xref: /illumos-gate/usr/src/uts/common/fs/nfs/nfs4_srv.c (revision 7dacfc4494f6d14358974ef2830b5cd8c66a84de)
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 	if (error)
927 		return (puterrno4(error));
928 
929 	/*
930 	 * If the vnode is in a pseudo filesystem, or if the security flavor
931 	 * used in the request is valid but not an explicitly shared flavor,
932 	 * or the access bit indicates that this is a limited access,
933 	 * check whether this vnode is visible.
934 	 */
935 	if (!different_export &&
936 	    (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) ||
937 	    cs->access & CS_ACCESS_LIMITED)) {
938 		if (! nfs_visible(cs->exi, vp, &different_export)) {
939 			VN_RELE(vp);
940 			return (puterrno4(ENOENT));
941 		}
942 	}
943 
944 	/*
945 	 * If it's a mountpoint, then traverse it.
946 	 */
947 	if (vn_ismntpt(vp)) {
948 		tvp = vp;
949 		if ((error = traverse(&tvp)) != 0) {
950 			VN_RELE(vp);
951 			return (puterrno4(error));
952 		}
953 		/* remember that we had to traverse mountpoint */
954 		did_traverse = TRUE;
955 		vp = tvp;
956 		different_export = 1;
957 	} else if (vp->v_vfsp != dvp->v_vfsp) {
958 		/*
959 		 * If vp isn't a mountpoint and the vfs ptrs aren't the same,
960 		 * then vp is probably an LOFS object.  We don't need the
961 		 * realvp, we just need to know that we might have crossed
962 		 * a server fs boundary and need to call checkexport4.
963 		 * (LOFS lookup hides server fs mountpoints, and actually calls
964 		 * traverse)
965 		 */
966 		different_export = 1;
967 		did_traverse = FALSE;
968 	}
969 
970 	/*
971 	 * Get the export information for it.
972 	 */
973 	if (different_export) {
974 
975 		bzero(&fid, sizeof (fid));
976 		fid.fid_len = MAXFIDSZ;
977 		error = vop_fid_pseudo(vp, &fid);
978 		if (error) {
979 			VN_RELE(vp);
980 			return (puterrno4(error));
981 		}
982 
983 		if (dotdot)
984 			exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE);
985 		else
986 			exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
987 
988 		if (exi == NULL) {
989 			if (did_traverse == TRUE) {
990 				/*
991 				 * If this vnode is a mounted-on vnode,
992 				 * but the mounted-on file system is not
993 				 * exported, send back the secinfo for
994 				 * the exported node that the mounted-on
995 				 * vnode lives in.
996 				 */
997 				exi = cs->exi;
998 			} else {
999 				VN_RELE(vp);
1000 				return (puterrno4(EACCES));
1001 			}
1002 		}
1003 	} else {
1004 		exi = cs->exi;
1005 	}
1006 	ASSERT(exi != NULL);
1007 
1008 
1009 	/*
1010 	 * Create the secinfo result based on the security information
1011 	 * from the exportinfo structure (exi).
1012 	 *
1013 	 * Return all flavors for a pseudo node.
1014 	 * For a real export node, return the flavor that the client
1015 	 * has access with.
1016 	 */
1017 	ASSERT(RW_LOCK_HELD(&exported_lock));
1018 	if (PSEUDO(exi)) {
1019 		count = exi->exi_export.ex_seccnt; /* total sec count */
1020 		resok_val = kmem_alloc(count * sizeof (secinfo4), KM_SLEEP);
1021 		secp = exi->exi_export.ex_secinfo;
1022 
1023 		for (i = 0; i < count; i++) {
1024 			si = &secp[i].s_secinfo;
1025 			resok_val[i].flavor = si->sc_rpcnum;
1026 			if (resok_val[i].flavor == RPCSEC_GSS) {
1027 				rpcsec_gss_info *info;
1028 
1029 				info = &resok_val[i].flavor_info;
1030 				info->qop = si->sc_qop;
1031 				info->service = (rpc_gss_svc_t)si->sc_service;
1032 
1033 				/* get oid opaque data */
1034 				info->oid.sec_oid4_len =
1035 				    si->sc_gss_mech_type->length;
1036 				info->oid.sec_oid4_val = kmem_alloc(
1037 				    si->sc_gss_mech_type->length, KM_SLEEP);
1038 				bcopy(
1039 				    si->sc_gss_mech_type->elements,
1040 				    info->oid.sec_oid4_val,
1041 				    info->oid.sec_oid4_len);
1042 			}
1043 		}
1044 		resp->SECINFO4resok_len = count;
1045 		resp->SECINFO4resok_val = resok_val;
1046 	} else {
1047 		int ret_cnt = 0, k = 0;
1048 		int *flavor_list;
1049 
1050 		count = exi->exi_export.ex_seccnt; /* total sec count */
1051 		secp = exi->exi_export.ex_secinfo;
1052 
1053 		flavor_list = kmem_alloc(count * sizeof (int), KM_SLEEP);
1054 		/* find out which flavors to return */
1055 		for (i = 0; i < count; i ++) {
1056 			int access, flavor, perm;
1057 
1058 			flavor = secp[i].s_secinfo.sc_nfsnum;
1059 			perm = secp[i].s_flags;
1060 
1061 			access = nfsauth4_secinfo_access(exi, cs->req,
1062 			    flavor, perm);
1063 
1064 			if (! (access & NFSAUTH_DENIED) &&
1065 			    ! (access & NFSAUTH_WRONGSEC)) {
1066 				flavor_list[ret_cnt] = flavor;
1067 				ret_cnt++;
1068 			}
1069 		}
1070 
1071 		/* Create the returning SECINFO value */
1072 		resok_val = kmem_alloc(ret_cnt * sizeof (secinfo4), KM_SLEEP);
1073 
1074 		for (i = 0; i < count; i++) {
1075 		/* If the flavor is in the flavor list, fill in resok_val. */
1076 			si = &secp[i].s_secinfo;
1077 			if (in_flavor_list(si->sc_nfsnum,
1078 			    flavor_list, ret_cnt)) {
1079 				resok_val[k].flavor = si->sc_rpcnum;
1080 				if (resok_val[k].flavor == RPCSEC_GSS) {
1081 					rpcsec_gss_info *info;
1082 
1083 					info = &resok_val[k].flavor_info;
1084 					info->qop = si->sc_qop;
1085 					info->service = (rpc_gss_svc_t)
1086 					    si->sc_service;
1087 
1088 					/* get oid opaque data */
1089 					info->oid.sec_oid4_len =
1090 					    si->sc_gss_mech_type->length;
1091 					info->oid.sec_oid4_val = kmem_alloc(
1092 					    si->sc_gss_mech_type->length,
1093 					    KM_SLEEP);
1094 					bcopy(si->sc_gss_mech_type->elements,
1095 					    info->oid.sec_oid4_val,
1096 					    info->oid.sec_oid4_len);
1097 				}
1098 				k++;
1099 			}
1100 		if (k >= ret_cnt)
1101 			break;
1102 		}
1103 		resp->SECINFO4resok_len = ret_cnt;
1104 		resp->SECINFO4resok_val = resok_val;
1105 		kmem_free(flavor_list, count * sizeof (int));
1106 	}
1107 
1108 	VN_RELE(vp);
1109 	return (NFS4_OK);
1110 }
1111 
1112 /*
1113  * SECINFO (Operation 33): Obtain required security information on
1114  * the component name in the format of (security-mechanism-oid, qop, service)
1115  * triplets.
1116  */
1117 /* ARGSUSED */
1118 static void
1119 rfs4_op_secinfo(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1120 	struct compound_state *cs)
1121 {
1122 	SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo;
1123 	utf8string *utfnm = &argop->nfs_argop4_u.opsecinfo.name;
1124 	uint_t len;
1125 	char *nm;
1126 
1127 	/*
1128 	 * Current file handle (cfh) should have been set before getting
1129 	 * into this function. If not, return error.
1130 	 */
1131 	if (cs->vp == NULL) {
1132 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1133 		return;
1134 	}
1135 
1136 	if (cs->vp->v_type != VDIR) {
1137 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
1138 		return;
1139 	}
1140 
1141 	/*
1142 	 * Verify the component name. If failed, error out, but
1143 	 * do not error out if the component name is a "..".
1144 	 * SECINFO will return its parents secinfo data for SECINFO "..".
1145 	 */
1146 	if (!utf8_dir_verify(utfnm)) {
1147 		if (utfnm->utf8string_len != 2 ||
1148 		    utfnm->utf8string_val[0] != '.' ||
1149 		    utfnm->utf8string_val[1] != '.') {
1150 			*cs->statusp = resp->status = NFS4ERR_INVAL;
1151 			return;
1152 		}
1153 	}
1154 
1155 	nm = utf8_to_str(utfnm, &len, NULL);
1156 	if (nm == NULL) {
1157 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1158 		return;
1159 	}
1160 
1161 	if (len > MAXNAMELEN) {
1162 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1163 		kmem_free(nm, len);
1164 		return;
1165 	}
1166 
1167 	*cs->statusp = resp->status = do_rfs4_op_secinfo(cs, nm, resp);
1168 
1169 	kmem_free(nm, len);
1170 }
1171 
1172 /*
1173  * Free SECINFO result.
1174  */
1175 /* ARGSUSED */
1176 static void
1177 rfs4_op_secinfo_free(nfs_resop4 *resop)
1178 {
1179 	SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo;
1180 	int count, i;
1181 	secinfo4 *resok_val;
1182 
1183 	/* If this is not an Ok result, nothing to free. */
1184 	if (resp->status != NFS4_OK) {
1185 		return;
1186 	}
1187 
1188 	count = resp->SECINFO4resok_len;
1189 	resok_val = resp->SECINFO4resok_val;
1190 
1191 	for (i = 0; i < count; i++) {
1192 		if (resok_val[i].flavor == RPCSEC_GSS) {
1193 			rpcsec_gss_info *info;
1194 
1195 			info = &resok_val[i].flavor_info;
1196 			kmem_free(info->oid.sec_oid4_val,
1197 			    info->oid.sec_oid4_len);
1198 		}
1199 	}
1200 	kmem_free(resok_val, count * sizeof (secinfo4));
1201 	resp->SECINFO4resok_len = 0;
1202 	resp->SECINFO4resok_val = NULL;
1203 }
1204 
1205 /* ARGSUSED */
1206 static void
1207 rfs4_op_access(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1208 	struct compound_state *cs)
1209 {
1210 	ACCESS4args *args = &argop->nfs_argop4_u.opaccess;
1211 	ACCESS4res *resp = &resop->nfs_resop4_u.opaccess;
1212 	int error;
1213 	vnode_t *vp;
1214 	struct vattr va;
1215 	int checkwriteperm;
1216 	cred_t *cr = cs->cr;
1217 	bslabel_t *clabel, *slabel;
1218 	ts_label_t *tslabel;
1219 	boolean_t admin_low_client;
1220 
1221 #if 0	/* XXX allow access even if !cs->access. Eventually only pseudo fs */
1222 	if (cs->access == CS_ACCESS_DENIED) {
1223 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1224 		return;
1225 	}
1226 #endif
1227 	if (cs->vp == NULL) {
1228 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1229 		return;
1230 	}
1231 
1232 	ASSERT(cr != NULL);
1233 
1234 	vp = cs->vp;
1235 
1236 	/*
1237 	 * If the file system is exported read only, it is not appropriate
1238 	 * to check write permissions for regular files and directories.
1239 	 * Special files are interpreted by the client, so the underlying
1240 	 * permissions are sent back to the client for interpretation.
1241 	 */
1242 	if (rdonly4(cs->exi, cs->vp, req) &&
1243 	    (vp->v_type == VREG || vp->v_type == VDIR))
1244 		checkwriteperm = 0;
1245 	else
1246 		checkwriteperm = 1;
1247 
1248 	/*
1249 	 * XXX
1250 	 * We need the mode so that we can correctly determine access
1251 	 * permissions relative to a mandatory lock file.  Access to
1252 	 * mandatory lock files is denied on the server, so it might
1253 	 * as well be reflected to the server during the open.
1254 	 */
1255 	va.va_mask = AT_MODE;
1256 	error = VOP_GETATTR(vp, &va, 0, cr);
1257 	if (error) {
1258 		*cs->statusp = resp->status = puterrno4(error);
1259 		return;
1260 	}
1261 	resp->access = 0;
1262 	resp->supported = 0;
1263 
1264 	if (is_system_labeled()) {
1265 		ASSERT(req->rq_label != NULL);
1266 		clabel = req->rq_label;
1267 		DTRACE_PROBE2(tx__rfs4__log__info__opaccess__clabel, char *,
1268 		    "got client label from request(1)",
1269 		    struct svc_req *, req);
1270 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
1271 			if ((tslabel = nfs_getflabel(vp)) == NULL) {
1272 				*cs->statusp = resp->status = puterrno4(EACCES);
1273 				return;
1274 			}
1275 			slabel = label2bslabel(tslabel);
1276 			DTRACE_PROBE3(tx__rfs4__log__info__opaccess__slabel,
1277 			    char *, "got server label(1) for vp(2)",
1278 			    bslabel_t *, slabel, vnode_t *, vp);
1279 
1280 			admin_low_client = B_FALSE;
1281 		} else
1282 			admin_low_client = B_TRUE;
1283 	}
1284 
1285 	if (args->access & ACCESS4_READ) {
1286 		error = VOP_ACCESS(vp, VREAD, 0, cr);
1287 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1288 		    (!is_system_labeled() || admin_low_client ||
1289 		    bldominates(clabel, slabel)))
1290 			resp->access |= ACCESS4_READ;
1291 		resp->supported |= ACCESS4_READ;
1292 	}
1293 	if ((args->access & ACCESS4_LOOKUP) && vp->v_type == VDIR) {
1294 		error = VOP_ACCESS(vp, VEXEC, 0, cr);
1295 		if (!error && (!is_system_labeled() || admin_low_client ||
1296 		    bldominates(clabel, slabel)))
1297 			resp->access |= ACCESS4_LOOKUP;
1298 		resp->supported |= ACCESS4_LOOKUP;
1299 	}
1300 	if (checkwriteperm &&
1301 	    (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND))) {
1302 		error = VOP_ACCESS(vp, VWRITE, 0, cr);
1303 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1304 		    (!is_system_labeled() || admin_low_client ||
1305 		    blequal(clabel, slabel)))
1306 			resp->access |=
1307 			    (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND));
1308 		resp->supported |= (ACCESS4_MODIFY|ACCESS4_EXTEND);
1309 	}
1310 
1311 	if (checkwriteperm &&
1312 	    (args->access & ACCESS4_DELETE) && vp->v_type == VDIR) {
1313 		error = VOP_ACCESS(vp, VWRITE, 0, cr);
1314 		if (!error && (!is_system_labeled() || admin_low_client ||
1315 		    blequal(clabel, slabel)))
1316 			resp->access |= ACCESS4_DELETE;
1317 		resp->supported |= ACCESS4_DELETE;
1318 	}
1319 	if (args->access & ACCESS4_EXECUTE && vp->v_type != VDIR) {
1320 		error = VOP_ACCESS(vp, VEXEC, 0, cr);
1321 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1322 		    (!is_system_labeled() || admin_low_client ||
1323 		    bldominates(clabel, slabel)))
1324 			resp->access |= ACCESS4_EXECUTE;
1325 		resp->supported |= ACCESS4_EXECUTE;
1326 	}
1327 
1328 	if (is_system_labeled() && !admin_low_client)
1329 		label_rele(tslabel);
1330 
1331 	*cs->statusp = resp->status = NFS4_OK;
1332 }
1333 
1334 /* ARGSUSED */
1335 static void
1336 rfs4_op_commit(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1337 	struct compound_state *cs)
1338 {
1339 	COMMIT4args *args = &argop->nfs_argop4_u.opcommit;
1340 	COMMIT4res *resp = &resop->nfs_resop4_u.opcommit;
1341 	int error;
1342 	vnode_t *vp = cs->vp;
1343 	cred_t *cr = cs->cr;
1344 	vattr_t va;
1345 
1346 	if (vp == NULL) {
1347 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1348 		return;
1349 	}
1350 	if (cs->access == CS_ACCESS_DENIED) {
1351 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1352 		return;
1353 	}
1354 
1355 	if (args->offset + args->count < args->offset) {
1356 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1357 		return;
1358 	}
1359 
1360 	va.va_mask = AT_UID;
1361 	error = VOP_GETATTR(vp, &va, 0, cr);
1362 
1363 	/*
1364 	 * If we can't get the attributes, then we can't do the
1365 	 * right access checking.  So, we'll fail the request.
1366 	 */
1367 	if (error) {
1368 		*cs->statusp = resp->status = puterrno4(error);
1369 		return;
1370 	}
1371 	if (rdonly4(cs->exi, cs->vp, req)) {
1372 		*cs->statusp = resp->status = NFS4ERR_ROFS;
1373 		return;
1374 	}
1375 
1376 	if (vp->v_type != VREG) {
1377 		if (vp->v_type == VDIR)
1378 			resp->status = NFS4ERR_ISDIR;
1379 		else
1380 			resp->status = NFS4ERR_INVAL;
1381 		*cs->statusp = resp->status;
1382 		return;
1383 	}
1384 
1385 	if (crgetuid(cr) != va.va_uid &&
1386 	    (error = VOP_ACCESS(vp, VWRITE, 0, cs->cr))) {
1387 		*cs->statusp = resp->status = puterrno4(error);
1388 		return;
1389 	}
1390 
1391 	error = VOP_PUTPAGE(vp, args->offset, args->count, 0, cr);
1392 	if (!error)
1393 		error = VOP_FSYNC(vp, FNODSYNC, cr);
1394 
1395 	if (error) {
1396 		*cs->statusp = resp->status = puterrno4(error);
1397 		return;
1398 	}
1399 
1400 	*cs->statusp = resp->status = NFS4_OK;
1401 	resp->writeverf = Write4verf;
1402 }
1403 
1404 /*
1405  * rfs4_op_mknod is called from rfs4_op_create after all initial verification
1406  * was completed. It does the nfsv4 create for special files.
1407  */
1408 /* ARGSUSED */
1409 static vnode_t *
1410 do_rfs4_op_mknod(CREATE4args *args, CREATE4res *resp, struct svc_req *req,
1411 	struct compound_state *cs, vattr_t *vap, char *nm)
1412 {
1413 	int error;
1414 	cred_t *cr = cs->cr;
1415 	vnode_t *dvp = cs->vp;
1416 	vnode_t *vp = NULL;
1417 	int mode;
1418 	enum vcexcl excl;
1419 
1420 	switch (args->type) {
1421 	case NF4CHR:
1422 	case NF4BLK:
1423 		if (secpolicy_sys_devices(cr) != 0) {
1424 			*cs->statusp = resp->status = NFS4ERR_PERM;
1425 			return (NULL);
1426 		}
1427 		if (args->type == NF4CHR)
1428 			vap->va_type = VCHR;
1429 		else
1430 			vap->va_type = VBLK;
1431 		vap->va_rdev = makedevice(args->ftype4_u.devdata.specdata1,
1432 		    args->ftype4_u.devdata.specdata2);
1433 		vap->va_mask |= AT_RDEV;
1434 		break;
1435 	case NF4SOCK:
1436 		vap->va_type = VSOCK;
1437 		break;
1438 	case NF4FIFO:
1439 		vap->va_type = VFIFO;
1440 		break;
1441 	default:
1442 		*cs->statusp = resp->status = NFS4ERR_BADTYPE;
1443 		return (NULL);
1444 	}
1445 
1446 	/*
1447 	 * Must specify the mode.
1448 	 */
1449 	if (!(vap->va_mask & AT_MODE)) {
1450 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1451 		return (NULL);
1452 	}
1453 
1454 	excl = EXCL;
1455 
1456 	mode = 0;
1457 
1458 	error = VOP_CREATE(dvp, nm, vap, excl, mode, &vp, cr, 0);
1459 	if (error) {
1460 		*cs->statusp = resp->status = puterrno4(error);
1461 		return (NULL);
1462 	}
1463 	return (vp);
1464 }
1465 
1466 /*
1467  * nfsv4 create is used to create non-regular files. For regular files,
1468  * use nfsv4 open.
1469  */
1470 /* ARGSUSED */
1471 static void
1472 rfs4_op_create(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1473 	struct compound_state *cs)
1474 {
1475 	CREATE4args *args = &argop->nfs_argop4_u.opcreate;
1476 	CREATE4res *resp = &resop->nfs_resop4_u.opcreate;
1477 	int error;
1478 	struct vattr bva, iva, iva2, ava, *vap;
1479 	cred_t *cr = cs->cr;
1480 	vnode_t *dvp = cs->vp;
1481 	vnode_t *vp = NULL;
1482 	char *nm, *lnm;
1483 	uint_t len, llen;
1484 	int syncval = 0;
1485 	struct nfs4_svgetit_arg sarg;
1486 	struct nfs4_ntov_table ntov;
1487 	struct statvfs64 sb;
1488 	nfsstat4 status;
1489 
1490 	resp->attrset = 0;
1491 
1492 	if (dvp == NULL) {
1493 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1494 		return;
1495 	}
1496 
1497 	/*
1498 	 * If there is an unshared filesystem mounted on this vnode,
1499 	 * do not allow to create an object in this directory.
1500 	 */
1501 	if (vn_ismntpt(dvp)) {
1502 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1503 		return;
1504 	}
1505 
1506 	/* Verify that type is correct */
1507 	switch (args->type) {
1508 	case NF4LNK:
1509 	case NF4BLK:
1510 	case NF4CHR:
1511 	case NF4SOCK:
1512 	case NF4FIFO:
1513 	case NF4DIR:
1514 		break;
1515 	default:
1516 		*cs->statusp = resp->status = NFS4ERR_BADTYPE;
1517 		return;
1518 	};
1519 
1520 	if (cs->access == CS_ACCESS_DENIED) {
1521 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1522 		return;
1523 	}
1524 	if (dvp->v_type != VDIR) {
1525 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
1526 		return;
1527 	}
1528 	if (!utf8_dir_verify(&args->objname)) {
1529 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1530 		return;
1531 	}
1532 
1533 	if (rdonly4(cs->exi, cs->vp, req)) {
1534 		*cs->statusp = resp->status = NFS4ERR_ROFS;
1535 		return;
1536 	}
1537 
1538 	/*
1539 	 * Name of newly created object
1540 	 */
1541 	nm = utf8_to_fn(&args->objname, &len, NULL);
1542 	if (nm == NULL) {
1543 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1544 		return;
1545 	}
1546 
1547 	if (len > MAXNAMELEN) {
1548 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1549 		kmem_free(nm, len);
1550 		return;
1551 	}
1552 
1553 	resp->attrset = 0;
1554 
1555 	sarg.sbp = &sb;
1556 	nfs4_ntov_table_init(&ntov);
1557 
1558 	status = do_rfs4_set_attrs(&resp->attrset,
1559 	    &args->createattrs, cs, &sarg, &ntov, NFS4ATTR_SETIT);
1560 
1561 	if (sarg.vap->va_mask == 0 && status == NFS4_OK)
1562 		status = NFS4ERR_INVAL;
1563 
1564 	if (status != NFS4_OK) {
1565 		*cs->statusp = resp->status = status;
1566 		kmem_free(nm, len);
1567 		nfs4_ntov_table_free(&ntov, &sarg);
1568 		resp->attrset = 0;
1569 		return;
1570 	}
1571 
1572 	/* Get "before" change value */
1573 	bva.va_mask = AT_CTIME|AT_SEQ;
1574 	error = VOP_GETATTR(dvp, &bva, 0, cr);
1575 	if (error) {
1576 		*cs->statusp = resp->status = puterrno4(error);
1577 		kmem_free(nm, len);
1578 		nfs4_ntov_table_free(&ntov, &sarg);
1579 		resp->attrset = 0;
1580 		return;
1581 	}
1582 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bva.va_ctime)
1583 
1584 	vap = sarg.vap;
1585 
1586 	/*
1587 	 * Set default initial values for attributes when not specified
1588 	 * in createattrs.
1589 	 */
1590 	if ((vap->va_mask & AT_UID) == 0) {
1591 		vap->va_uid = crgetuid(cr);
1592 		vap->va_mask |= AT_UID;
1593 	}
1594 	if ((vap->va_mask & AT_GID) == 0) {
1595 		vap->va_gid = crgetgid(cr);
1596 		vap->va_mask |= AT_GID;
1597 	}
1598 
1599 	vap->va_mask |= AT_TYPE;
1600 	switch (args->type) {
1601 	case NF4DIR:
1602 		vap->va_type = VDIR;
1603 		if ((vap->va_mask & AT_MODE) == 0) {
1604 			vap->va_mode = 0700;	/* default: owner rwx only */
1605 			vap->va_mask |= AT_MODE;
1606 		}
1607 		error = VOP_MKDIR(dvp, nm, vap, &vp, cr);
1608 		if (error)
1609 			break;
1610 
1611 		/*
1612 		 * Get the initial "after" sequence number, if it fails,
1613 		 * set to zero
1614 		 */
1615 		iva.va_mask = AT_SEQ;
1616 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr))
1617 			iva.va_seq = 0;
1618 		break;
1619 	case NF4LNK:
1620 		vap->va_type = VLNK;
1621 		if ((vap->va_mask & AT_MODE) == 0) {
1622 			vap->va_mode = 0700;	/* default: owner rwx only */
1623 			vap->va_mask |= AT_MODE;
1624 		}
1625 
1626 		/*
1627 		 * symlink names must be treated as data
1628 		 */
1629 		lnm = utf8_to_str(&args->ftype4_u.linkdata, &llen, NULL);
1630 
1631 		if (lnm == NULL) {
1632 			*cs->statusp = resp->status = NFS4ERR_INVAL;
1633 			kmem_free(nm, len);
1634 			nfs4_ntov_table_free(&ntov, &sarg);
1635 			resp->attrset = 0;
1636 			return;
1637 		}
1638 
1639 		if (llen > MAXPATHLEN) {
1640 			*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1641 			kmem_free(nm, len);
1642 			kmem_free(lnm, llen);
1643 			nfs4_ntov_table_free(&ntov, &sarg);
1644 			resp->attrset = 0;
1645 			return;
1646 		}
1647 
1648 		error = VOP_SYMLINK(dvp, nm, vap, lnm, cr);
1649 		if (lnm != NULL)
1650 			kmem_free(lnm, llen);
1651 		if (error)
1652 			break;
1653 
1654 		/*
1655 		 * Get the initial "after" sequence number, if it fails,
1656 		 * set to zero
1657 		 */
1658 		iva.va_mask = AT_SEQ;
1659 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr))
1660 			iva.va_seq = 0;
1661 
1662 		error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr);
1663 		if (error)
1664 			break;
1665 
1666 		/*
1667 		 * va_seq is not safe over VOP calls, check it again
1668 		 * if it has changed zero out iva to force atomic = FALSE.
1669 		 */
1670 		iva2.va_mask = AT_SEQ;
1671 		if (VOP_GETATTR(dvp, &iva2, 0, cs->cr) ||
1672 		    iva2.va_seq != iva.va_seq)
1673 			iva.va_seq = 0;
1674 		break;
1675 	default:
1676 		/*
1677 		 * probably a special file.
1678 		 */
1679 		if ((vap->va_mask & AT_MODE) == 0) {
1680 			vap->va_mode = 0600;	/* default: owner rw only */
1681 			vap->va_mask |= AT_MODE;
1682 		}
1683 		syncval = FNODSYNC;
1684 		/*
1685 		 * We know this will only generate one VOP call
1686 		 */
1687 		vp = do_rfs4_op_mknod(args, resp, req, cs, vap, nm);
1688 
1689 		if (vp == NULL) {
1690 			kmem_free(nm, len);
1691 			nfs4_ntov_table_free(&ntov, &sarg);
1692 			resp->attrset = 0;
1693 			return;
1694 		}
1695 
1696 		/*
1697 		 * Get the initial "after" sequence number, if it fails,
1698 		 * set to zero
1699 		 */
1700 		iva.va_mask = AT_SEQ;
1701 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr))
1702 			iva.va_seq = 0;
1703 
1704 		break;
1705 	}
1706 	kmem_free(nm, len);
1707 
1708 	if (error) {
1709 		*cs->statusp = resp->status = puterrno4(error);
1710 	}
1711 
1712 	/*
1713 	 * Force modified data and metadata out to stable storage.
1714 	 */
1715 	(void) VOP_FSYNC(dvp, 0, cr);
1716 
1717 	if (resp->status != NFS4_OK) {
1718 		if (vp != NULL)
1719 			VN_RELE(vp);
1720 		nfs4_ntov_table_free(&ntov, &sarg);
1721 		resp->attrset = 0;
1722 		return;
1723 	}
1724 
1725 	/*
1726 	 * Finish setup of cinfo response, "before" value already set.
1727 	 * Get "after" change value, if it fails, simply return the
1728 	 * before value.
1729 	 */
1730 	ava.va_mask = AT_CTIME|AT_SEQ;
1731 	if (VOP_GETATTR(dvp, &ava, 0, cr)) {
1732 		ava.va_ctime = bva.va_ctime;
1733 		ava.va_seq = 0;
1734 	}
1735 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, ava.va_ctime);
1736 
1737 	/*
1738 	 * True verification that object was created with correct
1739 	 * attrs is impossible.  The attrs could have been changed
1740 	 * immediately after object creation.  If attributes did
1741 	 * not verify, the only recourse for the server is to
1742 	 * destroy the object.  Maybe if some attrs (like gid)
1743 	 * are set incorrectly, the object should be destroyed;
1744 	 * however, seems bad as a default policy.  Do we really
1745 	 * want to destroy an object over one of the times not
1746 	 * verifying correctly?  For these reasons, the server
1747 	 * currently sets bits in attrset for createattrs
1748 	 * that were set; however, no verification is done.
1749 	 *
1750 	 * vmask_to_nmask accounts for vattr bits set on create
1751 	 *	[do_rfs4_set_attrs() only sets resp bits for
1752 	 *	 non-vattr/vfs bits.]
1753 	 * Mask off any bits set by default so as not to return
1754 	 * more attrset bits than were requested in createattrs
1755 	 */
1756 	nfs4_vmask_to_nmask(sarg.vap->va_mask, &resp->attrset);
1757 	resp->attrset &= args->createattrs.attrmask;
1758 	nfs4_ntov_table_free(&ntov, &sarg);
1759 
1760 	error = makefh4(&cs->fh, vp, cs->exi);
1761 	if (error) {
1762 		*cs->statusp = resp->status = puterrno4(error);
1763 	}
1764 
1765 	/*
1766 	 * The cinfo.atomic = TRUE only if we got no errors, we have
1767 	 * non-zero va_seq's, and it has incremented by exactly one
1768 	 * during the creation and it didn't change during the VOP_LOOKUP
1769 	 * or VOP_FSYNC.
1770 	 */
1771 	if (!error && bva.va_seq && iva.va_seq && ava.va_seq &&
1772 	    iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq)
1773 		resp->cinfo.atomic = TRUE;
1774 	else
1775 		resp->cinfo.atomic = FALSE;
1776 
1777 	(void) VOP_FSYNC(vp, syncval, cr);
1778 
1779 	if (resp->status != NFS4_OK) {
1780 		VN_RELE(vp);
1781 		return;
1782 	}
1783 	if (cs->vp)
1784 		VN_RELE(cs->vp);
1785 
1786 	cs->vp = vp;
1787 	*cs->statusp = resp->status = NFS4_OK;
1788 }
1789 
1790 
1791 /*ARGSUSED*/
1792 static void
1793 rfs4_op_delegreturn(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1794 	struct compound_state *cs)
1795 {
1796 	DELEGRETURN4args *args = &argop->nfs_argop4_u.opdelegreturn;
1797 	DELEGRETURN4res *resp = &resop->nfs_resop4_u.opdelegreturn;
1798 	rfs4_deleg_state_t *dsp;
1799 	nfsstat4 status;
1800 
1801 	status = rfs4_get_deleg_state(&args->deleg_stateid, &dsp);
1802 	resp->status = *cs->statusp = status;
1803 	if (status != NFS4_OK)
1804 		return;
1805 
1806 	/* Ensure specified filehandle matches */
1807 	if (cs->vp != dsp->finfo->vp) {
1808 		resp->status = *cs->statusp = NFS4ERR_BAD_STATEID;
1809 	} else
1810 		rfs4_return_deleg(dsp, FALSE);
1811 
1812 	rfs4_update_lease(dsp->client);
1813 
1814 	rfs4_deleg_state_rele(dsp);
1815 }
1816 
1817 /*
1818  * Check to see if a given "flavor" is an explicitly shared flavor.
1819  * The assumption of this routine is the "flavor" is already a valid
1820  * flavor in the secinfo list of "exi".
1821  *
1822  *	e.g.
1823  *		# share -o sec=flavor1 /export
1824  *		# share -o sec=flavor2 /export/home
1825  *
1826  *		flavor2 is not an explicitly shared flavor for /export,
1827  *		however it is in the secinfo list for /export thru the
1828  *		server namespace setup.
1829  */
1830 int
1831 is_exported_sec(int flavor, struct exportinfo *exi)
1832 {
1833 	int	i;
1834 	struct secinfo *sp;
1835 
1836 	sp = exi->exi_export.ex_secinfo;
1837 	for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
1838 		if (flavor == sp[i].s_secinfo.sc_nfsnum ||
1839 		    sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) {
1840 			return (SEC_REF_EXPORTED(&sp[i]));
1841 		}
1842 	}
1843 
1844 	/* Should not reach this point based on the assumption */
1845 	return (0);
1846 }
1847 
1848 /*
1849  * Check if the security flavor used in the request matches what is
1850  * required at the export point or at the root pseudo node (exi_root).
1851  *
1852  * returns 1 if there's a match or if exported with AUTH_NONE; 0 otherwise.
1853  *
1854  */
1855 static int
1856 secinfo_match_or_authnone(struct compound_state *cs)
1857 {
1858 	int	i;
1859 	struct secinfo *sp;
1860 
1861 	/*
1862 	 * Check cs->nfsflavor (from the request) against
1863 	 * the current export data in cs->exi.
1864 	 */
1865 	sp = cs->exi->exi_export.ex_secinfo;
1866 	for (i = 0; i < cs->exi->exi_export.ex_seccnt; i++) {
1867 		if (cs->nfsflavor == sp[i].s_secinfo.sc_nfsnum ||
1868 		    sp[i].s_secinfo.sc_nfsnum == AUTH_NONE)
1869 			return (1);
1870 	}
1871 
1872 	return (0);
1873 }
1874 
1875 /*
1876  * Check the access authority for the client and return the correct error.
1877  */
1878 nfsstat4
1879 call_checkauth4(struct compound_state *cs, struct svc_req *req)
1880 {
1881 	int	authres;
1882 
1883 	/*
1884 	 * First, check if the security flavor used in the request
1885 	 * are among the flavors set in the server namespace.
1886 	 */
1887 	if (!secinfo_match_or_authnone(cs)) {
1888 		*cs->statusp = NFS4ERR_WRONGSEC;
1889 		return (*cs->statusp);
1890 	}
1891 
1892 	authres = checkauth4(cs, req);
1893 
1894 	if (authres > 0) {
1895 		*cs->statusp = NFS4_OK;
1896 		if (! (cs->access & CS_ACCESS_LIMITED))
1897 			cs->access = CS_ACCESS_OK;
1898 	} else if (authres == 0) {
1899 		*cs->statusp = NFS4ERR_ACCESS;
1900 	} else if (authres == -2) {
1901 		*cs->statusp = NFS4ERR_WRONGSEC;
1902 	} else {
1903 		*cs->statusp = NFS4ERR_DELAY;
1904 	}
1905 	return (*cs->statusp);
1906 }
1907 
1908 /*
1909  * bitmap4_to_attrmask is called by getattr and readdir.
1910  * It sets up the vattr mask and determines whether vfsstat call is needed
1911  * based on the input bitmap.
1912  * Returns nfsv4 status.
1913  */
1914 static nfsstat4
1915 bitmap4_to_attrmask(bitmap4 breq, struct nfs4_svgetit_arg *sargp)
1916 {
1917 	int i;
1918 	uint_t	va_mask;
1919 	struct statvfs64 *sbp = sargp->sbp;
1920 
1921 	sargp->sbp = NULL;
1922 	sargp->flag = 0;
1923 	sargp->rdattr_error = NFS4_OK;
1924 	sargp->mntdfid_set = FALSE;
1925 	if (sargp->cs->vp)
1926 		sargp->xattr = get_fh4_flag(&sargp->cs->fh,
1927 		    FH4_ATTRDIR | FH4_NAMEDATTR);
1928 	else
1929 		sargp->xattr = 0;
1930 
1931 	/*
1932 	 * Set rdattr_error_req to true if return error per
1933 	 * failed entry rather than fail the readdir.
1934 	 */
1935 	if (breq & FATTR4_RDATTR_ERROR_MASK)
1936 		sargp->rdattr_error_req = 1;
1937 	else
1938 		sargp->rdattr_error_req = 0;
1939 
1940 	/*
1941 	 * generate the va_mask
1942 	 * Handle the easy cases first
1943 	 */
1944 	switch (breq) {
1945 	case NFS4_NTOV_ATTR_MASK:
1946 		sargp->vap->va_mask = NFS4_NTOV_ATTR_AT_MASK;
1947 		return (NFS4_OK);
1948 
1949 	case NFS4_FS_ATTR_MASK:
1950 		sargp->vap->va_mask = NFS4_FS_ATTR_AT_MASK;
1951 		sargp->sbp = sbp;
1952 		return (NFS4_OK);
1953 
1954 	case NFS4_NTOV_ATTR_CACHE_MASK:
1955 		sargp->vap->va_mask = NFS4_NTOV_ATTR_CACHE_AT_MASK;
1956 		return (NFS4_OK);
1957 
1958 	case FATTR4_LEASE_TIME_MASK:
1959 		sargp->vap->va_mask = 0;
1960 		return (NFS4_OK);
1961 
1962 	default:
1963 		va_mask = 0;
1964 		for (i = 0; i < nfs4_ntov_map_size; i++) {
1965 			if ((breq & nfs4_ntov_map[i].fbit) &&
1966 			    nfs4_ntov_map[i].vbit)
1967 				va_mask |= nfs4_ntov_map[i].vbit;
1968 		}
1969 
1970 		/*
1971 		 * Check is vfsstat is needed
1972 		 */
1973 		if (breq & NFS4_FS_ATTR_MASK)
1974 			sargp->sbp = sbp;
1975 
1976 		sargp->vap->va_mask = va_mask;
1977 		return (NFS4_OK);
1978 	}
1979 	/* NOTREACHED */
1980 }
1981 
1982 /*
1983  * bitmap4_get_sysattrs is called by getattr and readdir.
1984  * It calls both VOP_GETATTR and VFS_STATVFS calls to get the attrs.
1985  * Returns nfsv4 status.
1986  */
1987 static nfsstat4
1988 bitmap4_get_sysattrs(struct nfs4_svgetit_arg *sargp)
1989 {
1990 	int error;
1991 	struct compound_state *cs = sargp->cs;
1992 	vnode_t *vp = cs->vp;
1993 
1994 	if (sargp->sbp != NULL) {
1995 		if (error = VFS_STATVFS(vp->v_vfsp, sargp->sbp)) {
1996 			sargp->sbp = NULL;	/* to identify error */
1997 			return (puterrno4(error));
1998 		}
1999 	}
2000 
2001 	return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr));
2002 }
2003 
2004 static void
2005 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp)
2006 {
2007 	ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size,
2008 	    KM_SLEEP);
2009 	ntovp->attrcnt = 0;
2010 	ntovp->vfsstat = FALSE;
2011 }
2012 
2013 static void
2014 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp,
2015 	struct nfs4_svgetit_arg *sargp)
2016 {
2017 	int i;
2018 	union nfs4_attr_u *na;
2019 	uint8_t *amap;
2020 
2021 	/*
2022 	 * XXX Should do the same checks for whether the bit is set
2023 	 */
2024 	for (i = 0, na = ntovp->na, amap = ntovp->amap;
2025 	    i < ntovp->attrcnt; i++, na++, amap++) {
2026 		(void) (*nfs4_ntov_map[*amap].sv_getit)(
2027 		    NFS4ATTR_FREEIT, sargp, na);
2028 	}
2029 	if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) {
2030 		/*
2031 		 * xdr_free for getattr will be done later
2032 		 */
2033 		for (i = 0, na = ntovp->na, amap = ntovp->amap;
2034 		    i < ntovp->attrcnt; i++, na++, amap++) {
2035 			xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na);
2036 		}
2037 	}
2038 	kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size);
2039 }
2040 
2041 /*
2042  * do_rfs4_op_getattr gets the system attrs and converts into fattr4.
2043  */
2044 static nfsstat4
2045 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp,
2046 	struct nfs4_svgetit_arg *sargp)
2047 {
2048 	int error = 0;
2049 	int i, k;
2050 	struct nfs4_ntov_table ntov;
2051 	XDR xdr;
2052 	ulong_t xdr_size;
2053 	char *xdr_attrs;
2054 	nfsstat4 status = NFS4_OK;
2055 	nfsstat4 prev_rdattr_error = sargp->rdattr_error;
2056 	union nfs4_attr_u *na;
2057 	uint8_t *amap;
2058 
2059 	sargp->op = NFS4ATTR_GETIT;
2060 	sargp->flag = 0;
2061 
2062 	fattrp->attrmask = 0;
2063 	/* if no bits requested, then return empty fattr4 */
2064 	if (breq == 0) {
2065 		fattrp->attrlist4_len = 0;
2066 		fattrp->attrlist4 = NULL;
2067 		return (NFS4_OK);
2068 	}
2069 
2070 	/*
2071 	 * return NFS4ERR_INVAL when client requests write-only attrs
2072 	 */
2073 	if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK))
2074 		return (NFS4ERR_INVAL);
2075 
2076 	nfs4_ntov_table_init(&ntov);
2077 	na = ntov.na;
2078 	amap = ntov.amap;
2079 
2080 	/*
2081 	 * Now loop to get or verify the attrs
2082 	 */
2083 	for (i = 0; i < nfs4_ntov_map_size; i++) {
2084 		if (breq & nfs4_ntov_map[i].fbit) {
2085 			if ((*nfs4_ntov_map[i].sv_getit)(
2086 			    NFS4ATTR_SUPPORTED, sargp, NULL) == 0) {
2087 
2088 				error = (*nfs4_ntov_map[i].sv_getit)(
2089 				    NFS4ATTR_GETIT, sargp, na);
2090 
2091 				/*
2092 				 * Possible error values:
2093 				 * >0 if sv_getit failed to
2094 				 * get the attr; 0 if succeeded;
2095 				 * <0 if rdattr_error and the
2096 				 * attribute cannot be returned.
2097 				 */
2098 				if (error && !(sargp->rdattr_error_req))
2099 					goto done;
2100 				/*
2101 				 * If error then just for entry
2102 				 */
2103 				if (error == 0) {
2104 					fattrp->attrmask |=
2105 					    nfs4_ntov_map[i].fbit;
2106 					*amap++ =
2107 					    (uint8_t)nfs4_ntov_map[i].nval;
2108 					na++;
2109 					(ntov.attrcnt)++;
2110 				} else if ((error > 0) &&
2111 				    (sargp->rdattr_error == NFS4_OK)) {
2112 					sargp->rdattr_error = puterrno4(error);
2113 				}
2114 				error = 0;
2115 			}
2116 		}
2117 	}
2118 
2119 	/*
2120 	 * If rdattr_error was set after the return value for it was assigned,
2121 	 * update it.
2122 	 */
2123 	if (prev_rdattr_error != sargp->rdattr_error) {
2124 		na = ntov.na;
2125 		amap = ntov.amap;
2126 		for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2127 			k = *amap;
2128 			if (k < FATTR4_RDATTR_ERROR) {
2129 				continue;
2130 			}
2131 			if ((k == FATTR4_RDATTR_ERROR) &&
2132 			    ((*nfs4_ntov_map[k].sv_getit)(
2133 			    NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) {
2134 
2135 				(void) (*nfs4_ntov_map[k].sv_getit)(
2136 				    NFS4ATTR_GETIT, sargp, na);
2137 			}
2138 			break;
2139 		}
2140 	}
2141 
2142 	xdr_size = 0;
2143 	na = ntov.na;
2144 	amap = ntov.amap;
2145 	for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2146 		xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na);
2147 	}
2148 
2149 	fattrp->attrlist4_len = xdr_size;
2150 	if (xdr_size) {
2151 		/* freed by rfs4_op_getattr_free() */
2152 		fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP);
2153 
2154 		xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE);
2155 
2156 		na = ntov.na;
2157 		amap = ntov.amap;
2158 		for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2159 			if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) {
2160 				cmn_err(CE_WARN, "do_rfs4_op_getattr: xdr "
2161 				    "encode of attribute %d failed\n", *amap);
2162 				status = NFS4ERR_SERVERFAULT;
2163 				break;
2164 			}
2165 		}
2166 		/* xdrmem_destroy(&xdrs); */	/* NO-OP */
2167 	} else {
2168 		fattrp->attrlist4 = NULL;
2169 	}
2170 done:
2171 
2172 	nfs4_ntov_table_free(&ntov, sargp);
2173 
2174 	if (error != 0)
2175 		status = puterrno4(error);
2176 
2177 	return (status);
2178 }
2179 
2180 /* ARGSUSED */
2181 static void
2182 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2183 	struct compound_state *cs)
2184 {
2185 	GETATTR4args *args = &argop->nfs_argop4_u.opgetattr;
2186 	GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr;
2187 	struct nfs4_svgetit_arg sarg;
2188 	struct statvfs64 sb;
2189 	nfsstat4 status;
2190 
2191 	if (cs->vp == NULL) {
2192 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2193 		return;
2194 	}
2195 
2196 	if (cs->access == CS_ACCESS_DENIED) {
2197 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2198 		return;
2199 	}
2200 
2201 	sarg.sbp = &sb;
2202 	sarg.cs = cs;
2203 
2204 	status = bitmap4_to_attrmask(args->attr_request, &sarg);
2205 	if (status == NFS4_OK) {
2206 		status = bitmap4_get_sysattrs(&sarg);
2207 		if (status == NFS4_OK)
2208 			status = do_rfs4_op_getattr(args->attr_request,
2209 			    &resp->obj_attributes, &sarg);
2210 	}
2211 	*cs->statusp = resp->status = status;
2212 }
2213 
2214 static void
2215 rfs4_op_getattr_free(nfs_resop4 *resop)
2216 {
2217 	GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr;
2218 
2219 	nfs4_fattr4_free(&resp->obj_attributes);
2220 }
2221 
2222 /* ARGSUSED */
2223 static void
2224 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2225 	struct compound_state *cs)
2226 {
2227 	GETFH4res *resp = &resop->nfs_resop4_u.opgetfh;
2228 
2229 	if (cs->vp == NULL) {
2230 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2231 		return;
2232 	}
2233 	if (cs->access == CS_ACCESS_DENIED) {
2234 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2235 		return;
2236 	}
2237 
2238 	resp->object.nfs_fh4_val =
2239 	    kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP);
2240 	nfs_fh4_copy(&cs->fh, &resp->object);
2241 	*cs->statusp = resp->status = NFS4_OK;
2242 }
2243 
2244 static void
2245 rfs4_op_getfh_free(nfs_resop4 *resop)
2246 {
2247 	GETFH4res *resp = &resop->nfs_resop4_u.opgetfh;
2248 
2249 	if (resp->status == NFS4_OK &&
2250 	    resp->object.nfs_fh4_val != NULL) {
2251 		kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len);
2252 		resp->object.nfs_fh4_val = NULL;
2253 		resp->object.nfs_fh4_len = 0;
2254 	}
2255 }
2256 
2257 /*
2258  * illegal: args: void
2259  *	    res : status (NFS4ERR_OP_ILLEGAL)
2260  */
2261 /* ARGSUSED */
2262 static void
2263 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop,
2264 	struct svc_req *req, struct compound_state *cs)
2265 {
2266 	ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal;
2267 
2268 	resop->resop = OP_ILLEGAL;
2269 	*cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL;
2270 }
2271 
2272 /*
2273  * link: args: SAVED_FH: file, CURRENT_FH: target directory
2274  *	 res: status. If success - CURRENT_FH unchanged, return change_info
2275  */
2276 /* ARGSUSED */
2277 static void
2278 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2279 	struct compound_state *cs)
2280 {
2281 	LINK4args *args = &argop->nfs_argop4_u.oplink;
2282 	LINK4res *resp = &resop->nfs_resop4_u.oplink;
2283 	int error;
2284 	vnode_t *vp;
2285 	vnode_t *dvp;
2286 	struct vattr bdva, idva, adva;
2287 	char *nm;
2288 	uint_t  len;
2289 
2290 	/* SAVED_FH: source object */
2291 	vp = cs->saved_vp;
2292 	if (vp == NULL) {
2293 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2294 		return;
2295 	}
2296 
2297 	/* CURRENT_FH: target directory */
2298 	dvp = cs->vp;
2299 	if (dvp == NULL) {
2300 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2301 		return;
2302 	}
2303 
2304 	/*
2305 	 * If there is a non-shared filesystem mounted on this vnode,
2306 	 * do not allow to link any file in this directory.
2307 	 */
2308 	if (vn_ismntpt(dvp)) {
2309 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2310 		return;
2311 	}
2312 
2313 	if (cs->access == CS_ACCESS_DENIED) {
2314 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2315 		return;
2316 	}
2317 
2318 	/* Check source object's type validity */
2319 	if (vp->v_type == VDIR) {
2320 		*cs->statusp = resp->status = NFS4ERR_ISDIR;
2321 		return;
2322 	}
2323 
2324 	/* Check target directory's type */
2325 	if (dvp->v_type != VDIR) {
2326 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2327 		return;
2328 	}
2329 
2330 	if (cs->saved_exi != cs->exi) {
2331 		*cs->statusp = resp->status = NFS4ERR_XDEV;
2332 		return;
2333 	}
2334 
2335 	if (!utf8_dir_verify(&args->newname)) {
2336 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2337 		return;
2338 	}
2339 
2340 	nm = utf8_to_fn(&args->newname, &len, NULL);
2341 	if (nm == NULL) {
2342 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2343 		return;
2344 	}
2345 
2346 	if (len > MAXNAMELEN) {
2347 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
2348 		kmem_free(nm, len);
2349 		return;
2350 	}
2351 
2352 	if (rdonly4(cs->exi, cs->vp, req)) {
2353 		*cs->statusp = resp->status = NFS4ERR_ROFS;
2354 		kmem_free(nm, len);
2355 		return;
2356 	}
2357 
2358 	/* Get "before" change value */
2359 	bdva.va_mask = AT_CTIME|AT_SEQ;
2360 	error = VOP_GETATTR(dvp, &bdva, 0, cs->cr);
2361 	if (error) {
2362 		*cs->statusp = resp->status = puterrno4(error);
2363 		kmem_free(nm, len);
2364 		return;
2365 	}
2366 
2367 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime)
2368 
2369 	error = VOP_LINK(dvp, vp, nm, cs->cr);
2370 
2371 	kmem_free(nm, len);
2372 
2373 	/*
2374 	 * Get the initial "after" sequence number, if it fails, set to zero
2375 	 */
2376 	idva.va_mask = AT_SEQ;
2377 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr))
2378 		idva.va_seq = 0;
2379 
2380 	/*
2381 	 * Force modified data and metadata out to stable storage.
2382 	 */
2383 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr);
2384 	(void) VOP_FSYNC(dvp, 0, cs->cr);
2385 
2386 	if (error) {
2387 		*cs->statusp = resp->status = puterrno4(error);
2388 		return;
2389 	}
2390 
2391 	/*
2392 	 * Get "after" change value, if it fails, simply return the
2393 	 * before value.
2394 	 */
2395 	adva.va_mask = AT_CTIME|AT_SEQ;
2396 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr)) {
2397 		adva.va_ctime = bdva.va_ctime;
2398 		adva.va_seq = 0;
2399 	}
2400 
2401 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
2402 
2403 	/*
2404 	 * The cinfo.atomic = TRUE only if we have
2405 	 * non-zero va_seq's, and it has incremented by exactly one
2406 	 * during the VOP_LINK and it didn't change during the VOP_FSYNC.
2407 	 */
2408 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
2409 	    idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq)
2410 		resp->cinfo.atomic = TRUE;
2411 	else
2412 		resp->cinfo.atomic = FALSE;
2413 
2414 	*cs->statusp = resp->status = NFS4_OK;
2415 }
2416 
2417 /*
2418  * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work.
2419  */
2420 
2421 /* ARGSUSED */
2422 static nfsstat4
2423 do_rfs4_op_lookup(char *nm, uint_t buflen, struct svc_req *req,
2424 	struct compound_state *cs)
2425 {
2426 	int error;
2427 	int different_export = 0;
2428 	vnode_t *vp, *tvp, *pre_tvp = NULL, *oldvp = NULL;
2429 	struct exportinfo *exi = NULL, *pre_exi = NULL;
2430 	nfsstat4 stat;
2431 	fid_t fid;
2432 	int attrdir, dotdot, walk;
2433 	bool_t is_newvp = FALSE;
2434 
2435 	if (cs->vp->v_flag & V_XATTRDIR) {
2436 		attrdir = 1;
2437 		ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR));
2438 	} else {
2439 		attrdir = 0;
2440 		ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR));
2441 	}
2442 
2443 	dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0');
2444 
2445 	/*
2446 	 * If dotdotting, then need to check whether it's
2447 	 * above the root of a filesystem, or above an
2448 	 * export point.
2449 	 */
2450 	if (dotdot) {
2451 
2452 		/*
2453 		 * If dotdotting at the root of a filesystem, then
2454 		 * need to traverse back to the mounted-on filesystem
2455 		 * and do the dotdot lookup there.
2456 		 */
2457 		if (cs->vp->v_flag & VROOT) {
2458 
2459 			/*
2460 			 * If at the system root, then can
2461 			 * go up no further.
2462 			 */
2463 			if (VN_CMP(cs->vp, rootdir))
2464 				return (puterrno4(ENOENT));
2465 
2466 			/*
2467 			 * Traverse back to the mounted-on filesystem
2468 			 */
2469 			cs->vp = untraverse(cs->vp);
2470 
2471 			/*
2472 			 * Set the different_export flag so we remember
2473 			 * to pick up a new exportinfo entry for
2474 			 * this new filesystem.
2475 			 */
2476 			different_export = 1;
2477 		} else {
2478 
2479 			/*
2480 			 * If dotdotting above an export point then set
2481 			 * the different_export to get new export info.
2482 			 */
2483 			different_export = nfs_exported(cs->exi, cs->vp);
2484 		}
2485 	}
2486 
2487 	error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr);
2488 	if (error)
2489 		return (puterrno4(error));
2490 
2491 	/*
2492 	 * If the vnode is in a pseudo filesystem, check whether it is visible.
2493 	 *
2494 	 * XXX if the vnode is a symlink and it is not visible in
2495 	 * a pseudo filesystem, return ENOENT (not following symlink).
2496 	 * V4 client can not mount such symlink. This is a regression
2497 	 * from V2/V3.
2498 	 *
2499 	 * In the same exported filesystem, if the security flavor used
2500 	 * is not an explicitly shared flavor, limit the view to the visible
2501 	 * list entries only. This is not a WRONGSEC case because it's already
2502 	 * checked via PUTROOTFH/PUTPUBFH or PUTFH.
2503 	 */
2504 	if (!different_export &&
2505 	    (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) ||
2506 	    cs->access & CS_ACCESS_LIMITED)) {
2507 		if (! nfs_visible(cs->exi, vp, &different_export)) {
2508 			VN_RELE(vp);
2509 			return (puterrno4(ENOENT));
2510 		}
2511 	}
2512 
2513 	/*
2514 	 * If it's a mountpoint, then traverse it.
2515 	 */
2516 	if (vn_ismntpt(vp)) {
2517 		pre_exi = cs->exi;	/* save pre-traversed exportinfo */
2518 		pre_tvp = vp;		/* save pre-traversed vnode	*/
2519 
2520 		/*
2521 		 * hold pre_tvp to counteract rele by traverse.  We will
2522 		 * need pre_tvp below if checkexport4 fails
2523 		 */
2524 		VN_HOLD(pre_tvp);
2525 		tvp = vp;
2526 		if ((error = traverse(&tvp)) != 0) {
2527 			VN_RELE(vp);
2528 			VN_RELE(pre_tvp);
2529 			return (puterrno4(error));
2530 		}
2531 		vp = tvp;
2532 		different_export = 1;
2533 	} else if (vp->v_vfsp != cs->vp->v_vfsp) {
2534 		/*
2535 		 * The vfsp comparison is to handle the case where
2536 		 * a LOFS mount is shared.  lo_lookup traverses mount points,
2537 		 * and NFS is unaware of local fs transistions because
2538 		 * v_vfsmountedhere isn't set.  For this special LOFS case,
2539 		 * the dir and the obj returned by lookup will have different
2540 		 * vfs ptrs.
2541 		 */
2542 		different_export = 1;
2543 	}
2544 
2545 	if (different_export) {
2546 
2547 		bzero(&fid, sizeof (fid));
2548 		fid.fid_len = MAXFIDSZ;
2549 		error = vop_fid_pseudo(vp, &fid);
2550 		if (error) {
2551 			VN_RELE(vp);
2552 			if (pre_tvp)
2553 				VN_RELE(pre_tvp);
2554 			return (puterrno4(error));
2555 		}
2556 
2557 		if (dotdot)
2558 			exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE);
2559 		else
2560 			exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
2561 
2562 		if (exi == NULL) {
2563 			if (pre_tvp) {
2564 				/*
2565 				 * If this vnode is a mounted-on vnode,
2566 				 * but the mounted-on file system is not
2567 				 * exported, send back the filehandle for
2568 				 * the mounted-on vnode, not the root of
2569 				 * the mounted-on file system.
2570 				 */
2571 				VN_RELE(vp);
2572 				vp = pre_tvp;
2573 				exi = pre_exi;
2574 			} else {
2575 				VN_RELE(vp);
2576 				return (puterrno4(EACCES));
2577 			}
2578 		} else if (pre_tvp) {
2579 			/* we're done with pre_tvp now. release extra hold */
2580 			VN_RELE(pre_tvp);
2581 		}
2582 
2583 		cs->exi = exi;
2584 
2585 		/*
2586 		 * Now we do a checkauth4. The reason is that
2587 		 * this client/user may not have access to the new
2588 		 * exported file system, and if he does,
2589 		 * the client/user may be mapped to a different uid.
2590 		 *
2591 		 * We start with a new cr, because the checkauth4 done
2592 		 * in the PUT*FH operation over wrote the cred's uid,
2593 		 * gid, etc, and we want the real thing before calling
2594 		 * checkauth4()
2595 		 */
2596 		crfree(cs->cr);
2597 		cs->cr = crdup(cs->basecr);
2598 
2599 		if (cs->vp)
2600 			oldvp = cs->vp;
2601 		cs->vp = vp;
2602 		is_newvp = TRUE;
2603 
2604 		stat = call_checkauth4(cs, req);
2605 		if (stat != NFS4_OK) {
2606 			VN_RELE(cs->vp);
2607 			cs->vp = oldvp;
2608 			return (stat);
2609 		}
2610 	}
2611 
2612 	/*
2613 	 * After various NFS checks, do a label check on the path
2614 	 * component. The label on this path should either be the
2615 	 * global zone's label or a zone's label. We are only
2616 	 * interested in the zone's label because exported files
2617 	 * in global zone is accessible (though read-only) to
2618 	 * clients. The exportability/visibility check is already
2619 	 * done before reaching this code.
2620 	 */
2621 	if (is_system_labeled()) {
2622 		bslabel_t *clabel;
2623 
2624 		ASSERT(req->rq_label != NULL);
2625 		clabel = req->rq_label;
2626 		DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *,
2627 		    "got client label from request(1)", struct svc_req *, req);
2628 
2629 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
2630 			if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) {
2631 				error = EACCES;
2632 				goto err_out;
2633 			}
2634 		} else {
2635 			/*
2636 			 * We grant access to admin_low label clients
2637 			 * only if the client is trusted, i.e. also
2638 			 * running Solaris Trusted Extension.
2639 			 */
2640 			struct sockaddr	*ca;
2641 			int		addr_type;
2642 			void		*ipaddr;
2643 			tsol_tpc_t	*tp;
2644 
2645 			ca = (struct sockaddr *)svc_getrpccaller(
2646 			    req->rq_xprt)->buf;
2647 			if (ca->sa_family == AF_INET) {
2648 				addr_type = IPV4_VERSION;
2649 				ipaddr = &((struct sockaddr_in *)ca)->sin_addr;
2650 			} else if (ca->sa_family == AF_INET6) {
2651 				addr_type = IPV6_VERSION;
2652 				ipaddr = &((struct sockaddr_in6 *)
2653 				    ca)->sin6_addr;
2654 			}
2655 			tp = find_tpc(ipaddr, addr_type, B_FALSE);
2656 			if (tp == NULL || tp->tpc_tp.tp_doi !=
2657 			    l_admin_low->tsl_doi || tp->tpc_tp.host_type !=
2658 			    SUN_CIPSO) {
2659 				if (tp != NULL)
2660 					TPC_RELE(tp);
2661 				error = EACCES;
2662 				goto err_out;
2663 			}
2664 			TPC_RELE(tp);
2665 		}
2666 	}
2667 
2668 	error = makefh4(&cs->fh, vp, cs->exi);
2669 
2670 err_out:
2671 	if (error) {
2672 		if (is_newvp) {
2673 			VN_RELE(cs->vp);
2674 			cs->vp = oldvp;
2675 		} else
2676 			VN_RELE(vp);
2677 		return (puterrno4(error));
2678 	}
2679 
2680 	if (!is_newvp) {
2681 		if (cs->vp)
2682 			VN_RELE(cs->vp);
2683 		cs->vp = vp;
2684 	} else if (oldvp)
2685 		VN_RELE(oldvp);
2686 
2687 	/*
2688 	 * if did lookup on attrdir and didn't lookup .., set named
2689 	 * attr fh flag
2690 	 */
2691 	if (attrdir && ! dotdot)
2692 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
2693 
2694 	/* Assume false for now, open proc will set this */
2695 	cs->mandlock = FALSE;
2696 
2697 	return (NFS4_OK);
2698 }
2699 
2700 /* ARGSUSED */
2701 static void
2702 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2703 	struct compound_state *cs)
2704 {
2705 	LOOKUP4args *args = &argop->nfs_argop4_u.oplookup;
2706 	LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup;
2707 	char *nm;
2708 	uint_t len;
2709 
2710 	if (cs->vp == NULL) {
2711 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2712 		return;
2713 	}
2714 
2715 	if (cs->vp->v_type == VLNK) {
2716 		*cs->statusp = resp->status = NFS4ERR_SYMLINK;
2717 		return;
2718 	}
2719 
2720 	if (cs->vp->v_type != VDIR) {
2721 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2722 		return;
2723 	}
2724 
2725 	if (!utf8_dir_verify(&args->objname)) {
2726 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2727 		return;
2728 	}
2729 
2730 	nm = utf8_to_str(&args->objname, &len, NULL);
2731 	if (nm == NULL) {
2732 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2733 		return;
2734 	}
2735 
2736 	if (len > MAXNAMELEN) {
2737 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
2738 		kmem_free(nm, len);
2739 		return;
2740 	}
2741 
2742 	*cs->statusp = resp->status = do_rfs4_op_lookup(nm, len, req, cs);
2743 
2744 	kmem_free(nm, len);
2745 }
2746 
2747 /* ARGSUSED */
2748 static void
2749 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
2750 	struct compound_state *cs)
2751 {
2752 	LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp;
2753 
2754 	if (cs->vp == NULL) {
2755 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2756 		return;
2757 	}
2758 
2759 	if (cs->vp->v_type != VDIR) {
2760 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2761 		return;
2762 	}
2763 
2764 	*cs->statusp = resp->status = do_rfs4_op_lookup("..", 3, req, cs);
2765 
2766 	/*
2767 	 * From NFSV4 Specification, LOOKUPP should not check for
2768 	 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead.
2769 	 */
2770 	if (resp->status == NFS4ERR_WRONGSEC) {
2771 		*cs->statusp = resp->status = NFS4_OK;
2772 	}
2773 }
2774 
2775 
2776 /*ARGSUSED2*/
2777 static void
2778 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2779 	struct compound_state *cs)
2780 {
2781 	OPENATTR4args	*args = &argop->nfs_argop4_u.opopenattr;
2782 	OPENATTR4res	*resp = &resop->nfs_resop4_u.opopenattr;
2783 	vnode_t		*avp = NULL;
2784 	int		lookup_flags = LOOKUP_XATTR, error;
2785 	int		exp_ro = 0;
2786 
2787 	if (cs->vp == NULL) {
2788 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2789 		return;
2790 	}
2791 
2792 	/*
2793 	 * Make a couple of checks made by copen()
2794 	 *
2795 	 * Check to make sure underlying fs supports xattrs.  This
2796 	 * is required because solaris filesystem implementations
2797 	 * (UFS/TMPFS) don't enforce the noxattr mount option
2798 	 * in VOP_LOOKUP(LOOKUP_XATTR).  If fs doesn't support this
2799 	 * pathconf cmd or if fs supports cmd but doesn't claim
2800 	 * support for xattr, return NOTSUPP.  It would be better
2801 	 * to use VOP_PATHCONF( _PC_XATTR_ENABLED) for this; however,
2802 	 * that cmd is not available to VOP_PATHCONF interface
2803 	 * (it's only implemented inside pathconf syscall)...
2804 	 *
2805 	 * Verify permission to put attributes on files (access
2806 	 * checks from copen).
2807 	 */
2808 
2809 	if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0) {
2810 		error = ENOTSUP;
2811 		goto error_out;
2812 	}
2813 
2814 	if ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr) != 0) &&
2815 	    (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr) != 0) &&
2816 	    (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr) != 0)) {
2817 		error = EACCES;
2818 		goto error_out;
2819 	}
2820 
2821 	/*
2822 	 * The CREATE_XATTR_DIR VOP flag cannot be specified if
2823 	 * the file system is exported read-only -- regardless of
2824 	 * createdir flag.  Otherwise the attrdir would be created
2825 	 * (assuming server fs isn't mounted readonly locally).  If
2826 	 * VOP_LOOKUP returns ENOENT in this case, the error will
2827 	 * be translated into EROFS.  ENOSYS is mapped to ENOTSUP
2828 	 * because specfs has no VOP_LOOKUP op, so the macro would
2829 	 * return ENOSYS.  EINVAL is returned by all (current)
2830 	 * Solaris file system implementations when any of their
2831 	 * restrictions are violated (xattr(dir) can't have xattrdir).
2832 	 * Returning NOTSUPP is more appropriate in this case
2833 	 * because the object will never be able to have an attrdir.
2834 	 */
2835 	if (args->createdir && ! (exp_ro = rdonly4(cs->exi, cs->vp, req)))
2836 		lookup_flags |= CREATE_XATTR_DIR;
2837 
2838 	error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr);
2839 
2840 	if (error) {
2841 		if (error == ENOENT && args->createdir && exp_ro)
2842 			error = EROFS;
2843 		else if (error == EINVAL || error == ENOSYS)
2844 			error = ENOTSUP;
2845 		goto error_out;
2846 	}
2847 
2848 	ASSERT(avp->v_flag & V_XATTRDIR);
2849 
2850 	error = makefh4(&cs->fh, avp, cs->exi);
2851 
2852 	if (error) {
2853 		VN_RELE(avp);
2854 		goto error_out;
2855 	}
2856 
2857 	VN_RELE(cs->vp);
2858 	cs->vp = avp;
2859 
2860 	/*
2861 	 * There is no requirement for an attrdir fh flag
2862 	 * because the attrdir has a vnode flag to distinguish
2863 	 * it from regular (non-xattr) directories.  The
2864 	 * FH4_ATTRDIR flag is set for future sanity checks.
2865 	 */
2866 	set_fh4_flag(&cs->fh, FH4_ATTRDIR);
2867 	*cs->statusp = resp->status = NFS4_OK;
2868 	return;
2869 
2870 error_out:
2871 
2872 	*cs->statusp = resp->status = puterrno4(error);
2873 }
2874 
2875 static int
2876 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred)
2877 {
2878 	int error;
2879 	int i;
2880 	clock_t delaytime;
2881 	caller_context_t ct;
2882 
2883 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
2884 
2885 	/*
2886 	 * Don't block on mandatory locks. If this routine returns
2887 	 * EAGAIN, the caller should return NFS4ERR_LOCKED.
2888 	 */
2889 	uio->uio_fmode = FNONBLOCK;
2890 
2891 	ct.cc_sysid = 0;
2892 	ct.cc_pid = 0;
2893 	ct.cc_caller_id = nfs4_srv_caller_id;
2894 
2895 	for (i = 0; i < rfs4_maxlock_tries; i++) {
2896 
2897 
2898 		if (direction == FREAD) {
2899 			(void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, &ct);
2900 			error = VOP_READ(vp, uio, ioflag, cred, &ct);
2901 			VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, &ct);
2902 		} else {
2903 			(void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, &ct);
2904 			error = VOP_WRITE(vp, uio, ioflag, cred, &ct);
2905 			VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, &ct);
2906 		}
2907 
2908 		if (error != EAGAIN)
2909 			break;
2910 
2911 		if (i < rfs4_maxlock_tries - 1) {
2912 			delay(delaytime);
2913 			delaytime *= 2;
2914 		}
2915 	}
2916 
2917 	return (error);
2918 }
2919 
2920 /* ARGSUSED */
2921 static void
2922 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2923 	struct compound_state *cs)
2924 {
2925 	READ4args *args = &argop->nfs_argop4_u.opread;
2926 	READ4res *resp = &resop->nfs_resop4_u.opread;
2927 	int error;
2928 	int verror;
2929 	vnode_t *vp;
2930 	struct vattr va;
2931 	struct iovec iov;
2932 	struct uio uio;
2933 	u_offset_t offset;
2934 	bool_t *deleg = &cs->deleg;
2935 	nfsstat4 stat;
2936 	int in_crit = 0;
2937 	mblk_t *mp;
2938 	int alloc_err = 0;
2939 
2940 	vp = cs->vp;
2941 	if (vp == NULL) {
2942 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2943 		return;
2944 	}
2945 	if (cs->access == CS_ACCESS_DENIED) {
2946 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2947 		return;
2948 	}
2949 
2950 	/*
2951 	 * Enter the critical region before calling VOP_RWLOCK
2952 	 * to avoid a deadlock with write requests.
2953 	 */
2954 	if (nbl_need_check(vp)) {
2955 		nbl_start_crit(vp, RW_READER);
2956 		in_crit = 1;
2957 		if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0)) {
2958 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
2959 			goto out;
2960 		}
2961 	}
2962 
2963 	if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE,
2964 	    deleg, TRUE)) != NFS4_OK) {
2965 		*cs->statusp = resp->status = stat;
2966 		goto out;
2967 	}
2968 
2969 	va.va_mask = AT_MODE|AT_SIZE|AT_UID;
2970 	verror = VOP_GETATTR(vp, &va, 0, cs->cr);
2971 
2972 	/*
2973 	 * If we can't get the attributes, then we can't do the
2974 	 * right access checking.  So, we'll fail the request.
2975 	 */
2976 	if (verror) {
2977 		*cs->statusp = resp->status = puterrno4(verror);
2978 		goto out;
2979 	}
2980 
2981 	if (vp->v_type != VREG) {
2982 		*cs->statusp = resp->status =
2983 		    ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
2984 		goto out;
2985 	}
2986 
2987 	if (crgetuid(cs->cr) != va.va_uid &&
2988 	    (error = VOP_ACCESS(vp, VREAD, 0, cs->cr)) &&
2989 	    (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr))) {
2990 		*cs->statusp = resp->status = puterrno4(error);
2991 		goto out;
2992 	}
2993 
2994 	if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */
2995 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2996 		goto out;
2997 	}
2998 
2999 	offset = args->offset;
3000 	if (offset >= va.va_size) {
3001 		*cs->statusp = resp->status = NFS4_OK;
3002 		resp->eof = TRUE;
3003 		resp->data_len = 0;
3004 		resp->data_val = NULL;
3005 		resp->mblk = NULL;
3006 		*cs->statusp = resp->status = NFS4_OK;
3007 		goto out;
3008 	}
3009 
3010 	if (args->count == 0) {
3011 		*cs->statusp = resp->status = NFS4_OK;
3012 		resp->eof = FALSE;
3013 		resp->data_len = 0;
3014 		resp->data_val = NULL;
3015 		resp->mblk = NULL;
3016 		goto out;
3017 	}
3018 
3019 	/*
3020 	 * Do not allocate memory more than maximum allowed
3021 	 * transfer size
3022 	 */
3023 	if (args->count > rfs4_tsize(req))
3024 		args->count = rfs4_tsize(req);
3025 
3026 	/*
3027 	 * mp will contain the data to be sent out in the read reply.
3028 	 * It will be freed after the reply has been sent.
3029 	 * Let's roundup the data to a BYTES_PER_XDR_UNIT multiple,
3030 	 * so that the call to xdrmblk_putmblk() never fails.
3031 	 * If the first alloc of the requested size fails, then
3032 	 * decrease the size to something more reasonable and wait
3033 	 * for the allocation to occur.
3034 	 */
3035 	mp = allocb(RNDUP(args->count), BPRI_MED);
3036 	if (mp == NULL) {
3037 		if (args->count > MAXBSIZE)
3038 			args->count = MAXBSIZE;
3039 		mp = allocb_wait(RNDUP(args->count), BPRI_MED,
3040 		    STR_NOSIG, &alloc_err);
3041 	}
3042 	ASSERT(mp != NULL);
3043 	ASSERT(alloc_err == 0);
3044 
3045 	iov.iov_base = (caddr_t)mp->b_datap->db_base;
3046 	iov.iov_len = args->count;
3047 	uio.uio_iov = &iov;
3048 	uio.uio_iovcnt = 1;
3049 	uio.uio_segflg = UIO_SYSSPACE;
3050 	uio.uio_extflg = UIO_COPY_CACHED;
3051 	uio.uio_loffset = args->offset;
3052 	uio.uio_resid = args->count;
3053 
3054 	error = do_io(FREAD, vp, &uio, 0, cs->cr);
3055 
3056 	va.va_mask = AT_SIZE;
3057 	verror = VOP_GETATTR(vp, &va, 0, cs->cr);
3058 
3059 	if (error) {
3060 		freeb(mp);
3061 		*cs->statusp = resp->status = puterrno4(error);
3062 		goto out;
3063 	}
3064 
3065 	*cs->statusp = resp->status = NFS4_OK;
3066 
3067 	ASSERT(uio.uio_resid >= 0);
3068 	resp->data_len = args->count - uio.uio_resid;
3069 	resp->data_val = (char *)mp->b_datap->db_base;
3070 	resp->mblk = mp;
3071 
3072 	if (!verror && offset + resp->data_len == va.va_size)
3073 		resp->eof = TRUE;
3074 	else
3075 		resp->eof = FALSE;
3076 
3077 out:
3078 	if (in_crit)
3079 		nbl_end_crit(vp);
3080 }
3081 
3082 static void
3083 rfs4_op_read_free(nfs_resop4 *resop)
3084 {
3085 	READ4res *resp = &resop->nfs_resop4_u.opread;
3086 
3087 	if (resp->status == NFS4_OK && resp->mblk != NULL) {
3088 		freeb(resp->mblk);
3089 		resp->mblk = NULL;
3090 		resp->data_val = NULL;
3091 		resp->data_len = 0;
3092 	}
3093 }
3094 
3095 static void
3096 rfs4_op_readdir_free(nfs_resop4 *resop)
3097 {
3098 	READDIR4res *resp = &resop->nfs_resop4_u.opreaddir;
3099 
3100 	if (resp->status == NFS4_OK && resp->mblk != NULL) {
3101 		freeb(resp->mblk);
3102 		resp->mblk = NULL;
3103 		resp->data_len = 0;
3104 	}
3105 }
3106 
3107 
3108 /* ARGSUSED */
3109 static void
3110 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
3111 	struct compound_state *cs)
3112 {
3113 	PUTPUBFH4res *resp = &resop->nfs_resop4_u.opputpubfh;
3114 	int error;
3115 	vnode_t *vp;
3116 	struct exportinfo *exi, *sav_exi;
3117 	nfs_fh4_fmt_t *fh_fmtp;
3118 
3119 	if (cs->vp) {
3120 		VN_RELE(cs->vp);
3121 		cs->vp = NULL;
3122 	}
3123 
3124 	if (cs->cr)
3125 		crfree(cs->cr);
3126 
3127 	cs->cr = crdup(cs->basecr);
3128 
3129 	vp = exi_public->exi_vp;
3130 	if (vp == NULL) {
3131 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
3132 		return;
3133 	}
3134 
3135 	if (is_system_labeled()) {
3136 		bslabel_t *clabel;
3137 
3138 		ASSERT(req->rq_label != NULL);
3139 		clabel = req->rq_label;
3140 		DTRACE_PROBE2(tx__rfs4__log__info__opputpubfh__clabel, char *,
3141 		    "got client label from request(1)",
3142 		    struct svc_req *, req);
3143 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
3144 			if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) {
3145 				*cs->statusp = resp->status =
3146 				    NFS4ERR_SERVERFAULT;
3147 				return;
3148 			}
3149 		}
3150 	}
3151 
3152 	error = makefh4(&cs->fh, vp, exi_public);
3153 	if (error != 0) {
3154 		*cs->statusp = resp->status = puterrno4(error);
3155 		return;
3156 	}
3157 	sav_exi = cs->exi;
3158 	if (exi_public == exi_root) {
3159 		/*
3160 		 * No filesystem is actually shared public, so we default
3161 		 * to exi_root. In this case, we must check whether root
3162 		 * is exported.
3163 		 */
3164 		fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val;
3165 
3166 		/*
3167 		 * if root filesystem is exported, the exportinfo struct that we
3168 		 * should use is what checkexport4 returns, because root_exi is
3169 		 * actually a mostly empty struct.
3170 		 */
3171 		exi = checkexport4(&fh_fmtp->fh4_fsid,
3172 		    (fid_t *)&fh_fmtp->fh4_xlen, NULL);
3173 		cs->exi = ((exi != NULL) ? exi : exi_public);
3174 	} else {
3175 		/*
3176 		 * it's a properly shared filesystem
3177 		 */
3178 		cs->exi = exi_public;
3179 	}
3180 
3181 	VN_HOLD(vp);
3182 	cs->vp = vp;
3183 
3184 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3185 		VN_RELE(cs->vp);
3186 		cs->vp = NULL;
3187 		cs->exi = sav_exi;
3188 		return;
3189 	}
3190 
3191 	*cs->statusp = resp->status = NFS4_OK;
3192 }
3193 
3194 /*
3195  * XXX - issue with put*fh operations. Suppose /export/home is exported.
3196  * Suppose an NFS client goes to mount /export/home/joe. If /export, home,
3197  * or joe have restrictive search permissions, then we shouldn't let
3198  * the client get a file handle. This is easy to enforce. However, we
3199  * don't know what security flavor should be used until we resolve the
3200  * path name. Another complication is uid mapping. If root is
3201  * the user, then it will be mapped to the anonymous user by default,
3202  * but we won't know that till we've resolved the path name. And we won't
3203  * know what the anonymous user is.
3204  * Luckily, SECINFO is specified to take a full filename.
3205  * So what we will have to in rfs4_op_lookup is check that flavor of
3206  * the target object matches that of the request, and if root was the
3207  * caller, check for the root= and anon= options, and if necessary,
3208  * repeat the lookup using the right cred_t. But that's not done yet.
3209  */
3210 /* ARGSUSED */
3211 static void
3212 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3213 	struct compound_state *cs)
3214 {
3215 	PUTFH4args *args = &argop->nfs_argop4_u.opputfh;
3216 	PUTFH4res *resp = &resop->nfs_resop4_u.opputfh;
3217 	nfs_fh4_fmt_t *fh_fmtp;
3218 
3219 	if (cs->vp) {
3220 		VN_RELE(cs->vp);
3221 		cs->vp = NULL;
3222 	}
3223 
3224 	if (cs->cr) {
3225 		crfree(cs->cr);
3226 		cs->cr = NULL;
3227 	}
3228 
3229 
3230 	if (args->object.nfs_fh4_len < NFS_FH4_LEN) {
3231 		*cs->statusp = resp->status = NFS4ERR_BADHANDLE;
3232 		return;
3233 	}
3234 
3235 	fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val;
3236 	cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen,
3237 	    NULL);
3238 
3239 	if (cs->exi == NULL) {
3240 		*cs->statusp = resp->status = NFS4ERR_STALE;
3241 		return;
3242 	}
3243 
3244 	cs->cr = crdup(cs->basecr);
3245 
3246 	ASSERT(cs->cr != NULL);
3247 
3248 	if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) {
3249 		*cs->statusp = resp->status;
3250 		return;
3251 	}
3252 
3253 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3254 		VN_RELE(cs->vp);
3255 		cs->vp = NULL;
3256 		return;
3257 	}
3258 
3259 	nfs_fh4_copy(&args->object, &cs->fh);
3260 	*cs->statusp = resp->status = NFS4_OK;
3261 	cs->deleg = FALSE;
3262 }
3263 
3264 /* ARGSUSED */
3265 static void
3266 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3267 	struct compound_state *cs)
3268 
3269 {
3270 	PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh;
3271 	int error;
3272 	fid_t fid;
3273 	struct exportinfo *exi, *sav_exi;
3274 
3275 	if (cs->vp) {
3276 		VN_RELE(cs->vp);
3277 		cs->vp = NULL;
3278 	}
3279 
3280 	if (cs->cr)
3281 		crfree(cs->cr);
3282 
3283 	cs->cr = crdup(cs->basecr);
3284 
3285 	/*
3286 	 * Using rootdir, the system root vnode,
3287 	 * get its fid.
3288 	 */
3289 	bzero(&fid, sizeof (fid));
3290 	fid.fid_len = MAXFIDSZ;
3291 	error = vop_fid_pseudo(rootdir, &fid);
3292 	if (error != 0) {
3293 		*cs->statusp = resp->status = puterrno4(error);
3294 		return;
3295 	}
3296 
3297 	/*
3298 	 * Then use the root fsid & fid it to find out if it's exported
3299 	 *
3300 	 * If the server root isn't exported directly, then
3301 	 * it should at least be a pseudo export based on
3302 	 * one or more exports further down in the server's
3303 	 * file tree.
3304 	 */
3305 	exi = checkexport4(&rootdir->v_vfsp->vfs_fsid, &fid, NULL);
3306 	if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) {
3307 		NFS4_DEBUG(rfs4_debug,
3308 			(CE_WARN, "rfs4_op_putrootfh: export check failure"));
3309 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
3310 		return;
3311 	}
3312 
3313 	/*
3314 	 * Now make a filehandle based on the root
3315 	 * export and root vnode.
3316 	 */
3317 	error = makefh4(&cs->fh, rootdir, exi);
3318 	if (error != 0) {
3319 		*cs->statusp = resp->status = puterrno4(error);
3320 		return;
3321 	}
3322 
3323 	sav_exi = cs->exi;
3324 	cs->exi = exi;
3325 
3326 	VN_HOLD(rootdir);
3327 	cs->vp = rootdir;
3328 
3329 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3330 		VN_RELE(rootdir);
3331 		cs->vp = NULL;
3332 		cs->exi = sav_exi;
3333 		return;
3334 	}
3335 
3336 	*cs->statusp = resp->status = NFS4_OK;
3337 	cs->deleg = FALSE;
3338 }
3339 
3340 /*
3341  * A directory entry is a valid nfsv4 entry if
3342  * - it has a non-zero ino
3343  * - it is not a dot or dotdot name
3344  * - it is visible in a pseudo export or in a real export that can
3345  *   only have a limited view.
3346  */
3347 static bool_t
3348 valid_nfs4_entry(struct exportinfo *exi, struct dirent64 *dp,
3349 		int *expseudo, int check_visible)
3350 {
3351 	if (dp->d_ino == 0 || NFS_IS_DOTNAME(dp->d_name)) {
3352 		*expseudo = 0;
3353 		return (FALSE);
3354 	}
3355 
3356 	if (! check_visible) {
3357 		*expseudo = 0;
3358 		return (TRUE);
3359 	}
3360 
3361 	return (nfs_visible_inode(exi, dp->d_ino, expseudo));
3362 }
3363 
3364 /*
3365  * set_rdattr_params sets up the variables used to manage what information
3366  * to get for each directory entry.
3367  */
3368 static nfsstat4
3369 set_rdattr_params(struct nfs4_svgetit_arg *sargp,
3370 		bitmap4 attrs, bool_t *need_to_lookup)
3371 {
3372 	uint_t	va_mask;
3373 	nfsstat4 status;
3374 	bitmap4 objbits;
3375 
3376 	status = bitmap4_to_attrmask(attrs, sargp);
3377 	if (status != NFS4_OK) {
3378 		/*
3379 		 * could not even figure attr mask
3380 		 */
3381 		return (status);
3382 	}
3383 	va_mask = sargp->vap->va_mask;
3384 
3385 	/*
3386 	 * dirent's d_ino is always correct value for mounted_on_fileid.
3387 	 * mntdfid_set is set once here, but mounted_on_fileid is
3388 	 * set in main dirent processing loop for each dirent.
3389 	 * The mntdfid_set is a simple optimization that lets the
3390 	 * server attr code avoid work when caller is readdir.
3391 	 */
3392 	sargp->mntdfid_set = TRUE;
3393 
3394 	/*
3395 	 * Lookup entry only if client asked for any of the following:
3396 	 * a) vattr attrs
3397 	 * b) vfs attrs
3398 	 * c) attrs w/per-object scope requested (change, filehandle, etc)
3399 	 *    other than mounted_on_fileid (which we can take from dirent)
3400 	 */
3401 	objbits = attrs ? attrs & NFS4_VP_ATTR_MASK : 0;
3402 
3403 	if (va_mask || sargp->sbp || (objbits & ~FATTR4_MOUNTED_ON_FILEID_MASK))
3404 		*need_to_lookup = TRUE;
3405 	else
3406 		*need_to_lookup = FALSE;
3407 
3408 	if (sargp->sbp == NULL)
3409 		return (NFS4_OK);
3410 
3411 	/*
3412 	 * If filesystem attrs are requested, get them now from the
3413 	 * directory vp, as most entries will have same filesystem. The only
3414 	 * exception are mounted over entries but we handle
3415 	 * those as we go (XXX mounted over detection not yet implemented).
3416 	 */
3417 	sargp->vap->va_mask = 0;	/* to avoid VOP_GETATTR */
3418 	status = bitmap4_get_sysattrs(sargp);
3419 	sargp->vap->va_mask = va_mask;
3420 
3421 	if ((status != NFS4_OK) && sargp->rdattr_error_req) {
3422 		/*
3423 		 * Failed to get filesystem attributes.
3424 		 * Return a rdattr_error for each entry, but don't fail.
3425 		 * However, don't get any obj-dependent attrs.
3426 		 */
3427 		sargp->rdattr_error = status;	/* for rdattr_error */
3428 		*need_to_lookup = FALSE;
3429 		/*
3430 		 * At least get fileid for regular readdir output
3431 		 */
3432 		sargp->vap->va_mask &= AT_NODEID;
3433 		status = NFS4_OK;
3434 	}
3435 
3436 	return (status);
3437 }
3438 
3439 /*
3440  * readlink: args: CURRENT_FH.
3441  *	res: status. If success - CURRENT_FH unchanged, return linktext.
3442  */
3443 
3444 /* ARGSUSED */
3445 static void
3446 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3447 	struct compound_state *cs)
3448 {
3449 	READLINK4res *resp = &resop->nfs_resop4_u.opreadlink;
3450 	int error;
3451 	vnode_t *vp;
3452 	struct iovec iov;
3453 	struct vattr va;
3454 	struct uio uio;
3455 	char *data;
3456 
3457 	/* CURRENT_FH: directory */
3458 	vp = cs->vp;
3459 	if (vp == NULL) {
3460 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3461 		return;
3462 	}
3463 
3464 	if (cs->access == CS_ACCESS_DENIED) {
3465 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3466 		return;
3467 	}
3468 
3469 	if (vp->v_type == VDIR) {
3470 		*cs->statusp = resp->status = NFS4ERR_ISDIR;
3471 		return;
3472 	}
3473 
3474 	if (vp->v_type != VLNK) {
3475 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3476 		return;
3477 	}
3478 
3479 	va.va_mask = AT_MODE;
3480 	error = VOP_GETATTR(vp, &va, 0, cs->cr);
3481 	if (error) {
3482 		*cs->statusp = resp->status = puterrno4(error);
3483 		return;
3484 	}
3485 
3486 	if (MANDLOCK(vp, va.va_mode)) {
3487 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3488 		return;
3489 	}
3490 
3491 	data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP);
3492 
3493 	iov.iov_base = data;
3494 	iov.iov_len = MAXPATHLEN;
3495 	uio.uio_iov = &iov;
3496 	uio.uio_iovcnt = 1;
3497 	uio.uio_segflg = UIO_SYSSPACE;
3498 	uio.uio_extflg = UIO_COPY_CACHED;
3499 	uio.uio_loffset = 0;
3500 	uio.uio_resid = MAXPATHLEN;
3501 
3502 	error = VOP_READLINK(vp, &uio, cs->cr);
3503 
3504 	if (error) {
3505 		kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1);
3506 		*cs->statusp = resp->status = puterrno4(error);
3507 		return;
3508 	}
3509 
3510 	*(data + MAXPATHLEN - uio.uio_resid) = '\0';
3511 
3512 	/*
3513 	 * treat link name as data
3514 	 */
3515 	(void) str_to_utf8(data, &resp->link);
3516 
3517 	kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1);
3518 	*cs->statusp = resp->status = NFS4_OK;
3519 }
3520 
3521 static void
3522 rfs4_op_readlink_free(nfs_resop4 *resop)
3523 {
3524 	READLINK4res *resp = &resop->nfs_resop4_u.opreadlink;
3525 	utf8string *symlink = &resp->link;
3526 
3527 	if (symlink->utf8string_val) {
3528 		UTF8STRING_FREE(*symlink)
3529 	}
3530 }
3531 
3532 /*
3533  * release_lockowner:
3534  *	Release any state associated with the supplied
3535  *	lockowner. Note if any lo_state is holding locks we will not
3536  *	rele that lo_state and thus the lockowner will not be destroyed.
3537  *	A client using lock after the lock owner stateid has been released
3538  *	will suffer the consequence of NFS4ERR_BAD_STATEID and would have
3539  *	to reissue the lock with new_lock_owner set to TRUE.
3540  *	args: lock_owner
3541  *	res:  status
3542  */
3543 /* ARGSUSED */
3544 static void
3545 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop,
3546 	struct svc_req *req, struct compound_state *cs)
3547 {
3548 	RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner;
3549 	RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner;
3550 	rfs4_lockowner_t *lo;
3551 	rfs4_openowner_t *oop;
3552 	rfs4_state_t *sp;
3553 	rfs4_lo_state_t *lsp;
3554 	rfs4_client_t *cp;
3555 	bool_t create = FALSE;
3556 	locklist_t *llist;
3557 	sysid_t sysid;
3558 
3559 	/* Make sure there is a clientid around for this request */
3560 	cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE);
3561 
3562 	if (cp == NULL) {
3563 		*cs->statusp = resp->status =
3564 		    rfs4_check_clientid(&ap->lock_owner.clientid, 0);
3565 		return;
3566 	}
3567 	rfs4_client_rele(cp);
3568 
3569 	lo = rfs4_findlockowner(&ap->lock_owner, &create);
3570 	if (lo == NULL) {
3571 		*cs->statusp = resp->status = NFS4_OK;
3572 		return;
3573 	}
3574 	ASSERT(lo->client != NULL);
3575 
3576 	/*
3577 	 * Check for EXPIRED client. If so will reap state with in a lease
3578 	 * period or on next set_clientid_confirm step
3579 	 */
3580 	if (rfs4_lease_expired(lo->client)) {
3581 		rfs4_lockowner_rele(lo);
3582 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
3583 		return;
3584 	}
3585 
3586 	/*
3587 	 * If no sysid has been assigned, then no locks exist; just return.
3588 	 */
3589 	rfs4_dbe_lock(lo->client->dbe);
3590 	if (lo->client->sysidt == LM_NOSYSID) {
3591 		rfs4_lockowner_rele(lo);
3592 		rfs4_dbe_unlock(lo->client->dbe);
3593 		return;
3594 	}
3595 
3596 	sysid = lo->client->sysidt;
3597 	rfs4_dbe_unlock(lo->client->dbe);
3598 
3599 	/*
3600 	 * Mark the lockowner invalid.
3601 	 */
3602 	rfs4_dbe_hide(lo->dbe);
3603 
3604 	/*
3605 	 * sysid-pid pair should now not be used since the lockowner is
3606 	 * invalid. If the client were to instantiate the lockowner again
3607 	 * it would be assigned a new pid. Thus we can get the list of
3608 	 * current locks.
3609 	 */
3610 
3611 	llist = flk_get_active_locks(sysid, lo->pid);
3612 	/* If we are still holding locks fail */
3613 	if (llist != NULL) {
3614 
3615 		*cs->statusp = resp->status = NFS4ERR_LOCKS_HELD;
3616 
3617 		flk_free_locklist(llist);
3618 		/*
3619 		 * We need to unhide the lockowner so the client can
3620 		 * try it again. The bad thing here is if the client
3621 		 * has a logic error that took it here in the first place
3622 		 * he probably has lost accounting of the locks that it
3623 		 * is holding. So we may have dangling state until the
3624 		 * open owner state is reaped via close. One scenario
3625 		 * that could possibly occur is that the client has
3626 		 * sent the unlock request(s) in separate threads
3627 		 * and has not waited for the replies before sending the
3628 		 * RELEASE_LOCKOWNER request. Presumably, it would expect
3629 		 * and deal appropriately with NFS4ERR_LOCKS_HELD, by
3630 		 * reissuing the request.
3631 		 */
3632 		rfs4_dbe_unhide(lo->dbe);
3633 		rfs4_lockowner_rele(lo);
3634 		return;
3635 	}
3636 
3637 	/*
3638 	 * For the corresponding client we need to check each open
3639 	 * owner for any opens that have lockowner state associated
3640 	 * with this lockowner.
3641 	 */
3642 
3643 	rfs4_dbe_lock(lo->client->dbe);
3644 	for (oop = lo->client->openownerlist.next->oop; oop != NULL;
3645 	    oop = oop->openownerlist.next->oop) {
3646 
3647 		rfs4_dbe_lock(oop->dbe);
3648 		for (sp = oop->ownerstateids.next->sp; sp != NULL;
3649 		    sp = sp->ownerstateids.next->sp) {
3650 
3651 			rfs4_dbe_lock(sp->dbe);
3652 			for (lsp = sp->lockownerlist.next->lsp;
3653 			    lsp != NULL; lsp = lsp->lockownerlist.next->lsp) {
3654 				if (lsp->locker == lo) {
3655 					rfs4_dbe_lock(lsp->dbe);
3656 					rfs4_dbe_invalidate(lsp->dbe);
3657 					rfs4_dbe_unlock(lsp->dbe);
3658 				}
3659 			}
3660 			rfs4_dbe_unlock(sp->dbe);
3661 		}
3662 		rfs4_dbe_unlock(oop->dbe);
3663 	}
3664 	rfs4_dbe_unlock(lo->client->dbe);
3665 
3666 	rfs4_lockowner_rele(lo);
3667 
3668 	*cs->statusp = resp->status = NFS4_OK;
3669 }
3670 
3671 /*
3672  * short utility function to lookup a file and recall the delegation
3673  */
3674 static rfs4_file_t *
3675 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp,
3676 	int *lkup_error, cred_t *cr)
3677 {
3678 	vnode_t *vp;
3679 	rfs4_file_t *fp = NULL;
3680 	bool_t fcreate = FALSE;
3681 	int error;
3682 
3683 	if (vpp)
3684 		*vpp = NULL;
3685 
3686 	if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr)) == 0) {
3687 		if (vp->v_type == VREG)
3688 			fp = rfs4_findfile(vp, NULL, &fcreate);
3689 		if (vpp)
3690 			*vpp = vp;
3691 		else
3692 			VN_RELE(vp);
3693 	}
3694 
3695 	if (lkup_error)
3696 		*lkup_error = error;
3697 
3698 	return (fp);
3699 }
3700 
3701 /*
3702  * remove: args: CURRENT_FH: directory; name.
3703  *	res: status. If success - CURRENT_FH unchanged, return change_info
3704  *		for directory.
3705  */
3706 /* ARGSUSED */
3707 static void
3708 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3709 	struct compound_state *cs)
3710 {
3711 	REMOVE4args *args = &argop->nfs_argop4_u.opremove;
3712 	REMOVE4res *resp = &resop->nfs_resop4_u.opremove;
3713 	int error;
3714 	vnode_t *dvp, *vp;
3715 	struct vattr bdva, idva, adva;
3716 	char *nm;
3717 	uint_t len;
3718 	rfs4_file_t *fp;
3719 	int in_crit = 0;
3720 	bslabel_t *clabel;
3721 
3722 	/* CURRENT_FH: directory */
3723 	dvp = cs->vp;
3724 	if (dvp == NULL) {
3725 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3726 		return;
3727 	}
3728 
3729 	if (cs->access == CS_ACCESS_DENIED) {
3730 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3731 		return;
3732 	}
3733 
3734 	/*
3735 	 * If there is an unshared filesystem mounted on this vnode,
3736 	 * Do not allow to remove anything in this directory.
3737 	 */
3738 	if (vn_ismntpt(dvp)) {
3739 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3740 		return;
3741 	}
3742 
3743 	if (dvp->v_type != VDIR) {
3744 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
3745 		return;
3746 	}
3747 
3748 	if (!utf8_dir_verify(&args->target)) {
3749 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3750 		return;
3751 	}
3752 
3753 	/*
3754 	 * Lookup the file so that we can check if it's a directory
3755 	 */
3756 	nm = utf8_to_fn(&args->target, &len, NULL);
3757 	if (nm == NULL) {
3758 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3759 		return;
3760 	}
3761 
3762 	if (len > MAXNAMELEN) {
3763 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
3764 		kmem_free(nm, len);
3765 		return;
3766 	}
3767 
3768 	if (rdonly4(cs->exi, cs->vp, req)) {
3769 		*cs->statusp = resp->status = NFS4ERR_ROFS;
3770 		kmem_free(nm, len);
3771 		return;
3772 	}
3773 
3774 	/*
3775 	 * Lookup the file to determine type and while we are see if
3776 	 * there is a file struct around and check for delegation.
3777 	 * We don't need to acquire va_seq before this lookup, if
3778 	 * it causes an update, cinfo.before will not match, which will
3779 	 * trigger a cache flush even if atomic is TRUE.
3780 	 */
3781 	if (fp = rfs4_lookup_and_findfile(dvp, nm, &vp, &error, cs->cr)) {
3782 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
3783 		    NULL)) {
3784 			VN_RELE(vp);
3785 			rfs4_file_rele(fp);
3786 			*cs->statusp = resp->status = NFS4ERR_DELAY;
3787 			kmem_free(nm, len);
3788 			return;
3789 		}
3790 	}
3791 
3792 	/* Didn't find anything to remove */
3793 	if (vp == NULL) {
3794 		*cs->statusp = resp->status = error;
3795 		kmem_free(nm, len);
3796 		return;
3797 	}
3798 
3799 	if (nbl_need_check(vp)) {
3800 		nbl_start_crit(vp, RW_READER);
3801 		in_crit = 1;
3802 		if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0)) {
3803 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
3804 			kmem_free(nm, len);
3805 			nbl_end_crit(vp);
3806 			VN_RELE(vp);
3807 			if (fp) {
3808 				rfs4_clear_dont_grant(fp);
3809 				rfs4_file_rele(fp);
3810 			}
3811 			return;
3812 		}
3813 	}
3814 
3815 	/* check label before allowing removal */
3816 	if (is_system_labeled()) {
3817 		ASSERT(req->rq_label != NULL);
3818 		clabel = req->rq_label;
3819 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
3820 		    "got client label from request(1)",
3821 		    struct svc_req *, req);
3822 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
3823 			if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK)) {
3824 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
3825 				kmem_free(nm, len);
3826 				if (in_crit)
3827 					nbl_end_crit(vp);
3828 				VN_RELE(vp);
3829 				if (fp) {
3830 					rfs4_clear_dont_grant(fp);
3831 					rfs4_file_rele(fp);
3832 				}
3833 				return;
3834 			}
3835 		}
3836 	}
3837 
3838 	/* Get dir "before" change value */
3839 	bdva.va_mask = AT_CTIME|AT_SEQ;
3840 	error = VOP_GETATTR(dvp, &bdva, 0, cs->cr);
3841 	if (error) {
3842 		*cs->statusp = resp->status = puterrno4(error);
3843 		kmem_free(nm, len);
3844 		return;
3845 	}
3846 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime)
3847 
3848 	/* Actually do the REMOVE operation */
3849 	if (vp->v_type == VDIR) {
3850 		/*
3851 		 * Can't remove a directory that has a mounted-on filesystem.
3852 		 */
3853 		if (vn_ismntpt(vp)) {
3854 			error = EACCES;
3855 		} else {
3856 			/*
3857 			 * System V defines rmdir to return EEXIST,
3858 			 * not * ENOTEMPTY, if the directory is not
3859 			 * empty.  A System V NFS server needs to map
3860 			 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to
3861 			 * transmit over the wire.
3862 			 */
3863 			if ((error = VOP_RMDIR(dvp, nm, rootdir, cs->cr))
3864 			    == EEXIST)
3865 				error = ENOTEMPTY;
3866 		}
3867 	} else {
3868 		if ((error = VOP_REMOVE(dvp, nm, cs->cr)) == 0 &&
3869 		    fp != NULL) {
3870 			struct vattr va;
3871 			vnode_t *tvp;
3872 
3873 			rfs4_dbe_lock(fp->dbe);
3874 			tvp = fp->vp;
3875 			if (tvp)
3876 				VN_HOLD(tvp);
3877 			rfs4_dbe_unlock(fp->dbe);
3878 
3879 			if (tvp) {
3880 				/*
3881 				 * This is va_seq safe because we are not
3882 				 * manipulating dvp.
3883 				 */
3884 				va.va_mask = AT_NLINK;
3885 				if (!VOP_GETATTR(tvp, &va, 0, cs->cr) &&
3886 				    va.va_nlink == 0) {
3887 					/* Remove state on file remove */
3888 					if (in_crit) {
3889 						nbl_end_crit(vp);
3890 						in_crit = 0;
3891 					}
3892 					rfs4_close_all_state(fp);
3893 				}
3894 				VN_RELE(tvp);
3895 			}
3896 		}
3897 	}
3898 
3899 	if (in_crit)
3900 		nbl_end_crit(vp);
3901 	VN_RELE(vp);
3902 
3903 	if (fp) {
3904 		rfs4_clear_dont_grant(fp);
3905 		rfs4_file_rele(fp);
3906 	}
3907 	kmem_free(nm, len);
3908 
3909 	if (error) {
3910 		*cs->statusp = resp->status = puterrno4(error);
3911 		return;
3912 	}
3913 
3914 	/*
3915 	 * Get the initial "after" sequence number, if it fails, set to zero
3916 	 */
3917 	idva.va_mask = AT_SEQ;
3918 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr))
3919 		idva.va_seq = 0;
3920 
3921 	/*
3922 	 * Force modified data and metadata out to stable storage.
3923 	 */
3924 	(void) VOP_FSYNC(dvp, 0, cs->cr);
3925 
3926 	/*
3927 	 * Get "after" change value, if it fails, simply return the
3928 	 * before value.
3929 	 */
3930 	adva.va_mask = AT_CTIME|AT_SEQ;
3931 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr)) {
3932 		adva.va_ctime = bdva.va_ctime;
3933 		adva.va_seq = 0;
3934 	}
3935 
3936 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
3937 
3938 	/*
3939 	 * The cinfo.atomic = TRUE only if we have
3940 	 * non-zero va_seq's, and it has incremented by exactly one
3941 	 * during the VOP_REMOVE/RMDIR and it didn't change during
3942 	 * the VOP_FSYNC.
3943 	 */
3944 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
3945 	    idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq)
3946 		resp->cinfo.atomic = TRUE;
3947 	else
3948 		resp->cinfo.atomic = FALSE;
3949 
3950 	*cs->statusp = resp->status = NFS4_OK;
3951 }
3952 
3953 /*
3954  * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory,
3955  *		oldname and newname.
3956  *	res: status. If success - CURRENT_FH unchanged, return change_info
3957  *		for both from and target directories.
3958  */
3959 /* ARGSUSED */
3960 static void
3961 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3962 	struct compound_state *cs)
3963 {
3964 	RENAME4args *args = &argop->nfs_argop4_u.oprename;
3965 	RENAME4res *resp = &resop->nfs_resop4_u.oprename;
3966 	int error;
3967 	vnode_t *odvp;
3968 	vnode_t *ndvp;
3969 	vnode_t *srcvp, *targvp;
3970 	struct vattr obdva, oidva, oadva;
3971 	struct vattr nbdva, nidva, nadva;
3972 	char *onm, *nnm;
3973 	uint_t olen, nlen;
3974 	rfs4_file_t *fp, *sfp;
3975 	int in_crit_src, in_crit_targ;
3976 	int fp_rele_grant_hold, sfp_rele_grant_hold;
3977 	bslabel_t *clabel;
3978 
3979 	fp = sfp = NULL;
3980 	srcvp = targvp = NULL;
3981 	in_crit_src = in_crit_targ = 0;
3982 	fp_rele_grant_hold = sfp_rele_grant_hold = 0;
3983 
3984 	/* CURRENT_FH: target directory */
3985 	ndvp = cs->vp;
3986 	if (ndvp == NULL) {
3987 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3988 		return;
3989 	}
3990 
3991 	/* SAVED_FH: from directory */
3992 	odvp = cs->saved_vp;
3993 	if (odvp == NULL) {
3994 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3995 		return;
3996 	}
3997 
3998 	if (cs->access == CS_ACCESS_DENIED) {
3999 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4000 		return;
4001 	}
4002 
4003 	/*
4004 	 * If there is an unshared filesystem mounted on this vnode,
4005 	 * do not allow to rename objects in this directory.
4006 	 */
4007 	if (vn_ismntpt(odvp)) {
4008 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4009 		return;
4010 	}
4011 
4012 	/*
4013 	 * If there is an unshared filesystem mounted on this vnode,
4014 	 * do not allow to rename to this directory.
4015 	 */
4016 	if (vn_ismntpt(ndvp)) {
4017 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4018 		return;
4019 	}
4020 
4021 	if (odvp->v_type != VDIR || ndvp->v_type != VDIR) {
4022 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
4023 		return;
4024 	}
4025 
4026 	if (cs->saved_exi != cs->exi) {
4027 		*cs->statusp = resp->status = NFS4ERR_XDEV;
4028 		return;
4029 	}
4030 
4031 	if (!utf8_dir_verify(&args->oldname)) {
4032 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4033 		return;
4034 	}
4035 
4036 	if (!utf8_dir_verify(&args->newname)) {
4037 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4038 		return;
4039 	}
4040 
4041 	onm = utf8_to_fn(&args->oldname, &olen, NULL);
4042 	if (onm == NULL) {
4043 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4044 		return;
4045 	}
4046 
4047 	nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4048 	if (nnm == NULL) {
4049 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4050 		kmem_free(onm, olen);
4051 		return;
4052 	}
4053 
4054 	if (olen > MAXNAMELEN || nlen > MAXNAMELEN) {
4055 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
4056 		kmem_free(onm, olen);
4057 		kmem_free(nnm, nlen);
4058 		return;
4059 	}
4060 
4061 
4062 	if (rdonly4(cs->exi, cs->vp, req)) {
4063 		*cs->statusp = resp->status = NFS4ERR_ROFS;
4064 		kmem_free(onm, olen);
4065 		kmem_free(nnm, nlen);
4066 		return;
4067 	}
4068 
4069 	/* check label of the target dir */
4070 	if (is_system_labeled()) {
4071 		ASSERT(req->rq_label != NULL);
4072 		clabel = req->rq_label;
4073 		DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *,
4074 		    "got client label from request(1)",
4075 		    struct svc_req *, req);
4076 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
4077 			if (!do_rfs_label_check(clabel, ndvp,
4078 			    EQUALITY_CHECK)) {
4079 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
4080 				return;
4081 			}
4082 		}
4083 	}
4084 
4085 	/*
4086 	 * Is the source a file and have a delegation?
4087 	 * We don't need to acquire va_seq before these lookups, if
4088 	 * it causes an update, cinfo.before will not match, which will
4089 	 * trigger a cache flush even if atomic is TRUE.
4090 	 */
4091 	if (sfp = rfs4_lookup_and_findfile(odvp, onm, &srcvp, &error, cs->cr)) {
4092 		if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE,
4093 		    NULL)) {
4094 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4095 			goto err_out;
4096 		}
4097 	}
4098 
4099 	if (srcvp == NULL) {
4100 		*cs->statusp = resp->status = puterrno4(error);
4101 		kmem_free(onm, olen);
4102 		kmem_free(nnm, nlen);
4103 		return;
4104 	}
4105 
4106 	sfp_rele_grant_hold = 1;
4107 
4108 	/* Does the destination exist and a file and have a delegation? */
4109 	if (fp = rfs4_lookup_and_findfile(ndvp, nnm, &targvp, NULL, cs->cr)) {
4110 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
4111 		    NULL)) {
4112 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4113 			goto err_out;
4114 		}
4115 	}
4116 	fp_rele_grant_hold = 1;
4117 
4118 
4119 	/* Check for NBMAND lock on both source and target */
4120 	if (nbl_need_check(srcvp)) {
4121 		nbl_start_crit(srcvp, RW_READER);
4122 		in_crit_src = 1;
4123 		if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0)) {
4124 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4125 			goto err_out;
4126 		}
4127 	}
4128 
4129 	if (targvp && nbl_need_check(targvp)) {
4130 		nbl_start_crit(targvp, RW_READER);
4131 		in_crit_targ = 1;
4132 		if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0)) {
4133 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4134 			goto err_out;
4135 		}
4136 	}
4137 
4138 	/* Get source "before" change value */
4139 	obdva.va_mask = AT_CTIME|AT_SEQ;
4140 	error = VOP_GETATTR(odvp, &obdva, 0, cs->cr);
4141 	if (!error) {
4142 		nbdva.va_mask = AT_CTIME|AT_SEQ;
4143 		error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr);
4144 	}
4145 	if (error) {
4146 		*cs->statusp = resp->status = puterrno4(error);
4147 		goto err_out;
4148 	}
4149 
4150 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime)
4151 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime)
4152 
4153 	if ((error = VOP_RENAME(odvp, onm, ndvp, nnm, cs->cr)) == 0 &&
4154 	    fp != NULL) {
4155 		struct vattr va;
4156 		vnode_t *tvp;
4157 
4158 		rfs4_dbe_lock(fp->dbe);
4159 		tvp = fp->vp;
4160 		if (tvp)
4161 			VN_HOLD(tvp);
4162 		rfs4_dbe_unlock(fp->dbe);
4163 
4164 		if (tvp) {
4165 			va.va_mask = AT_NLINK;
4166 			if (!VOP_GETATTR(tvp, &va, 0, cs->cr) &&
4167 			    va.va_nlink == 0) {
4168 				/* The file is gone and so should the state */
4169 				if (in_crit_targ) {
4170 					nbl_end_crit(targvp);
4171 					in_crit_targ = 0;
4172 				}
4173 				rfs4_close_all_state(fp);
4174 			}
4175 			VN_RELE(tvp);
4176 		}
4177 	}
4178 	if (error == 0) {
4179 		char *tmp;
4180 
4181 		/* fix the path name for the renamed file */
4182 		mutex_enter(&srcvp->v_lock);
4183 		tmp = srcvp->v_path;
4184 		srcvp->v_path = NULL;
4185 		mutex_exit(&srcvp->v_lock);
4186 		vn_setpath(rootdir, ndvp, srcvp, nnm, nlen - 1);
4187 		if (tmp != NULL)
4188 			kmem_free(tmp, strlen(tmp) + 1);
4189 	}
4190 
4191 	if (in_crit_src)
4192 		nbl_end_crit(srcvp);
4193 	if (srcvp)
4194 		VN_RELE(srcvp);
4195 	if (in_crit_targ)
4196 		nbl_end_crit(targvp);
4197 	if (targvp)
4198 		VN_RELE(targvp);
4199 
4200 	if (sfp) {
4201 		rfs4_clear_dont_grant(sfp);
4202 		rfs4_file_rele(sfp);
4203 	}
4204 	if (fp) {
4205 		rfs4_clear_dont_grant(fp);
4206 		rfs4_file_rele(fp);
4207 	}
4208 
4209 	kmem_free(onm, olen);
4210 	kmem_free(nnm, nlen);
4211 
4212 	/*
4213 	 * Get the initial "after" sequence number, if it fails, set to zero
4214 	 */
4215 	oidva.va_mask = AT_SEQ;
4216 	if (VOP_GETATTR(odvp, &oidva, 0, cs->cr))
4217 		oidva.va_seq = 0;
4218 
4219 	nidva.va_mask = AT_SEQ;
4220 	if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr))
4221 		nidva.va_seq = 0;
4222 
4223 	/*
4224 	 * Force modified data and metadata out to stable storage.
4225 	 */
4226 	(void) VOP_FSYNC(odvp, 0, cs->cr);
4227 	(void) VOP_FSYNC(ndvp, 0, cs->cr);
4228 
4229 	if (error) {
4230 		*cs->statusp = resp->status = puterrno4(error);
4231 		return;
4232 	}
4233 
4234 	/*
4235 	 * Get "after" change values, if it fails, simply return the
4236 	 * before value.
4237 	 */
4238 	oadva.va_mask = AT_CTIME|AT_SEQ;
4239 	if (VOP_GETATTR(odvp, &oadva, 0, cs->cr)) {
4240 		oadva.va_ctime = obdva.va_ctime;
4241 		oadva.va_seq = 0;
4242 	}
4243 
4244 	nadva.va_mask = AT_CTIME|AT_SEQ;
4245 	if (VOP_GETATTR(odvp, &nadva, 0, cs->cr)) {
4246 		nadva.va_ctime = nbdva.va_ctime;
4247 		nadva.va_seq = 0;
4248 	}
4249 
4250 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime)
4251 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime)
4252 
4253 	/*
4254 	 * The cinfo.atomic = TRUE only if we have
4255 	 * non-zero va_seq's, and it has incremented by exactly one
4256 	 * during the VOP_RENAME and it didn't change during the VOP_FSYNC.
4257 	 */
4258 	if (obdva.va_seq && oidva.va_seq && oadva.va_seq &&
4259 	    oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq)
4260 		resp->source_cinfo.atomic = TRUE;
4261 	else
4262 		resp->source_cinfo.atomic = FALSE;
4263 
4264 	if (nbdva.va_seq && nidva.va_seq && nadva.va_seq &&
4265 	    nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq)
4266 		resp->target_cinfo.atomic = TRUE;
4267 	else
4268 		resp->target_cinfo.atomic = FALSE;
4269 
4270 #ifdef	VOLATILE_FH_TEST
4271 	{
4272 	extern void add_volrnm_fh(struct exportinfo *, vnode_t *);
4273 
4274 	/*
4275 	 * Add the renamed file handle to the volatile rename list
4276 	 */
4277 	if (cs->exi->exi_export.ex_flags & EX_VOLRNM) {
4278 		/* file handles may expire on rename */
4279 		vnode_t *vp;
4280 
4281 		nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4282 		/*
4283 		 * Already know that nnm will be a valid string
4284 		 */
4285 		error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr);
4286 		kmem_free(nnm, nlen);
4287 		if (!error) {
4288 			add_volrnm_fh(cs->exi, vp);
4289 			VN_RELE(vp);
4290 		}
4291 	}
4292 	}
4293 #endif	/* VOLATILE_FH_TEST */
4294 
4295 	*cs->statusp = resp->status = NFS4_OK;
4296 	return;
4297 
4298 err_out:
4299 	kmem_free(onm, olen);
4300 	kmem_free(nnm, nlen);
4301 
4302 	if (in_crit_src) nbl_end_crit(srcvp);
4303 	if (in_crit_targ) nbl_end_crit(targvp);
4304 	if (targvp) VN_RELE(targvp);
4305 	if (srcvp) VN_RELE(srcvp);
4306 	if (sfp) {
4307 		if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp);
4308 		rfs4_file_rele(sfp);
4309 	}
4310 	if (fp) {
4311 		if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp);
4312 		rfs4_file_rele(fp);
4313 	}
4314 }
4315 
4316 /* ARGSUSED */
4317 static void
4318 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4319 	struct compound_state *cs)
4320 {
4321 	RENEW4args *args = &argop->nfs_argop4_u.oprenew;
4322 	RENEW4res *resp = &resop->nfs_resop4_u.oprenew;
4323 	rfs4_client_t *cp;
4324 
4325 	if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) {
4326 		*cs->statusp = resp->status =
4327 		    rfs4_check_clientid(&args->clientid, 0);
4328 		return;
4329 	}
4330 
4331 	if (rfs4_lease_expired(cp)) {
4332 		rfs4_client_rele(cp);
4333 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
4334 		return;
4335 	}
4336 
4337 	rfs4_update_lease(cp);
4338 
4339 	mutex_enter(cp->cbinfo.cb_lock);
4340 	if (cp->cbinfo.cb_notified_of_cb_path_down == FALSE) {
4341 		cp->cbinfo.cb_notified_of_cb_path_down = TRUE;
4342 		*cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN;
4343 	} else {
4344 		*cs->statusp = resp->status = NFS4_OK;
4345 	}
4346 	mutex_exit(cp->cbinfo.cb_lock);
4347 
4348 	rfs4_client_rele(cp);
4349 
4350 }
4351 
4352 /* ARGSUSED */
4353 static void
4354 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
4355 	struct compound_state *cs)
4356 {
4357 	RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh;
4358 
4359 	/* No need to check cs->access - we are not accessing any object */
4360 	if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) {
4361 		*cs->statusp = resp->status = NFS4ERR_RESTOREFH;
4362 		return;
4363 	}
4364 	if (cs->vp != NULL) {
4365 		VN_RELE(cs->vp);
4366 	}
4367 	cs->vp = cs->saved_vp;
4368 	cs->saved_vp = NULL;
4369 	cs->exi = cs->saved_exi;
4370 	nfs_fh4_copy(&cs->saved_fh, &cs->fh);
4371 	*cs->statusp = resp->status = NFS4_OK;
4372 	cs->deleg = FALSE;
4373 }
4374 
4375 /* ARGSUSED */
4376 static void
4377 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4378 	struct compound_state *cs)
4379 {
4380 	SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh;
4381 
4382 	/* No need to check cs->access - we are not accessing any object */
4383 	if (cs->vp == NULL) {
4384 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4385 		return;
4386 	}
4387 	if (cs->saved_vp != NULL) {
4388 		VN_RELE(cs->saved_vp);
4389 	}
4390 	cs->saved_vp = cs->vp;
4391 	VN_HOLD(cs->saved_vp);
4392 	cs->saved_exi = cs->exi;
4393 	/*
4394 	 * since SAVEFH is fairly rare, don't alloc space for its fh
4395 	 * unless necessary.
4396 	 */
4397 	if (cs->saved_fh.nfs_fh4_val == NULL) {
4398 		cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP);
4399 	}
4400 	nfs_fh4_copy(&cs->fh, &cs->saved_fh);
4401 	*cs->statusp = resp->status = NFS4_OK;
4402 }
4403 
4404 /*
4405  * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to
4406  * return the bitmap of attrs that were set successfully. It is also
4407  * called by Verify/Nverify to test the vattr/vfsstat attrs. It should
4408  * always be called only after rfs4_do_set_attrs().
4409  *
4410  * Verify that the attributes are same as the expected ones. sargp->vap
4411  * and sargp->sbp contain the input attributes as translated from fattr4.
4412  *
4413  * This function verifies only the attrs that correspond to a vattr or
4414  * vfsstat struct. That is because of the extra step needed to get the
4415  * corresponding system structs. Other attributes have already been set or
4416  * verified by do_rfs4_set_attrs.
4417  *
4418  * Return 0 if all attrs match, -1 if some don't, error if error processing.
4419  */
4420 static int
4421 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp,
4422 	bitmap4 *resp, struct nfs4_ntov_table *ntovp)
4423 {
4424 	int error, ret_error = 0;
4425 	int i, k;
4426 	uint_t sva_mask = sargp->vap->va_mask;
4427 	uint_t vbit;
4428 	union nfs4_attr_u *na;
4429 	uint8_t *amap;
4430 	bool_t getsb = ntovp->vfsstat;
4431 
4432 	if (sva_mask != 0) {
4433 		/*
4434 		 * Okay to overwrite sargp->vap because we verify based
4435 		 * on the incoming values.
4436 		 */
4437 		ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0,
4438 		    sargp->cs->cr);
4439 		if (ret_error) {
4440 			if (resp == NULL)
4441 				return (ret_error);
4442 			/*
4443 			 * Must return bitmap of successful attrs
4444 			 */
4445 			sva_mask = 0;	/* to prevent checking vap later */
4446 		} else {
4447 			/*
4448 			 * Some file systems clobber va_mask. it is probably
4449 			 * wrong of them to do so, nonethless we practice
4450 			 * defensive coding.
4451 			 * See bug id 4276830.
4452 			 */
4453 			sargp->vap->va_mask = sva_mask;
4454 		}
4455 	}
4456 
4457 	if (getsb) {
4458 		/*
4459 		 * Now get the superblock and loop on the bitmap, as there is
4460 		 * no simple way of translating from superblock to bitmap4.
4461 		 */
4462 		ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp);
4463 		if (ret_error) {
4464 			if (resp == NULL)
4465 				goto errout;
4466 			getsb = FALSE;
4467 		}
4468 	}
4469 
4470 	/*
4471 	 * Now loop and verify each attribute which getattr returned
4472 	 * whether it's the same as the input.
4473 	 */
4474 	if (resp == NULL && !getsb && (sva_mask == 0))
4475 		goto errout;
4476 
4477 	na = ntovp->na;
4478 	amap = ntovp->amap;
4479 	k = 0;
4480 	for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) {
4481 		k = *amap;
4482 		ASSERT(nfs4_ntov_map[k].nval == k);
4483 		vbit = nfs4_ntov_map[k].vbit;
4484 
4485 		/*
4486 		 * If vattr attribute but VOP_GETATTR failed, or it's
4487 		 * superblock attribute but VFS_STATVFS failed, skip
4488 		 */
4489 		if (vbit) {
4490 			if ((vbit & sva_mask) == 0)
4491 				continue;
4492 		} else if (!(getsb && nfs4_ntov_map[k].vfsstat)) {
4493 			continue;
4494 		}
4495 		error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na);
4496 		if (resp != NULL) {
4497 			if (error)
4498 				ret_error = -1;	/* not all match */
4499 			else	/* update response bitmap */
4500 				*resp |= nfs4_ntov_map[k].fbit;
4501 			continue;
4502 		}
4503 		if (error) {
4504 			ret_error = -1;	/* not all match */
4505 			break;
4506 		}
4507 	}
4508 errout:
4509 	return (ret_error);
4510 }
4511 
4512 /*
4513  * Decode the attribute to be set/verified. If the attr requires a sys op
4514  * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't
4515  * call the sv_getit function for it, because the sys op hasn't yet been done.
4516  * Return 0 for success, error code if failed.
4517  *
4518  * Note: the decoded arg is not freed here but in nfs4_ntov_table_free.
4519  */
4520 static int
4521 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp,
4522 	int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap)
4523 {
4524 	int error = 0;
4525 	bool_t set_later;
4526 
4527 	sargp->vap->va_mask |= nfs4_ntov_map[k].vbit;
4528 
4529 	if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) {
4530 		set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat;
4531 		/*
4532 		 * don't verify yet if a vattr or sb dependent attr,
4533 		 * because we don't have their sys values yet.
4534 		 * Will be done later.
4535 		 */
4536 		if (! (set_later && (cmd == NFS4ATTR_VERIT))) {
4537 			/*
4538 			 * ACLs are a special case, since setting the MODE
4539 			 * conflicts with setting the ACL.  We delay setting
4540 			 * the ACL until all other attributes have been set.
4541 			 * The ACL gets set in do_rfs4_op_setattr().
4542 			 */
4543 			if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) {
4544 				error = (*nfs4_ntov_map[k].sv_getit)(cmd,
4545 				    sargp, nap);
4546 				if (error) {
4547 					xdr_free(nfs4_ntov_map[k].xfunc,
4548 					    (caddr_t)nap);
4549 				}
4550 			}
4551 		}
4552 	} else {
4553 #ifdef  DEBUG
4554 		cmn_err(CE_NOTE, "decode_fattr4_attr: error "
4555 		    "decoding attribute %d\n", k);
4556 #endif
4557 		error = EINVAL;
4558 	}
4559 	if (!error && resp_bval && !set_later) {
4560 		*resp_bval |= nfs4_ntov_map[k].fbit;
4561 	}
4562 
4563 	return (error);
4564 }
4565 
4566 /*
4567  * Set vattr based on incoming fattr4 attrs - used by setattr.
4568  * Set response mask. Ignore any values that are not writable vattr attrs.
4569  */
4570 static nfsstat4
4571 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
4572 		struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp,
4573 		nfs4_attr_cmd_t cmd)
4574 {
4575 	int error = 0;
4576 	int i;
4577 	char *attrs = fattrp->attrlist4;
4578 	uint32_t attrslen = fattrp->attrlist4_len;
4579 	XDR xdr;
4580 	nfsstat4 status = NFS4_OK;
4581 	vnode_t *vp = cs->vp;
4582 	union nfs4_attr_u *na;
4583 	uint8_t *amap;
4584 
4585 #ifndef lint
4586 	/*
4587 	 * Make sure that maximum attribute number can be expressed as an
4588 	 * 8 bit quantity.
4589 	 */
4590 	ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1));
4591 #endif
4592 
4593 	if (vp == NULL) {
4594 		if (resp)
4595 			*resp = 0;
4596 		return (NFS4ERR_NOFILEHANDLE);
4597 	}
4598 	if (cs->access == CS_ACCESS_DENIED) {
4599 		if (resp)
4600 			*resp = 0;
4601 		return (NFS4ERR_ACCESS);
4602 	}
4603 
4604 	sargp->op = cmd;
4605 	sargp->cs = cs;
4606 	sargp->flag = 0;	/* may be set later */
4607 	sargp->vap->va_mask = 0;
4608 	sargp->rdattr_error = NFS4_OK;
4609 	sargp->rdattr_error_req = FALSE;
4610 	/* sargp->sbp is set by the caller */
4611 
4612 	xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE);
4613 
4614 	na = ntovp->na;
4615 	amap = ntovp->amap;
4616 
4617 	/*
4618 	 * The following loop iterates on the nfs4_ntov_map checking
4619 	 * if the fbit is set in the requested bitmap.
4620 	 * If set then we process the arguments using the
4621 	 * rfs4_fattr4 conversion functions to populate the setattr
4622 	 * vattr and va_mask. Any settable attrs that are not using vattr
4623 	 * will be set in this loop.
4624 	 */
4625 	for (i = 0; i < nfs4_ntov_map_size; i++) {
4626 		if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) {
4627 			continue;
4628 		}
4629 		/*
4630 		 * If setattr, must be a writable attr.
4631 		 * If verify/nverify, must be a readable attr.
4632 		 */
4633 		if ((error = (*nfs4_ntov_map[i].sv_getit)(
4634 		    NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) {
4635 			/*
4636 			 * Client tries to set/verify an
4637 			 * unsupported attribute, tries to set
4638 			 * a read only attr or verify a write
4639 			 * only one - error!
4640 			 */
4641 			break;
4642 		}
4643 		/*
4644 		 * Decode the attribute to set/verify
4645 		 */
4646 		error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval,
4647 		    &xdr, resp ? resp : NULL, na);
4648 		if (error)
4649 			break;
4650 		*amap++ = (uint8_t)nfs4_ntov_map[i].nval;
4651 		na++;
4652 		(ntovp->attrcnt)++;
4653 		if (nfs4_ntov_map[i].vfsstat)
4654 			ntovp->vfsstat = TRUE;
4655 	}
4656 
4657 	if (error != 0)
4658 		status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP :
4659 		    puterrno4(error));
4660 	/* xdrmem_destroy(&xdrs); */	/* NO-OP */
4661 	return (status);
4662 }
4663 
4664 static nfsstat4
4665 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
4666 		stateid4 *stateid)
4667 {
4668 	int error = 0;
4669 	struct nfs4_svgetit_arg sarg;
4670 	bool_t trunc;
4671 
4672 	nfsstat4 status = NFS4_OK;
4673 	cred_t *cr = cs->cr;
4674 	vnode_t *vp = cs->vp;
4675 	struct nfs4_ntov_table ntov;
4676 	struct statvfs64 sb;
4677 	struct vattr bva;
4678 	struct flock64 bf;
4679 	int in_crit = 0;
4680 	uint_t saved_mask = 0;
4681 	caller_context_t ct;
4682 
4683 	*resp = 0;
4684 	sarg.sbp = &sb;
4685 	nfs4_ntov_table_init(&ntov);
4686 	status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov,
4687 	    NFS4ATTR_SETIT);
4688 	if (status != NFS4_OK) {
4689 		/*
4690 		 * failed set attrs
4691 		 */
4692 		goto done;
4693 	}
4694 	if ((sarg.vap->va_mask == 0) &&
4695 	    (! (fattrp->attrmask & FATTR4_ACL_MASK))) {
4696 		/*
4697 		 * no further work to be done
4698 		 */
4699 		goto done;
4700 	}
4701 
4702 	/*
4703 	 * If we got a request to set the ACL and the MODE, only
4704 	 * allow changing VSUID, VSGID, and VSVTX.  Attempting
4705 	 * to change any other bits, along with setting an ACL,
4706 	 * gives NFS4ERR_INVAL.
4707 	 */
4708 	if ((fattrp->attrmask & FATTR4_ACL_MASK) &&
4709 	    (fattrp->attrmask & FATTR4_MODE_MASK)) {
4710 		vattr_t va;
4711 
4712 		va.va_mask = AT_MODE;
4713 		error = VOP_GETATTR(vp, &va, 0, cs->cr);
4714 		if (error) {
4715 			status = puterrno4(error);
4716 			goto done;
4717 		}
4718 		if ((sarg.vap->va_mode ^ va.va_mode) &
4719 		    ~(VSUID | VSGID | VSVTX)) {
4720 			status = NFS4ERR_INVAL;
4721 			goto done;
4722 		}
4723 	}
4724 
4725 	/* Check stateid only if size has been set */
4726 	if (sarg.vap->va_mask & AT_SIZE) {
4727 		trunc = (sarg.vap->va_size == 0);
4728 		status = rfs4_check_stateid(FWRITE, cs->vp, stateid,
4729 		    trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE);
4730 		if (status != NFS4_OK)
4731 			goto done;
4732 	}
4733 
4734 	ct.cc_sysid = 0;
4735 	ct.cc_pid = 0;
4736 	ct.cc_caller_id = nfs4_srv_caller_id;
4737 
4738 	/* XXX start of possible race with delegations */
4739 
4740 	/*
4741 	 * We need to specially handle size changes because it is
4742 	 * possible for the client to create a file with read-only
4743 	 * modes, but with the file opened for writing. If the client
4744 	 * then tries to set the file size, e.g. ftruncate(3C),
4745 	 * fcntl(F_FREESP), the normal access checking done in
4746 	 * VOP_SETATTR would prevent the client from doing it even though
4747 	 * it should be allowed to do so.  To get around this, we do the
4748 	 * access checking for ourselves and use VOP_SPACE which doesn't
4749 	 * do the access checking.
4750 	 * Also the client should not be allowed to change the file
4751 	 * size if there is a conflicting non-blocking mandatory lock in
4752 	 * the region of the change.
4753 	 */
4754 	if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) {
4755 		u_offset_t offset;
4756 		ssize_t length;
4757 
4758 		/*
4759 		 * ufs_setattr clears AT_SIZE from vap->va_mask, but
4760 		 * before returning, sarg.vap->va_mask is used to
4761 		 * generate the setattr reply bitmap.  We also clear
4762 		 * AT_SIZE below before calling VOP_SPACE.  For both
4763 		 * of these cases, the va_mask needs to be saved here
4764 		 * and restored after calling VOP_SETATTR.
4765 		 */
4766 		saved_mask = sarg.vap->va_mask;
4767 
4768 		/*
4769 		 * Check any possible conflict due to NBMAND locks.
4770 		 * Get into critical region before VOP_GETATTR, so the
4771 		 * size attribute is valid when checking conflicts.
4772 		 */
4773 		if (nbl_need_check(vp)) {
4774 			nbl_start_crit(vp, RW_READER);
4775 			in_crit = 1;
4776 		}
4777 
4778 		bva.va_mask = AT_UID|AT_SIZE;
4779 		if (error = VOP_GETATTR(vp, &bva, 0, cr)) {
4780 			status = puterrno4(error);
4781 			goto done;
4782 		}
4783 
4784 		if (in_crit) {
4785 			if (sarg.vap->va_size < bva.va_size) {
4786 				offset = sarg.vap->va_size;
4787 				length = bva.va_size - sarg.vap->va_size;
4788 			} else {
4789 				offset = bva.va_size;
4790 				length = sarg.vap->va_size - bva.va_size;
4791 			}
4792 			if (nbl_conflict(vp, NBL_WRITE, offset, length, 0)) {
4793 				status = NFS4ERR_LOCKED;
4794 				goto done;
4795 			}
4796 		}
4797 
4798 		if (crgetuid(cr) == bva.va_uid) {
4799 			sarg.vap->va_mask &= ~AT_SIZE;
4800 			bf.l_type = F_WRLCK;
4801 			bf.l_whence = 0;
4802 			bf.l_start = (off64_t)sarg.vap->va_size;
4803 			bf.l_len = 0;
4804 			bf.l_sysid = 0;
4805 			bf.l_pid = 0;
4806 			error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE,
4807 			    (offset_t)sarg.vap->va_size, cr, &ct);
4808 		}
4809 	}
4810 
4811 	if (!error && sarg.vap->va_mask != 0)
4812 		error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct);
4813 
4814 	/* restore va_mask -- ufs_setattr clears AT_SIZE */
4815 	if (saved_mask & AT_SIZE)
4816 		sarg.vap->va_mask |= AT_SIZE;
4817 
4818 	/*
4819 	 * If an ACL was being set, it has been delayed until now,
4820 	 * in order to set the mode (via the VOP_SETATTR() above) first.
4821 	 */
4822 	if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) {
4823 		int i;
4824 
4825 		for (i = 0; i < NFS4_MAXNUM_ATTRS; i++)
4826 			if (ntov.amap[i] == FATTR4_ACL)
4827 				break;
4828 		if (i < NFS4_MAXNUM_ATTRS) {
4829 			error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)(
4830 			    NFS4ATTR_SETIT, &sarg, &ntov.na[i]);
4831 			if (error == 0) {
4832 				*resp |= FATTR4_ACL_MASK;
4833 			} else if (error == ENOTSUP) {
4834 				(void) rfs4_verify_attr(&sarg, resp, &ntov);
4835 				status = NFS4ERR_ATTRNOTSUPP;
4836 				goto done;
4837 			}
4838 		} else {
4839 			NFS4_DEBUG(rfs4_debug,
4840 			    (CE_NOTE, "do_rfs4_op_setattr: "
4841 			    "unable to find ACL in fattr4"));
4842 			error = EINVAL;
4843 		}
4844 	}
4845 
4846 	if (error) {
4847 		status = puterrno4(error);
4848 
4849 		/*
4850 		 * Set the response bitmap when setattr failed.
4851 		 * If VOP_SETATTR partially succeeded, test by doing a
4852 		 * VOP_GETATTR on the object and comparing the data
4853 		 * to the setattr arguments.
4854 		 */
4855 		(void) rfs4_verify_attr(&sarg, resp, &ntov);
4856 	} else {
4857 		/*
4858 		 * Force modified metadata out to stable storage.
4859 		 */
4860 		(void) VOP_FSYNC(vp, FNODSYNC, cr);
4861 		/*
4862 		 * Set response bitmap
4863 		 */
4864 		nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp);
4865 	}
4866 
4867 /* Return early and already have a NFSv4 error */
4868 done:
4869 	/*
4870 	 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr
4871 	 * conversion sets both readable and writeable NFS4 attrs
4872 	 * for AT_MTIME and AT_ATIME.  The line below masks out
4873 	 * unrequested attrs from the setattr result bitmap.  This
4874 	 * is placed after the done: label to catch the ATTRNOTSUP
4875 	 * case.
4876 	 */
4877 	*resp &= fattrp->attrmask;
4878 
4879 	if (in_crit)
4880 		nbl_end_crit(vp);
4881 
4882 	nfs4_ntov_table_free(&ntov, &sarg);
4883 
4884 	return (status);
4885 }
4886 
4887 /* ARGSUSED */
4888 static void
4889 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4890 	struct compound_state *cs)
4891 {
4892 	SETATTR4args *args = &argop->nfs_argop4_u.opsetattr;
4893 	SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr;
4894 	bslabel_t *clabel;
4895 
4896 	if (cs->vp == NULL) {
4897 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4898 		return;
4899 	}
4900 
4901 	/*
4902 	 * If there is an unshared filesystem mounted on this vnode,
4903 	 * do not allow to setattr on this vnode.
4904 	 */
4905 	if (vn_ismntpt(cs->vp)) {
4906 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4907 		return;
4908 	}
4909 
4910 	resp->attrsset = 0;
4911 
4912 	if (rdonly4(cs->exi, cs->vp, req)) {
4913 		*cs->statusp = resp->status = NFS4ERR_ROFS;
4914 		return;
4915 	}
4916 
4917 	/* check label before setting attributes */
4918 	if (is_system_labeled()) {
4919 		ASSERT(req->rq_label != NULL);
4920 		clabel = req->rq_label;
4921 		DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *,
4922 		    "got client label from request(1)",
4923 		    struct svc_req *, req);
4924 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
4925 			if (!do_rfs_label_check(clabel, cs->vp,
4926 			    EQUALITY_CHECK)) {
4927 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
4928 				return;
4929 			}
4930 		}
4931 	}
4932 
4933 	*cs->statusp = resp->status =
4934 	    do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs,
4935 	    &args->stateid);
4936 }
4937 
4938 /* ARGSUSED */
4939 static void
4940 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4941 	struct compound_state *cs)
4942 {
4943 	/*
4944 	 * verify and nverify are exactly the same, except that nverify
4945 	 * succeeds when some argument changed, and verify succeeds when
4946 	 * when none changed.
4947 	 */
4948 
4949 	VERIFY4args  *args = &argop->nfs_argop4_u.opverify;
4950 	VERIFY4res *resp = &resop->nfs_resop4_u.opverify;
4951 
4952 	int error;
4953 	struct nfs4_svgetit_arg sarg;
4954 	struct statvfs64 sb;
4955 	struct nfs4_ntov_table ntov;
4956 
4957 	if (cs->vp == NULL) {
4958 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4959 		return;
4960 	}
4961 
4962 	sarg.sbp = &sb;
4963 	nfs4_ntov_table_init(&ntov);
4964 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
4965 	    &sarg, &ntov, NFS4ATTR_VERIT);
4966 	if (resp->status != NFS4_OK) {
4967 		/*
4968 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
4969 		 * so could return -1 for "no match".
4970 		 */
4971 		if (resp->status == -1)
4972 			resp->status = NFS4ERR_NOT_SAME;
4973 		goto done;
4974 	}
4975 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
4976 	switch (error) {
4977 	case 0:
4978 		resp->status = NFS4_OK;
4979 		break;
4980 	case -1:
4981 		resp->status = NFS4ERR_NOT_SAME;
4982 		break;
4983 	default:
4984 		resp->status = puterrno4(error);
4985 		break;
4986 	}
4987 done:
4988 	*cs->statusp = resp->status;
4989 	nfs4_ntov_table_free(&ntov, &sarg);
4990 }
4991 
4992 /* ARGSUSED */
4993 static void
4994 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4995 	struct compound_state *cs)
4996 {
4997 	/*
4998 	 * verify and nverify are exactly the same, except that nverify
4999 	 * succeeds when some argument changed, and verify succeeds when
5000 	 * when none changed.
5001 	 */
5002 
5003 	NVERIFY4args  *args = &argop->nfs_argop4_u.opnverify;
5004 	NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify;
5005 
5006 	int error;
5007 	struct nfs4_svgetit_arg sarg;
5008 	struct statvfs64 sb;
5009 	struct nfs4_ntov_table ntov;
5010 
5011 	if (cs->vp == NULL) {
5012 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5013 		return;
5014 	}
5015 	sarg.sbp = &sb;
5016 	nfs4_ntov_table_init(&ntov);
5017 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
5018 	    &sarg, &ntov, NFS4ATTR_VERIT);
5019 	if (resp->status != NFS4_OK) {
5020 		/*
5021 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
5022 		 * so could return -1 for "no match".
5023 		 */
5024 		if (resp->status == -1)
5025 			resp->status = NFS4_OK;
5026 		goto done;
5027 	}
5028 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
5029 	switch (error) {
5030 	case 0:
5031 		resp->status = NFS4ERR_SAME;
5032 		break;
5033 	case -1:
5034 		resp->status = NFS4_OK;
5035 		break;
5036 	default:
5037 		resp->status = puterrno4(error);
5038 		break;
5039 	}
5040 done:
5041 	*cs->statusp = resp->status;
5042 	nfs4_ntov_table_free(&ntov, &sarg);
5043 }
5044 
5045 /*
5046  * XXX - This should live in an NFS header file.
5047  */
5048 #define	MAX_IOVECS	12
5049 
5050 /* ARGSUSED */
5051 static void
5052 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5053 	struct compound_state *cs)
5054 {
5055 	WRITE4args  *args = &argop->nfs_argop4_u.opwrite;
5056 	WRITE4res *resp = &resop->nfs_resop4_u.opwrite;
5057 	int error;
5058 	vnode_t *vp;
5059 	struct vattr bva;
5060 	u_offset_t rlimit;
5061 	struct uio uio;
5062 	struct iovec iov[MAX_IOVECS];
5063 	struct iovec *iovp;
5064 	int iovcnt;
5065 	int ioflag;
5066 	cred_t *savecred, *cr;
5067 	bool_t *deleg = &cs->deleg;
5068 	nfsstat4 stat;
5069 	int in_crit = 0;
5070 
5071 	vp = cs->vp;
5072 	if (vp == NULL) {
5073 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5074 		return;
5075 	}
5076 	if (cs->access == CS_ACCESS_DENIED) {
5077 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5078 		return;
5079 	}
5080 
5081 	cr = cs->cr;
5082 
5083 	/*
5084 	 * We have to enter the critical region before calling VOP_RWLOCK
5085 	 * to avoid a deadlock with ufs.
5086 	 */
5087 	if (nbl_need_check(vp)) {
5088 		nbl_start_crit(vp, RW_READER);
5089 		in_crit = 1;
5090 		if (nbl_conflict(vp, NBL_WRITE,
5091 		    args->offset, args->data_len, 0)) {
5092 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
5093 			goto out;
5094 		}
5095 	}
5096 
5097 	if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE,
5098 	    deleg, TRUE)) != NFS4_OK) {
5099 		*cs->statusp = resp->status = stat;
5100 		goto out;
5101 	}
5102 
5103 	bva.va_mask = AT_MODE | AT_UID;
5104 	error = VOP_GETATTR(vp, &bva, 0, cr);
5105 
5106 	/*
5107 	 * If we can't get the attributes, then we can't do the
5108 	 * right access checking.  So, we'll fail the request.
5109 	 */
5110 	if (error) {
5111 		*cs->statusp = resp->status = puterrno4(error);
5112 		goto out;
5113 	}
5114 
5115 	if (rdonly4(cs->exi, cs->vp, req)) {
5116 		*cs->statusp = resp->status = NFS4ERR_ROFS;
5117 		goto out;
5118 	}
5119 
5120 	if (vp->v_type != VREG) {
5121 		*cs->statusp = resp->status =
5122 		    ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
5123 		goto out;
5124 	}
5125 
5126 	if (crgetuid(cr) != bva.va_uid &&
5127 	    (error = VOP_ACCESS(vp, VWRITE, 0, cr))) {
5128 		*cs->statusp = resp->status = puterrno4(error);
5129 		goto out;
5130 	}
5131 
5132 	if (MANDLOCK(vp, bva.va_mode)) {
5133 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5134 		goto out;
5135 	}
5136 
5137 	if (args->data_len == 0) {
5138 		*cs->statusp = resp->status = NFS4_OK;
5139 		resp->count = 0;
5140 		resp->committed = args->stable;
5141 		resp->writeverf = Write4verf;
5142 		goto out;
5143 	}
5144 
5145 	if (args->mblk != NULL) {
5146 		mblk_t *m;
5147 		uint_t bytes, round_len;
5148 
5149 		iovcnt = 0;
5150 		bytes = 0;
5151 		round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT);
5152 		for (m = args->mblk;
5153 		    m != NULL && bytes < round_len;
5154 		    m = m->b_cont) {
5155 			iovcnt++;
5156 			bytes += MBLKL(m);
5157 		}
5158 #ifdef DEBUG
5159 		/* should have ended on an mblk boundary */
5160 		if (bytes != round_len) {
5161 			printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n",
5162 			    bytes, round_len, args->data_len);
5163 			printf("args=%p, args->mblk=%p, m=%p", (void *)args,
5164 			    (void *)args->mblk, (void *)m);
5165 			ASSERT(bytes == round_len);
5166 		}
5167 #endif
5168 		if (iovcnt <= MAX_IOVECS) {
5169 			iovp = iov;
5170 		} else {
5171 			iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP);
5172 		}
5173 		mblk_to_iov(args->mblk, iovcnt, iovp);
5174 	} else {
5175 		iovcnt = 1;
5176 		iovp = iov;
5177 		iovp->iov_base = args->data_val;
5178 		iovp->iov_len = args->data_len;
5179 	}
5180 
5181 	uio.uio_iov = iovp;
5182 	uio.uio_iovcnt = iovcnt;
5183 
5184 	uio.uio_segflg = UIO_SYSSPACE;
5185 	uio.uio_extflg = UIO_COPY_DEFAULT;
5186 	uio.uio_loffset = args->offset;
5187 	uio.uio_resid = args->data_len;
5188 	uio.uio_llimit = curproc->p_fsz_ctl;
5189 	rlimit = uio.uio_llimit - args->offset;
5190 	if (rlimit < (u_offset_t)uio.uio_resid)
5191 		uio.uio_resid = (int)rlimit;
5192 
5193 	if (args->stable == UNSTABLE4)
5194 		ioflag = 0;
5195 	else if (args->stable == FILE_SYNC4)
5196 		ioflag = FSYNC;
5197 	else if (args->stable == DATA_SYNC4)
5198 		ioflag = FDSYNC;
5199 	else {
5200 		if (iovp != iov)
5201 			kmem_free(iovp, sizeof (*iovp) * iovcnt);
5202 		*cs->statusp = resp->status = NFS4ERR_INVAL;
5203 		goto out;
5204 	}
5205 
5206 	/*
5207 	 * We're changing creds because VM may fault and we need
5208 	 * the cred of the current thread to be used if quota
5209 	 * checking is enabled.
5210 	 */
5211 	savecred = curthread->t_cred;
5212 	curthread->t_cred = cr;
5213 	error = do_io(FWRITE, vp, &uio, ioflag, cr);
5214 	curthread->t_cred = savecred;
5215 
5216 	if (iovp != iov)
5217 		kmem_free(iovp, sizeof (*iovp) * iovcnt);
5218 
5219 	if (error) {
5220 		*cs->statusp = resp->status = puterrno4(error);
5221 		goto out;
5222 	}
5223 
5224 	*cs->statusp = resp->status = NFS4_OK;
5225 	resp->count = args->data_len - uio.uio_resid;
5226 
5227 	if (ioflag == 0)
5228 		resp->committed = UNSTABLE4;
5229 	else
5230 		resp->committed = FILE_SYNC4;
5231 
5232 	resp->writeverf = Write4verf;
5233 
5234 out:
5235 	if (in_crit)
5236 		nbl_end_crit(vp);
5237 }
5238 
5239 
5240 /* XXX put in a header file */
5241 extern int	sec_svc_getcred(struct svc_req *, cred_t *,  caddr_t *, int *);
5242 
5243 void
5244 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi,
5245 	struct svc_req *req, cred_t *cr)
5246 {
5247 	uint_t i;
5248 	struct compound_state cs;
5249 
5250 	rfs4_init_compound_state(&cs);
5251 	/*
5252 	 * Form a reply tag by copying over the reqeuest tag.
5253 	 */
5254 	resp->tag.utf8string_val =
5255 	    kmem_alloc(args->tag.utf8string_len, KM_SLEEP);
5256 	resp->tag.utf8string_len = args->tag.utf8string_len;
5257 	bcopy(args->tag.utf8string_val, resp->tag.utf8string_val,
5258 	    resp->tag.utf8string_len);
5259 
5260 	cs.statusp = &resp->status;
5261 
5262 	/*
5263 	 * XXX for now, minorversion should be zero
5264 	 */
5265 	if (args->minorversion != NFS4_MINORVERSION) {
5266 		resp->array_len = 0;
5267 		resp->array = NULL;
5268 		resp->status = NFS4ERR_MINOR_VERS_MISMATCH;
5269 		return;
5270 	}
5271 
5272 	resp->array_len = args->array_len;
5273 	resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4),
5274 	    KM_SLEEP);
5275 
5276 	ASSERT(exi == NULL);
5277 	ASSERT(cr == NULL);
5278 
5279 	cr = crget();
5280 	ASSERT(cr != NULL);
5281 
5282 	if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) {
5283 		crfree(cr);
5284 		return;
5285 	}
5286 
5287 	cs.basecr = cr;
5288 
5289 	cs.req = req;
5290 
5291 	/*
5292 	 * For now, NFS4 compound processing must be protected by
5293 	 * exported_lock because it can access more than one exportinfo
5294 	 * per compound and share/unshare can now change multiple
5295 	 * exinfo structs.  The NFS2/3 code only refs 1 exportinfo
5296 	 * per proc (excluding public exinfo), and exi_count design
5297 	 * is sufficient to protect concurrent execution of NFS2/3
5298 	 * ops along with unexport.  This lock will be removed as
5299 	 * part of the NFSv4 phase 2 namespace redesign work.
5300 	 */
5301 	rw_enter(&exported_lock, RW_READER);
5302 
5303 	/*
5304 	 * If this is the first compound we've seen, we need to start all
5305 	 * new instances' grace periods.
5306 	 */
5307 	if (rfs4_seen_first_compound == 0) {
5308 		rfs4_grace_start_new();
5309 		/*
5310 		 * This must be set after rfs4_grace_start_new(), otherwise
5311 		 * another thread could proceed past here before the former
5312 		 * is finished.
5313 		 */
5314 		rfs4_seen_first_compound = 1;
5315 	}
5316 
5317 	for (i = 0; i < args->array_len && cs.cont; i++) {
5318 		nfs_argop4 *argop;
5319 		nfs_resop4 *resop;
5320 		uint_t op;
5321 
5322 		argop = &args->array[i];
5323 		resop = &resp->array[i];
5324 		resop->resop = argop->argop;
5325 		op = (uint_t)resop->resop;
5326 
5327 		if (op < rfsv4disp_cnt) {
5328 			/*
5329 			 * Count the individual ops here; NULL and COMPOUND
5330 			 * are counted in common_dispatch()
5331 			 */
5332 			rfsproccnt_v4_ptr[op].value.ui64++;
5333 
5334 			NFS4_DEBUG(rfs4_debug > 1,
5335 			    (CE_NOTE, "Executing %s", rfs4_op_string[op]));
5336 			(*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs);
5337 			NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d",
5338 			    rfs4_op_string[op], *cs.statusp));
5339 			if (*cs.statusp != NFS4_OK)
5340 				cs.cont = FALSE;
5341 		} else {
5342 			/*
5343 			 * This is effectively dead code since XDR code
5344 			 * will have already returned BADXDR if op doesn't
5345 			 * decode to legal value.  This only done for a
5346 			 * day when XDR code doesn't verify v4 opcodes.
5347 			 */
5348 			op = OP_ILLEGAL;
5349 			rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++;
5350 
5351 			rfs4_op_illegal(argop, resop, req, &cs);
5352 			cs.cont = FALSE;
5353 		}
5354 
5355 		/*
5356 		 * If not at last op, and if we are to stop, then
5357 		 * compact the results array.
5358 		 */
5359 		if ((i + 1) < args->array_len && !cs.cont) {
5360 			nfs_resop4 *new_res = kmem_alloc(
5361 			    (i+1) * sizeof (nfs_resop4), KM_SLEEP);
5362 			bcopy(resp->array,
5363 			    new_res, (i+1) * sizeof (nfs_resop4));
5364 			kmem_free(resp->array,
5365 			    args->array_len * sizeof (nfs_resop4));
5366 
5367 			resp->array_len =  i + 1;
5368 			resp->array = new_res;
5369 		}
5370 	}
5371 
5372 	rw_exit(&exported_lock);
5373 
5374 	if (cs.vp)
5375 		VN_RELE(cs.vp);
5376 	if (cs.saved_vp)
5377 		VN_RELE(cs.saved_vp);
5378 	if (cs.saved_fh.nfs_fh4_val)
5379 		kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE);
5380 
5381 	if (cs.basecr)
5382 		crfree(cs.basecr);
5383 	if (cs.cr)
5384 		crfree(cs.cr);
5385 	/*
5386 	 * done with this compound request, free the label
5387 	 */
5388 
5389 	if (req->rq_label != NULL) {
5390 		kmem_free(req->rq_label, sizeof (bslabel_t));
5391 		req->rq_label = NULL;
5392 	}
5393 }
5394 
5395 /*
5396  * XXX because of what appears to be duplicate calls to rfs4_compound_free
5397  * XXX zero out the tag and array values. Need to investigate why the
5398  * XXX calls occur, but at least prevent the panic for now.
5399  */
5400 void
5401 rfs4_compound_free(COMPOUND4res *resp)
5402 {
5403 	uint_t i;
5404 
5405 	if (resp->tag.utf8string_val) {
5406 		UTF8STRING_FREE(resp->tag)
5407 	}
5408 
5409 	for (i = 0; i < resp->array_len; i++) {
5410 		nfs_resop4 *resop;
5411 		uint_t op;
5412 
5413 		resop = &resp->array[i];
5414 		op = (uint_t)resop->resop;
5415 		if (op < rfsv4disp_cnt) {
5416 			(*rfsv4disptab[op].dis_resfree)(resop);
5417 		}
5418 	}
5419 	if (resp->array != NULL) {
5420 		kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4));
5421 	}
5422 }
5423 
5424 /*
5425  * Process the value of the compound request rpc flags, as a bit-AND
5426  * of the individual per-op flags (idempotent, allowork, publicfh_ok)
5427  */
5428 void
5429 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp)
5430 {
5431 	int i;
5432 	int flag = RPC_ALL;
5433 
5434 	for (i = 0; flag && i < args->array_len; i++) {
5435 		uint_t op;
5436 
5437 		op = (uint_t)args->array[i].argop;
5438 
5439 		if (op < rfsv4disp_cnt)
5440 			flag &= rfsv4disptab[op].dis_flags;
5441 		else
5442 			flag = 0;
5443 	}
5444 	*flagp = flag;
5445 }
5446 
5447 nfsstat4
5448 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp)
5449 {
5450 	nfsstat4 e;
5451 
5452 	rfs4_dbe_lock(cp->dbe);
5453 
5454 	if (cp->sysidt != LM_NOSYSID) {
5455 		*sp = cp->sysidt;
5456 		e = NFS4_OK;
5457 
5458 	} else if ((cp->sysidt = lm_alloc_sysidt()) != LM_NOSYSID) {
5459 		*sp = cp->sysidt;
5460 		e = NFS4_OK;
5461 
5462 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
5463 		    "rfs4_client_sysid: allocated 0x%x\n", *sp));
5464 	} else
5465 		e = NFS4ERR_DELAY;
5466 
5467 	rfs4_dbe_unlock(cp->dbe);
5468 	return (e);
5469 }
5470 
5471 #if defined(DEBUG) && ! defined(lint)
5472 static void lock_print(char *str, int operation, struct flock64 *flk)
5473 {
5474 	char *op, *type;
5475 
5476 	switch (operation) {
5477 	case F_GETLK: op = "F_GETLK";
5478 		break;
5479 	case F_SETLK: op = "F_SETLK";
5480 		break;
5481 	default: op = "F_UNKNOWN";
5482 		break;
5483 	}
5484 	switch (flk->l_type) {
5485 	case F_UNLCK: type = "F_UNLCK";
5486 		break;
5487 	case F_RDLCK: type = "F_RDLCK";
5488 		break;
5489 	case F_WRLCK: type = "F_WRLCK";
5490 		break;
5491 	default: type = "F_UNKNOWN";
5492 		break;
5493 	}
5494 
5495 	ASSERT(flk->l_whence == 0);
5496 	cmn_err(CE_NOTE, "%s:  %s, type = %s, off = %llx len = %llx pid = %d",
5497 	    str, op, type, (longlong_t)flk->l_start,
5498 	    flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid);
5499 }
5500 
5501 #define	LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f)
5502 #else
5503 #define	LOCK_PRINT(d, s, t, f)
5504 #endif
5505 
5506 /*ARGSUSED*/
5507 static bool_t
5508 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs)
5509 {
5510 	return (TRUE);
5511 }
5512 
5513 /*
5514  * Look up the pathname using the vp in cs as the directory vnode.
5515  * cs->vp will be the vnode for the file on success
5516  */
5517 
5518 static nfsstat4
5519 rfs4_lookup(component4 *component, struct svc_req *req,
5520 	    struct compound_state *cs)
5521 {
5522 	char *nm;
5523 	uint32_t len;
5524 	nfsstat4 status;
5525 
5526 	if (cs->vp == NULL) {
5527 		return (NFS4ERR_NOFILEHANDLE);
5528 	}
5529 	if (cs->vp->v_type != VDIR) {
5530 		return (NFS4ERR_NOTDIR);
5531 	}
5532 
5533 	if (!utf8_dir_verify(component))
5534 		return (NFS4ERR_INVAL);
5535 
5536 	nm = utf8_to_fn(component, &len, NULL);
5537 	if (nm == NULL) {
5538 		return (NFS4ERR_INVAL);
5539 	}
5540 
5541 	if (len > MAXNAMELEN) {
5542 		kmem_free(nm, len);
5543 		return (NFS4ERR_NAMETOOLONG);
5544 	}
5545 
5546 	status = do_rfs4_op_lookup(nm, len, req, cs);
5547 
5548 	kmem_free(nm, len);
5549 
5550 	return (status);
5551 }
5552 
5553 static nfsstat4
5554 rfs4_lookupfile(component4 *component, struct svc_req *req,
5555 		struct compound_state *cs, uint32_t access,
5556 		change_info4 *cinfo)
5557 {
5558 	nfsstat4 status;
5559 	vnode_t *dvp = cs->vp;
5560 	vattr_t bva, ava, fva;
5561 	int error;
5562 
5563 	/* Get "before" change value */
5564 	bva.va_mask = AT_CTIME|AT_SEQ;
5565 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr);
5566 	if (error)
5567 		return (puterrno4(error));
5568 
5569 	/* rfs4_lookup may VN_RELE directory */
5570 	VN_HOLD(dvp);
5571 
5572 	status = rfs4_lookup(component, req, cs);
5573 	if (status != NFS4_OK) {
5574 		VN_RELE(dvp);
5575 		return (status);
5576 	}
5577 
5578 	/*
5579 	 * Get "after" change value, if it fails, simply return the
5580 	 * before value.
5581 	 */
5582 	ava.va_mask = AT_CTIME|AT_SEQ;
5583 	if (VOP_GETATTR(dvp, &ava, 0, cs->cr)) {
5584 		ava.va_ctime = bva.va_ctime;
5585 		ava.va_seq = 0;
5586 	}
5587 	VN_RELE(dvp);
5588 
5589 	/*
5590 	 * Validate the file is a file
5591 	 */
5592 	fva.va_mask = AT_TYPE|AT_MODE;
5593 	error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr);
5594 	if (error)
5595 		return (puterrno4(error));
5596 
5597 	if (fva.va_type != VREG) {
5598 		if (fva.va_type == VDIR)
5599 			return (NFS4ERR_ISDIR);
5600 		if (fva.va_type == VLNK)
5601 			return (NFS4ERR_SYMLINK);
5602 		return (NFS4ERR_INVAL);
5603 	}
5604 
5605 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime);
5606 	NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
5607 
5608 	/*
5609 	 * It is undefined if VOP_LOOKUP will change va_seq, so
5610 	 * cinfo.atomic = TRUE only if we have
5611 	 * non-zero va_seq's, and they have not changed.
5612 	 */
5613 	if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq)
5614 		cinfo->atomic = TRUE;
5615 	else
5616 		cinfo->atomic = FALSE;
5617 
5618 	/* Check for mandatory locking */
5619 	cs->mandlock = MANDLOCK(cs->vp, fva.va_mode);
5620 	return (check_open_access(access, cs, req));
5621 }
5622 
5623 static nfsstat4
5624 create_vnode(vnode_t *dvp, char *nm,  vattr_t *vap, createmode4 mode,
5625 	    timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created)
5626 {
5627 	int error;
5628 	nfsstat4 status = NFS4_OK;
5629 	vattr_t va;
5630 
5631 tryagain:
5632 
5633 	/*
5634 	 * The file open mode used is VWRITE.  If the client needs
5635 	 * some other semantic, then it should do the access checking
5636 	 * itself.  It would have been nice to have the file open mode
5637 	 * passed as part of the arguments.
5638 	 */
5639 
5640 	*created = TRUE;
5641 	error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0);
5642 
5643 	if (error) {
5644 		*created = FALSE;
5645 
5646 		/*
5647 		 * If we got something other than file already exists
5648 		 * then just return this error.  Otherwise, we got
5649 		 * EEXIST.  If we were doing a GUARDED create, then
5650 		 * just return this error.  Otherwise, we need to
5651 		 * make sure that this wasn't a duplicate of an
5652 		 * exclusive create request.
5653 		 *
5654 		 * The assumption is made that a non-exclusive create
5655 		 * request will never return EEXIST.
5656 		 */
5657 
5658 		if (error != EEXIST || mode == GUARDED4) {
5659 			status = puterrno4(error);
5660 			return (status);
5661 		}
5662 		error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr);
5663 
5664 		if (error) {
5665 			/*
5666 			 * We couldn't find the file that we thought that
5667 			 * we just created.  So, we'll just try creating
5668 			 * it again.
5669 			 */
5670 			if (error == ENOENT)
5671 				goto tryagain;
5672 
5673 			status = puterrno4(error);
5674 			return (status);
5675 		}
5676 
5677 		if (mode == UNCHECKED4) {
5678 			/* existing object must be regular file */
5679 			if ((*vpp)->v_type != VREG) {
5680 				if ((*vpp)->v_type == VDIR)
5681 					status = NFS4ERR_ISDIR;
5682 				else if ((*vpp)->v_type == VLNK)
5683 					status = NFS4ERR_SYMLINK;
5684 				else
5685 					status = NFS4ERR_INVAL;
5686 				VN_RELE(*vpp);
5687 				return (status);
5688 			}
5689 
5690 			return (NFS4_OK);
5691 		}
5692 
5693 		/* Check for duplicate request */
5694 		ASSERT(mtime != 0);
5695 		va.va_mask = AT_MTIME;
5696 		error = VOP_GETATTR(*vpp, &va, 0, cr);
5697 		if (!error) {
5698 			/* We found the file */
5699 			if (va.va_mtime.tv_sec != mtime->tv_sec ||
5700 			    va.va_mtime.tv_nsec != mtime->tv_nsec) {
5701 				/* but its not our creation */
5702 				VN_RELE(*vpp);
5703 				return (NFS4ERR_EXIST);
5704 			}
5705 			*created = TRUE; /* retrans of create == created */
5706 			return (NFS4_OK);
5707 		}
5708 		VN_RELE(*vpp);
5709 		return (NFS4ERR_EXIST);
5710 	}
5711 
5712 	return (NFS4_OK);
5713 }
5714 
5715 static nfsstat4
5716 check_open_access(uint32_t access,
5717 		struct compound_state *cs, struct svc_req *req)
5718 {
5719 	int error;
5720 	vnode_t *vp;
5721 	bool_t readonly;
5722 	cred_t *cr = cs->cr;
5723 
5724 	/* For now we don't allow mandatory locking as per V2/V3 */
5725 	if (cs->access == CS_ACCESS_DENIED || cs->mandlock) {
5726 		return (NFS4ERR_ACCESS);
5727 	}
5728 
5729 	vp = cs->vp;
5730 	ASSERT(cr != NULL && vp->v_type == VREG);
5731 
5732 	/*
5733 	 * If the file system is exported read only and we are trying
5734 	 * to open for write, then return NFS4ERR_ROFS
5735 	 */
5736 
5737 	readonly = rdonly4(cs->exi, cs->vp, req);
5738 
5739 	if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly)
5740 		return (NFS4ERR_ROFS);
5741 
5742 	if (access & OPEN4_SHARE_ACCESS_READ) {
5743 		if ((VOP_ACCESS(vp, VREAD, 0, cr) != 0) &&
5744 		    (VOP_ACCESS(vp, VEXEC, 0, cr) != 0)) {
5745 			return (NFS4ERR_ACCESS);
5746 		}
5747 	}
5748 
5749 	if (access & OPEN4_SHARE_ACCESS_WRITE) {
5750 		error = VOP_ACCESS(vp, VWRITE, 0, cr);
5751 		if (error)
5752 			return (NFS4ERR_ACCESS);
5753 	}
5754 
5755 	return (NFS4_OK);
5756 }
5757 
5758 static nfsstat4
5759 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs,
5760 		change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid)
5761 {
5762 	struct nfs4_svgetit_arg sarg;
5763 	struct nfs4_ntov_table ntov;
5764 
5765 	bool_t ntov_table_init = FALSE;
5766 	struct statvfs64 sb;
5767 	nfsstat4 status;
5768 	vnode_t *vp;
5769 	vattr_t bva, ava, iva, cva, *vap;
5770 	vnode_t *dvp;
5771 	timespec32_t *mtime;
5772 	char *nm = NULL;
5773 	uint_t buflen;
5774 	bool_t created;
5775 	bool_t setsize = FALSE;
5776 	len_t reqsize;
5777 	int error;
5778 	bool_t trunc;
5779 	caller_context_t ct;
5780 	component4 *component;
5781 	bslabel_t *clabel;
5782 
5783 	sarg.sbp = &sb;
5784 
5785 	dvp = cs->vp;
5786 
5787 	/* Check if the file system is read only */
5788 	if (rdonly4(cs->exi, dvp, req))
5789 		return (NFS4ERR_ROFS);
5790 
5791 	/* check the label of including directory */
5792 	if (is_system_labeled()) {
5793 		ASSERT(req->rq_label != NULL);
5794 		clabel = req->rq_label;
5795 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
5796 		    "got client label from request(1)",
5797 		    struct svc_req *, req);
5798 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
5799 			if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK)) {
5800 				return (NFS4ERR_ACCESS);
5801 			}
5802 		}
5803 	}
5804 
5805 	/*
5806 	 * Get the last component of path name in nm. cs will reference
5807 	 * the including directory on success.
5808 	 */
5809 	component = &args->open_claim4_u.file;
5810 	if (!utf8_dir_verify(component))
5811 		return (NFS4ERR_INVAL);
5812 
5813 	nm = utf8_to_fn(component, &buflen, NULL);
5814 
5815 	if (nm == NULL)
5816 		return (NFS4ERR_RESOURCE);
5817 
5818 	if (buflen > MAXNAMELEN) {
5819 		kmem_free(nm, buflen);
5820 		return (NFS4ERR_NAMETOOLONG);
5821 	}
5822 
5823 	bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ;
5824 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr);
5825 	if (error) {
5826 		kmem_free(nm, buflen);
5827 		return (puterrno4(error));
5828 	}
5829 
5830 	if (bva.va_type != VDIR) {
5831 		kmem_free(nm, buflen);
5832 		return (NFS4ERR_NOTDIR);
5833 	}
5834 
5835 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime)
5836 
5837 	switch (args->mode) {
5838 	case GUARDED4:
5839 		/*FALLTHROUGH*/
5840 	case UNCHECKED4:
5841 		nfs4_ntov_table_init(&ntov);
5842 		ntov_table_init = TRUE;
5843 
5844 		*attrset = 0;
5845 		status = do_rfs4_set_attrs(attrset,
5846 		    &args->createhow4_u.createattrs,
5847 		    cs, &sarg, &ntov, NFS4ATTR_SETIT);
5848 
5849 		if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) &&
5850 		    sarg.vap->va_type != VREG) {
5851 			if (sarg.vap->va_type == VDIR)
5852 				status = NFS4ERR_ISDIR;
5853 			else if (sarg.vap->va_type == VLNK)
5854 				status = NFS4ERR_SYMLINK;
5855 			else
5856 				status = NFS4ERR_INVAL;
5857 		}
5858 
5859 		if (status != NFS4_OK) {
5860 			kmem_free(nm, buflen);
5861 			nfs4_ntov_table_free(&ntov, &sarg);
5862 			*attrset = 0;
5863 			return (status);
5864 		}
5865 
5866 		vap = sarg.vap;
5867 		vap->va_type = VREG;
5868 		vap->va_mask |= AT_TYPE;
5869 
5870 		if ((vap->va_mask & AT_MODE) == 0) {
5871 			vap->va_mask |= AT_MODE;
5872 			vap->va_mode = (mode_t)0600;
5873 		}
5874 
5875 		if (vap->va_mask & AT_SIZE) {
5876 
5877 			/* Disallow create with a non-zero size */
5878 
5879 			if ((reqsize = sarg.vap->va_size) != 0) {
5880 				kmem_free(nm, buflen);
5881 				nfs4_ntov_table_free(&ntov, &sarg);
5882 				*attrset = 0;
5883 				return (NFS4ERR_INVAL);
5884 			}
5885 			setsize = TRUE;
5886 		}
5887 		break;
5888 
5889 	case EXCLUSIVE4:
5890 		/* prohibit EXCL create of named attributes */
5891 		if (dvp->v_flag & V_XATTRDIR) {
5892 			kmem_free(nm, buflen);
5893 			*attrset = 0;
5894 			return (NFS4ERR_INVAL);
5895 		}
5896 
5897 		cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE;
5898 		cva.va_type = VREG;
5899 		/*
5900 		 * Ensure no time overflows. Assumes underlying
5901 		 * filesystem supports at least 32 bits.
5902 		 * Truncate nsec to usec resolution to allow valid
5903 		 * compares even if the underlying filesystem truncates.
5904 		 */
5905 		mtime = (timespec32_t *)&args->createhow4_u.createverf;
5906 		cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX;
5907 		cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000;
5908 		cva.va_mode = (mode_t)0;
5909 		vap = &cva;
5910 		break;
5911 	}
5912 
5913 	status = create_vnode(dvp, nm, vap, args->mode, mtime,
5914 	    cs->cr, &vp, &created);
5915 	kmem_free(nm, buflen);
5916 
5917 	if (status != NFS4_OK) {
5918 		if (ntov_table_init)
5919 			nfs4_ntov_table_free(&ntov, &sarg);
5920 		*attrset = 0;
5921 		return (status);
5922 	}
5923 
5924 	trunc = (setsize && !created);
5925 
5926 	if (args->mode != EXCLUSIVE4) {
5927 		bitmap4 createmask = args->createhow4_u.createattrs.attrmask;
5928 
5929 		/*
5930 		 * True verification that object was created with correct
5931 		 * attrs is impossible.  The attrs could have been changed
5932 		 * immediately after object creation.  If attributes did
5933 		 * not verify, the only recourse for the server is to
5934 		 * destroy the object.  Maybe if some attrs (like gid)
5935 		 * are set incorrectly, the object should be destroyed;
5936 		 * however, seems bad as a default policy.  Do we really
5937 		 * want to destroy an object over one of the times not
5938 		 * verifying correctly?  For these reasons, the server
5939 		 * currently sets bits in attrset for createattrs
5940 		 * that were set; however, no verification is done.
5941 		 *
5942 		 * vmask_to_nmask accounts for vattr bits set on create
5943 		 *	[do_rfs4_set_attrs() only sets resp bits for
5944 		 *	 non-vattr/vfs bits.]
5945 		 * Mask off any bits we set by default so as not to return
5946 		 * more attrset bits than were requested in createattrs
5947 		 */
5948 		if (created) {
5949 			nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset);
5950 			*attrset &= createmask;
5951 		} else {
5952 			/*
5953 			 * We did not create the vnode (we tried but it
5954 			 * already existed).  In this case, the only createattr
5955 			 * that the spec allows the server to set is size,
5956 			 * and even then, it can only be set if it is 0.
5957 			 */
5958 			*attrset = 0;
5959 			if (trunc)
5960 				*attrset = FATTR4_SIZE_MASK;
5961 		}
5962 	}
5963 	if (ntov_table_init)
5964 		nfs4_ntov_table_free(&ntov, &sarg);
5965 
5966 	/*
5967 	 * Get the initial "after" sequence number, if it fails,
5968 	 * set to zero, time to before.
5969 	 */
5970 	iva.va_mask = AT_CTIME|AT_SEQ;
5971 	if (VOP_GETATTR(dvp, &iva, 0, cs->cr)) {
5972 		iva.va_seq = 0;
5973 		iva.va_ctime = bva.va_ctime;
5974 	}
5975 
5976 	/*
5977 	 * create_vnode attempts to create the file exclusive,
5978 	 * if it already exists the VOP_CREATE will fail and
5979 	 * may not increase va_seq. It is atomic if
5980 	 * we haven't changed the directory, but if it has changed
5981 	 * we don't know what changed it.
5982 	 */
5983 	if (!created) {
5984 		if (bva.va_seq && iva.va_seq &&
5985 		    bva.va_seq == iva.va_seq)
5986 			cinfo->atomic = TRUE;
5987 		else
5988 			cinfo->atomic = FALSE;
5989 		NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime);
5990 	} else {
5991 		/*
5992 		 * The entry was created, we need to sync the
5993 		 * directory metadata.
5994 		 */
5995 		(void) VOP_FSYNC(dvp, 0, cs->cr);
5996 
5997 		/*
5998 		 * Get "after" change value, if it fails, simply return the
5999 		 * before value.
6000 		 */
6001 		ava.va_mask = AT_CTIME|AT_SEQ;
6002 		if (VOP_GETATTR(dvp, &ava, 0, cs->cr)) {
6003 			ava.va_ctime = bva.va_ctime;
6004 			ava.va_seq = 0;
6005 		}
6006 
6007 		NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
6008 
6009 		/*
6010 		 * The cinfo->atomic = TRUE only if we have
6011 		 * non-zero va_seq's, and it has incremented by exactly one
6012 		 * during the create_vnode and it didn't
6013 		 * change during the VOP_FSYNC.
6014 		 */
6015 		if (bva.va_seq && iva.va_seq && ava.va_seq &&
6016 		    iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq)
6017 			cinfo->atomic = TRUE;
6018 		else
6019 			cinfo->atomic = FALSE;
6020 	}
6021 
6022 	/* Check for mandatory locking and that the size gets set. */
6023 	cva.va_mask = AT_MODE;
6024 	if (setsize)
6025 		cva.va_mask |= AT_SIZE;
6026 
6027 	/* Assume the worst */
6028 	cs->mandlock = TRUE;
6029 
6030 	if (VOP_GETATTR(vp, &cva, 0, cs->cr) == 0) {
6031 		cs->mandlock = MANDLOCK(cs->vp, cva.va_mode);
6032 
6033 		/*
6034 		 * Truncate the file if necessary; this would be
6035 		 * the case for create over an existing file.
6036 		 */
6037 
6038 		if (trunc) {
6039 			int in_crit = 0;
6040 			rfs4_file_t *fp;
6041 			bool_t create = FALSE;
6042 
6043 			/*
6044 			 * We are writing over an existing file.
6045 			 * Check to see if we need to recall a delegation.
6046 			 */
6047 			rfs4_hold_deleg_policy();
6048 			if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) {
6049 				if (rfs4_check_delegated_byfp(FWRITE, fp,
6050 				    (reqsize == 0), FALSE, FALSE, &clientid)) {
6051 					rfs4_file_rele(fp);
6052 					rfs4_rele_deleg_policy();
6053 					VN_RELE(vp);
6054 					*attrset = 0;
6055 					return (NFS4ERR_DELAY);
6056 				}
6057 				rfs4_file_rele(fp);
6058 			}
6059 			rfs4_rele_deleg_policy();
6060 
6061 			if (nbl_need_check(vp)) {
6062 				in_crit = 1;
6063 
6064 				ASSERT(reqsize == 0);
6065 
6066 				nbl_start_crit(vp, RW_READER);
6067 				if (nbl_conflict(vp, NBL_WRITE, 0,
6068 				    cva.va_size, 0)) {
6069 					in_crit = 0;
6070 					nbl_end_crit(vp);
6071 					VN_RELE(vp);
6072 					*attrset = 0;
6073 					return (NFS4ERR_ACCESS);
6074 				}
6075 			}
6076 			ct.cc_sysid = 0;
6077 			ct.cc_pid = 0;
6078 			ct.cc_caller_id = nfs4_srv_caller_id;
6079 
6080 			cva.va_mask = AT_SIZE;
6081 			cva.va_size = reqsize;
6082 			(void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct);
6083 			if (in_crit)
6084 				nbl_end_crit(vp);
6085 		}
6086 	}
6087 
6088 	error = makefh4(&cs->fh, vp, cs->exi);
6089 
6090 	/*
6091 	 * Force modified data and metadata out to stable storage.
6092 	 */
6093 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr);
6094 
6095 	if (error) {
6096 		VN_RELE(vp);
6097 		*attrset = 0;
6098 		return (puterrno4(error));
6099 	}
6100 
6101 	/* if parent dir is attrdir, set namedattr fh flag */
6102 	if (dvp->v_flag & V_XATTRDIR)
6103 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
6104 
6105 	if (cs->vp)
6106 		VN_RELE(cs->vp);
6107 
6108 	cs->vp = vp;
6109 
6110 	/*
6111 	 * if we did not create the file, we will need to check
6112 	 * the access bits on the file
6113 	 */
6114 
6115 	if (!created) {
6116 		if (setsize)
6117 			args->share_access |= OPEN4_SHARE_ACCESS_WRITE;
6118 		status = check_open_access(args->share_access, cs, req);
6119 		if (status != NFS4_OK)
6120 			*attrset = 0;
6121 	}
6122 	return (status);
6123 }
6124 
6125 /*ARGSUSED*/
6126 static void
6127 rfs4_do_open(struct compound_state *cs, struct svc_req *req,
6128 		rfs4_openowner_t *oo, delegreq_t deleg,
6129 		uint32_t access, uint32_t deny,
6130 		OPEN4res *resp)
6131 {
6132 	/* XXX Currently not using req  */
6133 	rfs4_state_t *state;
6134 	rfs4_file_t *file;
6135 	bool_t screate = TRUE;
6136 	bool_t fcreate = TRUE;
6137 	uint32_t amodes;
6138 	uint32_t dmodes;
6139 	rfs4_deleg_state_t *dsp;
6140 	struct shrlock shr;
6141 	struct shr_locowner shr_loco;
6142 	sysid_t sysid;
6143 	nfsstat4 status;
6144 	int fflags = 0;
6145 	int recall = 0;
6146 	int err;
6147 
6148 	/* get the file struct and hold a lock on it during initial open */
6149 	file = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate);
6150 	if (file == NULL) {
6151 		NFS4_DEBUG(rfs4_debug,
6152 		    (CE_NOTE, "rfs4_do_open: can't find file"));
6153 		resp->status = NFS4ERR_SERVERFAULT;
6154 		return;
6155 	}
6156 
6157 	state = rfs4_findstate_by_owner_file(oo, file, &screate);
6158 	if (state == NULL) {
6159 		NFS4_DEBUG(rfs4_debug,
6160 		    (CE_NOTE, "rfs4_do_open: can't find state"));
6161 		resp->status = NFS4ERR_RESOURCE;
6162 		/* No need to keep any reference */
6163 		rfs4_file_rele_withunlock(file);
6164 		return;
6165 	}
6166 
6167 	/* try to get the sysid before continuing */
6168 	if ((status = rfs4_client_sysid(oo->client, &sysid)) != NFS4_OK) {
6169 		resp->status = status;
6170 		rfs4_file_rele(file);
6171 		/* Not a fully formed open; "close" it */
6172 		if (screate == TRUE)
6173 			rfs4_state_close(state, FALSE, FALSE, cs->cr);
6174 		rfs4_state_rele(state);
6175 		return;
6176 	}
6177 
6178 	/*
6179 	 * Calculate the new deny and access mode that this open is adding to
6180 	 * the file for this open owner;
6181 	 */
6182 	dmodes = (deny & ~state->share_deny);
6183 	amodes = (access & ~state->share_access);
6184 
6185 	/*
6186 	 * Check to see the client has already sent an open for this
6187 	 * open owner on this file with the same share/deny modes.
6188 	 * If so, we don't need to check for a conflict and we don't
6189 	 * need to add another shrlock.  If not, then we need to
6190 	 * check for conflicts in deny and access before checking for
6191 	 * conflicts in delegation.  We don't want to recall a
6192 	 * delegation based on an open that will eventually fail based
6193 	 * on shares modes.
6194 	 */
6195 
6196 	if (dmodes || amodes) {
6197 		shr.s_access = (short)access;
6198 		shr.s_deny = (short)deny;
6199 		shr.s_pid = rfs4_dbe_getid(oo->dbe);
6200 		shr.s_sysid = sysid;
6201 		shr_loco.sl_pid = shr.s_pid;
6202 		shr_loco.sl_id = shr.s_sysid;
6203 		shr.s_owner = (caddr_t)&shr_loco;
6204 		shr.s_own_len = sizeof (shr_loco);
6205 
6206 		fflags = 0;
6207 		if (access & OPEN4_SHARE_ACCESS_READ)
6208 			fflags |= FREAD;
6209 		if (access & OPEN4_SHARE_ACCESS_WRITE)
6210 			fflags |= FWRITE;
6211 
6212 		if ((err = vop_shrlock(cs->vp, F_SHARE, &shr, fflags)) != 0) {
6213 
6214 			resp->status = err == EAGAIN ?
6215 			    NFS4ERR_SHARE_DENIED : puterrno4(err);
6216 
6217 			rfs4_file_rele(file);
6218 			/* Not a fully formed open; "close" it */
6219 			if (screate == TRUE)
6220 				rfs4_state_close(state, FALSE, FALSE, cs->cr);
6221 			rfs4_state_rele(state);
6222 			return;
6223 		}
6224 	}
6225 
6226 	rfs4_dbe_lock(state->dbe);
6227 	rfs4_dbe_lock(file->dbe);
6228 
6229 	/*
6230 	 * Check to see if this file is delegated and if so, if a
6231 	 * recall needs to be done.
6232 	 */
6233 	if (rfs4_check_recall(state, access)) {
6234 		rfs4_dbe_unlock(file->dbe);
6235 		rfs4_dbe_unlock(state->dbe);
6236 		rfs4_recall_deleg(file, FALSE, state->owner->client);
6237 		delay(NFS4_DELEGATION_CONFLICT_DELAY);
6238 		rfs4_dbe_lock(state->dbe);
6239 		rfs4_dbe_lock(file->dbe);
6240 		/* Let's see if the delegation was returned */
6241 		if (rfs4_check_recall(state, access)) {
6242 			rfs4_dbe_unlock(file->dbe);
6243 			rfs4_dbe_unlock(state->dbe);
6244 			rfs4_file_rele(file);
6245 			rfs4_update_lease(state->owner->client);
6246 			/* recalculate flags to match what was added */
6247 			fflags = 0;
6248 			if (amodes & OPEN4_SHARE_ACCESS_READ)
6249 				fflags |= FREAD;
6250 			if (amodes & OPEN4_SHARE_ACCESS_WRITE)
6251 				fflags |= FWRITE;
6252 			(void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags);
6253 			/* Not a fully formed open; "close" it */
6254 			if (screate == TRUE)
6255 				rfs4_state_close(state, FALSE, FALSE, cs->cr);
6256 			rfs4_state_rele(state);
6257 			resp->status = NFS4ERR_DELAY;
6258 			return;
6259 		}
6260 	}
6261 
6262 	if (dmodes & OPEN4_SHARE_DENY_READ)
6263 		file->deny_read++;
6264 	if (dmodes & OPEN4_SHARE_DENY_WRITE)
6265 		file->deny_write++;
6266 	file->share_deny |= deny;
6267 	state->share_deny |= deny;
6268 
6269 	if (amodes & OPEN4_SHARE_ACCESS_READ)
6270 		file->access_read++;
6271 	if (amodes & OPEN4_SHARE_ACCESS_WRITE)
6272 		file->access_write++;
6273 	file->share_access |= access;
6274 	state->share_access |= access;
6275 
6276 	/*
6277 	 * Check for delegation here. if the deleg argument is not
6278 	 * DELEG_ANY, then this is a reclaim from a client and
6279 	 * we must honor the delegation requested. If necessary we can
6280 	 * set the recall flag.
6281 	 */
6282 
6283 	dsp = rfs4_grant_delegation(deleg, state, &recall);
6284 
6285 	cs->deleg = (file->dinfo->dtype == OPEN_DELEGATE_WRITE);
6286 
6287 	next_stateid(&state->stateid);
6288 
6289 	resp->stateid = state->stateid.stateid;
6290 
6291 	rfs4_dbe_unlock(file->dbe);
6292 	rfs4_dbe_unlock(state->dbe);
6293 
6294 	if (dsp) {
6295 		rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall);
6296 		rfs4_deleg_state_rele(dsp);
6297 	}
6298 
6299 	rfs4_file_rele(file);
6300 	rfs4_state_rele(state);
6301 
6302 	resp->status = NFS4_OK;
6303 }
6304 
6305 /*ARGSUSED*/
6306 static void
6307 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req,
6308 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6309 {
6310 	change_info4 *cinfo = &resp->cinfo;
6311 	bitmap4 *attrset = &resp->attrset;
6312 
6313 	if (args->opentype == OPEN4_NOCREATE)
6314 		resp->status = rfs4_lookupfile(&args->open_claim4_u.file,
6315 		    req, cs, args->share_access, cinfo);
6316 	else {
6317 		/* inhibit delegation grants during exclusive create */
6318 
6319 		if (args->mode == EXCLUSIVE4)
6320 			rfs4_disable_delegation();
6321 
6322 		resp->status = rfs4_createfile(args, req, cs, cinfo, attrset,
6323 		    oo->client->clientid);
6324 	}
6325 
6326 	if (resp->status == NFS4_OK) {
6327 
6328 		/* cs->vp cs->fh now reference the desired file */
6329 
6330 		rfs4_do_open(cs, req, oo, DELEG_ANY, args->share_access,
6331 		    args->share_deny, resp);
6332 
6333 		/*
6334 		 * If rfs4_createfile set attrset, we must
6335 		 * clear this attrset before the response is copied.
6336 		 */
6337 		if (resp->status != NFS4_OK && resp->attrset) {
6338 			resp->attrset = 0;
6339 		}
6340 	}
6341 	else
6342 		*cs->statusp = resp->status;
6343 
6344 	if (args->mode == EXCLUSIVE4)
6345 		rfs4_enable_delegation();
6346 }
6347 
6348 /*ARGSUSED*/
6349 static void
6350 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req,
6351 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6352 {
6353 	change_info4 *cinfo = &resp->cinfo;
6354 	vattr_t va;
6355 	vtype_t v_type = cs->vp->v_type;
6356 	int error = 0;
6357 
6358 	/* Verify that we have a regular file */
6359 	if (v_type != VREG) {
6360 		if (v_type == VDIR)
6361 			resp->status = NFS4ERR_ISDIR;
6362 		else if (v_type == VLNK)
6363 			resp->status = NFS4ERR_SYMLINK;
6364 		else
6365 			resp->status = NFS4ERR_INVAL;
6366 		return;
6367 	}
6368 
6369 	va.va_mask = AT_MODE|AT_UID;
6370 	error = VOP_GETATTR(cs->vp, &va, 0, cs->cr);
6371 	if (error) {
6372 		resp->status = puterrno4(error);
6373 		return;
6374 	}
6375 
6376 	cs->mandlock = MANDLOCK(cs->vp, va.va_mode);
6377 
6378 	/*
6379 	 * Check if we have access to the file, Note the the file
6380 	 * could have originally been open UNCHECKED or GUARDED
6381 	 * with mode bits that will now fail, but there is nothing
6382 	 * we can really do about that except in the case that the
6383 	 * owner of the file is the one requesting the open.
6384 	 */
6385 	if (crgetuid(cs->cr) != va.va_uid) {
6386 		resp->status = check_open_access(args->share_access, cs, req);
6387 		if (resp->status != NFS4_OK) {
6388 			return;
6389 		}
6390 	}
6391 
6392 	/*
6393 	 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero
6394 	 */
6395 	cinfo->before = 0;
6396 	cinfo->after = 0;
6397 	cinfo->atomic = FALSE;
6398 
6399 	rfs4_do_open(cs, req, oo,
6400 	    NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type),
6401 	    args->share_access, args->share_deny, resp);
6402 }
6403 
6404 static void
6405 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req,
6406 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6407 {
6408 	int error;
6409 	nfsstat4 status;
6410 	stateid4 stateid =
6411 	    args->open_claim4_u.delegate_cur_info.delegate_stateid;
6412 	rfs4_deleg_state_t *dsp;
6413 
6414 	/*
6415 	 * Find the state info from the stateid and confirm that the
6416 	 * file is delegated.  If the state openowner is the same as
6417 	 * the supplied openowner we're done. If not, get the file
6418 	 * info from the found state info. Use that file info to
6419 	 * create the state for this lock owner. Note solaris doen't
6420 	 * really need the pathname to find the file. We may want to
6421 	 * lookup the pathname and make sure that the vp exist and
6422 	 * matches the vp in the file structure. However it is
6423 	 * possible that the pathname nolonger exists (local process
6424 	 * unlinks the file), so this may not be that useful.
6425 	 */
6426 
6427 	status = rfs4_get_deleg_state(&stateid, &dsp);
6428 	if (status != NFS4_OK) {
6429 		resp->status = status;
6430 		return;
6431 	}
6432 
6433 	ASSERT(dsp->finfo->dinfo->dtype != OPEN_DELEGATE_NONE);
6434 
6435 	/*
6436 	 * New lock owner, create state. Since this was probably called
6437 	 * in response to a CB_RECALL we set deleg to DELEG_NONE
6438 	 */
6439 
6440 	ASSERT(cs->vp != NULL);
6441 	VN_RELE(cs->vp);
6442 	VN_HOLD(dsp->finfo->vp);
6443 	cs->vp = dsp->finfo->vp;
6444 
6445 	if (error = makefh4(&cs->fh, cs->vp, cs->exi)) {
6446 		rfs4_deleg_state_rele(dsp);
6447 		*cs->statusp = resp->status = puterrno4(error);
6448 		return;
6449 	}
6450 
6451 	/* Mark progress for delegation returns */
6452 	dsp->finfo->dinfo->time_lastwrite = gethrestime_sec();
6453 	rfs4_deleg_state_rele(dsp);
6454 	rfs4_do_open(cs, req, oo, DELEG_NONE,
6455 	    args->share_access, args->share_deny, resp);
6456 }
6457 
6458 /*ARGSUSED*/
6459 static void
6460 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req,
6461 			OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6462 {
6463 	/*
6464 	 * Lookup the pathname, it must already exist since this file
6465 	 * was delegated.
6466 	 *
6467 	 * Find the file and state info for this vp and open owner pair.
6468 	 *	check that they are in fact delegated.
6469 	 *	check that the state access and deny modes are the same.
6470 	 *
6471 	 * Return the delgation possibly seting the recall flag.
6472 	 */
6473 	rfs4_file_t *file;
6474 	rfs4_state_t *state;
6475 	bool_t create = FALSE;
6476 	bool_t dcreate = FALSE;
6477 	rfs4_deleg_state_t *dsp;
6478 	nfsace4 *ace;
6479 
6480 
6481 	/* Note we ignore oflags */
6482 	resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev,
6483 	    req, cs, args->share_access, &resp->cinfo);
6484 
6485 	if (resp->status != NFS4_OK) {
6486 		return;
6487 	}
6488 
6489 	/* get the file struct and hold a lock on it during initial open */
6490 	file = rfs4_findfile_withlock(cs->vp, NULL, &create);
6491 	if (file == NULL) {
6492 		NFS4_DEBUG(rfs4_debug,
6493 		    (CE_NOTE, "rfs4_do_opendelprev: can't find file"));
6494 		resp->status = NFS4ERR_SERVERFAULT;
6495 		return;
6496 	}
6497 
6498 	state = rfs4_findstate_by_owner_file(oo, file, &create);
6499 	if (state == NULL) {
6500 		NFS4_DEBUG(rfs4_debug,
6501 		    (CE_NOTE, "rfs4_do_opendelprev: can't find state"));
6502 		resp->status = NFS4ERR_SERVERFAULT;
6503 		rfs4_file_rele_withunlock(file);
6504 		return;
6505 	}
6506 
6507 	rfs4_dbe_lock(state->dbe);
6508 	rfs4_dbe_lock(file->dbe);
6509 	if (args->share_access != state->share_access ||
6510 	    args->share_deny != state->share_deny ||
6511 	    state->finfo->dinfo->dtype == OPEN_DELEGATE_NONE) {
6512 		NFS4_DEBUG(rfs4_debug,
6513 		    (CE_NOTE, "rfs4_do_opendelprev: state mixup"));
6514 		rfs4_dbe_unlock(file->dbe);
6515 		rfs4_dbe_unlock(state->dbe);
6516 		rfs4_file_rele(file);
6517 		rfs4_state_rele(state);
6518 		resp->status = NFS4ERR_SERVERFAULT;
6519 		return;
6520 	}
6521 	rfs4_dbe_unlock(file->dbe);
6522 	rfs4_dbe_unlock(state->dbe);
6523 
6524 	dsp = rfs4_finddeleg(state, &dcreate);
6525 	if (dsp == NULL) {
6526 		rfs4_state_rele(state);
6527 		rfs4_file_rele(file);
6528 		resp->status = NFS4ERR_SERVERFAULT;
6529 		return;
6530 	}
6531 
6532 	next_stateid(&state->stateid);
6533 
6534 	resp->stateid = state->stateid.stateid;
6535 
6536 	resp->delegation.delegation_type = dsp->dtype;
6537 
6538 	if (dsp->dtype == OPEN_DELEGATE_READ) {
6539 		open_read_delegation4 *rv =
6540 		    &resp->delegation.open_delegation4_u.read;
6541 
6542 		rv->stateid = dsp->delegid.stateid;
6543 		rv->recall = FALSE; /* no policy in place to set to TRUE */
6544 		ace = &rv->permissions;
6545 	} else {
6546 		open_write_delegation4 *rv =
6547 		    &resp->delegation.open_delegation4_u.write;
6548 
6549 		rv->stateid = dsp->delegid.stateid;
6550 		rv->recall = FALSE;  /* no policy in place to set to TRUE */
6551 		ace = &rv->permissions;
6552 		rv->space_limit.limitby = NFS_LIMIT_SIZE;
6553 		rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX;
6554 	}
6555 
6556 	/* XXX For now */
6557 	ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE;
6558 	ace->flag = 0;
6559 	ace->access_mask = 0;
6560 	ace->who.utf8string_len = 0;
6561 	ace->who.utf8string_val = 0;
6562 
6563 	rfs4_deleg_state_rele(dsp);
6564 	rfs4_state_rele(state);
6565 	rfs4_file_rele(file);
6566 }
6567 
6568 typedef enum {
6569 	NFS4_CHKSEQ_OKAY = 0,
6570 	NFS4_CHKSEQ_REPLAY = 1,
6571 	NFS4_CHKSEQ_BAD = 2
6572 } rfs4_chkseq_t;
6573 
6574 /*
6575  * Generic function for sequence number checks.
6576  */
6577 static rfs4_chkseq_t
6578 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop,
6579 		seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres)
6580 {
6581 	/* Same sequence ids and matching operations? */
6582 	if (seqid == rqst_seq && resop->resop == lastop->resop) {
6583 		if (copyres == TRUE) {
6584 			rfs4_free_reply(resop);
6585 			rfs4_copy_reply(resop, lastop);
6586 		}
6587 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
6588 		    "Replayed SEQID %d\n", seqid));
6589 		return (NFS4_CHKSEQ_REPLAY);
6590 	}
6591 
6592 	/* If the incoming sequence is not the next expected then it is bad */
6593 	if (rqst_seq != seqid + 1) {
6594 		if (rqst_seq == seqid) {
6595 			NFS4_DEBUG(rfs4_debug,
6596 			    (CE_NOTE, "BAD SEQID: Replayed sequence id "
6597 			    "but last op was %d current op is %d\n",
6598 			    lastop->resop, resop->resop));
6599 			return (NFS4_CHKSEQ_BAD);
6600 		}
6601 		NFS4_DEBUG(rfs4_debug,
6602 		    (CE_NOTE, "BAD SEQID: got %u expecting %u\n",
6603 		    rqst_seq, seqid));
6604 		return (NFS4_CHKSEQ_BAD);
6605 	}
6606 
6607 	/* Everything okay -- next expected */
6608 	return (NFS4_CHKSEQ_OKAY);
6609 }
6610 
6611 
6612 static rfs4_chkseq_t
6613 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop)
6614 {
6615 	rfs4_chkseq_t rc;
6616 
6617 	rfs4_dbe_lock(op->dbe);
6618 	rc = rfs4_check_seqid(op->open_seqid, op->reply, seqid, resop, TRUE);
6619 	rfs4_dbe_unlock(op->dbe);
6620 
6621 	if (rc == NFS4_CHKSEQ_OKAY)
6622 		rfs4_update_lease(op->client);
6623 
6624 	return (rc);
6625 }
6626 
6627 static rfs4_chkseq_t
6628 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op,
6629 	nfs_resop4 *resop)
6630 {
6631 	rfs4_chkseq_t rc;
6632 
6633 	rfs4_dbe_lock(op->dbe);
6634 	rc = rfs4_check_seqid(op->open_seqid, op->reply,
6635 	    olo_seqid, resop, FALSE);
6636 	rfs4_dbe_unlock(op->dbe);
6637 
6638 	return (rc);
6639 }
6640 
6641 static rfs4_chkseq_t
6642 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lp, nfs_resop4 *resop)
6643 {
6644 	rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY;
6645 
6646 	rfs4_dbe_lock(lp->dbe);
6647 	if (!lp->skip_seqid_check)
6648 		rc = rfs4_check_seqid(lp->seqid, lp->reply, seqid, resop, TRUE);
6649 	rfs4_dbe_unlock(lp->dbe);
6650 
6651 	return (rc);
6652 }
6653 
6654 static void
6655 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop,
6656 	    struct svc_req *req, struct compound_state *cs)
6657 {
6658 	OPEN4args *args = &argop->nfs_argop4_u.opopen;
6659 	OPEN4res *resp = &resop->nfs_resop4_u.opopen;
6660 	open_owner4 *owner = &args->owner;
6661 	open_claim_type4 claim = args->claim;
6662 	rfs4_client_t *cp;
6663 	rfs4_openowner_t *oo;
6664 	bool_t create;
6665 	bool_t replay = FALSE;
6666 	int can_reclaim;
6667 
6668 
6669 	if (cs->vp == NULL) {
6670 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
6671 		return;
6672 	}
6673 
6674 	/*
6675 	 * Need to check clientid and lease expiration first based on
6676 	 * error ordering and incrementing sequence id.
6677 	 */
6678 	cp = rfs4_findclient_by_id(owner->clientid, FALSE);
6679 	if (cp == NULL) {
6680 		*cs->statusp = resp->status =
6681 		    rfs4_check_clientid(&owner->clientid, 0);
6682 		return;
6683 	}
6684 
6685 	if (rfs4_lease_expired(cp)) {
6686 		rfs4_client_close(cp);
6687 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
6688 		return;
6689 	}
6690 	can_reclaim = cp->can_reclaim;
6691 
6692 	/*
6693 	 * Find the open_owner for use from this point forward.  Take
6694 	 * care in updating the sequence id based on the type of error
6695 	 * being returned.
6696 	 */
6697 retry:
6698 	create = TRUE;
6699 	oo = rfs4_findopenowner(owner, &create, args->seqid);
6700 	if (oo == NULL) {
6701 		*cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
6702 		rfs4_client_rele(cp);
6703 		return;
6704 	}
6705 
6706 	/* Hold off access to the sequence space while the open is done */
6707 	rfs4_sw_enter(&oo->oo_sw);
6708 
6709 	/*
6710 	 * If the open_owner existed before at the server, then check
6711 	 * the sequence id.
6712 	 */
6713 	if (!create && !oo->postpone_confirm) {
6714 		switch (rfs4_check_open_seqid(args->seqid, oo, resop)) {
6715 		case NFS4_CHKSEQ_BAD:
6716 			if ((args->seqid > oo->open_seqid) &&
6717 			    oo->need_confirm) {
6718 				rfs4_free_opens(oo, TRUE, FALSE);
6719 				rfs4_sw_exit(&oo->oo_sw);
6720 				rfs4_openowner_rele(oo);
6721 				goto retry;
6722 			}
6723 			resp->status = NFS4ERR_BAD_SEQID;
6724 			goto out;
6725 		case NFS4_CHKSEQ_REPLAY: /* replay of previous request */
6726 			replay = TRUE;
6727 			goto out;
6728 		default:
6729 			break;
6730 		}
6731 
6732 		/*
6733 		 * Sequence was ok and open owner exists
6734 		 * check to see if we have yet to see an
6735 		 * open_confirm.
6736 		 */
6737 		if (oo->need_confirm) {
6738 			rfs4_free_opens(oo, TRUE, FALSE);
6739 			rfs4_sw_exit(&oo->oo_sw);
6740 			rfs4_openowner_rele(oo);
6741 			goto retry;
6742 		}
6743 	}
6744 	/* Grace only applies to regular-type OPENs */
6745 	if (rfs4_clnt_in_grace(cp) &&
6746 	    (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) {
6747 		*cs->statusp = resp->status = NFS4ERR_GRACE;
6748 		goto out;
6749 	}
6750 
6751 	/*
6752 	 * If previous state at the server existed then can_reclaim
6753 	 * will be set. If not reply NFS4ERR_NO_GRACE to the
6754 	 * client.
6755 	 */
6756 	if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) {
6757 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
6758 		goto out;
6759 	}
6760 
6761 
6762 	/*
6763 	 * Reject the open if the client has missed the grace period
6764 	 */
6765 	if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) {
6766 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
6767 		goto out;
6768 	}
6769 
6770 	/* Couple of up-front bookkeeping items */
6771 	if (oo->need_confirm) {
6772 		/*
6773 		 * If this is a reclaim OPEN then we should not ask
6774 		 * for a confirmation of the open_owner per the
6775 		 * protocol specification.
6776 		 */
6777 		if (claim == CLAIM_PREVIOUS)
6778 			oo->need_confirm = FALSE;
6779 		else
6780 			resp->rflags |= OPEN4_RESULT_CONFIRM;
6781 	}
6782 	resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX;
6783 
6784 	/*
6785 	 * If there is an unshared filesystem mounted on this vnode,
6786 	 * do not allow to open/create in this directory.
6787 	 */
6788 	if (vn_ismntpt(cs->vp)) {
6789 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
6790 		goto out;
6791 	}
6792 
6793 	/*
6794 	 * access must READ, WRITE, or BOTH.  No access is invalid.
6795 	 * deny can be READ, WRITE, BOTH, or NONE.
6796 	 * bits not defined for access/deny are invalid.
6797 	 */
6798 	if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) ||
6799 	    (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) ||
6800 	    (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) {
6801 		*cs->statusp = resp->status = NFS4ERR_INVAL;
6802 		goto out;
6803 	}
6804 
6805 
6806 	/*
6807 	 * make sure attrset is zero before response is built.
6808 	 */
6809 	resp->attrset = 0;
6810 
6811 	switch (claim) {
6812 	case CLAIM_NULL:
6813 		rfs4_do_opennull(cs, req, args, oo, resp);
6814 		break;
6815 	case CLAIM_PREVIOUS:
6816 		rfs4_do_openprev(cs, req, args, oo, resp);
6817 		break;
6818 	case CLAIM_DELEGATE_CUR:
6819 		rfs4_do_opendelcur(cs, req, args, oo, resp);
6820 		break;
6821 	case CLAIM_DELEGATE_PREV:
6822 		rfs4_do_opendelprev(cs, req, args, oo, resp);
6823 		break;
6824 	default:
6825 		resp->status = NFS4ERR_INVAL;
6826 		break;
6827 	}
6828 
6829 out:
6830 	rfs4_client_rele(cp);
6831 
6832 	/* Catch sequence id handling here to make it a little easier */
6833 	switch (resp->status) {
6834 	case NFS4ERR_BADXDR:
6835 	case NFS4ERR_BAD_SEQID:
6836 	case NFS4ERR_BAD_STATEID:
6837 	case NFS4ERR_NOFILEHANDLE:
6838 	case NFS4ERR_RESOURCE:
6839 	case NFS4ERR_STALE_CLIENTID:
6840 	case NFS4ERR_STALE_STATEID:
6841 		/*
6842 		 * The protocol states that if any of these errors are
6843 		 * being returned, the sequence id should not be
6844 		 * incremented.  Any other return requires an
6845 		 * increment.
6846 		 */
6847 		break;
6848 	default:
6849 		/* Always update the lease in this case */
6850 		rfs4_update_lease(oo->client);
6851 
6852 		/* Regular response - copy the result */
6853 		if (!replay)
6854 			rfs4_update_open_resp(oo, resop, &cs->fh);
6855 
6856 		/*
6857 		 * REPLAY case: Only if the previous response was OK
6858 		 * do we copy the filehandle.  If not OK, no
6859 		 * filehandle to copy.
6860 		 */
6861 		if (replay == TRUE &&
6862 		    resp->status == NFS4_OK &&
6863 		    oo->reply_fh.nfs_fh4_val) {
6864 			/*
6865 			 * If this is a replay, we must restore the
6866 			 * current filehandle/vp to that of what was
6867 			 * returned originally.  Try our best to do
6868 			 * it.
6869 			 */
6870 			nfs_fh4_fmt_t *fh_fmtp =
6871 			    (nfs_fh4_fmt_t *)oo->reply_fh.nfs_fh4_val;
6872 
6873 			cs->exi = checkexport4(&fh_fmtp->fh4_fsid,
6874 			    (fid_t *)&fh_fmtp->fh4_xlen, NULL);
6875 
6876 			if (cs->exi == NULL) {
6877 				resp->status = NFS4ERR_STALE;
6878 				goto finish;
6879 			}
6880 
6881 			VN_RELE(cs->vp);
6882 
6883 			cs->vp = nfs4_fhtovp(&oo->reply_fh, cs->exi,
6884 			    &resp->status);
6885 
6886 			if (cs->vp == NULL)
6887 				goto finish;
6888 
6889 			nfs_fh4_copy(&oo->reply_fh, &cs->fh);
6890 		}
6891 
6892 		/*
6893 		 * If this was a replay, no need to update the
6894 		 * sequence id. If the open_owner was not created on
6895 		 * this pass, then update.  The first use of an
6896 		 * open_owner will not bump the sequence id.
6897 		 */
6898 		if (replay == FALSE && !create)
6899 			rfs4_update_open_sequence(oo);
6900 		/*
6901 		 * If the client is receiving an error and the
6902 		 * open_owner needs to be confirmed, there is no way
6903 		 * to notify the client of this fact ignoring the fact
6904 		 * that the server has no method of returning a
6905 		 * stateid to confirm.  Therefore, the server needs to
6906 		 * mark this open_owner in a way as to avoid the
6907 		 * sequence id checking the next time the client uses
6908 		 * this open_owner.
6909 		 */
6910 		if (resp->status != NFS4_OK && oo->need_confirm)
6911 			oo->postpone_confirm = TRUE;
6912 		/*
6913 		 * If OK response then clear the postpone flag and
6914 		 * reset the sequence id to keep in sync with the
6915 		 * client.
6916 		 */
6917 		if (resp->status == NFS4_OK && oo->postpone_confirm) {
6918 			oo->postpone_confirm = FALSE;
6919 			oo->open_seqid = args->seqid;
6920 		}
6921 		break;
6922 	}
6923 
6924 finish:
6925 	*cs->statusp = resp->status;
6926 
6927 	rfs4_sw_exit(&oo->oo_sw);
6928 	rfs4_openowner_rele(oo);
6929 }
6930 
6931 /*ARGSUSED*/
6932 void
6933 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
6934 		    struct svc_req *req, struct compound_state *cs)
6935 {
6936 	OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm;
6937 	OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm;
6938 	rfs4_state_t *sp;
6939 	nfsstat4 status;
6940 
6941 	if (cs->vp == NULL) {
6942 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
6943 		return;
6944 	}
6945 
6946 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
6947 	if (status != NFS4_OK) {
6948 		*cs->statusp = resp->status = status;
6949 		return;
6950 	}
6951 
6952 	/* Ensure specified filehandle matches */
6953 	if (cs->vp != sp->finfo->vp) {
6954 		rfs4_state_rele(sp);
6955 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
6956 		return;
6957 	}
6958 
6959 	/* hold off other access to open_owner while we tinker */
6960 	rfs4_sw_enter(&sp->owner->oo_sw);
6961 
6962 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
6963 	case NFS4_CHECK_STATEID_OKAY:
6964 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
6965 		    resop) != 0) {
6966 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
6967 			break;
6968 		}
6969 		/*
6970 		 * If it is the appropriate stateid and determined to
6971 		 * be "OKAY" then this means that the stateid does not
6972 		 * need to be confirmed and the client is in error for
6973 		 * sending an OPEN_CONFIRM.
6974 		 */
6975 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
6976 		break;
6977 	case NFS4_CHECK_STATEID_OLD:
6978 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
6979 		break;
6980 	case NFS4_CHECK_STATEID_BAD:
6981 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
6982 		break;
6983 	case NFS4_CHECK_STATEID_EXPIRED:
6984 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
6985 		break;
6986 	case NFS4_CHECK_STATEID_CLOSED:
6987 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
6988 		break;
6989 	case NFS4_CHECK_STATEID_REPLAY:
6990 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
6991 		case NFS4_CHKSEQ_OKAY:
6992 			/*
6993 			 * This is replayed stateid; if seqid matches
6994 			 * next expected, then client is using wrong seqid.
6995 			 */
6996 			/* fall through */
6997 		case NFS4_CHKSEQ_BAD:
6998 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
6999 			break;
7000 		case NFS4_CHKSEQ_REPLAY:
7001 			/*
7002 			 * Note this case is the duplicate case so
7003 			 * resp->status is already set.
7004 			 */
7005 			*cs->statusp = resp->status;
7006 			rfs4_update_lease(sp->owner->client);
7007 			break;
7008 		}
7009 		break;
7010 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7011 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7012 		    resop) != NFS4_CHKSEQ_OKAY) {
7013 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7014 			break;
7015 		}
7016 		*cs->statusp = resp->status = NFS4_OK;
7017 
7018 		next_stateid(&sp->stateid);
7019 		resp->open_stateid = sp->stateid.stateid;
7020 		sp->owner->need_confirm = FALSE;
7021 		rfs4_update_lease(sp->owner->client);
7022 		rfs4_update_open_sequence(sp->owner);
7023 		rfs4_update_open_resp(sp->owner, resop, NULL);
7024 		break;
7025 	default:
7026 		ASSERT(FALSE);
7027 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7028 		break;
7029 	}
7030 	rfs4_sw_exit(&sp->owner->oo_sw);
7031 	rfs4_state_rele(sp);
7032 }
7033 
7034 /*ARGSUSED*/
7035 void
7036 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop,
7037 		    struct svc_req *req, struct compound_state *cs)
7038 {
7039 	OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade;
7040 	OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade;
7041 	uint32_t access = args->share_access;
7042 	uint32_t deny = args->share_deny;
7043 	nfsstat4 status;
7044 	rfs4_state_t *sp;
7045 	rfs4_file_t *fp;
7046 
7047 	if (cs->vp == NULL) {
7048 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7049 		return;
7050 	}
7051 
7052 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
7053 	if (status != NFS4_OK) {
7054 		*cs->statusp = resp->status = status;
7055 		return;
7056 	}
7057 
7058 	/* Ensure specified filehandle matches */
7059 	if (cs->vp != sp->finfo->vp) {
7060 		rfs4_state_rele(sp);
7061 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7062 		return;
7063 	}
7064 
7065 	/* hold off other access to open_owner while we tinker */
7066 	rfs4_sw_enter(&sp->owner->oo_sw);
7067 
7068 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7069 	case NFS4_CHECK_STATEID_OKAY:
7070 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7071 		    resop) != NFS4_CHKSEQ_OKAY) {
7072 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7073 			goto end;
7074 		}
7075 		break;
7076 	case NFS4_CHECK_STATEID_OLD:
7077 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7078 		goto end;
7079 	case NFS4_CHECK_STATEID_BAD:
7080 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7081 		goto end;
7082 	case NFS4_CHECK_STATEID_EXPIRED:
7083 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7084 		goto end;
7085 	case NFS4_CHECK_STATEID_CLOSED:
7086 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7087 		goto end;
7088 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7089 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7090 		goto end;
7091 	case NFS4_CHECK_STATEID_REPLAY:
7092 		/* Check the sequence id for the open owner */
7093 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
7094 		case NFS4_CHKSEQ_OKAY:
7095 			/*
7096 			 * This is replayed stateid; if seqid matches
7097 			 * next expected, then client is using wrong seqid.
7098 			 */
7099 			/* fall through */
7100 		case NFS4_CHKSEQ_BAD:
7101 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7102 			goto end;
7103 		case NFS4_CHKSEQ_REPLAY:
7104 			/*
7105 			 * Note this case is the duplicate case so
7106 			 * resp->status is already set.
7107 			 */
7108 			*cs->statusp = resp->status;
7109 			rfs4_update_lease(sp->owner->client);
7110 			goto end;
7111 		}
7112 		break;
7113 	default:
7114 		ASSERT(FALSE);
7115 		break;
7116 	}
7117 
7118 	rfs4_dbe_lock(sp->dbe);
7119 	/*
7120 	 * Check that the new access modes and deny modes are valid.
7121 	 * Check that no invalid bits are set.
7122 	 */
7123 	if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) ||
7124 	    (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_READ))) {
7125 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7126 		rfs4_update_open_sequence(sp->owner);
7127 		rfs4_dbe_unlock(sp->dbe);
7128 		goto end;
7129 	}
7130 
7131 	/*
7132 	 * The new modes must be a subset of the current modes and
7133 	 * the access must specify at least one mode. To test that
7134 	 * the new mode is a subset of the current modes we bitwise
7135 	 * AND them together and check that the result equals the new
7136 	 * mode. For example:
7137 	 * New mode, access == R and current mode, sp->share_access  == RW
7138 	 * access & sp->share_access == R == access, so the new access mode
7139 	 * is valid. Consider access == RW, sp->share_access = R
7140 	 * access & sp->share_access == R != access, so the new access mode
7141 	 * is invalid.
7142 	 */
7143 	if ((access & sp->share_access) != access ||
7144 	    (deny & sp->share_deny) != deny ||
7145 	    (access &
7146 	    (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) {
7147 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7148 		rfs4_update_open_sequence(sp->owner);
7149 		rfs4_dbe_unlock(sp->dbe);
7150 		goto end;
7151 	}
7152 
7153 	/*
7154 	 * Release any share locks associated with this stateID.
7155 	 * Strictly speaking, this violates the spec because the
7156 	 * spec effectively requires that open downgrade be atomic.
7157 	 * At present, fs_shrlock does not have this capability.
7158 	 */
7159 	rfs4_dbe_unlock(sp->dbe);
7160 	rfs4_unshare(sp);
7161 	rfs4_dbe_lock(sp->dbe);
7162 
7163 	fp = sp->finfo;
7164 	rfs4_dbe_lock(fp->dbe);
7165 
7166 	/*
7167 	 * If the current mode has deny read and the new mode
7168 	 * does not, decrement the number of deny read mode bits
7169 	 * and if it goes to zero turn off the deny read bit
7170 	 * on the file.
7171 	 */
7172 	if ((sp->share_deny & OPEN4_SHARE_DENY_READ) &&
7173 	    (deny & OPEN4_SHARE_DENY_READ) == 0) {
7174 		fp->deny_read--;
7175 		if (fp->deny_read == 0)
7176 			fp->share_deny &= ~OPEN4_SHARE_DENY_READ;
7177 	}
7178 
7179 	/*
7180 	 * If the current mode has deny write and the new mode
7181 	 * does not, decrement the number of deny write mode bits
7182 	 * and if it goes to zero turn off the deny write bit
7183 	 * on the file.
7184 	 */
7185 	if ((sp->share_deny & OPEN4_SHARE_DENY_WRITE) &&
7186 	    (deny & OPEN4_SHARE_DENY_WRITE) == 0) {
7187 		fp->deny_write--;
7188 		if (fp->deny_write == 0)
7189 			fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE;
7190 	}
7191 
7192 	/*
7193 	 * If the current mode has access read and the new mode
7194 	 * does not, decrement the number of access read mode bits
7195 	 * and if it goes to zero turn off the access read bit
7196 	 * on the file.
7197 	 */
7198 	if ((sp->share_access & OPEN4_SHARE_ACCESS_READ) &&
7199 	    (access & OPEN4_SHARE_ACCESS_READ) == 0) {
7200 		fp->access_read--;
7201 		if (fp->access_read == 0)
7202 			fp->share_access &= ~OPEN4_SHARE_ACCESS_READ;
7203 	}
7204 
7205 	/*
7206 	 * If the current mode has access write and the new mode
7207 	 * does not, decrement the number of access write mode bits
7208 	 * and if it goes to zero turn off the access write bit
7209 	 * on the file.
7210 	 */
7211 	if ((sp->share_access & OPEN4_SHARE_ACCESS_WRITE) &&
7212 	    (access & OPEN4_SHARE_ACCESS_WRITE) == 0) {
7213 		fp->access_write--;
7214 		if (fp->access_write == 0)
7215 			fp->share_deny &= ~OPEN4_SHARE_ACCESS_WRITE;
7216 	}
7217 
7218 	/* Set the new access and deny modes */
7219 	sp->share_access = access;
7220 	sp->share_deny = deny;
7221 	/* Check that the file is still accessible */
7222 	ASSERT(fp->share_access);
7223 
7224 	rfs4_dbe_unlock(fp->dbe);
7225 
7226 	rfs4_dbe_unlock(sp->dbe);
7227 	if ((status = rfs4_share(sp)) != NFS4_OK) {
7228 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7229 		rfs4_update_open_sequence(sp->owner);
7230 		goto end;
7231 	}
7232 
7233 	rfs4_dbe_lock(sp->dbe);
7234 
7235 	/* Update the stateid */
7236 	next_stateid(&sp->stateid);
7237 	resp->open_stateid = sp->stateid.stateid;
7238 
7239 	rfs4_dbe_unlock(sp->dbe);
7240 
7241 	*cs->statusp = resp->status = NFS4_OK;
7242 	/* Update the lease */
7243 	rfs4_update_lease(sp->owner->client);
7244 	/* And the sequence */
7245 	rfs4_update_open_sequence(sp->owner);
7246 	rfs4_update_open_resp(sp->owner, resop, NULL);
7247 
7248 end:
7249 	rfs4_sw_exit(&sp->owner->oo_sw);
7250 	rfs4_state_rele(sp);
7251 }
7252 
7253 /*
7254  * The logic behind this function is detailed in the NFSv4 RFC in the
7255  * SETCLIENTID operation description under IMPLEMENTATION.  Refer to
7256  * that section for explicit guidance to server behavior for
7257  * SETCLIENTID.
7258  */
7259 void
7260 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop,
7261 		    struct svc_req *req, struct compound_state *cs)
7262 {
7263 	SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid;
7264 	SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid;
7265 	rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed;
7266 	bool_t create = TRUE;
7267 	char *addr, *netid;
7268 	int len;
7269 
7270 retry:
7271 	newcp = cp_confirmed = cp_unconfirmed = NULL;
7272 
7273 	/*
7274 	 * In search of an EXISTING client matching the incoming
7275 	 * request to establish a new client identifier at the server
7276 	 */
7277 	create = TRUE;
7278 	cp = rfs4_findclient(&args->client, &create, NULL);
7279 
7280 	/* Should never happen */
7281 	ASSERT(cp != NULL);
7282 
7283 	if (cp == NULL) {
7284 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
7285 		return;
7286 	}
7287 
7288 	/*
7289 	 * Easiest case. Client identifier is newly created and is
7290 	 * unconfirmed.  Also note that for this case, no other
7291 	 * entries exist for the client identifier.  Nothing else to
7292 	 * check.  Just setup the response and respond.
7293 	 */
7294 	if (create) {
7295 		*cs->statusp = res->status = NFS4_OK;
7296 		res->SETCLIENTID4res_u.resok4.clientid = cp->clientid;
7297 		res->SETCLIENTID4res_u.resok4.setclientid_confirm =
7298 		    cp->confirm_verf;
7299 		/* Setup callback information; CB_NULL confirmation later */
7300 		rfs4_client_setcb(cp, &args->callback, args->callback_ident);
7301 
7302 		rfs4_client_rele(cp);
7303 		return;
7304 	}
7305 
7306 	/*
7307 	 * An existing, confirmed client may exist but it may not have
7308 	 * been active for at least one lease period.  If so, then
7309 	 * "close" the client and create a new client identifier
7310 	 */
7311 	if (rfs4_lease_expired(cp)) {
7312 		rfs4_client_close(cp);
7313 		goto retry;
7314 	}
7315 
7316 	if (cp->need_confirm == TRUE)
7317 		cp_unconfirmed = cp;
7318 	else
7319 		cp_confirmed = cp;
7320 
7321 	cp = NULL;
7322 
7323 	/*
7324 	 * We have a confirmed client, now check for an
7325 	 * unconfimred entry
7326 	 */
7327 	if (cp_confirmed) {
7328 		/* If creds don't match then client identifier is inuse */
7329 		if (!creds_ok(cp_confirmed->cr_set, req, cs)) {
7330 			rfs4_cbinfo_t *cbp;
7331 			/*
7332 			 * Some one else has established this client
7333 			 * id. Try and say * who they are. We will use
7334 			 * the call back address supplied by * the
7335 			 * first client.
7336 			 */
7337 			*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
7338 
7339 			addr = netid = NULL;
7340 
7341 			cbp = &cp_confirmed->cbinfo;
7342 			if (cbp->cb_callback.cb_location.r_addr &&
7343 			    cbp->cb_callback.cb_location.r_netid) {
7344 				cb_client4 *cbcp = &cbp->cb_callback;
7345 
7346 				len = strlen(cbcp->cb_location.r_addr)+1;
7347 				addr = kmem_alloc(len, KM_SLEEP);
7348 				bcopy(cbcp->cb_location.r_addr, addr, len);
7349 				len = strlen(cbcp->cb_location.r_netid)+1;
7350 				netid = kmem_alloc(len, KM_SLEEP);
7351 				bcopy(cbcp->cb_location.r_netid, netid, len);
7352 			}
7353 
7354 			res->SETCLIENTID4res_u.client_using.r_addr = addr;
7355 			res->SETCLIENTID4res_u.client_using.r_netid = netid;
7356 
7357 			rfs4_client_rele(cp_confirmed);
7358 		}
7359 
7360 		/*
7361 		 * Confirmed, creds match, and verifier matches; must
7362 		 * be an update of the callback info
7363 		 */
7364 		if (cp_confirmed->nfs_client.verifier ==
7365 		    args->client.verifier) {
7366 			/* Setup callback information */
7367 			rfs4_client_setcb(cp_confirmed, &args->callback,
7368 			    args->callback_ident);
7369 
7370 			/* everything okay -- move ahead */
7371 			*cs->statusp = res->status = NFS4_OK;
7372 			res->SETCLIENTID4res_u.resok4.clientid =
7373 			    cp_confirmed->clientid;
7374 
7375 			/* update the confirm_verifier and return it */
7376 			rfs4_client_scv_next(cp_confirmed);
7377 			res->SETCLIENTID4res_u.resok4.setclientid_confirm =
7378 			    cp_confirmed->confirm_verf;
7379 
7380 			rfs4_client_rele(cp_confirmed);
7381 			return;
7382 		}
7383 
7384 		/*
7385 		 * Creds match but the verifier doesn't.  Must search
7386 		 * for an unconfirmed client that would be replaced by
7387 		 * this request.
7388 		 */
7389 		create = FALSE;
7390 		cp_unconfirmed = rfs4_findclient(&args->client, &create,
7391 		    cp_confirmed);
7392 	}
7393 
7394 	/*
7395 	 * At this point, we have taken care of the brand new client
7396 	 * struct, INUSE case, update of an existing, and confirmed
7397 	 * client struct.
7398 	 */
7399 
7400 	/*
7401 	 * check to see if things have changed while we originally
7402 	 * picked up the client struct.  If they have, then return and
7403 	 * retry the processing of this SETCLIENTID request.
7404 	 */
7405 	if (cp_unconfirmed) {
7406 		rfs4_dbe_lock(cp_unconfirmed->dbe);
7407 		if (!cp_unconfirmed->need_confirm) {
7408 			rfs4_dbe_unlock(cp_unconfirmed->dbe);
7409 			rfs4_client_rele(cp_unconfirmed);
7410 			if (cp_confirmed)
7411 				rfs4_client_rele(cp_confirmed);
7412 			goto retry;
7413 		}
7414 		/* do away with the old unconfirmed one */
7415 		rfs4_dbe_invalidate(cp_unconfirmed->dbe);
7416 		rfs4_dbe_unlock(cp_unconfirmed->dbe);
7417 		rfs4_client_rele(cp_unconfirmed);
7418 		cp_unconfirmed = NULL;
7419 	}
7420 
7421 	/*
7422 	 * This search will temporarily hide the confirmed client
7423 	 * struct while a new client struct is created as the
7424 	 * unconfirmed one.
7425 	 */
7426 	create = TRUE;
7427 	newcp = rfs4_findclient(&args->client, &create, cp_confirmed);
7428 
7429 	ASSERT(newcp != NULL);
7430 
7431 	if (newcp == NULL) {
7432 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
7433 		rfs4_client_rele(cp_confirmed);
7434 		return;
7435 	}
7436 
7437 	/*
7438 	 * If one was not created, then a similar request must be in
7439 	 * process so release and start over with this one
7440 	 */
7441 	if (create != TRUE) {
7442 		rfs4_client_rele(newcp);
7443 		if (cp_confirmed)
7444 			rfs4_client_rele(cp_confirmed);
7445 		goto retry;
7446 	}
7447 
7448 	*cs->statusp = res->status = NFS4_OK;
7449 	res->SETCLIENTID4res_u.resok4.clientid = newcp->clientid;
7450 	res->SETCLIENTID4res_u.resok4.setclientid_confirm = newcp->confirm_verf;
7451 	/* Setup callback information; CB_NULL confirmation later */
7452 	rfs4_client_setcb(newcp, &args->callback, args->callback_ident);
7453 
7454 	newcp->cp_confirmed = cp_confirmed;
7455 
7456 	rfs4_client_rele(newcp);
7457 }
7458 
7459 /*ARGSUSED*/
7460 void
7461 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
7462 			    struct svc_req *req, struct compound_state *cs)
7463 {
7464 	SETCLIENTID_CONFIRM4args *args =
7465 	    &argop->nfs_argop4_u.opsetclientid_confirm;
7466 	SETCLIENTID_CONFIRM4res *res =
7467 	    &resop->nfs_resop4_u.opsetclientid_confirm;
7468 	rfs4_client_t *cp, *cptoclose = NULL;
7469 
7470 	*cs->statusp = res->status = NFS4_OK;
7471 
7472 	cp = rfs4_findclient_by_id(args->clientid, TRUE);
7473 
7474 	if (cp == NULL) {
7475 		*cs->statusp = res->status =
7476 		    rfs4_check_clientid(&args->clientid, 1);
7477 		return;
7478 	}
7479 
7480 	if (!creds_ok(cp, req, cs)) {
7481 		*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
7482 		rfs4_client_rele(cp);
7483 		return;
7484 	}
7485 
7486 	/* If the verifier doesn't match, the record doesn't match */
7487 	if (cp->confirm_verf != args->setclientid_confirm) {
7488 		*cs->statusp = res->status = NFS4ERR_STALE_CLIENTID;
7489 		rfs4_client_rele(cp);
7490 		return;
7491 	}
7492 
7493 	rfs4_dbe_lock(cp->dbe);
7494 	cp->need_confirm = FALSE;
7495 	if (cp->cp_confirmed) {
7496 		cptoclose = cp->cp_confirmed;
7497 		cptoclose->ss_remove = 1;
7498 		cp->cp_confirmed = NULL;
7499 	}
7500 
7501 	/*
7502 	 * Update the client's associated server instance, if it's changed
7503 	 * since the client was created.
7504 	 */
7505 	if (rfs4_servinst(cp) != rfs4_cur_servinst)
7506 		rfs4_servinst_assign(cp, rfs4_cur_servinst);
7507 
7508 	/*
7509 	 * Record clientid in stable storage.
7510 	 * Must be done after server instance has been assigned.
7511 	 */
7512 	rfs4_ss_clid(cp, req);
7513 
7514 	rfs4_dbe_unlock(cp->dbe);
7515 
7516 	if (cptoclose)
7517 		/* don't need to rele, client_close does it */
7518 		rfs4_client_close(cptoclose);
7519 
7520 	/* If needed, initiate CB_NULL call for callback path */
7521 	rfs4_deleg_cb_check(cp);
7522 	rfs4_update_lease(cp);
7523 
7524 	/*
7525 	 * Check to see if client can perform reclaims
7526 	 */
7527 	rfs4_ss_chkclid(cp);
7528 
7529 	rfs4_client_rele(cp);
7530 }
7531 
7532 
7533 /*ARGSUSED*/
7534 void
7535 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop,
7536 	    struct svc_req *req, struct compound_state *cs)
7537 {
7538 	/* XXX Currently not using req arg */
7539 	CLOSE4args *args = &argop->nfs_argop4_u.opclose;
7540 	CLOSE4res *resp = &resop->nfs_resop4_u.opclose;
7541 	rfs4_state_t *sp;
7542 	nfsstat4 status;
7543 
7544 	if (cs->vp == NULL) {
7545 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7546 		return;
7547 	}
7548 
7549 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID);
7550 	if (status != NFS4_OK) {
7551 		*cs->statusp = resp->status = status;
7552 		return;
7553 	}
7554 
7555 	/* Ensure specified filehandle matches */
7556 	if (cs->vp != sp->finfo->vp) {
7557 		rfs4_state_rele(sp);
7558 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7559 		return;
7560 	}
7561 
7562 	/* hold off other access to open_owner while we tinker */
7563 	rfs4_sw_enter(&sp->owner->oo_sw);
7564 
7565 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7566 	case NFS4_CHECK_STATEID_OKAY:
7567 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7568 		    resop) != NFS4_CHKSEQ_OKAY) {
7569 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7570 			goto end;
7571 		}
7572 		break;
7573 	case NFS4_CHECK_STATEID_OLD:
7574 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7575 		goto end;
7576 	case NFS4_CHECK_STATEID_BAD:
7577 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7578 		goto end;
7579 	case NFS4_CHECK_STATEID_EXPIRED:
7580 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7581 		goto end;
7582 	case NFS4_CHECK_STATEID_CLOSED:
7583 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7584 		goto end;
7585 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7586 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7587 		goto end;
7588 	case NFS4_CHECK_STATEID_REPLAY:
7589 		/* Check the sequence id for the open owner */
7590 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
7591 		case NFS4_CHKSEQ_OKAY:
7592 			/*
7593 			 * This is replayed stateid; if seqid matches
7594 			 * next expected, then client is using wrong seqid.
7595 			 */
7596 			/* FALL THROUGH */
7597 		case NFS4_CHKSEQ_BAD:
7598 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7599 			goto end;
7600 		case NFS4_CHKSEQ_REPLAY:
7601 			/*
7602 			 * Note this case is the duplicate case so
7603 			 * resp->status is already set.
7604 			 */
7605 			*cs->statusp = resp->status;
7606 			rfs4_update_lease(sp->owner->client);
7607 			goto end;
7608 		}
7609 		break;
7610 	default:
7611 		ASSERT(FALSE);
7612 		break;
7613 	}
7614 
7615 	rfs4_dbe_lock(sp->dbe);
7616 
7617 	/* Update the stateid. */
7618 	next_stateid(&sp->stateid);
7619 	resp->open_stateid = sp->stateid.stateid;
7620 
7621 	rfs4_dbe_unlock(sp->dbe);
7622 
7623 	rfs4_update_lease(sp->owner->client);
7624 	rfs4_update_open_sequence(sp->owner);
7625 	rfs4_update_open_resp(sp->owner, resop, NULL);
7626 
7627 	rfs4_state_close(sp, FALSE, FALSE, cs->cr);
7628 
7629 	*cs->statusp = resp->status = status;
7630 
7631 end:
7632 	rfs4_sw_exit(&sp->owner->oo_sw);
7633 	rfs4_state_rele(sp);
7634 }
7635 
7636 /*
7637  * Manage the counts on the file struct and close all file locks
7638  */
7639 /*ARGSUSED*/
7640 void
7641 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr,
7642 	bool_t close_of_client)
7643 {
7644 	rfs4_file_t *fp = sp->finfo;
7645 	rfs4_lo_state_t *lsp;
7646 	struct shrlock shr;
7647 	struct shr_locowner shr_loco;
7648 	int fflags, s_access, s_deny;
7649 
7650 	fflags = s_access = s_deny = 0;
7651 	/*
7652 	 * Decrement the count for each access and deny bit that this
7653 	 * state has contributed to the file. If the file counts go to zero
7654 	 * clear the appropriate bit in the appropriate mask.
7655 	 */
7656 
7657 	if (sp->share_access & OPEN4_SHARE_ACCESS_READ) {
7658 		fp->access_read--;
7659 		fflags |= FREAD;
7660 		s_access |= F_RDACC;
7661 		if (fp->access_read == 0)
7662 			fp->share_access &= ~OPEN4_SHARE_ACCESS_READ;
7663 	}
7664 	if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) {
7665 		fp->access_write--;
7666 		fflags |= FWRITE;
7667 		s_access |= F_WRACC;
7668 		if (fp->access_write == 0)
7669 			fp->share_access &= ~OPEN4_SHARE_ACCESS_WRITE;
7670 	}
7671 	if (sp->share_deny & OPEN4_SHARE_DENY_READ) {
7672 		fp->deny_read--;
7673 		s_deny |= F_RDDNY;
7674 		if (fp->deny_read == 0)
7675 			fp->share_deny &= ~OPEN4_SHARE_DENY_READ;
7676 	}
7677 	if (sp->share_deny & OPEN4_SHARE_DENY_WRITE) {
7678 		fp->deny_write--;
7679 		s_deny |= F_WRDNY;
7680 		if (fp->deny_write == 0)
7681 			fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE;
7682 	}
7683 
7684 	/*
7685 	 * If this call is part of the larger closing down of client
7686 	 * state then it is just easier to release all locks
7687 	 * associated with this client instead of going through each
7688 	 * individual file and cleaning locks there.
7689 	 */
7690 	if (close_of_client) {
7691 		if (sp->owner->client->unlksys_completed == FALSE &&
7692 		    sp->lockownerlist.next->lsp != NULL &&
7693 		    sp->owner->client->sysidt != LM_NOSYSID) {
7694 			/* Is the PxFS kernel module loaded? */
7695 			if (lm_remove_file_locks != NULL) {
7696 				int new_sysid;
7697 
7698 				/* Encode the cluster nodeid in new sysid */
7699 				new_sysid = sp->owner->client->sysidt;
7700 				lm_set_nlmid_flk(&new_sysid);
7701 
7702 				/*
7703 				 * This PxFS routine removes file locks for a
7704 				 * client over all nodes of a cluster.
7705 				 */
7706 				NFS4_DEBUG(rfs4_debug, (CE_NOTE,
7707 				    "lm_remove_file_locks(sysid=0x%x)\n",
7708 				    new_sysid));
7709 				(*lm_remove_file_locks)(new_sysid);
7710 			} else {
7711 				struct flock64 flk;
7712 
7713 				/* Release all locks for this client */
7714 				flk.l_type = F_UNLKSYS;
7715 				flk.l_whence = 0;
7716 				flk.l_start = 0;
7717 				flk.l_len = 0;
7718 				flk.l_sysid = sp->owner->client->sysidt;
7719 				flk.l_pid = 0;
7720 				(void) VOP_FRLOCK(sp->finfo->vp, F_SETLK, &flk,
7721 				    F_REMOTELOCK | FREAD | FWRITE,
7722 				    (u_offset_t)0, NULL, CRED());
7723 			}
7724 
7725 			sp->owner->client->unlksys_completed = TRUE;
7726 		}
7727 	}
7728 
7729 	/*
7730 	 * Release all locks on this file by this lock owner or at
7731 	 * least mark the locks as having been released
7732 	 */
7733 	for (lsp = sp->lockownerlist.next->lsp; lsp != NULL;
7734 	    lsp = lsp->lockownerlist.next->lsp) {
7735 
7736 		lsp->locks_cleaned = TRUE;
7737 
7738 		/* Was this already taken care of above? */
7739 		if (!close_of_client &&
7740 		    sp->owner->client->sysidt != LM_NOSYSID)
7741 			(void) cleanlocks(sp->finfo->vp, lsp->locker->pid,
7742 			    lsp->locker->client->sysidt);
7743 	}
7744 
7745 	/*
7746 	 * Release any shrlocks associated with this open state ID.
7747 	 * This must be done before the rfs4_state gets marked closed.
7748 	 */
7749 	if (sp->owner->client->sysidt != LM_NOSYSID) {
7750 		shr.s_access = s_access;
7751 		shr.s_deny = s_deny;
7752 		shr.s_pid = rfs4_dbe_getid(sp->owner->dbe);
7753 		shr.s_sysid = sp->owner->client->sysidt;
7754 		shr_loco.sl_pid = shr.s_pid;
7755 		shr_loco.sl_id = shr.s_sysid;
7756 		shr.s_owner = (caddr_t)&shr_loco;
7757 		shr.s_own_len = sizeof (shr_loco);
7758 		(void) vop_shrlock(sp->finfo->vp, F_UNSHARE, &shr, fflags);
7759 	}
7760 }
7761 
7762 /*
7763  * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure.
7764  */
7765 static nfsstat4
7766 lock_denied(LOCK4denied *dp, struct flock64 *flk)
7767 {
7768 	rfs4_lockowner_t *lo;
7769 	rfs4_client_t *cp;
7770 	uint32_t len;
7771 
7772 	lo = rfs4_findlockowner_by_pid(flk->l_pid);
7773 	if (lo != NULL) {
7774 		cp = lo->client;
7775 		if (rfs4_lease_expired(cp)) {
7776 			rfs4_lockowner_rele(lo);
7777 			rfs4_dbe_hold(cp->dbe);
7778 			rfs4_client_close(cp);
7779 			return (NFS4ERR_EXPIRED);
7780 		}
7781 		dp->owner.clientid = lo->owner.clientid;
7782 		len = lo->owner.owner_len;
7783 		dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
7784 		bcopy(lo->owner.owner_val, dp->owner.owner_val, len);
7785 		dp->owner.owner_len = len;
7786 		rfs4_lockowner_rele(lo);
7787 		goto finish;
7788 	}
7789 
7790 	/*
7791 	 * Its not a NFS4 lock. We take advantage that the upper 32 bits
7792 	 * of the client id contain the boot time for a NFS4 lock. So we
7793 	 * fabricate and identity by setting clientid to the sysid, and
7794 	 * the lock owner to the pid.
7795 	 */
7796 	dp->owner.clientid = flk->l_sysid;
7797 	len = sizeof (pid_t);
7798 	dp->owner.owner_len = len;
7799 	dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
7800 	bcopy(&flk->l_pid, dp->owner.owner_val, len);
7801 finish:
7802 	dp->offset = flk->l_start;
7803 	dp->length = flk->l_len;
7804 
7805 	if (flk->l_type == F_RDLCK)
7806 		dp->locktype = READ_LT;
7807 	else if (flk->l_type == F_WRLCK)
7808 		dp->locktype = WRITE_LT;
7809 	else
7810 		return (NFS4ERR_INVAL);	/* no mapping from POSIX ltype to v4 */
7811 
7812 	return (NFS4_OK);
7813 }
7814 
7815 static int
7816 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred)
7817 {
7818 	int error;
7819 	struct flock64 flk;
7820 	int i;
7821 	clock_t delaytime;
7822 
7823 retry:
7824 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
7825 
7826 	for (i = 0; i < rfs4_maxlock_tries; i++) {
7827 		LOCK_PRINT(rfs4_debug, "setlock", F_SETLK, flock);
7828 		error = VOP_FRLOCK(vp, F_SETLK,
7829 		    flock, flag, (u_offset_t)0, NULL, cred);
7830 
7831 		if (error != EAGAIN && error != EACCES)
7832 			break;
7833 
7834 		if (i < rfs4_maxlock_tries - 1) {
7835 			delay(delaytime);
7836 			delaytime *= 2;
7837 		}
7838 	}
7839 
7840 	if (error == EAGAIN || error == EACCES) {
7841 		/* Get the owner of the lock */
7842 		flk = *flock;
7843 		LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk);
7844 		if (VOP_FRLOCK(vp, F_GETLK,
7845 		    &flk,  flag, (u_offset_t)0, NULL, cred) == 0) {
7846 			if (flk.l_type == F_UNLCK) {
7847 				/* No longer locked, retry */
7848 				goto retry;
7849 			}
7850 			*flock = flk;
7851 			LOCK_PRINT(rfs4_debug, "setlock(blocking lock)",
7852 			    F_GETLK, &flk);
7853 		}
7854 	}
7855 
7856 	return (error);
7857 }
7858 
7859 /*ARGSUSED*/
7860 static nfsstat4
7861 rfs4_do_lock(rfs4_lo_state_t *lp, nfs_lock_type4 locktype,
7862 	    seqid4 seqid, offset4 offset,
7863 	    length4 length, cred_t *cred, nfs_resop4 *resop)
7864 {
7865 	nfsstat4 status;
7866 	rfs4_lockowner_t *lo = lp->locker;
7867 	rfs4_state_t *sp = lp->state;
7868 	struct flock64 flock;
7869 	int16_t ltype;
7870 	int flag;
7871 	int error;
7872 	sysid_t sysid;
7873 	LOCK4res *lres;
7874 
7875 	if (rfs4_lease_expired(lo->client)) {
7876 		return (NFS4ERR_EXPIRED);
7877 	}
7878 
7879 	if ((status = rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK)
7880 		return (status);
7881 
7882 	/* Check for zero length. To lock to end of file use all ones for V4 */
7883 	if (length == 0)
7884 		return (NFS4ERR_INVAL);
7885 	else if (length == (length4)(~0))
7886 		length = 0;		/* Posix to end of file  */
7887 
7888 retry:
7889 	rfs4_dbe_lock(sp->dbe);
7890 
7891 
7892 	if (resop->resop != OP_LOCKU) {
7893 		switch (locktype) {
7894 		case READ_LT:
7895 		case READW_LT:
7896 			if ((sp->share_access
7897 			    & OPEN4_SHARE_ACCESS_READ) == 0) {
7898 				rfs4_dbe_unlock(sp->dbe);
7899 
7900 				return (NFS4ERR_OPENMODE);
7901 			}
7902 			ltype = F_RDLCK;
7903 			break;
7904 		case WRITE_LT:
7905 		case WRITEW_LT:
7906 			if ((sp->share_access
7907 			    & OPEN4_SHARE_ACCESS_WRITE) == 0) {
7908 				rfs4_dbe_unlock(sp->dbe);
7909 
7910 				return (NFS4ERR_OPENMODE);
7911 			}
7912 			ltype = F_WRLCK;
7913 			break;
7914 		}
7915 	} else
7916 		ltype = F_UNLCK;
7917 
7918 	flock.l_type = ltype;
7919 	flock.l_whence = 0;		/* SEEK_SET */
7920 	flock.l_start = offset;
7921 	flock.l_len = length;
7922 	flock.l_sysid = sysid;
7923 	flock.l_pid = lp->locker->pid;
7924 
7925 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
7926 	if (flock.l_len < 0 || flock.l_start < 0) {
7927 		rfs4_dbe_unlock(sp->dbe);
7928 		return (NFS4ERR_INVAL);
7929 	}
7930 
7931 	/*
7932 	 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and
7933 	 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE.
7934 	 */
7935 	flag = (int)sp->share_access | F_REMOTELOCK;
7936 
7937 	error = setlock(sp->finfo->vp, &flock, flag, cred);
7938 	if (error == 0) {
7939 		rfs4_dbe_lock(lp->dbe);
7940 		next_stateid(&lp->lockid);
7941 		rfs4_dbe_unlock(lp->dbe);
7942 	}
7943 
7944 	rfs4_dbe_unlock(sp->dbe);
7945 
7946 	/*
7947 	 * N.B. We map error values to nfsv4 errors. This is differrent
7948 	 * than puterrno4 routine.
7949 	 */
7950 	switch (error) {
7951 	case 0:
7952 		status = NFS4_OK;
7953 		break;
7954 	case EAGAIN:
7955 	case EACCES:		/* Old value */
7956 		/* Can only get here if op is OP_LOCK */
7957 		ASSERT(resop->resop == OP_LOCK);
7958 		lres = &resop->nfs_resop4_u.oplock;
7959 		status = NFS4ERR_DENIED;
7960 		if (lock_denied(&lres->LOCK4res_u.denied, &flock)
7961 		    == NFS4ERR_EXPIRED)
7962 			goto retry;
7963 		break;
7964 	case ENOLCK:
7965 		status = NFS4ERR_DELAY;
7966 		break;
7967 	case EOVERFLOW:
7968 		status = NFS4ERR_INVAL;
7969 		break;
7970 	case EINVAL:
7971 		status = NFS4ERR_NOTSUPP;
7972 		break;
7973 	default:
7974 		cmn_err(CE_WARN, "rfs4_do_lock: unexpected errno (%d)", error);
7975 		status = NFS4ERR_SERVERFAULT;
7976 		break;
7977 	}
7978 
7979 	return (status);
7980 }
7981 
7982 /*ARGSUSED*/
7983 void
7984 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop,
7985 	    struct svc_req *req, struct compound_state *cs)
7986 {
7987 	/* XXX Currently not using req arg */
7988 	LOCK4args *args = &argop->nfs_argop4_u.oplock;
7989 	LOCK4res *resp = &resop->nfs_resop4_u.oplock;
7990 	nfsstat4 status;
7991 	stateid4 *stateid;
7992 	rfs4_lockowner_t *lo;
7993 	rfs4_client_t *cp;
7994 	rfs4_state_t *sp = NULL;
7995 	rfs4_lo_state_t *lsp = NULL;
7996 	bool_t ls_sw_held = FALSE;
7997 	bool_t create = TRUE;
7998 	bool_t lcreate = TRUE;
7999 	bool_t dup_lock = FALSE;
8000 	int rc;
8001 
8002 	if (cs->vp == NULL) {
8003 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8004 		return;
8005 	}
8006 
8007 	if (args->locker.new_lock_owner) {
8008 		/* Create a new lockowner for this instance */
8009 		open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner;
8010 
8011 		NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner"));
8012 
8013 		stateid = &olo->open_stateid;
8014 		status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID);
8015 		if (status != NFS4_OK) {
8016 			NFS4_DEBUG(rfs4_debug,
8017 			    (CE_NOTE, "Get state failed in lock %d", status));
8018 			*cs->statusp = resp->status = status;
8019 			return;
8020 		}
8021 
8022 		/* Ensure specified filehandle matches */
8023 		if (cs->vp != sp->finfo->vp) {
8024 			rfs4_state_rele(sp);
8025 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8026 			return;
8027 		}
8028 
8029 		/* hold off other access to open_owner while we tinker */
8030 		rfs4_sw_enter(&sp->owner->oo_sw);
8031 
8032 		switch (rc = rfs4_check_stateid_seqid(sp, stateid)) {
8033 		case NFS4_CHECK_STATEID_OLD:
8034 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8035 			goto end;
8036 		case NFS4_CHECK_STATEID_BAD:
8037 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8038 			goto end;
8039 		case NFS4_CHECK_STATEID_EXPIRED:
8040 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8041 			goto end;
8042 		case NFS4_CHECK_STATEID_UNCONFIRMED:
8043 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8044 			goto end;
8045 		case NFS4_CHECK_STATEID_CLOSED:
8046 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8047 			goto end;
8048 		case NFS4_CHECK_STATEID_OKAY:
8049 		case NFS4_CHECK_STATEID_REPLAY:
8050 			switch (rfs4_check_olo_seqid(olo->open_seqid,
8051 			    sp->owner, resop)) {
8052 			case NFS4_CHKSEQ_OKAY:
8053 				if (rc == NFS4_CHECK_STATEID_OKAY)
8054 					break;
8055 				/*
8056 				 * This is replayed stateid; if seqid
8057 				 * matches next expected, then client
8058 				 * is using wrong seqid.
8059 				 */
8060 				/* FALLTHROUGH */
8061 			case NFS4_CHKSEQ_BAD:
8062 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8063 				goto end;
8064 			case NFS4_CHKSEQ_REPLAY:
8065 				/* This is a duplicate LOCK request */
8066 				dup_lock = TRUE;
8067 
8068 				/*
8069 				 * For a duplicate we do not want to
8070 				 * create a new lockowner as it should
8071 				 * already exist.
8072 				 * Turn off the lockowner create flag.
8073 				 */
8074 				lcreate = FALSE;
8075 			}
8076 			break;
8077 		}
8078 
8079 		lo = rfs4_findlockowner(&olo->lock_owner, &lcreate);
8080 		if (lo == NULL) {
8081 			NFS4_DEBUG(rfs4_debug,
8082 			    (CE_NOTE, "rfs4_op_lock: no lock owner"));
8083 			*cs->statusp = resp->status = NFS4ERR_RESOURCE;
8084 			goto end;
8085 		}
8086 
8087 		lsp = rfs4_findlo_state_by_owner(lo, sp, &create);
8088 		if (lsp == NULL) {
8089 			rfs4_update_lease(sp->owner->client);
8090 			/*
8091 			 * Only update theh open_seqid if this is not
8092 			 * a duplicate request
8093 			 */
8094 			if (dup_lock == FALSE) {
8095 				rfs4_update_open_sequence(sp->owner);
8096 			}
8097 
8098 			NFS4_DEBUG(rfs4_debug,
8099 			    (CE_NOTE, "rfs4_op_lock: no state"));
8100 			*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
8101 			rfs4_update_open_resp(sp->owner, resop, NULL);
8102 			rfs4_lockowner_rele(lo);
8103 			goto end;
8104 		}
8105 
8106 		/*
8107 		 * This is the new_lock_owner branch and the client is
8108 		 * supposed to be associating a new lock_owner with
8109 		 * the open file at this point.  If we find that a
8110 		 * lock_owner/state association already exists and a
8111 		 * successful LOCK request was returned to the client,
8112 		 * an error is returned to the client since this is
8113 		 * not appropriate.  The client should be using the
8114 		 * existing lock_owner branch.
8115 		 */
8116 		if (dup_lock == FALSE && create == FALSE) {
8117 			if (lsp->lock_completed == TRUE) {
8118 				*cs->statusp =
8119 				    resp->status = NFS4ERR_BAD_SEQID;
8120 				rfs4_lockowner_rele(lo);
8121 				goto end;
8122 			}
8123 		}
8124 
8125 		rfs4_update_lease(sp->owner->client);
8126 
8127 		/*
8128 		 * Only update theh open_seqid if this is not
8129 		 * a duplicate request
8130 		 */
8131 		if (dup_lock == FALSE) {
8132 			rfs4_update_open_sequence(sp->owner);
8133 		}
8134 
8135 		/*
8136 		 * If this is a duplicate lock request, just copy the
8137 		 * previously saved reply and return.
8138 		 */
8139 		if (dup_lock == TRUE) {
8140 			/* verify that lock_seqid's match */
8141 			if (lsp->seqid != olo->lock_seqid) {
8142 				NFS4_DEBUG(rfs4_debug,
8143 				    (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad"
8144 				    "lsp->seqid=%d old->seqid=%d",
8145 				    lsp->seqid, olo->lock_seqid));
8146 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8147 			} else {
8148 				rfs4_copy_reply(resop, lsp->reply);
8149 				/*
8150 				 * Make sure to copy the just
8151 				 * retrieved reply status into the
8152 				 * overall compound status
8153 				 */
8154 				*cs->statusp = resp->status;
8155 			}
8156 			rfs4_lockowner_rele(lo);
8157 			goto end;
8158 		}
8159 
8160 		rfs4_dbe_lock(lsp->dbe);
8161 
8162 		/* Make sure to update the lock sequence id */
8163 		lsp->seqid = olo->lock_seqid;
8164 
8165 		NFS4_DEBUG(rfs4_debug,
8166 		    (CE_NOTE, "Lock seqid established as %d", lsp->seqid));
8167 
8168 		/*
8169 		 * This is used to signify the newly created lockowner
8170 		 * stateid and its sequence number.  The checks for
8171 		 * sequence number and increment don't occur on the
8172 		 * very first lock request for a lockowner.
8173 		 */
8174 		lsp->skip_seqid_check = TRUE;
8175 
8176 		/* hold off other access to lsp while we tinker */
8177 		rfs4_sw_enter(&lsp->ls_sw);
8178 		ls_sw_held = TRUE;
8179 
8180 		rfs4_dbe_unlock(lsp->dbe);
8181 
8182 		rfs4_lockowner_rele(lo);
8183 	} else {
8184 		stateid = &args->locker.locker4_u.lock_owner.lock_stateid;
8185 		/* get lsp and hold the lock on the underlying file struct */
8186 		if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE))
8187 		    != NFS4_OK) {
8188 			*cs->statusp = resp->status = status;
8189 			return;
8190 		}
8191 		create = FALSE;	/* We didn't create lsp */
8192 
8193 		/* Ensure specified filehandle matches */
8194 		if (cs->vp != lsp->state->finfo->vp) {
8195 			rfs4_lo_state_rele(lsp, TRUE);
8196 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8197 			return;
8198 		}
8199 
8200 		/* hold off other access to lsp while we tinker */
8201 		rfs4_sw_enter(&lsp->ls_sw);
8202 		ls_sw_held = TRUE;
8203 
8204 		switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
8205 		/*
8206 		 * The stateid looks like it was okay (expected to be
8207 		 * the next one)
8208 		 */
8209 		case NFS4_CHECK_STATEID_OKAY:
8210 			/*
8211 			 * The sequence id is now checked.  Determine
8212 			 * if this is a replay or if it is in the
8213 			 * expected (next) sequence.  In the case of a
8214 			 * replay, there are two replay conditions
8215 			 * that may occur.  The first is the normal
8216 			 * condition where a LOCK is done with a
8217 			 * NFS4_OK response and the stateid is
8218 			 * updated.  That case is handled below when
8219 			 * the stateid is identified as a REPLAY.  The
8220 			 * second is the case where an error is
8221 			 * returned, like NFS4ERR_DENIED, and the
8222 			 * sequence number is updated but the stateid
8223 			 * is not updated.  This second case is dealt
8224 			 * with here.  So it may seem odd that the
8225 			 * stateid is okay but the sequence id is a
8226 			 * replay but it is okay.
8227 			 */
8228 			switch (rfs4_check_lock_seqid(
8229 			    args->locker.locker4_u.lock_owner.lock_seqid,
8230 			    lsp, resop)) {
8231 			case NFS4_CHKSEQ_REPLAY:
8232 				if (resp->status != NFS4_OK) {
8233 					/*
8234 					 * Here is our replay and need
8235 					 * to verify that the last
8236 					 * response was an error.
8237 					 */
8238 					*cs->statusp = resp->status;
8239 					goto end;
8240 				}
8241 				/*
8242 				 * This is done since the sequence id
8243 				 * looked like a replay but it didn't
8244 				 * pass our check so a BAD_SEQID is
8245 				 * returned as a result.
8246 				 */
8247 				/*FALLTHROUGH*/
8248 			case NFS4_CHKSEQ_BAD:
8249 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8250 				goto end;
8251 			case NFS4_CHKSEQ_OKAY:
8252 				/* Everything looks okay move ahead */
8253 				break;
8254 			}
8255 			break;
8256 		case NFS4_CHECK_STATEID_OLD:
8257 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8258 			goto end;
8259 		case NFS4_CHECK_STATEID_BAD:
8260 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8261 			goto end;
8262 		case NFS4_CHECK_STATEID_EXPIRED:
8263 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8264 			goto end;
8265 		case NFS4_CHECK_STATEID_CLOSED:
8266 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8267 			goto end;
8268 		case NFS4_CHECK_STATEID_REPLAY:
8269 			switch (rfs4_check_lock_seqid(
8270 			    args->locker.locker4_u.lock_owner.lock_seqid,
8271 			    lsp, resop)) {
8272 			case NFS4_CHKSEQ_OKAY:
8273 				/*
8274 				 * This is a replayed stateid; if
8275 				 * seqid matches the next expected,
8276 				 * then client is using wrong seqid.
8277 				 */
8278 			case NFS4_CHKSEQ_BAD:
8279 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8280 				goto end;
8281 			case NFS4_CHKSEQ_REPLAY:
8282 				rfs4_update_lease(lsp->locker->client);
8283 				*cs->statusp = status = resp->status;
8284 				goto end;
8285 			}
8286 			break;
8287 		default:
8288 			ASSERT(FALSE);
8289 			break;
8290 		}
8291 
8292 		rfs4_update_lock_sequence(lsp);
8293 		rfs4_update_lease(lsp->locker->client);
8294 	}
8295 
8296 	/*
8297 	 * NFS4 only allows locking on regular files, so
8298 	 * verify type of object.
8299 	 */
8300 	if (cs->vp->v_type != VREG) {
8301 		if (cs->vp->v_type == VDIR)
8302 			status = NFS4ERR_ISDIR;
8303 		else
8304 			status = NFS4ERR_INVAL;
8305 		goto out;
8306 	}
8307 
8308 	cp = lsp->state->owner->client;
8309 
8310 	if (rfs4_clnt_in_grace(cp) && !args->reclaim) {
8311 		status = NFS4ERR_GRACE;
8312 		goto out;
8313 	}
8314 
8315 	if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->can_reclaim) {
8316 		status = NFS4ERR_NO_GRACE;
8317 		goto out;
8318 	}
8319 
8320 	if (!rfs4_clnt_in_grace(cp) && args->reclaim) {
8321 		status = NFS4ERR_NO_GRACE;
8322 		goto out;
8323 	}
8324 
8325 	if (lsp->state->finfo->dinfo->dtype == OPEN_DELEGATE_WRITE)
8326 		cs->deleg = TRUE;
8327 
8328 	status = rfs4_do_lock(lsp, args->locktype,
8329 	    args->locker.locker4_u.lock_owner.lock_seqid,
8330 	    args->offset, args->length, cs->cr, resop);
8331 
8332 out:
8333 	lsp->skip_seqid_check = FALSE;
8334 
8335 	*cs->statusp = resp->status = status;
8336 
8337 	if (status == NFS4_OK) {
8338 		resp->LOCK4res_u.lock_stateid = lsp->lockid.stateid;
8339 		lsp->lock_completed = TRUE;
8340 	}
8341 	/*
8342 	 * Only update the "OPEN" response here if this was a new
8343 	 * lock_owner
8344 	 */
8345 	if (sp)
8346 		rfs4_update_open_resp(sp->owner, resop, NULL);
8347 
8348 	rfs4_update_lock_resp(lsp, resop);
8349 
8350 end:
8351 	if (lsp) {
8352 		if (ls_sw_held)
8353 			rfs4_sw_exit(&lsp->ls_sw);
8354 		/*
8355 		 * If an sp obtained, then the lsp does not represent
8356 		 * a lock on the file struct.
8357 		 */
8358 		if (sp != NULL)
8359 			rfs4_lo_state_rele(lsp, FALSE);
8360 		else
8361 			rfs4_lo_state_rele(lsp, TRUE);
8362 	}
8363 	if (sp) {
8364 		rfs4_sw_exit(&sp->owner->oo_sw);
8365 		rfs4_state_rele(sp);
8366 	}
8367 }
8368 
8369 /* free function for LOCK/LOCKT */
8370 static void
8371 lock_denied_free(nfs_resop4 *resop)
8372 {
8373 	LOCK4denied *dp = NULL;
8374 
8375 	switch (resop->resop) {
8376 	case OP_LOCK:
8377 		if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED)
8378 			dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied;
8379 		break;
8380 	case OP_LOCKT:
8381 		if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED)
8382 			dp = &resop->nfs_resop4_u.oplockt.denied;
8383 		break;
8384 	default:
8385 		break;
8386 	}
8387 
8388 	if (dp)
8389 		kmem_free(dp->owner.owner_val, dp->owner.owner_len);
8390 }
8391 
8392 /*ARGSUSED*/
8393 void
8394 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop,
8395 	    struct svc_req *req, struct compound_state *cs)
8396 {
8397 	/* XXX Currently not using req arg */
8398 	LOCKU4args *args = &argop->nfs_argop4_u.oplocku;
8399 	LOCKU4res *resp = &resop->nfs_resop4_u.oplocku;
8400 	nfsstat4 status;
8401 	stateid4 *stateid = &args->lock_stateid;
8402 	rfs4_lo_state_t *lsp;
8403 
8404 	if (cs->vp == NULL) {
8405 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8406 		return;
8407 	}
8408 
8409 	if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) {
8410 		*cs->statusp = resp->status = status;
8411 		return;
8412 	}
8413 
8414 	/* Ensure specified filehandle matches */
8415 	if (cs->vp != lsp->state->finfo->vp) {
8416 		rfs4_lo_state_rele(lsp, TRUE);
8417 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8418 		return;
8419 	}
8420 
8421 	/* hold off other access to lsp while we tinker */
8422 	rfs4_sw_enter(&lsp->ls_sw);
8423 
8424 	switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
8425 	case NFS4_CHECK_STATEID_OKAY:
8426 		if (rfs4_check_lock_seqid(args->seqid, lsp, resop)
8427 		    != NFS4_CHKSEQ_OKAY) {
8428 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8429 			goto end;
8430 		}
8431 		break;
8432 	case NFS4_CHECK_STATEID_OLD:
8433 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8434 		goto end;
8435 	case NFS4_CHECK_STATEID_BAD:
8436 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8437 		goto end;
8438 	case NFS4_CHECK_STATEID_EXPIRED:
8439 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8440 		goto end;
8441 	case NFS4_CHECK_STATEID_CLOSED:
8442 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8443 		goto end;
8444 	case NFS4_CHECK_STATEID_REPLAY:
8445 		switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) {
8446 		case NFS4_CHKSEQ_OKAY:
8447 				/*
8448 				 * This is a replayed stateid; if
8449 				 * seqid matches the next expected,
8450 				 * then client is using wrong seqid.
8451 				 */
8452 		case NFS4_CHKSEQ_BAD:
8453 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8454 			goto end;
8455 		case NFS4_CHKSEQ_REPLAY:
8456 			rfs4_update_lease(lsp->locker->client);
8457 			*cs->statusp = status = resp->status;
8458 			goto end;
8459 		}
8460 		break;
8461 	default:
8462 		ASSERT(FALSE);
8463 		break;
8464 	}
8465 
8466 	rfs4_update_lock_sequence(lsp);
8467 	rfs4_update_lease(lsp->locker->client);
8468 
8469 	/*
8470 	 * NFS4 only allows locking on regular files, so
8471 	 * verify type of object.
8472 	 */
8473 	if (cs->vp->v_type != VREG) {
8474 		if (cs->vp->v_type == VDIR)
8475 			status = NFS4ERR_ISDIR;
8476 		else
8477 			status = NFS4ERR_INVAL;
8478 		goto out;
8479 	}
8480 
8481 	if (rfs4_clnt_in_grace(lsp->state->owner->client)) {
8482 		status = NFS4ERR_GRACE;
8483 		goto out;
8484 	}
8485 
8486 	status = rfs4_do_lock(lsp, args->locktype,
8487 	    args->seqid, args->offset, args->length, cs->cr, resop);
8488 
8489 out:
8490 	*cs->statusp = resp->status = status;
8491 
8492 	if (status == NFS4_OK)
8493 		resp->lock_stateid = lsp->lockid.stateid;
8494 
8495 	rfs4_update_lock_resp(lsp, resop);
8496 
8497 end:
8498 	rfs4_sw_exit(&lsp->ls_sw);
8499 	rfs4_lo_state_rele(lsp, TRUE);
8500 }
8501 
8502 /*
8503  * LOCKT is a best effort routine, the client can not be guaranteed that
8504  * the status return is still in effect by the time the reply is received.
8505  * They are numerous race conditions in this routine, but we are not required
8506  * and can not be accurate.
8507  */
8508 /*ARGSUSED*/
8509 void
8510 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop,
8511 	    struct svc_req *req, struct compound_state *cs)
8512 {
8513 	LOCKT4args *args = &argop->nfs_argop4_u.oplockt;
8514 	LOCKT4res *resp = &resop->nfs_resop4_u.oplockt;
8515 	rfs4_lockowner_t *lo;
8516 	rfs4_client_t *cp;
8517 	bool_t create = FALSE;
8518 	struct flock64 flk;
8519 	int error;
8520 	int flag = FREAD | FWRITE;
8521 	int ltype;
8522 	length4 posix_length;
8523 	sysid_t sysid;
8524 	pid_t pid;
8525 
8526 	if (cs->vp == NULL) {
8527 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8528 		return;
8529 	}
8530 
8531 	/*
8532 	 * NFS4 only allows locking on regular files, so
8533 	 * verify type of object.
8534 	 */
8535 	if (cs->vp->v_type != VREG) {
8536 		if (cs->vp->v_type == VDIR)
8537 			*cs->statusp = resp->status = NFS4ERR_ISDIR;
8538 		else
8539 			*cs->statusp = resp->status =  NFS4ERR_INVAL;
8540 		return;
8541 	}
8542 
8543 	/*
8544 	 * Check out the clientid to ensure the server knows about it
8545 	 * so that we correctly inform the client of a server reboot.
8546 	 */
8547 	if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE))
8548 	    == NULL) {
8549 		*cs->statusp = resp->status =
8550 		    rfs4_check_clientid(&args->owner.clientid, 0);
8551 		return;
8552 	}
8553 	if (rfs4_lease_expired(cp)) {
8554 		rfs4_client_close(cp);
8555 		/*
8556 		 * Protocol doesn't allow returning NFS4ERR_STALE as
8557 		 * other operations do on this check so STALE_CLIENTID
8558 		 * is returned instead
8559 		 */
8560 		*cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
8561 		return;
8562 	}
8563 
8564 	if (rfs4_clnt_in_grace(cp)) {
8565 		*cs->statusp = resp->status = NFS4ERR_GRACE;
8566 		return;
8567 	}
8568 	rfs4_client_rele(cp);
8569 
8570 	resp->status = NFS4_OK;
8571 
8572 	switch (args->locktype) {
8573 	case READ_LT:
8574 	case READW_LT:
8575 		ltype = F_RDLCK;
8576 		break;
8577 	case WRITE_LT:
8578 	case WRITEW_LT:
8579 		ltype = F_WRLCK;
8580 		break;
8581 	}
8582 
8583 	posix_length = args->length;
8584 	/* Check for zero length. To lock to end of file use all ones for V4 */
8585 	if (posix_length == 0) {
8586 		*cs->statusp = resp->status = NFS4ERR_INVAL;
8587 		return;
8588 	} else if (posix_length == (length4)(~0)) {
8589 		posix_length = 0;	/* Posix to end of file  */
8590 	}
8591 
8592 	/* Find or create a lockowner */
8593 	lo = rfs4_findlockowner(&args->owner, &create);
8594 
8595 	if (lo) {
8596 		pid = lo->pid;
8597 		if ((resp->status =
8598 		    rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK)
8599 		goto out;
8600 	} else {
8601 		pid = 0;
8602 		sysid = lockt_sysid;
8603 	}
8604 retry:
8605 	flk.l_type = ltype;
8606 	flk.l_whence = 0;		/* SEEK_SET */
8607 	flk.l_start = args->offset;
8608 	flk.l_len = posix_length;
8609 	flk.l_sysid = sysid;
8610 	flk.l_pid = pid;
8611 	flag |= F_REMOTELOCK;
8612 
8613 	LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk);
8614 
8615 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
8616 	if (flk.l_len < 0 || flk.l_start < 0) {
8617 		resp->status = NFS4ERR_INVAL;
8618 		goto out;
8619 	}
8620 	error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0,
8621 	    NULL, cs->cr);
8622 
8623 	/*
8624 	 * N.B. We map error values to nfsv4 errors. This is differrent
8625 	 * than puterrno4 routine.
8626 	 */
8627 	switch (error) {
8628 	case 0:
8629 		if (flk.l_type == F_UNLCK)
8630 			resp->status = NFS4_OK;
8631 		else {
8632 			if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED)
8633 				goto retry;
8634 			resp->status = NFS4ERR_DENIED;
8635 		}
8636 		break;
8637 	case EOVERFLOW:
8638 		resp->status = NFS4ERR_INVAL;
8639 		break;
8640 	case EINVAL:
8641 		resp->status = NFS4ERR_NOTSUPP;
8642 		break;
8643 	default:
8644 		cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)",
8645 		    error);
8646 		resp->status = NFS4ERR_SERVERFAULT;
8647 		break;
8648 	}
8649 
8650 out:
8651 	if (lo)
8652 		rfs4_lockowner_rele(lo);
8653 	*cs->statusp = resp->status;
8654 }
8655 
8656 static int
8657 vop_shrlock(vnode_t *vp, int cmd, struct shrlock *sp, int fflags)
8658 {
8659 	int err;
8660 
8661 	if (cmd == F_UNSHARE && sp->s_deny == 0 && sp->s_access == 0)
8662 		return (0);
8663 
8664 	err = VOP_SHRLOCK(vp, cmd, sp, fflags, CRED());
8665 
8666 	NFS4_DEBUG(rfs4_shrlock_debug,
8667 	    (CE_NOTE, "rfs4_shrlock %s vp=%p acc=%d dny=%d sysid=%d "
8668 	    "pid=%d err=%d\n", cmd == F_SHARE ? "SHARE" : "UNSHR",
8669 	    (void *) vp, sp->s_access, sp->s_deny, sp->s_sysid, sp->s_pid,
8670 	    err));
8671 
8672 	return (err);
8673 }
8674 
8675 static int
8676 rfs4_shrlock(rfs4_state_t *sp, int cmd)
8677 {
8678 	struct shrlock shr;
8679 	struct shr_locowner shr_loco;
8680 	int fflags;
8681 
8682 	fflags = shr.s_access = shr.s_deny = 0;
8683 
8684 	if (sp->share_access & OPEN4_SHARE_ACCESS_READ) {
8685 		fflags |= FREAD;
8686 		shr.s_access |= F_RDACC;
8687 	}
8688 	if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) {
8689 		fflags |= FWRITE;
8690 		shr.s_access |= F_WRACC;
8691 	}
8692 	if (sp->share_deny & OPEN4_SHARE_DENY_READ)
8693 		shr.s_deny |= F_RDDNY;
8694 	if (sp->share_deny & OPEN4_SHARE_DENY_WRITE)
8695 		shr.s_deny |= F_WRDNY;
8696 
8697 	shr.s_pid = rfs4_dbe_getid(sp->owner->dbe);
8698 	shr.s_sysid = sp->owner->client->sysidt;
8699 	shr_loco.sl_pid = shr.s_pid;
8700 	shr_loco.sl_id = shr.s_sysid;
8701 	shr.s_owner = (caddr_t)&shr_loco;
8702 	shr.s_own_len = sizeof (shr_loco);
8703 	return (vop_shrlock(sp->finfo->vp, cmd, &shr, fflags));
8704 }
8705 
8706 static int
8707 rfs4_share(rfs4_state_t *sp)
8708 {
8709 	return (rfs4_shrlock(sp, F_SHARE));
8710 }
8711 
8712 void
8713 rfs4_unshare(rfs4_state_t *sp)
8714 {
8715 	(void) rfs4_shrlock(sp, F_UNSHARE);
8716 }
8717