xref: /titanic_50/usr/src/uts/common/fs/nfs/nfs4_srv.c (revision b2b3ca14272ffe2a6fc37bab2ab65b8f6702d750)
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 2009 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 #include <sys/param.h>
32 #include <sys/types.h>
33 #include <sys/systm.h>
34 #include <sys/cred.h>
35 #include <sys/buf.h>
36 #include <sys/vfs.h>
37 #include <sys/vfs_opreg.h>
38 #include <sys/vnode.h>
39 #include <sys/uio.h>
40 #include <sys/errno.h>
41 #include <sys/sysmacros.h>
42 #include <sys/statvfs.h>
43 #include <sys/kmem.h>
44 #include <sys/dirent.h>
45 #include <sys/cmn_err.h>
46 #include <sys/debug.h>
47 #include <sys/systeminfo.h>
48 #include <sys/flock.h>
49 #include <sys/pathname.h>
50 #include <sys/nbmlock.h>
51 #include <sys/share.h>
52 #include <sys/atomic.h>
53 #include <sys/policy.h>
54 #include <sys/fem.h>
55 #include <sys/sdt.h>
56 #include <sys/ddi.h>
57 #include <sys/zone.h>
58 
59 #include <rpc/types.h>
60 #include <rpc/auth.h>
61 #include <rpc/rpcsec_gss.h>
62 #include <rpc/svc.h>
63 
64 #include <nfs/nfs.h>
65 #include <nfs/export.h>
66 #include <nfs/nfs_cmd.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 extern struct svc_ops rdma_svc_ops;
85 /* End of Tunables */
86 
87 static int rdma_setup_read_data4(READ4args *, READ4res *);
88 
89 /*
90  * Used to bump the stateid4.seqid value and show changes in the stateid
91  */
92 #define	next_stateid(sp) (++(sp)->bits.chgseq)
93 
94 /*
95  * RFS4_MINLEN_ENTRY4: XDR-encoded size of smallest possible dirent.
96  *	This is used to return NFS4ERR_TOOSMALL when clients specify
97  *	maxcount that isn't large enough to hold the smallest possible
98  *	XDR encoded dirent.
99  *
100  *	    sizeof cookie (8 bytes) +
101  *	    sizeof name_len (4 bytes) +
102  *	    sizeof smallest (padded) name (4 bytes) +
103  *	    sizeof bitmap4_len (12 bytes) +   NOTE: we always encode len=2 bm4
104  *	    sizeof attrlist4_len (4 bytes) +
105  *	    sizeof next boolean (4 bytes)
106  *
107  * RFS4_MINLEN_RDDIR4: XDR-encoded size of READDIR op reply containing
108  * the smallest possible entry4 (assumes no attrs requested).
109  *	sizeof nfsstat4 (4 bytes) +
110  *	sizeof verifier4 (8 bytes) +
111  *	sizeof entry4list bool (4 bytes) +
112  *	sizeof entry4 	(36 bytes) +
113  *	sizeof eof bool  (4 bytes)
114  *
115  * RFS4_MINLEN_RDDIR_BUF: minimum length of buffer server will provide to
116  *	VOP_READDIR.  Its value is the size of the maximum possible dirent
117  *	for solaris.  The DIRENT64_RECLEN macro returns	the size of dirent
118  *	required for a given name length.  MAXNAMELEN is the maximum
119  *	filename length allowed in Solaris.  The first two DIRENT64_RECLEN()
120  *	macros are to allow for . and .. entries -- just a minor tweak to try
121  *	and guarantee that buffer we give to VOP_READDIR will be large enough
122  *	to hold ., .., and the largest possible solaris dirent64.
123  */
124 #define	RFS4_MINLEN_ENTRY4 36
125 #define	RFS4_MINLEN_RDDIR4 (4 + NFS4_VERIFIER_SIZE + 4 + RFS4_MINLEN_ENTRY4 + 4)
126 #define	RFS4_MINLEN_RDDIR_BUF \
127 	(DIRENT64_RECLEN(1) + DIRENT64_RECLEN(2) + DIRENT64_RECLEN(MAXNAMELEN))
128 
129 /*
130  * It would be better to pad to 4 bytes since that's what XDR would do,
131  * but the dirents UFS gives us are already padded to 8, so just take
132  * what we're given.  Dircount is only a hint anyway.  Currently the
133  * solaris kernel is ASCII only, so there's no point in calling the
134  * UTF8 functions.
135  *
136  * dirent64: named padded to provide 8 byte struct alignment
137  *	d_ino(8) + d_off(8) + d_reclen(2) + d_name(namelen + null(1) + pad)
138  *
139  * cookie: uint64_t   +  utf8namelen: uint_t  +   utf8name padded to 8 bytes
140  *
141  */
142 #define	DIRENT64_TO_DIRCOUNT(dp) \
143 	(3 * BYTES_PER_XDR_UNIT + DIRENT64_NAMELEN((dp)->d_reclen))
144 
145 time_t rfs4_start_time;			/* Initialized in rfs4_srvrinit */
146 
147 static sysid_t lockt_sysid;		/* dummy sysid for all LOCKT calls */
148 
149 u_longlong_t	nfs4_srv_caller_id;
150 uint_t		nfs4_srv_vkey = 0;
151 
152 verifier4	Write4verf;
153 verifier4	Readdir4verf;
154 
155 void	rfs4_init_compound_state(struct compound_state *);
156 
157 static void	nullfree(caddr_t);
158 static void	rfs4_op_inval(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
159 			struct compound_state *);
160 static void	rfs4_op_access(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
161 			struct compound_state *);
162 static void	rfs4_op_close(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
163 			struct compound_state *);
164 static void	rfs4_op_commit(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
165 			struct compound_state *);
166 static void	rfs4_op_create(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
167 			struct compound_state *);
168 static void	rfs4_op_create_free(nfs_resop4 *resop);
169 static void	rfs4_op_delegreturn(nfs_argop4 *, nfs_resop4 *,
170 			struct svc_req *, struct compound_state *);
171 static void	rfs4_op_delegpurge(nfs_argop4 *, nfs_resop4 *,
172 			struct svc_req *, struct compound_state *);
173 static void	rfs4_op_getattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
174 			struct compound_state *);
175 static void	rfs4_op_getattr_free(nfs_resop4 *);
176 static void	rfs4_op_getfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
177 			struct compound_state *);
178 static void	rfs4_op_getfh_free(nfs_resop4 *);
179 static void	rfs4_op_illegal(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
180 			struct compound_state *);
181 static void	rfs4_op_link(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
182 			struct compound_state *);
183 static void	rfs4_op_lock(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
184 			struct compound_state *);
185 static void	lock_denied_free(nfs_resop4 *);
186 static void	rfs4_op_locku(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
187 			struct compound_state *);
188 static void	rfs4_op_lockt(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
189 			struct compound_state *);
190 static void	rfs4_op_lookup(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
191 			struct compound_state *);
192 static void	rfs4_op_lookupp(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
193 			struct compound_state *);
194 static void	rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop,
195 				struct svc_req *req, struct compound_state *cs);
196 static void	rfs4_op_nverify(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
197 			struct compound_state *);
198 static void	rfs4_op_open(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
199 			struct compound_state *);
200 static void	rfs4_op_open_confirm(nfs_argop4 *, nfs_resop4 *,
201 			struct svc_req *, struct compound_state *);
202 static void	rfs4_op_open_downgrade(nfs_argop4 *, nfs_resop4 *,
203 			struct svc_req *, struct compound_state *);
204 static void	rfs4_op_putfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
205 			struct compound_state *);
206 static void	rfs4_op_putpubfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
207 			struct compound_state *);
208 static void	rfs4_op_putrootfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
209 			struct compound_state *);
210 static void	rfs4_op_read(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
211 			struct compound_state *);
212 static void	rfs4_op_read_free(nfs_resop4 *);
213 static void	rfs4_op_readdir_free(nfs_resop4 *resop);
214 static void	rfs4_op_readlink(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
215 			struct compound_state *);
216 static void	rfs4_op_readlink_free(nfs_resop4 *);
217 static void	rfs4_op_release_lockowner(nfs_argop4 *, nfs_resop4 *,
218 			struct svc_req *, struct compound_state *);
219 static void	rfs4_op_remove(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
220 			struct compound_state *);
221 static void	rfs4_op_rename(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
222 			struct compound_state *);
223 static void	rfs4_op_renew(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
224 			struct compound_state *);
225 static void	rfs4_op_restorefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
226 			struct compound_state *);
227 static void	rfs4_op_savefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
228 			struct compound_state *);
229 static void	rfs4_op_setattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
230 			struct compound_state *);
231 static void	rfs4_op_verify(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
232 			struct compound_state *);
233 static void	rfs4_op_write(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
234 			struct compound_state *);
235 static void	rfs4_op_setclientid(nfs_argop4 *, nfs_resop4 *,
236 			struct svc_req *, struct compound_state *);
237 static void	rfs4_op_setclientid_confirm(nfs_argop4 *, nfs_resop4 *,
238 			struct svc_req *req, struct compound_state *);
239 static void	rfs4_op_secinfo(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
240 			struct compound_state *);
241 static void	rfs4_op_secinfo_free(nfs_resop4 *);
242 
243 static nfsstat4 check_open_access(uint32_t,
244 				struct compound_state *, struct svc_req *);
245 nfsstat4 rfs4_client_sysid(rfs4_client_t *, sysid_t *);
246 static int	vop_shrlock(vnode_t *, int, struct shrlock *, int);
247 static int 	rfs4_shrlock(rfs4_state_t *, int);
248 static int	rfs4_share(rfs4_state_t *);
249 void rfs4_ss_clid(rfs4_client_t *, struct svc_req *);
250 
251 /*
252  * translation table for attrs
253  */
254 struct nfs4_ntov_table {
255 	union nfs4_attr_u *na;
256 	uint8_t amap[NFS4_MAXNUM_ATTRS];
257 	int attrcnt;
258 	bool_t vfsstat;
259 };
260 
261 static void	nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp);
262 static void	nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp,
263 				    struct nfs4_svgetit_arg *sargp);
264 
265 static nfsstat4	do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp,
266 		    struct compound_state *cs, struct nfs4_svgetit_arg *sargp,
267 		    struct nfs4_ntov_table *ntovp, nfs4_attr_cmd_t cmd);
268 
269 fem_t		*deleg_rdops;
270 fem_t		*deleg_wrops;
271 
272 rfs4_servinst_t *rfs4_cur_servinst = NULL;	/* current server instance */
273 kmutex_t	rfs4_servinst_lock;	/* protects linked list */
274 int		rfs4_seen_first_compound;	/* set first time we see one */
275 
276 /*
277  * NFS4 op dispatch table
278  */
279 
280 struct rfsv4disp {
281 	void	(*dis_proc)();		/* proc to call */
282 	void	(*dis_resfree)();	/* frees space allocated by proc */
283 	int	dis_flags;		/* RPC_IDEMPOTENT, etc... */
284 };
285 
286 static struct rfsv4disp rfsv4disptab[] = {
287 	/*
288 	 * NFS VERSION 4
289 	 */
290 
291 	/* RFS_NULL = 0 */
292 	{rfs4_op_illegal, nullfree, 0},
293 
294 	/* UNUSED = 1 */
295 	{rfs4_op_illegal, nullfree, 0},
296 
297 	/* UNUSED = 2 */
298 	{rfs4_op_illegal, nullfree, 0},
299 
300 	/* OP_ACCESS = 3 */
301 	{rfs4_op_access, nullfree, RPC_IDEMPOTENT},
302 
303 	/* OP_CLOSE = 4 */
304 	{rfs4_op_close, nullfree, 0},
305 
306 	/* OP_COMMIT = 5 */
307 	{rfs4_op_commit, nullfree, RPC_IDEMPOTENT},
308 
309 	/* OP_CREATE = 6 */
310 	{rfs4_op_create, nullfree, 0},
311 
312 	/* OP_DELEGPURGE = 7 */
313 	{rfs4_op_delegpurge, nullfree, 0},
314 
315 	/* OP_DELEGRETURN = 8 */
316 	{rfs4_op_delegreturn, nullfree, 0},
317 
318 	/* OP_GETATTR = 9 */
319 	{rfs4_op_getattr, rfs4_op_getattr_free, RPC_IDEMPOTENT},
320 
321 	/* OP_GETFH = 10 */
322 	{rfs4_op_getfh, rfs4_op_getfh_free, RPC_ALL},
323 
324 	/* OP_LINK = 11 */
325 	{rfs4_op_link, nullfree, 0},
326 
327 	/* OP_LOCK = 12 */
328 	{rfs4_op_lock, lock_denied_free, 0},
329 
330 	/* OP_LOCKT = 13 */
331 	{rfs4_op_lockt, lock_denied_free, 0},
332 
333 	/* OP_LOCKU = 14 */
334 	{rfs4_op_locku, nullfree, 0},
335 
336 	/* OP_LOOKUP = 15 */
337 	{rfs4_op_lookup, nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK)},
338 
339 	/* OP_LOOKUPP = 16 */
340 	{rfs4_op_lookupp, nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK)},
341 
342 	/* OP_NVERIFY = 17 */
343 	{rfs4_op_nverify, nullfree, RPC_IDEMPOTENT},
344 
345 	/* OP_OPEN = 18 */
346 	{rfs4_op_open, rfs4_free_reply, 0},
347 
348 	/* OP_OPENATTR = 19 */
349 	{rfs4_op_openattr, nullfree, 0},
350 
351 	/* OP_OPEN_CONFIRM = 20 */
352 	{rfs4_op_open_confirm, nullfree, 0},
353 
354 	/* OP_OPEN_DOWNGRADE = 21 */
355 	{rfs4_op_open_downgrade, nullfree, 0},
356 
357 	/* OP_OPEN_PUTFH = 22 */
358 	{rfs4_op_putfh, nullfree, RPC_ALL},
359 
360 	/* OP_PUTPUBFH = 23 */
361 	{rfs4_op_putpubfh, nullfree, RPC_ALL},
362 
363 	/* OP_PUTROOTFH = 24 */
364 	{rfs4_op_putrootfh, nullfree, RPC_ALL},
365 
366 	/* OP_READ = 25 */
367 	{rfs4_op_read, rfs4_op_read_free, RPC_IDEMPOTENT},
368 
369 	/* OP_READDIR = 26 */
370 	{rfs4_op_readdir, rfs4_op_readdir_free, RPC_IDEMPOTENT},
371 
372 	/* OP_READLINK = 27 */
373 	{rfs4_op_readlink, rfs4_op_readlink_free, RPC_IDEMPOTENT},
374 
375 	/* OP_REMOVE = 28 */
376 	{rfs4_op_remove, nullfree, 0},
377 
378 	/* OP_RENAME = 29 */
379 	{rfs4_op_rename, nullfree, 0},
380 
381 	/* OP_RENEW = 30 */
382 	{rfs4_op_renew, nullfree, 0},
383 
384 	/* OP_RESTOREFH = 31 */
385 	{rfs4_op_restorefh, nullfree, RPC_ALL},
386 
387 	/* OP_SAVEFH = 32 */
388 	{rfs4_op_savefh, nullfree, RPC_ALL},
389 
390 	/* OP_SECINFO = 33 */
391 	{rfs4_op_secinfo, rfs4_op_secinfo_free, 0},
392 
393 	/* OP_SETATTR = 34 */
394 	{rfs4_op_setattr, nullfree, 0},
395 
396 	/* OP_SETCLIENTID = 35 */
397 	{rfs4_op_setclientid, nullfree, 0},
398 
399 	/* OP_SETCLIENTID_CONFIRM = 36 */
400 	{rfs4_op_setclientid_confirm, nullfree, 0},
401 
402 	/* OP_VERIFY = 37 */
403 	{rfs4_op_verify, nullfree, RPC_IDEMPOTENT},
404 
405 	/* OP_WRITE = 38 */
406 	{rfs4_op_write, nullfree, 0},
407 
408 	/* OP_RELEASE_LOCKOWNER = 39 */
409 	{rfs4_op_release_lockowner, nullfree, 0},
410 };
411 
412 static uint_t rfsv4disp_cnt = sizeof (rfsv4disptab) / sizeof (rfsv4disptab[0]);
413 
414 #define	OP_ILLEGAL_IDX (rfsv4disp_cnt)
415 
416 #ifdef DEBUG
417 
418 int		rfs4_fillone_debug = 0;
419 int		rfs4_shrlock_debug = 0;
420 int		rfs4_no_stub_access = 1;
421 int		rfs4_rddir_debug = 0;
422 
423 static char    *rfs4_op_string[] = {
424 	"rfs4_op_null",
425 	"rfs4_op_1 unused",
426 	"rfs4_op_2 unused",
427 	"rfs4_op_access",
428 	"rfs4_op_close",
429 	"rfs4_op_commit",
430 	"rfs4_op_create",
431 	"rfs4_op_delegpurge",
432 	"rfs4_op_delegreturn",
433 	"rfs4_op_getattr",
434 	"rfs4_op_getfh",
435 	"rfs4_op_link",
436 	"rfs4_op_lock",
437 	"rfs4_op_lockt",
438 	"rfs4_op_locku",
439 	"rfs4_op_lookup",
440 	"rfs4_op_lookupp",
441 	"rfs4_op_nverify",
442 	"rfs4_op_open",
443 	"rfs4_op_openattr",
444 	"rfs4_op_open_confirm",
445 	"rfs4_op_open_downgrade",
446 	"rfs4_op_putfh",
447 	"rfs4_op_putpubfh",
448 	"rfs4_op_putrootfh",
449 	"rfs4_op_read",
450 	"rfs4_op_readdir",
451 	"rfs4_op_readlink",
452 	"rfs4_op_remove",
453 	"rfs4_op_rename",
454 	"rfs4_op_renew",
455 	"rfs4_op_restorefh",
456 	"rfs4_op_savefh",
457 	"rfs4_op_secinfo",
458 	"rfs4_op_setattr",
459 	"rfs4_op_setclientid",
460 	"rfs4_op_setclient_confirm",
461 	"rfs4_op_verify",
462 	"rfs4_op_write",
463 	"rfs4_op_release_lockowner",
464 	"rfs4_op_illegal"
465 };
466 #endif
467 
468 void	rfs4_ss_chkclid(rfs4_client_t *);
469 
470 extern size_t   strlcpy(char *dst, const char *src, size_t dstsize);
471 
472 #ifdef	nextdp
473 #undef nextdp
474 #endif
475 #define	nextdp(dp)	((struct dirent64 *)((char *)(dp) + (dp)->d_reclen))
476 
477 static const fs_operation_def_t nfs4_rd_deleg_tmpl[] = {
478 	VOPNAME_OPEN,		{ .femop_open = deleg_rd_open },
479 	VOPNAME_WRITE,		{ .femop_write = deleg_rd_write },
480 	VOPNAME_SETATTR,	{ .femop_setattr = deleg_rd_setattr },
481 	VOPNAME_RWLOCK,		{ .femop_rwlock = deleg_rd_rwlock },
482 	VOPNAME_SPACE,		{ .femop_space = deleg_rd_space },
483 	VOPNAME_SETSECATTR,	{ .femop_setsecattr = deleg_rd_setsecattr },
484 	VOPNAME_VNEVENT,	{ .femop_vnevent = deleg_rd_vnevent },
485 	NULL,			NULL
486 };
487 static const fs_operation_def_t nfs4_wr_deleg_tmpl[] = {
488 	VOPNAME_OPEN,		{ .femop_open = deleg_wr_open },
489 	VOPNAME_READ,		{ .femop_read = deleg_wr_read },
490 	VOPNAME_WRITE,		{ .femop_write = deleg_wr_write },
491 	VOPNAME_SETATTR,	{ .femop_setattr = deleg_wr_setattr },
492 	VOPNAME_RWLOCK,		{ .femop_rwlock = deleg_wr_rwlock },
493 	VOPNAME_SPACE,		{ .femop_space = deleg_wr_space },
494 	VOPNAME_SETSECATTR,	{ .femop_setsecattr = deleg_wr_setsecattr },
495 	VOPNAME_VNEVENT,	{ .femop_vnevent = deleg_wr_vnevent },
496 	NULL,			NULL
497 };
498 
499 int
500 rfs4_srvrinit(void)
501 {
502 	timespec32_t verf;
503 	int error;
504 	extern void rfs4_attr_init();
505 	extern krwlock_t rfs4_deleg_policy_lock;
506 
507 	/*
508 	 * The following algorithm attempts to find a unique verifier
509 	 * to be used as the write verifier returned from the server
510 	 * to the client.  It is important that this verifier change
511 	 * whenever the server reboots.  Of secondary importance, it
512 	 * is important for the verifier to be unique between two
513 	 * different servers.
514 	 *
515 	 * Thus, an attempt is made to use the system hostid and the
516 	 * current time in seconds when the nfssrv kernel module is
517 	 * loaded.  It is assumed that an NFS server will not be able
518 	 * to boot and then to reboot in less than a second.  If the
519 	 * hostid has not been set, then the current high resolution
520 	 * time is used.  This will ensure different verifiers each
521 	 * time the server reboots and minimize the chances that two
522 	 * different servers will have the same verifier.
523 	 * XXX - this is broken on LP64 kernels.
524 	 */
525 	verf.tv_sec = (time_t)zone_get_hostid(NULL);
526 	if (verf.tv_sec != 0) {
527 		verf.tv_nsec = gethrestime_sec();
528 	} else {
529 		timespec_t tverf;
530 
531 		gethrestime(&tverf);
532 		verf.tv_sec = (time_t)tverf.tv_sec;
533 		verf.tv_nsec = tverf.tv_nsec;
534 	}
535 
536 	Write4verf = *(uint64_t *)&verf;
537 
538 	rfs4_attr_init();
539 	mutex_init(&rfs4_deleg_lock, NULL, MUTEX_DEFAULT, NULL);
540 
541 	/* Used to manage create/destroy of server state */
542 	mutex_init(&rfs4_state_lock, NULL, MUTEX_DEFAULT, NULL);
543 
544 	/* Used to manage access to server instance linked list */
545 	mutex_init(&rfs4_servinst_lock, NULL, MUTEX_DEFAULT, NULL);
546 
547 	/* Used to manage access to rfs4_deleg_policy */
548 	rw_init(&rfs4_deleg_policy_lock, NULL, RW_DEFAULT, NULL);
549 
550 	error = fem_create("deleg_rdops", nfs4_rd_deleg_tmpl, &deleg_rdops);
551 	if (error != 0) {
552 		rfs4_disable_delegation();
553 	} else {
554 		error = fem_create("deleg_wrops", nfs4_wr_deleg_tmpl,
555 		    &deleg_wrops);
556 		if (error != 0) {
557 			rfs4_disable_delegation();
558 			fem_free(deleg_rdops);
559 		}
560 	}
561 
562 	nfs4_srv_caller_id = fs_new_caller_id();
563 
564 	lockt_sysid = lm_alloc_sysidt();
565 
566 	vsd_create(&nfs4_srv_vkey, NULL);
567 
568 	return (0);
569 }
570 
571 void
572 rfs4_srvrfini(void)
573 {
574 	extern krwlock_t rfs4_deleg_policy_lock;
575 
576 	if (lockt_sysid != LM_NOSYSID) {
577 		lm_free_sysidt(lockt_sysid);
578 		lockt_sysid = LM_NOSYSID;
579 	}
580 
581 	mutex_destroy(&rfs4_deleg_lock);
582 	mutex_destroy(&rfs4_state_lock);
583 	rw_destroy(&rfs4_deleg_policy_lock);
584 
585 	fem_free(deleg_rdops);
586 	fem_free(deleg_wrops);
587 }
588 
589 void
590 rfs4_init_compound_state(struct compound_state *cs)
591 {
592 	bzero(cs, sizeof (*cs));
593 	cs->cont = TRUE;
594 	cs->access = CS_ACCESS_DENIED;
595 	cs->deleg = FALSE;
596 	cs->mandlock = FALSE;
597 	cs->fh.nfs_fh4_val = cs->fhbuf;
598 }
599 
600 void
601 rfs4_grace_start(rfs4_servinst_t *sip)
602 {
603 	rw_enter(&sip->rwlock, RW_WRITER);
604 	sip->start_time = (time_t)TICK_TO_SEC(lbolt);
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 (((time_t)TICK_TO_SEC(lbolt)) < 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 	struct sockaddr *ca;
1134 	char *name = NULL;
1135 
1136 	DTRACE_NFSV4_2(op__secinfo__start, struct compound_state *, cs,
1137 	    SECINFO4args *, args);
1138 
1139 	/*
1140 	 * Current file handle (cfh) should have been set before getting
1141 	 * into this function. If not, return error.
1142 	 */
1143 	if (cs->vp == NULL) {
1144 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1145 		goto out;
1146 	}
1147 
1148 	if (cs->vp->v_type != VDIR) {
1149 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
1150 		goto out;
1151 	}
1152 
1153 	/*
1154 	 * Verify the component name. If failed, error out, but
1155 	 * do not error out if the component name is a "..".
1156 	 * SECINFO will return its parents secinfo data for SECINFO "..".
1157 	 */
1158 	if (!utf8_dir_verify(utfnm)) {
1159 		if (utfnm->utf8string_len != 2 ||
1160 		    utfnm->utf8string_val[0] != '.' ||
1161 		    utfnm->utf8string_val[1] != '.') {
1162 			*cs->statusp = resp->status = NFS4ERR_INVAL;
1163 			goto out;
1164 		}
1165 	}
1166 
1167 	nm = utf8_to_str(utfnm, &len, NULL);
1168 	if (nm == NULL) {
1169 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1170 		goto out;
1171 	}
1172 
1173 	if (len > MAXNAMELEN) {
1174 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1175 		kmem_free(nm, len);
1176 		goto out;
1177 	}
1178 	/* If necessary, convert to UTF-8 for illbehaved clients */
1179 
1180 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
1181 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
1182 	    MAXPATHLEN  + 1);
1183 
1184 	if (name == NULL) {
1185 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1186 		kmem_free(nm, len);
1187 		goto out;
1188 	}
1189 
1190 
1191 	*cs->statusp = resp->status = do_rfs4_op_secinfo(cs, name, resp);
1192 
1193 	if (name != nm)
1194 		kmem_free(name, MAXPATHLEN + 1);
1195 	kmem_free(nm, len);
1196 
1197 out:
1198 	DTRACE_NFSV4_2(op__secinfo__done, struct compound_state *, cs,
1199 	    SECINFO4res *, resp);
1200 }
1201 
1202 /*
1203  * Free SECINFO result.
1204  */
1205 /* ARGSUSED */
1206 static void
1207 rfs4_op_secinfo_free(nfs_resop4 *resop)
1208 {
1209 	SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo;
1210 	int count, i;
1211 	secinfo4 *resok_val;
1212 
1213 	/* If this is not an Ok result, nothing to free. */
1214 	if (resp->status != NFS4_OK) {
1215 		return;
1216 	}
1217 
1218 	count = resp->SECINFO4resok_len;
1219 	resok_val = resp->SECINFO4resok_val;
1220 
1221 	for (i = 0; i < count; i++) {
1222 		if (resok_val[i].flavor == RPCSEC_GSS) {
1223 			rpcsec_gss_info *info;
1224 
1225 			info = &resok_val[i].flavor_info;
1226 			kmem_free(info->oid.sec_oid4_val,
1227 			    info->oid.sec_oid4_len);
1228 		}
1229 	}
1230 	kmem_free(resok_val, count * sizeof (secinfo4));
1231 	resp->SECINFO4resok_len = 0;
1232 	resp->SECINFO4resok_val = NULL;
1233 }
1234 
1235 /* ARGSUSED */
1236 static void
1237 rfs4_op_access(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1238 	struct compound_state *cs)
1239 {
1240 	ACCESS4args *args = &argop->nfs_argop4_u.opaccess;
1241 	ACCESS4res *resp = &resop->nfs_resop4_u.opaccess;
1242 	int error;
1243 	vnode_t *vp;
1244 	struct vattr va;
1245 	int checkwriteperm;
1246 	cred_t *cr = cs->cr;
1247 	bslabel_t *clabel, *slabel;
1248 	ts_label_t *tslabel;
1249 	boolean_t admin_low_client;
1250 
1251 	DTRACE_NFSV4_2(op__access__start, struct compound_state *, cs,
1252 	    ACCESS4args *, args);
1253 
1254 #if 0	/* XXX allow access even if !cs->access. Eventually only pseudo fs */
1255 	if (cs->access == CS_ACCESS_DENIED) {
1256 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1257 		goto out;
1258 	}
1259 #endif
1260 	if (cs->vp == NULL) {
1261 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1262 		goto out;
1263 	}
1264 
1265 	ASSERT(cr != NULL);
1266 
1267 	vp = cs->vp;
1268 
1269 	/*
1270 	 * If the file system is exported read only, it is not appropriate
1271 	 * to check write permissions for regular files and directories.
1272 	 * Special files are interpreted by the client, so the underlying
1273 	 * permissions are sent back to the client for interpretation.
1274 	 */
1275 	if (rdonly4(cs->exi, cs->vp, req) &&
1276 	    (vp->v_type == VREG || vp->v_type == VDIR))
1277 		checkwriteperm = 0;
1278 	else
1279 		checkwriteperm = 1;
1280 
1281 	/*
1282 	 * XXX
1283 	 * We need the mode so that we can correctly determine access
1284 	 * permissions relative to a mandatory lock file.  Access to
1285 	 * mandatory lock files is denied on the server, so it might
1286 	 * as well be reflected to the server during the open.
1287 	 */
1288 	va.va_mask = AT_MODE;
1289 	error = VOP_GETATTR(vp, &va, 0, cr, NULL);
1290 	if (error) {
1291 		*cs->statusp = resp->status = puterrno4(error);
1292 		goto out;
1293 	}
1294 	resp->access = 0;
1295 	resp->supported = 0;
1296 
1297 	if (is_system_labeled()) {
1298 		ASSERT(req->rq_label != NULL);
1299 		clabel = req->rq_label;
1300 		DTRACE_PROBE2(tx__rfs4__log__info__opaccess__clabel, char *,
1301 		    "got client label from request(1)",
1302 		    struct svc_req *, req);
1303 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
1304 			if ((tslabel = nfs_getflabel(vp, cs->exi)) == NULL) {
1305 				*cs->statusp = resp->status = puterrno4(EACCES);
1306 				goto out;
1307 			}
1308 			slabel = label2bslabel(tslabel);
1309 			DTRACE_PROBE3(tx__rfs4__log__info__opaccess__slabel,
1310 			    char *, "got server label(1) for vp(2)",
1311 			    bslabel_t *, slabel, vnode_t *, vp);
1312 
1313 			admin_low_client = B_FALSE;
1314 		} else
1315 			admin_low_client = B_TRUE;
1316 	}
1317 
1318 	if (args->access & ACCESS4_READ) {
1319 		error = VOP_ACCESS(vp, VREAD, 0, cr, NULL);
1320 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1321 		    (!is_system_labeled() || admin_low_client ||
1322 		    bldominates(clabel, slabel)))
1323 			resp->access |= ACCESS4_READ;
1324 		resp->supported |= ACCESS4_READ;
1325 	}
1326 	if ((args->access & ACCESS4_LOOKUP) && vp->v_type == VDIR) {
1327 		error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL);
1328 		if (!error && (!is_system_labeled() || admin_low_client ||
1329 		    bldominates(clabel, slabel)))
1330 			resp->access |= ACCESS4_LOOKUP;
1331 		resp->supported |= ACCESS4_LOOKUP;
1332 	}
1333 	if (checkwriteperm &&
1334 	    (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND))) {
1335 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
1336 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1337 		    (!is_system_labeled() || admin_low_client ||
1338 		    blequal(clabel, slabel)))
1339 			resp->access |=
1340 			    (args->access & (ACCESS4_MODIFY | ACCESS4_EXTEND));
1341 		resp->supported |= (ACCESS4_MODIFY | ACCESS4_EXTEND);
1342 	}
1343 
1344 	if (checkwriteperm &&
1345 	    (args->access & ACCESS4_DELETE) && vp->v_type == VDIR) {
1346 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
1347 		if (!error && (!is_system_labeled() || admin_low_client ||
1348 		    blequal(clabel, slabel)))
1349 			resp->access |= ACCESS4_DELETE;
1350 		resp->supported |= ACCESS4_DELETE;
1351 	}
1352 	if (args->access & ACCESS4_EXECUTE && vp->v_type != VDIR) {
1353 		error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL);
1354 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1355 		    (!is_system_labeled() || admin_low_client ||
1356 		    bldominates(clabel, slabel)))
1357 			resp->access |= ACCESS4_EXECUTE;
1358 		resp->supported |= ACCESS4_EXECUTE;
1359 	}
1360 
1361 	if (is_system_labeled() && !admin_low_client)
1362 		label_rele(tslabel);
1363 
1364 	*cs->statusp = resp->status = NFS4_OK;
1365 out:
1366 	DTRACE_NFSV4_2(op__access__done, struct compound_state *, cs,
1367 	    ACCESS4res *, resp);
1368 }
1369 
1370 /* ARGSUSED */
1371 static void
1372 rfs4_op_commit(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1373 	struct compound_state *cs)
1374 {
1375 	COMMIT4args *args = &argop->nfs_argop4_u.opcommit;
1376 	COMMIT4res *resp = &resop->nfs_resop4_u.opcommit;
1377 	int error;
1378 	vnode_t *vp = cs->vp;
1379 	cred_t *cr = cs->cr;
1380 	vattr_t va;
1381 
1382 	DTRACE_NFSV4_2(op__commit__start, struct compound_state *, cs,
1383 	    COMMIT4args *, args);
1384 
1385 	if (vp == NULL) {
1386 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1387 		goto out;
1388 	}
1389 	if (cs->access == CS_ACCESS_DENIED) {
1390 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1391 		goto out;
1392 	}
1393 
1394 	if (args->offset + args->count < args->offset) {
1395 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1396 		goto out;
1397 	}
1398 
1399 	va.va_mask = AT_UID;
1400 	error = VOP_GETATTR(vp, &va, 0, cr, NULL);
1401 
1402 	/*
1403 	 * If we can't get the attributes, then we can't do the
1404 	 * right access checking.  So, we'll fail the request.
1405 	 */
1406 	if (error) {
1407 		*cs->statusp = resp->status = puterrno4(error);
1408 		goto out;
1409 	}
1410 	if (rdonly4(cs->exi, cs->vp, req)) {
1411 		*cs->statusp = resp->status = NFS4ERR_ROFS;
1412 		goto out;
1413 	}
1414 
1415 	if (vp->v_type != VREG) {
1416 		if (vp->v_type == VDIR)
1417 			resp->status = NFS4ERR_ISDIR;
1418 		else
1419 			resp->status = NFS4ERR_INVAL;
1420 		*cs->statusp = resp->status;
1421 		goto out;
1422 	}
1423 
1424 	if (crgetuid(cr) != va.va_uid &&
1425 	    (error = VOP_ACCESS(vp, VWRITE, 0, cs->cr, NULL))) {
1426 		*cs->statusp = resp->status = puterrno4(error);
1427 		goto out;
1428 	}
1429 
1430 	error = VOP_PUTPAGE(vp, args->offset, args->count, 0, cr, NULL);
1431 	if (!error)
1432 		error = VOP_FSYNC(vp, FNODSYNC, cr, NULL);
1433 
1434 	if (error) {
1435 		*cs->statusp = resp->status = puterrno4(error);
1436 		goto out;
1437 	}
1438 
1439 	*cs->statusp = resp->status = NFS4_OK;
1440 	resp->writeverf = Write4verf;
1441 out:
1442 	DTRACE_NFSV4_2(op__commit__done, struct compound_state *, cs,
1443 	    COMMIT4res *, resp);
1444 }
1445 
1446 /*
1447  * rfs4_op_mknod is called from rfs4_op_create after all initial verification
1448  * was completed. It does the nfsv4 create for special files.
1449  */
1450 /* ARGSUSED */
1451 static vnode_t *
1452 do_rfs4_op_mknod(CREATE4args *args, CREATE4res *resp, struct svc_req *req,
1453 	struct compound_state *cs, vattr_t *vap, char *nm)
1454 {
1455 	int error;
1456 	cred_t *cr = cs->cr;
1457 	vnode_t *dvp = cs->vp;
1458 	vnode_t *vp = NULL;
1459 	int mode;
1460 	enum vcexcl excl;
1461 
1462 	switch (args->type) {
1463 	case NF4CHR:
1464 	case NF4BLK:
1465 		if (secpolicy_sys_devices(cr) != 0) {
1466 			*cs->statusp = resp->status = NFS4ERR_PERM;
1467 			return (NULL);
1468 		}
1469 		if (args->type == NF4CHR)
1470 			vap->va_type = VCHR;
1471 		else
1472 			vap->va_type = VBLK;
1473 		vap->va_rdev = makedevice(args->ftype4_u.devdata.specdata1,
1474 		    args->ftype4_u.devdata.specdata2);
1475 		vap->va_mask |= AT_RDEV;
1476 		break;
1477 	case NF4SOCK:
1478 		vap->va_type = VSOCK;
1479 		break;
1480 	case NF4FIFO:
1481 		vap->va_type = VFIFO;
1482 		break;
1483 	default:
1484 		*cs->statusp = resp->status = NFS4ERR_BADTYPE;
1485 		return (NULL);
1486 	}
1487 
1488 	/*
1489 	 * Must specify the mode.
1490 	 */
1491 	if (!(vap->va_mask & AT_MODE)) {
1492 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1493 		return (NULL);
1494 	}
1495 
1496 	excl = EXCL;
1497 
1498 	mode = 0;
1499 
1500 	error = VOP_CREATE(dvp, nm, vap, excl, mode, &vp, cr, 0, NULL, NULL);
1501 	if (error) {
1502 		*cs->statusp = resp->status = puterrno4(error);
1503 		return (NULL);
1504 	}
1505 	return (vp);
1506 }
1507 
1508 /*
1509  * nfsv4 create is used to create non-regular files. For regular files,
1510  * use nfsv4 open.
1511  */
1512 /* ARGSUSED */
1513 static void
1514 rfs4_op_create(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1515 	struct compound_state *cs)
1516 {
1517 	CREATE4args *args = &argop->nfs_argop4_u.opcreate;
1518 	CREATE4res *resp = &resop->nfs_resop4_u.opcreate;
1519 	int error;
1520 	struct vattr bva, iva, iva2, ava, *vap;
1521 	cred_t *cr = cs->cr;
1522 	vnode_t *dvp = cs->vp;
1523 	vnode_t *vp = NULL;
1524 	vnode_t *realvp;
1525 	char *nm, *lnm;
1526 	uint_t len, llen;
1527 	int syncval = 0;
1528 	struct nfs4_svgetit_arg sarg;
1529 	struct nfs4_ntov_table ntov;
1530 	struct statvfs64 sb;
1531 	nfsstat4 status;
1532 	struct sockaddr *ca;
1533 	char *name = NULL;
1534 	char *lname = NULL;
1535 
1536 	DTRACE_NFSV4_2(op__create__start, struct compound_state *, cs,
1537 	    CREATE4args *, args);
1538 
1539 	resp->attrset = 0;
1540 
1541 	if (dvp == NULL) {
1542 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1543 		goto out;
1544 	}
1545 
1546 	/*
1547 	 * If there is an unshared filesystem mounted on this vnode,
1548 	 * do not allow to create an object in this directory.
1549 	 */
1550 	if (vn_ismntpt(dvp)) {
1551 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1552 		goto out;
1553 	}
1554 
1555 	/* Verify that type is correct */
1556 	switch (args->type) {
1557 	case NF4LNK:
1558 	case NF4BLK:
1559 	case NF4CHR:
1560 	case NF4SOCK:
1561 	case NF4FIFO:
1562 	case NF4DIR:
1563 		break;
1564 	default:
1565 		*cs->statusp = resp->status = NFS4ERR_BADTYPE;
1566 		goto out;
1567 	};
1568 
1569 	if (cs->access == CS_ACCESS_DENIED) {
1570 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1571 		goto out;
1572 	}
1573 	if (dvp->v_type != VDIR) {
1574 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
1575 		goto out;
1576 	}
1577 	if (!utf8_dir_verify(&args->objname)) {
1578 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1579 		goto out;
1580 	}
1581 
1582 	if (rdonly4(cs->exi, cs->vp, req)) {
1583 		*cs->statusp = resp->status = NFS4ERR_ROFS;
1584 		goto out;
1585 	}
1586 
1587 	/*
1588 	 * Name of newly created object
1589 	 */
1590 	nm = utf8_to_fn(&args->objname, &len, NULL);
1591 	if (nm == NULL) {
1592 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1593 		goto out;
1594 	}
1595 
1596 	if (len > MAXNAMELEN) {
1597 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1598 		kmem_free(nm, len);
1599 		goto out;
1600 	}
1601 
1602 	/* If necessary, convert to UTF-8 for poorly behaved clients */
1603 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
1604 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
1605 	    MAXPATHLEN  + 1);
1606 
1607 	if (name == NULL) {
1608 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1609 		kmem_free(nm, len);
1610 		goto out;
1611 	}
1612 
1613 	resp->attrset = 0;
1614 
1615 	sarg.sbp = &sb;
1616 	nfs4_ntov_table_init(&ntov);
1617 
1618 	status = do_rfs4_set_attrs(&resp->attrset,
1619 	    &args->createattrs, cs, &sarg, &ntov, NFS4ATTR_SETIT);
1620 
1621 	if (sarg.vap->va_mask == 0 && status == NFS4_OK)
1622 		status = NFS4ERR_INVAL;
1623 
1624 	if (status != NFS4_OK) {
1625 		*cs->statusp = resp->status = status;
1626 		kmem_free(nm, len);
1627 		nfs4_ntov_table_free(&ntov, &sarg);
1628 		resp->attrset = 0;
1629 		goto out;
1630 	}
1631 
1632 	/* Get "before" change value */
1633 	bva.va_mask = AT_CTIME|AT_SEQ;
1634 	error = VOP_GETATTR(dvp, &bva, 0, cr, NULL);
1635 	if (error) {
1636 		*cs->statusp = resp->status = puterrno4(error);
1637 		kmem_free(nm, len);
1638 		nfs4_ntov_table_free(&ntov, &sarg);
1639 		resp->attrset = 0;
1640 		goto out;
1641 	}
1642 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bva.va_ctime)
1643 
1644 	vap = sarg.vap;
1645 
1646 	/*
1647 	 * Set default initial values for attributes when not specified
1648 	 * in createattrs.
1649 	 */
1650 	if ((vap->va_mask & AT_UID) == 0) {
1651 		vap->va_uid = crgetuid(cr);
1652 		vap->va_mask |= AT_UID;
1653 	}
1654 	if ((vap->va_mask & AT_GID) == 0) {
1655 		vap->va_gid = crgetgid(cr);
1656 		vap->va_mask |= AT_GID;
1657 	}
1658 
1659 	vap->va_mask |= AT_TYPE;
1660 	switch (args->type) {
1661 	case NF4DIR:
1662 		vap->va_type = VDIR;
1663 		if ((vap->va_mask & AT_MODE) == 0) {
1664 			vap->va_mode = 0700;	/* default: owner rwx only */
1665 			vap->va_mask |= AT_MODE;
1666 		}
1667 		error = VOP_MKDIR(dvp, nm, vap, &vp, cr, NULL, 0, NULL);
1668 		if (error)
1669 			break;
1670 
1671 		/*
1672 		 * Get the initial "after" sequence number, if it fails,
1673 		 * set to zero
1674 		 */
1675 		iva.va_mask = AT_SEQ;
1676 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL))
1677 			iva.va_seq = 0;
1678 		break;
1679 	case NF4LNK:
1680 		vap->va_type = VLNK;
1681 		if ((vap->va_mask & AT_MODE) == 0) {
1682 			vap->va_mode = 0700;	/* default: owner rwx only */
1683 			vap->va_mask |= AT_MODE;
1684 		}
1685 
1686 		/*
1687 		 * symlink names must be treated as data
1688 		 */
1689 		lnm = utf8_to_str(&args->ftype4_u.linkdata, &llen, NULL);
1690 
1691 		if (lnm == NULL) {
1692 			*cs->statusp = resp->status = NFS4ERR_INVAL;
1693 			if (name != nm)
1694 				kmem_free(name, MAXPATHLEN + 1);
1695 			kmem_free(nm, len);
1696 			nfs4_ntov_table_free(&ntov, &sarg);
1697 			resp->attrset = 0;
1698 			goto out;
1699 		}
1700 
1701 		if (llen > MAXPATHLEN) {
1702 			*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1703 			if (name != nm)
1704 				kmem_free(name, MAXPATHLEN + 1);
1705 			kmem_free(nm, len);
1706 			kmem_free(lnm, llen);
1707 			nfs4_ntov_table_free(&ntov, &sarg);
1708 			resp->attrset = 0;
1709 			goto out;
1710 		}
1711 
1712 		lname = nfscmd_convname(ca, cs->exi, lnm,
1713 		    NFSCMD_CONV_INBOUND, MAXPATHLEN  + 1);
1714 
1715 		if (lname == NULL) {
1716 			*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
1717 			if (name != nm)
1718 				kmem_free(name, MAXPATHLEN + 1);
1719 			kmem_free(nm, len);
1720 			kmem_free(lnm, llen);
1721 			nfs4_ntov_table_free(&ntov, &sarg);
1722 			resp->attrset = 0;
1723 			goto out;
1724 		}
1725 
1726 		error = VOP_SYMLINK(dvp, nm, vap, lnm, cr, NULL, 0);
1727 		if (lname != lnm)
1728 			kmem_free(lname, MAXPATHLEN + 1);
1729 		if (lnm != NULL)
1730 			kmem_free(lnm, llen);
1731 		if (error)
1732 			break;
1733 
1734 		/*
1735 		 * Get the initial "after" sequence number, if it fails,
1736 		 * set to zero
1737 		 */
1738 		iva.va_mask = AT_SEQ;
1739 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL))
1740 			iva.va_seq = 0;
1741 
1742 		error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr,
1743 		    NULL, NULL, NULL);
1744 		if (error)
1745 			break;
1746 
1747 		/*
1748 		 * va_seq is not safe over VOP calls, check it again
1749 		 * if it has changed zero out iva to force atomic = FALSE.
1750 		 */
1751 		iva2.va_mask = AT_SEQ;
1752 		if (VOP_GETATTR(dvp, &iva2, 0, cs->cr, NULL) ||
1753 		    iva2.va_seq != iva.va_seq)
1754 			iva.va_seq = 0;
1755 		break;
1756 	default:
1757 		/*
1758 		 * probably a special file.
1759 		 */
1760 		if ((vap->va_mask & AT_MODE) == 0) {
1761 			vap->va_mode = 0600;	/* default: owner rw only */
1762 			vap->va_mask |= AT_MODE;
1763 		}
1764 		syncval = FNODSYNC;
1765 		/*
1766 		 * We know this will only generate one VOP call
1767 		 */
1768 		vp = do_rfs4_op_mknod(args, resp, req, cs, vap, nm);
1769 
1770 		if (vp == NULL) {
1771 			if (name != nm)
1772 				kmem_free(name, MAXPATHLEN + 1);
1773 			kmem_free(nm, len);
1774 			nfs4_ntov_table_free(&ntov, &sarg);
1775 			resp->attrset = 0;
1776 			goto out;
1777 		}
1778 
1779 		/*
1780 		 * Get the initial "after" sequence number, if it fails,
1781 		 * set to zero
1782 		 */
1783 		iva.va_mask = AT_SEQ;
1784 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL))
1785 			iva.va_seq = 0;
1786 
1787 		break;
1788 	}
1789 	if (name != nm)
1790 		kmem_free(name, MAXPATHLEN + 1);
1791 	kmem_free(nm, len);
1792 
1793 	if (error) {
1794 		*cs->statusp = resp->status = puterrno4(error);
1795 	}
1796 
1797 	/*
1798 	 * Force modified data and metadata out to stable storage.
1799 	 */
1800 	(void) VOP_FSYNC(dvp, 0, cr, NULL);
1801 
1802 	if (resp->status != NFS4_OK) {
1803 		if (vp != NULL)
1804 			VN_RELE(vp);
1805 		nfs4_ntov_table_free(&ntov, &sarg);
1806 		resp->attrset = 0;
1807 		goto out;
1808 	}
1809 
1810 	/*
1811 	 * Finish setup of cinfo response, "before" value already set.
1812 	 * Get "after" change value, if it fails, simply return the
1813 	 * before value.
1814 	 */
1815 	ava.va_mask = AT_CTIME|AT_SEQ;
1816 	if (VOP_GETATTR(dvp, &ava, 0, cr, NULL)) {
1817 		ava.va_ctime = bva.va_ctime;
1818 		ava.va_seq = 0;
1819 	}
1820 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, ava.va_ctime);
1821 
1822 	/*
1823 	 * True verification that object was created with correct
1824 	 * attrs is impossible.  The attrs could have been changed
1825 	 * immediately after object creation.  If attributes did
1826 	 * not verify, the only recourse for the server is to
1827 	 * destroy the object.  Maybe if some attrs (like gid)
1828 	 * are set incorrectly, the object should be destroyed;
1829 	 * however, seems bad as a default policy.  Do we really
1830 	 * want to destroy an object over one of the times not
1831 	 * verifying correctly?  For these reasons, the server
1832 	 * currently sets bits in attrset for createattrs
1833 	 * that were set; however, no verification is done.
1834 	 *
1835 	 * vmask_to_nmask accounts for vattr bits set on create
1836 	 *	[do_rfs4_set_attrs() only sets resp bits for
1837 	 *	 non-vattr/vfs bits.]
1838 	 * Mask off any bits set by default so as not to return
1839 	 * more attrset bits than were requested in createattrs
1840 	 */
1841 	nfs4_vmask_to_nmask(sarg.vap->va_mask, &resp->attrset);
1842 	resp->attrset &= args->createattrs.attrmask;
1843 	nfs4_ntov_table_free(&ntov, &sarg);
1844 
1845 	error = makefh4(&cs->fh, vp, cs->exi);
1846 	if (error) {
1847 		*cs->statusp = resp->status = puterrno4(error);
1848 	}
1849 
1850 	/*
1851 	 * The cinfo.atomic = TRUE only if we got no errors, we have
1852 	 * non-zero va_seq's, and it has incremented by exactly one
1853 	 * during the creation and it didn't change during the VOP_LOOKUP
1854 	 * or VOP_FSYNC.
1855 	 */
1856 	if (!error && bva.va_seq && iva.va_seq && ava.va_seq &&
1857 	    iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq)
1858 		resp->cinfo.atomic = TRUE;
1859 	else
1860 		resp->cinfo.atomic = FALSE;
1861 
1862 	/*
1863 	 * Force modified metadata out to stable storage.
1864 	 *
1865 	 * if a underlying vp exists, pass it to VOP_FSYNC
1866 	 */
1867 	if (VOP_REALVP(vp, &realvp, NULL) == 0)
1868 		(void) VOP_FSYNC(realvp, syncval, cr, NULL);
1869 	else
1870 		(void) VOP_FSYNC(vp, syncval, cr, NULL);
1871 
1872 	if (resp->status != NFS4_OK) {
1873 		VN_RELE(vp);
1874 		goto out;
1875 	}
1876 	if (cs->vp)
1877 		VN_RELE(cs->vp);
1878 
1879 	cs->vp = vp;
1880 	*cs->statusp = resp->status = NFS4_OK;
1881 out:
1882 	DTRACE_NFSV4_2(op__create__done, struct compound_state *, cs,
1883 	    CREATE4res *, resp);
1884 }
1885 
1886 /*ARGSUSED*/
1887 static void
1888 rfs4_op_delegpurge(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1889     struct compound_state *cs)
1890 {
1891 	DTRACE_NFSV4_2(op__delegpurge__start, struct compound_state *, cs,
1892 	    DELEGPURGE4args *, &argop->nfs_argop4_u.opdelegpurge);
1893 
1894 	rfs4_op_inval(argop, resop, req, cs);
1895 
1896 	DTRACE_NFSV4_2(op__delegpurge__done, struct compound_state *, cs,
1897 	    DELEGPURGE4res *, &resop->nfs_resop4_u.opdelegpurge);
1898 }
1899 
1900 /*ARGSUSED*/
1901 static void
1902 rfs4_op_delegreturn(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1903 	struct compound_state *cs)
1904 {
1905 	DELEGRETURN4args *args = &argop->nfs_argop4_u.opdelegreturn;
1906 	DELEGRETURN4res *resp = &resop->nfs_resop4_u.opdelegreturn;
1907 	rfs4_deleg_state_t *dsp;
1908 	nfsstat4 status;
1909 
1910 	DTRACE_NFSV4_2(op__delegreturn__start, struct compound_state *, cs,
1911 	    DELEGRETURN4args *, args);
1912 
1913 	status = rfs4_get_deleg_state(&args->deleg_stateid, &dsp);
1914 	resp->status = *cs->statusp = status;
1915 	if (status != NFS4_OK)
1916 		goto out;
1917 
1918 	/* Ensure specified filehandle matches */
1919 	if (cs->vp != dsp->finfo->vp) {
1920 		resp->status = *cs->statusp = NFS4ERR_BAD_STATEID;
1921 	} else
1922 		rfs4_return_deleg(dsp, FALSE);
1923 
1924 	rfs4_update_lease(dsp->client);
1925 
1926 	rfs4_deleg_state_rele(dsp);
1927 out:
1928 	DTRACE_NFSV4_2(op__delegreturn__done, struct compound_state *, cs,
1929 	    DELEGRETURN4res *, resp);
1930 }
1931 
1932 /*
1933  * Check to see if a given "flavor" is an explicitly shared flavor.
1934  * The assumption of this routine is the "flavor" is already a valid
1935  * flavor in the secinfo list of "exi".
1936  *
1937  *	e.g.
1938  *		# share -o sec=flavor1 /export
1939  *		# share -o sec=flavor2 /export/home
1940  *
1941  *		flavor2 is not an explicitly shared flavor for /export,
1942  *		however it is in the secinfo list for /export thru the
1943  *		server namespace setup.
1944  */
1945 int
1946 is_exported_sec(int flavor, struct exportinfo *exi)
1947 {
1948 	int	i;
1949 	struct secinfo *sp;
1950 
1951 	sp = exi->exi_export.ex_secinfo;
1952 	for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
1953 		if (flavor == sp[i].s_secinfo.sc_nfsnum ||
1954 		    sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) {
1955 			return (SEC_REF_EXPORTED(&sp[i]));
1956 		}
1957 	}
1958 
1959 	/* Should not reach this point based on the assumption */
1960 	return (0);
1961 }
1962 
1963 /*
1964  * Check if the security flavor used in the request matches what is
1965  * required at the export point or at the root pseudo node (exi_root).
1966  *
1967  * returns 1 if there's a match or if exported with AUTH_NONE; 0 otherwise.
1968  *
1969  */
1970 static int
1971 secinfo_match_or_authnone(struct compound_state *cs)
1972 {
1973 	int	i;
1974 	struct secinfo *sp;
1975 
1976 	/*
1977 	 * Check cs->nfsflavor (from the request) against
1978 	 * the current export data in cs->exi.
1979 	 */
1980 	sp = cs->exi->exi_export.ex_secinfo;
1981 	for (i = 0; i < cs->exi->exi_export.ex_seccnt; i++) {
1982 		if (cs->nfsflavor == sp[i].s_secinfo.sc_nfsnum ||
1983 		    sp[i].s_secinfo.sc_nfsnum == AUTH_NONE)
1984 			return (1);
1985 	}
1986 
1987 	return (0);
1988 }
1989 
1990 /*
1991  * Check the access authority for the client and return the correct error.
1992  */
1993 nfsstat4
1994 call_checkauth4(struct compound_state *cs, struct svc_req *req)
1995 {
1996 	int	authres;
1997 
1998 	/*
1999 	 * First, check if the security flavor used in the request
2000 	 * are among the flavors set in the server namespace.
2001 	 */
2002 	if (!secinfo_match_or_authnone(cs)) {
2003 		*cs->statusp = NFS4ERR_WRONGSEC;
2004 		return (*cs->statusp);
2005 	}
2006 
2007 	authres = checkauth4(cs, req);
2008 
2009 	if (authres > 0) {
2010 		*cs->statusp = NFS4_OK;
2011 		if (! (cs->access & CS_ACCESS_LIMITED))
2012 			cs->access = CS_ACCESS_OK;
2013 	} else if (authres == 0) {
2014 		*cs->statusp = NFS4ERR_ACCESS;
2015 	} else if (authres == -2) {
2016 		*cs->statusp = NFS4ERR_WRONGSEC;
2017 	} else {
2018 		*cs->statusp = NFS4ERR_DELAY;
2019 	}
2020 	return (*cs->statusp);
2021 }
2022 
2023 /*
2024  * bitmap4_to_attrmask is called by getattr and readdir.
2025  * It sets up the vattr mask and determines whether vfsstat call is needed
2026  * based on the input bitmap.
2027  * Returns nfsv4 status.
2028  */
2029 static nfsstat4
2030 bitmap4_to_attrmask(bitmap4 breq, struct nfs4_svgetit_arg *sargp)
2031 {
2032 	int i;
2033 	uint_t	va_mask;
2034 	struct statvfs64 *sbp = sargp->sbp;
2035 
2036 	sargp->sbp = NULL;
2037 	sargp->flag = 0;
2038 	sargp->rdattr_error = NFS4_OK;
2039 	sargp->mntdfid_set = FALSE;
2040 	if (sargp->cs->vp)
2041 		sargp->xattr = get_fh4_flag(&sargp->cs->fh,
2042 		    FH4_ATTRDIR | FH4_NAMEDATTR);
2043 	else
2044 		sargp->xattr = 0;
2045 
2046 	/*
2047 	 * Set rdattr_error_req to true if return error per
2048 	 * failed entry rather than fail the readdir.
2049 	 */
2050 	if (breq & FATTR4_RDATTR_ERROR_MASK)
2051 		sargp->rdattr_error_req = 1;
2052 	else
2053 		sargp->rdattr_error_req = 0;
2054 
2055 	/*
2056 	 * generate the va_mask
2057 	 * Handle the easy cases first
2058 	 */
2059 	switch (breq) {
2060 	case NFS4_NTOV_ATTR_MASK:
2061 		sargp->vap->va_mask = NFS4_NTOV_ATTR_AT_MASK;
2062 		return (NFS4_OK);
2063 
2064 	case NFS4_FS_ATTR_MASK:
2065 		sargp->vap->va_mask = NFS4_FS_ATTR_AT_MASK;
2066 		sargp->sbp = sbp;
2067 		return (NFS4_OK);
2068 
2069 	case NFS4_NTOV_ATTR_CACHE_MASK:
2070 		sargp->vap->va_mask = NFS4_NTOV_ATTR_CACHE_AT_MASK;
2071 		return (NFS4_OK);
2072 
2073 	case FATTR4_LEASE_TIME_MASK:
2074 		sargp->vap->va_mask = 0;
2075 		return (NFS4_OK);
2076 
2077 	default:
2078 		va_mask = 0;
2079 		for (i = 0; i < nfs4_ntov_map_size; i++) {
2080 			if ((breq & nfs4_ntov_map[i].fbit) &&
2081 			    nfs4_ntov_map[i].vbit)
2082 				va_mask |= nfs4_ntov_map[i].vbit;
2083 		}
2084 
2085 		/*
2086 		 * Check is vfsstat is needed
2087 		 */
2088 		if (breq & NFS4_FS_ATTR_MASK)
2089 			sargp->sbp = sbp;
2090 
2091 		sargp->vap->va_mask = va_mask;
2092 		return (NFS4_OK);
2093 	}
2094 	/* NOTREACHED */
2095 }
2096 
2097 /*
2098  * bitmap4_get_sysattrs is called by getattr and readdir.
2099  * It calls both VOP_GETATTR and VFS_STATVFS calls to get the attrs.
2100  * Returns nfsv4 status.
2101  */
2102 static nfsstat4
2103 bitmap4_get_sysattrs(struct nfs4_svgetit_arg *sargp)
2104 {
2105 	int error;
2106 	struct compound_state *cs = sargp->cs;
2107 	vnode_t *vp = cs->vp;
2108 
2109 	if (sargp->sbp != NULL) {
2110 		if (error = VFS_STATVFS(vp->v_vfsp, sargp->sbp)) {
2111 			sargp->sbp = NULL;	/* to identify error */
2112 			return (puterrno4(error));
2113 		}
2114 	}
2115 
2116 	return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr));
2117 }
2118 
2119 static void
2120 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp)
2121 {
2122 	ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size,
2123 	    KM_SLEEP);
2124 	ntovp->attrcnt = 0;
2125 	ntovp->vfsstat = FALSE;
2126 }
2127 
2128 static void
2129 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp,
2130 	struct nfs4_svgetit_arg *sargp)
2131 {
2132 	int i;
2133 	union nfs4_attr_u *na;
2134 	uint8_t *amap;
2135 
2136 	/*
2137 	 * XXX Should do the same checks for whether the bit is set
2138 	 */
2139 	for (i = 0, na = ntovp->na, amap = ntovp->amap;
2140 	    i < ntovp->attrcnt; i++, na++, amap++) {
2141 		(void) (*nfs4_ntov_map[*amap].sv_getit)(
2142 		    NFS4ATTR_FREEIT, sargp, na);
2143 	}
2144 	if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) {
2145 		/*
2146 		 * xdr_free for getattr will be done later
2147 		 */
2148 		for (i = 0, na = ntovp->na, amap = ntovp->amap;
2149 		    i < ntovp->attrcnt; i++, na++, amap++) {
2150 			xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na);
2151 		}
2152 	}
2153 	kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size);
2154 }
2155 
2156 /*
2157  * do_rfs4_op_getattr gets the system attrs and converts into fattr4.
2158  */
2159 static nfsstat4
2160 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp,
2161 	struct nfs4_svgetit_arg *sargp)
2162 {
2163 	int error = 0;
2164 	int i, k;
2165 	struct nfs4_ntov_table ntov;
2166 	XDR xdr;
2167 	ulong_t xdr_size;
2168 	char *xdr_attrs;
2169 	nfsstat4 status = NFS4_OK;
2170 	nfsstat4 prev_rdattr_error = sargp->rdattr_error;
2171 	union nfs4_attr_u *na;
2172 	uint8_t *amap;
2173 
2174 	sargp->op = NFS4ATTR_GETIT;
2175 	sargp->flag = 0;
2176 
2177 	fattrp->attrmask = 0;
2178 	/* if no bits requested, then return empty fattr4 */
2179 	if (breq == 0) {
2180 		fattrp->attrlist4_len = 0;
2181 		fattrp->attrlist4 = NULL;
2182 		return (NFS4_OK);
2183 	}
2184 
2185 	/*
2186 	 * return NFS4ERR_INVAL when client requests write-only attrs
2187 	 */
2188 	if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK))
2189 		return (NFS4ERR_INVAL);
2190 
2191 	nfs4_ntov_table_init(&ntov);
2192 	na = ntov.na;
2193 	amap = ntov.amap;
2194 
2195 	/*
2196 	 * Now loop to get or verify the attrs
2197 	 */
2198 	for (i = 0; i < nfs4_ntov_map_size; i++) {
2199 		if (breq & nfs4_ntov_map[i].fbit) {
2200 			if ((*nfs4_ntov_map[i].sv_getit)(
2201 			    NFS4ATTR_SUPPORTED, sargp, NULL) == 0) {
2202 
2203 				error = (*nfs4_ntov_map[i].sv_getit)(
2204 				    NFS4ATTR_GETIT, sargp, na);
2205 
2206 				/*
2207 				 * Possible error values:
2208 				 * >0 if sv_getit failed to
2209 				 * get the attr; 0 if succeeded;
2210 				 * <0 if rdattr_error and the
2211 				 * attribute cannot be returned.
2212 				 */
2213 				if (error && !(sargp->rdattr_error_req))
2214 					goto done;
2215 				/*
2216 				 * If error then just for entry
2217 				 */
2218 				if (error == 0) {
2219 					fattrp->attrmask |=
2220 					    nfs4_ntov_map[i].fbit;
2221 					*amap++ =
2222 					    (uint8_t)nfs4_ntov_map[i].nval;
2223 					na++;
2224 					(ntov.attrcnt)++;
2225 				} else if ((error > 0) &&
2226 				    (sargp->rdattr_error == NFS4_OK)) {
2227 					sargp->rdattr_error = puterrno4(error);
2228 				}
2229 				error = 0;
2230 			}
2231 		}
2232 	}
2233 
2234 	/*
2235 	 * If rdattr_error was set after the return value for it was assigned,
2236 	 * update it.
2237 	 */
2238 	if (prev_rdattr_error != sargp->rdattr_error) {
2239 		na = ntov.na;
2240 		amap = ntov.amap;
2241 		for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2242 			k = *amap;
2243 			if (k < FATTR4_RDATTR_ERROR) {
2244 				continue;
2245 			}
2246 			if ((k == FATTR4_RDATTR_ERROR) &&
2247 			    ((*nfs4_ntov_map[k].sv_getit)(
2248 			    NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) {
2249 
2250 				(void) (*nfs4_ntov_map[k].sv_getit)(
2251 				    NFS4ATTR_GETIT, sargp, na);
2252 			}
2253 			break;
2254 		}
2255 	}
2256 
2257 	xdr_size = 0;
2258 	na = ntov.na;
2259 	amap = ntov.amap;
2260 	for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2261 		xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na);
2262 	}
2263 
2264 	fattrp->attrlist4_len = xdr_size;
2265 	if (xdr_size) {
2266 		/* freed by rfs4_op_getattr_free() */
2267 		fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP);
2268 
2269 		xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE);
2270 
2271 		na = ntov.na;
2272 		amap = ntov.amap;
2273 		for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2274 			if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) {
2275 				DTRACE_PROBE1(nfss__e__getattr4_encfail,
2276 				    int, *amap);
2277 				status = NFS4ERR_SERVERFAULT;
2278 				break;
2279 			}
2280 		}
2281 		/* xdrmem_destroy(&xdrs); */	/* NO-OP */
2282 	} else {
2283 		fattrp->attrlist4 = NULL;
2284 	}
2285 done:
2286 
2287 	nfs4_ntov_table_free(&ntov, sargp);
2288 
2289 	if (error != 0)
2290 		status = puterrno4(error);
2291 
2292 	return (status);
2293 }
2294 
2295 /* ARGSUSED */
2296 static void
2297 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2298 	struct compound_state *cs)
2299 {
2300 	GETATTR4args *args = &argop->nfs_argop4_u.opgetattr;
2301 	GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr;
2302 	struct nfs4_svgetit_arg sarg;
2303 	struct statvfs64 sb;
2304 	nfsstat4 status;
2305 
2306 	DTRACE_NFSV4_2(op__getattr__start, struct compound_state *, cs,
2307 	    GETATTR4args *, args);
2308 
2309 	if (cs->vp == NULL) {
2310 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2311 		goto out;
2312 	}
2313 
2314 	if (cs->access == CS_ACCESS_DENIED) {
2315 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2316 		goto out;
2317 	}
2318 
2319 	sarg.sbp = &sb;
2320 	sarg.cs = cs;
2321 
2322 	status = bitmap4_to_attrmask(args->attr_request, &sarg);
2323 	if (status == NFS4_OK) {
2324 		status = bitmap4_get_sysattrs(&sarg);
2325 		if (status == NFS4_OK)
2326 			status = do_rfs4_op_getattr(args->attr_request,
2327 			    &resp->obj_attributes, &sarg);
2328 	}
2329 	*cs->statusp = resp->status = status;
2330 out:
2331 	DTRACE_NFSV4_2(op__getattr__done, struct compound_state *, cs,
2332 	    GETATTR4res *, resp);
2333 }
2334 
2335 static void
2336 rfs4_op_getattr_free(nfs_resop4 *resop)
2337 {
2338 	GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr;
2339 
2340 	nfs4_fattr4_free(&resp->obj_attributes);
2341 }
2342 
2343 /* ARGSUSED */
2344 static void
2345 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2346 	struct compound_state *cs)
2347 {
2348 	GETFH4res *resp = &resop->nfs_resop4_u.opgetfh;
2349 
2350 	DTRACE_NFSV4_1(op__getfh__start, struct compound_state *, cs);
2351 
2352 	if (cs->vp == NULL) {
2353 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2354 		goto out;
2355 	}
2356 	if (cs->access == CS_ACCESS_DENIED) {
2357 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2358 		goto out;
2359 	}
2360 
2361 	resp->object.nfs_fh4_val =
2362 	    kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP);
2363 	nfs_fh4_copy(&cs->fh, &resp->object);
2364 	*cs->statusp = resp->status = NFS4_OK;
2365 out:
2366 	DTRACE_NFSV4_2(op__getfh__done, struct compound_state *, cs,
2367 	    GETFH4res *, resp);
2368 }
2369 
2370 static void
2371 rfs4_op_getfh_free(nfs_resop4 *resop)
2372 {
2373 	GETFH4res *resp = &resop->nfs_resop4_u.opgetfh;
2374 
2375 	if (resp->status == NFS4_OK &&
2376 	    resp->object.nfs_fh4_val != NULL) {
2377 		kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len);
2378 		resp->object.nfs_fh4_val = NULL;
2379 		resp->object.nfs_fh4_len = 0;
2380 	}
2381 }
2382 
2383 /*
2384  * illegal: args: void
2385  *	    res : status (NFS4ERR_OP_ILLEGAL)
2386  */
2387 /* ARGSUSED */
2388 static void
2389 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop,
2390 	struct svc_req *req, struct compound_state *cs)
2391 {
2392 	ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal;
2393 
2394 	resop->resop = OP_ILLEGAL;
2395 	*cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL;
2396 }
2397 
2398 /*
2399  * link: args: SAVED_FH: file, CURRENT_FH: target directory
2400  *	 res: status. If success - CURRENT_FH unchanged, return change_info
2401  */
2402 /* ARGSUSED */
2403 static void
2404 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2405 	struct compound_state *cs)
2406 {
2407 	LINK4args *args = &argop->nfs_argop4_u.oplink;
2408 	LINK4res *resp = &resop->nfs_resop4_u.oplink;
2409 	int error;
2410 	vnode_t *vp;
2411 	vnode_t *dvp;
2412 	struct vattr bdva, idva, adva;
2413 	char *nm;
2414 	uint_t  len;
2415 	struct sockaddr *ca;
2416 	char *name = NULL;
2417 
2418 	DTRACE_NFSV4_2(op__link__start, struct compound_state *, cs,
2419 	    LINK4args *, args);
2420 
2421 	/* SAVED_FH: source object */
2422 	vp = cs->saved_vp;
2423 	if (vp == NULL) {
2424 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2425 		goto out;
2426 	}
2427 
2428 	/* CURRENT_FH: target directory */
2429 	dvp = cs->vp;
2430 	if (dvp == NULL) {
2431 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2432 		goto out;
2433 	}
2434 
2435 	/*
2436 	 * If there is a non-shared filesystem mounted on this vnode,
2437 	 * do not allow to link any file in this directory.
2438 	 */
2439 	if (vn_ismntpt(dvp)) {
2440 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2441 		goto out;
2442 	}
2443 
2444 	if (cs->access == CS_ACCESS_DENIED) {
2445 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2446 		goto out;
2447 	}
2448 
2449 	/* Check source object's type validity */
2450 	if (vp->v_type == VDIR) {
2451 		*cs->statusp = resp->status = NFS4ERR_ISDIR;
2452 		goto out;
2453 	}
2454 
2455 	/* Check target directory's type */
2456 	if (dvp->v_type != VDIR) {
2457 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2458 		goto out;
2459 	}
2460 
2461 	if (cs->saved_exi != cs->exi) {
2462 		*cs->statusp = resp->status = NFS4ERR_XDEV;
2463 		goto out;
2464 	}
2465 
2466 	if (!utf8_dir_verify(&args->newname)) {
2467 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2468 		goto out;
2469 	}
2470 
2471 	nm = utf8_to_fn(&args->newname, &len, NULL);
2472 	if (nm == NULL) {
2473 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2474 		goto out;
2475 	}
2476 
2477 	if (len > MAXNAMELEN) {
2478 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
2479 		kmem_free(nm, len);
2480 		goto out;
2481 	}
2482 
2483 	if (rdonly4(cs->exi, cs->vp, req)) {
2484 		*cs->statusp = resp->status = NFS4ERR_ROFS;
2485 		kmem_free(nm, len);
2486 		goto out;
2487 	}
2488 
2489 	/* Get "before" change value */
2490 	bdva.va_mask = AT_CTIME|AT_SEQ;
2491 	error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL);
2492 	if (error) {
2493 		*cs->statusp = resp->status = puterrno4(error);
2494 		kmem_free(nm, len);
2495 		goto out;
2496 	}
2497 
2498 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
2499 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
2500 	    MAXPATHLEN  + 1);
2501 
2502 	if (name == NULL) {
2503 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2504 		kmem_free(nm, len);
2505 		goto out;
2506 	}
2507 
2508 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime)
2509 
2510 	error = VOP_LINK(dvp, vp, name, cs->cr, NULL, 0);
2511 
2512 	if (nm != name)
2513 		kmem_free(name, MAXPATHLEN + 1);
2514 	kmem_free(nm, len);
2515 
2516 	/*
2517 	 * Get the initial "after" sequence number, if it fails, set to zero
2518 	 */
2519 	idva.va_mask = AT_SEQ;
2520 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL))
2521 		idva.va_seq = 0;
2522 
2523 	/*
2524 	 * Force modified data and metadata out to stable storage.
2525 	 */
2526 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL);
2527 	(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
2528 
2529 	if (error) {
2530 		*cs->statusp = resp->status = puterrno4(error);
2531 		goto out;
2532 	}
2533 
2534 	/*
2535 	 * Get "after" change value, if it fails, simply return the
2536 	 * before value.
2537 	 */
2538 	adva.va_mask = AT_CTIME|AT_SEQ;
2539 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) {
2540 		adva.va_ctime = bdva.va_ctime;
2541 		adva.va_seq = 0;
2542 	}
2543 
2544 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
2545 
2546 	/*
2547 	 * The cinfo.atomic = TRUE only if we have
2548 	 * non-zero va_seq's, and it has incremented by exactly one
2549 	 * during the VOP_LINK and it didn't change during the VOP_FSYNC.
2550 	 */
2551 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
2552 	    idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq)
2553 		resp->cinfo.atomic = TRUE;
2554 	else
2555 		resp->cinfo.atomic = FALSE;
2556 
2557 	*cs->statusp = resp->status = NFS4_OK;
2558 out:
2559 	DTRACE_NFSV4_2(op__link__done, struct compound_state *, cs,
2560 	    LINK4res *, resp);
2561 }
2562 
2563 /*
2564  * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work.
2565  */
2566 
2567 /* ARGSUSED */
2568 static nfsstat4
2569 do_rfs4_op_lookup(char *nm, uint_t buflen, struct svc_req *req,
2570 	struct compound_state *cs)
2571 {
2572 	int error;
2573 	int different_export = 0;
2574 	vnode_t *vp, *tvp, *pre_tvp = NULL, *oldvp = NULL;
2575 	struct exportinfo *exi = NULL, *pre_exi = NULL;
2576 	nfsstat4 stat;
2577 	fid_t fid;
2578 	int attrdir, dotdot, walk;
2579 	bool_t is_newvp = FALSE;
2580 
2581 	if (cs->vp->v_flag & V_XATTRDIR) {
2582 		attrdir = 1;
2583 		ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR));
2584 	} else {
2585 		attrdir = 0;
2586 		ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR));
2587 	}
2588 
2589 	dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0');
2590 
2591 	/*
2592 	 * If dotdotting, then need to check whether it's
2593 	 * above the root of a filesystem, or above an
2594 	 * export point.
2595 	 */
2596 	if (dotdot) {
2597 
2598 		/*
2599 		 * If dotdotting at the root of a filesystem, then
2600 		 * need to traverse back to the mounted-on filesystem
2601 		 * and do the dotdot lookup there.
2602 		 */
2603 		if (cs->vp->v_flag & VROOT) {
2604 
2605 			/*
2606 			 * If at the system root, then can
2607 			 * go up no further.
2608 			 */
2609 			if (VN_CMP(cs->vp, rootdir))
2610 				return (puterrno4(ENOENT));
2611 
2612 			/*
2613 			 * Traverse back to the mounted-on filesystem
2614 			 */
2615 			cs->vp = untraverse(cs->vp);
2616 
2617 			/*
2618 			 * Set the different_export flag so we remember
2619 			 * to pick up a new exportinfo entry for
2620 			 * this new filesystem.
2621 			 */
2622 			different_export = 1;
2623 		} else {
2624 
2625 			/*
2626 			 * If dotdotting above an export point then set
2627 			 * the different_export to get new export info.
2628 			 */
2629 			different_export = nfs_exported(cs->exi, cs->vp);
2630 		}
2631 	}
2632 
2633 	error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr,
2634 	    NULL, NULL, NULL);
2635 	if (error)
2636 		return (puterrno4(error));
2637 
2638 	/*
2639 	 * If the vnode is in a pseudo filesystem, check whether it is visible.
2640 	 *
2641 	 * XXX if the vnode is a symlink and it is not visible in
2642 	 * a pseudo filesystem, return ENOENT (not following symlink).
2643 	 * V4 client can not mount such symlink. This is a regression
2644 	 * from V2/V3.
2645 	 *
2646 	 * In the same exported filesystem, if the security flavor used
2647 	 * is not an explicitly shared flavor, limit the view to the visible
2648 	 * list entries only. This is not a WRONGSEC case because it's already
2649 	 * checked via PUTROOTFH/PUTPUBFH or PUTFH.
2650 	 */
2651 	if (!different_export &&
2652 	    (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) ||
2653 	    cs->access & CS_ACCESS_LIMITED)) {
2654 		if (! nfs_visible(cs->exi, vp, &different_export)) {
2655 			VN_RELE(vp);
2656 			return (puterrno4(ENOENT));
2657 		}
2658 	}
2659 
2660 	/*
2661 	 * If it's a mountpoint, then traverse it.
2662 	 */
2663 	if (vn_ismntpt(vp)) {
2664 		pre_exi = cs->exi;	/* save pre-traversed exportinfo */
2665 		pre_tvp = vp;		/* save pre-traversed vnode	*/
2666 
2667 		/*
2668 		 * hold pre_tvp to counteract rele by traverse.  We will
2669 		 * need pre_tvp below if checkexport4 fails
2670 		 */
2671 		VN_HOLD(pre_tvp);
2672 		tvp = vp;
2673 		if ((error = traverse(&tvp)) != 0) {
2674 			VN_RELE(vp);
2675 			VN_RELE(pre_tvp);
2676 			return (puterrno4(error));
2677 		}
2678 		vp = tvp;
2679 		different_export = 1;
2680 	} else if (vp->v_vfsp != cs->vp->v_vfsp) {
2681 		/*
2682 		 * The vfsp comparison is to handle the case where
2683 		 * a LOFS mount is shared.  lo_lookup traverses mount points,
2684 		 * and NFS is unaware of local fs transistions because
2685 		 * v_vfsmountedhere isn't set.  For this special LOFS case,
2686 		 * the dir and the obj returned by lookup will have different
2687 		 * vfs ptrs.
2688 		 */
2689 		different_export = 1;
2690 	}
2691 
2692 	if (different_export) {
2693 
2694 		bzero(&fid, sizeof (fid));
2695 		fid.fid_len = MAXFIDSZ;
2696 		error = vop_fid_pseudo(vp, &fid);
2697 		if (error) {
2698 			VN_RELE(vp);
2699 			if (pre_tvp)
2700 				VN_RELE(pre_tvp);
2701 			return (puterrno4(error));
2702 		}
2703 
2704 		if (dotdot)
2705 			exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE);
2706 		else
2707 			exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
2708 
2709 		if (exi == NULL) {
2710 			if (pre_tvp) {
2711 				/*
2712 				 * If this vnode is a mounted-on vnode,
2713 				 * but the mounted-on file system is not
2714 				 * exported, send back the filehandle for
2715 				 * the mounted-on vnode, not the root of
2716 				 * the mounted-on file system.
2717 				 */
2718 				VN_RELE(vp);
2719 				vp = pre_tvp;
2720 				exi = pre_exi;
2721 			} else {
2722 				VN_RELE(vp);
2723 				return (puterrno4(EACCES));
2724 			}
2725 		} else if (pre_tvp) {
2726 			/* we're done with pre_tvp now. release extra hold */
2727 			VN_RELE(pre_tvp);
2728 		}
2729 
2730 		cs->exi = exi;
2731 
2732 		/*
2733 		 * Now we do a checkauth4. The reason is that
2734 		 * this client/user may not have access to the new
2735 		 * exported file system, and if he does,
2736 		 * the client/user may be mapped to a different uid.
2737 		 *
2738 		 * We start with a new cr, because the checkauth4 done
2739 		 * in the PUT*FH operation over wrote the cred's uid,
2740 		 * gid, etc, and we want the real thing before calling
2741 		 * checkauth4()
2742 		 */
2743 		crfree(cs->cr);
2744 		cs->cr = crdup(cs->basecr);
2745 
2746 		if (cs->vp)
2747 			oldvp = cs->vp;
2748 		cs->vp = vp;
2749 		is_newvp = TRUE;
2750 
2751 		stat = call_checkauth4(cs, req);
2752 		if (stat != NFS4_OK) {
2753 			VN_RELE(cs->vp);
2754 			cs->vp = oldvp;
2755 			return (stat);
2756 		}
2757 	}
2758 
2759 	/*
2760 	 * After various NFS checks, do a label check on the path
2761 	 * component. The label on this path should either be the
2762 	 * global zone's label or a zone's label. We are only
2763 	 * interested in the zone's label because exported files
2764 	 * in global zone is accessible (though read-only) to
2765 	 * clients. The exportability/visibility check is already
2766 	 * done before reaching this code.
2767 	 */
2768 	if (is_system_labeled()) {
2769 		bslabel_t *clabel;
2770 
2771 		ASSERT(req->rq_label != NULL);
2772 		clabel = req->rq_label;
2773 		DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *,
2774 		    "got client label from request(1)", struct svc_req *, req);
2775 
2776 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
2777 			if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK,
2778 			    cs->exi)) {
2779 				error = EACCES;
2780 				goto err_out;
2781 			}
2782 		} else {
2783 			/*
2784 			 * We grant access to admin_low label clients
2785 			 * only if the client is trusted, i.e. also
2786 			 * running Solaris Trusted Extension.
2787 			 */
2788 			struct sockaddr	*ca;
2789 			int		addr_type;
2790 			void		*ipaddr;
2791 			tsol_tpc_t	*tp;
2792 
2793 			ca = (struct sockaddr *)svc_getrpccaller(
2794 			    req->rq_xprt)->buf;
2795 			if (ca->sa_family == AF_INET) {
2796 				addr_type = IPV4_VERSION;
2797 				ipaddr = &((struct sockaddr_in *)ca)->sin_addr;
2798 			} else if (ca->sa_family == AF_INET6) {
2799 				addr_type = IPV6_VERSION;
2800 				ipaddr = &((struct sockaddr_in6 *)
2801 				    ca)->sin6_addr;
2802 			}
2803 			tp = find_tpc(ipaddr, addr_type, B_FALSE);
2804 			if (tp == NULL || tp->tpc_tp.tp_doi !=
2805 			    l_admin_low->tsl_doi || tp->tpc_tp.host_type !=
2806 			    SUN_CIPSO) {
2807 				if (tp != NULL)
2808 					TPC_RELE(tp);
2809 				error = EACCES;
2810 				goto err_out;
2811 			}
2812 			TPC_RELE(tp);
2813 		}
2814 	}
2815 
2816 	error = makefh4(&cs->fh, vp, cs->exi);
2817 
2818 err_out:
2819 	if (error) {
2820 		if (is_newvp) {
2821 			VN_RELE(cs->vp);
2822 			cs->vp = oldvp;
2823 		} else
2824 			VN_RELE(vp);
2825 		return (puterrno4(error));
2826 	}
2827 
2828 	if (!is_newvp) {
2829 		if (cs->vp)
2830 			VN_RELE(cs->vp);
2831 		cs->vp = vp;
2832 	} else if (oldvp)
2833 		VN_RELE(oldvp);
2834 
2835 	/*
2836 	 * if did lookup on attrdir and didn't lookup .., set named
2837 	 * attr fh flag
2838 	 */
2839 	if (attrdir && ! dotdot)
2840 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
2841 
2842 	/* Assume false for now, open proc will set this */
2843 	cs->mandlock = FALSE;
2844 
2845 	return (NFS4_OK);
2846 }
2847 
2848 /* ARGSUSED */
2849 static void
2850 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2851 	struct compound_state *cs)
2852 {
2853 	LOOKUP4args *args = &argop->nfs_argop4_u.oplookup;
2854 	LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup;
2855 	char *nm;
2856 	uint_t len;
2857 	struct sockaddr *ca;
2858 	char *name = NULL;
2859 
2860 	DTRACE_NFSV4_2(op__lookup__start, struct compound_state *, cs,
2861 	    LOOKUP4args *, args);
2862 
2863 	if (cs->vp == NULL) {
2864 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2865 		goto out;
2866 	}
2867 
2868 	if (cs->vp->v_type == VLNK) {
2869 		*cs->statusp = resp->status = NFS4ERR_SYMLINK;
2870 		goto out;
2871 	}
2872 
2873 	if (cs->vp->v_type != VDIR) {
2874 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2875 		goto out;
2876 	}
2877 
2878 	if (!utf8_dir_verify(&args->objname)) {
2879 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2880 		goto out;
2881 	}
2882 
2883 	nm = utf8_to_str(&args->objname, &len, NULL);
2884 	if (nm == NULL) {
2885 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2886 		goto out;
2887 	}
2888 
2889 	if (len > MAXNAMELEN) {
2890 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
2891 		kmem_free(nm, len);
2892 		goto out;
2893 	}
2894 
2895 	/* If necessary, convert to UTF-8 for illbehaved clients */
2896 
2897 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
2898 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
2899 	    MAXPATHLEN  + 1);
2900 
2901 	if (name == NULL) {
2902 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2903 		kmem_free(nm, len);
2904 		goto out;
2905 	}
2906 
2907 	*cs->statusp = resp->status = do_rfs4_op_lookup(name, len, req, cs);
2908 
2909 	if (name != nm)
2910 		kmem_free(name, MAXPATHLEN + 1);
2911 	kmem_free(nm, len);
2912 
2913 out:
2914 	DTRACE_NFSV4_2(op__lookup__done, struct compound_state *, cs,
2915 	    LOOKUP4res *, resp);
2916 }
2917 
2918 /* ARGSUSED */
2919 static void
2920 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
2921 	struct compound_state *cs)
2922 {
2923 	LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp;
2924 
2925 	DTRACE_NFSV4_1(op__lookupp__start, struct compound_state *, cs);
2926 
2927 	if (cs->vp == NULL) {
2928 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2929 		goto out;
2930 	}
2931 
2932 	if (cs->vp->v_type != VDIR) {
2933 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2934 		goto out;
2935 	}
2936 
2937 	*cs->statusp = resp->status = do_rfs4_op_lookup("..", 3, req, cs);
2938 
2939 	/*
2940 	 * From NFSV4 Specification, LOOKUPP should not check for
2941 	 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead.
2942 	 */
2943 	if (resp->status == NFS4ERR_WRONGSEC) {
2944 		*cs->statusp = resp->status = NFS4_OK;
2945 	}
2946 
2947 out:
2948 	DTRACE_NFSV4_2(op__lookupp__done, struct compound_state *, cs,
2949 	    LOOKUPP4res *, resp);
2950 }
2951 
2952 
2953 /*ARGSUSED2*/
2954 static void
2955 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2956 	struct compound_state *cs)
2957 {
2958 	OPENATTR4args	*args = &argop->nfs_argop4_u.opopenattr;
2959 	OPENATTR4res	*resp = &resop->nfs_resop4_u.opopenattr;
2960 	vnode_t		*avp = NULL;
2961 	int		lookup_flags = LOOKUP_XATTR, error;
2962 	int		exp_ro = 0;
2963 
2964 	DTRACE_NFSV4_2(op__openattr__start, struct compound_state *, cs,
2965 	    OPENATTR4args *, args);
2966 
2967 	if (cs->vp == NULL) {
2968 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2969 		goto out;
2970 	}
2971 
2972 	if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0 &&
2973 	    !vfs_has_feature(cs->vp->v_vfsp, VFSFT_SYSATTR_VIEWS)) {
2974 		*cs->statusp = resp->status = puterrno4(ENOTSUP);
2975 		goto out;
2976 	}
2977 
2978 	/*
2979 	 * If file system supports passing ACE mask to VOP_ACCESS then
2980 	 * check for ACE_READ_NAMED_ATTRS, otherwise do legacy checks
2981 	 */
2982 
2983 	if (vfs_has_feature(cs->vp->v_vfsp, VFSFT_ACEMASKONACCESS))
2984 		error = VOP_ACCESS(cs->vp, ACE_READ_NAMED_ATTRS,
2985 		    V_ACE_MASK, cs->cr, NULL);
2986 	else
2987 		error = ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr, NULL) != 0) &&
2988 		    (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr, NULL) != 0) &&
2989 		    (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr, NULL) != 0));
2990 
2991 	if (error) {
2992 		*cs->statusp = resp->status = puterrno4(EACCES);
2993 		goto out;
2994 	}
2995 
2996 	/*
2997 	 * The CREATE_XATTR_DIR VOP flag cannot be specified if
2998 	 * the file system is exported read-only -- regardless of
2999 	 * createdir flag.  Otherwise the attrdir would be created
3000 	 * (assuming server fs isn't mounted readonly locally).  If
3001 	 * VOP_LOOKUP returns ENOENT in this case, the error will
3002 	 * be translated into EROFS.  ENOSYS is mapped to ENOTSUP
3003 	 * because specfs has no VOP_LOOKUP op, so the macro would
3004 	 * return ENOSYS.  EINVAL is returned by all (current)
3005 	 * Solaris file system implementations when any of their
3006 	 * restrictions are violated (xattr(dir) can't have xattrdir).
3007 	 * Returning NOTSUPP is more appropriate in this case
3008 	 * because the object will never be able to have an attrdir.
3009 	 */
3010 	if (args->createdir && ! (exp_ro = rdonly4(cs->exi, cs->vp, req)))
3011 		lookup_flags |= CREATE_XATTR_DIR;
3012 
3013 	error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr,
3014 	    NULL, NULL, NULL);
3015 
3016 	if (error) {
3017 		if (error == ENOENT && args->createdir && exp_ro)
3018 			*cs->statusp = resp->status = puterrno4(EROFS);
3019 		else if (error == EINVAL || error == ENOSYS)
3020 			*cs->statusp = resp->status = puterrno4(ENOTSUP);
3021 		else
3022 			*cs->statusp = resp->status = puterrno4(error);
3023 		goto out;
3024 	}
3025 
3026 	ASSERT(avp->v_flag & V_XATTRDIR);
3027 
3028 	error = makefh4(&cs->fh, avp, cs->exi);
3029 
3030 	if (error) {
3031 		VN_RELE(avp);
3032 		*cs->statusp = resp->status = puterrno4(error);
3033 		goto out;
3034 	}
3035 
3036 	VN_RELE(cs->vp);
3037 	cs->vp = avp;
3038 
3039 	/*
3040 	 * There is no requirement for an attrdir fh flag
3041 	 * because the attrdir has a vnode flag to distinguish
3042 	 * it from regular (non-xattr) directories.  The
3043 	 * FH4_ATTRDIR flag is set for future sanity checks.
3044 	 */
3045 	set_fh4_flag(&cs->fh, FH4_ATTRDIR);
3046 	*cs->statusp = resp->status = NFS4_OK;
3047 
3048 out:
3049 	DTRACE_NFSV4_2(op__openattr__done, struct compound_state *, cs,
3050 	    OPENATTR4res *, resp);
3051 }
3052 
3053 static int
3054 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred,
3055     caller_context_t *ct)
3056 {
3057 	int error;
3058 	int i;
3059 	clock_t delaytime;
3060 
3061 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
3062 
3063 	/*
3064 	 * Don't block on mandatory locks. If this routine returns
3065 	 * EAGAIN, the caller should return NFS4ERR_LOCKED.
3066 	 */
3067 	uio->uio_fmode = FNONBLOCK;
3068 
3069 	for (i = 0; i < rfs4_maxlock_tries; i++) {
3070 
3071 
3072 		if (direction == FREAD) {
3073 			(void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, ct);
3074 			error = VOP_READ(vp, uio, ioflag, cred, ct);
3075 			VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, ct);
3076 		} else {
3077 			(void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, ct);
3078 			error = VOP_WRITE(vp, uio, ioflag, cred, ct);
3079 			VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, ct);
3080 		}
3081 
3082 		if (error != EAGAIN)
3083 			break;
3084 
3085 		if (i < rfs4_maxlock_tries - 1) {
3086 			delay(delaytime);
3087 			delaytime *= 2;
3088 		}
3089 	}
3090 
3091 	return (error);
3092 }
3093 
3094 /* ARGSUSED */
3095 static void
3096 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3097 	struct compound_state *cs)
3098 {
3099 	READ4args *args = &argop->nfs_argop4_u.opread;
3100 	READ4res *resp = &resop->nfs_resop4_u.opread;
3101 	int error;
3102 	int verror;
3103 	vnode_t *vp;
3104 	struct vattr va;
3105 	struct iovec iov;
3106 	struct uio uio;
3107 	u_offset_t offset;
3108 	bool_t *deleg = &cs->deleg;
3109 	nfsstat4 stat;
3110 	int in_crit = 0;
3111 	mblk_t *mp;
3112 	int alloc_err = 0;
3113 	caller_context_t ct;
3114 
3115 	DTRACE_NFSV4_2(op__read__start, struct compound_state *, cs,
3116 	    READ4args, args);
3117 
3118 	vp = cs->vp;
3119 	if (vp == NULL) {
3120 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3121 		goto out;
3122 	}
3123 	if (cs->access == CS_ACCESS_DENIED) {
3124 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3125 		goto out;
3126 	}
3127 
3128 	if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE,
3129 	    deleg, TRUE, &ct)) != NFS4_OK) {
3130 		*cs->statusp = resp->status = stat;
3131 		goto out;
3132 	}
3133 
3134 	/*
3135 	 * Enter the critical region before calling VOP_RWLOCK
3136 	 * to avoid a deadlock with write requests.
3137 	 */
3138 	if (nbl_need_check(vp)) {
3139 		nbl_start_crit(vp, RW_READER);
3140 		in_crit = 1;
3141 		if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0,
3142 		    &ct)) {
3143 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
3144 			goto out;
3145 		}
3146 	}
3147 
3148 	if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE,
3149 	    deleg, TRUE, &ct)) != NFS4_OK) {
3150 		*cs->statusp = resp->status = stat;
3151 		goto out;
3152 	}
3153 
3154 	va.va_mask = AT_MODE|AT_SIZE|AT_UID;
3155 	verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct);
3156 
3157 	/*
3158 	 * If we can't get the attributes, then we can't do the
3159 	 * right access checking.  So, we'll fail the request.
3160 	 */
3161 	if (verror) {
3162 		*cs->statusp = resp->status = puterrno4(verror);
3163 		goto out;
3164 	}
3165 
3166 	if (vp->v_type != VREG) {
3167 		*cs->statusp = resp->status =
3168 		    ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
3169 		goto out;
3170 	}
3171 
3172 	if (crgetuid(cs->cr) != va.va_uid &&
3173 	    (error = VOP_ACCESS(vp, VREAD, 0, cs->cr, &ct)) &&
3174 	    (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr, &ct))) {
3175 		*cs->statusp = resp->status = puterrno4(error);
3176 		goto out;
3177 	}
3178 
3179 	if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */
3180 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3181 		goto out;
3182 	}
3183 
3184 	offset = args->offset;
3185 	if (offset >= va.va_size) {
3186 		*cs->statusp = resp->status = NFS4_OK;
3187 		resp->eof = TRUE;
3188 		resp->data_len = 0;
3189 		resp->data_val = NULL;
3190 		resp->mblk = NULL;
3191 		/* RDMA */
3192 		resp->wlist = args->wlist;
3193 		resp->wlist_len = resp->data_len;
3194 		*cs->statusp = resp->status = NFS4_OK;
3195 		if (resp->wlist)
3196 			clist_zero_len(resp->wlist);
3197 		goto out;
3198 	}
3199 
3200 	if (args->count == 0) {
3201 		*cs->statusp = resp->status = NFS4_OK;
3202 		resp->eof = FALSE;
3203 		resp->data_len = 0;
3204 		resp->data_val = NULL;
3205 		resp->mblk = NULL;
3206 		/* RDMA */
3207 		resp->wlist = args->wlist;
3208 		resp->wlist_len = resp->data_len;
3209 		if (resp->wlist)
3210 			clist_zero_len(resp->wlist);
3211 		goto out;
3212 	}
3213 
3214 	/*
3215 	 * Do not allocate memory more than maximum allowed
3216 	 * transfer size
3217 	 */
3218 	if (args->count > rfs4_tsize(req))
3219 		args->count = rfs4_tsize(req);
3220 
3221 	/*
3222 	 * If returning data via RDMA Write, then grab the chunk list. If we
3223 	 * aren't returning READ data w/RDMA_WRITE, then grab a mblk.
3224 	 */
3225 	if (args->wlist) {
3226 		mp = NULL;
3227 		(void) rdma_get_wchunk(req, &iov, args->wlist);
3228 	} else {
3229 		/*
3230 		 * mp will contain the data to be sent out in the read reply.
3231 		 * It will be freed after the reply has been sent. Let's
3232 		 * roundup the data to a BYTES_PER_XDR_UNIT multiple, so that
3233 		 * the call to xdrmblk_putmblk() never fails. If the first
3234 		 * alloc of the requested size fails, then decrease the size to
3235 		 * something more reasonable and wait for the allocation to
3236 		 * occur.
3237 		 */
3238 		mp = allocb(RNDUP(args->count), BPRI_MED);
3239 		if (mp == NULL) {
3240 			if (args->count > MAXBSIZE)
3241 				args->count = MAXBSIZE;
3242 			mp = allocb_wait(RNDUP(args->count), BPRI_MED,
3243 			    STR_NOSIG, &alloc_err);
3244 		}
3245 		ASSERT(mp != NULL);
3246 		ASSERT(alloc_err == 0);
3247 
3248 		iov.iov_base = (caddr_t)mp->b_datap->db_base;
3249 		iov.iov_len = args->count;
3250 	}
3251 
3252 	uio.uio_iov = &iov;
3253 	uio.uio_iovcnt = 1;
3254 	uio.uio_segflg = UIO_SYSSPACE;
3255 	uio.uio_extflg = UIO_COPY_CACHED;
3256 	uio.uio_loffset = args->offset;
3257 	uio.uio_resid = args->count;
3258 
3259 	error = do_io(FREAD, vp, &uio, 0, cs->cr, &ct);
3260 
3261 	va.va_mask = AT_SIZE;
3262 	verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct);
3263 
3264 	if (error) {
3265 		freeb(mp);
3266 		*cs->statusp = resp->status = puterrno4(error);
3267 		goto out;
3268 	}
3269 
3270 	*cs->statusp = resp->status = NFS4_OK;
3271 
3272 	ASSERT(uio.uio_resid >= 0);
3273 	resp->data_len = args->count - uio.uio_resid;
3274 	if (mp) {
3275 		resp->data_val = (char *)mp->b_datap->db_base;
3276 	} else {
3277 		resp->data_val = (caddr_t)iov.iov_base;
3278 	}
3279 	resp->mblk = mp;
3280 
3281 	if (!verror && offset + resp->data_len == va.va_size)
3282 		resp->eof = TRUE;
3283 	else
3284 		resp->eof = FALSE;
3285 
3286 	if (args->wlist) {
3287 		if (!rdma_setup_read_data4(args, resp)) {
3288 			*cs->statusp = resp->status = NFS4ERR_INVAL;
3289 		}
3290 	} else {
3291 		resp->wlist = NULL;
3292 	}
3293 
3294 out:
3295 	if (in_crit)
3296 		nbl_end_crit(vp);
3297 
3298 	DTRACE_NFSV4_2(op__read__done, struct compound_state *, cs,
3299 	    READ4res *, resp);
3300 }
3301 
3302 static void
3303 rfs4_op_read_free(nfs_resop4 *resop)
3304 {
3305 	READ4res	*resp = &resop->nfs_resop4_u.opread;
3306 
3307 	if (resp->status == NFS4_OK && resp->mblk != NULL) {
3308 		freeb(resp->mblk);
3309 		resp->mblk = NULL;
3310 		resp->data_val = NULL;
3311 		resp->data_len = 0;
3312 	}
3313 }
3314 
3315 static void
3316 rfs4_op_readdir_free(nfs_resop4 * resop)
3317 {
3318 	READDIR4res    *resp = &resop->nfs_resop4_u.opreaddir;
3319 
3320 	if (resp->status == NFS4_OK && resp->mblk != NULL) {
3321 		freeb(resp->mblk);
3322 		resp->mblk = NULL;
3323 		resp->data_len = 0;
3324 	}
3325 }
3326 
3327 
3328 /* ARGSUSED */
3329 static void
3330 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
3331     struct compound_state *cs)
3332 {
3333 	PUTPUBFH4res	*resp = &resop->nfs_resop4_u.opputpubfh;
3334 	int		error;
3335 	vnode_t		*vp;
3336 	struct exportinfo *exi, *sav_exi;
3337 	nfs_fh4_fmt_t	*fh_fmtp;
3338 
3339 	DTRACE_NFSV4_1(op__putpubfh__start, struct compound_state *, cs);
3340 
3341 	if (cs->vp) {
3342 		VN_RELE(cs->vp);
3343 		cs->vp = NULL;
3344 	}
3345 
3346 	if (cs->cr)
3347 		crfree(cs->cr);
3348 
3349 	cs->cr = crdup(cs->basecr);
3350 
3351 	vp = exi_public->exi_vp;
3352 	if (vp == NULL) {
3353 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
3354 		goto out;
3355 	}
3356 
3357 	error = makefh4(&cs->fh, vp, exi_public);
3358 	if (error != 0) {
3359 		*cs->statusp = resp->status = puterrno4(error);
3360 		goto out;
3361 	}
3362 	sav_exi = cs->exi;
3363 	if (exi_public == exi_root) {
3364 		/*
3365 		 * No filesystem is actually shared public, so we default
3366 		 * to exi_root. In this case, we must check whether root
3367 		 * is exported.
3368 		 */
3369 		fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val;
3370 
3371 		/*
3372 		 * if root filesystem is exported, the exportinfo struct that we
3373 		 * should use is what checkexport4 returns, because root_exi is
3374 		 * actually a mostly empty struct.
3375 		 */
3376 		exi = checkexport4(&fh_fmtp->fh4_fsid,
3377 		    (fid_t *)&fh_fmtp->fh4_xlen, NULL);
3378 		cs->exi = ((exi != NULL) ? exi : exi_public);
3379 	} else {
3380 		/*
3381 		 * it's a properly shared filesystem
3382 		 */
3383 		cs->exi = exi_public;
3384 	}
3385 
3386 	if (is_system_labeled()) {
3387 		bslabel_t *clabel;
3388 
3389 		ASSERT(req->rq_label != NULL);
3390 		clabel = req->rq_label;
3391 		DTRACE_PROBE2(tx__rfs4__log__info__opputpubfh__clabel, char *,
3392 		    "got client label from request(1)",
3393 		    struct svc_req *, req);
3394 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
3395 			if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK,
3396 			    cs->exi)) {
3397 				*cs->statusp = resp->status =
3398 				    NFS4ERR_SERVERFAULT;
3399 				goto out;
3400 			}
3401 		}
3402 	}
3403 
3404 	VN_HOLD(vp);
3405 	cs->vp = vp;
3406 
3407 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3408 		VN_RELE(cs->vp);
3409 		cs->vp = NULL;
3410 		cs->exi = sav_exi;
3411 		goto out;
3412 	}
3413 
3414 	*cs->statusp = resp->status = NFS4_OK;
3415 out:
3416 	DTRACE_NFSV4_2(op__putpubfh__done, struct compound_state *, cs,
3417 	    PUTPUBFH4res *, resp);
3418 }
3419 
3420 /*
3421  * XXX - issue with put*fh operations. Suppose /export/home is exported.
3422  * Suppose an NFS client goes to mount /export/home/joe. If /export, home,
3423  * or joe have restrictive search permissions, then we shouldn't let
3424  * the client get a file handle. This is easy to enforce. However, we
3425  * don't know what security flavor should be used until we resolve the
3426  * path name. Another complication is uid mapping. If root is
3427  * the user, then it will be mapped to the anonymous user by default,
3428  * but we won't know that till we've resolved the path name. And we won't
3429  * know what the anonymous user is.
3430  * Luckily, SECINFO is specified to take a full filename.
3431  * So what we will have to in rfs4_op_lookup is check that flavor of
3432  * the target object matches that of the request, and if root was the
3433  * caller, check for the root= and anon= options, and if necessary,
3434  * repeat the lookup using the right cred_t. But that's not done yet.
3435  */
3436 /* ARGSUSED */
3437 static void
3438 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3439 	struct compound_state *cs)
3440 {
3441 	PUTFH4args *args = &argop->nfs_argop4_u.opputfh;
3442 	PUTFH4res *resp = &resop->nfs_resop4_u.opputfh;
3443 	nfs_fh4_fmt_t *fh_fmtp;
3444 
3445 	DTRACE_NFSV4_2(op__putfh__start, struct compound_state *, cs,
3446 	    PUTFH4args *, args);
3447 
3448 	if (cs->vp) {
3449 		VN_RELE(cs->vp);
3450 		cs->vp = NULL;
3451 	}
3452 
3453 	if (cs->cr) {
3454 		crfree(cs->cr);
3455 		cs->cr = NULL;
3456 	}
3457 
3458 
3459 	if (args->object.nfs_fh4_len < NFS_FH4_LEN) {
3460 		*cs->statusp = resp->status = NFS4ERR_BADHANDLE;
3461 		goto out;
3462 	}
3463 
3464 	fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val;
3465 	cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen,
3466 	    NULL);
3467 
3468 	if (cs->exi == NULL) {
3469 		*cs->statusp = resp->status = NFS4ERR_STALE;
3470 		goto out;
3471 	}
3472 
3473 	cs->cr = crdup(cs->basecr);
3474 
3475 	ASSERT(cs->cr != NULL);
3476 
3477 	if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) {
3478 		*cs->statusp = resp->status;
3479 		goto out;
3480 	}
3481 
3482 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3483 		VN_RELE(cs->vp);
3484 		cs->vp = NULL;
3485 		goto out;
3486 	}
3487 
3488 	nfs_fh4_copy(&args->object, &cs->fh);
3489 	*cs->statusp = resp->status = NFS4_OK;
3490 	cs->deleg = FALSE;
3491 
3492 out:
3493 	DTRACE_NFSV4_2(op__putfh__done, struct compound_state *, cs,
3494 	    PUTFH4res *, resp);
3495 }
3496 
3497 /* ARGSUSED */
3498 static void
3499 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3500 	struct compound_state *cs)
3501 
3502 {
3503 	PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh;
3504 	int error;
3505 	fid_t fid;
3506 	struct exportinfo *exi, *sav_exi;
3507 
3508 	DTRACE_NFSV4_1(op__putrootfh__start, struct compound_state *, cs);
3509 
3510 	if (cs->vp) {
3511 		VN_RELE(cs->vp);
3512 		cs->vp = NULL;
3513 	}
3514 
3515 	if (cs->cr)
3516 		crfree(cs->cr);
3517 
3518 	cs->cr = crdup(cs->basecr);
3519 
3520 	/*
3521 	 * Using rootdir, the system root vnode,
3522 	 * get its fid.
3523 	 */
3524 	bzero(&fid, sizeof (fid));
3525 	fid.fid_len = MAXFIDSZ;
3526 	error = vop_fid_pseudo(rootdir, &fid);
3527 	if (error != 0) {
3528 		*cs->statusp = resp->status = puterrno4(error);
3529 		goto out;
3530 	}
3531 
3532 	/*
3533 	 * Then use the root fsid & fid it to find out if it's exported
3534 	 *
3535 	 * If the server root isn't exported directly, then
3536 	 * it should at least be a pseudo export based on
3537 	 * one or more exports further down in the server's
3538 	 * file tree.
3539 	 */
3540 	exi = checkexport4(&rootdir->v_vfsp->vfs_fsid, &fid, NULL);
3541 	if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) {
3542 		NFS4_DEBUG(rfs4_debug,
3543 			(CE_WARN, "rfs4_op_putrootfh: export check failure"));
3544 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
3545 		goto out;
3546 	}
3547 
3548 	/*
3549 	 * Now make a filehandle based on the root
3550 	 * export and root vnode.
3551 	 */
3552 	error = makefh4(&cs->fh, rootdir, exi);
3553 	if (error != 0) {
3554 		*cs->statusp = resp->status = puterrno4(error);
3555 		goto out;
3556 	}
3557 
3558 	sav_exi = cs->exi;
3559 	cs->exi = exi;
3560 
3561 	VN_HOLD(rootdir);
3562 	cs->vp = rootdir;
3563 
3564 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3565 		VN_RELE(rootdir);
3566 		cs->vp = NULL;
3567 		cs->exi = sav_exi;
3568 		goto out;
3569 	}
3570 
3571 	*cs->statusp = resp->status = NFS4_OK;
3572 	cs->deleg = FALSE;
3573 out:
3574 	DTRACE_NFSV4_2(op__putrootfh__done, struct compound_state *, cs,
3575 	    PUTROOTFH4res *, resp);
3576 }
3577 
3578 /*
3579  * A directory entry is a valid nfsv4 entry if
3580  * - it has a non-zero ino
3581  * - it is not a dot or dotdot name
3582  * - it is visible in a pseudo export or in a real export that can
3583  *   only have a limited view.
3584  */
3585 static bool_t
3586 valid_nfs4_entry(struct exportinfo *exi, struct dirent64 *dp,
3587 		int *expseudo, int check_visible)
3588 {
3589 	if (dp->d_ino == 0 || NFS_IS_DOTNAME(dp->d_name)) {
3590 		*expseudo = 0;
3591 		return (FALSE);
3592 	}
3593 
3594 	if (! check_visible) {
3595 		*expseudo = 0;
3596 		return (TRUE);
3597 	}
3598 
3599 	return (nfs_visible_inode(exi, dp->d_ino, expseudo));
3600 }
3601 
3602 /*
3603  * set_rdattr_params sets up the variables used to manage what information
3604  * to get for each directory entry.
3605  */
3606 static nfsstat4
3607 set_rdattr_params(struct nfs4_svgetit_arg *sargp,
3608 		bitmap4 attrs, bool_t *need_to_lookup)
3609 {
3610 	uint_t	va_mask;
3611 	nfsstat4 status;
3612 	bitmap4 objbits;
3613 
3614 	status = bitmap4_to_attrmask(attrs, sargp);
3615 	if (status != NFS4_OK) {
3616 		/*
3617 		 * could not even figure attr mask
3618 		 */
3619 		return (status);
3620 	}
3621 	va_mask = sargp->vap->va_mask;
3622 
3623 	/*
3624 	 * dirent's d_ino is always correct value for mounted_on_fileid.
3625 	 * mntdfid_set is set once here, but mounted_on_fileid is
3626 	 * set in main dirent processing loop for each dirent.
3627 	 * The mntdfid_set is a simple optimization that lets the
3628 	 * server attr code avoid work when caller is readdir.
3629 	 */
3630 	sargp->mntdfid_set = TRUE;
3631 
3632 	/*
3633 	 * Lookup entry only if client asked for any of the following:
3634 	 * a) vattr attrs
3635 	 * b) vfs attrs
3636 	 * c) attrs w/per-object scope requested (change, filehandle, etc)
3637 	 *    other than mounted_on_fileid (which we can take from dirent)
3638 	 */
3639 	objbits = attrs ? attrs & NFS4_VP_ATTR_MASK : 0;
3640 
3641 	if (va_mask || sargp->sbp || (objbits & ~FATTR4_MOUNTED_ON_FILEID_MASK))
3642 		*need_to_lookup = TRUE;
3643 	else
3644 		*need_to_lookup = FALSE;
3645 
3646 	if (sargp->sbp == NULL)
3647 		return (NFS4_OK);
3648 
3649 	/*
3650 	 * If filesystem attrs are requested, get them now from the
3651 	 * directory vp, as most entries will have same filesystem. The only
3652 	 * exception are mounted over entries but we handle
3653 	 * those as we go (XXX mounted over detection not yet implemented).
3654 	 */
3655 	sargp->vap->va_mask = 0;	/* to avoid VOP_GETATTR */
3656 	status = bitmap4_get_sysattrs(sargp);
3657 	sargp->vap->va_mask = va_mask;
3658 
3659 	if ((status != NFS4_OK) && sargp->rdattr_error_req) {
3660 		/*
3661 		 * Failed to get filesystem attributes.
3662 		 * Return a rdattr_error for each entry, but don't fail.
3663 		 * However, don't get any obj-dependent attrs.
3664 		 */
3665 		sargp->rdattr_error = status;	/* for rdattr_error */
3666 		*need_to_lookup = FALSE;
3667 		/*
3668 		 * At least get fileid for regular readdir output
3669 		 */
3670 		sargp->vap->va_mask &= AT_NODEID;
3671 		status = NFS4_OK;
3672 	}
3673 
3674 	return (status);
3675 }
3676 
3677 /*
3678  * readlink: args: CURRENT_FH.
3679  *	res: status. If success - CURRENT_FH unchanged, return linktext.
3680  */
3681 
3682 /* ARGSUSED */
3683 static void
3684 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3685 	struct compound_state *cs)
3686 {
3687 	READLINK4res *resp = &resop->nfs_resop4_u.opreadlink;
3688 	int error;
3689 	vnode_t *vp;
3690 	struct iovec iov;
3691 	struct vattr va;
3692 	struct uio uio;
3693 	char *data;
3694 	struct sockaddr *ca;
3695 	char *name = NULL;
3696 
3697 	DTRACE_NFSV4_1(op__readlink__start, struct compound_state *, cs);
3698 
3699 	/* CURRENT_FH: directory */
3700 	vp = cs->vp;
3701 	if (vp == NULL) {
3702 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3703 		goto out;
3704 	}
3705 
3706 	if (cs->access == CS_ACCESS_DENIED) {
3707 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3708 		goto out;
3709 	}
3710 
3711 	if (vp->v_type == VDIR) {
3712 		*cs->statusp = resp->status = NFS4ERR_ISDIR;
3713 		goto out;
3714 	}
3715 
3716 	if (vp->v_type != VLNK) {
3717 		*cs->statusp = resp->status = NFS4ERR_INVAL;
3718 		goto out;
3719 	}
3720 
3721 	va.va_mask = AT_MODE;
3722 	error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL);
3723 	if (error) {
3724 		*cs->statusp = resp->status = puterrno4(error);
3725 		goto out;
3726 	}
3727 
3728 	if (MANDLOCK(vp, va.va_mode)) {
3729 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3730 		goto out;
3731 	}
3732 
3733 	data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP);
3734 
3735 	iov.iov_base = data;
3736 	iov.iov_len = MAXPATHLEN;
3737 	uio.uio_iov = &iov;
3738 	uio.uio_iovcnt = 1;
3739 	uio.uio_segflg = UIO_SYSSPACE;
3740 	uio.uio_extflg = UIO_COPY_CACHED;
3741 	uio.uio_loffset = 0;
3742 	uio.uio_resid = MAXPATHLEN;
3743 
3744 	error = VOP_READLINK(vp, &uio, cs->cr, NULL);
3745 
3746 	if (error) {
3747 		kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1);
3748 		*cs->statusp = resp->status = puterrno4(error);
3749 		goto out;
3750 	}
3751 
3752 	*(data + MAXPATHLEN - uio.uio_resid) = '\0';
3753 
3754 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
3755 	name = nfscmd_convname(ca, cs->exi, data, NFSCMD_CONV_OUTBOUND,
3756 	    MAXPATHLEN  + 1);
3757 
3758 	if (name == NULL) {
3759 		/*
3760 		 * Even though the conversion failed, we return
3761 		 * something. We just don't translate it.
3762 		 */
3763 		name = data;
3764 	}
3765 
3766 	/*
3767 	 * treat link name as data
3768 	 */
3769 	(void) str_to_utf8(name, &resp->link);
3770 
3771 	if (name != data)
3772 		kmem_free(name, MAXPATHLEN + 1);
3773 	kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1);
3774 	*cs->statusp = resp->status = NFS4_OK;
3775 
3776 out:
3777 	DTRACE_NFSV4_2(op__readlink__done, struct compound_state *, cs,
3778 	    READLINK4res *, resp);
3779 }
3780 
3781 static void
3782 rfs4_op_readlink_free(nfs_resop4 *resop)
3783 {
3784 	READLINK4res *resp = &resop->nfs_resop4_u.opreadlink;
3785 	utf8string *symlink = &resp->link;
3786 
3787 	if (symlink->utf8string_val) {
3788 		UTF8STRING_FREE(*symlink)
3789 	}
3790 }
3791 
3792 /*
3793  * release_lockowner:
3794  *	Release any state associated with the supplied
3795  *	lockowner. Note if any lo_state is holding locks we will not
3796  *	rele that lo_state and thus the lockowner will not be destroyed.
3797  *	A client using lock after the lock owner stateid has been released
3798  *	will suffer the consequence of NFS4ERR_BAD_STATEID and would have
3799  *	to reissue the lock with new_lock_owner set to TRUE.
3800  *	args: lock_owner
3801  *	res:  status
3802  */
3803 /* ARGSUSED */
3804 static void
3805 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop,
3806 	struct svc_req *req, struct compound_state *cs)
3807 {
3808 	RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner;
3809 	RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner;
3810 	rfs4_lockowner_t *lo;
3811 	rfs4_openowner_t *oop;
3812 	rfs4_state_t *sp;
3813 	rfs4_lo_state_t *lsp;
3814 	rfs4_client_t *cp;
3815 	bool_t create = FALSE;
3816 	locklist_t *llist;
3817 	sysid_t sysid;
3818 
3819 	DTRACE_NFSV4_2(op__release__lockowner__start, struct compound_state *,
3820 	    cs, RELEASE_LOCKOWNER4args *, ap);
3821 
3822 	/* Make sure there is a clientid around for this request */
3823 	cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE);
3824 
3825 	if (cp == NULL) {
3826 		*cs->statusp = resp->status =
3827 		    rfs4_check_clientid(&ap->lock_owner.clientid, 0);
3828 		goto out;
3829 	}
3830 	rfs4_client_rele(cp);
3831 
3832 	lo = rfs4_findlockowner(&ap->lock_owner, &create);
3833 	if (lo == NULL) {
3834 		*cs->statusp = resp->status = NFS4_OK;
3835 		goto out;
3836 	}
3837 	ASSERT(lo->client != NULL);
3838 
3839 	/*
3840 	 * Check for EXPIRED client. If so will reap state with in a lease
3841 	 * period or on next set_clientid_confirm step
3842 	 */
3843 	if (rfs4_lease_expired(lo->client)) {
3844 		rfs4_lockowner_rele(lo);
3845 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
3846 		goto out;
3847 	}
3848 
3849 	/*
3850 	 * If no sysid has been assigned, then no locks exist; just return.
3851 	 */
3852 	rfs4_dbe_lock(lo->client->dbe);
3853 	if (lo->client->sysidt == LM_NOSYSID) {
3854 		rfs4_lockowner_rele(lo);
3855 		rfs4_dbe_unlock(lo->client->dbe);
3856 		goto out;
3857 	}
3858 
3859 	sysid = lo->client->sysidt;
3860 	rfs4_dbe_unlock(lo->client->dbe);
3861 
3862 	/*
3863 	 * Mark the lockowner invalid.
3864 	 */
3865 	rfs4_dbe_hide(lo->dbe);
3866 
3867 	/*
3868 	 * sysid-pid pair should now not be used since the lockowner is
3869 	 * invalid. If the client were to instantiate the lockowner again
3870 	 * it would be assigned a new pid. Thus we can get the list of
3871 	 * current locks.
3872 	 */
3873 
3874 	llist = flk_get_active_locks(sysid, lo->pid);
3875 	/* If we are still holding locks fail */
3876 	if (llist != NULL) {
3877 
3878 		*cs->statusp = resp->status = NFS4ERR_LOCKS_HELD;
3879 
3880 		flk_free_locklist(llist);
3881 		/*
3882 		 * We need to unhide the lockowner so the client can
3883 		 * try it again. The bad thing here is if the client
3884 		 * has a logic error that took it here in the first place
3885 		 * he probably has lost accounting of the locks that it
3886 		 * is holding. So we may have dangling state until the
3887 		 * open owner state is reaped via close. One scenario
3888 		 * that could possibly occur is that the client has
3889 		 * sent the unlock request(s) in separate threads
3890 		 * and has not waited for the replies before sending the
3891 		 * RELEASE_LOCKOWNER request. Presumably, it would expect
3892 		 * and deal appropriately with NFS4ERR_LOCKS_HELD, by
3893 		 * reissuing the request.
3894 		 */
3895 		rfs4_dbe_unhide(lo->dbe);
3896 		rfs4_lockowner_rele(lo);
3897 		goto out;
3898 	}
3899 
3900 	/*
3901 	 * For the corresponding client we need to check each open
3902 	 * owner for any opens that have lockowner state associated
3903 	 * with this lockowner.
3904 	 */
3905 
3906 	rfs4_dbe_lock(lo->client->dbe);
3907 	for (oop = lo->client->openownerlist.next->oop; oop != NULL;
3908 	    oop = oop->openownerlist.next->oop) {
3909 
3910 		rfs4_dbe_lock(oop->dbe);
3911 		for (sp = oop->ownerstateids.next->sp; sp != NULL;
3912 		    sp = sp->ownerstateids.next->sp) {
3913 
3914 			rfs4_dbe_lock(sp->dbe);
3915 			for (lsp = sp->lockownerlist.next->lsp;
3916 			    lsp != NULL; lsp = lsp->lockownerlist.next->lsp) {
3917 				if (lsp->locker == lo) {
3918 					rfs4_dbe_lock(lsp->dbe);
3919 					rfs4_dbe_invalidate(lsp->dbe);
3920 					rfs4_dbe_unlock(lsp->dbe);
3921 				}
3922 			}
3923 			rfs4_dbe_unlock(sp->dbe);
3924 		}
3925 		rfs4_dbe_unlock(oop->dbe);
3926 	}
3927 	rfs4_dbe_unlock(lo->client->dbe);
3928 
3929 	rfs4_lockowner_rele(lo);
3930 
3931 	*cs->statusp = resp->status = NFS4_OK;
3932 
3933 out:
3934 	DTRACE_NFSV4_2(op__release__lockowner__done, struct compound_state *,
3935 	    cs, RELEASE_LOCKOWNER4res *, resp);
3936 }
3937 
3938 /*
3939  * short utility function to lookup a file and recall the delegation
3940  */
3941 static rfs4_file_t *
3942 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp,
3943 	int *lkup_error, cred_t *cr)
3944 {
3945 	vnode_t *vp;
3946 	rfs4_file_t *fp = NULL;
3947 	bool_t fcreate = FALSE;
3948 	int error;
3949 
3950 	if (vpp)
3951 		*vpp = NULL;
3952 
3953 	if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, NULL, NULL,
3954 	    NULL)) == 0) {
3955 		if (vp->v_type == VREG)
3956 			fp = rfs4_findfile(vp, NULL, &fcreate);
3957 		if (vpp)
3958 			*vpp = vp;
3959 		else
3960 			VN_RELE(vp);
3961 	}
3962 
3963 	if (lkup_error)
3964 		*lkup_error = error;
3965 
3966 	return (fp);
3967 }
3968 
3969 /*
3970  * remove: args: CURRENT_FH: directory; name.
3971  *	res: status. If success - CURRENT_FH unchanged, return change_info
3972  *		for directory.
3973  */
3974 /* ARGSUSED */
3975 static void
3976 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3977 	struct compound_state *cs)
3978 {
3979 	REMOVE4args *args = &argop->nfs_argop4_u.opremove;
3980 	REMOVE4res *resp = &resop->nfs_resop4_u.opremove;
3981 	int error;
3982 	vnode_t *dvp, *vp;
3983 	struct vattr bdva, idva, adva;
3984 	char *nm;
3985 	uint_t len;
3986 	rfs4_file_t *fp;
3987 	int in_crit = 0;
3988 	bslabel_t *clabel;
3989 	struct sockaddr *ca;
3990 	char *name = NULL;
3991 
3992 	DTRACE_NFSV4_2(op__remove__start, struct compound_state *, cs,
3993 	    REMOVE4args *, args);
3994 
3995 	/* CURRENT_FH: directory */
3996 	dvp = cs->vp;
3997 	if (dvp == NULL) {
3998 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3999 		goto out;
4000 	}
4001 
4002 	if (cs->access == CS_ACCESS_DENIED) {
4003 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4004 		goto out;
4005 	}
4006 
4007 	/*
4008 	 * If there is an unshared filesystem mounted on this vnode,
4009 	 * Do not allow to remove anything in this directory.
4010 	 */
4011 	if (vn_ismntpt(dvp)) {
4012 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4013 		goto out;
4014 	}
4015 
4016 	if (dvp->v_type != VDIR) {
4017 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
4018 		goto out;
4019 	}
4020 
4021 	if (!utf8_dir_verify(&args->target)) {
4022 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4023 		goto out;
4024 	}
4025 
4026 	/*
4027 	 * Lookup the file so that we can check if it's a directory
4028 	 */
4029 	nm = utf8_to_fn(&args->target, &len, NULL);
4030 	if (nm == NULL) {
4031 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4032 		goto out;
4033 	}
4034 
4035 	if (len > MAXNAMELEN) {
4036 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
4037 		kmem_free(nm, len);
4038 		goto out;
4039 	}
4040 
4041 	if (rdonly4(cs->exi, cs->vp, req)) {
4042 		*cs->statusp = resp->status = NFS4ERR_ROFS;
4043 		kmem_free(nm, len);
4044 		goto out;
4045 	}
4046 
4047 	/* If necessary, convert to UTF-8 for illbehaved clients */
4048 
4049 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
4050 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
4051 	    MAXPATHLEN  + 1);
4052 
4053 	if (name == NULL) {
4054 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4055 		kmem_free(nm, len);
4056 		goto out;
4057 	}
4058 
4059 	/*
4060 	 * Lookup the file to determine type and while we are see if
4061 	 * there is a file struct around and check for delegation.
4062 	 * We don't need to acquire va_seq before this lookup, if
4063 	 * it causes an update, cinfo.before will not match, which will
4064 	 * trigger a cache flush even if atomic is TRUE.
4065 	 */
4066 	if (fp = rfs4_lookup_and_findfile(dvp, name, &vp, &error, cs->cr)) {
4067 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
4068 		    NULL)) {
4069 			VN_RELE(vp);
4070 			rfs4_file_rele(fp);
4071 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4072 			if (nm != name)
4073 				kmem_free(name, MAXPATHLEN + 1);
4074 			kmem_free(nm, len);
4075 			goto out;
4076 		}
4077 	}
4078 
4079 	/* Didn't find anything to remove */
4080 	if (vp == NULL) {
4081 		*cs->statusp = resp->status = error;
4082 		if (nm != name)
4083 			kmem_free(name, MAXPATHLEN + 1);
4084 		kmem_free(nm, len);
4085 		goto out;
4086 	}
4087 
4088 	if (nbl_need_check(vp)) {
4089 		nbl_start_crit(vp, RW_READER);
4090 		in_crit = 1;
4091 		if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0, NULL)) {
4092 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4093 			if (nm != name)
4094 				kmem_free(name, MAXPATHLEN + 1);
4095 			kmem_free(nm, len);
4096 			nbl_end_crit(vp);
4097 			VN_RELE(vp);
4098 			if (fp) {
4099 				rfs4_clear_dont_grant(fp);
4100 				rfs4_file_rele(fp);
4101 			}
4102 			goto out;
4103 		}
4104 	}
4105 
4106 	/* check label before allowing removal */
4107 	if (is_system_labeled()) {
4108 		ASSERT(req->rq_label != NULL);
4109 		clabel = req->rq_label;
4110 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
4111 		    "got client label from request(1)",
4112 		    struct svc_req *, req);
4113 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
4114 			if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK,
4115 			    cs->exi)) {
4116 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
4117 				if (name != nm)
4118 					kmem_free(name, MAXPATHLEN + 1);
4119 				kmem_free(nm, len);
4120 				if (in_crit)
4121 					nbl_end_crit(vp);
4122 				VN_RELE(vp);
4123 				if (fp) {
4124 					rfs4_clear_dont_grant(fp);
4125 					rfs4_file_rele(fp);
4126 				}
4127 				goto out;
4128 			}
4129 		}
4130 	}
4131 
4132 	/* Get dir "before" change value */
4133 	bdva.va_mask = AT_CTIME|AT_SEQ;
4134 	error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL);
4135 	if (error) {
4136 		*cs->statusp = resp->status = puterrno4(error);
4137 		if (nm != name)
4138 			kmem_free(name, MAXPATHLEN + 1);
4139 		kmem_free(nm, len);
4140 		if (in_crit)
4141 			nbl_end_crit(vp);
4142 		VN_RELE(vp);
4143 		if (fp) {
4144 			rfs4_clear_dont_grant(fp);
4145 			rfs4_file_rele(fp);
4146 		}
4147 		goto out;
4148 	}
4149 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime)
4150 
4151 	/* Actually do the REMOVE operation */
4152 	if (vp->v_type == VDIR) {
4153 		/*
4154 		 * Can't remove a directory that has a mounted-on filesystem.
4155 		 */
4156 		if (vn_ismntpt(vp)) {
4157 			error = EACCES;
4158 		} else {
4159 			/*
4160 			 * System V defines rmdir to return EEXIST,
4161 			 * not * ENOTEMPTY, if the directory is not
4162 			 * empty.  A System V NFS server needs to map
4163 			 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to
4164 			 * transmit over the wire.
4165 			 */
4166 			if ((error = VOP_RMDIR(dvp, nm, rootdir, cs->cr,
4167 			    NULL, 0)) == EEXIST)
4168 				error = ENOTEMPTY;
4169 		}
4170 	} else {
4171 		if ((error = VOP_REMOVE(dvp, name, cs->cr, NULL, 0)) == 0 &&
4172 		    fp != NULL) {
4173 			struct vattr va;
4174 			vnode_t *tvp;
4175 
4176 			rfs4_dbe_lock(fp->dbe);
4177 			tvp = fp->vp;
4178 			if (tvp)
4179 				VN_HOLD(tvp);
4180 			rfs4_dbe_unlock(fp->dbe);
4181 
4182 			if (tvp) {
4183 				/*
4184 				 * This is va_seq safe because we are not
4185 				 * manipulating dvp.
4186 				 */
4187 				va.va_mask = AT_NLINK;
4188 				if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) &&
4189 				    va.va_nlink == 0) {
4190 					/* Remove state on file remove */
4191 					if (in_crit) {
4192 						nbl_end_crit(vp);
4193 						in_crit = 0;
4194 					}
4195 					rfs4_close_all_state(fp);
4196 				}
4197 				VN_RELE(tvp);
4198 			}
4199 		}
4200 	}
4201 
4202 	if (in_crit)
4203 		nbl_end_crit(vp);
4204 	VN_RELE(vp);
4205 
4206 	if (fp) {
4207 		rfs4_clear_dont_grant(fp);
4208 		rfs4_file_rele(fp);
4209 	}
4210 	if (nm != name)
4211 		kmem_free(name, MAXPATHLEN + 1);
4212 	kmem_free(nm, len);
4213 
4214 	if (error) {
4215 		*cs->statusp = resp->status = puterrno4(error);
4216 		goto out;
4217 	}
4218 
4219 	/*
4220 	 * Get the initial "after" sequence number, if it fails, set to zero
4221 	 */
4222 	idva.va_mask = AT_SEQ;
4223 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL))
4224 		idva.va_seq = 0;
4225 
4226 	/*
4227 	 * Force modified data and metadata out to stable storage.
4228 	 */
4229 	(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
4230 
4231 	/*
4232 	 * Get "after" change value, if it fails, simply return the
4233 	 * before value.
4234 	 */
4235 	adva.va_mask = AT_CTIME|AT_SEQ;
4236 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) {
4237 		adva.va_ctime = bdva.va_ctime;
4238 		adva.va_seq = 0;
4239 	}
4240 
4241 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
4242 
4243 	/*
4244 	 * The cinfo.atomic = TRUE only if we have
4245 	 * non-zero va_seq's, and it has incremented by exactly one
4246 	 * during the VOP_REMOVE/RMDIR and it didn't change during
4247 	 * the VOP_FSYNC.
4248 	 */
4249 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
4250 	    idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq)
4251 		resp->cinfo.atomic = TRUE;
4252 	else
4253 		resp->cinfo.atomic = FALSE;
4254 
4255 	*cs->statusp = resp->status = NFS4_OK;
4256 
4257 out:
4258 	DTRACE_NFSV4_2(op__remove__done, struct compound_state *, cs,
4259 	    REMOVE4res *, resp);
4260 }
4261 
4262 /*
4263  * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory,
4264  *		oldname and newname.
4265  *	res: status. If success - CURRENT_FH unchanged, return change_info
4266  *		for both from and target directories.
4267  */
4268 /* ARGSUSED */
4269 static void
4270 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4271 	struct compound_state *cs)
4272 {
4273 	RENAME4args *args = &argop->nfs_argop4_u.oprename;
4274 	RENAME4res *resp = &resop->nfs_resop4_u.oprename;
4275 	int error;
4276 	vnode_t *odvp;
4277 	vnode_t *ndvp;
4278 	vnode_t *srcvp, *targvp;
4279 	struct vattr obdva, oidva, oadva;
4280 	struct vattr nbdva, nidva, nadva;
4281 	char *onm, *nnm;
4282 	uint_t olen, nlen;
4283 	rfs4_file_t *fp, *sfp;
4284 	int in_crit_src, in_crit_targ;
4285 	int fp_rele_grant_hold, sfp_rele_grant_hold;
4286 	bslabel_t *clabel;
4287 	struct sockaddr *ca;
4288 	char *converted_onm = NULL;
4289 	char *converted_nnm = NULL;
4290 
4291 	DTRACE_NFSV4_2(op__rename__start, struct compound_state *, cs,
4292 	    RENAME4args *, args);
4293 
4294 	fp = sfp = NULL;
4295 	srcvp = targvp = NULL;
4296 	in_crit_src = in_crit_targ = 0;
4297 	fp_rele_grant_hold = sfp_rele_grant_hold = 0;
4298 
4299 	/* CURRENT_FH: target directory */
4300 	ndvp = cs->vp;
4301 	if (ndvp == NULL) {
4302 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4303 		goto out;
4304 	}
4305 
4306 	/* SAVED_FH: from directory */
4307 	odvp = cs->saved_vp;
4308 	if (odvp == NULL) {
4309 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4310 		goto out;
4311 	}
4312 
4313 	if (cs->access == CS_ACCESS_DENIED) {
4314 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4315 		goto out;
4316 	}
4317 
4318 	/*
4319 	 * If there is an unshared filesystem mounted on this vnode,
4320 	 * do not allow to rename objects in this directory.
4321 	 */
4322 	if (vn_ismntpt(odvp)) {
4323 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4324 		goto out;
4325 	}
4326 
4327 	/*
4328 	 * If there is an unshared filesystem mounted on this vnode,
4329 	 * do not allow to rename to this directory.
4330 	 */
4331 	if (vn_ismntpt(ndvp)) {
4332 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4333 		goto out;
4334 	}
4335 
4336 	if (odvp->v_type != VDIR || ndvp->v_type != VDIR) {
4337 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
4338 		goto out;
4339 	}
4340 
4341 	if (cs->saved_exi != cs->exi) {
4342 		*cs->statusp = resp->status = NFS4ERR_XDEV;
4343 		goto out;
4344 	}
4345 
4346 	if (!utf8_dir_verify(&args->oldname)) {
4347 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4348 		goto out;
4349 	}
4350 
4351 	if (!utf8_dir_verify(&args->newname)) {
4352 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4353 		goto out;
4354 	}
4355 
4356 	onm = utf8_to_fn(&args->oldname, &olen, NULL);
4357 	if (onm == NULL) {
4358 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4359 		goto out;
4360 	}
4361 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
4362 	nlen = MAXPATHLEN + 1;
4363 	converted_onm = nfscmd_convname(ca, cs->exi, onm, NFSCMD_CONV_INBOUND,
4364 	    nlen);
4365 
4366 	if (converted_onm == NULL) {
4367 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4368 		kmem_free(onm, olen);
4369 		goto out;
4370 	}
4371 
4372 	nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4373 	if (nnm == NULL) {
4374 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4375 		if (onm != converted_onm)
4376 			kmem_free(converted_onm, MAXPATHLEN + 1);
4377 		kmem_free(onm, olen);
4378 		goto out;
4379 	}
4380 	converted_nnm = nfscmd_convname(ca, cs->exi, nnm, NFSCMD_CONV_INBOUND,
4381 	    MAXPATHLEN  + 1);
4382 
4383 	if (converted_nnm == NULL) {
4384 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4385 		kmem_free(nnm, nlen);
4386 		nnm = NULL;
4387 		if (onm != converted_onm)
4388 			kmem_free(converted_onm, MAXPATHLEN + 1);
4389 		kmem_free(onm, olen);
4390 		goto out;
4391 	}
4392 
4393 
4394 	if (olen > MAXNAMELEN || nlen > MAXNAMELEN) {
4395 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
4396 		kmem_free(onm, olen);
4397 		kmem_free(nnm, nlen);
4398 		goto out;
4399 	}
4400 
4401 
4402 	if (rdonly4(cs->exi, cs->vp, req)) {
4403 		*cs->statusp = resp->status = NFS4ERR_ROFS;
4404 		if (onm != converted_onm)
4405 			kmem_free(converted_onm, MAXPATHLEN + 1);
4406 		kmem_free(onm, olen);
4407 		if (nnm != converted_nnm)
4408 			kmem_free(converted_nnm, MAXPATHLEN + 1);
4409 		kmem_free(nnm, nlen);
4410 		goto out;
4411 	}
4412 
4413 	/* check label of the target dir */
4414 	if (is_system_labeled()) {
4415 		ASSERT(req->rq_label != NULL);
4416 		clabel = req->rq_label;
4417 		DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *,
4418 		    "got client label from request(1)",
4419 		    struct svc_req *, req);
4420 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
4421 			if (!do_rfs_label_check(clabel, ndvp,
4422 			    EQUALITY_CHECK, cs->exi)) {
4423 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
4424 				goto err_out;
4425 			}
4426 		}
4427 	}
4428 
4429 	/*
4430 	 * Is the source a file and have a delegation?
4431 	 * We don't need to acquire va_seq before these lookups, if
4432 	 * it causes an update, cinfo.before will not match, which will
4433 	 * trigger a cache flush even if atomic is TRUE.
4434 	 */
4435 	if (sfp = rfs4_lookup_and_findfile(odvp, converted_onm, &srcvp,
4436 	    &error, cs->cr)) {
4437 		if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE,
4438 		    NULL)) {
4439 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4440 			goto err_out;
4441 		}
4442 	}
4443 
4444 	if (srcvp == NULL) {
4445 		*cs->statusp = resp->status = puterrno4(error);
4446 		if (onm != converted_onm)
4447 			kmem_free(converted_onm, MAXPATHLEN + 1);
4448 		kmem_free(onm, olen);
4449 		if (nnm != converted_nnm)
4450 			kmem_free(converted_onm, MAXPATHLEN + 1);
4451 		kmem_free(nnm, nlen);
4452 		goto out;
4453 	}
4454 
4455 	sfp_rele_grant_hold = 1;
4456 
4457 	/* Does the destination exist and a file and have a delegation? */
4458 	if (fp = rfs4_lookup_and_findfile(ndvp, converted_nnm, &targvp,
4459 	    NULL, cs->cr)) {
4460 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
4461 		    NULL)) {
4462 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4463 			goto err_out;
4464 		}
4465 	}
4466 	fp_rele_grant_hold = 1;
4467 
4468 
4469 	/* Check for NBMAND lock on both source and target */
4470 	if (nbl_need_check(srcvp)) {
4471 		nbl_start_crit(srcvp, RW_READER);
4472 		in_crit_src = 1;
4473 		if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) {
4474 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4475 			goto err_out;
4476 		}
4477 	}
4478 
4479 	if (targvp && nbl_need_check(targvp)) {
4480 		nbl_start_crit(targvp, RW_READER);
4481 		in_crit_targ = 1;
4482 		if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) {
4483 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4484 			goto err_out;
4485 		}
4486 	}
4487 
4488 	/* Get source "before" change value */
4489 	obdva.va_mask = AT_CTIME|AT_SEQ;
4490 	error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL);
4491 	if (!error) {
4492 		nbdva.va_mask = AT_CTIME|AT_SEQ;
4493 		error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL);
4494 	}
4495 	if (error) {
4496 		*cs->statusp = resp->status = puterrno4(error);
4497 		goto err_out;
4498 	}
4499 
4500 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime)
4501 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime)
4502 
4503 	if ((error = VOP_RENAME(odvp, converted_onm, ndvp, converted_nnm,
4504 	    cs->cr, NULL, 0)) == 0 && fp != NULL) {
4505 		struct vattr va;
4506 		vnode_t *tvp;
4507 
4508 		rfs4_dbe_lock(fp->dbe);
4509 		tvp = fp->vp;
4510 		if (tvp)
4511 			VN_HOLD(tvp);
4512 		rfs4_dbe_unlock(fp->dbe);
4513 
4514 		if (tvp) {
4515 			va.va_mask = AT_NLINK;
4516 			if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) &&
4517 			    va.va_nlink == 0) {
4518 				/* The file is gone and so should the state */
4519 				if (in_crit_targ) {
4520 					nbl_end_crit(targvp);
4521 					in_crit_targ = 0;
4522 				}
4523 				rfs4_close_all_state(fp);
4524 			}
4525 			VN_RELE(tvp);
4526 		}
4527 	}
4528 	if (error == 0)
4529 		vn_renamepath(ndvp, srcvp, nnm, nlen - 1);
4530 
4531 	if (in_crit_src)
4532 		nbl_end_crit(srcvp);
4533 	if (srcvp)
4534 		VN_RELE(srcvp);
4535 	if (in_crit_targ)
4536 		nbl_end_crit(targvp);
4537 	if (targvp)
4538 		VN_RELE(targvp);
4539 
4540 	if (sfp) {
4541 		rfs4_clear_dont_grant(sfp);
4542 		rfs4_file_rele(sfp);
4543 	}
4544 	if (fp) {
4545 		rfs4_clear_dont_grant(fp);
4546 		rfs4_file_rele(fp);
4547 	}
4548 
4549 	if (converted_onm != onm)
4550 		kmem_free(converted_onm, MAXPATHLEN + 1);
4551 	kmem_free(onm, olen);
4552 	if (converted_nnm != nnm)
4553 		kmem_free(converted_nnm, MAXPATHLEN + 1);
4554 	kmem_free(nnm, nlen);
4555 
4556 	/*
4557 	 * Get the initial "after" sequence number, if it fails, set to zero
4558 	 */
4559 	oidva.va_mask = AT_SEQ;
4560 	if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL))
4561 		oidva.va_seq = 0;
4562 
4563 	nidva.va_mask = AT_SEQ;
4564 	if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL))
4565 		nidva.va_seq = 0;
4566 
4567 	/*
4568 	 * Force modified data and metadata out to stable storage.
4569 	 */
4570 	(void) VOP_FSYNC(odvp, 0, cs->cr, NULL);
4571 	(void) VOP_FSYNC(ndvp, 0, cs->cr, NULL);
4572 
4573 	if (error) {
4574 		*cs->statusp = resp->status = puterrno4(error);
4575 		goto out;
4576 	}
4577 
4578 	/*
4579 	 * Get "after" change values, if it fails, simply return the
4580 	 * before value.
4581 	 */
4582 	oadva.va_mask = AT_CTIME|AT_SEQ;
4583 	if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) {
4584 		oadva.va_ctime = obdva.va_ctime;
4585 		oadva.va_seq = 0;
4586 	}
4587 
4588 	nadva.va_mask = AT_CTIME|AT_SEQ;
4589 	if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) {
4590 		nadva.va_ctime = nbdva.va_ctime;
4591 		nadva.va_seq = 0;
4592 	}
4593 
4594 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime)
4595 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime)
4596 
4597 	/*
4598 	 * The cinfo.atomic = TRUE only if we have
4599 	 * non-zero va_seq's, and it has incremented by exactly one
4600 	 * during the VOP_RENAME and it didn't change during the VOP_FSYNC.
4601 	 */
4602 	if (obdva.va_seq && oidva.va_seq && oadva.va_seq &&
4603 	    oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq)
4604 		resp->source_cinfo.atomic = TRUE;
4605 	else
4606 		resp->source_cinfo.atomic = FALSE;
4607 
4608 	if (nbdva.va_seq && nidva.va_seq && nadva.va_seq &&
4609 	    nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq)
4610 		resp->target_cinfo.atomic = TRUE;
4611 	else
4612 		resp->target_cinfo.atomic = FALSE;
4613 
4614 #ifdef	VOLATILE_FH_TEST
4615 	{
4616 	extern void add_volrnm_fh(struct exportinfo *, vnode_t *);
4617 
4618 	/*
4619 	 * Add the renamed file handle to the volatile rename list
4620 	 */
4621 	if (cs->exi->exi_export.ex_flags & EX_VOLRNM) {
4622 		/* file handles may expire on rename */
4623 		vnode_t *vp;
4624 
4625 		nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4626 		/*
4627 		 * Already know that nnm will be a valid string
4628 		 */
4629 		error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr,
4630 		    NULL, NULL, NULL);
4631 		kmem_free(nnm, nlen);
4632 		if (!error) {
4633 			add_volrnm_fh(cs->exi, vp);
4634 			VN_RELE(vp);
4635 		}
4636 	}
4637 	}
4638 #endif	/* VOLATILE_FH_TEST */
4639 
4640 	*cs->statusp = resp->status = NFS4_OK;
4641 out:
4642 	DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs,
4643 	    RENAME4res *, resp);
4644 	return;
4645 
4646 err_out:
4647 	if (onm != converted_onm)
4648 		kmem_free(converted_onm, MAXPATHLEN + 1);
4649 	if (onm != NULL)
4650 		kmem_free(onm, olen);
4651 	if (nnm != converted_nnm)
4652 		kmem_free(converted_nnm, MAXPATHLEN + 1);
4653 	if (nnm != NULL)
4654 		kmem_free(nnm, nlen);
4655 
4656 	if (in_crit_src) nbl_end_crit(srcvp);
4657 	if (in_crit_targ) nbl_end_crit(targvp);
4658 	if (targvp) VN_RELE(targvp);
4659 	if (srcvp) VN_RELE(srcvp);
4660 	if (sfp) {
4661 		if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp);
4662 		rfs4_file_rele(sfp);
4663 	}
4664 	if (fp) {
4665 		if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp);
4666 		rfs4_file_rele(fp);
4667 	}
4668 
4669 	DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs,
4670 	    RENAME4res *, resp);
4671 }
4672 
4673 /* ARGSUSED */
4674 static void
4675 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4676 	struct compound_state *cs)
4677 {
4678 	RENEW4args *args = &argop->nfs_argop4_u.oprenew;
4679 	RENEW4res *resp = &resop->nfs_resop4_u.oprenew;
4680 	rfs4_client_t *cp;
4681 
4682 	DTRACE_NFSV4_2(op__renew__start, struct compound_state *, cs,
4683 	    RENEW4args *, args);
4684 
4685 	if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) {
4686 		*cs->statusp = resp->status =
4687 		    rfs4_check_clientid(&args->clientid, 0);
4688 		goto out;
4689 	}
4690 
4691 	if (rfs4_lease_expired(cp)) {
4692 		rfs4_client_rele(cp);
4693 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
4694 		goto out;
4695 	}
4696 
4697 	rfs4_update_lease(cp);
4698 
4699 	mutex_enter(cp->cbinfo.cb_lock);
4700 	if (cp->cbinfo.cb_notified_of_cb_path_down == FALSE) {
4701 		cp->cbinfo.cb_notified_of_cb_path_down = TRUE;
4702 		*cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN;
4703 	} else {
4704 		*cs->statusp = resp->status = NFS4_OK;
4705 	}
4706 	mutex_exit(cp->cbinfo.cb_lock);
4707 
4708 	rfs4_client_rele(cp);
4709 
4710 out:
4711 	DTRACE_NFSV4_2(op__renew__done, struct compound_state *, cs,
4712 	    RENEW4res *, resp);
4713 }
4714 
4715 /* ARGSUSED */
4716 static void
4717 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
4718 	struct compound_state *cs)
4719 {
4720 	RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh;
4721 
4722 	DTRACE_NFSV4_1(op__restorefh__start, struct compound_state *, cs);
4723 
4724 	/* No need to check cs->access - we are not accessing any object */
4725 	if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) {
4726 		*cs->statusp = resp->status = NFS4ERR_RESTOREFH;
4727 		goto out;
4728 	}
4729 	if (cs->vp != NULL) {
4730 		VN_RELE(cs->vp);
4731 	}
4732 	cs->vp = cs->saved_vp;
4733 	cs->saved_vp = NULL;
4734 	cs->exi = cs->saved_exi;
4735 	nfs_fh4_copy(&cs->saved_fh, &cs->fh);
4736 	*cs->statusp = resp->status = NFS4_OK;
4737 	cs->deleg = FALSE;
4738 
4739 out:
4740 	DTRACE_NFSV4_2(op__restorefh__done, struct compound_state *, cs,
4741 	    RESTOREFH4res *, resp);
4742 }
4743 
4744 /* ARGSUSED */
4745 static void
4746 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4747 	struct compound_state *cs)
4748 {
4749 	SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh;
4750 
4751 	DTRACE_NFSV4_1(op__savefh__start, struct compound_state *, cs);
4752 
4753 	/* No need to check cs->access - we are not accessing any object */
4754 	if (cs->vp == NULL) {
4755 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4756 		goto out;
4757 	}
4758 	if (cs->saved_vp != NULL) {
4759 		VN_RELE(cs->saved_vp);
4760 	}
4761 	cs->saved_vp = cs->vp;
4762 	VN_HOLD(cs->saved_vp);
4763 	cs->saved_exi = cs->exi;
4764 	/*
4765 	 * since SAVEFH is fairly rare, don't alloc space for its fh
4766 	 * unless necessary.
4767 	 */
4768 	if (cs->saved_fh.nfs_fh4_val == NULL) {
4769 		cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP);
4770 	}
4771 	nfs_fh4_copy(&cs->fh, &cs->saved_fh);
4772 	*cs->statusp = resp->status = NFS4_OK;
4773 
4774 out:
4775 	DTRACE_NFSV4_2(op__savefh__done, struct compound_state *, cs,
4776 	    SAVEFH4res *, resp);
4777 }
4778 
4779 /*
4780  * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to
4781  * return the bitmap of attrs that were set successfully. It is also
4782  * called by Verify/Nverify to test the vattr/vfsstat attrs. It should
4783  * always be called only after rfs4_do_set_attrs().
4784  *
4785  * Verify that the attributes are same as the expected ones. sargp->vap
4786  * and sargp->sbp contain the input attributes as translated from fattr4.
4787  *
4788  * This function verifies only the attrs that correspond to a vattr or
4789  * vfsstat struct. That is because of the extra step needed to get the
4790  * corresponding system structs. Other attributes have already been set or
4791  * verified by do_rfs4_set_attrs.
4792  *
4793  * Return 0 if all attrs match, -1 if some don't, error if error processing.
4794  */
4795 static int
4796 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp,
4797 	bitmap4 *resp, struct nfs4_ntov_table *ntovp)
4798 {
4799 	int error, ret_error = 0;
4800 	int i, k;
4801 	uint_t sva_mask = sargp->vap->va_mask;
4802 	uint_t vbit;
4803 	union nfs4_attr_u *na;
4804 	uint8_t *amap;
4805 	bool_t getsb = ntovp->vfsstat;
4806 
4807 	if (sva_mask != 0) {
4808 		/*
4809 		 * Okay to overwrite sargp->vap because we verify based
4810 		 * on the incoming values.
4811 		 */
4812 		ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0,
4813 		    sargp->cs->cr, NULL);
4814 		if (ret_error) {
4815 			if (resp == NULL)
4816 				return (ret_error);
4817 			/*
4818 			 * Must return bitmap of successful attrs
4819 			 */
4820 			sva_mask = 0;	/* to prevent checking vap later */
4821 		} else {
4822 			/*
4823 			 * Some file systems clobber va_mask. it is probably
4824 			 * wrong of them to do so, nonethless we practice
4825 			 * defensive coding.
4826 			 * See bug id 4276830.
4827 			 */
4828 			sargp->vap->va_mask = sva_mask;
4829 		}
4830 	}
4831 
4832 	if (getsb) {
4833 		/*
4834 		 * Now get the superblock and loop on the bitmap, as there is
4835 		 * no simple way of translating from superblock to bitmap4.
4836 		 */
4837 		ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp);
4838 		if (ret_error) {
4839 			if (resp == NULL)
4840 				goto errout;
4841 			getsb = FALSE;
4842 		}
4843 	}
4844 
4845 	/*
4846 	 * Now loop and verify each attribute which getattr returned
4847 	 * whether it's the same as the input.
4848 	 */
4849 	if (resp == NULL && !getsb && (sva_mask == 0))
4850 		goto errout;
4851 
4852 	na = ntovp->na;
4853 	amap = ntovp->amap;
4854 	k = 0;
4855 	for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) {
4856 		k = *amap;
4857 		ASSERT(nfs4_ntov_map[k].nval == k);
4858 		vbit = nfs4_ntov_map[k].vbit;
4859 
4860 		/*
4861 		 * If vattr attribute but VOP_GETATTR failed, or it's
4862 		 * superblock attribute but VFS_STATVFS failed, skip
4863 		 */
4864 		if (vbit) {
4865 			if ((vbit & sva_mask) == 0)
4866 				continue;
4867 		} else if (!(getsb && nfs4_ntov_map[k].vfsstat)) {
4868 			continue;
4869 		}
4870 		error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na);
4871 		if (resp != NULL) {
4872 			if (error)
4873 				ret_error = -1;	/* not all match */
4874 			else	/* update response bitmap */
4875 				*resp |= nfs4_ntov_map[k].fbit;
4876 			continue;
4877 		}
4878 		if (error) {
4879 			ret_error = -1;	/* not all match */
4880 			break;
4881 		}
4882 	}
4883 errout:
4884 	return (ret_error);
4885 }
4886 
4887 /*
4888  * Decode the attribute to be set/verified. If the attr requires a sys op
4889  * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't
4890  * call the sv_getit function for it, because the sys op hasn't yet been done.
4891  * Return 0 for success, error code if failed.
4892  *
4893  * Note: the decoded arg is not freed here but in nfs4_ntov_table_free.
4894  */
4895 static int
4896 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp,
4897 	int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap)
4898 {
4899 	int error = 0;
4900 	bool_t set_later;
4901 
4902 	sargp->vap->va_mask |= nfs4_ntov_map[k].vbit;
4903 
4904 	if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) {
4905 		set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat;
4906 		/*
4907 		 * don't verify yet if a vattr or sb dependent attr,
4908 		 * because we don't have their sys values yet.
4909 		 * Will be done later.
4910 		 */
4911 		if (! (set_later && (cmd == NFS4ATTR_VERIT))) {
4912 			/*
4913 			 * ACLs are a special case, since setting the MODE
4914 			 * conflicts with setting the ACL.  We delay setting
4915 			 * the ACL until all other attributes have been set.
4916 			 * The ACL gets set in do_rfs4_op_setattr().
4917 			 */
4918 			if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) {
4919 				error = (*nfs4_ntov_map[k].sv_getit)(cmd,
4920 				    sargp, nap);
4921 				if (error) {
4922 					xdr_free(nfs4_ntov_map[k].xfunc,
4923 					    (caddr_t)nap);
4924 				}
4925 			}
4926 		}
4927 	} else {
4928 #ifdef  DEBUG
4929 		cmn_err(CE_NOTE, "decode_fattr4_attr: error "
4930 		    "decoding attribute %d\n", k);
4931 #endif
4932 		error = EINVAL;
4933 	}
4934 	if (!error && resp_bval && !set_later) {
4935 		*resp_bval |= nfs4_ntov_map[k].fbit;
4936 	}
4937 
4938 	return (error);
4939 }
4940 
4941 /*
4942  * Set vattr based on incoming fattr4 attrs - used by setattr.
4943  * Set response mask. Ignore any values that are not writable vattr attrs.
4944  */
4945 static nfsstat4
4946 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
4947 		struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp,
4948 		nfs4_attr_cmd_t cmd)
4949 {
4950 	int error = 0;
4951 	int i;
4952 	char *attrs = fattrp->attrlist4;
4953 	uint32_t attrslen = fattrp->attrlist4_len;
4954 	XDR xdr;
4955 	nfsstat4 status = NFS4_OK;
4956 	vnode_t *vp = cs->vp;
4957 	union nfs4_attr_u *na;
4958 	uint8_t *amap;
4959 
4960 #ifndef lint
4961 	/*
4962 	 * Make sure that maximum attribute number can be expressed as an
4963 	 * 8 bit quantity.
4964 	 */
4965 	ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1));
4966 #endif
4967 
4968 	if (vp == NULL) {
4969 		if (resp)
4970 			*resp = 0;
4971 		return (NFS4ERR_NOFILEHANDLE);
4972 	}
4973 	if (cs->access == CS_ACCESS_DENIED) {
4974 		if (resp)
4975 			*resp = 0;
4976 		return (NFS4ERR_ACCESS);
4977 	}
4978 
4979 	sargp->op = cmd;
4980 	sargp->cs = cs;
4981 	sargp->flag = 0;	/* may be set later */
4982 	sargp->vap->va_mask = 0;
4983 	sargp->rdattr_error = NFS4_OK;
4984 	sargp->rdattr_error_req = FALSE;
4985 	/* sargp->sbp is set by the caller */
4986 
4987 	xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE);
4988 
4989 	na = ntovp->na;
4990 	amap = ntovp->amap;
4991 
4992 	/*
4993 	 * The following loop iterates on the nfs4_ntov_map checking
4994 	 * if the fbit is set in the requested bitmap.
4995 	 * If set then we process the arguments using the
4996 	 * rfs4_fattr4 conversion functions to populate the setattr
4997 	 * vattr and va_mask. Any settable attrs that are not using vattr
4998 	 * will be set in this loop.
4999 	 */
5000 	for (i = 0; i < nfs4_ntov_map_size; i++) {
5001 		if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) {
5002 			continue;
5003 		}
5004 		/*
5005 		 * If setattr, must be a writable attr.
5006 		 * If verify/nverify, must be a readable attr.
5007 		 */
5008 		if ((error = (*nfs4_ntov_map[i].sv_getit)(
5009 		    NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) {
5010 			/*
5011 			 * Client tries to set/verify an
5012 			 * unsupported attribute, tries to set
5013 			 * a read only attr or verify a write
5014 			 * only one - error!
5015 			 */
5016 			break;
5017 		}
5018 		/*
5019 		 * Decode the attribute to set/verify
5020 		 */
5021 		error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval,
5022 		    &xdr, resp ? resp : NULL, na);
5023 		if (error)
5024 			break;
5025 		*amap++ = (uint8_t)nfs4_ntov_map[i].nval;
5026 		na++;
5027 		(ntovp->attrcnt)++;
5028 		if (nfs4_ntov_map[i].vfsstat)
5029 			ntovp->vfsstat = TRUE;
5030 	}
5031 
5032 	if (error != 0)
5033 		status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP :
5034 		    puterrno4(error));
5035 	/* xdrmem_destroy(&xdrs); */	/* NO-OP */
5036 	return (status);
5037 }
5038 
5039 static nfsstat4
5040 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
5041 		stateid4 *stateid)
5042 {
5043 	int error = 0;
5044 	struct nfs4_svgetit_arg sarg;
5045 	bool_t trunc;
5046 
5047 	nfsstat4 status = NFS4_OK;
5048 	cred_t *cr = cs->cr;
5049 	vnode_t *vp = cs->vp;
5050 	struct nfs4_ntov_table ntov;
5051 	struct statvfs64 sb;
5052 	struct vattr bva;
5053 	struct flock64 bf;
5054 	int in_crit = 0;
5055 	uint_t saved_mask = 0;
5056 	caller_context_t ct;
5057 
5058 	*resp = 0;
5059 	sarg.sbp = &sb;
5060 	nfs4_ntov_table_init(&ntov);
5061 	status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov,
5062 	    NFS4ATTR_SETIT);
5063 	if (status != NFS4_OK) {
5064 		/*
5065 		 * failed set attrs
5066 		 */
5067 		goto done;
5068 	}
5069 	if ((sarg.vap->va_mask == 0) &&
5070 	    (! (fattrp->attrmask & FATTR4_ACL_MASK))) {
5071 		/*
5072 		 * no further work to be done
5073 		 */
5074 		goto done;
5075 	}
5076 
5077 	/*
5078 	 * If we got a request to set the ACL and the MODE, only
5079 	 * allow changing VSUID, VSGID, and VSVTX.  Attempting
5080 	 * to change any other bits, along with setting an ACL,
5081 	 * gives NFS4ERR_INVAL.
5082 	 */
5083 	if ((fattrp->attrmask & FATTR4_ACL_MASK) &&
5084 	    (fattrp->attrmask & FATTR4_MODE_MASK)) {
5085 		vattr_t va;
5086 
5087 		va.va_mask = AT_MODE;
5088 		error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL);
5089 		if (error) {
5090 			status = puterrno4(error);
5091 			goto done;
5092 		}
5093 		if ((sarg.vap->va_mode ^ va.va_mode) &
5094 		    ~(VSUID | VSGID | VSVTX)) {
5095 			status = NFS4ERR_INVAL;
5096 			goto done;
5097 		}
5098 	}
5099 
5100 	/* Check stateid only if size has been set */
5101 	if (sarg.vap->va_mask & AT_SIZE) {
5102 		trunc = (sarg.vap->va_size == 0);
5103 		status = rfs4_check_stateid(FWRITE, cs->vp, stateid,
5104 		    trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct);
5105 		if (status != NFS4_OK)
5106 			goto done;
5107 	} else {
5108 		ct.cc_sysid = 0;
5109 		ct.cc_pid = 0;
5110 		ct.cc_caller_id = nfs4_srv_caller_id;
5111 		ct.cc_flags = CC_DONTBLOCK;
5112 	}
5113 
5114 	/* XXX start of possible race with delegations */
5115 
5116 	/*
5117 	 * We need to specially handle size changes because it is
5118 	 * possible for the client to create a file with read-only
5119 	 * modes, but with the file opened for writing. If the client
5120 	 * then tries to set the file size, e.g. ftruncate(3C),
5121 	 * fcntl(F_FREESP), the normal access checking done in
5122 	 * VOP_SETATTR would prevent the client from doing it even though
5123 	 * it should be allowed to do so.  To get around this, we do the
5124 	 * access checking for ourselves and use VOP_SPACE which doesn't
5125 	 * do the access checking.
5126 	 * Also the client should not be allowed to change the file
5127 	 * size if there is a conflicting non-blocking mandatory lock in
5128 	 * the region of the change.
5129 	 */
5130 	if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) {
5131 		u_offset_t offset;
5132 		ssize_t length;
5133 
5134 		/*
5135 		 * ufs_setattr clears AT_SIZE from vap->va_mask, but
5136 		 * before returning, sarg.vap->va_mask is used to
5137 		 * generate the setattr reply bitmap.  We also clear
5138 		 * AT_SIZE below before calling VOP_SPACE.  For both
5139 		 * of these cases, the va_mask needs to be saved here
5140 		 * and restored after calling VOP_SETATTR.
5141 		 */
5142 		saved_mask = sarg.vap->va_mask;
5143 
5144 		/*
5145 		 * Check any possible conflict due to NBMAND locks.
5146 		 * Get into critical region before VOP_GETATTR, so the
5147 		 * size attribute is valid when checking conflicts.
5148 		 */
5149 		if (nbl_need_check(vp)) {
5150 			nbl_start_crit(vp, RW_READER);
5151 			in_crit = 1;
5152 		}
5153 
5154 		bva.va_mask = AT_UID|AT_SIZE;
5155 		if (error = VOP_GETATTR(vp, &bva, 0, cr, &ct)) {
5156 			status = puterrno4(error);
5157 			goto done;
5158 		}
5159 
5160 		if (in_crit) {
5161 			if (sarg.vap->va_size < bva.va_size) {
5162 				offset = sarg.vap->va_size;
5163 				length = bva.va_size - sarg.vap->va_size;
5164 			} else {
5165 				offset = bva.va_size;
5166 				length = sarg.vap->va_size - bva.va_size;
5167 			}
5168 			if (nbl_conflict(vp, NBL_WRITE, offset, length, 0,
5169 			    &ct)) {
5170 				status = NFS4ERR_LOCKED;
5171 				goto done;
5172 			}
5173 		}
5174 
5175 		if (crgetuid(cr) == bva.va_uid) {
5176 			sarg.vap->va_mask &= ~AT_SIZE;
5177 			bf.l_type = F_WRLCK;
5178 			bf.l_whence = 0;
5179 			bf.l_start = (off64_t)sarg.vap->va_size;
5180 			bf.l_len = 0;
5181 			bf.l_sysid = 0;
5182 			bf.l_pid = 0;
5183 			error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE,
5184 			    (offset_t)sarg.vap->va_size, cr, &ct);
5185 		}
5186 	}
5187 
5188 	if (!error && sarg.vap->va_mask != 0)
5189 		error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct);
5190 
5191 	/* restore va_mask -- ufs_setattr clears AT_SIZE */
5192 	if (saved_mask & AT_SIZE)
5193 		sarg.vap->va_mask |= AT_SIZE;
5194 
5195 	/*
5196 	 * If an ACL was being set, it has been delayed until now,
5197 	 * in order to set the mode (via the VOP_SETATTR() above) first.
5198 	 */
5199 	if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) {
5200 		int i;
5201 
5202 		for (i = 0; i < NFS4_MAXNUM_ATTRS; i++)
5203 			if (ntov.amap[i] == FATTR4_ACL)
5204 				break;
5205 		if (i < NFS4_MAXNUM_ATTRS) {
5206 			error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)(
5207 			    NFS4ATTR_SETIT, &sarg, &ntov.na[i]);
5208 			if (error == 0) {
5209 				*resp |= FATTR4_ACL_MASK;
5210 			} else if (error == ENOTSUP) {
5211 				(void) rfs4_verify_attr(&sarg, resp, &ntov);
5212 				status = NFS4ERR_ATTRNOTSUPP;
5213 				goto done;
5214 			}
5215 		} else {
5216 			NFS4_DEBUG(rfs4_debug,
5217 			    (CE_NOTE, "do_rfs4_op_setattr: "
5218 			    "unable to find ACL in fattr4"));
5219 			error = EINVAL;
5220 		}
5221 	}
5222 
5223 	if (error) {
5224 		/* check if a monitor detected a delegation conflict */
5225 		if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK))
5226 			status = NFS4ERR_DELAY;
5227 		else
5228 			status = puterrno4(error);
5229 
5230 		/*
5231 		 * Set the response bitmap when setattr failed.
5232 		 * If VOP_SETATTR partially succeeded, test by doing a
5233 		 * VOP_GETATTR on the object and comparing the data
5234 		 * to the setattr arguments.
5235 		 */
5236 		(void) rfs4_verify_attr(&sarg, resp, &ntov);
5237 	} else {
5238 		/*
5239 		 * Force modified metadata out to stable storage.
5240 		 */
5241 		(void) VOP_FSYNC(vp, FNODSYNC, cr, &ct);
5242 		/*
5243 		 * Set response bitmap
5244 		 */
5245 		nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp);
5246 	}
5247 
5248 /* Return early and already have a NFSv4 error */
5249 done:
5250 	/*
5251 	 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr
5252 	 * conversion sets both readable and writeable NFS4 attrs
5253 	 * for AT_MTIME and AT_ATIME.  The line below masks out
5254 	 * unrequested attrs from the setattr result bitmap.  This
5255 	 * is placed after the done: label to catch the ATTRNOTSUP
5256 	 * case.
5257 	 */
5258 	*resp &= fattrp->attrmask;
5259 
5260 	if (in_crit)
5261 		nbl_end_crit(vp);
5262 
5263 	nfs4_ntov_table_free(&ntov, &sarg);
5264 
5265 	return (status);
5266 }
5267 
5268 /* ARGSUSED */
5269 static void
5270 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5271 	struct compound_state *cs)
5272 {
5273 	SETATTR4args *args = &argop->nfs_argop4_u.opsetattr;
5274 	SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr;
5275 	bslabel_t *clabel;
5276 
5277 	DTRACE_NFSV4_2(op__setattr__start, struct compound_state *, cs,
5278 	    SETATTR4args *, args);
5279 
5280 	if (cs->vp == NULL) {
5281 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5282 		goto out;
5283 	}
5284 
5285 	/*
5286 	 * If there is an unshared filesystem mounted on this vnode,
5287 	 * do not allow to setattr on this vnode.
5288 	 */
5289 	if (vn_ismntpt(cs->vp)) {
5290 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5291 		goto out;
5292 	}
5293 
5294 	resp->attrsset = 0;
5295 
5296 	if (rdonly4(cs->exi, cs->vp, req)) {
5297 		*cs->statusp = resp->status = NFS4ERR_ROFS;
5298 		goto out;
5299 	}
5300 
5301 	/* check label before setting attributes */
5302 	if (is_system_labeled()) {
5303 		ASSERT(req->rq_label != NULL);
5304 		clabel = req->rq_label;
5305 		DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *,
5306 		    "got client label from request(1)",
5307 		    struct svc_req *, req);
5308 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
5309 			if (!do_rfs_label_check(clabel, cs->vp,
5310 			    EQUALITY_CHECK, cs->exi)) {
5311 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
5312 				goto out;
5313 			}
5314 		}
5315 	}
5316 
5317 	*cs->statusp = resp->status =
5318 	    do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs,
5319 	    &args->stateid);
5320 
5321 out:
5322 	DTRACE_NFSV4_2(op__setattr__done, struct compound_state *, cs,
5323 	    SETATTR4res *, resp);
5324 }
5325 
5326 /* ARGSUSED */
5327 static void
5328 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5329 	struct compound_state *cs)
5330 {
5331 	/*
5332 	 * verify and nverify are exactly the same, except that nverify
5333 	 * succeeds when some argument changed, and verify succeeds when
5334 	 * when none changed.
5335 	 */
5336 
5337 	VERIFY4args  *args = &argop->nfs_argop4_u.opverify;
5338 	VERIFY4res *resp = &resop->nfs_resop4_u.opverify;
5339 
5340 	int error;
5341 	struct nfs4_svgetit_arg sarg;
5342 	struct statvfs64 sb;
5343 	struct nfs4_ntov_table ntov;
5344 
5345 	DTRACE_NFSV4_2(op__verify__start, struct compound_state *, cs,
5346 	    VERIFY4args *, args);
5347 
5348 	if (cs->vp == NULL) {
5349 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5350 		goto out;
5351 	}
5352 
5353 	sarg.sbp = &sb;
5354 	nfs4_ntov_table_init(&ntov);
5355 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
5356 	    &sarg, &ntov, NFS4ATTR_VERIT);
5357 	if (resp->status != NFS4_OK) {
5358 		/*
5359 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
5360 		 * so could return -1 for "no match".
5361 		 */
5362 		if (resp->status == -1)
5363 			resp->status = NFS4ERR_NOT_SAME;
5364 		goto done;
5365 	}
5366 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
5367 	switch (error) {
5368 	case 0:
5369 		resp->status = NFS4_OK;
5370 		break;
5371 	case -1:
5372 		resp->status = NFS4ERR_NOT_SAME;
5373 		break;
5374 	default:
5375 		resp->status = puterrno4(error);
5376 		break;
5377 	}
5378 done:
5379 	*cs->statusp = resp->status;
5380 	nfs4_ntov_table_free(&ntov, &sarg);
5381 out:
5382 	DTRACE_NFSV4_2(op__verify__done, struct compound_state *, cs,
5383 	    VERIFY4res *, resp);
5384 }
5385 
5386 /* ARGSUSED */
5387 static void
5388 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5389 	struct compound_state *cs)
5390 {
5391 	/*
5392 	 * verify and nverify are exactly the same, except that nverify
5393 	 * succeeds when some argument changed, and verify succeeds when
5394 	 * when none changed.
5395 	 */
5396 
5397 	NVERIFY4args  *args = &argop->nfs_argop4_u.opnverify;
5398 	NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify;
5399 
5400 	int error;
5401 	struct nfs4_svgetit_arg sarg;
5402 	struct statvfs64 sb;
5403 	struct nfs4_ntov_table ntov;
5404 
5405 	DTRACE_NFSV4_2(op__nverify__start, struct compound_state *, cs,
5406 	    NVERIFY4args *, args);
5407 
5408 	if (cs->vp == NULL) {
5409 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5410 		DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs,
5411 		    NVERIFY4res *, resp);
5412 		return;
5413 	}
5414 	sarg.sbp = &sb;
5415 	nfs4_ntov_table_init(&ntov);
5416 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
5417 	    &sarg, &ntov, NFS4ATTR_VERIT);
5418 	if (resp->status != NFS4_OK) {
5419 		/*
5420 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
5421 		 * so could return -1 for "no match".
5422 		 */
5423 		if (resp->status == -1)
5424 			resp->status = NFS4_OK;
5425 		goto done;
5426 	}
5427 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
5428 	switch (error) {
5429 	case 0:
5430 		resp->status = NFS4ERR_SAME;
5431 		break;
5432 	case -1:
5433 		resp->status = NFS4_OK;
5434 		break;
5435 	default:
5436 		resp->status = puterrno4(error);
5437 		break;
5438 	}
5439 done:
5440 	*cs->statusp = resp->status;
5441 	nfs4_ntov_table_free(&ntov, &sarg);
5442 
5443 	DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs,
5444 	    NVERIFY4res *, resp);
5445 }
5446 
5447 /*
5448  * XXX - This should live in an NFS header file.
5449  */
5450 #define	MAX_IOVECS	12
5451 
5452 /* ARGSUSED */
5453 static void
5454 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5455 	struct compound_state *cs)
5456 {
5457 	WRITE4args *args = &argop->nfs_argop4_u.opwrite;
5458 	WRITE4res *resp = &resop->nfs_resop4_u.opwrite;
5459 	int error;
5460 	vnode_t *vp;
5461 	struct vattr bva;
5462 	u_offset_t rlimit;
5463 	struct uio uio;
5464 	struct iovec iov[MAX_IOVECS];
5465 	struct iovec *iovp;
5466 	int iovcnt;
5467 	int ioflag;
5468 	cred_t *savecred, *cr;
5469 	bool_t *deleg = &cs->deleg;
5470 	nfsstat4 stat;
5471 	int in_crit = 0;
5472 	caller_context_t ct;
5473 
5474 	DTRACE_NFSV4_2(op__write__start, struct compound_state *, cs,
5475 	    WRITE4args *, args);
5476 
5477 	vp = cs->vp;
5478 	if (vp == NULL) {
5479 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5480 		goto out;
5481 	}
5482 	if (cs->access == CS_ACCESS_DENIED) {
5483 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5484 		goto out;
5485 	}
5486 
5487 	cr = cs->cr;
5488 
5489 	if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE,
5490 	    deleg, TRUE, &ct)) != NFS4_OK) {
5491 		*cs->statusp = resp->status = stat;
5492 		goto out;
5493 	}
5494 
5495 	/*
5496 	 * We have to enter the critical region before calling VOP_RWLOCK
5497 	 * to avoid a deadlock with ufs.
5498 	 */
5499 	if (nbl_need_check(vp)) {
5500 		nbl_start_crit(vp, RW_READER);
5501 		in_crit = 1;
5502 		if (nbl_conflict(vp, NBL_WRITE,
5503 		    args->offset, args->data_len, 0, &ct)) {
5504 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
5505 			goto out;
5506 		}
5507 	}
5508 
5509 	bva.va_mask = AT_MODE | AT_UID;
5510 	error = VOP_GETATTR(vp, &bva, 0, cr, &ct);
5511 
5512 	/*
5513 	 * If we can't get the attributes, then we can't do the
5514 	 * right access checking.  So, we'll fail the request.
5515 	 */
5516 	if (error) {
5517 		*cs->statusp = resp->status = puterrno4(error);
5518 		goto out;
5519 	}
5520 
5521 	if (rdonly4(cs->exi, cs->vp, req)) {
5522 		*cs->statusp = resp->status = NFS4ERR_ROFS;
5523 		goto out;
5524 	}
5525 
5526 	if (vp->v_type != VREG) {
5527 		*cs->statusp = resp->status =
5528 		    ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
5529 		goto out;
5530 	}
5531 
5532 	if (crgetuid(cr) != bva.va_uid &&
5533 	    (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) {
5534 		*cs->statusp = resp->status = puterrno4(error);
5535 		goto out;
5536 	}
5537 
5538 	if (MANDLOCK(vp, bva.va_mode)) {
5539 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5540 		goto out;
5541 	}
5542 
5543 	if (args->data_len == 0) {
5544 		*cs->statusp = resp->status = NFS4_OK;
5545 		resp->count = 0;
5546 		resp->committed = args->stable;
5547 		resp->writeverf = Write4verf;
5548 		goto out;
5549 	}
5550 
5551 	if (args->mblk != NULL) {
5552 		mblk_t *m;
5553 		uint_t bytes, round_len;
5554 
5555 		iovcnt = 0;
5556 		bytes = 0;
5557 		round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT);
5558 		for (m = args->mblk;
5559 		    m != NULL && bytes < round_len;
5560 		    m = m->b_cont) {
5561 			iovcnt++;
5562 			bytes += MBLKL(m);
5563 		}
5564 #ifdef DEBUG
5565 		/* should have ended on an mblk boundary */
5566 		if (bytes != round_len) {
5567 			printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n",
5568 			    bytes, round_len, args->data_len);
5569 			printf("args=%p, args->mblk=%p, m=%p", (void *)args,
5570 			    (void *)args->mblk, (void *)m);
5571 			ASSERT(bytes == round_len);
5572 		}
5573 #endif
5574 		if (iovcnt <= MAX_IOVECS) {
5575 			iovp = iov;
5576 		} else {
5577 			iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP);
5578 		}
5579 		mblk_to_iov(args->mblk, iovcnt, iovp);
5580 	} else if (args->rlist != NULL) {
5581 		iovcnt = 1;
5582 		iovp = iov;
5583 		iovp->iov_base = (char *)((args->rlist)->u.c_daddr3);
5584 		iovp->iov_len = args->data_len;
5585 	} else {
5586 		iovcnt = 1;
5587 		iovp = iov;
5588 		iovp->iov_base = args->data_val;
5589 		iovp->iov_len = args->data_len;
5590 	}
5591 
5592 	uio.uio_iov = iovp;
5593 	uio.uio_iovcnt = iovcnt;
5594 
5595 	uio.uio_segflg = UIO_SYSSPACE;
5596 	uio.uio_extflg = UIO_COPY_DEFAULT;
5597 	uio.uio_loffset = args->offset;
5598 	uio.uio_resid = args->data_len;
5599 	uio.uio_llimit = curproc->p_fsz_ctl;
5600 	rlimit = uio.uio_llimit - args->offset;
5601 	if (rlimit < (u_offset_t)uio.uio_resid)
5602 		uio.uio_resid = (int)rlimit;
5603 
5604 	if (args->stable == UNSTABLE4)
5605 		ioflag = 0;
5606 	else if (args->stable == FILE_SYNC4)
5607 		ioflag = FSYNC;
5608 	else if (args->stable == DATA_SYNC4)
5609 		ioflag = FDSYNC;
5610 	else {
5611 		if (iovp != iov)
5612 			kmem_free(iovp, sizeof (*iovp) * iovcnt);
5613 		*cs->statusp = resp->status = NFS4ERR_INVAL;
5614 		goto out;
5615 	}
5616 
5617 	/*
5618 	 * We're changing creds because VM may fault and we need
5619 	 * the cred of the current thread to be used if quota
5620 	 * checking is enabled.
5621 	 */
5622 	savecred = curthread->t_cred;
5623 	curthread->t_cred = cr;
5624 	error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct);
5625 	curthread->t_cred = savecred;
5626 
5627 	if (iovp != iov)
5628 		kmem_free(iovp, sizeof (*iovp) * iovcnt);
5629 
5630 	if (error) {
5631 		*cs->statusp = resp->status = puterrno4(error);
5632 		goto out;
5633 	}
5634 
5635 	*cs->statusp = resp->status = NFS4_OK;
5636 	resp->count = args->data_len - uio.uio_resid;
5637 
5638 	if (ioflag == 0)
5639 		resp->committed = UNSTABLE4;
5640 	else
5641 		resp->committed = FILE_SYNC4;
5642 
5643 	resp->writeverf = Write4verf;
5644 
5645 out:
5646 	if (in_crit)
5647 		nbl_end_crit(vp);
5648 
5649 	DTRACE_NFSV4_2(op__write__done, struct compound_state *, cs,
5650 	    WRITE4res *, resp);
5651 }
5652 
5653 
5654 /* XXX put in a header file */
5655 extern int	sec_svc_getcred(struct svc_req *, cred_t *,  caddr_t *, int *);
5656 
5657 void
5658 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi,
5659 	struct svc_req *req, cred_t *cr, int *rv)
5660 {
5661 	uint_t i;
5662 	struct compound_state cs;
5663 
5664 	if (rv != NULL)
5665 		*rv = 0;
5666 	rfs4_init_compound_state(&cs);
5667 	/*
5668 	 * Form a reply tag by copying over the reqeuest tag.
5669 	 */
5670 	resp->tag.utf8string_val =
5671 	    kmem_alloc(args->tag.utf8string_len, KM_SLEEP);
5672 	resp->tag.utf8string_len = args->tag.utf8string_len;
5673 	bcopy(args->tag.utf8string_val, resp->tag.utf8string_val,
5674 	    resp->tag.utf8string_len);
5675 
5676 	cs.statusp = &resp->status;
5677 	cs.req = req;
5678 
5679 	/*
5680 	 * XXX for now, minorversion should be zero
5681 	 */
5682 	if (args->minorversion != NFS4_MINORVERSION) {
5683 		DTRACE_NFSV4_2(compound__start, struct compound_state *,
5684 		    &cs, COMPOUND4args *, args);
5685 		resp->array_len = 0;
5686 		resp->array = NULL;
5687 		resp->status = NFS4ERR_MINOR_VERS_MISMATCH;
5688 		DTRACE_NFSV4_2(compound__done, struct compound_state *,
5689 		    &cs, COMPOUND4res *, resp);
5690 		return;
5691 	}
5692 
5693 	ASSERT(exi == NULL);
5694 	ASSERT(cr == NULL);
5695 
5696 	cr = crget();
5697 	ASSERT(cr != NULL);
5698 
5699 	if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) {
5700 		DTRACE_NFSV4_2(compound__start, struct compound_state *,
5701 		    &cs, COMPOUND4args *, args);
5702 		crfree(cr);
5703 		DTRACE_NFSV4_2(compound__done, struct compound_state *,
5704 		    &cs, COMPOUND4res *, resp);
5705 		svcerr_badcred(req->rq_xprt);
5706 		if (rv != NULL)
5707 			*rv = 1;
5708 		return;
5709 	}
5710 	resp->array_len = args->array_len;
5711 	resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4),
5712 	    KM_SLEEP);
5713 
5714 	cs.basecr = cr;
5715 
5716 	DTRACE_NFSV4_2(compound__start, struct compound_state *, &cs,
5717 	    COMPOUND4args *, args);
5718 
5719 	/*
5720 	 * For now, NFS4 compound processing must be protected by
5721 	 * exported_lock because it can access more than one exportinfo
5722 	 * per compound and share/unshare can now change multiple
5723 	 * exinfo structs.  The NFS2/3 code only refs 1 exportinfo
5724 	 * per proc (excluding public exinfo), and exi_count design
5725 	 * is sufficient to protect concurrent execution of NFS2/3
5726 	 * ops along with unexport.  This lock will be removed as
5727 	 * part of the NFSv4 phase 2 namespace redesign work.
5728 	 */
5729 	rw_enter(&exported_lock, RW_READER);
5730 
5731 	/*
5732 	 * If this is the first compound we've seen, we need to start all
5733 	 * new instances' grace periods.
5734 	 */
5735 	if (rfs4_seen_first_compound == 0) {
5736 		rfs4_grace_start_new();
5737 		/*
5738 		 * This must be set after rfs4_grace_start_new(), otherwise
5739 		 * another thread could proceed past here before the former
5740 		 * is finished.
5741 		 */
5742 		rfs4_seen_first_compound = 1;
5743 	}
5744 
5745 	for (i = 0; i < args->array_len && cs.cont; i++) {
5746 		nfs_argop4 *argop;
5747 		nfs_resop4 *resop;
5748 		uint_t op;
5749 
5750 		argop = &args->array[i];
5751 		resop = &resp->array[i];
5752 		resop->resop = argop->argop;
5753 		op = (uint_t)resop->resop;
5754 
5755 		if (op < rfsv4disp_cnt) {
5756 			/*
5757 			 * Count the individual ops here; NULL and COMPOUND
5758 			 * are counted in common_dispatch()
5759 			 */
5760 			rfsproccnt_v4_ptr[op].value.ui64++;
5761 
5762 			NFS4_DEBUG(rfs4_debug > 1,
5763 			    (CE_NOTE, "Executing %s", rfs4_op_string[op]));
5764 			(*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs);
5765 			NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d",
5766 			    rfs4_op_string[op], *cs.statusp));
5767 			if (*cs.statusp != NFS4_OK)
5768 				cs.cont = FALSE;
5769 		} else {
5770 			/*
5771 			 * This is effectively dead code since XDR code
5772 			 * will have already returned BADXDR if op doesn't
5773 			 * decode to legal value.  This only done for a
5774 			 * day when XDR code doesn't verify v4 opcodes.
5775 			 */
5776 			op = OP_ILLEGAL;
5777 			rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++;
5778 
5779 			rfs4_op_illegal(argop, resop, req, &cs);
5780 			cs.cont = FALSE;
5781 		}
5782 
5783 		/*
5784 		 * If not at last op, and if we are to stop, then
5785 		 * compact the results array.
5786 		 */
5787 		if ((i + 1) < args->array_len && !cs.cont) {
5788 			nfs_resop4 *new_res = kmem_alloc(
5789 			    (i+1) * sizeof (nfs_resop4), KM_SLEEP);
5790 			bcopy(resp->array,
5791 			    new_res, (i+1) * sizeof (nfs_resop4));
5792 			kmem_free(resp->array,
5793 			    args->array_len * sizeof (nfs_resop4));
5794 
5795 			resp->array_len =  i + 1;
5796 			resp->array = new_res;
5797 		}
5798 	}
5799 
5800 	rw_exit(&exported_lock);
5801 
5802 	DTRACE_NFSV4_2(compound__done, struct compound_state *, &cs,
5803 	    COMPOUND4res *, resp);
5804 
5805 	if (cs.vp)
5806 		VN_RELE(cs.vp);
5807 	if (cs.saved_vp)
5808 		VN_RELE(cs.saved_vp);
5809 	if (cs.saved_fh.nfs_fh4_val)
5810 		kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE);
5811 
5812 	if (cs.basecr)
5813 		crfree(cs.basecr);
5814 	if (cs.cr)
5815 		crfree(cs.cr);
5816 	/*
5817 	 * done with this compound request, free the label
5818 	 */
5819 
5820 	if (req->rq_label != NULL) {
5821 		kmem_free(req->rq_label, sizeof (bslabel_t));
5822 		req->rq_label = NULL;
5823 	}
5824 }
5825 
5826 /*
5827  * XXX because of what appears to be duplicate calls to rfs4_compound_free
5828  * XXX zero out the tag and array values. Need to investigate why the
5829  * XXX calls occur, but at least prevent the panic for now.
5830  */
5831 void
5832 rfs4_compound_free(COMPOUND4res *resp)
5833 {
5834 	uint_t i;
5835 
5836 	if (resp->tag.utf8string_val) {
5837 		UTF8STRING_FREE(resp->tag)
5838 	}
5839 
5840 	for (i = 0; i < resp->array_len; i++) {
5841 		nfs_resop4 *resop;
5842 		uint_t op;
5843 
5844 		resop = &resp->array[i];
5845 		op = (uint_t)resop->resop;
5846 		if (op < rfsv4disp_cnt) {
5847 			(*rfsv4disptab[op].dis_resfree)(resop);
5848 		}
5849 	}
5850 	if (resp->array != NULL) {
5851 		kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4));
5852 	}
5853 }
5854 
5855 /*
5856  * Process the value of the compound request rpc flags, as a bit-AND
5857  * of the individual per-op flags (idempotent, allowork, publicfh_ok)
5858  */
5859 void
5860 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp)
5861 {
5862 	int i;
5863 	int flag = RPC_ALL;
5864 
5865 	for (i = 0; flag && i < args->array_len; i++) {
5866 		uint_t op;
5867 
5868 		op = (uint_t)args->array[i].argop;
5869 
5870 		if (op < rfsv4disp_cnt)
5871 			flag &= rfsv4disptab[op].dis_flags;
5872 		else
5873 			flag = 0;
5874 	}
5875 	*flagp = flag;
5876 }
5877 
5878 nfsstat4
5879 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp)
5880 {
5881 	nfsstat4 e;
5882 
5883 	rfs4_dbe_lock(cp->dbe);
5884 
5885 	if (cp->sysidt != LM_NOSYSID) {
5886 		*sp = cp->sysidt;
5887 		e = NFS4_OK;
5888 
5889 	} else if ((cp->sysidt = lm_alloc_sysidt()) != LM_NOSYSID) {
5890 		*sp = cp->sysidt;
5891 		e = NFS4_OK;
5892 
5893 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
5894 		    "rfs4_client_sysid: allocated 0x%x\n", *sp));
5895 	} else
5896 		e = NFS4ERR_DELAY;
5897 
5898 	rfs4_dbe_unlock(cp->dbe);
5899 	return (e);
5900 }
5901 
5902 #if defined(DEBUG) && ! defined(lint)
5903 static void lock_print(char *str, int operation, struct flock64 *flk)
5904 {
5905 	char *op, *type;
5906 
5907 	switch (operation) {
5908 	case F_GETLK: op = "F_GETLK";
5909 		break;
5910 	case F_SETLK: op = "F_SETLK";
5911 		break;
5912 	case F_SETLK_NBMAND: op = "F_SETLK_NBMAND";
5913 		break;
5914 	default: op = "F_UNKNOWN";
5915 		break;
5916 	}
5917 	switch (flk->l_type) {
5918 	case F_UNLCK: type = "F_UNLCK";
5919 		break;
5920 	case F_RDLCK: type = "F_RDLCK";
5921 		break;
5922 	case F_WRLCK: type = "F_WRLCK";
5923 		break;
5924 	default: type = "F_UNKNOWN";
5925 		break;
5926 	}
5927 
5928 	ASSERT(flk->l_whence == 0);
5929 	cmn_err(CE_NOTE, "%s:  %s, type = %s, off = %llx len = %llx pid = %d",
5930 	    str, op, type, (longlong_t)flk->l_start,
5931 	    flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid);
5932 }
5933 
5934 #define	LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f)
5935 #else
5936 #define	LOCK_PRINT(d, s, t, f)
5937 #endif
5938 
5939 /*ARGSUSED*/
5940 static bool_t
5941 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs)
5942 {
5943 	return (TRUE);
5944 }
5945 
5946 /*
5947  * Look up the pathname using the vp in cs as the directory vnode.
5948  * cs->vp will be the vnode for the file on success
5949  */
5950 
5951 static nfsstat4
5952 rfs4_lookup(component4 *component, struct svc_req *req,
5953 	    struct compound_state *cs)
5954 {
5955 	char *nm;
5956 	uint32_t len;
5957 	nfsstat4 status;
5958 
5959 	if (cs->vp == NULL) {
5960 		return (NFS4ERR_NOFILEHANDLE);
5961 	}
5962 	if (cs->vp->v_type != VDIR) {
5963 		return (NFS4ERR_NOTDIR);
5964 	}
5965 
5966 	if (!utf8_dir_verify(component))
5967 		return (NFS4ERR_INVAL);
5968 
5969 	nm = utf8_to_fn(component, &len, NULL);
5970 	if (nm == NULL) {
5971 		return (NFS4ERR_INVAL);
5972 	}
5973 
5974 	if (len > MAXNAMELEN) {
5975 		kmem_free(nm, len);
5976 		return (NFS4ERR_NAMETOOLONG);
5977 	}
5978 
5979 	status = do_rfs4_op_lookup(nm, len, req, cs);
5980 
5981 	kmem_free(nm, len);
5982 
5983 	return (status);
5984 }
5985 
5986 static nfsstat4
5987 rfs4_lookupfile(component4 *component, struct svc_req *req,
5988 		struct compound_state *cs, uint32_t access,
5989 		change_info4 *cinfo)
5990 {
5991 	nfsstat4 status;
5992 	vnode_t *dvp = cs->vp;
5993 	vattr_t bva, ava, fva;
5994 	int error;
5995 
5996 	/* Get "before" change value */
5997 	bva.va_mask = AT_CTIME|AT_SEQ;
5998 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL);
5999 	if (error)
6000 		return (puterrno4(error));
6001 
6002 	/* rfs4_lookup may VN_RELE directory */
6003 	VN_HOLD(dvp);
6004 
6005 	status = rfs4_lookup(component, req, cs);
6006 	if (status != NFS4_OK) {
6007 		VN_RELE(dvp);
6008 		return (status);
6009 	}
6010 
6011 	/*
6012 	 * Get "after" change value, if it fails, simply return the
6013 	 * before value.
6014 	 */
6015 	ava.va_mask = AT_CTIME|AT_SEQ;
6016 	if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) {
6017 		ava.va_ctime = bva.va_ctime;
6018 		ava.va_seq = 0;
6019 	}
6020 	VN_RELE(dvp);
6021 
6022 	/*
6023 	 * Validate the file is a file
6024 	 */
6025 	fva.va_mask = AT_TYPE|AT_MODE;
6026 	error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL);
6027 	if (error)
6028 		return (puterrno4(error));
6029 
6030 	if (fva.va_type != VREG) {
6031 		if (fva.va_type == VDIR)
6032 			return (NFS4ERR_ISDIR);
6033 		if (fva.va_type == VLNK)
6034 			return (NFS4ERR_SYMLINK);
6035 		return (NFS4ERR_INVAL);
6036 	}
6037 
6038 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime);
6039 	NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
6040 
6041 	/*
6042 	 * It is undefined if VOP_LOOKUP will change va_seq, so
6043 	 * cinfo.atomic = TRUE only if we have
6044 	 * non-zero va_seq's, and they have not changed.
6045 	 */
6046 	if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq)
6047 		cinfo->atomic = TRUE;
6048 	else
6049 		cinfo->atomic = FALSE;
6050 
6051 	/* Check for mandatory locking */
6052 	cs->mandlock = MANDLOCK(cs->vp, fva.va_mode);
6053 	return (check_open_access(access, cs, req));
6054 }
6055 
6056 static nfsstat4
6057 create_vnode(vnode_t *dvp, char *nm,  vattr_t *vap, createmode4 mode,
6058 	    timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created)
6059 {
6060 	int error;
6061 	nfsstat4 status = NFS4_OK;
6062 	vattr_t va;
6063 
6064 tryagain:
6065 
6066 	/*
6067 	 * The file open mode used is VWRITE.  If the client needs
6068 	 * some other semantic, then it should do the access checking
6069 	 * itself.  It would have been nice to have the file open mode
6070 	 * passed as part of the arguments.
6071 	 */
6072 
6073 	*created = TRUE;
6074 	error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL);
6075 
6076 	if (error) {
6077 		*created = FALSE;
6078 
6079 		/*
6080 		 * If we got something other than file already exists
6081 		 * then just return this error.  Otherwise, we got
6082 		 * EEXIST.  If we were doing a GUARDED create, then
6083 		 * just return this error.  Otherwise, we need to
6084 		 * make sure that this wasn't a duplicate of an
6085 		 * exclusive create request.
6086 		 *
6087 		 * The assumption is made that a non-exclusive create
6088 		 * request will never return EEXIST.
6089 		 */
6090 
6091 		if (error != EEXIST || mode == GUARDED4) {
6092 			status = puterrno4(error);
6093 			return (status);
6094 		}
6095 		error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr,
6096 		    NULL, NULL, NULL);
6097 
6098 		if (error) {
6099 			/*
6100 			 * We couldn't find the file that we thought that
6101 			 * we just created.  So, we'll just try creating
6102 			 * it again.
6103 			 */
6104 			if (error == ENOENT)
6105 				goto tryagain;
6106 
6107 			status = puterrno4(error);
6108 			return (status);
6109 		}
6110 
6111 		if (mode == UNCHECKED4) {
6112 			/* existing object must be regular file */
6113 			if ((*vpp)->v_type != VREG) {
6114 				if ((*vpp)->v_type == VDIR)
6115 					status = NFS4ERR_ISDIR;
6116 				else if ((*vpp)->v_type == VLNK)
6117 					status = NFS4ERR_SYMLINK;
6118 				else
6119 					status = NFS4ERR_INVAL;
6120 				VN_RELE(*vpp);
6121 				return (status);
6122 			}
6123 
6124 			return (NFS4_OK);
6125 		}
6126 
6127 		/* Check for duplicate request */
6128 		ASSERT(mtime != 0);
6129 		va.va_mask = AT_MTIME;
6130 		error = VOP_GETATTR(*vpp, &va, 0, cr, NULL);
6131 		if (!error) {
6132 			/* We found the file */
6133 			if (va.va_mtime.tv_sec != mtime->tv_sec ||
6134 			    va.va_mtime.tv_nsec != mtime->tv_nsec) {
6135 				/* but its not our creation */
6136 				VN_RELE(*vpp);
6137 				return (NFS4ERR_EXIST);
6138 			}
6139 			*created = TRUE; /* retrans of create == created */
6140 			return (NFS4_OK);
6141 		}
6142 		VN_RELE(*vpp);
6143 		return (NFS4ERR_EXIST);
6144 	}
6145 
6146 	return (NFS4_OK);
6147 }
6148 
6149 static nfsstat4
6150 check_open_access(uint32_t access,
6151 		struct compound_state *cs, struct svc_req *req)
6152 {
6153 	int error;
6154 	vnode_t *vp;
6155 	bool_t readonly;
6156 	cred_t *cr = cs->cr;
6157 
6158 	/* For now we don't allow mandatory locking as per V2/V3 */
6159 	if (cs->access == CS_ACCESS_DENIED || cs->mandlock) {
6160 		return (NFS4ERR_ACCESS);
6161 	}
6162 
6163 	vp = cs->vp;
6164 	ASSERT(cr != NULL && vp->v_type == VREG);
6165 
6166 	/*
6167 	 * If the file system is exported read only and we are trying
6168 	 * to open for write, then return NFS4ERR_ROFS
6169 	 */
6170 
6171 	readonly = rdonly4(cs->exi, cs->vp, req);
6172 
6173 	if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly)
6174 		return (NFS4ERR_ROFS);
6175 
6176 	if (access & OPEN4_SHARE_ACCESS_READ) {
6177 		if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) &&
6178 		    (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) {
6179 			return (NFS4ERR_ACCESS);
6180 		}
6181 	}
6182 
6183 	if (access & OPEN4_SHARE_ACCESS_WRITE) {
6184 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
6185 		if (error)
6186 			return (NFS4ERR_ACCESS);
6187 	}
6188 
6189 	return (NFS4_OK);
6190 }
6191 
6192 static nfsstat4
6193 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs,
6194 		change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid)
6195 {
6196 	struct nfs4_svgetit_arg sarg;
6197 	struct nfs4_ntov_table ntov;
6198 
6199 	bool_t ntov_table_init = FALSE;
6200 	struct statvfs64 sb;
6201 	nfsstat4 status;
6202 	vnode_t *vp;
6203 	vattr_t bva, ava, iva, cva, *vap;
6204 	vnode_t *dvp;
6205 	timespec32_t *mtime;
6206 	char *nm = NULL;
6207 	uint_t buflen;
6208 	bool_t created;
6209 	bool_t setsize = FALSE;
6210 	len_t reqsize;
6211 	int error;
6212 	bool_t trunc;
6213 	caller_context_t ct;
6214 	component4 *component;
6215 	bslabel_t *clabel;
6216 	struct sockaddr *ca;
6217 	char *name = NULL;
6218 
6219 	sarg.sbp = &sb;
6220 
6221 	dvp = cs->vp;
6222 
6223 	/* Check if the file system is read only */
6224 	if (rdonly4(cs->exi, dvp, req))
6225 		return (NFS4ERR_ROFS);
6226 
6227 	/* check the label of including directory */
6228 	if (is_system_labeled()) {
6229 		ASSERT(req->rq_label != NULL);
6230 		clabel = req->rq_label;
6231 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
6232 		    "got client label from request(1)",
6233 		    struct svc_req *, req);
6234 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
6235 			if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK,
6236 			    cs->exi)) {
6237 				return (NFS4ERR_ACCESS);
6238 			}
6239 		}
6240 	}
6241 
6242 	/*
6243 	 * Get the last component of path name in nm. cs will reference
6244 	 * the including directory on success.
6245 	 */
6246 	component = &args->open_claim4_u.file;
6247 	if (!utf8_dir_verify(component))
6248 		return (NFS4ERR_INVAL);
6249 
6250 	nm = utf8_to_fn(component, &buflen, NULL);
6251 
6252 	if (nm == NULL)
6253 		return (NFS4ERR_RESOURCE);
6254 
6255 	if (buflen > MAXNAMELEN) {
6256 		kmem_free(nm, buflen);
6257 		return (NFS4ERR_NAMETOOLONG);
6258 	}
6259 
6260 	bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ;
6261 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL);
6262 	if (error) {
6263 		kmem_free(nm, buflen);
6264 		return (puterrno4(error));
6265 	}
6266 
6267 	if (bva.va_type != VDIR) {
6268 		kmem_free(nm, buflen);
6269 		return (NFS4ERR_NOTDIR);
6270 	}
6271 
6272 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime)
6273 
6274 	switch (args->mode) {
6275 	case GUARDED4:
6276 		/*FALLTHROUGH*/
6277 	case UNCHECKED4:
6278 		nfs4_ntov_table_init(&ntov);
6279 		ntov_table_init = TRUE;
6280 
6281 		*attrset = 0;
6282 		status = do_rfs4_set_attrs(attrset,
6283 		    &args->createhow4_u.createattrs,
6284 		    cs, &sarg, &ntov, NFS4ATTR_SETIT);
6285 
6286 		if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) &&
6287 		    sarg.vap->va_type != VREG) {
6288 			if (sarg.vap->va_type == VDIR)
6289 				status = NFS4ERR_ISDIR;
6290 			else if (sarg.vap->va_type == VLNK)
6291 				status = NFS4ERR_SYMLINK;
6292 			else
6293 				status = NFS4ERR_INVAL;
6294 		}
6295 
6296 		if (status != NFS4_OK) {
6297 			kmem_free(nm, buflen);
6298 			nfs4_ntov_table_free(&ntov, &sarg);
6299 			*attrset = 0;
6300 			return (status);
6301 		}
6302 
6303 		vap = sarg.vap;
6304 		vap->va_type = VREG;
6305 		vap->va_mask |= AT_TYPE;
6306 
6307 		if ((vap->va_mask & AT_MODE) == 0) {
6308 			vap->va_mask |= AT_MODE;
6309 			vap->va_mode = (mode_t)0600;
6310 		}
6311 
6312 		if (vap->va_mask & AT_SIZE) {
6313 
6314 			/* Disallow create with a non-zero size */
6315 
6316 			if ((reqsize = sarg.vap->va_size) != 0) {
6317 				kmem_free(nm, buflen);
6318 				nfs4_ntov_table_free(&ntov, &sarg);
6319 				*attrset = 0;
6320 				return (NFS4ERR_INVAL);
6321 			}
6322 			setsize = TRUE;
6323 		}
6324 		break;
6325 
6326 	case EXCLUSIVE4:
6327 		/* prohibit EXCL create of named attributes */
6328 		if (dvp->v_flag & V_XATTRDIR) {
6329 			kmem_free(nm, buflen);
6330 			*attrset = 0;
6331 			return (NFS4ERR_INVAL);
6332 		}
6333 
6334 		cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE;
6335 		cva.va_type = VREG;
6336 		/*
6337 		 * Ensure no time overflows. Assumes underlying
6338 		 * filesystem supports at least 32 bits.
6339 		 * Truncate nsec to usec resolution to allow valid
6340 		 * compares even if the underlying filesystem truncates.
6341 		 */
6342 		mtime = (timespec32_t *)&args->createhow4_u.createverf;
6343 		cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX;
6344 		cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000;
6345 		cva.va_mode = (mode_t)0;
6346 		vap = &cva;
6347 		break;
6348 	}
6349 
6350 	/* If necessary, convert to UTF-8 for illbehaved clients */
6351 
6352 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
6353 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
6354 	    MAXPATHLEN  + 1);
6355 
6356 	if (name == NULL) {
6357 		kmem_free(nm, buflen);
6358 		return (NFS4ERR_SERVERFAULT);
6359 	}
6360 
6361 	status = create_vnode(dvp, name, vap, args->mode, mtime,
6362 	    cs->cr, &vp, &created);
6363 	if (nm != name)
6364 		kmem_free(name, MAXPATHLEN + 1);
6365 	kmem_free(nm, buflen);
6366 
6367 	if (status != NFS4_OK) {
6368 		if (ntov_table_init)
6369 			nfs4_ntov_table_free(&ntov, &sarg);
6370 		*attrset = 0;
6371 		return (status);
6372 	}
6373 
6374 	trunc = (setsize && !created);
6375 
6376 	if (args->mode != EXCLUSIVE4) {
6377 		bitmap4 createmask = args->createhow4_u.createattrs.attrmask;
6378 
6379 		/*
6380 		 * True verification that object was created with correct
6381 		 * attrs is impossible.  The attrs could have been changed
6382 		 * immediately after object creation.  If attributes did
6383 		 * not verify, the only recourse for the server is to
6384 		 * destroy the object.  Maybe if some attrs (like gid)
6385 		 * are set incorrectly, the object should be destroyed;
6386 		 * however, seems bad as a default policy.  Do we really
6387 		 * want to destroy an object over one of the times not
6388 		 * verifying correctly?  For these reasons, the server
6389 		 * currently sets bits in attrset for createattrs
6390 		 * that were set; however, no verification is done.
6391 		 *
6392 		 * vmask_to_nmask accounts for vattr bits set on create
6393 		 *	[do_rfs4_set_attrs() only sets resp bits for
6394 		 *	 non-vattr/vfs bits.]
6395 		 * Mask off any bits we set by default so as not to return
6396 		 * more attrset bits than were requested in createattrs
6397 		 */
6398 		if (created) {
6399 			nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset);
6400 			*attrset &= createmask;
6401 		} else {
6402 			/*
6403 			 * We did not create the vnode (we tried but it
6404 			 * already existed).  In this case, the only createattr
6405 			 * that the spec allows the server to set is size,
6406 			 * and even then, it can only be set if it is 0.
6407 			 */
6408 			*attrset = 0;
6409 			if (trunc)
6410 				*attrset = FATTR4_SIZE_MASK;
6411 		}
6412 	}
6413 	if (ntov_table_init)
6414 		nfs4_ntov_table_free(&ntov, &sarg);
6415 
6416 	/*
6417 	 * Get the initial "after" sequence number, if it fails,
6418 	 * set to zero, time to before.
6419 	 */
6420 	iva.va_mask = AT_CTIME|AT_SEQ;
6421 	if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) {
6422 		iva.va_seq = 0;
6423 		iva.va_ctime = bva.va_ctime;
6424 	}
6425 
6426 	/*
6427 	 * create_vnode attempts to create the file exclusive,
6428 	 * if it already exists the VOP_CREATE will fail and
6429 	 * may not increase va_seq. It is atomic if
6430 	 * we haven't changed the directory, but if it has changed
6431 	 * we don't know what changed it.
6432 	 */
6433 	if (!created) {
6434 		if (bva.va_seq && iva.va_seq &&
6435 		    bva.va_seq == iva.va_seq)
6436 			cinfo->atomic = TRUE;
6437 		else
6438 			cinfo->atomic = FALSE;
6439 		NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime);
6440 	} else {
6441 		/*
6442 		 * The entry was created, we need to sync the
6443 		 * directory metadata.
6444 		 */
6445 		(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
6446 
6447 		/*
6448 		 * Get "after" change value, if it fails, simply return the
6449 		 * before value.
6450 		 */
6451 		ava.va_mask = AT_CTIME|AT_SEQ;
6452 		if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) {
6453 			ava.va_ctime = bva.va_ctime;
6454 			ava.va_seq = 0;
6455 		}
6456 
6457 		NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
6458 
6459 		/*
6460 		 * The cinfo->atomic = TRUE only if we have
6461 		 * non-zero va_seq's, and it has incremented by exactly one
6462 		 * during the create_vnode and it didn't
6463 		 * change during the VOP_FSYNC.
6464 		 */
6465 		if (bva.va_seq && iva.va_seq && ava.va_seq &&
6466 		    iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq)
6467 			cinfo->atomic = TRUE;
6468 		else
6469 			cinfo->atomic = FALSE;
6470 	}
6471 
6472 	/* Check for mandatory locking and that the size gets set. */
6473 	cva.va_mask = AT_MODE;
6474 	if (setsize)
6475 		cva.va_mask |= AT_SIZE;
6476 
6477 	/* Assume the worst */
6478 	cs->mandlock = TRUE;
6479 
6480 	if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) {
6481 		cs->mandlock = MANDLOCK(cs->vp, cva.va_mode);
6482 
6483 		/*
6484 		 * Truncate the file if necessary; this would be
6485 		 * the case for create over an existing file.
6486 		 */
6487 
6488 		if (trunc) {
6489 			int in_crit = 0;
6490 			rfs4_file_t *fp;
6491 			bool_t create = FALSE;
6492 
6493 			/*
6494 			 * We are writing over an existing file.
6495 			 * Check to see if we need to recall a delegation.
6496 			 */
6497 			rfs4_hold_deleg_policy();
6498 			if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) {
6499 				if (rfs4_check_delegated_byfp(FWRITE, fp,
6500 				    (reqsize == 0), FALSE, FALSE, &clientid)) {
6501 					rfs4_file_rele(fp);
6502 					rfs4_rele_deleg_policy();
6503 					VN_RELE(vp);
6504 					*attrset = 0;
6505 					return (NFS4ERR_DELAY);
6506 				}
6507 				rfs4_file_rele(fp);
6508 			}
6509 			rfs4_rele_deleg_policy();
6510 
6511 			if (nbl_need_check(vp)) {
6512 				in_crit = 1;
6513 
6514 				ASSERT(reqsize == 0);
6515 
6516 				nbl_start_crit(vp, RW_READER);
6517 				if (nbl_conflict(vp, NBL_WRITE, 0,
6518 				    cva.va_size, 0, NULL)) {
6519 					in_crit = 0;
6520 					nbl_end_crit(vp);
6521 					VN_RELE(vp);
6522 					*attrset = 0;
6523 					return (NFS4ERR_ACCESS);
6524 				}
6525 			}
6526 			ct.cc_sysid = 0;
6527 			ct.cc_pid = 0;
6528 			ct.cc_caller_id = nfs4_srv_caller_id;
6529 			ct.cc_flags = CC_DONTBLOCK;
6530 
6531 			cva.va_mask = AT_SIZE;
6532 			cva.va_size = reqsize;
6533 			(void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct);
6534 			if (in_crit)
6535 				nbl_end_crit(vp);
6536 		}
6537 	}
6538 
6539 	error = makefh4(&cs->fh, vp, cs->exi);
6540 
6541 	/*
6542 	 * Force modified data and metadata out to stable storage.
6543 	 */
6544 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL);
6545 
6546 	if (error) {
6547 		VN_RELE(vp);
6548 		*attrset = 0;
6549 		return (puterrno4(error));
6550 	}
6551 
6552 	/* if parent dir is attrdir, set namedattr fh flag */
6553 	if (dvp->v_flag & V_XATTRDIR)
6554 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
6555 
6556 	if (cs->vp)
6557 		VN_RELE(cs->vp);
6558 
6559 	cs->vp = vp;
6560 
6561 	/*
6562 	 * if we did not create the file, we will need to check
6563 	 * the access bits on the file
6564 	 */
6565 
6566 	if (!created) {
6567 		if (setsize)
6568 			args->share_access |= OPEN4_SHARE_ACCESS_WRITE;
6569 		status = check_open_access(args->share_access, cs, req);
6570 		if (status != NFS4_OK)
6571 			*attrset = 0;
6572 	}
6573 	return (status);
6574 }
6575 
6576 /*ARGSUSED*/
6577 static void
6578 rfs4_do_open(struct compound_state *cs, struct svc_req *req,
6579 		rfs4_openowner_t *oo, delegreq_t deleg,
6580 		uint32_t access, uint32_t deny,
6581 		OPEN4res *resp, int deleg_cur)
6582 {
6583 	/* XXX Currently not using req  */
6584 	rfs4_state_t *state;
6585 	rfs4_file_t *file;
6586 	bool_t screate = TRUE;
6587 	bool_t fcreate = TRUE;
6588 	uint32_t amodes;
6589 	uint32_t dmodes;
6590 	rfs4_deleg_state_t *dsp;
6591 	struct shrlock shr;
6592 	struct shr_locowner shr_loco;
6593 	sysid_t sysid;
6594 	nfsstat4 status;
6595 	caller_context_t ct;
6596 	int fflags = 0;
6597 	int recall = 0;
6598 	int err;
6599 	int cmd;
6600 
6601 	/* get the file struct and hold a lock on it during initial open */
6602 	file = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate);
6603 	if (file == NULL) {
6604 		NFS4_DEBUG(rfs4_debug,
6605 		    (CE_NOTE, "rfs4_do_open: can't find file"));
6606 		resp->status = NFS4ERR_SERVERFAULT;
6607 		return;
6608 	}
6609 
6610 	state = rfs4_findstate_by_owner_file(oo, file, &screate);
6611 	if (state == NULL) {
6612 		NFS4_DEBUG(rfs4_debug,
6613 		    (CE_NOTE, "rfs4_do_open: can't find state"));
6614 		resp->status = NFS4ERR_RESOURCE;
6615 		/* No need to keep any reference */
6616 		rfs4_file_rele_withunlock(file);
6617 		return;
6618 	}
6619 
6620 	/* try to get the sysid before continuing */
6621 	if ((status = rfs4_client_sysid(oo->client, &sysid)) != NFS4_OK) {
6622 		resp->status = status;
6623 		rfs4_file_rele(file);
6624 		/* Not a fully formed open; "close" it */
6625 		if (screate == TRUE)
6626 			rfs4_state_close(state, FALSE, FALSE, cs->cr);
6627 		rfs4_state_rele(state);
6628 		return;
6629 	}
6630 
6631 	/* Calculate the fflags for this OPEN. */
6632 	if (access & OPEN4_SHARE_ACCESS_READ)
6633 		fflags |= FREAD;
6634 	if (access & OPEN4_SHARE_ACCESS_WRITE)
6635 		fflags |= FWRITE;
6636 
6637 	/*
6638 	 * Calculate the new deny and access mode that this open is adding to
6639 	 * the file for this open owner;
6640 	 */
6641 	dmodes = (deny & ~state->share_deny);
6642 	amodes = (access & ~state->share_access);
6643 
6644 	/*
6645 	 * Check to see the client has already sent an open for this
6646 	 * open owner on this file with the same share/deny modes.
6647 	 * If so, we don't need to check for a conflict and we don't
6648 	 * need to add another shrlock.  If not, then we need to
6649 	 * check for conflicts in deny and access before checking for
6650 	 * conflicts in delegation.  We don't want to recall a
6651 	 * delegation based on an open that will eventually fail based
6652 	 * on shares modes.
6653 	 */
6654 
6655 	if (dmodes || amodes) {
6656 		shr.s_access = (short)access;
6657 		shr.s_deny = (short)deny;
6658 		shr.s_pid = rfs4_dbe_getid(oo->dbe);
6659 		shr.s_sysid = sysid;
6660 		shr_loco.sl_pid = shr.s_pid;
6661 		shr_loco.sl_id = shr.s_sysid;
6662 		shr.s_owner = (caddr_t)&shr_loco;
6663 		shr.s_own_len = sizeof (shr_loco);
6664 
6665 		cmd = nbl_need_check(cs->vp) ? F_SHARE_NBMAND : F_SHARE;
6666 		if ((err = vop_shrlock(cs->vp, cmd, &shr, fflags)) != 0) {
6667 
6668 			resp->status = err == EAGAIN ?
6669 			    NFS4ERR_SHARE_DENIED : puterrno4(err);
6670 
6671 			rfs4_file_rele(file);
6672 			/* Not a fully formed open; "close" it */
6673 			if (screate == TRUE)
6674 				rfs4_state_close(state, FALSE, FALSE, cs->cr);
6675 			rfs4_state_rele(state);
6676 			return;
6677 		}
6678 	}
6679 
6680 	rfs4_dbe_lock(state->dbe);
6681 	rfs4_dbe_lock(file->dbe);
6682 
6683 	/*
6684 	 * Check to see if this file is delegated and if so, if a
6685 	 * recall needs to be done.
6686 	 */
6687 	if (rfs4_check_recall(state, access)) {
6688 		rfs4_dbe_unlock(file->dbe);
6689 		rfs4_dbe_unlock(state->dbe);
6690 		rfs4_recall_deleg(file, FALSE, state->owner->client);
6691 		delay(NFS4_DELEGATION_CONFLICT_DELAY);
6692 		rfs4_dbe_lock(state->dbe);
6693 		rfs4_dbe_lock(file->dbe);
6694 		/* Let's see if the delegation was returned */
6695 		if (rfs4_check_recall(state, access)) {
6696 			rfs4_dbe_unlock(file->dbe);
6697 			rfs4_dbe_unlock(state->dbe);
6698 			rfs4_file_rele(file);
6699 			rfs4_update_lease(state->owner->client);
6700 			(void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags);
6701 			/* Not a fully formed open; "close" it */
6702 			if (screate == TRUE)
6703 				rfs4_state_close(state, FALSE, FALSE, cs->cr);
6704 			rfs4_state_rele(state);
6705 			resp->status = NFS4ERR_DELAY;
6706 			return;
6707 		}
6708 	}
6709 	/*
6710 	 * the share check passed and any delegation conflict has been
6711 	 * taken care of, now call vop_open.
6712 	 * if this is the first open then call vop_open with fflags.
6713 	 * if not, call vn_open_upgrade with just the upgrade flags.
6714 	 *
6715 	 * if the file has been opened already, it will have the current
6716 	 * access mode in the state struct.  if it has no share access, then
6717 	 * this is a new open.
6718 	 *
6719 	 * However, if this is open with CLAIM_DLEGATE_CUR, then don't
6720 	 * call VOP_OPEN(), just do the open upgrade.
6721 	 */
6722 	if (((state->share_access & OPEN4_SHARE_ACCESS_BOTH) == 0) &&
6723 	    !deleg_cur) {
6724 		ct.cc_sysid = sysid;
6725 		ct.cc_pid = shr.s_pid;
6726 		ct.cc_caller_id = nfs4_srv_caller_id;
6727 		ct.cc_flags = CC_DONTBLOCK;
6728 		err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct);
6729 		if (err) {
6730 			rfs4_dbe_unlock(file->dbe);
6731 			rfs4_dbe_unlock(state->dbe);
6732 			rfs4_file_rele(file);
6733 			(void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags);
6734 			/* Not a fully formed open; "close" it */
6735 			if (screate == TRUE)
6736 				rfs4_state_close(state, FALSE, FALSE, cs->cr);
6737 			rfs4_state_rele(state);
6738 			/* check if a monitor detected a delegation conflict */
6739 			if (err == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK))
6740 				resp->status = NFS4ERR_DELAY;
6741 			else
6742 				resp->status = NFS4ERR_SERVERFAULT;
6743 			return;
6744 		}
6745 	} else { /* open upgrade */
6746 		/*
6747 		 * calculate the fflags for the new mode that is being added
6748 		 * by this upgrade.
6749 		 */
6750 		fflags = 0;
6751 		if (amodes & OPEN4_SHARE_ACCESS_READ)
6752 			fflags |= FREAD;
6753 		if (amodes & OPEN4_SHARE_ACCESS_WRITE)
6754 			fflags |= FWRITE;
6755 		vn_open_upgrade(cs->vp, fflags);
6756 	}
6757 
6758 	if (dmodes & OPEN4_SHARE_DENY_READ)
6759 		file->deny_read++;
6760 	if (dmodes & OPEN4_SHARE_DENY_WRITE)
6761 		file->deny_write++;
6762 	file->share_deny |= deny;
6763 	state->share_deny |= deny;
6764 
6765 	if (amodes & OPEN4_SHARE_ACCESS_READ)
6766 		file->access_read++;
6767 	if (amodes & OPEN4_SHARE_ACCESS_WRITE)
6768 		file->access_write++;
6769 	file->share_access |= access;
6770 	state->share_access |= access;
6771 
6772 	/*
6773 	 * Check for delegation here. if the deleg argument is not
6774 	 * DELEG_ANY, then this is a reclaim from a client and
6775 	 * we must honor the delegation requested. If necessary we can
6776 	 * set the recall flag.
6777 	 */
6778 
6779 	dsp = rfs4_grant_delegation(deleg, state, &recall);
6780 
6781 	cs->deleg = (file->dinfo->dtype == OPEN_DELEGATE_WRITE);
6782 
6783 	next_stateid(&state->stateid);
6784 
6785 	resp->stateid = state->stateid.stateid;
6786 
6787 	rfs4_dbe_unlock(file->dbe);
6788 	rfs4_dbe_unlock(state->dbe);
6789 
6790 	if (dsp) {
6791 		rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall);
6792 		rfs4_deleg_state_rele(dsp);
6793 	}
6794 
6795 	rfs4_file_rele(file);
6796 	rfs4_state_rele(state);
6797 
6798 	resp->status = NFS4_OK;
6799 }
6800 
6801 /*ARGSUSED*/
6802 static void
6803 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req,
6804 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6805 {
6806 	change_info4 *cinfo = &resp->cinfo;
6807 	bitmap4 *attrset = &resp->attrset;
6808 
6809 	if (args->opentype == OPEN4_NOCREATE)
6810 		resp->status = rfs4_lookupfile(&args->open_claim4_u.file,
6811 		    req, cs, args->share_access, cinfo);
6812 	else {
6813 		/* inhibit delegation grants during exclusive create */
6814 
6815 		if (args->mode == EXCLUSIVE4)
6816 			rfs4_disable_delegation();
6817 
6818 		resp->status = rfs4_createfile(args, req, cs, cinfo, attrset,
6819 		    oo->client->clientid);
6820 	}
6821 
6822 	if (resp->status == NFS4_OK) {
6823 
6824 		/* cs->vp cs->fh now reference the desired file */
6825 
6826 		rfs4_do_open(cs, req, oo, DELEG_ANY, args->share_access,
6827 		    args->share_deny, resp, 0);
6828 
6829 		/*
6830 		 * If rfs4_createfile set attrset, we must
6831 		 * clear this attrset before the response is copied.
6832 		 */
6833 		if (resp->status != NFS4_OK && resp->attrset) {
6834 			resp->attrset = 0;
6835 		}
6836 	}
6837 	else
6838 		*cs->statusp = resp->status;
6839 
6840 	if (args->mode == EXCLUSIVE4)
6841 		rfs4_enable_delegation();
6842 }
6843 
6844 /*ARGSUSED*/
6845 static void
6846 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req,
6847 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6848 {
6849 	change_info4 *cinfo = &resp->cinfo;
6850 	vattr_t va;
6851 	vtype_t v_type = cs->vp->v_type;
6852 	int error = 0;
6853 
6854 	/* Verify that we have a regular file */
6855 	if (v_type != VREG) {
6856 		if (v_type == VDIR)
6857 			resp->status = NFS4ERR_ISDIR;
6858 		else if (v_type == VLNK)
6859 			resp->status = NFS4ERR_SYMLINK;
6860 		else
6861 			resp->status = NFS4ERR_INVAL;
6862 		return;
6863 	}
6864 
6865 	va.va_mask = AT_MODE|AT_UID;
6866 	error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL);
6867 	if (error) {
6868 		resp->status = puterrno4(error);
6869 		return;
6870 	}
6871 
6872 	cs->mandlock = MANDLOCK(cs->vp, va.va_mode);
6873 
6874 	/*
6875 	 * Check if we have access to the file, Note the the file
6876 	 * could have originally been open UNCHECKED or GUARDED
6877 	 * with mode bits that will now fail, but there is nothing
6878 	 * we can really do about that except in the case that the
6879 	 * owner of the file is the one requesting the open.
6880 	 */
6881 	if (crgetuid(cs->cr) != va.va_uid) {
6882 		resp->status = check_open_access(args->share_access, cs, req);
6883 		if (resp->status != NFS4_OK) {
6884 			return;
6885 		}
6886 	}
6887 
6888 	/*
6889 	 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero
6890 	 */
6891 	cinfo->before = 0;
6892 	cinfo->after = 0;
6893 	cinfo->atomic = FALSE;
6894 
6895 	rfs4_do_open(cs, req, oo,
6896 	    NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type),
6897 	    args->share_access, args->share_deny, resp, 0);
6898 }
6899 
6900 static void
6901 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req,
6902 		OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6903 {
6904 	int error;
6905 	nfsstat4 status;
6906 	stateid4 stateid =
6907 	    args->open_claim4_u.delegate_cur_info.delegate_stateid;
6908 	rfs4_deleg_state_t *dsp;
6909 
6910 	/*
6911 	 * Find the state info from the stateid and confirm that the
6912 	 * file is delegated.  If the state openowner is the same as
6913 	 * the supplied openowner we're done. If not, get the file
6914 	 * info from the found state info. Use that file info to
6915 	 * create the state for this lock owner. Note solaris doen't
6916 	 * really need the pathname to find the file. We may want to
6917 	 * lookup the pathname and make sure that the vp exist and
6918 	 * matches the vp in the file structure. However it is
6919 	 * possible that the pathname nolonger exists (local process
6920 	 * unlinks the file), so this may not be that useful.
6921 	 */
6922 
6923 	status = rfs4_get_deleg_state(&stateid, &dsp);
6924 	if (status != NFS4_OK) {
6925 		resp->status = status;
6926 		return;
6927 	}
6928 
6929 	ASSERT(dsp->finfo->dinfo->dtype != OPEN_DELEGATE_NONE);
6930 
6931 	/*
6932 	 * New lock owner, create state. Since this was probably called
6933 	 * in response to a CB_RECALL we set deleg to DELEG_NONE
6934 	 */
6935 
6936 	ASSERT(cs->vp != NULL);
6937 	VN_RELE(cs->vp);
6938 	VN_HOLD(dsp->finfo->vp);
6939 	cs->vp = dsp->finfo->vp;
6940 
6941 	if (error = makefh4(&cs->fh, cs->vp, cs->exi)) {
6942 		rfs4_deleg_state_rele(dsp);
6943 		*cs->statusp = resp->status = puterrno4(error);
6944 		return;
6945 	}
6946 
6947 	/* Mark progress for delegation returns */
6948 	dsp->finfo->dinfo->time_lastwrite = gethrestime_sec();
6949 	rfs4_deleg_state_rele(dsp);
6950 	rfs4_do_open(cs, req, oo, DELEG_NONE,
6951 	    args->share_access, args->share_deny, resp, 1);
6952 }
6953 
6954 /*ARGSUSED*/
6955 static void
6956 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req,
6957 			OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
6958 {
6959 	/*
6960 	 * Lookup the pathname, it must already exist since this file
6961 	 * was delegated.
6962 	 *
6963 	 * Find the file and state info for this vp and open owner pair.
6964 	 *	check that they are in fact delegated.
6965 	 *	check that the state access and deny modes are the same.
6966 	 *
6967 	 * Return the delgation possibly seting the recall flag.
6968 	 */
6969 	rfs4_file_t *file;
6970 	rfs4_state_t *state;
6971 	bool_t create = FALSE;
6972 	bool_t dcreate = FALSE;
6973 	rfs4_deleg_state_t *dsp;
6974 	nfsace4 *ace;
6975 
6976 
6977 	/* Note we ignore oflags */
6978 	resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev,
6979 	    req, cs, args->share_access, &resp->cinfo);
6980 
6981 	if (resp->status != NFS4_OK) {
6982 		return;
6983 	}
6984 
6985 	/* get the file struct and hold a lock on it during initial open */
6986 	file = rfs4_findfile_withlock(cs->vp, NULL, &create);
6987 	if (file == NULL) {
6988 		NFS4_DEBUG(rfs4_debug,
6989 		    (CE_NOTE, "rfs4_do_opendelprev: can't find file"));
6990 		resp->status = NFS4ERR_SERVERFAULT;
6991 		return;
6992 	}
6993 
6994 	state = rfs4_findstate_by_owner_file(oo, file, &create);
6995 	if (state == NULL) {
6996 		NFS4_DEBUG(rfs4_debug,
6997 		    (CE_NOTE, "rfs4_do_opendelprev: can't find state"));
6998 		resp->status = NFS4ERR_SERVERFAULT;
6999 		rfs4_file_rele_withunlock(file);
7000 		return;
7001 	}
7002 
7003 	rfs4_dbe_lock(state->dbe);
7004 	rfs4_dbe_lock(file->dbe);
7005 	if (args->share_access != state->share_access ||
7006 	    args->share_deny != state->share_deny ||
7007 	    state->finfo->dinfo->dtype == OPEN_DELEGATE_NONE) {
7008 		NFS4_DEBUG(rfs4_debug,
7009 		    (CE_NOTE, "rfs4_do_opendelprev: state mixup"));
7010 		rfs4_dbe_unlock(file->dbe);
7011 		rfs4_dbe_unlock(state->dbe);
7012 		rfs4_file_rele(file);
7013 		rfs4_state_rele(state);
7014 		resp->status = NFS4ERR_SERVERFAULT;
7015 		return;
7016 	}
7017 	rfs4_dbe_unlock(file->dbe);
7018 	rfs4_dbe_unlock(state->dbe);
7019 
7020 	dsp = rfs4_finddeleg(state, &dcreate);
7021 	if (dsp == NULL) {
7022 		rfs4_state_rele(state);
7023 		rfs4_file_rele(file);
7024 		resp->status = NFS4ERR_SERVERFAULT;
7025 		return;
7026 	}
7027 
7028 	next_stateid(&state->stateid);
7029 
7030 	resp->stateid = state->stateid.stateid;
7031 
7032 	resp->delegation.delegation_type = dsp->dtype;
7033 
7034 	if (dsp->dtype == OPEN_DELEGATE_READ) {
7035 		open_read_delegation4 *rv =
7036 		    &resp->delegation.open_delegation4_u.read;
7037 
7038 		rv->stateid = dsp->delegid.stateid;
7039 		rv->recall = FALSE; /* no policy in place to set to TRUE */
7040 		ace = &rv->permissions;
7041 	} else {
7042 		open_write_delegation4 *rv =
7043 		    &resp->delegation.open_delegation4_u.write;
7044 
7045 		rv->stateid = dsp->delegid.stateid;
7046 		rv->recall = FALSE;  /* no policy in place to set to TRUE */
7047 		ace = &rv->permissions;
7048 		rv->space_limit.limitby = NFS_LIMIT_SIZE;
7049 		rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX;
7050 	}
7051 
7052 	/* XXX For now */
7053 	ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE;
7054 	ace->flag = 0;
7055 	ace->access_mask = 0;
7056 	ace->who.utf8string_len = 0;
7057 	ace->who.utf8string_val = 0;
7058 
7059 	rfs4_deleg_state_rele(dsp);
7060 	rfs4_state_rele(state);
7061 	rfs4_file_rele(file);
7062 }
7063 
7064 typedef enum {
7065 	NFS4_CHKSEQ_OKAY = 0,
7066 	NFS4_CHKSEQ_REPLAY = 1,
7067 	NFS4_CHKSEQ_BAD = 2
7068 } rfs4_chkseq_t;
7069 
7070 /*
7071  * Generic function for sequence number checks.
7072  */
7073 static rfs4_chkseq_t
7074 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop,
7075 		seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres)
7076 {
7077 	/* Same sequence ids and matching operations? */
7078 	if (seqid == rqst_seq && resop->resop == lastop->resop) {
7079 		if (copyres == TRUE) {
7080 			rfs4_free_reply(resop);
7081 			rfs4_copy_reply(resop, lastop);
7082 		}
7083 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
7084 		    "Replayed SEQID %d\n", seqid));
7085 		return (NFS4_CHKSEQ_REPLAY);
7086 	}
7087 
7088 	/* If the incoming sequence is not the next expected then it is bad */
7089 	if (rqst_seq != seqid + 1) {
7090 		if (rqst_seq == seqid) {
7091 			NFS4_DEBUG(rfs4_debug,
7092 			    (CE_NOTE, "BAD SEQID: Replayed sequence id "
7093 			    "but last op was %d current op is %d\n",
7094 			    lastop->resop, resop->resop));
7095 			return (NFS4_CHKSEQ_BAD);
7096 		}
7097 		NFS4_DEBUG(rfs4_debug,
7098 		    (CE_NOTE, "BAD SEQID: got %u expecting %u\n",
7099 		    rqst_seq, seqid));
7100 		return (NFS4_CHKSEQ_BAD);
7101 	}
7102 
7103 	/* Everything okay -- next expected */
7104 	return (NFS4_CHKSEQ_OKAY);
7105 }
7106 
7107 
7108 static rfs4_chkseq_t
7109 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop)
7110 {
7111 	rfs4_chkseq_t rc;
7112 
7113 	rfs4_dbe_lock(op->dbe);
7114 	rc = rfs4_check_seqid(op->open_seqid, op->reply, seqid, resop, TRUE);
7115 	rfs4_dbe_unlock(op->dbe);
7116 
7117 	if (rc == NFS4_CHKSEQ_OKAY)
7118 		rfs4_update_lease(op->client);
7119 
7120 	return (rc);
7121 }
7122 
7123 static rfs4_chkseq_t
7124 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op,
7125 	nfs_resop4 *resop)
7126 {
7127 	rfs4_chkseq_t rc;
7128 
7129 	rfs4_dbe_lock(op->dbe);
7130 	rc = rfs4_check_seqid(op->open_seqid, op->reply,
7131 	    olo_seqid, resop, FALSE);
7132 	rfs4_dbe_unlock(op->dbe);
7133 
7134 	return (rc);
7135 }
7136 
7137 static rfs4_chkseq_t
7138 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lp, nfs_resop4 *resop)
7139 {
7140 	rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY;
7141 
7142 	rfs4_dbe_lock(lp->dbe);
7143 	if (!lp->skip_seqid_check)
7144 		rc = rfs4_check_seqid(lp->seqid, lp->reply, seqid, resop, TRUE);
7145 	rfs4_dbe_unlock(lp->dbe);
7146 
7147 	return (rc);
7148 }
7149 
7150 static void
7151 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop,
7152 	    struct svc_req *req, struct compound_state *cs)
7153 {
7154 	OPEN4args *args = &argop->nfs_argop4_u.opopen;
7155 	OPEN4res *resp = &resop->nfs_resop4_u.opopen;
7156 	open_owner4 *owner = &args->owner;
7157 	open_claim_type4 claim = args->claim;
7158 	rfs4_client_t *cp;
7159 	rfs4_openowner_t *oo;
7160 	bool_t create;
7161 	bool_t replay = FALSE;
7162 	int can_reclaim;
7163 
7164 	DTRACE_NFSV4_2(op__open__start, struct compound_state *, cs,
7165 	    OPEN4args *, args);
7166 
7167 	if (cs->vp == NULL) {
7168 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7169 		goto end;
7170 	}
7171 
7172 	/*
7173 	 * Need to check clientid and lease expiration first based on
7174 	 * error ordering and incrementing sequence id.
7175 	 */
7176 	cp = rfs4_findclient_by_id(owner->clientid, FALSE);
7177 	if (cp == NULL) {
7178 		*cs->statusp = resp->status =
7179 		    rfs4_check_clientid(&owner->clientid, 0);
7180 		goto end;
7181 	}
7182 
7183 	if (rfs4_lease_expired(cp)) {
7184 		rfs4_client_close(cp);
7185 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7186 		goto end;
7187 	}
7188 	can_reclaim = cp->can_reclaim;
7189 
7190 	/*
7191 	 * Find the open_owner for use from this point forward.  Take
7192 	 * care in updating the sequence id based on the type of error
7193 	 * being returned.
7194 	 */
7195 retry:
7196 	create = TRUE;
7197 	oo = rfs4_findopenowner(owner, &create, args->seqid);
7198 	if (oo == NULL) {
7199 		*cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
7200 		rfs4_client_rele(cp);
7201 		goto end;
7202 	}
7203 
7204 	/* Hold off access to the sequence space while the open is done */
7205 	rfs4_sw_enter(&oo->oo_sw);
7206 
7207 	/*
7208 	 * If the open_owner existed before at the server, then check
7209 	 * the sequence id.
7210 	 */
7211 	if (!create && !oo->postpone_confirm) {
7212 		switch (rfs4_check_open_seqid(args->seqid, oo, resop)) {
7213 		case NFS4_CHKSEQ_BAD:
7214 			if ((args->seqid > oo->open_seqid) &&
7215 			    oo->need_confirm) {
7216 				rfs4_free_opens(oo, TRUE, FALSE);
7217 				rfs4_sw_exit(&oo->oo_sw);
7218 				rfs4_openowner_rele(oo);
7219 				goto retry;
7220 			}
7221 			resp->status = NFS4ERR_BAD_SEQID;
7222 			goto out;
7223 		case NFS4_CHKSEQ_REPLAY: /* replay of previous request */
7224 			replay = TRUE;
7225 			goto out;
7226 		default:
7227 			break;
7228 		}
7229 
7230 		/*
7231 		 * Sequence was ok and open owner exists
7232 		 * check to see if we have yet to see an
7233 		 * open_confirm.
7234 		 */
7235 		if (oo->need_confirm) {
7236 			rfs4_free_opens(oo, TRUE, FALSE);
7237 			rfs4_sw_exit(&oo->oo_sw);
7238 			rfs4_openowner_rele(oo);
7239 			goto retry;
7240 		}
7241 	}
7242 	/* Grace only applies to regular-type OPENs */
7243 	if (rfs4_clnt_in_grace(cp) &&
7244 	    (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) {
7245 		*cs->statusp = resp->status = NFS4ERR_GRACE;
7246 		goto out;
7247 	}
7248 
7249 	/*
7250 	 * If previous state at the server existed then can_reclaim
7251 	 * will be set. If not reply NFS4ERR_NO_GRACE to the
7252 	 * client.
7253 	 */
7254 	if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) {
7255 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
7256 		goto out;
7257 	}
7258 
7259 
7260 	/*
7261 	 * Reject the open if the client has missed the grace period
7262 	 */
7263 	if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) {
7264 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
7265 		goto out;
7266 	}
7267 
7268 	/* Couple of up-front bookkeeping items */
7269 	if (oo->need_confirm) {
7270 		/*
7271 		 * If this is a reclaim OPEN then we should not ask
7272 		 * for a confirmation of the open_owner per the
7273 		 * protocol specification.
7274 		 */
7275 		if (claim == CLAIM_PREVIOUS)
7276 			oo->need_confirm = FALSE;
7277 		else
7278 			resp->rflags |= OPEN4_RESULT_CONFIRM;
7279 	}
7280 	resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX;
7281 
7282 	/*
7283 	 * If there is an unshared filesystem mounted on this vnode,
7284 	 * do not allow to open/create in this directory.
7285 	 */
7286 	if (vn_ismntpt(cs->vp)) {
7287 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
7288 		goto out;
7289 	}
7290 
7291 	/*
7292 	 * access must READ, WRITE, or BOTH.  No access is invalid.
7293 	 * deny can be READ, WRITE, BOTH, or NONE.
7294 	 * bits not defined for access/deny are invalid.
7295 	 */
7296 	if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) ||
7297 	    (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) ||
7298 	    (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) {
7299 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7300 		goto out;
7301 	}
7302 
7303 
7304 	/*
7305 	 * make sure attrset is zero before response is built.
7306 	 */
7307 	resp->attrset = 0;
7308 
7309 	switch (claim) {
7310 	case CLAIM_NULL:
7311 		rfs4_do_opennull(cs, req, args, oo, resp);
7312 		break;
7313 	case CLAIM_PREVIOUS:
7314 		rfs4_do_openprev(cs, req, args, oo, resp);
7315 		break;
7316 	case CLAIM_DELEGATE_CUR:
7317 		rfs4_do_opendelcur(cs, req, args, oo, resp);
7318 		break;
7319 	case CLAIM_DELEGATE_PREV:
7320 		rfs4_do_opendelprev(cs, req, args, oo, resp);
7321 		break;
7322 	default:
7323 		resp->status = NFS4ERR_INVAL;
7324 		break;
7325 	}
7326 
7327 out:
7328 	rfs4_client_rele(cp);
7329 
7330 	/* Catch sequence id handling here to make it a little easier */
7331 	switch (resp->status) {
7332 	case NFS4ERR_BADXDR:
7333 	case NFS4ERR_BAD_SEQID:
7334 	case NFS4ERR_BAD_STATEID:
7335 	case NFS4ERR_NOFILEHANDLE:
7336 	case NFS4ERR_RESOURCE:
7337 	case NFS4ERR_STALE_CLIENTID:
7338 	case NFS4ERR_STALE_STATEID:
7339 		/*
7340 		 * The protocol states that if any of these errors are
7341 		 * being returned, the sequence id should not be
7342 		 * incremented.  Any other return requires an
7343 		 * increment.
7344 		 */
7345 		break;
7346 	default:
7347 		/* Always update the lease in this case */
7348 		rfs4_update_lease(oo->client);
7349 
7350 		/* Regular response - copy the result */
7351 		if (!replay)
7352 			rfs4_update_open_resp(oo, resop, &cs->fh);
7353 
7354 		/*
7355 		 * REPLAY case: Only if the previous response was OK
7356 		 * do we copy the filehandle.  If not OK, no
7357 		 * filehandle to copy.
7358 		 */
7359 		if (replay == TRUE &&
7360 		    resp->status == NFS4_OK &&
7361 		    oo->reply_fh.nfs_fh4_val) {
7362 			/*
7363 			 * If this is a replay, we must restore the
7364 			 * current filehandle/vp to that of what was
7365 			 * returned originally.  Try our best to do
7366 			 * it.
7367 			 */
7368 			nfs_fh4_fmt_t *fh_fmtp =
7369 			    (nfs_fh4_fmt_t *)oo->reply_fh.nfs_fh4_val;
7370 
7371 			cs->exi = checkexport4(&fh_fmtp->fh4_fsid,
7372 			    (fid_t *)&fh_fmtp->fh4_xlen, NULL);
7373 
7374 			if (cs->exi == NULL) {
7375 				resp->status = NFS4ERR_STALE;
7376 				goto finish;
7377 			}
7378 
7379 			VN_RELE(cs->vp);
7380 
7381 			cs->vp = nfs4_fhtovp(&oo->reply_fh, cs->exi,
7382 			    &resp->status);
7383 
7384 			if (cs->vp == NULL)
7385 				goto finish;
7386 
7387 			nfs_fh4_copy(&oo->reply_fh, &cs->fh);
7388 		}
7389 
7390 		/*
7391 		 * If this was a replay, no need to update the
7392 		 * sequence id. If the open_owner was not created on
7393 		 * this pass, then update.  The first use of an
7394 		 * open_owner will not bump the sequence id.
7395 		 */
7396 		if (replay == FALSE && !create)
7397 			rfs4_update_open_sequence(oo);
7398 		/*
7399 		 * If the client is receiving an error and the
7400 		 * open_owner needs to be confirmed, there is no way
7401 		 * to notify the client of this fact ignoring the fact
7402 		 * that the server has no method of returning a
7403 		 * stateid to confirm.  Therefore, the server needs to
7404 		 * mark this open_owner in a way as to avoid the
7405 		 * sequence id checking the next time the client uses
7406 		 * this open_owner.
7407 		 */
7408 		if (resp->status != NFS4_OK && oo->need_confirm)
7409 			oo->postpone_confirm = TRUE;
7410 		/*
7411 		 * If OK response then clear the postpone flag and
7412 		 * reset the sequence id to keep in sync with the
7413 		 * client.
7414 		 */
7415 		if (resp->status == NFS4_OK && oo->postpone_confirm) {
7416 			oo->postpone_confirm = FALSE;
7417 			oo->open_seqid = args->seqid;
7418 		}
7419 		break;
7420 	}
7421 
7422 finish:
7423 	*cs->statusp = resp->status;
7424 
7425 	rfs4_sw_exit(&oo->oo_sw);
7426 	rfs4_openowner_rele(oo);
7427 
7428 end:
7429 	DTRACE_NFSV4_2(op__open__done, struct compound_state *, cs,
7430 	    OPEN4res *, resp);
7431 }
7432 
7433 /*ARGSUSED*/
7434 void
7435 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
7436 		    struct svc_req *req, struct compound_state *cs)
7437 {
7438 	OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm;
7439 	OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm;
7440 	rfs4_state_t *sp;
7441 	nfsstat4 status;
7442 
7443 	DTRACE_NFSV4_2(op__open__confirm__start, struct compound_state *, cs,
7444 	    OPEN_CONFIRM4args *, args);
7445 
7446 	if (cs->vp == NULL) {
7447 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7448 		goto out;
7449 	}
7450 
7451 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
7452 	if (status != NFS4_OK) {
7453 		*cs->statusp = resp->status = status;
7454 		goto out;
7455 	}
7456 
7457 	/* Ensure specified filehandle matches */
7458 	if (cs->vp != sp->finfo->vp) {
7459 		rfs4_state_rele(sp);
7460 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7461 		goto out;
7462 	}
7463 
7464 	/* hold off other access to open_owner while we tinker */
7465 	rfs4_sw_enter(&sp->owner->oo_sw);
7466 
7467 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7468 	case NFS4_CHECK_STATEID_OKAY:
7469 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7470 		    resop) != 0) {
7471 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7472 			break;
7473 		}
7474 		/*
7475 		 * If it is the appropriate stateid and determined to
7476 		 * be "OKAY" then this means that the stateid does not
7477 		 * need to be confirmed and the client is in error for
7478 		 * sending an OPEN_CONFIRM.
7479 		 */
7480 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7481 		break;
7482 	case NFS4_CHECK_STATEID_OLD:
7483 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7484 		break;
7485 	case NFS4_CHECK_STATEID_BAD:
7486 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7487 		break;
7488 	case NFS4_CHECK_STATEID_EXPIRED:
7489 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7490 		break;
7491 	case NFS4_CHECK_STATEID_CLOSED:
7492 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7493 		break;
7494 	case NFS4_CHECK_STATEID_REPLAY:
7495 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
7496 		case NFS4_CHKSEQ_OKAY:
7497 			/*
7498 			 * This is replayed stateid; if seqid matches
7499 			 * next expected, then client is using wrong seqid.
7500 			 */
7501 			/* fall through */
7502 		case NFS4_CHKSEQ_BAD:
7503 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7504 			break;
7505 		case NFS4_CHKSEQ_REPLAY:
7506 			/*
7507 			 * Note this case is the duplicate case so
7508 			 * resp->status is already set.
7509 			 */
7510 			*cs->statusp = resp->status;
7511 			rfs4_update_lease(sp->owner->client);
7512 			break;
7513 		}
7514 		break;
7515 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7516 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7517 		    resop) != NFS4_CHKSEQ_OKAY) {
7518 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7519 			break;
7520 		}
7521 		*cs->statusp = resp->status = NFS4_OK;
7522 
7523 		next_stateid(&sp->stateid);
7524 		resp->open_stateid = sp->stateid.stateid;
7525 		sp->owner->need_confirm = FALSE;
7526 		rfs4_update_lease(sp->owner->client);
7527 		rfs4_update_open_sequence(sp->owner);
7528 		rfs4_update_open_resp(sp->owner, resop, NULL);
7529 		break;
7530 	default:
7531 		ASSERT(FALSE);
7532 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7533 		break;
7534 	}
7535 	rfs4_sw_exit(&sp->owner->oo_sw);
7536 	rfs4_state_rele(sp);
7537 
7538 out:
7539 	DTRACE_NFSV4_2(op__open__confirm__done, struct compound_state *, cs,
7540 	    OPEN_CONFIRM4res *, resp);
7541 }
7542 
7543 /*ARGSUSED*/
7544 void
7545 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop,
7546 		    struct svc_req *req, struct compound_state *cs)
7547 {
7548 	OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade;
7549 	OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade;
7550 	uint32_t access = args->share_access;
7551 	uint32_t deny = args->share_deny;
7552 	nfsstat4 status;
7553 	rfs4_state_t *sp;
7554 	rfs4_file_t *fp;
7555 	int fflags = 0;
7556 
7557 	DTRACE_NFSV4_2(op__open__downgrade__start, struct compound_state *, cs,
7558 	    OPEN_DOWNGRADE4args *, args);
7559 
7560 	if (cs->vp == NULL) {
7561 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7562 		goto out;
7563 	}
7564 
7565 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
7566 	if (status != NFS4_OK) {
7567 		*cs->statusp = resp->status = status;
7568 		goto out;
7569 	}
7570 
7571 	/* Ensure specified filehandle matches */
7572 	if (cs->vp != sp->finfo->vp) {
7573 		rfs4_state_rele(sp);
7574 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7575 		goto out;
7576 	}
7577 
7578 	/* hold off other access to open_owner while we tinker */
7579 	rfs4_sw_enter(&sp->owner->oo_sw);
7580 
7581 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7582 	case NFS4_CHECK_STATEID_OKAY:
7583 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
7584 		    resop) != NFS4_CHKSEQ_OKAY) {
7585 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7586 			goto end;
7587 		}
7588 		break;
7589 	case NFS4_CHECK_STATEID_OLD:
7590 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7591 		goto end;
7592 	case NFS4_CHECK_STATEID_BAD:
7593 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7594 		goto end;
7595 	case NFS4_CHECK_STATEID_EXPIRED:
7596 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7597 		goto end;
7598 	case NFS4_CHECK_STATEID_CLOSED:
7599 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7600 		goto end;
7601 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7602 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7603 		goto end;
7604 	case NFS4_CHECK_STATEID_REPLAY:
7605 		/* Check the sequence id for the open owner */
7606 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
7607 		case NFS4_CHKSEQ_OKAY:
7608 			/*
7609 			 * This is replayed stateid; if seqid matches
7610 			 * next expected, then client is using wrong seqid.
7611 			 */
7612 			/* fall through */
7613 		case NFS4_CHKSEQ_BAD:
7614 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7615 			goto end;
7616 		case NFS4_CHKSEQ_REPLAY:
7617 			/*
7618 			 * Note this case is the duplicate case so
7619 			 * resp->status is already set.
7620 			 */
7621 			*cs->statusp = resp->status;
7622 			rfs4_update_lease(sp->owner->client);
7623 			goto end;
7624 		}
7625 		break;
7626 	default:
7627 		ASSERT(FALSE);
7628 		break;
7629 	}
7630 
7631 	rfs4_dbe_lock(sp->dbe);
7632 	/*
7633 	 * Check that the new access modes and deny modes are valid.
7634 	 * Check that no invalid bits are set.
7635 	 */
7636 	if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) ||
7637 	    (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) {
7638 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7639 		rfs4_update_open_sequence(sp->owner);
7640 		rfs4_dbe_unlock(sp->dbe);
7641 		goto end;
7642 	}
7643 
7644 	/*
7645 	 * The new modes must be a subset of the current modes and
7646 	 * the access must specify at least one mode. To test that
7647 	 * the new mode is a subset of the current modes we bitwise
7648 	 * AND them together and check that the result equals the new
7649 	 * mode. For example:
7650 	 * New mode, access == R and current mode, sp->share_access  == RW
7651 	 * access & sp->share_access == R == access, so the new access mode
7652 	 * is valid. Consider access == RW, sp->share_access = R
7653 	 * access & sp->share_access == R != access, so the new access mode
7654 	 * is invalid.
7655 	 */
7656 	if ((access & sp->share_access) != access ||
7657 	    (deny & sp->share_deny) != deny ||
7658 	    (access &
7659 	    (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) {
7660 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7661 		rfs4_update_open_sequence(sp->owner);
7662 		rfs4_dbe_unlock(sp->dbe);
7663 		goto end;
7664 	}
7665 
7666 	/*
7667 	 * Release any share locks associated with this stateID.
7668 	 * Strictly speaking, this violates the spec because the
7669 	 * spec effectively requires that open downgrade be atomic.
7670 	 * At present, fs_shrlock does not have this capability.
7671 	 */
7672 	rfs4_dbe_unlock(sp->dbe);
7673 	rfs4_unshare(sp);
7674 	rfs4_dbe_lock(sp->dbe);
7675 
7676 	fp = sp->finfo;
7677 	rfs4_dbe_lock(fp->dbe);
7678 
7679 	/*
7680 	 * If the current mode has deny read and the new mode
7681 	 * does not, decrement the number of deny read mode bits
7682 	 * and if it goes to zero turn off the deny read bit
7683 	 * on the file.
7684 	 */
7685 	if ((sp->share_deny & OPEN4_SHARE_DENY_READ) &&
7686 	    (deny & OPEN4_SHARE_DENY_READ) == 0) {
7687 		fp->deny_read--;
7688 		if (fp->deny_read == 0)
7689 			fp->share_deny &= ~OPEN4_SHARE_DENY_READ;
7690 	}
7691 
7692 	/*
7693 	 * If the current mode has deny write and the new mode
7694 	 * does not, decrement the number of deny write mode bits
7695 	 * and if it goes to zero turn off the deny write bit
7696 	 * on the file.
7697 	 */
7698 	if ((sp->share_deny & OPEN4_SHARE_DENY_WRITE) &&
7699 	    (deny & OPEN4_SHARE_DENY_WRITE) == 0) {
7700 		fp->deny_write--;
7701 		if (fp->deny_write == 0)
7702 			fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE;
7703 	}
7704 
7705 	/*
7706 	 * If the current mode has access read and the new mode
7707 	 * does not, decrement the number of access read mode bits
7708 	 * and if it goes to zero turn off the access read bit
7709 	 * on the file.  set fflags to FREAD for the call to
7710 	 * vn_open_downgrade().
7711 	 */
7712 	if ((sp->share_access & OPEN4_SHARE_ACCESS_READ) &&
7713 	    (access & OPEN4_SHARE_ACCESS_READ) == 0) {
7714 		fp->access_read--;
7715 		if (fp->access_read == 0)
7716 			fp->share_access &= ~OPEN4_SHARE_ACCESS_READ;
7717 		fflags |= FREAD;
7718 	}
7719 
7720 	/*
7721 	 * If the current mode has access write and the new mode
7722 	 * does not, decrement the number of access write mode bits
7723 	 * and if it goes to zero turn off the access write bit
7724 	 * on the file.  set fflags to FWRITE for the call to
7725 	 * vn_open_downgrade().
7726 	 */
7727 	if ((sp->share_access & OPEN4_SHARE_ACCESS_WRITE) &&
7728 	    (access & OPEN4_SHARE_ACCESS_WRITE) == 0) {
7729 		fp->access_write--;
7730 		if (fp->access_write == 0)
7731 			fp->share_deny &= ~OPEN4_SHARE_ACCESS_WRITE;
7732 		fflags |= FWRITE;
7733 	}
7734 
7735 	/* Set the new access and deny modes */
7736 	sp->share_access = access;
7737 	sp->share_deny = deny;
7738 	/* Check that the file is still accessible */
7739 	ASSERT(fp->share_access);
7740 
7741 	rfs4_dbe_unlock(fp->dbe);
7742 
7743 	rfs4_dbe_unlock(sp->dbe);
7744 
7745 	if ((status = rfs4_share(sp)) != NFS4_OK) {
7746 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7747 		rfs4_update_open_sequence(sp->owner);
7748 		goto end;
7749 	}
7750 
7751 	/*
7752 	 * we successfully downgraded the share lock, now we need to downgrade
7753 	 * the open.  it is possible that the downgrade was only for a deny
7754 	 * mode and we have nothing else to do.
7755 	 */
7756 	if ((fflags & (FREAD|FWRITE)) != 0)
7757 		vn_open_downgrade(cs->vp, fflags);
7758 
7759 	rfs4_dbe_lock(sp->dbe);
7760 
7761 	/* Update the stateid */
7762 	next_stateid(&sp->stateid);
7763 	resp->open_stateid = sp->stateid.stateid;
7764 
7765 	rfs4_dbe_unlock(sp->dbe);
7766 
7767 	*cs->statusp = resp->status = NFS4_OK;
7768 	/* Update the lease */
7769 	rfs4_update_lease(sp->owner->client);
7770 	/* And the sequence */
7771 	rfs4_update_open_sequence(sp->owner);
7772 	rfs4_update_open_resp(sp->owner, resop, NULL);
7773 
7774 end:
7775 	rfs4_sw_exit(&sp->owner->oo_sw);
7776 	rfs4_state_rele(sp);
7777 out:
7778 	DTRACE_NFSV4_2(op__open__downgrade__done, struct compound_state *, cs,
7779 	    OPEN_DOWNGRADE4res *, resp);
7780 }
7781 
7782 /*
7783  * The logic behind this function is detailed in the NFSv4 RFC in the
7784  * SETCLIENTID operation description under IMPLEMENTATION.  Refer to
7785  * that section for explicit guidance to server behavior for
7786  * SETCLIENTID.
7787  */
7788 void
7789 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop,
7790 		    struct svc_req *req, struct compound_state *cs)
7791 {
7792 	SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid;
7793 	SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid;
7794 	rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed;
7795 	bool_t create = TRUE;
7796 	char *addr, *netid;
7797 	int len;
7798 
7799 	DTRACE_NFSV4_2(op__setclientid__start, struct compound_state *, cs,
7800 	    SETCLIENTID4args *, args);
7801 retry:
7802 	newcp = cp_confirmed = cp_unconfirmed = NULL;
7803 
7804 	/*
7805 	 * In search of an EXISTING client matching the incoming
7806 	 * request to establish a new client identifier at the server
7807 	 */
7808 	create = TRUE;
7809 	cp = rfs4_findclient(&args->client, &create, NULL);
7810 
7811 	/* Should never happen */
7812 	ASSERT(cp != NULL);
7813 
7814 	if (cp == NULL) {
7815 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
7816 		goto out;
7817 	}
7818 
7819 	/*
7820 	 * Easiest case. Client identifier is newly created and is
7821 	 * unconfirmed.  Also note that for this case, no other
7822 	 * entries exist for the client identifier.  Nothing else to
7823 	 * check.  Just setup the response and respond.
7824 	 */
7825 	if (create) {
7826 		*cs->statusp = res->status = NFS4_OK;
7827 		res->SETCLIENTID4res_u.resok4.clientid = cp->clientid;
7828 		res->SETCLIENTID4res_u.resok4.setclientid_confirm =
7829 		    cp->confirm_verf;
7830 		/* Setup callback information; CB_NULL confirmation later */
7831 		rfs4_client_setcb(cp, &args->callback, args->callback_ident);
7832 
7833 		rfs4_client_rele(cp);
7834 		goto out;
7835 	}
7836 
7837 	/*
7838 	 * An existing, confirmed client may exist but it may not have
7839 	 * been active for at least one lease period.  If so, then
7840 	 * "close" the client and create a new client identifier
7841 	 */
7842 	if (rfs4_lease_expired(cp)) {
7843 		rfs4_client_close(cp);
7844 		goto retry;
7845 	}
7846 
7847 	if (cp->need_confirm == TRUE)
7848 		cp_unconfirmed = cp;
7849 	else
7850 		cp_confirmed = cp;
7851 
7852 	cp = NULL;
7853 
7854 	/*
7855 	 * We have a confirmed client, now check for an
7856 	 * unconfimred entry
7857 	 */
7858 	if (cp_confirmed) {
7859 		/* If creds don't match then client identifier is inuse */
7860 		if (!creds_ok(cp_confirmed->cr_set, req, cs)) {
7861 			rfs4_cbinfo_t *cbp;
7862 			/*
7863 			 * Some one else has established this client
7864 			 * id. Try and say * who they are. We will use
7865 			 * the call back address supplied by * the
7866 			 * first client.
7867 			 */
7868 			*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
7869 
7870 			addr = netid = NULL;
7871 
7872 			cbp = &cp_confirmed->cbinfo;
7873 			if (cbp->cb_callback.cb_location.r_addr &&
7874 			    cbp->cb_callback.cb_location.r_netid) {
7875 				cb_client4 *cbcp = &cbp->cb_callback;
7876 
7877 				len = strlen(cbcp->cb_location.r_addr)+1;
7878 				addr = kmem_alloc(len, KM_SLEEP);
7879 				bcopy(cbcp->cb_location.r_addr, addr, len);
7880 				len = strlen(cbcp->cb_location.r_netid)+1;
7881 				netid = kmem_alloc(len, KM_SLEEP);
7882 				bcopy(cbcp->cb_location.r_netid, netid, len);
7883 			}
7884 
7885 			res->SETCLIENTID4res_u.client_using.r_addr = addr;
7886 			res->SETCLIENTID4res_u.client_using.r_netid = netid;
7887 
7888 			rfs4_client_rele(cp_confirmed);
7889 		}
7890 
7891 		/*
7892 		 * Confirmed, creds match, and verifier matches; must
7893 		 * be an update of the callback info
7894 		 */
7895 		if (cp_confirmed->nfs_client.verifier ==
7896 		    args->client.verifier) {
7897 			/* Setup callback information */
7898 			rfs4_client_setcb(cp_confirmed, &args->callback,
7899 			    args->callback_ident);
7900 
7901 			/* everything okay -- move ahead */
7902 			*cs->statusp = res->status = NFS4_OK;
7903 			res->SETCLIENTID4res_u.resok4.clientid =
7904 			    cp_confirmed->clientid;
7905 
7906 			/* update the confirm_verifier and return it */
7907 			rfs4_client_scv_next(cp_confirmed);
7908 			res->SETCLIENTID4res_u.resok4.setclientid_confirm =
7909 			    cp_confirmed->confirm_verf;
7910 
7911 			rfs4_client_rele(cp_confirmed);
7912 			goto out;
7913 		}
7914 
7915 		/*
7916 		 * Creds match but the verifier doesn't.  Must search
7917 		 * for an unconfirmed client that would be replaced by
7918 		 * this request.
7919 		 */
7920 		create = FALSE;
7921 		cp_unconfirmed = rfs4_findclient(&args->client, &create,
7922 		    cp_confirmed);
7923 	}
7924 
7925 	/*
7926 	 * At this point, we have taken care of the brand new client
7927 	 * struct, INUSE case, update of an existing, and confirmed
7928 	 * client struct.
7929 	 */
7930 
7931 	/*
7932 	 * check to see if things have changed while we originally
7933 	 * picked up the client struct.  If they have, then return and
7934 	 * retry the processing of this SETCLIENTID request.
7935 	 */
7936 	if (cp_unconfirmed) {
7937 		rfs4_dbe_lock(cp_unconfirmed->dbe);
7938 		if (!cp_unconfirmed->need_confirm) {
7939 			rfs4_dbe_unlock(cp_unconfirmed->dbe);
7940 			rfs4_client_rele(cp_unconfirmed);
7941 			if (cp_confirmed)
7942 				rfs4_client_rele(cp_confirmed);
7943 			goto retry;
7944 		}
7945 		/* do away with the old unconfirmed one */
7946 		rfs4_dbe_invalidate(cp_unconfirmed->dbe);
7947 		rfs4_dbe_unlock(cp_unconfirmed->dbe);
7948 		rfs4_client_rele(cp_unconfirmed);
7949 		cp_unconfirmed = NULL;
7950 	}
7951 
7952 	/*
7953 	 * This search will temporarily hide the confirmed client
7954 	 * struct while a new client struct is created as the
7955 	 * unconfirmed one.
7956 	 */
7957 	create = TRUE;
7958 	newcp = rfs4_findclient(&args->client, &create, cp_confirmed);
7959 
7960 	ASSERT(newcp != NULL);
7961 
7962 	if (newcp == NULL) {
7963 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
7964 		rfs4_client_rele(cp_confirmed);
7965 		goto out;
7966 	}
7967 
7968 	/*
7969 	 * If one was not created, then a similar request must be in
7970 	 * process so release and start over with this one
7971 	 */
7972 	if (create != TRUE) {
7973 		rfs4_client_rele(newcp);
7974 		if (cp_confirmed)
7975 			rfs4_client_rele(cp_confirmed);
7976 		goto retry;
7977 	}
7978 
7979 	*cs->statusp = res->status = NFS4_OK;
7980 	res->SETCLIENTID4res_u.resok4.clientid = newcp->clientid;
7981 	res->SETCLIENTID4res_u.resok4.setclientid_confirm = newcp->confirm_verf;
7982 	/* Setup callback information; CB_NULL confirmation later */
7983 	rfs4_client_setcb(newcp, &args->callback, args->callback_ident);
7984 
7985 	newcp->cp_confirmed = cp_confirmed;
7986 
7987 	rfs4_client_rele(newcp);
7988 
7989 out:
7990 	DTRACE_NFSV4_2(op__setclientid__done, struct compound_state *, cs,
7991 	    SETCLIENTID4res *, res);
7992 }
7993 
7994 /*ARGSUSED*/
7995 void
7996 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
7997 			    struct svc_req *req, struct compound_state *cs)
7998 {
7999 	SETCLIENTID_CONFIRM4args *args =
8000 	    &argop->nfs_argop4_u.opsetclientid_confirm;
8001 	SETCLIENTID_CONFIRM4res *res =
8002 	    &resop->nfs_resop4_u.opsetclientid_confirm;
8003 	rfs4_client_t *cp, *cptoclose = NULL;
8004 
8005 	DTRACE_NFSV4_2(op__setclientid__confirm__start,
8006 	    struct compound_state *, cs,
8007 	    SETCLIENTID_CONFIRM4args *, args);
8008 
8009 	*cs->statusp = res->status = NFS4_OK;
8010 
8011 	cp = rfs4_findclient_by_id(args->clientid, TRUE);
8012 
8013 	if (cp == NULL) {
8014 		*cs->statusp = res->status =
8015 		    rfs4_check_clientid(&args->clientid, 1);
8016 		goto out;
8017 	}
8018 
8019 	if (!creds_ok(cp, req, cs)) {
8020 		*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
8021 		rfs4_client_rele(cp);
8022 		goto out;
8023 	}
8024 
8025 	/* If the verifier doesn't match, the record doesn't match */
8026 	if (cp->confirm_verf != args->setclientid_confirm) {
8027 		*cs->statusp = res->status = NFS4ERR_STALE_CLIENTID;
8028 		rfs4_client_rele(cp);
8029 		goto out;
8030 	}
8031 
8032 	rfs4_dbe_lock(cp->dbe);
8033 	cp->need_confirm = FALSE;
8034 	if (cp->cp_confirmed) {
8035 		cptoclose = cp->cp_confirmed;
8036 		cptoclose->ss_remove = 1;
8037 		cp->cp_confirmed = NULL;
8038 	}
8039 
8040 	/*
8041 	 * Update the client's associated server instance, if it's changed
8042 	 * since the client was created.
8043 	 */
8044 	if (rfs4_servinst(cp) != rfs4_cur_servinst)
8045 		rfs4_servinst_assign(cp, rfs4_cur_servinst);
8046 
8047 	/*
8048 	 * Record clientid in stable storage.
8049 	 * Must be done after server instance has been assigned.
8050 	 */
8051 	rfs4_ss_clid(cp, req);
8052 
8053 	rfs4_dbe_unlock(cp->dbe);
8054 
8055 	if (cptoclose)
8056 		/* don't need to rele, client_close does it */
8057 		rfs4_client_close(cptoclose);
8058 
8059 	/* If needed, initiate CB_NULL call for callback path */
8060 	rfs4_deleg_cb_check(cp);
8061 	rfs4_update_lease(cp);
8062 
8063 	/*
8064 	 * Check to see if client can perform reclaims
8065 	 */
8066 	rfs4_ss_chkclid(cp);
8067 
8068 	rfs4_client_rele(cp);
8069 
8070 out:
8071 	DTRACE_NFSV4_2(op__setclientid__confirm__done,
8072 	    struct compound_state *, cs,
8073 	    SETCLIENTID_CONFIRM4 *, res);
8074 }
8075 
8076 
8077 /*ARGSUSED*/
8078 void
8079 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop,
8080 	    struct svc_req *req, struct compound_state *cs)
8081 {
8082 	CLOSE4args *args = &argop->nfs_argop4_u.opclose;
8083 	CLOSE4res *resp = &resop->nfs_resop4_u.opclose;
8084 	rfs4_state_t *sp;
8085 	nfsstat4 status;
8086 
8087 	DTRACE_NFSV4_2(op__close__start, struct compound_state *, cs,
8088 	    CLOSE4args *, args);
8089 
8090 	if (cs->vp == NULL) {
8091 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8092 		goto out;
8093 	}
8094 
8095 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID);
8096 	if (status != NFS4_OK) {
8097 		*cs->statusp = resp->status = status;
8098 		goto out;
8099 	}
8100 
8101 	/* Ensure specified filehandle matches */
8102 	if (cs->vp != sp->finfo->vp) {
8103 		rfs4_state_rele(sp);
8104 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8105 		goto out;
8106 	}
8107 
8108 	/* hold off other access to open_owner while we tinker */
8109 	rfs4_sw_enter(&sp->owner->oo_sw);
8110 
8111 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
8112 	case NFS4_CHECK_STATEID_OKAY:
8113 		if (rfs4_check_open_seqid(args->seqid, sp->owner,
8114 		    resop) != NFS4_CHKSEQ_OKAY) {
8115 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8116 			goto end;
8117 		}
8118 		break;
8119 	case NFS4_CHECK_STATEID_OLD:
8120 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8121 		goto end;
8122 	case NFS4_CHECK_STATEID_BAD:
8123 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8124 		goto end;
8125 	case NFS4_CHECK_STATEID_EXPIRED:
8126 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8127 		goto end;
8128 	case NFS4_CHECK_STATEID_CLOSED:
8129 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8130 		goto end;
8131 	case NFS4_CHECK_STATEID_UNCONFIRMED:
8132 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8133 		goto end;
8134 	case NFS4_CHECK_STATEID_REPLAY:
8135 		/* Check the sequence id for the open owner */
8136 		switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) {
8137 		case NFS4_CHKSEQ_OKAY:
8138 			/*
8139 			 * This is replayed stateid; if seqid matches
8140 			 * next expected, then client is using wrong seqid.
8141 			 */
8142 			/* FALL THROUGH */
8143 		case NFS4_CHKSEQ_BAD:
8144 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8145 			goto end;
8146 		case NFS4_CHKSEQ_REPLAY:
8147 			/*
8148 			 * Note this case is the duplicate case so
8149 			 * resp->status is already set.
8150 			 */
8151 			*cs->statusp = resp->status;
8152 			rfs4_update_lease(sp->owner->client);
8153 			goto end;
8154 		}
8155 		break;
8156 	default:
8157 		ASSERT(FALSE);
8158 		break;
8159 	}
8160 
8161 	rfs4_dbe_lock(sp->dbe);
8162 
8163 	/* Update the stateid. */
8164 	next_stateid(&sp->stateid);
8165 	resp->open_stateid = sp->stateid.stateid;
8166 
8167 	rfs4_dbe_unlock(sp->dbe);
8168 
8169 	rfs4_update_lease(sp->owner->client);
8170 	rfs4_update_open_sequence(sp->owner);
8171 	rfs4_update_open_resp(sp->owner, resop, NULL);
8172 
8173 	rfs4_state_close(sp, FALSE, FALSE, cs->cr);
8174 
8175 	*cs->statusp = resp->status = status;
8176 
8177 end:
8178 	rfs4_sw_exit(&sp->owner->oo_sw);
8179 	rfs4_state_rele(sp);
8180 out:
8181 	DTRACE_NFSV4_2(op__close__done, struct compound_state *, cs,
8182 	    CLOSE4res *, resp);
8183 }
8184 
8185 /*
8186  * Manage the counts on the file struct and close all file locks
8187  */
8188 /*ARGSUSED*/
8189 void
8190 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr,
8191 	bool_t close_of_client)
8192 {
8193 	rfs4_file_t *fp = sp->finfo;
8194 	rfs4_lo_state_t *lsp;
8195 	struct shrlock shr;
8196 	struct shr_locowner shr_loco;
8197 	int fflags, s_access, s_deny;
8198 
8199 	fflags = s_access = s_deny = 0;
8200 	/*
8201 	 * Decrement the count for each access and deny bit that this
8202 	 * state has contributed to the file. If the file counts go to zero
8203 	 * clear the appropriate bit in the appropriate mask.
8204 	 */
8205 
8206 	if (sp->share_access & OPEN4_SHARE_ACCESS_READ) {
8207 		fp->access_read--;
8208 		fflags |= FREAD;
8209 		s_access |= F_RDACC;
8210 		if (fp->access_read == 0)
8211 			fp->share_access &= ~OPEN4_SHARE_ACCESS_READ;
8212 	}
8213 	if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) {
8214 		fp->access_write--;
8215 		fflags |= FWRITE;
8216 		s_access |= F_WRACC;
8217 		if (fp->access_write == 0)
8218 			fp->share_access &= ~OPEN4_SHARE_ACCESS_WRITE;
8219 	}
8220 	if (sp->share_deny & OPEN4_SHARE_DENY_READ) {
8221 		fp->deny_read--;
8222 		s_deny |= F_RDDNY;
8223 		if (fp->deny_read == 0)
8224 			fp->share_deny &= ~OPEN4_SHARE_DENY_READ;
8225 	}
8226 	if (sp->share_deny & OPEN4_SHARE_DENY_WRITE) {
8227 		fp->deny_write--;
8228 		s_deny |= F_WRDNY;
8229 		if (fp->deny_write == 0)
8230 			fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE;
8231 	}
8232 
8233 	/*
8234 	 * If this call is part of the larger closing down of client
8235 	 * state then it is just easier to release all locks
8236 	 * associated with this client instead of going through each
8237 	 * individual file and cleaning locks there.
8238 	 */
8239 	if (close_of_client) {
8240 		if (sp->owner->client->unlksys_completed == FALSE &&
8241 		    sp->lockownerlist.next->lsp != NULL &&
8242 		    sp->owner->client->sysidt != LM_NOSYSID) {
8243 			/* Is the PxFS kernel module loaded? */
8244 			if (lm_remove_file_locks != NULL) {
8245 				int new_sysid;
8246 
8247 				/* Encode the cluster nodeid in new sysid */
8248 				new_sysid = sp->owner->client->sysidt;
8249 				lm_set_nlmid_flk(&new_sysid);
8250 
8251 				/*
8252 				 * This PxFS routine removes file locks for a
8253 				 * client over all nodes of a cluster.
8254 				 */
8255 				NFS4_DEBUG(rfs4_debug, (CE_NOTE,
8256 				    "lm_remove_file_locks(sysid=0x%x)\n",
8257 				    new_sysid));
8258 				(*lm_remove_file_locks)(new_sysid);
8259 			} else {
8260 				struct flock64 flk;
8261 
8262 				/* Release all locks for this client */
8263 				flk.l_type = F_UNLKSYS;
8264 				flk.l_whence = 0;
8265 				flk.l_start = 0;
8266 				flk.l_len = 0;
8267 				flk.l_sysid = sp->owner->client->sysidt;
8268 				flk.l_pid = 0;
8269 				(void) VOP_FRLOCK(sp->finfo->vp, F_SETLK, &flk,
8270 				    F_REMOTELOCK | FREAD | FWRITE,
8271 				    (u_offset_t)0, NULL, CRED(), NULL);
8272 			}
8273 
8274 			sp->owner->client->unlksys_completed = TRUE;
8275 		}
8276 	}
8277 
8278 	/*
8279 	 * Release all locks on this file by this lock owner or at
8280 	 * least mark the locks as having been released
8281 	 */
8282 	for (lsp = sp->lockownerlist.next->lsp; lsp != NULL;
8283 	    lsp = lsp->lockownerlist.next->lsp) {
8284 
8285 		lsp->locks_cleaned = TRUE;
8286 
8287 		/* Was this already taken care of above? */
8288 		if (!close_of_client &&
8289 		    sp->owner->client->sysidt != LM_NOSYSID)
8290 			(void) cleanlocks(sp->finfo->vp, lsp->locker->pid,
8291 			    lsp->locker->client->sysidt);
8292 	}
8293 
8294 	/*
8295 	 * Release any shrlocks associated with this open state ID.
8296 	 * This must be done before the rfs4_state gets marked closed.
8297 	 */
8298 	if (sp->owner->client->sysidt != LM_NOSYSID) {
8299 		shr.s_access = s_access;
8300 		shr.s_deny = s_deny;
8301 		shr.s_pid = rfs4_dbe_getid(sp->owner->dbe);
8302 		shr.s_sysid = sp->owner->client->sysidt;
8303 		shr_loco.sl_pid = shr.s_pid;
8304 		shr_loco.sl_id = shr.s_sysid;
8305 		shr.s_owner = (caddr_t)&shr_loco;
8306 		shr.s_own_len = sizeof (shr_loco);
8307 		(void) vop_shrlock(sp->finfo->vp, F_UNSHARE, &shr, fflags);
8308 	}
8309 
8310 	(void) VOP_CLOSE(fp->vp, fflags, 1, (offset_t)0, cr, NULL);
8311 }
8312 
8313 /*
8314  * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure.
8315  */
8316 static nfsstat4
8317 lock_denied(LOCK4denied *dp, struct flock64 *flk)
8318 {
8319 	rfs4_lockowner_t *lo;
8320 	rfs4_client_t *cp;
8321 	uint32_t len;
8322 
8323 	lo = rfs4_findlockowner_by_pid(flk->l_pid);
8324 	if (lo != NULL) {
8325 		cp = lo->client;
8326 		if (rfs4_lease_expired(cp)) {
8327 			rfs4_lockowner_rele(lo);
8328 			rfs4_dbe_hold(cp->dbe);
8329 			rfs4_client_close(cp);
8330 			return (NFS4ERR_EXPIRED);
8331 		}
8332 		dp->owner.clientid = lo->owner.clientid;
8333 		len = lo->owner.owner_len;
8334 		dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
8335 		bcopy(lo->owner.owner_val, dp->owner.owner_val, len);
8336 		dp->owner.owner_len = len;
8337 		rfs4_lockowner_rele(lo);
8338 		goto finish;
8339 	}
8340 
8341 	/*
8342 	 * Its not a NFS4 lock. We take advantage that the upper 32 bits
8343 	 * of the client id contain the boot time for a NFS4 lock. So we
8344 	 * fabricate and identity by setting clientid to the sysid, and
8345 	 * the lock owner to the pid.
8346 	 */
8347 	dp->owner.clientid = flk->l_sysid;
8348 	len = sizeof (pid_t);
8349 	dp->owner.owner_len = len;
8350 	dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
8351 	bcopy(&flk->l_pid, dp->owner.owner_val, len);
8352 finish:
8353 	dp->offset = flk->l_start;
8354 	dp->length = flk->l_len;
8355 
8356 	if (flk->l_type == F_RDLCK)
8357 		dp->locktype = READ_LT;
8358 	else if (flk->l_type == F_WRLCK)
8359 		dp->locktype = WRITE_LT;
8360 	else
8361 		return (NFS4ERR_INVAL);	/* no mapping from POSIX ltype to v4 */
8362 
8363 	return (NFS4_OK);
8364 }
8365 
8366 static int
8367 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred)
8368 {
8369 	int error;
8370 	struct flock64 flk;
8371 	int i;
8372 	clock_t delaytime;
8373 	int cmd;
8374 
8375 	cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK;
8376 retry:
8377 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
8378 
8379 	for (i = 0; i < rfs4_maxlock_tries; i++) {
8380 		LOCK_PRINT(rfs4_debug, "setlock", cmd, flock);
8381 		error = VOP_FRLOCK(vp, cmd,
8382 		    flock, flag, (u_offset_t)0, NULL, cred, NULL);
8383 
8384 		if (error != EAGAIN && error != EACCES)
8385 			break;
8386 
8387 		if (i < rfs4_maxlock_tries - 1) {
8388 			delay(delaytime);
8389 			delaytime *= 2;
8390 		}
8391 	}
8392 
8393 	if (error == EAGAIN || error == EACCES) {
8394 		/* Get the owner of the lock */
8395 		flk = *flock;
8396 		LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk);
8397 		if (VOP_FRLOCK(vp, F_GETLK, &flk, flag,
8398 		    (u_offset_t)0, NULL, cred, NULL) == 0) {
8399 			if (flk.l_type == F_UNLCK) {
8400 				/* No longer locked, retry */
8401 				goto retry;
8402 			}
8403 			*flock = flk;
8404 			LOCK_PRINT(rfs4_debug, "setlock(blocking lock)",
8405 			    F_GETLK, &flk);
8406 		}
8407 	}
8408 
8409 	return (error);
8410 }
8411 
8412 /*ARGSUSED*/
8413 static nfsstat4
8414 rfs4_do_lock(rfs4_lo_state_t *lp, nfs_lock_type4 locktype,
8415 	    seqid4 seqid, offset4 offset,
8416 	    length4 length, cred_t *cred, nfs_resop4 *resop)
8417 {
8418 	nfsstat4 status;
8419 	rfs4_lockowner_t *lo = lp->locker;
8420 	rfs4_state_t *sp = lp->state;
8421 	struct flock64 flock;
8422 	int16_t ltype;
8423 	int flag;
8424 	int error;
8425 	sysid_t sysid;
8426 	LOCK4res *lres;
8427 
8428 	if (rfs4_lease_expired(lo->client)) {
8429 		return (NFS4ERR_EXPIRED);
8430 	}
8431 
8432 	if ((status = rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK)
8433 		return (status);
8434 
8435 	/* Check for zero length. To lock to end of file use all ones for V4 */
8436 	if (length == 0)
8437 		return (NFS4ERR_INVAL);
8438 	else if (length == (length4)(~0))
8439 		length = 0;		/* Posix to end of file  */
8440 
8441 retry:
8442 	rfs4_dbe_lock(sp->dbe);
8443 
8444 
8445 	if (resop->resop != OP_LOCKU) {
8446 		switch (locktype) {
8447 		case READ_LT:
8448 		case READW_LT:
8449 			if ((sp->share_access
8450 			    & OPEN4_SHARE_ACCESS_READ) == 0) {
8451 				rfs4_dbe_unlock(sp->dbe);
8452 
8453 				return (NFS4ERR_OPENMODE);
8454 			}
8455 			ltype = F_RDLCK;
8456 			break;
8457 		case WRITE_LT:
8458 		case WRITEW_LT:
8459 			if ((sp->share_access
8460 			    & OPEN4_SHARE_ACCESS_WRITE) == 0) {
8461 				rfs4_dbe_unlock(sp->dbe);
8462 
8463 				return (NFS4ERR_OPENMODE);
8464 			}
8465 			ltype = F_WRLCK;
8466 			break;
8467 		}
8468 	} else
8469 		ltype = F_UNLCK;
8470 
8471 	flock.l_type = ltype;
8472 	flock.l_whence = 0;		/* SEEK_SET */
8473 	flock.l_start = offset;
8474 	flock.l_len = length;
8475 	flock.l_sysid = sysid;
8476 	flock.l_pid = lp->locker->pid;
8477 
8478 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
8479 	if (flock.l_len < 0 || flock.l_start < 0) {
8480 		rfs4_dbe_unlock(sp->dbe);
8481 		return (NFS4ERR_INVAL);
8482 	}
8483 
8484 	/*
8485 	 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and
8486 	 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE.
8487 	 */
8488 	flag = (int)sp->share_access | F_REMOTELOCK;
8489 
8490 	error = setlock(sp->finfo->vp, &flock, flag, cred);
8491 	if (error == 0) {
8492 		rfs4_dbe_lock(lp->dbe);
8493 		next_stateid(&lp->lockid);
8494 		rfs4_dbe_unlock(lp->dbe);
8495 	}
8496 
8497 	rfs4_dbe_unlock(sp->dbe);
8498 
8499 	/*
8500 	 * N.B. We map error values to nfsv4 errors. This is differrent
8501 	 * than puterrno4 routine.
8502 	 */
8503 	switch (error) {
8504 	case 0:
8505 		status = NFS4_OK;
8506 		break;
8507 	case EAGAIN:
8508 	case EACCES:		/* Old value */
8509 		/* Can only get here if op is OP_LOCK */
8510 		ASSERT(resop->resop == OP_LOCK);
8511 		lres = &resop->nfs_resop4_u.oplock;
8512 		status = NFS4ERR_DENIED;
8513 		if (lock_denied(&lres->LOCK4res_u.denied, &flock)
8514 		    == NFS4ERR_EXPIRED)
8515 			goto retry;
8516 		break;
8517 	case ENOLCK:
8518 		status = NFS4ERR_DELAY;
8519 		break;
8520 	case EOVERFLOW:
8521 		status = NFS4ERR_INVAL;
8522 		break;
8523 	case EINVAL:
8524 		status = NFS4ERR_NOTSUPP;
8525 		break;
8526 	default:
8527 		status = NFS4ERR_SERVERFAULT;
8528 		break;
8529 	}
8530 
8531 	return (status);
8532 }
8533 
8534 /*ARGSUSED*/
8535 void
8536 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop,
8537 	    struct svc_req *req, struct compound_state *cs)
8538 {
8539 	LOCK4args *args = &argop->nfs_argop4_u.oplock;
8540 	LOCK4res *resp = &resop->nfs_resop4_u.oplock;
8541 	nfsstat4 status;
8542 	stateid4 *stateid;
8543 	rfs4_lockowner_t *lo;
8544 	rfs4_client_t *cp;
8545 	rfs4_state_t *sp = NULL;
8546 	rfs4_lo_state_t *lsp = NULL;
8547 	bool_t ls_sw_held = FALSE;
8548 	bool_t create = TRUE;
8549 	bool_t lcreate = TRUE;
8550 	bool_t dup_lock = FALSE;
8551 	int rc;
8552 
8553 	DTRACE_NFSV4_2(op__lock__start, struct compound_state *, cs,
8554 	    LOCK4args *, args);
8555 
8556 	if (cs->vp == NULL) {
8557 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8558 		DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8559 		    cs, LOCK4res *, resp);
8560 		return;
8561 	}
8562 
8563 	if (args->locker.new_lock_owner) {
8564 		/* Create a new lockowner for this instance */
8565 		open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner;
8566 
8567 		NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner"));
8568 
8569 		stateid = &olo->open_stateid;
8570 		status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID);
8571 		if (status != NFS4_OK) {
8572 			NFS4_DEBUG(rfs4_debug,
8573 			    (CE_NOTE, "Get state failed in lock %d", status));
8574 			*cs->statusp = resp->status = status;
8575 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8576 			    cs, LOCK4res *, resp);
8577 			return;
8578 		}
8579 
8580 		/* Ensure specified filehandle matches */
8581 		if (cs->vp != sp->finfo->vp) {
8582 			rfs4_state_rele(sp);
8583 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8584 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8585 			    cs, LOCK4res *, resp);
8586 			return;
8587 		}
8588 
8589 		/* hold off other access to open_owner while we tinker */
8590 		rfs4_sw_enter(&sp->owner->oo_sw);
8591 
8592 		switch (rc = rfs4_check_stateid_seqid(sp, stateid)) {
8593 		case NFS4_CHECK_STATEID_OLD:
8594 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8595 			goto end;
8596 		case NFS4_CHECK_STATEID_BAD:
8597 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8598 			goto end;
8599 		case NFS4_CHECK_STATEID_EXPIRED:
8600 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8601 			goto end;
8602 		case NFS4_CHECK_STATEID_UNCONFIRMED:
8603 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8604 			goto end;
8605 		case NFS4_CHECK_STATEID_CLOSED:
8606 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8607 			goto end;
8608 		case NFS4_CHECK_STATEID_OKAY:
8609 		case NFS4_CHECK_STATEID_REPLAY:
8610 			switch (rfs4_check_olo_seqid(olo->open_seqid,
8611 			    sp->owner, resop)) {
8612 			case NFS4_CHKSEQ_OKAY:
8613 				if (rc == NFS4_CHECK_STATEID_OKAY)
8614 					break;
8615 				/*
8616 				 * This is replayed stateid; if seqid
8617 				 * matches next expected, then client
8618 				 * is using wrong seqid.
8619 				 */
8620 				/* FALLTHROUGH */
8621 			case NFS4_CHKSEQ_BAD:
8622 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8623 				goto end;
8624 			case NFS4_CHKSEQ_REPLAY:
8625 				/* This is a duplicate LOCK request */
8626 				dup_lock = TRUE;
8627 
8628 				/*
8629 				 * For a duplicate we do not want to
8630 				 * create a new lockowner as it should
8631 				 * already exist.
8632 				 * Turn off the lockowner create flag.
8633 				 */
8634 				lcreate = FALSE;
8635 			}
8636 			break;
8637 		}
8638 
8639 		lo = rfs4_findlockowner(&olo->lock_owner, &lcreate);
8640 		if (lo == NULL) {
8641 			NFS4_DEBUG(rfs4_debug,
8642 			    (CE_NOTE, "rfs4_op_lock: no lock owner"));
8643 			*cs->statusp = resp->status = NFS4ERR_RESOURCE;
8644 			goto end;
8645 		}
8646 
8647 		lsp = rfs4_findlo_state_by_owner(lo, sp, &create);
8648 		if (lsp == NULL) {
8649 			rfs4_update_lease(sp->owner->client);
8650 			/*
8651 			 * Only update theh open_seqid if this is not
8652 			 * a duplicate request
8653 			 */
8654 			if (dup_lock == FALSE) {
8655 				rfs4_update_open_sequence(sp->owner);
8656 			}
8657 
8658 			NFS4_DEBUG(rfs4_debug,
8659 			    (CE_NOTE, "rfs4_op_lock: no state"));
8660 			*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
8661 			rfs4_update_open_resp(sp->owner, resop, NULL);
8662 			rfs4_lockowner_rele(lo);
8663 			goto end;
8664 		}
8665 
8666 		/*
8667 		 * This is the new_lock_owner branch and the client is
8668 		 * supposed to be associating a new lock_owner with
8669 		 * the open file at this point.  If we find that a
8670 		 * lock_owner/state association already exists and a
8671 		 * successful LOCK request was returned to the client,
8672 		 * an error is returned to the client since this is
8673 		 * not appropriate.  The client should be using the
8674 		 * existing lock_owner branch.
8675 		 */
8676 		if (dup_lock == FALSE && create == FALSE) {
8677 			if (lsp->lock_completed == TRUE) {
8678 				*cs->statusp =
8679 				    resp->status = NFS4ERR_BAD_SEQID;
8680 				rfs4_lockowner_rele(lo);
8681 				goto end;
8682 			}
8683 		}
8684 
8685 		rfs4_update_lease(sp->owner->client);
8686 
8687 		/*
8688 		 * Only update theh open_seqid if this is not
8689 		 * a duplicate request
8690 		 */
8691 		if (dup_lock == FALSE) {
8692 			rfs4_update_open_sequence(sp->owner);
8693 		}
8694 
8695 		/*
8696 		 * If this is a duplicate lock request, just copy the
8697 		 * previously saved reply and return.
8698 		 */
8699 		if (dup_lock == TRUE) {
8700 			/* verify that lock_seqid's match */
8701 			if (lsp->seqid != olo->lock_seqid) {
8702 				NFS4_DEBUG(rfs4_debug,
8703 				    (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad"
8704 				    "lsp->seqid=%d old->seqid=%d",
8705 				    lsp->seqid, olo->lock_seqid));
8706 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8707 			} else {
8708 				rfs4_copy_reply(resop, lsp->reply);
8709 				/*
8710 				 * Make sure to copy the just
8711 				 * retrieved reply status into the
8712 				 * overall compound status
8713 				 */
8714 				*cs->statusp = resp->status;
8715 			}
8716 			rfs4_lockowner_rele(lo);
8717 			goto end;
8718 		}
8719 
8720 		rfs4_dbe_lock(lsp->dbe);
8721 
8722 		/* Make sure to update the lock sequence id */
8723 		lsp->seqid = olo->lock_seqid;
8724 
8725 		NFS4_DEBUG(rfs4_debug,
8726 		    (CE_NOTE, "Lock seqid established as %d", lsp->seqid));
8727 
8728 		/*
8729 		 * This is used to signify the newly created lockowner
8730 		 * stateid and its sequence number.  The checks for
8731 		 * sequence number and increment don't occur on the
8732 		 * very first lock request for a lockowner.
8733 		 */
8734 		lsp->skip_seqid_check = TRUE;
8735 
8736 		/* hold off other access to lsp while we tinker */
8737 		rfs4_sw_enter(&lsp->ls_sw);
8738 		ls_sw_held = TRUE;
8739 
8740 		rfs4_dbe_unlock(lsp->dbe);
8741 
8742 		rfs4_lockowner_rele(lo);
8743 	} else {
8744 		stateid = &args->locker.locker4_u.lock_owner.lock_stateid;
8745 		/* get lsp and hold the lock on the underlying file struct */
8746 		if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE))
8747 		    != NFS4_OK) {
8748 			*cs->statusp = resp->status = status;
8749 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8750 			    cs, LOCK4res *, resp);
8751 			return;
8752 		}
8753 		create = FALSE;	/* We didn't create lsp */
8754 
8755 		/* Ensure specified filehandle matches */
8756 		if (cs->vp != lsp->state->finfo->vp) {
8757 			rfs4_lo_state_rele(lsp, TRUE);
8758 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8759 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8760 			    cs, LOCK4res *, resp);
8761 			return;
8762 		}
8763 
8764 		/* hold off other access to lsp while we tinker */
8765 		rfs4_sw_enter(&lsp->ls_sw);
8766 		ls_sw_held = TRUE;
8767 
8768 		switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
8769 		/*
8770 		 * The stateid looks like it was okay (expected to be
8771 		 * the next one)
8772 		 */
8773 		case NFS4_CHECK_STATEID_OKAY:
8774 			/*
8775 			 * The sequence id is now checked.  Determine
8776 			 * if this is a replay or if it is in the
8777 			 * expected (next) sequence.  In the case of a
8778 			 * replay, there are two replay conditions
8779 			 * that may occur.  The first is the normal
8780 			 * condition where a LOCK is done with a
8781 			 * NFS4_OK response and the stateid is
8782 			 * updated.  That case is handled below when
8783 			 * the stateid is identified as a REPLAY.  The
8784 			 * second is the case where an error is
8785 			 * returned, like NFS4ERR_DENIED, and the
8786 			 * sequence number is updated but the stateid
8787 			 * is not updated.  This second case is dealt
8788 			 * with here.  So it may seem odd that the
8789 			 * stateid is okay but the sequence id is a
8790 			 * replay but it is okay.
8791 			 */
8792 			switch (rfs4_check_lock_seqid(
8793 			    args->locker.locker4_u.lock_owner.lock_seqid,
8794 			    lsp, resop)) {
8795 			case NFS4_CHKSEQ_REPLAY:
8796 				if (resp->status != NFS4_OK) {
8797 					/*
8798 					 * Here is our replay and need
8799 					 * to verify that the last
8800 					 * response was an error.
8801 					 */
8802 					*cs->statusp = resp->status;
8803 					goto end;
8804 				}
8805 				/*
8806 				 * This is done since the sequence id
8807 				 * looked like a replay but it didn't
8808 				 * pass our check so a BAD_SEQID is
8809 				 * returned as a result.
8810 				 */
8811 				/*FALLTHROUGH*/
8812 			case NFS4_CHKSEQ_BAD:
8813 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8814 				goto end;
8815 			case NFS4_CHKSEQ_OKAY:
8816 				/* Everything looks okay move ahead */
8817 				break;
8818 			}
8819 			break;
8820 		case NFS4_CHECK_STATEID_OLD:
8821 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8822 			goto end;
8823 		case NFS4_CHECK_STATEID_BAD:
8824 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8825 			goto end;
8826 		case NFS4_CHECK_STATEID_EXPIRED:
8827 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8828 			goto end;
8829 		case NFS4_CHECK_STATEID_CLOSED:
8830 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8831 			goto end;
8832 		case NFS4_CHECK_STATEID_REPLAY:
8833 			switch (rfs4_check_lock_seqid(
8834 			    args->locker.locker4_u.lock_owner.lock_seqid,
8835 			    lsp, resop)) {
8836 			case NFS4_CHKSEQ_OKAY:
8837 				/*
8838 				 * This is a replayed stateid; if
8839 				 * seqid matches the next expected,
8840 				 * then client is using wrong seqid.
8841 				 */
8842 			case NFS4_CHKSEQ_BAD:
8843 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8844 				goto end;
8845 			case NFS4_CHKSEQ_REPLAY:
8846 				rfs4_update_lease(lsp->locker->client);
8847 				*cs->statusp = status = resp->status;
8848 				goto end;
8849 			}
8850 			break;
8851 		default:
8852 			ASSERT(FALSE);
8853 			break;
8854 		}
8855 
8856 		rfs4_update_lock_sequence(lsp);
8857 		rfs4_update_lease(lsp->locker->client);
8858 	}
8859 
8860 	/*
8861 	 * NFS4 only allows locking on regular files, so
8862 	 * verify type of object.
8863 	 */
8864 	if (cs->vp->v_type != VREG) {
8865 		if (cs->vp->v_type == VDIR)
8866 			status = NFS4ERR_ISDIR;
8867 		else
8868 			status = NFS4ERR_INVAL;
8869 		goto out;
8870 	}
8871 
8872 	cp = lsp->state->owner->client;
8873 
8874 	if (rfs4_clnt_in_grace(cp) && !args->reclaim) {
8875 		status = NFS4ERR_GRACE;
8876 		goto out;
8877 	}
8878 
8879 	if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->can_reclaim) {
8880 		status = NFS4ERR_NO_GRACE;
8881 		goto out;
8882 	}
8883 
8884 	if (!rfs4_clnt_in_grace(cp) && args->reclaim) {
8885 		status = NFS4ERR_NO_GRACE;
8886 		goto out;
8887 	}
8888 
8889 	if (lsp->state->finfo->dinfo->dtype == OPEN_DELEGATE_WRITE)
8890 		cs->deleg = TRUE;
8891 
8892 	status = rfs4_do_lock(lsp, args->locktype,
8893 	    args->locker.locker4_u.lock_owner.lock_seqid,
8894 	    args->offset, args->length, cs->cr, resop);
8895 
8896 out:
8897 	lsp->skip_seqid_check = FALSE;
8898 
8899 	*cs->statusp = resp->status = status;
8900 
8901 	if (status == NFS4_OK) {
8902 		resp->LOCK4res_u.lock_stateid = lsp->lockid.stateid;
8903 		lsp->lock_completed = TRUE;
8904 	}
8905 	/*
8906 	 * Only update the "OPEN" response here if this was a new
8907 	 * lock_owner
8908 	 */
8909 	if (sp)
8910 		rfs4_update_open_resp(sp->owner, resop, NULL);
8911 
8912 	rfs4_update_lock_resp(lsp, resop);
8913 
8914 end:
8915 	if (lsp) {
8916 		if (ls_sw_held)
8917 			rfs4_sw_exit(&lsp->ls_sw);
8918 		/*
8919 		 * If an sp obtained, then the lsp does not represent
8920 		 * a lock on the file struct.
8921 		 */
8922 		if (sp != NULL)
8923 			rfs4_lo_state_rele(lsp, FALSE);
8924 		else
8925 			rfs4_lo_state_rele(lsp, TRUE);
8926 	}
8927 	if (sp) {
8928 		rfs4_sw_exit(&sp->owner->oo_sw);
8929 		rfs4_state_rele(sp);
8930 	}
8931 
8932 	DTRACE_NFSV4_2(op__lock__done, struct compound_state *, cs,
8933 	    LOCK4res *, resp);
8934 }
8935 
8936 /* free function for LOCK/LOCKT */
8937 static void
8938 lock_denied_free(nfs_resop4 *resop)
8939 {
8940 	LOCK4denied *dp = NULL;
8941 
8942 	switch (resop->resop) {
8943 	case OP_LOCK:
8944 		if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED)
8945 			dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied;
8946 		break;
8947 	case OP_LOCKT:
8948 		if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED)
8949 			dp = &resop->nfs_resop4_u.oplockt.denied;
8950 		break;
8951 	default:
8952 		break;
8953 	}
8954 
8955 	if (dp)
8956 		kmem_free(dp->owner.owner_val, dp->owner.owner_len);
8957 }
8958 
8959 /*ARGSUSED*/
8960 void
8961 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop,
8962 	    struct svc_req *req, struct compound_state *cs)
8963 {
8964 	LOCKU4args *args = &argop->nfs_argop4_u.oplocku;
8965 	LOCKU4res *resp = &resop->nfs_resop4_u.oplocku;
8966 	nfsstat4 status;
8967 	stateid4 *stateid = &args->lock_stateid;
8968 	rfs4_lo_state_t *lsp;
8969 
8970 	DTRACE_NFSV4_2(op__locku__start, struct compound_state *, cs,
8971 	    LOCKU4args *, args);
8972 
8973 	if (cs->vp == NULL) {
8974 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8975 		DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
8976 		    LOCKU4res *, resp);
8977 		return;
8978 	}
8979 
8980 	if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) {
8981 		*cs->statusp = resp->status = status;
8982 		DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
8983 		    LOCKU4res *, resp);
8984 		return;
8985 	}
8986 
8987 	/* Ensure specified filehandle matches */
8988 	if (cs->vp != lsp->state->finfo->vp) {
8989 		rfs4_lo_state_rele(lsp, TRUE);
8990 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8991 		DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
8992 		    LOCKU4res *, resp);
8993 		return;
8994 	}
8995 
8996 	/* hold off other access to lsp while we tinker */
8997 	rfs4_sw_enter(&lsp->ls_sw);
8998 
8999 	switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
9000 	case NFS4_CHECK_STATEID_OKAY:
9001 		if (rfs4_check_lock_seqid(args->seqid, lsp, resop)
9002 		    != NFS4_CHKSEQ_OKAY) {
9003 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9004 			goto end;
9005 		}
9006 		break;
9007 	case NFS4_CHECK_STATEID_OLD:
9008 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9009 		goto end;
9010 	case NFS4_CHECK_STATEID_BAD:
9011 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9012 		goto end;
9013 	case NFS4_CHECK_STATEID_EXPIRED:
9014 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
9015 		goto end;
9016 	case NFS4_CHECK_STATEID_CLOSED:
9017 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9018 		goto end;
9019 	case NFS4_CHECK_STATEID_REPLAY:
9020 		switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) {
9021 		case NFS4_CHKSEQ_OKAY:
9022 				/*
9023 				 * This is a replayed stateid; if
9024 				 * seqid matches the next expected,
9025 				 * then client is using wrong seqid.
9026 				 */
9027 		case NFS4_CHKSEQ_BAD:
9028 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9029 			goto end;
9030 		case NFS4_CHKSEQ_REPLAY:
9031 			rfs4_update_lease(lsp->locker->client);
9032 			*cs->statusp = status = resp->status;
9033 			goto end;
9034 		}
9035 		break;
9036 	default:
9037 		ASSERT(FALSE);
9038 		break;
9039 	}
9040 
9041 	rfs4_update_lock_sequence(lsp);
9042 	rfs4_update_lease(lsp->locker->client);
9043 
9044 	/*
9045 	 * NFS4 only allows locking on regular files, so
9046 	 * verify type of object.
9047 	 */
9048 	if (cs->vp->v_type != VREG) {
9049 		if (cs->vp->v_type == VDIR)
9050 			status = NFS4ERR_ISDIR;
9051 		else
9052 			status = NFS4ERR_INVAL;
9053 		goto out;
9054 	}
9055 
9056 	if (rfs4_clnt_in_grace(lsp->state->owner->client)) {
9057 		status = NFS4ERR_GRACE;
9058 		goto out;
9059 	}
9060 
9061 	status = rfs4_do_lock(lsp, args->locktype,
9062 	    args->seqid, args->offset, args->length, cs->cr, resop);
9063 
9064 out:
9065 	*cs->statusp = resp->status = status;
9066 
9067 	if (status == NFS4_OK)
9068 		resp->lock_stateid = lsp->lockid.stateid;
9069 
9070 	rfs4_update_lock_resp(lsp, resop);
9071 
9072 end:
9073 	rfs4_sw_exit(&lsp->ls_sw);
9074 	rfs4_lo_state_rele(lsp, TRUE);
9075 
9076 	DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
9077 	    LOCKU4res *, resp);
9078 }
9079 
9080 /*
9081  * LOCKT is a best effort routine, the client can not be guaranteed that
9082  * the status return is still in effect by the time the reply is received.
9083  * They are numerous race conditions in this routine, but we are not required
9084  * and can not be accurate.
9085  */
9086 /*ARGSUSED*/
9087 void
9088 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop,
9089 	    struct svc_req *req, struct compound_state *cs)
9090 {
9091 	LOCKT4args *args = &argop->nfs_argop4_u.oplockt;
9092 	LOCKT4res *resp = &resop->nfs_resop4_u.oplockt;
9093 	rfs4_lockowner_t *lo;
9094 	rfs4_client_t *cp;
9095 	bool_t create = FALSE;
9096 	struct flock64 flk;
9097 	int error;
9098 	int flag = FREAD | FWRITE;
9099 	int ltype;
9100 	length4 posix_length;
9101 	sysid_t sysid;
9102 	pid_t pid;
9103 
9104 	DTRACE_NFSV4_2(op__lockt__start, struct compound_state *, cs,
9105 	    LOCKT4args *, args);
9106 
9107 	if (cs->vp == NULL) {
9108 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
9109 		goto out;
9110 	}
9111 
9112 	/*
9113 	 * NFS4 only allows locking on regular files, so
9114 	 * verify type of object.
9115 	 */
9116 	if (cs->vp->v_type != VREG) {
9117 		if (cs->vp->v_type == VDIR)
9118 			*cs->statusp = resp->status = NFS4ERR_ISDIR;
9119 		else
9120 			*cs->statusp = resp->status =  NFS4ERR_INVAL;
9121 		goto out;
9122 	}
9123 
9124 	/*
9125 	 * Check out the clientid to ensure the server knows about it
9126 	 * so that we correctly inform the client of a server reboot.
9127 	 */
9128 	if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE))
9129 	    == NULL) {
9130 		*cs->statusp = resp->status =
9131 		    rfs4_check_clientid(&args->owner.clientid, 0);
9132 		goto out;
9133 	}
9134 	if (rfs4_lease_expired(cp)) {
9135 		rfs4_client_close(cp);
9136 		/*
9137 		 * Protocol doesn't allow returning NFS4ERR_STALE as
9138 		 * other operations do on this check so STALE_CLIENTID
9139 		 * is returned instead
9140 		 */
9141 		*cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
9142 		goto out;
9143 	}
9144 
9145 	if (rfs4_clnt_in_grace(cp) && !(cp->can_reclaim)) {
9146 		*cs->statusp = resp->status = NFS4ERR_GRACE;
9147 		rfs4_client_rele(cp);
9148 		goto out;
9149 	}
9150 	rfs4_client_rele(cp);
9151 
9152 	resp->status = NFS4_OK;
9153 
9154 	switch (args->locktype) {
9155 	case READ_LT:
9156 	case READW_LT:
9157 		ltype = F_RDLCK;
9158 		break;
9159 	case WRITE_LT:
9160 	case WRITEW_LT:
9161 		ltype = F_WRLCK;
9162 		break;
9163 	}
9164 
9165 	posix_length = args->length;
9166 	/* Check for zero length. To lock to end of file use all ones for V4 */
9167 	if (posix_length == 0) {
9168 		*cs->statusp = resp->status = NFS4ERR_INVAL;
9169 		goto out;
9170 	} else if (posix_length == (length4)(~0)) {
9171 		posix_length = 0;	/* Posix to end of file  */
9172 	}
9173 
9174 	/* Find or create a lockowner */
9175 	lo = rfs4_findlockowner(&args->owner, &create);
9176 
9177 	if (lo) {
9178 		pid = lo->pid;
9179 		if ((resp->status =
9180 		    rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK)
9181 			goto err;
9182 	} else {
9183 		pid = 0;
9184 		sysid = lockt_sysid;
9185 	}
9186 retry:
9187 	flk.l_type = ltype;
9188 	flk.l_whence = 0;		/* SEEK_SET */
9189 	flk.l_start = args->offset;
9190 	flk.l_len = posix_length;
9191 	flk.l_sysid = sysid;
9192 	flk.l_pid = pid;
9193 	flag |= F_REMOTELOCK;
9194 
9195 	LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk);
9196 
9197 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
9198 	if (flk.l_len < 0 || flk.l_start < 0) {
9199 		resp->status = NFS4ERR_INVAL;
9200 		goto err;
9201 	}
9202 	error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0,
9203 	    NULL, cs->cr, NULL);
9204 
9205 	/*
9206 	 * N.B. We map error values to nfsv4 errors. This is differrent
9207 	 * than puterrno4 routine.
9208 	 */
9209 	switch (error) {
9210 	case 0:
9211 		if (flk.l_type == F_UNLCK)
9212 			resp->status = NFS4_OK;
9213 		else {
9214 			if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED)
9215 				goto retry;
9216 			resp->status = NFS4ERR_DENIED;
9217 		}
9218 		break;
9219 	case EOVERFLOW:
9220 		resp->status = NFS4ERR_INVAL;
9221 		break;
9222 	case EINVAL:
9223 		resp->status = NFS4ERR_NOTSUPP;
9224 		break;
9225 	default:
9226 		cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)",
9227 		    error);
9228 		resp->status = NFS4ERR_SERVERFAULT;
9229 		break;
9230 	}
9231 
9232 err:
9233 	if (lo)
9234 		rfs4_lockowner_rele(lo);
9235 	*cs->statusp = resp->status;
9236 out:
9237 	DTRACE_NFSV4_2(op__lockt__done, struct compound_state *, cs,
9238 	    LOCKT4res *, resp);
9239 }
9240 
9241 static int
9242 vop_shrlock(vnode_t *vp, int cmd, struct shrlock *sp, int fflags)
9243 {
9244 	int err;
9245 
9246 	if (cmd == F_UNSHARE && sp->s_deny == 0 && sp->s_access == 0)
9247 		return (0);
9248 
9249 	err = VOP_SHRLOCK(vp, cmd, sp, fflags, CRED(), NULL);
9250 
9251 	NFS4_DEBUG(rfs4_shrlock_debug,
9252 	    (CE_NOTE, "rfs4_shrlock %s vp=%p acc=%d dny=%d sysid=%d "
9253 	    "pid=%d err=%d\n", cmd == F_UNSHARE ? "UNSHR" : "SHARE",
9254 	    (void *) vp, sp->s_access, sp->s_deny, sp->s_sysid, sp->s_pid,
9255 	    err));
9256 
9257 	return (err);
9258 }
9259 
9260 static int
9261 rfs4_shrlock(rfs4_state_t *sp, int cmd)
9262 {
9263 	struct shrlock shr;
9264 	struct shr_locowner shr_loco;
9265 	int fflags;
9266 
9267 	fflags = shr.s_access = shr.s_deny = 0;
9268 
9269 	if (sp->share_access & OPEN4_SHARE_ACCESS_READ) {
9270 		fflags |= FREAD;
9271 		shr.s_access |= F_RDACC;
9272 	}
9273 	if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) {
9274 		fflags |= FWRITE;
9275 		shr.s_access |= F_WRACC;
9276 	}
9277 	if (sp->share_deny & OPEN4_SHARE_DENY_READ)
9278 		shr.s_deny |= F_RDDNY;
9279 	if (sp->share_deny & OPEN4_SHARE_DENY_WRITE)
9280 		shr.s_deny |= F_WRDNY;
9281 
9282 	shr.s_pid = rfs4_dbe_getid(sp->owner->dbe);
9283 	shr.s_sysid = sp->owner->client->sysidt;
9284 	shr_loco.sl_pid = shr.s_pid;
9285 	shr_loco.sl_id = shr.s_sysid;
9286 	shr.s_owner = (caddr_t)&shr_loco;
9287 	shr.s_own_len = sizeof (shr_loco);
9288 	return (vop_shrlock(sp->finfo->vp, cmd, &shr, fflags));
9289 }
9290 
9291 static int
9292 rfs4_share(rfs4_state_t *sp)
9293 {
9294 	int cmd;
9295 	cmd = nbl_need_check(sp->finfo->vp) ? F_SHARE_NBMAND : F_SHARE;
9296 	return (rfs4_shrlock(sp, cmd));
9297 }
9298 
9299 void
9300 rfs4_unshare(rfs4_state_t *sp)
9301 {
9302 	(void) rfs4_shrlock(sp, F_UNSHARE);
9303 }
9304 
9305 static int
9306 rdma_setup_read_data4(READ4args *args, READ4res *rok)
9307 {
9308 	struct clist	*wcl;
9309 	count4		count = rok->data_len;
9310 	int		wlist_len;
9311 
9312 	wcl = args->wlist;
9313 	if (rdma_setup_read_chunks(wcl, count, &wlist_len) == FALSE) {
9314 		return (FALSE);
9315 	}
9316 	wcl = args->wlist;
9317 	rok->wlist_len = wlist_len;
9318 	rok->wlist = wcl;
9319 	return (TRUE);
9320 }
9321