) {
6819 	case NFS4ERR_BADXDR:
6820 	case NFS4ERR_BAD_SEQID:
6821 	case NFS4ERR_BAD_STATEID:
6822 	case NFS4ERR_NOFILEHANDLE:
6823 	case NFS4ERR_RESOURCE:
6824 	case NFS4ERR_STALE_CLIENTID:
6825 	case NFS4ERR_STALE_STATEID:
6826 		/*
6827 		 * The protocol states that if any of these errors are
6828 		 * being returned, the sequence id should not be
6829 		 * incremented.  Any other return requires an
6830 		 * increment.
6831 		 */
6832 		break;
6833 	default:
6834 		/* Always update the lease in this case */
6835 		rfs4_update_lease(oo->client);
6836 
6837 		/* Regular response - copy the result */
6838 		if (!replay)
6839 			rfs4_update_open_resp(oo, resop, &cs->fh);
6840 
6841 		/*
6842 		 * REPLAY case: Only if the previous response was OK
6843 		 * do we copy the filehandle.  If not OK, no
6844 		 * filehandle to copy.
6845 		 */
6846 		if (replay == TRUE &&
6847 		    resp->status == NFS4_OK &&
6848 		    oo->reply_fh.nfs_fh4_val) {
6849 			/*
6850 			 * If this is a replay, we must restore the
6851 			 * current filehandle/vp to that of what was
6852 			 * returned originally.  Try our best to do
6853 			 * it.
6854 			 */
6855 			nfs_fh4_fmt_t *fh_fmtp =
6856 				(nfs_fh4_fmt_t *)oo->reply_fh.nfs_fh4_val;
6857 
6858 			cs->exi = checkexport4(&fh_fmtp->fh4_fsid,
6859 				(fid_t *)&fh_fmtp->fh4_xlen, NULL);
6860 
6861 			if (cs->exi == NULL) {
6862 				resp->status = NFS4ERR_STALE;
6863 				goto finish;
6864 			}
6865 
6866 			VN_RELE(cs->vp);
6867 
6868 			cs->vp = nfs4_fhtovp(&oo->reply_fh, cs->exi,
6869 				&resp->status);
6870 
6871 			if (cs->vp == NULL)
6872 				goto finish;
6873 
6874 			nfs_fh4_copy(&oo->reply_fh, &cs->fh);
6875 		}
6876 
6877 		/*
6878 		 * If this was a replay, no need to update the
6879 		 * sequence id. If the open_owner was not created on
6880 		 * this pass, then update.  The first use of an
6881 		 * open_owner will not bump the sequence id.
6882 		 */
6883 		if (replay == FALSE && !create)
6884 			rfs4_update_open_sequence(oo);
6885 		/*
6886 		 * If the client is receiving an error and the
6887 		 * open_owner needs to be confirmed, there is no way
6888 		 * to notify the client of this fact ignoring the fact
6889 		 * that the server has no method of returning a
6890 		 * stateid to confirm.  Therefore, the server needs to
6891 		 * mark this open_owner in a way as to avoid the
6892 		 * sequence id checking the next time the client uses
6893 		 * this open_owner.
6894 		 */
6895 		if (resp->status != NFS4_OK && oo->need_confirm)
6896 			oo->postpone_confirm = TRUE;
6897 		/*
6898 		 * If OK response then clear the postpone flag and
6899 		 * reset the sequence id to keep in sync with the
6900 		 * client.
6901 		 */
6902 		if (resp->status == NFS4_OK && oo->postpone_confirm) {
6903 			oo->postpone_confirm = FALSE;
6904 			oo->open_seqid = args->seqid;
6905 		}
6906 		break;
6907 	}
6908 
6909 finish:
6910 	*cs->statusp = resp->status;
6911 
6912 	rfs4_sw_exit(&oo->oo_sw);
6913 	rfs4_openowner_rele(oo);
6914 }
6915 
6916 /*ARGSUSED*/
6917 void
6918 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
6919 		    struct svc_req *req, struct compound_state *cs)
6920 {
6921 	OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm;
6922 	OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm;
6923 	rfs4_state_t *sp;
6924 	nfsstat4 status;
6925 
6926 	if (cs->vp == NULL) {
6927 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
6928 		return;
6929 	}
6930 
6931 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
6932 	if (status != NFS4_OK) {
6933 		*cs->statusp = resp->status = status;
6934 		return;
6935 	}
6936 
6937 	/* Ensure specified filehandle matches */
6938 	if (cs->vp != sp->finfo->vp) {
6939 		rfs4_state_rele(sp);
6940 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
6941 		return;
6942 	}
6943 
6944 	/* hold off other access to open_owner while we tinker */
6945 	rfs4_sw_enter(&sp->owner->oo_sw);
6946 
6947 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
6948 	case NFS4_CHECK_STATEID_OKAY:
6949 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
6950 			resop) != 0) {
6951 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
6952 			break;
6953 		}
6954 		/*
6955 		 * If it is the appropriate stateid and determined to
6956 		 * be "OKAY" then this means that the stateid does not
6957 		 * need to be confirmed and the client is in error for
6958 		 * sending an OPEN_CONFIRM.
6959 		 */
6960 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
6961 		break;
6962 	case NFS4_CHECK_STATEID_OLD:
6963 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
6964 		break;
6965 	case NFS4_CHECK_STATEID_BAD:
6966 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
6967 		break;
6968 	case NFS4_CHECK_STATEID_EXPIRED:
6969 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
6970 		break;
6971 	case NFS4_CHECK_STATEID_CLOSED:
6972 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
6973 		break;
6974 	case NFS4_CHECK_STATEID_REPLAY:
6975 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
6976 		case NFS4_CHKSEQ_OKAY:
6977 			/*
6978 			 * This is replayed stateid; if seqid matches
6979 			 * next expected, then client is using wrong seqid.
6980 			 */
6981 			/* fall through */
6982 		case NFS4_CHKSEQ_BAD:
6983 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
6984 			break;
6985 		case NFS4_CHKSEQ_REPLAY:
6986 			/*
6987 			 * Note this case is the duplicate case so
6988 			 * resp->status is already set.
6989 			 */
6990 			*cs->statusp = resp->status;
6991 			rfs4_update_lease(sp->owner->client);
6992 			break;
6993 		}
6994 		break;
6995 	case NFS4_CHECK_STATEID_UNCONFIRMED:
6996 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
6997 			resop) != NFS4_CHKSEQ_OKAY) {
6998 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
6999 			break;
7000 		}
7001 		*cs->statusp = resp->status = NFS4_OK;
7002 
7003 		next_stateid(&sp->stateid);
7004 		resp->open_stateid = sp->stateid.stateid;
7005 		sp->owner->need_confirm = FALSE;
7006 		rfs4_update_lease(sp->owner->client);
7007 		rfs4_update_open_sequence(sp->owner);
7008 		rfs4_update_open_resp(sp->owner, resop, NULL);
7009 		break;
7010 	default:
7011 		ASSERT(FALSE);
7012 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7013 		break;
7014 	}
7015 	rfs4_sw_exit(&sp->owner->oo_sw);
7016 	rfs4_state_rele(sp);
7017 }
7018 
7019 /*ARGSUSED*/
7020 void
7021 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop,
7022 		    struct svc_req *req, struct compound_state *cs)
7023 {
7024 	OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade;
7025 	OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade;
7026 	uint32_t access = args->share_access;
7027 	uint32_t deny = args->share_deny;
7028 	nfsstat4 status;
7029 	rfs4_state_t *sp;
7030 	rfs4_file_t *fp;
7031 
7032 	if (cs->vp == NULL) {
7033 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7034 		return;
7035 	}
7036 
7037 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
7038 	if (status != NFS4_OK) {
7039 		*cs->statusp = resp->status = status;
7040 		return;
7041 	}
7042 
7043 	/* Ensure specified filehandle matches */
7044 	if (cs->vp != sp->finfo->vp) {
7045 		rfs4_state_rele(sp);
7046 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7047 		return;
7048 	}
7049 
7050 	/* hold off other access to open_owner while we tinker */
7051 	rfs4_sw_enter(&sp->owner->oo_sw);
7052 
7053 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7054 	case NFS4_CHECK_STATEID_OKAY:
7055 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7056 			resop) != NFS4_CHKSEQ_OKAY) {
7057 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7058 			goto end;
7059 		}
7060 		break;
7061 	case NFS4_CHECK_STATEID_OLD:
7062 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7063 		goto end;
7064 	case NFS4_CHECK_STATEID_BAD:
7065 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7066 		goto end;
7067 	case NFS4_CHECK_STATEID_EXPIRED:
7068 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7069 		goto end;
7070 	case NFS4_CHECK_STATEID_CLOSED:
7071 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7072 		goto end;
7073 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7074 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7075 		goto end;
7076 	case NFS4_CHECK_STATEID_REPLAY:
7077 		/* Check the sequence id for the open owner */
7078 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
7079 		case NFS4_CHKSEQ_OKAY:
7080 			/*
7081 			 * This is replayed stateid; if seqid matches
7082 			 * next expected, then client is using wrong seqid.
7083 			 */
7084 			/* fall through */
7085 		case NFS4_CHKSEQ_BAD:
7086 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7087 			goto end;
7088 		case NFS4_CHKSEQ_REPLAY:
7089 			/*
7090 			 * Note this case is the duplicate case so
7091 			 * resp->status is already set.
7092 			 */
7093 			*cs->statusp = resp->status;
7094 			rfs4_update_lease(sp->owner->client);
7095 			goto end;
7096 		}
7097 		break;
7098 	default:
7099 		ASSERT(FALSE);
7100 		break;
7101 	}
7102 
7103 	rfs4_dbe_lock(sp->dbe);
7104 	/*
7105 	 * Check that the new access modes and deny modes are valid.
7106 	 * Check that no invalid bits are set.
7107 	 */
7108 	if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) ||
7109 	    (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_READ))) {
7110 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7111 		rfs4_update_open_sequence(sp->owner);
7112 		rfs4_dbe_unlock(sp->dbe);
7113 		goto end;
7114 	}
7115 
7116 	/*
7117 	 * The new modes must be a subset of the current modes and
7118 	 * the access must specify at least one mode. To test that
7119 	 * the new mode is a subset of the current modes we bitwise
7120 	 * AND them together and check that the result equals the new
7121 	 * mode. For example:
7122 	 * New mode, access == R and current mode, sp->share_access  == RW
7123 	 * access & sp->share_access == R == access, so the new access mode
7124 	 * is valid. Consider access == RW, sp->share_access = R
7125 	 * access & sp->share_access == R != access, so the new access mode
7126 	 * is invalid.
7127 	 */
7128 	if ((access & sp->share_access) != access ||
7129 	    (deny & sp->share_deny) != deny ||
7130 	    (access &
7131 	    (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) {
7132 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7133 		rfs4_update_open_sequence(sp->owner);
7134 		rfs4_dbe_unlock(sp->dbe);
7135 		goto end;
7136 	}
7137 
7138 	/*
7139 	 * Release any share locks associated with this stateID.
7140 	 * Strictly speaking, this violates the spec because the
7141 	 * spec effectively requires that open downgrade be atomic.
7142 	 * At present, fs_shrlock does not have this capability.
7143 	 */
7144 	rfs4_dbe_unlock(sp->dbe);
7145 	rfs4_unshare(sp);
7146 	rfs4_dbe_lock(sp->dbe);
7147 
7148 	fp = sp->finfo;
7149 	rfs4_dbe_lock(fp->dbe);
7150 
7151 	/*
7152 	 * If the current mode has deny read and the new mode
7153 	 * does not, decrement the number of deny read mode bits
7154 	 * and if it goes to zero turn off the deny read bit
7155 	 * on the file.
7156 	 */
7157 	if ((sp->share_deny & OPEN4_SHARE_DENY_READ) &&
7158 	    (deny & OPEN4_SHARE_DENY_READ) == 0) {
7159 		fp->deny_read--;
7160 		if (fp->deny_read == 0)
7161 			fp->share_deny &= ~OPEN4_SHARE_DENY_READ;
7162 	}
7163 
7164 	/*
7165 	 * If the current mode has deny write and the new mode
7166 	 * does not, decrement the number of deny write mode bits
7167 	 * and if it goes to zero turn off the deny write bit
7168 	 * on the file.
7169 	 */
7170 	if ((sp->share_deny & OPEN4_SHARE_DENY_WRITE) &&
7171 	    (deny & OPEN4_SHARE_DENY_WRITE) == 0) {
7172 		fp->deny_write--;
7173 		if (fp->deny_write == 0)
7174 			fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE;
7175 	}
7176 
7177 	/*
7178 	 * If the current mode has access read and the new mode
7179 	 * does not, decrement the number of access read mode bits
7180 	 * and if it goes to zero turn off the access read bit
7181 	 * on the file.
7182 	 */
7183 	if ((sp->share_access & OPEN4_SHARE_ACCESS_READ) &&
7184 	    (access & OPEN4_SHARE_ACCESS_READ) == 0) {
7185 		fp->access_read--;
7186 		if (fp->access_read == 0)
7187 			fp->share_access &= ~OPEN4_SHARE_ACCESS_READ;
7188 	}
7189 
7190 	/*
7191 	 * If the current mode has access write and the new mode
7192 	 * does not, decrement the number of access write mode bits
7193 	 * and if it goes to zero turn off the access write bit
7194 	 * on the file.
7195 	 */
7196 	if ((sp->share_access & OPEN4_SHARE_ACCESS_WRITE) &&
7197 	    (access & OPEN4_SHARE_ACCESS_WRITE) == 0) {
7198 		fp->access_write--;
7199 		if (fp->access_write == 0)
7200 			fp->share_deny &= ~OPEN4_SHARE_ACCESS_WRITE;
7201 	}
7202 
7203 	/* Set the new access and deny modes */
7204 	sp->share_access = access;
7205 	sp->share_deny = deny;
7206 	/* Check that the file is still accessible */
7207 	ASSERT(fp->share_access);
7208 
7209 	rfs4_dbe_unlock(fp->dbe);
7210 
7211 	rfs4_dbe_unlock(sp->dbe);
7212 	if ((status = rfs4_share(sp)) != NFS4_OK) {
7213 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7214 		rfs4_update_open_sequence(sp->owner);
7215 		goto end;
7216 	}
7217 
7218 	rfs4_dbe_lock(sp->dbe);
7219 
7220 	/* Update the stateid */
7221 	next_stateid(&sp->stateid);
7222 	resp->open_stateid = sp->stateid.stateid;
7223 
7224 	rfs4_dbe_unlock(sp->dbe);
7225 
7226 	*cs->statusp = resp->status = NFS4_OK;
7227 	/* Update the lease */
7228 	rfs4_update_lease(sp->owner->client);
7229 	/* And the sequence */
7230 	rfs4_update_open_sequence(sp->owner);
7231 	rfs4_update_open_resp(sp->owner, resop, NULL);
7232 
7233 end:
7234 	rfs4_sw_exit(&sp->owner->oo_sw);
7235 	rfs4_state_rele(sp);
7236 }
7237 
7238 /*
7239  * The logic behind this function is detailed in the NFSv4 RFC in the
7240  * SETCLIENTID operation description under IMPLEMENTATION.  Refer to
7241  * that section for explicit guidance to server behavior for
7242  * SETCLIENTID.
7243  */
7244 void
7245 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop,
7246 		    struct svc_req *req, struct compound_state *cs)
7247 {
7248 	SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid;
7249 	SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid;
7250 	rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed;
7251 	bool_t create = TRUE;
7252 	char *addr, *netid;
7253 	int len;
7254 
7255 retry:
7256 	newcp = cp_confirmed = cp_unconfirmed = NULL;
7257 
7258 	/*
7259 	 * In search of an EXISTING client matching the incoming
7260 	 * request to establish a new client identifier at the server
7261 	 */
7262 	create = TRUE;
7263 	cp = rfs4_findclient(&args->client, &create, NULL);
7264 
7265 	/* Should never happen */
7266 	ASSERT(cp != NULL);
7267 
7268 	if (cp == NULL) {
7269 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
7270 		return;
7271 	}
7272 
7273 	/*
7274 	 * Easiest case. Client identifier is newly created and is
7275 	 * unconfirmed.  Also note that for this case, no other
7276 	 * entries exist for the client identifier.  Nothing else to
7277 	 * check.  Just setup the response and respond.
7278 	 */
7279 	if (create) {
7280 		*cs->statusp = res->status = NFS4_OK;
7281 		res->SETCLIENTID4res_u.resok4.clientid = cp->clientid;
7282 		res->SETCLIENTID4res_u.resok4.setclientid_confirm =
7283 							cp->confirm_verf;
7284 		/* Setup callback information; CB_NULL confirmation later */
7285 		rfs4_client_setcb(cp, &args->callback, args->callback_ident);
7286 
7287 		rfs4_client_rele(cp);
7288 		return;
7289 	}
7290 
7291 	/*
7292 	 * An existing, confirmed client may exist but it may not have
7293 	 * been active for at least one lease period.  If so, then
7294 	 * "close" the client and create a new client identifier
7295 	 */
7296 	if (rfs4_lease_expired(cp)) {
7297 		rfs4_client_close(cp);
7298 		goto retry;
7299 	}
7300 
7301 	if (cp->need_confirm == TRUE)
7302 		cp_unconfirmed = cp;
7303 	else
7304 		cp_confirmed = cp;
7305 
7306 	cp = NULL;
7307 
7308 	/*
7309 	 * We have a confirmed client, now check for an
7310 	 * unconfimred entry
7311 	 */
7312 	if (cp_confirmed) {
7313 		/* If creds don't match then client identifier is inuse */
7314 		if (!creds_ok(cp_confirmed->cr_set, req, cs)) {
7315 			rfs4_cbinfo_t *cbp;
7316 			/*
7317 			 * Some one else has established this client
7318 			 * id. Try and say * who they are. We will use
7319 			 * the call back address supplied by * the
7320 			 * first client.
7321 			 */
7322 			*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
7323 
7324 			addr = netid = NULL;
7325 
7326 			cbp = &cp_confirmed->cbinfo;
7327 			if (cbp->cb_callback.cb_location.r_addr &&
7328 			    cbp->cb_callback.cb_location.r_netid) {
7329 				cb_client4 *cbcp = &cbp->cb_callback;
7330 
7331 				len = strlen(cbcp->cb_location.r_addr)+1;
7332 				addr = kmem_alloc(len, KM_SLEEP);
7333 				bcopy(cbcp->cb_location.r_addr, addr, len);
7334 				len = strlen(cbcp->cb_location.r_netid)+1;
7335 				netid = kmem_alloc(len, KM_SLEEP);
7336 				bcopy(cbcp->cb_location.r_netid, netid, len);
7337 			}
7338 
7339 			res->SETCLIENTID4res_u.client_using.r_addr = addr;
7340 			res->SETCLIENTID4res_u.client_using.r_netid = netid;
7341 
7342 			rfs4_client_rele(cp_confirmed);
7343 		}
7344 
7345 		/*
7346 		 * Confirmed, creds match, and verifier matches; must
7347 		 * be an update of the callback info
7348 		 */
7349 		if (cp_confirmed->nfs_client.verifier ==
7350 						args->client.verifier) {
7351 			/* Setup callback information */
7352 			rfs4_client_setcb(cp_confirmed, &args->callback,
7353 						args->callback_ident);
7354 
7355 			/* everything okay -- move ahead */
7356 			*cs->statusp = res->status = NFS4_OK;
7357 			res->SETCLIENTID4res_u.resok4.clientid =
7358 				cp_confirmed->clientid;
7359 
7360 			/* update the confirm_verifier and return it */
7361 			rfs4_client_scv_next(cp_confirmed);
7362 			res->SETCLIENTID4res_u.resok4.setclientid_confirm =
7363 						cp_confirmed->confirm_verf;
7364 
7365 			rfs4_client_rele(cp_confirmed);
7366 			return;
7367 		}
7368 
7369 		/*
7370 		 * Creds match but the verifier doesn't.  Must search
7371 		 * for an unconfirmed client that would be replaced by
7372 		 * this request.
7373 		 */
7374 		create = FALSE;
7375 		cp_unconfirmed = rfs4_findclient(&args->client, &create,
7376 						cp_confirmed);
7377 	}
7378 
7379 	/*
7380 	 * At this point, we have taken care of the brand new client
7381 	 * struct, INUSE case, update of an existing, and confirmed
7382 	 * client struct.
7383 	 */
7384 
7385 	/*
7386 	 * check to see if things have changed while we originally
7387 	 * picked up the client struct.  If they have, then return and
7388 	 * retry the processing of this SETCLIENTID request.
7389 	 */
7390 	if (cp_unconfirmed) {
7391 		rfs4_dbe_lock(cp_unconfirmed->dbe);
7392 		if (!cp_unconfirmed->need_confirm) {
7393 			rfs4_dbe_unlock(cp_unconfirmed->dbe);
7394 			rfs4_client_rele(cp_unconfirmed);
7395 			if (cp_confirmed)
7396 				rfs4_client_rele(cp_confirmed);
7397 			goto retry;
7398 		}
7399 		/* do away with the old unconfirmed one */
7400 		rfs4_dbe_invalidate(cp_unconfirmed->dbe);
7401 		rfs4_dbe_unlock(cp_unconfirmed->dbe);
7402 		rfs4_client_rele(cp_unconfirmed);
7403 		cp_unconfirmed = NULL;
7404 	}
7405 
7406 	/*
7407 	 * This search will temporarily hide the confirmed client
7408 	 * struct while a new client struct is created as the
7409 	 * unconfirmed one.
7410 	 */
7411 	create = TRUE;
7412 	newcp = rfs4_findclient(&args->client, &create, cp_confirmed);
7413 
7414 	ASSERT(newcp != NULL);
7415 
7416 	if (newcp == NULL) {
7417 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
7418 		rfs4_client_rele(cp_confirmed);
7419 		return;
7420 	}
7421 
7422 	/*
7423 	 * If one was not created, then a similar request must be in
7424 	 * process so release and start over with this one
7425 	 */
7426 	if (create != TRUE) {
7427 		rfs4_client_rele(newcp);
7428 		if (cp_confirmed)
7429 			rfs4_client_rele(cp_confirmed);
7430 		goto retry;
7431 	}
7432 
7433 	*cs->statusp = res->status = NFS4_OK;
7434 	res->SETCLIENTID4res_u.resok4.clientid = newcp->clientid;
7435 	res->SETCLIENTID4res_u.resok4.setclientid_confirm =
7436 							newcp->confirm_verf;
7437 	/* Setup callback information; CB_NULL confirmation later */
7438 	rfs4_client_setcb(newcp, &args->callback,
7439 				args->callback_ident);
7440 
7441 	newcp->cp_confirmed = cp_confirmed;
7442 
7443 	rfs4_client_rele(newcp);
7444 }
7445 
7446 /*ARGSUSED*/
7447 void
7448 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
7449 			    struct svc_req *req, struct compound_state *cs)
7450 {
7451 	SETCLIENTID_CONFIRM4args *args =
7452 		&argop->nfs_argop4_u.opsetclientid_confirm;
7453 	SETCLIENTID_CONFIRM4res *res =
7454 		&resop->nfs_resop4_u.opsetclientid_confirm;
7455 	rfs4_client_t *cp, *cptoclose = NULL;
7456 
7457 	*cs->statusp = res->status = NFS4_OK;
7458 
7459 	cp = rfs4_findclient_by_id(args->clientid, TRUE);
7460 
7461 	if (cp == NULL) {
7462 		*cs->statusp = res->status =
7463 			rfs4_check_clientid(&args->clientid, 1);
7464 		return;
7465 	}
7466 
7467 	if (!creds_ok(cp, req, cs)) {
7468 		*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
7469 		rfs4_client_rele(cp);
7470 		return;
7471 	}
7472 
7473 	/* If the verifier doesn't match, the record doesn't match */
7474 	if (cp->confirm_verf != args->setclientid_confirm) {
7475 		*cs->statusp = res->status = NFS4ERR_STALE_CLIENTID;
7476 		rfs4_client_rele(cp);
7477 		return;
7478 	}
7479 
7480 	rfs4_dbe_lock(cp->dbe);
7481 	cp->need_confirm = FALSE;
7482 	if (cp->cp_confirmed) {
7483 		cptoclose = cp->cp_confirmed;
7484 		cptoclose->ss_remove = 1;
7485 		cp->cp_confirmed = NULL;
7486 	}
7487 
7488 	/*
7489 	 * Record clientid in stable storage
7490 	 */
7491 	rfs4_ss_clid(cp, req);
7492 
7493 	rfs4_dbe_unlock(cp->dbe);
7494 
7495 	if (cptoclose)
7496 		/* don't need to rele, client_close does it */
7497 		rfs4_client_close(cptoclose);
7498 
7499 	/* If needed, initiate CB_NULL call for callback path */
7500 	rfs4_deleg_cb_check(cp);
7501 	rfs4_update_lease(cp);
7502 
7503 	/*
7504 	 * Update the client's associated server instance, if it's changed
7505 	 * since the client was created.
7506 	 */
7507 	if (rfs4_servinst(cp) != rfs4_cur_servinst)
7508 		rfs4_servinst_assign(cp, rfs4_cur_servinst);
7509 
7510 	/*
7511 	 * Check to see if client can perform reclaims
7512 	 */
7513 	rfs4_ss_chkclid(cp);
7514 
7515 	rfs4_client_rele(cp);
7516 }
7517 
7518 
7519 /*ARGSUSED*/
7520 void
7521 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop,
7522 	    struct svc_req *req, struct compound_state *cs)
7523 {
7524 	/* XXX Currently not using req arg */
7525 	CLOSE4args *args = &argop->nfs_argop4_u.opclose;
7526 	CLOSE4res *resp = &resop->nfs_resop4_u.opclose;
7527 	rfs4_state_t *sp;
7528 	nfsstat4 status;
7529 
7530 	if (cs->vp == NULL) {
7531 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7532 		return;
7533 	}
7534 
7535 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID);
7536 	if (status != NFS4_OK) {
7537 		*cs->statusp = resp->status = status;
7538 		return;
7539 	}
7540 
7541 	/* Ensure specified filehandle matches */
7542 	if (cs->vp != sp->finfo->vp) {
7543 		rfs4_state_rele(sp);
7544 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7545 		return;
7546 	}
7547 
7548 	/* hold off other access to open_owner while we tinker */
7549 	rfs4_sw_enter(&sp->owner->oo_sw);
7550 
7551 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7552 	case NFS4_CHECK_STATEID_OKAY:
7553 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7554 			resop) != NFS4_CHKSEQ_OKAY) {
7555 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7556 			goto end;
7557 		}
7558 		break;
7559 	case NFS4_CHECK_STATEID_OLD:
7560 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7561 		goto end;
7562 	case NFS4_CHECK_STATEID_BAD:
7563 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7564 		goto end;
7565 	case NFS4_CHECK_STATEID_EXPIRED:
7566 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7567 		goto end;
7568 	case NFS4_CHECK_STATEID_CLOSED:
7569 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7570 		goto end;
7571 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7572 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7573 		goto end;
7574 	case NFS4_CHECK_STATEID_REPLAY:
7575 		/* Check the sequence id for the open owner */
7576 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
7577 		case NFS4_CHKSEQ_OKAY:
7578 			/*
7579 			 * This is replayed stateid; if seqid matches
7580 			 * next expected, then client is using wrong seqid.
7581 			 */
7582 			/* FALL THROUGH */
7583 		case NFS4_CHKSEQ_BAD:
7584 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7585 			goto end;
7586 		case NFS4_CHKSEQ_REPLAY:
7587 			/*
7588 			 * Note this case is the duplicate case so
7589 			 * resp->status is already set.
7590 			 */
7591 			*cs->statusp = resp->status;
7592 			rfs4_update_lease(sp->owner->client);
7593 			goto end;
7594 		}
7595 		break;
7596 	default:
7597 		ASSERT(FALSE);
7598 		break;
7599 	}
7600 
7601 	rfs4_dbe_lock(sp->dbe);
7602 
7603 	/* Update the stateid. */
7604 	next_stateid(&sp->stateid);
7605 	resp->open_stateid = sp->stateid.stateid;
7606 
7607 	rfs4_dbe_unlock(sp->dbe);
7608 
7609 	rfs4_update_lease(sp->owner->client);
7610 	rfs4_update_open_sequence(sp->owner);
7611 	rfs4_update_open_resp(sp->owner, resop, NULL);
7612 
7613 	rfs4_state_close(sp, FALSE, FALSE, cs->cr);
7614 
7615 	*cs->statusp = resp->status = status;
7616 
7617 end:
7618 	rfs4_sw_exit(&sp->owner->oo_sw);
7619 	rfs4_state_rele(sp);
7620 }
7621 
7622 /*
7623  * Manage the counts on the file struct and close all file locks
7624  */
7625 /*ARGSUSED*/
7626 void
7627 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr,
7628 	bool_t close_of_client)
7629 {
7630 	rfs4_file_t *fp = sp->finfo;
7631 	rfs4_lo_state_t *lsp;
7632 	struct shrlock shr;
7633 	struct shr_locowner shr_loco;
7634 	int fflags, s_access, s_deny;
7635 
7636 	fflags = s_access = s_deny = 0;
7637 	/*
7638 	 * Decrement the count for each access and deny bit that this
7639 	 * state has contributed to the file. If the file counts go to zero
7640 	 * clear the appropriate bit in the appropriate mask.
7641 	 */
7642 
7643 	if (sp->share_access & OPEN4_SHARE_ACCESS_READ) {
7644 		fp->access_read--;
7645 		fflags |= FREAD;
7646 		s_access |= F_RDACC;
7647 		if (fp->access_read == 0)
7648 			fp->share_access &= ~OPEN4_SHARE_ACCESS_READ;
7649 	}
7650 	if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) {
7651 		fp->access_write--;
7652 		fflags |= FWRITE;
7653 		s_access |= F_WRACC;
7654 		if (fp->access_write == 0)
7655 			fp->share_access &= ~OPEN4_SHARE_ACCESS_WRITE;
7656 	}
7657 	if (sp->share_deny & OPEN4_SHARE_DENY_READ) {
7658 		fp->deny_read--;
7659 		s_deny |= F_RDDNY;
7660 		if (fp->deny_read == 0)
7661 			fp->share_deny &= ~OPEN4_SHARE_DENY_READ;
7662 	}
7663 	if (sp->share_deny & OPEN4_SHARE_DENY_WRITE) {
7664 		fp->deny_write--;
7665 		s_deny |= F_WRDNY;
7666 		if (fp->deny_write == 0)
7667 			fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE;
7668 	}
7669 
7670 	/*
7671 	 * If this call is part of the larger closing down of client
7672 	 * state then it is just easier to release all locks
7673 	 * associated with this client instead of going through each
7674 	 * individual file and cleaning locks there.
7675 	 */
7676 	if (close_of_client) {
7677 		if (sp->owner->client->unlksys_completed == FALSE &&
7678 		    sp->lockownerlist.next->lsp != NULL &&
7679 			sp->owner->client->sysidt != LM_NOSYSID) {
7680 			/* Is the PxFS kernel module loaded? */
7681 			if (lm_remove_file_locks != NULL) {
7682 				int new_sysid;
7683 
7684 				/* Encode the cluster nodeid in new sysid */
7685 				new_sysid = sp->owner->client->sysidt;
7686 				lm_set_nlmid_flk(&new_sysid);
7687 
7688 				/*
7689 				 * This PxFS routine removes file locks for a
7690 				 * client over all nodes of a cluster.
7691 				 */
7692 				NFS4_DEBUG(rfs4_debug, (CE_NOTE,
7693 				    "lm_remove_file_locks(sysid=0x%x)\n",
7694 				    new_sysid));
7695 				(*lm_remove_file_locks)(new_sysid);
7696 			} else {
7697 				struct flock64 flk;
7698 
7699 				/* Release all locks for this client */
7700 				flk.l_type = F_UNLKSYS;
7701 				flk.l_whence = 0;
7702 				flk.l_start = 0;
7703 				flk.l_len = 0;
7704 				flk.l_sysid = sp->owner->client->sysidt;
7705 				flk.l_pid = 0;
7706 				(void) VOP_FRLOCK(sp->finfo->vp, F_SETLK, &flk,
7707 				    F_REMOTELOCK | FREAD | FWRITE,
7708 				    (u_offset_t)0, NULL, CRED());
7709 			}
7710 
7711 			sp->owner->client->unlksys_completed = TRUE;
7712 		}
7713 	}
7714 
7715 	/*
7716 	 * Release all locks on this file by this lock owner or at
7717 	 * least mark the locks as having been released
7718 	 */
7719 	for (lsp = sp->lockownerlist.next->lsp; lsp != NULL;
7720 		lsp = lsp->lockownerlist.next->lsp) {
7721 
7722 		lsp->locks_cleaned = TRUE;
7723 
7724 		/* Was this already taken care of above? */
7725 		if (!close_of_client &&
7726 		    sp->owner->client->sysidt != LM_NOSYSID)
7727 			(void) cleanlocks(sp->finfo->vp, lsp->locker->pid,
7728 				lsp->locker->client->sysidt);
7729 	}
7730 
7731 	/*
7732 	 * Release any shrlocks associated with this open state ID.
7733 	 * This must be done before the rfs4_state gets marked closed.
7734 	 */
7735 	if (sp->owner->client->sysidt != LM_NOSYSID) {
7736 		shr.s_access = s_access;
7737 		shr.s_deny = s_deny;
7738 		shr.s_pid = rfs4_dbe_getid(sp->owner->dbe);
7739 		shr.s_sysid = sp->owner->client->sysidt;
7740 		shr_loco.sl_pid = shr.s_pid;
7741 		shr_loco.sl_id = shr.s_sysid;
7742 		shr.s_owner = (caddr_t)&shr_loco;
7743 		shr.s_own_len = sizeof (shr_loco);
7744 		(void) vop_shrlock(sp->finfo->vp, F_UNSHARE, &shr, fflags);
7745 	}
7746 }
7747 
7748 /*
7749  * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure.
7750  */
7751 static nfsstat4
7752 lock_denied(LOCK4denied *dp, struct flock64 *flk)
7753 {
7754 	rfs4_lockowner_t *lo;
7755 	rfs4_client_t *cp;
7756 	uint32_t len;
7757 
7758 	lo = rfs4_findlockowner_by_pid(flk->l_pid);
7759 	if (lo != NULL) {
7760 		cp = lo->client;
7761 		if (rfs4_lease_expired(cp)) {
7762 			rfs4_lockowner_rele(lo);
7763 			rfs4_dbe_hold(cp->dbe);
7764 			rfs4_client_close(cp);
7765 			return (NFS4ERR_EXPIRED);
7766 		}
7767 		dp->owner.clientid = lo->owner.clientid;
7768 		len = lo->owner.owner_len;
7769 		dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
7770 		bcopy(lo->owner.owner_val, dp->owner.owner_val, len);
7771 		dp->owner.owner_len = len;
7772 		rfs4_lockowner_rele(lo);
7773 		goto finish;
7774 	}
7775 
7776 	/*
7777 	 * Its not a NFS4 lock. We take advantage that the upper 32 bits
7778 	 * of the client id contain the boot time for a NFS4 lock. So we
7779 	 * fabricate and identity by setting clientid to the sysid, and
7780 	 * the lock owner to the pid.
7781 	 */
7782 	dp->owner.clientid = flk->l_sysid;
7783 	len = sizeof (pid_t);
7784 	dp->owner.owner_len = len;
7785 	dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
7786 	bcopy(&flk->l_pid, dp->owner.owner_val, len);
7787 finish:
7788 	dp->offset = flk->l_start;
7789 	dp->length = flk->l_len;
7790 
7791 	if (flk->l_type == F_RDLCK)
7792 		dp->locktype = READ_LT;
7793 	else if (flk->l_type == F_WRLCK)
7794 		dp->locktype = WRITE_LT;
7795 	else
7796 		return (NFS4ERR_INVAL);	/* no mapping from POSIX ltype to v4 */
7797 
7798 	return (NFS4_OK);
7799 }
7800 
7801 static int
7802 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred)
7803 {
7804 	int error;
7805 	struct flock64 flk;
7806 	int i;
7807 	clock_t delaytime;
7808 
7809 retry:
7810 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
7811 
7812 	for (i = 0; i < rfs4_maxlock_tries; i++) {
7813 		LOCK_PRINT(rfs4_debug, "setlock", F_SETLK, flock);
7814 		error = VOP_FRLOCK(vp, F_SETLK,
7815 				flock, flag, (u_offset_t)0, NULL, cred);
7816 
7817 		if (error != EAGAIN && error != EACCES)
7818 			break;
7819 
7820 		if (i < rfs4_maxlock_tries - 1) {
7821 			delay(delaytime);
7822 			delaytime *= 2;
7823 		}
7824 	}
7825 
7826 	if (error == EAGAIN || error == EACCES) {
7827 		/* Get the owner of the lock */
7828 		flk = *flock;
7829 		LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk);
7830 		if (VOP_FRLOCK(vp, F_GETLK,
7831 			    &flk,  flag, (u_offset_t)0, NULL, cred) == 0) {
7832 			if (flk.l_type == F_UNLCK) {
7833 				/* No longer locked, retry */
7834 				goto retry;
7835 			}
7836 			*flock = flk;
7837 			LOCK_PRINT(rfs4_debug, "setlock(blocking lock)",
7838 				F_GETLK, &flk);
7839 		}
7840 	}
7841 
7842 	return (error);
7843 }
7844 
7845 /*ARGSUSED*/
7846 static nfsstat4
7847 rfs4_do_lock(rfs4_lo_state_t *lp, nfs_lock_type4 locktype,
7848 	    seqid4 seqid, offset4 offset,
7849 	    length4 length, cred_t *cred, nfs_resop4 *resop)
7850 {
7851 	nfsstat4 status;
7852 	rfs4_lockowner_t *lo = lp->locker;
7853 	rfs4_state_t *sp = lp->state;
7854 	struct flock64 flock;
7855 	int16_t ltype;
7856 	int flag;
7857 	int error;
7858 	sysid_t sysid;
7859 	LOCK4res *lres;
7860 
7861 	if (rfs4_lease_expired(lo->client)) {
7862 		return (NFS4ERR_EXPIRED);
7863 	}
7864 
7865 	if ((status = rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK)
7866 		return (status);
7867 
7868 	/* Check for zero length. To lock to end of file use all ones for V4 */
7869 	if (length == 0)
7870 		return (NFS4ERR_INVAL);
7871 	else if (length == (length4)(~0))
7872 		length = 0;		/* Posix to end of file  */
7873 
7874 retry:
7875 	rfs4_dbe_lock(sp->dbe);
7876 
7877 
7878 	if (resop->resop != OP_LOCKU) {
7879 		switch (locktype) {
7880 		case READ_LT:
7881 		case READW_LT:
7882 			if ((sp->share_access
7883 			    & OPEN4_SHARE_ACCESS_READ) == 0) {
7884 				rfs4_dbe_unlock(sp->dbe);
7885 
7886 				return (NFS4ERR_OPENMODE);
7887 			}
7888 			ltype = F_RDLCK;
7889 			break;
7890 		case WRITE_LT:
7891 		case WRITEW_LT:
7892 			if ((sp->share_access
7893 			    & OPEN4_SHARE_ACCESS_WRITE) == 0) {
7894 				rfs4_dbe_unlock(sp->dbe);
7895 
7896 				return (NFS4ERR_OPENMODE);
7897 			}
7898 			ltype = F_WRLCK;
7899 			break;
7900 		}
7901 	} else
7902 		ltype = F_UNLCK;
7903 
7904 	flock.l_type = ltype;
7905 	flock.l_whence = 0;		/* SEEK_SET */
7906 	flock.l_start = offset;
7907 	flock.l_len = length;
7908 	flock.l_sysid = sysid;
7909 	flock.l_pid = lp->locker->pid;
7910 
7911 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
7912 	if (flock.l_len < 0 || flock.l_start < 0) {
7913 		rfs4_dbe_unlock(sp->dbe);
7914 		return (NFS4ERR_INVAL);
7915 	}
7916 
7917 	/*
7918 	 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and
7919 	 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE.
7920 	 */
7921 	flag = (int)sp->share_access | F_REMOTELOCK;
7922 
7923 	error = setlock(sp->finfo->vp, &flock, flag, cred);
7924 	if (error == 0) {
7925 		rfs4_dbe_lock(lp->dbe);
7926 		next_stateid(&lp->lockid);
7927 		rfs4_dbe_unlock(lp->dbe);
7928 	}
7929 
7930 	rfs4_dbe_unlock(sp->dbe);
7931 
7932 	/*
7933 	 * N.B. We map error values to nfsv4 errors. This is differrent
7934 	 * than puterrno4 routine.
7935 	 */
7936 	switch (error) {
7937 	case 0:
7938 		status = NFS4_OK;
7939 		break;
7940 	case EAGAIN:
7941 	case EACCES:		/* Old value */
7942 		/* Can only get here if op is OP_LOCK */
7943 		ASSERT(resop->resop == OP_LOCK);
7944 		lres = &resop->nfs_resop4_u.oplock;
7945 		status = NFS4ERR_DENIED;
7946 		if (lock_denied(&lres->LOCK4res_u.denied, &flock)
7947 			== NFS4ERR_EXPIRED)
7948 			goto retry;
7949 		break;
7950 	case ENOLCK:
7951 		status = NFS4ERR_DELAY;
7952 		break;
7953 	case EOVERFLOW:
7954 		status = NFS4ERR_INVAL;
7955 		break;
7956 	case EINVAL:
7957 		status = NFS4ERR_NOTSUPP;
7958 		break;
7959 	default:
7960 		cmn_err(CE_WARN, "rfs4_do_lock: unexpected errno (%d)",
7961 			error);
7962 		status = NFS4ERR_SERVERFAULT;
7963 		break;
7964 	}
7965 
7966 	return (status);
7967 }
7968 
7969 /*ARGSUSED*/
7970 void
7971 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop,
7972 	    struct svc_req *req, struct compound_state *cs)
7973 {
7974 	/* XXX Currently not using req arg */
7975 	LOCK4args *args = &argop->nfs_argop4_u.oplock;
7976 	LOCK4res *resp = &resop->nfs_resop4_u.oplock;
7977 	nfsstat4 status;
7978 	stateid4 *stateid;
7979 	rfs4_lockowner_t *lo;
7980 	rfs4_client_t *cp;
7981 	rfs4_state_t *sp = NULL;
7982 	rfs4_lo_state_t *lsp = NULL;
7983 	bool_t ls_sw_held = FALSE;
7984 	bool_t create = TRUE;
7985 	bool_t lcreate = TRUE;
7986 	bool_t dup_lock = FALSE;
7987 	int rc;
7988 
7989 	if (cs->vp == NULL) {
7990 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7991 		return;
7992 	}
7993 
7994 	if (args->locker.new_lock_owner) {
7995 		/* Create a new lockowner for this instance */
7996 		open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner;
7997 
7998 		NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner"));
7999 
8000 		stateid = &olo->open_stateid;
8001 		status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID);
8002 		if (status != NFS4_OK) {
8003 			NFS4_DEBUG(rfs4_debug,
8004 				(CE_NOTE, "Get state failed in lock %d",
8005 				status));
8006 			*cs->statusp = resp->status = status;
8007 			return;
8008 		}
8009 
8010 		/* Ensure specified filehandle matches */
8011 		if (cs->vp != sp->finfo->vp) {
8012 			rfs4_state_rele(sp);
8013 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8014 			return;
8015 		}
8016 
8017 		/* hold off other access to open_owner while we tinker */
8018 		rfs4_sw_enter(&sp->owner->oo_sw);
8019 
8020 		switch (rc = rfs4_check_stateid_seqid(sp, stateid)) {
8021 		case NFS4_CHECK_STATEID_OLD:
8022 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8023 			goto end;
8024 		case NFS4_CHECK_STATEID_BAD:
8025 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8026 			goto end;
8027 		case NFS4_CHECK_STATEID_EXPIRED:
8028 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8029 			goto end;
8030 		case NFS4_CHECK_STATEID_UNCONFIRMED:
8031 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8032 			goto end;
8033 		case NFS4_CHECK_STATEID_CLOSED:
8034 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8035 			goto end;
8036 		case NFS4_CHECK_STATEID_OKAY:
8037 		case NFS4_CHECK_STATEID_REPLAY:
8038 			switch (rfs4_check_olo_seqid(olo->open_seqid,
8039 				sp->owner, resop)) {
8040 			case NFS4_CHKSEQ_OKAY:
8041 				if (rc == NFS4_CHECK_STATEID_OKAY)
8042 					break;
8043 				/*
8044 				 * This is replayed stateid; if seqid
8045 				 * matches next expected, then client
8046 				 * is using wrong seqid.
8047 				 */
8048 				/* FALLTHROUGH */
8049 			case NFS4_CHKSEQ_BAD:
8050 				*cs->statusp = resp->status =
8051 					NFS4ERR_BAD_SEQID;
8052 				goto end;
8053 			case NFS4_CHKSEQ_REPLAY:
8054 				/* This is a duplicate LOCK request */
8055 				dup_lock = TRUE;
8056 
8057 				/*
8058 				 * For a duplicate we do not want to
8059 				 * create a new lockowner as it should
8060 				 * already exist.
8061 				 * Turn off the lockowner create flag.
8062 				 */
8063 				lcreate = FALSE;
8064 			}
8065 			break;
8066 		}
8067 
8068 		lo = rfs4_findlockowner(&olo->lock_owner, &lcreate);
8069 		if (lo == NULL) {
8070 			NFS4_DEBUG(rfs4_debug,
8071 				(CE_NOTE, "rfs4_op_lock: no lock owner"));
8072 			*cs->statusp = resp->status = NFS4ERR_RESOURCE;
8073 			goto end;
8074 		}
8075 
8076 		lsp = rfs4_findlo_state_by_owner(lo, sp, &create);
8077 		if (lsp == NULL) {
8078 			rfs4_update_lease(sp->owner->client);
8079 			/*
8080 			 * Only update theh open_seqid if this is not
8081 			 * a duplicate request
8082 			 */
8083 			if (dup_lock == FALSE) {
8084 				rfs4_update_open_sequence(sp->owner);
8085 			}
8086 
8087 			NFS4_DEBUG(rfs4_debug,
8088 				(CE_NOTE, "rfs4_op_lock: no state"));
8089 			*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
8090 			rfs4_update_open_resp(sp->owner, resop, NULL);
8091 			rfs4_lockowner_rele(lo);
8092 			goto end;
8093 		}
8094 
8095 		/*
8096 		 * This is the new_lock_owner branch and the client is
8097 		 * supposed to be associating a new lock_owner with
8098 		 * the open file at this point.  If we find that a
8099 		 * lock_owner/state association already exists and a
8100 		 * successful LOCK request was returned to the client,
8101 		 * an error is returned to the client since this is
8102 		 * not appropriate.  The client should be using the
8103 		 * existing lock_owner branch.
8104 		 */
8105 		if (dup_lock == FALSE && create == FALSE) {
8106 			if (lsp->lock_completed == TRUE) {
8107 				*cs->statusp =
8108 					resp->status = NFS4ERR_BAD_SEQID;
8109 				rfs4_lockowner_rele(lo);
8110 				goto end;
8111 			}
8112 		}
8113 
8114 		rfs4_update_lease(sp->owner->client);
8115 
8116 		/*
8117 		 * Only update theh open_seqid if this is not
8118 		 * a duplicate request
8119 		 */
8120 		if (dup_lock == FALSE) {
8121 			rfs4_update_open_sequence(sp->owner);
8122 		}
8123 
8124 		/*
8125 		 * If this is a duplicate lock request, just copy the
8126 		 * previously saved reply and return.
8127 		 */
8128 		if (dup_lock == TRUE) {
8129 			/* verify that lock_seqid's match */
8130 			if (lsp->seqid != olo->lock_seqid) {
8131 				NFS4_DEBUG(rfs4_debug,
8132 				(CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad"
8133 				"lsp->seqid=%d old->seqid=%d",
8134 				lsp->seqid, olo->lock_seqid));
8135 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8136 			} else {
8137 				rfs4_copy_reply(resop, lsp->reply);
8138 				/*
8139 				 * Make sure to copy the just
8140 				 * retrieved reply status into the
8141 				 * overall compound status
8142 				 */
8143 				*cs->statusp = resp->status;
8144 			}
8145 			rfs4_lockowner_rele(lo);
8146 			goto end;
8147 		}
8148 
8149 		rfs4_dbe_lock(lsp->dbe);
8150 
8151 		/* Make sure to update the lock sequence id */
8152 		lsp->seqid = olo->lock_seqid;
8153 
8154 		NFS4_DEBUG(rfs4_debug,
8155 			(CE_NOTE, "Lock seqid established as %d", lsp->seqid));
8156 
8157 		/*
8158 		 * This is used to signify the newly created lockowner
8159 		 * stateid and its sequence number.  The checks for
8160 		 * sequence number and increment don't occur on the
8161 		 * very first lock request for a lockowner.
8162 		 */
8163 		lsp->skip_seqid_check = TRUE;
8164 
8165 		/* hold off other access to lsp while we tinker */
8166 		rfs4_sw_enter(&lsp->ls_sw);
8167 		ls_sw_held = TRUE;
8168 
8169 		rfs4_dbe_unlock(lsp->dbe);
8170 
8171 		rfs4_lockowner_rele(lo);
8172 	} else {
8173 		stateid = &args->locker.locker4_u.lock_owner.lock_stateid;
8174 		/* get lsp and hold the lock on the underlying file struct */
8175 		if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE))
8176 		    != NFS4_OK) {
8177 			*cs->statusp = resp->status = status;
8178 			return;
8179 		}
8180 		create = FALSE;	/* We didn't create lsp */
8181 
8182 		/* Ensure specified filehandle matches */
8183 		if (cs->vp != lsp->state->finfo->vp) {
8184 			rfs4_lo_state_rele(lsp, TRUE);
8185 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8186 			return;
8187 		}
8188 
8189 		/* hold off other access to lsp while we tinker */
8190 		rfs4_sw_enter(&lsp->ls_sw);
8191 		ls_sw_held = TRUE;
8192 
8193 		switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
8194 		/*
8195 		 * The stateid looks like it was okay (expected to be
8196 		 * the next one)
8197 		 */
8198 		case NFS4_CHECK_STATEID_OKAY:
8199 			/*
8200 			 * The sequence id is now checked.  Determine
8201 			 * if this is a replay or if it is in the
8202 			 * expected (next) sequence.  In the case of a
8203 			 * replay, there are two replay conditions
8204 			 * that may occur.  The first is the normal
8205 			 * condition where a LOCK is done with a
8206 			 * NFS4_OK response and the stateid is
8207 			 * updated.  That case is handled below when
8208 			 * the stateid is identified as a REPLAY.  The
8209 			 * second is the case where an error is
8210 			 * returned, like NFS4ERR_DENIED, and the
8211 			 * sequence number is updated but the stateid
8212 			 * is not updated.  This second case is dealt
8213 			 * with here.  So it may seem odd that the
8214 			 * stateid is okay but the sequence id is a
8215 			 * replay but it is okay.
8216 			 */
8217 			switch (rfs4_check_lock_seqid(
8218 				args->locker.locker4_u.lock_owner.lock_seqid,
8219 				lsp, resop)) {
8220 			case NFS4_CHKSEQ_REPLAY:
8221 				if (resp->status != NFS4_OK) {
8222 					/*
8223 					 * Here is our replay and need
8224 					 * to verify that the last
8225 					 * response was an error.
8226 					 */
8227 					*cs->statusp = resp->status;
8228 					goto end;
8229 				}
8230 				/*
8231 				 * This is done since the sequence id
8232 				 * looked like a replay but it didn't
8233 				 * pass our check so a BAD_SEQID is
8234 				 * returned as a result.
8235 				 */
8236 				/*FALLTHROUGH*/
8237 			case NFS4_CHKSEQ_BAD:
8238 				*cs->statusp = resp->status =
8239 					NFS4ERR_BAD_SEQID;
8240 				goto end;
8241 			case NFS4_CHKSEQ_OKAY:
8242 				/* Everything looks okay move ahead */
8243 				break;
8244 			}
8245 			break;
8246 		case NFS4_CHECK_STATEID_OLD:
8247 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8248 			goto end;
8249 		case NFS4_CHECK_STATEID_BAD:
8250 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8251 			goto end;
8252 		case NFS4_CHECK_STATEID_EXPIRED:
8253 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8254 			goto end;
8255 		case NFS4_CHECK_STATEID_CLOSED:
8256 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8257 			goto end;
8258 		case NFS4_CHECK_STATEID_REPLAY:
8259 			switch (rfs4_check_lock_seqid(
8260 				args->locker.locker4_u.lock_owner.lock_seqid,
8261 				lsp, resop)) {
8262 			case NFS4_CHKSEQ_OKAY:
8263 				/*
8264 				 * This is a replayed stateid; if
8265 				 * seqid matches the next expected,
8266 				 * then client is using wrong seqid.
8267 				 */
8268 			case NFS4_CHKSEQ_BAD:
8269 				*cs->statusp = resp->status =
8270 					NFS4ERR_BAD_SEQID;
8271 				goto end;
8272 			case NFS4_CHKSEQ_REPLAY:
8273 				rfs4_update_lease(lsp->locker->client);
8274 				*cs->statusp = status = resp->status;
8275 				goto end;
8276 			}
8277 			break;
8278 		default:
8279 			ASSERT(FALSE);
8280 			break;
8281 		}
8282 
8283 		rfs4_update_lock_sequence(lsp);
8284 		rfs4_update_lease(lsp->locker->client);
8285 	}
8286 
8287 	/*
8288 	 * NFS4 only allows locking on regular files, so
8289 	 * verify type of object.
8290 	 */
8291 	if (cs->vp->v_type != VREG) {
8292 		if (cs->vp->v_type == VDIR)
8293 			status = NFS4ERR_ISDIR;
8294 		else
8295 			status = NFS4ERR_INVAL;
8296 		goto out;
8297 	}
8298 
8299 	cp = lsp->state->owner->client;
8300 
8301 	if (rfs4_clnt_in_grace(cp) && !args->reclaim) {
8302 		status = NFS4ERR_GRACE;
8303 		goto out;
8304 	}
8305 
8306 	if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->can_reclaim) {
8307 		status = NFS4ERR_NO_GRACE;
8308 		goto out;
8309 	}
8310 
8311 	if (!rfs4_clnt_in_grace(cp) && args->reclaim) {
8312 		status = NFS4ERR_NO_GRACE;
8313 		goto out;
8314 	}
8315 
8316 	if (lsp->state->finfo->dinfo->dtype == OPEN_DELEGATE_WRITE)
8317 		cs->deleg = TRUE;
8318 
8319 	status = rfs4_do_lock(lsp, args->locktype,
8320 				args->locker.locker4_u.lock_owner.lock_seqid,
8321 				args->offset,
8322 				args->length, cs->cr, resop);
8323 
8324 out:
8325 	lsp->skip_seqid_check = FALSE;
8326 
8327 	*cs->statusp = resp->status = status;
8328 
8329 	if (status == NFS4_OK) {
8330 		resp->LOCK4res_u.lock_stateid = lsp->lockid.stateid;
8331 		lsp->lock_completed = TRUE;
8332 	}
8333 	/*
8334 	 * Only update the "OPEN" response here if this was a new
8335 	 * lock_owner
8336 	 */
8337 	if (sp)
8338 		rfs4_update_open_resp(sp->owner, resop, NULL);
8339 
8340 	rfs4_update_lock_resp(lsp, resop);
8341 
8342 end:
8343 	if (lsp) {
8344 		if (ls_sw_held)
8345 			rfs4_sw_exit(&lsp->ls_sw);
8346 		/*
8347 		 * If an sp obtained, then the lsp does not represent
8348 		 * a lock on the file struct.
8349 		 */
8350 		if (sp != NULL)
8351 			rfs4_lo_state_rele(lsp, FALSE);
8352 		else
8353 			rfs4_lo_state_rele(lsp, TRUE);
8354 	}
8355 	if (sp) {
8356 		rfs4_sw_exit(&sp->owner->oo_sw);
8357 		rfs4_state_rele(sp);
8358 	}
8359 }
8360 
8361 /* free function for LOCK/LOCKT */
8362 static void
8363 lock_denied_free(nfs_resop4 *resop)
8364 {
8365 	LOCK4denied *dp = NULL;
8366 
8367 	switch (resop->resop) {
8368 	case OP_LOCK:
8369 		if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED)
8370 			dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied;
8371 		break;
8372 	case OP_LOCKT:
8373 		if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED)
8374 			dp = &resop->nfs_resop4_u.oplockt.denied;
8375 		break;
8376 	default:
8377 		break;
8378 	}
8379 
8380 	if (dp)
8381 		kmem_free(dp->owner.owner_val, dp->owner.owner_len);
8382 }
8383 
8384 /*ARGSUSED*/
8385 void
8386 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop,
8387 	    struct svc_req *req, struct compound_state *cs)
8388 {
8389 	/* XXX Currently not using req arg */
8390 	LOCKU4args *args = &argop->nfs_argop4_u.oplocku;
8391 	LOCKU4res *resp = &resop->nfs_resop4_u.oplocku;
8392 	nfsstat4 status;
8393 	stateid4 *stateid = &args->lock_stateid;
8394 	rfs4_lo_state_t *lsp;
8395 
8396 	if (cs->vp == NULL) {
8397 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8398 		return;
8399 	}
8400 
8401 	if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) {
8402 		*cs->statusp = resp->status = status;
8403 		return;
8404 	}
8405 
8406 	/* Ensure specified filehandle matches */
8407 	if (cs->vp != lsp->state->finfo->vp) {
8408 		rfs4_lo_state_rele(lsp, TRUE);
8409 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8410 		return;
8411 	}
8412 
8413 	/* hold off other access to lsp while we tinker */
8414 	rfs4_sw_enter(&lsp->ls_sw);
8415 
8416 	switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
8417 	case NFS4_CHECK_STATEID_OKAY:
8418 		if (rfs4_check_lock_seqid(args->seqid, lsp, resop)
8419 		    != NFS4_CHKSEQ_OKAY) {
8420 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8421 			goto end;
8422 		}
8423 		break;
8424 	case NFS4_CHECK_STATEID_OLD:
8425 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8426 		goto end;
8427 	case NFS4_CHECK_STATEID_BAD:
8428 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8429 		goto end;
8430 	case NFS4_CHECK_STATEID_EXPIRED:
8431 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8432 		goto end;
8433 	case NFS4_CHECK_STATEID_CLOSED:
8434 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8435 		goto end;
8436 	case NFS4_CHECK_STATEID_REPLAY:
8437 		switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) {
8438 		case NFS4_CHKSEQ_OKAY:
8439 				/*
8440 				 * This is a replayed stateid; if
8441 				 * seqid matches the next expected,
8442 				 * then client is using wrong seqid.
8443 				 */
8444 		case NFS4_CHKSEQ_BAD:
8445 			*cs->statusp = resp->status =
8446 				NFS4ERR_BAD_SEQID;
8447 			goto end;
8448 		case NFS4_CHKSEQ_REPLAY:
8449 			rfs4_update_lease(lsp->locker->client);
8450 			*cs->statusp = status = resp->status;
8451 			goto end;
8452 		}
8453 		break;
8454 	default:
8455 		ASSERT(FALSE);
8456 		break;
8457 	}
8458 
8459 	rfs4_update_lock_sequence(lsp);
8460 	rfs4_update_lease(lsp->locker->client);
8461 
8462 	/*
8463 	 * NFS4 only allows locking on regular files, so
8464 	 * verify type of object.
8465 	 */
8466 	if (cs->vp->v_type != VREG) {
8467 		if (cs->vp->v_type == VDIR)
8468 			status = NFS4ERR_ISDIR;
8469 		else
8470 			status = NFS4ERR_INVAL;
8471 		goto out;
8472 	}
8473 
8474 	if (rfs4_clnt_in_grace(lsp->state->owner->client)) {
8475 		status = NFS4ERR_GRACE;
8476 		goto out;
8477 	}
8478 
8479 	status = rfs4_do_lock(lsp, args->locktype,
8480 			    args->seqid, args->offset,
8481 			    args->length, cs->cr, resop);
8482 
8483 out:
8484 	*cs->statusp = resp->status = status;
8485 
8486 	if (status == NFS4_OK)
8487 		resp->lock_stateid = lsp->lockid.stateid;
8488 
8489 	rfs4_update_lock_resp(lsp, resop);
8490 
8491 end:
8492 	rfs4_sw_exit(&lsp->ls_sw);
8493 	rfs4_lo_state_rele(lsp, TRUE);
8494 }
8495 
8496 /*
8497  * LOCKT is a best effort routine, the client can not be guaranteed that
8498  * the status return is still in effect by the time the reply is received.
8499  * They are numerous race conditions in this routine, but we are not required
8500  * and can not be accurate.
8501  */
8502 /*ARGSUSED*/
8503 void
8504 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop,
8505 	    struct svc_req *req, struct compound_state *cs)
8506 {
8507 	LOCKT4args *args = &argop->nfs_argop4_u.oplockt;
8508 	LOCKT4res *resp = &resop->nfs_resop4_u.oplockt;
8509 	rfs4_lockowner_t *lo;
8510 	rfs4_client_t *cp;
8511 	bool_t create = FALSE;
8512 	struct flock64 flk;
8513 	int error;
8514 	int flag = FREAD | FWRITE;
8515 	int ltype;
8516 	length4 posix_length;
8517 	sysid_t sysid;
8518 	pid_t pid;
8519 
8520 	if (cs->vp == NULL) {
8521 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8522 		return;
8523 	}
8524 
8525 	/*
8526 	 * NFS4 only allows locking on regular files, so
8527 	 * verify type of object.
8528 	 */
8529 	if (cs->vp->v_type != VREG) {
8530 		if (cs->vp->v_type == VDIR)
8531 			*cs->statusp = resp->status = NFS4ERR_ISDIR;
8532 		else
8533 			*cs->statusp = resp->status =  NFS4ERR_INVAL;
8534 		return;
8535 	}
8536 
8537 	/*
8538 	 * Check out the clientid to ensure the server knows about it
8539 	 * so that we correctly inform the client of a server reboot.
8540 	 */
8541 	if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE))
8542 	    == NULL) {
8543 		*cs->statusp = resp->status =
8544 			rfs4_check_clientid(&args->owner.clientid, 0);
8545 		return;
8546 	}
8547 	if (rfs4_lease_expired(cp)) {
8548 		rfs4_client_close(cp);
8549 		/*
8550 		 * Protocol doesn't allow returning NFS4ERR_STALE as
8551 		 * other operations do on this check so STALE_CLIENTID
8552 		 * is returned instead
8553 		 */
8554 		*cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
8555 		return;
8556 	}
8557 
8558 	if (rfs4_clnt_in_grace(cp)) {
8559 		*cs->statusp = resp->status = NFS4ERR_GRACE;
8560 		return;
8561 	}
8562 	rfs4_client_rele(cp);
8563 
8564 	resp->status = NFS4_OK;
8565 
8566 	switch (args->locktype) {
8567 	case READ_LT:
8568 	case READW_LT:
8569 		ltype = F_RDLCK;
8570 		break;
8571 	case WRITE_LT:
8572 	case WRITEW_LT:
8573 		ltype = F_WRLCK;
8574 		break;
8575 	}
8576 
8577 	posix_length = args->length;
8578 	/* Check for zero length. To lock to end of file use all ones for V4 */
8579 	if (posix_length == 0) {
8580 		*cs->statusp = resp->status = NFS4ERR_INVAL;
8581 		return;
8582 	} else if (posix_length == (length4)(~0)) {
8583 		posix_length = 0;	/* Posix to end of file  */
8584 	}
8585 
8586 	/* Find or create a lockowner */
8587 	lo = rfs4_findlockowner(&args->owner, &create);
8588 
8589 	if (lo) {
8590 		pid = lo->pid;
8591 		if ((resp->status =
8592 			rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK)
8593 		goto out;
8594 	} else {
8595 		pid = 0;
8596 		sysid = lockt_sysid;
8597 	}
8598 retry:
8599 	flk.l_type = ltype;
8600 	flk.l_whence = 0;		/* SEEK_SET */
8601 	flk.l_start = args->offset;
8602 	flk.l_len = posix_length;
8603 	flk.l_sysid = sysid;
8604 	flk.l_pid = pid;
8605 	flag |= F_REMOTELOCK;
8606 
8607 	LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk);
8608 
8609 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
8610 	if (flk.l_len < 0 || flk.l_start < 0) {
8611 		resp->status = NFS4ERR_INVAL;
8612 		goto out;
8613 	}
8614 	error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0,
8615 	    NULL, cs->cr);
8616 
8617 	/*
8618 	 * N.B. We map error values to nfsv4 errors. This is differrent
8619 	 * than puterrno4 routine.
8620 	 */
8621 	switch (error) {
8622 	case 0:
8623 		if (flk.l_type == F_UNLCK)
8624 			resp->status = NFS4_OK;
8625 		else {
8626 			if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED)
8627 				goto retry;
8628 			resp->status = NFS4ERR_DENIED;
8629 		}
8630 		break;
8631 	case EOVERFLOW:
8632 		resp->status = NFS4ERR_INVAL;
8633 		break;
8634 	case EINVAL:
8635 		resp->status = NFS4ERR_NOTSUPP;
8636 		break;
8637 	default:
8638 		cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)",
8639 			error);
8640 		resp->status = NFS4ERR_SERVERFAULT;
8641 		break;
8642 	}
8643 
8644 out:
8645 	if (lo)
8646 		rfs4_lockowner_rele(lo);
8647 	*cs->statusp = resp->status;
8648 }
8649 
8650 static int
8651 vop_shrlock(vnode_t *vp, int cmd, struct shrlock *sp, int fflags)
8652 {
8653 	int err;
8654 
8655 	if (cmd == F_UNSHARE && sp->s_deny == 0 && sp->s_access == 0)
8656 		return (0);
8657 
8658 	err = VOP_SHRLOCK(vp, cmd, sp, fflags, CRED());
8659 
8660 	NFS4_DEBUG(rfs4_shrlock_debug,
8661 		(CE_NOTE, "rfs4_shrlock %s vp=%p acc=%d dny=%d sysid=%d "
8662 		"pid=%d err=%d\n", cmd == F_SHARE ? "SHARE" : "UNSHR",
8663 		(void *) vp, sp->s_access, sp->s_deny, sp->s_sysid, sp->s_pid,
8664 		err));
8665 
8666 	return (err);
8667 }
8668 
8669 static int
8670 rfs4_shrlock(rfs4_state_t *sp, int cmd)
8671 {
8672 	struct shrlock shr;
8673 	struct shr_locowner shr_loco;
8674 	int fflags;
8675 
8676 	fflags = shr.s_access = shr.s_deny = 0;
8677 
8678 	if (sp->share_access & OPEN4_SHARE_ACCESS_READ) {
8679 		fflags |= FREAD;
8680 		shr.s_access |= F_RDACC;
8681 	}
8682 	if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) {
8683 		fflags |= FWRITE;
8684 		shr.s_access |= F_WRACC;
8685 	}
8686 	if (sp->share_deny & OPEN4_SHARE_DENY_READ)
8687 		shr.s_deny |= F_RDDNY;
8688 	if (sp->share_deny & OPEN4_SHARE_DENY_WRITE)
8689 		shr.s_deny |= F_WRDNY;
8690 
8691 	shr.s_pid = rfs4_dbe_getid(sp->owner->dbe);
8692 	shr.s_sysid = sp->owner->client->sysidt;
8693 	shr_loco.sl_pid = shr.s_pid;
8694 	shr_loco.sl_id = shr.s_sysid;
8695 	shr.s_owner = (caddr_t)&shr_loco;
8696 	shr.s_own_len = sizeof (shr_loco);
8697 	return (vop_shrlock(sp->finfo->vp, cmd, &shr, fflags));
8698 }
8699 
8700 static int
8701 rfs4_share(rfs4_state_t *sp)
8702 {
8703 	return (rfs4_shrlock(sp, F_SHARE));
8704 }
8705 
8706 void
8707 rfs4_unshare(rfs4_state_t *sp)
8708 {
8709 	(void) rfs4_shrlock(sp, F_UNSHARE);
8710 }
8711