xref: /titanic_50/usr/src/uts/common/fs/nfs/nfs4_srv.c (revision 582271e8d649568c83e9a016cc0d54265389c5d9)
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 2007 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/vfs_opreg.h>
40 #include <sys/vnode.h>
41 #include <sys/uio.h>
42 #include <sys/errno.h>
43 #include <sys/sysmacros.h>
44 #include <sys/statvfs.h>
45 #include <sys/kmem.h>
46 #include <sys/dirent.h>
47 #include <sys/cmn_err.h>
48 #include <sys/debug.h>
49 #include <sys/systeminfo.h>
50 #include <sys/flock.h>
51 #include <sys/pathname.h>
52 #include <sys/nbmlock.h>
53 #include <sys/share.h>
54 #include <sys/atomic.h>
55 #include <sys/policy.h>
56 #include <sys/fem.h>
57 #include <sys/sdt.h>
58 #include <sys/ddi.h>
59 
60 #include <rpc/types.h>
61 #include <rpc/auth.h>
62 #include <rpc/rpcsec_gss.h>
63 #include <rpc/svc.h>
64 
65 #include <nfs/nfs.h>
66 #include <nfs/export.h>
67 #include <nfs/lm.h>
68 #include <nfs/nfs4.h>
69 
70 #include <sys/strsubr.h>
71 #include <sys/strsun.h>
72 
73 #include <inet/common.h>
74 #include <inet/ip.h>
75 #include <inet/ip6.h>
76 
77 #include <sys/tsol/label.h>
78 #include <sys/tsol/tndb.h>
79 
80 #define	RFS4_MAXLOCK_TRIES 4	/* Try to get the lock this many times */
81 static int rfs4_maxlock_tries = RFS4_MAXLOCK_TRIES;
82 #define	RFS4_LOCK_DELAY 10	/* Milliseconds */
83 static clock_t rfs4_lock_delay = RFS4_LOCK_DELAY;
84 
85 /* End of Tunables */
86 
87 /*
88  * Used to bump the stateid4.seqid value and show changes in the stateid
89  */
90 #define	next_stateid(sp) (++(sp)->bits.chgseq)
91 
92 /*
93  * RFS4_MINLEN_ENTRY4: XDR-encoded size of smallest possible dirent.
94  *	This is used to return NFS4ERR_TOOSMALL when clients specify
95  *	maxcount that isn't large enough to hold the smallest possible
96  *	XDR encoded dirent.
97  *
98  *	    sizeof cookie (8 bytes) +
99  *	    sizeof name_len (4 bytes) +
100  *	    sizeof smallest (padded) name (4 bytes) +
101  *	    sizeof bitmap4_len (12 bytes) +   NOTE: we always encode len=2 bm4
102  *	    sizeof attrlist4_len (4 bytes) +
103  *	    sizeof next boolean (4 bytes)
104  *
105  * RFS4_MINLEN_RDDIR4: XDR-encoded size of READDIR op reply containing
106  * the smallest possible entry4 (assumes no attrs requested).
107  *	sizeof nfsstat4 (4 bytes) +
108  *	sizeof verifier4 (8 bytes) +
109  *	sizeof entry4list bool (4 bytes) +
110  *	sizeof entry4 	(36 bytes) +
111  *	sizeof eof bool  (4 bytes)
112  *
113  * RFS4_MINLEN_RDDIR_BUF: minimum length of buffer server will provide to
114  *	VOP_READDIR.  Its value is the size of the maximum possible dirent
115  *	for solaris.  The DIRENT64_RECLEN macro returns	the size of dirent
116  *	required for a given name length.  MAXNAMELEN is the maximum
117  *	filename length allowed in Solaris.  The first two DIRENT64_RECLEN()
118  *	macros are to allow for . and .. entries -- just a minor tweak to try
119  *	and guarantee that buffer we give to VOP_READDIR will be large enough
120  *	to hold ., .., and the largest possible solaris dirent64.
121  */
122 #define	RFS4_MINLEN_ENTRY4 36
123 #define	RFS4_MINLEN_RDDIR4 (4 + NFS4_VERIFIER_SIZE + 4 + RFS4_MINLEN_ENTRY4 + 4)
124 #define	RFS4_MINLEN_RDDIR_BUF \
125 	(DIRENT64_RECLEN(1) + DIRENT64_RECLEN(2) + DIRENT64_RECLEN(MAXNAMELEN))
126 
127 /*
128  * It would be better to pad to 4 bytes since that's what XDR would do,
129  * but the dirents UFS gives us are already padded to 8, so just take
130  * what we're given.  Dircount is only a hint anyway.  Currently the
131  * solaris kernel is ASCII only, so there's no point in calling the
132  * UTF8 functions.
133  *
134  * dirent64: named padded to provide 8 byte struct alignment
135  *	d_ino(8) + d_off(8) + d_reclen(2) + d_name(namelen + null(1) + pad)
136  *
137  * cookie: uint64_t   +  utf8namelen: uint_t  +   utf8name padded to 8 bytes
138  *
139  */
140 #define	DIRENT64_TO_DIRCOUNT(dp) \
141 	(3 * BYTES_PER_XDR_UNIT + DIRENT64_NAMELEN((dp)->d_reclen))
142 
143 time_t rfs4_start_time;			/* Initialized in rfs4_srvrinit */
144 
145 static sysid_t lockt_sysid;		/* dummy sysid for all LOCKT calls */
146 
147 u_longlong_t nfs4_srv_caller_id;
148 uint_t nfs4_srv_vkey = 0;
149 
150 verifier4	Write4verf;
151 verifier4	Readdir4verf;
152 
153 void		rfs4_init_compound_state(struct compound_state *);
154 
155 static void	nullfree(caddr_t);
156 static void	rfs4_op_inval(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
157 			struct compound_state *);
158 static void	rfs4_op_access(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
159 			struct compound_state *);
160 static void	rfs4_op_close(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
161 			struct compound_state *);
162 static void	rfs4_op_commit(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
163 			struct compound_state *);
164 static void	rfs4_op_create(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
165 			struct compound_state *);
166 static void	rfs4_op_create_free(nfs_resop4 *resop);
167 static void	rfs4_op_delegreturn(nfs_argop4 *, nfs_resop4 *,
168 				    struct svc_req *, struct compound_state *);
169 static void	rfs4_op_getattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
170 			struct compound_state *);
171 static void	rfs4_op_getattr_free(nfs_resop4 *);
172 static void	rfs4_op_getfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
173 			struct compound_state *);
174 static void	rfs4_op_getfh_free(nfs_resop4 *);
175 static void	rfs4_op_illegal(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
176 			struct compound_state *);
177 static void	rfs4_op_link(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
178 			struct compound_state *);
179 static void	rfs4_op_lock(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
180 			struct compound_state *);
181 static void	lock_denied_free(nfs_resop4 *);
182 static void	rfs4_op_locku(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
183 			struct compound_state *);
184 static void	rfs4_op_lockt(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
185 			struct compound_state *);
186 static void	rfs4_op_lookup(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
187 			struct compound_state *);
188 static void	rfs4_op_lookupp(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
189 			struct compound_state *);
190 static void	rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop,
191 				struct svc_req *req, struct compound_state *cs);
192 static void	rfs4_op_nverify(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
193 			struct compound_state *);
194 static void	rfs4_op_open(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
195 			struct compound_state *);
196 static void	rfs4_op_open_confirm(nfs_argop4 *, nfs_resop4 *,
197 			struct svc_req *, struct compound_state *);
198 static void	rfs4_op_open_downgrade(nfs_argop4 *, nfs_resop4 *,
199 			struct svc_req *, struct compound_state *);
200 static void	rfs4_op_putfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
201 			struct compound_state *);
202 static void	rfs4_op_putpubfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
203 			struct compound_state *);
204 static void	rfs4_op_putrootfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
205 			struct compound_state *);
206 static void	rfs4_op_read(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
207 			struct compound_state *);
208 static void	rfs4_op_read_free(nfs_resop4 *);
209 static void	rfs4_op_readdir_free(nfs_resop4 *resop);
210 static void	rfs4_op_readlink(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
211 			struct compound_state *);
212 static void	rfs4_op_readlink_free(nfs_resop4 *);
213 static void	rfs4_op_release_lockowner(nfs_argop4 *, nfs_resop4 *,
214 			struct svc_req *, struct compound_state *);
215 static void	rfs4_op_remove(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
216 			struct compound_state *);
217 static void	rfs4_op_rename(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
218 			struct compound_state *);
219 static void	rfs4_op_renew(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
220 			struct compound_state *);
221 static void	rfs4_op_restorefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
222 			struct compound_state *);
223 static void	rfs4_op_savefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
224 			struct compound_state *);
225 static void	rfs4_op_setattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
226 			struct compound_state *);
227 static void	rfs4_op_verify(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
228 			struct compound_state *);
229 static void	rfs4_op_write(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
230 			struct compound_state *);
231 static void	rfs4_op_setclientid(nfs_argop4 *, nfs_resop4 *,
232 			struct svc_req *, struct compound_state *);
233 static void	rfs4_op_setclientid_confirm(nfs_argop4 *, nfs_resop4 *,
234 			struct svc_req *req, struct compound_state *);
235 static void	rfs4_op_secinfo(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
236 			struct compound_state *);
237 static void	rfs4_op_secinfo_free(nfs_resop4 *);
238 
239 static nfsstat4 check_open_access(uint32_t,
240 				struct compound_state *, struct svc_req *);
241 nfsstat4 rfs4_client_sysid(rfs4_client_t *, sysid_t *);
242 static int	vop_shrlock(vnode_t *, int, struct shrlock *, int);
243 static int 	rfs4_shrlock(rfs4_state_t *, int);
244 static int	rfs4_share(rfs4_state_t *);
245 void rfs4_ss_clid(rfs4_client_t *, struct svc_req *);
246 
247 /*
248  * translation table for attrs
249  */
250 struct nfs4_ntov_table {
251 	union nfs4_attr_u *na;
252 	uint8_t amap[NFS4_MAXNUM_ATTRS];
253 	int attrcnt;
254 	bool_t vfsstat;
255 };
256 
257 static void	nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp);
258 static void	nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp,
259 				    struct nfs4_svgetit_arg *sargp);
260 
261 static nfsstat4	do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp,
262 		    struct compound_state *cs, struct nfs4_svgetit_arg *sargp,
263 		    struct nfs4_ntov_table *ntovp, nfs4_attr_cmd_t cmd);
264 
265 fem_t	*deleg_rdops;
266 fem_t	*deleg_wrops;
267 
268 rfs4_servinst_t	*rfs4_cur_servinst = NULL;	/* current server instance */
269 kmutex_t	rfs4_servinst_lock;		/* protects linked list */
270 int		rfs4_seen_first_compound;	/* set first time we see one */
271 
272 /*
273  * NFS4 op dispatch table
274  */
275 
276 struct rfsv4disp {
277 	void	(*dis_proc)();		/* proc to call */
278 	void	(*dis_resfree)();	/* frees space allocated by proc */
279 	int	dis_flags;		/* RPC_IDEMPOTENT, etc... */
280 };
281 
282 static struct rfsv4disp rfsv4disptab[] = {
283 	/*
284 	 * NFS VERSION 4
285 	 */
286 
287 	/* RFS_NULL = 0 */
288 	{rfs4_op_illegal, nullfree, 0},
289 
290 	/* UNUSED = 1 */
291 	{rfs4_op_illegal, nullfree, 0},
292 
293 	/* UNUSED = 2 */
294 	{rfs4_op_illegal, nullfree, 0},
295 
296 	/* OP_ACCESS = 3 */
297 	{rfs4_op_access, nullfree, RPC_IDEMPOTENT},
298 
299 	/* OP_CLOSE = 4 */
300 	{rfs4_op_close, nullfree, 0},
301 
302 	/* OP_COMMIT = 5 */
303 	{rfs4_op_commit, nullfree, RPC_IDEMPOTENT},
304 
305 	/* OP_CREATE = 6 */
306 	{rfs4_op_create, nullfree, 0},
307 
308 	/* OP_DELEGPURGE = 7 */
309 	{rfs4_op_inval, nullfree, 0},
310 
311 	/* OP_DELEGRETURN = 8 */
312 	{rfs4_op_delegreturn, nullfree, 0},
313 
314 	/* OP_GETATTR = 9 */
315 	{rfs4_op_getattr, rfs4_op_getattr_free, RPC_IDEMPOTENT},
316 
317 	/* OP_GETFH = 10 */
318 	{rfs4_op_getfh, rfs4_op_getfh_free, RPC_ALL},
319 
320 	/* OP_LINK = 11 */
321 	{rfs4_op_link, nullfree, 0},
322 
323 	/* OP_LOCK = 12 */
324 	{rfs4_op_lock, lock_denied_free, 0},
325 
326 	/* OP_LOCKT = 13 */
327 	{rfs4_op_lockt, lock_denied_free, 0},
328 
329 	/* OP_LOCKU = 14 */
330 	{rfs4_op_locku, nullfree, 0},
331 
332 	/* OP_LOOKUP = 15 */
333 	{rfs4_op_lookup, nullfree, (RPC_IDEMPOTENT|RPC_PUBLICFH_OK)},
334 
335 	/* OP_LOOKUPP = 16 */
336 	{rfs4_op_lookupp, nullfree, (RPC_IDEMPOTENT|RPC_PUBLICFH_OK)},
337 
338 	/* OP_NVERIFY = 17 */
339 	{rfs4_op_nverify, nullfree, RPC_IDEMPOTENT},
340 
341 	/* OP_OPEN = 18 */
342 	{rfs4_op_open, rfs4_free_reply, 0},
343 
344 	/* OP_OPENATTR = 19 */
345 	{rfs4_op_openattr, nullfree, 0},
346 
347 	/* OP_OPEN_CONFIRM = 20 */
348 	{rfs4_op_open_confirm, nullfree, 0},
349 
350 	/* OP_OPEN_DOWNGRADE = 21 */
351 	{rfs4_op_open_downgrade, nullfree, 0},
352 
353 	/* OP_OPEN_PUTFH = 22 */
354 	{rfs4_op_putfh, nullfree, RPC_ALL},
355 
356 	/* OP_PUTPUBFH = 23 */
357 	{rfs4_op_putpubfh, nullfree, RPC_ALL},
358 
359 	/* OP_PUTROOTFH = 24 */
360 	{rfs4_op_putrootfh, nullfree, RPC_ALL},
361 
362 	/* OP_READ = 25 */
363 	{rfs4_op_read, rfs4_op_read_free, RPC_IDEMPOTENT},
364 
365 	/* OP_READDIR = 26 */
366 	{rfs4_op_readdir, rfs4_op_readdir_free, RPC_IDEMPOTENT},
367 
368 	/* OP_READLINK = 27 */
369 	{rfs4_op_readlink, rfs4_op_readlink_free, RPC_IDEMPOTENT},
370 
371 	/* OP_REMOVE = 28 */
372 	{rfs4_op_remove, nullfree, 0},
373 
374 	/* OP_RENAME = 29 */
375 	{rfs4_op_rename, nullfree, 0},
376 
377 	/* OP_RENEW = 30 */
378 	{rfs4_op_renew, nullfree, 0},
379 
380 	/* OP_RESTOREFH = 31 */
381 	{rfs4_op_restorefh, nullfree, RPC_ALL},
382 
383 	/* OP_SAVEFH = 32 */
384 	{rfs4_op_savefh, nullfree, RPC_ALL},
385 
386 	/* OP_SECINFO = 33 */
387 	{rfs4_op_secinfo, rfs4_op_secinfo_free, 0},
388 
389 	/* OP_SETATTR = 34 */
390 	{rfs4_op_setattr, nullfree, 0},
391 
392 	/* OP_SETCLIENTID = 35 */
393 	{rfs4_op_setclientid, nullfree, 0},
394 
395 	/* OP_SETCLIENTID_CONFIRM = 36 */
396 	{rfs4_op_setclientid_confirm, nullfree, 0},
397 
398 	/* OP_VERIFY = 37 */
399 	{rfs4_op_verify, nullfree, RPC_IDEMPOTENT},
400 
401 	/* OP_WRITE = 38 */
402 	{rfs4_op_write, nullfree, 0},
403 
404 	/* OP_RELEASE_LOCKOWNER = 39 */
405 	{rfs4_op_release_lockowner, nullfree, 0},
406 };
407 
408 static uint_t rfsv4disp_cnt = sizeof (rfsv4disptab) / sizeof (rfsv4disptab[0]);
409 
410 #define	OP_ILLEGAL_IDX (rfsv4disp_cnt)
411 
412 #ifdef DEBUG
413 
414 int rfs4_fillone_debug = 0;
415 int rfs4_shrlock_debug = 0;
416 int rfs4_no_stub_access = 1;
417 int rfs4_rddir_debug = 0;
418 
419 static char *rfs4_op_string[] = {
420 	"rfs4_op_null",
421 	"rfs4_op_1 unused",
422 	"rfs4_op_2 unused",
423 	"rfs4_op_access",
424 	"rfs4_op_close",
425 	"rfs4_op_commit",
426 	"rfs4_op_create",
427 	"rfs4_op_delegpurge",
428 	"rfs4_op_delegreturn",
429 	"rfs4_op_getattr",
430 	"rfs4_op_getfh",
431 	"rfs4_op_link",
432 	"rfs4_op_lock",
433 	"rfs4_op_lockt",
434 	"rfs4_op_locku",
435 	"rfs4_op_lookup",
436 	"rfs4_op_lookupp",
437 	"rfs4_op_nverify",
438 	"rfs4_op_open",
439 	"rfs4_op_openattr",
440 	"rfs4_op_open_confirm",
441 	"rfs4_op_open_downgrade",
442 	"rfs4_op_putfh",
443 	"rfs4_op_putpubfh",
444 	"rfs4_op_putrootfh",
445 	"rfs4_op_read",
446 	"rfs4_op_readdir",
447 	"rfs4_op_readlink",
448 	"rfs4_op_remove",
449 	"rfs4_op_rename",
450 	"rfs4_op_renew",
451 	"rfs4_op_restorefh",
452 	"rfs4_op_savefh",
453 	"rfs4_op_secinfo",
454 	"rfs4_op_setattr",
455 	"rfs4_op_setclientid",
456 	"rfs4_op_setclient_confirm",
457 	"rfs4_op_verify",
458 	"rfs4_op_write",
459 	"rfs4_op_release_lockowner",
460 	"rfs4_op_illegal"
461 };
462 #endif
463 
464 void rfs4_ss_chkclid(rfs4_client_t *);
465 
466 extern size_t strlcpy(char *dst, const char *src, size_t dstsize);
467 
468 #ifdef	nextdp
469 #undef nextdp
470 #endif
471 #define	nextdp(dp)	((struct dirent64 *)((char *)(dp) + (dp)->d_reclen))
472 
473 static const fs_operation_def_t nfs4_rd_deleg_tmpl[] = {
474 	VOPNAME_OPEN,		{ .femop_open = deleg_rdopen },
475 	VOPNAME_WRITE,		{ .femop_write = deleg_write },
476 	VOPNAME_SETATTR,	{ .femop_setattr = deleg_setattr },
477 	VOPNAME_RWLOCK,		{ .femop_rwlock = deleg_rd_rwlock },
478 	VOPNAME_SPACE,		{ .femop_space = deleg_space },
479 	VOPNAME_SETSECATTR,	{ .femop_setsecattr = deleg_setsecattr },
480 	VOPNAME_VNEVENT,	{ .femop_vnevent = deleg_vnevent },
481 	NULL,			NULL
482 };
483 static const fs_operation_def_t nfs4_wr_deleg_tmpl[] = {
484 	VOPNAME_OPEN,		{ .femop_open = deleg_wropen },
485 	VOPNAME_READ,		{ .femop_read = deleg_read },
486 	VOPNAME_WRITE,		{ .femop_write = deleg_write },
487 	VOPNAME_SETATTR,	{ .femop_setattr = deleg_setattr },
488 	VOPNAME_RWLOCK,		{ .femop_rwlock = deleg_wr_rwlock },
489 	VOPNAME_SPACE,		{ .femop_space = deleg_space },
490 	VOPNAME_SETSECATTR,	{ .femop_setsecattr = deleg_setsecattr },
491 	VOPNAME_VNEVENT,	{ .femop_vnevent = deleg_vnevent },
492 	NULL,			NULL
493 };
494 
495 int
496 rfs4_srvrinit(void)
497 {
498 	timespec32_t verf;
499 	int error;
500 	extern void rfs4_attr_init();
501 	extern krwlock_t rfs4_deleg_policy_lock;
502 
503 	/*
504 	 * The following algorithm attempts to find a unique verifier
505 	 * to be used as the write verifier returned from the server
506 	 * to the client.  It is important that this verifier change
507 	 * whenever the server reboots.  Of secondary importance, it
508 	 * is important for the verifier to be unique between two
509 	 * different servers.
510 	 *
511 	 * Thus, an attempt is made to use the system hostid and the
512 	 * current time in seconds when the nfssrv kernel module is
513 	 * loaded.  It is assumed that an NFS server will not be able
514 	 * to boot and then to reboot in less than a second.  If the
515 	 * hostid has not been set, then the current high resolution
516 	 * time is used.  This will ensure different verifiers each
517 	 * time the server reboots and minimize the chances that two
518 	 * different servers will have the same verifier.
519 	 * XXX - this is broken on LP64 kernels.
520 	 */
521 	verf.tv_sec = (time_t)nfs_atoi(hw_serial);
522 	if (verf.tv_sec != 0) {
523 		verf.tv_nsec = gethrestime_sec();
524 	} else {
525 		timespec_t tverf;
526 
527 		gethrestime(&tverf);
528 		verf.tv_sec = (time_t)tverf.tv_sec;
529 		verf.tv_nsec = tverf.tv_nsec;
530 	}
531 
532 	Write4verf = *(uint64_t *)&verf;
533 
534 	rfs4_attr_init();
535 	mutex_init(&rfs4_deleg_lock, NULL, MUTEX_DEFAULT, NULL);
536 
537 	/* Used to manage create/destroy of server state */
538 	mutex_init(&rfs4_state_lock, NULL, MUTEX_DEFAULT, NULL);
539 
540 	/* Used to manage access to server instance linked list */
541 	mutex_init(&rfs4_servinst_lock, NULL, MUTEX_DEFAULT, NULL);
542 
543 	/* Used to manage access to rfs4_deleg_policy */
544 	rw_init(&rfs4_deleg_policy_lock, NULL, RW_DEFAULT, NULL);
545 
546 	error = fem_create("deleg_rdops", nfs4_rd_deleg_tmpl, &deleg_rdops);
547 	if (error != 0) {
548 		rfs4_disable_delegation();
549 	} else {
550 		error = fem_create("deleg_wrops", nfs4_wr_deleg_tmpl,
551 		    &deleg_wrops);
552 		if (error != 0) {
553 			rfs4_disable_delegation();
554 			fem_free(deleg_rdops);
555 		}
556 	}
557 
558 	nfs4_srv_caller_id = fs_new_caller_id();
559 
560 	lockt_sysid = lm_alloc_sysidt();
561 
562 	vsd_create(&nfs4_srv_vkey, NULL);
563 
564 	return (0);
565 }
566 
567 void
568 rfs4_srvrfini(void)
569 {
570 	extern krwlock_t rfs4_deleg_policy_lock;
571 
572 	if (lockt_sysid != LM_NOSYSID) {
573 		lm_free_sysidt(lockt_sysid);
574 		lockt_sysid = LM_NOSYSID;
575 	}
576 
577 	mutex_destroy(&rfs4_deleg_lock);
578 	mutex_destroy(&rfs4_state_lock);
579 	rw_destroy(&rfs4_deleg_policy_lock);
580 
581 	fem_free(deleg_rdops);
582 	fem_free(deleg_wrops);
583 }
584 
585 void
586 rfs4_init_compound_state(struct compound_state *cs)
587 {
588 	bzero(cs, sizeof (*cs));
589 	cs->cont = TRUE;
590 	cs->access = CS_ACCESS_DENIED;
591 	cs->deleg = FALSE;
592 	cs->mandlock = FALSE;
593 	cs->fh.nfs_fh4_val = cs->fhbuf;
594 }
595 
596 void
597 rfs4_grace_start(rfs4_servinst_t *sip)
598 {
599 	time_t now = gethrestime_sec();
600 
601 	rw_enter(&sip->rwlock, RW_WRITER);
602 	sip->start_time = now;
603 	sip->grace_period = rfs4_grace_period;
604 	rw_exit(&sip->rwlock);
605 }
606 
607 /*
608  * returns true if the instance's grace period has never been started
609  */
610 int
611 rfs4_servinst_grace_new(rfs4_servinst_t *sip)
612 {
613 	time_t start_time;
614 
615 	rw_enter(&sip->rwlock, RW_READER);
616 	start_time = sip->start_time;
617 	rw_exit(&sip->rwlock);
618 
619 	return (start_time == 0);
620 }
621 
622 /*
623  * Indicates if server instance is within the
624  * grace period.
625  */
626 int
627 rfs4_servinst_in_grace(rfs4_servinst_t *sip)
628 {
629 	time_t grace_expiry;
630 
631 	rw_enter(&sip->rwlock, RW_READER);
632 	grace_expiry = sip->start_time + sip->grace_period;
633 	rw_exit(&sip->rwlock);
634 
635 	return (gethrestime_sec() < grace_expiry);
636 }
637 
638 int
639 rfs4_clnt_in_grace(rfs4_client_t *cp)
640 {
641 	ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0);
642 
643 	return (rfs4_servinst_in_grace(cp->server_instance));
644 }
645 
646 /*
647  * reset all currently active grace periods
648  */
649 void
650 rfs4_grace_reset_all(void)
651 {
652 	rfs4_servinst_t *sip;
653 
654 	mutex_enter(&rfs4_servinst_lock);
655 	for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev)
656 		if (rfs4_servinst_in_grace(sip))
657 			rfs4_grace_start(sip);
658 	mutex_exit(&rfs4_servinst_lock);
659 }
660 
661 /*
662  * start any new instances' grace periods
663  */
664 void
665 rfs4_grace_start_new(void)
666 {
667 	rfs4_servinst_t *sip;
668 
669 	mutex_enter(&rfs4_servinst_lock);
670 	for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev)
671 		if (rfs4_servinst_grace_new(sip))
672 			rfs4_grace_start(sip);
673 	mutex_exit(&rfs4_servinst_lock);
674 }
675 
676 static rfs4_dss_path_t *
677 rfs4_dss_newpath(rfs4_servinst_t *sip, char *path, unsigned index)
678 {
679 	size_t len;
680 	rfs4_dss_path_t *dss_path;
681 
682 	dss_path = kmem_alloc(sizeof (rfs4_dss_path_t), KM_SLEEP);
683 
684 	/*
685 	 * Take a copy of the string, since the original may be overwritten.
686 	 * Sadly, no strdup() in the kernel.
687 	 */
688 	/* allow for NUL */
689 	len = strlen(path) + 1;
690 	dss_path->path = kmem_alloc(len, KM_SLEEP);
691 	(void) strlcpy(dss_path->path, path, len);
692 
693 	/* associate with servinst */
694 	dss_path->sip = sip;
695 	dss_path->index = index;
696 
697 	/*
698 	 * Add to list of served paths.
699 	 * No locking required, as we're only ever called at startup.
700 	 */
701 	if (rfs4_dss_pathlist == NULL) {
702 		/* this is the first dss_path_t */
703 
704 		/* needed for insque/remque */
705 		dss_path->next = dss_path->prev = dss_path;
706 
707 		rfs4_dss_pathlist = dss_path;
708 	} else {
709 		insque(dss_path, rfs4_dss_pathlist);
710 	}
711 
712 	return (dss_path);
713 }
714 
715 /*
716  * Create a new server instance, and make it the currently active instance.
717  * Note that starting the grace period too early will reduce the clients'
718  * recovery window.
719  */
720 void
721 rfs4_servinst_create(int start_grace, int dss_npaths, char **dss_paths)
722 {
723 	unsigned i;
724 	rfs4_servinst_t *sip;
725 	rfs4_oldstate_t *oldstate;
726 
727 	sip = kmem_alloc(sizeof (rfs4_servinst_t), KM_SLEEP);
728 	rw_init(&sip->rwlock, NULL, RW_DEFAULT, NULL);
729 
730 	sip->start_time = (time_t)0;
731 	sip->grace_period = (time_t)0;
732 	sip->next = NULL;
733 	sip->prev = NULL;
734 
735 	rw_init(&sip->oldstate_lock, NULL, RW_DEFAULT, NULL);
736 	/*
737 	 * This initial dummy entry is required to setup for insque/remque.
738 	 * It must be skipped over whenever the list is traversed.
739 	 */
740 	oldstate = kmem_alloc(sizeof (rfs4_oldstate_t), KM_SLEEP);
741 	/* insque/remque require initial list entry to be self-terminated */
742 	oldstate->next = oldstate;
743 	oldstate->prev = oldstate;
744 	sip->oldstate = oldstate;
745 
746 
747 	sip->dss_npaths = dss_npaths;
748 	sip->dss_paths = kmem_alloc(dss_npaths *
749 	    sizeof (rfs4_dss_path_t *), KM_SLEEP);
750 
751 	for (i = 0; i < dss_npaths; i++) {
752 		sip->dss_paths[i] = rfs4_dss_newpath(sip, dss_paths[i], i);
753 	}
754 
755 	mutex_enter(&rfs4_servinst_lock);
756 	if (rfs4_cur_servinst != NULL) {
757 		/* add to linked list */
758 		sip->prev = rfs4_cur_servinst;
759 		rfs4_cur_servinst->next = sip;
760 	}
761 	if (start_grace)
762 		rfs4_grace_start(sip);
763 	/* make the new instance "current" */
764 	rfs4_cur_servinst = sip;
765 
766 	mutex_exit(&rfs4_servinst_lock);
767 }
768 
769 /*
770  * In future, we might add a rfs4_servinst_destroy(sip) but, for now, destroy
771  * all instances directly.
772  */
773 void
774 rfs4_servinst_destroy_all(void)
775 {
776 	rfs4_servinst_t *sip, *prev, *current;
777 #ifdef DEBUG
778 	int n = 0;
779 #endif
780 
781 	mutex_enter(&rfs4_servinst_lock);
782 	ASSERT(rfs4_cur_servinst != NULL);
783 	current = rfs4_cur_servinst;
784 	rfs4_cur_servinst = NULL;
785 	for (sip = current; sip != NULL; sip = prev) {
786 		prev = sip->prev;
787 		rw_destroy(&sip->rwlock);
788 		if (sip->oldstate)
789 			kmem_free(sip->oldstate, sizeof (rfs4_oldstate_t));
790 		if (sip->dss_paths)
791 			kmem_free(sip->dss_paths,
792 			    sip->dss_npaths * sizeof (rfs4_dss_path_t *));
793 		kmem_free(sip, sizeof (rfs4_servinst_t));
794 #ifdef DEBUG
795 		n++;
796 #endif
797 	}
798 	mutex_exit(&rfs4_servinst_lock);
799 }
800 
801 /*
802  * Assign the current server instance to a client_t.
803  * Should be called with cp->dbe held.
804  */
805 void
806 rfs4_servinst_assign(rfs4_client_t *cp, rfs4_servinst_t *sip)
807 {
808 	ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0);
809 
810 	/*
811 	 * The lock ensures that if the current instance is in the process
812 	 * of changing, we will see the new one.
813 	 */
814 	mutex_enter(&rfs4_servinst_lock);
815 	cp->server_instance = sip;
816 	mutex_exit(&rfs4_servinst_lock);
817 }
818 
819 rfs4_servinst_t *
820 rfs4_servinst(rfs4_client_t *cp)
821 {
822 	ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0);
823 
824 	return (cp->server_instance);
825 }
826 
827 /* ARGSUSED */
828 static void
829 nullfree(caddr_t resop)
830 {
831 }
832 
833 /*
834  * This is a fall-through for invalid or not implemented (yet) ops
835  */
836 /* ARGSUSED */
837 static void
838 rfs4_op_inval(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
839 	struct compound_state *cs)
840 {
841 	*cs->statusp = *((nfsstat4 *)&(resop)->nfs_resop4_u) = NFS4ERR_INVAL;
842 }
843 
844 /*
845  * Check if the security flavor, nfsnum, is in the flavor_list.
846  */
847 bool_t
848 in_flavor_list(int nfsnum, int *flavor_list, int count)
849 {
850 	int i;
851 
852 	for (i = 0; i < count; i++) {
853 		if (nfsnum == flavor_list[i])
854 			return (TRUE);
855 	}
856 	return (FALSE);
857 }
858 
859 /*
860  * Used by rfs4_op_secinfo to get the security information from the
861  * export structure associated with the component.
862  */
863 /* ARGSUSED */
864 static nfsstat4
865 do_rfs4_op_secinfo(struct compound_state *cs, char *nm, SECINFO4res *resp)
866 {
867 	int error, different_export = 0;
868 	vnode_t *dvp, *vp, *tvp;
869 	struct exportinfo *exi = NULL;
870 	fid_t fid;
871 	uint_t count, i;
872 	secinfo4 *resok_val;
873 	struct secinfo *secp;
874 	seconfig_t *si;
875 	bool_t did_traverse;
876 	int dotdot, walk;
877 
878 	dvp = cs->vp;
879 	dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0');
880 
881 	/*
882 	 * If dotdotting, then need to check whether it's above the
883 	 * root of a filesystem, or above an export point.
884 	 */
885 	if (dotdot) {
886 
887 		/*
888 		 * If dotdotting at the root of a filesystem, then
889 		 * need to traverse back to the mounted-on filesystem
890 		 * and do the dotdot lookup there.
891 		 */
892 		if (cs->vp->v_flag & VROOT) {
893 
894 			/*
895 			 * If at the system root, then can
896 			 * go up no further.
897 			 */
898 			if (VN_CMP(dvp, rootdir))
899 				return (puterrno4(ENOENT));
900 
901 			/*
902 			 * Traverse back to the mounted-on filesystem
903 			 */
904 			dvp = untraverse(cs->vp);
905 
906 			/*
907 			 * Set the different_export flag so we remember
908 			 * to pick up a new exportinfo entry for
909 			 * this new filesystem.
910 			 */
911 			different_export = 1;
912 		} else {
913 
914 			/*
915 			 * If dotdotting above an export point then set
916 			 * the different_export to get new export info.
917 			 */
918 			different_export = nfs_exported(cs->exi, cs->vp);
919 		}
920 	}
921 
922 	/*
923 	 * Get the vnode for the component "nm".
924 	 */
925 	error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cs->cr,
926 	    NULL, NULL, NULL);
927 	if (error)
928 		return (puterrno4(error));
929 
930 	/*
931 	 * If the vnode is in a pseudo filesystem, or if the security flavor
932 	 * used in the request is valid but not an explicitly shared flavor,
933 	 * or the access bit indicates that this is a limited access,
934 	 * check whether this vnode is visible.
935 	 */
936 	if (!different_export &&
937 	    (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) ||
938 	    cs->access & CS_ACCESS_LIMITED)) {
939 		if (! nfs_visible(cs->exi, vp, &different_export)) {
940 			VN_RELE(vp);
941 			return (puterrno4(ENOENT));
942 		}
943 	}
944 
945 	/*
946 	 * If it's a mountpoint, then traverse it.
947 	 */
948 	if (vn_ismntpt(vp)) {
949 		tvp = vp;
950 		if ((error = traverse(&tvp)) != 0) {
951 			VN_RELE(vp);
952 			return (puterrno4(error));
953 		}
954 		/* remember that we had to traverse mountpoint */
955 		did_traverse = TRUE;
956 		vp = tvp;
957 		different_export = 1;
958 	} else if (vp->v_vfsp != dvp->v_vfsp) {
959 		/*
960 		 * If vp isn't a mountpoint and the vfs ptrs aren't the same,
961 		 * then vp is probably an LOFS object.  We don't need the
962 		 * realvp, we just need to know that we might have crossed
963 		 * a server fs boundary and need to call checkexport4.
964 		 * (LOFS lookup hides server fs mountpoints, and actually calls
965 		 * traverse)
966 		 */
967 		different_export = 1;
968 		did_traverse = FALSE;
969 	}
970 
971 	/*
972 	 * Get the export information for it.
973 	 */
974 	if (different_export) {
975 
976 		bzero(&fid, sizeof (fid));
977 		fid.fid_len = MAXFIDSZ;
978 		error = vop_fid_pseudo(vp, &fid);
979 		if (error) {
980 			VN_RELE(vp);
981 			return (puterrno4(error));
982 		}
983 
984 		if (dotdot)
985 			exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE);
986 		else
987 			exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
988 
989 		if (exi == NULL) {
990 			if (did_traverse == TRUE) {
991 				/*
992 				 * If this vnode is a mounted-on vnode,
993 				 * but the mounted-on file system is not
994 				 * exported, send back the secinfo for
995 				 * the exported node that the mounted-on
996 				 * vnode lives in.
997 				 */
998 				exi = cs->exi;
999 			} else {
1000 				VN_RELE(vp);
1001 				return (puterrno4(EACCES));
1002 			}
1003 		}
1004 	} else {
1005 		exi = cs->exi;
1006 	}
1007 	ASSERT(exi != NULL);
1008 
1009 
1010 	/*
1011 	 * Create the secinfo result based on the security information
1012 	 * from the exportinfo structure (exi).
1013 	 *
1014 	 * Return all flavors for a pseudo node.
1015 	 * For a real export node, return the flavor that the client
1016 	 * has access with.
1017 	 */
1018 	ASSERT(RW_LOCK_HELD(&exported_lock));
1019 	if (PSEUDO(exi)) {
1020 		count = exi->exi_export.ex_seccnt; /* total sec count */
1021 		resok_val = kmem_alloc(count * sizeof (secinfo4), KM_SLEEP);
1022 		secp = exi->exi_export.ex_secinfo;
1023 
1024 		for (i = 0; i < count; i++) {
1025 			si = &secp[i].s_secinfo;
1026 			resok_val[i].flavor = si->sc_rpcnum;
1027 			if (resok_val[i].flavor == RPCSEC_GSS) {
1028 				rpcsec_gss_info *info;
1029 
1030 				info = &resok_val[i].flavor_info;
1031 				info->qop = si->sc_qop;
1032 				info->service = (rpc_gss_svc_t)si->sc_service;
1033 
1034 				/* get oid opaque data */
1035 				info->oid.sec_oid4_len =
1036 				    si->sc_gss_mech_type->length;
1037 				info->oid.sec_oid4_val = kmem_alloc(
1038 				    si->sc_gss_mech_type->length, KM_SLEEP);
1039 				bcopy(
1040 				    si->sc_gss_mech_type->elements,
1041 				    info->oid.sec_oid4_val,
1042 				    info->oid.sec_oid4_len);
1043 			}
1044 		}
1045 		resp->SECINFO4resok_len = count;
1046 		resp->SECINFO4resok_val = resok_val;
1047 	} else {
1048 		int ret_cnt = 0, k = 0;
1049 		int *flavor_list;
1050 
1051 		count = exi->exi_export.ex_seccnt; /* total sec count */
1052 		secp = exi->exi_export.ex_secinfo;
1053 
1054 		flavor_list = kmem_alloc(count * sizeof (int), KM_SLEEP);
1055 		/* find out which flavors to return */
1056 		for (i = 0; i < count; i ++) {
1057 			int access, flavor, perm;
1058 
1059 			flavor = secp[i].s_secinfo.sc_nfsnum;
1060 			perm = secp[i].s_flags;
1061 
1062 			access = nfsauth4_secinfo_access(exi, cs->req,
1063 			    flavor, perm);
1064 
1065 			if (! (access & NFSAUTH_DENIED) &&
1066 			    ! (access & NFSAUTH_WRONGSEC)) {
1067 				flavor_list[ret_cnt] = flavor;
1068 				ret_cnt++;
1069 			}
1070 		}
1071 
1072 		/* Create the returning SECINFO value */
1073 		resok_val = kmem_alloc(ret_cnt * sizeof (secinfo4), KM_SLEEP);
1074 
1075 		for (i = 0; i < count; i++) {
1076 		/* If the flavor is in the flavor list, fill in resok_val. */
1077 			si = &secp[i].s_secinfo;
1078 			if (in_flavor_list(si->sc_nfsnum,
1079 			    flavor_list, ret_cnt)) {
1080 				resok_val[k].flavor = si->sc_rpcnum;
1081 				if (resok_val[k].flavor == RPCSEC_GSS) {
1082 					rpcsec_gss_info *info;
1083 
1084 					info = &resok_val[k].flavor_info;
1085 					info->qop = si->sc_qop;
1086 					info->service = (rpc_gss_svc_t)
1087 					    si->sc_service;
1088 
1089 					/* get oid opaque data */
1090 					info->oid.sec_oid4_len =
1091 					    si->sc_gss_mech_type->length;
1092 					info->oid.sec_oid4_val = kmem_alloc(
1093 					    si->sc_gss_mech_type->length,
1094 					    KM_SLEEP);
1095 					bcopy(si->sc_gss_mech_type->elements,
1096 					    info->oid.sec_oid4_val,
1097 					    info->oid.sec_oid4_len);
1098 				}
1099 				k++;
1100 			}
1101 		if (k >= ret_cnt)
1102 			break;
1103 		}
1104 		resp->SECINFO4resok_len = ret_cnt;
1105 		resp->SECINFO4resok_val = resok_val;
1106 		kmem_free(flavor_list, count * sizeof (int));
1107 	}
1108 
1109 	VN_RELE(vp);
1110 	return (NFS4_OK);
1111 }
1112 
1113 /*
1114  * SECINFO (Operation 33): Obtain required security information on
1115  * the component name in the format of (security-mechanism-oid, qop, service)
1116  * triplets.
1117  */
1118 /* ARGSUSED */
1119 static void
1120 rfs4_op_secinfo(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1121 	struct compound_state *cs)
1122 {
1123 	SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo;
1124 	utf8string *utfnm = &argop->nfs_argop4_u.opsecinfo.name;
1125 	uint_t len;
1126 	char *nm;
1127 
1128 	/*
1129 	 * Current file handle (cfh) should have been set before getting
1130 	 * into this function. If not, return error.
1131 	 */
1132 	if (cs->vp == NULL) {
1133 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1134 		return;
1135 	}
1136 
1137 	if (cs->vp->v_type != VDIR) {
1138 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
1139 		return;
1140 	}
1141 
1142 	/*
1143 	 * Verify the component name. If failed, error out, but
1144 	 * do not error out if the component name is a "..".
1145 	 * SECINFO will return its parents secinfo data for SECINFO "..".
1146 	 */
1147 	if (!utf8_dir_verify(utfnm)) {
1148 		if (utfnm->utf8string_len != 2 ||
1149 		    utfnm->utf8string_val[0] != '.' ||
1150 		    utfnm->utf8string_val[1] != '.') {
1151 			*cs->statusp = resp->status = NFS4ERR_INVAL;
1152 			return;
1153 		}
1154 	}
1155 
1156 	nm = utf8_to_str(utfnm, &len, NULL);
1157 	if (nm == NULL) {
1158 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1159 		return;
1160 	}
1161 
1162 	if (len > MAXNAMELEN) {
1163 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1164 		kmem_free(nm, len);
1165 		return;
1166 	}
1167 
1168 	*cs->statusp = resp->status = do_rfs4_op_secinfo(cs, nm, resp);
1169 
1170 	kmem_free(nm, len);
1171 }
1172 
1173 /*
1174  * Free SECINFO result.
1175  */
1176 /* ARGSUSED */
1177 static void
1178 rfs4_op_secinfo_free(nfs_resop4 *resop)
1179 {
1180 	SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo;
1181 	int count, i;
1182 	secinfo4 *resok_val;
1183 
1184 	/* If this is not an Ok result, nothing to free. */
1185 	if (resp->status != NFS4_OK) {
1186 		return;
1187 	}
1188 
1189 	count = resp->SECINFO4resok_len;
1190 	resok_val = resp->SECINFO4resok_val;
1191 
1192 	for (i = 0; i < count; i++) {
1193 		if (resok_val[i].flavor == RPCSEC_GSS) {
1194 			rpcsec_gss_info *info;
1195 
1196 			info = &resok_val[i].flavor_info;
1197 			kmem_free(info->oid.sec_oid4_val,
1198 			    info->oid.sec_oid4_len);
1199 		}
1200 	}
1201 	kmem_free(resok_val, count * sizeof (secinfo4));
1202 	resp->SECINFO4resok_len = 0;
1203 	resp->SECINFO4resok_val = NULL;
1204 }
1205 
1206 /* ARGSUSED */
1207 static void
1208 rfs4_op_access(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1209 	struct compound_state *cs)
1210 {
1211 	ACCESS4args *args = &argop->nfs_argop4_u.opaccess;
1212 	ACCESS4res *resp = &resop->nfs_resop4_u.opaccess;
1213 	int error;
1214 	vnode_t *vp;
1215 	struct vattr va;
1216 	int checkwriteperm;
1217 	cred_t *cr = cs->cr;
1218 	bslabel_t *clabel, *slabel;
1219 	ts_label_t *tslabel;
1220 	boolean_t admin_low_client;
1221 
1222 #if 0	/* XXX allow access even if !cs->access. Eventually only pseudo fs */
1223 	if (cs->access == CS_ACCESS_DENIED) {
1224 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1225 		return;
1226 	}
1227 #endif
1228 	if (cs->vp == NULL) {
1229 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1230 		return;
1231 	}
1232 
1233 	ASSERT(cr != NULL);
1234 
1235 	vp = cs->vp;
1236 
1237 	/*
1238 	 * If the file system is exported read only, it is not appropriate
1239 	 * to check write permissions for regular files and directories.
1240 	 * Special files are interpreted by the client, so the underlying
1241 	 * permissions are sent back to the client for interpretation.
1242 	 */
1243 	if (rdonly4(cs->exi, cs->vp, req) &&
1244 	    (vp->v_type == VREG || vp->v_type == VDIR))
1245 		checkwriteperm = 0;
1246 	else
1247 		checkwriteperm = 1;
1248 
1249 	/*
1250 	 * XXX
1251 	 * We need the mode so that we can correctly determine access
1252 	 * permissions relative to a mandatory lock file.  Access to
1253 	 * mandatory lock files is denied on the server, so it might
1254 	 * as well be reflected to the server during the open.
1255 	 */
1256 	va.va_mask = AT_MODE;
1257 	error = VOP_GETATTR(vp, &va, 0, cr, NULL);
1258 	if (error) {
1259 		*cs->statusp = resp->status = puterrno4(error);
1260 		return;
1261 	}
1262 	resp->access = 0;
1263 	resp->supported = 0;
1264 
1265 	if (is_system_labeled()) {
1266 		ASSERT(req->rq_label != NULL);
1267 		clabel = req->rq_label;
1268 		DTRACE_PROBE2(tx__rfs4__log__info__opaccess__clabel, char *,
1269 		    "got client label from request(1)",
1270 		    struct svc_req *, req);
1271 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
1272 			if ((tslabel = nfs_getflabel(vp)) == NULL) {
1273 				*cs->statusp = resp->status = puterrno4(EACCES);
1274 				return;
1275 			}
1276 			slabel = label2bslabel(tslabel);
1277 			DTRACE_PROBE3(tx__rfs4__log__info__opaccess__slabel,
1278 			    char *, "got server label(1) for vp(2)",
1279 			    bslabel_t *, slabel, vnode_t *, vp);
1280 
1281 			admin_low_client = B_FALSE;
1282 		} else
1283 			admin_low_client = B_TRUE;
1284 	}
1285 
1286 	if (args->access & ACCESS4_READ) {
1287 		error = VOP_ACCESS(vp, VREAD, 0, cr, NULL);
1288 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1289 		    (!is_system_labeled() || admin_low_client ||
1290 		    bldominates(clabel, slabel)))
1291 			resp->access |= ACCESS4_READ;
1292 		resp->supported |= ACCESS4_READ;
1293 	}
1294 	if ((args->access & ACCESS4_LOOKUP) && vp->v_type == VDIR) {
1295 		error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL);
1296 		if (!error && (!is_system_labeled() || admin_low_client ||
1297 		    bldominates(clabel, slabel)))
1298 			resp->access |= ACCESS4_LOOKUP;
1299 		resp->supported |= ACCESS4_LOOKUP;
1300 	}
1301 	if (checkwriteperm &&
1302 	    (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND))) {
1303 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
1304 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1305 		    (!is_system_labeled() || admin_low_client ||
1306 		    blequal(clabel, slabel)))
1307 			resp->access |=
1308 			    (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND));
1309 		resp->supported |= (ACCESS4_MODIFY|ACCESS4_EXTEND);
1310 	}
1311 
1312 	if (checkwriteperm &&
1313 	    (args->access & ACCESS4_DELETE) && vp->v_type == VDIR) {
1314 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
1315 		if (!error && (!is_system_labeled() || admin_low_client ||
1316 		    blequal(clabel, slabel)))
1317 			resp->access |= ACCESS4_DELETE;
1318 		resp->supported |= ACCESS4_DELETE;
1319 	}
1320 	if (args->access & ACCESS4_EXECUTE && vp->v_type != VDIR) {
1321 		error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL);
1322 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1323 		    (!is_system_labeled() || admin_low_client ||
1324 		    bldominates(clabel, slabel)))
1325 			resp->access |= ACCESS4_EXECUTE;
1326 		resp->supported |= ACCESS4_EXECUTE;
1327 	}
1328 
1329 	if (is_system_labeled() && !admin_low_client)
1330 		label_rele(tslabel);
1331 
1332 	*cs->statusp = resp->status = NFS4_OK;
1333 }
1334 
1335 /* ARGSUSED */
1336 static void
1337 rfs4_op_commit(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1338 	struct compound_state *cs)
1339 {
1340 	COMMIT4args *args = &argop->nfs_argop4_u.opcommit;
1341 	COMMIT4res *resp = &resop->nfs_resop4_u.opcommit;
1342 	int error;
1343 	vnode_t *vp = cs->vp;
1344 	cred_t *cr = cs->cr;
1345 	vattr_t va;
1346 
1347 	if (vp == NULL) {
1348 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1349 		return;
1350 	}
1351 	if (cs->access == CS_ACCESS_DENIED) {
1352 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1353 		return;
1354 	}
1355 
1356 	if (args->offset + args->count < args->offset) {
1357 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1358 		return;
1359 	}
1360 
1361 	va.va_mask = AT_UID;
1362 	error = VOP_GETATTR(vp, &va, 0, cr, NULL);
1363 
1364 	/*
1365 	 * If we can't get the attributes, then we can't do the
1366 	 * right access checking.  So, we'll fail the request.
1367 	 */
1368 	if (error) {
1369 		*cs->statusp = resp->status = puterrno4(error);
1370 		return;
1371 	}
1372 	if (rdonly4(cs->exi, cs->vp, req)) {
1373 		*cs->statusp = resp->status = NFS4ERR_ROFS;
1374 		return;
1375 	}
1376 
1377 	if (vp->v_type != VREG) {
1378 		if (vp->v_type == VDIR)
1379 			resp->status = NFS4ERR_ISDIR;
1380 		else
1381 			resp->status = NFS4ERR_INVAL;
1382 		*cs->statusp = resp->status;
1383 		return;
1384 	}
1385 
1386 	if (crgetuid(cr) != va.va_uid &&
1387 	    (error = VOP_ACCESS(vp, VWRITE, 0, cs->cr, NULL))) {
1388 		*cs->statusp = resp->status = puterrno4(error);
1389 		return;
1390 	}
1391 
1392 	error = VOP_PUTPAGE(vp, args->offset, args->count, 0, cr, NULL);
1393 	if (!error)
1394 		error = VOP_FSYNC(vp, FNODSYNC, cr, NULL);
1395 
1396 	if (error) {
1397 		*cs->statusp = resp->status = puterrno4(error);
1398 		return;
1399 	}
1400 
1401 	*cs->statusp = resp->status = NFS4_OK;
1402 	resp->writeverf = Write4verf;
1403 }
1404 
1405 /*
1406  * rfs4_op_mknod is called from rfs4_op_create after all initial verification
1407  * was completed. It does the nfsv4 create for special files.
1408  */
1409 /* ARGSUSED */
1410 static vnode_t *
1411 do_rfs4_op_mknod(CREATE4args *args, CREATE4res *resp, struct svc_req *req,
1412 	struct compound_state *cs, vattr_t *vap, char *nm)
1413 {
1414 	int error;
1415 	cred_t *cr = cs->cr;
1416 	vnode_t *dvp = cs->vp;
1417 	vnode_t *vp = NULL;
1418 	int mode;
1419 	enum vcexcl excl;
1420 
1421 	switch (args->type) {
1422 	case NF4CHR:
1423 	case NF4BLK:
1424 		if (secpolicy_sys_devices(cr) != 0) {
1425 			*cs->statusp = resp->status = NFS4ERR_PERM;
1426 			return (NULL);
1427 		}
1428 		if (args->type == NF4CHR)
1429 			vap->va_type = VCHR;
1430 		else
1431 			vap->va_type = VBLK;
1432 		vap->va_rdev = makedevice(args->ftype4_u.devdata.specdata1,
1433 		    args->ftype4_u.devdata.specdata2);
1434 		vap->va_mask |= AT_RDEV;
1435 		break;
1436 	case NF4SOCK:
1437 		vap->va_type = VSOCK;
1438 		break;
1439 	case NF4FIFO:
1440 		vap->va_type = VFIFO;
1441 		break;
1442 	default:
1443 		*cs->statusp = resp->status = NFS4ERR_BADTYPE;
1444 		return (NULL);
1445 	}
1446 
1447 	/*
1448 	 * Must specify the mode.
1449 	 */
1450 	if (!(vap->va_mask & AT_MODE)) {
1451 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1452 		return (NULL);
1453 	}
1454 
1455 	excl = EXCL;
1456 
1457 	mode = 0;
1458 
1459 	error = VOP_CREATE(dvp, nm, vap, excl, mode, &vp, cr, 0, NULL, NULL);
1460 	if (error) {
1461 		*cs->statusp = resp->status = puterrno4(error);
1462 		return (NULL);
1463 	}
1464 	return (vp);
1465 }
1466 
1467 /*
1468  * nfsv4 create is used to create non-regular files. For regular files,
1469  * use nfsv4 open.
1470  */
1471 /* ARGSUSED */
1472 static void
1473 rfs4_op_create(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1474 	struct compound_state *cs)
1475 {
1476 	CREATE4args *args = &argop->nfs_argop4_u.opcreate;
1477 	CREATE4res *resp = &resop->nfs_resop4_u.opcreate;
1478 	int error;
1479 	struct vattr bva, iva, iva2, ava, *vap;
1480 	cred_t *cr = cs->cr;
1481 	vnode_t *dvp = cs->vp;
1482 	vnode_t *vp = NULL;
1483 	char *nm, *lnm;
1484 	uint_t len, llen;
1485 	int syncval = 0;
1486 	struct nfs4_svgetit_arg sarg;
1487 	struct nfs4_ntov_table ntov;
1488 	struct statvfs64 sb;
1489 	nfsstat4 status;
1490 
1491 	resp->attrset = 0;
1492 
1493 	if (dvp == NULL) {
1494 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1495 		return;
1496 	}
1497 
1498 	/*
1499 	 * If there is an unshared filesystem mounted on this vnode,
1500 	 * do not allow to create an object in this directory.
1501 	 */
1502 	if (vn_ismntpt(dvp)) {
1503 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1504 		return;
1505 	}
1506 
1507 	/* Verify that type is correct */
1508 	switch (args->type) {
1509 	case NF4LNK:
1510 	case NF4BLK:
1511 	case NF4CHR:
1512 	case NF4SOCK:
1513 	case NF4FIFO:
1514 	case NF4DIR:
1515 		break;
1516 	default:
1517 		*cs->statusp = resp->status = NFS4ERR_BADTYPE;
1518 		return;
1519 	};
1520 
1521 	if (cs->access == CS_ACCESS_DENIED) {
1522 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1523 		return;
1524 	}
1525 	if (dvp->v_type != VDIR) {
1526 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
1527 		return;
1528 	}
1529 	if (!utf8_dir_verify(&args->objname)) {
1530 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1531 		return;
1532 	}
1533 
1534 	if (rdonly4(cs->exi, cs->vp, req)) {
1535 		*cs->statusp = resp->status = NFS4ERR_ROFS;
1536 		return;
1537 	}
1538 
1539 	/*
1540 	 * Name of newly created object
1541 	 */
1542 	nm = utf8_to_fn(&args->objname, &len, NULL);
1543 	if (nm == NULL) {
1544 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1545 		return;
1546 	}
1547 
1548 	if (len > MAXNAMELEN) {
1549 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1550 		kmem_free(nm, len);
1551 		return;
1552 	}
1553 
1554 	resp->attrset = 0;
1555 
1556 	sarg.sbp = &sb;
1557 	nfs4_ntov_table_init(&ntov);
1558 
1559 	status = do_rfs4_set_attrs(&resp->attrset,
1560 	    &args->createattrs, cs, &sarg, &ntov, NFS4ATTR_SETIT);
1561 
1562 	if (sarg.vap->va_mask == 0 && status == NFS4_OK)
1563 		status = NFS4ERR_INVAL;
1564 
1565 	if (status != NFS4_OK) {
1566 		*cs->statusp = resp->status = status;
1567 		kmem_free(nm, len);
1568 		nfs4_ntov_table_free(&ntov, &sarg);
1569 		resp->attrset = 0;
1570 		return;
1571 	}
1572 
1573 	/* Get "before" change value */
1574 	bva.va_mask = AT_CTIME|AT_SEQ;
1575 	error = VOP_GETATTR(dvp, &bva, 0, cr, NULL);
1576 	if (error) {
1577 		*cs->statusp = resp->status = puterrno4(error);
1578 		kmem_free(nm, len);
1579 		nfs4_ntov_table_free(&ntov, &sarg);
1580 		resp->attrset = 0;
1581 		return;
1582 	}
1583 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bva.va_ctime)
1584 
1585 	vap = sarg.vap;
1586 
1587 	/*
1588 	 * Set default initial values for attributes when not specified
1589 	 * in createattrs.
1590 	 */
1591 	if ((vap->va_mask & AT_UID) == 0) {
1592 		vap->va_uid = crgetuid(cr);
1593 		vap->va_mask |= AT_UID;
1594 	}
1595 	if ((vap->va_mask & AT_GID) == 0) {
1596 		vap->va_gid = crgetgid(cr);
1597 		vap->va_mask |= AT_GID;
1598 	}
1599 
1600 	vap->va_mask |= AT_TYPE;
1601 	switch (args->type) {
1602 	case NF4DIR:
1603 		vap->va_type = VDIR;
1604 		if ((vap->va_mask & AT_MODE) == 0) {
1605 			vap->va_mode = 0700;	/* default: owner rwx only */
1606 			vap->va_mask |= AT_MODE;
1607 		}
1608 		error = VOP_MKDIR(dvp, nm, vap, &vp, cr, NULL, 0, NULL);
1609 		if (error)
1610 			break;
1611 
1612 		/*
1613 		 * Get the initial "after" sequence number, if it fails,
1614 		 * set to zero
1615 		 */
1616 		iva.va_mask = AT_SEQ;
1617 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL))
1618 			iva.va_seq = 0;
1619 		break;
1620 	case NF4LNK:
1621 		vap->va_type = VLNK;
1622 		if ((vap->va_mask & AT_MODE) == 0) {
1623 			vap->va_mode = 0700;	/* default: owner rwx only */
1624 			vap->va_mask |= AT_MODE;
1625 		}
1626 
1627 		/*
1628 		 * symlink names must be treated as data
1629 		 */
1630 		lnm = utf8_to_str(&args->ftype4_u.linkdata, &llen, NULL);
1631 
1632 		if (lnm == NULL) {
1633 			*cs->statusp = resp->status = NFS4ERR_INVAL;
1634 			kmem_free(nm, len);
1635 			nfs4_ntov_table_free(&ntov, &sarg);
1636 			resp->attrset = 0;
1637 			return;
1638 		}
1639 
1640 		if (llen > MAXPATHLEN) {
1641 			*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1642 			kmem_free(nm, len);
1643 			kmem_free(lnm, llen);
1644 			nfs4_ntov_table_free(&ntov, &sarg);
1645 			resp->attrset = 0;
1646 			return;
1647 		}
1648 
1649 		error = VOP_SYMLINK(dvp, nm, vap, lnm, cr, NULL, 0);
1650 		if (lnm != NULL)
1651 			kmem_free(lnm, llen);
1652 		if (error)
1653 			break;
1654 
1655 		/*
1656 		 * Get the initial "after" sequence number, if it fails,
1657 		 * set to zero
1658 		 */
1659 		iva.va_mask = AT_SEQ;
1660 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL))
1661 			iva.va_seq = 0;
1662 
1663 		error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr,
1664 		    NULL, NULL, NULL);
1665 		if (error)
1666 			break;
1667 
1668 		/*
1669 		 * va_seq is not safe over VOP calls, check it again
1670 		 * if it has changed zero out iva to force atomic = FALSE.
1671 		 */
1672 		iva2.va_mask = AT_SEQ;
1673 		if (VOP_GETATTR(dvp, &iva2, 0, cs->cr, NULL) ||
1674 		    iva2.va_seq != iva.va_seq)
1675 			iva.va_seq = 0;
1676 		break;
1677 	default:
1678 		/*
1679 		 * probably a special file.
1680 		 */
1681 		if ((vap->va_mask & AT_MODE) == 0) {
1682 			vap->va_mode = 0600;	/* default: owner rw only */
1683 			vap->va_mask |= AT_MODE;
1684 		}
1685 		syncval = FNODSYNC;
1686 		/*
1687 		 * We know this will only generate one VOP call
1688 		 */
1689 		vp = do_rfs4_op_mknod(args, resp, req, cs, vap, nm);
1690 
1691 		if (vp == NULL) {
1692 			kmem_free(nm, len);
1693 			nfs4_ntov_table_free(&ntov, &sarg);
1694 			resp->attrset = 0;
1695 			return;
1696 		}
1697 
1698 		/*
1699 		 * Get the initial "after" sequence number, if it fails,
1700 		 * set to zero
1701 		 */
1702 		iva.va_mask = AT_SEQ;
1703 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL))
1704 			iva.va_seq = 0;
1705 
1706 		break;
1707 	}
1708 	kmem_free(nm, len);
1709 
1710 	if (error) {
1711 		*cs->statusp = resp->status = puterrno4(error);
1712 	}
1713 
1714 	/*
1715 	 * Force modified data and metadata out to stable storage.
1716 	 */
1717 	(void) VOP_FSYNC(dvp, 0, cr, NULL);
1718 
1719 	if (resp->status != NFS4_OK) {
1720 		if (vp != NULL)
1721 			VN_RELE(vp);
1722 		nfs4_ntov_table_free(&ntov, &sarg);
1723 		resp->attrset = 0;
1724 		return;
1725 	}
1726 
1727 	/*
1728 	 * Finish setup of cinfo response, "before" value already set.
1729 	 * Get "after" change value, if it fails, simply return the
1730 	 * before value.
1731 	 */
1732 	ava.va_mask = AT_CTIME|AT_SEQ;
1733 	if (VOP_GETATTR(dvp, &ava, 0, cr, NULL)) {
1734 		ava.va_ctime = bva.va_ctime;
1735 		ava.va_seq = 0;
1736 	}
1737 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, ava.va_ctime);
1738 
1739 	/*
1740 	 * True verification that object was created with correct
1741 	 * attrs is impossible.  The attrs could have been changed
1742 	 * immediately after object creation.  If attributes did
1743 	 * not verify, the only recourse for the server is to
1744 	 * destroy the object.  Maybe if some attrs (like gid)
1745 	 * are set incorrectly, the object should be destroyed;
1746 	 * however, seems bad as a default policy.  Do we really
1747 	 * want to destroy an object over one of the times not
1748 	 * verifying correctly?  For these reasons, the server
1749 	 * currently sets bits in attrset for createattrs
1750 	 * that were set; however, no verification is done.
1751 	 *
1752 	 * vmask_to_nmask accounts for vattr bits set on create
1753 	 *	[do_rfs4_set_attrs() only sets resp bits for
1754 	 *	 non-vattr/vfs bits.]
1755 	 * Mask off any bits set by default so as not to return
1756 	 * more attrset bits than were requested in createattrs
1757 	 */
1758 	nfs4_vmask_to_nmask(sarg.vap->va_mask, &resp->attrset);
1759 	resp->attrset &= args->createattrs.attrmask;
1760 	nfs4_ntov_table_free(&ntov, &sarg);
1761 
1762 	error = makefh4(&cs->fh, vp, cs->exi);
1763 	if (error) {
1764 		*cs->statusp = resp->status = puterrno4(error);
1765 	}
1766 
1767 	/*
1768 	 * The cinfo.atomic = TRUE only if we got no errors, we have
1769 	 * non-zero va_seq's, and it has incremented by exactly one
1770 	 * during the creation and it didn't change during the VOP_LOOKUP
1771 	 * or VOP_FSYNC.
1772 	 */
1773 	if (!error && bva.va_seq && iva.va_seq && ava.va_seq &&
1774 	    iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq)
1775 		resp->cinfo.atomic = TRUE;
1776 	else
1777 		resp->cinfo.atomic = FALSE;
1778 
1779 	(void) VOP_FSYNC(vp, syncval, cr, NULL);
1780 
1781 	if (resp->status != NFS4_OK) {
1782 		VN_RELE(vp);
1783 		return;
1784 	}
1785 	if (cs->vp)
1786 		VN_RELE(cs->vp);
1787 
1788 	cs->vp = vp;
1789 	*cs->statusp = resp->status = NFS4_OK;
1790 }
1791 
1792 
1793 /*ARGSUSED*/
1794 static void
1795 rfs4_op_delegreturn(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1796 	struct compound_state *cs)
1797 {
1798 	DELEGRETURN4args *args = &argop->nfs_argop4_u.opdelegreturn;
1799 	DELEGRETURN4res *resp = &resop->nfs_resop4_u.opdelegreturn;
1800 	rfs4_deleg_state_t *dsp;
1801 	nfsstat4 status;
1802 
1803 	status = rfs4_get_deleg_state(&args->deleg_stateid, &dsp);
1804 	resp->status = *cs->statusp = status;
1805 	if (status != NFS4_OK)
1806 		return;
1807 
1808 	/* Ensure specified filehandle matches */
1809 	if (cs->vp != dsp->finfo->vp) {
1810 		resp->status = *cs->statusp = NFS4ERR_BAD_STATEID;
1811 	} else
1812 		rfs4_return_deleg(dsp, FALSE);
1813 
1814 	rfs4_update_lease(dsp->client);
1815 
1816 	rfs4_deleg_state_rele(dsp);
1817 }
1818 
1819 /*
1820  * Check to see if a given "flavor" is an explicitly shared flavor.
1821  * The assumption of this routine is the "flavor" is already a valid
1822  * flavor in the secinfo list of "exi".
1823  *
1824  *	e.g.
1825  *		# share -o sec=flavor1 /export
1826  *		# share -o sec=flavor2 /export/home
1827  *
1828  *		flavor2 is not an explicitly shared flavor for /export,
1829  *		however it is in the secinfo list for /export thru the
1830  *		server namespace setup.
1831  */
1832 int
1833 is_exported_sec(int flavor, struct exportinfo *exi)
1834 {
1835 	int	i;
1836 	struct secinfo *sp;
1837 
1838 	sp = exi->exi_export.ex_secinfo;
1839 	for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
1840 		if (flavor == sp[i].s_secinfo.sc_nfsnum ||
1841 		    sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) {
1842 			return (SEC_REF_EXPORTED(&sp[i]));
1843 		}
1844 	}
1845 
1846 	/* Should not reach this point based on the assumption */
1847 	return (0);
1848 }
1849 
1850 /*
1851  * Check if the security flavor used in the request matches what is
1852  * required at the export point or at the root pseudo node (exi_root).
1853  *
1854  * returns 1 if there's a match or if exported with AUTH_NONE; 0 otherwise.
1855  *
1856  */
1857 static int
1858 secinfo_match_or_authnone(struct compound_state *cs)
1859 {
1860 	int	i;
1861 	struct secinfo *sp;
1862 
1863 	/*
1864 	 * Check cs->nfsflavor (from the request) against
1865 	 * the current export data in cs->exi.
1866 	 */
1867 	sp = cs->exi->exi_export.ex_secinfo;
1868 	for (i = 0; i < cs->exi->exi_export.ex_seccnt; i++) {
1869 		if (cs->nfsflavor == sp[i].s_secinfo.sc_nfsnum ||
1870 		    sp[i].s_secinfo.sc_nfsnum == AUTH_NONE)
1871 			return (1);
1872 	}
1873 
1874 	return (0);
1875 }
1876 
1877 /*
1878  * Check the access authority for the client and return the correct error.
1879  */
1880 nfsstat4
1881 call_checkauth4(struct compound_state *cs, struct svc_req *req)
1882 {
1883 	int	authres;
1884 
1885 	/*
1886 	 * First, check if the security flavor used in the request
1887 	 * are among the flavors set in the server namespace.
1888 	 */
1889 	if (!secinfo_match_or_authnone(cs)) {
1890 		*cs->statusp = NFS4ERR_WRONGSEC;
1891 		return (*cs->statusp);
1892 	}
1893 
1894 	authres = checkauth4(cs, req);
1895 
1896 	if (authres > 0) {
1897 		*cs->statusp = NFS4_OK;
1898 		if (! (cs->access & CS_ACCESS_LIMITED))
1899 			cs->access = CS_ACCESS_OK;
1900 	} else if (authres == 0) {
1901 		*cs->statusp = NFS4ERR_ACCESS;
1902 	} else if (authres == -2) {
1903 		*cs->statusp = NFS4ERR_WRONGSEC;
1904 	} else {
1905 		*cs->statusp = NFS4ERR_DELAY;
1906 	}
1907 	return (*cs->statusp);
1908 }
1909 
1910 /*
1911  * bitmap4_to_attrmask is called by getattr and readdir.
1912  * It sets up the vattr mask and determines whether vfsstat call is needed
1913  * based on the input bitmap.
1914  * Returns nfsv4 status.
1915  */
1916 static nfsstat4
1917 bitmap4_to_attrmask(bitmap4 breq, struct nfs4_svgetit_arg *sargp)
1918 {
1919 	int i;
1920 	uint_t	va_mask;
1921 	struct statvfs64 *sbp = sargp->sbp;
1922 
1923 	sargp->sbp = NULL;
1924 	sargp->flag = 0;
1925 	sargp->rdattr_error = NFS4_OK;
1926 	sargp->mntdfid_set = FALSE;
1927 	if (sargp->cs->vp)
1928 		sargp->xattr = get_fh4_flag(&sargp->cs->fh,
1929 		    FH4_ATTRDIR | FH4_NAMEDATTR);
1930 	else
1931 		sargp->xattr = 0;
1932 
1933 	/*
1934 	 * Set rdattr_error_req to true if return error per
1935 	 * failed entry rather than fail the readdir.
1936 	 */
1937 	if (breq & FATTR4_RDATTR_ERROR_MASK)
1938 		sargp->rdattr_error_req = 1;
1939 	else
1940 		sargp->rdattr_error_req = 0;
1941 
1942 	/*
1943 	 * generate the va_mask
1944 	 * Handle the easy cases first
1945 	 */
1946 	switch (breq) {
1947 	case NFS4_NTOV_ATTR_MASK:
1948 		sargp->vap->va_mask = NFS4_NTOV_ATTR_AT_MASK;
1949 		return (NFS4_OK);
1950 
1951 	case NFS4_FS_ATTR_MASK:
1952 		sargp->vap->va_mask = NFS4_FS_ATTR_AT_MASK;
1953 		sargp->sbp = sbp;
1954 		return (NFS4_OK);
1955 
1956 	case NFS4_NTOV_ATTR_CACHE_MASK:
1957 		sargp->vap->va_mask = NFS4_NTOV_ATTR_CACHE_AT_MASK;
1958 		return (NFS4_OK);
1959 
1960 	case FATTR4_LEASE_TIME_MASK:
1961 		sargp->vap->va_mask = 0;
1962 		return (NFS4_OK);
1963 
1964 	default:
1965 		va_mask = 0;
1966 		for (i = 0; i < nfs4_ntov_map_size; i++) {
1967 			if ((breq & nfs4_ntov_map[i].fbit) &&
1968 			    nfs4_ntov_map[i].vbit)
1969 				va_mask |= nfs4_ntov_map[i].vbit;
1970 		}
1971 
1972 		/*
1973 		 * Check is vfsstat is needed
1974 		 */
1975 		if (breq & NFS4_FS_ATTR_MASK)
1976 			sargp->sbp = sbp;
1977 
1978 		sargp->vap->va_mask = va_mask;
1979 		return (NFS4_OK);
1980 	}
1981 	/* NOTREACHED */
1982 }
1983 
1984 /*
1985  * bitmap4_get_sysattrs is called by getattr and readdir.
1986  * It calls both VOP_GETATTR and VFS_STATVFS calls to get the attrs.
1987  * Returns nfsv4 status.
1988  */
1989 static nfsstat4
1990 bitmap4_get_sysattrs(struct nfs4_svgetit_arg *sargp)
1991 {
1992 	int error;
1993 	struct compound_state *cs = sargp->cs;
1994 	vnode_t *vp = cs->vp;
1995 
1996 	if (sargp->sbp != NULL) {
1997 		if (error = VFS_STATVFS(vp->v_vfsp, sargp->sbp)) {
1998 			sargp->sbp = NULL;	/* to identify error */
1999 			return (puterrno4(error));
2000 		}
2001 	}
2002 
2003 	return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr));
2004 }
2005 
2006 static void
2007 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp)
2008 {
2009 	ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size,
2010 	    KM_SLEEP);
2011 	ntovp->attrcnt = 0;
2012 	ntovp->vfsstat = FALSE;
2013 }
2014 
2015 static void
2016 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp,
2017 	struct nfs4_svgetit_arg *sargp)
2018 {
2019 	int i;
2020 	union nfs4_attr_u *na;
2021 	uint8_t *amap;
2022 
2023 	/*
2024 	 * XXX Should do the same checks for whether the bit is set
2025 	 */
2026 	for (i = 0, na = ntovp->na, amap = ntovp->amap;
2027 	    i < ntovp->attrcnt; i++, na++, amap++) {
2028 		(void) (*nfs4_ntov_map[*amap].sv_getit)(
2029 		    NFS4ATTR_FREEIT, sargp, na);
2030 	}
2031 	if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) {
2032 		/*
2033 		 * xdr_free for getattr will be done later
2034 		 */
2035 		for (i = 0, na = ntovp->na, amap = ntovp->amap;
2036 		    i < ntovp->attrcnt; i++, na++, amap++) {
2037 			xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na);
2038 		}
2039 	}
2040 	kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size);
2041 }
2042 
2043 /*
2044  * do_rfs4_op_getattr gets the system attrs and converts into fattr4.
2045  */
2046 static nfsstat4
2047 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp,
2048 	struct nfs4_svgetit_arg *sargp)
2049 {
2050 	int error = 0;
2051 	int i, k;
2052 	struct nfs4_ntov_table ntov;
2053 	XDR xdr;
2054 	ulong_t xdr_size;
2055 	char *xdr_attrs;
2056 	nfsstat4 status = NFS4_OK;
2057 	nfsstat4 prev_rdattr_error = sargp->rdattr_error;
2058 	union nfs4_attr_u *na;
2059 	uint8_t *amap;
2060 
2061 	sargp->op = NFS4ATTR_GETIT;
2062 	sargp->flag = 0;
2063 
2064 	fattrp->attrmask = 0;
2065 	/* if no bits requested, then return empty fattr4 */
2066 	if (breq == 0) {
2067 		fattrp->attrlist4_len = 0;
2068 		fattrp->attrlist4 = NULL;
2069 		return (NFS4_OK);
2070 	}
2071 
2072 	/*
2073 	 * return NFS4ERR_INVAL when client requests write-only attrs
2074 	 */
2075 	if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK))
2076 		return (NFS4ERR_INVAL);
2077 
2078 	nfs4_ntov_table_init(&ntov);
2079 	na = ntov.na;
2080 	amap = ntov.amap;
2081 
2082 	/*
2083 	 * Now loop to get or verify the attrs
2084 	 */
2085 	for (i = 0; i < nfs4_ntov_map_size; i++) {
2086 		if (breq & nfs4_ntov_map[i].fbit) {
2087 			if ((*nfs4_ntov_map[i].sv_getit)(
2088 			    NFS4ATTR_SUPPORTED, sargp, NULL) == 0) {
2089 
2090 				error = (*nfs4_ntov_map[i].sv_getit)(
2091 				    NFS4ATTR_GETIT, sargp, na);
2092 
2093 				/*
2094 				 * Possible error values:
2095 				 * >0 if sv_getit failed to
2096 				 * get the attr; 0 if succeeded;
2097 				 * <0 if rdattr_error and the
2098 				 * attribute cannot be returned.
2099 				 */
2100 				if (error && !(sargp->rdattr_error_req))
2101 					goto done;
2102 				/*
2103 				 * If error then just for entry
2104 				 */
2105 				if (error == 0) {
2106 					fattrp->attrmask |=
2107 					    nfs4_ntov_map[i].fbit;
2108 					*amap++ =
2109 					    (uint8_t)nfs4_ntov_map[i].nval;
2110 					na++;
2111 					(ntov.attrcnt)++;
2112 				} else if ((error > 0) &&
2113 				    (sargp->rdattr_error == NFS4_OK)) {
2114 					sargp->rdattr_error = puterrno4(error);
2115 				}
2116 				error = 0;
2117 			}
2118 		}
2119 	}
2120 
2121 	/*
2122 	 * If rdattr_error was set after the return value for it was assigned,
2123 	 * update it.
2124 	 */
2125 	if (prev_rdattr_error != sargp->rdattr_error) {
2126 		na = ntov.na;
2127 		amap = ntov.amap;
2128 		for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2129 			k = *amap;
2130 			if (k < FATTR4_RDATTR_ERROR) {
2131 				continue;
2132 			}
2133 			if ((k == FATTR4_RDATTR_ERROR) &&
2134 			    ((*nfs4_ntov_map[k].sv_getit)(
2135 			    NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) {
2136 
2137 				(void) (*nfs4_ntov_map[k].sv_getit)(
2138 				    NFS4ATTR_GETIT, sargp, na);
2139 			}
2140 			break;
2141 		}
2142 	}
2143 
2144 	xdr_size = 0;
2145 	na = ntov.na;
2146 	amap = ntov.amap;
2147 	for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2148 		xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na);
2149 	}
2150 
2151 	fattrp->attrlist4_len = xdr_size;
2152 	if (xdr_size) {
2153 		/* freed by rfs4_op_getattr_free() */
2154 		fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP);
2155 
2156 		xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE);
2157 
2158 		na = ntov.na;
2159 		amap = ntov.amap;
2160 		for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2161 			if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) {
2162 				cmn_err(CE_WARN, "do_rfs4_op_getattr: xdr "
2163 				    "encode of attribute %d failed\n", *amap);
2164 				status = NFS4ERR_SERVERFAULT;
2165 				break;
2166 			}
2167 		}
2168 		/* xdrmem_destroy(&xdrs); */	/* NO-OP */
2169 	} else {
2170 		fattrp->attrlist4 = NULL;
2171 	}
2172 done:
2173 
2174 	nfs4_ntov_table_free(&ntov, sargp);
2175 
2176 	if (error != 0)
2177 		status = puterrno4(error);
2178 
2179 	return (status);
2180 }
2181 
2182 /* ARGSUSED */
2183 static void
2184 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2185 	struct compound_state *cs)
2186 {
2187 	GETATTR4args *args = &argop->nfs_argop4_u.opgetattr;
2188 	GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr;
2189 	struct nfs4_svgetit_arg sarg;
2190 	struct statvfs64 sb;
2191 	nfsstat4 status;
2192 
2193 	if (cs->vp == NULL) {
2194 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2195 		return;
2196 	}
2197 
2198 	if (cs->access == CS_ACCESS_DENIED) {
2199 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2200 		return;
2201 	}
2202 
2203 	sarg.sbp = &sb;
2204 	sarg.cs = cs;
2205 
2206 	status = bitmap4_to_attrmask(args->attr_request, &sarg);
2207 	if (status == NFS4_OK) {
2208 		status = bitmap4_get_sysattrs(&sarg);
2209 		if (status == NFS4_OK)
2210 			status = do_rfs4_op_getattr(args->attr_request,
2211 			    &resp->obj_attributes, &sarg);
2212 	}
2213 	*cs->statusp = resp->status = status;
2214 }
2215 
2216 static void
2217 rfs4_op_getattr_free(nfs_resop4 *resop)
2218 {
2219 	GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr;
2220 
2221 	nfs4_fattr4_free(&resp->obj_attributes);
2222 }
2223 
2224 /* ARGSUSED */
2225 static void
2226 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2227 	struct compound_state *cs)
2228 {
2229 	GETFH4res *resp = &resop->nfs_resop4_u.opgetfh;
2230 
2231 	if (cs->vp == NULL) {
2232 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2233 		return;
2234 	}
2235 	if (cs->access == CS_ACCESS_DENIED) {
2236 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2237 		return;
2238 	}
2239 
2240 	resp->object.nfs_fh4_val =
2241 	    kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP);
2242 	nfs_fh4_copy(&cs->fh, &resp->object);
2243 	*cs->statusp = resp->status = NFS4_OK;
2244 }
2245 
2246 static void
2247 rfs4_op_getfh_free(nfs_resop4 *resop)
2248 {
2249 	GETFH4res *resp = &resop->nfs_resop4_u.opgetfh;
2250 
2251 	if (resp->status == NFS4_OK &&
2252 	    resp->object.nfs_fh4_val != NULL) {
2253 		kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len);
2254 		resp->object.nfs_fh4_val = NULL;
2255 		resp->object.nfs_fh4_len = 0;
2256 	}
2257 }
2258 
2259 /*
2260  * illegal: args: void
2261  *	    res : status (NFS4ERR_OP_ILLEGAL)
2262  */
2263 /* ARGSUSED */
2264 static void
2265 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop,
2266 	struct svc_req *req, struct compound_state *cs)
2267 {
2268 	ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal;
2269 
2270 	resop->resop = OP_ILLEGAL;
2271 	*cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL;
2272 }
2273 
2274 /*
2275  * link: args: SAVED_FH: file, CURRENT_FH: target directory
2276  *	 res: status. If success - CURRENT_FH unchanged, return change_info
2277  */
2278 /* ARGSUSED */
2279 static void
2280 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2281 	struct compound_state *cs)
2282 {
2283 	LINK4args *args = &argop->nfs_argop4_u.oplink;
2284 	LINK4res *resp = &resop->nfs_resop4_u.oplink;
2285 	int error;
2286 	vnode_t *vp;
2287 	vnode_t *dvp;
2288 	struct vattr bdva, idva, adva;
2289 	char *nm;
2290 	uint_t  len;
2291 
2292 	/* SAVED_FH: source object */
2293 	vp = cs->saved_vp;
2294 	if (vp == NULL) {
2295 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2296 		return;
2297 	}
2298 
2299 	/* CURRENT_FH: target directory */
2300 	dvp = cs->vp;
2301 	if (dvp == NULL) {
2302 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2303 		return;
2304 	}
2305 
2306 	/*
2307 	 * If there is a non-shared filesystem mounted on this vnode,
2308 	 * do not allow to link any file in this directory.
2309 	 */
2310 	if (vn_ismntpt(dvp)) {
2311 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2312 		return;
2313 	}
2314 
2315 	if (cs->access == CS_ACCESS_DENIED) {
2316 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2317 		return;
2318 	}
2319 
2320 	/* Check source object's type validity */
2321 	if (vp->v_type == VDIR) {
2322 		*cs->statusp = resp->status = NFS4ERR_ISDIR;
2323 		return;
2324 	}
2325 
2326 	/* Check target directory's type */
2327 	if (dvp->v_type != VDIR) {
2328 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2329 		return;
2330 	}
2331 
2332 	if (cs->saved_exi != cs->exi) {
2333 		*cs->statusp = resp->status = NFS4ERR_XDEV;
2334 		return;
2335 	}
2336 
2337 	if (!utf8_dir_verify(&args->newname)) {
2338 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2339 		return;
2340 	}
2341 
2342 	nm = utf8_to_fn(&args->newname, &len, NULL);
2343 	if (nm == NULL) {
2344 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2345 		return;
2346 	}
2347 
2348 	if (len > MAXNAMELEN) {
2349 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
2350 		kmem_free(nm, len);
2351 		return;
2352 	}
2353 
2354 	if (rdonly4(cs->exi, cs->vp, req)) {
2355 		*cs->statusp = resp->status = NFS4ERR_ROFS;
2356 		kmem_free(nm, len);
2357 		return;
2358 	}
2359 
2360 	/* Get "before" change value */
2361 	bdva.va_mask = AT_CTIME|AT_SEQ;
2362 	error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL);
2363 	if (error) {
2364 		*cs->statusp = resp->status = puterrno4(error);
2365 		kmem_free(nm, len);
2366 		return;
2367 	}
2368 
2369 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime)
2370 
2371 	error = VOP_LINK(dvp, vp, nm, cs->cr, NULL, 0);
2372 
2373 	kmem_free(nm, len);
2374 
2375 	/*
2376 	 * Get the initial "after" sequence number, if it fails, set to zero
2377 	 */
2378 	idva.va_mask = AT_SEQ;
2379 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL))
2380 		idva.va_seq = 0;
2381 
2382 	/*
2383 	 * Force modified data and metadata out to stable storage.
2384 	 */
2385 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL);
2386 	(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
2387 
2388 	if (error) {
2389 		*cs->statusp = resp->status = puterrno4(error);
2390 		return;
2391 	}
2392 
2393 	/*
2394 	 * Get "after" change value, if it fails, simply return the
2395 	 * before value.
2396 	 */
2397 	adva.va_mask = AT_CTIME|AT_SEQ;
2398 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) {
2399 		adva.va_ctime = bdva.va_ctime;
2400 		adva.va_seq = 0;
2401 	}
2402 
2403 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
2404 
2405 	/*
2406 	 * The cinfo.atomic = TRUE only if we have
2407 	 * non-zero va_seq's, and it has incremented by exactly one
2408 	 * during the VOP_LINK and it didn't change during the VOP_FSYNC.
2409 	 */
2410 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
2411 	    idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq)
2412 		resp->cinfo.atomic = TRUE;
2413 	else
2414 		resp->cinfo.atomic = FALSE;
2415 
2416 	*cs->statusp = resp->status = NFS4_OK;
2417 }
2418 
2419 /*
2420  * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work.
2421  */
2422 
2423 /* ARGSUSED */
2424 static nfsstat4
2425 do_rfs4_op_lookup(char *nm, uint_t buflen, struct svc_req *req,
2426 	struct compound_state *cs)
2427 {
2428 	int error;
2429 	int different_export = 0;
2430 	vnode_t *vp, *tvp, *pre_tvp = NULL, *oldvp = NULL;
2431 	struct exportinfo *exi = NULL, *pre_exi = NULL;
2432 	nfsstat4 stat;
2433 	fid_t fid;
2434 	int attrdir, dotdot, walk;
2435 	bool_t is_newvp = FALSE;
2436 
2437 	if (cs->vp->v_flag & V_XATTRDIR) {
2438 		attrdir = 1;
2439 		ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR));
2440 	} else {
2441 		attrdir = 0;
2442 		ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR));
2443 	}
2444 
2445 	dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0');
2446 
2447 	/*
2448 	 * If dotdotting, then need to check whether it's
2449 	 * above the root of a filesystem, or above an
2450 	 * export point.
2451 	 */
2452 	if (dotdot) {
2453 
2454 		/*
2455 		 * If dotdotting at the root of a filesystem, then
2456 		 * need to traverse back to the mounted-on filesystem
2457 		 * and do the dotdot lookup there.
2458 		 */
2459 		if (cs->vp->v_flag & VROOT) {
2460 
2461 			/*
2462 			 * If at the system root, then can
2463 			 * go up no further.
2464 			 */
2465 			if (VN_CMP(cs->vp, rootdir))
2466 				return (puterrno4(ENOENT));
2467 
2468 			/*
2469 			 * Traverse back to the mounted-on filesystem
2470 			 */
2471 			cs->vp = untraverse(cs->vp);
2472 
2473 			/*
2474 			 * Set the different_export flag so we remember
2475 			 * to pick up a new exportinfo entry for
2476 			 * this new filesystem.
2477 			 */
2478 			different_export = 1;
2479 		} else {
2480 
2481 			/*
2482 			 * If dotdotting above an export point then set
2483 			 * the different_export to get new export info.
2484 			 */
2485 			different_export = nfs_exported(cs->exi, cs->vp);
2486 		}
2487 	}
2488 
2489 	error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr,
2490 	    NULL, NULL, NULL);
2491 	if (error)
2492 		return (puterrno4(error));
2493 
2494 	/*
2495 	 * If the vnode is in a pseudo filesystem, check whether it is visible.
2496 	 *
2497 	 * XXX if the vnode is a symlink and it is not visible in
2498 	 * a pseudo filesystem, return ENOENT (not following symlink).
2499 	 * V4 client can not mount such symlink. This is a regression
2500 	 * from V2/V3.
2501 	 *
2502 	 * In the same exported filesystem, if the security flavor used
2503 	 * is not an explicitly shared flavor, limit the view to the visible
2504 	 * list entries only. This is not a WRONGSEC case because it's already
2505 	 * checked via PUTROOTFH/PUTPUBFH or PUTFH.
2506 	 */
2507 	if (!different_export &&
2508 	    (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) ||
2509 	    cs->access & CS_ACCESS_LIMITED)) {
2510 		if (! nfs_visible(cs->exi, vp, &different_export)) {
2511 			VN_RELE(vp);
2512 			return (puterrno4(ENOENT));
2513 		}
2514 	}
2515 
2516 	/*
2517 	 * If it's a mountpoint, then traverse it.
2518 	 */
2519 	if (vn_ismntpt(vp)) {
2520 		pre_exi = cs->exi;	/* save pre-traversed exportinfo */
2521 		pre_tvp = vp;		/* save pre-traversed vnode	*/
2522 
2523 		/*
2524 		 * hold pre_tvp to counteract rele by traverse.  We will
2525 		 * need pre_tvp below if checkexport4 fails
2526 		 */
2527 		VN_HOLD(pre_tvp);
2528 		tvp = vp;
2529 		if ((error = traverse(&tvp)) != 0) {
2530 			VN_RELE(vp);
2531 			VN_RELE(pre_tvp);
2532 			return (puterrno4(error));
2533 		}
2534 		vp = tvp;
2535 		different_export = 1;
2536 	} else if (vp->v_vfsp != cs->vp->v_vfsp) {
2537 		/*
2538 		 * The vfsp comparison is to handle the case where
2539 		 * a LOFS mount is shared.  lo_lookup traverses mount points,
2540 		 * and NFS is unaware of local fs transistions because
2541 		 * v_vfsmountedhere isn't set.  For this special LOFS case,
2542 		 * the dir and the obj returned by lookup will have different
2543 		 * vfs ptrs.
2544 		 */
2545 		different_export = 1;
2546 	}
2547 
2548 	if (different_export) {
2549 
2550 		bzero(&fid, sizeof (fid));
2551 		fid.fid_len = MAXFIDSZ;
2552 		error = vop_fid_pseudo(vp, &fid);
2553 		if (error) {
2554 			VN_RELE(vp);
2555 			if (pre_tvp)
2556 				VN_RELE(pre_tvp);
2557 			return (puterrno4(error));
2558 		}
2559 
2560 		if (dotdot)
2561 			exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE);
2562 		else
2563 			exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
2564 
2565 		if (exi == NULL) {
2566 			if (pre_tvp) {
2567 				/*
2568 				 * If this vnode is a mounted-on vnode,
2569 				 * but the mounted-on file system is not
2570 				 * exported, send back the filehandle for
2571 				 * the mounted-on vnode, not the root of
2572 				 * the mounted-on file system.
2573 				 */
2574 				VN_RELE(vp);
2575 				vp = pre_tvp;
2576 				exi = pre_exi;
2577 			} else {
2578 				VN_RELE(vp);
2579 				return (puterrno4(EACCES));
2580 			}
2581 		} else if (pre_tvp) {
2582 			/* we're done with pre_tvp now. release extra hold */
2583 			VN_RELE(pre_tvp);
2584 		}
2585 
2586 		cs->exi = exi;
2587 
2588 		/*
2589 		 * Now we do a checkauth4. The reason is that
2590 		 * this client/user may not have access to the new
2591 		 * exported file system, and if he does,
2592 		 * the client/user may be mapped to a different uid.
2593 		 *
2594 		 * We start with a new cr, because the checkauth4 done
2595 		 * in the PUT*FH operation over wrote the cred's uid,
2596 		 * gid, etc, and we want the real thing before calling
2597 		 * checkauth4()
2598 		 */
2599 		crfree(cs->cr);
2600 		cs->cr = crdup(cs->basecr);
2601 
2602 		if (cs->vp)
2603 			oldvp = cs->vp;
2604 		cs->vp = vp;
2605 		is_newvp = TRUE;
2606 
2607 		stat = call_checkauth4(cs, req);
2608 		if (stat != NFS4_OK) {
2609 			VN_RELE(cs->vp);
2610 			cs->vp = oldvp;
2611 			return (stat);
2612 		}
2613 	}
2614 
2615 	/*
2616 	 * After various NFS checks, do a label check on the path
2617 	 * component. The label on this path should either be the
2618 	 * global zone's label or a zone's label. We are only
2619 	 * interested in the zone's label because exported files
2620 	 * in global zone is accessible (though read-only) to
2621 	 * clients. The exportability/visibility check is already
2622 	 * done before reaching this code.
2623 	 */
2624 	if (is_system_labeled()) {
2625 		bslabel_t *clabel;
2626 
2627 		ASSERT(req->rq_label != NULL);
2628 		clabel = req->rq_label;
2629 		DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *,
2630 		    "got client label from request(1)", struct svc_req *, req);
2631 
2632 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
2633 			if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) {
2634 				error = EACCES;
2635 				goto err_out;
2636 			}
2637 		} else {
2638 			/*
2639 			 * We grant access to admin_low label clients
2640 			 * only if the client is trusted, i.e. also
2641 			 * running Solaris Trusted Extension.
2642 			 */
2643 			struct sockaddr	*ca;
2644 			int		addr_type;
2645 			void		*ipaddr;
2646 			tsol_tpc_t	*tp;
2647 
2648 			ca = (struct sockaddr *)svc_getrpccaller(
2649 			    req->rq_xprt)->buf;
2650 			if (ca->sa_family == AF_INET) {
2651 				addr_type = IPV4_VERSION;
2652 				ipaddr = &((struct sockaddr_in *)ca)->sin_addr;
2653 			} else if (ca->sa_family == AF_INET6) {
2654 				addr_type = IPV6_VERSION;
2655 				ipaddr = &((struct sockaddr_in6 *)
2656 				    ca)->sin6_addr;
2657 			}
2658 			tp = find_tpc(ipaddr, addr_type, B_FALSE);
2659 			if (tp == NULL || tp->tpc_tp.tp_doi !=
2660 			    l_admin_low->tsl_doi || tp->tpc_tp.host_type !=
2661 			    SUN_CIPSO) {
2662 				if (tp != NULL)
2663 					TPC_RELE(tp);
2664 				error = EACCES;
2665 				goto err_out;
2666 			}
2667 			TPC_RELE(tp);
2668 		}
2669 	}
2670 
2671 	error = makefh4(&cs->fh, vp, cs->exi);
2672 
2673 err_out:
2674 	if (error) {
2675 		if (is_newvp) {
2676 			VN_RELE(cs->vp);
2677 			cs->vp = oldvp;
2678 		} else
2679 			VN_RELE(vp);
2680 		return (puterrno4(error));
2681 	}
2682 
2683 	if (!is_newvp) {
2684 		if (cs->vp)
2685 			VN_RELE(cs->vp);
2686 		cs->vp = vp;
2687 	} else if (oldvp)
2688 		VN_RELE(oldvp);
2689 
2690 	/*
2691 	 * if did lookup on attrdir and didn't lookup .., set named
2692 	 * attr fh flag
2693 	 */
2694 	if (attrdir && ! dotdot)
2695 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
2696 
2697 	/* Assume false for now, open proc will set this */
2698 	cs->mandlock = FALSE;
2699 
2700 	return (NFS4_OK);
2701 }
2702 
2703 /* ARGSUSED */
2704 static void
2705 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2706 	struct compound_state *cs)
2707 {
2708 	LOOKUP4args *args = &argop->nfs_argop4_u.oplookup;
2709 	LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup;
2710 	char *nm;
2711 	uint_t len;
2712 
2713 	if (cs->vp == NULL) {
2714 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2715 		return;
2716 	}
2717 
2718 	if (cs->vp->v_type == VLNK) {
2719 		*cs->statusp = resp->status = NFS4ERR_SYMLINK;
2720 		return;
2721 	}
2722 
2723 	if (cs->vp->v_type != VDIR) {
2724 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2725 		return;
2726 	}
2727 
2728 	if (!utf8_dir_verify(&args->objname)) {
2729 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2730 		return;
2731 	}
2732 
2733 	nm = utf8_to_str(&args->objname, &len, NULL);
2734 	if (nm == NULL) {
2735 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2736 		return;
2737 	}
2738 
2739 	if (len > MAXNAMELEN) {
2740 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
2741 		kmem_free(nm, len);
2742 		return;
2743 	}
2744 
2745 	*cs->statusp = resp->status = do_rfs4_op_lookup(nm, len, req, cs);
2746 
2747 	kmem_free(nm, len);
2748 }
2749 
2750 /* ARGSUSED */
2751 static void
2752 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
2753 	struct compound_state *cs)
2754 {
2755 	LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp;
2756 
2757 	if (cs->vp == NULL) {
2758 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2759 		return;
2760 	}
2761 
2762 	if (cs->vp->v_type != VDIR) {
2763 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2764 		return;
2765 	}
2766 
2767 	*cs->statusp = resp->status = do_rfs4_op_lookup("..", 3, req, cs);
2768 
2769 	/*
2770 	 * From NFSV4 Specification, LOOKUPP should not check for
2771 	 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead.
2772 	 */
2773 	if (resp->status == NFS4ERR_WRONGSEC) {
2774 		*cs->statusp = resp->status = NFS4_OK;
2775 	}
2776 }
2777 
2778 
2779 /*ARGSUSED2*/
2780 static void
2781 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2782 	struct compound_state *cs)
2783 {
2784 	OPENATTR4args	*args = &argop->nfs_argop4_u.opopenattr;
2785 	OPENATTR4res	*resp = &resop->nfs_resop4_u.opopenattr;
2786 	vnode_t		*avp = NULL;
2787 	int		lookup_flags = LOOKUP_XATTR, error;
2788 	int		exp_ro = 0;
2789 
2790 	if (cs->vp == NULL) {
2791 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2792 		return;
2793 	}
2794 
2795 	/*
2796 	 * Make a couple of checks made by copen()
2797 	 *
2798 	 * Check to make sure underlying fs supports xattrs.  This
2799 	 * is required because solaris filesystem implementations
2800 	 * (UFS/TMPFS) don't enforce the noxattr mount option
2801 	 * in VOP_LOOKUP(LOOKUP_XATTR).  If fs doesn't support this
2802 	 * pathconf cmd or if fs supports cmd but doesn't claim
2803 	 * support for xattr, return NOTSUPP.  It would be better
2804 	 * to use VOP_PATHCONF( _PC_XATTR_ENABLED) for this; however,
2805 	 * that cmd is not available to VOP_PATHCONF interface
2806 	 * (it's only implemented inside pathconf syscall)...
2807 	 *
2808 	 * Verify permission to put attributes on files (access
2809 	 * checks from copen).
2810 	 */
2811 
2812 	if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0 &&
2813 	    !vfs_has_feature(cs->vp->v_vfsp, VFSFT_XVATTR)) {
2814 		error = ENOTSUP;
2815 		goto error_out;
2816 	}
2817 
2818 	if ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr, NULL) != 0) &&
2819 	    (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr, NULL) != 0) &&
2820 	    (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr, NULL) != 0)) {
2821 		error = EACCES;
2822 		goto error_out;
2823 	}
2824 
2825 	/*
2826 	 * The CREATE_XATTR_DIR VOP flag cannot be specified if
2827 	 * the file system is exported read-only -- regardless of
2828 	 * createdir flag.  Otherwise the attrdir would be created
2829 	 * (assuming server fs isn't mounted readonly locally).  If
2830 	 * VOP_LOOKUP returns ENOENT in this case, the error will
2831 	 * be translated into EROFS.  ENOSYS is mapped to ENOTSUP
2832 	 * because specfs has no VOP_LOOKUP op, so the macro would
2833 	 * return ENOSYS.  EINVAL is returned by all (current)
2834 	 * Solaris file system implementations when any of their
2835 	 * restrictions are violated (xattr(dir) can't have xattrdir).
2836 	 * Returning NOTSUPP is more appropriate in this case
2837 	 * because the object will never be able to have an attrdir.
2838 	 */
2839 	if (args->createdir && ! (exp_ro = rdonly4(cs->exi, cs->vp, req)))
2840 		lookup_flags |= CREATE_XATTR_DIR;
2841 
2842 	error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr,
2843 	    NULL, NULL, NULL);
2844 
2845 	if (error) {
2846 		if (error == ENOENT && args->createdir && exp_ro)
2847 			error = EROFS;
2848 		else if (error == EINVAL || error == ENOSYS)
2849 			error = ENOTSUP;
2850 		goto error_out;
2851 	}
2852 
2853 	ASSERT(avp->v_flag & V_XATTRDIR);
2854 
2855 	error = makefh4(&cs->fh, avp, cs->exi);
2856 
2857 	if (error) {
2858 		VN_RELE(avp);
2859 		goto error_out;
2860 	}
2861 
2862 	VN_RELE(cs->vp);
2863 	cs->vp = avp;
2864 
2865 	/*
2866 	 * There is no requirement for an attrdir fh flag
2867 	 * because the attrdir has a vnode flag to distinguish
2868 	 * it from regular (non-xattr) directories.  The
2869 	 * FH4_ATTRDIR flag is set for future sanity checks.
2870 	 */
2871 	set_fh4_flag(&cs->fh, FH4_ATTRDIR);
2872 	*cs->statusp = resp->status = NFS4_OK;
2873 	return;
2874 
2875 error_out:
2876 
2877 	*cs->statusp = resp->status = puterrno4(error);
2878 }
2879 
2880 static int
2881 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred,
2882     caller_context_t *ct)
2883 {
2884 	int error;
2885 	int i;
2886 	clock_t delaytime;
2887 
2888 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
2889 
2890 	/*
2891 	 * Don't block on mandatory locks. If this routine returns
2892 	 * EAGAIN, the caller should return NFS4ERR_LOCKED.
2893 	 */
2894 	uio->uio_fmode = FNONBLOCK;
2895 
2896 	for (i = 0; i < rfs4_maxlock_tries; i++) {
2897 
2898 
2899 		if (direction == FREAD) {
2900 			(void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, ct);
2901 			error = VOP_READ(vp, uio, ioflag, cred, ct);
2902 			VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, ct);
2903 		} else {
2904 			(void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, ct);
2905 			error = VOP_WRITE(vp, uio, ioflag, cred, ct);
2906 			VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, ct);
2907 		}
2908 
2909 		if (error != EAGAIN)
2910 			break;
2911 
2912 		if (i < rfs4_maxlock_tries - 1) {
2913 			delay(delaytime);
2914 			delaytime *= 2;
2915 		}
2916 	}
2917 
2918 	return (error);
2919 }
2920 
2921 /* ARGSUSED */
2922 static void
2923 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2924 	struct compound_state *cs)
2925 {
2926 	READ4args *args = &argop->nfs_argop4_u.opread;
2927 	READ4res *resp = &resop->nfs_resop4_u.opread;
2928 	int error;
2929 	int verror;
2930 	vnode_t *vp;
2931 	struct vattr va;
2932 	struct iovec iov;
2933 	struct uio uio;
2934 	u_offset_t offset;
2935 	bool_t *deleg = &cs->deleg;
2936 	nfsstat4 stat;
2937 	int in_crit = 0;
2938 	mblk_t *mp;
2939 	int alloc_err = 0;
2940 	caller_context_t ct;
2941 
2942 	vp = cs->vp;
2943 	if (vp == NULL) {
2944 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2945 		return;
2946 	}
2947 	if (cs->access == CS_ACCESS_DENIED) {
2948 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2949 		return;
2950 	}
2951 
2952 	if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE,
2953 	    deleg, TRUE, &ct)) != NFS4_OK) {
2954 		*cs->statusp = resp->status = stat;
2955 		goto out;
2956 	}
2957 
2958 	/*
2959 	 * Enter the critical region before calling VOP_RWLOCK
2960 	 * to avoid a deadlock with write requests.
2961 	 */
2962 	if (nbl_need_check(vp)) {
2963 		nbl_start_crit(vp, RW_READER);
2964 		in_crit = 1;
2965 		if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0,
2966 		    &ct)) {
2967 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
2968 			goto out;
2969 		}
2970 	}
2971 
2972 	if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE,
2973 	    deleg, TRUE, &ct)) != NFS4_OK) {
2974 		*cs->statusp = resp->status = stat;
2975 		goto out;
2976 	}
2977 
2978 	va.va_mask = AT_MODE|AT_SIZE|AT_UID;
2979 	verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct);
2980 
2981 	/*
2982 	 * If we can't get the attributes, then we can't do the
2983 	 * right access checking.  So, we'll fail the request.
2984 	 */
2985 	if (verror) {
2986 		*cs->statusp = resp->status = puterrno4(verror);
2987 		goto out;
2988 	}
2989 
2990 	if (vp->v_type != VREG) {
2991 		*cs->statusp = resp->status =
2992 		    ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
2993 		goto out;
2994 	}
2995 
2996 	if (crgetuid(cs->cr) != va.va_uid &&
2997 	    (error = VOP_ACCESS(vp, VREAD, 0, cs->cr, &ct)) &&
2998 	    (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr, &ct))) {
2999 		*cs->statusp = resp->status = puterrno4(error);
3000 		goto out;
3001 	}
3002 
3003 	if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */
3004 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3005 		goto out;
3006 	}
3007 
3008 	offset = args->offset;
3009 	if (offset >= va.va_size) {
3010 		*cs->statusp = resp->status = NFS4_OK;
3011 		resp->eof = TRUE;
3012 		resp->data_len = 0;
3013 		resp->data_val = NULL;
3014 		resp->mblk = NULL;
3015 		*cs->statusp = resp->status = NFS4_OK;
3016 		goto out;
3017 	}
3018 
3019 	if (args->count == 0) {
3020 		*cs->statusp = resp->status = NFS4_OK;
3021 		resp->eof = FALSE;
3022 		resp->data_len = 0;
3023 		resp->data_val = NULL;
3024 		resp->mblk = NULL;
3025 		goto out;
3026 	}
3027 
3028 	/*
3029 	 * Do not allocate memory more than maximum allowed
3030 	 * transfer size
3031 	 */
3032 	if (args->count > rfs4_tsize(req))
3033 		args->count = rfs4_tsize(req);
3034 
3035 	/*
3036 	 * mp will contain the data to be sent out in the read reply.
3037 	 * It will be freed after the reply has been sent.
3038 	 * Let's roundup the data to a BYTES_PER_XDR_UNIT multiple,
3039 	 * so that the call to xdrmblk_putmblk() never fails.
3040 	 * If the first alloc of the requested size fails, then
3041 	 * decrease the size to something more reasonable and wait
3042 	 * for the allocation to occur.
3043 	 */
3044 	mp = allocb(RNDUP(args->count), BPRI_MED);
3045 	if (mp == NULL) {
3046 		if (args->count > MAXBSIZE)
3047 			args->count = MAXBSIZE;
3048 		mp = allocb_wait(RNDUP(args->count), BPRI_MED,
3049 		    STR_NOSIG, &alloc_err);
3050 	}
3051 	ASSERT(mp != NULL);
3052 	ASSERT(alloc_err == 0);
3053 
3054 	iov.iov_base = (caddr_t)mp->b_datap->db_base;
3055 	iov.iov_len = args->count;
3056 	uio.uio_iov = &iov;
3057 	uio.uio_iovcnt = 1;
3058 	uio.uio_segflg = UIO_SYSSPACE;
3059 	uio.uio_extflg = UIO_COPY_CACHED;
3060 	uio.uio_loffset = args->offset;
3061 	uio.uio_resid = args->count;
3062 
3063 	error = do_io(FREAD, vp, &uio, 0, cs->cr, &ct);
3064 
3065 	va.va_mask = AT_SIZE;
3066 	verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct);
3067 
3068 	if (error) {
3069 		freeb(mp);
3070 		*cs->statusp = resp->status = puterrno4(error);
3071 		goto out;
3072 	}
3073 
3074 	*cs->statusp = resp->status = NFS4_OK;
3075 
3076 	ASSERT(uio.uio_resid >= 0);
3077 	resp->data_len = args->count - uio.uio_resid;
3078 	resp->data_val = (char *)mp->b_datap->db_base;
3079 	resp->mblk = mp;
3080 
3081 	if (!verror && offset + resp->data_len == va.va_size)
3082 		resp->eof = TRUE;
3083 	else
3084 		resp->eof = FALSE;
3085 
3086 out:
3087 	if (in_crit)
3088 		nbl_end_crit(vp);
3089 }
3090 
3091 static void
3092 rfs4_op_read_free(nfs_resop4 *resop)
3093 {
3094 	READ4res *resp = &resop->nfs_resop4_u.opread;
3095 
3096 	if (resp->status == NFS4_OK && resp->mblk != NULL) {
3097 		freeb(resp->mblk);
3098 		resp->mblk = NULL;
3099 		resp->data_val = NULL;
3100 		resp->data_len = 0;
3101 	}
3102 }
3103 
3104 static void
3105 rfs4_op_readdir_free(nfs_resop4 *resop)
3106 {
3107 	READDIR4res *resp = &resop->nfs_resop4_u.opreaddir;
3108 
3109 	if (resp->status == NFS4_OK && resp->mblk != NULL) {
3110 		freeb(resp->mblk);
3111 		resp->mblk = NULL;
3112 		resp->data_len = 0;
3113 	}
3114 }
3115 
3116 
3117 /* ARGSUSED */
3118 static void
3119 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
3120 	struct compound_state *cs)
3121 {
3122 	PUTPUBFH4res *resp = &resop->nfs_resop4_u.opputpubfh;
3123 	int error;
3124 	vnode_t *vp;
3125 	struct exportinfo *exi, *sav_exi;
3126 	nfs_fh4_fmt_t *fh_fmtp;
3127 
3128 	if (cs->vp) {
3129 		VN_RELE(cs->vp);
3130 		cs->vp = NULL;
3131 	}
3132 
3133 	if (cs->cr)
3134 		crfree(cs->cr);
3135 
3136 	cs->cr = crdup(cs->basecr);
3137 
3138 	vp = exi_public->exi_vp;
3139 	if (vp == NULL) {
3140 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
3141 		return;
3142 	}
3143 
3144 	if (is_system_labeled()) {
3145 		bslabel_t *clabel;
3146 
3147 		ASSERT(req->rq_label != NULL);
3148 		clabel = req->rq_label;
3149 		DTRACE_PROBE2(tx__rfs4__log__info__opputpubfh__clabel, char *,
3150 		    "got client label from request(1)",
3151 		    struct svc_req *, req);
3152 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
3153 			if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) {
3154 				*cs->statusp = resp->status =
3155 				    NFS4ERR_SERVERFAULT;
3156 				return;
3157 			}
3158 		}
3159 	}
3160 
3161 	error = makefh4(&cs->fh, vp, exi_public);
3162 	if (error != 0) {
3163 		*cs->statusp = resp->status = puterrno4(error);
3164 		return;
3165 	}
3166 	sav_exi = cs->exi;
3167 	if (exi_public == exi_root) {
3168 		/*
3169 		 * No filesystem is actually shared public, so we default
3170 		 * to exi_root. In this case, we must check whether root
3171 		 * is exported.
3172 		 */
3173 		fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val;
3174 
3175 		/*
3176 		 * if root filesystem is exported, the exportinfo struct that we
3177 		 * should use is what checkexport4 returns, because root_exi is
3178 		 * actually a mostly empty struct.
3179 		 */
3180 		exi = checkexport4(&fh_fmtp->fh4_fsid,
3181 		    (fid_t *)&fh_fmtp->fh4_xlen, NULL);
3182 		cs->exi = ((exi != NULL) ? exi : exi_public);
3183 	} else {
3184 		/*
3185 		 * it's a properly shared filesystem
3186 		 */
3187 		cs->exi = exi_public;
3188 	}
3189 
3190 	VN_HOLD(vp);
3191 	cs->vp = vp;
3192 
3193 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3194 		VN_RELE(cs->vp);
3195 		cs->vp = NULL;
3196 		cs->exi = sav_exi;
3197 		return;
3198 	}
3199 
3200 	*cs->statusp = resp->status = NFS4_OK;
3201 }
3202 
3203 /*
3204  * XXX - issue with put*fh operations. Suppose /export/home is exported.
3205  * Suppose an NFS client goes to mount /export/home/joe. If /export, home,
3206  * or joe have restrictive search permissions, then we shouldn't let
3207  * the client get a file handle. This is easy to enforce. However, we
3208  * don't know what security flavor should be used until we resolve the
3209  * path name. Another complication is uid mapping. If root is
3210  * the user, then it will be mapped to the anonymous user by default,
3211  * but we won't know that till we've resolved the path name. And we won't
3212  * know what the anonymous user is.
3213  * Luckily, SECINFO is specified to take a full filename.
3214  * So what we will have to in rfs4_op_lookup is check that flavor of
3215  * the target object matches that of the request, and if root was the
3216  * caller, check for the root= and anon= options, and if necessary,
3217  * repeat the lookup using the right cred_t. But that's not done yet.
3218  */
3219 /* ARGSUSED */
3220 static void
3221 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3222 	struct compound_state *cs)
3223 {
3224 	PUTFH4args *args = &argop->nfs_argop4_u.opputfh;
3225 	PUTFH4res *resp = &resop->nfs_resop4_u.opputfh;
3226 	nfs_fh4_fmt_t *fh_fmtp;
3227 
3228 	if (cs->vp) {
3229 		VN_RELE(cs->vp);
3230 		cs->vp = NULL;
3231 	}
3232 
3233 	if (cs->cr) {
3234 		crfree(cs->cr);
3235 		cs->cr = NULL;
3236 	}
3237 
3238 
3239 	if (args->object.nfs_fh4_len < NFS_FH4_LEN) {
3240 		*cs->statusp = resp->status = NFS4ERR_BADHANDLE;
3241 		return;
3242 	}
3243 
3244 	fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val;
3245 	cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen,
3246 	    NULL);
3247 
3248 	if (cs->exi == NULL) {
3249 		*cs->statusp = resp->status = NFS4ERR_STALE;
3250 		return;
3251 	}
3252 
3253 	cs->cr = crdup(cs->basecr);
3254 
3255 	ASSERT(cs->cr != NULL);
3256 
3257 	if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) {
3258 		*cs->statusp = resp->status;
3259 		return;
3260 	}
3261 
3262 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3263 		VN_RELE(cs->vp);
3264 		cs->vp = NULL;
3265 		return;
3266 	}
3267 
3268 	nfs_fh4_copy(&args->object, &cs->fh);
3269 	*cs->statusp = resp->status = NFS4_OK;
3270 	cs->deleg = FALSE;
3271 }
3272 
3273 /* ARGSUSED */
3274 static void
3275 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3276 	struct compound_state *cs)
3277 
3278 {
3279 	PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh;
3280 	int error;
3281 	fid_t fid;
3282 	struct exportinfo *exi, *sav_exi;
3283 
3284 	if (cs->vp) {
3285 		VN_RELE(cs->vp);
3286 		cs->vp = NULL;
3287 	}
3288 
3289 	if (cs->cr)
3290 		crfree(cs->cr);
3291 
3292 	cs->cr = crdup(cs->basecr);
3293 
3294 	/*
3295 	 * Using rootdir, the system root vnode,
3296 	 * get its fid.
3297 	 */
3298 	bzero(&fid, sizeof (fid));
3299 	fid.fid_len = MAXFIDSZ;
3300 	error = vop_fid_pseudo(rootdir, &fid);
3301 	if (error != 0) {
3302 		*cs->statusp = resp->status = puterrno4(error);
3303 		return;
3304 	}
3305 
3306 	/*
3307 	 * Then use the root fsid & fid it to find out if it's exported
3308 	 *
3309 	 * If the server root isn't exported directly, then
3310 	 * it should at least be a pseudo export based on
3311 	 * one or more exports further down in the server's
3312 	 * file tree.
3313 	 */
3314 	exi = checkexport4(&rootdir->v_vfsp->vfs_fsid, &fid, NULL);
3315 	if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) {
3316 		NFS4_DEBUG(rfs4_debug,
3317 			(CE_WARN, "rfs4_op_putrootfh: export check failure"));
3318 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
3319 		return;
3320 	}
3321 
3322 	/*
3323 	 * Now make a filehandle based on the root
3324 	 * export and root vnode.
3325 	 */
3326 	error = makefh4(&cs->fh, rootdir, exi);
3327 	if (error != 0) {
3328 		*cs->statusp = resp->status = puterrno4(error);
3329 		return;
3330 	}
3331 
3332 	sav_exi = cs->exi;
3333 	cs->exi = exi;
3334 
3335 	VN_HOLD(rootdir);
3336 	cs->vp = rootdir;
3337 
3338 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3339 		VN_RELE(rootdir);
3340 		cs->vp = NULL;
3341 		cs->exi = sav_exi;
3342 		return;
3343 	}
3344 
3345 	*cs->statusp = resp->status = NFS4_OK;
3346 	cs->deleg = FALSE;
3347 }
3348 
3349 /*
3350  * A directory entry is a valid nfsv4 entry if
3351  * - it has a non-zero ino
3352  * - it is not a dot or dotdot name
3353  * - it is visible in a pseudo export or in a real export that can
3354  *   only have a limited view.
3355  */
3356 static bool_t
3357 valid_nfs4_entry(struct exportinfo *exi, struct dirent64 *dp,
3358 		int *expseudo, int check_visible)
3359 {
3360 	if (dp->d_ino == 0 || NFS_IS_DOTNAME(dp->d_name)) {
3361 		*expseudo = 0;
3362 		return (FALSE);
3363 	}
3364 
3365 	if (! check_visible) {
3366 		*expseudo = 0;
3367 		return (TRUE);
3368 	}
3369 
3370 	return (nfs_visible_inode(exi, dp->d_ino, expseudo));
3371 }
3372 
3373 /*
3374  * set_rdattr_params sets up the variables used to manage what information
3375  * to get for each directory entry.
3376  */
3377 static nfsstat4
3378 set_rdattr_params(struct nfs4_svgetit_arg *sargp,
3379 		bitmap4 attrs, bool_t *need_to_lookup)
3380 {
3381 	uint_t	va_mask;
3382 	nfsstat4 status;
3383 	bitmap4 objbits;
3384 
3385 	status = bitmap4_to_attrmask(attrs, sargp);
3386 	if (status != NFS4_OK) {
3387 		/*
3388 		 * could not even figure attr mask
3389 		 */
3390 		return (status);
3391 	}
3392 	va_mask = sargp->vap->va_mask;
3393 
3394 	/*
3395 	 * dirent's d_ino is always correct value for mounted_on_fileid.
3396 	 * mntdfid_set is set once here, but mounted_on_fileid is
3397 	 * set in main dirent processing loop for each dirent.
3398 	 * The mntdfid_set is a simple optimization that lets the
3399 	 * server attr code avoid work when caller is readdir.
3400 	 */
3401 	sargp->mntdfid_set = TRUE;
3402 
3403 	/*
3404 	 * Lookup entry only if client asked for any of the following:
3405 	 * a) vattr attrs
3406 	 * b) vfs attrs
3407 	 * c) attrs w/per-object scope requested (change, filehandle, etc)
3408 	 *    other than mounted_on_fileid (which we can take from dirent)
3409 	 */
3410 	objbits = attrs ? attrs & NFS4_VP_ATTR_MASK : 0;
3411 
3412 	if (va_mask || sargp->sbp || (objbits & ~FATTR4_MOUNTED_ON_FILEID_MASK))
3413 		*need_to_lookup = TRUE;
3414 	else
3415 		*need_to_lookup = FALSE;
3416 
3417 	if (sargp->sbp == NULL)
3418 		return (NFS4_OK);
3419 
3420 	/*
3421 	 * If filesystem attrs are requested, get them now from the
3422 	 * directory vp, as most entries will have same filesystem. The only
3423 	 * exception are mounted over entries but we handle
3424 	 * those as we go (XXX mounted over detection not yet implemented).
3425 	 */
3426 	sargp->vap->va_mask = 0;	/* to avoid VOP_GETATTR */
3427 	status = bitmap4_get_sysattrs(sargp);
3428 	sargp->vap->va_mask = va_mask;
3429 
3430 	if ((status != NFS4_OK) && sargp->rdattr_error_req) {
3431 		/*
3432 		 * Failed to get filesystem attributes.
3433 		 * Return a rdattr_error for each entry, but don't fail.
3434 		 * However, don't get any obj-dependent attrs.
3435 		 */
3436 		sargp->rdattr_error = status;	/* for rdattr_error */
3437 		*need_to_lookup = FALSE;
3438 		/*
3439 		 * At least get fileid for regular readdir output
3440 		 */
3441 		sargp->vap->va_mask &= AT_NODEID;
3442 		status = NFS4_OK;
3443 	}
3444 
3445 	return (status);
3446 }
3447 
3448 /*
3449  * readlink: args: CURRENT_FH.
3450  *	res: status. If success - CURRENT_FH unchanged, return linktext.
3451  */
3452 
3453 /* ARGSUSED */
3454 static void
3455 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3456 	struct compound_state *cs)
3457 {
3458 	READLINK4res *resp = &resop->nfs_resop4_u.opreadlink;
3459 	int error;
3460 	vnode_t *vp;
3461 	struct iovec iov;
3462 	struct vattr va;
3463 	struct uio uio;
3464 	char *data;
3465 
3466 	/* CURRENT_FH: directory */
3467 	vp = cs->vp;
3468 	if (vp == NULL) {
3469 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3470 		return;
3471 	}
3472 
3473 	if (cs->access == CS_ACCESS_DENIED) {
3474 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3475 		return;
3476 	}
3477 
3478 	if (vp->v_type == VDIR) {
3479 		*cs->statusp = resp->status = NFS4ERR_ISDIR;
3480 		return;
3481 	}
3482 
3483 	if (vp->v_type != VLNK) {
3484 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3485 		return;
3486 	}
3487 
3488 	va.va_mask = AT_MODE;
3489 	error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL);
3490 	if (error) {
3491 		*cs->statusp = resp->status = puterrno4(error);
3492 		return;
3493 	}
3494 
3495 	if (MANDLOCK(vp, va.va_mode)) {
3496 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3497 		return;
3498 	}
3499 
3500 	data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP);
3501 
3502 	iov.iov_base = data;
3503 	iov.iov_len = MAXPATHLEN;
3504 	uio.uio_iov = &iov;
3505 	uio.uio_iovcnt = 1;
3506 	uio.uio_segflg = UIO_SYSSPACE;
3507 	uio.uio_extflg = UIO_COPY_CACHED;
3508 	uio.uio_loffset = 0;
3509 	uio.uio_resid = MAXPATHLEN;
3510 
3511 	error = VOP_READLINK(vp, &uio, cs->cr, NULL);
3512 
3513 	if (error) {
3514 		kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1);
3515 		*cs->statusp = resp->status = puterrno4(error);
3516 		return;
3517 	}
3518 
3519 	*(data + MAXPATHLEN - uio.uio_resid) = '\0';
3520 
3521 	/*
3522 	 * treat link name as data
3523 	 */
3524 	(void) str_to_utf8(data, &resp->link);
3525 
3526 	kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1);
3527 	*cs->statusp = resp->status = NFS4_OK;
3528 }
3529 
3530 static void
3531 rfs4_op_readlink_free(nfs_resop4 *resop)
3532 {
3533 	READLINK4res *resp = &resop->nfs_resop4_u.opreadlink;
3534 	utf8string *symlink = &resp->link;
3535 
3536 	if (symlink->utf8string_val) {
3537 		UTF8STRING_FREE(*symlink)
3538 	}
3539 }
3540 
3541 /*
3542  * release_lockowner:
3543  *	Release any state associated with the supplied
3544  *	lockowner. Note if any lo_state is holding locks we will not
3545  *	rele that lo_state and thus the lockowner will not be destroyed.
3546  *	A client using lock after the lock owner stateid has been released
3547  *	will suffer the consequence of NFS4ERR_BAD_STATEID and would have
3548  *	to reissue the lock with new_lock_owner set to TRUE.
3549  *	args: lock_owner
3550  *	res:  status
3551  */
3552 /* ARGSUSED */
3553 static void
3554 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop,
3555 	struct svc_req *req, struct compound_state *cs)
3556 {
3557 	RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner;
3558 	RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner;
3559 	rfs4_lockowner_t *lo;
3560 	rfs4_openowner_t *oop;
3561 	rfs4_state_t *sp;
3562 	rfs4_lo_state_t *lsp;
3563 	rfs4_client_t *cp;
3564 	bool_t create = FALSE;
3565 	locklist_t *llist;
3566 	sysid_t sysid;
3567 
3568 	/* Make sure there is a clientid around for this request */
3569 	cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE);
3570 
3571 	if (cp == NULL) {
3572 		*cs->statusp = resp->status =
3573 		    rfs4_check_clientid(&ap->lock_owner.clientid, 0);
3574 		return;
3575 	}
3576 	rfs4_client_rele(cp);
3577 
3578 	lo = rfs4_findlockowner(&ap->lock_owner, &create);
3579 	if (lo == NULL) {
3580 		*cs->statusp = resp->status = NFS4_OK;
3581 		return;
3582 	}
3583 	ASSERT(lo->client != NULL);
3584 
3585 	/*
3586 	 * Check for EXPIRED client. If so will reap state with in a lease
3587 	 * period or on next set_clientid_confirm step
3588 	 */
3589 	if (rfs4_lease_expired(lo->client)) {
3590 		rfs4_lockowner_rele(lo);
3591 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
3592 		return;
3593 	}
3594 
3595 	/*
3596 	 * If no sysid has been assigned, then no locks exist; just return.
3597 	 */
3598 	rfs4_dbe_lock(lo->client->dbe);
3599 	if (lo->client->sysidt == LM_NOSYSID) {
3600 		rfs4_lockowner_rele(lo);
3601 		rfs4_dbe_unlock(lo->client->dbe);
3602 		return;
3603 	}
3604 
3605 	sysid = lo->client->sysidt;
3606 	rfs4_dbe_unlock(lo->client->dbe);
3607 
3608 	/*
3609 	 * Mark the lockowner invalid.
3610 	 */
3611 	rfs4_dbe_hide(lo->dbe);
3612 
3613 	/*
3614 	 * sysid-pid pair should now not be used since the lockowner is
3615 	 * invalid. If the client were to instantiate the lockowner again
3616 	 * it would be assigned a new pid. Thus we can get the list of
3617 	 * current locks.
3618 	 */
3619 
3620 	llist = flk_get_active_locks(sysid, lo->pid);
3621 	/* If we are still holding locks fail */
3622 	if (llist != NULL) {
3623 
3624 		*cs->statusp = resp->status = NFS4ERR_LOCKS_HELD;
3625 
3626 		flk_free_locklist(llist);
3627 		/*
3628 		 * We need to unhide the lockowner so the client can
3629 		 * try it again. The bad thing here is if the client
3630 		 * has a logic error that took it here in the first place
3631 		 * he probably has lost accounting of the locks that it
3632 		 * is holding. So we may have dangling state until the
3633 		 * open owner state is reaped via close. One scenario
3634 		 * that could possibly occur is that the client has
3635 		 * sent the unlock request(s) in separate threads
3636 		 * and has not waited for the replies before sending the
3637 		 * RELEASE_LOCKOWNER request. Presumably, it would expect
3638 		 * and deal appropriately with NFS4ERR_LOCKS_HELD, by
3639 		 * reissuing the request.
3640 		 */
3641 		rfs4_dbe_unhide(lo->dbe);
3642 		rfs4_lockowner_rele(lo);
3643 		return;
3644 	}
3645 
3646 	/*
3647 	 * For the corresponding client we need to check each open
3648 	 * owner for any opens that have lockowner state associated
3649 	 * with this lockowner.
3650 	 */
3651 
3652 	rfs4_dbe_lock(lo->client->dbe);
3653 	for (oop = lo->client->openownerlist.next->oop; oop != NULL;
3654 	    oop = oop->openownerlist.next->oop) {
3655 
3656 		rfs4_dbe_lock(oop->dbe);
3657 		for (sp = oop->ownerstateids.next->sp; sp != NULL;
3658 		    sp = sp->ownerstateids.next->sp) {
3659 
3660 			rfs4_dbe_lock(sp->dbe);
3661 			for (lsp = sp->lockownerlist.next->lsp;
3662 			    lsp != NULL; lsp = lsp->lockownerlist.next->lsp) {
3663 				if (lsp->locker == lo) {
3664 					rfs4_dbe_lock(lsp->dbe);
3665 					rfs4_dbe_invalidate(lsp->dbe);
3666 					rfs4_dbe_unlock(lsp->dbe);
3667 				}
3668 			}
3669 			rfs4_dbe_unlock(sp->dbe);
3670 		}
3671 		rfs4_dbe_unlock(oop->dbe);
3672 	}
3673 	rfs4_dbe_unlock(lo->client->dbe);
3674 
3675 	rfs4_lockowner_rele(lo);
3676 
3677 	*cs->statusp = resp->status = NFS4_OK;
3678 }
3679 
3680 /*
3681  * short utility function to lookup a file and recall the delegation
3682  */
3683 static rfs4_file_t *
3684 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp,
3685 	int *lkup_error, cred_t *cr)
3686 {
3687 	vnode_t *vp;
3688 	rfs4_file_t *fp = NULL;
3689 	bool_t fcreate = FALSE;
3690 	int error;
3691 
3692 	if (vpp)
3693 		*vpp = NULL;
3694 
3695 	if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, NULL, NULL,
3696 	    NULL)) == 0) {
3697 		if (vp->v_type == VREG)
3698 			fp = rfs4_findfile(vp, NULL, &fcreate);
3699 		if (vpp)
3700 			*vpp = vp;
3701 		else
3702 			VN_RELE(vp);
3703 	}
3704 
3705 	if (lkup_error)
3706 		*lkup_error = error;
3707 
3708 	return (fp);
3709 }
3710 
3711 /*
3712  * remove: args: CURRENT_FH: directory; name.
3713  *	res: status. If success - CURRENT_FH unchanged, return change_info
3714  *		for directory.
3715  */
3716 /* ARGSUSED */
3717 static void
3718 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3719 	struct compound_state *cs)
3720 {
3721 	REMOVE4args *args = &argop->nfs_argop4_u.opremove;
3722 	REMOVE4res *resp = &resop->nfs_resop4_u.opremove;
3723 	int error;
3724 	vnode_t *dvp, *vp;
3725 	struct vattr bdva, idva, adva;
3726 	char *nm;
3727 	uint_t len;
3728 	rfs4_file_t *fp;
3729 	int in_crit = 0;
3730 	bslabel_t *clabel;
3731 
3732 	/* CURRENT_FH: directory */
3733 	dvp = cs->vp;
3734 	if (dvp == NULL) {
3735 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3736 		return;
3737 	}
3738 
3739 	if (cs->access == CS_ACCESS_DENIED) {
3740 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3741 		return;
3742 	}
3743 
3744 	/*
3745 	 * If there is an unshared filesystem mounted on this vnode,
3746 	 * Do not allow to remove anything in this directory.
3747 	 */
3748 	if (vn_ismntpt(dvp)) {
3749 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3750 		return;
3751 	}
3752 
3753 	if (dvp->v_type != VDIR) {
3754 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
3755 		return;
3756 	}
3757 
3758 	if (!utf8_dir_verify(&args->target)) {
3759 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3760 		return;
3761 	}
3762 
3763 	/*
3764 	 * Lookup the file so that we can check if it's a directory
3765 	 */
3766 	nm = utf8_to_fn(&args->target, &len, NULL);
3767 	if (nm == NULL) {
3768 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3769 		return;
3770 	}
3771 
3772 	if (len > MAXNAMELEN) {
3773 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
3774 		kmem_free(nm, len);
3775 		return;
3776 	}
3777 
3778 	if (rdonly4(cs->exi, cs->vp, req)) {
3779 		*cs->statusp = resp->status = NFS4ERR_ROFS;
3780 		kmem_free(nm, len);
3781 		return;
3782 	}
3783 
3784 	/*
3785 	 * Lookup the file to determine type and while we are see if
3786 	 * there is a file struct around and check for delegation.
3787 	 * We don't need to acquire va_seq before this lookup, if
3788 	 * it causes an update, cinfo.before will not match, which will
3789 	 * trigger a cache flush even if atomic is TRUE.
3790 	 */
3791 	if (fp = rfs4_lookup_and_findfile(dvp, nm, &vp, &error, cs->cr)) {
3792 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
3793 		    NULL)) {
3794 			VN_RELE(vp);
3795 			rfs4_file_rele(fp);
3796 			*cs->statusp = resp->status = NFS4ERR_DELAY;
3797 			kmem_free(nm, len);
3798 			return;
3799 		}
3800 	}
3801 
3802 	/* Didn't find anything to remove */
3803 	if (vp == NULL) {
3804 		*cs->statusp = resp->status = error;
3805 		kmem_free(nm, len);
3806 		return;
3807 	}
3808 
3809 	if (nbl_need_check(vp)) {
3810 		nbl_start_crit(vp, RW_READER);
3811 		in_crit = 1;
3812 		if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0, NULL)) {
3813 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
3814 			kmem_free(nm, len);
3815 			nbl_end_crit(vp);
3816 			VN_RELE(vp);
3817 			if (fp) {
3818 				rfs4_clear_dont_grant(fp);
3819 				rfs4_file_rele(fp);
3820 			}
3821 			return;
3822 		}
3823 	}
3824 
3825 	/* check label before allowing removal */
3826 	if (is_system_labeled()) {
3827 		ASSERT(req->rq_label != NULL);
3828 		clabel = req->rq_label;
3829 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
3830 		    "got client label from request(1)",
3831 		    struct svc_req *, req);
3832 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
3833 			if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK)) {
3834 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
3835 				kmem_free(nm, len);
3836 				if (in_crit)
3837 					nbl_end_crit(vp);
3838 				VN_RELE(vp);
3839 				if (fp) {
3840 					rfs4_clear_dont_grant(fp);
3841 					rfs4_file_rele(fp);
3842 				}
3843 				return;
3844 			}
3845 		}
3846 	}
3847 
3848 	/* Get dir "before" change value */
3849 	bdva.va_mask = AT_CTIME|AT_SEQ;
3850 	error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL);
3851 	if (error) {
3852 		*cs->statusp = resp->status = puterrno4(error);
3853 		kmem_free(nm, len);
3854 		return;
3855 	}
3856 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime)
3857 
3858 	/* Actually do the REMOVE operation */
3859 	if (vp->v_type == VDIR) {
3860 		/*
3861 		 * Can't remove a directory that has a mounted-on filesystem.
3862 		 */
3863 		if (vn_ismntpt(vp)) {
3864 			error = EACCES;
3865 		} else {
3866 			/*
3867 			 * System V defines rmdir to return EEXIST,
3868 			 * not * ENOTEMPTY, if the directory is not
3869 			 * empty.  A System V NFS server needs to map
3870 			 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to
3871 			 * transmit over the wire.
3872 			 */
3873 			if ((error = VOP_RMDIR(dvp, nm, rootdir, cs->cr,
3874 			    NULL, 0)) == EEXIST)
3875 				error = ENOTEMPTY;
3876 		}
3877 	} else {
3878 		if ((error = VOP_REMOVE(dvp, nm, cs->cr, NULL, 0)) == 0 &&
3879 		    fp != NULL) {
3880 			struct vattr va;
3881 			vnode_t *tvp;
3882 
3883 			rfs4_dbe_lock(fp->dbe);
3884 			tvp = fp->vp;
3885 			if (tvp)
3886 				VN_HOLD(tvp);
3887 			rfs4_dbe_unlock(fp->dbe);
3888 
3889 			if (tvp) {
3890 				/*
3891 				 * This is va_seq safe because we are not
3892 				 * manipulating dvp.
3893 				 */
3894 				va.va_mask = AT_NLINK;
3895 				if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) &&
3896 				    va.va_nlink == 0) {
3897 					/* Remove state on file remove */
3898 					if (in_crit) {
3899 						nbl_end_crit(vp);
3900 						in_crit = 0;
3901 					}
3902 					rfs4_close_all_state(fp);
3903 				}
3904 				VN_RELE(tvp);
3905 			}
3906 		}
3907 	}
3908 
3909 	if (in_crit)
3910 		nbl_end_crit(vp);
3911 	VN_RELE(vp);
3912 
3913 	if (fp) {
3914 		rfs4_clear_dont_grant(fp);
3915 		rfs4_file_rele(fp);
3916 	}
3917 	kmem_free(nm, len);
3918 
3919 	if (error) {
3920 		*cs->statusp = resp->status = puterrno4(error);
3921 		return;
3922 	}
3923 
3924 	/*
3925 	 * Get the initial "after" sequence number, if it fails, set to zero
3926 	 */
3927 	idva.va_mask = AT_SEQ;
3928 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL))
3929 		idva.va_seq = 0;
3930 
3931 	/*
3932 	 * Force modified data and metadata out to stable storage.
3933 	 */
3934 	(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
3935 
3936 	/*
3937 	 * Get "after" change value, if it fails, simply return the
3938 	 * before value.
3939 	 */
3940 	adva.va_mask = AT_CTIME|AT_SEQ;
3941 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) {
3942 		adva.va_ctime = bdva.va_ctime;
3943 		adva.va_seq = 0;
3944 	}
3945 
3946 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
3947 
3948 	/*
3949 	 * The cinfo.atomic = TRUE only if we have
3950 	 * non-zero va_seq's, and it has incremented by exactly one
3951 	 * during the VOP_REMOVE/RMDIR and it didn't change during
3952 	 * the VOP_FSYNC.
3953 	 */
3954 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
3955 	    idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq)
3956 		resp->cinfo.atomic = TRUE;
3957 	else
3958 		resp->cinfo.atomic = FALSE;
3959 
3960 	*cs->statusp = resp->status = NFS4_OK;
3961 }
3962 
3963 /*
3964  * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory,
3965  *		oldname and newname.
3966  *	res: status. If success - CURRENT_FH unchanged, return change_info
3967  *		for both from and target directories.
3968  */
3969 /* ARGSUSED */
3970 static void
3971 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3972 	struct compound_state *cs)
3973 {
3974 	RENAME4args *args = &argop->nfs_argop4_u.oprename;
3975 	RENAME4res *resp = &resop->nfs_resop4_u.oprename;
3976 	int error;
3977 	vnode_t *odvp;
3978 	vnode_t *ndvp;
3979 	vnode_t *srcvp, *targvp;
3980 	struct vattr obdva, oidva, oadva;
3981 	struct vattr nbdva, nidva, nadva;
3982 	char *onm, *nnm;
3983 	uint_t olen, nlen;
3984 	rfs4_file_t *fp, *sfp;
3985 	int in_crit_src, in_crit_targ;
3986 	int fp_rele_grant_hold, sfp_rele_grant_hold;
3987 	bslabel_t *clabel;
3988 
3989 	fp = sfp = NULL;
3990 	srcvp = targvp = NULL;
3991 	in_crit_src = in_crit_targ = 0;
3992 	fp_rele_grant_hold = sfp_rele_grant_hold = 0;
3993 
3994 	/* CURRENT_FH: target directory */
3995 	ndvp = cs->vp;
3996 	if (ndvp == NULL) {
3997 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3998 		return;
3999 	}
4000 
4001 	/* SAVED_FH: from directory */
4002 	odvp = cs->saved_vp;
4003 	if (odvp == NULL) {
4004 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4005 		return;
4006 	}
4007 
4008 	if (cs->access == CS_ACCESS_DENIED) {
4009 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4010 		return;
4011 	}
4012 
4013 	/*
4014 	 * If there is an unshared filesystem mounted on this vnode,
4015 	 * do not allow to rename objects in this directory.
4016 	 */
4017 	if (vn_ismntpt(odvp)) {
4018 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4019 		return;
4020 	}
4021 
4022 	/*
4023 	 * If there is an unshared filesystem mounted on this vnode,
4024 	 * do not allow to rename to this directory.
4025 	 */
4026 	if (vn_ismntpt(ndvp)) {
4027 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4028 		return;
4029 	}
4030 
4031 	if (odvp->v_type != VDIR || ndvp->v_type != VDIR) {
4032 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
4033 		return;
4034 	}
4035 
4036 	if (cs->saved_exi != cs->exi) {
4037 		*cs->statusp = resp->status = NFS4ERR_XDEV;
4038 		return;
4039 	}
4040 
4041 	if (!utf8_dir_verify(&args->oldname)) {
4042 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4043 		return;
4044 	}
4045 
4046 	if (!utf8_dir_verify(&args->newname)) {
4047 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4048 		return;
4049 	}
4050 
4051 	onm = utf8_to_fn(&args->oldname, &olen, NULL);
4052 	if (onm == NULL) {
4053 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4054 		return;
4055 	}
4056 
4057 	nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4058 	if (nnm == NULL) {
4059 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4060 		kmem_free(onm, olen);
4061 		return;
4062 	}
4063 
4064 	if (olen > MAXNAMELEN || nlen > MAXNAMELEN) {
4065 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
4066 		kmem_free(onm, olen);
4067 		kmem_free(nnm, nlen);
4068 		return;
4069 	}
4070 
4071 
4072 	if (rdonly4(cs->exi, cs->vp, req)) {
4073 		*cs->statusp = resp->status = NFS4ERR_ROFS;
4074 		kmem_free(onm, olen);
4075 		kmem_free(nnm, nlen);
4076 		return;
4077 	}
4078 
4079 	/* check label of the target dir */
4080 	if (is_system_labeled()) {
4081 		ASSERT(req->rq_label != NULL);
4082 		clabel = req->rq_label;
4083 		DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *,
4084 		    "got client label from request(1)",
4085 		    struct svc_req *, req);
4086 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
4087 			if (!do_rfs_label_check(clabel, ndvp,
4088 			    EQUALITY_CHECK)) {
4089 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
4090 				return;
4091 			}
4092 		}
4093 	}
4094 
4095 	/*
4096 	 * Is the source a file and have a delegation?
4097 	 * We don't need to acquire va_seq before these lookups, if
4098 	 * it causes an update, cinfo.before will not match, which will
4099 	 * trigger a cache flush even if atomic is TRUE.
4100 	 */
4101 	if (sfp = rfs4_lookup_and_findfile(odvp, onm, &srcvp, &error, cs->cr)) {
4102 		if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE,
4103 		    NULL)) {
4104 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4105 			goto err_out;
4106 		}
4107 	}
4108 
4109 	if (srcvp == NULL) {
4110 		*cs->statusp = resp->status = puterrno4(error);
4111 		kmem_free(onm, olen);
4112 		kmem_free(nnm, nlen);
4113 		return;
4114 	}
4115 
4116 	sfp_rele_grant_hold = 1;
4117 
4118 	/* Does the destination exist and a file and have a delegation? */
4119 	if (fp = rfs4_lookup_and_findfile(ndvp, nnm, &targvp, NULL, cs->cr)) {
4120 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
4121 		    NULL)) {
4122 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4123 			goto err_out;
4124 		}
4125 	}
4126 	fp_rele_grant_hold = 1;
4127 
4128 
4129 	/* Check for NBMAND lock on both source and target */
4130 	if (nbl_need_check(srcvp)) {
4131 		nbl_start_crit(srcvp, RW_READER);
4132 		in_crit_src = 1;
4133 		if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) {
4134 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4135 			goto err_out;
4136 		}
4137 	}
4138 
4139 	if (targvp && nbl_need_check(targvp)) {
4140 		nbl_start_crit(targvp, RW_READER);
4141 		in_crit_targ = 1;
4142 		if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) {
4143 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4144 			goto err_out;
4145 		}
4146 	}
4147 
4148 	/* Get source "before" change value */
4149 	obdva.va_mask = AT_CTIME|AT_SEQ;
4150 	error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL);
4151 	if (!error) {
4152 		nbdva.va_mask = AT_CTIME|AT_SEQ;
4153 		error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL);
4154 	}
4155 	if (error) {
4156 		*cs->statusp = resp->status = puterrno4(error);
4157 		goto err_out;
4158 	}
4159 
4160 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime)
4161 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime)
4162 
4163 	if ((error = VOP_RENAME(odvp, onm, ndvp, nnm, cs->cr, NULL, 0)) == 0 &&
4164 	    fp != NULL) {
4165 		struct vattr va;
4166 		vnode_t *tvp;
4167 
4168 		rfs4_dbe_lock(fp->dbe);
4169 		tvp = fp->vp;
4170 		if (tvp)
4171 			VN_HOLD(tvp);
4172 		rfs4_dbe_unlock(fp->dbe);
4173 
4174 		if (tvp) {
4175 			va.va_mask = AT_NLINK;
4176 			if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) &&
4177 			    va.va_nlink == 0) {
4178 				/* The file is gone and so should the state */
4179 				if (in_crit_targ) {
4180 					nbl_end_crit(targvp);
4181 					in_crit_targ = 0;
4182 				}
4183 				rfs4_close_all_state(fp);
4184 			}
4185 			VN_RELE(tvp);
4186 		}
4187 	}
4188 	if (error == 0) {
4189 		char *tmp;
4190 
4191 		/* fix the path name for the renamed file */
4192 		mutex_enter(&srcvp->v_lock);
4193 		tmp = srcvp->v_path;
4194 		srcvp->v_path = NULL;
4195 		mutex_exit(&srcvp->v_lock);
4196 		vn_setpath(rootdir, ndvp, srcvp, nnm, nlen - 1);
4197 		if (tmp != NULL)
4198 			kmem_free(tmp, strlen(tmp) + 1);
4199 	}
4200 
4201 	if (in_crit_src)
4202 		nbl_end_crit(srcvp);
4203 	if (srcvp)
4204 		VN_RELE(srcvp);
4205 	if (in_crit_targ)
4206 		nbl_end_crit(targvp);
4207 	if (targvp)
4208 		VN_RELE(targvp);
4209 
4210 	if (sfp) {
4211 		rfs4_clear_dont_grant(sfp);
4212 		rfs4_file_rele(sfp);
4213 	}
4214 	if (fp) {
4215 		rfs4_clear_dont_grant(fp);
4216 		rfs4_file_rele(fp);
4217 	}
4218 
4219 	kmem_free(onm, olen);
4220 	kmem_free(nnm, nlen);
4221 
4222 	/*
4223 	 * Get the initial "after" sequence number, if it fails, set to zero
4224 	 */
4225 	oidva.va_mask = AT_SEQ;
4226 	if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL))
4227 		oidva.va_seq = 0;
4228 
4229 	nidva.va_mask = AT_SEQ;
4230 	if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL))
4231 		nidva.va_seq = 0;
4232 
4233 	/*
4234 	 * Force modified data and metadata out to stable storage.
4235 	 */
4236 	(void) VOP_FSYNC(odvp, 0, cs->cr, NULL);
4237 	(void) VOP_FSYNC(ndvp, 0, cs->cr, NULL);
4238 
4239 	if (error) {
4240 		*cs->statusp = resp->status = puterrno4(error);
4241 		return;
4242 	}
4243 
4244 	/*
4245 	 * Get "after" change values, if it fails, simply return the
4246 	 * before value.
4247 	 */
4248 	oadva.va_mask = AT_CTIME|AT_SEQ;
4249 	if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) {
4250 		oadva.va_ctime = obdva.va_ctime;
4251 		oadva.va_seq = 0;
4252 	}
4253 
4254 	nadva.va_mask = AT_CTIME|AT_SEQ;
4255 	if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) {
4256 		nadva.va_ctime = nbdva.va_ctime;
4257 		nadva.va_seq = 0;
4258 	}
4259 
4260 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime)
4261 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime)
4262 
4263 	/*
4264 	 * The cinfo.atomic = TRUE only if we have
4265 	 * non-zero va_seq's, and it has incremented by exactly one
4266 	 * during the VOP_RENAME and it didn't change during the VOP_FSYNC.
4267 	 */
4268 	if (obdva.va_seq && oidva.va_seq && oadva.va_seq &&
4269 	    oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq)
4270 		resp->source_cinfo.atomic = TRUE;
4271 	else
4272 		resp->source_cinfo.atomic = FALSE;
4273 
4274 	if (nbdva.va_seq && nidva.va_seq && nadva.va_seq &&
4275 	    nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq)
4276 		resp->target_cinfo.atomic = TRUE;
4277 	else
4278 		resp->target_cinfo.atomic = FALSE;
4279 
4280 #ifdef	VOLATILE_FH_TEST
4281 	{
4282 	extern void add_volrnm_fh(struct exportinfo *, vnode_t *);
4283 
4284 	/*
4285 	 * Add the renamed file handle to the volatile rename list
4286 	 */
4287 	if (cs->exi->exi_export.ex_flags & EX_VOLRNM) {
4288 		/* file handles may expire on rename */
4289 		vnode_t *vp;
4290 
4291 		nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4292 		/*
4293 		 * Already know that nnm will be a valid string
4294 		 */
4295 		error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr,
4296 		    NULL, NULL, NULL);
4297 		kmem_free(nnm, nlen);
4298 		if (!error) {
4299 			add_volrnm_fh(cs->exi, vp);
4300 			VN_RELE(vp);
4301 		}
4302 	}
4303 	}
4304 #endif	/* VOLATILE_FH_TEST */
4305 
4306 	*cs->statusp = resp->status = NFS4_OK;
4307 	return;
4308 
4309 err_out:
4310 	kmem_free(onm, olen);
4311 	kmem_free(nnm, nlen);
4312 
4313 	if (in_crit_src) nbl_end_crit(srcvp);
4314 	if (in_crit_targ) nbl_end_crit(targvp);
4315 	if (targvp) VN_RELE(targvp);
4316 	if (srcvp) VN_RELE(srcvp);
4317 	if (sfp) {
4318 		if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp);
4319 		rfs4_file_rele(sfp);
4320 	}
4321 	if (fp) {
4322 		if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp);
4323 		rfs4_file_rele(fp);
4324 	}
4325 }
4326 
4327 /* ARGSUSED */
4328 static void
4329 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4330 	struct compound_state *cs)
4331 {
4332 	RENEW4args *args = &argop->nfs_argop4_u.oprenew;
4333 	RENEW4res *resp = &resop->nfs_resop4_u.oprenew;
4334 	rfs4_client_t *cp;
4335 
4336 	if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) {
4337 		*cs->statusp = resp->status =
4338 		    rfs4_check_clientid(&args->clientid, 0);
4339 		return;
4340 	}
4341 
4342 	if (rfs4_lease_expired(cp)) {
4343 		rfs4_client_rele(cp);
4344 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
4345 		return;
4346 	}
4347 
4348 	rfs4_update_lease(cp);
4349 
4350 	mutex_enter(cp->cbinfo.cb_lock);
4351 	if (cp->cbinfo.cb_notified_of_cb_path_down == FALSE) {
4352 		cp->cbinfo.cb_notified_of_cb_path_down = TRUE;
4353 		*cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN;
4354 	} else {
4355 		*cs->statusp = resp->status = NFS4_OK;
4356 	}
4357 	mutex_exit(cp->cbinfo.cb_lock);
4358 
4359 	rfs4_client_rele(cp);
4360 
4361 }
4362 
4363 /* ARGSUSED */
4364 static void
4365 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
4366 	struct compound_state *cs)
4367 {
4368 	RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh;
4369 
4370 	/* No need to check cs->access - we are not accessing any object */
4371 	if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) {
4372 		*cs->statusp = resp->status = NFS4ERR_RESTOREFH;
4373 		return;
4374 	}
4375 	if (cs->vp != NULL) {
4376 		VN_RELE(cs->vp);
4377 	}
4378 	cs->vp = cs->saved_vp;
4379 	cs->saved_vp = NULL;
4380 	cs->exi = cs->saved_exi;
4381 	nfs_fh4_copy(&cs->saved_fh, &cs->fh);
4382 	*cs->statusp = resp->status = NFS4_OK;
4383 	cs->deleg = FALSE;
4384 }
4385 
4386 /* ARGSUSED */
4387 static void
4388 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4389 	struct compound_state *cs)
4390 {
4391 	SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh;
4392 
4393 	/* No need to check cs->access - we are not accessing any object */
4394 	if (cs->vp == NULL) {
4395 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4396 		return;
4397 	}
4398 	if (cs->saved_vp != NULL) {
4399 		VN_RELE(cs->saved_vp);
4400 	}
4401 	cs->saved_vp = cs->vp;
4402 	VN_HOLD(cs->saved_vp);
4403 	cs->saved_exi = cs->exi;
4404 	/*
4405 	 * since SAVEFH is fairly rare, don't alloc space for its fh
4406 	 * unless necessary.
4407 	 */
4408 	if (cs->saved_fh.nfs_fh4_val == NULL) {
4409 		cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP);
4410 	}
4411 	nfs_fh4_copy(&cs->fh, &cs->saved_fh);
4412 	*cs->statusp = resp->status = NFS4_OK;
4413 }
4414 
4415 /*
4416  * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to
4417  * return the bitmap of attrs that were set successfully. It is also
4418  * called by Verify/Nverify to test the vattr/vfsstat attrs. It should
4419  * always be called only after rfs4_do_set_attrs().
4420  *
4421  * Verify that the attributes are same as the expected ones. sargp->vap
4422  * and sargp->sbp contain the input attributes as translated from fattr4.
4423  *
4424  * This function verifies only the attrs that correspond to a vattr or
4425  * vfsstat struct. That is because of the extra step needed to get the
4426  * corresponding system structs. Other attributes have already been set or
4427  * verified by do_rfs4_set_attrs.
4428  *
4429  * Return 0 if all attrs match, -1 if some don't, error if error processing.
4430  */
4431 static int
4432 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp,
4433 	bitmap4 *resp, struct nfs4_ntov_table *ntovp)
4434 {
4435 	int error, ret_error = 0;
4436 	int i, k;
4437 	uint_t sva_mask = sargp->vap->va_mask;
4438 	uint_t vbit;
4439 	union nfs4_attr_u *na;
4440 	uint8_t *amap;
4441 	bool_t getsb = ntovp->vfsstat;
4442 
4443 	if (sva_mask != 0) {
4444 		/*
4445 		 * Okay to overwrite sargp->vap because we verify based
4446 		 * on the incoming values.
4447 		 */
4448 		ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0,
4449 		    sargp->cs->cr, NULL);
4450 		if (ret_error) {
4451 			if (resp == NULL)
4452 				return (ret_error);
4453 			/*
4454 			 * Must return bitmap of successful attrs
4455 			 */
4456 			sva_mask = 0;	/* to prevent checking vap later */
4457 		} else {
4458 			/*
4459 			 * Some file systems clobber va_mask. it is probably
4460 			 * wrong of them to do so, nonethless we practice
4461 			 * defensive coding.
4462 			 * See bug id 4276830.
4463 			 */
4464 			sargp->vap->va_mask = sva_mask;
4465 		}
4466 	}
4467 
4468 	if (getsb) {
4469 		/*
4470 		 * Now get the superblock and loop on the bitmap, as there is
4471 		 * no simple way of translating from superblock to bitmap4.
4472 		 */
4473 		ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp);
4474 		if (ret_error) {
4475 			if (resp == NULL)
4476 				goto errout;
4477 			getsb = FALSE;
4478 		}
4479 	}
4480 
4481 	/*
4482 	 * Now loop and verify each attribute which getattr returned
4483 	 * whether it's the same as the input.
4484 	 */
4485 	if (resp == NULL && !getsb && (sva_mask == 0))
4486 		goto errout;
4487 
4488 	na = ntovp->na;
4489 	amap = ntovp->amap;
4490 	k = 0;
4491 	for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) {
4492 		k = *amap;
4493 		ASSERT(nfs4_ntov_map[k].nval == k);
4494 		vbit = nfs4_ntov_map[k].vbit;
4495 
4496 		/*
4497 		 * If vattr attribute but VOP_GETATTR failed, or it's
4498 		 * superblock attribute but VFS_STATVFS failed, skip
4499 		 */
4500 		if (vbit) {
4501 			if ((vbit & sva_mask) == 0)
4502 				continue;
4503 		} else if (!(getsb && nfs4_ntov_map[k].vfsstat)) {
4504 			continue;
4505 		}
4506 		error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na);
4507 		if (resp != NULL) {
4508 			if (error)
4509 				ret_error = -1;	/* not all match */
4510 			else	/* update response bitmap */
4511 				*resp |= nfs4_ntov_map[k].fbit;
4512 			continue;
4513 		}
4514 		if (error) {
4515 			ret_error = -1;	/* not all match */
4516 			break;
4517 		}
4518 	}
4519 errout:
4520 	return (ret_error);
4521 }
4522 
4523 /*
4524  * Decode the attribute to be set/verified. If the attr requires a sys op
4525  * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't
4526  * call the sv_getit function for it, because the sys op hasn't yet been done.
4527  * Return 0 for success, error code if failed.
4528  *
4529  * Note: the decoded arg is not freed here but in nfs4_ntov_table_free.
4530  */
4531 static int
4532 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp,
4533 	int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap)
4534 {
4535 	int error = 0;
4536 	bool_t set_later;
4537 
4538 	sargp->vap->va_mask |= nfs4_ntov_map[k].vbit;
4539 
4540 	if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) {
4541 		set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat;
4542 		/*
4543 		 * don't verify yet if a vattr or sb dependent attr,
4544 		 * because we don't have their sys values yet.
4545 		 * Will be done later.
4546 		 */
4547 		if (! (set_later && (cmd == NFS4ATTR_VERIT))) {
4548 			/*
4549 			 * ACLs are a special case, since setting the MODE
4550 			 * conflicts with setting the ACL.  We delay setting
4551 			 * the ACL until all other attributes have been set.
4552 			 * The ACL gets set in do_rfs4_op_setattr().
4553 			 */
4554 			if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) {
4555 				error = (*nfs4_ntov_map[k].sv_getit)(cmd,
4556 				    sargp, nap);
4557 				if (error) {
4558 					xdr_free(nfs4_ntov_map[k].xfunc,
4559 					    (caddr_t)nap);
4560 				}
4561 			}
4562 		}
4563 	} else {
4564 #ifdef  DEBUG
4565 		cmn_err(CE_NOTE, "decode_fattr4_attr: error "
4566 		    "decoding attribute %d\n", k);
4567 #endif
4568 		error = EINVAL;
4569 	}
4570 	if (!error && resp_bval && !set_later) {
4571 		*resp_bval |= nfs4_ntov_map[k].fbit;
4572 	}
4573 
4574 	return (error);
4575 }
4576 
4577 /*
4578  * Set vattr based on incoming fattr4 attrs - used by setattr.
4579  * Set response mask. Ignore any values that are not writable vattr attrs.
4580  */
4581 static nfsstat4
4582 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
4583 		struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp,
4584 		nfs4_attr_cmd_t cmd)
4585 {
4586 	int error = 0;
4587 	int i;
4588 	char *attrs = fattrp->attrlist4;
4589 	uint32_t attrslen = fattrp->attrlist4_len;
4590 	XDR xdr;
4591 	nfsstat4 status = NFS4_OK;
4592 	vnode_t *vp = cs->vp;
4593 	union nfs4_attr_u *na;
4594 	uint8_t *amap;
4595 
4596 #ifndef lint
4597 	/*
4598 	 * Make sure that maximum attribute number can be expressed as an
4599 	 * 8 bit quantity.
4600 	 */
4601 	ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1));
4602 #endif
4603 
4604 	if (vp == NULL) {
4605 		if (resp)
4606 			*resp = 0;
4607 		return (NFS4ERR_NOFILEHANDLE);
4608 	}
4609 	if (cs->access == CS_ACCESS_DENIED) {
4610 		if (resp)
4611 			*resp = 0;
4612 		return (NFS4ERR_ACCESS);
4613 	}
4614 
4615 	sargp->op = cmd;
4616 	sargp->cs = cs;
4617 	sargp->flag = 0;	/* may be set later */
4618 	sargp->vap->va_mask = 0;
4619 	sargp->rdattr_error = NFS4_OK;
4620 	sargp->rdattr_error_req = FALSE;
4621 	/* sargp->sbp is set by the caller */
4622 
4623 	xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE);
4624 
4625 	na = ntovp->na;
4626 	amap = ntovp->amap;
4627 
4628 	/*
4629 	 * The following loop iterates on the nfs4_ntov_map checking
4630 	 * if the fbit is set in the requested bitmap.
4631 	 * If set then we process the arguments using the
4632 	 * rfs4_fattr4 conversion functions to populate the setattr
4633 	 * vattr and va_mask. Any settable attrs that are not using vattr
4634 	 * will be set in this loop.
4635 	 */
4636 	for (i = 0; i < nfs4_ntov_map_size; i++) {
4637 		if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) {
4638 			continue;
4639 		}
4640 		/*
4641 		 * If setattr, must be a writable attr.
4642 		 * If verify/nverify, must be a readable attr.
4643 		 */
4644 		if ((error = (*nfs4_ntov_map[i].sv_getit)(
4645 		    NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) {
4646 			/*
4647 			 * Client tries to set/verify an
4648 			 * unsupported attribute, tries to set
4649 			 * a read only attr or verify a write
4650 			 * only one - error!
4651 			 */
4652 			break;
4653 		}
4654 		/*
4655 		 * Decode the attribute to set/verify
4656 		 */
4657 		error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval,
4658 		    &xdr, resp ? resp : NULL, na);
4659 		if (error)
4660 			break;
4661 		*amap++ = (uint8_t)nfs4_ntov_map[i].nval;
4662 		na++;
4663 		(ntovp->attrcnt)++;
4664 		if (nfs4_ntov_map[i].vfsstat)
4665 			ntovp->vfsstat = TRUE;
4666 	}
4667 
4668 	if (error != 0)
4669 		status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP :
4670 		    puterrno4(error));
4671 	/* xdrmem_destroy(&xdrs); */	/* NO-OP */
4672 	return (status);
4673 }
4674 
4675 static nfsstat4
4676 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
4677 		stateid4 *stateid)
4678 {
4679 	int error = 0;
4680 	struct nfs4_svgetit_arg sarg;
4681 	bool_t trunc;
4682 
4683 	nfsstat4 status = NFS4_OK;
4684 	cred_t *cr = cs->cr;
4685 	vnode_t *vp = cs->vp;
4686 	struct nfs4_ntov_table ntov;
4687 	struct statvfs64 sb;
4688 	struct vattr bva;
4689 	struct flock64 bf;
4690 	int in_crit = 0;
4691 	uint_t saved_mask = 0;
4692 	caller_context_t ct;
4693 
4694 	*resp = 0;
4695 	sarg.sbp = &sb;
4696 	nfs4_ntov_table_init(&ntov);
4697 	status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov,
4698 	    NFS4ATTR_SETIT);
4699 	if (status != NFS4_OK) {
4700 		/*
4701 		 * failed set attrs
4702 		 */
4703 		goto done;
4704 	}
4705 	if ((sarg.vap->va_mask == 0) &&
4706 	    (! (fattrp->attrmask & FATTR4_ACL_MASK))) {
4707 		/*
4708 		 * no further work to be done
4709 		 */
4710 		goto done;
4711 	}
4712 
4713 	/*
4714 	 * If we got a request to set the ACL and the MODE, only
4715 	 * allow changing VSUID, VSGID, and VSVTX.  Attempting
4716 	 * to change any other bits, along with setting an ACL,
4717 	 * gives NFS4ERR_INVAL.
4718 	 */
4719 	if ((fattrp->attrmask & FATTR4_ACL_MASK) &&
4720 	    (fattrp->attrmask & FATTR4_MODE_MASK)) {
4721 		vattr_t va;
4722 
4723 		va.va_mask = AT_MODE;
4724 		error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL);
4725 		if (error) {
4726 			status = puterrno4(error);
4727 			goto done;
4728 		}
4729 		if ((sarg.vap->va_mode ^ va.va_mode) &
4730 		    ~(VSUID | VSGID | VSVTX)) {
4731 			status = NFS4ERR_INVAL;
4732 			goto done;
4733 		}
4734 	}
4735 
4736 	/* Check stateid only if size has been set */
4737 	if (sarg.vap->va_mask & AT_SIZE) {
4738 		trunc = (sarg.vap->va_size == 0);
4739 		status = rfs4_check_stateid(FWRITE, cs->vp, stateid,
4740 		    trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct);
4741 		if (status != NFS4_OK)
4742 			goto done;
4743 	} else {
4744 		ct.cc_sysid = 0;
4745 		ct.cc_pid = 0;
4746 		ct.cc_caller_id = nfs4_srv_caller_id;
4747 	}
4748 
4749 	/* XXX start of possible race with delegations */
4750 
4751 	/*
4752 	 * We need to specially handle size changes because it is
4753 	 * possible for the client to create a file with read-only
4754 	 * modes, but with the file opened for writing. If the client
4755 	 * then tries to set the file size, e.g. ftruncate(3C),
4756 	 * fcntl(F_FREESP), the normal access checking done in
4757 	 * VOP_SETATTR would prevent the client from doing it even though
4758 	 * it should be allowed to do so.  To get around this, we do the
4759 	 * access checking for ourselves and use VOP_SPACE which doesn't
4760 	 * do the access checking.
4761 	 * Also the client should not be allowed to change the file
4762 	 * size if there is a conflicting non-blocking mandatory lock in
4763 	 * the region of the change.
4764 	 */
4765 	if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) {
4766 		u_offset_t offset;
4767 		ssize_t length;
4768 
4769 		/*
4770 		 * ufs_setattr clears AT_SIZE from vap->va_mask, but
4771 		 * before returning, sarg.vap->va_mask is used to
4772 		 * generate the setattr reply bitmap.  We also clear
4773 		 * AT_SIZE below before calling VOP_SPACE.  For both
4774 		 * of these cases, the va_mask needs to be saved here
4775 		 * and restored after calling VOP_SETATTR.
4776 		 */
4777 		saved_mask = sarg.vap->va_mask;
4778 
4779 		/*
4780 		 * Check any possible conflict due to NBMAND locks.
4781 		 * Get into critical region before VOP_GETATTR, so the
4782 		 * size attribute is valid when checking conflicts.
4783 		 */
4784 		if (nbl_need_check(vp)) {
4785 			nbl_start_crit(vp, RW_READER);
4786 			in_crit = 1;
4787 		}
4788 
4789 		bva.va_mask = AT_UID|AT_SIZE;
4790 		if (error = VOP_GETATTR(vp, &bva, 0, cr, &ct)) {
4791 			status = puterrno4(error);
4792 			goto done;
4793 		}
4794 
4795 		if (in_crit) {
4796 			if (sarg.vap->va_size < bva.va_size) {
4797 				offset = sarg.vap->va_size;
4798 				length = bva.va_size - sarg.vap->va_size;
4799 			} else {
4800 				offset = bva.va_size;
4801 				length = sarg.vap->va_size - bva.va_size;
4802 			}
4803 			if (nbl_conflict(vp, NBL_WRITE, offset, length, 0,
4804 			    &ct)) {
4805 				status = NFS4ERR_LOCKED;
4806 				goto done;
4807 			}
4808 		}
4809 
4810 		if (crgetuid(cr) == bva.va_uid) {
4811 			sarg.vap->va_mask &= ~AT_SIZE;
4812 			bf.l_type = F_WRLCK;
4813 			bf.l_whence = 0;
4814 			bf.l_start = (off64_t)sarg.vap->va_size;
4815 			bf.l_len = 0;
4816 			bf.l_sysid = 0;
4817 			bf.l_pid = 0;
4818 			error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE,
4819 			    (offset_t)sarg.vap->va_size, cr, &ct);
4820 		}
4821 	}
4822 
4823 	if (!error && sarg.vap->va_mask != 0)
4824 		error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct);
4825 
4826 	/* restore va_mask -- ufs_setattr clears AT_SIZE */
4827 	if (saved_mask & AT_SIZE)
4828 		sarg.vap->va_mask |= AT_SIZE;
4829 
4830 	/*
4831 	 * If an ACL was being set, it has been delayed until now,
4832 	 * in order to set the mode (via the VOP_SETATTR() above) first.
4833 	 */
4834 	if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) {
4835 		int i;
4836 
4837 		for (i = 0; i < NFS4_MAXNUM_ATTRS; i++)
4838 			if (ntov.amap[i] == FATTR4_ACL)
4839 				break;
4840 		if (i < NFS4_MAXNUM_ATTRS) {
4841 			error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)(
4842 			    NFS4ATTR_SETIT, &sarg, &ntov.na[i]);
4843 			if (error == 0) {
4844 				*resp |= FATTR4_ACL_MASK;
4845 			} else if (error == ENOTSUP) {
4846 				(void) rfs4_verify_attr(&sarg, resp, &ntov);
4847 				status = NFS4ERR_ATTRNOTSUPP;
4848 				goto done;
4849 			}
4850 		} else {
4851 			NFS4_DEBUG(rfs4_debug,
4852 			    (CE_NOTE, "do_rfs4_op_setattr: "
4853 			    "unable to find ACL in fattr4"));
4854 			error = EINVAL;
4855 		}
4856 	}
4857 
4858 	if (error) {
4859 		status = puterrno4(error);
4860 
4861 		/*
4862 		 * Set the response bitmap when setattr failed.
4863 		 * If VOP_SETATTR partially succeeded, test by doing a
4864 		 * VOP_GETATTR on the object and comparing the data
4865 		 * to the setattr arguments.
4866 		 */
4867 		(void) rfs4_verify_attr(&sarg, resp, &ntov);
4868 	} else {
4869 		/*
4870 		 * Force modified metadata out to stable storage.
4871 		 */
4872 		(void) VOP_FSYNC(vp, FNODSYNC, cr, &ct);
4873 		/*
4874 		 * Set response bitmap
4875 		 */
4876 		nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp);
4877 	}
4878 
4879 /* Return early and already have a NFSv4 error */
4880 done:
4881 	/*
4882 	 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr
4883 	 * conversion sets both readable and writeable NFS4 attrs
4884 	 * for AT_MTIME and AT_ATIME.  The line below masks out
4885 	 * unrequested attrs from the setattr result bitmap.  This
4886 	 * is placed after the done: label to catch the ATTRNOTSUP
4887 	 * case.
4888 	 */
4889 	*resp &= fattrp->attrmask;
4890 
4891 	if (in_crit)
4892 		nbl_end_crit(vp);
4893 
4894 	nfs4_ntov_table_free(&ntov, &sarg);
4895 
4896 	return (status);
4897 }
4898 
4899 /* ARGSUSED */
4900 static void
4901 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4902 	struct compound_state *cs)
4903 {
4904 	SETATTR4args *args = &argop->nfs_argop4_u.opsetattr;
4905 	SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr;
4906 	bslabel_t *clabel;
4907 
4908 	if (cs->vp == NULL) {
4909 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4910 		return;
4911 	}
4912 
4913 	/*
4914 	 * If there is an unshared filesystem mounted on this vnode,
4915 	 * do not allow to setattr on this vnode.
4916 	 */
4917 	if (vn_ismntpt(cs->vp)) {
4918 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4919 		return;
4920 	}
4921 
4922 	resp->attrsset = 0;
4923 
4924 	if (rdonly4(cs->exi, cs->vp, req)) {
4925 		*cs->statusp = resp->status = NFS4ERR_ROFS;
4926 		return;
4927 	}
4928 
4929 	/* check label before setting attributes */
4930 	if (is_system_labeled()) {
4931 		ASSERT(req->rq_label != NULL);
4932 		clabel = req->rq_label;
4933 		DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *,
4934 		    "got client label from request(1)",
4935 		    struct svc_req *, req);
4936 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
4937 			if (!do_rfs_label_check(clabel, cs->vp,
4938 			    EQUALITY_CHECK)) {
4939 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
4940 				return;
4941 			}
4942 		}
4943 	}
4944 
4945 	*cs->statusp = resp->status =
4946 	    do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs,
4947 	    &args->stateid);
4948 }
4949 
4950 /* ARGSUSED */
4951 static void
4952 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4953 	struct compound_state *cs)
4954 {
4955 	/*
4956 	 * verify and nverify are exactly the same, except that nverify
4957 	 * succeeds when some argument changed, and verify succeeds when
4958 	 * when none changed.
4959 	 */
4960 
4961 	VERIFY4args  *args = &argop->nfs_argop4_u.opverify;
4962 	VERIFY4res *resp = &resop->nfs_resop4_u.opverify;
4963 
4964 	int error;
4965 	struct nfs4_svgetit_arg sarg;
4966 	struct statvfs64 sb;
4967 	struct nfs4_ntov_table ntov;
4968 
4969 	if (cs->vp == NULL) {
4970 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4971 		return;
4972 	}
4973 
4974 	sarg.sbp = &sb;
4975 	nfs4_ntov_table_init(&ntov);
4976 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
4977 	    &sarg, &ntov, NFS4ATTR_VERIT);
4978 	if (resp->status != NFS4_OK) {
4979 		/*
4980 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
4981 		 * so could return -1 for "no match".
4982 		 */
4983 		if (resp->status == -1)
4984 			resp->status = NFS4ERR_NOT_SAME;
4985 		goto done;
4986 	}
4987 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
4988 	switch (error) {
4989 	case 0:
4990 		resp->status = NFS4_OK;
4991 		break;
4992 	case -1:
4993 		resp->status = NFS4ERR_NOT_SAME;
4994 		break;
4995 	default:
4996 		resp->status = puterrno4(error);
4997 		break;
4998 	}
4999 done:
5000 	*cs->statusp = resp->status;
5001 	nfs4_ntov_table_free(&ntov, &sarg);
5002 }
5003 
5004 /* ARGSUSED */
5005 static void
5006 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5007 	struct compound_state *cs)
5008 {
5009 	/*
5010 	 * verify and nverify are exactly the same, except that nverify
5011 	 * succeeds when some argument changed, and verify succeeds when
5012 	 * when none changed.
5013 	 */
5014 
5015 	NVERIFY4args  *args = &argop->nfs_argop4_u.opnverify;
5016 	NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify;
5017 
5018 	int error;
5019 	struct nfs4_svgetit_arg sarg;
5020 	struct statvfs64 sb;
5021 	struct nfs4_ntov_table ntov;
5022 
5023 	if (cs->vp == NULL) {
5024 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5025 		return;
5026 	}
5027 	sarg.sbp = &sb;
5028 	nfs4_ntov_table_init(&ntov);
5029 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
5030 	    &sarg, &ntov, NFS4ATTR_VERIT);
5031 	if (resp->status != NFS4_OK) {
5032 		/*
5033 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
5034 		 * so could return -1 for "no match".
5035 		 */
5036 		if (resp->status == -1)
5037 			resp->status = NFS4_OK;
5038 		goto done;
5039 	}
5040 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
5041 	switch (error) {
5042 	case 0:
5043 		resp->status = NFS4ERR_SAME;
5044 		break;
5045 	case -1:
5046 		resp->status = NFS4_OK;
5047 		break;
5048 	default:
5049 		resp->status = puterrno4(error);
5050 		break;
5051 	}
5052 done:
5053 	*cs->statusp = resp->status;
5054 	nfs4_ntov_table_free(&ntov, &sarg);
5055 }
5056 
5057 /*
5058  * XXX - This should live in an NFS header file.
5059  */
5060 #define	MAX_IOVECS	12
5061 
5062 /* ARGSUSED */
5063 static void
5064 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5065 	struct compound_state *cs)
5066 {
5067 	WRITE4args  *args = &argop->nfs_argop4_u.opwrite;
5068 	WRITE4res *resp = &resop->nfs_resop4_u.opwrite;
5069 	int error;
5070 	vnode_t *vp;
5071 	struct vattr bva;
5072 	u_offset_t rlimit;
5073 	struct uio uio;
5074 	struct iovec iov[MAX_IOVECS];
5075 	struct iovec *iovp;
5076 	int iovcnt;
5077 	int ioflag;
5078 	cred_t *savecred, *cr;
5079 	bool_t *deleg = &cs->deleg;
5080 	nfsstat4 stat;
5081 	int in_crit = 0;
5082 	caller_context_t ct;
5083 
5084 	vp = cs->vp;
5085 	if (vp == NULL) {
5086 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5087 		return;
5088 	}
5089 	if (cs->access == CS_ACCESS_DENIED) {
5090 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5091 		return;
5092 	}
5093 
5094 	cr = cs->cr;
5095 
5096 	if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE,
5097 	    deleg, TRUE, &ct)) != NFS4_OK) {
5098 		*cs->statusp = resp->status = stat;
5099 		goto out;
5100 	}
5101 
5102 	/*
5103 	 * We have to enter the critical region before calling VOP_RWLOCK
5104 	 * to avoid a deadlock with ufs.
5105 	 */
5106 	if (nbl_need_check(vp)) {
5107 		nbl_start_crit(vp, RW_READER);
5108 		in_crit = 1;
5109 		if (nbl_conflict(vp, NBL_WRITE,
5110 		    args->offset, args->data_len, 0, &ct)) {
5111 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
5112 			goto out;
5113 		}
5114 	}
5115 
5116 	bva.va_mask = AT_MODE | AT_UID;
5117 	error = VOP_GETATTR(vp, &bva, 0, cr, &ct);
5118 
5119 	/*
5120 	 * If we can't get the attributes, then we can't do the
5121 	 * right access checking.  So, we'll fail the request.
5122 	 */
5123 	if (error) {
5124 		*cs->statusp = resp->status = puterrno4(error);
5125 		goto out;
5126 	}
5127 
5128 	if (rdonly4(cs->exi, cs->vp, req)) {
5129 		*cs->statusp = resp->status = NFS4ERR_ROFS;
5130 		goto out;
5131 	}
5132 
5133 	if (vp->v_type != VREG) {
5134 		*cs->statusp = resp->status =
5135 		    ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
5136 		goto out;
5137 	}
5138 
5139 	if (crgetuid(cr) != bva.va_uid &&
5140 	    (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) {
5141 		*cs->statusp = resp->status = puterrno4(error);
5142 		goto out;
5143 	}
5144 
5145 	if (MANDLOCK(vp, bva.va_mode)) {
5146 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5147 		goto out;
5148 	}
5149 
5150 	if (args->data_len == 0) {
5151 		*cs->statusp = resp->status = NFS4_OK;
5152 		resp->count = 0;
5153 		resp->committed = args->stable;
5154 		resp->writeverf = Write4verf;
5155 		goto out;
5156 	}
5157 
5158 	if (args->mblk != NULL) {
5159 		mblk_t *m;
5160 		uint_t bytes, round_len;
5161 
5162 		iovcnt = 0;
5163 		bytes = 0;
5164 		round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT);
5165 		for (m = args->mblk;
5166 		    m != NULL && bytes < round_len;
5167 		    m = m->b_cont) {
5168 			iovcnt++;
5169 			bytes += MBLKL(m);
5170 		}
5171 #ifdef DEBUG
5172 		/* should have ended on an mblk boundary */
5173 		if (bytes != round_len) {
5174 			printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n",
5175 			    bytes, round_len, args->data_len);
5176 			printf("args=%p, args->mblk=%p, m=%p", (void *)args,
5177 			    (void *)args->mblk, (void *)m);
5178 			ASSERT(bytes == round_len);
5179 		}
5180 #endif
5181 		if (iovcnt <= MAX_IOVECS) {
5182 			iovp = iov;
5183 		} else {
5184 			iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP);
5185 		}
5186 		mblk_to_iov(args->mblk, iovcnt, iovp);
5187 	} else {
5188 		iovcnt = 1;
5189 		iovp = iov;
5190 		iovp->iov_base = args->data_val;
5191 		iovp->iov_len = args->data_len;
5192 	}
5193 
5194 	uio.uio_iov = iovp;
5195 	uio.uio_iovcnt = iovcnt;
5196 
5197 	uio.uio_segflg = UIO_SYSSPACE;
5198 	uio.uio_extflg = UIO_COPY_DEFAULT;
5199 	uio.uio_loffset = args->offset;
5200 	uio.uio_resid = args->data_len;
5201 	uio.uio_llimit = curproc->p_fsz_ctl;
5202 	rlimit = uio.uio_llimit - args->offset;
5203 	if (rlimit < (u_offset_t)uio.uio_resid)
5204 		uio.uio_resid = (int)rlimit;
5205 
5206 	if (args->stable == UNSTABLE4)
5207 		ioflag = 0;
5208 	else if (args->stable == FILE_SYNC4)
5209 		ioflag = FSYNC;
5210 	else if (args->stable == DATA_SYNC4)
5211 		ioflag = FDSYNC;
5212 	else {
5213 		if (iovp != iov)
5214 			kmem_free(iovp, sizeof (*iovp) * iovcnt);
5215 		*cs->statusp = resp->status = NFS4ERR_INVAL;
5216 		goto out;
5217 	}
5218 
5219 	/*
5220 	 * We're changing creds because VM may fault and we need
5221 	 * the cred of the current thread to be used if quota
5222 	 * checking is enabled.
5223 	 */
5224 	savecred = curthread->t_cred;
5225 	curthread->t_cred = cr;
5226 	error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct);
5227 	curthread->t_cred = savecred;
5228 
5229 	if (iovp != iov)
5230 		kmem_free(iovp, sizeof (*iovp) * iovcnt);
5231 
5232 	if (error) {
5233 		*cs->statusp = resp->status = puterrno4(error);
5234 		goto out;
5235 	}
5236 
5237 	*cs->statusp = resp->status = NFS4_OK;
5238 	resp->count = args->data_len - uio.uio_resid;
5239 
5240 	if (ioflag == 0)
5241 		resp->committed = UNSTABLE4;
5242 	else
5243 		resp->committed = FILE_SYNC4;
5244 
5245 	resp->writeverf = Write4verf;
5246 
5247 out:
5248 	if (in_crit)
5249 		nbl_end_crit(vp);
5250 }
5251 
5252 
5253 /* XXX put in a header file */
5254 extern int	sec_svc_getcred(struct svc_req *, cred_t *,  caddr_t *, int *);
5255 
5256 void
5257 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi,
5258 	struct svc_req *req, cred_t *cr)
5259 {
5260 	uint_t i;
5261 	struct compound_state cs;
5262 
5263 	rfs4_init_compound_state(&cs);
5264 	/*
5265 	 * Form a reply tag by copying over the reqeuest tag.
5266 	 */
5267 	resp->tag.utf8string_val =
5268 	    kmem_alloc(args->tag.utf8string_len, KM_SLEEP);
5269 	resp->tag.utf8string_len = args->tag.utf8string_len;
5270 	bcopy(args->tag.utf8string_val, resp->tag.utf8string_val,
5271 	    resp->tag.utf8string_len);
5272 
5273 	cs.statusp = &resp->status;
5274 
5275 	/*
5276 	 * XXX for now, minorversion should be zero
5277 	 */
5278 	if (args->minorversion != NFS4_MINORVERSION) {
5279 		resp->array_len = 0;
5280 		resp->array = NULL;
5281 		resp->status = NFS4ERR_MINOR_VERS_MISMATCH;
5282 		return;
5283 	}
5284 
5285 	resp->array_len = args->array_len;
5286 	resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4),
5287 	    KM_SLEEP);
5288 
5289 	ASSERT(exi == NULL);
5290 	ASSERT(cr == NULL);
5291 
5292 	cr = crget();
5293 	ASSERT(cr != NULL);
5294 
5295 	if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) {
5296 		crfree(cr);
5297 		return;
5298 	}
5299 
5300 	cs.basecr = cr;
5301 
5302 	cs.req = req;
5303 
5304 	/*
5305 	 * For now, NFS4 compound processing must be protected by
5306 	 * exported_lock because it can access more than one exportinfo
5307 	 * per compound and share/unshare can now change multiple
5308 	 * exinfo structs.  The NFS2/3 code only refs 1 exportinfo
5309 	 * per proc (excluding public exinfo), and exi_count design
5310 	 * is sufficient to protect concurrent execution of NFS2/3
5311 	 * ops along with unexport.  This lock will be removed as
5312 	 * part of the NFSv4 phase 2 namespace redesign work.
5313 	 */
5314 	rw_enter(&exported_lock, RW_READER);
5315 
5316 	/*
5317 	 * If this is the first compound we've seen, we need to start all
5318 	 * new instances' grace periods.
5319 	 */
5320 	if (rfs4_seen_first_compound == 0) {
5321 		rfs4_grace_start_new();
5322 		/*
5323 		 * This must be set after rfs4_grace_start_new(), otherwise
5324 		 * another thread could proceed past here before the former
5325 		 * is finished.
5326 		 */
5327 		rfs4_seen_first_compound = 1;
5328 	}
5329 
5330 	for (i = 0; i < args->array_len && cs.cont; i++) {
5331 		nfs_argop4 *argop;
5332 		nfs_resop4 *resop;
5333 		uint_t op;
5334 
5335 		argop = &args->array[i];
5336 		resop = &resp->array[i];
5337 		resop->resop = argop->argop;
5338 		op = (uint_t)resop->resop;
5339 
5340 		if (op < rfsv4disp_cnt) {
5341 			/*
5342 			 * Count the individual ops here; NULL and COMPOUND
5343 			 * are counted in common_dispatch()
5344 			 */
5345 			rfsproccnt_v4_ptr[op].value.ui64++;
5346 
5347 			NFS4_DEBUG(rfs4_debug > 1,
5348 			    (CE_NOTE, "Executing %s", rfs4_op_string[op]));
5349 			(*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs);
5350 			NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d",
5351 			    rfs4_op_string[op], *cs.statusp));
5352 			if (*cs.statusp != NFS4_OK)
5353 				cs.cont = FALSE;
5354 		} else {
5355 			/*
5356 			 * This is effectively dead code since XDR code
5357 			 * will have already returned BADXDR if op doesn't
5358 			 * decode to legal value.  This only done for a
5359 			 * day when XDR code doesn't verify v4 opcodes.
5360 			 */
5361 			op = OP_ILLEGAL;
5362 			rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++;
5363 
5364 			rfs4_op_illegal(argop, resop, req, &cs);
5365 			cs.cont = FALSE;
5366 		}
5367 
5368 		/*
5369 		 * If not at last op, and if we are to stop, then
5370 		 * compact the results array.
5371 		 */
5372 		if ((i + 1) < args->array_len && !cs.cont) {
5373 			nfs_resop4 *new_res = kmem_alloc(
5374 			    (i+1) * sizeof (nfs_resop4), KM_SLEEP);
5375 			bcopy(resp->array,
5376 			    new_res, (i+1) * sizeof (nfs_resop4));
5377 			kmem_free(resp->array,
5378 			    args->array_len * sizeof (nfs_resop4));
5379 
5380 			resp->array_len =  i + 1;
5381 			resp->array = new_res;
5382 		}
5383 	}
5384 
5385 	rw_exit(&exported_lock);
5386 
5387 	if (cs.vp)
5388 		VN_RELE(cs.vp);
5389 	if (cs.saved_vp)
5390 		VN_RELE(cs.saved_vp);
5391 	if (cs.saved_fh.nfs_fh4_val)
5392 		kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE);
5393 
5394 	if (cs.basecr)
5395 		crfree(cs.basecr);
5396 	if (cs.cr)
5397 		crfree(cs.cr);
5398 	/*
5399 	 * done with this compound request, free the label
5400 	 */
5401 
5402 	if (req->rq_label != NULL) {
5403 		kmem_free(req->rq_label, sizeof (bslabel_t));
5404 		req->rq_label = NULL;
5405 	}
5406 }
5407 
5408 /*
5409  * XXX because of what appears to be duplicate calls to rfs4_compound_free
5410  * XXX zero out the tag and array values. Need to investigate why the
5411  * XXX calls occur, but at least prevent the panic for now.
5412  */
5413 void
5414 rfs4_compound_free(COMPOUND4res *resp)
5415 {
5416 	uint_t i;
5417 
5418 	if (resp->tag.utf8string_val) {
5419 		UTF8STRING_FREE(resp->tag)
5420 	}
5421 
5422 	for (i = 0; i < resp->array_len; i++) {
5423 		nfs_resop4 *resop;
5424 		uint_t op;
5425 
5426 		resop = &resp->array[i];
5427 		op = (uint_t)resop->resop;
5428 		if (op < rfsv4disp_cnt) {
5429 			(*rfsv4disptab[op].dis_resfree)(resop);
5430 		}
5431 	}
5432 	if (resp->array != NULL) {
5433 		kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4));
5434 	}
5435 }
5436 
5437 /*
5438  * Process the value of the compound request rpc flags, as a bit-AND
5439  * of the individual per-op flags (idempotent, allowork, publicfh_ok)
5440  */
5441 void
5442 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp)
5443 {
5444 	int i;
5445 	int flag = RPC_ALL;
5446 
5447 	for (i = 0; flag && i < args->array_len; i++) {
5448 		uint_t op;
5449 
5450 		op = (uint_t)args->array[i].argop;
5451 
5452 		if (op < rfsv4disp_cnt)
5453 			flag &= rfsv4disptab[op].dis_flags;
5454 		else
5455 			flag = 0;
5456 	}
5457 	*flagp = flag;
5458 }
5459 
5460 nfsstat4
5461 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp)
5462 {
5463 	nfsstat4 e;
5464 
5465 	rfs4_dbe_lock(cp->dbe);
5466 
5467 	if (cp->sysidt != LM_NOSYSID) {
5468 		*sp = cp->sysidt;
5469 		e = NFS4_OK;
5470 
5471 	} else if ((cp->sysidt = lm_alloc_sysidt()) != LM_NOSYSID) {
5472 		*sp = cp->sysidt;
5473 		e = NFS4_OK;
5474 
5475 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
5476 		    "rfs4_client_sysid: allocated 0x%x\n", *sp));
5477 	} else
5478 		e = NFS4ERR_DELAY;
5479 
5480 	rfs4_dbe_unlock(cp->dbe);
5481 	return (e);
5482 }
5483 
5484 #if defined(DEBUG) && ! defined(lint)
5485 static void lock_print(char *str, int operation, struct flock64 *flk)
5486 {
5487 	char *op, *type;
5488 
5489 	switch (operation) {
5490 	case F_GETLK: op = "F_GETLK";
5491 		break;
5492 	case F_SETLK: op = "F_SETLK";
5493 		break;
5494 	case F_SETLK_NBMAND: op = "F_SETLK_NBMAND";
5495 		break;
5496 	default: op = "F_UNKNOWN";
5497 		break;
5498 	}
5499 	switch (flk->l_type) {
5500 	case F_UNLCK: type = "F_UNLCK";
5501 		break;
5502 	case F_RDLCK: type = "F_RDLCK";
5503 		break;
5504 	case F_WRLCK: type = "F_WRLCK";
5505 		break;
5506 	default: type = "F_UNKNOWN";
5507 		break;
5508 	}
5509 
5510 	ASSERT(flk->l_whence == 0);
5511 	cmn_err(CE_NOTE, "%s:  %s, type = %s, off = %llx len = %llx pid = %d",
5512 	    str, op, type, (longlong_t)flk->l_start,
5513 	    flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid);
5514 }
5515 
5516 #define	LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f)
5517 #else
5518 #define	LOCK_PRINT(d, s, t, f)
5519 #endif
5520 
5521 /*ARGSUSED*/
5522 static bool_t
5523 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs)
5524 {
5525 	return (TRUE);
5526 }
5527 
5528 /*
5529  * Look up the pathname using the vp in cs as the directory vnode.
5530  * cs->vp will be the vnode for the file on success
5531  */
5532 
5533 static nfsstat4
5534 rfs4_lookup(component4 *component, struct svc_req *req,
5535 	    struct compound_state *cs)
5536 {
5537 	char *nm;
5538 	uint32_t len;
5539 	nfsstat4 status;
5540 
5541 	if (cs->vp == NULL) {
5542 		return (NFS4ERR_NOFILEHANDLE);
5543 	}
5544 	if (cs->vp->v_type != VDIR) {
5545 		return (NFS4ERR_NOTDIR);
5546 	}
5547 
5548 	if (!utf8_dir_verify(component))
5549 		return (NFS4ERR_INVAL);
5550 
5551 	nm = utf8_to_fn(component, &len, NULL);
5552 	if (nm == NULL) {
5553 		return (NFS4ERR_INVAL);
5554 	}
5555 
5556 	if (len > MAXNAMELEN) {
5557 		kmem_free(nm, len);
5558 		return (NFS4ERR_NAMETOOLONG);
5559 	}
5560 
5561 	status = do_rfs4_op_lookup(nm, len, req, cs);
5562 
5563 	kmem_free(nm, len);
5564 
5565 	return (status);
5566 }
5567 
5568 static nfsstat4
5569 rfs4_lookupfile(component4 *component, struct svc_req *req,
5570 		struct compound_state *cs, uint32_t access,
5571 		change_info4 *cinfo)
5572 {
5573 	nfsstat4 status;
5574 	vnode_t *dvp = cs->vp;
5575 	vattr_t bva, ava, fva;
5576 	int error;
5577 
5578 	/* Get "before" change value */
5579 	bva.va_mask = AT_CTIME|AT_SEQ;
5580 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL);
5581 	if (error)
5582 		return (puterrno4(error));
5583 
5584 	/* rfs4_lookup may VN_RELE directory */
5585 	VN_HOLD(dvp);
5586 
5587 	status = rfs4_lookup(component, req, cs);
5588 	if (status != NFS4_OK) {
5589 		VN_RELE(dvp);
5590 		return (status);
5591 	}
5592 
5593 	/*
5594 	 * Get "after" change value, if it fails, simply return the
5595 	 * before value.
5596 	 */
5597 	ava.va_mask = AT_CTIME|AT_SEQ;
5598 	if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) {
5599 		ava.va_ctime = bva.va_ctime;
5600 		ava.va_seq = 0;
5601 	}
5602 	VN_RELE(dvp);
5603 
5604 	/*
5605 	 * Validate the file is a file
5606 	 */
5607 	fva.va_mask = AT_TYPE|AT_MODE;
5608 	error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL);
5609 	if (error)
5610 		return (puterrno4(error));
5611 
5612 	if (fva.va_type != VREG) {
5613 		if (fva.va_type == VDIR)
5614 			return (NFS4ERR_ISDIR);
5615 		if (fva.va_type == VLNK)
5616 			return (NFS4ERR_SYMLINK);
5617 		return (NFS4ERR_INVAL);
5618 	}
5619 
5620 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime);
5621 	NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
5622 
5623 	/*
5624 	 * It is undefined if VOP_LOOKUP will change va_seq, so
5625 	 * cinfo.atomic = TRUE only if we have
5626 	 * non-zero va_seq's, and they have not changed.
5627 	 */
5628 	if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq)
5629 		cinfo->atomic = TRUE;
5630 	else
5631 		cinfo->atomic = FALSE;
5632 
5633 	/* Check for mandatory locking */
5634 	cs->mandlock = MANDLOCK(cs->vp, fva.va_mode);
5635 	return (check_open_access(access, cs, req));
5636 }
5637 
5638 static nfsstat4
5639 create_vnode(vnode_t *dvp, char *nm,  vattr_t *vap, createmode4 mode,
5640 	    timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created)
5641 {
5642 	int error;
5643 	nfsstat4 status = NFS4_OK;
5644 	vattr_t va;
5645 
5646 tryagain:
5647 
5648 	/*
5649 	 * The file open mode used is VWRITE.  If the client needs
5650 	 * some other semantic, then it should do the access checking
5651 	 * itself.  It would have been nice to have the file open mode
5652 	 * passed as part of the arguments.
5653 	 */
5654 
5655 	*created = TRUE;
5656 	error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL);
5657 
5658 	if (error) {
5659 		*created = FALSE;
5660 
5661 		/*
5662 		 * If we got something other than file already exists
5663 		 * then just return this error.  Otherwise, we got
5664 		 * EEXIST.  If we were doing a GUARDED create, then
5665 		 * just return this error.  Otherwise, we need to
5666 		 * make sure that this wasn't a duplicate of an
5667 		 * exclusive create request.
5668 		 *
5669 		 * The assumption is made that a non-exclusive create
5670 		 * request will never return EEXIST.
5671 		 */
5672 
5673 		if (error != EEXIST || mode == GUARDED4) {
5674 			status = puterrno4(error);
5675 			return (status);
5676 		}
5677 		error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr,
5678 		    NULL, NULL, NULL);
5679 
5680 		if (error) {
5681 			/*
5682 			 * We couldn't find the file that we thought that
5683 			 * we just created.  So, we'll just try creating
5684 			 * it again.
5685 			 */
5686 			if (error == ENOENT)
5687 				goto tryagain;
5688 
5689 			status = puterrno4(error);
5690 			return (status);
5691 		}
5692 
5693 		if (mode == UNCHECKED4) {
5694 			/* existing object must be regular file */
5695 			if ((*vpp)->v_type != VREG) {
5696 				if ((*vpp)->v_type == VDIR)
5697 					status = NFS4ERR_ISDIR;
5698 				else if ((*vpp)->v_type == VLNK)
5699 					status = NFS4ERR_SYMLINK;
5700 				else
5701 					status = NFS4ERR_INVAL;
5702 				VN_RELE(*vpp);
5703 				return (status);
5704 			}
5705 
5706 			return (NFS4_OK);
5707 		}
5708 
5709 		/* Check for duplicate request */
5710 		ASSERT(mtime != 0);
5711 		va.va_mask = AT_MTIME;
5712 		error = VOP_GETATTR(*vpp, &va, 0, cr, NULL);
5713 		if (!error) {
5714 			/* We found the file */
5715 			if (va.va_mtime.tv_sec != mtime->tv_sec ||
5716 			    va.va_mtime.tv_nsec != mtime->tv_nsec) {
5717 				/* but its not our creation */
5718 				VN_RELE(*vpp);
5719 				return (NFS4ERR_EXIST);
5720 			}
5721 			*created = TRUE; /* retrans of create == created */
5722 			return (NFS4_OK);
5723 		}
5724 		VN_RELE(*vpp);
5725 		return (NFS4ERR_EXIST);
5726 	}
5727 
5728 	return (NFS4_OK);
5729 }
5730 
5731 static nfsstat4
5732 check_open_access(uint32_t access,
5733 		struct compound_state *cs, struct svc_req *req)
5734 {
5735 	int error;
5736 	vnode_t *vp;
5737 	bool_t readonly;
5738 	cred_t *cr = cs->cr;
5739 
5740 	/* For now we don't allow mandatory locking as per V2/V3 */
5741 	if (cs->access == CS_ACCESS_DENIED || cs->mandlock) {
5742 		return (NFS4ERR_ACCESS);
5743 	}
5744 
5745 	vp = cs->vp;
5746 	ASSERT(cr != NULL && vp->v_type == VREG);
5747 
5748 	/*
5749 	 * If the file system is exported read only and we are trying
5750 	 * to open for write, then return NFS4ERR_ROFS
5751 	 */
5752 
5753 	readonly = rdonly4(cs->exi, cs->vp, req);
5754 
5755 	if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly)
5756 		return (NFS4ERR_ROFS);
5757 
5758 	if (access & OPEN4_SHARE_ACCESS_READ) {
5759 		if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) &&
5760 		    (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) {
5761 			return (NFS4ERR_ACCESS);
5762 		}
5763 	}
5764 
5765 	if (access & OPEN4_SHARE_ACCESS_WRITE) {
5766 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
5767 		if (error)
5768 			return (NFS4ERR_ACCESS);
5769 	}
5770 
5771 	return (NFS4_OK);
5772 }
5773 
5774 static nfsstat4
5775 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs,
5776 		change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid)
5777 {
5778 	struct nfs4_svgetit_arg sarg;
5779 	struct nfs4_ntov_table ntov;
5780 
5781 	bool_t ntov_table_init = FALSE;
5782 	struct statvfs64 sb;
5783 	nfsstat4 status;
5784 	vnode_t *vp;
5785 	vattr_t bva, ava, iva, cva, *vap;
5786 	vnode_t *dvp;
5787 	timespec32_t *mtime;
5788 	char *nm = NULL;
5789 	uint_t buflen;
5790 	bool_t created;
5791 	bool_t setsize = FALSE;
5792 	len_t reqsize;
5793 	int error;
5794 	bool_t trunc;
5795 	caller_context_t ct;
5796 	component4 *component;
5797 	bslabel_t *clabel;
5798 
5799 	sarg.sbp = &sb;
5800 
5801 	dvp = cs->vp;
5802 
5803 	/* Check if the file system is read only */
5804 	if (rdonly4(cs->exi, dvp, req))
5805 		return (NFS4ERR_ROFS);
5806 
5807 	/* check the label of including directory */
5808 	if (is_system_labeled()) {
5809 		ASSERT(req->rq_label != NULL);
5810 		clabel = req->rq_label;
5811 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
5812 		    "got client label from request(1)",
5813 		    struct svc_req *, req);
5814 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
5815 			if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK)) {
5816 				return (NFS4ERR_ACCESS);
5817 			}
5818 		}
5819 	}
5820 
5821 	/*
5822 	 * Get the last component of path name in nm. cs will reference
5823 	 * the including directory on success.
5824 	 */
5825 	component = &args->open_claim4_u.file;
5826 	if (!utf8_dir_verify(component))
5827 		return (NFS4ERR_INVAL);
5828 
5829 	nm = utf8_to_fn(component, &buflen, NULL);
5830 
5831 	if (nm == NULL)
5832 		return (NFS4ERR_RESOURCE);
5833 
5834 	if (buflen > MAXNAMELEN) {
5835 		kmem_free(nm, buflen);
5836 		return (NFS4ERR_NAMETOOLONG);
5837 	}
5838 
5839 	bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ;
5840 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL);
5841 	if (error) {
5842 		kmem_free(nm, buflen);
5843 		return (puterrno4(error));
5844 	}
5845 
5846 	if (bva.va_type != VDIR) {
5847 		kmem_free(nm, buflen);
5848 		return (NFS4ERR_NOTDIR);
5849 	}
5850 
5851 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime)
5852 
5853 	switch (args->mode) {
5854 	case GUARDED4:
5855 		/*FALLTHROUGH*/
5856 	case UNCHECKED4:
5857 		nfs4_ntov_table_init(&ntov);
5858 		ntov_table_init = TRUE;
5859 
5860 		*attrset = 0;
5861 		status = do_rfs4_set_attrs(attrset,
5862 		    &args->createhow4_u.createattrs,
5863 		    cs, &sarg, &ntov, NFS4ATTR_SETIT);
5864 
5865 		if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) &&
5866 		    sarg.vap->va_type != VREG) {
5867 			if (sarg.vap->va_type == VDIR)
5868 				status = NFS4ERR_ISDIR;
5869 			else if (sarg.vap->va_type == VLNK)
5870 				status = NFS4ERR_SYMLINK;
5871 			else
5872 				status = NFS4ERR_INVAL;
5873 		}
5874 
5875 		if (status != NFS4_OK) {
5876 			kmem_free(nm, buflen);
5877 			nfs4_ntov_table_free(&ntov, &sarg);
5878 			*attrset = 0;
5879 			return (status);
5880 		}
5881 
5882 		vap = sarg.vap;
5883 		vap->va_type = VREG;
5884 		vap->va_mask |= AT_TYPE;
5885 
5886 		if ((vap->va_mask & AT_MODE) == 0) {
5887 			vap->va_mask |= AT_MODE;
5888 			vap->va_mode = (mode_t)0600;
5889 		}
5890 
5891 		if (vap->va_mask & AT_SIZE) {
5892 
5893 			/* Disallow create with a non-zero size */
5894 
5895 			if ((reqsize = sarg.vap->va_size) != 0) {
5896 				kmem_free(nm, buflen);
5897 				nfs4_ntov_table_free(&ntov, &sarg);
5898 				*attrset = 0;
5899 				return (NFS4ERR_INVAL);
5900 			}
5901 			setsize = TRUE;
5902 		}
5903 		break;
5904 
5905 	case EXCLUSIVE4:
5906 		/* prohibit EXCL create of named attributes */
5907 		if (dvp->v_flag & V_XATTRDIR) {
5908 			kmem_free(nm, buflen);
5909 			*attrset = 0;
5910 			return (NFS4ERR_INVAL);
5911 		}
5912 
5913 		cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE;
5914 		cva.va_type = VREG;
5915 		/*
5916 		 * Ensure no time overflows. Assumes underlying
5917 		 * filesystem supports at least 32 bits.
5918 		 * Truncate nsec to usec resolution to allow valid
5919 		 * compares even if the underlying filesystem truncates.
5920 		 */
5921 		mtime = (timespec32_t *)&args->createhow4_u.createverf;
5922 		cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX;
5923 		cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000;
5924 		cva.va_mode = (mode_t)0;
5925 		vap = &cva;
5926 		break;
5927 	}
5928 
5929 	status = create_vnode(dvp, nm, vap, args->mode, mtime,
5930 	    cs->cr, &vp, &created);
5931 	kmem_free(nm, buflen);
5932 
5933 	if (status != NFS4_OK) {
5934 		if (ntov_table_init)
5935 			nfs4_ntov_table_free(&ntov, &sarg);
5936 		*attrset = 0;
5937 		return (status);
5938 	}
5939 
5940 	trunc = (setsize && !created);
5941 
5942 	if (args->mode != EXCLUSIVE4) {
5943 		bitmap4 createmask = args->createhow4_u.createattrs.attrmask;
5944 
5945 		/*
5946 		 * True verification that object was created with correct
5947 		 * attrs is impossible.  The attrs could have been changed
5948 		 * immediately after object creation.  If attributes did
5949 		 * not verify, the only recourse for the server is to
5950 		 * destroy the object.  Maybe if some attrs (like gid)
5951 		 * are set incorrectly, the object should be destroyed;
5952 		 * however, seems bad as a default policy.  Do we really
5953 		 * want to destroy an object over one of the times not
5954 		 * verifying correctly?  For these reasons, the server
5955 		 * currently sets bits in attrset for createattrs
5956 		 * that were set; however, no verification is done.
5957 		 *
5958 		 * vmask_to_nmask accounts for vattr bits set on create
5959 		 *	[do_rfs4_set_attrs() only sets resp bits for
5960 		 *	 non-vattr/vfs bits.]
5961 		 * Mask off any bits we set by default so as not to return
5962 		 * more attrset bits than were requested in createattrs
5963 		 */
5964 		if (created) {
5965 			nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset);
5966 			*attrset &= createmask;
5967 		} else {
5968 			/*
5969 			 * We did not create the vnode (we tried but it
5970 			 * already existed).  In this case, the only createattr
5971 			 * that the spec allows the server to set is size,
5972 			 * and even then, it can only be set if it is 0.
5973 			 */
5974 			*attrset = 0;
5975 			if (trunc)
5976 				*attrset = FATTR4_SIZE_MASK;
5977 		}
5978 	}
5979 	if (ntov_table_init)
5980 		nfs4_ntov_table_free(&ntov, &sarg);
5981 
5982 	/*
5983 	 * Get the initial "after" sequence number, if it fails,
5984 	 * set to zero, time to before.
5985 	 */
5986 	iva.va_mask = AT_CTIME|AT_SEQ;
5987 	if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) {
5988 		iva.va_seq = 0;
5989 		iva.va_ctime = bva.va_ctime;
5990 	}
5991 
5992 	/*
5993 	 * create_vnode attempts to create the file exclusive,
5994 	 * if it already exists the VOP_CREATE will fail and
5995 	 * may not increase va_seq. It is atomic if
5996 	 * we haven't changed the directory, but if it has changed
5997 	 * we don't know what changed it.
5998 	 */
5999 	if (!created) {
6000 		if (bva.va_seq && iva.va_seq &&
6001 		    bva.va_seq == iva.va_seq)
6002 			cinfo->atomic = TRUE;
6003 		else
6004 			cinfo->atomic = FALSE;
6005 		NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime);
6006 	} else {
6007 		/*
6008 		 * The entry was created, we need to sync the
6009 		 * directory metadata.
6010 		 */
6011 		(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
6012 
6013 		/*
6014 		 * Get "after" change value, if it fails, simply return the
6015 		 * before value.
6016 		 */
6017 		ava.va_mask = AT_CTIME|AT_SEQ;
6018 		if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) {
6019 			ava.va_ctime = bva.va_ctime;
6020 			ava.va_seq = 0;
6021 		}
6022 
6023 		NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
6024 
6025 		/*
6026 		 * The cinfo->atomic = TRUE only if we have
6027 		 * non-zero va_seq's, and it has incremented by exactly one
6028 		 * during the create_vnode and it didn't
6029 		 * change during the VOP_FSYNC.
6030 		 */
6031 		if (bva.va_seq && iva.va_seq && ava.va_seq &&
6032 		    iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq)
6033 			cinfo->atomic = TRUE;
6034 		else
6035 			cinfo->atomic = FALSE;
6036 	}
6037 
6038 	/* Check for mandatory locking and that the size gets set. */
6039 	cva.va_mask = AT_MODE;
6040 	if (setsize)
6041 		cva.va_mask |= AT_SIZE;
6042 
6043 	/* Assume the worst */
6044 	cs->mandlock = TRUE;
6045 
6046 	if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) {
6047 		cs->mandlock = MANDLOCK(cs->vp, cva.va_mode);
6048 
6049 		/*
6050 		 * Truncate the file if necessary; this would be
6051 		 * the case for create over an existing file.
6052 		 */
6053 
6054 		if (trunc) {
6055 			int in_crit = 0;
6056 			rfs4_file_t *fp;
6057 			bool_t create = FALSE;
6058 
6059 			/*
6060 			 * We are writing over an existing file.
6061 			 * Check to see if we need to recall a delegation.
6062 			 */
6063 			rfs4_hold_deleg_policy();
6064 			if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) {
6065 				if (rfs4_check_delegated_byfp(FWRITE, fp,
6066 				    (reqsize == 0), FALSE, FALSE, &clientid)) {
6067 					rfs4_file_rele(fp);
6068 					rfs4_rele_deleg_policy();
6069 					VN_RELE(vp);
6070 					*attrset = 0;
6071 					return (NFS4ERR_DELAY);
6072 				}
6073 				rfs4_file_rele(fp);
6074 			}
6075 			rfs4_rele_deleg_policy();
6076 
6077 			if (nbl_need_check(vp)) {
6078 				in_crit = 1;
6079 
6080 				ASSERT(reqsize == 0);
6081 
6082 				nbl_start_crit(vp, RW_READER);
6083 				if (nbl_conflict(vp, NBL_WRITE, 0,
6084 				    cva.va_size, 0, NULL)) {
6085 					in_crit = 0;
6086 					nbl_end_crit(vp);
6087 					VN_RELE(vp);
6088 					*attrset = 0;
6089 					return (NFS4ERR_ACCESS);
6090 				}
6091 			}
6092 			ct.cc_sysid = 0;
6093 			ct.cc_pid = 0;
6094 			ct.cc_caller_id = nfs4_srv_caller_id;
6095 
6096 			cva.va_mask = AT_SIZE;
6097 			cva.va_size = reqsize;
6098 			(void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct);
6099 			if (in_crit)
6100 				nbl_end_crit(vp);
6101 		}
6102 	}
6103 
6104 	error = makefh4(&cs->fh, vp, cs->exi);
6105 
6106 	/*
6107 	 * Force modified data and metadata out to stable storage.
6108 	 */
6109 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL);
6110 
6111 	if (error) {
6112 		VN_RELE(vp);
6113 		*attrset = 0;
6114 		return (puterrno4(error));
6115 	}
6116 
6117 	/* if parent dir is attrdir, set namedattr fh flag */
6118 	if (dvp->v_flag & V_XATTRDIR)
6119 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
6120 
6121 	if (cs->vp)
6122 		VN_RELE(cs->vp);
6123 
6124 	cs->vp = vp;
6125 
6126 	/*
6127 	 * if we did not create the file, we will need to check
6128 	 * the access bits on the file
6129 	 */
6130 
6131 	if (!created) {
6132 		if (setsize)
6133 			args->share_access |= OPEN4_SHARE_ACCESS_WRITE;
6134 		status = check_open_access(args->share_access, cs, req);
6135 		if (status != NFS4_OK)
6136 			*attrset = 0;
6137 	}
6138 	return (status);
6139 }
6140 
6141 /*ARGSUSED*/
6142 static void
6143 rfs4_do_open(struct compound_state *cs, struct svc_req *req,
6144 		rfs4_openowner_t *oo, delegreq_t deleg,
6145 		uint32_t access, uint32_t deny,
6146 		OPEN4res *resp, int deleg_cur)
6147 {
6148 	/* XXX Currently not using req  */
6149 	rfs4_state_t *state;
6150 	rfs4_file_t *file;
6151 	bool_t screate = TRUE;
6152 	bool_t fcreate = TRUE;
6153 	uint32_t amodes;
6154 	uint32_t dmodes;
6155 	rfs4_deleg_state_t *dsp;
6156 	struct shrlock shr;
6157 	struct shr_locowner shr_loco;
6158 	sysid_t sysid;
6159 	nfsstat4 status;
6160 	caller_context_t ct;
6161 	int fflags = 0;
6162 	int recall = 0;
6163 	int err;
6164 	int cmd;
6165 
6166 	/* get the file struct and hold a lock on it during initial open */
6167 	file = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate);
6168 	if (file == NULL) {
6169 		NFS4_DEBUG(rfs4_debug,
6170 		    (CE_NOTE, "rfs4_do_open: can't find file"));
6171 		resp->status = NFS4ERR_SERVERFAULT;
6172 		return;
6173 	}
6174 
6175 	state = rfs4_findstate_by_owner_file(oo, file, &screate);
6176 	if (state == NULL) {
6177 		NFS4_DEBUG(rfs4_debug,
6178 		    (CE_NOTE, "rfs4_do_open: can't find state"));
6179 		resp->status = NFS4ERR_RESOURCE;
6180 		/* No need to keep any reference */
6181 		rfs4_file_rele_withunlock(file);
6182 		return;
6183 	}
6184 
6185 	/* try to get the sysid before continuing */
6186 	if ((status = rfs4_client_sysid(oo->client, &sysid)) != NFS4_OK) {
6187 		resp->status = status;
6188 		rfs4_file_rele(file);
6189 		/* Not a fully formed open; "close" it */
6190 		if (screate == TRUE)
6191 			rfs4_state_close(state, FALSE, FALSE, cs->cr);
6192 		rfs4_state_rele(state);
6193 		return;
6194 	}
6195 
6196 	/* Calculate the fflags for this OPEN. */
6197 	if (access & OPEN4_SHARE_ACCESS_READ)
6198 		fflags |= FREAD;
6199 	if (access & OPEN4_SHARE_ACCESS_WRITE)
6200 		fflags |= FWRITE;
6201 
6202 	/*
6203 	 * Calculate the new deny and access mode that this open is adding to
6204 	 * the file for this open owner;
6205 	 */
6206 	dmodes = (deny & ~state->share_deny);
6207 	amodes = (access & ~state->share_access);
6208 
6209 	/*
6210 	 * Check to see the client has already sent an open for this
6211 	 * open owner on this file with the same share/deny modes.
6212 	 * If so, we don't need to check for a conflict and we don't
6213 	 * need to add another shrlock.  If not, then we need to
6214 	 * check for conflicts in deny and access before checking for
6215 	 * conflicts in delegation.  We don't want to recall a
6216 	 * delegation based on an open that will eventually fail based
6217 	 * on shares modes.
6218 	 */
6219 
6220 	if (dmodes || amodes) {
6221 		shr.s_access = (short)access;
6222 		shr.s_deny = (short)deny;
6223 		shr.s_pid = rfs4_dbe_getid(oo->dbe);
6224 		shr.s_sysid = sysid;
6225 		shr_loco.sl_pid = shr.s_pid;
6226 		shr_loco.sl_id = shr.s_sysid;
6227 		shr.s_owner = (caddr_t)&shr_loco;
6228 		shr.s_own_len = sizeof (shr_loco);
6229 
6230 		cmd = nbl_need_check(cs->vp) ? F_SHARE_NBMAND : F_SHARE;
6231 		if ((err = vop_shrlock(cs->vp, cmd, &shr, fflags)) != 0) {
6232 
6233 			resp->status = err == EAGAIN ?
6234 			    NFS4ERR_SHARE_DENIED : puterrno4(err);
6235 
6236 			rfs4_file_rele(file);
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 			return;
6242 		}
6243 	}
6244 
6245 	rfs4_dbe_lock(state->dbe);
6246 	rfs4_dbe_lock(file->dbe);
6247 
6248 	/*
6249 	 * Check to see if this file is delegated and if so, if a
6250 	 * recall needs to be done.
6251 	 */
6252 	if (rfs4_check_recall(state, access)) {
6253 		rfs4_dbe_unlock(file->dbe);
6254 		rfs4_dbe_unlock(state->dbe);
6255 		rfs4_recall_deleg(file, FALSE, state->owner->client);
6256 		delay(NFS4_DELEGATION_CONFLICT_DELAY);
6257 		rfs4_dbe_lock(state->dbe);
6258 		rfs4_dbe_lock(file->dbe);
6259 		/* Let's see if the delegation was returned */
6260 		if (rfs4_check_recall(state, access)) {
6261 			rfs4_dbe_unlock(file->dbe);
6262 			rfs4_dbe_unlock(state->dbe);
6263 			rfs4_file_rele(file);
6264 			rfs4_update_lease(state->owner->client);
6265 			(void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags);
6266 			/* Not a fully formed open; "close" it */
6267 			if (screate == TRUE)
6268 				rfs4_state_close(state, FALSE, FALSE, cs->cr);
6269 			rfs4_state_rele(state);
6270 			resp->status = NFS4ERR_DELAY;
6271 			return;
6272 		}
6273 	}
6274 	/*
6275 	 * the share check passed and any delegation conflict has been
6276 	 * taken care of, now call vop_open.
6277 	 * if this is the first open then call vop_open with fflags.
6278 	 * if not, call vn_open_upgrade with just the upgrade flags.
6279 	 *
6280 	 * if the file has been opened already, it will have the current
6281 	 * access mode in the state struct.  if it has no share access, then
6282 	 * this is a new open.
6283 	 *
6284 	 * However, if this is open with CLAIM_DLEGATE_CUR, then don't
6285 	 * call VOP_OPEN(), just do the open upgrade.
6286 	 */
6287 	if (((state->share_access & OPEN4_SHARE_ACCESS_BOTH) == 0) &&
6288 	    !deleg_cur) {
6289 		ct.cc_sysid = sysid;
6290 		ct.cc_pid = shr.s_pid;
6291 		ct.cc_caller_id = nfs4_srv_caller_id;
6292 		err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct);
6293 		if (err) {
6294 			rfs4_dbe_unlock(file->dbe);
6295 			rfs4_dbe_unlock(state->dbe);
6296 			rfs4_file_rele(file);
6297 			(void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags);
6298 			/* Not a fully formed open; "close" it */
6299 			if (screate == TRUE)
6300 				rfs4_state_close(state, FALSE, FALSE, cs->cr);
6301 			rfs4_state_rele(state);
6302 			resp->status = NFS4ERR_SERVERFAULT;
6303 			return;
6304 		}
6305 	} else { /* open upgrade */
6306 		/*
6307 		 * calculate the fflags for the new mode that is being added
6308 		 * by this upgrade.
6309 		 */
6310 		fflags = 0;
6311 		if (amodes & OPEN4_SHARE_ACCESS_READ)
6312 			fflags |= FREAD;
6313 		if (amodes & OPEN4_SHARE_ACCESS_WRITE)
6314 			fflags |= FWRITE;
6315 		vn_open_upgrade(cs->vp, fflags);
6316 	}
6317 
6318 	if (dmodes & OPEN4_SHARE_DENY_READ)
6319 		file->deny_read++;
6320 	if (dmodes & OPEN4_SHARE_DENY_WRITE)
6321 		file->deny_write++;
6322 	file->share_deny |= deny;
6323 	state->share_deny |= deny;
6324 
6325 	if (amodes & OPEN4_SHARE_ACCESS_READ)
6326 		file->access_read++;
6327 	if (amodes & OPEN4_SHARE_ACCESS_WRITE)
6328 		file->access_write++;
6329 	file->share_access |= access;
6330 	state->share_access |= access;
6331 
6332 	/*
6333 	 * Check for delegation here. if the deleg argument is not
6334 	 * DELEG_ANY, then this is a reclaim from a client and
6335 	 * we must honor the delegation requested. If necessary we can
6336 	 * set the recall flag.
6337 	 */
6338 
6339 	dsp = rfs4_grant_delegation(deleg, state, &recall);
6340 
6341 	cs->deleg = (file->dinfo->dtype == OPEN_DELEGATE_WRITE);
6342 
6343 	next_stateid(&state->stateid);
6344 
6345 	resp->stateid = state->stateid.stateid;
6346 
6347 	rfs4_dbe_unlock(file->dbe);
6348 	rfs4_dbe_unlock(state->dbe);
6349 
6350 	if (dsp) {
6351 		rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall);
6352 		rfs4_deleg_state_rele(dsp);
6353 	}
6354 
6355 	rfs4_file_rele(file);
6356 	rfs4_state_rele(state);
6357 
6358 	resp->status = NFS4_OK;
6359 }
6360 
6361 /*ARGSUSED*/
6362 static void
6363 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req,
6364 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6365 {
6366 	change_info4 *cinfo = &resp->cinfo;
6367 	bitmap4 *attrset = &resp->attrset;
6368 
6369 	if (args->opentype == OPEN4_NOCREATE)
6370 		resp->status = rfs4_lookupfile(&args->open_claim4_u.file,
6371 		    req, cs, args->share_access, cinfo);
6372 	else {
6373 		/* inhibit delegation grants during exclusive create */
6374 
6375 		if (args->mode == EXCLUSIVE4)
6376 			rfs4_disable_delegation();
6377 
6378 		resp->status = rfs4_createfile(args, req, cs, cinfo, attrset,
6379 		    oo->client->clientid);
6380 	}
6381 
6382 	if (resp->status == NFS4_OK) {
6383 
6384 		/* cs->vp cs->fh now reference the desired file */
6385 
6386 		rfs4_do_open(cs, req, oo, DELEG_ANY, args->share_access,
6387 		    args->share_deny, resp, 0);
6388 
6389 		/*
6390 		 * If rfs4_createfile set attrset, we must
6391 		 * clear this attrset before the response is copied.
6392 		 */
6393 		if (resp->status != NFS4_OK && resp->attrset) {
6394 			resp->attrset = 0;
6395 		}
6396 	}
6397 	else
6398 		*cs->statusp = resp->status;
6399 
6400 	if (args->mode == EXCLUSIVE4)
6401 		rfs4_enable_delegation();
6402 }
6403 
6404 /*ARGSUSED*/
6405 static void
6406 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req,
6407 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6408 {
6409 	change_info4 *cinfo = &resp->cinfo;
6410 	vattr_t va;
6411 	vtype_t v_type = cs->vp->v_type;
6412 	int error = 0;
6413 
6414 	/* Verify that we have a regular file */
6415 	if (v_type != VREG) {
6416 		if (v_type == VDIR)
6417 			resp->status = NFS4ERR_ISDIR;
6418 		else if (v_type == VLNK)
6419 			resp->status = NFS4ERR_SYMLINK;
6420 		else
6421 			resp->status = NFS4ERR_INVAL;
6422 		return;
6423 	}
6424 
6425 	va.va_mask = AT_MODE|AT_UID;
6426 	error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL);
6427 	if (error) {
6428 		resp->status = puterrno4(error);
6429 		return;
6430 	}
6431 
6432 	cs->mandlock = MANDLOCK(cs->vp, va.va_mode);
6433 
6434 	/*
6435 	 * Check if we have access to the file, Note the the file
6436 	 * could have originally been open UNCHECKED or GUARDED
6437 	 * with mode bits that will now fail, but there is nothing
6438 	 * we can really do about that except in the case that the
6439 	 * owner of the file is the one requesting the open.
6440 	 */
6441 	if (crgetuid(cs->cr) != va.va_uid) {
6442 		resp->status = check_open_access(args->share_access, cs, req);
6443 		if (resp->status != NFS4_OK) {
6444 			return;
6445 		}
6446 	}
6447 
6448 	/*
6449 	 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero
6450 	 */
6451 	cinfo->before = 0;
6452 	cinfo->after = 0;
6453 	cinfo->atomic = FALSE;
6454 
6455 	rfs4_do_open(cs, req, oo,
6456 	    NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type),
6457 	    args->share_access, args->share_deny, resp, 0);
6458 }
6459 
6460 static void
6461 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req,
6462 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6463 {
6464 	int error;
6465 	nfsstat4 status;
6466 	stateid4 stateid =
6467 	    args->open_claim4_u.delegate_cur_info.delegate_stateid;
6468 	rfs4_deleg_state_t *dsp;
6469 
6470 	/*
6471 	 * Find the state info from the stateid and confirm that the
6472 	 * file is delegated.  If the state openowner is the same as
6473 	 * the supplied openowner we're done. If not, get the file
6474 	 * info from the found state info. Use that file info to
6475 	 * create the state for this lock owner. Note solaris doen't
6476 	 * really need the pathname to find the file. We may want to
6477 	 * lookup the pathname and make sure that the vp exist and
6478 	 * matches the vp in the file structure. However it is
6479 	 * possible that the pathname nolonger exists (local process
6480 	 * unlinks the file), so this may not be that useful.
6481 	 */
6482 
6483 	status = rfs4_get_deleg_state(&stateid, &dsp);
6484 	if (status != NFS4_OK) {
6485 		resp->status = status;
6486 		return;
6487 	}
6488 
6489 	ASSERT(dsp->finfo->dinfo->dtype != OPEN_DELEGATE_NONE);
6490 
6491 	/*
6492 	 * New lock owner, create state. Since this was probably called
6493 	 * in response to a CB_RECALL we set deleg to DELEG_NONE
6494 	 */
6495 
6496 	ASSERT(cs->vp != NULL);
6497 	VN_RELE(cs->vp);
6498 	VN_HOLD(dsp->finfo->vp);
6499 	cs->vp = dsp->finfo->vp;
6500 
6501 	if (error = makefh4(&cs->fh, cs->vp, cs->exi)) {
6502 		rfs4_deleg_state_rele(dsp);
6503 		*cs->statusp = resp->status = puterrno4(error);
6504 		return;
6505 	}
6506 
6507 	/* Mark progress for delegation returns */
6508 	dsp->finfo->dinfo->time_lastwrite = gethrestime_sec();
6509 	rfs4_deleg_state_rele(dsp);
6510 	rfs4_do_open(cs, req, oo, DELEG_NONE,
6511 	    args->share_access, args->share_deny, resp, 1);
6512 }
6513 
6514 /*ARGSUSED*/
6515 static void
6516 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req,
6517 			OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6518 {
6519 	/*
6520 	 * Lookup the pathname, it must already exist since this file
6521 	 * was delegated.
6522 	 *
6523 	 * Find the file and state info for this vp and open owner pair.
6524 	 *	check that they are in fact delegated.
6525 	 *	check that the state access and deny modes are the same.
6526 	 *
6527 	 * Return the delgation possibly seting the recall flag.
6528 	 */
6529 	rfs4_file_t *file;
6530 	rfs4_state_t *state;
6531 	bool_t create = FALSE;
6532 	bool_t dcreate = FALSE;
6533 	rfs4_deleg_state_t *dsp;
6534 	nfsace4 *ace;
6535 
6536 
6537 	/* Note we ignore oflags */
6538 	resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev,
6539 	    req, cs, args->share_access, &resp->cinfo);
6540 
6541 	if (resp->status != NFS4_OK) {
6542 		return;
6543 	}
6544 
6545 	/* get the file struct and hold a lock on it during initial open */
6546 	file = rfs4_findfile_withlock(cs->vp, NULL, &create);
6547 	if (file == NULL) {
6548 		NFS4_DEBUG(rfs4_debug,
6549 		    (CE_NOTE, "rfs4_do_opendelprev: can't find file"));
6550 		resp->status = NFS4ERR_SERVERFAULT;
6551 		return;
6552 	}
6553 
6554 	state = rfs4_findstate_by_owner_file(oo, file, &create);
6555 	if (state == NULL) {
6556 		NFS4_DEBUG(rfs4_debug,
6557 		    (CE_NOTE, "rfs4_do_opendelprev: can't find state"));
6558 		resp->status = NFS4ERR_SERVERFAULT;
6559 		rfs4_file_rele_withunlock(file);
6560 		return;
6561 	}
6562 
6563 	rfs4_dbe_lock(state->dbe);
6564 	rfs4_dbe_lock(file->dbe);
6565 	if (args->share_access != state->share_access ||
6566 	    args->share_deny != state->share_deny ||
6567 	    state->finfo->dinfo->dtype == OPEN_DELEGATE_NONE) {
6568 		NFS4_DEBUG(rfs4_debug,
6569 		    (CE_NOTE, "rfs4_do_opendelprev: state mixup"));
6570 		rfs4_dbe_unlock(file->dbe);
6571 		rfs4_dbe_unlock(state->dbe);
6572 		rfs4_file_rele(file);
6573 		rfs4_state_rele(state);
6574 		resp->status = NFS4ERR_SERVERFAULT;
6575 		return;
6576 	}
6577 	rfs4_dbe_unlock(file->dbe);
6578 	rfs4_dbe_unlock(state->dbe);
6579 
6580 	dsp = rfs4_finddeleg(state, &dcreate);
6581 	if (dsp == NULL) {
6582 		rfs4_state_rele(state);
6583 		rfs4_file_rele(file);
6584 		resp->status = NFS4ERR_SERVERFAULT;
6585 		return;
6586 	}
6587 
6588 	next_stateid(&state->stateid);
6589 
6590 	resp->stateid = state->stateid.stateid;
6591 
6592 	resp->delegation.delegation_type = dsp->dtype;
6593 
6594 	if (dsp->dtype == OPEN_DELEGATE_READ) {
6595 		open_read_delegation4 *rv =
6596 		    &resp->delegation.open_delegation4_u.read;
6597 
6598 		rv->stateid = dsp->delegid.stateid;
6599 		rv->recall = FALSE; /* no policy in place to set to TRUE */
6600 		ace = &rv->permissions;
6601 	} else {
6602 		open_write_delegation4 *rv =
6603 		    &resp->delegation.open_delegation4_u.write;
6604 
6605 		rv->stateid = dsp->delegid.stateid;
6606 		rv->recall = FALSE;  /* no policy in place to set to TRUE */
6607 		ace = &rv->permissions;
6608 		rv->space_limit.limitby = NFS_LIMIT_SIZE;
6609 		rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX;
6610 	}
6611 
6612 	/* XXX For now */
6613 	ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE;
6614 	ace->flag = 0;
6615 	ace->access_mask = 0;
6616 	ace->who.utf8string_len = 0;
6617 	ace->who.utf8string_val = 0;
6618 
6619 	rfs4_deleg_state_rele(dsp);
6620 	rfs4_state_rele(state);
6621 	rfs4_file_rele(file);
6622 }
6623 
6624 typedef enum {
6625 	NFS4_CHKSEQ_OKAY = 0,
6626 	NFS4_CHKSEQ_REPLAY = 1,
6627 	NFS4_CHKSEQ_BAD = 2
6628 } rfs4_chkseq_t;
6629 
6630 /*
6631  * Generic function for sequence number checks.
6632  */
6633 static rfs4_chkseq_t
6634 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop,
6635 		seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres)
6636 {
6637 	/* Same sequence ids and matching operations? */
6638 	if (seqid == rqst_seq && resop->resop == lastop->resop) {
6639 		if (copyres == TRUE) {
6640 			rfs4_free_reply(resop);
6641 			rfs4_copy_reply(resop, lastop);
6642 		}
6643 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
6644 		    "Replayed SEQID %d\n", seqid));
6645 		return (NFS4_CHKSEQ_REPLAY);
6646 	}
6647 
6648 	/* If the incoming sequence is not the next expected then it is bad */
6649 	if (rqst_seq != seqid + 1) {
6650 		if (rqst_seq == seqid) {
6651 			NFS4_DEBUG(rfs4_debug,
6652 			    (CE_NOTE, "BAD SEQID: Replayed sequence id "
6653 			    "but last op was %d current op is %d\n",
6654 			    lastop->resop, resop->resop));
6655 			return (NFS4_CHKSEQ_BAD);
6656 		}
6657 		NFS4_DEBUG(rfs4_debug,
6658 		    (CE_NOTE, "BAD SEQID: got %u expecting %u\n",
6659 		    rqst_seq, seqid));
6660 		return (NFS4_CHKSEQ_BAD);
6661 	}
6662 
6663 	/* Everything okay -- next expected */
6664 	return (NFS4_CHKSEQ_OKAY);
6665 }
6666 
6667 
6668 static rfs4_chkseq_t
6669 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop)
6670 {
6671 	rfs4_chkseq_t rc;
6672 
6673 	rfs4_dbe_lock(op->dbe);
6674 	rc = rfs4_check_seqid(op->open_seqid, op->reply, seqid, resop, TRUE);
6675 	rfs4_dbe_unlock(op->dbe);
6676 
6677 	if (rc == NFS4_CHKSEQ_OKAY)
6678 		rfs4_update_lease(op->client);
6679 
6680 	return (rc);
6681 }
6682 
6683 static rfs4_chkseq_t
6684 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op,
6685 	nfs_resop4 *resop)
6686 {
6687 	rfs4_chkseq_t rc;
6688 
6689 	rfs4_dbe_lock(op->dbe);
6690 	rc = rfs4_check_seqid(op->open_seqid, op->reply,
6691 	    olo_seqid, resop, FALSE);
6692 	rfs4_dbe_unlock(op->dbe);
6693 
6694 	return (rc);
6695 }
6696 
6697 static rfs4_chkseq_t
6698 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lp, nfs_resop4 *resop)
6699 {
6700 	rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY;
6701 
6702 	rfs4_dbe_lock(lp->dbe);
6703 	if (!lp->skip_seqid_check)
6704 		rc = rfs4_check_seqid(lp->seqid, lp->reply, seqid, resop, TRUE);
6705 	rfs4_dbe_unlock(lp->dbe);
6706 
6707 	return (rc);
6708 }
6709 
6710 static void
6711 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop,
6712 	    struct svc_req *req, struct compound_state *cs)
6713 {
6714 	OPEN4args *args = &argop->nfs_argop4_u.opopen;
6715 	OPEN4res *resp = &resop->nfs_resop4_u.opopen;
6716 	open_owner4 *owner = &args->owner;
6717 	open_claim_type4 claim = args->claim;
6718 	rfs4_client_t *cp;
6719 	rfs4_openowner_t *oo;
6720 	bool_t create;
6721 	bool_t replay = FALSE;
6722 	int can_reclaim;
6723 
6724 
6725 	if (cs->vp == NULL) {
6726 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
6727 		return;
6728 	}
6729 
6730 	/*
6731 	 * Need to check clientid and lease expiration first based on
6732 	 * error ordering and incrementing sequence id.
6733 	 */
6734 	cp = rfs4_findclient_by_id(owner->clientid, FALSE);
6735 	if (cp == NULL) {
6736 		*cs->statusp = resp->status =
6737 		    rfs4_check_clientid(&owner->clientid, 0);
6738 		return;
6739 	}
6740 
6741 	if (rfs4_lease_expired(cp)) {
6742 		rfs4_client_close(cp);
6743 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
6744 		return;
6745 	}
6746 	can_reclaim = cp->can_reclaim;
6747 
6748 	/*
6749 	 * Find the open_owner for use from this point forward.  Take
6750 	 * care in updating the sequence id based on the type of error
6751 	 * being returned.
6752 	 */
6753 retry:
6754 	create = TRUE;
6755 	oo = rfs4_findopenowner(owner, &create, args->seqid);
6756 	if (oo == NULL) {
6757 		*cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
6758 		rfs4_client_rele(cp);
6759 		return;
6760 	}
6761 
6762 	/* Hold off access to the sequence space while the open is done */
6763 	rfs4_sw_enter(&oo->oo_sw);
6764 
6765 	/*
6766 	 * If the open_owner existed before at the server, then check
6767 	 * the sequence id.
6768 	 */
6769 	if (!create && !oo->postpone_confirm) {
6770 		switch (rfs4_check_open_seqid(args->seqid, oo, resop)) {
6771 		case NFS4_CHKSEQ_BAD:
6772 			if ((args->seqid > oo->open_seqid) &&
6773 			    oo->need_confirm) {
6774 				rfs4_free_opens(oo, TRUE, FALSE);
6775 				rfs4_sw_exit(&oo->oo_sw);
6776 				rfs4_openowner_rele(oo);
6777 				goto retry;
6778 			}
6779 			resp->status = NFS4ERR_BAD_SEQID;
6780 			goto out;
6781 		case NFS4_CHKSEQ_REPLAY: /* replay of previous request */
6782 			replay = TRUE;
6783 			goto out;
6784 		default:
6785 			break;
6786 		}
6787 
6788 		/*
6789 		 * Sequence was ok and open owner exists
6790 		 * check to see if we have yet to see an
6791 		 * open_confirm.
6792 		 */
6793 		if (oo->need_confirm) {
6794 			rfs4_free_opens(oo, TRUE, FALSE);
6795 			rfs4_sw_exit(&oo->oo_sw);
6796 			rfs4_openowner_rele(oo);
6797 			goto retry;
6798 		}
6799 	}
6800 	/* Grace only applies to regular-type OPENs */
6801 	if (rfs4_clnt_in_grace(cp) &&
6802 	    (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) {
6803 		*cs->statusp = resp->status = NFS4ERR_GRACE;
6804 		goto out;
6805 	}
6806 
6807 	/*
6808 	 * If previous state at the server existed then can_reclaim
6809 	 * will be set. If not reply NFS4ERR_NO_GRACE to the
6810 	 * client.
6811 	 */
6812 	if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) {
6813 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
6814 		goto out;
6815 	}
6816 
6817 
6818 	/*
6819 	 * Reject the open if the client has missed the grace period
6820 	 */
6821 	if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) {
6822 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
6823 		goto out;
6824 	}
6825 
6826 	/* Couple of up-front bookkeeping items */
6827 	if (oo->need_confirm) {
6828 		/*
6829 		 * If this is a reclaim OPEN then we should not ask
6830 		 * for a confirmation of the open_owner per the
6831 		 * protocol specification.
6832 		 */
6833 		if (claim == CLAIM_PREVIOUS)
6834 			oo->need_confirm = FALSE;
6835 		else
6836 			resp->rflags |= OPEN4_RESULT_CONFIRM;
6837 	}
6838 	resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX;
6839 
6840 	/*
6841 	 * If there is an unshared filesystem mounted on this vnode,
6842 	 * do not allow to open/create in this directory.
6843 	 */
6844 	if (vn_ismntpt(cs->vp)) {
6845 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
6846 		goto out;
6847 	}
6848 
6849 	/*
6850 	 * access must READ, WRITE, or BOTH.  No access is invalid.
6851 	 * deny can be READ, WRITE, BOTH, or NONE.
6852 	 * bits not defined for access/deny are invalid.
6853 	 */
6854 	if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) ||
6855 	    (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) ||
6856 	    (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) {
6857 		*cs->statusp = resp->status = NFS4ERR_INVAL;
6858 		goto out;
6859 	}
6860 
6861 
6862 	/*
6863 	 * make sure attrset is zero before response is built.
6864 	 */
6865 	resp->attrset = 0;
6866 
6867 	switch (claim) {
6868 	case CLAIM_NULL:
6869 		rfs4_do_opennull(cs, req, args, oo, resp);
6870 		break;
6871 	case CLAIM_PREVIOUS:
6872 		rfs4_do_openprev(cs, req, args, oo, resp);
6873 		break;
6874 	case CLAIM_DELEGATE_CUR:
6875 		rfs4_do_opendelcur(cs, req, args, oo, resp);
6876 		break;
6877 	case CLAIM_DELEGATE_PREV:
6878 		rfs4_do_opendelprev(cs, req, args, oo, resp);
6879 		break;
6880 	default:
6881 		resp->status = NFS4ERR_INVAL;
6882 		break;
6883 	}
6884 
6885 out:
6886 	rfs4_client_rele(cp);
6887 
6888 	/* Catch sequence id handling here to make it a little easier */
6889 	switch (resp->status) {
6890 	case NFS4ERR_BADXDR:
6891 	case NFS4ERR_BAD_SEQID:
6892 	case NFS4ERR_BAD_STATEID:
6893 	case NFS4ERR_NOFILEHANDLE:
6894 	case NFS4ERR_RESOURCE:
6895 	case NFS4ERR_STALE_CLIENTID:
6896 	case NFS4ERR_STALE_STATEID:
6897 		/*
6898 		 * The protocol states that if any of these errors are
6899 		 * being returned, the sequence id should not be
6900 		 * incremented.  Any other return requires an
6901 		 * increment.
6902 		 */
6903 		break;
6904 	default:
6905 		/* Always update the lease in this case */
6906 		rfs4_update_lease(oo->client);
6907 
6908 		/* Regular response - copy the result */
6909 		if (!replay)
6910 			rfs4_update_open_resp(oo, resop, &cs->fh);
6911 
6912 		/*
6913 		 * REPLAY case: Only if the previous response was OK
6914 		 * do we copy the filehandle.  If not OK, no
6915 		 * filehandle to copy.
6916 		 */
6917 		if (replay == TRUE &&
6918 		    resp->status == NFS4_OK &&
6919 		    oo->reply_fh.nfs_fh4_val) {
6920 			/*
6921 			 * If this is a replay, we must restore the
6922 			 * current filehandle/vp to that of what was
6923 			 * returned originally.  Try our best to do
6924 			 * it.
6925 			 */
6926 			nfs_fh4_fmt_t *fh_fmtp =
6927 			    (nfs_fh4_fmt_t *)oo->reply_fh.nfs_fh4_val;
6928 
6929 			cs->exi = checkexport4(&fh_fmtp->fh4_fsid,
6930 			    (fid_t *)&fh_fmtp->fh4_xlen, NULL);
6931 
6932 			if (cs->exi == NULL) {
6933 				resp->status = NFS4ERR_STALE;
6934 				goto finish;
6935 			}
6936 
6937 			VN_RELE(cs->vp);
6938 
6939 			cs->vp = nfs4_fhtovp(&oo->reply_fh, cs->exi,
6940 			    &resp->status);
6941 
6942 			if (cs->vp == NULL)
6943 				goto finish;
6944 
6945 			nfs_fh4_copy(&oo->reply_fh, &cs->fh);
6946 		}
6947 
6948 		/*
6949 		 * If this was a replay, no need to update the
6950 		 * sequence id. If the open_owner was not created on
6951 		 * this pass, then update.  The first use of an
6952 		 * open_owner will not bump the sequence id.
6953 		 */
6954 		if (replay == FALSE && !create)
6955 			rfs4_update_open_sequence(oo);
6956 		/*
6957 		 * If the client is receiving an error and the
6958 		 * open_owner needs to be confirmed, there is no way
6959 		 * to notify the client of this fact ignoring the fact
6960 		 * that the server has no method of returning a
6961 		 * stateid to confirm.  Therefore, the server needs to
6962 		 * mark this open_owner in a way as to avoid the
6963 		 * sequence id checking the next time the client uses
6964 		 * this open_owner.
6965 		 */
6966 		if (resp->status != NFS4_OK && oo->need_confirm)
6967 			oo->postpone_confirm = TRUE;
6968 		/*
6969 		 * If OK response then clear the postpone flag and
6970 		 * reset the sequence id to keep in sync with the
6971 		 * client.
6972 		 */
6973 		if (resp->status == NFS4_OK && oo->postpone_confirm) {
6974 			oo->postpone_confirm = FALSE;
6975 			oo->open_seqid = args->seqid;
6976 		}
6977 		break;
6978 	}
6979 
6980 finish:
6981 	*cs->statusp = resp->status;
6982 
6983 	rfs4_sw_exit(&oo->oo_sw);
6984 	rfs4_openowner_rele(oo);
6985 }
6986 
6987 /*ARGSUSED*/
6988 void
6989 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
6990 		    struct svc_req *req, struct compound_state *cs)
6991 {
6992 	OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm;
6993 	OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm;
6994 	rfs4_state_t *sp;
6995 	nfsstat4 status;
6996 
6997 	if (cs->vp == NULL) {
6998 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
6999 		return;
7000 	}
7001 
7002 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
7003 	if (status != NFS4_OK) {
7004 		*cs->statusp = resp->status = status;
7005 		return;
7006 	}
7007 
7008 	/* Ensure specified filehandle matches */
7009 	if (cs->vp != sp->finfo->vp) {
7010 		rfs4_state_rele(sp);
7011 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7012 		return;
7013 	}
7014 
7015 	/* hold off other access to open_owner while we tinker */
7016 	rfs4_sw_enter(&sp->owner->oo_sw);
7017 
7018 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7019 	case NFS4_CHECK_STATEID_OKAY:
7020 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7021 		    resop) != 0) {
7022 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7023 			break;
7024 		}
7025 		/*
7026 		 * If it is the appropriate stateid and determined to
7027 		 * be "OKAY" then this means that the stateid does not
7028 		 * need to be confirmed and the client is in error for
7029 		 * sending an OPEN_CONFIRM.
7030 		 */
7031 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7032 		break;
7033 	case NFS4_CHECK_STATEID_OLD:
7034 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7035 		break;
7036 	case NFS4_CHECK_STATEID_BAD:
7037 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7038 		break;
7039 	case NFS4_CHECK_STATEID_EXPIRED:
7040 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7041 		break;
7042 	case NFS4_CHECK_STATEID_CLOSED:
7043 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7044 		break;
7045 	case NFS4_CHECK_STATEID_REPLAY:
7046 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
7047 		case NFS4_CHKSEQ_OKAY:
7048 			/*
7049 			 * This is replayed stateid; if seqid matches
7050 			 * next expected, then client is using wrong seqid.
7051 			 */
7052 			/* fall through */
7053 		case NFS4_CHKSEQ_BAD:
7054 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7055 			break;
7056 		case NFS4_CHKSEQ_REPLAY:
7057 			/*
7058 			 * Note this case is the duplicate case so
7059 			 * resp->status is already set.
7060 			 */
7061 			*cs->statusp = resp->status;
7062 			rfs4_update_lease(sp->owner->client);
7063 			break;
7064 		}
7065 		break;
7066 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7067 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7068 		    resop) != NFS4_CHKSEQ_OKAY) {
7069 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7070 			break;
7071 		}
7072 		*cs->statusp = resp->status = NFS4_OK;
7073 
7074 		next_stateid(&sp->stateid);
7075 		resp->open_stateid = sp->stateid.stateid;
7076 		sp->owner->need_confirm = FALSE;
7077 		rfs4_update_lease(sp->owner->client);
7078 		rfs4_update_open_sequence(sp->owner);
7079 		rfs4_update_open_resp(sp->owner, resop, NULL);
7080 		break;
7081 	default:
7082 		ASSERT(FALSE);
7083 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7084 		break;
7085 	}
7086 	rfs4_sw_exit(&sp->owner->oo_sw);
7087 	rfs4_state_rele(sp);
7088 }
7089 
7090 /*ARGSUSED*/
7091 void
7092 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop,
7093 		    struct svc_req *req, struct compound_state *cs)
7094 {
7095 	OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade;
7096 	OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade;
7097 	uint32_t access = args->share_access;
7098 	uint32_t deny = args->share_deny;
7099 	nfsstat4 status;
7100 	rfs4_state_t *sp;
7101 	rfs4_file_t *fp;
7102 	int fflags = 0;
7103 
7104 	if (cs->vp == NULL) {
7105 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7106 		return;
7107 	}
7108 
7109 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
7110 	if (status != NFS4_OK) {
7111 		*cs->statusp = resp->status = status;
7112 		return;
7113 	}
7114 
7115 	/* Ensure specified filehandle matches */
7116 	if (cs->vp != sp->finfo->vp) {
7117 		rfs4_state_rele(sp);
7118 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7119 		return;
7120 	}
7121 
7122 	/* hold off other access to open_owner while we tinker */
7123 	rfs4_sw_enter(&sp->owner->oo_sw);
7124 
7125 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7126 	case NFS4_CHECK_STATEID_OKAY:
7127 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7128 		    resop) != NFS4_CHKSEQ_OKAY) {
7129 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7130 			goto end;
7131 		}
7132 		break;
7133 	case NFS4_CHECK_STATEID_OLD:
7134 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7135 		goto end;
7136 	case NFS4_CHECK_STATEID_BAD:
7137 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7138 		goto end;
7139 	case NFS4_CHECK_STATEID_EXPIRED:
7140 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7141 		goto end;
7142 	case NFS4_CHECK_STATEID_CLOSED:
7143 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7144 		goto end;
7145 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7146 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7147 		goto end;
7148 	case NFS4_CHECK_STATEID_REPLAY:
7149 		/* Check the sequence id for the open owner */
7150 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
7151 		case NFS4_CHKSEQ_OKAY:
7152 			/*
7153 			 * This is replayed stateid; if seqid matches
7154 			 * next expected, then client is using wrong seqid.
7155 			 */
7156 			/* fall through */
7157 		case NFS4_CHKSEQ_BAD:
7158 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7159 			goto end;
7160 		case NFS4_CHKSEQ_REPLAY:
7161 			/*
7162 			 * Note this case is the duplicate case so
7163 			 * resp->status is already set.
7164 			 */
7165 			*cs->statusp = resp->status;
7166 			rfs4_update_lease(sp->owner->client);
7167 			goto end;
7168 		}
7169 		break;
7170 	default:
7171 		ASSERT(FALSE);
7172 		break;
7173 	}
7174 
7175 	rfs4_dbe_lock(sp->dbe);
7176 	/*
7177 	 * Check that the new access modes and deny modes are valid.
7178 	 * Check that no invalid bits are set.
7179 	 */
7180 	if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) ||
7181 	    (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) {
7182 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7183 		rfs4_update_open_sequence(sp->owner);
7184 		rfs4_dbe_unlock(sp->dbe);
7185 		goto end;
7186 	}
7187 
7188 	/*
7189 	 * The new modes must be a subset of the current modes and
7190 	 * the access must specify at least one mode. To test that
7191 	 * the new mode is a subset of the current modes we bitwise
7192 	 * AND them together and check that the result equals the new
7193 	 * mode. For example:
7194 	 * New mode, access == R and current mode, sp->share_access  == RW
7195 	 * access & sp->share_access == R == access, so the new access mode
7196 	 * is valid. Consider access == RW, sp->share_access = R
7197 	 * access & sp->share_access == R != access, so the new access mode
7198 	 * is invalid.
7199 	 */
7200 	if ((access & sp->share_access) != access ||
7201 	    (deny & sp->share_deny) != deny ||
7202 	    (access &
7203 	    (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) {
7204 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7205 		rfs4_update_open_sequence(sp->owner);
7206 		rfs4_dbe_unlock(sp->dbe);
7207 		goto end;
7208 	}
7209 
7210 	/*
7211 	 * Release any share locks associated with this stateID.
7212 	 * Strictly speaking, this violates the spec because the
7213 	 * spec effectively requires that open downgrade be atomic.
7214 	 * At present, fs_shrlock does not have this capability.
7215 	 */
7216 	rfs4_dbe_unlock(sp->dbe);
7217 	rfs4_unshare(sp);
7218 	rfs4_dbe_lock(sp->dbe);
7219 
7220 	fp = sp->finfo;
7221 	rfs4_dbe_lock(fp->dbe);
7222 
7223 	/*
7224 	 * If the current mode has deny read and the new mode
7225 	 * does not, decrement the number of deny read mode bits
7226 	 * and if it goes to zero turn off the deny read bit
7227 	 * on the file.
7228 	 */
7229 	if ((sp->share_deny & OPEN4_SHARE_DENY_READ) &&
7230 	    (deny & OPEN4_SHARE_DENY_READ) == 0) {
7231 		fp->deny_read--;
7232 		if (fp->deny_read == 0)
7233 			fp->share_deny &= ~OPEN4_SHARE_DENY_READ;
7234 	}
7235 
7236 	/*
7237 	 * If the current mode has deny write and the new mode
7238 	 * does not, decrement the number of deny write mode bits
7239 	 * and if it goes to zero turn off the deny write bit
7240 	 * on the file.
7241 	 */
7242 	if ((sp->share_deny & OPEN4_SHARE_DENY_WRITE) &&
7243 	    (deny & OPEN4_SHARE_DENY_WRITE) == 0) {
7244 		fp->deny_write--;
7245 		if (fp->deny_write == 0)
7246 			fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE;
7247 	}
7248 
7249 	/*
7250 	 * If the current mode has access read and the new mode
7251 	 * does not, decrement the number of access read mode bits
7252 	 * and if it goes to zero turn off the access read bit
7253 	 * on the file.  set fflags to FREAD for the call to
7254 	 * vn_open_downgrade().
7255 	 */
7256 	if ((sp->share_access & OPEN4_SHARE_ACCESS_READ) &&
7257 	    (access & OPEN4_SHARE_ACCESS_READ) == 0) {
7258 		fp->access_read--;
7259 		if (fp->access_read == 0)
7260 			fp->share_access &= ~OPEN4_SHARE_ACCESS_READ;
7261 		fflags |= FREAD;
7262 	}
7263 
7264 	/*
7265 	 * If the current mode has access write and the new mode
7266 	 * does not, decrement the number of access write mode bits
7267 	 * and if it goes to zero turn off the access write bit
7268 	 * on the file.  set fflags to FWRITE for the call to
7269 	 * vn_open_downgrade().
7270 	 */
7271 	if ((sp->share_access & OPEN4_SHARE_ACCESS_WRITE) &&
7272 	    (access & OPEN4_SHARE_ACCESS_WRITE) == 0) {
7273 		fp->access_write--;
7274 		if (fp->access_write == 0)
7275 			fp->share_deny &= ~OPEN4_SHARE_ACCESS_WRITE;
7276 		fflags |= FWRITE;
7277 	}
7278 
7279 	/* Set the new access and deny modes */
7280 	sp->share_access = access;
7281 	sp->share_deny = deny;
7282 	/* Check that the file is still accessible */
7283 	ASSERT(fp->share_access);
7284 
7285 	rfs4_dbe_unlock(fp->dbe);
7286 
7287 	rfs4_dbe_unlock(sp->dbe);
7288 
7289 	if ((status = rfs4_share(sp)) != NFS4_OK) {
7290 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7291 		rfs4_update_open_sequence(sp->owner);
7292 		goto end;
7293 	}
7294 
7295 	/*
7296 	 * we successfully downgraded the share lock, now we need to downgrade
7297 	 * the open.  it is possible that the downgrade was only for a deny
7298 	 * mode and we have nothing else to do.
7299 	 */
7300 	if ((fflags & (FREAD|FWRITE)) != 0)
7301 		vn_open_downgrade(cs->vp, fflags);
7302 
7303 	rfs4_dbe_lock(sp->dbe);
7304 
7305 	/* Update the stateid */
7306 	next_stateid(&sp->stateid);
7307 	resp->open_stateid = sp->stateid.stateid;
7308 
7309 	rfs4_dbe_unlock(sp->dbe);
7310 
7311 	*cs->statusp = resp->status = NFS4_OK;
7312 	/* Update the lease */
7313 	rfs4_update_lease(sp->owner->client);
7314 	/* And the sequence */
7315 	rfs4_update_open_sequence(sp->owner);
7316 	rfs4_update_open_resp(sp->owner, resop, NULL);
7317 
7318 end:
7319 	rfs4_sw_exit(&sp->owner->oo_sw);
7320 	rfs4_state_rele(sp);
7321 }
7322 
7323 /*
7324  * The logic behind this function is detailed in the NFSv4 RFC in the
7325  * SETCLIENTID operation description under IMPLEMENTATION.  Refer to
7326  * that section for explicit guidance to server behavior for
7327  * SETCLIENTID.
7328  */
7329 void
7330 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop,
7331 		    struct svc_req *req, struct compound_state *cs)
7332 {
7333 	SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid;
7334 	SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid;
7335 	rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed;
7336 	bool_t create = TRUE;
7337 	char *addr, *netid;
7338 	int len;
7339 
7340 retry:
7341 	newcp = cp_confirmed = cp_unconfirmed = NULL;
7342 
7343 	/*
7344 	 * In search of an EXISTING client matching the incoming
7345 	 * request to establish a new client identifier at the server
7346 	 */
7347 	create = TRUE;
7348 	cp = rfs4_findclient(&args->client, &create, NULL);
7349 
7350 	/* Should never happen */
7351 	ASSERT(cp != NULL);
7352 
7353 	if (cp == NULL) {
7354 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
7355 		return;
7356 	}
7357 
7358 	/*
7359 	 * Easiest case. Client identifier is newly created and is
7360 	 * unconfirmed.  Also note that for this case, no other
7361 	 * entries exist for the client identifier.  Nothing else to
7362 	 * check.  Just setup the response and respond.
7363 	 */
7364 	if (create) {
7365 		*cs->statusp = res->status = NFS4_OK;
7366 		res->SETCLIENTID4res_u.resok4.clientid = cp->clientid;
7367 		res->SETCLIENTID4res_u.resok4.setclientid_confirm =
7368 		    cp->confirm_verf;
7369 		/* Setup callback information; CB_NULL confirmation later */
7370 		rfs4_client_setcb(cp, &args->callback, args->callback_ident);
7371 
7372 		rfs4_client_rele(cp);
7373 		return;
7374 	}
7375 
7376 	/*
7377 	 * An existing, confirmed client may exist but it may not have
7378 	 * been active for at least one lease period.  If so, then
7379 	 * "close" the client and create a new client identifier
7380 	 */
7381 	if (rfs4_lease_expired(cp)) {
7382 		rfs4_client_close(cp);
7383 		goto retry;
7384 	}
7385 
7386 	if (cp->need_confirm == TRUE)
7387 		cp_unconfirmed = cp;
7388 	else
7389 		cp_confirmed = cp;
7390 
7391 	cp = NULL;
7392 
7393 	/*
7394 	 * We have a confirmed client, now check for an
7395 	 * unconfimred entry
7396 	 */
7397 	if (cp_confirmed) {
7398 		/* If creds don't match then client identifier is inuse */
7399 		if (!creds_ok(cp_confirmed->cr_set, req, cs)) {
7400 			rfs4_cbinfo_t *cbp;
7401 			/*
7402 			 * Some one else has established this client
7403 			 * id. Try and say * who they are. We will use
7404 			 * the call back address supplied by * the
7405 			 * first client.
7406 			 */
7407 			*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
7408 
7409 			addr = netid = NULL;
7410 
7411 			cbp = &cp_confirmed->cbinfo;
7412 			if (cbp->cb_callback.cb_location.r_addr &&
7413 			    cbp->cb_callback.cb_location.r_netid) {
7414 				cb_client4 *cbcp = &cbp->cb_callback;
7415 
7416 				len = strlen(cbcp->cb_location.r_addr)+1;
7417 				addr = kmem_alloc(len, KM_SLEEP);
7418 				bcopy(cbcp->cb_location.r_addr, addr, len);
7419 				len = strlen(cbcp->cb_location.r_netid)+1;
7420 				netid = kmem_alloc(len, KM_SLEEP);
7421 				bcopy(cbcp->cb_location.r_netid, netid, len);
7422 			}
7423 
7424 			res->SETCLIENTID4res_u.client_using.r_addr = addr;
7425 			res->SETCLIENTID4res_u.client_using.r_netid = netid;
7426 
7427 			rfs4_client_rele(cp_confirmed);
7428 		}
7429 
7430 		/*
7431 		 * Confirmed, creds match, and verifier matches; must
7432 		 * be an update of the callback info
7433 		 */
7434 		if (cp_confirmed->nfs_client.verifier ==
7435 		    args->client.verifier) {
7436 			/* Setup callback information */
7437 			rfs4_client_setcb(cp_confirmed, &args->callback,
7438 			    args->callback_ident);
7439 
7440 			/* everything okay -- move ahead */
7441 			*cs->statusp = res->status = NFS4_OK;
7442 			res->SETCLIENTID4res_u.resok4.clientid =
7443 			    cp_confirmed->clientid;
7444 
7445 			/* update the confirm_verifier and return it */
7446 			rfs4_client_scv_next(cp_confirmed);
7447 			res->SETCLIENTID4res_u.resok4.setclientid_confirm =
7448 			    cp_confirmed->confirm_verf;
7449 
7450 			rfs4_client_rele(cp_confirmed);
7451 			return;
7452 		}
7453 
7454 		/*
7455 		 * Creds match but the verifier doesn't.  Must search
7456 		 * for an unconfirmed client that would be replaced by
7457 		 * this request.
7458 		 */
7459 		create = FALSE;
7460 		cp_unconfirmed = rfs4_findclient(&args->client, &create,
7461 		    cp_confirmed);
7462 	}
7463 
7464 	/*
7465 	 * At this point, we have taken care of the brand new client
7466 	 * struct, INUSE case, update of an existing, and confirmed
7467 	 * client struct.
7468 	 */
7469 
7470 	/*
7471 	 * check to see if things have changed while we originally
7472 	 * picked up the client struct.  If they have, then return and
7473 	 * retry the processing of this SETCLIENTID request.
7474 	 */
7475 	if (cp_unconfirmed) {
7476 		rfs4_dbe_lock(cp_unconfirmed->dbe);
7477 		if (!cp_unconfirmed->need_confirm) {
7478 			rfs4_dbe_unlock(cp_unconfirmed->dbe);
7479 			rfs4_client_rele(cp_unconfirmed);
7480 			if (cp_confirmed)
7481 				rfs4_client_rele(cp_confirmed);
7482 			goto retry;
7483 		}
7484 		/* do away with the old unconfirmed one */
7485 		rfs4_dbe_invalidate(cp_unconfirmed->dbe);
7486 		rfs4_dbe_unlock(cp_unconfirmed->dbe);
7487 		rfs4_client_rele(cp_unconfirmed);
7488 		cp_unconfirmed = NULL;
7489 	}
7490 
7491 	/*
7492 	 * This search will temporarily hide the confirmed client
7493 	 * struct while a new client struct is created as the
7494 	 * unconfirmed one.
7495 	 */
7496 	create = TRUE;
7497 	newcp = rfs4_findclient(&args->client, &create, cp_confirmed);
7498 
7499 	ASSERT(newcp != NULL);
7500 
7501 	if (newcp == NULL) {
7502 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
7503 		rfs4_client_rele(cp_confirmed);
7504 		return;
7505 	}
7506 
7507 	/*
7508 	 * If one was not created, then a similar request must be in
7509 	 * process so release and start over with this one
7510 	 */
7511 	if (create != TRUE) {
7512 		rfs4_client_rele(newcp);
7513 		if (cp_confirmed)
7514 			rfs4_client_rele(cp_confirmed);
7515 		goto retry;
7516 	}
7517 
7518 	*cs->statusp = res->status = NFS4_OK;
7519 	res->SETCLIENTID4res_u.resok4.clientid = newcp->clientid;
7520 	res->SETCLIENTID4res_u.resok4.setclientid_confirm = newcp->confirm_verf;
7521 	/* Setup callback information; CB_NULL confirmation later */
7522 	rfs4_client_setcb(newcp, &args->callback, args->callback_ident);
7523 
7524 	newcp->cp_confirmed = cp_confirmed;
7525 
7526 	rfs4_client_rele(newcp);
7527 }
7528 
7529 /*ARGSUSED*/
7530 void
7531 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
7532 			    struct svc_req *req, struct compound_state *cs)
7533 {
7534 	SETCLIENTID_CONFIRM4args *args =
7535 	    &argop->nfs_argop4_u.opsetclientid_confirm;
7536 	SETCLIENTID_CONFIRM4res *res =
7537 	    &resop->nfs_resop4_u.opsetclientid_confirm;
7538 	rfs4_client_t *cp, *cptoclose = NULL;
7539 
7540 	*cs->statusp = res->status = NFS4_OK;
7541 
7542 	cp = rfs4_findclient_by_id(args->clientid, TRUE);
7543 
7544 	if (cp == NULL) {
7545 		*cs->statusp = res->status =
7546 		    rfs4_check_clientid(&args->clientid, 1);
7547 		return;
7548 	}
7549 
7550 	if (!creds_ok(cp, req, cs)) {
7551 		*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
7552 		rfs4_client_rele(cp);
7553 		return;
7554 	}
7555 
7556 	/* If the verifier doesn't match, the record doesn't match */
7557 	if (cp->confirm_verf != args->setclientid_confirm) {
7558 		*cs->statusp = res->status = NFS4ERR_STALE_CLIENTID;
7559 		rfs4_client_rele(cp);
7560 		return;
7561 	}
7562 
7563 	rfs4_dbe_lock(cp->dbe);
7564 	cp->need_confirm = FALSE;
7565 	if (cp->cp_confirmed) {
7566 		cptoclose = cp->cp_confirmed;
7567 		cptoclose->ss_remove = 1;
7568 		cp->cp_confirmed = NULL;
7569 	}
7570 
7571 	/*
7572 	 * Update the client's associated server instance, if it's changed
7573 	 * since the client was created.
7574 	 */
7575 	if (rfs4_servinst(cp) != rfs4_cur_servinst)
7576 		rfs4_servinst_assign(cp, rfs4_cur_servinst);
7577 
7578 	/*
7579 	 * Record clientid in stable storage.
7580 	 * Must be done after server instance has been assigned.
7581 	 */
7582 	rfs4_ss_clid(cp, req);
7583 
7584 	rfs4_dbe_unlock(cp->dbe);
7585 
7586 	if (cptoclose)
7587 		/* don't need to rele, client_close does it */
7588 		rfs4_client_close(cptoclose);
7589 
7590 	/* If needed, initiate CB_NULL call for callback path */
7591 	rfs4_deleg_cb_check(cp);
7592 	rfs4_update_lease(cp);
7593 
7594 	/*
7595 	 * Check to see if client can perform reclaims
7596 	 */
7597 	rfs4_ss_chkclid(cp);
7598 
7599 	rfs4_client_rele(cp);
7600 }
7601 
7602 
7603 /*ARGSUSED*/
7604 void
7605 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop,
7606 	    struct svc_req *req, struct compound_state *cs)
7607 {
7608 	/* XXX Currently not using req arg */
7609 	CLOSE4args *args = &argop->nfs_argop4_u.opclose;
7610 	CLOSE4res *resp = &resop->nfs_resop4_u.opclose;
7611 	rfs4_state_t *sp;
7612 	nfsstat4 status;
7613 
7614 	if (cs->vp == NULL) {
7615 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7616 		return;
7617 	}
7618 
7619 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID);
7620 	if (status != NFS4_OK) {
7621 		*cs->statusp = resp->status = status;
7622 		return;
7623 	}
7624 
7625 	/* Ensure specified filehandle matches */
7626 	if (cs->vp != sp->finfo->vp) {
7627 		rfs4_state_rele(sp);
7628 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7629 		return;
7630 	}
7631 
7632 	/* hold off other access to open_owner while we tinker */
7633 	rfs4_sw_enter(&sp->owner->oo_sw);
7634 
7635 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7636 	case NFS4_CHECK_STATEID_OKAY:
7637 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7638 		    resop) != NFS4_CHKSEQ_OKAY) {
7639 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7640 			goto end;
7641 		}
7642 		break;
7643 	case NFS4_CHECK_STATEID_OLD:
7644 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7645 		goto end;
7646 	case NFS4_CHECK_STATEID_BAD:
7647 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7648 		goto end;
7649 	case NFS4_CHECK_STATEID_EXPIRED:
7650 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7651 		goto end;
7652 	case NFS4_CHECK_STATEID_CLOSED:
7653 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7654 		goto end;
7655 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7656 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7657 		goto end;
7658 	case NFS4_CHECK_STATEID_REPLAY:
7659 		/* Check the sequence id for the open owner */
7660 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
7661 		case NFS4_CHKSEQ_OKAY:
7662 			/*
7663 			 * This is replayed stateid; if seqid matches
7664 			 * next expected, then client is using wrong seqid.
7665 			 */
7666 			/* FALL THROUGH */
7667 		case NFS4_CHKSEQ_BAD:
7668 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7669 			goto end;
7670 		case NFS4_CHKSEQ_REPLAY:
7671 			/*
7672 			 * Note this case is the duplicate case so
7673 			 * resp->status is already set.
7674 			 */
7675 			*cs->statusp = resp->status;
7676 			rfs4_update_lease(sp->owner->client);
7677 			goto end;
7678 		}
7679 		break;
7680 	default:
7681 		ASSERT(FALSE);
7682 		break;
7683 	}
7684 
7685 	rfs4_dbe_lock(sp->dbe);
7686 
7687 	/* Update the stateid. */
7688 	next_stateid(&sp->stateid);
7689 	resp->open_stateid = sp->stateid.stateid;
7690 
7691 	rfs4_dbe_unlock(sp->dbe);
7692 
7693 	rfs4_update_lease(sp->owner->client);
7694 	rfs4_update_open_sequence(sp->owner);
7695 	rfs4_update_open_resp(sp->owner, resop, NULL);
7696 
7697 	rfs4_state_close(sp, FALSE, FALSE, cs->cr);
7698 
7699 	*cs->statusp = resp->status = status;
7700 
7701 end:
7702 	rfs4_sw_exit(&sp->owner->oo_sw);
7703 	rfs4_state_rele(sp);
7704 }
7705 
7706 /*
7707  * Manage the counts on the file struct and close all file locks
7708  */
7709 /*ARGSUSED*/
7710 void
7711 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr,
7712 	bool_t close_of_client)
7713 {
7714 	rfs4_file_t *fp = sp->finfo;
7715 	rfs4_lo_state_t *lsp;
7716 	struct shrlock shr;
7717 	struct shr_locowner shr_loco;
7718 	int fflags, s_access, s_deny;
7719 
7720 	fflags = s_access = s_deny = 0;
7721 	/*
7722 	 * Decrement the count for each access and deny bit that this
7723 	 * state has contributed to the file. If the file counts go to zero
7724 	 * clear the appropriate bit in the appropriate mask.
7725 	 */
7726 
7727 	if (sp->share_access & OPEN4_SHARE_ACCESS_READ) {
7728 		fp->access_read--;
7729 		fflags |= FREAD;
7730 		s_access |= F_RDACC;
7731 		if (fp->access_read == 0)
7732 			fp->share_access &= ~OPEN4_SHARE_ACCESS_READ;
7733 	}
7734 	if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) {
7735 		fp->access_write--;
7736 		fflags |= FWRITE;
7737 		s_access |= F_WRACC;
7738 		if (fp->access_write == 0)
7739 			fp->share_access &= ~OPEN4_SHARE_ACCESS_WRITE;
7740 	}
7741 	if (sp->share_deny & OPEN4_SHARE_DENY_READ) {
7742 		fp->deny_read--;
7743 		s_deny |= F_RDDNY;
7744 		if (fp->deny_read == 0)
7745 			fp->share_deny &= ~OPEN4_SHARE_DENY_READ;
7746 	}
7747 	if (sp->share_deny & OPEN4_SHARE_DENY_WRITE) {
7748 		fp->deny_write--;
7749 		s_deny |= F_WRDNY;
7750 		if (fp->deny_write == 0)
7751 			fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE;
7752 	}
7753 
7754 	/*
7755 	 * If this call is part of the larger closing down of client
7756 	 * state then it is just easier to release all locks
7757 	 * associated with this client instead of going through each
7758 	 * individual file and cleaning locks there.
7759 	 */
7760 	if (close_of_client) {
7761 		if (sp->owner->client->unlksys_completed == FALSE &&
7762 		    sp->lockownerlist.next->lsp != NULL &&
7763 		    sp->owner->client->sysidt != LM_NOSYSID) {
7764 			/* Is the PxFS kernel module loaded? */
7765 			if (lm_remove_file_locks != NULL) {
7766 				int new_sysid;
7767 
7768 				/* Encode the cluster nodeid in new sysid */
7769 				new_sysid = sp->owner->client->sysidt;
7770 				lm_set_nlmid_flk(&new_sysid);
7771 
7772 				/*
7773 				 * This PxFS routine removes file locks for a
7774 				 * client over all nodes of a cluster.
7775 				 */
7776 				NFS4_DEBUG(rfs4_debug, (CE_NOTE,
7777 				    "lm_remove_file_locks(sysid=0x%x)\n",
7778 				    new_sysid));
7779 				(*lm_remove_file_locks)(new_sysid);
7780 			} else {
7781 				struct flock64 flk;
7782 
7783 				/* Release all locks for this client */
7784 				flk.l_type = F_UNLKSYS;
7785 				flk.l_whence = 0;
7786 				flk.l_start = 0;
7787 				flk.l_len = 0;
7788 				flk.l_sysid = sp->owner->client->sysidt;
7789 				flk.l_pid = 0;
7790 				(void) VOP_FRLOCK(sp->finfo->vp, F_SETLK, &flk,
7791 				    F_REMOTELOCK | FREAD | FWRITE,
7792 				    (u_offset_t)0, NULL, CRED(), NULL);
7793 			}
7794 
7795 			sp->owner->client->unlksys_completed = TRUE;
7796 		}
7797 	}
7798 
7799 	/*
7800 	 * Release all locks on this file by this lock owner or at
7801 	 * least mark the locks as having been released
7802 	 */
7803 	for (lsp = sp->lockownerlist.next->lsp; lsp != NULL;
7804 	    lsp = lsp->lockownerlist.next->lsp) {
7805 
7806 		lsp->locks_cleaned = TRUE;
7807 
7808 		/* Was this already taken care of above? */
7809 		if (!close_of_client &&
7810 		    sp->owner->client->sysidt != LM_NOSYSID)
7811 			(void) cleanlocks(sp->finfo->vp, lsp->locker->pid,
7812 			    lsp->locker->client->sysidt);
7813 	}
7814 
7815 	/*
7816 	 * Release any shrlocks associated with this open state ID.
7817 	 * This must be done before the rfs4_state gets marked closed.
7818 	 */
7819 	if (sp->owner->client->sysidt != LM_NOSYSID) {
7820 		shr.s_access = s_access;
7821 		shr.s_deny = s_deny;
7822 		shr.s_pid = rfs4_dbe_getid(sp->owner->dbe);
7823 		shr.s_sysid = sp->owner->client->sysidt;
7824 		shr_loco.sl_pid = shr.s_pid;
7825 		shr_loco.sl_id = shr.s_sysid;
7826 		shr.s_owner = (caddr_t)&shr_loco;
7827 		shr.s_own_len = sizeof (shr_loco);
7828 		(void) vop_shrlock(sp->finfo->vp, F_UNSHARE, &shr, fflags);
7829 	}
7830 
7831 	(void) VOP_CLOSE(fp->vp, fflags, 1, (offset_t)0, cr, NULL);
7832 }
7833 
7834 /*
7835  * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure.
7836  */
7837 static nfsstat4
7838 lock_denied(LOCK4denied *dp, struct flock64 *flk)
7839 {
7840 	rfs4_lockowner_t *lo;
7841 	rfs4_client_t *cp;
7842 	uint32_t len;
7843 
7844 	lo = rfs4_findlockowner_by_pid(flk->l_pid);
7845 	if (lo != NULL) {
7846 		cp = lo->client;
7847 		if (rfs4_lease_expired(cp)) {
7848 			rfs4_lockowner_rele(lo);
7849 			rfs4_dbe_hold(cp->dbe);
7850 			rfs4_client_close(cp);
7851 			return (NFS4ERR_EXPIRED);
7852 		}
7853 		dp->owner.clientid = lo->owner.clientid;
7854 		len = lo->owner.owner_len;
7855 		dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
7856 		bcopy(lo->owner.owner_val, dp->owner.owner_val, len);
7857 		dp->owner.owner_len = len;
7858 		rfs4_lockowner_rele(lo);
7859 		goto finish;
7860 	}
7861 
7862 	/*
7863 	 * Its not a NFS4 lock. We take advantage that the upper 32 bits
7864 	 * of the client id contain the boot time for a NFS4 lock. So we
7865 	 * fabricate and identity by setting clientid to the sysid, and
7866 	 * the lock owner to the pid.
7867 	 */
7868 	dp->owner.clientid = flk->l_sysid;
7869 	len = sizeof (pid_t);
7870 	dp->owner.owner_len = len;
7871 	dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
7872 	bcopy(&flk->l_pid, dp->owner.owner_val, len);
7873 finish:
7874 	dp->offset = flk->l_start;
7875 	dp->length = flk->l_len;
7876 
7877 	if (flk->l_type == F_RDLCK)
7878 		dp->locktype = READ_LT;
7879 	else if (flk->l_type == F_WRLCK)
7880 		dp->locktype = WRITE_LT;
7881 	else
7882 		return (NFS4ERR_INVAL);	/* no mapping from POSIX ltype to v4 */
7883 
7884 	return (NFS4_OK);
7885 }
7886 
7887 static int
7888 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred)
7889 {
7890 	int error;
7891 	struct flock64 flk;
7892 	int i;
7893 	clock_t delaytime;
7894 	int cmd;
7895 
7896 	cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK;
7897 retry:
7898 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
7899 
7900 	for (i = 0; i < rfs4_maxlock_tries; i++) {
7901 		LOCK_PRINT(rfs4_debug, "setlock", cmd, flock);
7902 		error = VOP_FRLOCK(vp, cmd,
7903 		    flock, flag, (u_offset_t)0, NULL, cred, NULL);
7904 
7905 		if (error != EAGAIN && error != EACCES)
7906 			break;
7907 
7908 		if (i < rfs4_maxlock_tries - 1) {
7909 			delay(delaytime);
7910 			delaytime *= 2;
7911 		}
7912 	}
7913 
7914 	if (error == EAGAIN || error == EACCES) {
7915 		/* Get the owner of the lock */
7916 		flk = *flock;
7917 		LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk);
7918 		if (VOP_FRLOCK(vp, F_GETLK, &flk, flag,
7919 		    (u_offset_t)0, NULL, cred, NULL) == 0) {
7920 			if (flk.l_type == F_UNLCK) {
7921 				/* No longer locked, retry */
7922 				goto retry;
7923 			}
7924 			*flock = flk;
7925 			LOCK_PRINT(rfs4_debug, "setlock(blocking lock)",
7926 			    F_GETLK, &flk);
7927 		}
7928 	}
7929 
7930 	return (error);
7931 }
7932 
7933 /*ARGSUSED*/
7934 static nfsstat4
7935 rfs4_do_lock(rfs4_lo_state_t *lp, nfs_lock_type4 locktype,
7936 	    seqid4 seqid, offset4 offset,
7937 	    length4 length, cred_t *cred, nfs_resop4 *resop)
7938 {
7939 	nfsstat4 status;
7940 	rfs4_lockowner_t *lo = lp->locker;
7941 	rfs4_state_t *sp = lp->state;
7942 	struct flock64 flock;
7943 	int16_t ltype;
7944 	int flag;
7945 	int error;
7946 	sysid_t sysid;
7947 	LOCK4res *lres;
7948 
7949 	if (rfs4_lease_expired(lo->client)) {
7950 		return (NFS4ERR_EXPIRED);
7951 	}
7952 
7953 	if ((status = rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK)
7954 		return (status);
7955 
7956 	/* Check for zero length. To lock to end of file use all ones for V4 */
7957 	if (length == 0)
7958 		return (NFS4ERR_INVAL);
7959 	else if (length == (length4)(~0))
7960 		length = 0;		/* Posix to end of file  */
7961 
7962 retry:
7963 	rfs4_dbe_lock(sp->dbe);
7964 
7965 
7966 	if (resop->resop != OP_LOCKU) {
7967 		switch (locktype) {
7968 		case READ_LT:
7969 		case READW_LT:
7970 			if ((sp->share_access
7971 			    & OPEN4_SHARE_ACCESS_READ) == 0) {
7972 				rfs4_dbe_unlock(sp->dbe);
7973 
7974 				return (NFS4ERR_OPENMODE);
7975 			}
7976 			ltype = F_RDLCK;
7977 			break;
7978 		case WRITE_LT:
7979 		case WRITEW_LT:
7980 			if ((sp->share_access
7981 			    & OPEN4_SHARE_ACCESS_WRITE) == 0) {
7982 				rfs4_dbe_unlock(sp->dbe);
7983 
7984 				return (NFS4ERR_OPENMODE);
7985 			}
7986 			ltype = F_WRLCK;
7987 			break;
7988 		}
7989 	} else
7990 		ltype = F_UNLCK;
7991 
7992 	flock.l_type = ltype;
7993 	flock.l_whence = 0;		/* SEEK_SET */
7994 	flock.l_start = offset;
7995 	flock.l_len = length;
7996 	flock.l_sysid = sysid;
7997 	flock.l_pid = lp->locker->pid;
7998 
7999 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
8000 	if (flock.l_len < 0 || flock.l_start < 0) {
8001 		rfs4_dbe_unlock(sp->dbe);
8002 		return (NFS4ERR_INVAL);
8003 	}
8004 
8005 	/*
8006 	 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and
8007 	 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE.
8008 	 */
8009 	flag = (int)sp->share_access | F_REMOTELOCK;
8010 
8011 	error = setlock(sp->finfo->vp, &flock, flag, cred);
8012 	if (error == 0) {
8013 		rfs4_dbe_lock(lp->dbe);
8014 		next_stateid(&lp->lockid);
8015 		rfs4_dbe_unlock(lp->dbe);
8016 	}
8017 
8018 	rfs4_dbe_unlock(sp->dbe);
8019 
8020 	/*
8021 	 * N.B. We map error values to nfsv4 errors. This is differrent
8022 	 * than puterrno4 routine.
8023 	 */
8024 	switch (error) {
8025 	case 0:
8026 		status = NFS4_OK;
8027 		break;
8028 	case EAGAIN:
8029 	case EACCES:		/* Old value */
8030 		/* Can only get here if op is OP_LOCK */
8031 		ASSERT(resop->resop == OP_LOCK);
8032 		lres = &resop->nfs_resop4_u.oplock;
8033 		status = NFS4ERR_DENIED;
8034 		if (lock_denied(&lres->LOCK4res_u.denied, &flock)
8035 		    == NFS4ERR_EXPIRED)
8036 			goto retry;
8037 		break;
8038 	case ENOLCK:
8039 		status = NFS4ERR_DELAY;
8040 		break;
8041 	case EOVERFLOW:
8042 		status = NFS4ERR_INVAL;
8043 		break;
8044 	case EINVAL:
8045 		status = NFS4ERR_NOTSUPP;
8046 		break;
8047 	default:
8048 		cmn_err(CE_WARN, "rfs4_do_lock: unexpected errno (%d)", error);
8049 		status = NFS4ERR_SERVERFAULT;
8050 		break;
8051 	}
8052 
8053 	return (status);
8054 }
8055 
8056 /*ARGSUSED*/
8057 void
8058 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop,
8059 	    struct svc_req *req, struct compound_state *cs)
8060 {
8061 	/* XXX Currently not using req arg */
8062 	LOCK4args *args = &argop->nfs_argop4_u.oplock;
8063 	LOCK4res *resp = &resop->nfs_resop4_u.oplock;
8064 	nfsstat4 status;
8065 	stateid4 *stateid;
8066 	rfs4_lockowner_t *lo;
8067 	rfs4_client_t *cp;
8068 	rfs4_state_t *sp = NULL;
8069 	rfs4_lo_state_t *lsp = NULL;
8070 	bool_t ls_sw_held = FALSE;
8071 	bool_t create = TRUE;
8072 	bool_t lcreate = TRUE;
8073 	bool_t dup_lock = FALSE;
8074 	int rc;
8075 
8076 	if (cs->vp == NULL) {
8077 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8078 		return;
8079 	}
8080 
8081 	if (args->locker.new_lock_owner) {
8082 		/* Create a new lockowner for this instance */
8083 		open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner;
8084 
8085 		NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner"));
8086 
8087 		stateid = &olo->open_stateid;
8088 		status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID);
8089 		if (status != NFS4_OK) {
8090 			NFS4_DEBUG(rfs4_debug,
8091 			    (CE_NOTE, "Get state failed in lock %d", status));
8092 			*cs->statusp = resp->status = status;
8093 			return;
8094 		}
8095 
8096 		/* Ensure specified filehandle matches */
8097 		if (cs->vp != sp->finfo->vp) {
8098 			rfs4_state_rele(sp);
8099 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8100 			return;
8101 		}
8102 
8103 		/* hold off other access to open_owner while we tinker */
8104 		rfs4_sw_enter(&sp->owner->oo_sw);
8105 
8106 		switch (rc = rfs4_check_stateid_seqid(sp, stateid)) {
8107 		case NFS4_CHECK_STATEID_OLD:
8108 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8109 			goto end;
8110 		case NFS4_CHECK_STATEID_BAD:
8111 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8112 			goto end;
8113 		case NFS4_CHECK_STATEID_EXPIRED:
8114 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8115 			goto end;
8116 		case NFS4_CHECK_STATEID_UNCONFIRMED:
8117 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8118 			goto end;
8119 		case NFS4_CHECK_STATEID_CLOSED:
8120 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8121 			goto end;
8122 		case NFS4_CHECK_STATEID_OKAY:
8123 		case NFS4_CHECK_STATEID_REPLAY:
8124 			switch (rfs4_check_olo_seqid(olo->open_seqid,
8125 			    sp->owner, resop)) {
8126 			case NFS4_CHKSEQ_OKAY:
8127 				if (rc == NFS4_CHECK_STATEID_OKAY)
8128 					break;
8129 				/*
8130 				 * This is replayed stateid; if seqid
8131 				 * matches next expected, then client
8132 				 * is using wrong seqid.
8133 				 */
8134 				/* FALLTHROUGH */
8135 			case NFS4_CHKSEQ_BAD:
8136 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8137 				goto end;
8138 			case NFS4_CHKSEQ_REPLAY:
8139 				/* This is a duplicate LOCK request */
8140 				dup_lock = TRUE;
8141 
8142 				/*
8143 				 * For a duplicate we do not want to
8144 				 * create a new lockowner as it should
8145 				 * already exist.
8146 				 * Turn off the lockowner create flag.
8147 				 */
8148 				lcreate = FALSE;
8149 			}
8150 			break;
8151 		}
8152 
8153 		lo = rfs4_findlockowner(&olo->lock_owner, &lcreate);
8154 		if (lo == NULL) {
8155 			NFS4_DEBUG(rfs4_debug,
8156 			    (CE_NOTE, "rfs4_op_lock: no lock owner"));
8157 			*cs->statusp = resp->status = NFS4ERR_RESOURCE;
8158 			goto end;
8159 		}
8160 
8161 		lsp = rfs4_findlo_state_by_owner(lo, sp, &create);
8162 		if (lsp == NULL) {
8163 			rfs4_update_lease(sp->owner->client);
8164 			/*
8165 			 * Only update theh open_seqid if this is not
8166 			 * a duplicate request
8167 			 */
8168 			if (dup_lock == FALSE) {
8169 				rfs4_update_open_sequence(sp->owner);
8170 			}
8171 
8172 			NFS4_DEBUG(rfs4_debug,
8173 			    (CE_NOTE, "rfs4_op_lock: no state"));
8174 			*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
8175 			rfs4_update_open_resp(sp->owner, resop, NULL);
8176 			rfs4_lockowner_rele(lo);
8177 			goto end;
8178 		}
8179 
8180 		/*
8181 		 * This is the new_lock_owner branch and the client is
8182 		 * supposed to be associating a new lock_owner with
8183 		 * the open file at this point.  If we find that a
8184 		 * lock_owner/state association already exists and a
8185 		 * successful LOCK request was returned to the client,
8186 		 * an error is returned to the client since this is
8187 		 * not appropriate.  The client should be using the
8188 		 * existing lock_owner branch.
8189 		 */
8190 		if (dup_lock == FALSE && create == FALSE) {
8191 			if (lsp->lock_completed == TRUE) {
8192 				*cs->statusp =
8193 				    resp->status = NFS4ERR_BAD_SEQID;
8194 				rfs4_lockowner_rele(lo);
8195 				goto end;
8196 			}
8197 		}
8198 
8199 		rfs4_update_lease(sp->owner->client);
8200 
8201 		/*
8202 		 * Only update theh open_seqid if this is not
8203 		 * a duplicate request
8204 		 */
8205 		if (dup_lock == FALSE) {
8206 			rfs4_update_open_sequence(sp->owner);
8207 		}
8208 
8209 		/*
8210 		 * If this is a duplicate lock request, just copy the
8211 		 * previously saved reply and return.
8212 		 */
8213 		if (dup_lock == TRUE) {
8214 			/* verify that lock_seqid's match */
8215 			if (lsp->seqid != olo->lock_seqid) {
8216 				NFS4_DEBUG(rfs4_debug,
8217 				    (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad"
8218 				    "lsp->seqid=%d old->seqid=%d",
8219 				    lsp->seqid, olo->lock_seqid));
8220 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8221 			} else {
8222 				rfs4_copy_reply(resop, lsp->reply);
8223 				/*
8224 				 * Make sure to copy the just
8225 				 * retrieved reply status into the
8226 				 * overall compound status
8227 				 */
8228 				*cs->statusp = resp->status;
8229 			}
8230 			rfs4_lockowner_rele(lo);
8231 			goto end;
8232 		}
8233 
8234 		rfs4_dbe_lock(lsp->dbe);
8235 
8236 		/* Make sure to update the lock sequence id */
8237 		lsp->seqid = olo->lock_seqid;
8238 
8239 		NFS4_DEBUG(rfs4_debug,
8240 		    (CE_NOTE, "Lock seqid established as %d", lsp->seqid));
8241 
8242 		/*
8243 		 * This is used to signify the newly created lockowner
8244 		 * stateid and its sequence number.  The checks for
8245 		 * sequence number and increment don't occur on the
8246 		 * very first lock request for a lockowner.
8247 		 */
8248 		lsp->skip_seqid_check = TRUE;
8249 
8250 		/* hold off other access to lsp while we tinker */
8251 		rfs4_sw_enter(&lsp->ls_sw);
8252 		ls_sw_held = TRUE;
8253 
8254 		rfs4_dbe_unlock(lsp->dbe);
8255 
8256 		rfs4_lockowner_rele(lo);
8257 	} else {
8258 		stateid = &args->locker.locker4_u.lock_owner.lock_stateid;
8259 		/* get lsp and hold the lock on the underlying file struct */
8260 		if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE))
8261 		    != NFS4_OK) {
8262 			*cs->statusp = resp->status = status;
8263 			return;
8264 		}
8265 		create = FALSE;	/* We didn't create lsp */
8266 
8267 		/* Ensure specified filehandle matches */
8268 		if (cs->vp != lsp->state->finfo->vp) {
8269 			rfs4_lo_state_rele(lsp, TRUE);
8270 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8271 			return;
8272 		}
8273 
8274 		/* hold off other access to lsp while we tinker */
8275 		rfs4_sw_enter(&lsp->ls_sw);
8276 		ls_sw_held = TRUE;
8277 
8278 		switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
8279 		/*
8280 		 * The stateid looks like it was okay (expected to be
8281 		 * the next one)
8282 		 */
8283 		case NFS4_CHECK_STATEID_OKAY:
8284 			/*
8285 			 * The sequence id is now checked.  Determine
8286 			 * if this is a replay or if it is in the
8287 			 * expected (next) sequence.  In the case of a
8288 			 * replay, there are two replay conditions
8289 			 * that may occur.  The first is the normal
8290 			 * condition where a LOCK is done with a
8291 			 * NFS4_OK response and the stateid is
8292 			 * updated.  That case is handled below when
8293 			 * the stateid is identified as a REPLAY.  The
8294 			 * second is the case where an error is
8295 			 * returned, like NFS4ERR_DENIED, and the
8296 			 * sequence number is updated but the stateid
8297 			 * is not updated.  This second case is dealt
8298 			 * with here.  So it may seem odd that the
8299 			 * stateid is okay but the sequence id is a
8300 			 * replay but it is okay.
8301 			 */
8302 			switch (rfs4_check_lock_seqid(
8303 			    args->locker.locker4_u.lock_owner.lock_seqid,
8304 			    lsp, resop)) {
8305 			case NFS4_CHKSEQ_REPLAY:
8306 				if (resp->status != NFS4_OK) {
8307 					/*
8308 					 * Here is our replay and need
8309 					 * to verify that the last
8310 					 * response was an error.
8311 					 */
8312 					*cs->statusp = resp->status;
8313 					goto end;
8314 				}
8315 				/*
8316 				 * This is done since the sequence id
8317 				 * looked like a replay but it didn't
8318 				 * pass our check so a BAD_SEQID is
8319 				 * returned as a result.
8320 				 */
8321 				/*FALLTHROUGH*/
8322 			case NFS4_CHKSEQ_BAD:
8323 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8324 				goto end;
8325 			case NFS4_CHKSEQ_OKAY:
8326 				/* Everything looks okay move ahead */
8327 				break;
8328 			}
8329 			break;
8330 		case NFS4_CHECK_STATEID_OLD:
8331 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8332 			goto end;
8333 		case NFS4_CHECK_STATEID_BAD:
8334 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8335 			goto end;
8336 		case NFS4_CHECK_STATEID_EXPIRED:
8337 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8338 			goto end;
8339 		case NFS4_CHECK_STATEID_CLOSED:
8340 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8341 			goto end;
8342 		case NFS4_CHECK_STATEID_REPLAY:
8343 			switch (rfs4_check_lock_seqid(
8344 			    args->locker.locker4_u.lock_owner.lock_seqid,
8345 			    lsp, resop)) {
8346 			case NFS4_CHKSEQ_OKAY:
8347 				/*
8348 				 * This is a replayed stateid; if
8349 				 * seqid matches the next expected,
8350 				 * then client is using wrong seqid.
8351 				 */
8352 			case NFS4_CHKSEQ_BAD:
8353 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8354 				goto end;
8355 			case NFS4_CHKSEQ_REPLAY:
8356 				rfs4_update_lease(lsp->locker->client);
8357 				*cs->statusp = status = resp->status;
8358 				goto end;
8359 			}
8360 			break;
8361 		default:
8362 			ASSERT(FALSE);
8363 			break;
8364 		}
8365 
8366 		rfs4_update_lock_sequence(lsp);
8367 		rfs4_update_lease(lsp->locker->client);
8368 	}
8369 
8370 	/*
8371 	 * NFS4 only allows locking on regular files, so
8372 	 * verify type of object.
8373 	 */
8374 	if (cs->vp->v_type != VREG) {
8375 		if (cs->vp->v_type == VDIR)
8376 			status = NFS4ERR_ISDIR;
8377 		else
8378 			status = NFS4ERR_INVAL;
8379 		goto out;
8380 	}
8381 
8382 	cp = lsp->state->owner->client;
8383 
8384 	if (rfs4_clnt_in_grace(cp) && !args->reclaim) {
8385 		status = NFS4ERR_GRACE;
8386 		goto out;
8387 	}
8388 
8389 	if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->can_reclaim) {
8390 		status = NFS4ERR_NO_GRACE;
8391 		goto out;
8392 	}
8393 
8394 	if (!rfs4_clnt_in_grace(cp) && args->reclaim) {
8395 		status = NFS4ERR_NO_GRACE;
8396 		goto out;
8397 	}
8398 
8399 	if (lsp->state->finfo->dinfo->dtype == OPEN_DELEGATE_WRITE)
8400 		cs->deleg = TRUE;
8401 
8402 	status = rfs4_do_lock(lsp, args->locktype,
8403 	    args->locker.locker4_u.lock_owner.lock_seqid,
8404 	    args->offset, args->length, cs->cr, resop);
8405 
8406 out:
8407 	lsp->skip_seqid_check = FALSE;
8408 
8409 	*cs->statusp = resp->status = status;
8410 
8411 	if (status == NFS4_OK) {
8412 		resp->LOCK4res_u.lock_stateid = lsp->lockid.stateid;
8413 		lsp->lock_completed = TRUE;
8414 	}
8415 	/*
8416 	 * Only update the "OPEN" response here if this was a new
8417 	 * lock_owner
8418 	 */
8419 	if (sp)
8420 		rfs4_update_open_resp(sp->owner, resop, NULL);
8421 
8422 	rfs4_update_lock_resp(lsp, resop);
8423 
8424 end:
8425 	if (lsp) {
8426 		if (ls_sw_held)
8427 			rfs4_sw_exit(&lsp->ls_sw);
8428 		/*
8429 		 * If an sp obtained, then the lsp does not represent
8430 		 * a lock on the file struct.
8431 		 */
8432 		if (sp != NULL)
8433 			rfs4_lo_state_rele(lsp, FALSE);
8434 		else
8435 			rfs4_lo_state_rele(lsp, TRUE);
8436 	}
8437 	if (sp) {
8438 		rfs4_sw_exit(&sp->owner->oo_sw);
8439 		rfs4_state_rele(sp);
8440 	}
8441 }
8442 
8443 /* free function for LOCK/LOCKT */
8444 static void
8445 lock_denied_free(nfs_resop4 *resop)
8446 {
8447 	LOCK4denied *dp = NULL;
8448 
8449 	switch (resop->resop) {
8450 	case OP_LOCK:
8451 		if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED)
8452 			dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied;
8453 		break;
8454 	case OP_LOCKT:
8455 		if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED)
8456 			dp = &resop->nfs_resop4_u.oplockt.denied;
8457 		break;
8458 	default:
8459 		break;
8460 	}
8461 
8462 	if (dp)
8463 		kmem_free(dp->owner.owner_val, dp->owner.owner_len);
8464 }
8465 
8466 /*ARGSUSED*/
8467 void
8468 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop,
8469 	    struct svc_req *req, struct compound_state *cs)
8470 {
8471 	/* XXX Currently not using req arg */
8472 	LOCKU4args *args = &argop->nfs_argop4_u.oplocku;
8473 	LOCKU4res *resp = &resop->nfs_resop4_u.oplocku;
8474 	nfsstat4 status;
8475 	stateid4 *stateid = &args->lock_stateid;
8476 	rfs4_lo_state_t *lsp;
8477 
8478 	if (cs->vp == NULL) {
8479 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8480 		return;
8481 	}
8482 
8483 	if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) {
8484 		*cs->statusp = resp->status = status;
8485 		return;
8486 	}
8487 
8488 	/* Ensure specified filehandle matches */
8489 	if (cs->vp != lsp->state->finfo->vp) {
8490 		rfs4_lo_state_rele(lsp, TRUE);
8491 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8492 		return;
8493 	}
8494 
8495 	/* hold off other access to lsp while we tinker */
8496 	rfs4_sw_enter(&lsp->ls_sw);
8497 
8498 	switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
8499 	case NFS4_CHECK_STATEID_OKAY:
8500 		if (rfs4_check_lock_seqid(args->seqid, lsp, resop)
8501 		    != NFS4_CHKSEQ_OKAY) {
8502 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8503 			goto end;
8504 		}
8505 		break;
8506 	case NFS4_CHECK_STATEID_OLD:
8507 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8508 		goto end;
8509 	case NFS4_CHECK_STATEID_BAD:
8510 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8511 		goto end;
8512 	case NFS4_CHECK_STATEID_EXPIRED:
8513 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8514 		goto end;
8515 	case NFS4_CHECK_STATEID_CLOSED:
8516 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8517 		goto end;
8518 	case NFS4_CHECK_STATEID_REPLAY:
8519 		switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) {
8520 		case NFS4_CHKSEQ_OKAY:
8521 				/*
8522 				 * This is a replayed stateid; if
8523 				 * seqid matches the next expected,
8524 				 * then client is using wrong seqid.
8525 				 */
8526 		case NFS4_CHKSEQ_BAD:
8527 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8528 			goto end;
8529 		case NFS4_CHKSEQ_REPLAY:
8530 			rfs4_update_lease(lsp->locker->client);
8531 			*cs->statusp = status = resp->status;
8532 			goto end;
8533 		}
8534 		break;
8535 	default:
8536 		ASSERT(FALSE);
8537 		break;
8538 	}
8539 
8540 	rfs4_update_lock_sequence(lsp);
8541 	rfs4_update_lease(lsp->locker->client);
8542 
8543 	/*
8544 	 * NFS4 only allows locking on regular files, so
8545 	 * verify type of object.
8546 	 */
8547 	if (cs->vp->v_type != VREG) {
8548 		if (cs->vp->v_type == VDIR)
8549 			status = NFS4ERR_ISDIR;
8550 		else
8551 			status = NFS4ERR_INVAL;
8552 		goto out;
8553 	}
8554 
8555 	if (rfs4_clnt_in_grace(lsp->state->owner->client)) {
8556 		status = NFS4ERR_GRACE;
8557 		goto out;
8558 	}
8559 
8560 	status = rfs4_do_lock(lsp, args->locktype,
8561 	    args->seqid, args->offset, args->length, cs->cr, resop);
8562 
8563 out:
8564 	*cs->statusp = resp->status = status;
8565 
8566 	if (status == NFS4_OK)
8567 		resp->lock_stateid = lsp->lockid.stateid;
8568 
8569 	rfs4_update_lock_resp(lsp, resop);
8570 
8571 end:
8572 	rfs4_sw_exit(&lsp->ls_sw);
8573 	rfs4_lo_state_rele(lsp, TRUE);
8574 }
8575 
8576 /*
8577  * LOCKT is a best effort routine, the client can not be guaranteed that
8578  * the status return is still in effect by the time the reply is received.
8579  * They are numerous race conditions in this routine, but we are not required
8580  * and can not be accurate.
8581  */
8582 /*ARGSUSED*/
8583 void
8584 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop,
8585 	    struct svc_req *req, struct compound_state *cs)
8586 {
8587 	LOCKT4args *args = &argop->nfs_argop4_u.oplockt;
8588 	LOCKT4res *resp = &resop->nfs_resop4_u.oplockt;
8589 	rfs4_lockowner_t *lo;
8590 	rfs4_client_t *cp;
8591 	bool_t create = FALSE;
8592 	struct flock64 flk;
8593 	int error;
8594 	int flag = FREAD | FWRITE;
8595 	int ltype;
8596 	length4 posix_length;
8597 	sysid_t sysid;
8598 	pid_t pid;
8599 
8600 	if (cs->vp == NULL) {
8601 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8602 		return;
8603 	}
8604 
8605 	/*
8606 	 * NFS4 only allows locking on regular files, so
8607 	 * verify type of object.
8608 	 */
8609 	if (cs->vp->v_type != VREG) {
8610 		if (cs->vp->v_type == VDIR)
8611 			*cs->statusp = resp->status = NFS4ERR_ISDIR;
8612 		else
8613 			*cs->statusp = resp->status =  NFS4ERR_INVAL;
8614 		return;
8615 	}
8616 
8617 	/*
8618 	 * Check out the clientid to ensure the server knows about it
8619 	 * so that we correctly inform the client of a server reboot.
8620 	 */
8621 	if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE))
8622 	    == NULL) {
8623 		*cs->statusp = resp->status =
8624 		    rfs4_check_clientid(&args->owner.clientid, 0);
8625 		return;
8626 	}
8627 	if (rfs4_lease_expired(cp)) {
8628 		rfs4_client_close(cp);
8629 		/*
8630 		 * Protocol doesn't allow returning NFS4ERR_STALE as
8631 		 * other operations do on this check so STALE_CLIENTID
8632 		 * is returned instead
8633 		 */
8634 		*cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
8635 		return;
8636 	}
8637 
8638 	if (rfs4_clnt_in_grace(cp)) {
8639 		*cs->statusp = resp->status = NFS4ERR_GRACE;
8640 		return;
8641 	}
8642 	rfs4_client_rele(cp);
8643 
8644 	resp->status = NFS4_OK;
8645 
8646 	switch (args->locktype) {
8647 	case READ_LT:
8648 	case READW_LT:
8649 		ltype = F_RDLCK;
8650 		break;
8651 	case WRITE_LT:
8652 	case WRITEW_LT:
8653 		ltype = F_WRLCK;
8654 		break;
8655 	}
8656 
8657 	posix_length = args->length;
8658 	/* Check for zero length. To lock to end of file use all ones for V4 */
8659 	if (posix_length == 0) {
8660 		*cs->statusp = resp->status = NFS4ERR_INVAL;
8661 		return;
8662 	} else if (posix_length == (length4)(~0)) {
8663 		posix_length = 0;	/* Posix to end of file  */
8664 	}
8665 
8666 	/* Find or create a lockowner */
8667 	lo = rfs4_findlockowner(&args->owner, &create);
8668 
8669 	if (lo) {
8670 		pid = lo->pid;
8671 		if ((resp->status =
8672 		    rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK)
8673 		goto out;
8674 	} else {
8675 		pid = 0;
8676 		sysid = lockt_sysid;
8677 	}
8678 retry:
8679 	flk.l_type = ltype;
8680 	flk.l_whence = 0;		/* SEEK_SET */
8681 	flk.l_start = args->offset;
8682 	flk.l_len = posix_length;
8683 	flk.l_sysid = sysid;
8684 	flk.l_pid = pid;
8685 	flag |= F_REMOTELOCK;
8686 
8687 	LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk);
8688 
8689 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
8690 	if (flk.l_len < 0 || flk.l_start < 0) {
8691 		resp->status = NFS4ERR_INVAL;
8692 		goto out;
8693 	}
8694 	error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0,
8695 	    NULL, cs->cr, NULL);
8696 
8697 	/*
8698 	 * N.B. We map error values to nfsv4 errors. This is differrent
8699 	 * than puterrno4 routine.
8700 	 */
8701 	switch (error) {
8702 	case 0:
8703 		if (flk.l_type == F_UNLCK)
8704 			resp->status = NFS4_OK;
8705 		else {
8706 			if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED)
8707 				goto retry;
8708 			resp->status = NFS4ERR_DENIED;
8709 		}
8710 		break;
8711 	case EOVERFLOW:
8712 		resp->status = NFS4ERR_INVAL;
8713 		break;
8714 	case EINVAL:
8715 		resp->status = NFS4ERR_NOTSUPP;
8716 		break;
8717 	default:
8718 		cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)",
8719 		    error);
8720 		resp->status = NFS4ERR_SERVERFAULT;
8721 		break;
8722 	}
8723 
8724 out:
8725 	if (lo)
8726 		rfs4_lockowner_rele(lo);
8727 	*cs->statusp = resp->status;
8728 }
8729 
8730 static int
8731 vop_shrlock(vnode_t *vp, int cmd, struct shrlock *sp, int fflags)
8732 {
8733 	int err;
8734 
8735 	if (cmd == F_UNSHARE && sp->s_deny == 0 && sp->s_access == 0)
8736 		return (0);
8737 
8738 	err = VOP_SHRLOCK(vp, cmd, sp, fflags, CRED(), NULL);
8739 
8740 	NFS4_DEBUG(rfs4_shrlock_debug,
8741 	    (CE_NOTE, "rfs4_shrlock %s vp=%p acc=%d dny=%d sysid=%d "
8742 	    "pid=%d err=%d\n", cmd == F_UNSHARE ? "UNSHR" : "SHARE",
8743 	    (void *) vp, sp->s_access, sp->s_deny, sp->s_sysid, sp->s_pid,
8744 	    err));
8745 
8746 	return (err);
8747 }
8748 
8749 static int
8750 rfs4_shrlock(rfs4_state_t *sp, int cmd)
8751 {
8752 	struct shrlock shr;
8753 	struct shr_locowner shr_loco;
8754 	int fflags;
8755 
8756 	fflags = shr.s_access = shr.s_deny = 0;
8757 
8758 	if (sp->share_access & OPEN4_SHARE_ACCESS_READ) {
8759 		fflags |= FREAD;
8760 		shr.s_access |= F_RDACC;
8761 	}
8762 	if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) {
8763 		fflags |= FWRITE;
8764 		shr.s_access |= F_WRACC;
8765 	}
8766 	if (sp->share_deny & OPEN4_SHARE_DENY_READ)
8767 		shr.s_deny |= F_RDDNY;
8768 	if (sp->share_deny & OPEN4_SHARE_DENY_WRITE)
8769 		shr.s_deny |= F_WRDNY;
8770 
8771 	shr.s_pid = rfs4_dbe_getid(sp->owner->dbe);
8772 	shr.s_sysid = sp->owner->client->sysidt;
8773 	shr_loco.sl_pid = shr.s_pid;
8774 	shr_loco.sl_id = shr.s_sysid;
8775 	shr.s_owner = (caddr_t)&shr_loco;
8776 	shr.s_own_len = sizeof (shr_loco);
8777 	return (vop_shrlock(sp->finfo->vp, cmd, &shr, fflags));
8778 }
8779 
8780 static int
8781 rfs4_share(rfs4_state_t *sp)
8782 {
8783 	int cmd;
8784 	cmd = nbl_need_check(sp->finfo->vp) ? F_SHARE_NBMAND : F_SHARE;
8785 	return (rfs4_shrlock(sp, cmd));
8786 }
8787 
8788 void
8789 rfs4_unshare(rfs4_state_t *sp)
8790 {
8791 	(void) rfs4_shrlock(sp, F_UNSHARE);
8792 }
8793