xref: /titanic_44/usr/src/uts/common/fs/nfs/nfs4_srv.c (revision ee0eb9f2d12b8a03b5a193cd667735c1962e0c7b)
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 2008 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_delegpurge(nfs_argop4 *, nfs_resop4 *,
170 			struct svc_req *, struct compound_state *);
171 static void	rfs4_op_getattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
172 			struct compound_state *);
173 static void	rfs4_op_getattr_free(nfs_resop4 *);
174 static void	rfs4_op_getfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
175 			struct compound_state *);
176 static void	rfs4_op_getfh_free(nfs_resop4 *);
177 static void	rfs4_op_illegal(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
178 			struct compound_state *);
179 static void	rfs4_op_link(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
180 			struct compound_state *);
181 static void	rfs4_op_lock(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
182 			struct compound_state *);
183 static void	lock_denied_free(nfs_resop4 *);
184 static void	rfs4_op_locku(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
185 			struct compound_state *);
186 static void	rfs4_op_lockt(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
187 			struct compound_state *);
188 static void	rfs4_op_lookup(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
189 			struct compound_state *);
190 static void	rfs4_op_lookupp(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
191 			struct compound_state *);
192 static void	rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop,
193 				struct svc_req *req, struct compound_state *cs);
194 static void	rfs4_op_nverify(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
195 			struct compound_state *);
196 static void	rfs4_op_open(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
197 			struct compound_state *);
198 static void	rfs4_op_open_confirm(nfs_argop4 *, nfs_resop4 *,
199 			struct svc_req *, struct compound_state *);
200 static void	rfs4_op_open_downgrade(nfs_argop4 *, nfs_resop4 *,
201 			struct svc_req *, struct compound_state *);
202 static void	rfs4_op_putfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
203 			struct compound_state *);
204 static void	rfs4_op_putpubfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
205 			struct compound_state *);
206 static void	rfs4_op_putrootfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
207 			struct compound_state *);
208 static void	rfs4_op_read(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
209 			struct compound_state *);
210 static void	rfs4_op_read_free(nfs_resop4 *);
211 static void	rfs4_op_readdir_free(nfs_resop4 *resop);
212 static void	rfs4_op_readlink(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
213 			struct compound_state *);
214 static void	rfs4_op_readlink_free(nfs_resop4 *);
215 static void	rfs4_op_release_lockowner(nfs_argop4 *, nfs_resop4 *,
216 			struct svc_req *, struct compound_state *);
217 static void	rfs4_op_remove(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
218 			struct compound_state *);
219 static void	rfs4_op_rename(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
220 			struct compound_state *);
221 static void	rfs4_op_renew(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
222 			struct compound_state *);
223 static void	rfs4_op_restorefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
224 			struct compound_state *);
225 static void	rfs4_op_savefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
226 			struct compound_state *);
227 static void	rfs4_op_setattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
228 			struct compound_state *);
229 static void	rfs4_op_verify(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
230 			struct compound_state *);
231 static void	rfs4_op_write(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
232 			struct compound_state *);
233 static void	rfs4_op_setclientid(nfs_argop4 *, nfs_resop4 *,
234 			struct svc_req *, struct compound_state *);
235 static void	rfs4_op_setclientid_confirm(nfs_argop4 *, nfs_resop4 *,
236 			struct svc_req *req, struct compound_state *);
237 static void	rfs4_op_secinfo(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
238 			struct compound_state *);
239 static void	rfs4_op_secinfo_free(nfs_resop4 *);
240 
241 static nfsstat4 check_open_access(uint32_t,
242 				struct compound_state *, struct svc_req *);
243 nfsstat4 rfs4_client_sysid(rfs4_client_t *, sysid_t *);
244 static int	vop_shrlock(vnode_t *, int, struct shrlock *, int);
245 static int 	rfs4_shrlock(rfs4_state_t *, int);
246 static int	rfs4_share(rfs4_state_t *);
247 void rfs4_ss_clid(rfs4_client_t *, struct svc_req *);
248 
249 /*
250  * translation table for attrs
251  */
252 struct nfs4_ntov_table {
253 	union nfs4_attr_u *na;
254 	uint8_t amap[NFS4_MAXNUM_ATTRS];
255 	int attrcnt;
256 	bool_t vfsstat;
257 };
258 
259 static void	nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp);
260 static void	nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp,
261 				    struct nfs4_svgetit_arg *sargp);
262 
263 static nfsstat4	do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp,
264 		    struct compound_state *cs, struct nfs4_svgetit_arg *sargp,
265 		    struct nfs4_ntov_table *ntovp, nfs4_attr_cmd_t cmd);
266 
267 fem_t	*deleg_rdops;
268 fem_t	*deleg_wrops;
269 
270 rfs4_servinst_t	*rfs4_cur_servinst = NULL;	/* current server instance */
271 kmutex_t	rfs4_servinst_lock;		/* protects linked list */
272 int		rfs4_seen_first_compound;	/* set first time we see one */
273 
274 /*
275  * NFS4 op dispatch table
276  */
277 
278 struct rfsv4disp {
279 	void	(*dis_proc)();		/* proc to call */
280 	void	(*dis_resfree)();	/* frees space allocated by proc */
281 	int	dis_flags;		/* RPC_IDEMPOTENT, etc... */
282 };
283 
284 static struct rfsv4disp rfsv4disptab[] = {
285 	/*
286 	 * NFS VERSION 4
287 	 */
288 
289 	/* RFS_NULL = 0 */
290 	{rfs4_op_illegal, nullfree, 0},
291 
292 	/* UNUSED = 1 */
293 	{rfs4_op_illegal, nullfree, 0},
294 
295 	/* UNUSED = 2 */
296 	{rfs4_op_illegal, nullfree, 0},
297 
298 	/* OP_ACCESS = 3 */
299 	{rfs4_op_access, nullfree, RPC_IDEMPOTENT},
300 
301 	/* OP_CLOSE = 4 */
302 	{rfs4_op_close, nullfree, 0},
303 
304 	/* OP_COMMIT = 5 */
305 	{rfs4_op_commit, nullfree, RPC_IDEMPOTENT},
306 
307 	/* OP_CREATE = 6 */
308 	{rfs4_op_create, nullfree, 0},
309 
310 	/* OP_DELEGPURGE = 7 */
311 	{rfs4_op_delegpurge, nullfree, 0},
312 
313 	/* OP_DELEGRETURN = 8 */
314 	{rfs4_op_delegreturn, nullfree, 0},
315 
316 	/* OP_GETATTR = 9 */
317 	{rfs4_op_getattr, rfs4_op_getattr_free, RPC_IDEMPOTENT},
318 
319 	/* OP_GETFH = 10 */
320 	{rfs4_op_getfh, rfs4_op_getfh_free, RPC_ALL},
321 
322 	/* OP_LINK = 11 */
323 	{rfs4_op_link, nullfree, 0},
324 
325 	/* OP_LOCK = 12 */
326 	{rfs4_op_lock, lock_denied_free, 0},
327 
328 	/* OP_LOCKT = 13 */
329 	{rfs4_op_lockt, lock_denied_free, 0},
330 
331 	/* OP_LOCKU = 14 */
332 	{rfs4_op_locku, nullfree, 0},
333 
334 	/* OP_LOOKUP = 15 */
335 	{rfs4_op_lookup, nullfree, (RPC_IDEMPOTENT|RPC_PUBLICFH_OK)},
336 
337 	/* OP_LOOKUPP = 16 */
338 	{rfs4_op_lookupp, nullfree, (RPC_IDEMPOTENT|RPC_PUBLICFH_OK)},
339 
340 	/* OP_NVERIFY = 17 */
341 	{rfs4_op_nverify, nullfree, RPC_IDEMPOTENT},
342 
343 	/* OP_OPEN = 18 */
344 	{rfs4_op_open, rfs4_free_reply, 0},
345 
346 	/* OP_OPENATTR = 19 */
347 	{rfs4_op_openattr, nullfree, 0},
348 
349 	/* OP_OPEN_CONFIRM = 20 */
350 	{rfs4_op_open_confirm, nullfree, 0},
351 
352 	/* OP_OPEN_DOWNGRADE = 21 */
353 	{rfs4_op_open_downgrade, nullfree, 0},
354 
355 	/* OP_OPEN_PUTFH = 22 */
356 	{rfs4_op_putfh, nullfree, RPC_ALL},
357 
358 	/* OP_PUTPUBFH = 23 */
359 	{rfs4_op_putpubfh, nullfree, RPC_ALL},
360 
361 	/* OP_PUTROOTFH = 24 */
362 	{rfs4_op_putrootfh, nullfree, RPC_ALL},
363 
364 	/* OP_READ = 25 */
365 	{rfs4_op_read, rfs4_op_read_free, RPC_IDEMPOTENT},
366 
367 	/* OP_READDIR = 26 */
368 	{rfs4_op_readdir, rfs4_op_readdir_free, RPC_IDEMPOTENT},
369 
370 	/* OP_READLINK = 27 */
371 	{rfs4_op_readlink, rfs4_op_readlink_free, RPC_IDEMPOTENT},
372 
373 	/* OP_REMOVE = 28 */
374 	{rfs4_op_remove, nullfree, 0},
375 
376 	/* OP_RENAME = 29 */
377 	{rfs4_op_rename, nullfree, 0},
378 
379 	/* OP_RENEW = 30 */
380 	{rfs4_op_renew, nullfree, 0},
381 
382 	/* OP_RESTOREFH = 31 */
383 	{rfs4_op_restorefh, nullfree, RPC_ALL},
384 
385 	/* OP_SAVEFH = 32 */
386 	{rfs4_op_savefh, nullfree, RPC_ALL},
387 
388 	/* OP_SECINFO = 33 */
389 	{rfs4_op_secinfo, rfs4_op_secinfo_free, 0},
390 
391 	/* OP_SETATTR = 34 */
392 	{rfs4_op_setattr, nullfree, 0},
393 
394 	/* OP_SETCLIENTID = 35 */
395 	{rfs4_op_setclientid, nullfree, 0},
396 
397 	/* OP_SETCLIENTID_CONFIRM = 36 */
398 	{rfs4_op_setclientid_confirm, nullfree, 0},
399 
400 	/* OP_VERIFY = 37 */
401 	{rfs4_op_verify, nullfree, RPC_IDEMPOTENT},
402 
403 	/* OP_WRITE = 38 */
404 	{rfs4_op_write, nullfree, 0},
405 
406 	/* OP_RELEASE_LOCKOWNER = 39 */
407 	{rfs4_op_release_lockowner, nullfree, 0},
408 };
409 
410 static uint_t rfsv4disp_cnt = sizeof (rfsv4disptab) / sizeof (rfsv4disptab[0]);
411 
412 #define	OP_ILLEGAL_IDX (rfsv4disp_cnt)
413 
414 #ifdef DEBUG
415 
416 int rfs4_fillone_debug = 0;
417 int rfs4_shrlock_debug = 0;
418 int rfs4_no_stub_access = 1;
419 int rfs4_rddir_debug = 0;
420 
421 static char *rfs4_op_string[] = {
422 	"rfs4_op_null",
423 	"rfs4_op_1 unused",
424 	"rfs4_op_2 unused",
425 	"rfs4_op_access",
426 	"rfs4_op_close",
427 	"rfs4_op_commit",
428 	"rfs4_op_create",
429 	"rfs4_op_delegpurge",
430 	"rfs4_op_delegreturn",
431 	"rfs4_op_getattr",
432 	"rfs4_op_getfh",
433 	"rfs4_op_link",
434 	"rfs4_op_lock",
435 	"rfs4_op_lockt",
436 	"rfs4_op_locku",
437 	"rfs4_op_lookup",
438 	"rfs4_op_lookupp",
439 	"rfs4_op_nverify",
440 	"rfs4_op_open",
441 	"rfs4_op_openattr",
442 	"rfs4_op_open_confirm",
443 	"rfs4_op_open_downgrade",
444 	"rfs4_op_putfh",
445 	"rfs4_op_putpubfh",
446 	"rfs4_op_putrootfh",
447 	"rfs4_op_read",
448 	"rfs4_op_readdir",
449 	"rfs4_op_readlink",
450 	"rfs4_op_remove",
451 	"rfs4_op_rename",
452 	"rfs4_op_renew",
453 	"rfs4_op_restorefh",
454 	"rfs4_op_savefh",
455 	"rfs4_op_secinfo",
456 	"rfs4_op_setattr",
457 	"rfs4_op_setclientid",
458 	"rfs4_op_setclient_confirm",
459 	"rfs4_op_verify",
460 	"rfs4_op_write",
461 	"rfs4_op_release_lockowner",
462 	"rfs4_op_illegal"
463 };
464 #endif
465 
466 void rfs4_ss_chkclid(rfs4_client_t *);
467 
468 extern size_t strlcpy(char *dst, const char *src, size_t dstsize);
469 
470 #ifdef	nextdp
471 #undef nextdp
472 #endif
473 #define	nextdp(dp)	((struct dirent64 *)((char *)(dp) + (dp)->d_reclen))
474 
475 static const fs_operation_def_t nfs4_rd_deleg_tmpl[] = {
476 	VOPNAME_OPEN,		{ .femop_open = deleg_rd_open },
477 	VOPNAME_WRITE,		{ .femop_write = deleg_rd_write },
478 	VOPNAME_SETATTR,	{ .femop_setattr = deleg_rd_setattr },
479 	VOPNAME_RWLOCK,		{ .femop_rwlock = deleg_rd_rwlock },
480 	VOPNAME_SPACE,		{ .femop_space = deleg_rd_space },
481 	VOPNAME_SETSECATTR,	{ .femop_setsecattr = deleg_rd_setsecattr },
482 	VOPNAME_VNEVENT,	{ .femop_vnevent = deleg_rd_vnevent },
483 	NULL,			NULL
484 };
485 static const fs_operation_def_t nfs4_wr_deleg_tmpl[] = {
486 	VOPNAME_OPEN,		{ .femop_open = deleg_wr_open },
487 	VOPNAME_READ,		{ .femop_read = deleg_wr_read },
488 	VOPNAME_WRITE,		{ .femop_write = deleg_wr_write },
489 	VOPNAME_SETATTR,	{ .femop_setattr = deleg_wr_setattr },
490 	VOPNAME_RWLOCK,		{ .femop_rwlock = deleg_wr_rwlock },
491 	VOPNAME_SPACE,		{ .femop_space = deleg_wr_space },
492 	VOPNAME_SETSECATTR,	{ .femop_setsecattr = deleg_wr_setsecattr },
493 	VOPNAME_VNEVENT,	{ .femop_vnevent = deleg_wr_vnevent },
494 	NULL,			NULL
495 };
496 
497 int
498 rfs4_srvrinit(void)
499 {
500 	timespec32_t verf;
501 	int error;
502 	extern void rfs4_attr_init();
503 	extern krwlock_t rfs4_deleg_policy_lock;
504 
505 	/*
506 	 * The following algorithm attempts to find a unique verifier
507 	 * to be used as the write verifier returned from the server
508 	 * to the client.  It is important that this verifier change
509 	 * whenever the server reboots.  Of secondary importance, it
510 	 * is important for the verifier to be unique between two
511 	 * different servers.
512 	 *
513 	 * Thus, an attempt is made to use the system hostid and the
514 	 * current time in seconds when the nfssrv kernel module is
515 	 * loaded.  It is assumed that an NFS server will not be able
516 	 * to boot and then to reboot in less than a second.  If the
517 	 * hostid has not been set, then the current high resolution
518 	 * time is used.  This will ensure different verifiers each
519 	 * time the server reboots and minimize the chances that two
520 	 * different servers will have the same verifier.
521 	 * XXX - this is broken on LP64 kernels.
522 	 */
523 	verf.tv_sec = (time_t)nfs_atoi(hw_serial);
524 	if (verf.tv_sec != 0) {
525 		verf.tv_nsec = gethrestime_sec();
526 	} else {
527 		timespec_t tverf;
528 
529 		gethrestime(&tverf);
530 		verf.tv_sec = (time_t)tverf.tv_sec;
531 		verf.tv_nsec = tverf.tv_nsec;
532 	}
533 
534 	Write4verf = *(uint64_t *)&verf;
535 
536 	rfs4_attr_init();
537 	mutex_init(&rfs4_deleg_lock, NULL, MUTEX_DEFAULT, NULL);
538 
539 	/* Used to manage create/destroy of server state */
540 	mutex_init(&rfs4_state_lock, NULL, MUTEX_DEFAULT, NULL);
541 
542 	/* Used to manage access to server instance linked list */
543 	mutex_init(&rfs4_servinst_lock, NULL, MUTEX_DEFAULT, NULL);
544 
545 	/* Used to manage access to rfs4_deleg_policy */
546 	rw_init(&rfs4_deleg_policy_lock, NULL, RW_DEFAULT, NULL);
547 
548 	error = fem_create("deleg_rdops", nfs4_rd_deleg_tmpl, &deleg_rdops);
549 	if (error != 0) {
550 		rfs4_disable_delegation();
551 	} else {
552 		error = fem_create("deleg_wrops", nfs4_wr_deleg_tmpl,
553 		    &deleg_wrops);
554 		if (error != 0) {
555 			rfs4_disable_delegation();
556 			fem_free(deleg_rdops);
557 		}
558 	}
559 
560 	nfs4_srv_caller_id = fs_new_caller_id();
561 
562 	lockt_sysid = lm_alloc_sysidt();
563 
564 	vsd_create(&nfs4_srv_vkey, NULL);
565 
566 	return (0);
567 }
568 
569 void
570 rfs4_srvrfini(void)
571 {
572 	extern krwlock_t rfs4_deleg_policy_lock;
573 
574 	if (lockt_sysid != LM_NOSYSID) {
575 		lm_free_sysidt(lockt_sysid);
576 		lockt_sysid = LM_NOSYSID;
577 	}
578 
579 	mutex_destroy(&rfs4_deleg_lock);
580 	mutex_destroy(&rfs4_state_lock);
581 	rw_destroy(&rfs4_deleg_policy_lock);
582 
583 	fem_free(deleg_rdops);
584 	fem_free(deleg_wrops);
585 }
586 
587 void
588 rfs4_init_compound_state(struct compound_state *cs)
589 {
590 	bzero(cs, sizeof (*cs));
591 	cs->cont = TRUE;
592 	cs->access = CS_ACCESS_DENIED;
593 	cs->deleg = FALSE;
594 	cs->mandlock = FALSE;
595 	cs->fh.nfs_fh4_val = cs->fhbuf;
596 }
597 
598 void
599 rfs4_grace_start(rfs4_servinst_t *sip)
600 {
601 	time_t now = gethrestime_sec();
602 
603 	rw_enter(&sip->rwlock, RW_WRITER);
604 	sip->start_time = now;
605 	sip->grace_period = rfs4_grace_period;
606 	rw_exit(&sip->rwlock);
607 }
608 
609 /*
610  * returns true if the instance's grace period has never been started
611  */
612 int
613 rfs4_servinst_grace_new(rfs4_servinst_t *sip)
614 {
615 	time_t start_time;
616 
617 	rw_enter(&sip->rwlock, RW_READER);
618 	start_time = sip->start_time;
619 	rw_exit(&sip->rwlock);
620 
621 	return (start_time == 0);
622 }
623 
624 /*
625  * Indicates if server instance is within the
626  * grace period.
627  */
628 int
629 rfs4_servinst_in_grace(rfs4_servinst_t *sip)
630 {
631 	time_t grace_expiry;
632 
633 	rw_enter(&sip->rwlock, RW_READER);
634 	grace_expiry = sip->start_time + sip->grace_period;
635 	rw_exit(&sip->rwlock);
636 
637 	return (gethrestime_sec() < grace_expiry);
638 }
639 
640 int
641 rfs4_clnt_in_grace(rfs4_client_t *cp)
642 {
643 	ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0);
644 
645 	return (rfs4_servinst_in_grace(cp->server_instance));
646 }
647 
648 /*
649  * reset all currently active grace periods
650  */
651 void
652 rfs4_grace_reset_all(void)
653 {
654 	rfs4_servinst_t *sip;
655 
656 	mutex_enter(&rfs4_servinst_lock);
657 	for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev)
658 		if (rfs4_servinst_in_grace(sip))
659 			rfs4_grace_start(sip);
660 	mutex_exit(&rfs4_servinst_lock);
661 }
662 
663 /*
664  * start any new instances' grace periods
665  */
666 void
667 rfs4_grace_start_new(void)
668 {
669 	rfs4_servinst_t *sip;
670 
671 	mutex_enter(&rfs4_servinst_lock);
672 	for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev)
673 		if (rfs4_servinst_grace_new(sip))
674 			rfs4_grace_start(sip);
675 	mutex_exit(&rfs4_servinst_lock);
676 }
677 
678 static rfs4_dss_path_t *
679 rfs4_dss_newpath(rfs4_servinst_t *sip, char *path, unsigned index)
680 {
681 	size_t len;
682 	rfs4_dss_path_t *dss_path;
683 
684 	dss_path = kmem_alloc(sizeof (rfs4_dss_path_t), KM_SLEEP);
685 
686 	/*
687 	 * Take a copy of the string, since the original may be overwritten.
688 	 * Sadly, no strdup() in the kernel.
689 	 */
690 	/* allow for NUL */
691 	len = strlen(path) + 1;
692 	dss_path->path = kmem_alloc(len, KM_SLEEP);
693 	(void) strlcpy(dss_path->path, path, len);
694 
695 	/* associate with servinst */
696 	dss_path->sip = sip;
697 	dss_path->index = index;
698 
699 	/*
700 	 * Add to list of served paths.
701 	 * No locking required, as we're only ever called at startup.
702 	 */
703 	if (rfs4_dss_pathlist == NULL) {
704 		/* this is the first dss_path_t */
705 
706 		/* needed for insque/remque */
707 		dss_path->next = dss_path->prev = dss_path;
708 
709 		rfs4_dss_pathlist = dss_path;
710 	} else {
711 		insque(dss_path, rfs4_dss_pathlist);
712 	}
713 
714 	return (dss_path);
715 }
716 
717 /*
718  * Create a new server instance, and make it the currently active instance.
719  * Note that starting the grace period too early will reduce the clients'
720  * recovery window.
721  */
722 void
723 rfs4_servinst_create(int start_grace, int dss_npaths, char **dss_paths)
724 {
725 	unsigned i;
726 	rfs4_servinst_t *sip;
727 	rfs4_oldstate_t *oldstate;
728 
729 	sip = kmem_alloc(sizeof (rfs4_servinst_t), KM_SLEEP);
730 	rw_init(&sip->rwlock, NULL, RW_DEFAULT, NULL);
731 
732 	sip->start_time = (time_t)0;
733 	sip->grace_period = (time_t)0;
734 	sip->next = NULL;
735 	sip->prev = NULL;
736 
737 	rw_init(&sip->oldstate_lock, NULL, RW_DEFAULT, NULL);
738 	/*
739 	 * This initial dummy entry is required to setup for insque/remque.
740 	 * It must be skipped over whenever the list is traversed.
741 	 */
742 	oldstate = kmem_alloc(sizeof (rfs4_oldstate_t), KM_SLEEP);
743 	/* insque/remque require initial list entry to be self-terminated */
744 	oldstate->next = oldstate;
745 	oldstate->prev = oldstate;
746 	sip->oldstate = oldstate;
747 
748 
749 	sip->dss_npaths = dss_npaths;
750 	sip->dss_paths = kmem_alloc(dss_npaths *
751 	    sizeof (rfs4_dss_path_t *), KM_SLEEP);
752 
753 	for (i = 0; i < dss_npaths; i++) {
754 		sip->dss_paths[i] = rfs4_dss_newpath(sip, dss_paths[i], i);
755 	}
756 
757 	mutex_enter(&rfs4_servinst_lock);
758 	if (rfs4_cur_servinst != NULL) {
759 		/* add to linked list */
760 		sip->prev = rfs4_cur_servinst;
761 		rfs4_cur_servinst->next = sip;
762 	}
763 	if (start_grace)
764 		rfs4_grace_start(sip);
765 	/* make the new instance "current" */
766 	rfs4_cur_servinst = sip;
767 
768 	mutex_exit(&rfs4_servinst_lock);
769 }
770 
771 /*
772  * In future, we might add a rfs4_servinst_destroy(sip) but, for now, destroy
773  * all instances directly.
774  */
775 void
776 rfs4_servinst_destroy_all(void)
777 {
778 	rfs4_servinst_t *sip, *prev, *current;
779 #ifdef DEBUG
780 	int n = 0;
781 #endif
782 
783 	mutex_enter(&rfs4_servinst_lock);
784 	ASSERT(rfs4_cur_servinst != NULL);
785 	current = rfs4_cur_servinst;
786 	rfs4_cur_servinst = NULL;
787 	for (sip = current; sip != NULL; sip = prev) {
788 		prev = sip->prev;
789 		rw_destroy(&sip->rwlock);
790 		if (sip->oldstate)
791 			kmem_free(sip->oldstate, sizeof (rfs4_oldstate_t));
792 		if (sip->dss_paths)
793 			kmem_free(sip->dss_paths,
794 			    sip->dss_npaths * sizeof (rfs4_dss_path_t *));
795 		kmem_free(sip, sizeof (rfs4_servinst_t));
796 #ifdef DEBUG
797 		n++;
798 #endif
799 	}
800 	mutex_exit(&rfs4_servinst_lock);
801 }
802 
803 /*
804  * Assign the current server instance to a client_t.
805  * Should be called with cp->dbe held.
806  */
807 void
808 rfs4_servinst_assign(rfs4_client_t *cp, rfs4_servinst_t *sip)
809 {
810 	ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0);
811 
812 	/*
813 	 * The lock ensures that if the current instance is in the process
814 	 * of changing, we will see the new one.
815 	 */
816 	mutex_enter(&rfs4_servinst_lock);
817 	cp->server_instance = sip;
818 	mutex_exit(&rfs4_servinst_lock);
819 }
820 
821 rfs4_servinst_t *
822 rfs4_servinst(rfs4_client_t *cp)
823 {
824 	ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0);
825 
826 	return (cp->server_instance);
827 }
828 
829 /* ARGSUSED */
830 static void
831 nullfree(caddr_t resop)
832 {
833 }
834 
835 /*
836  * This is a fall-through for invalid or not implemented (yet) ops
837  */
838 /* ARGSUSED */
839 static void
840 rfs4_op_inval(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
841 	struct compound_state *cs)
842 {
843 	*cs->statusp = *((nfsstat4 *)&(resop)->nfs_resop4_u) = NFS4ERR_INVAL;
844 }
845 
846 /*
847  * Check if the security flavor, nfsnum, is in the flavor_list.
848  */
849 bool_t
850 in_flavor_list(int nfsnum, int *flavor_list, int count)
851 {
852 	int i;
853 
854 	for (i = 0; i < count; i++) {
855 		if (nfsnum == flavor_list[i])
856 			return (TRUE);
857 	}
858 	return (FALSE);
859 }
860 
861 /*
862  * Used by rfs4_op_secinfo to get the security information from the
863  * export structure associated with the component.
864  */
865 /* ARGSUSED */
866 static nfsstat4
867 do_rfs4_op_secinfo(struct compound_state *cs, char *nm, SECINFO4res *resp)
868 {
869 	int error, different_export = 0;
870 	vnode_t *dvp, *vp, *tvp;
871 	struct exportinfo *exi = NULL;
872 	fid_t fid;
873 	uint_t count, i;
874 	secinfo4 *resok_val;
875 	struct secinfo *secp;
876 	seconfig_t *si;
877 	bool_t did_traverse;
878 	int dotdot, walk;
879 
880 	dvp = cs->vp;
881 	dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0');
882 
883 	/*
884 	 * If dotdotting, then need to check whether it's above the
885 	 * root of a filesystem, or above an export point.
886 	 */
887 	if (dotdot) {
888 
889 		/*
890 		 * If dotdotting at the root of a filesystem, then
891 		 * need to traverse back to the mounted-on filesystem
892 		 * and do the dotdot lookup there.
893 		 */
894 		if (cs->vp->v_flag & VROOT) {
895 
896 			/*
897 			 * If at the system root, then can
898 			 * go up no further.
899 			 */
900 			if (VN_CMP(dvp, rootdir))
901 				return (puterrno4(ENOENT));
902 
903 			/*
904 			 * Traverse back to the mounted-on filesystem
905 			 */
906 			dvp = untraverse(cs->vp);
907 
908 			/*
909 			 * Set the different_export flag so we remember
910 			 * to pick up a new exportinfo entry for
911 			 * this new filesystem.
912 			 */
913 			different_export = 1;
914 		} else {
915 
916 			/*
917 			 * If dotdotting above an export point then set
918 			 * the different_export to get new export info.
919 			 */
920 			different_export = nfs_exported(cs->exi, cs->vp);
921 		}
922 	}
923 
924 	/*
925 	 * Get the vnode for the component "nm".
926 	 */
927 	error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cs->cr,
928 	    NULL, NULL, NULL);
929 	if (error)
930 		return (puterrno4(error));
931 
932 	/*
933 	 * If the vnode is in a pseudo filesystem, or if the security flavor
934 	 * used in the request is valid but not an explicitly shared flavor,
935 	 * or the access bit indicates that this is a limited access,
936 	 * check whether this vnode is visible.
937 	 */
938 	if (!different_export &&
939 	    (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) ||
940 	    cs->access & CS_ACCESS_LIMITED)) {
941 		if (! nfs_visible(cs->exi, vp, &different_export)) {
942 			VN_RELE(vp);
943 			return (puterrno4(ENOENT));
944 		}
945 	}
946 
947 	/*
948 	 * If it's a mountpoint, then traverse it.
949 	 */
950 	if (vn_ismntpt(vp)) {
951 		tvp = vp;
952 		if ((error = traverse(&tvp)) != 0) {
953 			VN_RELE(vp);
954 			return (puterrno4(error));
955 		}
956 		/* remember that we had to traverse mountpoint */
957 		did_traverse = TRUE;
958 		vp = tvp;
959 		different_export = 1;
960 	} else if (vp->v_vfsp != dvp->v_vfsp) {
961 		/*
962 		 * If vp isn't a mountpoint and the vfs ptrs aren't the same,
963 		 * then vp is probably an LOFS object.  We don't need the
964 		 * realvp, we just need to know that we might have crossed
965 		 * a server fs boundary and need to call checkexport4.
966 		 * (LOFS lookup hides server fs mountpoints, and actually calls
967 		 * traverse)
968 		 */
969 		different_export = 1;
970 		did_traverse = FALSE;
971 	}
972 
973 	/*
974 	 * Get the export information for it.
975 	 */
976 	if (different_export) {
977 
978 		bzero(&fid, sizeof (fid));
979 		fid.fid_len = MAXFIDSZ;
980 		error = vop_fid_pseudo(vp, &fid);
981 		if (error) {
982 			VN_RELE(vp);
983 			return (puterrno4(error));
984 		}
985 
986 		if (dotdot)
987 			exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE);
988 		else
989 			exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
990 
991 		if (exi == NULL) {
992 			if (did_traverse == TRUE) {
993 				/*
994 				 * If this vnode is a mounted-on vnode,
995 				 * but the mounted-on file system is not
996 				 * exported, send back the secinfo for
997 				 * the exported node that the mounted-on
998 				 * vnode lives in.
999 				 */
1000 				exi = cs->exi;
1001 			} else {
1002 				VN_RELE(vp);
1003 				return (puterrno4(EACCES));
1004 			}
1005 		}
1006 	} else {
1007 		exi = cs->exi;
1008 	}
1009 	ASSERT(exi != NULL);
1010 
1011 
1012 	/*
1013 	 * Create the secinfo result based on the security information
1014 	 * from the exportinfo structure (exi).
1015 	 *
1016 	 * Return all flavors for a pseudo node.
1017 	 * For a real export node, return the flavor that the client
1018 	 * has access with.
1019 	 */
1020 	ASSERT(RW_LOCK_HELD(&exported_lock));
1021 	if (PSEUDO(exi)) {
1022 		count = exi->exi_export.ex_seccnt; /* total sec count */
1023 		resok_val = kmem_alloc(count * sizeof (secinfo4), KM_SLEEP);
1024 		secp = exi->exi_export.ex_secinfo;
1025 
1026 		for (i = 0; i < count; i++) {
1027 			si = &secp[i].s_secinfo;
1028 			resok_val[i].flavor = si->sc_rpcnum;
1029 			if (resok_val[i].flavor == RPCSEC_GSS) {
1030 				rpcsec_gss_info *info;
1031 
1032 				info = &resok_val[i].flavor_info;
1033 				info->qop = si->sc_qop;
1034 				info->service = (rpc_gss_svc_t)si->sc_service;
1035 
1036 				/* get oid opaque data */
1037 				info->oid.sec_oid4_len =
1038 				    si->sc_gss_mech_type->length;
1039 				info->oid.sec_oid4_val = kmem_alloc(
1040 				    si->sc_gss_mech_type->length, KM_SLEEP);
1041 				bcopy(
1042 				    si->sc_gss_mech_type->elements,
1043 				    info->oid.sec_oid4_val,
1044 				    info->oid.sec_oid4_len);
1045 			}
1046 		}
1047 		resp->SECINFO4resok_len = count;
1048 		resp->SECINFO4resok_val = resok_val;
1049 	} else {
1050 		int ret_cnt = 0, k = 0;
1051 		int *flavor_list;
1052 
1053 		count = exi->exi_export.ex_seccnt; /* total sec count */
1054 		secp = exi->exi_export.ex_secinfo;
1055 
1056 		flavor_list = kmem_alloc(count * sizeof (int), KM_SLEEP);
1057 		/* find out which flavors to return */
1058 		for (i = 0; i < count; i ++) {
1059 			int access, flavor, perm;
1060 
1061 			flavor = secp[i].s_secinfo.sc_nfsnum;
1062 			perm = secp[i].s_flags;
1063 
1064 			access = nfsauth4_secinfo_access(exi, cs->req,
1065 			    flavor, perm);
1066 
1067 			if (! (access & NFSAUTH_DENIED) &&
1068 			    ! (access & NFSAUTH_WRONGSEC)) {
1069 				flavor_list[ret_cnt] = flavor;
1070 				ret_cnt++;
1071 			}
1072 		}
1073 
1074 		/* Create the returning SECINFO value */
1075 		resok_val = kmem_alloc(ret_cnt * sizeof (secinfo4), KM_SLEEP);
1076 
1077 		for (i = 0; i < count; i++) {
1078 			/*
1079 			 * If the flavor is in the flavor list,
1080 			 * fill in resok_val.
1081 			 */
1082 			si = &secp[i].s_secinfo;
1083 			if (in_flavor_list(si->sc_nfsnum,
1084 			    flavor_list, ret_cnt)) {
1085 				resok_val[k].flavor = si->sc_rpcnum;
1086 				if (resok_val[k].flavor == RPCSEC_GSS) {
1087 					rpcsec_gss_info *info;
1088 
1089 					info = &resok_val[k].flavor_info;
1090 					info->qop = si->sc_qop;
1091 					info->service = (rpc_gss_svc_t)
1092 					    si->sc_service;
1093 
1094 					/* get oid opaque data */
1095 					info->oid.sec_oid4_len =
1096 					    si->sc_gss_mech_type->length;
1097 					info->oid.sec_oid4_val = kmem_alloc(
1098 					    si->sc_gss_mech_type->length,
1099 					    KM_SLEEP);
1100 					bcopy(si->sc_gss_mech_type->elements,
1101 					    info->oid.sec_oid4_val,
1102 					    info->oid.sec_oid4_len);
1103 				}
1104 				k++;
1105 			}
1106 			if (k >= ret_cnt)
1107 				break;
1108 		}
1109 		resp->SECINFO4resok_len = ret_cnt;
1110 		resp->SECINFO4resok_val = resok_val;
1111 		kmem_free(flavor_list, count * sizeof (int));
1112 	}
1113 
1114 	VN_RELE(vp);
1115 	return (NFS4_OK);
1116 }
1117 
1118 /*
1119  * SECINFO (Operation 33): Obtain required security information on
1120  * the component name in the format of (security-mechanism-oid, qop, service)
1121  * triplets.
1122  */
1123 /* ARGSUSED */
1124 static void
1125 rfs4_op_secinfo(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1126 	struct compound_state *cs)
1127 {
1128 	SECINFO4args *args = &argop->nfs_argop4_u.opsecinfo;
1129 	SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo;
1130 	utf8string *utfnm = &args->name;
1131 	uint_t len;
1132 	char *nm;
1133 
1134 	DTRACE_NFSV4_2(op__secinfo__start, struct compound_state *, cs,
1135 	    SECINFO4args *, args);
1136 
1137 	/*
1138 	 * Current file handle (cfh) should have been set before getting
1139 	 * into this function. If not, return error.
1140 	 */
1141 	if (cs->vp == NULL) {
1142 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1143 		goto out;
1144 	}
1145 
1146 	if (cs->vp->v_type != VDIR) {
1147 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
1148 		goto out;
1149 	}
1150 
1151 	/*
1152 	 * Verify the component name. If failed, error out, but
1153 	 * do not error out if the component name is a "..".
1154 	 * SECINFO will return its parents secinfo data for SECINFO "..".
1155 	 */
1156 	if (!utf8_dir_verify(utfnm)) {
1157 		if (utfnm->utf8string_len != 2 ||
1158 		    utfnm->utf8string_val[0] != '.' ||
1159 		    utfnm->utf8string_val[1] != '.') {
1160 			*cs->statusp = resp->status = NFS4ERR_INVAL;
1161 			goto out;
1162 		}
1163 	}
1164 
1165 	nm = utf8_to_str(utfnm, &len, NULL);
1166 	if (nm == NULL) {
1167 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1168 		goto out;
1169 	}
1170 
1171 	if (len > MAXNAMELEN) {
1172 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1173 		kmem_free(nm, len);
1174 		goto out;
1175 	}
1176 
1177 	*cs->statusp = resp->status = do_rfs4_op_secinfo(cs, nm, resp);
1178 
1179 	kmem_free(nm, len);
1180 
1181 out:
1182 	DTRACE_NFSV4_2(op__secinfo__done, struct compound_state *, cs,
1183 	    SECINFO4res *, resp);
1184 }
1185 
1186 /*
1187  * Free SECINFO result.
1188  */
1189 /* ARGSUSED */
1190 static void
1191 rfs4_op_secinfo_free(nfs_resop4 *resop)
1192 {
1193 	SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo;
1194 	int count, i;
1195 	secinfo4 *resok_val;
1196 
1197 	/* If this is not an Ok result, nothing to free. */
1198 	if (resp->status != NFS4_OK) {
1199 		return;
1200 	}
1201 
1202 	count = resp->SECINFO4resok_len;
1203 	resok_val = resp->SECINFO4resok_val;
1204 
1205 	for (i = 0; i < count; i++) {
1206 		if (resok_val[i].flavor == RPCSEC_GSS) {
1207 			rpcsec_gss_info *info;
1208 
1209 			info = &resok_val[i].flavor_info;
1210 			kmem_free(info->oid.sec_oid4_val,
1211 			    info->oid.sec_oid4_len);
1212 		}
1213 	}
1214 	kmem_free(resok_val, count * sizeof (secinfo4));
1215 	resp->SECINFO4resok_len = 0;
1216 	resp->SECINFO4resok_val = NULL;
1217 }
1218 
1219 /* ARGSUSED */
1220 static void
1221 rfs4_op_access(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1222 	struct compound_state *cs)
1223 {
1224 	ACCESS4args *args = &argop->nfs_argop4_u.opaccess;
1225 	ACCESS4res *resp = &resop->nfs_resop4_u.opaccess;
1226 	int error;
1227 	vnode_t *vp;
1228 	struct vattr va;
1229 	int checkwriteperm;
1230 	cred_t *cr = cs->cr;
1231 	bslabel_t *clabel, *slabel;
1232 	ts_label_t *tslabel;
1233 	boolean_t admin_low_client;
1234 
1235 	DTRACE_NFSV4_2(op__access__start, struct compound_state *, cs,
1236 	    ACCESS4args *, args);
1237 
1238 #if 0	/* XXX allow access even if !cs->access. Eventually only pseudo fs */
1239 	if (cs->access == CS_ACCESS_DENIED) {
1240 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1241 		goto out;
1242 	}
1243 #endif
1244 	if (cs->vp == NULL) {
1245 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1246 		goto out;
1247 	}
1248 
1249 	ASSERT(cr != NULL);
1250 
1251 	vp = cs->vp;
1252 
1253 	/*
1254 	 * If the file system is exported read only, it is not appropriate
1255 	 * to check write permissions for regular files and directories.
1256 	 * Special files are interpreted by the client, so the underlying
1257 	 * permissions are sent back to the client for interpretation.
1258 	 */
1259 	if (rdonly4(cs->exi, cs->vp, req) &&
1260 	    (vp->v_type == VREG || vp->v_type == VDIR))
1261 		checkwriteperm = 0;
1262 	else
1263 		checkwriteperm = 1;
1264 
1265 	/*
1266 	 * XXX
1267 	 * We need the mode so that we can correctly determine access
1268 	 * permissions relative to a mandatory lock file.  Access to
1269 	 * mandatory lock files is denied on the server, so it might
1270 	 * as well be reflected to the server during the open.
1271 	 */
1272 	va.va_mask = AT_MODE;
1273 	error = VOP_GETATTR(vp, &va, 0, cr, NULL);
1274 	if (error) {
1275 		*cs->statusp = resp->status = puterrno4(error);
1276 		goto out;
1277 	}
1278 	resp->access = 0;
1279 	resp->supported = 0;
1280 
1281 	if (is_system_labeled()) {
1282 		ASSERT(req->rq_label != NULL);
1283 		clabel = req->rq_label;
1284 		DTRACE_PROBE2(tx__rfs4__log__info__opaccess__clabel, char *,
1285 		    "got client label from request(1)",
1286 		    struct svc_req *, req);
1287 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
1288 			if ((tslabel = nfs_getflabel(vp)) == NULL) {
1289 				*cs->statusp = resp->status = puterrno4(EACCES);
1290 				goto out;
1291 			}
1292 			slabel = label2bslabel(tslabel);
1293 			DTRACE_PROBE3(tx__rfs4__log__info__opaccess__slabel,
1294 			    char *, "got server label(1) for vp(2)",
1295 			    bslabel_t *, slabel, vnode_t *, vp);
1296 
1297 			admin_low_client = B_FALSE;
1298 		} else
1299 			admin_low_client = B_TRUE;
1300 	}
1301 
1302 	if (args->access & ACCESS4_READ) {
1303 		error = VOP_ACCESS(vp, VREAD, 0, cr, NULL);
1304 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1305 		    (!is_system_labeled() || admin_low_client ||
1306 		    bldominates(clabel, slabel)))
1307 			resp->access |= ACCESS4_READ;
1308 		resp->supported |= ACCESS4_READ;
1309 	}
1310 	if ((args->access & ACCESS4_LOOKUP) && vp->v_type == VDIR) {
1311 		error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL);
1312 		if (!error && (!is_system_labeled() || admin_low_client ||
1313 		    bldominates(clabel, slabel)))
1314 			resp->access |= ACCESS4_LOOKUP;
1315 		resp->supported |= ACCESS4_LOOKUP;
1316 	}
1317 	if (checkwriteperm &&
1318 	    (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND))) {
1319 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
1320 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1321 		    (!is_system_labeled() || admin_low_client ||
1322 		    blequal(clabel, slabel)))
1323 			resp->access |=
1324 			    (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND));
1325 		resp->supported |= (ACCESS4_MODIFY|ACCESS4_EXTEND);
1326 	}
1327 
1328 	if (checkwriteperm &&
1329 	    (args->access & ACCESS4_DELETE) && vp->v_type == VDIR) {
1330 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
1331 		if (!error && (!is_system_labeled() || admin_low_client ||
1332 		    blequal(clabel, slabel)))
1333 			resp->access |= ACCESS4_DELETE;
1334 		resp->supported |= ACCESS4_DELETE;
1335 	}
1336 	if (args->access & ACCESS4_EXECUTE && vp->v_type != VDIR) {
1337 		error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL);
1338 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1339 		    (!is_system_labeled() || admin_low_client ||
1340 		    bldominates(clabel, slabel)))
1341 			resp->access |= ACCESS4_EXECUTE;
1342 		resp->supported |= ACCESS4_EXECUTE;
1343 	}
1344 
1345 	if (is_system_labeled() && !admin_low_client)
1346 		label_rele(tslabel);
1347 
1348 	*cs->statusp = resp->status = NFS4_OK;
1349 out:
1350 	DTRACE_NFSV4_2(op__access__done, struct compound_state *, cs,
1351 	    ACCESS4res *, resp);
1352 }
1353 
1354 /* ARGSUSED */
1355 static void
1356 rfs4_op_commit(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1357 	struct compound_state *cs)
1358 {
1359 	COMMIT4args *args = &argop->nfs_argop4_u.opcommit;
1360 	COMMIT4res *resp = &resop->nfs_resop4_u.opcommit;
1361 	int error;
1362 	vnode_t *vp = cs->vp;
1363 	cred_t *cr = cs->cr;
1364 	vattr_t va;
1365 
1366 	DTRACE_NFSV4_2(op__commit__start, struct compound_state *, cs,
1367 	    COMMIT4args *, args);
1368 
1369 	if (vp == NULL) {
1370 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1371 		goto out;
1372 	}
1373 	if (cs->access == CS_ACCESS_DENIED) {
1374 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1375 		goto out;
1376 	}
1377 
1378 	if (args->offset + args->count < args->offset) {
1379 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1380 		goto out;
1381 	}
1382 
1383 	va.va_mask = AT_UID;
1384 	error = VOP_GETATTR(vp, &va, 0, cr, NULL);
1385 
1386 	/*
1387 	 * If we can't get the attributes, then we can't do the
1388 	 * right access checking.  So, we'll fail the request.
1389 	 */
1390 	if (error) {
1391 		*cs->statusp = resp->status = puterrno4(error);
1392 		goto out;
1393 	}
1394 	if (rdonly4(cs->exi, cs->vp, req)) {
1395 		*cs->statusp = resp->status = NFS4ERR_ROFS;
1396 		goto out;
1397 	}
1398 
1399 	if (vp->v_type != VREG) {
1400 		if (vp->v_type == VDIR)
1401 			resp->status = NFS4ERR_ISDIR;
1402 		else
1403 			resp->status = NFS4ERR_INVAL;
1404 		*cs->statusp = resp->status;
1405 		goto out;
1406 	}
1407 
1408 	if (crgetuid(cr) != va.va_uid &&
1409 	    (error = VOP_ACCESS(vp, VWRITE, 0, cs->cr, NULL))) {
1410 		*cs->statusp = resp->status = puterrno4(error);
1411 		goto out;
1412 	}
1413 
1414 	error = VOP_PUTPAGE(vp, args->offset, args->count, 0, cr, NULL);
1415 	if (!error)
1416 		error = VOP_FSYNC(vp, FNODSYNC, cr, NULL);
1417 
1418 	if (error) {
1419 		*cs->statusp = resp->status = puterrno4(error);
1420 		goto out;
1421 	}
1422 
1423 	*cs->statusp = resp->status = NFS4_OK;
1424 	resp->writeverf = Write4verf;
1425 out:
1426 	DTRACE_NFSV4_2(op__commit__done, struct compound_state *, cs,
1427 	    COMMIT4res *, resp);
1428 }
1429 
1430 /*
1431  * rfs4_op_mknod is called from rfs4_op_create after all initial verification
1432  * was completed. It does the nfsv4 create for special files.
1433  */
1434 /* ARGSUSED */
1435 static vnode_t *
1436 do_rfs4_op_mknod(CREATE4args *args, CREATE4res *resp, struct svc_req *req,
1437 	struct compound_state *cs, vattr_t *vap, char *nm)
1438 {
1439 	int error;
1440 	cred_t *cr = cs->cr;
1441 	vnode_t *dvp = cs->vp;
1442 	vnode_t *vp = NULL;
1443 	int mode;
1444 	enum vcexcl excl;
1445 
1446 	switch (args->type) {
1447 	case NF4CHR:
1448 	case NF4BLK:
1449 		if (secpolicy_sys_devices(cr) != 0) {
1450 			*cs->statusp = resp->status = NFS4ERR_PERM;
1451 			return (NULL);
1452 		}
1453 		if (args->type == NF4CHR)
1454 			vap->va_type = VCHR;
1455 		else
1456 			vap->va_type = VBLK;
1457 		vap->va_rdev = makedevice(args->ftype4_u.devdata.specdata1,
1458 		    args->ftype4_u.devdata.specdata2);
1459 		vap->va_mask |= AT_RDEV;
1460 		break;
1461 	case NF4SOCK:
1462 		vap->va_type = VSOCK;
1463 		break;
1464 	case NF4FIFO:
1465 		vap->va_type = VFIFO;
1466 		break;
1467 	default:
1468 		*cs->statusp = resp->status = NFS4ERR_BADTYPE;
1469 		return (NULL);
1470 	}
1471 
1472 	/*
1473 	 * Must specify the mode.
1474 	 */
1475 	if (!(vap->va_mask & AT_MODE)) {
1476 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1477 		return (NULL);
1478 	}
1479 
1480 	excl = EXCL;
1481 
1482 	mode = 0;
1483 
1484 	error = VOP_CREATE(dvp, nm, vap, excl, mode, &vp, cr, 0, NULL, NULL);
1485 	if (error) {
1486 		*cs->statusp = resp->status = puterrno4(error);
1487 		return (NULL);
1488 	}
1489 	return (vp);
1490 }
1491 
1492 /*
1493  * nfsv4 create is used to create non-regular files. For regular files,
1494  * use nfsv4 open.
1495  */
1496 /* ARGSUSED */
1497 static void
1498 rfs4_op_create(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1499 	struct compound_state *cs)
1500 {
1501 	CREATE4args *args = &argop->nfs_argop4_u.opcreate;
1502 	CREATE4res *resp = &resop->nfs_resop4_u.opcreate;
1503 	int error;
1504 	struct vattr bva, iva, iva2, ava, *vap;
1505 	cred_t *cr = cs->cr;
1506 	vnode_t *dvp = cs->vp;
1507 	vnode_t *vp = NULL;
1508 	vnode_t *realvp;
1509 	char *nm, *lnm;
1510 	uint_t len, llen;
1511 	int syncval = 0;
1512 	struct nfs4_svgetit_arg sarg;
1513 	struct nfs4_ntov_table ntov;
1514 	struct statvfs64 sb;
1515 	nfsstat4 status;
1516 
1517 	DTRACE_NFSV4_2(op__create__start, struct compound_state *, cs,
1518 	    CREATE4args *, args);
1519 
1520 	resp->attrset = 0;
1521 
1522 	if (dvp == NULL) {
1523 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1524 		goto out;
1525 	}
1526 
1527 	/*
1528 	 * If there is an unshared filesystem mounted on this vnode,
1529 	 * do not allow to create an object in this directory.
1530 	 */
1531 	if (vn_ismntpt(dvp)) {
1532 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1533 		goto out;
1534 	}
1535 
1536 	/* Verify that type is correct */
1537 	switch (args->type) {
1538 	case NF4LNK:
1539 	case NF4BLK:
1540 	case NF4CHR:
1541 	case NF4SOCK:
1542 	case NF4FIFO:
1543 	case NF4DIR:
1544 		break;
1545 	default:
1546 		*cs->statusp = resp->status = NFS4ERR_BADTYPE;
1547 		goto out;
1548 	};
1549 
1550 	if (cs->access == CS_ACCESS_DENIED) {
1551 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1552 		goto out;
1553 	}
1554 	if (dvp->v_type != VDIR) {
1555 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
1556 		goto out;
1557 	}
1558 	if (!utf8_dir_verify(&args->objname)) {
1559 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1560 		goto out;
1561 	}
1562 
1563 	if (rdonly4(cs->exi, cs->vp, req)) {
1564 		*cs->statusp = resp->status = NFS4ERR_ROFS;
1565 		goto out;
1566 	}
1567 
1568 	/*
1569 	 * Name of newly created object
1570 	 */
1571 	nm = utf8_to_fn(&args->objname, &len, NULL);
1572 	if (nm == NULL) {
1573 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1574 		goto out;
1575 	}
1576 
1577 	if (len > MAXNAMELEN) {
1578 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1579 		kmem_free(nm, len);
1580 		goto out;
1581 	}
1582 
1583 	resp->attrset = 0;
1584 
1585 	sarg.sbp = &sb;
1586 	nfs4_ntov_table_init(&ntov);
1587 
1588 	status = do_rfs4_set_attrs(&resp->attrset,
1589 	    &args->createattrs, cs, &sarg, &ntov, NFS4ATTR_SETIT);
1590 
1591 	if (sarg.vap->va_mask == 0 && status == NFS4_OK)
1592 		status = NFS4ERR_INVAL;
1593 
1594 	if (status != NFS4_OK) {
1595 		*cs->statusp = resp->status = status;
1596 		kmem_free(nm, len);
1597 		nfs4_ntov_table_free(&ntov, &sarg);
1598 		resp->attrset = 0;
1599 		goto out;
1600 	}
1601 
1602 	/* Get "before" change value */
1603 	bva.va_mask = AT_CTIME|AT_SEQ;
1604 	error = VOP_GETATTR(dvp, &bva, 0, cr, NULL);
1605 	if (error) {
1606 		*cs->statusp = resp->status = puterrno4(error);
1607 		kmem_free(nm, len);
1608 		nfs4_ntov_table_free(&ntov, &sarg);
1609 		resp->attrset = 0;
1610 		goto out;
1611 	}
1612 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bva.va_ctime)
1613 
1614 	vap = sarg.vap;
1615 
1616 	/*
1617 	 * Set default initial values for attributes when not specified
1618 	 * in createattrs.
1619 	 */
1620 	if ((vap->va_mask & AT_UID) == 0) {
1621 		vap->va_uid = crgetuid(cr);
1622 		vap->va_mask |= AT_UID;
1623 	}
1624 	if ((vap->va_mask & AT_GID) == 0) {
1625 		vap->va_gid = crgetgid(cr);
1626 		vap->va_mask |= AT_GID;
1627 	}
1628 
1629 	vap->va_mask |= AT_TYPE;
1630 	switch (args->type) {
1631 	case NF4DIR:
1632 		vap->va_type = VDIR;
1633 		if ((vap->va_mask & AT_MODE) == 0) {
1634 			vap->va_mode = 0700;	/* default: owner rwx only */
1635 			vap->va_mask |= AT_MODE;
1636 		}
1637 		error = VOP_MKDIR(dvp, nm, vap, &vp, cr, NULL, 0, NULL);
1638 		if (error)
1639 			break;
1640 
1641 		/*
1642 		 * Get the initial "after" sequence number, if it fails,
1643 		 * set to zero
1644 		 */
1645 		iva.va_mask = AT_SEQ;
1646 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL))
1647 			iva.va_seq = 0;
1648 		break;
1649 	case NF4LNK:
1650 		vap->va_type = VLNK;
1651 		if ((vap->va_mask & AT_MODE) == 0) {
1652 			vap->va_mode = 0700;	/* default: owner rwx only */
1653 			vap->va_mask |= AT_MODE;
1654 		}
1655 
1656 		/*
1657 		 * symlink names must be treated as data
1658 		 */
1659 		lnm = utf8_to_str(&args->ftype4_u.linkdata, &llen, NULL);
1660 
1661 		if (lnm == NULL) {
1662 			*cs->statusp = resp->status = NFS4ERR_INVAL;
1663 			kmem_free(nm, len);
1664 			nfs4_ntov_table_free(&ntov, &sarg);
1665 			resp->attrset = 0;
1666 			goto out;
1667 		}
1668 
1669 		if (llen > MAXPATHLEN) {
1670 			*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1671 			kmem_free(nm, len);
1672 			kmem_free(lnm, llen);
1673 			nfs4_ntov_table_free(&ntov, &sarg);
1674 			resp->attrset = 0;
1675 			goto out;
1676 		}
1677 
1678 		error = VOP_SYMLINK(dvp, nm, vap, lnm, cr, NULL, 0);
1679 		if (lnm != NULL)
1680 			kmem_free(lnm, llen);
1681 		if (error)
1682 			break;
1683 
1684 		/*
1685 		 * Get the initial "after" sequence number, if it fails,
1686 		 * set to zero
1687 		 */
1688 		iva.va_mask = AT_SEQ;
1689 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL))
1690 			iva.va_seq = 0;
1691 
1692 		error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr,
1693 		    NULL, NULL, NULL);
1694 		if (error)
1695 			break;
1696 
1697 		/*
1698 		 * va_seq is not safe over VOP calls, check it again
1699 		 * if it has changed zero out iva to force atomic = FALSE.
1700 		 */
1701 		iva2.va_mask = AT_SEQ;
1702 		if (VOP_GETATTR(dvp, &iva2, 0, cs->cr, NULL) ||
1703 		    iva2.va_seq != iva.va_seq)
1704 			iva.va_seq = 0;
1705 		break;
1706 	default:
1707 		/*
1708 		 * probably a special file.
1709 		 */
1710 		if ((vap->va_mask & AT_MODE) == 0) {
1711 			vap->va_mode = 0600;	/* default: owner rw only */
1712 			vap->va_mask |= AT_MODE;
1713 		}
1714 		syncval = FNODSYNC;
1715 		/*
1716 		 * We know this will only generate one VOP call
1717 		 */
1718 		vp = do_rfs4_op_mknod(args, resp, req, cs, vap, nm);
1719 
1720 		if (vp == NULL) {
1721 			kmem_free(nm, len);
1722 			nfs4_ntov_table_free(&ntov, &sarg);
1723 			resp->attrset = 0;
1724 			goto out;
1725 		}
1726 
1727 		/*
1728 		 * Get the initial "after" sequence number, if it fails,
1729 		 * set to zero
1730 		 */
1731 		iva.va_mask = AT_SEQ;
1732 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL))
1733 			iva.va_seq = 0;
1734 
1735 		break;
1736 	}
1737 	kmem_free(nm, len);
1738 
1739 	if (error) {
1740 		*cs->statusp = resp->status = puterrno4(error);
1741 	}
1742 
1743 	/*
1744 	 * Force modified data and metadata out to stable storage.
1745 	 */
1746 	(void) VOP_FSYNC(dvp, 0, cr, NULL);
1747 
1748 	if (resp->status != NFS4_OK) {
1749 		if (vp != NULL)
1750 			VN_RELE(vp);
1751 		nfs4_ntov_table_free(&ntov, &sarg);
1752 		resp->attrset = 0;
1753 		goto out;
1754 	}
1755 
1756 	/*
1757 	 * Finish setup of cinfo response, "before" value already set.
1758 	 * Get "after" change value, if it fails, simply return the
1759 	 * before value.
1760 	 */
1761 	ava.va_mask = AT_CTIME|AT_SEQ;
1762 	if (VOP_GETATTR(dvp, &ava, 0, cr, NULL)) {
1763 		ava.va_ctime = bva.va_ctime;
1764 		ava.va_seq = 0;
1765 	}
1766 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, ava.va_ctime);
1767 
1768 	/*
1769 	 * True verification that object was created with correct
1770 	 * attrs is impossible.  The attrs could have been changed
1771 	 * immediately after object creation.  If attributes did
1772 	 * not verify, the only recourse for the server is to
1773 	 * destroy the object.  Maybe if some attrs (like gid)
1774 	 * are set incorrectly, the object should be destroyed;
1775 	 * however, seems bad as a default policy.  Do we really
1776 	 * want to destroy an object over one of the times not
1777 	 * verifying correctly?  For these reasons, the server
1778 	 * currently sets bits in attrset for createattrs
1779 	 * that were set; however, no verification is done.
1780 	 *
1781 	 * vmask_to_nmask accounts for vattr bits set on create
1782 	 *	[do_rfs4_set_attrs() only sets resp bits for
1783 	 *	 non-vattr/vfs bits.]
1784 	 * Mask off any bits set by default so as not to return
1785 	 * more attrset bits than were requested in createattrs
1786 	 */
1787 	nfs4_vmask_to_nmask(sarg.vap->va_mask, &resp->attrset);
1788 	resp->attrset &= args->createattrs.attrmask;
1789 	nfs4_ntov_table_free(&ntov, &sarg);
1790 
1791 	error = makefh4(&cs->fh, vp, cs->exi);
1792 	if (error) {
1793 		*cs->statusp = resp->status = puterrno4(error);
1794 	}
1795 
1796 	/*
1797 	 * The cinfo.atomic = TRUE only if we got no errors, we have
1798 	 * non-zero va_seq's, and it has incremented by exactly one
1799 	 * during the creation and it didn't change during the VOP_LOOKUP
1800 	 * or VOP_FSYNC.
1801 	 */
1802 	if (!error && bva.va_seq && iva.va_seq && ava.va_seq &&
1803 	    iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq)
1804 		resp->cinfo.atomic = TRUE;
1805 	else
1806 		resp->cinfo.atomic = FALSE;
1807 
1808 	/*
1809 	 * Force modified metadata out to stable storage.
1810 	 *
1811 	 * if a underlying vp exists, pass it to VOP_FSYNC
1812 	 */
1813 	if (VOP_REALVP(vp, &realvp, NULL) == 0)
1814 		(void) VOP_FSYNC(realvp, syncval, cr, NULL);
1815 	else
1816 		(void) VOP_FSYNC(vp, syncval, cr, NULL);
1817 
1818 	if (resp->status != NFS4_OK) {
1819 		VN_RELE(vp);
1820 		goto out;
1821 	}
1822 	if (cs->vp)
1823 		VN_RELE(cs->vp);
1824 
1825 	cs->vp = vp;
1826 	*cs->statusp = resp->status = NFS4_OK;
1827 out:
1828 	DTRACE_NFSV4_2(op__create__done, struct compound_state *, cs,
1829 	    CREATE4res *, resp);
1830 }
1831 
1832 /*ARGSUSED*/
1833 static void
1834 rfs4_op_delegpurge(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1835     struct compound_state *cs)
1836 {
1837 	DTRACE_NFSV4_2(op__delegpurge__start, struct compound_state *, cs,
1838 	    DELEGPURGE4args *, &argop->nfs_argop4_u.opdelegpurge);
1839 
1840 	rfs4_op_inval(argop, resop, req, cs);
1841 
1842 	DTRACE_NFSV4_2(op__delegpurge__done, struct compound_state *, cs,
1843 	    DELEGPURGE4res *, &resop->nfs_resop4_u.opdelegpurge);
1844 }
1845 
1846 /*ARGSUSED*/
1847 static void
1848 rfs4_op_delegreturn(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1849 	struct compound_state *cs)
1850 {
1851 	DELEGRETURN4args *args = &argop->nfs_argop4_u.opdelegreturn;
1852 	DELEGRETURN4res *resp = &resop->nfs_resop4_u.opdelegreturn;
1853 	rfs4_deleg_state_t *dsp;
1854 	nfsstat4 status;
1855 
1856 	DTRACE_NFSV4_2(op__delegreturn__start, struct compound_state *, cs,
1857 	    DELEGRETURN4args *, args);
1858 
1859 	status = rfs4_get_deleg_state(&args->deleg_stateid, &dsp);
1860 	resp->status = *cs->statusp = status;
1861 	if (status != NFS4_OK)
1862 		goto out;
1863 
1864 	/* Ensure specified filehandle matches */
1865 	if (cs->vp != dsp->finfo->vp) {
1866 		resp->status = *cs->statusp = NFS4ERR_BAD_STATEID;
1867 	} else
1868 		rfs4_return_deleg(dsp, FALSE);
1869 
1870 	rfs4_update_lease(dsp->client);
1871 
1872 	rfs4_deleg_state_rele(dsp);
1873 out:
1874 	DTRACE_NFSV4_2(op__delegreturn__done, struct compound_state *, cs,
1875 	    DELEGRETURN4res *, resp);
1876 }
1877 
1878 /*
1879  * Check to see if a given "flavor" is an explicitly shared flavor.
1880  * The assumption of this routine is the "flavor" is already a valid
1881  * flavor in the secinfo list of "exi".
1882  *
1883  *	e.g.
1884  *		# share -o sec=flavor1 /export
1885  *		# share -o sec=flavor2 /export/home
1886  *
1887  *		flavor2 is not an explicitly shared flavor for /export,
1888  *		however it is in the secinfo list for /export thru the
1889  *		server namespace setup.
1890  */
1891 int
1892 is_exported_sec(int flavor, struct exportinfo *exi)
1893 {
1894 	int	i;
1895 	struct secinfo *sp;
1896 
1897 	sp = exi->exi_export.ex_secinfo;
1898 	for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
1899 		if (flavor == sp[i].s_secinfo.sc_nfsnum ||
1900 		    sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) {
1901 			return (SEC_REF_EXPORTED(&sp[i]));
1902 		}
1903 	}
1904 
1905 	/* Should not reach this point based on the assumption */
1906 	return (0);
1907 }
1908 
1909 /*
1910  * Check if the security flavor used in the request matches what is
1911  * required at the export point or at the root pseudo node (exi_root).
1912  *
1913  * returns 1 if there's a match or if exported with AUTH_NONE; 0 otherwise.
1914  *
1915  */
1916 static int
1917 secinfo_match_or_authnone(struct compound_state *cs)
1918 {
1919 	int	i;
1920 	struct secinfo *sp;
1921 
1922 	/*
1923 	 * Check cs->nfsflavor (from the request) against
1924 	 * the current export data in cs->exi.
1925 	 */
1926 	sp = cs->exi->exi_export.ex_secinfo;
1927 	for (i = 0; i < cs->exi->exi_export.ex_seccnt; i++) {
1928 		if (cs->nfsflavor == sp[i].s_secinfo.sc_nfsnum ||
1929 		    sp[i].s_secinfo.sc_nfsnum == AUTH_NONE)
1930 			return (1);
1931 	}
1932 
1933 	return (0);
1934 }
1935 
1936 /*
1937  * Check the access authority for the client and return the correct error.
1938  */
1939 nfsstat4
1940 call_checkauth4(struct compound_state *cs, struct svc_req *req)
1941 {
1942 	int	authres;
1943 
1944 	/*
1945 	 * First, check if the security flavor used in the request
1946 	 * are among the flavors set in the server namespace.
1947 	 */
1948 	if (!secinfo_match_or_authnone(cs)) {
1949 		*cs->statusp = NFS4ERR_WRONGSEC;
1950 		return (*cs->statusp);
1951 	}
1952 
1953 	authres = checkauth4(cs, req);
1954 
1955 	if (authres > 0) {
1956 		*cs->statusp = NFS4_OK;
1957 		if (! (cs->access & CS_ACCESS_LIMITED))
1958 			cs->access = CS_ACCESS_OK;
1959 	} else if (authres == 0) {
1960 		*cs->statusp = NFS4ERR_ACCESS;
1961 	} else if (authres == -2) {
1962 		*cs->statusp = NFS4ERR_WRONGSEC;
1963 	} else {
1964 		*cs->statusp = NFS4ERR_DELAY;
1965 	}
1966 	return (*cs->statusp);
1967 }
1968 
1969 /*
1970  * bitmap4_to_attrmask is called by getattr and readdir.
1971  * It sets up the vattr mask and determines whether vfsstat call is needed
1972  * based on the input bitmap.
1973  * Returns nfsv4 status.
1974  */
1975 static nfsstat4
1976 bitmap4_to_attrmask(bitmap4 breq, struct nfs4_svgetit_arg *sargp)
1977 {
1978 	int i;
1979 	uint_t	va_mask;
1980 	struct statvfs64 *sbp = sargp->sbp;
1981 
1982 	sargp->sbp = NULL;
1983 	sargp->flag = 0;
1984 	sargp->rdattr_error = NFS4_OK;
1985 	sargp->mntdfid_set = FALSE;
1986 	if (sargp->cs->vp)
1987 		sargp->xattr = get_fh4_flag(&sargp->cs->fh,
1988 		    FH4_ATTRDIR | FH4_NAMEDATTR);
1989 	else
1990 		sargp->xattr = 0;
1991 
1992 	/*
1993 	 * Set rdattr_error_req to true if return error per
1994 	 * failed entry rather than fail the readdir.
1995 	 */
1996 	if (breq & FATTR4_RDATTR_ERROR_MASK)
1997 		sargp->rdattr_error_req = 1;
1998 	else
1999 		sargp->rdattr_error_req = 0;
2000 
2001 	/*
2002 	 * generate the va_mask
2003 	 * Handle the easy cases first
2004 	 */
2005 	switch (breq) {
2006 	case NFS4_NTOV_ATTR_MASK:
2007 		sargp->vap->va_mask = NFS4_NTOV_ATTR_AT_MASK;
2008 		return (NFS4_OK);
2009 
2010 	case NFS4_FS_ATTR_MASK:
2011 		sargp->vap->va_mask = NFS4_FS_ATTR_AT_MASK;
2012 		sargp->sbp = sbp;
2013 		return (NFS4_OK);
2014 
2015 	case NFS4_NTOV_ATTR_CACHE_MASK:
2016 		sargp->vap->va_mask = NFS4_NTOV_ATTR_CACHE_AT_MASK;
2017 		return (NFS4_OK);
2018 
2019 	case FATTR4_LEASE_TIME_MASK:
2020 		sargp->vap->va_mask = 0;
2021 		return (NFS4_OK);
2022 
2023 	default:
2024 		va_mask = 0;
2025 		for (i = 0; i < nfs4_ntov_map_size; i++) {
2026 			if ((breq & nfs4_ntov_map[i].fbit) &&
2027 			    nfs4_ntov_map[i].vbit)
2028 				va_mask |= nfs4_ntov_map[i].vbit;
2029 		}
2030 
2031 		/*
2032 		 * Check is vfsstat is needed
2033 		 */
2034 		if (breq & NFS4_FS_ATTR_MASK)
2035 			sargp->sbp = sbp;
2036 
2037 		sargp->vap->va_mask = va_mask;
2038 		return (NFS4_OK);
2039 	}
2040 	/* NOTREACHED */
2041 }
2042 
2043 /*
2044  * bitmap4_get_sysattrs is called by getattr and readdir.
2045  * It calls both VOP_GETATTR and VFS_STATVFS calls to get the attrs.
2046  * Returns nfsv4 status.
2047  */
2048 static nfsstat4
2049 bitmap4_get_sysattrs(struct nfs4_svgetit_arg *sargp)
2050 {
2051 	int error;
2052 	struct compound_state *cs = sargp->cs;
2053 	vnode_t *vp = cs->vp;
2054 
2055 	if (sargp->sbp != NULL) {
2056 		if (error = VFS_STATVFS(vp->v_vfsp, sargp->sbp)) {
2057 			sargp->sbp = NULL;	/* to identify error */
2058 			return (puterrno4(error));
2059 		}
2060 	}
2061 
2062 	return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr));
2063 }
2064 
2065 static void
2066 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp)
2067 {
2068 	ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size,
2069 	    KM_SLEEP);
2070 	ntovp->attrcnt = 0;
2071 	ntovp->vfsstat = FALSE;
2072 }
2073 
2074 static void
2075 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp,
2076 	struct nfs4_svgetit_arg *sargp)
2077 {
2078 	int i;
2079 	union nfs4_attr_u *na;
2080 	uint8_t *amap;
2081 
2082 	/*
2083 	 * XXX Should do the same checks for whether the bit is set
2084 	 */
2085 	for (i = 0, na = ntovp->na, amap = ntovp->amap;
2086 	    i < ntovp->attrcnt; i++, na++, amap++) {
2087 		(void) (*nfs4_ntov_map[*amap].sv_getit)(
2088 		    NFS4ATTR_FREEIT, sargp, na);
2089 	}
2090 	if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) {
2091 		/*
2092 		 * xdr_free for getattr will be done later
2093 		 */
2094 		for (i = 0, na = ntovp->na, amap = ntovp->amap;
2095 		    i < ntovp->attrcnt; i++, na++, amap++) {
2096 			xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na);
2097 		}
2098 	}
2099 	kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size);
2100 }
2101 
2102 /*
2103  * do_rfs4_op_getattr gets the system attrs and converts into fattr4.
2104  */
2105 static nfsstat4
2106 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp,
2107 	struct nfs4_svgetit_arg *sargp)
2108 {
2109 	int error = 0;
2110 	int i, k;
2111 	struct nfs4_ntov_table ntov;
2112 	XDR xdr;
2113 	ulong_t xdr_size;
2114 	char *xdr_attrs;
2115 	nfsstat4 status = NFS4_OK;
2116 	nfsstat4 prev_rdattr_error = sargp->rdattr_error;
2117 	union nfs4_attr_u *na;
2118 	uint8_t *amap;
2119 
2120 	sargp->op = NFS4ATTR_GETIT;
2121 	sargp->flag = 0;
2122 
2123 	fattrp->attrmask = 0;
2124 	/* if no bits requested, then return empty fattr4 */
2125 	if (breq == 0) {
2126 		fattrp->attrlist4_len = 0;
2127 		fattrp->attrlist4 = NULL;
2128 		return (NFS4_OK);
2129 	}
2130 
2131 	/*
2132 	 * return NFS4ERR_INVAL when client requests write-only attrs
2133 	 */
2134 	if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK))
2135 		return (NFS4ERR_INVAL);
2136 
2137 	nfs4_ntov_table_init(&ntov);
2138 	na = ntov.na;
2139 	amap = ntov.amap;
2140 
2141 	/*
2142 	 * Now loop to get or verify the attrs
2143 	 */
2144 	for (i = 0; i < nfs4_ntov_map_size; i++) {
2145 		if (breq & nfs4_ntov_map[i].fbit) {
2146 			if ((*nfs4_ntov_map[i].sv_getit)(
2147 			    NFS4ATTR_SUPPORTED, sargp, NULL) == 0) {
2148 
2149 				error = (*nfs4_ntov_map[i].sv_getit)(
2150 				    NFS4ATTR_GETIT, sargp, na);
2151 
2152 				/*
2153 				 * Possible error values:
2154 				 * >0 if sv_getit failed to
2155 				 * get the attr; 0 if succeeded;
2156 				 * <0 if rdattr_error and the
2157 				 * attribute cannot be returned.
2158 				 */
2159 				if (error && !(sargp->rdattr_error_req))
2160 					goto done;
2161 				/*
2162 				 * If error then just for entry
2163 				 */
2164 				if (error == 0) {
2165 					fattrp->attrmask |=
2166 					    nfs4_ntov_map[i].fbit;
2167 					*amap++ =
2168 					    (uint8_t)nfs4_ntov_map[i].nval;
2169 					na++;
2170 					(ntov.attrcnt)++;
2171 				} else if ((error > 0) &&
2172 				    (sargp->rdattr_error == NFS4_OK)) {
2173 					sargp->rdattr_error = puterrno4(error);
2174 				}
2175 				error = 0;
2176 			}
2177 		}
2178 	}
2179 
2180 	/*
2181 	 * If rdattr_error was set after the return value for it was assigned,
2182 	 * update it.
2183 	 */
2184 	if (prev_rdattr_error != sargp->rdattr_error) {
2185 		na = ntov.na;
2186 		amap = ntov.amap;
2187 		for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2188 			k = *amap;
2189 			if (k < FATTR4_RDATTR_ERROR) {
2190 				continue;
2191 			}
2192 			if ((k == FATTR4_RDATTR_ERROR) &&
2193 			    ((*nfs4_ntov_map[k].sv_getit)(
2194 			    NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) {
2195 
2196 				(void) (*nfs4_ntov_map[k].sv_getit)(
2197 				    NFS4ATTR_GETIT, sargp, na);
2198 			}
2199 			break;
2200 		}
2201 	}
2202 
2203 	xdr_size = 0;
2204 	na = ntov.na;
2205 	amap = ntov.amap;
2206 	for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2207 		xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na);
2208 	}
2209 
2210 	fattrp->attrlist4_len = xdr_size;
2211 	if (xdr_size) {
2212 		/* freed by rfs4_op_getattr_free() */
2213 		fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP);
2214 
2215 		xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE);
2216 
2217 		na = ntov.na;
2218 		amap = ntov.amap;
2219 		for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2220 			if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) {
2221 				cmn_err(CE_WARN, "do_rfs4_op_getattr: xdr "
2222 				    "encode of attribute %d failed\n", *amap);
2223 				status = NFS4ERR_SERVERFAULT;
2224 				break;
2225 			}
2226 		}
2227 		/* xdrmem_destroy(&xdrs); */	/* NO-OP */
2228 	} else {
2229 		fattrp->attrlist4 = NULL;
2230 	}
2231 done:
2232 
2233 	nfs4_ntov_table_free(&ntov, sargp);
2234 
2235 	if (error != 0)
2236 		status = puterrno4(error);
2237 
2238 	return (status);
2239 }
2240 
2241 /* ARGSUSED */
2242 static void
2243 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2244 	struct compound_state *cs)
2245 {
2246 	GETATTR4args *args = &argop->nfs_argop4_u.opgetattr;
2247 	GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr;
2248 	struct nfs4_svgetit_arg sarg;
2249 	struct statvfs64 sb;
2250 	nfsstat4 status;
2251 
2252 	DTRACE_NFSV4_2(op__getattr__start, struct compound_state *, cs,
2253 	    GETATTR4args *, args);
2254 
2255 	if (cs->vp == NULL) {
2256 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2257 		goto out;
2258 	}
2259 
2260 	if (cs->access == CS_ACCESS_DENIED) {
2261 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2262 		goto out;
2263 	}
2264 
2265 	sarg.sbp = &sb;
2266 	sarg.cs = cs;
2267 
2268 	status = bitmap4_to_attrmask(args->attr_request, &sarg);
2269 	if (status == NFS4_OK) {
2270 		status = bitmap4_get_sysattrs(&sarg);
2271 		if (status == NFS4_OK)
2272 			status = do_rfs4_op_getattr(args->attr_request,
2273 			    &resp->obj_attributes, &sarg);
2274 	}
2275 	*cs->statusp = resp->status = status;
2276 out:
2277 	DTRACE_NFSV4_2(op__getattr__done, struct compound_state *, cs,
2278 	    GETATTR4res *, resp);
2279 }
2280 
2281 static void
2282 rfs4_op_getattr_free(nfs_resop4 *resop)
2283 {
2284 	GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr;
2285 
2286 	nfs4_fattr4_free(&resp->obj_attributes);
2287 }
2288 
2289 /* ARGSUSED */
2290 static void
2291 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2292 	struct compound_state *cs)
2293 {
2294 	GETFH4res *resp = &resop->nfs_resop4_u.opgetfh;
2295 
2296 	DTRACE_NFSV4_1(op__getfh__start, struct compound_state *, cs);
2297 
2298 	if (cs->vp == NULL) {
2299 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2300 		goto out;
2301 	}
2302 	if (cs->access == CS_ACCESS_DENIED) {
2303 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2304 		goto out;
2305 	}
2306 
2307 	resp->object.nfs_fh4_val =
2308 	    kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP);
2309 	nfs_fh4_copy(&cs->fh, &resp->object);
2310 	*cs->statusp = resp->status = NFS4_OK;
2311 out:
2312 	DTRACE_NFSV4_2(op__getfh__done, struct compound_state *, cs,
2313 	    GETFH4res *, resp);
2314 }
2315 
2316 static void
2317 rfs4_op_getfh_free(nfs_resop4 *resop)
2318 {
2319 	GETFH4res *resp = &resop->nfs_resop4_u.opgetfh;
2320 
2321 	if (resp->status == NFS4_OK &&
2322 	    resp->object.nfs_fh4_val != NULL) {
2323 		kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len);
2324 		resp->object.nfs_fh4_val = NULL;
2325 		resp->object.nfs_fh4_len = 0;
2326 	}
2327 }
2328 
2329 /*
2330  * illegal: args: void
2331  *	    res : status (NFS4ERR_OP_ILLEGAL)
2332  */
2333 /* ARGSUSED */
2334 static void
2335 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop,
2336 	struct svc_req *req, struct compound_state *cs)
2337 {
2338 	ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal;
2339 
2340 	resop->resop = OP_ILLEGAL;
2341 	*cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL;
2342 }
2343 
2344 /*
2345  * link: args: SAVED_FH: file, CURRENT_FH: target directory
2346  *	 res: status. If success - CURRENT_FH unchanged, return change_info
2347  */
2348 /* ARGSUSED */
2349 static void
2350 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2351 	struct compound_state *cs)
2352 {
2353 	LINK4args *args = &argop->nfs_argop4_u.oplink;
2354 	LINK4res *resp = &resop->nfs_resop4_u.oplink;
2355 	int error;
2356 	vnode_t *vp;
2357 	vnode_t *dvp;
2358 	struct vattr bdva, idva, adva;
2359 	char *nm;
2360 	uint_t  len;
2361 
2362 	DTRACE_NFSV4_2(op__link__start, struct compound_state *, cs,
2363 	    LINK4args *, args);
2364 
2365 	/* SAVED_FH: source object */
2366 	vp = cs->saved_vp;
2367 	if (vp == NULL) {
2368 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2369 		goto out;
2370 	}
2371 
2372 	/* CURRENT_FH: target directory */
2373 	dvp = cs->vp;
2374 	if (dvp == NULL) {
2375 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2376 		goto out;
2377 	}
2378 
2379 	/*
2380 	 * If there is a non-shared filesystem mounted on this vnode,
2381 	 * do not allow to link any file in this directory.
2382 	 */
2383 	if (vn_ismntpt(dvp)) {
2384 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2385 		goto out;
2386 	}
2387 
2388 	if (cs->access == CS_ACCESS_DENIED) {
2389 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2390 		goto out;
2391 	}
2392 
2393 	/* Check source object's type validity */
2394 	if (vp->v_type == VDIR) {
2395 		*cs->statusp = resp->status = NFS4ERR_ISDIR;
2396 		goto out;
2397 	}
2398 
2399 	/* Check target directory's type */
2400 	if (dvp->v_type != VDIR) {
2401 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2402 		goto out;
2403 	}
2404 
2405 	if (cs->saved_exi != cs->exi) {
2406 		*cs->statusp = resp->status = NFS4ERR_XDEV;
2407 		goto out;
2408 	}
2409 
2410 	if (!utf8_dir_verify(&args->newname)) {
2411 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2412 		goto out;
2413 	}
2414 
2415 	nm = utf8_to_fn(&args->newname, &len, NULL);
2416 	if (nm == NULL) {
2417 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2418 		goto out;
2419 	}
2420 
2421 	if (len > MAXNAMELEN) {
2422 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
2423 		kmem_free(nm, len);
2424 		goto out;
2425 	}
2426 
2427 	if (rdonly4(cs->exi, cs->vp, req)) {
2428 		*cs->statusp = resp->status = NFS4ERR_ROFS;
2429 		kmem_free(nm, len);
2430 		goto out;
2431 	}
2432 
2433 	/* Get "before" change value */
2434 	bdva.va_mask = AT_CTIME|AT_SEQ;
2435 	error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL);
2436 	if (error) {
2437 		*cs->statusp = resp->status = puterrno4(error);
2438 		kmem_free(nm, len);
2439 		goto out;
2440 	}
2441 
2442 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime)
2443 
2444 	error = VOP_LINK(dvp, vp, nm, cs->cr, NULL, 0);
2445 
2446 	kmem_free(nm, len);
2447 
2448 	/*
2449 	 * Get the initial "after" sequence number, if it fails, set to zero
2450 	 */
2451 	idva.va_mask = AT_SEQ;
2452 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL))
2453 		idva.va_seq = 0;
2454 
2455 	/*
2456 	 * Force modified data and metadata out to stable storage.
2457 	 */
2458 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL);
2459 	(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
2460 
2461 	if (error) {
2462 		*cs->statusp = resp->status = puterrno4(error);
2463 		goto out;
2464 	}
2465 
2466 	/*
2467 	 * Get "after" change value, if it fails, simply return the
2468 	 * before value.
2469 	 */
2470 	adva.va_mask = AT_CTIME|AT_SEQ;
2471 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) {
2472 		adva.va_ctime = bdva.va_ctime;
2473 		adva.va_seq = 0;
2474 	}
2475 
2476 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
2477 
2478 	/*
2479 	 * The cinfo.atomic = TRUE only if we have
2480 	 * non-zero va_seq's, and it has incremented by exactly one
2481 	 * during the VOP_LINK and it didn't change during the VOP_FSYNC.
2482 	 */
2483 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
2484 	    idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq)
2485 		resp->cinfo.atomic = TRUE;
2486 	else
2487 		resp->cinfo.atomic = FALSE;
2488 
2489 	*cs->statusp = resp->status = NFS4_OK;
2490 out:
2491 	DTRACE_NFSV4_2(op__link__done, struct compound_state *, cs,
2492 	    LINK4res *, resp);
2493 }
2494 
2495 /*
2496  * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work.
2497  */
2498 
2499 /* ARGSUSED */
2500 static nfsstat4
2501 do_rfs4_op_lookup(char *nm, uint_t buflen, struct svc_req *req,
2502 	struct compound_state *cs)
2503 {
2504 	int error;
2505 	int different_export = 0;
2506 	vnode_t *vp, *tvp, *pre_tvp = NULL, *oldvp = NULL;
2507 	struct exportinfo *exi = NULL, *pre_exi = NULL;
2508 	nfsstat4 stat;
2509 	fid_t fid;
2510 	int attrdir, dotdot, walk;
2511 	bool_t is_newvp = FALSE;
2512 
2513 	if (cs->vp->v_flag & V_XATTRDIR) {
2514 		attrdir = 1;
2515 		ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR));
2516 	} else {
2517 		attrdir = 0;
2518 		ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR));
2519 	}
2520 
2521 	dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0');
2522 
2523 	/*
2524 	 * If dotdotting, then need to check whether it's
2525 	 * above the root of a filesystem, or above an
2526 	 * export point.
2527 	 */
2528 	if (dotdot) {
2529 
2530 		/*
2531 		 * If dotdotting at the root of a filesystem, then
2532 		 * need to traverse back to the mounted-on filesystem
2533 		 * and do the dotdot lookup there.
2534 		 */
2535 		if (cs->vp->v_flag & VROOT) {
2536 
2537 			/*
2538 			 * If at the system root, then can
2539 			 * go up no further.
2540 			 */
2541 			if (VN_CMP(cs->vp, rootdir))
2542 				return (puterrno4(ENOENT));
2543 
2544 			/*
2545 			 * Traverse back to the mounted-on filesystem
2546 			 */
2547 			cs->vp = untraverse(cs->vp);
2548 
2549 			/*
2550 			 * Set the different_export flag so we remember
2551 			 * to pick up a new exportinfo entry for
2552 			 * this new filesystem.
2553 			 */
2554 			different_export = 1;
2555 		} else {
2556 
2557 			/*
2558 			 * If dotdotting above an export point then set
2559 			 * the different_export to get new export info.
2560 			 */
2561 			different_export = nfs_exported(cs->exi, cs->vp);
2562 		}
2563 	}
2564 
2565 	error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr,
2566 	    NULL, NULL, NULL);
2567 	if (error)
2568 		return (puterrno4(error));
2569 
2570 	/*
2571 	 * If the vnode is in a pseudo filesystem, check whether it is visible.
2572 	 *
2573 	 * XXX if the vnode is a symlink and it is not visible in
2574 	 * a pseudo filesystem, return ENOENT (not following symlink).
2575 	 * V4 client can not mount such symlink. This is a regression
2576 	 * from V2/V3.
2577 	 *
2578 	 * In the same exported filesystem, if the security flavor used
2579 	 * is not an explicitly shared flavor, limit the view to the visible
2580 	 * list entries only. This is not a WRONGSEC case because it's already
2581 	 * checked via PUTROOTFH/PUTPUBFH or PUTFH.
2582 	 */
2583 	if (!different_export &&
2584 	    (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) ||
2585 	    cs->access & CS_ACCESS_LIMITED)) {
2586 		if (! nfs_visible(cs->exi, vp, &different_export)) {
2587 			VN_RELE(vp);
2588 			return (puterrno4(ENOENT));
2589 		}
2590 	}
2591 
2592 	/*
2593 	 * If it's a mountpoint, then traverse it.
2594 	 */
2595 	if (vn_ismntpt(vp)) {
2596 		pre_exi = cs->exi;	/* save pre-traversed exportinfo */
2597 		pre_tvp = vp;		/* save pre-traversed vnode	*/
2598 
2599 		/*
2600 		 * hold pre_tvp to counteract rele by traverse.  We will
2601 		 * need pre_tvp below if checkexport4 fails
2602 		 */
2603 		VN_HOLD(pre_tvp);
2604 		tvp = vp;
2605 		if ((error = traverse(&tvp)) != 0) {
2606 			VN_RELE(vp);
2607 			VN_RELE(pre_tvp);
2608 			return (puterrno4(error));
2609 		}
2610 		vp = tvp;
2611 		different_export = 1;
2612 	} else if (vp->v_vfsp != cs->vp->v_vfsp) {
2613 		/*
2614 		 * The vfsp comparison is to handle the case where
2615 		 * a LOFS mount is shared.  lo_lookup traverses mount points,
2616 		 * and NFS is unaware of local fs transistions because
2617 		 * v_vfsmountedhere isn't set.  For this special LOFS case,
2618 		 * the dir and the obj returned by lookup will have different
2619 		 * vfs ptrs.
2620 		 */
2621 		different_export = 1;
2622 	}
2623 
2624 	if (different_export) {
2625 
2626 		bzero(&fid, sizeof (fid));
2627 		fid.fid_len = MAXFIDSZ;
2628 		error = vop_fid_pseudo(vp, &fid);
2629 		if (error) {
2630 			VN_RELE(vp);
2631 			if (pre_tvp)
2632 				VN_RELE(pre_tvp);
2633 			return (puterrno4(error));
2634 		}
2635 
2636 		if (dotdot)
2637 			exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE);
2638 		else
2639 			exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
2640 
2641 		if (exi == NULL) {
2642 			if (pre_tvp) {
2643 				/*
2644 				 * If this vnode is a mounted-on vnode,
2645 				 * but the mounted-on file system is not
2646 				 * exported, send back the filehandle for
2647 				 * the mounted-on vnode, not the root of
2648 				 * the mounted-on file system.
2649 				 */
2650 				VN_RELE(vp);
2651 				vp = pre_tvp;
2652 				exi = pre_exi;
2653 			} else {
2654 				VN_RELE(vp);
2655 				return (puterrno4(EACCES));
2656 			}
2657 		} else if (pre_tvp) {
2658 			/* we're done with pre_tvp now. release extra hold */
2659 			VN_RELE(pre_tvp);
2660 		}
2661 
2662 		cs->exi = exi;
2663 
2664 		/*
2665 		 * Now we do a checkauth4. The reason is that
2666 		 * this client/user may not have access to the new
2667 		 * exported file system, and if he does,
2668 		 * the client/user may be mapped to a different uid.
2669 		 *
2670 		 * We start with a new cr, because the checkauth4 done
2671 		 * in the PUT*FH operation over wrote the cred's uid,
2672 		 * gid, etc, and we want the real thing before calling
2673 		 * checkauth4()
2674 		 */
2675 		crfree(cs->cr);
2676 		cs->cr = crdup(cs->basecr);
2677 
2678 		if (cs->vp)
2679 			oldvp = cs->vp;
2680 		cs->vp = vp;
2681 		is_newvp = TRUE;
2682 
2683 		stat = call_checkauth4(cs, req);
2684 		if (stat != NFS4_OK) {
2685 			VN_RELE(cs->vp);
2686 			cs->vp = oldvp;
2687 			return (stat);
2688 		}
2689 	}
2690 
2691 	/*
2692 	 * After various NFS checks, do a label check on the path
2693 	 * component. The label on this path should either be the
2694 	 * global zone's label or a zone's label. We are only
2695 	 * interested in the zone's label because exported files
2696 	 * in global zone is accessible (though read-only) to
2697 	 * clients. The exportability/visibility check is already
2698 	 * done before reaching this code.
2699 	 */
2700 	if (is_system_labeled()) {
2701 		bslabel_t *clabel;
2702 
2703 		ASSERT(req->rq_label != NULL);
2704 		clabel = req->rq_label;
2705 		DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *,
2706 		    "got client label from request(1)", struct svc_req *, req);
2707 
2708 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
2709 			if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) {
2710 				error = EACCES;
2711 				goto err_out;
2712 			}
2713 		} else {
2714 			/*
2715 			 * We grant access to admin_low label clients
2716 			 * only if the client is trusted, i.e. also
2717 			 * running Solaris Trusted Extension.
2718 			 */
2719 			struct sockaddr	*ca;
2720 			int		addr_type;
2721 			void		*ipaddr;
2722 			tsol_tpc_t	*tp;
2723 
2724 			ca = (struct sockaddr *)svc_getrpccaller(
2725 			    req->rq_xprt)->buf;
2726 			if (ca->sa_family == AF_INET) {
2727 				addr_type = IPV4_VERSION;
2728 				ipaddr = &((struct sockaddr_in *)ca)->sin_addr;
2729 			} else if (ca->sa_family == AF_INET6) {
2730 				addr_type = IPV6_VERSION;
2731 				ipaddr = &((struct sockaddr_in6 *)
2732 				    ca)->sin6_addr;
2733 			}
2734 			tp = find_tpc(ipaddr, addr_type, B_FALSE);
2735 			if (tp == NULL || tp->tpc_tp.tp_doi !=
2736 			    l_admin_low->tsl_doi || tp->tpc_tp.host_type !=
2737 			    SUN_CIPSO) {
2738 				if (tp != NULL)
2739 					TPC_RELE(tp);
2740 				error = EACCES;
2741 				goto err_out;
2742 			}
2743 			TPC_RELE(tp);
2744 		}
2745 	}
2746 
2747 	error = makefh4(&cs->fh, vp, cs->exi);
2748 
2749 err_out:
2750 	if (error) {
2751 		if (is_newvp) {
2752 			VN_RELE(cs->vp);
2753 			cs->vp = oldvp;
2754 		} else
2755 			VN_RELE(vp);
2756 		return (puterrno4(error));
2757 	}
2758 
2759 	if (!is_newvp) {
2760 		if (cs->vp)
2761 			VN_RELE(cs->vp);
2762 		cs->vp = vp;
2763 	} else if (oldvp)
2764 		VN_RELE(oldvp);
2765 
2766 	/*
2767 	 * if did lookup on attrdir and didn't lookup .., set named
2768 	 * attr fh flag
2769 	 */
2770 	if (attrdir && ! dotdot)
2771 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
2772 
2773 	/* Assume false for now, open proc will set this */
2774 	cs->mandlock = FALSE;
2775 
2776 	return (NFS4_OK);
2777 }
2778 
2779 /* ARGSUSED */
2780 static void
2781 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2782 	struct compound_state *cs)
2783 {
2784 	LOOKUP4args *args = &argop->nfs_argop4_u.oplookup;
2785 	LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup;
2786 	char *nm;
2787 	uint_t len;
2788 
2789 	DTRACE_NFSV4_2(op__lookup__start, struct compound_state *, cs,
2790 	    LOOKUP4args *, args);
2791 
2792 	if (cs->vp == NULL) {
2793 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2794 		goto out;
2795 	}
2796 
2797 	if (cs->vp->v_type == VLNK) {
2798 		*cs->statusp = resp->status = NFS4ERR_SYMLINK;
2799 		goto out;
2800 	}
2801 
2802 	if (cs->vp->v_type != VDIR) {
2803 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2804 		goto out;
2805 	}
2806 
2807 	if (!utf8_dir_verify(&args->objname)) {
2808 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2809 		goto out;
2810 	}
2811 
2812 	nm = utf8_to_str(&args->objname, &len, NULL);
2813 	if (nm == NULL) {
2814 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2815 		goto out;
2816 	}
2817 
2818 	if (len > MAXNAMELEN) {
2819 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
2820 		kmem_free(nm, len);
2821 		goto out;
2822 	}
2823 
2824 	*cs->statusp = resp->status = do_rfs4_op_lookup(nm, len, req, cs);
2825 
2826 	kmem_free(nm, len);
2827 
2828 out:
2829 	DTRACE_NFSV4_2(op__lookup__done, struct compound_state *, cs,
2830 	    LOOKUP4res *, resp);
2831 }
2832 
2833 /* ARGSUSED */
2834 static void
2835 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
2836 	struct compound_state *cs)
2837 {
2838 	LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp;
2839 
2840 	DTRACE_NFSV4_1(op__lookupp__start, struct compound_state *, cs);
2841 
2842 	if (cs->vp == NULL) {
2843 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2844 		goto out;
2845 	}
2846 
2847 	if (cs->vp->v_type != VDIR) {
2848 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2849 		goto out;
2850 	}
2851 
2852 	*cs->statusp = resp->status = do_rfs4_op_lookup("..", 3, req, cs);
2853 
2854 	/*
2855 	 * From NFSV4 Specification, LOOKUPP should not check for
2856 	 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead.
2857 	 */
2858 	if (resp->status == NFS4ERR_WRONGSEC) {
2859 		*cs->statusp = resp->status = NFS4_OK;
2860 	}
2861 
2862 out:
2863 	DTRACE_NFSV4_2(op__lookupp__done, struct compound_state *, cs,
2864 	    LOOKUPP4res *, resp);
2865 }
2866 
2867 
2868 /*ARGSUSED2*/
2869 static void
2870 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2871 	struct compound_state *cs)
2872 {
2873 	OPENATTR4args	*args = &argop->nfs_argop4_u.opopenattr;
2874 	OPENATTR4res	*resp = &resop->nfs_resop4_u.opopenattr;
2875 	vnode_t		*avp = NULL;
2876 	int		lookup_flags = LOOKUP_XATTR, error;
2877 	int		exp_ro = 0;
2878 
2879 	DTRACE_NFSV4_2(op__openattr__start, struct compound_state *, cs,
2880 	    OPENATTR4args *, args);
2881 
2882 	if (cs->vp == NULL) {
2883 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2884 		goto out;
2885 	}
2886 
2887 	/*
2888 	 * Make a couple of checks made by copen()
2889 	 *
2890 	 * Check to make sure underlying fs supports xattrs.  This
2891 	 * is required because solaris filesystem implementations
2892 	 * (UFS/TMPFS) don't enforce the noxattr mount option
2893 	 * in VOP_LOOKUP(LOOKUP_XATTR).  If fs doesn't support this
2894 	 * pathconf cmd or if fs supports cmd but doesn't claim
2895 	 * support for xattr, return NOTSUPP.  It would be better
2896 	 * to use VOP_PATHCONF( _PC_XATTR_ENABLED) for this; however,
2897 	 * that cmd is not available to VOP_PATHCONF interface
2898 	 * (it's only implemented inside pathconf syscall)...
2899 	 *
2900 	 * Verify permission to put attributes on files (access
2901 	 * checks from copen).
2902 	 */
2903 
2904 	if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0 &&
2905 	    !vfs_has_feature(cs->vp->v_vfsp, VFSFT_XVATTR)) {
2906 		*cs->statusp = resp->status = puterrno4(ENOTSUP);
2907 		goto out;
2908 	}
2909 
2910 	/*
2911 	 * If file system supports passing ACE mask to VOP_ACCESS then
2912 	 * check for ACE_READ_NAMED_ATTRS, otherwise do legacy checks
2913 	 */
2914 
2915 	if (vfs_has_feature(cs->vp->v_vfsp, VFSFT_ACEMASKONACCESS))
2916 		error = VOP_ACCESS(cs->vp, ACE_READ_NAMED_ATTRS,
2917 		    V_ACE_MASK, cs->cr, NULL);
2918 	else
2919 		error = ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr, NULL) != 0) &&
2920 		    (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr, NULL) != 0) &&
2921 		    (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr, NULL) != 0));
2922 
2923 	if (error) {
2924 		*cs->statusp = resp->status = puterrno4(EACCES);
2925 		goto out;
2926 	}
2927 
2928 	/*
2929 	 * The CREATE_XATTR_DIR VOP flag cannot be specified if
2930 	 * the file system is exported read-only -- regardless of
2931 	 * createdir flag.  Otherwise the attrdir would be created
2932 	 * (assuming server fs isn't mounted readonly locally).  If
2933 	 * VOP_LOOKUP returns ENOENT in this case, the error will
2934 	 * be translated into EROFS.  ENOSYS is mapped to ENOTSUP
2935 	 * because specfs has no VOP_LOOKUP op, so the macro would
2936 	 * return ENOSYS.  EINVAL is returned by all (current)
2937 	 * Solaris file system implementations when any of their
2938 	 * restrictions are violated (xattr(dir) can't have xattrdir).
2939 	 * Returning NOTSUPP is more appropriate in this case
2940 	 * because the object will never be able to have an attrdir.
2941 	 */
2942 	if (args->createdir && ! (exp_ro = rdonly4(cs->exi, cs->vp, req)))
2943 		lookup_flags |= CREATE_XATTR_DIR;
2944 
2945 	error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr,
2946 	    NULL, NULL, NULL);
2947 
2948 	if (error) {
2949 		if (error == ENOENT && args->createdir && exp_ro)
2950 			*cs->statusp = resp->status = puterrno4(EROFS);
2951 		else if (error == EINVAL || error == ENOSYS)
2952 			*cs->statusp = resp->status = puterrno4(ENOTSUP);
2953 		else
2954 			*cs->statusp = resp->status = puterrno4(error);
2955 		goto out;
2956 	}
2957 
2958 	ASSERT(avp->v_flag & V_XATTRDIR);
2959 
2960 	error = makefh4(&cs->fh, avp, cs->exi);
2961 
2962 	if (error) {
2963 		VN_RELE(avp);
2964 		*cs->statusp = resp->status = puterrno4(error);
2965 		goto out;
2966 	}
2967 
2968 	VN_RELE(cs->vp);
2969 	cs->vp = avp;
2970 
2971 	/*
2972 	 * There is no requirement for an attrdir fh flag
2973 	 * because the attrdir has a vnode flag to distinguish
2974 	 * it from regular (non-xattr) directories.  The
2975 	 * FH4_ATTRDIR flag is set for future sanity checks.
2976 	 */
2977 	set_fh4_flag(&cs->fh, FH4_ATTRDIR);
2978 	*cs->statusp = resp->status = NFS4_OK;
2979 
2980 out:
2981 	DTRACE_NFSV4_2(op__openattr__done, struct compound_state *, cs,
2982 	    OPENATTR4res *, resp);
2983 }
2984 
2985 static int
2986 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred,
2987     caller_context_t *ct)
2988 {
2989 	int error;
2990 	int i;
2991 	clock_t delaytime;
2992 
2993 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
2994 
2995 	/*
2996 	 * Don't block on mandatory locks. If this routine returns
2997 	 * EAGAIN, the caller should return NFS4ERR_LOCKED.
2998 	 */
2999 	uio->uio_fmode = FNONBLOCK;
3000 
3001 	for (i = 0; i < rfs4_maxlock_tries; i++) {
3002 
3003 
3004 		if (direction == FREAD) {
3005 			(void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, ct);
3006 			error = VOP_READ(vp, uio, ioflag, cred, ct);
3007 			VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, ct);
3008 		} else {
3009 			(void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, ct);
3010 			error = VOP_WRITE(vp, uio, ioflag, cred, ct);
3011 			VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, ct);
3012 		}
3013 
3014 		if (error != EAGAIN)
3015 			break;
3016 
3017 		if (i < rfs4_maxlock_tries - 1) {
3018 			delay(delaytime);
3019 			delaytime *= 2;
3020 		}
3021 	}
3022 
3023 	return (error);
3024 }
3025 
3026 /* ARGSUSED */
3027 static void
3028 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3029 	struct compound_state *cs)
3030 {
3031 	READ4args *args = &argop->nfs_argop4_u.opread;
3032 	READ4res *resp = &resop->nfs_resop4_u.opread;
3033 	int error;
3034 	int verror;
3035 	vnode_t *vp;
3036 	struct vattr va;
3037 	struct iovec iov;
3038 	struct uio uio;
3039 	u_offset_t offset;
3040 	bool_t *deleg = &cs->deleg;
3041 	nfsstat4 stat;
3042 	int in_crit = 0;
3043 	mblk_t *mp;
3044 	int alloc_err = 0;
3045 	caller_context_t ct;
3046 
3047 	DTRACE_NFSV4_2(op__read__start, struct compound_state *, cs,
3048 	    READ4args, args);
3049 
3050 	vp = cs->vp;
3051 	if (vp == NULL) {
3052 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3053 		goto out;
3054 	}
3055 	if (cs->access == CS_ACCESS_DENIED) {
3056 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3057 		goto out;
3058 	}
3059 
3060 	if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE,
3061 	    deleg, TRUE, &ct)) != NFS4_OK) {
3062 		*cs->statusp = resp->status = stat;
3063 		goto out;
3064 	}
3065 
3066 	/*
3067 	 * Enter the critical region before calling VOP_RWLOCK
3068 	 * to avoid a deadlock with write requests.
3069 	 */
3070 	if (nbl_need_check(vp)) {
3071 		nbl_start_crit(vp, RW_READER);
3072 		in_crit = 1;
3073 		if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0,
3074 		    &ct)) {
3075 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
3076 			goto out;
3077 		}
3078 	}
3079 
3080 	if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE,
3081 	    deleg, TRUE, &ct)) != NFS4_OK) {
3082 		*cs->statusp = resp->status = stat;
3083 		goto out;
3084 	}
3085 
3086 	va.va_mask = AT_MODE|AT_SIZE|AT_UID;
3087 	verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct);
3088 
3089 	/*
3090 	 * If we can't get the attributes, then we can't do the
3091 	 * right access checking.  So, we'll fail the request.
3092 	 */
3093 	if (verror) {
3094 		*cs->statusp = resp->status = puterrno4(verror);
3095 		goto out;
3096 	}
3097 
3098 	if (vp->v_type != VREG) {
3099 		*cs->statusp = resp->status =
3100 		    ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
3101 		goto out;
3102 	}
3103 
3104 	if (crgetuid(cs->cr) != va.va_uid &&
3105 	    (error = VOP_ACCESS(vp, VREAD, 0, cs->cr, &ct)) &&
3106 	    (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr, &ct))) {
3107 		*cs->statusp = resp->status = puterrno4(error);
3108 		goto out;
3109 	}
3110 
3111 	if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */
3112 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3113 		goto out;
3114 	}
3115 
3116 	offset = args->offset;
3117 	if (offset >= va.va_size) {
3118 		*cs->statusp = resp->status = NFS4_OK;
3119 		resp->eof = TRUE;
3120 		resp->data_len = 0;
3121 		resp->data_val = NULL;
3122 		resp->mblk = NULL;
3123 		*cs->statusp = resp->status = NFS4_OK;
3124 		goto out;
3125 	}
3126 
3127 	if (args->count == 0) {
3128 		*cs->statusp = resp->status = NFS4_OK;
3129 		resp->eof = FALSE;
3130 		resp->data_len = 0;
3131 		resp->data_val = NULL;
3132 		resp->mblk = NULL;
3133 		goto out;
3134 	}
3135 
3136 	/*
3137 	 * Do not allocate memory more than maximum allowed
3138 	 * transfer size
3139 	 */
3140 	if (args->count > rfs4_tsize(req))
3141 		args->count = rfs4_tsize(req);
3142 
3143 	/*
3144 	 * mp will contain the data to be sent out in the read reply.
3145 	 * It will be freed after the reply has been sent.
3146 	 * Let's roundup the data to a BYTES_PER_XDR_UNIT multiple,
3147 	 * so that the call to xdrmblk_putmblk() never fails.
3148 	 * If the first alloc of the requested size fails, then
3149 	 * decrease the size to something more reasonable and wait
3150 	 * for the allocation to occur.
3151 	 */
3152 	mp = allocb(RNDUP(args->count), BPRI_MED);
3153 	if (mp == NULL) {
3154 		if (args->count > MAXBSIZE)
3155 			args->count = MAXBSIZE;
3156 		mp = allocb_wait(RNDUP(args->count), BPRI_MED,
3157 		    STR_NOSIG, &alloc_err);
3158 	}
3159 	ASSERT(mp != NULL);
3160 	ASSERT(alloc_err == 0);
3161 
3162 	iov.iov_base = (caddr_t)mp->b_datap->db_base;
3163 	iov.iov_len = args->count;
3164 	uio.uio_iov = &iov;
3165 	uio.uio_iovcnt = 1;
3166 	uio.uio_segflg = UIO_SYSSPACE;
3167 	uio.uio_extflg = UIO_COPY_CACHED;
3168 	uio.uio_loffset = args->offset;
3169 	uio.uio_resid = args->count;
3170 
3171 	error = do_io(FREAD, vp, &uio, 0, cs->cr, &ct);
3172 
3173 	va.va_mask = AT_SIZE;
3174 	verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct);
3175 
3176 	if (error) {
3177 		freeb(mp);
3178 		*cs->statusp = resp->status = puterrno4(error);
3179 		goto out;
3180 	}
3181 
3182 	*cs->statusp = resp->status = NFS4_OK;
3183 
3184 	ASSERT(uio.uio_resid >= 0);
3185 	resp->data_len = args->count - uio.uio_resid;
3186 	resp->data_val = (char *)mp->b_datap->db_base;
3187 	resp->mblk = mp;
3188 
3189 	if (!verror && offset + resp->data_len == va.va_size)
3190 		resp->eof = TRUE;
3191 	else
3192 		resp->eof = FALSE;
3193 
3194 out:
3195 	if (in_crit)
3196 		nbl_end_crit(vp);
3197 
3198 	DTRACE_NFSV4_2(op__read__done, struct compound_state *, cs,
3199 	    READ4res *, resp);
3200 }
3201 
3202 static void
3203 rfs4_op_read_free(nfs_resop4 *resop)
3204 {
3205 	READ4res *resp = &resop->nfs_resop4_u.opread;
3206 
3207 	if (resp->status == NFS4_OK && resp->mblk != NULL) {
3208 		freeb(resp->mblk);
3209 		resp->mblk = NULL;
3210 		resp->data_val = NULL;
3211 		resp->data_len = 0;
3212 	}
3213 }
3214 
3215 static void
3216 rfs4_op_readdir_free(nfs_resop4 *resop)
3217 {
3218 	READDIR4res *resp = &resop->nfs_resop4_u.opreaddir;
3219 
3220 	if (resp->status == NFS4_OK && resp->mblk != NULL) {
3221 		freeb(resp->mblk);
3222 		resp->mblk = NULL;
3223 		resp->data_len = 0;
3224 	}
3225 }
3226 
3227 
3228 /* ARGSUSED */
3229 static void
3230 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
3231 	struct compound_state *cs)
3232 {
3233 	PUTPUBFH4res *resp = &resop->nfs_resop4_u.opputpubfh;
3234 	int error;
3235 	vnode_t *vp;
3236 	struct exportinfo *exi, *sav_exi;
3237 	nfs_fh4_fmt_t *fh_fmtp;
3238 
3239 	DTRACE_NFSV4_1(op__putpubfh__start, struct compound_state *, cs);
3240 
3241 	if (cs->vp) {
3242 		VN_RELE(cs->vp);
3243 		cs->vp = NULL;
3244 	}
3245 
3246 	if (cs->cr)
3247 		crfree(cs->cr);
3248 
3249 	cs->cr = crdup(cs->basecr);
3250 
3251 	vp = exi_public->exi_vp;
3252 	if (vp == NULL) {
3253 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
3254 		goto out;
3255 	}
3256 
3257 	if (is_system_labeled()) {
3258 		bslabel_t *clabel;
3259 
3260 		ASSERT(req->rq_label != NULL);
3261 		clabel = req->rq_label;
3262 		DTRACE_PROBE2(tx__rfs4__log__info__opputpubfh__clabel, char *,
3263 		    "got client label from request(1)",
3264 		    struct svc_req *, req);
3265 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
3266 			if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) {
3267 				*cs->statusp = resp->status =
3268 				    NFS4ERR_SERVERFAULT;
3269 				return;
3270 			}
3271 		}
3272 	}
3273 
3274 	error = makefh4(&cs->fh, vp, exi_public);
3275 	if (error != 0) {
3276 		*cs->statusp = resp->status = puterrno4(error);
3277 		goto out;
3278 	}
3279 	sav_exi = cs->exi;
3280 	if (exi_public == exi_root) {
3281 		/*
3282 		 * No filesystem is actually shared public, so we default
3283 		 * to exi_root. In this case, we must check whether root
3284 		 * is exported.
3285 		 */
3286 		fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val;
3287 
3288 		/*
3289 		 * if root filesystem is exported, the exportinfo struct that we
3290 		 * should use is what checkexport4 returns, because root_exi is
3291 		 * actually a mostly empty struct.
3292 		 */
3293 		exi = checkexport4(&fh_fmtp->fh4_fsid,
3294 		    (fid_t *)&fh_fmtp->fh4_xlen, NULL);
3295 		cs->exi = ((exi != NULL) ? exi : exi_public);
3296 	} else {
3297 		/*
3298 		 * it's a properly shared filesystem
3299 		 */
3300 		cs->exi = exi_public;
3301 	}
3302 
3303 	VN_HOLD(vp);
3304 	cs->vp = vp;
3305 
3306 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3307 		VN_RELE(cs->vp);
3308 		cs->vp = NULL;
3309 		cs->exi = sav_exi;
3310 		goto out;
3311 	}
3312 
3313 	*cs->statusp = resp->status = NFS4_OK;
3314 out:
3315 	DTRACE_NFSV4_2(op__putpubfh__done, struct compound_state *, cs,
3316 	    PUTPUBFH4res *, resp);
3317 }
3318 
3319 /*
3320  * XXX - issue with put*fh operations. Suppose /export/home is exported.
3321  * Suppose an NFS client goes to mount /export/home/joe. If /export, home,
3322  * or joe have restrictive search permissions, then we shouldn't let
3323  * the client get a file handle. This is easy to enforce. However, we
3324  * don't know what security flavor should be used until we resolve the
3325  * path name. Another complication is uid mapping. If root is
3326  * the user, then it will be mapped to the anonymous user by default,
3327  * but we won't know that till we've resolved the path name. And we won't
3328  * know what the anonymous user is.
3329  * Luckily, SECINFO is specified to take a full filename.
3330  * So what we will have to in rfs4_op_lookup is check that flavor of
3331  * the target object matches that of the request, and if root was the
3332  * caller, check for the root= and anon= options, and if necessary,
3333  * repeat the lookup using the right cred_t. But that's not done yet.
3334  */
3335 /* ARGSUSED */
3336 static void
3337 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3338 	struct compound_state *cs)
3339 {
3340 	PUTFH4args *args = &argop->nfs_argop4_u.opputfh;
3341 	PUTFH4res *resp = &resop->nfs_resop4_u.opputfh;
3342 	nfs_fh4_fmt_t *fh_fmtp;
3343 
3344 	DTRACE_NFSV4_2(op__putfh__start, struct compound_state *, cs,
3345 	    PUTFH4args *, args);
3346 
3347 	if (cs->vp) {
3348 		VN_RELE(cs->vp);
3349 		cs->vp = NULL;
3350 	}
3351 
3352 	if (cs->cr) {
3353 		crfree(cs->cr);
3354 		cs->cr = NULL;
3355 	}
3356 
3357 
3358 	if (args->object.nfs_fh4_len < NFS_FH4_LEN) {
3359 		*cs->statusp = resp->status = NFS4ERR_BADHANDLE;
3360 		goto out;
3361 	}
3362 
3363 	fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val;
3364 	cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen,
3365 	    NULL);
3366 
3367 	if (cs->exi == NULL) {
3368 		*cs->statusp = resp->status = NFS4ERR_STALE;
3369 		goto out;
3370 	}
3371 
3372 	cs->cr = crdup(cs->basecr);
3373 
3374 	ASSERT(cs->cr != NULL);
3375 
3376 	if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) {
3377 		*cs->statusp = resp->status;
3378 		goto out;
3379 	}
3380 
3381 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3382 		VN_RELE(cs->vp);
3383 		cs->vp = NULL;
3384 		goto out;
3385 	}
3386 
3387 	nfs_fh4_copy(&args->object, &cs->fh);
3388 	*cs->statusp = resp->status = NFS4_OK;
3389 	cs->deleg = FALSE;
3390 
3391 out:
3392 	DTRACE_NFSV4_2(op__putfh__done, struct compound_state *, cs,
3393 	    PUTFH4res *, resp);
3394 }
3395 
3396 /* ARGSUSED */
3397 static void
3398 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3399 	struct compound_state *cs)
3400 
3401 {
3402 	PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh;
3403 	int error;
3404 	fid_t fid;
3405 	struct exportinfo *exi, *sav_exi;
3406 
3407 	DTRACE_NFSV4_1(op__putrootfh__start, struct compound_state *, cs);
3408 
3409 	if (cs->vp) {
3410 		VN_RELE(cs->vp);
3411 		cs->vp = NULL;
3412 	}
3413 
3414 	if (cs->cr)
3415 		crfree(cs->cr);
3416 
3417 	cs->cr = crdup(cs->basecr);
3418 
3419 	/*
3420 	 * Using rootdir, the system root vnode,
3421 	 * get its fid.
3422 	 */
3423 	bzero(&fid, sizeof (fid));
3424 	fid.fid_len = MAXFIDSZ;
3425 	error = vop_fid_pseudo(rootdir, &fid);
3426 	if (error != 0) {
3427 		*cs->statusp = resp->status = puterrno4(error);
3428 		goto out;
3429 	}
3430 
3431 	/*
3432 	 * Then use the root fsid & fid it to find out if it's exported
3433 	 *
3434 	 * If the server root isn't exported directly, then
3435 	 * it should at least be a pseudo export based on
3436 	 * one or more exports further down in the server's
3437 	 * file tree.
3438 	 */
3439 	exi = checkexport4(&rootdir->v_vfsp->vfs_fsid, &fid, NULL);
3440 	if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) {
3441 		NFS4_DEBUG(rfs4_debug,
3442 			(CE_WARN, "rfs4_op_putrootfh: export check failure"));
3443 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
3444 		goto out;
3445 	}
3446 
3447 	/*
3448 	 * Now make a filehandle based on the root
3449 	 * export and root vnode.
3450 	 */
3451 	error = makefh4(&cs->fh, rootdir, exi);
3452 	if (error != 0) {
3453 		*cs->statusp = resp->status = puterrno4(error);
3454 		goto out;
3455 	}
3456 
3457 	sav_exi = cs->exi;
3458 	cs->exi = exi;
3459 
3460 	VN_HOLD(rootdir);
3461 	cs->vp = rootdir;
3462 
3463 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3464 		VN_RELE(rootdir);
3465 		cs->vp = NULL;
3466 		cs->exi = sav_exi;
3467 		goto out;
3468 	}
3469 
3470 	*cs->statusp = resp->status = NFS4_OK;
3471 	cs->deleg = FALSE;
3472 out:
3473 	DTRACE_NFSV4_2(op__putrootfh__done, struct compound_state *, cs,
3474 	    PUTROOTFH4res *, resp);
3475 }
3476 
3477 /*
3478  * A directory entry is a valid nfsv4 entry if
3479  * - it has a non-zero ino
3480  * - it is not a dot or dotdot name
3481  * - it is visible in a pseudo export or in a real export that can
3482  *   only have a limited view.
3483  */
3484 static bool_t
3485 valid_nfs4_entry(struct exportinfo *exi, struct dirent64 *dp,
3486 		int *expseudo, int check_visible)
3487 {
3488 	if (dp->d_ino == 0 || NFS_IS_DOTNAME(dp->d_name)) {
3489 		*expseudo = 0;
3490 		return (FALSE);
3491 	}
3492 
3493 	if (! check_visible) {
3494 		*expseudo = 0;
3495 		return (TRUE);
3496 	}
3497 
3498 	return (nfs_visible_inode(exi, dp->d_ino, expseudo));
3499 }
3500 
3501 /*
3502  * set_rdattr_params sets up the variables used to manage what information
3503  * to get for each directory entry.
3504  */
3505 static nfsstat4
3506 set_rdattr_params(struct nfs4_svgetit_arg *sargp,
3507 		bitmap4 attrs, bool_t *need_to_lookup)
3508 {
3509 	uint_t	va_mask;
3510 	nfsstat4 status;
3511 	bitmap4 objbits;
3512 
3513 	status = bitmap4_to_attrmask(attrs, sargp);
3514 	if (status != NFS4_OK) {
3515 		/*
3516 		 * could not even figure attr mask
3517 		 */
3518 		return (status);
3519 	}
3520 	va_mask = sargp->vap->va_mask;
3521 
3522 	/*
3523 	 * dirent's d_ino is always correct value for mounted_on_fileid.
3524 	 * mntdfid_set is set once here, but mounted_on_fileid is
3525 	 * set in main dirent processing loop for each dirent.
3526 	 * The mntdfid_set is a simple optimization that lets the
3527 	 * server attr code avoid work when caller is readdir.
3528 	 */
3529 	sargp->mntdfid_set = TRUE;
3530 
3531 	/*
3532 	 * Lookup entry only if client asked for any of the following:
3533 	 * a) vattr attrs
3534 	 * b) vfs attrs
3535 	 * c) attrs w/per-object scope requested (change, filehandle, etc)
3536 	 *    other than mounted_on_fileid (which we can take from dirent)
3537 	 */
3538 	objbits = attrs ? attrs & NFS4_VP_ATTR_MASK : 0;
3539 
3540 	if (va_mask || sargp->sbp || (objbits & ~FATTR4_MOUNTED_ON_FILEID_MASK))
3541 		*need_to_lookup = TRUE;
3542 	else
3543 		*need_to_lookup = FALSE;
3544 
3545 	if (sargp->sbp == NULL)
3546 		return (NFS4_OK);
3547 
3548 	/*
3549 	 * If filesystem attrs are requested, get them now from the
3550 	 * directory vp, as most entries will have same filesystem. The only
3551 	 * exception are mounted over entries but we handle
3552 	 * those as we go (XXX mounted over detection not yet implemented).
3553 	 */
3554 	sargp->vap->va_mask = 0;	/* to avoid VOP_GETATTR */
3555 	status = bitmap4_get_sysattrs(sargp);
3556 	sargp->vap->va_mask = va_mask;
3557 
3558 	if ((status != NFS4_OK) && sargp->rdattr_error_req) {
3559 		/*
3560 		 * Failed to get filesystem attributes.
3561 		 * Return a rdattr_error for each entry, but don't fail.
3562 		 * However, don't get any obj-dependent attrs.
3563 		 */
3564 		sargp->rdattr_error = status;	/* for rdattr_error */
3565 		*need_to_lookup = FALSE;
3566 		/*
3567 		 * At least get fileid for regular readdir output
3568 		 */
3569 		sargp->vap->va_mask &= AT_NODEID;
3570 		status = NFS4_OK;
3571 	}
3572 
3573 	return (status);
3574 }
3575 
3576 /*
3577  * readlink: args: CURRENT_FH.
3578  *	res: status. If success - CURRENT_FH unchanged, return linktext.
3579  */
3580 
3581 /* ARGSUSED */
3582 static void
3583 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3584 	struct compound_state *cs)
3585 {
3586 	READLINK4res *resp = &resop->nfs_resop4_u.opreadlink;
3587 	int error;
3588 	vnode_t *vp;
3589 	struct iovec iov;
3590 	struct vattr va;
3591 	struct uio uio;
3592 	char *data;
3593 
3594 	DTRACE_NFSV4_1(op__readlink__start, struct compound_state *, cs);
3595 
3596 	/* CURRENT_FH: directory */
3597 	vp = cs->vp;
3598 	if (vp == NULL) {
3599 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3600 		goto out;
3601 	}
3602 
3603 	if (cs->access == CS_ACCESS_DENIED) {
3604 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3605 		goto out;
3606 	}
3607 
3608 	if (vp->v_type == VDIR) {
3609 		*cs->statusp = resp->status = NFS4ERR_ISDIR;
3610 		goto out;
3611 	}
3612 
3613 	if (vp->v_type != VLNK) {
3614 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3615 		goto out;
3616 	}
3617 
3618 	va.va_mask = AT_MODE;
3619 	error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL);
3620 	if (error) {
3621 		*cs->statusp = resp->status = puterrno4(error);
3622 		goto out;
3623 	}
3624 
3625 	if (MANDLOCK(vp, va.va_mode)) {
3626 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3627 		goto out;
3628 	}
3629 
3630 	data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP);
3631 
3632 	iov.iov_base = data;
3633 	iov.iov_len = MAXPATHLEN;
3634 	uio.uio_iov = &iov;
3635 	uio.uio_iovcnt = 1;
3636 	uio.uio_segflg = UIO_SYSSPACE;
3637 	uio.uio_extflg = UIO_COPY_CACHED;
3638 	uio.uio_loffset = 0;
3639 	uio.uio_resid = MAXPATHLEN;
3640 
3641 	error = VOP_READLINK(vp, &uio, cs->cr, NULL);
3642 
3643 	if (error) {
3644 		kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1);
3645 		*cs->statusp = resp->status = puterrno4(error);
3646 		goto out;
3647 	}
3648 
3649 	*(data + MAXPATHLEN - uio.uio_resid) = '\0';
3650 
3651 	/*
3652 	 * treat link name as data
3653 	 */
3654 	(void) str_to_utf8(data, &resp->link);
3655 
3656 	kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1);
3657 	*cs->statusp = resp->status = NFS4_OK;
3658 
3659 out:
3660 	DTRACE_NFSV4_2(op__readlink__done, struct compound_state *, cs,
3661 	    READLINK4res *, resp);
3662 }
3663 
3664 static void
3665 rfs4_op_readlink_free(nfs_resop4 *resop)
3666 {
3667 	READLINK4res *resp = &resop->nfs_resop4_u.opreadlink;
3668 	utf8string *symlink = &resp->link;
3669 
3670 	if (symlink->utf8string_val) {
3671 		UTF8STRING_FREE(*symlink)
3672 	}
3673 }
3674 
3675 /*
3676  * release_lockowner:
3677  *	Release any state associated with the supplied
3678  *	lockowner. Note if any lo_state is holding locks we will not
3679  *	rele that lo_state and thus the lockowner will not be destroyed.
3680  *	A client using lock after the lock owner stateid has been released
3681  *	will suffer the consequence of NFS4ERR_BAD_STATEID and would have
3682  *	to reissue the lock with new_lock_owner set to TRUE.
3683  *	args: lock_owner
3684  *	res:  status
3685  */
3686 /* ARGSUSED */
3687 static void
3688 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop,
3689 	struct svc_req *req, struct compound_state *cs)
3690 {
3691 	RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner;
3692 	RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner;
3693 	rfs4_lockowner_t *lo;
3694 	rfs4_openowner_t *oop;
3695 	rfs4_state_t *sp;
3696 	rfs4_lo_state_t *lsp;
3697 	rfs4_client_t *cp;
3698 	bool_t create = FALSE;
3699 	locklist_t *llist;
3700 	sysid_t sysid;
3701 
3702 	DTRACE_NFSV4_2(op__release__lockowner__start, struct compound_state *,
3703 	    cs, RELEASE_LOCKOWNER4args *, ap);
3704 
3705 	/* Make sure there is a clientid around for this request */
3706 	cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE);
3707 
3708 	if (cp == NULL) {
3709 		*cs->statusp = resp->status =
3710 		    rfs4_check_clientid(&ap->lock_owner.clientid, 0);
3711 		goto out;
3712 	}
3713 	rfs4_client_rele(cp);
3714 
3715 	lo = rfs4_findlockowner(&ap->lock_owner, &create);
3716 	if (lo == NULL) {
3717 		*cs->statusp = resp->status = NFS4_OK;
3718 		goto out;
3719 	}
3720 	ASSERT(lo->client != NULL);
3721 
3722 	/*
3723 	 * Check for EXPIRED client. If so will reap state with in a lease
3724 	 * period or on next set_clientid_confirm step
3725 	 */
3726 	if (rfs4_lease_expired(lo->client)) {
3727 		rfs4_lockowner_rele(lo);
3728 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
3729 		goto out;
3730 	}
3731 
3732 	/*
3733 	 * If no sysid has been assigned, then no locks exist; just return.
3734 	 */
3735 	rfs4_dbe_lock(lo->client->dbe);
3736 	if (lo->client->sysidt == LM_NOSYSID) {
3737 		rfs4_lockowner_rele(lo);
3738 		rfs4_dbe_unlock(lo->client->dbe);
3739 		goto out;
3740 	}
3741 
3742 	sysid = lo->client->sysidt;
3743 	rfs4_dbe_unlock(lo->client->dbe);
3744 
3745 	/*
3746 	 * Mark the lockowner invalid.
3747 	 */
3748 	rfs4_dbe_hide(lo->dbe);
3749 
3750 	/*
3751 	 * sysid-pid pair should now not be used since the lockowner is
3752 	 * invalid. If the client were to instantiate the lockowner again
3753 	 * it would be assigned a new pid. Thus we can get the list of
3754 	 * current locks.
3755 	 */
3756 
3757 	llist = flk_get_active_locks(sysid, lo->pid);
3758 	/* If we are still holding locks fail */
3759 	if (llist != NULL) {
3760 
3761 		*cs->statusp = resp->status = NFS4ERR_LOCKS_HELD;
3762 
3763 		flk_free_locklist(llist);
3764 		/*
3765 		 * We need to unhide the lockowner so the client can
3766 		 * try it again. The bad thing here is if the client
3767 		 * has a logic error that took it here in the first place
3768 		 * he probably has lost accounting of the locks that it
3769 		 * is holding. So we may have dangling state until the
3770 		 * open owner state is reaped via close. One scenario
3771 		 * that could possibly occur is that the client has
3772 		 * sent the unlock request(s) in separate threads
3773 		 * and has not waited for the replies before sending the
3774 		 * RELEASE_LOCKOWNER request. Presumably, it would expect
3775 		 * and deal appropriately with NFS4ERR_LOCKS_HELD, by
3776 		 * reissuing the request.
3777 		 */
3778 		rfs4_dbe_unhide(lo->dbe);
3779 		rfs4_lockowner_rele(lo);
3780 		goto out;
3781 	}
3782 
3783 	/*
3784 	 * For the corresponding client we need to check each open
3785 	 * owner for any opens that have lockowner state associated
3786 	 * with this lockowner.
3787 	 */
3788 
3789 	rfs4_dbe_lock(lo->client->dbe);
3790 	for (oop = lo->client->openownerlist.next->oop; oop != NULL;
3791 	    oop = oop->openownerlist.next->oop) {
3792 
3793 		rfs4_dbe_lock(oop->dbe);
3794 		for (sp = oop->ownerstateids.next->sp; sp != NULL;
3795 		    sp = sp->ownerstateids.next->sp) {
3796 
3797 			rfs4_dbe_lock(sp->dbe);
3798 			for (lsp = sp->lockownerlist.next->lsp;
3799 			    lsp != NULL; lsp = lsp->lockownerlist.next->lsp) {
3800 				if (lsp->locker == lo) {
3801 					rfs4_dbe_lock(lsp->dbe);
3802 					rfs4_dbe_invalidate(lsp->dbe);
3803 					rfs4_dbe_unlock(lsp->dbe);
3804 				}
3805 			}
3806 			rfs4_dbe_unlock(sp->dbe);
3807 		}
3808 		rfs4_dbe_unlock(oop->dbe);
3809 	}
3810 	rfs4_dbe_unlock(lo->client->dbe);
3811 
3812 	rfs4_lockowner_rele(lo);
3813 
3814 	*cs->statusp = resp->status = NFS4_OK;
3815 
3816 out:
3817 	DTRACE_NFSV4_2(op__release__lockowner__done, struct compound_state *,
3818 	    cs, RELEASE_LOCKOWNER4res *, resp);
3819 }
3820 
3821 /*
3822  * short utility function to lookup a file and recall the delegation
3823  */
3824 static rfs4_file_t *
3825 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp,
3826 	int *lkup_error, cred_t *cr)
3827 {
3828 	vnode_t *vp;
3829 	rfs4_file_t *fp = NULL;
3830 	bool_t fcreate = FALSE;
3831 	int error;
3832 
3833 	if (vpp)
3834 		*vpp = NULL;
3835 
3836 	if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, NULL, NULL,
3837 	    NULL)) == 0) {
3838 		if (vp->v_type == VREG)
3839 			fp = rfs4_findfile(vp, NULL, &fcreate);
3840 		if (vpp)
3841 			*vpp = vp;
3842 		else
3843 			VN_RELE(vp);
3844 	}
3845 
3846 	if (lkup_error)
3847 		*lkup_error = error;
3848 
3849 	return (fp);
3850 }
3851 
3852 /*
3853  * remove: args: CURRENT_FH: directory; name.
3854  *	res: status. If success - CURRENT_FH unchanged, return change_info
3855  *		for directory.
3856  */
3857 /* ARGSUSED */
3858 static void
3859 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3860 	struct compound_state *cs)
3861 {
3862 	REMOVE4args *args = &argop->nfs_argop4_u.opremove;
3863 	REMOVE4res *resp = &resop->nfs_resop4_u.opremove;
3864 	int error;
3865 	vnode_t *dvp, *vp;
3866 	struct vattr bdva, idva, adva;
3867 	char *nm;
3868 	uint_t len;
3869 	rfs4_file_t *fp;
3870 	int in_crit = 0;
3871 	bslabel_t *clabel;
3872 
3873 	DTRACE_NFSV4_2(op__remove__start, struct compound_state *, cs,
3874 	    REMOVE4args *, args);
3875 
3876 	/* CURRENT_FH: directory */
3877 	dvp = cs->vp;
3878 	if (dvp == NULL) {
3879 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3880 		goto out;
3881 	}
3882 
3883 	if (cs->access == CS_ACCESS_DENIED) {
3884 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3885 		goto out;
3886 	}
3887 
3888 	/*
3889 	 * If there is an unshared filesystem mounted on this vnode,
3890 	 * Do not allow to remove anything in this directory.
3891 	 */
3892 	if (vn_ismntpt(dvp)) {
3893 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3894 		goto out;
3895 	}
3896 
3897 	if (dvp->v_type != VDIR) {
3898 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
3899 		goto out;
3900 	}
3901 
3902 	if (!utf8_dir_verify(&args->target)) {
3903 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3904 		goto out;
3905 	}
3906 
3907 	/*
3908 	 * Lookup the file so that we can check if it's a directory
3909 	 */
3910 	nm = utf8_to_fn(&args->target, &len, NULL);
3911 	if (nm == NULL) {
3912 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3913 		goto out;
3914 	}
3915 
3916 	if (len > MAXNAMELEN) {
3917 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
3918 		kmem_free(nm, len);
3919 		goto out;
3920 	}
3921 
3922 	if (rdonly4(cs->exi, cs->vp, req)) {
3923 		*cs->statusp = resp->status = NFS4ERR_ROFS;
3924 		kmem_free(nm, len);
3925 		goto out;
3926 	}
3927 
3928 	/*
3929 	 * Lookup the file to determine type and while we are see if
3930 	 * there is a file struct around and check for delegation.
3931 	 * We don't need to acquire va_seq before this lookup, if
3932 	 * it causes an update, cinfo.before will not match, which will
3933 	 * trigger a cache flush even if atomic is TRUE.
3934 	 */
3935 	if (fp = rfs4_lookup_and_findfile(dvp, nm, &vp, &error, cs->cr)) {
3936 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
3937 		    NULL)) {
3938 			VN_RELE(vp);
3939 			rfs4_file_rele(fp);
3940 			*cs->statusp = resp->status = NFS4ERR_DELAY;
3941 			kmem_free(nm, len);
3942 			goto out;
3943 		}
3944 	}
3945 
3946 	/* Didn't find anything to remove */
3947 	if (vp == NULL) {
3948 		*cs->statusp = resp->status = error;
3949 		kmem_free(nm, len);
3950 		goto out;
3951 	}
3952 
3953 	if (nbl_need_check(vp)) {
3954 		nbl_start_crit(vp, RW_READER);
3955 		in_crit = 1;
3956 		if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0, NULL)) {
3957 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
3958 			kmem_free(nm, len);
3959 			nbl_end_crit(vp);
3960 			VN_RELE(vp);
3961 			if (fp) {
3962 				rfs4_clear_dont_grant(fp);
3963 				rfs4_file_rele(fp);
3964 			}
3965 			goto out;
3966 		}
3967 	}
3968 
3969 	/* check label before allowing removal */
3970 	if (is_system_labeled()) {
3971 		ASSERT(req->rq_label != NULL);
3972 		clabel = req->rq_label;
3973 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
3974 		    "got client label from request(1)",
3975 		    struct svc_req *, req);
3976 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
3977 			if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK)) {
3978 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
3979 				kmem_free(nm, len);
3980 				if (in_crit)
3981 					nbl_end_crit(vp);
3982 				VN_RELE(vp);
3983 				if (fp) {
3984 					rfs4_clear_dont_grant(fp);
3985 					rfs4_file_rele(fp);
3986 				}
3987 				goto out;
3988 			}
3989 		}
3990 	}
3991 
3992 	/* Get dir "before" change value */
3993 	bdva.va_mask = AT_CTIME|AT_SEQ;
3994 	error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL);
3995 	if (error) {
3996 		*cs->statusp = resp->status = puterrno4(error);
3997 		kmem_free(nm, len);
3998 		goto out;
3999 	}
4000 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime)
4001 
4002 	/* Actually do the REMOVE operation */
4003 	if (vp->v_type == VDIR) {
4004 		/*
4005 		 * Can't remove a directory that has a mounted-on filesystem.
4006 		 */
4007 		if (vn_ismntpt(vp)) {
4008 			error = EACCES;
4009 		} else {
4010 			/*
4011 			 * System V defines rmdir to return EEXIST,
4012 			 * not * ENOTEMPTY, if the directory is not
4013 			 * empty.  A System V NFS server needs to map
4014 			 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to
4015 			 * transmit over the wire.
4016 			 */
4017 			if ((error = VOP_RMDIR(dvp, nm, rootdir, cs->cr,
4018 			    NULL, 0)) == EEXIST)
4019 				error = ENOTEMPTY;
4020 		}
4021 	} else {
4022 		if ((error = VOP_REMOVE(dvp, nm, cs->cr, NULL, 0)) == 0 &&
4023 		    fp != NULL) {
4024 			struct vattr va;
4025 			vnode_t *tvp;
4026 
4027 			rfs4_dbe_lock(fp->dbe);
4028 			tvp = fp->vp;
4029 			if (tvp)
4030 				VN_HOLD(tvp);
4031 			rfs4_dbe_unlock(fp->dbe);
4032 
4033 			if (tvp) {
4034 				/*
4035 				 * This is va_seq safe because we are not
4036 				 * manipulating dvp.
4037 				 */
4038 				va.va_mask = AT_NLINK;
4039 				if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) &&
4040 				    va.va_nlink == 0) {
4041 					/* Remove state on file remove */
4042 					if (in_crit) {
4043 						nbl_end_crit(vp);
4044 						in_crit = 0;
4045 					}
4046 					rfs4_close_all_state(fp);
4047 				}
4048 				VN_RELE(tvp);
4049 			}
4050 		}
4051 	}
4052 
4053 	if (in_crit)
4054 		nbl_end_crit(vp);
4055 	VN_RELE(vp);
4056 
4057 	if (fp) {
4058 		rfs4_clear_dont_grant(fp);
4059 		rfs4_file_rele(fp);
4060 	}
4061 	kmem_free(nm, len);
4062 
4063 	if (error) {
4064 		*cs->statusp = resp->status = puterrno4(error);
4065 		goto out;
4066 	}
4067 
4068 	/*
4069 	 * Get the initial "after" sequence number, if it fails, set to zero
4070 	 */
4071 	idva.va_mask = AT_SEQ;
4072 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL))
4073 		idva.va_seq = 0;
4074 
4075 	/*
4076 	 * Force modified data and metadata out to stable storage.
4077 	 */
4078 	(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
4079 
4080 	/*
4081 	 * Get "after" change value, if it fails, simply return the
4082 	 * before value.
4083 	 */
4084 	adva.va_mask = AT_CTIME|AT_SEQ;
4085 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) {
4086 		adva.va_ctime = bdva.va_ctime;
4087 		adva.va_seq = 0;
4088 	}
4089 
4090 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
4091 
4092 	/*
4093 	 * The cinfo.atomic = TRUE only if we have
4094 	 * non-zero va_seq's, and it has incremented by exactly one
4095 	 * during the VOP_REMOVE/RMDIR and it didn't change during
4096 	 * the VOP_FSYNC.
4097 	 */
4098 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
4099 	    idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq)
4100 		resp->cinfo.atomic = TRUE;
4101 	else
4102 		resp->cinfo.atomic = FALSE;
4103 
4104 	*cs->statusp = resp->status = NFS4_OK;
4105 
4106 out:
4107 	DTRACE_NFSV4_2(op__remove__done, struct compound_state *, cs,
4108 	    REMOVE4res *, resp);
4109 }
4110 
4111 /*
4112  * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory,
4113  *		oldname and newname.
4114  *	res: status. If success - CURRENT_FH unchanged, return change_info
4115  *		for both from and target directories.
4116  */
4117 /* ARGSUSED */
4118 static void
4119 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4120 	struct compound_state *cs)
4121 {
4122 	RENAME4args *args = &argop->nfs_argop4_u.oprename;
4123 	RENAME4res *resp = &resop->nfs_resop4_u.oprename;
4124 	int error;
4125 	vnode_t *odvp;
4126 	vnode_t *ndvp;
4127 	vnode_t *srcvp, *targvp;
4128 	struct vattr obdva, oidva, oadva;
4129 	struct vattr nbdva, nidva, nadva;
4130 	char *onm, *nnm;
4131 	uint_t olen, nlen;
4132 	rfs4_file_t *fp, *sfp;
4133 	int in_crit_src, in_crit_targ;
4134 	int fp_rele_grant_hold, sfp_rele_grant_hold;
4135 	bslabel_t *clabel;
4136 
4137 	DTRACE_NFSV4_2(op__rename__start, struct compound_state *, cs,
4138 	    RENAME4args *, args);
4139 
4140 	fp = sfp = NULL;
4141 	srcvp = targvp = NULL;
4142 	in_crit_src = in_crit_targ = 0;
4143 	fp_rele_grant_hold = sfp_rele_grant_hold = 0;
4144 
4145 	/* CURRENT_FH: target directory */
4146 	ndvp = cs->vp;
4147 	if (ndvp == NULL) {
4148 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4149 		goto out;
4150 	}
4151 
4152 	/* SAVED_FH: from directory */
4153 	odvp = cs->saved_vp;
4154 	if (odvp == NULL) {
4155 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4156 		goto out;
4157 	}
4158 
4159 	if (cs->access == CS_ACCESS_DENIED) {
4160 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4161 		goto out;
4162 	}
4163 
4164 	/*
4165 	 * If there is an unshared filesystem mounted on this vnode,
4166 	 * do not allow to rename objects in this directory.
4167 	 */
4168 	if (vn_ismntpt(odvp)) {
4169 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4170 		goto out;
4171 	}
4172 
4173 	/*
4174 	 * If there is an unshared filesystem mounted on this vnode,
4175 	 * do not allow to rename to this directory.
4176 	 */
4177 	if (vn_ismntpt(ndvp)) {
4178 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4179 		goto out;
4180 	}
4181 
4182 	if (odvp->v_type != VDIR || ndvp->v_type != VDIR) {
4183 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
4184 		goto out;
4185 	}
4186 
4187 	if (cs->saved_exi != cs->exi) {
4188 		*cs->statusp = resp->status = NFS4ERR_XDEV;
4189 		goto out;
4190 	}
4191 
4192 	if (!utf8_dir_verify(&args->oldname)) {
4193 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4194 		goto out;
4195 	}
4196 
4197 	if (!utf8_dir_verify(&args->newname)) {
4198 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4199 		goto out;
4200 	}
4201 
4202 	onm = utf8_to_fn(&args->oldname, &olen, NULL);
4203 	if (onm == NULL) {
4204 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4205 		goto out;
4206 	}
4207 
4208 	nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4209 	if (nnm == NULL) {
4210 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4211 		kmem_free(onm, olen);
4212 		goto out;
4213 	}
4214 
4215 	if (olen > MAXNAMELEN || nlen > MAXNAMELEN) {
4216 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
4217 		kmem_free(onm, olen);
4218 		kmem_free(nnm, nlen);
4219 		goto out;
4220 	}
4221 
4222 
4223 	if (rdonly4(cs->exi, cs->vp, req)) {
4224 		*cs->statusp = resp->status = NFS4ERR_ROFS;
4225 		kmem_free(onm, olen);
4226 		kmem_free(nnm, nlen);
4227 		goto out;
4228 	}
4229 
4230 	/* check label of the target dir */
4231 	if (is_system_labeled()) {
4232 		ASSERT(req->rq_label != NULL);
4233 		clabel = req->rq_label;
4234 		DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *,
4235 		    "got client label from request(1)",
4236 		    struct svc_req *, req);
4237 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
4238 			if (!do_rfs_label_check(clabel, ndvp,
4239 			    EQUALITY_CHECK)) {
4240 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
4241 				goto out;
4242 			}
4243 		}
4244 	}
4245 
4246 	/*
4247 	 * Is the source a file and have a delegation?
4248 	 * We don't need to acquire va_seq before these lookups, if
4249 	 * it causes an update, cinfo.before will not match, which will
4250 	 * trigger a cache flush even if atomic is TRUE.
4251 	 */
4252 	if (sfp = rfs4_lookup_and_findfile(odvp, onm, &srcvp, &error, cs->cr)) {
4253 		if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE,
4254 		    NULL)) {
4255 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4256 			goto err_out;
4257 		}
4258 	}
4259 
4260 	if (srcvp == NULL) {
4261 		*cs->statusp = resp->status = puterrno4(error);
4262 		kmem_free(onm, olen);
4263 		kmem_free(nnm, nlen);
4264 		goto out;
4265 	}
4266 
4267 	sfp_rele_grant_hold = 1;
4268 
4269 	/* Does the destination exist and a file and have a delegation? */
4270 	if (fp = rfs4_lookup_and_findfile(ndvp, nnm, &targvp, NULL, cs->cr)) {
4271 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
4272 		    NULL)) {
4273 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4274 			goto err_out;
4275 		}
4276 	}
4277 	fp_rele_grant_hold = 1;
4278 
4279 
4280 	/* Check for NBMAND lock on both source and target */
4281 	if (nbl_need_check(srcvp)) {
4282 		nbl_start_crit(srcvp, RW_READER);
4283 		in_crit_src = 1;
4284 		if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) {
4285 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4286 			goto err_out;
4287 		}
4288 	}
4289 
4290 	if (targvp && nbl_need_check(targvp)) {
4291 		nbl_start_crit(targvp, RW_READER);
4292 		in_crit_targ = 1;
4293 		if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) {
4294 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4295 			goto err_out;
4296 		}
4297 	}
4298 
4299 	/* Get source "before" change value */
4300 	obdva.va_mask = AT_CTIME|AT_SEQ;
4301 	error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL);
4302 	if (!error) {
4303 		nbdva.va_mask = AT_CTIME|AT_SEQ;
4304 		error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL);
4305 	}
4306 	if (error) {
4307 		*cs->statusp = resp->status = puterrno4(error);
4308 		goto err_out;
4309 	}
4310 
4311 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime)
4312 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime)
4313 
4314 	if ((error = VOP_RENAME(odvp, onm, ndvp, nnm, cs->cr, NULL, 0)) == 0 &&
4315 	    fp != NULL) {
4316 		struct vattr va;
4317 		vnode_t *tvp;
4318 
4319 		rfs4_dbe_lock(fp->dbe);
4320 		tvp = fp->vp;
4321 		if (tvp)
4322 			VN_HOLD(tvp);
4323 		rfs4_dbe_unlock(fp->dbe);
4324 
4325 		if (tvp) {
4326 			va.va_mask = AT_NLINK;
4327 			if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) &&
4328 			    va.va_nlink == 0) {
4329 				/* The file is gone and so should the state */
4330 				if (in_crit_targ) {
4331 					nbl_end_crit(targvp);
4332 					in_crit_targ = 0;
4333 				}
4334 				rfs4_close_all_state(fp);
4335 			}
4336 			VN_RELE(tvp);
4337 		}
4338 	}
4339 	if (error == 0)
4340 		vn_renamepath(ndvp, srcvp, nnm, nlen - 1);
4341 
4342 	if (in_crit_src)
4343 		nbl_end_crit(srcvp);
4344 	if (srcvp)
4345 		VN_RELE(srcvp);
4346 	if (in_crit_targ)
4347 		nbl_end_crit(targvp);
4348 	if (targvp)
4349 		VN_RELE(targvp);
4350 
4351 	if (sfp) {
4352 		rfs4_clear_dont_grant(sfp);
4353 		rfs4_file_rele(sfp);
4354 	}
4355 	if (fp) {
4356 		rfs4_clear_dont_grant(fp);
4357 		rfs4_file_rele(fp);
4358 	}
4359 
4360 	kmem_free(onm, olen);
4361 	kmem_free(nnm, nlen);
4362 
4363 	/*
4364 	 * Get the initial "after" sequence number, if it fails, set to zero
4365 	 */
4366 	oidva.va_mask = AT_SEQ;
4367 	if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL))
4368 		oidva.va_seq = 0;
4369 
4370 	nidva.va_mask = AT_SEQ;
4371 	if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL))
4372 		nidva.va_seq = 0;
4373 
4374 	/*
4375 	 * Force modified data and metadata out to stable storage.
4376 	 */
4377 	(void) VOP_FSYNC(odvp, 0, cs->cr, NULL);
4378 	(void) VOP_FSYNC(ndvp, 0, cs->cr, NULL);
4379 
4380 	if (error) {
4381 		*cs->statusp = resp->status = puterrno4(error);
4382 		goto out;
4383 	}
4384 
4385 	/*
4386 	 * Get "after" change values, if it fails, simply return the
4387 	 * before value.
4388 	 */
4389 	oadva.va_mask = AT_CTIME|AT_SEQ;
4390 	if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) {
4391 		oadva.va_ctime = obdva.va_ctime;
4392 		oadva.va_seq = 0;
4393 	}
4394 
4395 	nadva.va_mask = AT_CTIME|AT_SEQ;
4396 	if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) {
4397 		nadva.va_ctime = nbdva.va_ctime;
4398 		nadva.va_seq = 0;
4399 	}
4400 
4401 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime)
4402 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime)
4403 
4404 	/*
4405 	 * The cinfo.atomic = TRUE only if we have
4406 	 * non-zero va_seq's, and it has incremented by exactly one
4407 	 * during the VOP_RENAME and it didn't change during the VOP_FSYNC.
4408 	 */
4409 	if (obdva.va_seq && oidva.va_seq && oadva.va_seq &&
4410 	    oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq)
4411 		resp->source_cinfo.atomic = TRUE;
4412 	else
4413 		resp->source_cinfo.atomic = FALSE;
4414 
4415 	if (nbdva.va_seq && nidva.va_seq && nadva.va_seq &&
4416 	    nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq)
4417 		resp->target_cinfo.atomic = TRUE;
4418 	else
4419 		resp->target_cinfo.atomic = FALSE;
4420 
4421 #ifdef	VOLATILE_FH_TEST
4422 	{
4423 	extern void add_volrnm_fh(struct exportinfo *, vnode_t *);
4424 
4425 	/*
4426 	 * Add the renamed file handle to the volatile rename list
4427 	 */
4428 	if (cs->exi->exi_export.ex_flags & EX_VOLRNM) {
4429 		/* file handles may expire on rename */
4430 		vnode_t *vp;
4431 
4432 		nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4433 		/*
4434 		 * Already know that nnm will be a valid string
4435 		 */
4436 		error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr,
4437 		    NULL, NULL, NULL);
4438 		kmem_free(nnm, nlen);
4439 		if (!error) {
4440 			add_volrnm_fh(cs->exi, vp);
4441 			VN_RELE(vp);
4442 		}
4443 	}
4444 	}
4445 #endif	/* VOLATILE_FH_TEST */
4446 
4447 	*cs->statusp = resp->status = NFS4_OK;
4448 out:
4449 	DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs,
4450 	    RENAME4res *, resp);
4451 	return;
4452 
4453 err_out:
4454 	kmem_free(onm, olen);
4455 	kmem_free(nnm, nlen);
4456 
4457 	if (in_crit_src) nbl_end_crit(srcvp);
4458 	if (in_crit_targ) nbl_end_crit(targvp);
4459 	if (targvp) VN_RELE(targvp);
4460 	if (srcvp) VN_RELE(srcvp);
4461 	if (sfp) {
4462 		if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp);
4463 		rfs4_file_rele(sfp);
4464 	}
4465 	if (fp) {
4466 		if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp);
4467 		rfs4_file_rele(fp);
4468 	}
4469 
4470 	DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs,
4471 	    RENAME4res *, resp);
4472 }
4473 
4474 /* ARGSUSED */
4475 static void
4476 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4477 	struct compound_state *cs)
4478 {
4479 	RENEW4args *args = &argop->nfs_argop4_u.oprenew;
4480 	RENEW4res *resp = &resop->nfs_resop4_u.oprenew;
4481 	rfs4_client_t *cp;
4482 
4483 	DTRACE_NFSV4_2(op__renew__start, struct compound_state *, cs,
4484 	    RENEW4args *, args);
4485 
4486 	if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) {
4487 		*cs->statusp = resp->status =
4488 		    rfs4_check_clientid(&args->clientid, 0);
4489 		goto out;
4490 	}
4491 
4492 	if (rfs4_lease_expired(cp)) {
4493 		rfs4_client_rele(cp);
4494 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
4495 		goto out;
4496 	}
4497 
4498 	rfs4_update_lease(cp);
4499 
4500 	mutex_enter(cp->cbinfo.cb_lock);
4501 	if (cp->cbinfo.cb_notified_of_cb_path_down == FALSE) {
4502 		cp->cbinfo.cb_notified_of_cb_path_down = TRUE;
4503 		*cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN;
4504 	} else {
4505 		*cs->statusp = resp->status = NFS4_OK;
4506 	}
4507 	mutex_exit(cp->cbinfo.cb_lock);
4508 
4509 	rfs4_client_rele(cp);
4510 
4511 out:
4512 	DTRACE_NFSV4_2(op__renew__done, struct compound_state *, cs,
4513 	    RENEW4res *, resp);
4514 }
4515 
4516 /* ARGSUSED */
4517 static void
4518 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
4519 	struct compound_state *cs)
4520 {
4521 	RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh;
4522 
4523 	DTRACE_NFSV4_1(op__restorefh__start, struct compound_state *, cs);
4524 
4525 	/* No need to check cs->access - we are not accessing any object */
4526 	if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) {
4527 		*cs->statusp = resp->status = NFS4ERR_RESTOREFH;
4528 		goto out;
4529 	}
4530 	if (cs->vp != NULL) {
4531 		VN_RELE(cs->vp);
4532 	}
4533 	cs->vp = cs->saved_vp;
4534 	cs->saved_vp = NULL;
4535 	cs->exi = cs->saved_exi;
4536 	nfs_fh4_copy(&cs->saved_fh, &cs->fh);
4537 	*cs->statusp = resp->status = NFS4_OK;
4538 	cs->deleg = FALSE;
4539 
4540 out:
4541 	DTRACE_NFSV4_2(op__restorefh__done, struct compound_state *, cs,
4542 	    RESTOREFH4res *, resp);
4543 }
4544 
4545 /* ARGSUSED */
4546 static void
4547 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4548 	struct compound_state *cs)
4549 {
4550 	SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh;
4551 
4552 	DTRACE_NFSV4_1(op__savefh__start, struct compound_state *, cs);
4553 
4554 	/* No need to check cs->access - we are not accessing any object */
4555 	if (cs->vp == NULL) {
4556 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4557 		goto out;
4558 	}
4559 	if (cs->saved_vp != NULL) {
4560 		VN_RELE(cs->saved_vp);
4561 	}
4562 	cs->saved_vp = cs->vp;
4563 	VN_HOLD(cs->saved_vp);
4564 	cs->saved_exi = cs->exi;
4565 	/*
4566 	 * since SAVEFH is fairly rare, don't alloc space for its fh
4567 	 * unless necessary.
4568 	 */
4569 	if (cs->saved_fh.nfs_fh4_val == NULL) {
4570 		cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP);
4571 	}
4572 	nfs_fh4_copy(&cs->fh, &cs->saved_fh);
4573 	*cs->statusp = resp->status = NFS4_OK;
4574 
4575 out:
4576 	DTRACE_NFSV4_2(op__savefh__done, struct compound_state *, cs,
4577 	    SAVEFH4res *, resp);
4578 }
4579 
4580 /*
4581  * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to
4582  * return the bitmap of attrs that were set successfully. It is also
4583  * called by Verify/Nverify to test the vattr/vfsstat attrs. It should
4584  * always be called only after rfs4_do_set_attrs().
4585  *
4586  * Verify that the attributes are same as the expected ones. sargp->vap
4587  * and sargp->sbp contain the input attributes as translated from fattr4.
4588  *
4589  * This function verifies only the attrs that correspond to a vattr or
4590  * vfsstat struct. That is because of the extra step needed to get the
4591  * corresponding system structs. Other attributes have already been set or
4592  * verified by do_rfs4_set_attrs.
4593  *
4594  * Return 0 if all attrs match, -1 if some don't, error if error processing.
4595  */
4596 static int
4597 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp,
4598 	bitmap4 *resp, struct nfs4_ntov_table *ntovp)
4599 {
4600 	int error, ret_error = 0;
4601 	int i, k;
4602 	uint_t sva_mask = sargp->vap->va_mask;
4603 	uint_t vbit;
4604 	union nfs4_attr_u *na;
4605 	uint8_t *amap;
4606 	bool_t getsb = ntovp->vfsstat;
4607 
4608 	if (sva_mask != 0) {
4609 		/*
4610 		 * Okay to overwrite sargp->vap because we verify based
4611 		 * on the incoming values.
4612 		 */
4613 		ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0,
4614 		    sargp->cs->cr, NULL);
4615 		if (ret_error) {
4616 			if (resp == NULL)
4617 				return (ret_error);
4618 			/*
4619 			 * Must return bitmap of successful attrs
4620 			 */
4621 			sva_mask = 0;	/* to prevent checking vap later */
4622 		} else {
4623 			/*
4624 			 * Some file systems clobber va_mask. it is probably
4625 			 * wrong of them to do so, nonethless we practice
4626 			 * defensive coding.
4627 			 * See bug id 4276830.
4628 			 */
4629 			sargp->vap->va_mask = sva_mask;
4630 		}
4631 	}
4632 
4633 	if (getsb) {
4634 		/*
4635 		 * Now get the superblock and loop on the bitmap, as there is
4636 		 * no simple way of translating from superblock to bitmap4.
4637 		 */
4638 		ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp);
4639 		if (ret_error) {
4640 			if (resp == NULL)
4641 				goto errout;
4642 			getsb = FALSE;
4643 		}
4644 	}
4645 
4646 	/*
4647 	 * Now loop and verify each attribute which getattr returned
4648 	 * whether it's the same as the input.
4649 	 */
4650 	if (resp == NULL && !getsb && (sva_mask == 0))
4651 		goto errout;
4652 
4653 	na = ntovp->na;
4654 	amap = ntovp->amap;
4655 	k = 0;
4656 	for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) {
4657 		k = *amap;
4658 		ASSERT(nfs4_ntov_map[k].nval == k);
4659 		vbit = nfs4_ntov_map[k].vbit;
4660 
4661 		/*
4662 		 * If vattr attribute but VOP_GETATTR failed, or it's
4663 		 * superblock attribute but VFS_STATVFS failed, skip
4664 		 */
4665 		if (vbit) {
4666 			if ((vbit & sva_mask) == 0)
4667 				continue;
4668 		} else if (!(getsb && nfs4_ntov_map[k].vfsstat)) {
4669 			continue;
4670 		}
4671 		error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na);
4672 		if (resp != NULL) {
4673 			if (error)
4674 				ret_error = -1;	/* not all match */
4675 			else	/* update response bitmap */
4676 				*resp |= nfs4_ntov_map[k].fbit;
4677 			continue;
4678 		}
4679 		if (error) {
4680 			ret_error = -1;	/* not all match */
4681 			break;
4682 		}
4683 	}
4684 errout:
4685 	return (ret_error);
4686 }
4687 
4688 /*
4689  * Decode the attribute to be set/verified. If the attr requires a sys op
4690  * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't
4691  * call the sv_getit function for it, because the sys op hasn't yet been done.
4692  * Return 0 for success, error code if failed.
4693  *
4694  * Note: the decoded arg is not freed here but in nfs4_ntov_table_free.
4695  */
4696 static int
4697 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp,
4698 	int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap)
4699 {
4700 	int error = 0;
4701 	bool_t set_later;
4702 
4703 	sargp->vap->va_mask |= nfs4_ntov_map[k].vbit;
4704 
4705 	if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) {
4706 		set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat;
4707 		/*
4708 		 * don't verify yet if a vattr or sb dependent attr,
4709 		 * because we don't have their sys values yet.
4710 		 * Will be done later.
4711 		 */
4712 		if (! (set_later && (cmd == NFS4ATTR_VERIT))) {
4713 			/*
4714 			 * ACLs are a special case, since setting the MODE
4715 			 * conflicts with setting the ACL.  We delay setting
4716 			 * the ACL until all other attributes have been set.
4717 			 * The ACL gets set in do_rfs4_op_setattr().
4718 			 */
4719 			if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) {
4720 				error = (*nfs4_ntov_map[k].sv_getit)(cmd,
4721 				    sargp, nap);
4722 				if (error) {
4723 					xdr_free(nfs4_ntov_map[k].xfunc,
4724 					    (caddr_t)nap);
4725 				}
4726 			}
4727 		}
4728 	} else {
4729 #ifdef  DEBUG
4730 		cmn_err(CE_NOTE, "decode_fattr4_attr: error "
4731 		    "decoding attribute %d\n", k);
4732 #endif
4733 		error = EINVAL;
4734 	}
4735 	if (!error && resp_bval && !set_later) {
4736 		*resp_bval |= nfs4_ntov_map[k].fbit;
4737 	}
4738 
4739 	return (error);
4740 }
4741 
4742 /*
4743  * Set vattr based on incoming fattr4 attrs - used by setattr.
4744  * Set response mask. Ignore any values that are not writable vattr attrs.
4745  */
4746 static nfsstat4
4747 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
4748 		struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp,
4749 		nfs4_attr_cmd_t cmd)
4750 {
4751 	int error = 0;
4752 	int i;
4753 	char *attrs = fattrp->attrlist4;
4754 	uint32_t attrslen = fattrp->attrlist4_len;
4755 	XDR xdr;
4756 	nfsstat4 status = NFS4_OK;
4757 	vnode_t *vp = cs->vp;
4758 	union nfs4_attr_u *na;
4759 	uint8_t *amap;
4760 
4761 #ifndef lint
4762 	/*
4763 	 * Make sure that maximum attribute number can be expressed as an
4764 	 * 8 bit quantity.
4765 	 */
4766 	ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1));
4767 #endif
4768 
4769 	if (vp == NULL) {
4770 		if (resp)
4771 			*resp = 0;
4772 		return (NFS4ERR_NOFILEHANDLE);
4773 	}
4774 	if (cs->access == CS_ACCESS_DENIED) {
4775 		if (resp)
4776 			*resp = 0;
4777 		return (NFS4ERR_ACCESS);
4778 	}
4779 
4780 	sargp->op = cmd;
4781 	sargp->cs = cs;
4782 	sargp->flag = 0;	/* may be set later */
4783 	sargp->vap->va_mask = 0;
4784 	sargp->rdattr_error = NFS4_OK;
4785 	sargp->rdattr_error_req = FALSE;
4786 	/* sargp->sbp is set by the caller */
4787 
4788 	xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE);
4789 
4790 	na = ntovp->na;
4791 	amap = ntovp->amap;
4792 
4793 	/*
4794 	 * The following loop iterates on the nfs4_ntov_map checking
4795 	 * if the fbit is set in the requested bitmap.
4796 	 * If set then we process the arguments using the
4797 	 * rfs4_fattr4 conversion functions to populate the setattr
4798 	 * vattr and va_mask. Any settable attrs that are not using vattr
4799 	 * will be set in this loop.
4800 	 */
4801 	for (i = 0; i < nfs4_ntov_map_size; i++) {
4802 		if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) {
4803 			continue;
4804 		}
4805 		/*
4806 		 * If setattr, must be a writable attr.
4807 		 * If verify/nverify, must be a readable attr.
4808 		 */
4809 		if ((error = (*nfs4_ntov_map[i].sv_getit)(
4810 		    NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) {
4811 			/*
4812 			 * Client tries to set/verify an
4813 			 * unsupported attribute, tries to set
4814 			 * a read only attr or verify a write
4815 			 * only one - error!
4816 			 */
4817 			break;
4818 		}
4819 		/*
4820 		 * Decode the attribute to set/verify
4821 		 */
4822 		error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval,
4823 		    &xdr, resp ? resp : NULL, na);
4824 		if (error)
4825 			break;
4826 		*amap++ = (uint8_t)nfs4_ntov_map[i].nval;
4827 		na++;
4828 		(ntovp->attrcnt)++;
4829 		if (nfs4_ntov_map[i].vfsstat)
4830 			ntovp->vfsstat = TRUE;
4831 	}
4832 
4833 	if (error != 0)
4834 		status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP :
4835 		    puterrno4(error));
4836 	/* xdrmem_destroy(&xdrs); */	/* NO-OP */
4837 	return (status);
4838 }
4839 
4840 static nfsstat4
4841 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
4842 		stateid4 *stateid)
4843 {
4844 	int error = 0;
4845 	struct nfs4_svgetit_arg sarg;
4846 	bool_t trunc;
4847 
4848 	nfsstat4 status = NFS4_OK;
4849 	cred_t *cr = cs->cr;
4850 	vnode_t *vp = cs->vp;
4851 	struct nfs4_ntov_table ntov;
4852 	struct statvfs64 sb;
4853 	struct vattr bva;
4854 	struct flock64 bf;
4855 	int in_crit = 0;
4856 	uint_t saved_mask = 0;
4857 	caller_context_t ct;
4858 
4859 	*resp = 0;
4860 	sarg.sbp = &sb;
4861 	nfs4_ntov_table_init(&ntov);
4862 	status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov,
4863 	    NFS4ATTR_SETIT);
4864 	if (status != NFS4_OK) {
4865 		/*
4866 		 * failed set attrs
4867 		 */
4868 		goto done;
4869 	}
4870 	if ((sarg.vap->va_mask == 0) &&
4871 	    (! (fattrp->attrmask & FATTR4_ACL_MASK))) {
4872 		/*
4873 		 * no further work to be done
4874 		 */
4875 		goto done;
4876 	}
4877 
4878 	/*
4879 	 * If we got a request to set the ACL and the MODE, only
4880 	 * allow changing VSUID, VSGID, and VSVTX.  Attempting
4881 	 * to change any other bits, along with setting an ACL,
4882 	 * gives NFS4ERR_INVAL.
4883 	 */
4884 	if ((fattrp->attrmask & FATTR4_ACL_MASK) &&
4885 	    (fattrp->attrmask & FATTR4_MODE_MASK)) {
4886 		vattr_t va;
4887 
4888 		va.va_mask = AT_MODE;
4889 		error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL);
4890 		if (error) {
4891 			status = puterrno4(error);
4892 			goto done;
4893 		}
4894 		if ((sarg.vap->va_mode ^ va.va_mode) &
4895 		    ~(VSUID | VSGID | VSVTX)) {
4896 			status = NFS4ERR_INVAL;
4897 			goto done;
4898 		}
4899 	}
4900 
4901 	/* Check stateid only if size has been set */
4902 	if (sarg.vap->va_mask & AT_SIZE) {
4903 		trunc = (sarg.vap->va_size == 0);
4904 		status = rfs4_check_stateid(FWRITE, cs->vp, stateid,
4905 		    trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct);
4906 		if (status != NFS4_OK)
4907 			goto done;
4908 	} else {
4909 		ct.cc_sysid = 0;
4910 		ct.cc_pid = 0;
4911 		ct.cc_caller_id = nfs4_srv_caller_id;
4912 		ct.cc_flags = CC_DONTBLOCK;
4913 	}
4914 
4915 	/* XXX start of possible race with delegations */
4916 
4917 	/*
4918 	 * We need to specially handle size changes because it is
4919 	 * possible for the client to create a file with read-only
4920 	 * modes, but with the file opened for writing. If the client
4921 	 * then tries to set the file size, e.g. ftruncate(3C),
4922 	 * fcntl(F_FREESP), the normal access checking done in
4923 	 * VOP_SETATTR would prevent the client from doing it even though
4924 	 * it should be allowed to do so.  To get around this, we do the
4925 	 * access checking for ourselves and use VOP_SPACE which doesn't
4926 	 * do the access checking.
4927 	 * Also the client should not be allowed to change the file
4928 	 * size if there is a conflicting non-blocking mandatory lock in
4929 	 * the region of the change.
4930 	 */
4931 	if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) {
4932 		u_offset_t offset;
4933 		ssize_t length;
4934 
4935 		/*
4936 		 * ufs_setattr clears AT_SIZE from vap->va_mask, but
4937 		 * before returning, sarg.vap->va_mask is used to
4938 		 * generate the setattr reply bitmap.  We also clear
4939 		 * AT_SIZE below before calling VOP_SPACE.  For both
4940 		 * of these cases, the va_mask needs to be saved here
4941 		 * and restored after calling VOP_SETATTR.
4942 		 */
4943 		saved_mask = sarg.vap->va_mask;
4944 
4945 		/*
4946 		 * Check any possible conflict due to NBMAND locks.
4947 		 * Get into critical region before VOP_GETATTR, so the
4948 		 * size attribute is valid when checking conflicts.
4949 		 */
4950 		if (nbl_need_check(vp)) {
4951 			nbl_start_crit(vp, RW_READER);
4952 			in_crit = 1;
4953 		}
4954 
4955 		bva.va_mask = AT_UID|AT_SIZE;
4956 		if (error = VOP_GETATTR(vp, &bva, 0, cr, &ct)) {
4957 			status = puterrno4(error);
4958 			goto done;
4959 		}
4960 
4961 		if (in_crit) {
4962 			if (sarg.vap->va_size < bva.va_size) {
4963 				offset = sarg.vap->va_size;
4964 				length = bva.va_size - sarg.vap->va_size;
4965 			} else {
4966 				offset = bva.va_size;
4967 				length = sarg.vap->va_size - bva.va_size;
4968 			}
4969 			if (nbl_conflict(vp, NBL_WRITE, offset, length, 0,
4970 			    &ct)) {
4971 				status = NFS4ERR_LOCKED;
4972 				goto done;
4973 			}
4974 		}
4975 
4976 		if (crgetuid(cr) == bva.va_uid) {
4977 			sarg.vap->va_mask &= ~AT_SIZE;
4978 			bf.l_type = F_WRLCK;
4979 			bf.l_whence = 0;
4980 			bf.l_start = (off64_t)sarg.vap->va_size;
4981 			bf.l_len = 0;
4982 			bf.l_sysid = 0;
4983 			bf.l_pid = 0;
4984 			error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE,
4985 			    (offset_t)sarg.vap->va_size, cr, &ct);
4986 		}
4987 	}
4988 
4989 	if (!error && sarg.vap->va_mask != 0)
4990 		error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct);
4991 
4992 	/* restore va_mask -- ufs_setattr clears AT_SIZE */
4993 	if (saved_mask & AT_SIZE)
4994 		sarg.vap->va_mask |= AT_SIZE;
4995 
4996 	/*
4997 	 * If an ACL was being set, it has been delayed until now,
4998 	 * in order to set the mode (via the VOP_SETATTR() above) first.
4999 	 */
5000 	if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) {
5001 		int i;
5002 
5003 		for (i = 0; i < NFS4_MAXNUM_ATTRS; i++)
5004 			if (ntov.amap[i] == FATTR4_ACL)
5005 				break;
5006 		if (i < NFS4_MAXNUM_ATTRS) {
5007 			error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)(
5008 			    NFS4ATTR_SETIT, &sarg, &ntov.na[i]);
5009 			if (error == 0) {
5010 				*resp |= FATTR4_ACL_MASK;
5011 			} else if (error == ENOTSUP) {
5012 				(void) rfs4_verify_attr(&sarg, resp, &ntov);
5013 				status = NFS4ERR_ATTRNOTSUPP;
5014 				goto done;
5015 			}
5016 		} else {
5017 			NFS4_DEBUG(rfs4_debug,
5018 			    (CE_NOTE, "do_rfs4_op_setattr: "
5019 			    "unable to find ACL in fattr4"));
5020 			error = EINVAL;
5021 		}
5022 	}
5023 
5024 	if (error) {
5025 		/* check if a monitor detected a delegation conflict */
5026 		if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK))
5027 			status = NFS4ERR_DELAY;
5028 		else
5029 			status = puterrno4(error);
5030 
5031 		/*
5032 		 * Set the response bitmap when setattr failed.
5033 		 * If VOP_SETATTR partially succeeded, test by doing a
5034 		 * VOP_GETATTR on the object and comparing the data
5035 		 * to the setattr arguments.
5036 		 */
5037 		(void) rfs4_verify_attr(&sarg, resp, &ntov);
5038 	} else {
5039 		/*
5040 		 * Force modified metadata out to stable storage.
5041 		 */
5042 		(void) VOP_FSYNC(vp, FNODSYNC, cr, &ct);
5043 		/*
5044 		 * Set response bitmap
5045 		 */
5046 		nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp);
5047 	}
5048 
5049 /* Return early and already have a NFSv4 error */
5050 done:
5051 	/*
5052 	 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr
5053 	 * conversion sets both readable and writeable NFS4 attrs
5054 	 * for AT_MTIME and AT_ATIME.  The line below masks out
5055 	 * unrequested attrs from the setattr result bitmap.  This
5056 	 * is placed after the done: label to catch the ATTRNOTSUP
5057 	 * case.
5058 	 */
5059 	*resp &= fattrp->attrmask;
5060 
5061 	if (in_crit)
5062 		nbl_end_crit(vp);
5063 
5064 	nfs4_ntov_table_free(&ntov, &sarg);
5065 
5066 	return (status);
5067 }
5068 
5069 /* ARGSUSED */
5070 static void
5071 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5072 	struct compound_state *cs)
5073 {
5074 	SETATTR4args *args = &argop->nfs_argop4_u.opsetattr;
5075 	SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr;
5076 	bslabel_t *clabel;
5077 
5078 	DTRACE_NFSV4_2(op__setattr__start, struct compound_state *, cs,
5079 	    SETATTR4args *, args);
5080 
5081 	if (cs->vp == NULL) {
5082 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5083 		goto out;
5084 	}
5085 
5086 	/*
5087 	 * If there is an unshared filesystem mounted on this vnode,
5088 	 * do not allow to setattr on this vnode.
5089 	 */
5090 	if (vn_ismntpt(cs->vp)) {
5091 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5092 		goto out;
5093 	}
5094 
5095 	resp->attrsset = 0;
5096 
5097 	if (rdonly4(cs->exi, cs->vp, req)) {
5098 		*cs->statusp = resp->status = NFS4ERR_ROFS;
5099 		goto out;
5100 	}
5101 
5102 	/* check label before setting attributes */
5103 	if (is_system_labeled()) {
5104 		ASSERT(req->rq_label != NULL);
5105 		clabel = req->rq_label;
5106 		DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *,
5107 		    "got client label from request(1)",
5108 		    struct svc_req *, req);
5109 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
5110 			if (!do_rfs_label_check(clabel, cs->vp,
5111 			    EQUALITY_CHECK)) {
5112 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
5113 				goto out;
5114 			}
5115 		}
5116 	}
5117 
5118 	*cs->statusp = resp->status =
5119 	    do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs,
5120 	    &args->stateid);
5121 
5122 out:
5123 	DTRACE_NFSV4_2(op__setattr__done, struct compound_state *, cs,
5124 	    SETATTR4res *, resp);
5125 }
5126 
5127 /* ARGSUSED */
5128 static void
5129 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5130 	struct compound_state *cs)
5131 {
5132 	/*
5133 	 * verify and nverify are exactly the same, except that nverify
5134 	 * succeeds when some argument changed, and verify succeeds when
5135 	 * when none changed.
5136 	 */
5137 
5138 	VERIFY4args  *args = &argop->nfs_argop4_u.opverify;
5139 	VERIFY4res *resp = &resop->nfs_resop4_u.opverify;
5140 
5141 	int error;
5142 	struct nfs4_svgetit_arg sarg;
5143 	struct statvfs64 sb;
5144 	struct nfs4_ntov_table ntov;
5145 
5146 	DTRACE_NFSV4_2(op__verify__start, struct compound_state *, cs,
5147 	    VERIFY4args *, args);
5148 
5149 	if (cs->vp == NULL) {
5150 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5151 		goto out;
5152 	}
5153 
5154 	sarg.sbp = &sb;
5155 	nfs4_ntov_table_init(&ntov);
5156 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
5157 	    &sarg, &ntov, NFS4ATTR_VERIT);
5158 	if (resp->status != NFS4_OK) {
5159 		/*
5160 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
5161 		 * so could return -1 for "no match".
5162 		 */
5163 		if (resp->status == -1)
5164 			resp->status = NFS4ERR_NOT_SAME;
5165 		goto done;
5166 	}
5167 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
5168 	switch (error) {
5169 	case 0:
5170 		resp->status = NFS4_OK;
5171 		break;
5172 	case -1:
5173 		resp->status = NFS4ERR_NOT_SAME;
5174 		break;
5175 	default:
5176 		resp->status = puterrno4(error);
5177 		break;
5178 	}
5179 done:
5180 	*cs->statusp = resp->status;
5181 	nfs4_ntov_table_free(&ntov, &sarg);
5182 out:
5183 	DTRACE_NFSV4_2(op__verify__done, struct compound_state *, cs,
5184 	    VERIFY4res *, resp);
5185 }
5186 
5187 /* ARGSUSED */
5188 static void
5189 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5190 	struct compound_state *cs)
5191 {
5192 	/*
5193 	 * verify and nverify are exactly the same, except that nverify
5194 	 * succeeds when some argument changed, and verify succeeds when
5195 	 * when none changed.
5196 	 */
5197 
5198 	NVERIFY4args  *args = &argop->nfs_argop4_u.opnverify;
5199 	NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify;
5200 
5201 	int error;
5202 	struct nfs4_svgetit_arg sarg;
5203 	struct statvfs64 sb;
5204 	struct nfs4_ntov_table ntov;
5205 
5206 	DTRACE_NFSV4_2(op__nverify__start, struct compound_state *, cs,
5207 	    NVERIFY4args *, args);
5208 
5209 	if (cs->vp == NULL) {
5210 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5211 		DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs,
5212 		    NVERIFY4res *, resp);
5213 		return;
5214 	}
5215 	sarg.sbp = &sb;
5216 	nfs4_ntov_table_init(&ntov);
5217 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
5218 	    &sarg, &ntov, NFS4ATTR_VERIT);
5219 	if (resp->status != NFS4_OK) {
5220 		/*
5221 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
5222 		 * so could return -1 for "no match".
5223 		 */
5224 		if (resp->status == -1)
5225 			resp->status = NFS4_OK;
5226 		goto done;
5227 	}
5228 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
5229 	switch (error) {
5230 	case 0:
5231 		resp->status = NFS4ERR_SAME;
5232 		break;
5233 	case -1:
5234 		resp->status = NFS4_OK;
5235 		break;
5236 	default:
5237 		resp->status = puterrno4(error);
5238 		break;
5239 	}
5240 done:
5241 	*cs->statusp = resp->status;
5242 	nfs4_ntov_table_free(&ntov, &sarg);
5243 
5244 	DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs,
5245 	    NVERIFY4res *, resp);
5246 }
5247 
5248 /*
5249  * XXX - This should live in an NFS header file.
5250  */
5251 #define	MAX_IOVECS	12
5252 
5253 /* ARGSUSED */
5254 static void
5255 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5256 	struct compound_state *cs)
5257 {
5258 	WRITE4args *args = &argop->nfs_argop4_u.opwrite;
5259 	WRITE4res *resp = &resop->nfs_resop4_u.opwrite;
5260 	int error;
5261 	vnode_t *vp;
5262 	struct vattr bva;
5263 	u_offset_t rlimit;
5264 	struct uio uio;
5265 	struct iovec iov[MAX_IOVECS];
5266 	struct iovec *iovp;
5267 	int iovcnt;
5268 	int ioflag;
5269 	cred_t *savecred, *cr;
5270 	bool_t *deleg = &cs->deleg;
5271 	nfsstat4 stat;
5272 	int in_crit = 0;
5273 	caller_context_t ct;
5274 
5275 	DTRACE_NFSV4_2(op__write__start, struct compound_state *, cs,
5276 	    WRITE4args *, args);
5277 
5278 	vp = cs->vp;
5279 	if (vp == NULL) {
5280 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5281 		goto out;
5282 	}
5283 	if (cs->access == CS_ACCESS_DENIED) {
5284 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5285 		goto out;
5286 	}
5287 
5288 	cr = cs->cr;
5289 
5290 	if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE,
5291 	    deleg, TRUE, &ct)) != NFS4_OK) {
5292 		*cs->statusp = resp->status = stat;
5293 		goto out;
5294 	}
5295 
5296 	/*
5297 	 * We have to enter the critical region before calling VOP_RWLOCK
5298 	 * to avoid a deadlock with ufs.
5299 	 */
5300 	if (nbl_need_check(vp)) {
5301 		nbl_start_crit(vp, RW_READER);
5302 		in_crit = 1;
5303 		if (nbl_conflict(vp, NBL_WRITE,
5304 		    args->offset, args->data_len, 0, &ct)) {
5305 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
5306 			goto out;
5307 		}
5308 	}
5309 
5310 	bva.va_mask = AT_MODE | AT_UID;
5311 	error = VOP_GETATTR(vp, &bva, 0, cr, &ct);
5312 
5313 	/*
5314 	 * If we can't get the attributes, then we can't do the
5315 	 * right access checking.  So, we'll fail the request.
5316 	 */
5317 	if (error) {
5318 		*cs->statusp = resp->status = puterrno4(error);
5319 		goto out;
5320 	}
5321 
5322 	if (rdonly4(cs->exi, cs->vp, req)) {
5323 		*cs->statusp = resp->status = NFS4ERR_ROFS;
5324 		goto out;
5325 	}
5326 
5327 	if (vp->v_type != VREG) {
5328 		*cs->statusp = resp->status =
5329 		    ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
5330 		goto out;
5331 	}
5332 
5333 	if (crgetuid(cr) != bva.va_uid &&
5334 	    (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) {
5335 		*cs->statusp = resp->status = puterrno4(error);
5336 		goto out;
5337 	}
5338 
5339 	if (MANDLOCK(vp, bva.va_mode)) {
5340 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5341 		goto out;
5342 	}
5343 
5344 	if (args->data_len == 0) {
5345 		*cs->statusp = resp->status = NFS4_OK;
5346 		resp->count = 0;
5347 		resp->committed = args->stable;
5348 		resp->writeverf = Write4verf;
5349 		goto out;
5350 	}
5351 
5352 	if (args->mblk != NULL) {
5353 		mblk_t *m;
5354 		uint_t bytes, round_len;
5355 
5356 		iovcnt = 0;
5357 		bytes = 0;
5358 		round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT);
5359 		for (m = args->mblk;
5360 		    m != NULL && bytes < round_len;
5361 		    m = m->b_cont) {
5362 			iovcnt++;
5363 			bytes += MBLKL(m);
5364 		}
5365 #ifdef DEBUG
5366 		/* should have ended on an mblk boundary */
5367 		if (bytes != round_len) {
5368 			printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n",
5369 			    bytes, round_len, args->data_len);
5370 			printf("args=%p, args->mblk=%p, m=%p", (void *)args,
5371 			    (void *)args->mblk, (void *)m);
5372 			ASSERT(bytes == round_len);
5373 		}
5374 #endif
5375 		if (iovcnt <= MAX_IOVECS) {
5376 			iovp = iov;
5377 		} else {
5378 			iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP);
5379 		}
5380 		mblk_to_iov(args->mblk, iovcnt, iovp);
5381 	} else {
5382 		iovcnt = 1;
5383 		iovp = iov;
5384 		iovp->iov_base = args->data_val;
5385 		iovp->iov_len = args->data_len;
5386 	}
5387 
5388 	uio.uio_iov = iovp;
5389 	uio.uio_iovcnt = iovcnt;
5390 
5391 	uio.uio_segflg = UIO_SYSSPACE;
5392 	uio.uio_extflg = UIO_COPY_DEFAULT;
5393 	uio.uio_loffset = args->offset;
5394 	uio.uio_resid = args->data_len;
5395 	uio.uio_llimit = curproc->p_fsz_ctl;
5396 	rlimit = uio.uio_llimit - args->offset;
5397 	if (rlimit < (u_offset_t)uio.uio_resid)
5398 		uio.uio_resid = (int)rlimit;
5399 
5400 	if (args->stable == UNSTABLE4)
5401 		ioflag = 0;
5402 	else if (args->stable == FILE_SYNC4)
5403 		ioflag = FSYNC;
5404 	else if (args->stable == DATA_SYNC4)
5405 		ioflag = FDSYNC;
5406 	else {
5407 		if (iovp != iov)
5408 			kmem_free(iovp, sizeof (*iovp) * iovcnt);
5409 		*cs->statusp = resp->status = NFS4ERR_INVAL;
5410 		goto out;
5411 	}
5412 
5413 	/*
5414 	 * We're changing creds because VM may fault and we need
5415 	 * the cred of the current thread to be used if quota
5416 	 * checking is enabled.
5417 	 */
5418 	savecred = curthread->t_cred;
5419 	curthread->t_cred = cr;
5420 	error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct);
5421 	curthread->t_cred = savecred;
5422 
5423 	if (iovp != iov)
5424 		kmem_free(iovp, sizeof (*iovp) * iovcnt);
5425 
5426 	if (error) {
5427 		*cs->statusp = resp->status = puterrno4(error);
5428 		goto out;
5429 	}
5430 
5431 	*cs->statusp = resp->status = NFS4_OK;
5432 	resp->count = args->data_len - uio.uio_resid;
5433 
5434 	if (ioflag == 0)
5435 		resp->committed = UNSTABLE4;
5436 	else
5437 		resp->committed = FILE_SYNC4;
5438 
5439 	resp->writeverf = Write4verf;
5440 
5441 out:
5442 	if (in_crit)
5443 		nbl_end_crit(vp);
5444 
5445 	DTRACE_NFSV4_2(op__write__done, struct compound_state *, cs,
5446 	    WRITE4res *, resp);
5447 }
5448 
5449 
5450 /* XXX put in a header file */
5451 extern int	sec_svc_getcred(struct svc_req *, cred_t *,  caddr_t *, int *);
5452 
5453 void
5454 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi,
5455 	struct svc_req *req, cred_t *cr, int *rv)
5456 {
5457 	uint_t i;
5458 	struct compound_state cs;
5459 
5460 	if (rv != NULL)
5461 		*rv = 0;
5462 	rfs4_init_compound_state(&cs);
5463 	/*
5464 	 * Form a reply tag by copying over the reqeuest tag.
5465 	 */
5466 	resp->tag.utf8string_val =
5467 	    kmem_alloc(args->tag.utf8string_len, KM_SLEEP);
5468 	resp->tag.utf8string_len = args->tag.utf8string_len;
5469 	bcopy(args->tag.utf8string_val, resp->tag.utf8string_val,
5470 	    resp->tag.utf8string_len);
5471 
5472 	cs.statusp = &resp->status;
5473 	cs.req = req;
5474 
5475 	/*
5476 	 * XXX for now, minorversion should be zero
5477 	 */
5478 	if (args->minorversion != NFS4_MINORVERSION) {
5479 		DTRACE_NFSV4_2(compound__start, struct compound_state *,
5480 		    &cs, COMPOUND4args *, args);
5481 		resp->array_len = 0;
5482 		resp->array = NULL;
5483 		resp->status = NFS4ERR_MINOR_VERS_MISMATCH;
5484 		DTRACE_NFSV4_2(compound__done, struct compound_state *,
5485 		    &cs, COMPOUND4res *, resp);
5486 		return;
5487 	}
5488 
5489 	ASSERT(exi == NULL);
5490 	ASSERT(cr == NULL);
5491 
5492 	cr = crget();
5493 	ASSERT(cr != NULL);
5494 
5495 	if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) {
5496 		DTRACE_NFSV4_2(compound__start, struct compound_state *,
5497 		    &cs, COMPOUND4args *, args);
5498 		crfree(cr);
5499 		DTRACE_NFSV4_2(compound__done, struct compound_state *,
5500 		    &cs, COMPOUND4res *, resp);
5501 		svcerr_badcred(req->rq_xprt);
5502 		if (rv != NULL)
5503 			*rv = 1;
5504 		return;
5505 	}
5506 	resp->array_len = args->array_len;
5507 	resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4),
5508 	    KM_SLEEP);
5509 
5510 	cs.basecr = cr;
5511 
5512 	DTRACE_NFSV4_2(compound__start, struct compound_state *, &cs,
5513 	    COMPOUND4args *, args);
5514 
5515 	/*
5516 	 * For now, NFS4 compound processing must be protected by
5517 	 * exported_lock because it can access more than one exportinfo
5518 	 * per compound and share/unshare can now change multiple
5519 	 * exinfo structs.  The NFS2/3 code only refs 1 exportinfo
5520 	 * per proc (excluding public exinfo), and exi_count design
5521 	 * is sufficient to protect concurrent execution of NFS2/3
5522 	 * ops along with unexport.  This lock will be removed as
5523 	 * part of the NFSv4 phase 2 namespace redesign work.
5524 	 */
5525 	rw_enter(&exported_lock, RW_READER);
5526 
5527 	/*
5528 	 * If this is the first compound we've seen, we need to start all
5529 	 * new instances' grace periods.
5530 	 */
5531 	if (rfs4_seen_first_compound == 0) {
5532 		rfs4_grace_start_new();
5533 		/*
5534 		 * This must be set after rfs4_grace_start_new(), otherwise
5535 		 * another thread could proceed past here before the former
5536 		 * is finished.
5537 		 */
5538 		rfs4_seen_first_compound = 1;
5539 	}
5540 
5541 	for (i = 0; i < args->array_len && cs.cont; i++) {
5542 		nfs_argop4 *argop;
5543 		nfs_resop4 *resop;
5544 		uint_t op;
5545 
5546 		argop = &args->array[i];
5547 		resop = &resp->array[i];
5548 		resop->resop = argop->argop;
5549 		op = (uint_t)resop->resop;
5550 
5551 		if (op < rfsv4disp_cnt) {
5552 			/*
5553 			 * Count the individual ops here; NULL and COMPOUND
5554 			 * are counted in common_dispatch()
5555 			 */
5556 			rfsproccnt_v4_ptr[op].value.ui64++;
5557 
5558 			NFS4_DEBUG(rfs4_debug > 1,
5559 			    (CE_NOTE, "Executing %s", rfs4_op_string[op]));
5560 			(*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs);
5561 			NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d",
5562 			    rfs4_op_string[op], *cs.statusp));
5563 			if (*cs.statusp != NFS4_OK)
5564 				cs.cont = FALSE;
5565 		} else {
5566 			/*
5567 			 * This is effectively dead code since XDR code
5568 			 * will have already returned BADXDR if op doesn't
5569 			 * decode to legal value.  This only done for a
5570 			 * day when XDR code doesn't verify v4 opcodes.
5571 			 */
5572 			op = OP_ILLEGAL;
5573 			rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++;
5574 
5575 			rfs4_op_illegal(argop, resop, req, &cs);
5576 			cs.cont = FALSE;
5577 		}
5578 
5579 		/*
5580 		 * If not at last op, and if we are to stop, then
5581 		 * compact the results array.
5582 		 */
5583 		if ((i + 1) < args->array_len && !cs.cont) {
5584 			nfs_resop4 *new_res = kmem_alloc(
5585 			    (i+1) * sizeof (nfs_resop4), KM_SLEEP);
5586 			bcopy(resp->array,
5587 			    new_res, (i+1) * sizeof (nfs_resop4));
5588 			kmem_free(resp->array,
5589 			    args->array_len * sizeof (nfs_resop4));
5590 
5591 			resp->array_len =  i + 1;
5592 			resp->array = new_res;
5593 		}
5594 	}
5595 
5596 	rw_exit(&exported_lock);
5597 
5598 	DTRACE_NFSV4_2(compound__done, struct compound_state *, &cs,
5599 	    COMPOUND4res *, resp);
5600 
5601 	if (cs.vp)
5602 		VN_RELE(cs.vp);
5603 	if (cs.saved_vp)
5604 		VN_RELE(cs.saved_vp);
5605 	if (cs.saved_fh.nfs_fh4_val)
5606 		kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE);
5607 
5608 	if (cs.basecr)
5609 		crfree(cs.basecr);
5610 	if (cs.cr)
5611 		crfree(cs.cr);
5612 	/*
5613 	 * done with this compound request, free the label
5614 	 */
5615 
5616 	if (req->rq_label != NULL) {
5617 		kmem_free(req->rq_label, sizeof (bslabel_t));
5618 		req->rq_label = NULL;
5619 	}
5620 }
5621 
5622 /*
5623  * XXX because of what appears to be duplicate calls to rfs4_compound_free
5624  * XXX zero out the tag and array values. Need to investigate why the
5625  * XXX calls occur, but at least prevent the panic for now.
5626  */
5627 void
5628 rfs4_compound_free(COMPOUND4res *resp)
5629 {
5630 	uint_t i;
5631 
5632 	if (resp->tag.utf8string_val) {
5633 		UTF8STRING_FREE(resp->tag)
5634 	}
5635 
5636 	for (i = 0; i < resp->array_len; i++) {
5637 		nfs_resop4 *resop;
5638 		uint_t op;
5639 
5640 		resop = &resp->array[i];
5641 		op = (uint_t)resop->resop;
5642 		if (op < rfsv4disp_cnt) {
5643 			(*rfsv4disptab[op].dis_resfree)(resop);
5644 		}
5645 	}
5646 	if (resp->array != NULL) {
5647 		kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4));
5648 	}
5649 }
5650 
5651 /*
5652  * Process the value of the compound request rpc flags, as a bit-AND
5653  * of the individual per-op flags (idempotent, allowork, publicfh_ok)
5654  */
5655 void
5656 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp)
5657 {
5658 	int i;
5659 	int flag = RPC_ALL;
5660 
5661 	for (i = 0; flag && i < args->array_len; i++) {
5662 		uint_t op;
5663 
5664 		op = (uint_t)args->array[i].argop;
5665 
5666 		if (op < rfsv4disp_cnt)
5667 			flag &= rfsv4disptab[op].dis_flags;
5668 		else
5669 			flag = 0;
5670 	}
5671 	*flagp = flag;
5672 }
5673 
5674 nfsstat4
5675 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp)
5676 {
5677 	nfsstat4 e;
5678 
5679 	rfs4_dbe_lock(cp->dbe);
5680 
5681 	if (cp->sysidt != LM_NOSYSID) {
5682 		*sp = cp->sysidt;
5683 		e = NFS4_OK;
5684 
5685 	} else if ((cp->sysidt = lm_alloc_sysidt()) != LM_NOSYSID) {
5686 		*sp = cp->sysidt;
5687 		e = NFS4_OK;
5688 
5689 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
5690 		    "rfs4_client_sysid: allocated 0x%x\n", *sp));
5691 	} else
5692 		e = NFS4ERR_DELAY;
5693 
5694 	rfs4_dbe_unlock(cp->dbe);
5695 	return (e);
5696 }
5697 
5698 #if defined(DEBUG) && ! defined(lint)
5699 static void lock_print(char *str, int operation, struct flock64 *flk)
5700 {
5701 	char *op, *type;
5702 
5703 	switch (operation) {
5704 	case F_GETLK: op = "F_GETLK";
5705 		break;
5706 	case F_SETLK: op = "F_SETLK";
5707 		break;
5708 	case F_SETLK_NBMAND: op = "F_SETLK_NBMAND";
5709 		break;
5710 	default: op = "F_UNKNOWN";
5711 		break;
5712 	}
5713 	switch (flk->l_type) {
5714 	case F_UNLCK: type = "F_UNLCK";
5715 		break;
5716 	case F_RDLCK: type = "F_RDLCK";
5717 		break;
5718 	case F_WRLCK: type = "F_WRLCK";
5719 		break;
5720 	default: type = "F_UNKNOWN";
5721 		break;
5722 	}
5723 
5724 	ASSERT(flk->l_whence == 0);
5725 	cmn_err(CE_NOTE, "%s:  %s, type = %s, off = %llx len = %llx pid = %d",
5726 	    str, op, type, (longlong_t)flk->l_start,
5727 	    flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid);
5728 }
5729 
5730 #define	LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f)
5731 #else
5732 #define	LOCK_PRINT(d, s, t, f)
5733 #endif
5734 
5735 /*ARGSUSED*/
5736 static bool_t
5737 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs)
5738 {
5739 	return (TRUE);
5740 }
5741 
5742 /*
5743  * Look up the pathname using the vp in cs as the directory vnode.
5744  * cs->vp will be the vnode for the file on success
5745  */
5746 
5747 static nfsstat4
5748 rfs4_lookup(component4 *component, struct svc_req *req,
5749 	    struct compound_state *cs)
5750 {
5751 	char *nm;
5752 	uint32_t len;
5753 	nfsstat4 status;
5754 
5755 	if (cs->vp == NULL) {
5756 		return (NFS4ERR_NOFILEHANDLE);
5757 	}
5758 	if (cs->vp->v_type != VDIR) {
5759 		return (NFS4ERR_NOTDIR);
5760 	}
5761 
5762 	if (!utf8_dir_verify(component))
5763 		return (NFS4ERR_INVAL);
5764 
5765 	nm = utf8_to_fn(component, &len, NULL);
5766 	if (nm == NULL) {
5767 		return (NFS4ERR_INVAL);
5768 	}
5769 
5770 	if (len > MAXNAMELEN) {
5771 		kmem_free(nm, len);
5772 		return (NFS4ERR_NAMETOOLONG);
5773 	}
5774 
5775 	status = do_rfs4_op_lookup(nm, len, req, cs);
5776 
5777 	kmem_free(nm, len);
5778 
5779 	return (status);
5780 }
5781 
5782 static nfsstat4
5783 rfs4_lookupfile(component4 *component, struct svc_req *req,
5784 		struct compound_state *cs, uint32_t access,
5785 		change_info4 *cinfo)
5786 {
5787 	nfsstat4 status;
5788 	vnode_t *dvp = cs->vp;
5789 	vattr_t bva, ava, fva;
5790 	int error;
5791 
5792 	/* Get "before" change value */
5793 	bva.va_mask = AT_CTIME|AT_SEQ;
5794 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL);
5795 	if (error)
5796 		return (puterrno4(error));
5797 
5798 	/* rfs4_lookup may VN_RELE directory */
5799 	VN_HOLD(dvp);
5800 
5801 	status = rfs4_lookup(component, req, cs);
5802 	if (status != NFS4_OK) {
5803 		VN_RELE(dvp);
5804 		return (status);
5805 	}
5806 
5807 	/*
5808 	 * Get "after" change value, if it fails, simply return the
5809 	 * before value.
5810 	 */
5811 	ava.va_mask = AT_CTIME|AT_SEQ;
5812 	if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) {
5813 		ava.va_ctime = bva.va_ctime;
5814 		ava.va_seq = 0;
5815 	}
5816 	VN_RELE(dvp);
5817 
5818 	/*
5819 	 * Validate the file is a file
5820 	 */
5821 	fva.va_mask = AT_TYPE|AT_MODE;
5822 	error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL);
5823 	if (error)
5824 		return (puterrno4(error));
5825 
5826 	if (fva.va_type != VREG) {
5827 		if (fva.va_type == VDIR)
5828 			return (NFS4ERR_ISDIR);
5829 		if (fva.va_type == VLNK)
5830 			return (NFS4ERR_SYMLINK);
5831 		return (NFS4ERR_INVAL);
5832 	}
5833 
5834 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime);
5835 	NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
5836 
5837 	/*
5838 	 * It is undefined if VOP_LOOKUP will change va_seq, so
5839 	 * cinfo.atomic = TRUE only if we have
5840 	 * non-zero va_seq's, and they have not changed.
5841 	 */
5842 	if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq)
5843 		cinfo->atomic = TRUE;
5844 	else
5845 		cinfo->atomic = FALSE;
5846 
5847 	/* Check for mandatory locking */
5848 	cs->mandlock = MANDLOCK(cs->vp, fva.va_mode);
5849 	return (check_open_access(access, cs, req));
5850 }
5851 
5852 static nfsstat4
5853 create_vnode(vnode_t *dvp, char *nm,  vattr_t *vap, createmode4 mode,
5854 	    timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created)
5855 {
5856 	int error;
5857 	nfsstat4 status = NFS4_OK;
5858 	vattr_t va;
5859 
5860 tryagain:
5861 
5862 	/*
5863 	 * The file open mode used is VWRITE.  If the client needs
5864 	 * some other semantic, then it should do the access checking
5865 	 * itself.  It would have been nice to have the file open mode
5866 	 * passed as part of the arguments.
5867 	 */
5868 
5869 	*created = TRUE;
5870 	error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL);
5871 
5872 	if (error) {
5873 		*created = FALSE;
5874 
5875 		/*
5876 		 * If we got something other than file already exists
5877 		 * then just return this error.  Otherwise, we got
5878 		 * EEXIST.  If we were doing a GUARDED create, then
5879 		 * just return this error.  Otherwise, we need to
5880 		 * make sure that this wasn't a duplicate of an
5881 		 * exclusive create request.
5882 		 *
5883 		 * The assumption is made that a non-exclusive create
5884 		 * request will never return EEXIST.
5885 		 */
5886 
5887 		if (error != EEXIST || mode == GUARDED4) {
5888 			status = puterrno4(error);
5889 			return (status);
5890 		}
5891 		error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr,
5892 		    NULL, NULL, NULL);
5893 
5894 		if (error) {
5895 			/*
5896 			 * We couldn't find the file that we thought that
5897 			 * we just created.  So, we'll just try creating
5898 			 * it again.
5899 			 */
5900 			if (error == ENOENT)
5901 				goto tryagain;
5902 
5903 			status = puterrno4(error);
5904 			return (status);
5905 		}
5906 
5907 		if (mode == UNCHECKED4) {
5908 			/* existing object must be regular file */
5909 			if ((*vpp)->v_type != VREG) {
5910 				if ((*vpp)->v_type == VDIR)
5911 					status = NFS4ERR_ISDIR;
5912 				else if ((*vpp)->v_type == VLNK)
5913 					status = NFS4ERR_SYMLINK;
5914 				else
5915 					status = NFS4ERR_INVAL;
5916 				VN_RELE(*vpp);
5917 				return (status);
5918 			}
5919 
5920 			return (NFS4_OK);
5921 		}
5922 
5923 		/* Check for duplicate request */
5924 		ASSERT(mtime != 0);
5925 		va.va_mask = AT_MTIME;
5926 		error = VOP_GETATTR(*vpp, &va, 0, cr, NULL);
5927 		if (!error) {
5928 			/* We found the file */
5929 			if (va.va_mtime.tv_sec != mtime->tv_sec ||
5930 			    va.va_mtime.tv_nsec != mtime->tv_nsec) {
5931 				/* but its not our creation */
5932 				VN_RELE(*vpp);
5933 				return (NFS4ERR_EXIST);
5934 			}
5935 			*created = TRUE; /* retrans of create == created */
5936 			return (NFS4_OK);
5937 		}
5938 		VN_RELE(*vpp);
5939 		return (NFS4ERR_EXIST);
5940 	}
5941 
5942 	return (NFS4_OK);
5943 }
5944 
5945 static nfsstat4
5946 check_open_access(uint32_t access,
5947 		struct compound_state *cs, struct svc_req *req)
5948 {
5949 	int error;
5950 	vnode_t *vp;
5951 	bool_t readonly;
5952 	cred_t *cr = cs->cr;
5953 
5954 	/* For now we don't allow mandatory locking as per V2/V3 */
5955 	if (cs->access == CS_ACCESS_DENIED || cs->mandlock) {
5956 		return (NFS4ERR_ACCESS);
5957 	}
5958 
5959 	vp = cs->vp;
5960 	ASSERT(cr != NULL && vp->v_type == VREG);
5961 
5962 	/*
5963 	 * If the file system is exported read only and we are trying
5964 	 * to open for write, then return NFS4ERR_ROFS
5965 	 */
5966 
5967 	readonly = rdonly4(cs->exi, cs->vp, req);
5968 
5969 	if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly)
5970 		return (NFS4ERR_ROFS);
5971 
5972 	if (access & OPEN4_SHARE_ACCESS_READ) {
5973 		if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) &&
5974 		    (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) {
5975 			return (NFS4ERR_ACCESS);
5976 		}
5977 	}
5978 
5979 	if (access & OPEN4_SHARE_ACCESS_WRITE) {
5980 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
5981 		if (error)
5982 			return (NFS4ERR_ACCESS);
5983 	}
5984 
5985 	return (NFS4_OK);
5986 }
5987 
5988 static nfsstat4
5989 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs,
5990 		change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid)
5991 {
5992 	struct nfs4_svgetit_arg sarg;
5993 	struct nfs4_ntov_table ntov;
5994 
5995 	bool_t ntov_table_init = FALSE;
5996 	struct statvfs64 sb;
5997 	nfsstat4 status;
5998 	vnode_t *vp;
5999 	vattr_t bva, ava, iva, cva, *vap;
6000 	vnode_t *dvp;
6001 	timespec32_t *mtime;
6002 	char *nm = NULL;
6003 	uint_t buflen;
6004 	bool_t created;
6005 	bool_t setsize = FALSE;
6006 	len_t reqsize;
6007 	int error;
6008 	bool_t trunc;
6009 	caller_context_t ct;
6010 	component4 *component;
6011 	bslabel_t *clabel;
6012 
6013 	sarg.sbp = &sb;
6014 
6015 	dvp = cs->vp;
6016 
6017 	/* Check if the file system is read only */
6018 	if (rdonly4(cs->exi, dvp, req))
6019 		return (NFS4ERR_ROFS);
6020 
6021 	/* check the label of including directory */
6022 	if (is_system_labeled()) {
6023 		ASSERT(req->rq_label != NULL);
6024 		clabel = req->rq_label;
6025 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
6026 		    "got client label from request(1)",
6027 		    struct svc_req *, req);
6028 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
6029 			if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK)) {
6030 				return (NFS4ERR_ACCESS);
6031 			}
6032 		}
6033 	}
6034 
6035 	/*
6036 	 * Get the last component of path name in nm. cs will reference
6037 	 * the including directory on success.
6038 	 */
6039 	component = &args->open_claim4_u.file;
6040 	if (!utf8_dir_verify(component))
6041 		return (NFS4ERR_INVAL);
6042 
6043 	nm = utf8_to_fn(component, &buflen, NULL);
6044 
6045 	if (nm == NULL)
6046 		return (NFS4ERR_RESOURCE);
6047 
6048 	if (buflen > MAXNAMELEN) {
6049 		kmem_free(nm, buflen);
6050 		return (NFS4ERR_NAMETOOLONG);
6051 	}
6052 
6053 	bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ;
6054 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL);
6055 	if (error) {
6056 		kmem_free(nm, buflen);
6057 		return (puterrno4(error));
6058 	}
6059 
6060 	if (bva.va_type != VDIR) {
6061 		kmem_free(nm, buflen);
6062 		return (NFS4ERR_NOTDIR);
6063 	}
6064 
6065 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime)
6066 
6067 	switch (args->mode) {
6068 	case GUARDED4:
6069 		/*FALLTHROUGH*/
6070 	case UNCHECKED4:
6071 		nfs4_ntov_table_init(&ntov);
6072 		ntov_table_init = TRUE;
6073 
6074 		*attrset = 0;
6075 		status = do_rfs4_set_attrs(attrset,
6076 		    &args->createhow4_u.createattrs,
6077 		    cs, &sarg, &ntov, NFS4ATTR_SETIT);
6078 
6079 		if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) &&
6080 		    sarg.vap->va_type != VREG) {
6081 			if (sarg.vap->va_type == VDIR)
6082 				status = NFS4ERR_ISDIR;
6083 			else if (sarg.vap->va_type == VLNK)
6084 				status = NFS4ERR_SYMLINK;
6085 			else
6086 				status = NFS4ERR_INVAL;
6087 		}
6088 
6089 		if (status != NFS4_OK) {
6090 			kmem_free(nm, buflen);
6091 			nfs4_ntov_table_free(&ntov, &sarg);
6092 			*attrset = 0;
6093 			return (status);
6094 		}
6095 
6096 		vap = sarg.vap;
6097 		vap->va_type = VREG;
6098 		vap->va_mask |= AT_TYPE;
6099 
6100 		if ((vap->va_mask & AT_MODE) == 0) {
6101 			vap->va_mask |= AT_MODE;
6102 			vap->va_mode = (mode_t)0600;
6103 		}
6104 
6105 		if (vap->va_mask & AT_SIZE) {
6106 
6107 			/* Disallow create with a non-zero size */
6108 
6109 			if ((reqsize = sarg.vap->va_size) != 0) {
6110 				kmem_free(nm, buflen);
6111 				nfs4_ntov_table_free(&ntov, &sarg);
6112 				*attrset = 0;
6113 				return (NFS4ERR_INVAL);
6114 			}
6115 			setsize = TRUE;
6116 		}
6117 		break;
6118 
6119 	case EXCLUSIVE4:
6120 		/* prohibit EXCL create of named attributes */
6121 		if (dvp->v_flag & V_XATTRDIR) {
6122 			kmem_free(nm, buflen);
6123 			*attrset = 0;
6124 			return (NFS4ERR_INVAL);
6125 		}
6126 
6127 		cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE;
6128 		cva.va_type = VREG;
6129 		/*
6130 		 * Ensure no time overflows. Assumes underlying
6131 		 * filesystem supports at least 32 bits.
6132 		 * Truncate nsec to usec resolution to allow valid
6133 		 * compares even if the underlying filesystem truncates.
6134 		 */
6135 		mtime = (timespec32_t *)&args->createhow4_u.createverf;
6136 		cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX;
6137 		cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000;
6138 		cva.va_mode = (mode_t)0;
6139 		vap = &cva;
6140 		break;
6141 	}
6142 
6143 	status = create_vnode(dvp, nm, vap, args->mode, mtime,
6144 	    cs->cr, &vp, &created);
6145 	kmem_free(nm, buflen);
6146 
6147 	if (status != NFS4_OK) {
6148 		if (ntov_table_init)
6149 			nfs4_ntov_table_free(&ntov, &sarg);
6150 		*attrset = 0;
6151 		return (status);
6152 	}
6153 
6154 	trunc = (setsize && !created);
6155 
6156 	if (args->mode != EXCLUSIVE4) {
6157 		bitmap4 createmask = args->createhow4_u.createattrs.attrmask;
6158 
6159 		/*
6160 		 * True verification that object was created with correct
6161 		 * attrs is impossible.  The attrs could have been changed
6162 		 * immediately after object creation.  If attributes did
6163 		 * not verify, the only recourse for the server is to
6164 		 * destroy the object.  Maybe if some attrs (like gid)
6165 		 * are set incorrectly, the object should be destroyed;
6166 		 * however, seems bad as a default policy.  Do we really
6167 		 * want to destroy an object over one of the times not
6168 		 * verifying correctly?  For these reasons, the server
6169 		 * currently sets bits in attrset for createattrs
6170 		 * that were set; however, no verification is done.
6171 		 *
6172 		 * vmask_to_nmask accounts for vattr bits set on create
6173 		 *	[do_rfs4_set_attrs() only sets resp bits for
6174 		 *	 non-vattr/vfs bits.]
6175 		 * Mask off any bits we set by default so as not to return
6176 		 * more attrset bits than were requested in createattrs
6177 		 */
6178 		if (created) {
6179 			nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset);
6180 			*attrset &= createmask;
6181 		} else {
6182 			/*
6183 			 * We did not create the vnode (we tried but it
6184 			 * already existed).  In this case, the only createattr
6185 			 * that the spec allows the server to set is size,
6186 			 * and even then, it can only be set if it is 0.
6187 			 */
6188 			*attrset = 0;
6189 			if (trunc)
6190 				*attrset = FATTR4_SIZE_MASK;
6191 		}
6192 	}
6193 	if (ntov_table_init)
6194 		nfs4_ntov_table_free(&ntov, &sarg);
6195 
6196 	/*
6197 	 * Get the initial "after" sequence number, if it fails,
6198 	 * set to zero, time to before.
6199 	 */
6200 	iva.va_mask = AT_CTIME|AT_SEQ;
6201 	if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) {
6202 		iva.va_seq = 0;
6203 		iva.va_ctime = bva.va_ctime;
6204 	}
6205 
6206 	/*
6207 	 * create_vnode attempts to create the file exclusive,
6208 	 * if it already exists the VOP_CREATE will fail and
6209 	 * may not increase va_seq. It is atomic if
6210 	 * we haven't changed the directory, but if it has changed
6211 	 * we don't know what changed it.
6212 	 */
6213 	if (!created) {
6214 		if (bva.va_seq && iva.va_seq &&
6215 		    bva.va_seq == iva.va_seq)
6216 			cinfo->atomic = TRUE;
6217 		else
6218 			cinfo->atomic = FALSE;
6219 		NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime);
6220 	} else {
6221 		/*
6222 		 * The entry was created, we need to sync the
6223 		 * directory metadata.
6224 		 */
6225 		(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
6226 
6227 		/*
6228 		 * Get "after" change value, if it fails, simply return the
6229 		 * before value.
6230 		 */
6231 		ava.va_mask = AT_CTIME|AT_SEQ;
6232 		if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) {
6233 			ava.va_ctime = bva.va_ctime;
6234 			ava.va_seq = 0;
6235 		}
6236 
6237 		NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
6238 
6239 		/*
6240 		 * The cinfo->atomic = TRUE only if we have
6241 		 * non-zero va_seq's, and it has incremented by exactly one
6242 		 * during the create_vnode and it didn't
6243 		 * change during the VOP_FSYNC.
6244 		 */
6245 		if (bva.va_seq && iva.va_seq && ava.va_seq &&
6246 		    iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq)
6247 			cinfo->atomic = TRUE;
6248 		else
6249 			cinfo->atomic = FALSE;
6250 	}
6251 
6252 	/* Check for mandatory locking and that the size gets set. */
6253 	cva.va_mask = AT_MODE;
6254 	if (setsize)
6255 		cva.va_mask |= AT_SIZE;
6256 
6257 	/* Assume the worst */
6258 	cs->mandlock = TRUE;
6259 
6260 	if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) {
6261 		cs->mandlock = MANDLOCK(cs->vp, cva.va_mode);
6262 
6263 		/*
6264 		 * Truncate the file if necessary; this would be
6265 		 * the case for create over an existing file.
6266 		 */
6267 
6268 		if (trunc) {
6269 			int in_crit = 0;
6270 			rfs4_file_t *fp;
6271 			bool_t create = FALSE;
6272 
6273 			/*
6274 			 * We are writing over an existing file.
6275 			 * Check to see if we need to recall a delegation.
6276 			 */
6277 			rfs4_hold_deleg_policy();
6278 			if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) {
6279 				if (rfs4_check_delegated_byfp(FWRITE, fp,
6280 				    (reqsize == 0), FALSE, FALSE, &clientid)) {
6281 					rfs4_file_rele(fp);
6282 					rfs4_rele_deleg_policy();
6283 					VN_RELE(vp);
6284 					*attrset = 0;
6285 					return (NFS4ERR_DELAY);
6286 				}
6287 				rfs4_file_rele(fp);
6288 			}
6289 			rfs4_rele_deleg_policy();
6290 
6291 			if (nbl_need_check(vp)) {
6292 				in_crit = 1;
6293 
6294 				ASSERT(reqsize == 0);
6295 
6296 				nbl_start_crit(vp, RW_READER);
6297 				if (nbl_conflict(vp, NBL_WRITE, 0,
6298 				    cva.va_size, 0, NULL)) {
6299 					in_crit = 0;
6300 					nbl_end_crit(vp);
6301 					VN_RELE(vp);
6302 					*attrset = 0;
6303 					return (NFS4ERR_ACCESS);
6304 				}
6305 			}
6306 			ct.cc_sysid = 0;
6307 			ct.cc_pid = 0;
6308 			ct.cc_caller_id = nfs4_srv_caller_id;
6309 			ct.cc_flags = CC_DONTBLOCK;
6310 
6311 			cva.va_mask = AT_SIZE;
6312 			cva.va_size = reqsize;
6313 			(void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct);
6314 			if (in_crit)
6315 				nbl_end_crit(vp);
6316 		}
6317 	}
6318 
6319 	error = makefh4(&cs->fh, vp, cs->exi);
6320 
6321 	/*
6322 	 * Force modified data and metadata out to stable storage.
6323 	 */
6324 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL);
6325 
6326 	if (error) {
6327 		VN_RELE(vp);
6328 		*attrset = 0;
6329 		return (puterrno4(error));
6330 	}
6331 
6332 	/* if parent dir is attrdir, set namedattr fh flag */
6333 	if (dvp->v_flag & V_XATTRDIR)
6334 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
6335 
6336 	if (cs->vp)
6337 		VN_RELE(cs->vp);
6338 
6339 	cs->vp = vp;
6340 
6341 	/*
6342 	 * if we did not create the file, we will need to check
6343 	 * the access bits on the file
6344 	 */
6345 
6346 	if (!created) {
6347 		if (setsize)
6348 			args->share_access |= OPEN4_SHARE_ACCESS_WRITE;
6349 		status = check_open_access(args->share_access, cs, req);
6350 		if (status != NFS4_OK)
6351 			*attrset = 0;
6352 	}
6353 	return (status);
6354 }
6355 
6356 /*ARGSUSED*/
6357 static void
6358 rfs4_do_open(struct compound_state *cs, struct svc_req *req,
6359 		rfs4_openowner_t *oo, delegreq_t deleg,
6360 		uint32_t access, uint32_t deny,
6361 		OPEN4res *resp, int deleg_cur)
6362 {
6363 	/* XXX Currently not using req  */
6364 	rfs4_state_t *state;
6365 	rfs4_file_t *file;
6366 	bool_t screate = TRUE;
6367 	bool_t fcreate = TRUE;
6368 	uint32_t amodes;
6369 	uint32_t dmodes;
6370 	rfs4_deleg_state_t *dsp;
6371 	struct shrlock shr;
6372 	struct shr_locowner shr_loco;
6373 	sysid_t sysid;
6374 	nfsstat4 status;
6375 	caller_context_t ct;
6376 	int fflags = 0;
6377 	int recall = 0;
6378 	int err;
6379 	int cmd;
6380 
6381 	/* get the file struct and hold a lock on it during initial open */
6382 	file = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate);
6383 	if (file == NULL) {
6384 		NFS4_DEBUG(rfs4_debug,
6385 		    (CE_NOTE, "rfs4_do_open: can't find file"));
6386 		resp->status = NFS4ERR_SERVERFAULT;
6387 		return;
6388 	}
6389 
6390 	state = rfs4_findstate_by_owner_file(oo, file, &screate);
6391 	if (state == NULL) {
6392 		NFS4_DEBUG(rfs4_debug,
6393 		    (CE_NOTE, "rfs4_do_open: can't find state"));
6394 		resp->status = NFS4ERR_RESOURCE;
6395 		/* No need to keep any reference */
6396 		rfs4_file_rele_withunlock(file);
6397 		return;
6398 	}
6399 
6400 	/* try to get the sysid before continuing */
6401 	if ((status = rfs4_client_sysid(oo->client, &sysid)) != NFS4_OK) {
6402 		resp->status = status;
6403 		rfs4_file_rele(file);
6404 		/* Not a fully formed open; "close" it */
6405 		if (screate == TRUE)
6406 			rfs4_state_close(state, FALSE, FALSE, cs->cr);
6407 		rfs4_state_rele(state);
6408 		return;
6409 	}
6410 
6411 	/* Calculate the fflags for this OPEN. */
6412 	if (access & OPEN4_SHARE_ACCESS_READ)
6413 		fflags |= FREAD;
6414 	if (access & OPEN4_SHARE_ACCESS_WRITE)
6415 		fflags |= FWRITE;
6416 
6417 	/*
6418 	 * Calculate the new deny and access mode that this open is adding to
6419 	 * the file for this open owner;
6420 	 */
6421 	dmodes = (deny & ~state->share_deny);
6422 	amodes = (access & ~state->share_access);
6423 
6424 	/*
6425 	 * Check to see the client has already sent an open for this
6426 	 * open owner on this file with the same share/deny modes.
6427 	 * If so, we don't need to check for a conflict and we don't
6428 	 * need to add another shrlock.  If not, then we need to
6429 	 * check for conflicts in deny and access before checking for
6430 	 * conflicts in delegation.  We don't want to recall a
6431 	 * delegation based on an open that will eventually fail based
6432 	 * on shares modes.
6433 	 */
6434 
6435 	if (dmodes || amodes) {
6436 		shr.s_access = (short)access;
6437 		shr.s_deny = (short)deny;
6438 		shr.s_pid = rfs4_dbe_getid(oo->dbe);
6439 		shr.s_sysid = sysid;
6440 		shr_loco.sl_pid = shr.s_pid;
6441 		shr_loco.sl_id = shr.s_sysid;
6442 		shr.s_owner = (caddr_t)&shr_loco;
6443 		shr.s_own_len = sizeof (shr_loco);
6444 
6445 		cmd = nbl_need_check(cs->vp) ? F_SHARE_NBMAND : F_SHARE;
6446 		if ((err = vop_shrlock(cs->vp, cmd, &shr, fflags)) != 0) {
6447 
6448 			resp->status = err == EAGAIN ?
6449 			    NFS4ERR_SHARE_DENIED : puterrno4(err);
6450 
6451 			rfs4_file_rele(file);
6452 			/* Not a fully formed open; "close" it */
6453 			if (screate == TRUE)
6454 				rfs4_state_close(state, FALSE, FALSE, cs->cr);
6455 			rfs4_state_rele(state);
6456 			return;
6457 		}
6458 	}
6459 
6460 	rfs4_dbe_lock(state->dbe);
6461 	rfs4_dbe_lock(file->dbe);
6462 
6463 	/*
6464 	 * Check to see if this file is delegated and if so, if a
6465 	 * recall needs to be done.
6466 	 */
6467 	if (rfs4_check_recall(state, access)) {
6468 		rfs4_dbe_unlock(file->dbe);
6469 		rfs4_dbe_unlock(state->dbe);
6470 		rfs4_recall_deleg(file, FALSE, state->owner->client);
6471 		delay(NFS4_DELEGATION_CONFLICT_DELAY);
6472 		rfs4_dbe_lock(state->dbe);
6473 		rfs4_dbe_lock(file->dbe);
6474 		/* Let's see if the delegation was returned */
6475 		if (rfs4_check_recall(state, access)) {
6476 			rfs4_dbe_unlock(file->dbe);
6477 			rfs4_dbe_unlock(state->dbe);
6478 			rfs4_file_rele(file);
6479 			rfs4_update_lease(state->owner->client);
6480 			(void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags);
6481 			/* Not a fully formed open; "close" it */
6482 			if (screate == TRUE)
6483 				rfs4_state_close(state, FALSE, FALSE, cs->cr);
6484 			rfs4_state_rele(state);
6485 			resp->status = NFS4ERR_DELAY;
6486 			return;
6487 		}
6488 	}
6489 	/*
6490 	 * the share check passed and any delegation conflict has been
6491 	 * taken care of, now call vop_open.
6492 	 * if this is the first open then call vop_open with fflags.
6493 	 * if not, call vn_open_upgrade with just the upgrade flags.
6494 	 *
6495 	 * if the file has been opened already, it will have the current
6496 	 * access mode in the state struct.  if it has no share access, then
6497 	 * this is a new open.
6498 	 *
6499 	 * However, if this is open with CLAIM_DLEGATE_CUR, then don't
6500 	 * call VOP_OPEN(), just do the open upgrade.
6501 	 */
6502 	if (((state->share_access & OPEN4_SHARE_ACCESS_BOTH) == 0) &&
6503 	    !deleg_cur) {
6504 		ct.cc_sysid = sysid;
6505 		ct.cc_pid = shr.s_pid;
6506 		ct.cc_caller_id = nfs4_srv_caller_id;
6507 		ct.cc_flags = CC_DONTBLOCK;
6508 		err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct);
6509 		if (err) {
6510 			rfs4_dbe_unlock(file->dbe);
6511 			rfs4_dbe_unlock(state->dbe);
6512 			rfs4_file_rele(file);
6513 			(void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags);
6514 			/* Not a fully formed open; "close" it */
6515 			if (screate == TRUE)
6516 				rfs4_state_close(state, FALSE, FALSE, cs->cr);
6517 			rfs4_state_rele(state);
6518 			/* check if a monitor detected a delegation conflict */
6519 			if (err == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK))
6520 				resp->status = NFS4ERR_DELAY;
6521 			else
6522 				resp->status = NFS4ERR_SERVERFAULT;
6523 			return;
6524 		}
6525 	} else { /* open upgrade */
6526 		/*
6527 		 * calculate the fflags for the new mode that is being added
6528 		 * by this upgrade.
6529 		 */
6530 		fflags = 0;
6531 		if (amodes & OPEN4_SHARE_ACCESS_READ)
6532 			fflags |= FREAD;
6533 		if (amodes & OPEN4_SHARE_ACCESS_WRITE)
6534 			fflags |= FWRITE;
6535 		vn_open_upgrade(cs->vp, fflags);
6536 	}
6537 
6538 	if (dmodes & OPEN4_SHARE_DENY_READ)
6539 		file->deny_read++;
6540 	if (dmodes & OPEN4_SHARE_DENY_WRITE)
6541 		file->deny_write++;
6542 	file->share_deny |= deny;
6543 	state->share_deny |= deny;
6544 
6545 	if (amodes & OPEN4_SHARE_ACCESS_READ)
6546 		file->access_read++;
6547 	if (amodes & OPEN4_SHARE_ACCESS_WRITE)
6548 		file->access_write++;
6549 	file->share_access |= access;
6550 	state->share_access |= access;
6551 
6552 	/*
6553 	 * Check for delegation here. if the deleg argument is not
6554 	 * DELEG_ANY, then this is a reclaim from a client and
6555 	 * we must honor the delegation requested. If necessary we can
6556 	 * set the recall flag.
6557 	 */
6558 
6559 	dsp = rfs4_grant_delegation(deleg, state, &recall);
6560 
6561 	cs->deleg = (file->dinfo->dtype == OPEN_DELEGATE_WRITE);
6562 
6563 	next_stateid(&state->stateid);
6564 
6565 	resp->stateid = state->stateid.stateid;
6566 
6567 	rfs4_dbe_unlock(file->dbe);
6568 	rfs4_dbe_unlock(state->dbe);
6569 
6570 	if (dsp) {
6571 		rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall);
6572 		rfs4_deleg_state_rele(dsp);
6573 	}
6574 
6575 	rfs4_file_rele(file);
6576 	rfs4_state_rele(state);
6577 
6578 	resp->status = NFS4_OK;
6579 }
6580 
6581 /*ARGSUSED*/
6582 static void
6583 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req,
6584 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6585 {
6586 	change_info4 *cinfo = &resp->cinfo;
6587 	bitmap4 *attrset = &resp->attrset;
6588 
6589 	if (args->opentype == OPEN4_NOCREATE)
6590 		resp->status = rfs4_lookupfile(&args->open_claim4_u.file,
6591 		    req, cs, args->share_access, cinfo);
6592 	else {
6593 		/* inhibit delegation grants during exclusive create */
6594 
6595 		if (args->mode == EXCLUSIVE4)
6596 			rfs4_disable_delegation();
6597 
6598 		resp->status = rfs4_createfile(args, req, cs, cinfo, attrset,
6599 		    oo->client->clientid);
6600 	}
6601 
6602 	if (resp->status == NFS4_OK) {
6603 
6604 		/* cs->vp cs->fh now reference the desired file */
6605 
6606 		rfs4_do_open(cs, req, oo, DELEG_ANY, args->share_access,
6607 		    args->share_deny, resp, 0);
6608 
6609 		/*
6610 		 * If rfs4_createfile set attrset, we must
6611 		 * clear this attrset before the response is copied.
6612 		 */
6613 		if (resp->status != NFS4_OK && resp->attrset) {
6614 			resp->attrset = 0;
6615 		}
6616 	}
6617 	else
6618 		*cs->statusp = resp->status;
6619 
6620 	if (args->mode == EXCLUSIVE4)
6621 		rfs4_enable_delegation();
6622 }
6623 
6624 /*ARGSUSED*/
6625 static void
6626 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req,
6627 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6628 {
6629 	change_info4 *cinfo = &resp->cinfo;
6630 	vattr_t va;
6631 	vtype_t v_type = cs->vp->v_type;
6632 	int error = 0;
6633 
6634 	/* Verify that we have a regular file */
6635 	if (v_type != VREG) {
6636 		if (v_type == VDIR)
6637 			resp->status = NFS4ERR_ISDIR;
6638 		else if (v_type == VLNK)
6639 			resp->status = NFS4ERR_SYMLINK;
6640 		else
6641 			resp->status = NFS4ERR_INVAL;
6642 		return;
6643 	}
6644 
6645 	va.va_mask = AT_MODE|AT_UID;
6646 	error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL);
6647 	if (error) {
6648 		resp->status = puterrno4(error);
6649 		return;
6650 	}
6651 
6652 	cs->mandlock = MANDLOCK(cs->vp, va.va_mode);
6653 
6654 	/*
6655 	 * Check if we have access to the file, Note the the file
6656 	 * could have originally been open UNCHECKED or GUARDED
6657 	 * with mode bits that will now fail, but there is nothing
6658 	 * we can really do about that except in the case that the
6659 	 * owner of the file is the one requesting the open.
6660 	 */
6661 	if (crgetuid(cs->cr) != va.va_uid) {
6662 		resp->status = check_open_access(args->share_access, cs, req);
6663 		if (resp->status != NFS4_OK) {
6664 			return;
6665 		}
6666 	}
6667 
6668 	/*
6669 	 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero
6670 	 */
6671 	cinfo->before = 0;
6672 	cinfo->after = 0;
6673 	cinfo->atomic = FALSE;
6674 
6675 	rfs4_do_open(cs, req, oo,
6676 	    NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type),
6677 	    args->share_access, args->share_deny, resp, 0);
6678 }
6679 
6680 static void
6681 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req,
6682 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6683 {
6684 	int error;
6685 	nfsstat4 status;
6686 	stateid4 stateid =
6687 	    args->open_claim4_u.delegate_cur_info.delegate_stateid;
6688 	rfs4_deleg_state_t *dsp;
6689 
6690 	/*
6691 	 * Find the state info from the stateid and confirm that the
6692 	 * file is delegated.  If the state openowner is the same as
6693 	 * the supplied openowner we're done. If not, get the file
6694 	 * info from the found state info. Use that file info to
6695 	 * create the state for this lock owner. Note solaris doen't
6696 	 * really need the pathname to find the file. We may want to
6697 	 * lookup the pathname and make sure that the vp exist and
6698 	 * matches the vp in the file structure. However it is
6699 	 * possible that the pathname nolonger exists (local process
6700 	 * unlinks the file), so this may not be that useful.
6701 	 */
6702 
6703 	status = rfs4_get_deleg_state(&stateid, &dsp);
6704 	if (status != NFS4_OK) {
6705 		resp->status = status;
6706 		return;
6707 	}
6708 
6709 	ASSERT(dsp->finfo->dinfo->dtype != OPEN_DELEGATE_NONE);
6710 
6711 	/*
6712 	 * New lock owner, create state. Since this was probably called
6713 	 * in response to a CB_RECALL we set deleg to DELEG_NONE
6714 	 */
6715 
6716 	ASSERT(cs->vp != NULL);
6717 	VN_RELE(cs->vp);
6718 	VN_HOLD(dsp->finfo->vp);
6719 	cs->vp = dsp->finfo->vp;
6720 
6721 	if (error = makefh4(&cs->fh, cs->vp, cs->exi)) {
6722 		rfs4_deleg_state_rele(dsp);
6723 		*cs->statusp = resp->status = puterrno4(error);
6724 		return;
6725 	}
6726 
6727 	/* Mark progress for delegation returns */
6728 	dsp->finfo->dinfo->time_lastwrite = gethrestime_sec();
6729 	rfs4_deleg_state_rele(dsp);
6730 	rfs4_do_open(cs, req, oo, DELEG_NONE,
6731 	    args->share_access, args->share_deny, resp, 1);
6732 }
6733 
6734 /*ARGSUSED*/
6735 static void
6736 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req,
6737 			OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6738 {
6739 	/*
6740 	 * Lookup the pathname, it must already exist since this file
6741 	 * was delegated.
6742 	 *
6743 	 * Find the file and state info for this vp and open owner pair.
6744 	 *	check that they are in fact delegated.
6745 	 *	check that the state access and deny modes are the same.
6746 	 *
6747 	 * Return the delgation possibly seting the recall flag.
6748 	 */
6749 	rfs4_file_t *file;
6750 	rfs4_state_t *state;
6751 	bool_t create = FALSE;
6752 	bool_t dcreate = FALSE;
6753 	rfs4_deleg_state_t *dsp;
6754 	nfsace4 *ace;
6755 
6756 
6757 	/* Note we ignore oflags */
6758 	resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev,
6759 	    req, cs, args->share_access, &resp->cinfo);
6760 
6761 	if (resp->status != NFS4_OK) {
6762 		return;
6763 	}
6764 
6765 	/* get the file struct and hold a lock on it during initial open */
6766 	file = rfs4_findfile_withlock(cs->vp, NULL, &create);
6767 	if (file == NULL) {
6768 		NFS4_DEBUG(rfs4_debug,
6769 		    (CE_NOTE, "rfs4_do_opendelprev: can't find file"));
6770 		resp->status = NFS4ERR_SERVERFAULT;
6771 		return;
6772 	}
6773 
6774 	state = rfs4_findstate_by_owner_file(oo, file, &create);
6775 	if (state == NULL) {
6776 		NFS4_DEBUG(rfs4_debug,
6777 		    (CE_NOTE, "rfs4_do_opendelprev: can't find state"));
6778 		resp->status = NFS4ERR_SERVERFAULT;
6779 		rfs4_file_rele_withunlock(file);
6780 		return;
6781 	}
6782 
6783 	rfs4_dbe_lock(state->dbe);
6784 	rfs4_dbe_lock(file->dbe);
6785 	if (args->share_access != state->share_access ||
6786 	    args->share_deny != state->share_deny ||
6787 	    state->finfo->dinfo->dtype == OPEN_DELEGATE_NONE) {
6788 		NFS4_DEBUG(rfs4_debug,
6789 		    (CE_NOTE, "rfs4_do_opendelprev: state mixup"));
6790 		rfs4_dbe_unlock(file->dbe);
6791 		rfs4_dbe_unlock(state->dbe);
6792 		rfs4_file_rele(file);
6793 		rfs4_state_rele(state);
6794 		resp->status = NFS4ERR_SERVERFAULT;
6795 		return;
6796 	}
6797 	rfs4_dbe_unlock(file->dbe);
6798 	rfs4_dbe_unlock(state->dbe);
6799 
6800 	dsp = rfs4_finddeleg(state, &dcreate);
6801 	if (dsp == NULL) {
6802 		rfs4_state_rele(state);
6803 		rfs4_file_rele(file);
6804 		resp->status = NFS4ERR_SERVERFAULT;
6805 		return;
6806 	}
6807 
6808 	next_stateid(&state->stateid);
6809 
6810 	resp->stateid = state->stateid.stateid;
6811 
6812 	resp->delegation.delegation_type = dsp->dtype;
6813 
6814 	if (dsp->dtype == OPEN_DELEGATE_READ) {
6815 		open_read_delegation4 *rv =
6816 		    &resp->delegation.open_delegation4_u.read;
6817 
6818 		rv->stateid = dsp->delegid.stateid;
6819 		rv->recall = FALSE; /* no policy in place to set to TRUE */
6820 		ace = &rv->permissions;
6821 	} else {
6822 		open_write_delegation4 *rv =
6823 		    &resp->delegation.open_delegation4_u.write;
6824 
6825 		rv->stateid = dsp->delegid.stateid;
6826 		rv->recall = FALSE;  /* no policy in place to set to TRUE */
6827 		ace = &rv->permissions;
6828 		rv->space_limit.limitby = NFS_LIMIT_SIZE;
6829 		rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX;
6830 	}
6831 
6832 	/* XXX For now */
6833 	ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE;
6834 	ace->flag = 0;
6835 	ace->access_mask = 0;
6836 	ace->who.utf8string_len = 0;
6837 	ace->who.utf8string_val = 0;
6838 
6839 	rfs4_deleg_state_rele(dsp);
6840 	rfs4_state_rele(state);
6841 	rfs4_file_rele(file);
6842 }
6843 
6844 typedef enum {
6845 	NFS4_CHKSEQ_OKAY = 0,
6846 	NFS4_CHKSEQ_REPLAY = 1,
6847 	NFS4_CHKSEQ_BAD = 2
6848 } rfs4_chkseq_t;
6849 
6850 /*
6851  * Generic function for sequence number checks.
6852  */
6853 static rfs4_chkseq_t
6854 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop,
6855 		seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres)
6856 {
6857 	/* Same sequence ids and matching operations? */
6858 	if (seqid == rqst_seq && resop->resop == lastop->resop) {
6859 		if (copyres == TRUE) {
6860 			rfs4_free_reply(resop);
6861 			rfs4_copy_reply(resop, lastop);
6862 		}
6863 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
6864 		    "Replayed SEQID %d\n", seqid));
6865 		return (NFS4_CHKSEQ_REPLAY);
6866 	}
6867 
6868 	/* If the incoming sequence is not the next expected then it is bad */
6869 	if (rqst_seq != seqid + 1) {
6870 		if (rqst_seq == seqid) {
6871 			NFS4_DEBUG(rfs4_debug,
6872 			    (CE_NOTE, "BAD SEQID: Replayed sequence id "
6873 			    "but last op was %d current op is %d\n",
6874 			    lastop->resop, resop->resop));
6875 			return (NFS4_CHKSEQ_BAD);
6876 		}
6877 		NFS4_DEBUG(rfs4_debug,
6878 		    (CE_NOTE, "BAD SEQID: got %u expecting %u\n",
6879 		    rqst_seq, seqid));
6880 		return (NFS4_CHKSEQ_BAD);
6881 	}
6882 
6883 	/* Everything okay -- next expected */
6884 	return (NFS4_CHKSEQ_OKAY);
6885 }
6886 
6887 
6888 static rfs4_chkseq_t
6889 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop)
6890 {
6891 	rfs4_chkseq_t rc;
6892 
6893 	rfs4_dbe_lock(op->dbe);
6894 	rc = rfs4_check_seqid(op->open_seqid, op->reply, seqid, resop, TRUE);
6895 	rfs4_dbe_unlock(op->dbe);
6896 
6897 	if (rc == NFS4_CHKSEQ_OKAY)
6898 		rfs4_update_lease(op->client);
6899 
6900 	return (rc);
6901 }
6902 
6903 static rfs4_chkseq_t
6904 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op,
6905 	nfs_resop4 *resop)
6906 {
6907 	rfs4_chkseq_t rc;
6908 
6909 	rfs4_dbe_lock(op->dbe);
6910 	rc = rfs4_check_seqid(op->open_seqid, op->reply,
6911 	    olo_seqid, resop, FALSE);
6912 	rfs4_dbe_unlock(op->dbe);
6913 
6914 	return (rc);
6915 }
6916 
6917 static rfs4_chkseq_t
6918 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lp, nfs_resop4 *resop)
6919 {
6920 	rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY;
6921 
6922 	rfs4_dbe_lock(lp->dbe);
6923 	if (!lp->skip_seqid_check)
6924 		rc = rfs4_check_seqid(lp->seqid, lp->reply, seqid, resop, TRUE);
6925 	rfs4_dbe_unlock(lp->dbe);
6926 
6927 	return (rc);
6928 }
6929 
6930 static void
6931 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop,
6932 	    struct svc_req *req, struct compound_state *cs)
6933 {
6934 	OPEN4args *args = &argop->nfs_argop4_u.opopen;
6935 	OPEN4res *resp = &resop->nfs_resop4_u.opopen;
6936 	open_owner4 *owner = &args->owner;
6937 	open_claim_type4 claim = args->claim;
6938 	rfs4_client_t *cp;
6939 	rfs4_openowner_t *oo;
6940 	bool_t create;
6941 	bool_t replay = FALSE;
6942 	int can_reclaim;
6943 
6944 	DTRACE_NFSV4_2(op__open__start, struct compound_state *, cs,
6945 	    OPEN4args *, args);
6946 
6947 	if (cs->vp == NULL) {
6948 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
6949 		goto end;
6950 	}
6951 
6952 	/*
6953 	 * Need to check clientid and lease expiration first based on
6954 	 * error ordering and incrementing sequence id.
6955 	 */
6956 	cp = rfs4_findclient_by_id(owner->clientid, FALSE);
6957 	if (cp == NULL) {
6958 		*cs->statusp = resp->status =
6959 		    rfs4_check_clientid(&owner->clientid, 0);
6960 		goto end;
6961 	}
6962 
6963 	if (rfs4_lease_expired(cp)) {
6964 		rfs4_client_close(cp);
6965 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
6966 		goto end;
6967 	}
6968 	can_reclaim = cp->can_reclaim;
6969 
6970 	/*
6971 	 * Find the open_owner for use from this point forward.  Take
6972 	 * care in updating the sequence id based on the type of error
6973 	 * being returned.
6974 	 */
6975 retry:
6976 	create = TRUE;
6977 	oo = rfs4_findopenowner(owner, &create, args->seqid);
6978 	if (oo == NULL) {
6979 		*cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
6980 		rfs4_client_rele(cp);
6981 		goto end;
6982 	}
6983 
6984 	/* Hold off access to the sequence space while the open is done */
6985 	rfs4_sw_enter(&oo->oo_sw);
6986 
6987 	/*
6988 	 * If the open_owner existed before at the server, then check
6989 	 * the sequence id.
6990 	 */
6991 	if (!create && !oo->postpone_confirm) {
6992 		switch (rfs4_check_open_seqid(args->seqid, oo, resop)) {
6993 		case NFS4_CHKSEQ_BAD:
6994 			if ((args->seqid > oo->open_seqid) &&
6995 			    oo->need_confirm) {
6996 				rfs4_free_opens(oo, TRUE, FALSE);
6997 				rfs4_sw_exit(&oo->oo_sw);
6998 				rfs4_openowner_rele(oo);
6999 				goto retry;
7000 			}
7001 			resp->status = NFS4ERR_BAD_SEQID;
7002 			goto out;
7003 		case NFS4_CHKSEQ_REPLAY: /* replay of previous request */
7004 			replay = TRUE;
7005 			goto out;
7006 		default:
7007 			break;
7008 		}
7009 
7010 		/*
7011 		 * Sequence was ok and open owner exists
7012 		 * check to see if we have yet to see an
7013 		 * open_confirm.
7014 		 */
7015 		if (oo->need_confirm) {
7016 			rfs4_free_opens(oo, TRUE, FALSE);
7017 			rfs4_sw_exit(&oo->oo_sw);
7018 			rfs4_openowner_rele(oo);
7019 			goto retry;
7020 		}
7021 	}
7022 	/* Grace only applies to regular-type OPENs */
7023 	if (rfs4_clnt_in_grace(cp) &&
7024 	    (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) {
7025 		*cs->statusp = resp->status = NFS4ERR_GRACE;
7026 		goto out;
7027 	}
7028 
7029 	/*
7030 	 * If previous state at the server existed then can_reclaim
7031 	 * will be set. If not reply NFS4ERR_NO_GRACE to the
7032 	 * client.
7033 	 */
7034 	if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) {
7035 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
7036 		goto out;
7037 	}
7038 
7039 
7040 	/*
7041 	 * Reject the open if the client has missed the grace period
7042 	 */
7043 	if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) {
7044 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
7045 		goto out;
7046 	}
7047 
7048 	/* Couple of up-front bookkeeping items */
7049 	if (oo->need_confirm) {
7050 		/*
7051 		 * If this is a reclaim OPEN then we should not ask
7052 		 * for a confirmation of the open_owner per the
7053 		 * protocol specification.
7054 		 */
7055 		if (claim == CLAIM_PREVIOUS)
7056 			oo->need_confirm = FALSE;
7057 		else
7058 			resp->rflags |= OPEN4_RESULT_CONFIRM;
7059 	}
7060 	resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX;
7061 
7062 	/*
7063 	 * If there is an unshared filesystem mounted on this vnode,
7064 	 * do not allow to open/create in this directory.
7065 	 */
7066 	if (vn_ismntpt(cs->vp)) {
7067 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
7068 		goto out;
7069 	}
7070 
7071 	/*
7072 	 * access must READ, WRITE, or BOTH.  No access is invalid.
7073 	 * deny can be READ, WRITE, BOTH, or NONE.
7074 	 * bits not defined for access/deny are invalid.
7075 	 */
7076 	if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) ||
7077 	    (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) ||
7078 	    (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) {
7079 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7080 		goto out;
7081 	}
7082 
7083 
7084 	/*
7085 	 * make sure attrset is zero before response is built.
7086 	 */
7087 	resp->attrset = 0;
7088 
7089 	switch (claim) {
7090 	case CLAIM_NULL:
7091 		rfs4_do_opennull(cs, req, args, oo, resp);
7092 		break;
7093 	case CLAIM_PREVIOUS:
7094 		rfs4_do_openprev(cs, req, args, oo, resp);
7095 		break;
7096 	case CLAIM_DELEGATE_CUR:
7097 		rfs4_do_opendelcur(cs, req, args, oo, resp);
7098 		break;
7099 	case CLAIM_DELEGATE_PREV:
7100 		rfs4_do_opendelprev(cs, req, args, oo, resp);
7101 		break;
7102 	default:
7103 		resp->status = NFS4ERR_INVAL;
7104 		break;
7105 	}
7106 
7107 out:
7108 	rfs4_client_rele(cp);
7109 
7110 	/* Catch sequence id handling here to make it a little easier */
7111 	switch (resp->status) {
7112 	case NFS4ERR_BADXDR:
7113 	case NFS4ERR_BAD_SEQID:
7114 	case NFS4ERR_BAD_STATEID:
7115 	case NFS4ERR_NOFILEHANDLE:
7116 	case NFS4ERR_RESOURCE:
7117 	case NFS4ERR_STALE_CLIENTID:
7118 	case NFS4ERR_STALE_STATEID:
7119 		/*
7120 		 * The protocol states that if any of these errors are
7121 		 * being returned, the sequence id should not be
7122 		 * incremented.  Any other return requires an
7123 		 * increment.
7124 		 */
7125 		break;
7126 	default:
7127 		/* Always update the lease in this case */
7128 		rfs4_update_lease(oo->client);
7129 
7130 		/* Regular response - copy the result */
7131 		if (!replay)
7132 			rfs4_update_open_resp(oo, resop, &cs->fh);
7133 
7134 		/*
7135 		 * REPLAY case: Only if the previous response was OK
7136 		 * do we copy the filehandle.  If not OK, no
7137 		 * filehandle to copy.
7138 		 */
7139 		if (replay == TRUE &&
7140 		    resp->status == NFS4_OK &&
7141 		    oo->reply_fh.nfs_fh4_val) {
7142 			/*
7143 			 * If this is a replay, we must restore the
7144 			 * current filehandle/vp to that of what was
7145 			 * returned originally.  Try our best to do
7146 			 * it.
7147 			 */
7148 			nfs_fh4_fmt_t *fh_fmtp =
7149 			    (nfs_fh4_fmt_t *)oo->reply_fh.nfs_fh4_val;
7150 
7151 			cs->exi = checkexport4(&fh_fmtp->fh4_fsid,
7152 			    (fid_t *)&fh_fmtp->fh4_xlen, NULL);
7153 
7154 			if (cs->exi == NULL) {
7155 				resp->status = NFS4ERR_STALE;
7156 				goto finish;
7157 			}
7158 
7159 			VN_RELE(cs->vp);
7160 
7161 			cs->vp = nfs4_fhtovp(&oo->reply_fh, cs->exi,
7162 			    &resp->status);
7163 
7164 			if (cs->vp == NULL)
7165 				goto finish;
7166 
7167 			nfs_fh4_copy(&oo->reply_fh, &cs->fh);
7168 		}
7169 
7170 		/*
7171 		 * If this was a replay, no need to update the
7172 		 * sequence id. If the open_owner was not created on
7173 		 * this pass, then update.  The first use of an
7174 		 * open_owner will not bump the sequence id.
7175 		 */
7176 		if (replay == FALSE && !create)
7177 			rfs4_update_open_sequence(oo);
7178 		/*
7179 		 * If the client is receiving an error and the
7180 		 * open_owner needs to be confirmed, there is no way
7181 		 * to notify the client of this fact ignoring the fact
7182 		 * that the server has no method of returning a
7183 		 * stateid to confirm.  Therefore, the server needs to
7184 		 * mark this open_owner in a way as to avoid the
7185 		 * sequence id checking the next time the client uses
7186 		 * this open_owner.
7187 		 */
7188 		if (resp->status != NFS4_OK && oo->need_confirm)
7189 			oo->postpone_confirm = TRUE;
7190 		/*
7191 		 * If OK response then clear the postpone flag and
7192 		 * reset the sequence id to keep in sync with the
7193 		 * client.
7194 		 */
7195 		if (resp->status == NFS4_OK && oo->postpone_confirm) {
7196 			oo->postpone_confirm = FALSE;
7197 			oo->open_seqid = args->seqid;
7198 		}
7199 		break;
7200 	}
7201 
7202 finish:
7203 	*cs->statusp = resp->status;
7204 
7205 	rfs4_sw_exit(&oo->oo_sw);
7206 	rfs4_openowner_rele(oo);
7207 
7208 end:
7209 	DTRACE_NFSV4_2(op__open__done, struct compound_state *, cs,
7210 	    OPEN4res *, resp);
7211 }
7212 
7213 /*ARGSUSED*/
7214 void
7215 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
7216 		    struct svc_req *req, struct compound_state *cs)
7217 {
7218 	OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm;
7219 	OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm;
7220 	rfs4_state_t *sp;
7221 	nfsstat4 status;
7222 
7223 	DTRACE_NFSV4_2(op__open__confirm__start, struct compound_state *, cs,
7224 	    OPEN_CONFIRM4args *, args);
7225 
7226 	if (cs->vp == NULL) {
7227 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7228 		goto out;
7229 	}
7230 
7231 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
7232 	if (status != NFS4_OK) {
7233 		*cs->statusp = resp->status = status;
7234 		goto out;
7235 	}
7236 
7237 	/* Ensure specified filehandle matches */
7238 	if (cs->vp != sp->finfo->vp) {
7239 		rfs4_state_rele(sp);
7240 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7241 		goto out;
7242 	}
7243 
7244 	/* hold off other access to open_owner while we tinker */
7245 	rfs4_sw_enter(&sp->owner->oo_sw);
7246 
7247 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7248 	case NFS4_CHECK_STATEID_OKAY:
7249 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7250 		    resop) != 0) {
7251 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7252 			break;
7253 		}
7254 		/*
7255 		 * If it is the appropriate stateid and determined to
7256 		 * be "OKAY" then this means that the stateid does not
7257 		 * need to be confirmed and the client is in error for
7258 		 * sending an OPEN_CONFIRM.
7259 		 */
7260 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7261 		break;
7262 	case NFS4_CHECK_STATEID_OLD:
7263 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7264 		break;
7265 	case NFS4_CHECK_STATEID_BAD:
7266 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7267 		break;
7268 	case NFS4_CHECK_STATEID_EXPIRED:
7269 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7270 		break;
7271 	case NFS4_CHECK_STATEID_CLOSED:
7272 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7273 		break;
7274 	case NFS4_CHECK_STATEID_REPLAY:
7275 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
7276 		case NFS4_CHKSEQ_OKAY:
7277 			/*
7278 			 * This is replayed stateid; if seqid matches
7279 			 * next expected, then client is using wrong seqid.
7280 			 */
7281 			/* fall through */
7282 		case NFS4_CHKSEQ_BAD:
7283 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7284 			break;
7285 		case NFS4_CHKSEQ_REPLAY:
7286 			/*
7287 			 * Note this case is the duplicate case so
7288 			 * resp->status is already set.
7289 			 */
7290 			*cs->statusp = resp->status;
7291 			rfs4_update_lease(sp->owner->client);
7292 			break;
7293 		}
7294 		break;
7295 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7296 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7297 		    resop) != NFS4_CHKSEQ_OKAY) {
7298 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7299 			break;
7300 		}
7301 		*cs->statusp = resp->status = NFS4_OK;
7302 
7303 		next_stateid(&sp->stateid);
7304 		resp->open_stateid = sp->stateid.stateid;
7305 		sp->owner->need_confirm = FALSE;
7306 		rfs4_update_lease(sp->owner->client);
7307 		rfs4_update_open_sequence(sp->owner);
7308 		rfs4_update_open_resp(sp->owner, resop, NULL);
7309 		break;
7310 	default:
7311 		ASSERT(FALSE);
7312 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7313 		break;
7314 	}
7315 	rfs4_sw_exit(&sp->owner->oo_sw);
7316 	rfs4_state_rele(sp);
7317 
7318 out:
7319 	DTRACE_NFSV4_2(op__open__confirm__done, struct compound_state *, cs,
7320 	    OPEN_CONFIRM4res *, resp);
7321 }
7322 
7323 /*ARGSUSED*/
7324 void
7325 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop,
7326 		    struct svc_req *req, struct compound_state *cs)
7327 {
7328 	OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade;
7329 	OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade;
7330 	uint32_t access = args->share_access;
7331 	uint32_t deny = args->share_deny;
7332 	nfsstat4 status;
7333 	rfs4_state_t *sp;
7334 	rfs4_file_t *fp;
7335 	int fflags = 0;
7336 
7337 	DTRACE_NFSV4_2(op__open__downgrade__start, struct compound_state *, cs,
7338 	    OPEN_DOWNGRADE4args *, args);
7339 
7340 	if (cs->vp == NULL) {
7341 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7342 		goto out;
7343 	}
7344 
7345 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
7346 	if (status != NFS4_OK) {
7347 		*cs->statusp = resp->status = status;
7348 		goto out;
7349 	}
7350 
7351 	/* Ensure specified filehandle matches */
7352 	if (cs->vp != sp->finfo->vp) {
7353 		rfs4_state_rele(sp);
7354 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7355 		goto out;
7356 	}
7357 
7358 	/* hold off other access to open_owner while we tinker */
7359 	rfs4_sw_enter(&sp->owner->oo_sw);
7360 
7361 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7362 	case NFS4_CHECK_STATEID_OKAY:
7363 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7364 		    resop) != NFS4_CHKSEQ_OKAY) {
7365 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7366 			goto end;
7367 		}
7368 		break;
7369 	case NFS4_CHECK_STATEID_OLD:
7370 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7371 		goto end;
7372 	case NFS4_CHECK_STATEID_BAD:
7373 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7374 		goto end;
7375 	case NFS4_CHECK_STATEID_EXPIRED:
7376 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7377 		goto end;
7378 	case NFS4_CHECK_STATEID_CLOSED:
7379 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7380 		goto end;
7381 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7382 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7383 		goto end;
7384 	case NFS4_CHECK_STATEID_REPLAY:
7385 		/* Check the sequence id for the open owner */
7386 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
7387 		case NFS4_CHKSEQ_OKAY:
7388 			/*
7389 			 * This is replayed stateid; if seqid matches
7390 			 * next expected, then client is using wrong seqid.
7391 			 */
7392 			/* fall through */
7393 		case NFS4_CHKSEQ_BAD:
7394 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7395 			goto end;
7396 		case NFS4_CHKSEQ_REPLAY:
7397 			/*
7398 			 * Note this case is the duplicate case so
7399 			 * resp->status is already set.
7400 			 */
7401 			*cs->statusp = resp->status;
7402 			rfs4_update_lease(sp->owner->client);
7403 			goto end;
7404 		}
7405 		break;
7406 	default:
7407 		ASSERT(FALSE);
7408 		break;
7409 	}
7410 
7411 	rfs4_dbe_lock(sp->dbe);
7412 	/*
7413 	 * Check that the new access modes and deny modes are valid.
7414 	 * Check that no invalid bits are set.
7415 	 */
7416 	if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) ||
7417 	    (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) {
7418 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7419 		rfs4_update_open_sequence(sp->owner);
7420 		rfs4_dbe_unlock(sp->dbe);
7421 		goto end;
7422 	}
7423 
7424 	/*
7425 	 * The new modes must be a subset of the current modes and
7426 	 * the access must specify at least one mode. To test that
7427 	 * the new mode is a subset of the current modes we bitwise
7428 	 * AND them together and check that the result equals the new
7429 	 * mode. For example:
7430 	 * New mode, access == R and current mode, sp->share_access  == RW
7431 	 * access & sp->share_access == R == access, so the new access mode
7432 	 * is valid. Consider access == RW, sp->share_access = R
7433 	 * access & sp->share_access == R != access, so the new access mode
7434 	 * is invalid.
7435 	 */
7436 	if ((access & sp->share_access) != access ||
7437 	    (deny & sp->share_deny) != deny ||
7438 	    (access &
7439 	    (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) {
7440 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7441 		rfs4_update_open_sequence(sp->owner);
7442 		rfs4_dbe_unlock(sp->dbe);
7443 		goto end;
7444 	}
7445 
7446 	/*
7447 	 * Release any share locks associated with this stateID.
7448 	 * Strictly speaking, this violates the spec because the
7449 	 * spec effectively requires that open downgrade be atomic.
7450 	 * At present, fs_shrlock does not have this capability.
7451 	 */
7452 	rfs4_dbe_unlock(sp->dbe);
7453 	rfs4_unshare(sp);
7454 	rfs4_dbe_lock(sp->dbe);
7455 
7456 	fp = sp->finfo;
7457 	rfs4_dbe_lock(fp->dbe);
7458 
7459 	/*
7460 	 * If the current mode has deny read and the new mode
7461 	 * does not, decrement the number of deny read mode bits
7462 	 * and if it goes to zero turn off the deny read bit
7463 	 * on the file.
7464 	 */
7465 	if ((sp->share_deny & OPEN4_SHARE_DENY_READ) &&
7466 	    (deny & OPEN4_SHARE_DENY_READ) == 0) {
7467 		fp->deny_read--;
7468 		if (fp->deny_read == 0)
7469 			fp->share_deny &= ~OPEN4_SHARE_DENY_READ;
7470 	}
7471 
7472 	/*
7473 	 * If the current mode has deny write and the new mode
7474 	 * does not, decrement the number of deny write mode bits
7475 	 * and if it goes to zero turn off the deny write bit
7476 	 * on the file.
7477 	 */
7478 	if ((sp->share_deny & OPEN4_SHARE_DENY_WRITE) &&
7479 	    (deny & OPEN4_SHARE_DENY_WRITE) == 0) {
7480 		fp->deny_write--;
7481 		if (fp->deny_write == 0)
7482 			fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE;
7483 	}
7484 
7485 	/*
7486 	 * If the current mode has access read and the new mode
7487 	 * does not, decrement the number of access read mode bits
7488 	 * and if it goes to zero turn off the access read bit
7489 	 * on the file.  set fflags to FREAD for the call to
7490 	 * vn_open_downgrade().
7491 	 */
7492 	if ((sp->share_access & OPEN4_SHARE_ACCESS_READ) &&
7493 	    (access & OPEN4_SHARE_ACCESS_READ) == 0) {
7494 		fp->access_read--;
7495 		if (fp->access_read == 0)
7496 			fp->share_access &= ~OPEN4_SHARE_ACCESS_READ;
7497 		fflags |= FREAD;
7498 	}
7499 
7500 	/*
7501 	 * If the current mode has access write and the new mode
7502 	 * does not, decrement the number of access write mode bits
7503 	 * and if it goes to zero turn off the access write bit
7504 	 * on the file.  set fflags to FWRITE for the call to
7505 	 * vn_open_downgrade().
7506 	 */
7507 	if ((sp->share_access & OPEN4_SHARE_ACCESS_WRITE) &&
7508 	    (access & OPEN4_SHARE_ACCESS_WRITE) == 0) {
7509 		fp->access_write--;
7510 		if (fp->access_write == 0)
7511 			fp->share_deny &= ~OPEN4_SHARE_ACCESS_WRITE;
7512 		fflags |= FWRITE;
7513 	}
7514 
7515 	/* Set the new access and deny modes */
7516 	sp->share_access = access;
7517 	sp->share_deny = deny;
7518 	/* Check that the file is still accessible */
7519 	ASSERT(fp->share_access);
7520 
7521 	rfs4_dbe_unlock(fp->dbe);
7522 
7523 	rfs4_dbe_unlock(sp->dbe);
7524 
7525 	if ((status = rfs4_share(sp)) != NFS4_OK) {
7526 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7527 		rfs4_update_open_sequence(sp->owner);
7528 		goto end;
7529 	}
7530 
7531 	/*
7532 	 * we successfully downgraded the share lock, now we need to downgrade
7533 	 * the open.  it is possible that the downgrade was only for a deny
7534 	 * mode and we have nothing else to do.
7535 	 */
7536 	if ((fflags & (FREAD|FWRITE)) != 0)
7537 		vn_open_downgrade(cs->vp, fflags);
7538 
7539 	rfs4_dbe_lock(sp->dbe);
7540 
7541 	/* Update the stateid */
7542 	next_stateid(&sp->stateid);
7543 	resp->open_stateid = sp->stateid.stateid;
7544 
7545 	rfs4_dbe_unlock(sp->dbe);
7546 
7547 	*cs->statusp = resp->status = NFS4_OK;
7548 	/* Update the lease */
7549 	rfs4_update_lease(sp->owner->client);
7550 	/* And the sequence */
7551 	rfs4_update_open_sequence(sp->owner);
7552 	rfs4_update_open_resp(sp->owner, resop, NULL);
7553 
7554 end:
7555 	rfs4_sw_exit(&sp->owner->oo_sw);
7556 	rfs4_state_rele(sp);
7557 out:
7558 	DTRACE_NFSV4_2(op__open__downgrade__done, struct compound_state *, cs,
7559 	    OPEN_DOWNGRADE4res *, resp);
7560 }
7561 
7562 /*
7563  * The logic behind this function is detailed in the NFSv4 RFC in the
7564  * SETCLIENTID operation description under IMPLEMENTATION.  Refer to
7565  * that section for explicit guidance to server behavior for
7566  * SETCLIENTID.
7567  */
7568 void
7569 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop,
7570 		    struct svc_req *req, struct compound_state *cs)
7571 {
7572 	SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid;
7573 	SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid;
7574 	rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed;
7575 	bool_t create = TRUE;
7576 	char *addr, *netid;
7577 	int len;
7578 
7579 	DTRACE_NFSV4_2(op__setclientid__start, struct compound_state *, cs,
7580 	    SETCLIENTID4args *, args);
7581 retry:
7582 	newcp = cp_confirmed = cp_unconfirmed = NULL;
7583 
7584 	/*
7585 	 * In search of an EXISTING client matching the incoming
7586 	 * request to establish a new client identifier at the server
7587 	 */
7588 	create = TRUE;
7589 	cp = rfs4_findclient(&args->client, &create, NULL);
7590 
7591 	/* Should never happen */
7592 	ASSERT(cp != NULL);
7593 
7594 	if (cp == NULL) {
7595 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
7596 		goto out;
7597 	}
7598 
7599 	/*
7600 	 * Easiest case. Client identifier is newly created and is
7601 	 * unconfirmed.  Also note that for this case, no other
7602 	 * entries exist for the client identifier.  Nothing else to
7603 	 * check.  Just setup the response and respond.
7604 	 */
7605 	if (create) {
7606 		*cs->statusp = res->status = NFS4_OK;
7607 		res->SETCLIENTID4res_u.resok4.clientid = cp->clientid;
7608 		res->SETCLIENTID4res_u.resok4.setclientid_confirm =
7609 		    cp->confirm_verf;
7610 		/* Setup callback information; CB_NULL confirmation later */
7611 		rfs4_client_setcb(cp, &args->callback, args->callback_ident);
7612 
7613 		rfs4_client_rele(cp);
7614 		goto out;
7615 	}
7616 
7617 	/*
7618 	 * An existing, confirmed client may exist but it may not have
7619 	 * been active for at least one lease period.  If so, then
7620 	 * "close" the client and create a new client identifier
7621 	 */
7622 	if (rfs4_lease_expired(cp)) {
7623 		rfs4_client_close(cp);
7624 		goto retry;
7625 	}
7626 
7627 	if (cp->need_confirm == TRUE)
7628 		cp_unconfirmed = cp;
7629 	else
7630 		cp_confirmed = cp;
7631 
7632 	cp = NULL;
7633 
7634 	/*
7635 	 * We have a confirmed client, now check for an
7636 	 * unconfimred entry
7637 	 */
7638 	if (cp_confirmed) {
7639 		/* If creds don't match then client identifier is inuse */
7640 		if (!creds_ok(cp_confirmed->cr_set, req, cs)) {
7641 			rfs4_cbinfo_t *cbp;
7642 			/*
7643 			 * Some one else has established this client
7644 			 * id. Try and say * who they are. We will use
7645 			 * the call back address supplied by * the
7646 			 * first client.
7647 			 */
7648 			*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
7649 
7650 			addr = netid = NULL;
7651 
7652 			cbp = &cp_confirmed->cbinfo;
7653 			if (cbp->cb_callback.cb_location.r_addr &&
7654 			    cbp->cb_callback.cb_location.r_netid) {
7655 				cb_client4 *cbcp = &cbp->cb_callback;
7656 
7657 				len = strlen(cbcp->cb_location.r_addr)+1;
7658 				addr = kmem_alloc(len, KM_SLEEP);
7659 				bcopy(cbcp->cb_location.r_addr, addr, len);
7660 				len = strlen(cbcp->cb_location.r_netid)+1;
7661 				netid = kmem_alloc(len, KM_SLEEP);
7662 				bcopy(cbcp->cb_location.r_netid, netid, len);
7663 			}
7664 
7665 			res->SETCLIENTID4res_u.client_using.r_addr = addr;
7666 			res->SETCLIENTID4res_u.client_using.r_netid = netid;
7667 
7668 			rfs4_client_rele(cp_confirmed);
7669 		}
7670 
7671 		/*
7672 		 * Confirmed, creds match, and verifier matches; must
7673 		 * be an update of the callback info
7674 		 */
7675 		if (cp_confirmed->nfs_client.verifier ==
7676 		    args->client.verifier) {
7677 			/* Setup callback information */
7678 			rfs4_client_setcb(cp_confirmed, &args->callback,
7679 			    args->callback_ident);
7680 
7681 			/* everything okay -- move ahead */
7682 			*cs->statusp = res->status = NFS4_OK;
7683 			res->SETCLIENTID4res_u.resok4.clientid =
7684 			    cp_confirmed->clientid;
7685 
7686 			/* update the confirm_verifier and return it */
7687 			rfs4_client_scv_next(cp_confirmed);
7688 			res->SETCLIENTID4res_u.resok4.setclientid_confirm =
7689 			    cp_confirmed->confirm_verf;
7690 
7691 			rfs4_client_rele(cp_confirmed);
7692 			goto out;
7693 		}
7694 
7695 		/*
7696 		 * Creds match but the verifier doesn't.  Must search
7697 		 * for an unconfirmed client that would be replaced by
7698 		 * this request.
7699 		 */
7700 		create = FALSE;
7701 		cp_unconfirmed = rfs4_findclient(&args->client, &create,
7702 		    cp_confirmed);
7703 	}
7704 
7705 	/*
7706 	 * At this point, we have taken care of the brand new client
7707 	 * struct, INUSE case, update of an existing, and confirmed
7708 	 * client struct.
7709 	 */
7710 
7711 	/*
7712 	 * check to see if things have changed while we originally
7713 	 * picked up the client struct.  If they have, then return and
7714 	 * retry the processing of this SETCLIENTID request.
7715 	 */
7716 	if (cp_unconfirmed) {
7717 		rfs4_dbe_lock(cp_unconfirmed->dbe);
7718 		if (!cp_unconfirmed->need_confirm) {
7719 			rfs4_dbe_unlock(cp_unconfirmed->dbe);
7720 			rfs4_client_rele(cp_unconfirmed);
7721 			if (cp_confirmed)
7722 				rfs4_client_rele(cp_confirmed);
7723 			goto retry;
7724 		}
7725 		/* do away with the old unconfirmed one */
7726 		rfs4_dbe_invalidate(cp_unconfirmed->dbe);
7727 		rfs4_dbe_unlock(cp_unconfirmed->dbe);
7728 		rfs4_client_rele(cp_unconfirmed);
7729 		cp_unconfirmed = NULL;
7730 	}
7731 
7732 	/*
7733 	 * This search will temporarily hide the confirmed client
7734 	 * struct while a new client struct is created as the
7735 	 * unconfirmed one.
7736 	 */
7737 	create = TRUE;
7738 	newcp = rfs4_findclient(&args->client, &create, cp_confirmed);
7739 
7740 	ASSERT(newcp != NULL);
7741 
7742 	if (newcp == NULL) {
7743 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
7744 		rfs4_client_rele(cp_confirmed);
7745 		goto out;
7746 	}
7747 
7748 	/*
7749 	 * If one was not created, then a similar request must be in
7750 	 * process so release and start over with this one
7751 	 */
7752 	if (create != TRUE) {
7753 		rfs4_client_rele(newcp);
7754 		if (cp_confirmed)
7755 			rfs4_client_rele(cp_confirmed);
7756 		goto retry;
7757 	}
7758 
7759 	*cs->statusp = res->status = NFS4_OK;
7760 	res->SETCLIENTID4res_u.resok4.clientid = newcp->clientid;
7761 	res->SETCLIENTID4res_u.resok4.setclientid_confirm = newcp->confirm_verf;
7762 	/* Setup callback information; CB_NULL confirmation later */
7763 	rfs4_client_setcb(newcp, &args->callback, args->callback_ident);
7764 
7765 	newcp->cp_confirmed = cp_confirmed;
7766 
7767 	rfs4_client_rele(newcp);
7768 
7769 out:
7770 	DTRACE_NFSV4_2(op__setclientid__done, struct compound_state *, cs,
7771 	    SETCLIENTID4res *, res);
7772 }
7773 
7774 /*ARGSUSED*/
7775 void
7776 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
7777 			    struct svc_req *req, struct compound_state *cs)
7778 {
7779 	SETCLIENTID_CONFIRM4args *args =
7780 	    &argop->nfs_argop4_u.opsetclientid_confirm;
7781 	SETCLIENTID_CONFIRM4res *res =
7782 	    &resop->nfs_resop4_u.opsetclientid_confirm;
7783 	rfs4_client_t *cp, *cptoclose = NULL;
7784 
7785 	DTRACE_NFSV4_2(op__setclientid__confirm__start,
7786 	    struct compound_state *, cs,
7787 	    SETCLIENTID_CONFIRM4args *, args);
7788 
7789 	*cs->statusp = res->status = NFS4_OK;
7790 
7791 	cp = rfs4_findclient_by_id(args->clientid, TRUE);
7792 
7793 	if (cp == NULL) {
7794 		*cs->statusp = res->status =
7795 		    rfs4_check_clientid(&args->clientid, 1);
7796 		goto out;
7797 	}
7798 
7799 	if (!creds_ok(cp, req, cs)) {
7800 		*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
7801 		rfs4_client_rele(cp);
7802 		goto out;
7803 	}
7804 
7805 	/* If the verifier doesn't match, the record doesn't match */
7806 	if (cp->confirm_verf != args->setclientid_confirm) {
7807 		*cs->statusp = res->status = NFS4ERR_STALE_CLIENTID;
7808 		rfs4_client_rele(cp);
7809 		goto out;
7810 	}
7811 
7812 	rfs4_dbe_lock(cp->dbe);
7813 	cp->need_confirm = FALSE;
7814 	if (cp->cp_confirmed) {
7815 		cptoclose = cp->cp_confirmed;
7816 		cptoclose->ss_remove = 1;
7817 		cp->cp_confirmed = NULL;
7818 	}
7819 
7820 	/*
7821 	 * Update the client's associated server instance, if it's changed
7822 	 * since the client was created.
7823 	 */
7824 	if (rfs4_servinst(cp) != rfs4_cur_servinst)
7825 		rfs4_servinst_assign(cp, rfs4_cur_servinst);
7826 
7827 	/*
7828 	 * Record clientid in stable storage.
7829 	 * Must be done after server instance has been assigned.
7830 	 */
7831 	rfs4_ss_clid(cp, req);
7832 
7833 	rfs4_dbe_unlock(cp->dbe);
7834 
7835 	if (cptoclose)
7836 		/* don't need to rele, client_close does it */
7837 		rfs4_client_close(cptoclose);
7838 
7839 	/* If needed, initiate CB_NULL call for callback path */
7840 	rfs4_deleg_cb_check(cp);
7841 	rfs4_update_lease(cp);
7842 
7843 	/*
7844 	 * Check to see if client can perform reclaims
7845 	 */
7846 	rfs4_ss_chkclid(cp);
7847 
7848 	rfs4_client_rele(cp);
7849 
7850 out:
7851 	DTRACE_NFSV4_2(op__setclientid__confirm__done,
7852 	    struct compound_state *, cs,
7853 	    SETCLIENTID_CONFIRM4 *, res);
7854 }
7855 
7856 
7857 /*ARGSUSED*/
7858 void
7859 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop,
7860 	    struct svc_req *req, struct compound_state *cs)
7861 {
7862 	CLOSE4args *args = &argop->nfs_argop4_u.opclose;
7863 	CLOSE4res *resp = &resop->nfs_resop4_u.opclose;
7864 	rfs4_state_t *sp;
7865 	nfsstat4 status;
7866 
7867 	DTRACE_NFSV4_2(op__close__start, struct compound_state *, cs,
7868 	    CLOSE4args *, args);
7869 
7870 	if (cs->vp == NULL) {
7871 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7872 		goto out;
7873 	}
7874 
7875 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID);
7876 	if (status != NFS4_OK) {
7877 		*cs->statusp = resp->status = status;
7878 		goto out;
7879 	}
7880 
7881 	/* Ensure specified filehandle matches */
7882 	if (cs->vp != sp->finfo->vp) {
7883 		rfs4_state_rele(sp);
7884 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7885 		goto out;
7886 	}
7887 
7888 	/* hold off other access to open_owner while we tinker */
7889 	rfs4_sw_enter(&sp->owner->oo_sw);
7890 
7891 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7892 	case NFS4_CHECK_STATEID_OKAY:
7893 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7894 		    resop) != NFS4_CHKSEQ_OKAY) {
7895 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7896 			goto end;
7897 		}
7898 		break;
7899 	case NFS4_CHECK_STATEID_OLD:
7900 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7901 		goto end;
7902 	case NFS4_CHECK_STATEID_BAD:
7903 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7904 		goto end;
7905 	case NFS4_CHECK_STATEID_EXPIRED:
7906 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7907 		goto end;
7908 	case NFS4_CHECK_STATEID_CLOSED:
7909 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7910 		goto end;
7911 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7912 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7913 		goto end;
7914 	case NFS4_CHECK_STATEID_REPLAY:
7915 		/* Check the sequence id for the open owner */
7916 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
7917 		case NFS4_CHKSEQ_OKAY:
7918 			/*
7919 			 * This is replayed stateid; if seqid matches
7920 			 * next expected, then client is using wrong seqid.
7921 			 */
7922 			/* FALL THROUGH */
7923 		case NFS4_CHKSEQ_BAD:
7924 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7925 			goto end;
7926 		case NFS4_CHKSEQ_REPLAY:
7927 			/*
7928 			 * Note this case is the duplicate case so
7929 			 * resp->status is already set.
7930 			 */
7931 			*cs->statusp = resp->status;
7932 			rfs4_update_lease(sp->owner->client);
7933 			goto end;
7934 		}
7935 		break;
7936 	default:
7937 		ASSERT(FALSE);
7938 		break;
7939 	}
7940 
7941 	rfs4_dbe_lock(sp->dbe);
7942 
7943 	/* Update the stateid. */
7944 	next_stateid(&sp->stateid);
7945 	resp->open_stateid = sp->stateid.stateid;
7946 
7947 	rfs4_dbe_unlock(sp->dbe);
7948 
7949 	rfs4_update_lease(sp->owner->client);
7950 	rfs4_update_open_sequence(sp->owner);
7951 	rfs4_update_open_resp(sp->owner, resop, NULL);
7952 
7953 	rfs4_state_close(sp, FALSE, FALSE, cs->cr);
7954 
7955 	*cs->statusp = resp->status = status;
7956 
7957 end:
7958 	rfs4_sw_exit(&sp->owner->oo_sw);
7959 	rfs4_state_rele(sp);
7960 out:
7961 	DTRACE_NFSV4_2(op__close__done, struct compound_state *, cs,
7962 	    CLOSE4res *, resp);
7963 }
7964 
7965 /*
7966  * Manage the counts on the file struct and close all file locks
7967  */
7968 /*ARGSUSED*/
7969 void
7970 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr,
7971 	bool_t close_of_client)
7972 {
7973 	rfs4_file_t *fp = sp->finfo;
7974 	rfs4_lo_state_t *lsp;
7975 	struct shrlock shr;
7976 	struct shr_locowner shr_loco;
7977 	int fflags, s_access, s_deny;
7978 
7979 	fflags = s_access = s_deny = 0;
7980 	/*
7981 	 * Decrement the count for each access and deny bit that this
7982 	 * state has contributed to the file. If the file counts go to zero
7983 	 * clear the appropriate bit in the appropriate mask.
7984 	 */
7985 
7986 	if (sp->share_access & OPEN4_SHARE_ACCESS_READ) {
7987 		fp->access_read--;
7988 		fflags |= FREAD;
7989 		s_access |= F_RDACC;
7990 		if (fp->access_read == 0)
7991 			fp->share_access &= ~OPEN4_SHARE_ACCESS_READ;
7992 	}
7993 	if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) {
7994 		fp->access_write--;
7995 		fflags |= FWRITE;
7996 		s_access |= F_WRACC;
7997 		if (fp->access_write == 0)
7998 			fp->share_access &= ~OPEN4_SHARE_ACCESS_WRITE;
7999 	}
8000 	if (sp->share_deny & OPEN4_SHARE_DENY_READ) {
8001 		fp->deny_read--;
8002 		s_deny |= F_RDDNY;
8003 		if (fp->deny_read == 0)
8004 			fp->share_deny &= ~OPEN4_SHARE_DENY_READ;
8005 	}
8006 	if (sp->share_deny & OPEN4_SHARE_DENY_WRITE) {
8007 		fp->deny_write--;
8008 		s_deny |= F_WRDNY;
8009 		if (fp->deny_write == 0)
8010 			fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE;
8011 	}
8012 
8013 	/*
8014 	 * If this call is part of the larger closing down of client
8015 	 * state then it is just easier to release all locks
8016 	 * associated with this client instead of going through each
8017 	 * individual file and cleaning locks there.
8018 	 */
8019 	if (close_of_client) {
8020 		if (sp->owner->client->unlksys_completed == FALSE &&
8021 		    sp->lockownerlist.next->lsp != NULL &&
8022 		    sp->owner->client->sysidt != LM_NOSYSID) {
8023 			/* Is the PxFS kernel module loaded? */
8024 			if (lm_remove_file_locks != NULL) {
8025 				int new_sysid;
8026 
8027 				/* Encode the cluster nodeid in new sysid */
8028 				new_sysid = sp->owner->client->sysidt;
8029 				lm_set_nlmid_flk(&new_sysid);
8030 
8031 				/*
8032 				 * This PxFS routine removes file locks for a
8033 				 * client over all nodes of a cluster.
8034 				 */
8035 				NFS4_DEBUG(rfs4_debug, (CE_NOTE,
8036 				    "lm_remove_file_locks(sysid=0x%x)\n",
8037 				    new_sysid));
8038 				(*lm_remove_file_locks)(new_sysid);
8039 			} else {
8040 				struct flock64 flk;
8041 
8042 				/* Release all locks for this client */
8043 				flk.l_type = F_UNLKSYS;
8044 				flk.l_whence = 0;
8045 				flk.l_start = 0;
8046 				flk.l_len = 0;
8047 				flk.l_sysid = sp->owner->client->sysidt;
8048 				flk.l_pid = 0;
8049 				(void) VOP_FRLOCK(sp->finfo->vp, F_SETLK, &flk,
8050 				    F_REMOTELOCK | FREAD | FWRITE,
8051 				    (u_offset_t)0, NULL, CRED(), NULL);
8052 			}
8053 
8054 			sp->owner->client->unlksys_completed = TRUE;
8055 		}
8056 	}
8057 
8058 	/*
8059 	 * Release all locks on this file by this lock owner or at
8060 	 * least mark the locks as having been released
8061 	 */
8062 	for (lsp = sp->lockownerlist.next->lsp; lsp != NULL;
8063 	    lsp = lsp->lockownerlist.next->lsp) {
8064 
8065 		lsp->locks_cleaned = TRUE;
8066 
8067 		/* Was this already taken care of above? */
8068 		if (!close_of_client &&
8069 		    sp->owner->client->sysidt != LM_NOSYSID)
8070 			(void) cleanlocks(sp->finfo->vp, lsp->locker->pid,
8071 			    lsp->locker->client->sysidt);
8072 	}
8073 
8074 	/*
8075 	 * Release any shrlocks associated with this open state ID.
8076 	 * This must be done before the rfs4_state gets marked closed.
8077 	 */
8078 	if (sp->owner->client->sysidt != LM_NOSYSID) {
8079 		shr.s_access = s_access;
8080 		shr.s_deny = s_deny;
8081 		shr.s_pid = rfs4_dbe_getid(sp->owner->dbe);
8082 		shr.s_sysid = sp->owner->client->sysidt;
8083 		shr_loco.sl_pid = shr.s_pid;
8084 		shr_loco.sl_id = shr.s_sysid;
8085 		shr.s_owner = (caddr_t)&shr_loco;
8086 		shr.s_own_len = sizeof (shr_loco);
8087 		(void) vop_shrlock(sp->finfo->vp, F_UNSHARE, &shr, fflags);
8088 	}
8089 
8090 	(void) VOP_CLOSE(fp->vp, fflags, 1, (offset_t)0, cr, NULL);
8091 }
8092 
8093 /*
8094  * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure.
8095  */
8096 static nfsstat4
8097 lock_denied(LOCK4denied *dp, struct flock64 *flk)
8098 {
8099 	rfs4_lockowner_t *lo;
8100 	rfs4_client_t *cp;
8101 	uint32_t len;
8102 
8103 	lo = rfs4_findlockowner_by_pid(flk->l_pid);
8104 	if (lo != NULL) {
8105 		cp = lo->client;
8106 		if (rfs4_lease_expired(cp)) {
8107 			rfs4_lockowner_rele(lo);
8108 			rfs4_dbe_hold(cp->dbe);
8109 			rfs4_client_close(cp);
8110 			return (NFS4ERR_EXPIRED);
8111 		}
8112 		dp->owner.clientid = lo->owner.clientid;
8113 		len = lo->owner.owner_len;
8114 		dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
8115 		bcopy(lo->owner.owner_val, dp->owner.owner_val, len);
8116 		dp->owner.owner_len = len;
8117 		rfs4_lockowner_rele(lo);
8118 		goto finish;
8119 	}
8120 
8121 	/*
8122 	 * Its not a NFS4 lock. We take advantage that the upper 32 bits
8123 	 * of the client id contain the boot time for a NFS4 lock. So we
8124 	 * fabricate and identity by setting clientid to the sysid, and
8125 	 * the lock owner to the pid.
8126 	 */
8127 	dp->owner.clientid = flk->l_sysid;
8128 	len = sizeof (pid_t);
8129 	dp->owner.owner_len = len;
8130 	dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
8131 	bcopy(&flk->l_pid, dp->owner.owner_val, len);
8132 finish:
8133 	dp->offset = flk->l_start;
8134 	dp->length = flk->l_len;
8135 
8136 	if (flk->l_type == F_RDLCK)
8137 		dp->locktype = READ_LT;
8138 	else if (flk->l_type == F_WRLCK)
8139 		dp->locktype = WRITE_LT;
8140 	else
8141 		return (NFS4ERR_INVAL);	/* no mapping from POSIX ltype to v4 */
8142 
8143 	return (NFS4_OK);
8144 }
8145 
8146 static int
8147 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred)
8148 {
8149 	int error;
8150 	struct flock64 flk;
8151 	int i;
8152 	clock_t delaytime;
8153 	int cmd;
8154 
8155 	cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK;
8156 retry:
8157 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
8158 
8159 	for (i = 0; i < rfs4_maxlock_tries; i++) {
8160 		LOCK_PRINT(rfs4_debug, "setlock", cmd, flock);
8161 		error = VOP_FRLOCK(vp, cmd,
8162 		    flock, flag, (u_offset_t)0, NULL, cred, NULL);
8163 
8164 		if (error != EAGAIN && error != EACCES)
8165 			break;
8166 
8167 		if (i < rfs4_maxlock_tries - 1) {
8168 			delay(delaytime);
8169 			delaytime *= 2;
8170 		}
8171 	}
8172 
8173 	if (error == EAGAIN || error == EACCES) {
8174 		/* Get the owner of the lock */
8175 		flk = *flock;
8176 		LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk);
8177 		if (VOP_FRLOCK(vp, F_GETLK, &flk, flag,
8178 		    (u_offset_t)0, NULL, cred, NULL) == 0) {
8179 			if (flk.l_type == F_UNLCK) {
8180 				/* No longer locked, retry */
8181 				goto retry;
8182 			}
8183 			*flock = flk;
8184 			LOCK_PRINT(rfs4_debug, "setlock(blocking lock)",
8185 			    F_GETLK, &flk);
8186 		}
8187 	}
8188 
8189 	return (error);
8190 }
8191 
8192 /*ARGSUSED*/
8193 static nfsstat4
8194 rfs4_do_lock(rfs4_lo_state_t *lp, nfs_lock_type4 locktype,
8195 	    seqid4 seqid, offset4 offset,
8196 	    length4 length, cred_t *cred, nfs_resop4 *resop)
8197 {
8198 	nfsstat4 status;
8199 	rfs4_lockowner_t *lo = lp->locker;
8200 	rfs4_state_t *sp = lp->state;
8201 	struct flock64 flock;
8202 	int16_t ltype;
8203 	int flag;
8204 	int error;
8205 	sysid_t sysid;
8206 	LOCK4res *lres;
8207 
8208 	if (rfs4_lease_expired(lo->client)) {
8209 		return (NFS4ERR_EXPIRED);
8210 	}
8211 
8212 	if ((status = rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK)
8213 		return (status);
8214 
8215 	/* Check for zero length. To lock to end of file use all ones for V4 */
8216 	if (length == 0)
8217 		return (NFS4ERR_INVAL);
8218 	else if (length == (length4)(~0))
8219 		length = 0;		/* Posix to end of file  */
8220 
8221 retry:
8222 	rfs4_dbe_lock(sp->dbe);
8223 
8224 
8225 	if (resop->resop != OP_LOCKU) {
8226 		switch (locktype) {
8227 		case READ_LT:
8228 		case READW_LT:
8229 			if ((sp->share_access
8230 			    & OPEN4_SHARE_ACCESS_READ) == 0) {
8231 				rfs4_dbe_unlock(sp->dbe);
8232 
8233 				return (NFS4ERR_OPENMODE);
8234 			}
8235 			ltype = F_RDLCK;
8236 			break;
8237 		case WRITE_LT:
8238 		case WRITEW_LT:
8239 			if ((sp->share_access
8240 			    & OPEN4_SHARE_ACCESS_WRITE) == 0) {
8241 				rfs4_dbe_unlock(sp->dbe);
8242 
8243 				return (NFS4ERR_OPENMODE);
8244 			}
8245 			ltype = F_WRLCK;
8246 			break;
8247 		}
8248 	} else
8249 		ltype = F_UNLCK;
8250 
8251 	flock.l_type = ltype;
8252 	flock.l_whence = 0;		/* SEEK_SET */
8253 	flock.l_start = offset;
8254 	flock.l_len = length;
8255 	flock.l_sysid = sysid;
8256 	flock.l_pid = lp->locker->pid;
8257 
8258 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
8259 	if (flock.l_len < 0 || flock.l_start < 0) {
8260 		rfs4_dbe_unlock(sp->dbe);
8261 		return (NFS4ERR_INVAL);
8262 	}
8263 
8264 	/*
8265 	 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and
8266 	 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE.
8267 	 */
8268 	flag = (int)sp->share_access | F_REMOTELOCK;
8269 
8270 	error = setlock(sp->finfo->vp, &flock, flag, cred);
8271 	if (error == 0) {
8272 		rfs4_dbe_lock(lp->dbe);
8273 		next_stateid(&lp->lockid);
8274 		rfs4_dbe_unlock(lp->dbe);
8275 	}
8276 
8277 	rfs4_dbe_unlock(sp->dbe);
8278 
8279 	/*
8280 	 * N.B. We map error values to nfsv4 errors. This is differrent
8281 	 * than puterrno4 routine.
8282 	 */
8283 	switch (error) {
8284 	case 0:
8285 		status = NFS4_OK;
8286 		break;
8287 	case EAGAIN:
8288 	case EACCES:		/* Old value */
8289 		/* Can only get here if op is OP_LOCK */
8290 		ASSERT(resop->resop == OP_LOCK);
8291 		lres = &resop->nfs_resop4_u.oplock;
8292 		status = NFS4ERR_DENIED;
8293 		if (lock_denied(&lres->LOCK4res_u.denied, &flock)
8294 		    == NFS4ERR_EXPIRED)
8295 			goto retry;
8296 		break;
8297 	case ENOLCK:
8298 		status = NFS4ERR_DELAY;
8299 		break;
8300 	case EOVERFLOW:
8301 		status = NFS4ERR_INVAL;
8302 		break;
8303 	case EINVAL:
8304 		status = NFS4ERR_NOTSUPP;
8305 		break;
8306 	default:
8307 		cmn_err(CE_WARN, "rfs4_do_lock: unexpected errno (%d)", error);
8308 		status = NFS4ERR_SERVERFAULT;
8309 		break;
8310 	}
8311 
8312 	return (status);
8313 }
8314 
8315 /*ARGSUSED*/
8316 void
8317 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop,
8318 	    struct svc_req *req, struct compound_state *cs)
8319 {
8320 	LOCK4args *args = &argop->nfs_argop4_u.oplock;
8321 	LOCK4res *resp = &resop->nfs_resop4_u.oplock;
8322 	nfsstat4 status;
8323 	stateid4 *stateid;
8324 	rfs4_lockowner_t *lo;
8325 	rfs4_client_t *cp;
8326 	rfs4_state_t *sp = NULL;
8327 	rfs4_lo_state_t *lsp = NULL;
8328 	bool_t ls_sw_held = FALSE;
8329 	bool_t create = TRUE;
8330 	bool_t lcreate = TRUE;
8331 	bool_t dup_lock = FALSE;
8332 	int rc;
8333 
8334 	DTRACE_NFSV4_2(op__lock__start, struct compound_state *, cs,
8335 	    LOCK4args *, args);
8336 
8337 	if (cs->vp == NULL) {
8338 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8339 		DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8340 		    cs, LOCK4res *, resp);
8341 		return;
8342 	}
8343 
8344 	if (args->locker.new_lock_owner) {
8345 		/* Create a new lockowner for this instance */
8346 		open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner;
8347 
8348 		NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner"));
8349 
8350 		stateid = &olo->open_stateid;
8351 		status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID);
8352 		if (status != NFS4_OK) {
8353 			NFS4_DEBUG(rfs4_debug,
8354 			    (CE_NOTE, "Get state failed in lock %d", status));
8355 			*cs->statusp = resp->status = status;
8356 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8357 			    cs, LOCK4res *, resp);
8358 			return;
8359 		}
8360 
8361 		/* Ensure specified filehandle matches */
8362 		if (cs->vp != sp->finfo->vp) {
8363 			rfs4_state_rele(sp);
8364 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8365 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8366 			    cs, LOCK4res *, resp);
8367 			return;
8368 		}
8369 
8370 		/* hold off other access to open_owner while we tinker */
8371 		rfs4_sw_enter(&sp->owner->oo_sw);
8372 
8373 		switch (rc = rfs4_check_stateid_seqid(sp, stateid)) {
8374 		case NFS4_CHECK_STATEID_OLD:
8375 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8376 			goto end;
8377 		case NFS4_CHECK_STATEID_BAD:
8378 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8379 			goto end;
8380 		case NFS4_CHECK_STATEID_EXPIRED:
8381 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8382 			goto end;
8383 		case NFS4_CHECK_STATEID_UNCONFIRMED:
8384 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8385 			goto end;
8386 		case NFS4_CHECK_STATEID_CLOSED:
8387 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8388 			goto end;
8389 		case NFS4_CHECK_STATEID_OKAY:
8390 		case NFS4_CHECK_STATEID_REPLAY:
8391 			switch (rfs4_check_olo_seqid(olo->open_seqid,
8392 			    sp->owner, resop)) {
8393 			case NFS4_CHKSEQ_OKAY:
8394 				if (rc == NFS4_CHECK_STATEID_OKAY)
8395 					break;
8396 				/*
8397 				 * This is replayed stateid; if seqid
8398 				 * matches next expected, then client
8399 				 * is using wrong seqid.
8400 				 */
8401 				/* FALLTHROUGH */
8402 			case NFS4_CHKSEQ_BAD:
8403 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8404 				goto end;
8405 			case NFS4_CHKSEQ_REPLAY:
8406 				/* This is a duplicate LOCK request */
8407 				dup_lock = TRUE;
8408 
8409 				/*
8410 				 * For a duplicate we do not want to
8411 				 * create a new lockowner as it should
8412 				 * already exist.
8413 				 * Turn off the lockowner create flag.
8414 				 */
8415 				lcreate = FALSE;
8416 			}
8417 			break;
8418 		}
8419 
8420 		lo = rfs4_findlockowner(&olo->lock_owner, &lcreate);
8421 		if (lo == NULL) {
8422 			NFS4_DEBUG(rfs4_debug,
8423 			    (CE_NOTE, "rfs4_op_lock: no lock owner"));
8424 			*cs->statusp = resp->status = NFS4ERR_RESOURCE;
8425 			goto end;
8426 		}
8427 
8428 		lsp = rfs4_findlo_state_by_owner(lo, sp, &create);
8429 		if (lsp == NULL) {
8430 			rfs4_update_lease(sp->owner->client);
8431 			/*
8432 			 * Only update theh open_seqid if this is not
8433 			 * a duplicate request
8434 			 */
8435 			if (dup_lock == FALSE) {
8436 				rfs4_update_open_sequence(sp->owner);
8437 			}
8438 
8439 			NFS4_DEBUG(rfs4_debug,
8440 			    (CE_NOTE, "rfs4_op_lock: no state"));
8441 			*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
8442 			rfs4_update_open_resp(sp->owner, resop, NULL);
8443 			rfs4_lockowner_rele(lo);
8444 			goto end;
8445 		}
8446 
8447 		/*
8448 		 * This is the new_lock_owner branch and the client is
8449 		 * supposed to be associating a new lock_owner with
8450 		 * the open file at this point.  If we find that a
8451 		 * lock_owner/state association already exists and a
8452 		 * successful LOCK request was returned to the client,
8453 		 * an error is returned to the client since this is
8454 		 * not appropriate.  The client should be using the
8455 		 * existing lock_owner branch.
8456 		 */
8457 		if (dup_lock == FALSE && create == FALSE) {
8458 			if (lsp->lock_completed == TRUE) {
8459 				*cs->statusp =
8460 				    resp->status = NFS4ERR_BAD_SEQID;
8461 				rfs4_lockowner_rele(lo);
8462 				goto end;
8463 			}
8464 		}
8465 
8466 		rfs4_update_lease(sp->owner->client);
8467 
8468 		/*
8469 		 * Only update theh open_seqid if this is not
8470 		 * a duplicate request
8471 		 */
8472 		if (dup_lock == FALSE) {
8473 			rfs4_update_open_sequence(sp->owner);
8474 		}
8475 
8476 		/*
8477 		 * If this is a duplicate lock request, just copy the
8478 		 * previously saved reply and return.
8479 		 */
8480 		if (dup_lock == TRUE) {
8481 			/* verify that lock_seqid's match */
8482 			if (lsp->seqid != olo->lock_seqid) {
8483 				NFS4_DEBUG(rfs4_debug,
8484 				    (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad"
8485 				    "lsp->seqid=%d old->seqid=%d",
8486 				    lsp->seqid, olo->lock_seqid));
8487 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8488 			} else {
8489 				rfs4_copy_reply(resop, lsp->reply);
8490 				/*
8491 				 * Make sure to copy the just
8492 				 * retrieved reply status into the
8493 				 * overall compound status
8494 				 */
8495 				*cs->statusp = resp->status;
8496 			}
8497 			rfs4_lockowner_rele(lo);
8498 			goto end;
8499 		}
8500 
8501 		rfs4_dbe_lock(lsp->dbe);
8502 
8503 		/* Make sure to update the lock sequence id */
8504 		lsp->seqid = olo->lock_seqid;
8505 
8506 		NFS4_DEBUG(rfs4_debug,
8507 		    (CE_NOTE, "Lock seqid established as %d", lsp->seqid));
8508 
8509 		/*
8510 		 * This is used to signify the newly created lockowner
8511 		 * stateid and its sequence number.  The checks for
8512 		 * sequence number and increment don't occur on the
8513 		 * very first lock request for a lockowner.
8514 		 */
8515 		lsp->skip_seqid_check = TRUE;
8516 
8517 		/* hold off other access to lsp while we tinker */
8518 		rfs4_sw_enter(&lsp->ls_sw);
8519 		ls_sw_held = TRUE;
8520 
8521 		rfs4_dbe_unlock(lsp->dbe);
8522 
8523 		rfs4_lockowner_rele(lo);
8524 	} else {
8525 		stateid = &args->locker.locker4_u.lock_owner.lock_stateid;
8526 		/* get lsp and hold the lock on the underlying file struct */
8527 		if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE))
8528 		    != NFS4_OK) {
8529 			*cs->statusp = resp->status = status;
8530 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8531 			    cs, LOCK4res *, resp);
8532 			return;
8533 		}
8534 		create = FALSE;	/* We didn't create lsp */
8535 
8536 		/* Ensure specified filehandle matches */
8537 		if (cs->vp != lsp->state->finfo->vp) {
8538 			rfs4_lo_state_rele(lsp, TRUE);
8539 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8540 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8541 			    cs, LOCK4res *, resp);
8542 			return;
8543 		}
8544 
8545 		/* hold off other access to lsp while we tinker */
8546 		rfs4_sw_enter(&lsp->ls_sw);
8547 		ls_sw_held = TRUE;
8548 
8549 		switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
8550 		/*
8551 		 * The stateid looks like it was okay (expected to be
8552 		 * the next one)
8553 		 */
8554 		case NFS4_CHECK_STATEID_OKAY:
8555 			/*
8556 			 * The sequence id is now checked.  Determine
8557 			 * if this is a replay or if it is in the
8558 			 * expected (next) sequence.  In the case of a
8559 			 * replay, there are two replay conditions
8560 			 * that may occur.  The first is the normal
8561 			 * condition where a LOCK is done with a
8562 			 * NFS4_OK response and the stateid is
8563 			 * updated.  That case is handled below when
8564 			 * the stateid is identified as a REPLAY.  The
8565 			 * second is the case where an error is
8566 			 * returned, like NFS4ERR_DENIED, and the
8567 			 * sequence number is updated but the stateid
8568 			 * is not updated.  This second case is dealt
8569 			 * with here.  So it may seem odd that the
8570 			 * stateid is okay but the sequence id is a
8571 			 * replay but it is okay.
8572 			 */
8573 			switch (rfs4_check_lock_seqid(
8574 			    args->locker.locker4_u.lock_owner.lock_seqid,
8575 			    lsp, resop)) {
8576 			case NFS4_CHKSEQ_REPLAY:
8577 				if (resp->status != NFS4_OK) {
8578 					/*
8579 					 * Here is our replay and need
8580 					 * to verify that the last
8581 					 * response was an error.
8582 					 */
8583 					*cs->statusp = resp->status;
8584 					goto end;
8585 				}
8586 				/*
8587 				 * This is done since the sequence id
8588 				 * looked like a replay but it didn't
8589 				 * pass our check so a BAD_SEQID is
8590 				 * returned as a result.
8591 				 */
8592 				/*FALLTHROUGH*/
8593 			case NFS4_CHKSEQ_BAD:
8594 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8595 				goto end;
8596 			case NFS4_CHKSEQ_OKAY:
8597 				/* Everything looks okay move ahead */
8598 				break;
8599 			}
8600 			break;
8601 		case NFS4_CHECK_STATEID_OLD:
8602 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8603 			goto end;
8604 		case NFS4_CHECK_STATEID_BAD:
8605 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8606 			goto end;
8607 		case NFS4_CHECK_STATEID_EXPIRED:
8608 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8609 			goto end;
8610 		case NFS4_CHECK_STATEID_CLOSED:
8611 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8612 			goto end;
8613 		case NFS4_CHECK_STATEID_REPLAY:
8614 			switch (rfs4_check_lock_seqid(
8615 			    args->locker.locker4_u.lock_owner.lock_seqid,
8616 			    lsp, resop)) {
8617 			case NFS4_CHKSEQ_OKAY:
8618 				/*
8619 				 * This is a replayed stateid; if
8620 				 * seqid matches the next expected,
8621 				 * then client is using wrong seqid.
8622 				 */
8623 			case NFS4_CHKSEQ_BAD:
8624 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8625 				goto end;
8626 			case NFS4_CHKSEQ_REPLAY:
8627 				rfs4_update_lease(lsp->locker->client);
8628 				*cs->statusp = status = resp->status;
8629 				goto end;
8630 			}
8631 			break;
8632 		default:
8633 			ASSERT(FALSE);
8634 			break;
8635 		}
8636 
8637 		rfs4_update_lock_sequence(lsp);
8638 		rfs4_update_lease(lsp->locker->client);
8639 	}
8640 
8641 	/*
8642 	 * NFS4 only allows locking on regular files, so
8643 	 * verify type of object.
8644 	 */
8645 	if (cs->vp->v_type != VREG) {
8646 		if (cs->vp->v_type == VDIR)
8647 			status = NFS4ERR_ISDIR;
8648 		else
8649 			status = NFS4ERR_INVAL;
8650 		goto out;
8651 	}
8652 
8653 	cp = lsp->state->owner->client;
8654 
8655 	if (rfs4_clnt_in_grace(cp) && !args->reclaim) {
8656 		status = NFS4ERR_GRACE;
8657 		goto out;
8658 	}
8659 
8660 	if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->can_reclaim) {
8661 		status = NFS4ERR_NO_GRACE;
8662 		goto out;
8663 	}
8664 
8665 	if (!rfs4_clnt_in_grace(cp) && args->reclaim) {
8666 		status = NFS4ERR_NO_GRACE;
8667 		goto out;
8668 	}
8669 
8670 	if (lsp->state->finfo->dinfo->dtype == OPEN_DELEGATE_WRITE)
8671 		cs->deleg = TRUE;
8672 
8673 	status = rfs4_do_lock(lsp, args->locktype,
8674 	    args->locker.locker4_u.lock_owner.lock_seqid,
8675 	    args->offset, args->length, cs->cr, resop);
8676 
8677 out:
8678 	lsp->skip_seqid_check = FALSE;
8679 
8680 	*cs->statusp = resp->status = status;
8681 
8682 	if (status == NFS4_OK) {
8683 		resp->LOCK4res_u.lock_stateid = lsp->lockid.stateid;
8684 		lsp->lock_completed = TRUE;
8685 	}
8686 	/*
8687 	 * Only update the "OPEN" response here if this was a new
8688 	 * lock_owner
8689 	 */
8690 	if (sp)
8691 		rfs4_update_open_resp(sp->owner, resop, NULL);
8692 
8693 	rfs4_update_lock_resp(lsp, resop);
8694 
8695 end:
8696 	if (lsp) {
8697 		if (ls_sw_held)
8698 			rfs4_sw_exit(&lsp->ls_sw);
8699 		/*
8700 		 * If an sp obtained, then the lsp does not represent
8701 		 * a lock on the file struct.
8702 		 */
8703 		if (sp != NULL)
8704 			rfs4_lo_state_rele(lsp, FALSE);
8705 		else
8706 			rfs4_lo_state_rele(lsp, TRUE);
8707 	}
8708 	if (sp) {
8709 		rfs4_sw_exit(&sp->owner->oo_sw);
8710 		rfs4_state_rele(sp);
8711 	}
8712 
8713 	DTRACE_NFSV4_2(op__lock__done, struct compound_state *, cs,
8714 	    LOCK4res *, resp);
8715 }
8716 
8717 /* free function for LOCK/LOCKT */
8718 static void
8719 lock_denied_free(nfs_resop4 *resop)
8720 {
8721 	LOCK4denied *dp = NULL;
8722 
8723 	switch (resop->resop) {
8724 	case OP_LOCK:
8725 		if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED)
8726 			dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied;
8727 		break;
8728 	case OP_LOCKT:
8729 		if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED)
8730 			dp = &resop->nfs_resop4_u.oplockt.denied;
8731 		break;
8732 	default:
8733 		break;
8734 	}
8735 
8736 	if (dp)
8737 		kmem_free(dp->owner.owner_val, dp->owner.owner_len);
8738 }
8739 
8740 /*ARGSUSED*/
8741 void
8742 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop,
8743 	    struct svc_req *req, struct compound_state *cs)
8744 {
8745 	LOCKU4args *args = &argop->nfs_argop4_u.oplocku;
8746 	LOCKU4res *resp = &resop->nfs_resop4_u.oplocku;
8747 	nfsstat4 status;
8748 	stateid4 *stateid = &args->lock_stateid;
8749 	rfs4_lo_state_t *lsp;
8750 
8751 	DTRACE_NFSV4_2(op__locku__start, struct compound_state *, cs,
8752 	    LOCKU4args *, args);
8753 
8754 	if (cs->vp == NULL) {
8755 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8756 		DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
8757 		    LOCKU4res *, resp);
8758 		return;
8759 	}
8760 
8761 	if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) {
8762 		*cs->statusp = resp->status = status;
8763 		DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
8764 		    LOCKU4res *, resp);
8765 		return;
8766 	}
8767 
8768 	/* Ensure specified filehandle matches */
8769 	if (cs->vp != lsp->state->finfo->vp) {
8770 		rfs4_lo_state_rele(lsp, TRUE);
8771 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8772 		DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
8773 		    LOCKU4res *, resp);
8774 		return;
8775 	}
8776 
8777 	/* hold off other access to lsp while we tinker */
8778 	rfs4_sw_enter(&lsp->ls_sw);
8779 
8780 	switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
8781 	case NFS4_CHECK_STATEID_OKAY:
8782 		if (rfs4_check_lock_seqid(args->seqid, lsp, resop)
8783 		    != NFS4_CHKSEQ_OKAY) {
8784 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8785 			goto end;
8786 		}
8787 		break;
8788 	case NFS4_CHECK_STATEID_OLD:
8789 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8790 		goto end;
8791 	case NFS4_CHECK_STATEID_BAD:
8792 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8793 		goto end;
8794 	case NFS4_CHECK_STATEID_EXPIRED:
8795 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8796 		goto end;
8797 	case NFS4_CHECK_STATEID_CLOSED:
8798 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8799 		goto end;
8800 	case NFS4_CHECK_STATEID_REPLAY:
8801 		switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) {
8802 		case NFS4_CHKSEQ_OKAY:
8803 				/*
8804 				 * This is a replayed stateid; if
8805 				 * seqid matches the next expected,
8806 				 * then client is using wrong seqid.
8807 				 */
8808 		case NFS4_CHKSEQ_BAD:
8809 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8810 			goto end;
8811 		case NFS4_CHKSEQ_REPLAY:
8812 			rfs4_update_lease(lsp->locker->client);
8813 			*cs->statusp = status = resp->status;
8814 			goto end;
8815 		}
8816 		break;
8817 	default:
8818 		ASSERT(FALSE);
8819 		break;
8820 	}
8821 
8822 	rfs4_update_lock_sequence(lsp);
8823 	rfs4_update_lease(lsp->locker->client);
8824 
8825 	/*
8826 	 * NFS4 only allows locking on regular files, so
8827 	 * verify type of object.
8828 	 */
8829 	if (cs->vp->v_type != VREG) {
8830 		if (cs->vp->v_type == VDIR)
8831 			status = NFS4ERR_ISDIR;
8832 		else
8833 			status = NFS4ERR_INVAL;
8834 		goto out;
8835 	}
8836 
8837 	if (rfs4_clnt_in_grace(lsp->state->owner->client)) {
8838 		status = NFS4ERR_GRACE;
8839 		goto out;
8840 	}
8841 
8842 	status = rfs4_do_lock(lsp, args->locktype,
8843 	    args->seqid, args->offset, args->length, cs->cr, resop);
8844 
8845 out:
8846 	*cs->statusp = resp->status = status;
8847 
8848 	if (status == NFS4_OK)
8849 		resp->lock_stateid = lsp->lockid.stateid;
8850 
8851 	rfs4_update_lock_resp(lsp, resop);
8852 
8853 end:
8854 	rfs4_sw_exit(&lsp->ls_sw);
8855 	rfs4_lo_state_rele(lsp, TRUE);
8856 
8857 	DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
8858 	    LOCKU4res *, resp);
8859 }
8860 
8861 /*
8862  * LOCKT is a best effort routine, the client can not be guaranteed that
8863  * the status return is still in effect by the time the reply is received.
8864  * They are numerous race conditions in this routine, but we are not required
8865  * and can not be accurate.
8866  */
8867 /*ARGSUSED*/
8868 void
8869 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop,
8870 	    struct svc_req *req, struct compound_state *cs)
8871 {
8872 	LOCKT4args *args = &argop->nfs_argop4_u.oplockt;
8873 	LOCKT4res *resp = &resop->nfs_resop4_u.oplockt;
8874 	rfs4_lockowner_t *lo;
8875 	rfs4_client_t *cp;
8876 	bool_t create = FALSE;
8877 	struct flock64 flk;
8878 	int error;
8879 	int flag = FREAD | FWRITE;
8880 	int ltype;
8881 	length4 posix_length;
8882 	sysid_t sysid;
8883 	pid_t pid;
8884 
8885 	DTRACE_NFSV4_2(op__lockt__start, struct compound_state *, cs,
8886 	    LOCKT4args *, args);
8887 
8888 	if (cs->vp == NULL) {
8889 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8890 		goto out;
8891 	}
8892 
8893 	/*
8894 	 * NFS4 only allows locking on regular files, so
8895 	 * verify type of object.
8896 	 */
8897 	if (cs->vp->v_type != VREG) {
8898 		if (cs->vp->v_type == VDIR)
8899 			*cs->statusp = resp->status = NFS4ERR_ISDIR;
8900 		else
8901 			*cs->statusp = resp->status =  NFS4ERR_INVAL;
8902 		goto out;
8903 	}
8904 
8905 	/*
8906 	 * Check out the clientid to ensure the server knows about it
8907 	 * so that we correctly inform the client of a server reboot.
8908 	 */
8909 	if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE))
8910 	    == NULL) {
8911 		*cs->statusp = resp->status =
8912 		    rfs4_check_clientid(&args->owner.clientid, 0);
8913 		goto out;
8914 	}
8915 	if (rfs4_lease_expired(cp)) {
8916 		rfs4_client_close(cp);
8917 		/*
8918 		 * Protocol doesn't allow returning NFS4ERR_STALE as
8919 		 * other operations do on this check so STALE_CLIENTID
8920 		 * is returned instead
8921 		 */
8922 		*cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
8923 		goto out;
8924 	}
8925 
8926 	if (rfs4_clnt_in_grace(cp) && !(cp->can_reclaim)) {
8927 		*cs->statusp = resp->status = NFS4ERR_GRACE;
8928 		rfs4_client_rele(cp);
8929 		goto out;
8930 	}
8931 	rfs4_client_rele(cp);
8932 
8933 	resp->status = NFS4_OK;
8934 
8935 	switch (args->locktype) {
8936 	case READ_LT:
8937 	case READW_LT:
8938 		ltype = F_RDLCK;
8939 		break;
8940 	case WRITE_LT:
8941 	case WRITEW_LT:
8942 		ltype = F_WRLCK;
8943 		break;
8944 	}
8945 
8946 	posix_length = args->length;
8947 	/* Check for zero length. To lock to end of file use all ones for V4 */
8948 	if (posix_length == 0) {
8949 		*cs->statusp = resp->status = NFS4ERR_INVAL;
8950 		goto out;
8951 	} else if (posix_length == (length4)(~0)) {
8952 		posix_length = 0;	/* Posix to end of file  */
8953 	}
8954 
8955 	/* Find or create a lockowner */
8956 	lo = rfs4_findlockowner(&args->owner, &create);
8957 
8958 	if (lo) {
8959 		pid = lo->pid;
8960 		if ((resp->status =
8961 		    rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK)
8962 			goto err;
8963 	} else {
8964 		pid = 0;
8965 		sysid = lockt_sysid;
8966 	}
8967 retry:
8968 	flk.l_type = ltype;
8969 	flk.l_whence = 0;		/* SEEK_SET */
8970 	flk.l_start = args->offset;
8971 	flk.l_len = posix_length;
8972 	flk.l_sysid = sysid;
8973 	flk.l_pid = pid;
8974 	flag |= F_REMOTELOCK;
8975 
8976 	LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk);
8977 
8978 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
8979 	if (flk.l_len < 0 || flk.l_start < 0) {
8980 		resp->status = NFS4ERR_INVAL;
8981 		goto err;
8982 	}
8983 	error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0,
8984 	    NULL, cs->cr, NULL);
8985 
8986 	/*
8987 	 * N.B. We map error values to nfsv4 errors. This is differrent
8988 	 * than puterrno4 routine.
8989 	 */
8990 	switch (error) {
8991 	case 0:
8992 		if (flk.l_type == F_UNLCK)
8993 			resp->status = NFS4_OK;
8994 		else {
8995 			if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED)
8996 				goto retry;
8997 			resp->status = NFS4ERR_DENIED;
8998 		}
8999 		break;
9000 	case EOVERFLOW:
9001 		resp->status = NFS4ERR_INVAL;
9002 		break;
9003 	case EINVAL:
9004 		resp->status = NFS4ERR_NOTSUPP;
9005 		break;
9006 	default:
9007 		cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)",
9008 		    error);
9009 		resp->status = NFS4ERR_SERVERFAULT;
9010 		break;
9011 	}
9012 
9013 err:
9014 	if (lo)
9015 		rfs4_lockowner_rele(lo);
9016 	*cs->statusp = resp->status;
9017 out:
9018 	DTRACE_NFSV4_2(op__lockt__done, struct compound_state *, cs,
9019 	    LOCKT4res *, resp);
9020 }
9021 
9022 static int
9023 vop_shrlock(vnode_t *vp, int cmd, struct shrlock *sp, int fflags)
9024 {
9025 	int err;
9026 
9027 	if (cmd == F_UNSHARE && sp->s_deny == 0 && sp->s_access == 0)
9028 		return (0);
9029 
9030 	err = VOP_SHRLOCK(vp, cmd, sp, fflags, CRED(), NULL);
9031 
9032 	NFS4_DEBUG(rfs4_shrlock_debug,
9033 	    (CE_NOTE, "rfs4_shrlock %s vp=%p acc=%d dny=%d sysid=%d "
9034 	    "pid=%d err=%d\n", cmd == F_UNSHARE ? "UNSHR" : "SHARE",
9035 	    (void *) vp, sp->s_access, sp->s_deny, sp->s_sysid, sp->s_pid,
9036 	    err));
9037 
9038 	return (err);
9039 }
9040 
9041 static int
9042 rfs4_shrlock(rfs4_state_t *sp, int cmd)
9043 {
9044 	struct shrlock shr;
9045 	struct shr_locowner shr_loco;
9046 	int fflags;
9047 
9048 	fflags = shr.s_access = shr.s_deny = 0;
9049 
9050 	if (sp->share_access & OPEN4_SHARE_ACCESS_READ) {
9051 		fflags |= FREAD;
9052 		shr.s_access |= F_RDACC;
9053 	}
9054 	if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) {
9055 		fflags |= FWRITE;
9056 		shr.s_access |= F_WRACC;
9057 	}
9058 	if (sp->share_deny & OPEN4_SHARE_DENY_READ)
9059 		shr.s_deny |= F_RDDNY;
9060 	if (sp->share_deny & OPEN4_SHARE_DENY_WRITE)
9061 		shr.s_deny |= F_WRDNY;
9062 
9063 	shr.s_pid = rfs4_dbe_getid(sp->owner->dbe);
9064 	shr.s_sysid = sp->owner->client->sysidt;
9065 	shr_loco.sl_pid = shr.s_pid;
9066 	shr_loco.sl_id = shr.s_sysid;
9067 	shr.s_owner = (caddr_t)&shr_loco;
9068 	shr.s_own_len = sizeof (shr_loco);
9069 	return (vop_shrlock(sp->finfo->vp, cmd, &shr, fflags));
9070 }
9071 
9072 static int
9073 rfs4_share(rfs4_state_t *sp)
9074 {
9075 	int cmd;
9076 	cmd = nbl_need_check(sp->finfo->vp) ? F_SHARE_NBMAND : F_SHARE;
9077 	return (rfs4_shrlock(sp, cmd));
9078 }
9079 
9080 void
9081 rfs4_unshare(rfs4_state_t *sp)
9082 {
9083 	(void) rfs4_shrlock(sp, F_UNSHARE);
9084 }
9085