xref: /titanic_51/usr/src/uts/common/fs/nfs/nfs4_srv.c (revision dec7d25083bb614c1dd67e18534dc6dcfa13ec5d)
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 /*
23  * Copyright 2016 Nexenta Systems, Inc.  All rights reserved.
24  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
25  * Copyright (c) 2012 by Delphix. All rights reserved.
26  */
27 
28 /*
29  *	Copyright (c) 1983,1984,1985,1986,1987,1988,1989  AT&T.
30  *	All Rights Reserved
31  */
32 
33 #include <sys/param.h>
34 #include <sys/types.h>
35 #include <sys/systm.h>
36 #include <sys/cred.h>
37 #include <sys/buf.h>
38 #include <sys/vfs.h>
39 #include <sys/vfs_opreg.h>
40 #include <sys/vnode.h>
41 #include <sys/uio.h>
42 #include <sys/errno.h>
43 #include <sys/sysmacros.h>
44 #include <sys/statvfs.h>
45 #include <sys/kmem.h>
46 #include <sys/dirent.h>
47 #include <sys/cmn_err.h>
48 #include <sys/debug.h>
49 #include <sys/systeminfo.h>
50 #include <sys/flock.h>
51 #include <sys/pathname.h>
52 #include <sys/nbmlock.h>
53 #include <sys/share.h>
54 #include <sys/atomic.h>
55 #include <sys/policy.h>
56 #include <sys/fem.h>
57 #include <sys/sdt.h>
58 #include <sys/ddi.h>
59 #include <sys/zone.h>
60 #include <sys/kstat.h>
61 
62 #include <fs/fs_reparse.h>
63 
64 #include <rpc/types.h>
65 #include <rpc/auth.h>
66 #include <rpc/rpcsec_gss.h>
67 #include <rpc/svc.h>
68 
69 #include <nfs/nfs.h>
70 #include <nfs/export.h>
71 #include <nfs/nfs_cmd.h>
72 #include <nfs/lm.h>
73 #include <nfs/nfs4.h>
74 
75 #include <sys/strsubr.h>
76 #include <sys/strsun.h>
77 
78 #include <inet/common.h>
79 #include <inet/ip.h>
80 #include <inet/ip6.h>
81 
82 #include <sys/tsol/label.h>
83 #include <sys/tsol/tndb.h>
84 
85 #define	RFS4_MAXLOCK_TRIES 4	/* Try to get the lock this many times */
86 static int rfs4_maxlock_tries = RFS4_MAXLOCK_TRIES;
87 #define	RFS4_LOCK_DELAY 10	/* Milliseconds */
88 static clock_t  rfs4_lock_delay = RFS4_LOCK_DELAY;
89 extern struct svc_ops rdma_svc_ops;
90 extern int nfs_loaned_buffers;
91 /* End of Tunables */
92 
93 static int rdma_setup_read_data4(READ4args *, READ4res *);
94 
95 /*
96  * Used to bump the stateid4.seqid value and show changes in the stateid
97  */
98 #define	next_stateid(sp) (++(sp)->bits.chgseq)
99 
100 /*
101  * RFS4_MINLEN_ENTRY4: XDR-encoded size of smallest possible dirent.
102  *	This is used to return NFS4ERR_TOOSMALL when clients specify
103  *	maxcount that isn't large enough to hold the smallest possible
104  *	XDR encoded dirent.
105  *
106  *	    sizeof cookie (8 bytes) +
107  *	    sizeof name_len (4 bytes) +
108  *	    sizeof smallest (padded) name (4 bytes) +
109  *	    sizeof bitmap4_len (12 bytes) +   NOTE: we always encode len=2 bm4
110  *	    sizeof attrlist4_len (4 bytes) +
111  *	    sizeof next boolean (4 bytes)
112  *
113  * RFS4_MINLEN_RDDIR4: XDR-encoded size of READDIR op reply containing
114  * the smallest possible entry4 (assumes no attrs requested).
115  *	sizeof nfsstat4 (4 bytes) +
116  *	sizeof verifier4 (8 bytes) +
117  *	sizeof entry4list bool (4 bytes) +
118  *	sizeof entry4 	(36 bytes) +
119  *	sizeof eof bool  (4 bytes)
120  *
121  * RFS4_MINLEN_RDDIR_BUF: minimum length of buffer server will provide to
122  *	VOP_READDIR.  Its value is the size of the maximum possible dirent
123  *	for solaris.  The DIRENT64_RECLEN macro returns	the size of dirent
124  *	required for a given name length.  MAXNAMELEN is the maximum
125  *	filename length allowed in Solaris.  The first two DIRENT64_RECLEN()
126  *	macros are to allow for . and .. entries -- just a minor tweak to try
127  *	and guarantee that buffer we give to VOP_READDIR will be large enough
128  *	to hold ., .., and the largest possible solaris dirent64.
129  */
130 #define	RFS4_MINLEN_ENTRY4 36
131 #define	RFS4_MINLEN_RDDIR4 (4 + NFS4_VERIFIER_SIZE + 4 + RFS4_MINLEN_ENTRY4 + 4)
132 #define	RFS4_MINLEN_RDDIR_BUF \
133 	(DIRENT64_RECLEN(1) + DIRENT64_RECLEN(2) + DIRENT64_RECLEN(MAXNAMELEN))
134 
135 /*
136  * It would be better to pad to 4 bytes since that's what XDR would do,
137  * but the dirents UFS gives us are already padded to 8, so just take
138  * what we're given.  Dircount is only a hint anyway.  Currently the
139  * solaris kernel is ASCII only, so there's no point in calling the
140  * UTF8 functions.
141  *
142  * dirent64: named padded to provide 8 byte struct alignment
143  *	d_ino(8) + d_off(8) + d_reclen(2) + d_name(namelen + null(1) + pad)
144  *
145  * cookie: uint64_t   +  utf8namelen: uint_t  +   utf8name padded to 8 bytes
146  *
147  */
148 #define	DIRENT64_TO_DIRCOUNT(dp) \
149 	(3 * BYTES_PER_XDR_UNIT + DIRENT64_NAMELEN((dp)->d_reclen))
150 
151 time_t rfs4_start_time;			/* Initialized in rfs4_srvrinit */
152 
153 static sysid_t lockt_sysid;		/* dummy sysid for all LOCKT calls */
154 
155 u_longlong_t	nfs4_srv_caller_id;
156 uint_t		nfs4_srv_vkey = 0;
157 
158 verifier4	Write4verf;
159 verifier4	Readdir4verf;
160 
161 void	rfs4_init_compound_state(struct compound_state *);
162 
163 static void	nullfree(caddr_t);
164 static void	rfs4_op_inval(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
165 			struct compound_state *);
166 static void	rfs4_op_access(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
167 			struct compound_state *);
168 static void	rfs4_op_close(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
169 			struct compound_state *);
170 static void	rfs4_op_commit(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
171 			struct compound_state *);
172 static void	rfs4_op_create(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
173 			struct compound_state *);
174 static void	rfs4_op_create_free(nfs_resop4 *resop);
175 static void	rfs4_op_delegreturn(nfs_argop4 *, nfs_resop4 *,
176 			struct svc_req *, struct compound_state *);
177 static void	rfs4_op_delegpurge(nfs_argop4 *, nfs_resop4 *,
178 			struct svc_req *, struct compound_state *);
179 static void	rfs4_op_getattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
180 			struct compound_state *);
181 static void	rfs4_op_getattr_free(nfs_resop4 *);
182 static void	rfs4_op_getfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
183 			struct compound_state *);
184 static void	rfs4_op_getfh_free(nfs_resop4 *);
185 static void	rfs4_op_illegal(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
186 			struct compound_state *);
187 static void	rfs4_op_link(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
188 			struct compound_state *);
189 static void	rfs4_op_lock(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
190 			struct compound_state *);
191 static void	lock_denied_free(nfs_resop4 *);
192 static void	rfs4_op_locku(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
193 			struct compound_state *);
194 static void	rfs4_op_lockt(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
195 			struct compound_state *);
196 static void	rfs4_op_lookup(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
197 			struct compound_state *);
198 static void	rfs4_op_lookupp(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
199 			struct compound_state *);
200 static void	rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop,
201 				struct svc_req *req, struct compound_state *cs);
202 static void	rfs4_op_nverify(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
203 			struct compound_state *);
204 static void	rfs4_op_open(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
205 			struct compound_state *);
206 static void	rfs4_op_open_confirm(nfs_argop4 *, nfs_resop4 *,
207 			struct svc_req *, struct compound_state *);
208 static void	rfs4_op_open_downgrade(nfs_argop4 *, nfs_resop4 *,
209 			struct svc_req *, struct compound_state *);
210 static void	rfs4_op_putfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
211 			struct compound_state *);
212 static void	rfs4_op_putpubfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
213 			struct compound_state *);
214 static void	rfs4_op_putrootfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
215 			struct compound_state *);
216 static void	rfs4_op_read(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
217 			struct compound_state *);
218 static void	rfs4_op_read_free(nfs_resop4 *);
219 static void	rfs4_op_readdir_free(nfs_resop4 *resop);
220 static void	rfs4_op_readlink(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
221 			struct compound_state *);
222 static void	rfs4_op_readlink_free(nfs_resop4 *);
223 static void	rfs4_op_release_lockowner(nfs_argop4 *, nfs_resop4 *,
224 			struct svc_req *, struct compound_state *);
225 static void	rfs4_op_remove(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
226 			struct compound_state *);
227 static void	rfs4_op_rename(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
228 			struct compound_state *);
229 static void	rfs4_op_renew(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
230 			struct compound_state *);
231 static void	rfs4_op_restorefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
232 			struct compound_state *);
233 static void	rfs4_op_savefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
234 			struct compound_state *);
235 static void	rfs4_op_setattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
236 			struct compound_state *);
237 static void	rfs4_op_verify(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
238 			struct compound_state *);
239 static void	rfs4_op_write(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
240 			struct compound_state *);
241 static void	rfs4_op_setclientid(nfs_argop4 *, nfs_resop4 *,
242 			struct svc_req *, struct compound_state *);
243 static void	rfs4_op_setclientid_confirm(nfs_argop4 *, nfs_resop4 *,
244 			struct svc_req *req, struct compound_state *);
245 static void	rfs4_op_secinfo(nfs_argop4 *, nfs_resop4 *, struct svc_req *,
246 			struct compound_state *);
247 static void	rfs4_op_secinfo_free(nfs_resop4 *);
248 
249 static nfsstat4 check_open_access(uint32_t,
250 				struct compound_state *, struct svc_req *);
251 nfsstat4 rfs4_client_sysid(rfs4_client_t *, sysid_t *);
252 void rfs4_ss_clid(rfs4_client_t *);
253 
254 /*
255  * translation table for attrs
256  */
257 struct nfs4_ntov_table {
258 	union nfs4_attr_u *na;
259 	uint8_t amap[NFS4_MAXNUM_ATTRS];
260 	int attrcnt;
261 	bool_t vfsstat;
262 };
263 
264 static void	nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp);
265 static void	nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp,
266 				    struct nfs4_svgetit_arg *sargp);
267 
268 static nfsstat4	do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp,
269 		    struct compound_state *cs, struct nfs4_svgetit_arg *sargp,
270 		    struct nfs4_ntov_table *ntovp, nfs4_attr_cmd_t cmd);
271 
272 fem_t		*deleg_rdops;
273 fem_t		*deleg_wrops;
274 
275 rfs4_servinst_t *rfs4_cur_servinst = NULL;	/* current server instance */
276 kmutex_t	rfs4_servinst_lock;	/* protects linked list */
277 int		rfs4_seen_first_compound;	/* set first time we see one */
278 
279 /*
280  * NFS4 op dispatch table
281  */
282 
283 struct rfsv4disp {
284 	void	(*dis_proc)();		/* proc to call */
285 	void	(*dis_resfree)();	/* frees space allocated by proc */
286 	int	dis_flags;		/* RPC_IDEMPOTENT, etc... */
287 	int	op_type;		/* operation type, see below */
288 };
289 
290 /*
291  * operation types; used primarily for the per-exportinfo kstat implementation
292  */
293 #define	NFS4_OP_NOFH	0	/* The operation does not operate with any */
294 				/* particular filehandle; we cannot associate */
295 				/* it with any exportinfo. */
296 
297 #define	NFS4_OP_CFH	1	/* The operation works with the current */
298 				/* filehandle; we associate the operation */
299 				/* with the exportinfo related to the current */
300 				/* filehandle (as set before the operation is */
301 				/* executed). */
302 
303 #define	NFS4_OP_SFH	2	/* The operation works with the saved */
304 				/* filehandle; we associate the operation */
305 				/* with the exportinfo related to the saved */
306 				/* filehandle (as set before the operation is */
307 				/* executed). */
308 
309 #define	NFS4_OP_POSTCFH	3	/* The operation ignores the current */
310 				/* filehandle, but sets the new current */
311 				/* filehandle instead; we associate the */
312 				/* operation with the exportinfo related to */
313 				/* the current filehandle as set after the */
314 				/* operation is successfuly executed.  Since */
315 				/* we do not know the particular exportinfo */
316 				/* (and thus the kstat) before the operation */
317 				/* is done, there is no simple way how to */
318 				/* update some I/O kstat statistics related */
319 				/* to kstat_queue(9F). */
320 
321 static struct rfsv4disp rfsv4disptab[] = {
322 	/*
323 	 * NFS VERSION 4
324 	 */
325 
326 	/* RFS_NULL = 0 */
327 	{rfs4_op_illegal, nullfree, 0, NFS4_OP_NOFH},
328 
329 	/* UNUSED = 1 */
330 	{rfs4_op_illegal, nullfree, 0, NFS4_OP_NOFH},
331 
332 	/* UNUSED = 2 */
333 	{rfs4_op_illegal, nullfree, 0, NFS4_OP_NOFH},
334 
335 	/* OP_ACCESS = 3 */
336 	{rfs4_op_access, nullfree, RPC_IDEMPOTENT, NFS4_OP_CFH},
337 
338 	/* OP_CLOSE = 4 */
339 	{rfs4_op_close, nullfree, 0, NFS4_OP_CFH},
340 
341 	/* OP_COMMIT = 5 */
342 	{rfs4_op_commit, nullfree, RPC_IDEMPOTENT, NFS4_OP_CFH},
343 
344 	/* OP_CREATE = 6 */
345 	{rfs4_op_create, nullfree, 0, NFS4_OP_CFH},
346 
347 	/* OP_DELEGPURGE = 7 */
348 	{rfs4_op_delegpurge, nullfree, 0, NFS4_OP_NOFH},
349 
350 	/* OP_DELEGRETURN = 8 */
351 	{rfs4_op_delegreturn, nullfree, 0, NFS4_OP_CFH},
352 
353 	/* OP_GETATTR = 9 */
354 	{rfs4_op_getattr, rfs4_op_getattr_free, RPC_IDEMPOTENT, NFS4_OP_CFH},
355 
356 	/* OP_GETFH = 10 */
357 	{rfs4_op_getfh, rfs4_op_getfh_free, RPC_ALL, NFS4_OP_CFH},
358 
359 	/* OP_LINK = 11 */
360 	{rfs4_op_link, nullfree, 0, NFS4_OP_CFH},
361 
362 	/* OP_LOCK = 12 */
363 	{rfs4_op_lock, lock_denied_free, 0, NFS4_OP_CFH},
364 
365 	/* OP_LOCKT = 13 */
366 	{rfs4_op_lockt, lock_denied_free, 0, NFS4_OP_CFH},
367 
368 	/* OP_LOCKU = 14 */
369 	{rfs4_op_locku, nullfree, 0, NFS4_OP_CFH},
370 
371 	/* OP_LOOKUP = 15 */
372 	{rfs4_op_lookup, nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK),
373 	    NFS4_OP_CFH},
374 
375 	/* OP_LOOKUPP = 16 */
376 	{rfs4_op_lookupp, nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK),
377 	    NFS4_OP_CFH},
378 
379 	/* OP_NVERIFY = 17 */
380 	{rfs4_op_nverify, nullfree, RPC_IDEMPOTENT, NFS4_OP_CFH},
381 
382 	/* OP_OPEN = 18 */
383 	{rfs4_op_open, rfs4_free_reply, 0, NFS4_OP_CFH},
384 
385 	/* OP_OPENATTR = 19 */
386 	{rfs4_op_openattr, nullfree, 0, NFS4_OP_CFH},
387 
388 	/* OP_OPEN_CONFIRM = 20 */
389 	{rfs4_op_open_confirm, nullfree, 0, NFS4_OP_CFH},
390 
391 	/* OP_OPEN_DOWNGRADE = 21 */
392 	{rfs4_op_open_downgrade, nullfree, 0, NFS4_OP_CFH},
393 
394 	/* OP_OPEN_PUTFH = 22 */
395 	{rfs4_op_putfh, nullfree, RPC_ALL, NFS4_OP_POSTCFH},
396 
397 	/* OP_PUTPUBFH = 23 */
398 	{rfs4_op_putpubfh, nullfree, RPC_ALL, NFS4_OP_POSTCFH},
399 
400 	/* OP_PUTROOTFH = 24 */
401 	{rfs4_op_putrootfh, nullfree, RPC_ALL, NFS4_OP_POSTCFH},
402 
403 	/* OP_READ = 25 */
404 	{rfs4_op_read, rfs4_op_read_free, RPC_IDEMPOTENT, NFS4_OP_CFH},
405 
406 	/* OP_READDIR = 26 */
407 	{rfs4_op_readdir, rfs4_op_readdir_free, RPC_IDEMPOTENT, NFS4_OP_CFH},
408 
409 	/* OP_READLINK = 27 */
410 	{rfs4_op_readlink, rfs4_op_readlink_free, RPC_IDEMPOTENT, NFS4_OP_CFH},
411 
412 	/* OP_REMOVE = 28 */
413 	{rfs4_op_remove, nullfree, 0, NFS4_OP_CFH},
414 
415 	/* OP_RENAME = 29 */
416 	{rfs4_op_rename, nullfree, 0, NFS4_OP_CFH},
417 
418 	/* OP_RENEW = 30 */
419 	{rfs4_op_renew, nullfree, 0, NFS4_OP_NOFH},
420 
421 	/* OP_RESTOREFH = 31 */
422 	{rfs4_op_restorefh, nullfree, RPC_ALL, NFS4_OP_SFH},
423 
424 	/* OP_SAVEFH = 32 */
425 	{rfs4_op_savefh, nullfree, RPC_ALL, NFS4_OP_CFH},
426 
427 	/* OP_SECINFO = 33 */
428 	{rfs4_op_secinfo, rfs4_op_secinfo_free, 0, NFS4_OP_CFH},
429 
430 	/* OP_SETATTR = 34 */
431 	{rfs4_op_setattr, nullfree, 0, NFS4_OP_CFH},
432 
433 	/* OP_SETCLIENTID = 35 */
434 	{rfs4_op_setclientid, nullfree, 0, NFS4_OP_NOFH},
435 
436 	/* OP_SETCLIENTID_CONFIRM = 36 */
437 	{rfs4_op_setclientid_confirm, nullfree, 0, NFS4_OP_NOFH},
438 
439 	/* OP_VERIFY = 37 */
440 	{rfs4_op_verify, nullfree, RPC_IDEMPOTENT, NFS4_OP_CFH},
441 
442 	/* OP_WRITE = 38 */
443 	{rfs4_op_write, nullfree, 0, NFS4_OP_CFH},
444 
445 	/* OP_RELEASE_LOCKOWNER = 39 */
446 	{rfs4_op_release_lockowner, nullfree, 0, NFS4_OP_NOFH},
447 };
448 
449 static uint_t rfsv4disp_cnt = sizeof (rfsv4disptab) / sizeof (rfsv4disptab[0]);
450 
451 #define	OP_ILLEGAL_IDX (rfsv4disp_cnt)
452 
453 #ifdef DEBUG
454 
455 int		rfs4_fillone_debug = 0;
456 int		rfs4_no_stub_access = 1;
457 int		rfs4_rddir_debug = 0;
458 
459 static char    *rfs4_op_string[] = {
460 	"rfs4_op_null",
461 	"rfs4_op_1 unused",
462 	"rfs4_op_2 unused",
463 	"rfs4_op_access",
464 	"rfs4_op_close",
465 	"rfs4_op_commit",
466 	"rfs4_op_create",
467 	"rfs4_op_delegpurge",
468 	"rfs4_op_delegreturn",
469 	"rfs4_op_getattr",
470 	"rfs4_op_getfh",
471 	"rfs4_op_link",
472 	"rfs4_op_lock",
473 	"rfs4_op_lockt",
474 	"rfs4_op_locku",
475 	"rfs4_op_lookup",
476 	"rfs4_op_lookupp",
477 	"rfs4_op_nverify",
478 	"rfs4_op_open",
479 	"rfs4_op_openattr",
480 	"rfs4_op_open_confirm",
481 	"rfs4_op_open_downgrade",
482 	"rfs4_op_putfh",
483 	"rfs4_op_putpubfh",
484 	"rfs4_op_putrootfh",
485 	"rfs4_op_read",
486 	"rfs4_op_readdir",
487 	"rfs4_op_readlink",
488 	"rfs4_op_remove",
489 	"rfs4_op_rename",
490 	"rfs4_op_renew",
491 	"rfs4_op_restorefh",
492 	"rfs4_op_savefh",
493 	"rfs4_op_secinfo",
494 	"rfs4_op_setattr",
495 	"rfs4_op_setclientid",
496 	"rfs4_op_setclient_confirm",
497 	"rfs4_op_verify",
498 	"rfs4_op_write",
499 	"rfs4_op_release_lockowner",
500 	"rfs4_op_illegal"
501 };
502 #endif
503 
504 void	rfs4_ss_chkclid(rfs4_client_t *);
505 
506 extern size_t   strlcpy(char *dst, const char *src, size_t dstsize);
507 
508 extern void	rfs4_free_fs_locations4(fs_locations4 *);
509 
510 #ifdef	nextdp
511 #undef nextdp
512 #endif
513 #define	nextdp(dp)	((struct dirent64 *)((char *)(dp) + (dp)->d_reclen))
514 
515 static const fs_operation_def_t nfs4_rd_deleg_tmpl[] = {
516 	VOPNAME_OPEN,		{ .femop_open = deleg_rd_open },
517 	VOPNAME_WRITE,		{ .femop_write = deleg_rd_write },
518 	VOPNAME_SETATTR,	{ .femop_setattr = deleg_rd_setattr },
519 	VOPNAME_RWLOCK,		{ .femop_rwlock = deleg_rd_rwlock },
520 	VOPNAME_SPACE,		{ .femop_space = deleg_rd_space },
521 	VOPNAME_SETSECATTR,	{ .femop_setsecattr = deleg_rd_setsecattr },
522 	VOPNAME_VNEVENT,	{ .femop_vnevent = deleg_rd_vnevent },
523 	NULL,			NULL
524 };
525 static const fs_operation_def_t nfs4_wr_deleg_tmpl[] = {
526 	VOPNAME_OPEN,		{ .femop_open = deleg_wr_open },
527 	VOPNAME_READ,		{ .femop_read = deleg_wr_read },
528 	VOPNAME_WRITE,		{ .femop_write = deleg_wr_write },
529 	VOPNAME_SETATTR,	{ .femop_setattr = deleg_wr_setattr },
530 	VOPNAME_RWLOCK,		{ .femop_rwlock = deleg_wr_rwlock },
531 	VOPNAME_SPACE,		{ .femop_space = deleg_wr_space },
532 	VOPNAME_SETSECATTR,	{ .femop_setsecattr = deleg_wr_setsecattr },
533 	VOPNAME_VNEVENT,	{ .femop_vnevent = deleg_wr_vnevent },
534 	NULL,			NULL
535 };
536 
537 int
538 rfs4_srvrinit(void)
539 {
540 	timespec32_t verf;
541 	int error;
542 	extern void rfs4_attr_init();
543 	extern krwlock_t rfs4_deleg_policy_lock;
544 
545 	/*
546 	 * The following algorithm attempts to find a unique verifier
547 	 * to be used as the write verifier returned from the server
548 	 * to the client.  It is important that this verifier change
549 	 * whenever the server reboots.  Of secondary importance, it
550 	 * is important for the verifier to be unique between two
551 	 * different servers.
552 	 *
553 	 * Thus, an attempt is made to use the system hostid and the
554 	 * current time in seconds when the nfssrv kernel module is
555 	 * loaded.  It is assumed that an NFS server will not be able
556 	 * to boot and then to reboot in less than a second.  If the
557 	 * hostid has not been set, then the current high resolution
558 	 * time is used.  This will ensure different verifiers each
559 	 * time the server reboots and minimize the chances that two
560 	 * different servers will have the same verifier.
561 	 * XXX - this is broken on LP64 kernels.
562 	 */
563 	verf.tv_sec = (time_t)zone_get_hostid(NULL);
564 	if (verf.tv_sec != 0) {
565 		verf.tv_nsec = gethrestime_sec();
566 	} else {
567 		timespec_t tverf;
568 
569 		gethrestime(&tverf);
570 		verf.tv_sec = (time_t)tverf.tv_sec;
571 		verf.tv_nsec = tverf.tv_nsec;
572 	}
573 
574 	Write4verf = *(uint64_t *)&verf;
575 
576 	rfs4_attr_init();
577 	mutex_init(&rfs4_deleg_lock, NULL, MUTEX_DEFAULT, NULL);
578 
579 	/* Used to manage create/destroy of server state */
580 	mutex_init(&rfs4_state_lock, NULL, MUTEX_DEFAULT, NULL);
581 
582 	/* Used to manage access to server instance linked list */
583 	mutex_init(&rfs4_servinst_lock, NULL, MUTEX_DEFAULT, NULL);
584 
585 	/* Used to manage access to rfs4_deleg_policy */
586 	rw_init(&rfs4_deleg_policy_lock, NULL, RW_DEFAULT, NULL);
587 
588 	error = fem_create("deleg_rdops", nfs4_rd_deleg_tmpl, &deleg_rdops);
589 	if (error != 0) {
590 		rfs4_disable_delegation();
591 	} else {
592 		error = fem_create("deleg_wrops", nfs4_wr_deleg_tmpl,
593 		    &deleg_wrops);
594 		if (error != 0) {
595 			rfs4_disable_delegation();
596 			fem_free(deleg_rdops);
597 		}
598 	}
599 
600 	nfs4_srv_caller_id = fs_new_caller_id();
601 
602 	lockt_sysid = lm_alloc_sysidt();
603 
604 	vsd_create(&nfs4_srv_vkey, NULL);
605 
606 	return (0);
607 }
608 
609 void
610 rfs4_srvrfini(void)
611 {
612 	extern krwlock_t rfs4_deleg_policy_lock;
613 
614 	if (lockt_sysid != LM_NOSYSID) {
615 		lm_free_sysidt(lockt_sysid);
616 		lockt_sysid = LM_NOSYSID;
617 	}
618 
619 	mutex_destroy(&rfs4_deleg_lock);
620 	mutex_destroy(&rfs4_state_lock);
621 	rw_destroy(&rfs4_deleg_policy_lock);
622 
623 	fem_free(deleg_rdops);
624 	fem_free(deleg_wrops);
625 }
626 
627 void
628 rfs4_init_compound_state(struct compound_state *cs)
629 {
630 	bzero(cs, sizeof (*cs));
631 	cs->cont = TRUE;
632 	cs->access = CS_ACCESS_DENIED;
633 	cs->deleg = FALSE;
634 	cs->mandlock = FALSE;
635 	cs->fh.nfs_fh4_val = cs->fhbuf;
636 }
637 
638 void
639 rfs4_grace_start(rfs4_servinst_t *sip)
640 {
641 	rw_enter(&sip->rwlock, RW_WRITER);
642 	sip->start_time = (time_t)TICK_TO_SEC(ddi_get_lbolt());
643 	sip->grace_period = rfs4_grace_period;
644 	rw_exit(&sip->rwlock);
645 }
646 
647 /*
648  * returns true if the instance's grace period has never been started
649  */
650 int
651 rfs4_servinst_grace_new(rfs4_servinst_t *sip)
652 {
653 	time_t start_time;
654 
655 	rw_enter(&sip->rwlock, RW_READER);
656 	start_time = sip->start_time;
657 	rw_exit(&sip->rwlock);
658 
659 	return (start_time == 0);
660 }
661 
662 /*
663  * Indicates if server instance is within the
664  * grace period.
665  */
666 int
667 rfs4_servinst_in_grace(rfs4_servinst_t *sip)
668 {
669 	time_t grace_expiry;
670 
671 	rw_enter(&sip->rwlock, RW_READER);
672 	grace_expiry = sip->start_time + sip->grace_period;
673 	rw_exit(&sip->rwlock);
674 
675 	return (((time_t)TICK_TO_SEC(ddi_get_lbolt())) < grace_expiry);
676 }
677 
678 int
679 rfs4_clnt_in_grace(rfs4_client_t *cp)
680 {
681 	ASSERT(rfs4_dbe_refcnt(cp->rc_dbe) > 0);
682 
683 	return (rfs4_servinst_in_grace(cp->rc_server_instance));
684 }
685 
686 /*
687  * reset all currently active grace periods
688  */
689 void
690 rfs4_grace_reset_all(void)
691 {
692 	rfs4_servinst_t *sip;
693 
694 	mutex_enter(&rfs4_servinst_lock);
695 	for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev)
696 		if (rfs4_servinst_in_grace(sip))
697 			rfs4_grace_start(sip);
698 	mutex_exit(&rfs4_servinst_lock);
699 }
700 
701 /*
702  * start any new instances' grace periods
703  */
704 void
705 rfs4_grace_start_new(void)
706 {
707 	rfs4_servinst_t *sip;
708 
709 	mutex_enter(&rfs4_servinst_lock);
710 	for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev)
711 		if (rfs4_servinst_grace_new(sip))
712 			rfs4_grace_start(sip);
713 	mutex_exit(&rfs4_servinst_lock);
714 }
715 
716 static rfs4_dss_path_t *
717 rfs4_dss_newpath(rfs4_servinst_t *sip, char *path, unsigned index)
718 {
719 	size_t len;
720 	rfs4_dss_path_t *dss_path;
721 
722 	dss_path = kmem_alloc(sizeof (rfs4_dss_path_t), KM_SLEEP);
723 
724 	/*
725 	 * Take a copy of the string, since the original may be overwritten.
726 	 * Sadly, no strdup() in the kernel.
727 	 */
728 	/* allow for NUL */
729 	len = strlen(path) + 1;
730 	dss_path->path = kmem_alloc(len, KM_SLEEP);
731 	(void) strlcpy(dss_path->path, path, len);
732 
733 	/* associate with servinst */
734 	dss_path->sip = sip;
735 	dss_path->index = index;
736 
737 	/*
738 	 * Add to list of served paths.
739 	 * No locking required, as we're only ever called at startup.
740 	 */
741 	if (rfs4_dss_pathlist == NULL) {
742 		/* this is the first dss_path_t */
743 
744 		/* needed for insque/remque */
745 		dss_path->next = dss_path->prev = dss_path;
746 
747 		rfs4_dss_pathlist = dss_path;
748 	} else {
749 		insque(dss_path, rfs4_dss_pathlist);
750 	}
751 
752 	return (dss_path);
753 }
754 
755 /*
756  * Create a new server instance, and make it the currently active instance.
757  * Note that starting the grace period too early will reduce the clients'
758  * recovery window.
759  */
760 void
761 rfs4_servinst_create(int start_grace, int dss_npaths, char **dss_paths)
762 {
763 	unsigned i;
764 	rfs4_servinst_t *sip;
765 	rfs4_oldstate_t *oldstate;
766 
767 	sip = kmem_alloc(sizeof (rfs4_servinst_t), KM_SLEEP);
768 	rw_init(&sip->rwlock, NULL, RW_DEFAULT, NULL);
769 
770 	sip->start_time = (time_t)0;
771 	sip->grace_period = (time_t)0;
772 	sip->next = NULL;
773 	sip->prev = NULL;
774 
775 	rw_init(&sip->oldstate_lock, NULL, RW_DEFAULT, NULL);
776 	/*
777 	 * This initial dummy entry is required to setup for insque/remque.
778 	 * It must be skipped over whenever the list is traversed.
779 	 */
780 	oldstate = kmem_alloc(sizeof (rfs4_oldstate_t), KM_SLEEP);
781 	/* insque/remque require initial list entry to be self-terminated */
782 	oldstate->next = oldstate;
783 	oldstate->prev = oldstate;
784 	sip->oldstate = oldstate;
785 
786 
787 	sip->dss_npaths = dss_npaths;
788 	sip->dss_paths = kmem_alloc(dss_npaths *
789 	    sizeof (rfs4_dss_path_t *), KM_SLEEP);
790 
791 	for (i = 0; i < dss_npaths; i++) {
792 		sip->dss_paths[i] = rfs4_dss_newpath(sip, dss_paths[i], i);
793 	}
794 
795 	mutex_enter(&rfs4_servinst_lock);
796 	if (rfs4_cur_servinst != NULL) {
797 		/* add to linked list */
798 		sip->prev = rfs4_cur_servinst;
799 		rfs4_cur_servinst->next = sip;
800 	}
801 	if (start_grace)
802 		rfs4_grace_start(sip);
803 	/* make the new instance "current" */
804 	rfs4_cur_servinst = sip;
805 
806 	mutex_exit(&rfs4_servinst_lock);
807 }
808 
809 /*
810  * In future, we might add a rfs4_servinst_destroy(sip) but, for now, destroy
811  * all instances directly.
812  */
813 void
814 rfs4_servinst_destroy_all(void)
815 {
816 	rfs4_servinst_t *sip, *prev, *current;
817 #ifdef DEBUG
818 	int n = 0;
819 #endif
820 
821 	mutex_enter(&rfs4_servinst_lock);
822 	ASSERT(rfs4_cur_servinst != NULL);
823 	current = rfs4_cur_servinst;
824 	rfs4_cur_servinst = NULL;
825 	for (sip = current; sip != NULL; sip = prev) {
826 		prev = sip->prev;
827 		rw_destroy(&sip->rwlock);
828 		if (sip->oldstate)
829 			kmem_free(sip->oldstate, sizeof (rfs4_oldstate_t));
830 		if (sip->dss_paths)
831 			kmem_free(sip->dss_paths,
832 			    sip->dss_npaths * sizeof (rfs4_dss_path_t *));
833 		kmem_free(sip, sizeof (rfs4_servinst_t));
834 #ifdef DEBUG
835 		n++;
836 #endif
837 	}
838 	mutex_exit(&rfs4_servinst_lock);
839 }
840 
841 /*
842  * Assign the current server instance to a client_t.
843  * Should be called with cp->rc_dbe held.
844  */
845 void
846 rfs4_servinst_assign(rfs4_client_t *cp, rfs4_servinst_t *sip)
847 {
848 	ASSERT(rfs4_dbe_refcnt(cp->rc_dbe) > 0);
849 
850 	/*
851 	 * The lock ensures that if the current instance is in the process
852 	 * of changing, we will see the new one.
853 	 */
854 	mutex_enter(&rfs4_servinst_lock);
855 	cp->rc_server_instance = sip;
856 	mutex_exit(&rfs4_servinst_lock);
857 }
858 
859 rfs4_servinst_t *
860 rfs4_servinst(rfs4_client_t *cp)
861 {
862 	ASSERT(rfs4_dbe_refcnt(cp->rc_dbe) > 0);
863 
864 	return (cp->rc_server_instance);
865 }
866 
867 /* ARGSUSED */
868 static void
869 nullfree(caddr_t resop)
870 {
871 }
872 
873 /*
874  * This is a fall-through for invalid or not implemented (yet) ops
875  */
876 /* ARGSUSED */
877 static void
878 rfs4_op_inval(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
879     struct compound_state *cs)
880 {
881 	*cs->statusp = *((nfsstat4 *)&(resop)->nfs_resop4_u) = NFS4ERR_INVAL;
882 }
883 
884 /*
885  * Check if the security flavor, nfsnum, is in the flavor_list.
886  */
887 bool_t
888 in_flavor_list(int nfsnum, int *flavor_list, int count)
889 {
890 	int i;
891 
892 	for (i = 0; i < count; i++) {
893 		if (nfsnum == flavor_list[i])
894 			return (TRUE);
895 	}
896 	return (FALSE);
897 }
898 
899 /*
900  * Used by rfs4_op_secinfo to get the security information from the
901  * export structure associated with the component.
902  */
903 /* ARGSUSED */
904 static nfsstat4
905 do_rfs4_op_secinfo(struct compound_state *cs, char *nm, SECINFO4res *resp)
906 {
907 	int error, different_export = 0;
908 	vnode_t *dvp, *vp;
909 	struct exportinfo *exi = NULL;
910 	fid_t fid;
911 	uint_t count, i;
912 	secinfo4 *resok_val;
913 	struct secinfo *secp;
914 	seconfig_t *si;
915 	bool_t did_traverse = FALSE;
916 	int dotdot, walk;
917 
918 	dvp = cs->vp;
919 	dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0');
920 
921 	/*
922 	 * If dotdotting, then need to check whether it's above the
923 	 * root of a filesystem, or above an export point.
924 	 */
925 	if (dotdot) {
926 
927 		/*
928 		 * If dotdotting at the root of a filesystem, then
929 		 * need to traverse back to the mounted-on filesystem
930 		 * and do the dotdot lookup there.
931 		 */
932 		if (cs->vp->v_flag & VROOT) {
933 
934 			/*
935 			 * If at the system root, then can
936 			 * go up no further.
937 			 */
938 			if (VN_CMP(dvp, rootdir))
939 				return (puterrno4(ENOENT));
940 
941 			/*
942 			 * Traverse back to the mounted-on filesystem
943 			 */
944 			dvp = untraverse(cs->vp);
945 
946 			/*
947 			 * Set the different_export flag so we remember
948 			 * to pick up a new exportinfo entry for
949 			 * this new filesystem.
950 			 */
951 			different_export = 1;
952 		} else {
953 
954 			/*
955 			 * If dotdotting above an export point then set
956 			 * the different_export to get new export info.
957 			 */
958 			different_export = nfs_exported(cs->exi, cs->vp);
959 		}
960 	}
961 
962 	/*
963 	 * Get the vnode for the component "nm".
964 	 */
965 	error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cs->cr,
966 	    NULL, NULL, NULL);
967 	if (error)
968 		return (puterrno4(error));
969 
970 	/*
971 	 * If the vnode is in a pseudo filesystem, or if the security flavor
972 	 * used in the request is valid but not an explicitly shared flavor,
973 	 * or the access bit indicates that this is a limited access,
974 	 * check whether this vnode is visible.
975 	 */
976 	if (!different_export &&
977 	    (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) ||
978 	    cs->access & CS_ACCESS_LIMITED)) {
979 		if (! nfs_visible(cs->exi, vp, &different_export)) {
980 			VN_RELE(vp);
981 			return (puterrno4(ENOENT));
982 		}
983 	}
984 
985 	/*
986 	 * If it's a mountpoint, then traverse it.
987 	 */
988 	if (vn_ismntpt(vp)) {
989 		if ((error = traverse(&vp)) != 0) {
990 			VN_RELE(vp);
991 			return (puterrno4(error));
992 		}
993 		/* remember that we had to traverse mountpoint */
994 		did_traverse = TRUE;
995 		different_export = 1;
996 	} else if (vp->v_vfsp != dvp->v_vfsp) {
997 		/*
998 		 * If vp isn't a mountpoint and the vfs ptrs aren't the same,
999 		 * then vp is probably an LOFS object.  We don't need the
1000 		 * realvp, we just need to know that we might have crossed
1001 		 * a server fs boundary and need to call checkexport4.
1002 		 * (LOFS lookup hides server fs mountpoints, and actually calls
1003 		 * traverse)
1004 		 */
1005 		different_export = 1;
1006 	}
1007 
1008 	/*
1009 	 * Get the export information for it.
1010 	 */
1011 	if (different_export) {
1012 
1013 		bzero(&fid, sizeof (fid));
1014 		fid.fid_len = MAXFIDSZ;
1015 		error = vop_fid_pseudo(vp, &fid);
1016 		if (error) {
1017 			VN_RELE(vp);
1018 			return (puterrno4(error));
1019 		}
1020 
1021 		if (dotdot)
1022 			exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE);
1023 		else
1024 			exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
1025 
1026 		if (exi == NULL) {
1027 			if (did_traverse == TRUE) {
1028 				/*
1029 				 * If this vnode is a mounted-on vnode,
1030 				 * but the mounted-on file system is not
1031 				 * exported, send back the secinfo for
1032 				 * the exported node that the mounted-on
1033 				 * vnode lives in.
1034 				 */
1035 				exi = cs->exi;
1036 			} else {
1037 				VN_RELE(vp);
1038 				return (puterrno4(EACCES));
1039 			}
1040 		}
1041 	} else {
1042 		exi = cs->exi;
1043 	}
1044 	ASSERT(exi != NULL);
1045 
1046 
1047 	/*
1048 	 * Create the secinfo result based on the security information
1049 	 * from the exportinfo structure (exi).
1050 	 *
1051 	 * Return all flavors for a pseudo node.
1052 	 * For a real export node, return the flavor that the client
1053 	 * has access with.
1054 	 */
1055 	ASSERT(RW_LOCK_HELD(&exported_lock));
1056 	if (PSEUDO(exi)) {
1057 		count = exi->exi_export.ex_seccnt; /* total sec count */
1058 		resok_val = kmem_alloc(count * sizeof (secinfo4), KM_SLEEP);
1059 		secp = exi->exi_export.ex_secinfo;
1060 
1061 		for (i = 0; i < count; i++) {
1062 			si = &secp[i].s_secinfo;
1063 			resok_val[i].flavor = si->sc_rpcnum;
1064 			if (resok_val[i].flavor == RPCSEC_GSS) {
1065 				rpcsec_gss_info *info;
1066 
1067 				info = &resok_val[i].flavor_info;
1068 				info->qop = si->sc_qop;
1069 				info->service = (rpc_gss_svc_t)si->sc_service;
1070 
1071 				/* get oid opaque data */
1072 				info->oid.sec_oid4_len =
1073 				    si->sc_gss_mech_type->length;
1074 				info->oid.sec_oid4_val = kmem_alloc(
1075 				    si->sc_gss_mech_type->length, KM_SLEEP);
1076 				bcopy(
1077 				    si->sc_gss_mech_type->elements,
1078 				    info->oid.sec_oid4_val,
1079 				    info->oid.sec_oid4_len);
1080 			}
1081 		}
1082 		resp->SECINFO4resok_len = count;
1083 		resp->SECINFO4resok_val = resok_val;
1084 	} else {
1085 		int ret_cnt = 0, k = 0;
1086 		int *flavor_list;
1087 
1088 		count = exi->exi_export.ex_seccnt; /* total sec count */
1089 		secp = exi->exi_export.ex_secinfo;
1090 
1091 		flavor_list = kmem_alloc(count * sizeof (int), KM_SLEEP);
1092 		/* find out which flavors to return */
1093 		for (i = 0; i < count; i ++) {
1094 			int access, flavor, perm;
1095 
1096 			flavor = secp[i].s_secinfo.sc_nfsnum;
1097 			perm = secp[i].s_flags;
1098 
1099 			access = nfsauth4_secinfo_access(exi, cs->req,
1100 			    flavor, perm, cs->basecr);
1101 
1102 			if (! (access & NFSAUTH_DENIED) &&
1103 			    ! (access & NFSAUTH_WRONGSEC)) {
1104 				flavor_list[ret_cnt] = flavor;
1105 				ret_cnt++;
1106 			}
1107 		}
1108 
1109 		/* Create the returning SECINFO value */
1110 		resok_val = kmem_alloc(ret_cnt * sizeof (secinfo4), KM_SLEEP);
1111 
1112 		for (i = 0; i < count; i++) {
1113 			/*
1114 			 * If the flavor is in the flavor list,
1115 			 * fill in resok_val.
1116 			 */
1117 			si = &secp[i].s_secinfo;
1118 			if (in_flavor_list(si->sc_nfsnum,
1119 			    flavor_list, ret_cnt)) {
1120 				resok_val[k].flavor = si->sc_rpcnum;
1121 				if (resok_val[k].flavor == RPCSEC_GSS) {
1122 					rpcsec_gss_info *info;
1123 
1124 					info = &resok_val[k].flavor_info;
1125 					info->qop = si->sc_qop;
1126 					info->service = (rpc_gss_svc_t)
1127 					    si->sc_service;
1128 
1129 					/* get oid opaque data */
1130 					info->oid.sec_oid4_len =
1131 					    si->sc_gss_mech_type->length;
1132 					info->oid.sec_oid4_val = kmem_alloc(
1133 					    si->sc_gss_mech_type->length,
1134 					    KM_SLEEP);
1135 					bcopy(si->sc_gss_mech_type->elements,
1136 					    info->oid.sec_oid4_val,
1137 					    info->oid.sec_oid4_len);
1138 				}
1139 				k++;
1140 			}
1141 			if (k >= ret_cnt)
1142 				break;
1143 		}
1144 		resp->SECINFO4resok_len = ret_cnt;
1145 		resp->SECINFO4resok_val = resok_val;
1146 		kmem_free(flavor_list, count * sizeof (int));
1147 	}
1148 
1149 	VN_RELE(vp);
1150 	return (NFS4_OK);
1151 }
1152 
1153 /*
1154  * SECINFO (Operation 33): Obtain required security information on
1155  * the component name in the format of (security-mechanism-oid, qop, service)
1156  * triplets.
1157  */
1158 /* ARGSUSED */
1159 static void
1160 rfs4_op_secinfo(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1161     struct compound_state *cs)
1162 {
1163 	SECINFO4args *args = &argop->nfs_argop4_u.opsecinfo;
1164 	SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo;
1165 	utf8string *utfnm = &args->name;
1166 	uint_t len;
1167 	char *nm;
1168 	struct sockaddr *ca;
1169 	char *name = NULL;
1170 	nfsstat4 status = NFS4_OK;
1171 
1172 	DTRACE_NFSV4_2(op__secinfo__start, struct compound_state *, cs,
1173 	    SECINFO4args *, args);
1174 
1175 	/*
1176 	 * Current file handle (cfh) should have been set before getting
1177 	 * into this function. If not, return error.
1178 	 */
1179 	if (cs->vp == NULL) {
1180 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1181 		goto out;
1182 	}
1183 
1184 	if (cs->vp->v_type != VDIR) {
1185 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
1186 		goto out;
1187 	}
1188 
1189 	/*
1190 	 * Verify the component name. If failed, error out, but
1191 	 * do not error out if the component name is a "..".
1192 	 * SECINFO will return its parents secinfo data for SECINFO "..".
1193 	 */
1194 	status = utf8_dir_verify(utfnm);
1195 	if (status != NFS4_OK) {
1196 		if (utfnm->utf8string_len != 2 ||
1197 		    utfnm->utf8string_val[0] != '.' ||
1198 		    utfnm->utf8string_val[1] != '.') {
1199 			*cs->statusp = resp->status = status;
1200 			goto out;
1201 		}
1202 	}
1203 
1204 	nm = utf8_to_str(utfnm, &len, NULL);
1205 	if (nm == NULL) {
1206 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1207 		goto out;
1208 	}
1209 
1210 	if (len > MAXNAMELEN) {
1211 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1212 		kmem_free(nm, len);
1213 		goto out;
1214 	}
1215 
1216 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
1217 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
1218 	    MAXPATHLEN  + 1);
1219 
1220 	if (name == NULL) {
1221 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1222 		kmem_free(nm, len);
1223 		goto out;
1224 	}
1225 
1226 
1227 	*cs->statusp = resp->status = do_rfs4_op_secinfo(cs, name, resp);
1228 
1229 	if (name != nm)
1230 		kmem_free(name, MAXPATHLEN + 1);
1231 	kmem_free(nm, len);
1232 
1233 out:
1234 	DTRACE_NFSV4_2(op__secinfo__done, struct compound_state *, cs,
1235 	    SECINFO4res *, resp);
1236 }
1237 
1238 /*
1239  * Free SECINFO result.
1240  */
1241 /* ARGSUSED */
1242 static void
1243 rfs4_op_secinfo_free(nfs_resop4 *resop)
1244 {
1245 	SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo;
1246 	int count, i;
1247 	secinfo4 *resok_val;
1248 
1249 	/* If this is not an Ok result, nothing to free. */
1250 	if (resp->status != NFS4_OK) {
1251 		return;
1252 	}
1253 
1254 	count = resp->SECINFO4resok_len;
1255 	resok_val = resp->SECINFO4resok_val;
1256 
1257 	for (i = 0; i < count; i++) {
1258 		if (resok_val[i].flavor == RPCSEC_GSS) {
1259 			rpcsec_gss_info *info;
1260 
1261 			info = &resok_val[i].flavor_info;
1262 			kmem_free(info->oid.sec_oid4_val,
1263 			    info->oid.sec_oid4_len);
1264 		}
1265 	}
1266 	kmem_free(resok_val, count * sizeof (secinfo4));
1267 	resp->SECINFO4resok_len = 0;
1268 	resp->SECINFO4resok_val = NULL;
1269 }
1270 
1271 /* ARGSUSED */
1272 static void
1273 rfs4_op_access(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1274     struct compound_state *cs)
1275 {
1276 	ACCESS4args *args = &argop->nfs_argop4_u.opaccess;
1277 	ACCESS4res *resp = &resop->nfs_resop4_u.opaccess;
1278 	int error;
1279 	vnode_t *vp;
1280 	struct vattr va;
1281 	int checkwriteperm;
1282 	cred_t *cr = cs->cr;
1283 	bslabel_t *clabel, *slabel;
1284 	ts_label_t *tslabel;
1285 	boolean_t admin_low_client;
1286 
1287 	DTRACE_NFSV4_2(op__access__start, struct compound_state *, cs,
1288 	    ACCESS4args *, args);
1289 
1290 #if 0	/* XXX allow access even if !cs->access. Eventually only pseudo fs */
1291 	if (cs->access == CS_ACCESS_DENIED) {
1292 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1293 		goto out;
1294 	}
1295 #endif
1296 	if (cs->vp == NULL) {
1297 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1298 		goto out;
1299 	}
1300 
1301 	ASSERT(cr != NULL);
1302 
1303 	vp = cs->vp;
1304 
1305 	/*
1306 	 * If the file system is exported read only, it is not appropriate
1307 	 * to check write permissions for regular files and directories.
1308 	 * Special files are interpreted by the client, so the underlying
1309 	 * permissions are sent back to the client for interpretation.
1310 	 */
1311 	if (rdonly4(req, cs) &&
1312 	    (vp->v_type == VREG || vp->v_type == VDIR))
1313 		checkwriteperm = 0;
1314 	else
1315 		checkwriteperm = 1;
1316 
1317 	/*
1318 	 * XXX
1319 	 * We need the mode so that we can correctly determine access
1320 	 * permissions relative to a mandatory lock file.  Access to
1321 	 * mandatory lock files is denied on the server, so it might
1322 	 * as well be reflected to the server during the open.
1323 	 */
1324 	va.va_mask = AT_MODE;
1325 	error = VOP_GETATTR(vp, &va, 0, cr, NULL);
1326 	if (error) {
1327 		*cs->statusp = resp->status = puterrno4(error);
1328 		goto out;
1329 	}
1330 	resp->access = 0;
1331 	resp->supported = 0;
1332 
1333 	if (is_system_labeled()) {
1334 		ASSERT(req->rq_label != NULL);
1335 		clabel = req->rq_label;
1336 		DTRACE_PROBE2(tx__rfs4__log__info__opaccess__clabel, char *,
1337 		    "got client label from request(1)",
1338 		    struct svc_req *, req);
1339 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
1340 			if ((tslabel = nfs_getflabel(vp, cs->exi)) == NULL) {
1341 				*cs->statusp = resp->status = puterrno4(EACCES);
1342 				goto out;
1343 			}
1344 			slabel = label2bslabel(tslabel);
1345 			DTRACE_PROBE3(tx__rfs4__log__info__opaccess__slabel,
1346 			    char *, "got server label(1) for vp(2)",
1347 			    bslabel_t *, slabel, vnode_t *, vp);
1348 
1349 			admin_low_client = B_FALSE;
1350 		} else
1351 			admin_low_client = B_TRUE;
1352 	}
1353 
1354 	if (args->access & ACCESS4_READ) {
1355 		error = VOP_ACCESS(vp, VREAD, 0, cr, NULL);
1356 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1357 		    (!is_system_labeled() || admin_low_client ||
1358 		    bldominates(clabel, slabel)))
1359 			resp->access |= ACCESS4_READ;
1360 		resp->supported |= ACCESS4_READ;
1361 	}
1362 	if ((args->access & ACCESS4_LOOKUP) && vp->v_type == VDIR) {
1363 		error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL);
1364 		if (!error && (!is_system_labeled() || admin_low_client ||
1365 		    bldominates(clabel, slabel)))
1366 			resp->access |= ACCESS4_LOOKUP;
1367 		resp->supported |= ACCESS4_LOOKUP;
1368 	}
1369 	if (checkwriteperm &&
1370 	    (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND))) {
1371 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
1372 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1373 		    (!is_system_labeled() || admin_low_client ||
1374 		    blequal(clabel, slabel)))
1375 			resp->access |=
1376 			    (args->access & (ACCESS4_MODIFY | ACCESS4_EXTEND));
1377 		resp->supported |=
1378 		    resp->access & (ACCESS4_MODIFY | ACCESS4_EXTEND);
1379 	}
1380 
1381 	if (checkwriteperm &&
1382 	    (args->access & ACCESS4_DELETE) && vp->v_type == VDIR) {
1383 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
1384 		if (!error && (!is_system_labeled() || admin_low_client ||
1385 		    blequal(clabel, slabel)))
1386 			resp->access |= ACCESS4_DELETE;
1387 		resp->supported |= ACCESS4_DELETE;
1388 	}
1389 	if (args->access & ACCESS4_EXECUTE && vp->v_type != VDIR) {
1390 		error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL);
1391 		if (!error && !MANDLOCK(vp, va.va_mode) &&
1392 		    (!is_system_labeled() || admin_low_client ||
1393 		    bldominates(clabel, slabel)))
1394 			resp->access |= ACCESS4_EXECUTE;
1395 		resp->supported |= ACCESS4_EXECUTE;
1396 	}
1397 
1398 	if (is_system_labeled() && !admin_low_client)
1399 		label_rele(tslabel);
1400 
1401 	*cs->statusp = resp->status = NFS4_OK;
1402 out:
1403 	DTRACE_NFSV4_2(op__access__done, struct compound_state *, cs,
1404 	    ACCESS4res *, resp);
1405 }
1406 
1407 /* ARGSUSED */
1408 static void
1409 rfs4_op_commit(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1410     struct compound_state *cs)
1411 {
1412 	COMMIT4args *args = &argop->nfs_argop4_u.opcommit;
1413 	COMMIT4res *resp = &resop->nfs_resop4_u.opcommit;
1414 	int error;
1415 	vnode_t *vp = cs->vp;
1416 	cred_t *cr = cs->cr;
1417 	vattr_t va;
1418 
1419 	DTRACE_NFSV4_2(op__commit__start, struct compound_state *, cs,
1420 	    COMMIT4args *, args);
1421 
1422 	if (vp == NULL) {
1423 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1424 		goto out;
1425 	}
1426 	if (cs->access == CS_ACCESS_DENIED) {
1427 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1428 		goto out;
1429 	}
1430 
1431 	if (args->offset + args->count < args->offset) {
1432 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1433 		goto out;
1434 	}
1435 
1436 	va.va_mask = AT_UID;
1437 	error = VOP_GETATTR(vp, &va, 0, cr, NULL);
1438 
1439 	/*
1440 	 * If we can't get the attributes, then we can't do the
1441 	 * right access checking.  So, we'll fail the request.
1442 	 */
1443 	if (error) {
1444 		*cs->statusp = resp->status = puterrno4(error);
1445 		goto out;
1446 	}
1447 	if (rdonly4(req, cs)) {
1448 		*cs->statusp = resp->status = NFS4ERR_ROFS;
1449 		goto out;
1450 	}
1451 
1452 	if (vp->v_type != VREG) {
1453 		if (vp->v_type == VDIR)
1454 			resp->status = NFS4ERR_ISDIR;
1455 		else
1456 			resp->status = NFS4ERR_INVAL;
1457 		*cs->statusp = resp->status;
1458 		goto out;
1459 	}
1460 
1461 	if (crgetuid(cr) != va.va_uid &&
1462 	    (error = VOP_ACCESS(vp, VWRITE, 0, cs->cr, NULL))) {
1463 		*cs->statusp = resp->status = puterrno4(error);
1464 		goto out;
1465 	}
1466 
1467 	error = VOP_FSYNC(vp, FSYNC, cr, NULL);
1468 
1469 	if (error) {
1470 		*cs->statusp = resp->status = puterrno4(error);
1471 		goto out;
1472 	}
1473 
1474 	*cs->statusp = resp->status = NFS4_OK;
1475 	resp->writeverf = Write4verf;
1476 out:
1477 	DTRACE_NFSV4_2(op__commit__done, struct compound_state *, cs,
1478 	    COMMIT4res *, resp);
1479 }
1480 
1481 /*
1482  * rfs4_op_mknod is called from rfs4_op_create after all initial verification
1483  * was completed. It does the nfsv4 create for special files.
1484  */
1485 /* ARGSUSED */
1486 static vnode_t *
1487 do_rfs4_op_mknod(CREATE4args *args, CREATE4res *resp, struct svc_req *req,
1488     struct compound_state *cs, vattr_t *vap, char *nm)
1489 {
1490 	int error;
1491 	cred_t *cr = cs->cr;
1492 	vnode_t *dvp = cs->vp;
1493 	vnode_t *vp = NULL;
1494 	int mode;
1495 	enum vcexcl excl;
1496 
1497 	switch (args->type) {
1498 	case NF4CHR:
1499 	case NF4BLK:
1500 		if (secpolicy_sys_devices(cr) != 0) {
1501 			*cs->statusp = resp->status = NFS4ERR_PERM;
1502 			return (NULL);
1503 		}
1504 		if (args->type == NF4CHR)
1505 			vap->va_type = VCHR;
1506 		else
1507 			vap->va_type = VBLK;
1508 		vap->va_rdev = makedevice(args->ftype4_u.devdata.specdata1,
1509 		    args->ftype4_u.devdata.specdata2);
1510 		vap->va_mask |= AT_RDEV;
1511 		break;
1512 	case NF4SOCK:
1513 		vap->va_type = VSOCK;
1514 		break;
1515 	case NF4FIFO:
1516 		vap->va_type = VFIFO;
1517 		break;
1518 	default:
1519 		*cs->statusp = resp->status = NFS4ERR_BADTYPE;
1520 		return (NULL);
1521 	}
1522 
1523 	/*
1524 	 * Must specify the mode.
1525 	 */
1526 	if (!(vap->va_mask & AT_MODE)) {
1527 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1528 		return (NULL);
1529 	}
1530 
1531 	excl = EXCL;
1532 
1533 	mode = 0;
1534 
1535 	error = VOP_CREATE(dvp, nm, vap, excl, mode, &vp, cr, 0, NULL, NULL);
1536 	if (error) {
1537 		*cs->statusp = resp->status = puterrno4(error);
1538 		return (NULL);
1539 	}
1540 	return (vp);
1541 }
1542 
1543 /*
1544  * nfsv4 create is used to create non-regular files. For regular files,
1545  * use nfsv4 open.
1546  */
1547 /* ARGSUSED */
1548 static void
1549 rfs4_op_create(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1550     struct compound_state *cs)
1551 {
1552 	CREATE4args *args = &argop->nfs_argop4_u.opcreate;
1553 	CREATE4res *resp = &resop->nfs_resop4_u.opcreate;
1554 	int error;
1555 	struct vattr bva, iva, iva2, ava, *vap;
1556 	cred_t *cr = cs->cr;
1557 	vnode_t *dvp = cs->vp;
1558 	vnode_t *vp = NULL;
1559 	vnode_t *realvp;
1560 	char *nm, *lnm;
1561 	uint_t len, llen;
1562 	int syncval = 0;
1563 	struct nfs4_svgetit_arg sarg;
1564 	struct nfs4_ntov_table ntov;
1565 	struct statvfs64 sb;
1566 	nfsstat4 status;
1567 	struct sockaddr *ca;
1568 	char *name = NULL;
1569 	char *lname = NULL;
1570 
1571 	DTRACE_NFSV4_2(op__create__start, struct compound_state *, cs,
1572 	    CREATE4args *, args);
1573 
1574 	resp->attrset = 0;
1575 
1576 	if (dvp == NULL) {
1577 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
1578 		goto out;
1579 	}
1580 
1581 	/*
1582 	 * If there is an unshared filesystem mounted on this vnode,
1583 	 * do not allow to create an object in this directory.
1584 	 */
1585 	if (vn_ismntpt(dvp)) {
1586 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1587 		goto out;
1588 	}
1589 
1590 	/* Verify that type is correct */
1591 	switch (args->type) {
1592 	case NF4LNK:
1593 	case NF4BLK:
1594 	case NF4CHR:
1595 	case NF4SOCK:
1596 	case NF4FIFO:
1597 	case NF4DIR:
1598 		break;
1599 	default:
1600 		*cs->statusp = resp->status = NFS4ERR_BADTYPE;
1601 		goto out;
1602 	};
1603 
1604 	if (cs->access == CS_ACCESS_DENIED) {
1605 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
1606 		goto out;
1607 	}
1608 	if (dvp->v_type != VDIR) {
1609 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
1610 		goto out;
1611 	}
1612 	status = utf8_dir_verify(&args->objname);
1613 	if (status != NFS4_OK) {
1614 		*cs->statusp = resp->status = status;
1615 		goto out;
1616 	}
1617 
1618 	if (rdonly4(req, cs)) {
1619 		*cs->statusp = resp->status = NFS4ERR_ROFS;
1620 		goto out;
1621 	}
1622 
1623 	/*
1624 	 * Name of newly created object
1625 	 */
1626 	nm = utf8_to_fn(&args->objname, &len, NULL);
1627 	if (nm == NULL) {
1628 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1629 		goto out;
1630 	}
1631 
1632 	if (len > MAXNAMELEN) {
1633 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1634 		kmem_free(nm, len);
1635 		goto out;
1636 	}
1637 
1638 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
1639 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
1640 	    MAXPATHLEN  + 1);
1641 
1642 	if (name == NULL) {
1643 		*cs->statusp = resp->status = NFS4ERR_INVAL;
1644 		kmem_free(nm, len);
1645 		goto out;
1646 	}
1647 
1648 	resp->attrset = 0;
1649 
1650 	sarg.sbp = &sb;
1651 	sarg.is_referral = B_FALSE;
1652 	nfs4_ntov_table_init(&ntov);
1653 
1654 	status = do_rfs4_set_attrs(&resp->attrset,
1655 	    &args->createattrs, cs, &sarg, &ntov, NFS4ATTR_SETIT);
1656 
1657 	if (sarg.vap->va_mask == 0 && status == NFS4_OK)
1658 		status = NFS4ERR_INVAL;
1659 
1660 	if (status != NFS4_OK) {
1661 		*cs->statusp = resp->status = status;
1662 		if (name != nm)
1663 			kmem_free(name, MAXPATHLEN + 1);
1664 		kmem_free(nm, len);
1665 		nfs4_ntov_table_free(&ntov, &sarg);
1666 		resp->attrset = 0;
1667 		goto out;
1668 	}
1669 
1670 	/* Get "before" change value */
1671 	bva.va_mask = AT_CTIME|AT_SEQ|AT_MODE;
1672 	error = VOP_GETATTR(dvp, &bva, 0, cr, NULL);
1673 	if (error) {
1674 		*cs->statusp = resp->status = puterrno4(error);
1675 		if (name != nm)
1676 			kmem_free(name, MAXPATHLEN + 1);
1677 		kmem_free(nm, len);
1678 		nfs4_ntov_table_free(&ntov, &sarg);
1679 		resp->attrset = 0;
1680 		goto out;
1681 	}
1682 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bva.va_ctime)
1683 
1684 	vap = sarg.vap;
1685 
1686 	/*
1687 	 * Set the default initial values for attributes when the parent
1688 	 * directory does not have the VSUID/VSGID bit set and they have
1689 	 * not been specified in createattrs.
1690 	 */
1691 	if (!(bva.va_mode & VSUID) && (vap->va_mask & AT_UID) == 0) {
1692 		vap->va_uid = crgetuid(cr);
1693 		vap->va_mask |= AT_UID;
1694 	}
1695 	if (!(bva.va_mode & VSGID) && (vap->va_mask & AT_GID) == 0) {
1696 		vap->va_gid = crgetgid(cr);
1697 		vap->va_mask |= AT_GID;
1698 	}
1699 
1700 	vap->va_mask |= AT_TYPE;
1701 	switch (args->type) {
1702 	case NF4DIR:
1703 		vap->va_type = VDIR;
1704 		if ((vap->va_mask & AT_MODE) == 0) {
1705 			vap->va_mode = 0700;	/* default: owner rwx only */
1706 			vap->va_mask |= AT_MODE;
1707 		}
1708 		error = VOP_MKDIR(dvp, name, vap, &vp, cr, NULL, 0, NULL);
1709 		if (error)
1710 			break;
1711 
1712 		/*
1713 		 * Get the initial "after" sequence number, if it fails,
1714 		 * set to zero
1715 		 */
1716 		iva.va_mask = AT_SEQ;
1717 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL))
1718 			iva.va_seq = 0;
1719 		break;
1720 	case NF4LNK:
1721 		vap->va_type = VLNK;
1722 		if ((vap->va_mask & AT_MODE) == 0) {
1723 			vap->va_mode = 0700;	/* default: owner rwx only */
1724 			vap->va_mask |= AT_MODE;
1725 		}
1726 
1727 		/*
1728 		 * symlink names must be treated as data
1729 		 */
1730 		lnm = utf8_to_str((utf8string *)&args->ftype4_u.linkdata,
1731 		    &llen, NULL);
1732 
1733 		if (lnm == NULL) {
1734 			*cs->statusp = resp->status = NFS4ERR_INVAL;
1735 			if (name != nm)
1736 				kmem_free(name, MAXPATHLEN + 1);
1737 			kmem_free(nm, len);
1738 			nfs4_ntov_table_free(&ntov, &sarg);
1739 			resp->attrset = 0;
1740 			goto out;
1741 		}
1742 
1743 		if (llen > MAXPATHLEN) {
1744 			*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
1745 			if (name != nm)
1746 				kmem_free(name, MAXPATHLEN + 1);
1747 			kmem_free(nm, len);
1748 			kmem_free(lnm, llen);
1749 			nfs4_ntov_table_free(&ntov, &sarg);
1750 			resp->attrset = 0;
1751 			goto out;
1752 		}
1753 
1754 		lname = nfscmd_convname(ca, cs->exi, lnm,
1755 		    NFSCMD_CONV_INBOUND, MAXPATHLEN  + 1);
1756 
1757 		if (lname == NULL) {
1758 			*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
1759 			if (name != nm)
1760 				kmem_free(name, MAXPATHLEN + 1);
1761 			kmem_free(nm, len);
1762 			kmem_free(lnm, llen);
1763 			nfs4_ntov_table_free(&ntov, &sarg);
1764 			resp->attrset = 0;
1765 			goto out;
1766 		}
1767 
1768 		error = VOP_SYMLINK(dvp, name, vap, lname, cr, NULL, 0);
1769 		if (lname != lnm)
1770 			kmem_free(lname, MAXPATHLEN + 1);
1771 		kmem_free(lnm, llen);
1772 		if (error)
1773 			break;
1774 
1775 		/*
1776 		 * Get the initial "after" sequence number, if it fails,
1777 		 * set to zero
1778 		 */
1779 		iva.va_mask = AT_SEQ;
1780 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL))
1781 			iva.va_seq = 0;
1782 
1783 		error = VOP_LOOKUP(dvp, name, &vp, NULL, 0, NULL, cr,
1784 		    NULL, NULL, NULL);
1785 		if (error)
1786 			break;
1787 
1788 		/*
1789 		 * va_seq is not safe over VOP calls, check it again
1790 		 * if it has changed zero out iva to force atomic = FALSE.
1791 		 */
1792 		iva2.va_mask = AT_SEQ;
1793 		if (VOP_GETATTR(dvp, &iva2, 0, cs->cr, NULL) ||
1794 		    iva2.va_seq != iva.va_seq)
1795 			iva.va_seq = 0;
1796 		break;
1797 	default:
1798 		/*
1799 		 * probably a special file.
1800 		 */
1801 		if ((vap->va_mask & AT_MODE) == 0) {
1802 			vap->va_mode = 0600;	/* default: owner rw only */
1803 			vap->va_mask |= AT_MODE;
1804 		}
1805 		syncval = FNODSYNC;
1806 		/*
1807 		 * We know this will only generate one VOP call
1808 		 */
1809 		vp = do_rfs4_op_mknod(args, resp, req, cs, vap, name);
1810 
1811 		if (vp == NULL) {
1812 			if (name != nm)
1813 				kmem_free(name, MAXPATHLEN + 1);
1814 			kmem_free(nm, len);
1815 			nfs4_ntov_table_free(&ntov, &sarg);
1816 			resp->attrset = 0;
1817 			goto out;
1818 		}
1819 
1820 		/*
1821 		 * Get the initial "after" sequence number, if it fails,
1822 		 * set to zero
1823 		 */
1824 		iva.va_mask = AT_SEQ;
1825 		if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL))
1826 			iva.va_seq = 0;
1827 
1828 		break;
1829 	}
1830 	if (name != nm)
1831 		kmem_free(name, MAXPATHLEN + 1);
1832 	kmem_free(nm, len);
1833 
1834 	if (error) {
1835 		*cs->statusp = resp->status = puterrno4(error);
1836 	}
1837 
1838 	/*
1839 	 * Force modified data and metadata out to stable storage.
1840 	 */
1841 	(void) VOP_FSYNC(dvp, 0, cr, NULL);
1842 
1843 	if (resp->status != NFS4_OK) {
1844 		if (vp != NULL)
1845 			VN_RELE(vp);
1846 		nfs4_ntov_table_free(&ntov, &sarg);
1847 		resp->attrset = 0;
1848 		goto out;
1849 	}
1850 
1851 	/*
1852 	 * Finish setup of cinfo response, "before" value already set.
1853 	 * Get "after" change value, if it fails, simply return the
1854 	 * before value.
1855 	 */
1856 	ava.va_mask = AT_CTIME|AT_SEQ;
1857 	if (VOP_GETATTR(dvp, &ava, 0, cr, NULL)) {
1858 		ava.va_ctime = bva.va_ctime;
1859 		ava.va_seq = 0;
1860 	}
1861 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, ava.va_ctime);
1862 
1863 	/*
1864 	 * True verification that object was created with correct
1865 	 * attrs is impossible.  The attrs could have been changed
1866 	 * immediately after object creation.  If attributes did
1867 	 * not verify, the only recourse for the server is to
1868 	 * destroy the object.  Maybe if some attrs (like gid)
1869 	 * are set incorrectly, the object should be destroyed;
1870 	 * however, seems bad as a default policy.  Do we really
1871 	 * want to destroy an object over one of the times not
1872 	 * verifying correctly?  For these reasons, the server
1873 	 * currently sets bits in attrset for createattrs
1874 	 * that were set; however, no verification is done.
1875 	 *
1876 	 * vmask_to_nmask accounts for vattr bits set on create
1877 	 *	[do_rfs4_set_attrs() only sets resp bits for
1878 	 *	 non-vattr/vfs bits.]
1879 	 * Mask off any bits set by default so as not to return
1880 	 * more attrset bits than were requested in createattrs
1881 	 */
1882 	nfs4_vmask_to_nmask(sarg.vap->va_mask, &resp->attrset);
1883 	resp->attrset &= args->createattrs.attrmask;
1884 	nfs4_ntov_table_free(&ntov, &sarg);
1885 
1886 	error = makefh4(&cs->fh, vp, cs->exi);
1887 	if (error) {
1888 		*cs->statusp = resp->status = puterrno4(error);
1889 	}
1890 
1891 	/*
1892 	 * The cinfo.atomic = TRUE only if we got no errors, we have
1893 	 * non-zero va_seq's, and it has incremented by exactly one
1894 	 * during the creation and it didn't change during the VOP_LOOKUP
1895 	 * or VOP_FSYNC.
1896 	 */
1897 	if (!error && bva.va_seq && iva.va_seq && ava.va_seq &&
1898 	    iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq)
1899 		resp->cinfo.atomic = TRUE;
1900 	else
1901 		resp->cinfo.atomic = FALSE;
1902 
1903 	/*
1904 	 * Force modified metadata out to stable storage.
1905 	 *
1906 	 * if a underlying vp exists, pass it to VOP_FSYNC
1907 	 */
1908 	if (VOP_REALVP(vp, &realvp, NULL) == 0)
1909 		(void) VOP_FSYNC(realvp, syncval, cr, NULL);
1910 	else
1911 		(void) VOP_FSYNC(vp, syncval, cr, NULL);
1912 
1913 	if (resp->status != NFS4_OK) {
1914 		VN_RELE(vp);
1915 		goto out;
1916 	}
1917 	if (cs->vp)
1918 		VN_RELE(cs->vp);
1919 
1920 	cs->vp = vp;
1921 	*cs->statusp = resp->status = NFS4_OK;
1922 out:
1923 	DTRACE_NFSV4_2(op__create__done, struct compound_state *, cs,
1924 	    CREATE4res *, resp);
1925 }
1926 
1927 /*ARGSUSED*/
1928 static void
1929 rfs4_op_delegpurge(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1930     struct compound_state *cs)
1931 {
1932 	DTRACE_NFSV4_2(op__delegpurge__start, struct compound_state *, cs,
1933 	    DELEGPURGE4args *, &argop->nfs_argop4_u.opdelegpurge);
1934 
1935 	rfs4_op_inval(argop, resop, req, cs);
1936 
1937 	DTRACE_NFSV4_2(op__delegpurge__done, struct compound_state *, cs,
1938 	    DELEGPURGE4res *, &resop->nfs_resop4_u.opdelegpurge);
1939 }
1940 
1941 /*ARGSUSED*/
1942 static void
1943 rfs4_op_delegreturn(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
1944     struct compound_state *cs)
1945 {
1946 	DELEGRETURN4args *args = &argop->nfs_argop4_u.opdelegreturn;
1947 	DELEGRETURN4res *resp = &resop->nfs_resop4_u.opdelegreturn;
1948 	rfs4_deleg_state_t *dsp;
1949 	nfsstat4 status;
1950 
1951 	DTRACE_NFSV4_2(op__delegreturn__start, struct compound_state *, cs,
1952 	    DELEGRETURN4args *, args);
1953 
1954 	status = rfs4_get_deleg_state(&args->deleg_stateid, &dsp);
1955 	resp->status = *cs->statusp = status;
1956 	if (status != NFS4_OK)
1957 		goto out;
1958 
1959 	/* Ensure specified filehandle matches */
1960 	if (cs->vp != dsp->rds_finfo->rf_vp) {
1961 		resp->status = *cs->statusp = NFS4ERR_BAD_STATEID;
1962 	} else
1963 		rfs4_return_deleg(dsp, FALSE);
1964 
1965 	rfs4_update_lease(dsp->rds_client);
1966 
1967 	rfs4_deleg_state_rele(dsp);
1968 out:
1969 	DTRACE_NFSV4_2(op__delegreturn__done, struct compound_state *, cs,
1970 	    DELEGRETURN4res *, resp);
1971 }
1972 
1973 /*
1974  * Check to see if a given "flavor" is an explicitly shared flavor.
1975  * The assumption of this routine is the "flavor" is already a valid
1976  * flavor in the secinfo list of "exi".
1977  *
1978  *	e.g.
1979  *		# share -o sec=flavor1 /export
1980  *		# share -o sec=flavor2 /export/home
1981  *
1982  *		flavor2 is not an explicitly shared flavor for /export,
1983  *		however it is in the secinfo list for /export thru the
1984  *		server namespace setup.
1985  */
1986 int
1987 is_exported_sec(int flavor, struct exportinfo *exi)
1988 {
1989 	int	i;
1990 	struct secinfo *sp;
1991 
1992 	sp = exi->exi_export.ex_secinfo;
1993 	for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
1994 		if (flavor == sp[i].s_secinfo.sc_nfsnum ||
1995 		    sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) {
1996 			return (SEC_REF_EXPORTED(&sp[i]));
1997 		}
1998 	}
1999 
2000 	/* Should not reach this point based on the assumption */
2001 	return (0);
2002 }
2003 
2004 /*
2005  * Check if the security flavor used in the request matches what is
2006  * required at the export point or at the root pseudo node (exi_root).
2007  *
2008  * returns 1 if there's a match or if exported with AUTH_NONE; 0 otherwise.
2009  *
2010  */
2011 static int
2012 secinfo_match_or_authnone(struct compound_state *cs)
2013 {
2014 	int	i;
2015 	struct secinfo *sp;
2016 
2017 	/*
2018 	 * Check cs->nfsflavor (from the request) against
2019 	 * the current export data in cs->exi.
2020 	 */
2021 	sp = cs->exi->exi_export.ex_secinfo;
2022 	for (i = 0; i < cs->exi->exi_export.ex_seccnt; i++) {
2023 		if (cs->nfsflavor == sp[i].s_secinfo.sc_nfsnum ||
2024 		    sp[i].s_secinfo.sc_nfsnum == AUTH_NONE)
2025 			return (1);
2026 	}
2027 
2028 	return (0);
2029 }
2030 
2031 /*
2032  * Check the access authority for the client and return the correct error.
2033  */
2034 nfsstat4
2035 call_checkauth4(struct compound_state *cs, struct svc_req *req)
2036 {
2037 	int	authres;
2038 
2039 	/*
2040 	 * First, check if the security flavor used in the request
2041 	 * are among the flavors set in the server namespace.
2042 	 */
2043 	if (!secinfo_match_or_authnone(cs)) {
2044 		*cs->statusp = NFS4ERR_WRONGSEC;
2045 		return (*cs->statusp);
2046 	}
2047 
2048 	authres = checkauth4(cs, req);
2049 
2050 	if (authres > 0) {
2051 		*cs->statusp = NFS4_OK;
2052 		if (! (cs->access & CS_ACCESS_LIMITED))
2053 			cs->access = CS_ACCESS_OK;
2054 	} else if (authres == 0) {
2055 		*cs->statusp = NFS4ERR_ACCESS;
2056 	} else if (authres == -2) {
2057 		*cs->statusp = NFS4ERR_WRONGSEC;
2058 	} else {
2059 		*cs->statusp = NFS4ERR_DELAY;
2060 	}
2061 	return (*cs->statusp);
2062 }
2063 
2064 /*
2065  * bitmap4_to_attrmask is called by getattr and readdir.
2066  * It sets up the vattr mask and determines whether vfsstat call is needed
2067  * based on the input bitmap.
2068  * Returns nfsv4 status.
2069  */
2070 static nfsstat4
2071 bitmap4_to_attrmask(bitmap4 breq, struct nfs4_svgetit_arg *sargp)
2072 {
2073 	int i;
2074 	uint_t	va_mask;
2075 	struct statvfs64 *sbp = sargp->sbp;
2076 
2077 	sargp->sbp = NULL;
2078 	sargp->flag = 0;
2079 	sargp->rdattr_error = NFS4_OK;
2080 	sargp->mntdfid_set = FALSE;
2081 	if (sargp->cs->vp)
2082 		sargp->xattr = get_fh4_flag(&sargp->cs->fh,
2083 		    FH4_ATTRDIR | FH4_NAMEDATTR);
2084 	else
2085 		sargp->xattr = 0;
2086 
2087 	/*
2088 	 * Set rdattr_error_req to true if return error per
2089 	 * failed entry rather than fail the readdir.
2090 	 */
2091 	if (breq & FATTR4_RDATTR_ERROR_MASK)
2092 		sargp->rdattr_error_req = 1;
2093 	else
2094 		sargp->rdattr_error_req = 0;
2095 
2096 	/*
2097 	 * generate the va_mask
2098 	 * Handle the easy cases first
2099 	 */
2100 	switch (breq) {
2101 	case NFS4_NTOV_ATTR_MASK:
2102 		sargp->vap->va_mask = NFS4_NTOV_ATTR_AT_MASK;
2103 		return (NFS4_OK);
2104 
2105 	case NFS4_FS_ATTR_MASK:
2106 		sargp->vap->va_mask = NFS4_FS_ATTR_AT_MASK;
2107 		sargp->sbp = sbp;
2108 		return (NFS4_OK);
2109 
2110 	case NFS4_NTOV_ATTR_CACHE_MASK:
2111 		sargp->vap->va_mask = NFS4_NTOV_ATTR_CACHE_AT_MASK;
2112 		return (NFS4_OK);
2113 
2114 	case FATTR4_LEASE_TIME_MASK:
2115 		sargp->vap->va_mask = 0;
2116 		return (NFS4_OK);
2117 
2118 	default:
2119 		va_mask = 0;
2120 		for (i = 0; i < nfs4_ntov_map_size; i++) {
2121 			if ((breq & nfs4_ntov_map[i].fbit) &&
2122 			    nfs4_ntov_map[i].vbit)
2123 				va_mask |= nfs4_ntov_map[i].vbit;
2124 		}
2125 
2126 		/*
2127 		 * Check is vfsstat is needed
2128 		 */
2129 		if (breq & NFS4_FS_ATTR_MASK)
2130 			sargp->sbp = sbp;
2131 
2132 		sargp->vap->va_mask = va_mask;
2133 		return (NFS4_OK);
2134 	}
2135 	/* NOTREACHED */
2136 }
2137 
2138 /*
2139  * bitmap4_get_sysattrs is called by getattr and readdir.
2140  * It calls both VOP_GETATTR and VFS_STATVFS calls to get the attrs.
2141  * Returns nfsv4 status.
2142  */
2143 static nfsstat4
2144 bitmap4_get_sysattrs(struct nfs4_svgetit_arg *sargp)
2145 {
2146 	int error;
2147 	struct compound_state *cs = sargp->cs;
2148 	vnode_t *vp = cs->vp;
2149 
2150 	if (sargp->sbp != NULL) {
2151 		if (error = VFS_STATVFS(vp->v_vfsp, sargp->sbp)) {
2152 			sargp->sbp = NULL;	/* to identify error */
2153 			return (puterrno4(error));
2154 		}
2155 	}
2156 
2157 	return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr));
2158 }
2159 
2160 static void
2161 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp)
2162 {
2163 	ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size,
2164 	    KM_SLEEP);
2165 	ntovp->attrcnt = 0;
2166 	ntovp->vfsstat = FALSE;
2167 }
2168 
2169 static void
2170 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp,
2171     struct nfs4_svgetit_arg *sargp)
2172 {
2173 	int i;
2174 	union nfs4_attr_u *na;
2175 	uint8_t *amap;
2176 
2177 	/*
2178 	 * XXX Should do the same checks for whether the bit is set
2179 	 */
2180 	for (i = 0, na = ntovp->na, amap = ntovp->amap;
2181 	    i < ntovp->attrcnt; i++, na++, amap++) {
2182 		(void) (*nfs4_ntov_map[*amap].sv_getit)(
2183 		    NFS4ATTR_FREEIT, sargp, na);
2184 	}
2185 	if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) {
2186 		/*
2187 		 * xdr_free for getattr will be done later
2188 		 */
2189 		for (i = 0, na = ntovp->na, amap = ntovp->amap;
2190 		    i < ntovp->attrcnt; i++, na++, amap++) {
2191 			xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na);
2192 		}
2193 	}
2194 	kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size);
2195 }
2196 
2197 /*
2198  * do_rfs4_op_getattr gets the system attrs and converts into fattr4.
2199  */
2200 static nfsstat4
2201 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp,
2202     struct nfs4_svgetit_arg *sargp)
2203 {
2204 	int error = 0;
2205 	int i, k;
2206 	struct nfs4_ntov_table ntov;
2207 	XDR xdr;
2208 	ulong_t xdr_size;
2209 	char *xdr_attrs;
2210 	nfsstat4 status = NFS4_OK;
2211 	nfsstat4 prev_rdattr_error = sargp->rdattr_error;
2212 	union nfs4_attr_u *na;
2213 	uint8_t *amap;
2214 
2215 	sargp->op = NFS4ATTR_GETIT;
2216 	sargp->flag = 0;
2217 
2218 	fattrp->attrmask = 0;
2219 	/* if no bits requested, then return empty fattr4 */
2220 	if (breq == 0) {
2221 		fattrp->attrlist4_len = 0;
2222 		fattrp->attrlist4 = NULL;
2223 		return (NFS4_OK);
2224 	}
2225 
2226 	/*
2227 	 * return NFS4ERR_INVAL when client requests write-only attrs
2228 	 */
2229 	if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK))
2230 		return (NFS4ERR_INVAL);
2231 
2232 	nfs4_ntov_table_init(&ntov);
2233 	na = ntov.na;
2234 	amap = ntov.amap;
2235 
2236 	/*
2237 	 * Now loop to get or verify the attrs
2238 	 */
2239 	for (i = 0; i < nfs4_ntov_map_size; i++) {
2240 		if (breq & nfs4_ntov_map[i].fbit) {
2241 			if ((*nfs4_ntov_map[i].sv_getit)(
2242 			    NFS4ATTR_SUPPORTED, sargp, NULL) == 0) {
2243 
2244 				error = (*nfs4_ntov_map[i].sv_getit)(
2245 				    NFS4ATTR_GETIT, sargp, na);
2246 
2247 				/*
2248 				 * Possible error values:
2249 				 * >0 if sv_getit failed to
2250 				 * get the attr; 0 if succeeded;
2251 				 * <0 if rdattr_error and the
2252 				 * attribute cannot be returned.
2253 				 */
2254 				if (error && !(sargp->rdattr_error_req))
2255 					goto done;
2256 				/*
2257 				 * If error then just for entry
2258 				 */
2259 				if (error == 0) {
2260 					fattrp->attrmask |=
2261 					    nfs4_ntov_map[i].fbit;
2262 					*amap++ =
2263 					    (uint8_t)nfs4_ntov_map[i].nval;
2264 					na++;
2265 					(ntov.attrcnt)++;
2266 				} else if ((error > 0) &&
2267 				    (sargp->rdattr_error == NFS4_OK)) {
2268 					sargp->rdattr_error = puterrno4(error);
2269 				}
2270 				error = 0;
2271 			}
2272 		}
2273 	}
2274 
2275 	/*
2276 	 * If rdattr_error was set after the return value for it was assigned,
2277 	 * update it.
2278 	 */
2279 	if (prev_rdattr_error != sargp->rdattr_error) {
2280 		na = ntov.na;
2281 		amap = ntov.amap;
2282 		for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2283 			k = *amap;
2284 			if (k < FATTR4_RDATTR_ERROR) {
2285 				continue;
2286 			}
2287 			if ((k == FATTR4_RDATTR_ERROR) &&
2288 			    ((*nfs4_ntov_map[k].sv_getit)(
2289 			    NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) {
2290 
2291 				(void) (*nfs4_ntov_map[k].sv_getit)(
2292 				    NFS4ATTR_GETIT, sargp, na);
2293 			}
2294 			break;
2295 		}
2296 	}
2297 
2298 	xdr_size = 0;
2299 	na = ntov.na;
2300 	amap = ntov.amap;
2301 	for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2302 		xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na);
2303 	}
2304 
2305 	fattrp->attrlist4_len = xdr_size;
2306 	if (xdr_size) {
2307 		/* freed by rfs4_op_getattr_free() */
2308 		fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP);
2309 
2310 		xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE);
2311 
2312 		na = ntov.na;
2313 		amap = ntov.amap;
2314 		for (i = 0; i < ntov.attrcnt; i++, na++, amap++) {
2315 			if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) {
2316 				DTRACE_PROBE1(nfss__e__getattr4_encfail,
2317 				    int, *amap);
2318 				status = NFS4ERR_SERVERFAULT;
2319 				break;
2320 			}
2321 		}
2322 		/* xdrmem_destroy(&xdrs); */	/* NO-OP */
2323 	} else {
2324 		fattrp->attrlist4 = NULL;
2325 	}
2326 done:
2327 
2328 	nfs4_ntov_table_free(&ntov, sargp);
2329 
2330 	if (error != 0)
2331 		status = puterrno4(error);
2332 
2333 	return (status);
2334 }
2335 
2336 /* ARGSUSED */
2337 static void
2338 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2339     struct compound_state *cs)
2340 {
2341 	GETATTR4args *args = &argop->nfs_argop4_u.opgetattr;
2342 	GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr;
2343 	struct nfs4_svgetit_arg sarg;
2344 	struct statvfs64 sb;
2345 	nfsstat4 status;
2346 
2347 	DTRACE_NFSV4_2(op__getattr__start, struct compound_state *, cs,
2348 	    GETATTR4args *, args);
2349 
2350 	if (cs->vp == NULL) {
2351 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2352 		goto out;
2353 	}
2354 
2355 	if (cs->access == CS_ACCESS_DENIED) {
2356 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2357 		goto out;
2358 	}
2359 
2360 	sarg.sbp = &sb;
2361 	sarg.cs = cs;
2362 	sarg.is_referral = B_FALSE;
2363 
2364 	status = bitmap4_to_attrmask(args->attr_request, &sarg);
2365 	if (status == NFS4_OK) {
2366 
2367 		status = bitmap4_get_sysattrs(&sarg);
2368 		if (status == NFS4_OK) {
2369 
2370 			/* Is this a referral? */
2371 			if (vn_is_nfs_reparse(cs->vp, cs->cr)) {
2372 				/* Older V4 Solaris client sees a link */
2373 				if (client_is_downrev(req))
2374 					sarg.vap->va_type = VLNK;
2375 				else
2376 					sarg.is_referral = B_TRUE;
2377 			}
2378 
2379 			status = do_rfs4_op_getattr(args->attr_request,
2380 			    &resp->obj_attributes, &sarg);
2381 		}
2382 	}
2383 	*cs->statusp = resp->status = status;
2384 out:
2385 	DTRACE_NFSV4_2(op__getattr__done, struct compound_state *, cs,
2386 	    GETATTR4res *, resp);
2387 }
2388 
2389 static void
2390 rfs4_op_getattr_free(nfs_resop4 *resop)
2391 {
2392 	GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr;
2393 
2394 	nfs4_fattr4_free(&resp->obj_attributes);
2395 }
2396 
2397 /* ARGSUSED */
2398 static void
2399 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2400     struct compound_state *cs)
2401 {
2402 	GETFH4res *resp = &resop->nfs_resop4_u.opgetfh;
2403 
2404 	DTRACE_NFSV4_1(op__getfh__start, struct compound_state *, cs);
2405 
2406 	if (cs->vp == NULL) {
2407 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2408 		goto out;
2409 	}
2410 	if (cs->access == CS_ACCESS_DENIED) {
2411 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2412 		goto out;
2413 	}
2414 
2415 	/* check for reparse point at the share point */
2416 	if (cs->exi->exi_moved || vn_is_nfs_reparse(cs->exi->exi_vp, cs->cr)) {
2417 		/* it's all bad */
2418 		cs->exi->exi_moved = 1;
2419 		*cs->statusp = resp->status = NFS4ERR_MOVED;
2420 		DTRACE_PROBE2(nfs4serv__func__referral__shared__moved,
2421 		    vnode_t *, cs->vp, char *, "rfs4_op_getfh");
2422 		return;
2423 	}
2424 
2425 	/* check for reparse point at vp */
2426 	if (vn_is_nfs_reparse(cs->vp, cs->cr) && !client_is_downrev(req)) {
2427 		/* it's not all bad */
2428 		*cs->statusp = resp->status = NFS4ERR_MOVED;
2429 		DTRACE_PROBE2(nfs4serv__func__referral__moved,
2430 		    vnode_t *, cs->vp, char *, "rfs4_op_getfh");
2431 		return;
2432 	}
2433 
2434 	resp->object.nfs_fh4_val =
2435 	    kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP);
2436 	nfs_fh4_copy(&cs->fh, &resp->object);
2437 	*cs->statusp = resp->status = NFS4_OK;
2438 out:
2439 	DTRACE_NFSV4_2(op__getfh__done, struct compound_state *, cs,
2440 	    GETFH4res *, resp);
2441 }
2442 
2443 static void
2444 rfs4_op_getfh_free(nfs_resop4 *resop)
2445 {
2446 	GETFH4res *resp = &resop->nfs_resop4_u.opgetfh;
2447 
2448 	if (resp->status == NFS4_OK &&
2449 	    resp->object.nfs_fh4_val != NULL) {
2450 		kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len);
2451 		resp->object.nfs_fh4_val = NULL;
2452 		resp->object.nfs_fh4_len = 0;
2453 	}
2454 }
2455 
2456 /*
2457  * illegal: args: void
2458  *	    res : status (NFS4ERR_OP_ILLEGAL)
2459  */
2460 /* ARGSUSED */
2461 static void
2462 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop,
2463     struct svc_req *req, struct compound_state *cs)
2464 {
2465 	ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal;
2466 
2467 	resop->resop = OP_ILLEGAL;
2468 	*cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL;
2469 }
2470 
2471 /*
2472  * link: args: SAVED_FH: file, CURRENT_FH: target directory
2473  *	 res: status. If success - CURRENT_FH unchanged, return change_info
2474  */
2475 /* ARGSUSED */
2476 static void
2477 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2478     struct compound_state *cs)
2479 {
2480 	LINK4args *args = &argop->nfs_argop4_u.oplink;
2481 	LINK4res *resp = &resop->nfs_resop4_u.oplink;
2482 	int error;
2483 	vnode_t *vp;
2484 	vnode_t *dvp;
2485 	struct vattr bdva, idva, adva;
2486 	char *nm;
2487 	uint_t  len;
2488 	struct sockaddr *ca;
2489 	char *name = NULL;
2490 	nfsstat4 status;
2491 
2492 	DTRACE_NFSV4_2(op__link__start, struct compound_state *, cs,
2493 	    LINK4args *, args);
2494 
2495 	/* SAVED_FH: source object */
2496 	vp = cs->saved_vp;
2497 	if (vp == NULL) {
2498 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2499 		goto out;
2500 	}
2501 
2502 	/* CURRENT_FH: target directory */
2503 	dvp = cs->vp;
2504 	if (dvp == NULL) {
2505 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2506 		goto out;
2507 	}
2508 
2509 	/*
2510 	 * If there is a non-shared filesystem mounted on this vnode,
2511 	 * do not allow to link any file in this directory.
2512 	 */
2513 	if (vn_ismntpt(dvp)) {
2514 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2515 		goto out;
2516 	}
2517 
2518 	if (cs->access == CS_ACCESS_DENIED) {
2519 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
2520 		goto out;
2521 	}
2522 
2523 	/* Check source object's type validity */
2524 	if (vp->v_type == VDIR) {
2525 		*cs->statusp = resp->status = NFS4ERR_ISDIR;
2526 		goto out;
2527 	}
2528 
2529 	/* Check target directory's type */
2530 	if (dvp->v_type != VDIR) {
2531 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2532 		goto out;
2533 	}
2534 
2535 	if (cs->saved_exi != cs->exi) {
2536 		*cs->statusp = resp->status = NFS4ERR_XDEV;
2537 		goto out;
2538 	}
2539 
2540 	status = utf8_dir_verify(&args->newname);
2541 	if (status != NFS4_OK) {
2542 		*cs->statusp = resp->status = status;
2543 		goto out;
2544 	}
2545 
2546 	nm = utf8_to_fn(&args->newname, &len, NULL);
2547 	if (nm == NULL) {
2548 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2549 		goto out;
2550 	}
2551 
2552 	if (len > MAXNAMELEN) {
2553 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
2554 		kmem_free(nm, len);
2555 		goto out;
2556 	}
2557 
2558 	if (rdonly4(req, cs)) {
2559 		*cs->statusp = resp->status = NFS4ERR_ROFS;
2560 		kmem_free(nm, len);
2561 		goto out;
2562 	}
2563 
2564 	/* Get "before" change value */
2565 	bdva.va_mask = AT_CTIME|AT_SEQ;
2566 	error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL);
2567 	if (error) {
2568 		*cs->statusp = resp->status = puterrno4(error);
2569 		kmem_free(nm, len);
2570 		goto out;
2571 	}
2572 
2573 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
2574 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
2575 	    MAXPATHLEN  + 1);
2576 
2577 	if (name == NULL) {
2578 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2579 		kmem_free(nm, len);
2580 		goto out;
2581 	}
2582 
2583 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime)
2584 
2585 	error = VOP_LINK(dvp, vp, name, cs->cr, NULL, 0);
2586 
2587 	if (nm != name)
2588 		kmem_free(name, MAXPATHLEN + 1);
2589 	kmem_free(nm, len);
2590 
2591 	/*
2592 	 * Get the initial "after" sequence number, if it fails, set to zero
2593 	 */
2594 	idva.va_mask = AT_SEQ;
2595 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL))
2596 		idva.va_seq = 0;
2597 
2598 	/*
2599 	 * Force modified data and metadata out to stable storage.
2600 	 */
2601 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL);
2602 	(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
2603 
2604 	if (error) {
2605 		*cs->statusp = resp->status = puterrno4(error);
2606 		goto out;
2607 	}
2608 
2609 	/*
2610 	 * Get "after" change value, if it fails, simply return the
2611 	 * before value.
2612 	 */
2613 	adva.va_mask = AT_CTIME|AT_SEQ;
2614 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) {
2615 		adva.va_ctime = bdva.va_ctime;
2616 		adva.va_seq = 0;
2617 	}
2618 
2619 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
2620 
2621 	/*
2622 	 * The cinfo.atomic = TRUE only if we have
2623 	 * non-zero va_seq's, and it has incremented by exactly one
2624 	 * during the VOP_LINK and it didn't change during the VOP_FSYNC.
2625 	 */
2626 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
2627 	    idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq)
2628 		resp->cinfo.atomic = TRUE;
2629 	else
2630 		resp->cinfo.atomic = FALSE;
2631 
2632 	*cs->statusp = resp->status = NFS4_OK;
2633 out:
2634 	DTRACE_NFSV4_2(op__link__done, struct compound_state *, cs,
2635 	    LINK4res *, resp);
2636 }
2637 
2638 /*
2639  * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work.
2640  */
2641 
2642 /* ARGSUSED */
2643 static nfsstat4
2644 do_rfs4_op_lookup(char *nm, struct svc_req *req, struct compound_state *cs)
2645 {
2646 	int error;
2647 	int different_export = 0;
2648 	vnode_t *vp, *pre_tvp = NULL, *oldvp = NULL;
2649 	struct exportinfo *exi = NULL, *pre_exi = NULL;
2650 	nfsstat4 stat;
2651 	fid_t fid;
2652 	int attrdir, dotdot, walk;
2653 	bool_t is_newvp = FALSE;
2654 
2655 	if (cs->vp->v_flag & V_XATTRDIR) {
2656 		attrdir = 1;
2657 		ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR));
2658 	} else {
2659 		attrdir = 0;
2660 		ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR));
2661 	}
2662 
2663 	dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0');
2664 
2665 	/*
2666 	 * If dotdotting, then need to check whether it's
2667 	 * above the root of a filesystem, or above an
2668 	 * export point.
2669 	 */
2670 	if (dotdot) {
2671 
2672 		/*
2673 		 * If dotdotting at the root of a filesystem, then
2674 		 * need to traverse back to the mounted-on filesystem
2675 		 * and do the dotdot lookup there.
2676 		 */
2677 		if (cs->vp->v_flag & VROOT) {
2678 
2679 			/*
2680 			 * If at the system root, then can
2681 			 * go up no further.
2682 			 */
2683 			if (VN_CMP(cs->vp, rootdir))
2684 				return (puterrno4(ENOENT));
2685 
2686 			/*
2687 			 * Traverse back to the mounted-on filesystem
2688 			 */
2689 			cs->vp = untraverse(cs->vp);
2690 
2691 			/*
2692 			 * Set the different_export flag so we remember
2693 			 * to pick up a new exportinfo entry for
2694 			 * this new filesystem.
2695 			 */
2696 			different_export = 1;
2697 		} else {
2698 
2699 			/*
2700 			 * If dotdotting above an export point then set
2701 			 * the different_export to get new export info.
2702 			 */
2703 			different_export = nfs_exported(cs->exi, cs->vp);
2704 		}
2705 	}
2706 
2707 	error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr,
2708 	    NULL, NULL, NULL);
2709 	if (error)
2710 		return (puterrno4(error));
2711 
2712 	/*
2713 	 * If the vnode is in a pseudo filesystem, check whether it is visible.
2714 	 *
2715 	 * XXX if the vnode is a symlink and it is not visible in
2716 	 * a pseudo filesystem, return ENOENT (not following symlink).
2717 	 * V4 client can not mount such symlink. This is a regression
2718 	 * from V2/V3.
2719 	 *
2720 	 * In the same exported filesystem, if the security flavor used
2721 	 * is not an explicitly shared flavor, limit the view to the visible
2722 	 * list entries only. This is not a WRONGSEC case because it's already
2723 	 * checked via PUTROOTFH/PUTPUBFH or PUTFH.
2724 	 */
2725 	if (!different_export &&
2726 	    (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) ||
2727 	    cs->access & CS_ACCESS_LIMITED)) {
2728 		if (! nfs_visible(cs->exi, vp, &different_export)) {
2729 			VN_RELE(vp);
2730 			return (puterrno4(ENOENT));
2731 		}
2732 	}
2733 
2734 	/*
2735 	 * If it's a mountpoint, then traverse it.
2736 	 */
2737 	if (vn_ismntpt(vp)) {
2738 		pre_exi = cs->exi;	/* save pre-traversed exportinfo */
2739 		pre_tvp = vp;		/* save pre-traversed vnode	*/
2740 
2741 		/*
2742 		 * hold pre_tvp to counteract rele by traverse.  We will
2743 		 * need pre_tvp below if checkexport4 fails
2744 		 */
2745 		VN_HOLD(pre_tvp);
2746 		if ((error = traverse(&vp)) != 0) {
2747 			VN_RELE(vp);
2748 			VN_RELE(pre_tvp);
2749 			return (puterrno4(error));
2750 		}
2751 		different_export = 1;
2752 	} else if (vp->v_vfsp != cs->vp->v_vfsp) {
2753 		/*
2754 		 * The vfsp comparison is to handle the case where
2755 		 * a LOFS mount is shared.  lo_lookup traverses mount points,
2756 		 * and NFS is unaware of local fs transistions because
2757 		 * v_vfsmountedhere isn't set.  For this special LOFS case,
2758 		 * the dir and the obj returned by lookup will have different
2759 		 * vfs ptrs.
2760 		 */
2761 		different_export = 1;
2762 	}
2763 
2764 	if (different_export) {
2765 
2766 		bzero(&fid, sizeof (fid));
2767 		fid.fid_len = MAXFIDSZ;
2768 		error = vop_fid_pseudo(vp, &fid);
2769 		if (error) {
2770 			VN_RELE(vp);
2771 			if (pre_tvp)
2772 				VN_RELE(pre_tvp);
2773 			return (puterrno4(error));
2774 		}
2775 
2776 		if (dotdot)
2777 			exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE);
2778 		else
2779 			exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
2780 
2781 		if (exi == NULL) {
2782 			if (pre_tvp) {
2783 				/*
2784 				 * If this vnode is a mounted-on vnode,
2785 				 * but the mounted-on file system is not
2786 				 * exported, send back the filehandle for
2787 				 * the mounted-on vnode, not the root of
2788 				 * the mounted-on file system.
2789 				 */
2790 				VN_RELE(vp);
2791 				vp = pre_tvp;
2792 				exi = pre_exi;
2793 			} else {
2794 				VN_RELE(vp);
2795 				return (puterrno4(EACCES));
2796 			}
2797 		} else if (pre_tvp) {
2798 			/* we're done with pre_tvp now. release extra hold */
2799 			VN_RELE(pre_tvp);
2800 		}
2801 
2802 		cs->exi = exi;
2803 
2804 		/*
2805 		 * Now we do a checkauth4. The reason is that
2806 		 * this client/user may not have access to the new
2807 		 * exported file system, and if he does,
2808 		 * the client/user may be mapped to a different uid.
2809 		 *
2810 		 * We start with a new cr, because the checkauth4 done
2811 		 * in the PUT*FH operation over wrote the cred's uid,
2812 		 * gid, etc, and we want the real thing before calling
2813 		 * checkauth4()
2814 		 */
2815 		crfree(cs->cr);
2816 		cs->cr = crdup(cs->basecr);
2817 
2818 		oldvp = cs->vp;
2819 		cs->vp = vp;
2820 		is_newvp = TRUE;
2821 
2822 		stat = call_checkauth4(cs, req);
2823 		if (stat != NFS4_OK) {
2824 			VN_RELE(cs->vp);
2825 			cs->vp = oldvp;
2826 			return (stat);
2827 		}
2828 	}
2829 
2830 	/*
2831 	 * After various NFS checks, do a label check on the path
2832 	 * component. The label on this path should either be the
2833 	 * global zone's label or a zone's label. We are only
2834 	 * interested in the zone's label because exported files
2835 	 * in global zone is accessible (though read-only) to
2836 	 * clients. The exportability/visibility check is already
2837 	 * done before reaching this code.
2838 	 */
2839 	if (is_system_labeled()) {
2840 		bslabel_t *clabel;
2841 
2842 		ASSERT(req->rq_label != NULL);
2843 		clabel = req->rq_label;
2844 		DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *,
2845 		    "got client label from request(1)", struct svc_req *, req);
2846 
2847 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
2848 			if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK,
2849 			    cs->exi)) {
2850 				error = EACCES;
2851 				goto err_out;
2852 			}
2853 		} else {
2854 			/*
2855 			 * We grant access to admin_low label clients
2856 			 * only if the client is trusted, i.e. also
2857 			 * running Solaris Trusted Extension.
2858 			 */
2859 			struct sockaddr	*ca;
2860 			int		addr_type;
2861 			void		*ipaddr;
2862 			tsol_tpc_t	*tp;
2863 
2864 			ca = (struct sockaddr *)svc_getrpccaller(
2865 			    req->rq_xprt)->buf;
2866 			if (ca->sa_family == AF_INET) {
2867 				addr_type = IPV4_VERSION;
2868 				ipaddr = &((struct sockaddr_in *)ca)->sin_addr;
2869 			} else if (ca->sa_family == AF_INET6) {
2870 				addr_type = IPV6_VERSION;
2871 				ipaddr = &((struct sockaddr_in6 *)
2872 				    ca)->sin6_addr;
2873 			}
2874 			tp = find_tpc(ipaddr, addr_type, B_FALSE);
2875 			if (tp == NULL || tp->tpc_tp.tp_doi !=
2876 			    l_admin_low->tsl_doi || tp->tpc_tp.host_type !=
2877 			    SUN_CIPSO) {
2878 				if (tp != NULL)
2879 					TPC_RELE(tp);
2880 				error = EACCES;
2881 				goto err_out;
2882 			}
2883 			TPC_RELE(tp);
2884 		}
2885 	}
2886 
2887 	error = makefh4(&cs->fh, vp, cs->exi);
2888 
2889 err_out:
2890 	if (error) {
2891 		if (is_newvp) {
2892 			VN_RELE(cs->vp);
2893 			cs->vp = oldvp;
2894 		} else
2895 			VN_RELE(vp);
2896 		return (puterrno4(error));
2897 	}
2898 
2899 	if (!is_newvp) {
2900 		if (cs->vp)
2901 			VN_RELE(cs->vp);
2902 		cs->vp = vp;
2903 	} else if (oldvp)
2904 		VN_RELE(oldvp);
2905 
2906 	/*
2907 	 * if did lookup on attrdir and didn't lookup .., set named
2908 	 * attr fh flag
2909 	 */
2910 	if (attrdir && ! dotdot)
2911 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
2912 
2913 	/* Assume false for now, open proc will set this */
2914 	cs->mandlock = FALSE;
2915 
2916 	return (NFS4_OK);
2917 }
2918 
2919 /* ARGSUSED */
2920 static void
2921 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
2922     struct compound_state *cs)
2923 {
2924 	LOOKUP4args *args = &argop->nfs_argop4_u.oplookup;
2925 	LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup;
2926 	char *nm;
2927 	uint_t len;
2928 	struct sockaddr *ca;
2929 	char *name = NULL;
2930 	nfsstat4 status;
2931 
2932 	DTRACE_NFSV4_2(op__lookup__start, struct compound_state *, cs,
2933 	    LOOKUP4args *, args);
2934 
2935 	if (cs->vp == NULL) {
2936 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
2937 		goto out;
2938 	}
2939 
2940 	if (cs->vp->v_type == VLNK) {
2941 		*cs->statusp = resp->status = NFS4ERR_SYMLINK;
2942 		goto out;
2943 	}
2944 
2945 	if (cs->vp->v_type != VDIR) {
2946 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
2947 		goto out;
2948 	}
2949 
2950 	status = utf8_dir_verify(&args->objname);
2951 	if (status != NFS4_OK) {
2952 		*cs->statusp = resp->status = status;
2953 		goto out;
2954 	}
2955 
2956 	nm = utf8_to_str(&args->objname, &len, NULL);
2957 	if (nm == NULL) {
2958 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2959 		goto out;
2960 	}
2961 
2962 	if (len > MAXNAMELEN) {
2963 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
2964 		kmem_free(nm, len);
2965 		goto out;
2966 	}
2967 
2968 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
2969 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
2970 	    MAXPATHLEN  + 1);
2971 
2972 	if (name == NULL) {
2973 		*cs->statusp = resp->status = NFS4ERR_INVAL;
2974 		kmem_free(nm, len);
2975 		goto out;
2976 	}
2977 
2978 	*cs->statusp = resp->status = do_rfs4_op_lookup(name, req, cs);
2979 
2980 	if (name != nm)
2981 		kmem_free(name, MAXPATHLEN + 1);
2982 	kmem_free(nm, len);
2983 
2984 out:
2985 	DTRACE_NFSV4_2(op__lookup__done, struct compound_state *, cs,
2986 	    LOOKUP4res *, resp);
2987 }
2988 
2989 /* ARGSUSED */
2990 static void
2991 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
2992     struct compound_state *cs)
2993 {
2994 	LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp;
2995 
2996 	DTRACE_NFSV4_1(op__lookupp__start, struct compound_state *, cs);
2997 
2998 	if (cs->vp == NULL) {
2999 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3000 		goto out;
3001 	}
3002 
3003 	if (cs->vp->v_type != VDIR) {
3004 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
3005 		goto out;
3006 	}
3007 
3008 	*cs->statusp = resp->status = do_rfs4_op_lookup("..", req, cs);
3009 
3010 	/*
3011 	 * From NFSV4 Specification, LOOKUPP should not check for
3012 	 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead.
3013 	 */
3014 	if (resp->status == NFS4ERR_WRONGSEC) {
3015 		*cs->statusp = resp->status = NFS4_OK;
3016 	}
3017 
3018 out:
3019 	DTRACE_NFSV4_2(op__lookupp__done, struct compound_state *, cs,
3020 	    LOOKUPP4res *, resp);
3021 }
3022 
3023 
3024 /*ARGSUSED2*/
3025 static void
3026 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3027     struct compound_state *cs)
3028 {
3029 	OPENATTR4args	*args = &argop->nfs_argop4_u.opopenattr;
3030 	OPENATTR4res	*resp = &resop->nfs_resop4_u.opopenattr;
3031 	vnode_t		*avp = NULL;
3032 	int		lookup_flags = LOOKUP_XATTR, error;
3033 	int		exp_ro = 0;
3034 
3035 	DTRACE_NFSV4_2(op__openattr__start, struct compound_state *, cs,
3036 	    OPENATTR4args *, args);
3037 
3038 	if (cs->vp == NULL) {
3039 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3040 		goto out;
3041 	}
3042 
3043 	if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0 &&
3044 	    !vfs_has_feature(cs->vp->v_vfsp, VFSFT_SYSATTR_VIEWS)) {
3045 		*cs->statusp = resp->status = puterrno4(ENOTSUP);
3046 		goto out;
3047 	}
3048 
3049 	/*
3050 	 * If file system supports passing ACE mask to VOP_ACCESS then
3051 	 * check for ACE_READ_NAMED_ATTRS, otherwise do legacy checks
3052 	 */
3053 
3054 	if (vfs_has_feature(cs->vp->v_vfsp, VFSFT_ACEMASKONACCESS))
3055 		error = VOP_ACCESS(cs->vp, ACE_READ_NAMED_ATTRS,
3056 		    V_ACE_MASK, cs->cr, NULL);
3057 	else
3058 		error = ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr, NULL) != 0) &&
3059 		    (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr, NULL) != 0) &&
3060 		    (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr, NULL) != 0));
3061 
3062 	if (error) {
3063 		*cs->statusp = resp->status = puterrno4(EACCES);
3064 		goto out;
3065 	}
3066 
3067 	/*
3068 	 * The CREATE_XATTR_DIR VOP flag cannot be specified if
3069 	 * the file system is exported read-only -- regardless of
3070 	 * createdir flag.  Otherwise the attrdir would be created
3071 	 * (assuming server fs isn't mounted readonly locally).  If
3072 	 * VOP_LOOKUP returns ENOENT in this case, the error will
3073 	 * be translated into EROFS.  ENOSYS is mapped to ENOTSUP
3074 	 * because specfs has no VOP_LOOKUP op, so the macro would
3075 	 * return ENOSYS.  EINVAL is returned by all (current)
3076 	 * Solaris file system implementations when any of their
3077 	 * restrictions are violated (xattr(dir) can't have xattrdir).
3078 	 * Returning NOTSUPP is more appropriate in this case
3079 	 * because the object will never be able to have an attrdir.
3080 	 */
3081 	if (args->createdir && ! (exp_ro = rdonly4(req, cs)))
3082 		lookup_flags |= CREATE_XATTR_DIR;
3083 
3084 	error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr,
3085 	    NULL, NULL, NULL);
3086 
3087 	if (error) {
3088 		if (error == ENOENT && args->createdir && exp_ro)
3089 			*cs->statusp = resp->status = puterrno4(EROFS);
3090 		else if (error == EINVAL || error == ENOSYS)
3091 			*cs->statusp = resp->status = puterrno4(ENOTSUP);
3092 		else
3093 			*cs->statusp = resp->status = puterrno4(error);
3094 		goto out;
3095 	}
3096 
3097 	ASSERT(avp->v_flag & V_XATTRDIR);
3098 
3099 	error = makefh4(&cs->fh, avp, cs->exi);
3100 
3101 	if (error) {
3102 		VN_RELE(avp);
3103 		*cs->statusp = resp->status = puterrno4(error);
3104 		goto out;
3105 	}
3106 
3107 	VN_RELE(cs->vp);
3108 	cs->vp = avp;
3109 
3110 	/*
3111 	 * There is no requirement for an attrdir fh flag
3112 	 * because the attrdir has a vnode flag to distinguish
3113 	 * it from regular (non-xattr) directories.  The
3114 	 * FH4_ATTRDIR flag is set for future sanity checks.
3115 	 */
3116 	set_fh4_flag(&cs->fh, FH4_ATTRDIR);
3117 	*cs->statusp = resp->status = NFS4_OK;
3118 
3119 out:
3120 	DTRACE_NFSV4_2(op__openattr__done, struct compound_state *, cs,
3121 	    OPENATTR4res *, resp);
3122 }
3123 
3124 static int
3125 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred,
3126     caller_context_t *ct)
3127 {
3128 	int error;
3129 	int i;
3130 	clock_t delaytime;
3131 
3132 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
3133 
3134 	/*
3135 	 * Don't block on mandatory locks. If this routine returns
3136 	 * EAGAIN, the caller should return NFS4ERR_LOCKED.
3137 	 */
3138 	uio->uio_fmode = FNONBLOCK;
3139 
3140 	for (i = 0; i < rfs4_maxlock_tries; i++) {
3141 
3142 
3143 		if (direction == FREAD) {
3144 			(void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, ct);
3145 			error = VOP_READ(vp, uio, ioflag, cred, ct);
3146 			VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, ct);
3147 		} else {
3148 			(void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, ct);
3149 			error = VOP_WRITE(vp, uio, ioflag, cred, ct);
3150 			VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, ct);
3151 		}
3152 
3153 		if (error != EAGAIN)
3154 			break;
3155 
3156 		if (i < rfs4_maxlock_tries - 1) {
3157 			delay(delaytime);
3158 			delaytime *= 2;
3159 		}
3160 	}
3161 
3162 	return (error);
3163 }
3164 
3165 /* ARGSUSED */
3166 static void
3167 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3168     struct compound_state *cs)
3169 {
3170 	READ4args *args = &argop->nfs_argop4_u.opread;
3171 	READ4res *resp = &resop->nfs_resop4_u.opread;
3172 	int error;
3173 	int verror;
3174 	vnode_t *vp;
3175 	struct vattr va;
3176 	struct iovec iov, *iovp = NULL;
3177 	int iovcnt;
3178 	struct uio uio;
3179 	u_offset_t offset;
3180 	bool_t *deleg = &cs->deleg;
3181 	nfsstat4 stat;
3182 	int in_crit = 0;
3183 	mblk_t *mp = NULL;
3184 	int alloc_err = 0;
3185 	int rdma_used = 0;
3186 	int loaned_buffers;
3187 	caller_context_t ct;
3188 	struct uio *uiop;
3189 
3190 	DTRACE_NFSV4_2(op__read__start, struct compound_state *, cs,
3191 	    READ4args, args);
3192 
3193 	vp = cs->vp;
3194 	if (vp == NULL) {
3195 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3196 		goto out;
3197 	}
3198 	if (cs->access == CS_ACCESS_DENIED) {
3199 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3200 		goto out;
3201 	}
3202 
3203 	if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE,
3204 	    deleg, TRUE, &ct)) != NFS4_OK) {
3205 		*cs->statusp = resp->status = stat;
3206 		goto out;
3207 	}
3208 
3209 	/*
3210 	 * Enter the critical region before calling VOP_RWLOCK
3211 	 * to avoid a deadlock with write requests.
3212 	 */
3213 	if (nbl_need_check(vp)) {
3214 		nbl_start_crit(vp, RW_READER);
3215 		in_crit = 1;
3216 		if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0,
3217 		    &ct)) {
3218 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
3219 			goto out;
3220 		}
3221 	}
3222 
3223 	if (args->wlist) {
3224 		if (args->count > clist_len(args->wlist)) {
3225 			*cs->statusp = resp->status = NFS4ERR_INVAL;
3226 			goto out;
3227 		}
3228 		rdma_used = 1;
3229 	}
3230 
3231 	/* use loaned buffers for TCP */
3232 	loaned_buffers = (nfs_loaned_buffers && !rdma_used) ? 1 : 0;
3233 
3234 	va.va_mask = AT_MODE|AT_SIZE|AT_UID;
3235 	verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct);
3236 
3237 	/*
3238 	 * If we can't get the attributes, then we can't do the
3239 	 * right access checking.  So, we'll fail the request.
3240 	 */
3241 	if (verror) {
3242 		*cs->statusp = resp->status = puterrno4(verror);
3243 		goto out;
3244 	}
3245 
3246 	if (vp->v_type != VREG) {
3247 		*cs->statusp = resp->status =
3248 		    ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
3249 		goto out;
3250 	}
3251 
3252 	if (crgetuid(cs->cr) != va.va_uid &&
3253 	    (error = VOP_ACCESS(vp, VREAD, 0, cs->cr, &ct)) &&
3254 	    (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr, &ct))) {
3255 		*cs->statusp = resp->status = puterrno4(error);
3256 		goto out;
3257 	}
3258 
3259 	if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */
3260 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3261 		goto out;
3262 	}
3263 
3264 	offset = args->offset;
3265 	if (offset >= va.va_size) {
3266 		*cs->statusp = resp->status = NFS4_OK;
3267 		resp->eof = TRUE;
3268 		resp->data_len = 0;
3269 		resp->data_val = NULL;
3270 		resp->mblk = NULL;
3271 		/* RDMA */
3272 		resp->wlist = args->wlist;
3273 		resp->wlist_len = resp->data_len;
3274 		*cs->statusp = resp->status = NFS4_OK;
3275 		if (resp->wlist)
3276 			clist_zero_len(resp->wlist);
3277 		goto out;
3278 	}
3279 
3280 	if (args->count == 0) {
3281 		*cs->statusp = resp->status = NFS4_OK;
3282 		resp->eof = FALSE;
3283 		resp->data_len = 0;
3284 		resp->data_val = NULL;
3285 		resp->mblk = NULL;
3286 		/* RDMA */
3287 		resp->wlist = args->wlist;
3288 		resp->wlist_len = resp->data_len;
3289 		if (resp->wlist)
3290 			clist_zero_len(resp->wlist);
3291 		goto out;
3292 	}
3293 
3294 	/*
3295 	 * Do not allocate memory more than maximum allowed
3296 	 * transfer size
3297 	 */
3298 	if (args->count > rfs4_tsize(req))
3299 		args->count = rfs4_tsize(req);
3300 
3301 	if (loaned_buffers) {
3302 		uiop = (uio_t *)rfs_setup_xuio(vp);
3303 		ASSERT(uiop != NULL);
3304 		uiop->uio_segflg = UIO_SYSSPACE;
3305 		uiop->uio_loffset = args->offset;
3306 		uiop->uio_resid = args->count;
3307 
3308 		/* Jump to do the read if successful */
3309 		if (!VOP_REQZCBUF(vp, UIO_READ, (xuio_t *)uiop, cs->cr, &ct)) {
3310 			/*
3311 			 * Need to hold the vnode until after VOP_RETZCBUF()
3312 			 * is called.
3313 			 */
3314 			VN_HOLD(vp);
3315 			goto doio_read;
3316 		}
3317 
3318 		DTRACE_PROBE2(nfss__i__reqzcbuf_failed, int,
3319 		    uiop->uio_loffset, int, uiop->uio_resid);
3320 
3321 		uiop->uio_extflg = 0;
3322 
3323 		/* failure to setup for zero copy */
3324 		rfs_free_xuio((void *)uiop);
3325 		loaned_buffers = 0;
3326 	}
3327 
3328 	/*
3329 	 * If returning data via RDMA Write, then grab the chunk list. If we
3330 	 * aren't returning READ data w/RDMA_WRITE, then grab a mblk.
3331 	 */
3332 	if (rdma_used) {
3333 		mp = NULL;
3334 		(void) rdma_get_wchunk(req, &iov, args->wlist);
3335 		uio.uio_iov = &iov;
3336 		uio.uio_iovcnt = 1;
3337 	} else {
3338 		/*
3339 		 * mp will contain the data to be sent out in the read reply.
3340 		 * It will be freed after the reply has been sent.
3341 		 */
3342 		mp = rfs_read_alloc(args->count, &iovp, &iovcnt);
3343 		ASSERT(mp != NULL);
3344 		ASSERT(alloc_err == 0);
3345 		uio.uio_iov = iovp;
3346 		uio.uio_iovcnt = iovcnt;
3347 	}
3348 
3349 	uio.uio_segflg = UIO_SYSSPACE;
3350 	uio.uio_extflg = UIO_COPY_CACHED;
3351 	uio.uio_loffset = args->offset;
3352 	uio.uio_resid = args->count;
3353 	uiop = &uio;
3354 
3355 doio_read:
3356 	error = do_io(FREAD, vp, uiop, 0, cs->cr, &ct);
3357 
3358 	va.va_mask = AT_SIZE;
3359 	verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct);
3360 
3361 	if (error) {
3362 		if (mp)
3363 			freemsg(mp);
3364 		*cs->statusp = resp->status = puterrno4(error);
3365 		goto out;
3366 	}
3367 
3368 	/* make mblk using zc buffers */
3369 	if (loaned_buffers) {
3370 		mp = uio_to_mblk(uiop);
3371 		ASSERT(mp != NULL);
3372 	}
3373 
3374 	*cs->statusp = resp->status = NFS4_OK;
3375 
3376 	ASSERT(uiop->uio_resid >= 0);
3377 	resp->data_len = args->count - uiop->uio_resid;
3378 	if (mp) {
3379 		resp->data_val = (char *)mp->b_datap->db_base;
3380 		rfs_rndup_mblks(mp, resp->data_len, loaned_buffers);
3381 	} else {
3382 		resp->data_val = (caddr_t)iov.iov_base;
3383 	}
3384 
3385 	resp->mblk = mp;
3386 
3387 	if (!verror && offset + resp->data_len == va.va_size)
3388 		resp->eof = TRUE;
3389 	else
3390 		resp->eof = FALSE;
3391 
3392 	if (rdma_used) {
3393 		if (!rdma_setup_read_data4(args, resp)) {
3394 			*cs->statusp = resp->status = NFS4ERR_INVAL;
3395 		}
3396 	} else {
3397 		resp->wlist = NULL;
3398 	}
3399 
3400 out:
3401 	if (in_crit)
3402 		nbl_end_crit(vp);
3403 
3404 	if (iovp != NULL)
3405 		kmem_free(iovp, iovcnt * sizeof (struct iovec));
3406 
3407 	DTRACE_NFSV4_2(op__read__done, struct compound_state *, cs,
3408 	    READ4res *, resp);
3409 }
3410 
3411 static void
3412 rfs4_op_read_free(nfs_resop4 *resop)
3413 {
3414 	READ4res	*resp = &resop->nfs_resop4_u.opread;
3415 
3416 	if (resp->status == NFS4_OK && resp->mblk != NULL) {
3417 		freemsg(resp->mblk);
3418 		resp->mblk = NULL;
3419 		resp->data_val = NULL;
3420 		resp->data_len = 0;
3421 	}
3422 }
3423 
3424 static void
3425 rfs4_op_readdir_free(nfs_resop4 * resop)
3426 {
3427 	READDIR4res    *resp = &resop->nfs_resop4_u.opreaddir;
3428 
3429 	if (resp->status == NFS4_OK && resp->mblk != NULL) {
3430 		freeb(resp->mblk);
3431 		resp->mblk = NULL;
3432 		resp->data_len = 0;
3433 	}
3434 }
3435 
3436 
3437 /* ARGSUSED */
3438 static void
3439 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
3440     struct compound_state *cs)
3441 {
3442 	PUTPUBFH4res	*resp = &resop->nfs_resop4_u.opputpubfh;
3443 	int		error;
3444 	vnode_t		*vp;
3445 	struct exportinfo *exi, *sav_exi;
3446 	nfs_fh4_fmt_t	*fh_fmtp;
3447 
3448 	DTRACE_NFSV4_1(op__putpubfh__start, struct compound_state *, cs);
3449 
3450 	if (cs->vp) {
3451 		VN_RELE(cs->vp);
3452 		cs->vp = NULL;
3453 	}
3454 
3455 	if (cs->cr)
3456 		crfree(cs->cr);
3457 
3458 	cs->cr = crdup(cs->basecr);
3459 
3460 	vp = exi_public->exi_vp;
3461 	if (vp == NULL) {
3462 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
3463 		goto out;
3464 	}
3465 
3466 	error = makefh4(&cs->fh, vp, exi_public);
3467 	if (error != 0) {
3468 		*cs->statusp = resp->status = puterrno4(error);
3469 		goto out;
3470 	}
3471 	sav_exi = cs->exi;
3472 	if (exi_public == exi_root) {
3473 		/*
3474 		 * No filesystem is actually shared public, so we default
3475 		 * to exi_root. In this case, we must check whether root
3476 		 * is exported.
3477 		 */
3478 		fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val;
3479 
3480 		/*
3481 		 * if root filesystem is exported, the exportinfo struct that we
3482 		 * should use is what checkexport4 returns, because root_exi is
3483 		 * actually a mostly empty struct.
3484 		 */
3485 		exi = checkexport4(&fh_fmtp->fh4_fsid,
3486 		    (fid_t *)&fh_fmtp->fh4_xlen, NULL);
3487 		cs->exi = ((exi != NULL) ? exi : exi_public);
3488 	} else {
3489 		/*
3490 		 * it's a properly shared filesystem
3491 		 */
3492 		cs->exi = exi_public;
3493 	}
3494 
3495 	if (is_system_labeled()) {
3496 		bslabel_t *clabel;
3497 
3498 		ASSERT(req->rq_label != NULL);
3499 		clabel = req->rq_label;
3500 		DTRACE_PROBE2(tx__rfs4__log__info__opputpubfh__clabel, char *,
3501 		    "got client label from request(1)",
3502 		    struct svc_req *, req);
3503 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
3504 			if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK,
3505 			    cs->exi)) {
3506 				*cs->statusp = resp->status =
3507 				    NFS4ERR_SERVERFAULT;
3508 				goto out;
3509 			}
3510 		}
3511 	}
3512 
3513 	VN_HOLD(vp);
3514 	cs->vp = vp;
3515 
3516 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3517 		VN_RELE(cs->vp);
3518 		cs->vp = NULL;
3519 		cs->exi = sav_exi;
3520 		goto out;
3521 	}
3522 
3523 	*cs->statusp = resp->status = NFS4_OK;
3524 out:
3525 	DTRACE_NFSV4_2(op__putpubfh__done, struct compound_state *, cs,
3526 	    PUTPUBFH4res *, resp);
3527 }
3528 
3529 /*
3530  * XXX - issue with put*fh operations. Suppose /export/home is exported.
3531  * Suppose an NFS client goes to mount /export/home/joe. If /export, home,
3532  * or joe have restrictive search permissions, then we shouldn't let
3533  * the client get a file handle. This is easy to enforce. However, we
3534  * don't know what security flavor should be used until we resolve the
3535  * path name. Another complication is uid mapping. If root is
3536  * the user, then it will be mapped to the anonymous user by default,
3537  * but we won't know that till we've resolved the path name. And we won't
3538  * know what the anonymous user is.
3539  * Luckily, SECINFO is specified to take a full filename.
3540  * So what we will have to in rfs4_op_lookup is check that flavor of
3541  * the target object matches that of the request, and if root was the
3542  * caller, check for the root= and anon= options, and if necessary,
3543  * repeat the lookup using the right cred_t. But that's not done yet.
3544  */
3545 /* ARGSUSED */
3546 static void
3547 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3548     struct compound_state *cs)
3549 {
3550 	PUTFH4args *args = &argop->nfs_argop4_u.opputfh;
3551 	PUTFH4res *resp = &resop->nfs_resop4_u.opputfh;
3552 	nfs_fh4_fmt_t *fh_fmtp;
3553 
3554 	DTRACE_NFSV4_2(op__putfh__start, struct compound_state *, cs,
3555 	    PUTFH4args *, args);
3556 
3557 	if (cs->vp) {
3558 		VN_RELE(cs->vp);
3559 		cs->vp = NULL;
3560 	}
3561 
3562 	if (cs->cr) {
3563 		crfree(cs->cr);
3564 		cs->cr = NULL;
3565 	}
3566 
3567 	if (args->object.nfs_fh4_len < NFS_FH4_LEN) {
3568 		*cs->statusp = resp->status = NFS4ERR_BADHANDLE;
3569 		goto out;
3570 	}
3571 
3572 	fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val;
3573 	cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen,
3574 	    NULL);
3575 
3576 	if (cs->exi == NULL) {
3577 		*cs->statusp = resp->status = NFS4ERR_STALE;
3578 		goto out;
3579 	}
3580 
3581 	cs->cr = crdup(cs->basecr);
3582 
3583 	ASSERT(cs->cr != NULL);
3584 
3585 	if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) {
3586 		*cs->statusp = resp->status;
3587 		goto out;
3588 	}
3589 
3590 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3591 		VN_RELE(cs->vp);
3592 		cs->vp = NULL;
3593 		goto out;
3594 	}
3595 
3596 	nfs_fh4_copy(&args->object, &cs->fh);
3597 	*cs->statusp = resp->status = NFS4_OK;
3598 	cs->deleg = FALSE;
3599 
3600 out:
3601 	DTRACE_NFSV4_2(op__putfh__done, struct compound_state *, cs,
3602 	    PUTFH4res *, resp);
3603 }
3604 
3605 /* ARGSUSED */
3606 static void
3607 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3608     struct compound_state *cs)
3609 {
3610 	PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh;
3611 	int error;
3612 	fid_t fid;
3613 	struct exportinfo *exi, *sav_exi;
3614 
3615 	DTRACE_NFSV4_1(op__putrootfh__start, struct compound_state *, cs);
3616 
3617 	if (cs->vp) {
3618 		VN_RELE(cs->vp);
3619 		cs->vp = NULL;
3620 	}
3621 
3622 	if (cs->cr)
3623 		crfree(cs->cr);
3624 
3625 	cs->cr = crdup(cs->basecr);
3626 
3627 	/*
3628 	 * Using rootdir, the system root vnode,
3629 	 * get its fid.
3630 	 */
3631 	bzero(&fid, sizeof (fid));
3632 	fid.fid_len = MAXFIDSZ;
3633 	error = vop_fid_pseudo(rootdir, &fid);
3634 	if (error != 0) {
3635 		*cs->statusp = resp->status = puterrno4(error);
3636 		goto out;
3637 	}
3638 
3639 	/*
3640 	 * Then use the root fsid & fid it to find out if it's exported
3641 	 *
3642 	 * If the server root isn't exported directly, then
3643 	 * it should at least be a pseudo export based on
3644 	 * one or more exports further down in the server's
3645 	 * file tree.
3646 	 */
3647 	exi = checkexport4(&rootdir->v_vfsp->vfs_fsid, &fid, NULL);
3648 	if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) {
3649 		NFS4_DEBUG(rfs4_debug,
3650 		    (CE_WARN, "rfs4_op_putrootfh: export check failure"));
3651 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
3652 		goto out;
3653 	}
3654 
3655 	/*
3656 	 * Now make a filehandle based on the root
3657 	 * export and root vnode.
3658 	 */
3659 	error = makefh4(&cs->fh, rootdir, exi);
3660 	if (error != 0) {
3661 		*cs->statusp = resp->status = puterrno4(error);
3662 		goto out;
3663 	}
3664 
3665 	sav_exi = cs->exi;
3666 	cs->exi = exi;
3667 
3668 	VN_HOLD(rootdir);
3669 	cs->vp = rootdir;
3670 
3671 	if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) {
3672 		VN_RELE(rootdir);
3673 		cs->vp = NULL;
3674 		cs->exi = sav_exi;
3675 		goto out;
3676 	}
3677 
3678 	*cs->statusp = resp->status = NFS4_OK;
3679 	cs->deleg = FALSE;
3680 out:
3681 	DTRACE_NFSV4_2(op__putrootfh__done, struct compound_state *, cs,
3682 	    PUTROOTFH4res *, resp);
3683 }
3684 
3685 /*
3686  * set_rdattr_params sets up the variables used to manage what information
3687  * to get for each directory entry.
3688  */
3689 static nfsstat4
3690 set_rdattr_params(struct nfs4_svgetit_arg *sargp,
3691     bitmap4 attrs, bool_t *need_to_lookup)
3692 {
3693 	uint_t	va_mask;
3694 	nfsstat4 status;
3695 	bitmap4 objbits;
3696 
3697 	status = bitmap4_to_attrmask(attrs, sargp);
3698 	if (status != NFS4_OK) {
3699 		/*
3700 		 * could not even figure attr mask
3701 		 */
3702 		return (status);
3703 	}
3704 	va_mask = sargp->vap->va_mask;
3705 
3706 	/*
3707 	 * dirent's d_ino is always correct value for mounted_on_fileid.
3708 	 * mntdfid_set is set once here, but mounted_on_fileid is
3709 	 * set in main dirent processing loop for each dirent.
3710 	 * The mntdfid_set is a simple optimization that lets the
3711 	 * server attr code avoid work when caller is readdir.
3712 	 */
3713 	sargp->mntdfid_set = TRUE;
3714 
3715 	/*
3716 	 * Lookup entry only if client asked for any of the following:
3717 	 * a) vattr attrs
3718 	 * b) vfs attrs
3719 	 * c) attrs w/per-object scope requested (change, filehandle, etc)
3720 	 *    other than mounted_on_fileid (which we can take from dirent)
3721 	 */
3722 	objbits = attrs ? attrs & NFS4_VP_ATTR_MASK : 0;
3723 
3724 	if (va_mask || sargp->sbp || (objbits & ~FATTR4_MOUNTED_ON_FILEID_MASK))
3725 		*need_to_lookup = TRUE;
3726 	else
3727 		*need_to_lookup = FALSE;
3728 
3729 	if (sargp->sbp == NULL)
3730 		return (NFS4_OK);
3731 
3732 	/*
3733 	 * If filesystem attrs are requested, get them now from the
3734 	 * directory vp, as most entries will have same filesystem. The only
3735 	 * exception are mounted over entries but we handle
3736 	 * those as we go (XXX mounted over detection not yet implemented).
3737 	 */
3738 	sargp->vap->va_mask = 0;	/* to avoid VOP_GETATTR */
3739 	status = bitmap4_get_sysattrs(sargp);
3740 	sargp->vap->va_mask = va_mask;
3741 
3742 	if ((status != NFS4_OK) && sargp->rdattr_error_req) {
3743 		/*
3744 		 * Failed to get filesystem attributes.
3745 		 * Return a rdattr_error for each entry, but don't fail.
3746 		 * However, don't get any obj-dependent attrs.
3747 		 */
3748 		sargp->rdattr_error = status;	/* for rdattr_error */
3749 		*need_to_lookup = FALSE;
3750 		/*
3751 		 * At least get fileid for regular readdir output
3752 		 */
3753 		sargp->vap->va_mask &= AT_NODEID;
3754 		status = NFS4_OK;
3755 	}
3756 
3757 	return (status);
3758 }
3759 
3760 /*
3761  * readlink: args: CURRENT_FH.
3762  *	res: status. If success - CURRENT_FH unchanged, return linktext.
3763  */
3764 
3765 /* ARGSUSED */
3766 static void
3767 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
3768     struct compound_state *cs)
3769 {
3770 	READLINK4res *resp = &resop->nfs_resop4_u.opreadlink;
3771 	int error;
3772 	vnode_t *vp;
3773 	struct iovec iov;
3774 	struct vattr va;
3775 	struct uio uio;
3776 	char *data;
3777 	struct sockaddr *ca;
3778 	char *name = NULL;
3779 	int is_referral;
3780 
3781 	DTRACE_NFSV4_1(op__readlink__start, struct compound_state *, cs);
3782 
3783 	/* CURRENT_FH: directory */
3784 	vp = cs->vp;
3785 	if (vp == NULL) {
3786 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
3787 		goto out;
3788 	}
3789 
3790 	if (cs->access == CS_ACCESS_DENIED) {
3791 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3792 		goto out;
3793 	}
3794 
3795 	/* Is it a referral? */
3796 	if (vn_is_nfs_reparse(vp, cs->cr) && client_is_downrev(req)) {
3797 
3798 		is_referral = 1;
3799 
3800 	} else {
3801 
3802 		is_referral = 0;
3803 
3804 		if (vp->v_type == VDIR) {
3805 			*cs->statusp = resp->status = NFS4ERR_ISDIR;
3806 			goto out;
3807 		}
3808 
3809 		if (vp->v_type != VLNK) {
3810 			*cs->statusp = resp->status = NFS4ERR_INVAL;
3811 			goto out;
3812 		}
3813 
3814 	}
3815 
3816 	va.va_mask = AT_MODE;
3817 	error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL);
3818 	if (error) {
3819 		*cs->statusp = resp->status = puterrno4(error);
3820 		goto out;
3821 	}
3822 
3823 	if (MANDLOCK(vp, va.va_mode)) {
3824 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
3825 		goto out;
3826 	}
3827 
3828 	data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP);
3829 
3830 	if (is_referral) {
3831 		char *s;
3832 		size_t strsz;
3833 
3834 		/* Get an artificial symlink based on a referral */
3835 		s = build_symlink(vp, cs->cr, &strsz);
3836 		global_svstat_ptr[4][NFS_REFERLINKS].value.ui64++;
3837 		DTRACE_PROBE2(nfs4serv__func__referral__reflink,
3838 		    vnode_t *, vp, char *, s);
3839 		if (s == NULL)
3840 			error = EINVAL;
3841 		else {
3842 			error = 0;
3843 			(void) strlcpy(data, s, MAXPATHLEN + 1);
3844 			kmem_free(s, strsz);
3845 		}
3846 
3847 	} else {
3848 
3849 		iov.iov_base = data;
3850 		iov.iov_len = MAXPATHLEN;
3851 		uio.uio_iov = &iov;
3852 		uio.uio_iovcnt = 1;
3853 		uio.uio_segflg = UIO_SYSSPACE;
3854 		uio.uio_extflg = UIO_COPY_CACHED;
3855 		uio.uio_loffset = 0;
3856 		uio.uio_resid = MAXPATHLEN;
3857 
3858 		error = VOP_READLINK(vp, &uio, cs->cr, NULL);
3859 
3860 		if (!error)
3861 			*(data + MAXPATHLEN - uio.uio_resid) = '\0';
3862 	}
3863 
3864 	if (error) {
3865 		kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1);
3866 		*cs->statusp = resp->status = puterrno4(error);
3867 		goto out;
3868 	}
3869 
3870 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
3871 	name = nfscmd_convname(ca, cs->exi, data, NFSCMD_CONV_OUTBOUND,
3872 	    MAXPATHLEN  + 1);
3873 
3874 	if (name == NULL) {
3875 		/*
3876 		 * Even though the conversion failed, we return
3877 		 * something. We just don't translate it.
3878 		 */
3879 		name = data;
3880 	}
3881 
3882 	/*
3883 	 * treat link name as data
3884 	 */
3885 	(void) str_to_utf8(name, (utf8string *)&resp->link);
3886 
3887 	if (name != data)
3888 		kmem_free(name, MAXPATHLEN + 1);
3889 	kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1);
3890 	*cs->statusp = resp->status = NFS4_OK;
3891 
3892 out:
3893 	DTRACE_NFSV4_2(op__readlink__done, struct compound_state *, cs,
3894 	    READLINK4res *, resp);
3895 }
3896 
3897 static void
3898 rfs4_op_readlink_free(nfs_resop4 *resop)
3899 {
3900 	READLINK4res *resp = &resop->nfs_resop4_u.opreadlink;
3901 	utf8string *symlink = (utf8string *)&resp->link;
3902 
3903 	if (symlink->utf8string_val) {
3904 		UTF8STRING_FREE(*symlink)
3905 	}
3906 }
3907 
3908 /*
3909  * release_lockowner:
3910  *	Release any state associated with the supplied
3911  *	lockowner. Note if any lo_state is holding locks we will not
3912  *	rele that lo_state and thus the lockowner will not be destroyed.
3913  *	A client using lock after the lock owner stateid has been released
3914  *	will suffer the consequence of NFS4ERR_BAD_STATEID and would have
3915  *	to reissue the lock with new_lock_owner set to TRUE.
3916  *	args: lock_owner
3917  *	res:  status
3918  */
3919 /* ARGSUSED */
3920 static void
3921 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop,
3922     struct svc_req *req, struct compound_state *cs)
3923 {
3924 	RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner;
3925 	RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner;
3926 	rfs4_lockowner_t *lo;
3927 	rfs4_openowner_t *oo;
3928 	rfs4_state_t *sp;
3929 	rfs4_lo_state_t *lsp;
3930 	rfs4_client_t *cp;
3931 	bool_t create = FALSE;
3932 	locklist_t *llist;
3933 	sysid_t sysid;
3934 
3935 	DTRACE_NFSV4_2(op__release__lockowner__start, struct compound_state *,
3936 	    cs, RELEASE_LOCKOWNER4args *, ap);
3937 
3938 	/* Make sure there is a clientid around for this request */
3939 	cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE);
3940 
3941 	if (cp == NULL) {
3942 		*cs->statusp = resp->status =
3943 		    rfs4_check_clientid(&ap->lock_owner.clientid, 0);
3944 		goto out;
3945 	}
3946 	rfs4_client_rele(cp);
3947 
3948 	lo = rfs4_findlockowner(&ap->lock_owner, &create);
3949 	if (lo == NULL) {
3950 		*cs->statusp = resp->status = NFS4_OK;
3951 		goto out;
3952 	}
3953 	ASSERT(lo->rl_client != NULL);
3954 
3955 	/*
3956 	 * Check for EXPIRED client. If so will reap state with in a lease
3957 	 * period or on next set_clientid_confirm step
3958 	 */
3959 	if (rfs4_lease_expired(lo->rl_client)) {
3960 		rfs4_lockowner_rele(lo);
3961 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
3962 		goto out;
3963 	}
3964 
3965 	/*
3966 	 * If no sysid has been assigned, then no locks exist; just return.
3967 	 */
3968 	rfs4_dbe_lock(lo->rl_client->rc_dbe);
3969 	if (lo->rl_client->rc_sysidt == LM_NOSYSID) {
3970 		rfs4_lockowner_rele(lo);
3971 		rfs4_dbe_unlock(lo->rl_client->rc_dbe);
3972 		goto out;
3973 	}
3974 
3975 	sysid = lo->rl_client->rc_sysidt;
3976 	rfs4_dbe_unlock(lo->rl_client->rc_dbe);
3977 
3978 	/*
3979 	 * Mark the lockowner invalid.
3980 	 */
3981 	rfs4_dbe_hide(lo->rl_dbe);
3982 
3983 	/*
3984 	 * sysid-pid pair should now not be used since the lockowner is
3985 	 * invalid. If the client were to instantiate the lockowner again
3986 	 * it would be assigned a new pid. Thus we can get the list of
3987 	 * current locks.
3988 	 */
3989 
3990 	llist = flk_get_active_locks(sysid, lo->rl_pid);
3991 	/* If we are still holding locks fail */
3992 	if (llist != NULL) {
3993 
3994 		*cs->statusp = resp->status = NFS4ERR_LOCKS_HELD;
3995 
3996 		flk_free_locklist(llist);
3997 		/*
3998 		 * We need to unhide the lockowner so the client can
3999 		 * try it again. The bad thing here is if the client
4000 		 * has a logic error that took it here in the first place
4001 		 * he probably has lost accounting of the locks that it
4002 		 * is holding. So we may have dangling state until the
4003 		 * open owner state is reaped via close. One scenario
4004 		 * that could possibly occur is that the client has
4005 		 * sent the unlock request(s) in separate threads
4006 		 * and has not waited for the replies before sending the
4007 		 * RELEASE_LOCKOWNER request. Presumably, it would expect
4008 		 * and deal appropriately with NFS4ERR_LOCKS_HELD, by
4009 		 * reissuing the request.
4010 		 */
4011 		rfs4_dbe_unhide(lo->rl_dbe);
4012 		rfs4_lockowner_rele(lo);
4013 		goto out;
4014 	}
4015 
4016 	/*
4017 	 * For the corresponding client we need to check each open
4018 	 * owner for any opens that have lockowner state associated
4019 	 * with this lockowner.
4020 	 */
4021 
4022 	rfs4_dbe_lock(lo->rl_client->rc_dbe);
4023 	for (oo = list_head(&lo->rl_client->rc_openownerlist); oo != NULL;
4024 	    oo = list_next(&lo->rl_client->rc_openownerlist, oo)) {
4025 
4026 		rfs4_dbe_lock(oo->ro_dbe);
4027 		for (sp = list_head(&oo->ro_statelist); sp != NULL;
4028 		    sp = list_next(&oo->ro_statelist, sp)) {
4029 
4030 			rfs4_dbe_lock(sp->rs_dbe);
4031 			for (lsp = list_head(&sp->rs_lostatelist);
4032 			    lsp != NULL;
4033 			    lsp = list_next(&sp->rs_lostatelist, lsp)) {
4034 				if (lsp->rls_locker == lo) {
4035 					rfs4_dbe_lock(lsp->rls_dbe);
4036 					rfs4_dbe_invalidate(lsp->rls_dbe);
4037 					rfs4_dbe_unlock(lsp->rls_dbe);
4038 				}
4039 			}
4040 			rfs4_dbe_unlock(sp->rs_dbe);
4041 		}
4042 		rfs4_dbe_unlock(oo->ro_dbe);
4043 	}
4044 	rfs4_dbe_unlock(lo->rl_client->rc_dbe);
4045 
4046 	rfs4_lockowner_rele(lo);
4047 
4048 	*cs->statusp = resp->status = NFS4_OK;
4049 
4050 out:
4051 	DTRACE_NFSV4_2(op__release__lockowner__done, struct compound_state *,
4052 	    cs, RELEASE_LOCKOWNER4res *, resp);
4053 }
4054 
4055 /*
4056  * short utility function to lookup a file and recall the delegation
4057  */
4058 static rfs4_file_t *
4059 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp,
4060     int *lkup_error, cred_t *cr)
4061 {
4062 	vnode_t *vp;
4063 	rfs4_file_t *fp = NULL;
4064 	bool_t fcreate = FALSE;
4065 	int error;
4066 
4067 	if (vpp)
4068 		*vpp = NULL;
4069 
4070 	if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, NULL, NULL,
4071 	    NULL)) == 0) {
4072 		if (vp->v_type == VREG)
4073 			fp = rfs4_findfile(vp, NULL, &fcreate);
4074 		if (vpp)
4075 			*vpp = vp;
4076 		else
4077 			VN_RELE(vp);
4078 	}
4079 
4080 	if (lkup_error)
4081 		*lkup_error = error;
4082 
4083 	return (fp);
4084 }
4085 
4086 /*
4087  * remove: args: CURRENT_FH: directory; name.
4088  *	res: status. If success - CURRENT_FH unchanged, return change_info
4089  *		for directory.
4090  */
4091 /* ARGSUSED */
4092 static void
4093 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4094     struct compound_state *cs)
4095 {
4096 	REMOVE4args *args = &argop->nfs_argop4_u.opremove;
4097 	REMOVE4res *resp = &resop->nfs_resop4_u.opremove;
4098 	int error;
4099 	vnode_t *dvp, *vp;
4100 	struct vattr bdva, idva, adva;
4101 	char *nm;
4102 	uint_t len;
4103 	rfs4_file_t *fp;
4104 	int in_crit = 0;
4105 	bslabel_t *clabel;
4106 	struct sockaddr *ca;
4107 	char *name = NULL;
4108 	nfsstat4 status;
4109 
4110 	DTRACE_NFSV4_2(op__remove__start, struct compound_state *, cs,
4111 	    REMOVE4args *, args);
4112 
4113 	/* CURRENT_FH: directory */
4114 	dvp = cs->vp;
4115 	if (dvp == NULL) {
4116 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4117 		goto out;
4118 	}
4119 
4120 	if (cs->access == CS_ACCESS_DENIED) {
4121 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4122 		goto out;
4123 	}
4124 
4125 	/*
4126 	 * If there is an unshared filesystem mounted on this vnode,
4127 	 * Do not allow to remove anything in this directory.
4128 	 */
4129 	if (vn_ismntpt(dvp)) {
4130 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4131 		goto out;
4132 	}
4133 
4134 	if (dvp->v_type != VDIR) {
4135 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
4136 		goto out;
4137 	}
4138 
4139 	status = utf8_dir_verify(&args->target);
4140 	if (status != NFS4_OK) {
4141 		*cs->statusp = resp->status = status;
4142 		goto out;
4143 	}
4144 
4145 	/*
4146 	 * Lookup the file so that we can check if it's a directory
4147 	 */
4148 	nm = utf8_to_fn(&args->target, &len, NULL);
4149 	if (nm == NULL) {
4150 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4151 		goto out;
4152 	}
4153 
4154 	if (len > MAXNAMELEN) {
4155 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
4156 		kmem_free(nm, len);
4157 		goto out;
4158 	}
4159 
4160 	if (rdonly4(req, cs)) {
4161 		*cs->statusp = resp->status = NFS4ERR_ROFS;
4162 		kmem_free(nm, len);
4163 		goto out;
4164 	}
4165 
4166 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
4167 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
4168 	    MAXPATHLEN  + 1);
4169 
4170 	if (name == NULL) {
4171 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4172 		kmem_free(nm, len);
4173 		goto out;
4174 	}
4175 
4176 	/*
4177 	 * Lookup the file to determine type and while we are see if
4178 	 * there is a file struct around and check for delegation.
4179 	 * We don't need to acquire va_seq before this lookup, if
4180 	 * it causes an update, cinfo.before will not match, which will
4181 	 * trigger a cache flush even if atomic is TRUE.
4182 	 */
4183 	if (fp = rfs4_lookup_and_findfile(dvp, name, &vp, &error, cs->cr)) {
4184 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
4185 		    NULL)) {
4186 			VN_RELE(vp);
4187 			rfs4_file_rele(fp);
4188 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4189 			if (nm != name)
4190 				kmem_free(name, MAXPATHLEN + 1);
4191 			kmem_free(nm, len);
4192 			goto out;
4193 		}
4194 	}
4195 
4196 	/* Didn't find anything to remove */
4197 	if (vp == NULL) {
4198 		*cs->statusp = resp->status = error;
4199 		if (nm != name)
4200 			kmem_free(name, MAXPATHLEN + 1);
4201 		kmem_free(nm, len);
4202 		goto out;
4203 	}
4204 
4205 	if (nbl_need_check(vp)) {
4206 		nbl_start_crit(vp, RW_READER);
4207 		in_crit = 1;
4208 		if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0, NULL)) {
4209 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4210 			if (nm != name)
4211 				kmem_free(name, MAXPATHLEN + 1);
4212 			kmem_free(nm, len);
4213 			nbl_end_crit(vp);
4214 			VN_RELE(vp);
4215 			if (fp) {
4216 				rfs4_clear_dont_grant(fp);
4217 				rfs4_file_rele(fp);
4218 			}
4219 			goto out;
4220 		}
4221 	}
4222 
4223 	/* check label before allowing removal */
4224 	if (is_system_labeled()) {
4225 		ASSERT(req->rq_label != NULL);
4226 		clabel = req->rq_label;
4227 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
4228 		    "got client label from request(1)",
4229 		    struct svc_req *, req);
4230 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
4231 			if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK,
4232 			    cs->exi)) {
4233 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
4234 				if (name != nm)
4235 					kmem_free(name, MAXPATHLEN + 1);
4236 				kmem_free(nm, len);
4237 				if (in_crit)
4238 					nbl_end_crit(vp);
4239 				VN_RELE(vp);
4240 				if (fp) {
4241 					rfs4_clear_dont_grant(fp);
4242 					rfs4_file_rele(fp);
4243 				}
4244 				goto out;
4245 			}
4246 		}
4247 	}
4248 
4249 	/* Get dir "before" change value */
4250 	bdva.va_mask = AT_CTIME|AT_SEQ;
4251 	error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL);
4252 	if (error) {
4253 		*cs->statusp = resp->status = puterrno4(error);
4254 		if (nm != name)
4255 			kmem_free(name, MAXPATHLEN + 1);
4256 		kmem_free(nm, len);
4257 		if (in_crit)
4258 			nbl_end_crit(vp);
4259 		VN_RELE(vp);
4260 		if (fp) {
4261 			rfs4_clear_dont_grant(fp);
4262 			rfs4_file_rele(fp);
4263 		}
4264 		goto out;
4265 	}
4266 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime)
4267 
4268 	/* Actually do the REMOVE operation */
4269 	if (vp->v_type == VDIR) {
4270 		/*
4271 		 * Can't remove a directory that has a mounted-on filesystem.
4272 		 */
4273 		if (vn_ismntpt(vp)) {
4274 			error = EACCES;
4275 		} else {
4276 			/*
4277 			 * System V defines rmdir to return EEXIST,
4278 			 * not ENOTEMPTY, if the directory is not
4279 			 * empty.  A System V NFS server needs to map
4280 			 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to
4281 			 * transmit over the wire.
4282 			 */
4283 			if ((error = VOP_RMDIR(dvp, name, rootdir, cs->cr,
4284 			    NULL, 0)) == EEXIST)
4285 				error = ENOTEMPTY;
4286 		}
4287 
4288 		if (in_crit)
4289 			nbl_end_crit(vp);
4290 		VN_RELE(vp);
4291 	} else {
4292 		if (!in_crit)
4293 			VN_RELE(vp);
4294 		if ((error = VOP_REMOVE(dvp, name, cs->cr, NULL, 0)) == 0 &&
4295 		    fp != NULL) {
4296 			struct vattr va;
4297 			vnode_t *tvp;
4298 
4299 			rfs4_dbe_lock(fp->rf_dbe);
4300 			tvp = fp->rf_vp;
4301 			if (tvp)
4302 				VN_HOLD(tvp);
4303 			rfs4_dbe_unlock(fp->rf_dbe);
4304 
4305 			if (tvp) {
4306 				/*
4307 				 * This is va_seq safe because we are not
4308 				 * manipulating dvp.
4309 				 */
4310 				va.va_mask = AT_NLINK;
4311 				if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) &&
4312 				    va.va_nlink == 0) {
4313 					/* Remove state on file remove */
4314 					if (in_crit) {
4315 						nbl_end_crit(vp);
4316 						VN_RELE(vp);
4317 						in_crit = 0;
4318 					}
4319 					rfs4_close_all_state(fp);
4320 				}
4321 				VN_RELE(tvp);
4322 			}
4323 		}
4324 
4325 		if (in_crit) {
4326 			nbl_end_crit(vp);
4327 			VN_RELE(vp);
4328 		}
4329 	}
4330 
4331 	if (fp) {
4332 		rfs4_clear_dont_grant(fp);
4333 		rfs4_file_rele(fp);
4334 	}
4335 	if (nm != name)
4336 		kmem_free(name, MAXPATHLEN + 1);
4337 	kmem_free(nm, len);
4338 
4339 	if (error) {
4340 		*cs->statusp = resp->status = puterrno4(error);
4341 		goto out;
4342 	}
4343 
4344 	/*
4345 	 * Get the initial "after" sequence number, if it fails, set to zero
4346 	 */
4347 	idva.va_mask = AT_SEQ;
4348 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL))
4349 		idva.va_seq = 0;
4350 
4351 	/*
4352 	 * Force modified data and metadata out to stable storage.
4353 	 */
4354 	(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
4355 
4356 	/*
4357 	 * Get "after" change value, if it fails, simply return the
4358 	 * before value.
4359 	 */
4360 	adva.va_mask = AT_CTIME|AT_SEQ;
4361 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) {
4362 		adva.va_ctime = bdva.va_ctime;
4363 		adva.va_seq = 0;
4364 	}
4365 
4366 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
4367 
4368 	/*
4369 	 * The cinfo.atomic = TRUE only if we have
4370 	 * non-zero va_seq's, and it has incremented by exactly one
4371 	 * during the VOP_REMOVE/RMDIR and it didn't change during
4372 	 * the VOP_FSYNC.
4373 	 */
4374 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
4375 	    idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq)
4376 		resp->cinfo.atomic = TRUE;
4377 	else
4378 		resp->cinfo.atomic = FALSE;
4379 
4380 	*cs->statusp = resp->status = NFS4_OK;
4381 
4382 out:
4383 	DTRACE_NFSV4_2(op__remove__done, struct compound_state *, cs,
4384 	    REMOVE4res *, resp);
4385 }
4386 
4387 /*
4388  * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory,
4389  *		oldname and newname.
4390  *	res: status. If success - CURRENT_FH unchanged, return change_info
4391  *		for both from and target directories.
4392  */
4393 /* ARGSUSED */
4394 static void
4395 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4396     struct compound_state *cs)
4397 {
4398 	RENAME4args *args = &argop->nfs_argop4_u.oprename;
4399 	RENAME4res *resp = &resop->nfs_resop4_u.oprename;
4400 	int error;
4401 	vnode_t *odvp;
4402 	vnode_t *ndvp;
4403 	vnode_t *srcvp, *targvp;
4404 	struct vattr obdva, oidva, oadva;
4405 	struct vattr nbdva, nidva, nadva;
4406 	char *onm, *nnm;
4407 	uint_t olen, nlen;
4408 	rfs4_file_t *fp, *sfp;
4409 	int in_crit_src, in_crit_targ;
4410 	int fp_rele_grant_hold, sfp_rele_grant_hold;
4411 	bslabel_t *clabel;
4412 	struct sockaddr *ca;
4413 	char *converted_onm = NULL;
4414 	char *converted_nnm = NULL;
4415 	nfsstat4 status;
4416 
4417 	DTRACE_NFSV4_2(op__rename__start, struct compound_state *, cs,
4418 	    RENAME4args *, args);
4419 
4420 	fp = sfp = NULL;
4421 	srcvp = targvp = NULL;
4422 	in_crit_src = in_crit_targ = 0;
4423 	fp_rele_grant_hold = sfp_rele_grant_hold = 0;
4424 
4425 	/* CURRENT_FH: target directory */
4426 	ndvp = cs->vp;
4427 	if (ndvp == NULL) {
4428 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4429 		goto out;
4430 	}
4431 
4432 	/* SAVED_FH: from directory */
4433 	odvp = cs->saved_vp;
4434 	if (odvp == NULL) {
4435 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4436 		goto out;
4437 	}
4438 
4439 	if (cs->access == CS_ACCESS_DENIED) {
4440 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4441 		goto out;
4442 	}
4443 
4444 	/*
4445 	 * If there is an unshared filesystem mounted on this vnode,
4446 	 * do not allow to rename objects in this directory.
4447 	 */
4448 	if (vn_ismntpt(odvp)) {
4449 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4450 		goto out;
4451 	}
4452 
4453 	/*
4454 	 * If there is an unshared filesystem mounted on this vnode,
4455 	 * do not allow to rename to this directory.
4456 	 */
4457 	if (vn_ismntpt(ndvp)) {
4458 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4459 		goto out;
4460 	}
4461 
4462 	if (odvp->v_type != VDIR || ndvp->v_type != VDIR) {
4463 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
4464 		goto out;
4465 	}
4466 
4467 	if (cs->saved_exi != cs->exi) {
4468 		*cs->statusp = resp->status = NFS4ERR_XDEV;
4469 		goto out;
4470 	}
4471 
4472 	status = utf8_dir_verify(&args->oldname);
4473 	if (status != NFS4_OK) {
4474 		*cs->statusp = resp->status = status;
4475 		goto out;
4476 	}
4477 
4478 	status = utf8_dir_verify(&args->newname);
4479 	if (status != NFS4_OK) {
4480 		*cs->statusp = resp->status = status;
4481 		goto out;
4482 	}
4483 
4484 	onm = utf8_to_fn(&args->oldname, &olen, NULL);
4485 	if (onm == NULL) {
4486 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4487 		goto out;
4488 	}
4489 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
4490 	nlen = MAXPATHLEN + 1;
4491 	converted_onm = nfscmd_convname(ca, cs->exi, onm, NFSCMD_CONV_INBOUND,
4492 	    nlen);
4493 
4494 	if (converted_onm == NULL) {
4495 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4496 		kmem_free(onm, olen);
4497 		goto out;
4498 	}
4499 
4500 	nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4501 	if (nnm == NULL) {
4502 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4503 		if (onm != converted_onm)
4504 			kmem_free(converted_onm, MAXPATHLEN + 1);
4505 		kmem_free(onm, olen);
4506 		goto out;
4507 	}
4508 	converted_nnm = nfscmd_convname(ca, cs->exi, nnm, NFSCMD_CONV_INBOUND,
4509 	    MAXPATHLEN  + 1);
4510 
4511 	if (converted_nnm == NULL) {
4512 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4513 		kmem_free(nnm, nlen);
4514 		nnm = NULL;
4515 		if (onm != converted_onm)
4516 			kmem_free(converted_onm, MAXPATHLEN + 1);
4517 		kmem_free(onm, olen);
4518 		goto out;
4519 	}
4520 
4521 
4522 	if (olen > MAXNAMELEN || nlen > MAXNAMELEN) {
4523 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
4524 		kmem_free(onm, olen);
4525 		kmem_free(nnm, nlen);
4526 		goto out;
4527 	}
4528 
4529 
4530 	if (rdonly4(req, cs)) {
4531 		*cs->statusp = resp->status = NFS4ERR_ROFS;
4532 		if (onm != converted_onm)
4533 			kmem_free(converted_onm, MAXPATHLEN + 1);
4534 		kmem_free(onm, olen);
4535 		if (nnm != converted_nnm)
4536 			kmem_free(converted_nnm, MAXPATHLEN + 1);
4537 		kmem_free(nnm, nlen);
4538 		goto out;
4539 	}
4540 
4541 	/* check label of the target dir */
4542 	if (is_system_labeled()) {
4543 		ASSERT(req->rq_label != NULL);
4544 		clabel = req->rq_label;
4545 		DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *,
4546 		    "got client label from request(1)",
4547 		    struct svc_req *, req);
4548 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
4549 			if (!do_rfs_label_check(clabel, ndvp,
4550 			    EQUALITY_CHECK, cs->exi)) {
4551 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
4552 				goto err_out;
4553 			}
4554 		}
4555 	}
4556 
4557 	/*
4558 	 * Is the source a file and have a delegation?
4559 	 * We don't need to acquire va_seq before these lookups, if
4560 	 * it causes an update, cinfo.before will not match, which will
4561 	 * trigger a cache flush even if atomic is TRUE.
4562 	 */
4563 	if (sfp = rfs4_lookup_and_findfile(odvp, converted_onm, &srcvp,
4564 	    &error, cs->cr)) {
4565 		if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE,
4566 		    NULL)) {
4567 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4568 			goto err_out;
4569 		}
4570 	}
4571 
4572 	if (srcvp == NULL) {
4573 		*cs->statusp = resp->status = puterrno4(error);
4574 		if (onm != converted_onm)
4575 			kmem_free(converted_onm, MAXPATHLEN + 1);
4576 		kmem_free(onm, olen);
4577 		if (nnm != converted_nnm)
4578 			kmem_free(converted_nnm, MAXPATHLEN + 1);
4579 		kmem_free(nnm, nlen);
4580 		goto out;
4581 	}
4582 
4583 	sfp_rele_grant_hold = 1;
4584 
4585 	/* Does the destination exist and a file and have a delegation? */
4586 	if (fp = rfs4_lookup_and_findfile(ndvp, converted_nnm, &targvp,
4587 	    NULL, cs->cr)) {
4588 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
4589 		    NULL)) {
4590 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4591 			goto err_out;
4592 		}
4593 	}
4594 	fp_rele_grant_hold = 1;
4595 
4596 
4597 	/* Check for NBMAND lock on both source and target */
4598 	if (nbl_need_check(srcvp)) {
4599 		nbl_start_crit(srcvp, RW_READER);
4600 		in_crit_src = 1;
4601 		if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) {
4602 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4603 			goto err_out;
4604 		}
4605 	}
4606 
4607 	if (targvp && nbl_need_check(targvp)) {
4608 		nbl_start_crit(targvp, RW_READER);
4609 		in_crit_targ = 1;
4610 		if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) {
4611 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4612 			goto err_out;
4613 		}
4614 	}
4615 
4616 	/* Get source "before" change value */
4617 	obdva.va_mask = AT_CTIME|AT_SEQ;
4618 	error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL);
4619 	if (!error) {
4620 		nbdva.va_mask = AT_CTIME|AT_SEQ;
4621 		error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL);
4622 	}
4623 	if (error) {
4624 		*cs->statusp = resp->status = puterrno4(error);
4625 		goto err_out;
4626 	}
4627 
4628 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime)
4629 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime)
4630 
4631 	if ((error = VOP_RENAME(odvp, converted_onm, ndvp, converted_nnm,
4632 	    cs->cr, NULL, 0)) == 0 && fp != NULL) {
4633 		struct vattr va;
4634 		vnode_t *tvp;
4635 
4636 		rfs4_dbe_lock(fp->rf_dbe);
4637 		tvp = fp->rf_vp;
4638 		if (tvp)
4639 			VN_HOLD(tvp);
4640 		rfs4_dbe_unlock(fp->rf_dbe);
4641 
4642 		if (tvp) {
4643 			va.va_mask = AT_NLINK;
4644 			if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) &&
4645 			    va.va_nlink == 0) {
4646 				/* The file is gone and so should the state */
4647 				if (in_crit_targ) {
4648 					nbl_end_crit(targvp);
4649 					in_crit_targ = 0;
4650 				}
4651 				rfs4_close_all_state(fp);
4652 			}
4653 			VN_RELE(tvp);
4654 		}
4655 	}
4656 	if (error == 0)
4657 		vn_renamepath(ndvp, srcvp, nnm, nlen - 1);
4658 
4659 	if (in_crit_src)
4660 		nbl_end_crit(srcvp);
4661 	if (srcvp)
4662 		VN_RELE(srcvp);
4663 	if (in_crit_targ)
4664 		nbl_end_crit(targvp);
4665 	if (targvp)
4666 		VN_RELE(targvp);
4667 
4668 	if (sfp) {
4669 		rfs4_clear_dont_grant(sfp);
4670 		rfs4_file_rele(sfp);
4671 	}
4672 	if (fp) {
4673 		rfs4_clear_dont_grant(fp);
4674 		rfs4_file_rele(fp);
4675 	}
4676 
4677 	if (converted_onm != onm)
4678 		kmem_free(converted_onm, MAXPATHLEN + 1);
4679 	kmem_free(onm, olen);
4680 	if (converted_nnm != nnm)
4681 		kmem_free(converted_nnm, MAXPATHLEN + 1);
4682 	kmem_free(nnm, nlen);
4683 
4684 	/*
4685 	 * Get the initial "after" sequence number, if it fails, set to zero
4686 	 */
4687 	oidva.va_mask = AT_SEQ;
4688 	if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL))
4689 		oidva.va_seq = 0;
4690 
4691 	nidva.va_mask = AT_SEQ;
4692 	if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL))
4693 		nidva.va_seq = 0;
4694 
4695 	/*
4696 	 * Force modified data and metadata out to stable storage.
4697 	 */
4698 	(void) VOP_FSYNC(odvp, 0, cs->cr, NULL);
4699 	(void) VOP_FSYNC(ndvp, 0, cs->cr, NULL);
4700 
4701 	if (error) {
4702 		*cs->statusp = resp->status = puterrno4(error);
4703 		goto out;
4704 	}
4705 
4706 	/*
4707 	 * Get "after" change values, if it fails, simply return the
4708 	 * before value.
4709 	 */
4710 	oadva.va_mask = AT_CTIME|AT_SEQ;
4711 	if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) {
4712 		oadva.va_ctime = obdva.va_ctime;
4713 		oadva.va_seq = 0;
4714 	}
4715 
4716 	nadva.va_mask = AT_CTIME|AT_SEQ;
4717 	if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) {
4718 		nadva.va_ctime = nbdva.va_ctime;
4719 		nadva.va_seq = 0;
4720 	}
4721 
4722 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime)
4723 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime)
4724 
4725 	/*
4726 	 * The cinfo.atomic = TRUE only if we have
4727 	 * non-zero va_seq's, and it has incremented by exactly one
4728 	 * during the VOP_RENAME and it didn't change during the VOP_FSYNC.
4729 	 */
4730 	if (obdva.va_seq && oidva.va_seq && oadva.va_seq &&
4731 	    oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq)
4732 		resp->source_cinfo.atomic = TRUE;
4733 	else
4734 		resp->source_cinfo.atomic = FALSE;
4735 
4736 	if (nbdva.va_seq && nidva.va_seq && nadva.va_seq &&
4737 	    nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq)
4738 		resp->target_cinfo.atomic = TRUE;
4739 	else
4740 		resp->target_cinfo.atomic = FALSE;
4741 
4742 #ifdef	VOLATILE_FH_TEST
4743 	{
4744 	extern void add_volrnm_fh(struct exportinfo *, vnode_t *);
4745 
4746 	/*
4747 	 * Add the renamed file handle to the volatile rename list
4748 	 */
4749 	if (cs->exi->exi_export.ex_flags & EX_VOLRNM) {
4750 		/* file handles may expire on rename */
4751 		vnode_t *vp;
4752 
4753 		nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4754 		/*
4755 		 * Already know that nnm will be a valid string
4756 		 */
4757 		error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr,
4758 		    NULL, NULL, NULL);
4759 		kmem_free(nnm, nlen);
4760 		if (!error) {
4761 			add_volrnm_fh(cs->exi, vp);
4762 			VN_RELE(vp);
4763 		}
4764 	}
4765 	}
4766 #endif	/* VOLATILE_FH_TEST */
4767 
4768 	*cs->statusp = resp->status = NFS4_OK;
4769 out:
4770 	DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs,
4771 	    RENAME4res *, resp);
4772 	return;
4773 
4774 err_out:
4775 	if (onm != converted_onm)
4776 		kmem_free(converted_onm, MAXPATHLEN + 1);
4777 	if (onm != NULL)
4778 		kmem_free(onm, olen);
4779 	if (nnm != converted_nnm)
4780 		kmem_free(converted_nnm, MAXPATHLEN + 1);
4781 	if (nnm != NULL)
4782 		kmem_free(nnm, nlen);
4783 
4784 	if (in_crit_src) nbl_end_crit(srcvp);
4785 	if (in_crit_targ) nbl_end_crit(targvp);
4786 	if (targvp) VN_RELE(targvp);
4787 	if (srcvp) VN_RELE(srcvp);
4788 	if (sfp) {
4789 		if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp);
4790 		rfs4_file_rele(sfp);
4791 	}
4792 	if (fp) {
4793 		if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp);
4794 		rfs4_file_rele(fp);
4795 	}
4796 
4797 	DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs,
4798 	    RENAME4res *, resp);
4799 }
4800 
4801 /* ARGSUSED */
4802 static void
4803 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4804     struct compound_state *cs)
4805 {
4806 	RENEW4args *args = &argop->nfs_argop4_u.oprenew;
4807 	RENEW4res *resp = &resop->nfs_resop4_u.oprenew;
4808 	rfs4_client_t *cp;
4809 
4810 	DTRACE_NFSV4_2(op__renew__start, struct compound_state *, cs,
4811 	    RENEW4args *, args);
4812 
4813 	if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) {
4814 		*cs->statusp = resp->status =
4815 		    rfs4_check_clientid(&args->clientid, 0);
4816 		goto out;
4817 	}
4818 
4819 	if (rfs4_lease_expired(cp)) {
4820 		rfs4_client_rele(cp);
4821 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
4822 		goto out;
4823 	}
4824 
4825 	rfs4_update_lease(cp);
4826 
4827 	mutex_enter(cp->rc_cbinfo.cb_lock);
4828 	if (cp->rc_cbinfo.cb_notified_of_cb_path_down == FALSE) {
4829 		cp->rc_cbinfo.cb_notified_of_cb_path_down = TRUE;
4830 		*cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN;
4831 	} else {
4832 		*cs->statusp = resp->status = NFS4_OK;
4833 	}
4834 	mutex_exit(cp->rc_cbinfo.cb_lock);
4835 
4836 	rfs4_client_rele(cp);
4837 
4838 out:
4839 	DTRACE_NFSV4_2(op__renew__done, struct compound_state *, cs,
4840 	    RENEW4res *, resp);
4841 }
4842 
4843 /* ARGSUSED */
4844 static void
4845 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
4846     struct compound_state *cs)
4847 {
4848 	RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh;
4849 
4850 	DTRACE_NFSV4_1(op__restorefh__start, struct compound_state *, cs);
4851 
4852 	/* No need to check cs->access - we are not accessing any object */
4853 	if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) {
4854 		*cs->statusp = resp->status = NFS4ERR_RESTOREFH;
4855 		goto out;
4856 	}
4857 	if (cs->vp != NULL) {
4858 		VN_RELE(cs->vp);
4859 	}
4860 	cs->vp = cs->saved_vp;
4861 	cs->saved_vp = NULL;
4862 	cs->exi = cs->saved_exi;
4863 	nfs_fh4_copy(&cs->saved_fh, &cs->fh);
4864 	*cs->statusp = resp->status = NFS4_OK;
4865 	cs->deleg = FALSE;
4866 
4867 out:
4868 	DTRACE_NFSV4_2(op__restorefh__done, struct compound_state *, cs,
4869 	    RESTOREFH4res *, resp);
4870 }
4871 
4872 /* ARGSUSED */
4873 static void
4874 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4875     struct compound_state *cs)
4876 {
4877 	SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh;
4878 
4879 	DTRACE_NFSV4_1(op__savefh__start, struct compound_state *, cs);
4880 
4881 	/* No need to check cs->access - we are not accessing any object */
4882 	if (cs->vp == NULL) {
4883 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4884 		goto out;
4885 	}
4886 	if (cs->saved_vp != NULL) {
4887 		VN_RELE(cs->saved_vp);
4888 	}
4889 	cs->saved_vp = cs->vp;
4890 	VN_HOLD(cs->saved_vp);
4891 	cs->saved_exi = cs->exi;
4892 	/*
4893 	 * since SAVEFH is fairly rare, don't alloc space for its fh
4894 	 * unless necessary.
4895 	 */
4896 	if (cs->saved_fh.nfs_fh4_val == NULL) {
4897 		cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP);
4898 	}
4899 	nfs_fh4_copy(&cs->fh, &cs->saved_fh);
4900 	*cs->statusp = resp->status = NFS4_OK;
4901 
4902 out:
4903 	DTRACE_NFSV4_2(op__savefh__done, struct compound_state *, cs,
4904 	    SAVEFH4res *, resp);
4905 }
4906 
4907 /*
4908  * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to
4909  * return the bitmap of attrs that were set successfully. It is also
4910  * called by Verify/Nverify to test the vattr/vfsstat attrs. It should
4911  * always be called only after rfs4_do_set_attrs().
4912  *
4913  * Verify that the attributes are same as the expected ones. sargp->vap
4914  * and sargp->sbp contain the input attributes as translated from fattr4.
4915  *
4916  * This function verifies only the attrs that correspond to a vattr or
4917  * vfsstat struct. That is because of the extra step needed to get the
4918  * corresponding system structs. Other attributes have already been set or
4919  * verified by do_rfs4_set_attrs.
4920  *
4921  * Return 0 if all attrs match, -1 if some don't, error if error processing.
4922  */
4923 static int
4924 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp,
4925     bitmap4 *resp, struct nfs4_ntov_table *ntovp)
4926 {
4927 	int error, ret_error = 0;
4928 	int i, k;
4929 	uint_t sva_mask = sargp->vap->va_mask;
4930 	uint_t vbit;
4931 	union nfs4_attr_u *na;
4932 	uint8_t *amap;
4933 	bool_t getsb = ntovp->vfsstat;
4934 
4935 	if (sva_mask != 0) {
4936 		/*
4937 		 * Okay to overwrite sargp->vap because we verify based
4938 		 * on the incoming values.
4939 		 */
4940 		ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0,
4941 		    sargp->cs->cr, NULL);
4942 		if (ret_error) {
4943 			if (resp == NULL)
4944 				return (ret_error);
4945 			/*
4946 			 * Must return bitmap of successful attrs
4947 			 */
4948 			sva_mask = 0;	/* to prevent checking vap later */
4949 		} else {
4950 			/*
4951 			 * Some file systems clobber va_mask. it is probably
4952 			 * wrong of them to do so, nonethless we practice
4953 			 * defensive coding.
4954 			 * See bug id 4276830.
4955 			 */
4956 			sargp->vap->va_mask = sva_mask;
4957 		}
4958 	}
4959 
4960 	if (getsb) {
4961 		/*
4962 		 * Now get the superblock and loop on the bitmap, as there is
4963 		 * no simple way of translating from superblock to bitmap4.
4964 		 */
4965 		ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp);
4966 		if (ret_error) {
4967 			if (resp == NULL)
4968 				goto errout;
4969 			getsb = FALSE;
4970 		}
4971 	}
4972 
4973 	/*
4974 	 * Now loop and verify each attribute which getattr returned
4975 	 * whether it's the same as the input.
4976 	 */
4977 	if (resp == NULL && !getsb && (sva_mask == 0))
4978 		goto errout;
4979 
4980 	na = ntovp->na;
4981 	amap = ntovp->amap;
4982 	k = 0;
4983 	for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) {
4984 		k = *amap;
4985 		ASSERT(nfs4_ntov_map[k].nval == k);
4986 		vbit = nfs4_ntov_map[k].vbit;
4987 
4988 		/*
4989 		 * If vattr attribute but VOP_GETATTR failed, or it's
4990 		 * superblock attribute but VFS_STATVFS failed, skip
4991 		 */
4992 		if (vbit) {
4993 			if ((vbit & sva_mask) == 0)
4994 				continue;
4995 		} else if (!(getsb && nfs4_ntov_map[k].vfsstat)) {
4996 			continue;
4997 		}
4998 		error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na);
4999 		if (resp != NULL) {
5000 			if (error)
5001 				ret_error = -1;	/* not all match */
5002 			else	/* update response bitmap */
5003 				*resp |= nfs4_ntov_map[k].fbit;
5004 			continue;
5005 		}
5006 		if (error) {
5007 			ret_error = -1;	/* not all match */
5008 			break;
5009 		}
5010 	}
5011 errout:
5012 	return (ret_error);
5013 }
5014 
5015 /*
5016  * Decode the attribute to be set/verified. If the attr requires a sys op
5017  * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't
5018  * call the sv_getit function for it, because the sys op hasn't yet been done.
5019  * Return 0 for success, error code if failed.
5020  *
5021  * Note: the decoded arg is not freed here but in nfs4_ntov_table_free.
5022  */
5023 static int
5024 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp,
5025     int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap)
5026 {
5027 	int error = 0;
5028 	bool_t set_later;
5029 
5030 	sargp->vap->va_mask |= nfs4_ntov_map[k].vbit;
5031 
5032 	if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) {
5033 		set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat;
5034 		/*
5035 		 * don't verify yet if a vattr or sb dependent attr,
5036 		 * because we don't have their sys values yet.
5037 		 * Will be done later.
5038 		 */
5039 		if (! (set_later && (cmd == NFS4ATTR_VERIT))) {
5040 			/*
5041 			 * ACLs are a special case, since setting the MODE
5042 			 * conflicts with setting the ACL.  We delay setting
5043 			 * the ACL until all other attributes have been set.
5044 			 * The ACL gets set in do_rfs4_op_setattr().
5045 			 */
5046 			if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) {
5047 				error = (*nfs4_ntov_map[k].sv_getit)(cmd,
5048 				    sargp, nap);
5049 				if (error) {
5050 					xdr_free(nfs4_ntov_map[k].xfunc,
5051 					    (caddr_t)nap);
5052 				}
5053 			}
5054 		}
5055 	} else {
5056 #ifdef  DEBUG
5057 		cmn_err(CE_NOTE, "decode_fattr4_attr: error "
5058 		    "decoding attribute %d\n", k);
5059 #endif
5060 		error = EINVAL;
5061 	}
5062 	if (!error && resp_bval && !set_later) {
5063 		*resp_bval |= nfs4_ntov_map[k].fbit;
5064 	}
5065 
5066 	return (error);
5067 }
5068 
5069 /*
5070  * Set vattr based on incoming fattr4 attrs - used by setattr.
5071  * Set response mask. Ignore any values that are not writable vattr attrs.
5072  */
5073 static nfsstat4
5074 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
5075     struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp,
5076     nfs4_attr_cmd_t cmd)
5077 {
5078 	int error = 0;
5079 	int i;
5080 	char *attrs = fattrp->attrlist4;
5081 	uint32_t attrslen = fattrp->attrlist4_len;
5082 	XDR xdr;
5083 	nfsstat4 status = NFS4_OK;
5084 	vnode_t *vp = cs->vp;
5085 	union nfs4_attr_u *na;
5086 	uint8_t *amap;
5087 
5088 #ifndef lint
5089 	/*
5090 	 * Make sure that maximum attribute number can be expressed as an
5091 	 * 8 bit quantity.
5092 	 */
5093 	ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1));
5094 #endif
5095 
5096 	if (vp == NULL) {
5097 		if (resp)
5098 			*resp = 0;
5099 		return (NFS4ERR_NOFILEHANDLE);
5100 	}
5101 	if (cs->access == CS_ACCESS_DENIED) {
5102 		if (resp)
5103 			*resp = 0;
5104 		return (NFS4ERR_ACCESS);
5105 	}
5106 
5107 	sargp->op = cmd;
5108 	sargp->cs = cs;
5109 	sargp->flag = 0;	/* may be set later */
5110 	sargp->vap->va_mask = 0;
5111 	sargp->rdattr_error = NFS4_OK;
5112 	sargp->rdattr_error_req = FALSE;
5113 	/* sargp->sbp is set by the caller */
5114 
5115 	xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE);
5116 
5117 	na = ntovp->na;
5118 	amap = ntovp->amap;
5119 
5120 	/*
5121 	 * The following loop iterates on the nfs4_ntov_map checking
5122 	 * if the fbit is set in the requested bitmap.
5123 	 * If set then we process the arguments using the
5124 	 * rfs4_fattr4 conversion functions to populate the setattr
5125 	 * vattr and va_mask. Any settable attrs that are not using vattr
5126 	 * will be set in this loop.
5127 	 */
5128 	for (i = 0; i < nfs4_ntov_map_size; i++) {
5129 		if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) {
5130 			continue;
5131 		}
5132 		/*
5133 		 * If setattr, must be a writable attr.
5134 		 * If verify/nverify, must be a readable attr.
5135 		 */
5136 		if ((error = (*nfs4_ntov_map[i].sv_getit)(
5137 		    NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) {
5138 			/*
5139 			 * Client tries to set/verify an
5140 			 * unsupported attribute, tries to set
5141 			 * a read only attr or verify a write
5142 			 * only one - error!
5143 			 */
5144 			break;
5145 		}
5146 		/*
5147 		 * Decode the attribute to set/verify
5148 		 */
5149 		error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval,
5150 		    &xdr, resp ? resp : NULL, na);
5151 		if (error)
5152 			break;
5153 		*amap++ = (uint8_t)nfs4_ntov_map[i].nval;
5154 		na++;
5155 		(ntovp->attrcnt)++;
5156 		if (nfs4_ntov_map[i].vfsstat)
5157 			ntovp->vfsstat = TRUE;
5158 	}
5159 
5160 	if (error != 0)
5161 		status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP :
5162 		    puterrno4(error));
5163 	/* xdrmem_destroy(&xdrs); */	/* NO-OP */
5164 	return (status);
5165 }
5166 
5167 static nfsstat4
5168 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
5169     stateid4 *stateid)
5170 {
5171 	int error = 0;
5172 	struct nfs4_svgetit_arg sarg;
5173 	bool_t trunc;
5174 
5175 	nfsstat4 status = NFS4_OK;
5176 	cred_t *cr = cs->cr;
5177 	vnode_t *vp = cs->vp;
5178 	struct nfs4_ntov_table ntov;
5179 	struct statvfs64 sb;
5180 	struct vattr bva;
5181 	struct flock64 bf;
5182 	int in_crit = 0;
5183 	uint_t saved_mask = 0;
5184 	caller_context_t ct;
5185 
5186 	*resp = 0;
5187 	sarg.sbp = &sb;
5188 	sarg.is_referral = B_FALSE;
5189 	nfs4_ntov_table_init(&ntov);
5190 	status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov,
5191 	    NFS4ATTR_SETIT);
5192 	if (status != NFS4_OK) {
5193 		/*
5194 		 * failed set attrs
5195 		 */
5196 		goto done;
5197 	}
5198 	if ((sarg.vap->va_mask == 0) &&
5199 	    (! (fattrp->attrmask & FATTR4_ACL_MASK))) {
5200 		/*
5201 		 * no further work to be done
5202 		 */
5203 		goto done;
5204 	}
5205 
5206 	/*
5207 	 * If we got a request to set the ACL and the MODE, only
5208 	 * allow changing VSUID, VSGID, and VSVTX.  Attempting
5209 	 * to change any other bits, along with setting an ACL,
5210 	 * gives NFS4ERR_INVAL.
5211 	 */
5212 	if ((fattrp->attrmask & FATTR4_ACL_MASK) &&
5213 	    (fattrp->attrmask & FATTR4_MODE_MASK)) {
5214 		vattr_t va;
5215 
5216 		va.va_mask = AT_MODE;
5217 		error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL);
5218 		if (error) {
5219 			status = puterrno4(error);
5220 			goto done;
5221 		}
5222 		if ((sarg.vap->va_mode ^ va.va_mode) &
5223 		    ~(VSUID | VSGID | VSVTX)) {
5224 			status = NFS4ERR_INVAL;
5225 			goto done;
5226 		}
5227 	}
5228 
5229 	/* Check stateid only if size has been set */
5230 	if (sarg.vap->va_mask & AT_SIZE) {
5231 		trunc = (sarg.vap->va_size == 0);
5232 		status = rfs4_check_stateid(FWRITE, cs->vp, stateid,
5233 		    trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct);
5234 		if (status != NFS4_OK)
5235 			goto done;
5236 	} else {
5237 		ct.cc_sysid = 0;
5238 		ct.cc_pid = 0;
5239 		ct.cc_caller_id = nfs4_srv_caller_id;
5240 		ct.cc_flags = CC_DONTBLOCK;
5241 	}
5242 
5243 	/* XXX start of possible race with delegations */
5244 
5245 	/*
5246 	 * We need to specially handle size changes because it is
5247 	 * possible for the client to create a file with read-only
5248 	 * modes, but with the file opened for writing. If the client
5249 	 * then tries to set the file size, e.g. ftruncate(3C),
5250 	 * fcntl(F_FREESP), the normal access checking done in
5251 	 * VOP_SETATTR would prevent the client from doing it even though
5252 	 * it should be allowed to do so.  To get around this, we do the
5253 	 * access checking for ourselves and use VOP_SPACE which doesn't
5254 	 * do the access checking.
5255 	 * Also the client should not be allowed to change the file
5256 	 * size if there is a conflicting non-blocking mandatory lock in
5257 	 * the region of the change.
5258 	 */
5259 	if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) {
5260 		u_offset_t offset;
5261 		ssize_t length;
5262 
5263 		/*
5264 		 * ufs_setattr clears AT_SIZE from vap->va_mask, but
5265 		 * before returning, sarg.vap->va_mask is used to
5266 		 * generate the setattr reply bitmap.  We also clear
5267 		 * AT_SIZE below before calling VOP_SPACE.  For both
5268 		 * of these cases, the va_mask needs to be saved here
5269 		 * and restored after calling VOP_SETATTR.
5270 		 */
5271 		saved_mask = sarg.vap->va_mask;
5272 
5273 		/*
5274 		 * Check any possible conflict due to NBMAND locks.
5275 		 * Get into critical region before VOP_GETATTR, so the
5276 		 * size attribute is valid when checking conflicts.
5277 		 */
5278 		if (nbl_need_check(vp)) {
5279 			nbl_start_crit(vp, RW_READER);
5280 			in_crit = 1;
5281 		}
5282 
5283 		bva.va_mask = AT_UID|AT_SIZE;
5284 		if (error = VOP_GETATTR(vp, &bva, 0, cr, &ct)) {
5285 			status = puterrno4(error);
5286 			goto done;
5287 		}
5288 
5289 		if (in_crit) {
5290 			if (sarg.vap->va_size < bva.va_size) {
5291 				offset = sarg.vap->va_size;
5292 				length = bva.va_size - sarg.vap->va_size;
5293 			} else {
5294 				offset = bva.va_size;
5295 				length = sarg.vap->va_size - bva.va_size;
5296 			}
5297 			if (nbl_conflict(vp, NBL_WRITE, offset, length, 0,
5298 			    &ct)) {
5299 				status = NFS4ERR_LOCKED;
5300 				goto done;
5301 			}
5302 		}
5303 
5304 		if (crgetuid(cr) == bva.va_uid) {
5305 			sarg.vap->va_mask &= ~AT_SIZE;
5306 			bf.l_type = F_WRLCK;
5307 			bf.l_whence = 0;
5308 			bf.l_start = (off64_t)sarg.vap->va_size;
5309 			bf.l_len = 0;
5310 			bf.l_sysid = 0;
5311 			bf.l_pid = 0;
5312 			error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE,
5313 			    (offset_t)sarg.vap->va_size, cr, &ct);
5314 		}
5315 	}
5316 
5317 	if (!error && sarg.vap->va_mask != 0)
5318 		error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct);
5319 
5320 	/* restore va_mask -- ufs_setattr clears AT_SIZE */
5321 	if (saved_mask & AT_SIZE)
5322 		sarg.vap->va_mask |= AT_SIZE;
5323 
5324 	/*
5325 	 * If an ACL was being set, it has been delayed until now,
5326 	 * in order to set the mode (via the VOP_SETATTR() above) first.
5327 	 */
5328 	if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) {
5329 		int i;
5330 
5331 		for (i = 0; i < NFS4_MAXNUM_ATTRS; i++)
5332 			if (ntov.amap[i] == FATTR4_ACL)
5333 				break;
5334 		if (i < NFS4_MAXNUM_ATTRS) {
5335 			error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)(
5336 			    NFS4ATTR_SETIT, &sarg, &ntov.na[i]);
5337 			if (error == 0) {
5338 				*resp |= FATTR4_ACL_MASK;
5339 			} else if (error == ENOTSUP) {
5340 				(void) rfs4_verify_attr(&sarg, resp, &ntov);
5341 				status = NFS4ERR_ATTRNOTSUPP;
5342 				goto done;
5343 			}
5344 		} else {
5345 			NFS4_DEBUG(rfs4_debug,
5346 			    (CE_NOTE, "do_rfs4_op_setattr: "
5347 			    "unable to find ACL in fattr4"));
5348 			error = EINVAL;
5349 		}
5350 	}
5351 
5352 	if (error) {
5353 		/* check if a monitor detected a delegation conflict */
5354 		if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK))
5355 			status = NFS4ERR_DELAY;
5356 		else
5357 			status = puterrno4(error);
5358 
5359 		/*
5360 		 * Set the response bitmap when setattr failed.
5361 		 * If VOP_SETATTR partially succeeded, test by doing a
5362 		 * VOP_GETATTR on the object and comparing the data
5363 		 * to the setattr arguments.
5364 		 */
5365 		(void) rfs4_verify_attr(&sarg, resp, &ntov);
5366 	} else {
5367 		/*
5368 		 * Force modified metadata out to stable storage.
5369 		 */
5370 		(void) VOP_FSYNC(vp, FNODSYNC, cr, &ct);
5371 		/*
5372 		 * Set response bitmap
5373 		 */
5374 		nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp);
5375 	}
5376 
5377 /* Return early and already have a NFSv4 error */
5378 done:
5379 	/*
5380 	 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr
5381 	 * conversion sets both readable and writeable NFS4 attrs
5382 	 * for AT_MTIME and AT_ATIME.  The line below masks out
5383 	 * unrequested attrs from the setattr result bitmap.  This
5384 	 * is placed after the done: label to catch the ATTRNOTSUP
5385 	 * case.
5386 	 */
5387 	*resp &= fattrp->attrmask;
5388 
5389 	if (in_crit)
5390 		nbl_end_crit(vp);
5391 
5392 	nfs4_ntov_table_free(&ntov, &sarg);
5393 
5394 	return (status);
5395 }
5396 
5397 /* ARGSUSED */
5398 static void
5399 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5400     struct compound_state *cs)
5401 {
5402 	SETATTR4args *args = &argop->nfs_argop4_u.opsetattr;
5403 	SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr;
5404 	bslabel_t *clabel;
5405 
5406 	DTRACE_NFSV4_2(op__setattr__start, struct compound_state *, cs,
5407 	    SETATTR4args *, args);
5408 
5409 	if (cs->vp == NULL) {
5410 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5411 		goto out;
5412 	}
5413 
5414 	/*
5415 	 * If there is an unshared filesystem mounted on this vnode,
5416 	 * do not allow to setattr on this vnode.
5417 	 */
5418 	if (vn_ismntpt(cs->vp)) {
5419 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5420 		goto out;
5421 	}
5422 
5423 	resp->attrsset = 0;
5424 
5425 	if (rdonly4(req, cs)) {
5426 		*cs->statusp = resp->status = NFS4ERR_ROFS;
5427 		goto out;
5428 	}
5429 
5430 	/* check label before setting attributes */
5431 	if (is_system_labeled()) {
5432 		ASSERT(req->rq_label != NULL);
5433 		clabel = req->rq_label;
5434 		DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *,
5435 		    "got client label from request(1)",
5436 		    struct svc_req *, req);
5437 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
5438 			if (!do_rfs_label_check(clabel, cs->vp,
5439 			    EQUALITY_CHECK, cs->exi)) {
5440 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
5441 				goto out;
5442 			}
5443 		}
5444 	}
5445 
5446 	*cs->statusp = resp->status =
5447 	    do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs,
5448 	    &args->stateid);
5449 
5450 out:
5451 	DTRACE_NFSV4_2(op__setattr__done, struct compound_state *, cs,
5452 	    SETATTR4res *, resp);
5453 }
5454 
5455 /* ARGSUSED */
5456 static void
5457 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5458     struct compound_state *cs)
5459 {
5460 	/*
5461 	 * verify and nverify are exactly the same, except that nverify
5462 	 * succeeds when some argument changed, and verify succeeds when
5463 	 * when none changed.
5464 	 */
5465 
5466 	VERIFY4args  *args = &argop->nfs_argop4_u.opverify;
5467 	VERIFY4res *resp = &resop->nfs_resop4_u.opverify;
5468 
5469 	int error;
5470 	struct nfs4_svgetit_arg sarg;
5471 	struct statvfs64 sb;
5472 	struct nfs4_ntov_table ntov;
5473 
5474 	DTRACE_NFSV4_2(op__verify__start, struct compound_state *, cs,
5475 	    VERIFY4args *, args);
5476 
5477 	if (cs->vp == NULL) {
5478 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5479 		goto out;
5480 	}
5481 
5482 	sarg.sbp = &sb;
5483 	sarg.is_referral = B_FALSE;
5484 	nfs4_ntov_table_init(&ntov);
5485 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
5486 	    &sarg, &ntov, NFS4ATTR_VERIT);
5487 	if (resp->status != NFS4_OK) {
5488 		/*
5489 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
5490 		 * so could return -1 for "no match".
5491 		 */
5492 		if (resp->status == -1)
5493 			resp->status = NFS4ERR_NOT_SAME;
5494 		goto done;
5495 	}
5496 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
5497 	switch (error) {
5498 	case 0:
5499 		resp->status = NFS4_OK;
5500 		break;
5501 	case -1:
5502 		resp->status = NFS4ERR_NOT_SAME;
5503 		break;
5504 	default:
5505 		resp->status = puterrno4(error);
5506 		break;
5507 	}
5508 done:
5509 	*cs->statusp = resp->status;
5510 	nfs4_ntov_table_free(&ntov, &sarg);
5511 out:
5512 	DTRACE_NFSV4_2(op__verify__done, struct compound_state *, cs,
5513 	    VERIFY4res *, resp);
5514 }
5515 
5516 /* ARGSUSED */
5517 static void
5518 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5519     struct compound_state *cs)
5520 {
5521 	/*
5522 	 * verify and nverify are exactly the same, except that nverify
5523 	 * succeeds when some argument changed, and verify succeeds when
5524 	 * when none changed.
5525 	 */
5526 
5527 	NVERIFY4args  *args = &argop->nfs_argop4_u.opnverify;
5528 	NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify;
5529 
5530 	int error;
5531 	struct nfs4_svgetit_arg sarg;
5532 	struct statvfs64 sb;
5533 	struct nfs4_ntov_table ntov;
5534 
5535 	DTRACE_NFSV4_2(op__nverify__start, struct compound_state *, cs,
5536 	    NVERIFY4args *, args);
5537 
5538 	if (cs->vp == NULL) {
5539 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5540 		DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs,
5541 		    NVERIFY4res *, resp);
5542 		return;
5543 	}
5544 	sarg.sbp = &sb;
5545 	sarg.is_referral = B_FALSE;
5546 	nfs4_ntov_table_init(&ntov);
5547 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
5548 	    &sarg, &ntov, NFS4ATTR_VERIT);
5549 	if (resp->status != NFS4_OK) {
5550 		/*
5551 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
5552 		 * so could return -1 for "no match".
5553 		 */
5554 		if (resp->status == -1)
5555 			resp->status = NFS4_OK;
5556 		goto done;
5557 	}
5558 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
5559 	switch (error) {
5560 	case 0:
5561 		resp->status = NFS4ERR_SAME;
5562 		break;
5563 	case -1:
5564 		resp->status = NFS4_OK;
5565 		break;
5566 	default:
5567 		resp->status = puterrno4(error);
5568 		break;
5569 	}
5570 done:
5571 	*cs->statusp = resp->status;
5572 	nfs4_ntov_table_free(&ntov, &sarg);
5573 
5574 	DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs,
5575 	    NVERIFY4res *, resp);
5576 }
5577 
5578 /*
5579  * XXX - This should live in an NFS header file.
5580  */
5581 #define	MAX_IOVECS	12
5582 
5583 /* ARGSUSED */
5584 static void
5585 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5586     struct compound_state *cs)
5587 {
5588 	WRITE4args *args = &argop->nfs_argop4_u.opwrite;
5589 	WRITE4res *resp = &resop->nfs_resop4_u.opwrite;
5590 	int error;
5591 	vnode_t *vp;
5592 	struct vattr bva;
5593 	u_offset_t rlimit;
5594 	struct uio uio;
5595 	struct iovec iov[MAX_IOVECS];
5596 	struct iovec *iovp;
5597 	int iovcnt;
5598 	int ioflag;
5599 	cred_t *savecred, *cr;
5600 	bool_t *deleg = &cs->deleg;
5601 	nfsstat4 stat;
5602 	int in_crit = 0;
5603 	caller_context_t ct;
5604 
5605 	DTRACE_NFSV4_2(op__write__start, struct compound_state *, cs,
5606 	    WRITE4args *, args);
5607 
5608 	vp = cs->vp;
5609 	if (vp == NULL) {
5610 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5611 		goto out;
5612 	}
5613 	if (cs->access == CS_ACCESS_DENIED) {
5614 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5615 		goto out;
5616 	}
5617 
5618 	cr = cs->cr;
5619 
5620 	if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE,
5621 	    deleg, TRUE, &ct)) != NFS4_OK) {
5622 		*cs->statusp = resp->status = stat;
5623 		goto out;
5624 	}
5625 
5626 	/*
5627 	 * We have to enter the critical region before calling VOP_RWLOCK
5628 	 * to avoid a deadlock with ufs.
5629 	 */
5630 	if (nbl_need_check(vp)) {
5631 		nbl_start_crit(vp, RW_READER);
5632 		in_crit = 1;
5633 		if (nbl_conflict(vp, NBL_WRITE,
5634 		    args->offset, args->data_len, 0, &ct)) {
5635 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
5636 			goto out;
5637 		}
5638 	}
5639 
5640 	bva.va_mask = AT_MODE | AT_UID;
5641 	error = VOP_GETATTR(vp, &bva, 0, cr, &ct);
5642 
5643 	/*
5644 	 * If we can't get the attributes, then we can't do the
5645 	 * right access checking.  So, we'll fail the request.
5646 	 */
5647 	if (error) {
5648 		*cs->statusp = resp->status = puterrno4(error);
5649 		goto out;
5650 	}
5651 
5652 	if (rdonly4(req, cs)) {
5653 		*cs->statusp = resp->status = NFS4ERR_ROFS;
5654 		goto out;
5655 	}
5656 
5657 	if (vp->v_type != VREG) {
5658 		*cs->statusp = resp->status =
5659 		    ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
5660 		goto out;
5661 	}
5662 
5663 	if (crgetuid(cr) != bva.va_uid &&
5664 	    (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) {
5665 		*cs->statusp = resp->status = puterrno4(error);
5666 		goto out;
5667 	}
5668 
5669 	if (MANDLOCK(vp, bva.va_mode)) {
5670 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5671 		goto out;
5672 	}
5673 
5674 	if (args->data_len == 0) {
5675 		*cs->statusp = resp->status = NFS4_OK;
5676 		resp->count = 0;
5677 		resp->committed = args->stable;
5678 		resp->writeverf = Write4verf;
5679 		goto out;
5680 	}
5681 
5682 	if (args->mblk != NULL) {
5683 		mblk_t *m;
5684 		uint_t bytes, round_len;
5685 
5686 		iovcnt = 0;
5687 		bytes = 0;
5688 		round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT);
5689 		for (m = args->mblk;
5690 		    m != NULL && bytes < round_len;
5691 		    m = m->b_cont) {
5692 			iovcnt++;
5693 			bytes += MBLKL(m);
5694 		}
5695 #ifdef DEBUG
5696 		/* should have ended on an mblk boundary */
5697 		if (bytes != round_len) {
5698 			printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n",
5699 			    bytes, round_len, args->data_len);
5700 			printf("args=%p, args->mblk=%p, m=%p", (void *)args,
5701 			    (void *)args->mblk, (void *)m);
5702 			ASSERT(bytes == round_len);
5703 		}
5704 #endif
5705 		if (iovcnt <= MAX_IOVECS) {
5706 			iovp = iov;
5707 		} else {
5708 			iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP);
5709 		}
5710 		mblk_to_iov(args->mblk, iovcnt, iovp);
5711 	} else if (args->rlist != NULL) {
5712 		iovcnt = 1;
5713 		iovp = iov;
5714 		iovp->iov_base = (char *)((args->rlist)->u.c_daddr3);
5715 		iovp->iov_len = args->data_len;
5716 	} else {
5717 		iovcnt = 1;
5718 		iovp = iov;
5719 		iovp->iov_base = args->data_val;
5720 		iovp->iov_len = args->data_len;
5721 	}
5722 
5723 	uio.uio_iov = iovp;
5724 	uio.uio_iovcnt = iovcnt;
5725 
5726 	uio.uio_segflg = UIO_SYSSPACE;
5727 	uio.uio_extflg = UIO_COPY_DEFAULT;
5728 	uio.uio_loffset = args->offset;
5729 	uio.uio_resid = args->data_len;
5730 	uio.uio_llimit = curproc->p_fsz_ctl;
5731 	rlimit = uio.uio_llimit - args->offset;
5732 	if (rlimit < (u_offset_t)uio.uio_resid)
5733 		uio.uio_resid = (int)rlimit;
5734 
5735 	if (args->stable == UNSTABLE4)
5736 		ioflag = 0;
5737 	else if (args->stable == FILE_SYNC4)
5738 		ioflag = FSYNC;
5739 	else if (args->stable == DATA_SYNC4)
5740 		ioflag = FDSYNC;
5741 	else {
5742 		if (iovp != iov)
5743 			kmem_free(iovp, sizeof (*iovp) * iovcnt);
5744 		*cs->statusp = resp->status = NFS4ERR_INVAL;
5745 		goto out;
5746 	}
5747 
5748 	/*
5749 	 * We're changing creds because VM may fault and we need
5750 	 * the cred of the current thread to be used if quota
5751 	 * checking is enabled.
5752 	 */
5753 	savecred = curthread->t_cred;
5754 	curthread->t_cred = cr;
5755 	error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct);
5756 	curthread->t_cred = savecred;
5757 
5758 	if (iovp != iov)
5759 		kmem_free(iovp, sizeof (*iovp) * iovcnt);
5760 
5761 	if (error) {
5762 		*cs->statusp = resp->status = puterrno4(error);
5763 		goto out;
5764 	}
5765 
5766 	*cs->statusp = resp->status = NFS4_OK;
5767 	resp->count = args->data_len - uio.uio_resid;
5768 
5769 	if (ioflag == 0)
5770 		resp->committed = UNSTABLE4;
5771 	else
5772 		resp->committed = FILE_SYNC4;
5773 
5774 	resp->writeverf = Write4verf;
5775 
5776 out:
5777 	if (in_crit)
5778 		nbl_end_crit(vp);
5779 
5780 	DTRACE_NFSV4_2(op__write__done, struct compound_state *, cs,
5781 	    WRITE4res *, resp);
5782 }
5783 
5784 
5785 /* XXX put in a header file */
5786 extern int	sec_svc_getcred(struct svc_req *, cred_t *,  caddr_t *, int *);
5787 
5788 void
5789 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi,
5790     struct svc_req *req, cred_t *cr, int *rv)
5791 {
5792 	uint_t i;
5793 	struct compound_state cs;
5794 
5795 	if (rv != NULL)
5796 		*rv = 0;
5797 	rfs4_init_compound_state(&cs);
5798 	/*
5799 	 * Form a reply tag by copying over the reqeuest tag.
5800 	 */
5801 	resp->tag.utf8string_val =
5802 	    kmem_alloc(args->tag.utf8string_len, KM_SLEEP);
5803 	resp->tag.utf8string_len = args->tag.utf8string_len;
5804 	bcopy(args->tag.utf8string_val, resp->tag.utf8string_val,
5805 	    resp->tag.utf8string_len);
5806 
5807 	cs.statusp = &resp->status;
5808 	cs.req = req;
5809 	resp->array = NULL;
5810 	resp->array_len = 0;
5811 
5812 	/*
5813 	 * XXX for now, minorversion should be zero
5814 	 */
5815 	if (args->minorversion != NFS4_MINORVERSION) {
5816 		DTRACE_NFSV4_2(compound__start, struct compound_state *,
5817 		    &cs, COMPOUND4args *, args);
5818 		resp->status = NFS4ERR_MINOR_VERS_MISMATCH;
5819 		DTRACE_NFSV4_2(compound__done, struct compound_state *,
5820 		    &cs, COMPOUND4res *, resp);
5821 		return;
5822 	}
5823 
5824 	if (args->array_len == 0) {
5825 		resp->status = NFS4_OK;
5826 		return;
5827 	}
5828 
5829 	ASSERT(exi == NULL);
5830 	ASSERT(cr == NULL);
5831 
5832 	cr = crget();
5833 	ASSERT(cr != NULL);
5834 
5835 	if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) {
5836 		DTRACE_NFSV4_2(compound__start, struct compound_state *,
5837 		    &cs, COMPOUND4args *, args);
5838 		crfree(cr);
5839 		DTRACE_NFSV4_2(compound__done, struct compound_state *,
5840 		    &cs, COMPOUND4res *, resp);
5841 		svcerr_badcred(req->rq_xprt);
5842 		if (rv != NULL)
5843 			*rv = 1;
5844 		return;
5845 	}
5846 	resp->array_len = args->array_len;
5847 	resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4),
5848 	    KM_SLEEP);
5849 
5850 	cs.basecr = cr;
5851 
5852 	DTRACE_NFSV4_2(compound__start, struct compound_state *, &cs,
5853 	    COMPOUND4args *, args);
5854 
5855 	/*
5856 	 * For now, NFS4 compound processing must be protected by
5857 	 * exported_lock because it can access more than one exportinfo
5858 	 * per compound and share/unshare can now change multiple
5859 	 * exinfo structs.  The NFS2/3 code only refs 1 exportinfo
5860 	 * per proc (excluding public exinfo), and exi_count design
5861 	 * is sufficient to protect concurrent execution of NFS2/3
5862 	 * ops along with unexport.  This lock will be removed as
5863 	 * part of the NFSv4 phase 2 namespace redesign work.
5864 	 */
5865 	rw_enter(&exported_lock, RW_READER);
5866 
5867 	/*
5868 	 * If this is the first compound we've seen, we need to start all
5869 	 * new instances' grace periods.
5870 	 */
5871 	if (rfs4_seen_first_compound == 0) {
5872 		rfs4_grace_start_new();
5873 		/*
5874 		 * This must be set after rfs4_grace_start_new(), otherwise
5875 		 * another thread could proceed past here before the former
5876 		 * is finished.
5877 		 */
5878 		rfs4_seen_first_compound = 1;
5879 	}
5880 
5881 	for (i = 0; i < args->array_len && cs.cont; i++) {
5882 		nfs_argop4 *argop;
5883 		nfs_resop4 *resop;
5884 		uint_t op;
5885 
5886 		argop = &args->array[i];
5887 		resop = &resp->array[i];
5888 		resop->resop = argop->argop;
5889 		op = (uint_t)resop->resop;
5890 
5891 		if (op < rfsv4disp_cnt) {
5892 			kstat_t *ksp = rfsprocio_v4_ptr[op];
5893 			kstat_t *exi_ksp = NULL;
5894 
5895 			/*
5896 			 * Count the individual ops here; NULL and COMPOUND
5897 			 * are counted in common_dispatch()
5898 			 */
5899 			rfsproccnt_v4_ptr[op].value.ui64++;
5900 
5901 			if (ksp != NULL) {
5902 				mutex_enter(ksp->ks_lock);
5903 				kstat_runq_enter(KSTAT_IO_PTR(ksp));
5904 				mutex_exit(ksp->ks_lock);
5905 			}
5906 
5907 			switch (rfsv4disptab[op].op_type) {
5908 			case NFS4_OP_CFH:
5909 				resop->exi = cs.exi;
5910 				break;
5911 			case NFS4_OP_SFH:
5912 				resop->exi = cs.saved_exi;
5913 				break;
5914 			default:
5915 				ASSERT(resop->exi == NULL);
5916 				break;
5917 			}
5918 
5919 			if (resop->exi != NULL) {
5920 				exi_ksp = resop->exi->exi_kstats->
5921 				    rfsprocio_v4_ptr[op];
5922 				if (exi_ksp != NULL) {
5923 					mutex_enter(exi_ksp->ks_lock);
5924 					kstat_runq_enter(KSTAT_IO_PTR(exi_ksp));
5925 					mutex_exit(exi_ksp->ks_lock);
5926 				}
5927 			}
5928 
5929 			NFS4_DEBUG(rfs4_debug > 1,
5930 			    (CE_NOTE, "Executing %s", rfs4_op_string[op]));
5931 			(*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs);
5932 			NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d",
5933 			    rfs4_op_string[op], *cs.statusp));
5934 			if (*cs.statusp != NFS4_OK)
5935 				cs.cont = FALSE;
5936 
5937 			if (rfsv4disptab[op].op_type == NFS4_OP_POSTCFH &&
5938 			    *cs.statusp == NFS4_OK &&
5939 			    (resop->exi = cs.exi) != NULL) {
5940 				exi_ksp = resop->exi->exi_kstats->
5941 				    rfsprocio_v4_ptr[op];
5942 			}
5943 
5944 			if (exi_ksp != NULL) {
5945 				mutex_enter(exi_ksp->ks_lock);
5946 				KSTAT_IO_PTR(exi_ksp)->nwritten +=
5947 				    argop->opsize;
5948 				KSTAT_IO_PTR(exi_ksp)->writes++;
5949 				if (rfsv4disptab[op].op_type != NFS4_OP_POSTCFH)
5950 					kstat_runq_exit(KSTAT_IO_PTR(exi_ksp));
5951 				mutex_exit(exi_ksp->ks_lock);
5952 
5953 				exi_hold(resop->exi);
5954 			} else {
5955 				resop->exi = NULL;
5956 			}
5957 
5958 			if (ksp != NULL) {
5959 				mutex_enter(ksp->ks_lock);
5960 				kstat_runq_exit(KSTAT_IO_PTR(ksp));
5961 				mutex_exit(ksp->ks_lock);
5962 			}
5963 		} else {
5964 			/*
5965 			 * This is effectively dead code since XDR code
5966 			 * will have already returned BADXDR if op doesn't
5967 			 * decode to legal value.  This only done for a
5968 			 * day when XDR code doesn't verify v4 opcodes.
5969 			 */
5970 			op = OP_ILLEGAL;
5971 			rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++;
5972 
5973 			rfs4_op_illegal(argop, resop, req, &cs);
5974 			cs.cont = FALSE;
5975 		}
5976 
5977 		/*
5978 		 * If not at last op, and if we are to stop, then
5979 		 * compact the results array.
5980 		 */
5981 		if ((i + 1) < args->array_len && !cs.cont) {
5982 			nfs_resop4 *new_res = kmem_alloc(
5983 			    (i + 1) * sizeof (nfs_resop4), KM_SLEEP);
5984 			bcopy(resp->array,
5985 			    new_res, (i + 1) * sizeof (nfs_resop4));
5986 			kmem_free(resp->array,
5987 			    args->array_len * sizeof (nfs_resop4));
5988 
5989 			resp->array_len = i + 1;
5990 			resp->array = new_res;
5991 		}
5992 	}
5993 
5994 	rw_exit(&exported_lock);
5995 
5996 	DTRACE_NFSV4_2(compound__done, struct compound_state *, &cs,
5997 	    COMPOUND4res *, resp);
5998 
5999 	if (cs.vp)
6000 		VN_RELE(cs.vp);
6001 	if (cs.saved_vp)
6002 		VN_RELE(cs.saved_vp);
6003 	if (cs.saved_fh.nfs_fh4_val)
6004 		kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE);
6005 
6006 	if (cs.basecr)
6007 		crfree(cs.basecr);
6008 	if (cs.cr)
6009 		crfree(cs.cr);
6010 	/*
6011 	 * done with this compound request, free the label
6012 	 */
6013 
6014 	if (req->rq_label != NULL) {
6015 		kmem_free(req->rq_label, sizeof (bslabel_t));
6016 		req->rq_label = NULL;
6017 	}
6018 }
6019 
6020 /*
6021  * XXX because of what appears to be duplicate calls to rfs4_compound_free
6022  * XXX zero out the tag and array values. Need to investigate why the
6023  * XXX calls occur, but at least prevent the panic for now.
6024  */
6025 void
6026 rfs4_compound_free(COMPOUND4res *resp)
6027 {
6028 	uint_t i;
6029 
6030 	if (resp->tag.utf8string_val) {
6031 		UTF8STRING_FREE(resp->tag)
6032 	}
6033 
6034 	for (i = 0; i < resp->array_len; i++) {
6035 		nfs_resop4 *resop;
6036 		uint_t op;
6037 
6038 		resop = &resp->array[i];
6039 		op = (uint_t)resop->resop;
6040 		if (op < rfsv4disp_cnt) {
6041 			(*rfsv4disptab[op].dis_resfree)(resop);
6042 		}
6043 	}
6044 	if (resp->array != NULL) {
6045 		kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4));
6046 	}
6047 }
6048 
6049 /*
6050  * Process the value of the compound request rpc flags, as a bit-AND
6051  * of the individual per-op flags (idempotent, allowork, publicfh_ok)
6052  */
6053 void
6054 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp)
6055 {
6056 	int i;
6057 	int flag = RPC_ALL;
6058 
6059 	for (i = 0; flag && i < args->array_len; i++) {
6060 		uint_t op;
6061 
6062 		op = (uint_t)args->array[i].argop;
6063 
6064 		if (op < rfsv4disp_cnt)
6065 			flag &= rfsv4disptab[op].dis_flags;
6066 		else
6067 			flag = 0;
6068 	}
6069 	*flagp = flag;
6070 }
6071 
6072 void
6073 rfs4_compound_kstat_args(COMPOUND4args *args)
6074 {
6075 	int i;
6076 
6077 	for (i = 0; i < args->array_len; i++) {
6078 		uint_t op = (uint_t)args->array[i].argop;
6079 
6080 		if (op < rfsv4disp_cnt) {
6081 			kstat_t *ksp = rfsprocio_v4_ptr[op];
6082 
6083 			if (ksp != NULL) {
6084 				mutex_enter(ksp->ks_lock);
6085 				KSTAT_IO_PTR(ksp)->nwritten +=
6086 				    args->array[i].opsize;
6087 				KSTAT_IO_PTR(ksp)->writes++;
6088 				mutex_exit(ksp->ks_lock);
6089 			}
6090 		}
6091 	}
6092 }
6093 
6094 void
6095 rfs4_compound_kstat_res(COMPOUND4res *res)
6096 {
6097 	int i;
6098 
6099 	for (i = 0; i < res->array_len; i++) {
6100 		uint_t op = (uint_t)res->array[i].resop;
6101 
6102 		if (op < rfsv4disp_cnt) {
6103 			kstat_t *ksp = rfsprocio_v4_ptr[op];
6104 			struct exportinfo *exi = res->array[i].exi;
6105 
6106 			if (ksp != NULL) {
6107 				mutex_enter(ksp->ks_lock);
6108 				KSTAT_IO_PTR(ksp)->nread +=
6109 				    res->array[i].opsize;
6110 				KSTAT_IO_PTR(ksp)->reads++;
6111 				mutex_exit(ksp->ks_lock);
6112 			}
6113 
6114 			if (exi != NULL) {
6115 				kstat_t *exi_ksp;
6116 
6117 				rw_enter(&exported_lock, RW_READER);
6118 
6119 				exi_ksp = exi->exi_kstats->rfsprocio_v4_ptr[op];
6120 				if (exi_ksp != NULL) {
6121 					mutex_enter(exi_ksp->ks_lock);
6122 					KSTAT_IO_PTR(exi_ksp)->nread +=
6123 					    res->array[i].opsize;
6124 					KSTAT_IO_PTR(exi_ksp)->reads++;
6125 					mutex_exit(exi_ksp->ks_lock);
6126 				}
6127 
6128 				rw_exit(&exported_lock);
6129 
6130 				exi_rele(exi);
6131 			}
6132 		}
6133 	}
6134 }
6135 
6136 nfsstat4
6137 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp)
6138 {
6139 	nfsstat4 e;
6140 
6141 	rfs4_dbe_lock(cp->rc_dbe);
6142 
6143 	if (cp->rc_sysidt != LM_NOSYSID) {
6144 		*sp = cp->rc_sysidt;
6145 		e = NFS4_OK;
6146 
6147 	} else if ((cp->rc_sysidt = lm_alloc_sysidt()) != LM_NOSYSID) {
6148 		*sp = cp->rc_sysidt;
6149 		e = NFS4_OK;
6150 
6151 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
6152 		    "rfs4_client_sysid: allocated 0x%x\n", *sp));
6153 	} else
6154 		e = NFS4ERR_DELAY;
6155 
6156 	rfs4_dbe_unlock(cp->rc_dbe);
6157 	return (e);
6158 }
6159 
6160 #if defined(DEBUG) && ! defined(lint)
6161 static void lock_print(char *str, int operation, struct flock64 *flk)
6162 {
6163 	char *op, *type;
6164 
6165 	switch (operation) {
6166 	case F_GETLK: op = "F_GETLK";
6167 		break;
6168 	case F_SETLK: op = "F_SETLK";
6169 		break;
6170 	case F_SETLK_NBMAND: op = "F_SETLK_NBMAND";
6171 		break;
6172 	default: op = "F_UNKNOWN";
6173 		break;
6174 	}
6175 	switch (flk->l_type) {
6176 	case F_UNLCK: type = "F_UNLCK";
6177 		break;
6178 	case F_RDLCK: type = "F_RDLCK";
6179 		break;
6180 	case F_WRLCK: type = "F_WRLCK";
6181 		break;
6182 	default: type = "F_UNKNOWN";
6183 		break;
6184 	}
6185 
6186 	ASSERT(flk->l_whence == 0);
6187 	cmn_err(CE_NOTE, "%s:  %s, type = %s, off = %llx len = %llx pid = %d",
6188 	    str, op, type, (longlong_t)flk->l_start,
6189 	    flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid);
6190 }
6191 
6192 #define	LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f)
6193 #else
6194 #define	LOCK_PRINT(d, s, t, f)
6195 #endif
6196 
6197 /*ARGSUSED*/
6198 static bool_t
6199 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs)
6200 {
6201 	return (TRUE);
6202 }
6203 
6204 /*
6205  * Look up the pathname using the vp in cs as the directory vnode.
6206  * cs->vp will be the vnode for the file on success
6207  */
6208 
6209 static nfsstat4
6210 rfs4_lookup(component4 *component, struct svc_req *req,
6211     struct compound_state *cs)
6212 {
6213 	char *nm;
6214 	uint32_t len;
6215 	nfsstat4 status;
6216 	struct sockaddr *ca;
6217 	char *name;
6218 
6219 	if (cs->vp == NULL) {
6220 		return (NFS4ERR_NOFILEHANDLE);
6221 	}
6222 	if (cs->vp->v_type != VDIR) {
6223 		return (NFS4ERR_NOTDIR);
6224 	}
6225 
6226 	status = utf8_dir_verify(component);
6227 	if (status != NFS4_OK)
6228 		return (status);
6229 
6230 	nm = utf8_to_fn(component, &len, NULL);
6231 	if (nm == NULL) {
6232 		return (NFS4ERR_INVAL);
6233 	}
6234 
6235 	if (len > MAXNAMELEN) {
6236 		kmem_free(nm, len);
6237 		return (NFS4ERR_NAMETOOLONG);
6238 	}
6239 
6240 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
6241 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
6242 	    MAXPATHLEN + 1);
6243 
6244 	if (name == NULL) {
6245 		kmem_free(nm, len);
6246 		return (NFS4ERR_INVAL);
6247 	}
6248 
6249 	status = do_rfs4_op_lookup(name, req, cs);
6250 
6251 	if (name != nm)
6252 		kmem_free(name, MAXPATHLEN + 1);
6253 
6254 	kmem_free(nm, len);
6255 
6256 	return (status);
6257 }
6258 
6259 static nfsstat4
6260 rfs4_lookupfile(component4 *component, struct svc_req *req,
6261     struct compound_state *cs, uint32_t access, change_info4 *cinfo)
6262 {
6263 	nfsstat4 status;
6264 	vnode_t *dvp = cs->vp;
6265 	vattr_t bva, ava, fva;
6266 	int error;
6267 
6268 	/* Get "before" change value */
6269 	bva.va_mask = AT_CTIME|AT_SEQ;
6270 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL);
6271 	if (error)
6272 		return (puterrno4(error));
6273 
6274 	/* rfs4_lookup may VN_RELE directory */
6275 	VN_HOLD(dvp);
6276 
6277 	status = rfs4_lookup(component, req, cs);
6278 	if (status != NFS4_OK) {
6279 		VN_RELE(dvp);
6280 		return (status);
6281 	}
6282 
6283 	/*
6284 	 * Get "after" change value, if it fails, simply return the
6285 	 * before value.
6286 	 */
6287 	ava.va_mask = AT_CTIME|AT_SEQ;
6288 	if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) {
6289 		ava.va_ctime = bva.va_ctime;
6290 		ava.va_seq = 0;
6291 	}
6292 	VN_RELE(dvp);
6293 
6294 	/*
6295 	 * Validate the file is a file
6296 	 */
6297 	fva.va_mask = AT_TYPE|AT_MODE;
6298 	error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL);
6299 	if (error)
6300 		return (puterrno4(error));
6301 
6302 	if (fva.va_type != VREG) {
6303 		if (fva.va_type == VDIR)
6304 			return (NFS4ERR_ISDIR);
6305 		if (fva.va_type == VLNK)
6306 			return (NFS4ERR_SYMLINK);
6307 		return (NFS4ERR_INVAL);
6308 	}
6309 
6310 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime);
6311 	NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
6312 
6313 	/*
6314 	 * It is undefined if VOP_LOOKUP will change va_seq, so
6315 	 * cinfo.atomic = TRUE only if we have
6316 	 * non-zero va_seq's, and they have not changed.
6317 	 */
6318 	if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq)
6319 		cinfo->atomic = TRUE;
6320 	else
6321 		cinfo->atomic = FALSE;
6322 
6323 	/* Check for mandatory locking */
6324 	cs->mandlock = MANDLOCK(cs->vp, fva.va_mode);
6325 	return (check_open_access(access, cs, req));
6326 }
6327 
6328 static nfsstat4
6329 create_vnode(vnode_t *dvp, char *nm,  vattr_t *vap, createmode4 mode,
6330     timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created)
6331 {
6332 	int error;
6333 	nfsstat4 status = NFS4_OK;
6334 	vattr_t va;
6335 
6336 tryagain:
6337 
6338 	/*
6339 	 * The file open mode used is VWRITE.  If the client needs
6340 	 * some other semantic, then it should do the access checking
6341 	 * itself.  It would have been nice to have the file open mode
6342 	 * passed as part of the arguments.
6343 	 */
6344 
6345 	*created = TRUE;
6346 	error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL);
6347 
6348 	if (error) {
6349 		*created = FALSE;
6350 
6351 		/*
6352 		 * If we got something other than file already exists
6353 		 * then just return this error.  Otherwise, we got
6354 		 * EEXIST.  If we were doing a GUARDED create, then
6355 		 * just return this error.  Otherwise, we need to
6356 		 * make sure that this wasn't a duplicate of an
6357 		 * exclusive create request.
6358 		 *
6359 		 * The assumption is made that a non-exclusive create
6360 		 * request will never return EEXIST.
6361 		 */
6362 
6363 		if (error != EEXIST || mode == GUARDED4) {
6364 			status = puterrno4(error);
6365 			return (status);
6366 		}
6367 		error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr,
6368 		    NULL, NULL, NULL);
6369 
6370 		if (error) {
6371 			/*
6372 			 * We couldn't find the file that we thought that
6373 			 * we just created.  So, we'll just try creating
6374 			 * it again.
6375 			 */
6376 			if (error == ENOENT)
6377 				goto tryagain;
6378 
6379 			status = puterrno4(error);
6380 			return (status);
6381 		}
6382 
6383 		if (mode == UNCHECKED4) {
6384 			/* existing object must be regular file */
6385 			if ((*vpp)->v_type != VREG) {
6386 				if ((*vpp)->v_type == VDIR)
6387 					status = NFS4ERR_ISDIR;
6388 				else if ((*vpp)->v_type == VLNK)
6389 					status = NFS4ERR_SYMLINK;
6390 				else
6391 					status = NFS4ERR_INVAL;
6392 				VN_RELE(*vpp);
6393 				return (status);
6394 			}
6395 
6396 			return (NFS4_OK);
6397 		}
6398 
6399 		/* Check for duplicate request */
6400 		ASSERT(mtime != 0);
6401 		va.va_mask = AT_MTIME;
6402 		error = VOP_GETATTR(*vpp, &va, 0, cr, NULL);
6403 		if (!error) {
6404 			/* We found the file */
6405 			if (va.va_mtime.tv_sec != mtime->tv_sec ||
6406 			    va.va_mtime.tv_nsec != mtime->tv_nsec) {
6407 				/* but its not our creation */
6408 				VN_RELE(*vpp);
6409 				return (NFS4ERR_EXIST);
6410 			}
6411 			*created = TRUE; /* retrans of create == created */
6412 			return (NFS4_OK);
6413 		}
6414 		VN_RELE(*vpp);
6415 		return (NFS4ERR_EXIST);
6416 	}
6417 
6418 	return (NFS4_OK);
6419 }
6420 
6421 static nfsstat4
6422 check_open_access(uint32_t access, struct compound_state *cs,
6423     struct svc_req *req)
6424 {
6425 	int error;
6426 	vnode_t *vp;
6427 	bool_t readonly;
6428 	cred_t *cr = cs->cr;
6429 
6430 	/* For now we don't allow mandatory locking as per V2/V3 */
6431 	if (cs->access == CS_ACCESS_DENIED || cs->mandlock) {
6432 		return (NFS4ERR_ACCESS);
6433 	}
6434 
6435 	vp = cs->vp;
6436 	ASSERT(cr != NULL && vp->v_type == VREG);
6437 
6438 	/*
6439 	 * If the file system is exported read only and we are trying
6440 	 * to open for write, then return NFS4ERR_ROFS
6441 	 */
6442 
6443 	readonly = rdonly4(req, cs);
6444 
6445 	if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly)
6446 		return (NFS4ERR_ROFS);
6447 
6448 	if (access & OPEN4_SHARE_ACCESS_READ) {
6449 		if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) &&
6450 		    (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) {
6451 			return (NFS4ERR_ACCESS);
6452 		}
6453 	}
6454 
6455 	if (access & OPEN4_SHARE_ACCESS_WRITE) {
6456 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
6457 		if (error)
6458 			return (NFS4ERR_ACCESS);
6459 	}
6460 
6461 	return (NFS4_OK);
6462 }
6463 
6464 static nfsstat4
6465 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs,
6466     change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid)
6467 {
6468 	struct nfs4_svgetit_arg sarg;
6469 	struct nfs4_ntov_table ntov;
6470 
6471 	bool_t ntov_table_init = FALSE;
6472 	struct statvfs64 sb;
6473 	nfsstat4 status;
6474 	vnode_t *vp;
6475 	vattr_t bva, ava, iva, cva, *vap;
6476 	vnode_t *dvp;
6477 	timespec32_t *mtime;
6478 	char *nm = NULL;
6479 	uint_t buflen;
6480 	bool_t created;
6481 	bool_t setsize = FALSE;
6482 	len_t reqsize;
6483 	int error;
6484 	bool_t trunc;
6485 	caller_context_t ct;
6486 	component4 *component;
6487 	bslabel_t *clabel;
6488 	struct sockaddr *ca;
6489 	char *name = NULL;
6490 
6491 	sarg.sbp = &sb;
6492 	sarg.is_referral = B_FALSE;
6493 
6494 	dvp = cs->vp;
6495 
6496 	/* Check if the file system is read only */
6497 	if (rdonly4(req, cs))
6498 		return (NFS4ERR_ROFS);
6499 
6500 	/* check the label of including directory */
6501 	if (is_system_labeled()) {
6502 		ASSERT(req->rq_label != NULL);
6503 		clabel = req->rq_label;
6504 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
6505 		    "got client label from request(1)",
6506 		    struct svc_req *, req);
6507 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
6508 			if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK,
6509 			    cs->exi)) {
6510 				return (NFS4ERR_ACCESS);
6511 			}
6512 		}
6513 	}
6514 
6515 	/*
6516 	 * Get the last component of path name in nm. cs will reference
6517 	 * the including directory on success.
6518 	 */
6519 	component = &args->open_claim4_u.file;
6520 	status = utf8_dir_verify(component);
6521 	if (status != NFS4_OK)
6522 		return (status);
6523 
6524 	nm = utf8_to_fn(component, &buflen, NULL);
6525 
6526 	if (nm == NULL)
6527 		return (NFS4ERR_RESOURCE);
6528 
6529 	if (buflen > MAXNAMELEN) {
6530 		kmem_free(nm, buflen);
6531 		return (NFS4ERR_NAMETOOLONG);
6532 	}
6533 
6534 	bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ;
6535 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL);
6536 	if (error) {
6537 		kmem_free(nm, buflen);
6538 		return (puterrno4(error));
6539 	}
6540 
6541 	if (bva.va_type != VDIR) {
6542 		kmem_free(nm, buflen);
6543 		return (NFS4ERR_NOTDIR);
6544 	}
6545 
6546 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime)
6547 
6548 	switch (args->mode) {
6549 	case GUARDED4:
6550 		/*FALLTHROUGH*/
6551 	case UNCHECKED4:
6552 		nfs4_ntov_table_init(&ntov);
6553 		ntov_table_init = TRUE;
6554 
6555 		*attrset = 0;
6556 		status = do_rfs4_set_attrs(attrset,
6557 		    &args->createhow4_u.createattrs,
6558 		    cs, &sarg, &ntov, NFS4ATTR_SETIT);
6559 
6560 		if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) &&
6561 		    sarg.vap->va_type != VREG) {
6562 			if (sarg.vap->va_type == VDIR)
6563 				status = NFS4ERR_ISDIR;
6564 			else if (sarg.vap->va_type == VLNK)
6565 				status = NFS4ERR_SYMLINK;
6566 			else
6567 				status = NFS4ERR_INVAL;
6568 		}
6569 
6570 		if (status != NFS4_OK) {
6571 			kmem_free(nm, buflen);
6572 			nfs4_ntov_table_free(&ntov, &sarg);
6573 			*attrset = 0;
6574 			return (status);
6575 		}
6576 
6577 		vap = sarg.vap;
6578 		vap->va_type = VREG;
6579 		vap->va_mask |= AT_TYPE;
6580 
6581 		if ((vap->va_mask & AT_MODE) == 0) {
6582 			vap->va_mask |= AT_MODE;
6583 			vap->va_mode = (mode_t)0600;
6584 		}
6585 
6586 		if (vap->va_mask & AT_SIZE) {
6587 
6588 			/* Disallow create with a non-zero size */
6589 
6590 			if ((reqsize = sarg.vap->va_size) != 0) {
6591 				kmem_free(nm, buflen);
6592 				nfs4_ntov_table_free(&ntov, &sarg);
6593 				*attrset = 0;
6594 				return (NFS4ERR_INVAL);
6595 			}
6596 			setsize = TRUE;
6597 		}
6598 		break;
6599 
6600 	case EXCLUSIVE4:
6601 		/* prohibit EXCL create of named attributes */
6602 		if (dvp->v_flag & V_XATTRDIR) {
6603 			kmem_free(nm, buflen);
6604 			*attrset = 0;
6605 			return (NFS4ERR_INVAL);
6606 		}
6607 
6608 		cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE;
6609 		cva.va_type = VREG;
6610 		/*
6611 		 * Ensure no time overflows. Assumes underlying
6612 		 * filesystem supports at least 32 bits.
6613 		 * Truncate nsec to usec resolution to allow valid
6614 		 * compares even if the underlying filesystem truncates.
6615 		 */
6616 		mtime = (timespec32_t *)&args->createhow4_u.createverf;
6617 		cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX;
6618 		cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000;
6619 		cva.va_mode = (mode_t)0;
6620 		vap = &cva;
6621 
6622 		/*
6623 		 * For EXCL create, attrset is set to the server attr
6624 		 * used to cache the client's verifier.
6625 		 */
6626 		*attrset = FATTR4_TIME_MODIFY_MASK;
6627 		break;
6628 	}
6629 
6630 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
6631 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
6632 	    MAXPATHLEN  + 1);
6633 
6634 	if (name == NULL) {
6635 		kmem_free(nm, buflen);
6636 		return (NFS4ERR_SERVERFAULT);
6637 	}
6638 
6639 	status = create_vnode(dvp, name, vap, args->mode, mtime,
6640 	    cs->cr, &vp, &created);
6641 	if (nm != name)
6642 		kmem_free(name, MAXPATHLEN + 1);
6643 	kmem_free(nm, buflen);
6644 
6645 	if (status != NFS4_OK) {
6646 		if (ntov_table_init)
6647 			nfs4_ntov_table_free(&ntov, &sarg);
6648 		*attrset = 0;
6649 		return (status);
6650 	}
6651 
6652 	trunc = (setsize && !created);
6653 
6654 	if (args->mode != EXCLUSIVE4) {
6655 		bitmap4 createmask = args->createhow4_u.createattrs.attrmask;
6656 
6657 		/*
6658 		 * True verification that object was created with correct
6659 		 * attrs is impossible.  The attrs could have been changed
6660 		 * immediately after object creation.  If attributes did
6661 		 * not verify, the only recourse for the server is to
6662 		 * destroy the object.  Maybe if some attrs (like gid)
6663 		 * are set incorrectly, the object should be destroyed;
6664 		 * however, seems bad as a default policy.  Do we really
6665 		 * want to destroy an object over one of the times not
6666 		 * verifying correctly?  For these reasons, the server
6667 		 * currently sets bits in attrset for createattrs
6668 		 * that were set; however, no verification is done.
6669 		 *
6670 		 * vmask_to_nmask accounts for vattr bits set on create
6671 		 *	[do_rfs4_set_attrs() only sets resp bits for
6672 		 *	 non-vattr/vfs bits.]
6673 		 * Mask off any bits we set by default so as not to return
6674 		 * more attrset bits than were requested in createattrs
6675 		 */
6676 		if (created) {
6677 			nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset);
6678 			*attrset &= createmask;
6679 		} else {
6680 			/*
6681 			 * We did not create the vnode (we tried but it
6682 			 * already existed).  In this case, the only createattr
6683 			 * that the spec allows the server to set is size,
6684 			 * and even then, it can only be set if it is 0.
6685 			 */
6686 			*attrset = 0;
6687 			if (trunc)
6688 				*attrset = FATTR4_SIZE_MASK;
6689 		}
6690 	}
6691 	if (ntov_table_init)
6692 		nfs4_ntov_table_free(&ntov, &sarg);
6693 
6694 	/*
6695 	 * Get the initial "after" sequence number, if it fails,
6696 	 * set to zero, time to before.
6697 	 */
6698 	iva.va_mask = AT_CTIME|AT_SEQ;
6699 	if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) {
6700 		iva.va_seq = 0;
6701 		iva.va_ctime = bva.va_ctime;
6702 	}
6703 
6704 	/*
6705 	 * create_vnode attempts to create the file exclusive,
6706 	 * if it already exists the VOP_CREATE will fail and
6707 	 * may not increase va_seq. It is atomic if
6708 	 * we haven't changed the directory, but if it has changed
6709 	 * we don't know what changed it.
6710 	 */
6711 	if (!created) {
6712 		if (bva.va_seq && iva.va_seq &&
6713 		    bva.va_seq == iva.va_seq)
6714 			cinfo->atomic = TRUE;
6715 		else
6716 			cinfo->atomic = FALSE;
6717 		NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime);
6718 	} else {
6719 		/*
6720 		 * The entry was created, we need to sync the
6721 		 * directory metadata.
6722 		 */
6723 		(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
6724 
6725 		/*
6726 		 * Get "after" change value, if it fails, simply return the
6727 		 * before value.
6728 		 */
6729 		ava.va_mask = AT_CTIME|AT_SEQ;
6730 		if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) {
6731 			ava.va_ctime = bva.va_ctime;
6732 			ava.va_seq = 0;
6733 		}
6734 
6735 		NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
6736 
6737 		/*
6738 		 * The cinfo->atomic = TRUE only if we have
6739 		 * non-zero va_seq's, and it has incremented by exactly one
6740 		 * during the create_vnode and it didn't
6741 		 * change during the VOP_FSYNC.
6742 		 */
6743 		if (bva.va_seq && iva.va_seq && ava.va_seq &&
6744 		    iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq)
6745 			cinfo->atomic = TRUE;
6746 		else
6747 			cinfo->atomic = FALSE;
6748 	}
6749 
6750 	/* Check for mandatory locking and that the size gets set. */
6751 	cva.va_mask = AT_MODE;
6752 	if (setsize)
6753 		cva.va_mask |= AT_SIZE;
6754 
6755 	/* Assume the worst */
6756 	cs->mandlock = TRUE;
6757 
6758 	if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) {
6759 		cs->mandlock = MANDLOCK(cs->vp, cva.va_mode);
6760 
6761 		/*
6762 		 * Truncate the file if necessary; this would be
6763 		 * the case for create over an existing file.
6764 		 */
6765 
6766 		if (trunc) {
6767 			int in_crit = 0;
6768 			rfs4_file_t *fp;
6769 			bool_t create = FALSE;
6770 
6771 			/*
6772 			 * We are writing over an existing file.
6773 			 * Check to see if we need to recall a delegation.
6774 			 */
6775 			rfs4_hold_deleg_policy();
6776 			if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) {
6777 				if (rfs4_check_delegated_byfp(FWRITE, fp,
6778 				    (reqsize == 0), FALSE, FALSE, &clientid)) {
6779 					rfs4_file_rele(fp);
6780 					rfs4_rele_deleg_policy();
6781 					VN_RELE(vp);
6782 					*attrset = 0;
6783 					return (NFS4ERR_DELAY);
6784 				}
6785 				rfs4_file_rele(fp);
6786 			}
6787 			rfs4_rele_deleg_policy();
6788 
6789 			if (nbl_need_check(vp)) {
6790 				in_crit = 1;
6791 
6792 				ASSERT(reqsize == 0);
6793 
6794 				nbl_start_crit(vp, RW_READER);
6795 				if (nbl_conflict(vp, NBL_WRITE, 0,
6796 				    cva.va_size, 0, NULL)) {
6797 					in_crit = 0;
6798 					nbl_end_crit(vp);
6799 					VN_RELE(vp);
6800 					*attrset = 0;
6801 					return (NFS4ERR_ACCESS);
6802 				}
6803 			}
6804 			ct.cc_sysid = 0;
6805 			ct.cc_pid = 0;
6806 			ct.cc_caller_id = nfs4_srv_caller_id;
6807 			ct.cc_flags = CC_DONTBLOCK;
6808 
6809 			cva.va_mask = AT_SIZE;
6810 			cva.va_size = reqsize;
6811 			(void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct);
6812 			if (in_crit)
6813 				nbl_end_crit(vp);
6814 		}
6815 	}
6816 
6817 	error = makefh4(&cs->fh, vp, cs->exi);
6818 
6819 	/*
6820 	 * Force modified data and metadata out to stable storage.
6821 	 */
6822 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL);
6823 
6824 	if (error) {
6825 		VN_RELE(vp);
6826 		*attrset = 0;
6827 		return (puterrno4(error));
6828 	}
6829 
6830 	/* if parent dir is attrdir, set namedattr fh flag */
6831 	if (dvp->v_flag & V_XATTRDIR)
6832 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
6833 
6834 	if (cs->vp)
6835 		VN_RELE(cs->vp);
6836 
6837 	cs->vp = vp;
6838 
6839 	/*
6840 	 * if we did not create the file, we will need to check
6841 	 * the access bits on the file
6842 	 */
6843 
6844 	if (!created) {
6845 		if (setsize)
6846 			args->share_access |= OPEN4_SHARE_ACCESS_WRITE;
6847 		status = check_open_access(args->share_access, cs, req);
6848 		if (status != NFS4_OK)
6849 			*attrset = 0;
6850 	}
6851 	return (status);
6852 }
6853 
6854 /*ARGSUSED*/
6855 static void
6856 rfs4_do_open(struct compound_state *cs, struct svc_req *req,
6857     rfs4_openowner_t *oo, delegreq_t deleg,
6858     uint32_t access, uint32_t deny,
6859     OPEN4res *resp, int deleg_cur)
6860 {
6861 	/* XXX Currently not using req  */
6862 	rfs4_state_t *sp;
6863 	rfs4_file_t *fp;
6864 	bool_t screate = TRUE;
6865 	bool_t fcreate = TRUE;
6866 	uint32_t open_a, share_a;
6867 	uint32_t open_d, share_d;
6868 	rfs4_deleg_state_t *dsp;
6869 	sysid_t sysid;
6870 	nfsstat4 status;
6871 	caller_context_t ct;
6872 	int fflags = 0;
6873 	int recall = 0;
6874 	int err;
6875 	int first_open;
6876 
6877 	/* get the file struct and hold a lock on it during initial open */
6878 	fp = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate);
6879 	if (fp == NULL) {
6880 		resp->status = NFS4ERR_RESOURCE;
6881 		DTRACE_PROBE1(nfss__e__do__open1, nfsstat4, resp->status);
6882 		return;
6883 	}
6884 
6885 	sp = rfs4_findstate_by_owner_file(oo, fp, &screate);
6886 	if (sp == NULL) {
6887 		resp->status = NFS4ERR_RESOURCE;
6888 		DTRACE_PROBE1(nfss__e__do__open2, nfsstat4, resp->status);
6889 		/* No need to keep any reference */
6890 		rw_exit(&fp->rf_file_rwlock);
6891 		rfs4_file_rele(fp);
6892 		return;
6893 	}
6894 
6895 	/* try to get the sysid before continuing */
6896 	if ((status = rfs4_client_sysid(oo->ro_client, &sysid)) != NFS4_OK) {
6897 		resp->status = status;
6898 		rfs4_file_rele(fp);
6899 		/* Not a fully formed open; "close" it */
6900 		if (screate == TRUE)
6901 			rfs4_state_close(sp, FALSE, FALSE, cs->cr);
6902 		rfs4_state_rele(sp);
6903 		return;
6904 	}
6905 
6906 	/* Calculate the fflags for this OPEN. */
6907 	if (access & OPEN4_SHARE_ACCESS_READ)
6908 		fflags |= FREAD;
6909 	if (access & OPEN4_SHARE_ACCESS_WRITE)
6910 		fflags |= FWRITE;
6911 
6912 	rfs4_dbe_lock(sp->rs_dbe);
6913 
6914 	/*
6915 	 * Calculate the new deny and access mode that this open is adding to
6916 	 * the file for this open owner;
6917 	 */
6918 	open_d = (deny & ~sp->rs_open_deny);
6919 	open_a = (access & ~sp->rs_open_access);
6920 
6921 	/*
6922 	 * Calculate the new share access and share deny modes that this open
6923 	 * is adding to the file for this open owner;
6924 	 */
6925 	share_a = (access & ~sp->rs_share_access);
6926 	share_d = (deny & ~sp->rs_share_deny);
6927 
6928 	first_open = (sp->rs_open_access & OPEN4_SHARE_ACCESS_BOTH) == 0;
6929 
6930 	/*
6931 	 * Check to see the client has already sent an open for this
6932 	 * open owner on this file with the same share/deny modes.
6933 	 * If so, we don't need to check for a conflict and we don't
6934 	 * need to add another shrlock.  If not, then we need to
6935 	 * check for conflicts in deny and access before checking for
6936 	 * conflicts in delegation.  We don't want to recall a
6937 	 * delegation based on an open that will eventually fail based
6938 	 * on shares modes.
6939 	 */
6940 
6941 	if (share_a || share_d) {
6942 		if ((err = rfs4_share(sp, access, deny)) != 0) {
6943 			rfs4_dbe_unlock(sp->rs_dbe);
6944 			resp->status = err;
6945 
6946 			rfs4_file_rele(fp);
6947 			/* Not a fully formed open; "close" it */
6948 			if (screate == TRUE)
6949 				rfs4_state_close(sp, FALSE, FALSE, cs->cr);
6950 			rfs4_state_rele(sp);
6951 			return;
6952 		}
6953 	}
6954 
6955 	rfs4_dbe_lock(fp->rf_dbe);
6956 
6957 	/*
6958 	 * Check to see if this file is delegated and if so, if a
6959 	 * recall needs to be done.
6960 	 */
6961 	if (rfs4_check_recall(sp, access)) {
6962 		rfs4_dbe_unlock(fp->rf_dbe);
6963 		rfs4_dbe_unlock(sp->rs_dbe);
6964 		rfs4_recall_deleg(fp, FALSE, sp->rs_owner->ro_client);
6965 		delay(NFS4_DELEGATION_CONFLICT_DELAY);
6966 		rfs4_dbe_lock(sp->rs_dbe);
6967 
6968 		/* if state closed while lock was dropped */
6969 		if (sp->rs_closed) {
6970 			if (share_a || share_d)
6971 				(void) rfs4_unshare(sp);
6972 			rfs4_dbe_unlock(sp->rs_dbe);
6973 			rfs4_file_rele(fp);
6974 			/* Not a fully formed open; "close" it */
6975 			if (screate == TRUE)
6976 				rfs4_state_close(sp, FALSE, FALSE, cs->cr);
6977 			rfs4_state_rele(sp);
6978 			resp->status = NFS4ERR_OLD_STATEID;
6979 			return;
6980 		}
6981 
6982 		rfs4_dbe_lock(fp->rf_dbe);
6983 		/* Let's see if the delegation was returned */
6984 		if (rfs4_check_recall(sp, access)) {
6985 			rfs4_dbe_unlock(fp->rf_dbe);
6986 			if (share_a || share_d)
6987 				(void) rfs4_unshare(sp);
6988 			rfs4_dbe_unlock(sp->rs_dbe);
6989 			rfs4_file_rele(fp);
6990 			rfs4_update_lease(sp->rs_owner->ro_client);
6991 
6992 			/* Not a fully formed open; "close" it */
6993 			if (screate == TRUE)
6994 				rfs4_state_close(sp, FALSE, FALSE, cs->cr);
6995 			rfs4_state_rele(sp);
6996 			resp->status = NFS4ERR_DELAY;
6997 			return;
6998 		}
6999 	}
7000 	/*
7001 	 * the share check passed and any delegation conflict has been
7002 	 * taken care of, now call vop_open.
7003 	 * if this is the first open then call vop_open with fflags.
7004 	 * if not, call vn_open_upgrade with just the upgrade flags.
7005 	 *
7006 	 * if the file has been opened already, it will have the current
7007 	 * access mode in the state struct.  if it has no share access, then
7008 	 * this is a new open.
7009 	 *
7010 	 * However, if this is open with CLAIM_DLEGATE_CUR, then don't
7011 	 * call VOP_OPEN(), just do the open upgrade.
7012 	 */
7013 	if (first_open && !deleg_cur) {
7014 		ct.cc_sysid = sysid;
7015 		ct.cc_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe);
7016 		ct.cc_caller_id = nfs4_srv_caller_id;
7017 		ct.cc_flags = CC_DONTBLOCK;
7018 		err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct);
7019 		if (err) {
7020 			rfs4_dbe_unlock(fp->rf_dbe);
7021 			if (share_a || share_d)
7022 				(void) rfs4_unshare(sp);
7023 			rfs4_dbe_unlock(sp->rs_dbe);
7024 			rfs4_file_rele(fp);
7025 
7026 			/* Not a fully formed open; "close" it */
7027 			if (screate == TRUE)
7028 				rfs4_state_close(sp, FALSE, FALSE, cs->cr);
7029 			rfs4_state_rele(sp);
7030 			/* check if a monitor detected a delegation conflict */
7031 			if (err == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK))
7032 				resp->status = NFS4ERR_DELAY;
7033 			else
7034 				resp->status = NFS4ERR_SERVERFAULT;
7035 			return;
7036 		}
7037 	} else { /* open upgrade */
7038 		/*
7039 		 * calculate the fflags for the new mode that is being added
7040 		 * by this upgrade.
7041 		 */
7042 		fflags = 0;
7043 		if (open_a & OPEN4_SHARE_ACCESS_READ)
7044 			fflags |= FREAD;
7045 		if (open_a & OPEN4_SHARE_ACCESS_WRITE)
7046 			fflags |= FWRITE;
7047 		vn_open_upgrade(cs->vp, fflags);
7048 	}
7049 	sp->rs_open_access |= access;
7050 	sp->rs_open_deny |= deny;
7051 
7052 	if (open_d & OPEN4_SHARE_DENY_READ)
7053 		fp->rf_deny_read++;
7054 	if (open_d & OPEN4_SHARE_DENY_WRITE)
7055 		fp->rf_deny_write++;
7056 	fp->rf_share_deny |= deny;
7057 
7058 	if (open_a & OPEN4_SHARE_ACCESS_READ)
7059 		fp->rf_access_read++;
7060 	if (open_a & OPEN4_SHARE_ACCESS_WRITE)
7061 		fp->rf_access_write++;
7062 	fp->rf_share_access |= access;
7063 
7064 	/*
7065 	 * Check for delegation here. if the deleg argument is not
7066 	 * DELEG_ANY, then this is a reclaim from a client and
7067 	 * we must honor the delegation requested. If necessary we can
7068 	 * set the recall flag.
7069 	 */
7070 
7071 	dsp = rfs4_grant_delegation(deleg, sp, &recall);
7072 
7073 	cs->deleg = (fp->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE);
7074 
7075 	next_stateid(&sp->rs_stateid);
7076 
7077 	resp->stateid = sp->rs_stateid.stateid;
7078 
7079 	rfs4_dbe_unlock(fp->rf_dbe);
7080 	rfs4_dbe_unlock(sp->rs_dbe);
7081 
7082 	if (dsp) {
7083 		rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall);
7084 		rfs4_deleg_state_rele(dsp);
7085 	}
7086 
7087 	rfs4_file_rele(fp);
7088 	rfs4_state_rele(sp);
7089 
7090 	resp->status = NFS4_OK;
7091 }
7092 
7093 /*ARGSUSED*/
7094 static void
7095 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req,
7096     OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
7097 {
7098 	change_info4 *cinfo = &resp->cinfo;
7099 	bitmap4 *attrset = &resp->attrset;
7100 
7101 	if (args->opentype == OPEN4_NOCREATE)
7102 		resp->status = rfs4_lookupfile(&args->open_claim4_u.file,
7103 		    req, cs, args->share_access, cinfo);
7104 	else {
7105 		/* inhibit delegation grants during exclusive create */
7106 
7107 		if (args->mode == EXCLUSIVE4)
7108 			rfs4_disable_delegation();
7109 
7110 		resp->status = rfs4_createfile(args, req, cs, cinfo, attrset,
7111 		    oo->ro_client->rc_clientid);
7112 	}
7113 
7114 	if (resp->status == NFS4_OK) {
7115 
7116 		/* cs->vp cs->fh now reference the desired file */
7117 
7118 		rfs4_do_open(cs, req, oo,
7119 		    oo->ro_need_confirm ? DELEG_NONE : DELEG_ANY,
7120 		    args->share_access, args->share_deny, resp, 0);
7121 
7122 		/*
7123 		 * If rfs4_createfile set attrset, we must
7124 		 * clear this attrset before the response is copied.
7125 		 */
7126 		if (resp->status != NFS4_OK && resp->attrset) {
7127 			resp->attrset = 0;
7128 		}
7129 	}
7130 	else
7131 		*cs->statusp = resp->status;
7132 
7133 	if (args->mode == EXCLUSIVE4)
7134 		rfs4_enable_delegation();
7135 }
7136 
7137 /*ARGSUSED*/
7138 static void
7139 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req,
7140     OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
7141 {
7142 	change_info4 *cinfo = &resp->cinfo;
7143 	vattr_t va;
7144 	vtype_t v_type = cs->vp->v_type;
7145 	int error = 0;
7146 
7147 	/* Verify that we have a regular file */
7148 	if (v_type != VREG) {
7149 		if (v_type == VDIR)
7150 			resp->status = NFS4ERR_ISDIR;
7151 		else if (v_type == VLNK)
7152 			resp->status = NFS4ERR_SYMLINK;
7153 		else
7154 			resp->status = NFS4ERR_INVAL;
7155 		return;
7156 	}
7157 
7158 	va.va_mask = AT_MODE|AT_UID;
7159 	error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL);
7160 	if (error) {
7161 		resp->status = puterrno4(error);
7162 		return;
7163 	}
7164 
7165 	cs->mandlock = MANDLOCK(cs->vp, va.va_mode);
7166 
7167 	/*
7168 	 * Check if we have access to the file, Note the the file
7169 	 * could have originally been open UNCHECKED or GUARDED
7170 	 * with mode bits that will now fail, but there is nothing
7171 	 * we can really do about that except in the case that the
7172 	 * owner of the file is the one requesting the open.
7173 	 */
7174 	if (crgetuid(cs->cr) != va.va_uid) {
7175 		resp->status = check_open_access(args->share_access, cs, req);
7176 		if (resp->status != NFS4_OK) {
7177 			return;
7178 		}
7179 	}
7180 
7181 	/*
7182 	 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero
7183 	 */
7184 	cinfo->before = 0;
7185 	cinfo->after = 0;
7186 	cinfo->atomic = FALSE;
7187 
7188 	rfs4_do_open(cs, req, oo,
7189 	    NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type),
7190 	    args->share_access, args->share_deny, resp, 0);
7191 }
7192 
7193 static void
7194 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req,
7195     OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
7196 {
7197 	int error;
7198 	nfsstat4 status;
7199 	stateid4 stateid =
7200 	    args->open_claim4_u.delegate_cur_info.delegate_stateid;
7201 	rfs4_deleg_state_t *dsp;
7202 
7203 	/*
7204 	 * Find the state info from the stateid and confirm that the
7205 	 * file is delegated.  If the state openowner is the same as
7206 	 * the supplied openowner we're done. If not, get the file
7207 	 * info from the found state info. Use that file info to
7208 	 * create the state for this lock owner. Note solaris doen't
7209 	 * really need the pathname to find the file. We may want to
7210 	 * lookup the pathname and make sure that the vp exist and
7211 	 * matches the vp in the file structure. However it is
7212 	 * possible that the pathname nolonger exists (local process
7213 	 * unlinks the file), so this may not be that useful.
7214 	 */
7215 
7216 	status = rfs4_get_deleg_state(&stateid, &dsp);
7217 	if (status != NFS4_OK) {
7218 		resp->status = status;
7219 		return;
7220 	}
7221 
7222 	ASSERT(dsp->rds_finfo->rf_dinfo.rd_dtype != OPEN_DELEGATE_NONE);
7223 
7224 	/*
7225 	 * New lock owner, create state. Since this was probably called
7226 	 * in response to a CB_RECALL we set deleg to DELEG_NONE
7227 	 */
7228 
7229 	ASSERT(cs->vp != NULL);
7230 	VN_RELE(cs->vp);
7231 	VN_HOLD(dsp->rds_finfo->rf_vp);
7232 	cs->vp = dsp->rds_finfo->rf_vp;
7233 
7234 	if (error = makefh4(&cs->fh, cs->vp, cs->exi)) {
7235 		rfs4_deleg_state_rele(dsp);
7236 		*cs->statusp = resp->status = puterrno4(error);
7237 		return;
7238 	}
7239 
7240 	/* Mark progress for delegation returns */
7241 	dsp->rds_finfo->rf_dinfo.rd_time_lastwrite = gethrestime_sec();
7242 	rfs4_deleg_state_rele(dsp);
7243 	rfs4_do_open(cs, req, oo, DELEG_NONE,
7244 	    args->share_access, args->share_deny, resp, 1);
7245 }
7246 
7247 /*ARGSUSED*/
7248 static void
7249 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req,
7250     OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
7251 {
7252 	/*
7253 	 * Lookup the pathname, it must already exist since this file
7254 	 * was delegated.
7255 	 *
7256 	 * Find the file and state info for this vp and open owner pair.
7257 	 *	check that they are in fact delegated.
7258 	 *	check that the state access and deny modes are the same.
7259 	 *
7260 	 * Return the delgation possibly seting the recall flag.
7261 	 */
7262 	rfs4_file_t *fp;
7263 	rfs4_state_t *sp;
7264 	bool_t create = FALSE;
7265 	bool_t dcreate = FALSE;
7266 	rfs4_deleg_state_t *dsp;
7267 	nfsace4 *ace;
7268 
7269 	/* Note we ignore oflags */
7270 	resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev,
7271 	    req, cs, args->share_access, &resp->cinfo);
7272 
7273 	if (resp->status != NFS4_OK) {
7274 		return;
7275 	}
7276 
7277 	/* get the file struct and hold a lock on it during initial open */
7278 	fp = rfs4_findfile_withlock(cs->vp, NULL, &create);
7279 	if (fp == NULL) {
7280 		resp->status = NFS4ERR_RESOURCE;
7281 		DTRACE_PROBE1(nfss__e__do_opendelprev1, nfsstat4, resp->status);
7282 		return;
7283 	}
7284 
7285 	sp = rfs4_findstate_by_owner_file(oo, fp, &create);
7286 	if (sp == NULL) {
7287 		resp->status = NFS4ERR_SERVERFAULT;
7288 		DTRACE_PROBE1(nfss__e__do_opendelprev2, nfsstat4, resp->status);
7289 		rw_exit(&fp->rf_file_rwlock);
7290 		rfs4_file_rele(fp);
7291 		return;
7292 	}
7293 
7294 	rfs4_dbe_lock(sp->rs_dbe);
7295 	rfs4_dbe_lock(fp->rf_dbe);
7296 	if (args->share_access != sp->rs_share_access ||
7297 	    args->share_deny != sp->rs_share_deny ||
7298 	    sp->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_NONE) {
7299 		NFS4_DEBUG(rfs4_debug,
7300 		    (CE_NOTE, "rfs4_do_opendelprev: state mixup"));
7301 		rfs4_dbe_unlock(fp->rf_dbe);
7302 		rfs4_dbe_unlock(sp->rs_dbe);
7303 		rfs4_file_rele(fp);
7304 		rfs4_state_rele(sp);
7305 		resp->status = NFS4ERR_SERVERFAULT;
7306 		return;
7307 	}
7308 	rfs4_dbe_unlock(fp->rf_dbe);
7309 	rfs4_dbe_unlock(sp->rs_dbe);
7310 
7311 	dsp = rfs4_finddeleg(sp, &dcreate);
7312 	if (dsp == NULL) {
7313 		rfs4_state_rele(sp);
7314 		rfs4_file_rele(fp);
7315 		resp->status = NFS4ERR_SERVERFAULT;
7316 		return;
7317 	}
7318 
7319 	next_stateid(&sp->rs_stateid);
7320 
7321 	resp->stateid = sp->rs_stateid.stateid;
7322 
7323 	resp->delegation.delegation_type = dsp->rds_dtype;
7324 
7325 	if (dsp->rds_dtype == OPEN_DELEGATE_READ) {
7326 		open_read_delegation4 *rv =
7327 		    &resp->delegation.open_delegation4_u.read;
7328 
7329 		rv->stateid = dsp->rds_delegid.stateid;
7330 		rv->recall = FALSE; /* no policy in place to set to TRUE */
7331 		ace = &rv->permissions;
7332 	} else {
7333 		open_write_delegation4 *rv =
7334 		    &resp->delegation.open_delegation4_u.write;
7335 
7336 		rv->stateid = dsp->rds_delegid.stateid;
7337 		rv->recall = FALSE;  /* no policy in place to set to TRUE */
7338 		ace = &rv->permissions;
7339 		rv->space_limit.limitby = NFS_LIMIT_SIZE;
7340 		rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX;
7341 	}
7342 
7343 	/* XXX For now */
7344 	ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE;
7345 	ace->flag = 0;
7346 	ace->access_mask = 0;
7347 	ace->who.utf8string_len = 0;
7348 	ace->who.utf8string_val = 0;
7349 
7350 	rfs4_deleg_state_rele(dsp);
7351 	rfs4_state_rele(sp);
7352 	rfs4_file_rele(fp);
7353 }
7354 
7355 typedef enum {
7356 	NFS4_CHKSEQ_OKAY = 0,
7357 	NFS4_CHKSEQ_REPLAY = 1,
7358 	NFS4_CHKSEQ_BAD = 2
7359 } rfs4_chkseq_t;
7360 
7361 /*
7362  * Generic function for sequence number checks.
7363  */
7364 static rfs4_chkseq_t
7365 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop,
7366     seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres)
7367 {
7368 	/* Same sequence ids and matching operations? */
7369 	if (seqid == rqst_seq && resop->resop == lastop->resop) {
7370 		if (copyres == TRUE) {
7371 			rfs4_free_reply(resop);
7372 			rfs4_copy_reply(resop, lastop);
7373 		}
7374 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
7375 		    "Replayed SEQID %d\n", seqid));
7376 		return (NFS4_CHKSEQ_REPLAY);
7377 	}
7378 
7379 	/* If the incoming sequence is not the next expected then it is bad */
7380 	if (rqst_seq != seqid + 1) {
7381 		if (rqst_seq == seqid) {
7382 			NFS4_DEBUG(rfs4_debug,
7383 			    (CE_NOTE, "BAD SEQID: Replayed sequence id "
7384 			    "but last op was %d current op is %d\n",
7385 			    lastop->resop, resop->resop));
7386 			return (NFS4_CHKSEQ_BAD);
7387 		}
7388 		NFS4_DEBUG(rfs4_debug,
7389 		    (CE_NOTE, "BAD SEQID: got %u expecting %u\n",
7390 		    rqst_seq, seqid));
7391 		return (NFS4_CHKSEQ_BAD);
7392 	}
7393 
7394 	/* Everything okay -- next expected */
7395 	return (NFS4_CHKSEQ_OKAY);
7396 }
7397 
7398 
7399 static rfs4_chkseq_t
7400 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop)
7401 {
7402 	rfs4_chkseq_t rc;
7403 
7404 	rfs4_dbe_lock(op->ro_dbe);
7405 	rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply, seqid, resop,
7406 	    TRUE);
7407 	rfs4_dbe_unlock(op->ro_dbe);
7408 
7409 	if (rc == NFS4_CHKSEQ_OKAY)
7410 		rfs4_update_lease(op->ro_client);
7411 
7412 	return (rc);
7413 }
7414 
7415 static rfs4_chkseq_t
7416 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op, nfs_resop4 *resop)
7417 {
7418 	rfs4_chkseq_t rc;
7419 
7420 	rfs4_dbe_lock(op->ro_dbe);
7421 	rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply,
7422 	    olo_seqid, resop, FALSE);
7423 	rfs4_dbe_unlock(op->ro_dbe);
7424 
7425 	return (rc);
7426 }
7427 
7428 static rfs4_chkseq_t
7429 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lsp, nfs_resop4 *resop)
7430 {
7431 	rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY;
7432 
7433 	rfs4_dbe_lock(lsp->rls_dbe);
7434 	if (!lsp->rls_skip_seqid_check)
7435 		rc = rfs4_check_seqid(lsp->rls_seqid, &lsp->rls_reply, seqid,
7436 		    resop, TRUE);
7437 	rfs4_dbe_unlock(lsp->rls_dbe);
7438 
7439 	return (rc);
7440 }
7441 
7442 static void
7443 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop,
7444     struct svc_req *req, struct compound_state *cs)
7445 {
7446 	OPEN4args *args = &argop->nfs_argop4_u.opopen;
7447 	OPEN4res *resp = &resop->nfs_resop4_u.opopen;
7448 	open_owner4 *owner = &args->owner;
7449 	open_claim_type4 claim = args->claim;
7450 	rfs4_client_t *cp;
7451 	rfs4_openowner_t *oo;
7452 	bool_t create;
7453 	bool_t replay = FALSE;
7454 	int can_reclaim;
7455 
7456 	DTRACE_NFSV4_2(op__open__start, struct compound_state *, cs,
7457 	    OPEN4args *, args);
7458 
7459 	if (cs->vp == NULL) {
7460 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7461 		goto end;
7462 	}
7463 
7464 	/*
7465 	 * Need to check clientid and lease expiration first based on
7466 	 * error ordering and incrementing sequence id.
7467 	 */
7468 	cp = rfs4_findclient_by_id(owner->clientid, FALSE);
7469 	if (cp == NULL) {
7470 		*cs->statusp = resp->status =
7471 		    rfs4_check_clientid(&owner->clientid, 0);
7472 		goto end;
7473 	}
7474 
7475 	if (rfs4_lease_expired(cp)) {
7476 		rfs4_client_close(cp);
7477 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7478 		goto end;
7479 	}
7480 	can_reclaim = cp->rc_can_reclaim;
7481 
7482 	/*
7483 	 * Find the open_owner for use from this point forward.  Take
7484 	 * care in updating the sequence id based on the type of error
7485 	 * being returned.
7486 	 */
7487 retry:
7488 	create = TRUE;
7489 	oo = rfs4_findopenowner(owner, &create, args->seqid);
7490 	if (oo == NULL) {
7491 		*cs->statusp = resp->status = NFS4ERR_RESOURCE;
7492 		rfs4_client_rele(cp);
7493 		goto end;
7494 	}
7495 
7496 	/* Hold off access to the sequence space while the open is done */
7497 	rfs4_sw_enter(&oo->ro_sw);
7498 
7499 	/*
7500 	 * If the open_owner existed before at the server, then check
7501 	 * the sequence id.
7502 	 */
7503 	if (!create && !oo->ro_postpone_confirm) {
7504 		switch (rfs4_check_open_seqid(args->seqid, oo, resop)) {
7505 		case NFS4_CHKSEQ_BAD:
7506 			if ((args->seqid > oo->ro_open_seqid) &&
7507 			    oo->ro_need_confirm) {
7508 				rfs4_free_opens(oo, TRUE, FALSE);
7509 				rfs4_sw_exit(&oo->ro_sw);
7510 				rfs4_openowner_rele(oo);
7511 				goto retry;
7512 			}
7513 			resp->status = NFS4ERR_BAD_SEQID;
7514 			goto out;
7515 		case NFS4_CHKSEQ_REPLAY: /* replay of previous request */
7516 			replay = TRUE;
7517 			goto out;
7518 		default:
7519 			break;
7520 		}
7521 
7522 		/*
7523 		 * Sequence was ok and open owner exists
7524 		 * check to see if we have yet to see an
7525 		 * open_confirm.
7526 		 */
7527 		if (oo->ro_need_confirm) {
7528 			rfs4_free_opens(oo, TRUE, FALSE);
7529 			rfs4_sw_exit(&oo->ro_sw);
7530 			rfs4_openowner_rele(oo);
7531 			goto retry;
7532 		}
7533 	}
7534 	/* Grace only applies to regular-type OPENs */
7535 	if (rfs4_clnt_in_grace(cp) &&
7536 	    (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) {
7537 		*cs->statusp = resp->status = NFS4ERR_GRACE;
7538 		goto out;
7539 	}
7540 
7541 	/*
7542 	 * If previous state at the server existed then can_reclaim
7543 	 * will be set. If not reply NFS4ERR_NO_GRACE to the
7544 	 * client.
7545 	 */
7546 	if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) {
7547 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
7548 		goto out;
7549 	}
7550 
7551 
7552 	/*
7553 	 * Reject the open if the client has missed the grace period
7554 	 */
7555 	if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) {
7556 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
7557 		goto out;
7558 	}
7559 
7560 	/* Couple of up-front bookkeeping items */
7561 	if (oo->ro_need_confirm) {
7562 		/*
7563 		 * If this is a reclaim OPEN then we should not ask
7564 		 * for a confirmation of the open_owner per the
7565 		 * protocol specification.
7566 		 */
7567 		if (claim == CLAIM_PREVIOUS)
7568 			oo->ro_need_confirm = FALSE;
7569 		else
7570 			resp->rflags |= OPEN4_RESULT_CONFIRM;
7571 	}
7572 	resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX;
7573 
7574 	/*
7575 	 * If there is an unshared filesystem mounted on this vnode,
7576 	 * do not allow to open/create in this directory.
7577 	 */
7578 	if (vn_ismntpt(cs->vp)) {
7579 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
7580 		goto out;
7581 	}
7582 
7583 	/*
7584 	 * access must READ, WRITE, or BOTH.  No access is invalid.
7585 	 * deny can be READ, WRITE, BOTH, or NONE.
7586 	 * bits not defined for access/deny are invalid.
7587 	 */
7588 	if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) ||
7589 	    (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) ||
7590 	    (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) {
7591 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7592 		goto out;
7593 	}
7594 
7595 
7596 	/*
7597 	 * make sure attrset is zero before response is built.
7598 	 */
7599 	resp->attrset = 0;
7600 
7601 	switch (claim) {
7602 	case CLAIM_NULL:
7603 		rfs4_do_opennull(cs, req, args, oo, resp);
7604 		break;
7605 	case CLAIM_PREVIOUS:
7606 		rfs4_do_openprev(cs, req, args, oo, resp);
7607 		break;
7608 	case CLAIM_DELEGATE_CUR:
7609 		rfs4_do_opendelcur(cs, req, args, oo, resp);
7610 		break;
7611 	case CLAIM_DELEGATE_PREV:
7612 		rfs4_do_opendelprev(cs, req, args, oo, resp);
7613 		break;
7614 	default:
7615 		resp->status = NFS4ERR_INVAL;
7616 		break;
7617 	}
7618 
7619 out:
7620 	rfs4_client_rele(cp);
7621 
7622 	/* Catch sequence id handling here to make it a little easier */
7623 	switch (resp->status) {
7624 	case NFS4ERR_BADXDR:
7625 	case NFS4ERR_BAD_SEQID:
7626 	case NFS4ERR_BAD_STATEID:
7627 	case NFS4ERR_NOFILEHANDLE:
7628 	case NFS4ERR_RESOURCE:
7629 	case NFS4ERR_STALE_CLIENTID:
7630 	case NFS4ERR_STALE_STATEID:
7631 		/*
7632 		 * The protocol states that if any of these errors are
7633 		 * being returned, the sequence id should not be
7634 		 * incremented.  Any other return requires an
7635 		 * increment.
7636 		 */
7637 		break;
7638 	default:
7639 		/* Always update the lease in this case */
7640 		rfs4_update_lease(oo->ro_client);
7641 
7642 		/* Regular response - copy the result */
7643 		if (!replay)
7644 			rfs4_update_open_resp(oo, resop, &cs->fh);
7645 
7646 		/*
7647 		 * REPLAY case: Only if the previous response was OK
7648 		 * do we copy the filehandle.  If not OK, no
7649 		 * filehandle to copy.
7650 		 */
7651 		if (replay == TRUE &&
7652 		    resp->status == NFS4_OK &&
7653 		    oo->ro_reply_fh.nfs_fh4_val) {
7654 			/*
7655 			 * If this is a replay, we must restore the
7656 			 * current filehandle/vp to that of what was
7657 			 * returned originally.  Try our best to do
7658 			 * it.
7659 			 */
7660 			nfs_fh4_fmt_t *fh_fmtp =
7661 			    (nfs_fh4_fmt_t *)oo->ro_reply_fh.nfs_fh4_val;
7662 
7663 			cs->exi = checkexport4(&fh_fmtp->fh4_fsid,
7664 			    (fid_t *)&fh_fmtp->fh4_xlen, NULL);
7665 
7666 			if (cs->exi == NULL) {
7667 				resp->status = NFS4ERR_STALE;
7668 				goto finish;
7669 			}
7670 
7671 			VN_RELE(cs->vp);
7672 
7673 			cs->vp = nfs4_fhtovp(&oo->ro_reply_fh, cs->exi,
7674 			    &resp->status);
7675 
7676 			if (cs->vp == NULL)
7677 				goto finish;
7678 
7679 			nfs_fh4_copy(&oo->ro_reply_fh, &cs->fh);
7680 		}
7681 
7682 		/*
7683 		 * If this was a replay, no need to update the
7684 		 * sequence id. If the open_owner was not created on
7685 		 * this pass, then update.  The first use of an
7686 		 * open_owner will not bump the sequence id.
7687 		 */
7688 		if (replay == FALSE && !create)
7689 			rfs4_update_open_sequence(oo);
7690 		/*
7691 		 * If the client is receiving an error and the
7692 		 * open_owner needs to be confirmed, there is no way
7693 		 * to notify the client of this fact ignoring the fact
7694 		 * that the server has no method of returning a
7695 		 * stateid to confirm.  Therefore, the server needs to
7696 		 * mark this open_owner in a way as to avoid the
7697 		 * sequence id checking the next time the client uses
7698 		 * this open_owner.
7699 		 */
7700 		if (resp->status != NFS4_OK && oo->ro_need_confirm)
7701 			oo->ro_postpone_confirm = TRUE;
7702 		/*
7703 		 * If OK response then clear the postpone flag and
7704 		 * reset the sequence id to keep in sync with the
7705 		 * client.
7706 		 */
7707 		if (resp->status == NFS4_OK && oo->ro_postpone_confirm) {
7708 			oo->ro_postpone_confirm = FALSE;
7709 			oo->ro_open_seqid = args->seqid;
7710 		}
7711 		break;
7712 	}
7713 
7714 finish:
7715 	*cs->statusp = resp->status;
7716 
7717 	rfs4_sw_exit(&oo->ro_sw);
7718 	rfs4_openowner_rele(oo);
7719 
7720 end:
7721 	DTRACE_NFSV4_2(op__open__done, struct compound_state *, cs,
7722 	    OPEN4res *, resp);
7723 }
7724 
7725 /*ARGSUSED*/
7726 void
7727 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
7728     struct svc_req *req, struct compound_state *cs)
7729 {
7730 	OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm;
7731 	OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm;
7732 	rfs4_state_t *sp;
7733 	nfsstat4 status;
7734 
7735 	DTRACE_NFSV4_2(op__open__confirm__start, struct compound_state *, cs,
7736 	    OPEN_CONFIRM4args *, args);
7737 
7738 	if (cs->vp == NULL) {
7739 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7740 		goto out;
7741 	}
7742 
7743 	if (cs->vp->v_type != VREG) {
7744 		*cs->statusp = resp->status =
7745 		    cs->vp->v_type == VDIR ? NFS4ERR_ISDIR : NFS4ERR_INVAL;
7746 		return;
7747 	}
7748 
7749 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
7750 	if (status != NFS4_OK) {
7751 		*cs->statusp = resp->status = status;
7752 		goto out;
7753 	}
7754 
7755 	/* Ensure specified filehandle matches */
7756 	if (cs->vp != sp->rs_finfo->rf_vp) {
7757 		rfs4_state_rele(sp);
7758 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7759 		goto out;
7760 	}
7761 
7762 	/* hold off other access to open_owner while we tinker */
7763 	rfs4_sw_enter(&sp->rs_owner->ro_sw);
7764 
7765 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7766 	case NFS4_CHECK_STATEID_OKAY:
7767 		if (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
7768 		    resop) != 0) {
7769 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7770 			break;
7771 		}
7772 		/*
7773 		 * If it is the appropriate stateid and determined to
7774 		 * be "OKAY" then this means that the stateid does not
7775 		 * need to be confirmed and the client is in error for
7776 		 * sending an OPEN_CONFIRM.
7777 		 */
7778 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7779 		break;
7780 	case NFS4_CHECK_STATEID_OLD:
7781 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7782 		break;
7783 	case NFS4_CHECK_STATEID_BAD:
7784 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7785 		break;
7786 	case NFS4_CHECK_STATEID_EXPIRED:
7787 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7788 		break;
7789 	case NFS4_CHECK_STATEID_CLOSED:
7790 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7791 		break;
7792 	case NFS4_CHECK_STATEID_REPLAY:
7793 		switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
7794 		    resop)) {
7795 		case NFS4_CHKSEQ_OKAY:
7796 			/*
7797 			 * This is replayed stateid; if seqid matches
7798 			 * next expected, then client is using wrong seqid.
7799 			 */
7800 			/* fall through */
7801 		case NFS4_CHKSEQ_BAD:
7802 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7803 			break;
7804 		case NFS4_CHKSEQ_REPLAY:
7805 			/*
7806 			 * Note this case is the duplicate case so
7807 			 * resp->status is already set.
7808 			 */
7809 			*cs->statusp = resp->status;
7810 			rfs4_update_lease(sp->rs_owner->ro_client);
7811 			break;
7812 		}
7813 		break;
7814 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7815 		if (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
7816 		    resop) != NFS4_CHKSEQ_OKAY) {
7817 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7818 			break;
7819 		}
7820 		*cs->statusp = resp->status = NFS4_OK;
7821 
7822 		next_stateid(&sp->rs_stateid);
7823 		resp->open_stateid = sp->rs_stateid.stateid;
7824 		sp->rs_owner->ro_need_confirm = FALSE;
7825 		rfs4_update_lease(sp->rs_owner->ro_client);
7826 		rfs4_update_open_sequence(sp->rs_owner);
7827 		rfs4_update_open_resp(sp->rs_owner, resop, NULL);
7828 		break;
7829 	default:
7830 		ASSERT(FALSE);
7831 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7832 		break;
7833 	}
7834 	rfs4_sw_exit(&sp->rs_owner->ro_sw);
7835 	rfs4_state_rele(sp);
7836 
7837 out:
7838 	DTRACE_NFSV4_2(op__open__confirm__done, struct compound_state *, cs,
7839 	    OPEN_CONFIRM4res *, resp);
7840 }
7841 
7842 /*ARGSUSED*/
7843 void
7844 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop,
7845     struct svc_req *req, struct compound_state *cs)
7846 {
7847 	OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade;
7848 	OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade;
7849 	uint32_t access = args->share_access;
7850 	uint32_t deny = args->share_deny;
7851 	nfsstat4 status;
7852 	rfs4_state_t *sp;
7853 	rfs4_file_t *fp;
7854 	int fflags = 0;
7855 
7856 	DTRACE_NFSV4_2(op__open__downgrade__start, struct compound_state *, cs,
7857 	    OPEN_DOWNGRADE4args *, args);
7858 
7859 	if (cs->vp == NULL) {
7860 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7861 		goto out;
7862 	}
7863 
7864 	if (cs->vp->v_type != VREG) {
7865 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7866 		return;
7867 	}
7868 
7869 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
7870 	if (status != NFS4_OK) {
7871 		*cs->statusp = resp->status = status;
7872 		goto out;
7873 	}
7874 
7875 	/* Ensure specified filehandle matches */
7876 	if (cs->vp != sp->rs_finfo->rf_vp) {
7877 		rfs4_state_rele(sp);
7878 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7879 		goto out;
7880 	}
7881 
7882 	/* hold off other access to open_owner while we tinker */
7883 	rfs4_sw_enter(&sp->rs_owner->ro_sw);
7884 
7885 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7886 	case NFS4_CHECK_STATEID_OKAY:
7887 		if (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
7888 		    resop) != NFS4_CHKSEQ_OKAY) {
7889 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7890 			goto end;
7891 		}
7892 		break;
7893 	case NFS4_CHECK_STATEID_OLD:
7894 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7895 		goto end;
7896 	case NFS4_CHECK_STATEID_BAD:
7897 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7898 		goto end;
7899 	case NFS4_CHECK_STATEID_EXPIRED:
7900 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7901 		goto end;
7902 	case NFS4_CHECK_STATEID_CLOSED:
7903 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7904 		goto end;
7905 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7906 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7907 		goto end;
7908 	case NFS4_CHECK_STATEID_REPLAY:
7909 		/* Check the sequence id for the open owner */
7910 		switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
7911 		    resop)) {
7912 		case NFS4_CHKSEQ_OKAY:
7913 			/*
7914 			 * This is replayed stateid; if seqid matches
7915 			 * next expected, then client is using wrong seqid.
7916 			 */
7917 			/* fall through */
7918 		case NFS4_CHKSEQ_BAD:
7919 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7920 			goto end;
7921 		case NFS4_CHKSEQ_REPLAY:
7922 			/*
7923 			 * Note this case is the duplicate case so
7924 			 * resp->status is already set.
7925 			 */
7926 			*cs->statusp = resp->status;
7927 			rfs4_update_lease(sp->rs_owner->ro_client);
7928 			goto end;
7929 		}
7930 		break;
7931 	default:
7932 		ASSERT(FALSE);
7933 		break;
7934 	}
7935 
7936 	rfs4_dbe_lock(sp->rs_dbe);
7937 	/*
7938 	 * Check that the new access modes and deny modes are valid.
7939 	 * Check that no invalid bits are set.
7940 	 */
7941 	if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) ||
7942 	    (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) {
7943 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7944 		rfs4_update_open_sequence(sp->rs_owner);
7945 		rfs4_dbe_unlock(sp->rs_dbe);
7946 		goto end;
7947 	}
7948 
7949 	/*
7950 	 * The new modes must be a subset of the current modes and
7951 	 * the access must specify at least one mode. To test that
7952 	 * the new mode is a subset of the current modes we bitwise
7953 	 * AND them together and check that the result equals the new
7954 	 * mode. For example:
7955 	 * New mode, access == R and current mode, sp->rs_open_access  == RW
7956 	 * access & sp->rs_open_access == R == access, so the new access mode
7957 	 * is valid. Consider access == RW, sp->rs_open_access = R
7958 	 * access & sp->rs_open_access == R != access, so the new access mode
7959 	 * is invalid.
7960 	 */
7961 	if ((access & sp->rs_open_access) != access ||
7962 	    (deny & sp->rs_open_deny) != deny ||
7963 	    (access &
7964 	    (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) {
7965 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7966 		rfs4_update_open_sequence(sp->rs_owner);
7967 		rfs4_dbe_unlock(sp->rs_dbe);
7968 		goto end;
7969 	}
7970 
7971 	/*
7972 	 * Release any share locks associated with this stateID.
7973 	 * Strictly speaking, this violates the spec because the
7974 	 * spec effectively requires that open downgrade be atomic.
7975 	 * At present, fs_shrlock does not have this capability.
7976 	 */
7977 	(void) rfs4_unshare(sp);
7978 
7979 	status = rfs4_share(sp, access, deny);
7980 	if (status != NFS4_OK) {
7981 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7982 		rfs4_update_open_sequence(sp->rs_owner);
7983 		rfs4_dbe_unlock(sp->rs_dbe);
7984 		goto end;
7985 	}
7986 
7987 	fp = sp->rs_finfo;
7988 	rfs4_dbe_lock(fp->rf_dbe);
7989 
7990 	/*
7991 	 * If the current mode has deny read and the new mode
7992 	 * does not, decrement the number of deny read mode bits
7993 	 * and if it goes to zero turn off the deny read bit
7994 	 * on the file.
7995 	 */
7996 	if ((sp->rs_open_deny & OPEN4_SHARE_DENY_READ) &&
7997 	    (deny & OPEN4_SHARE_DENY_READ) == 0) {
7998 		fp->rf_deny_read--;
7999 		if (fp->rf_deny_read == 0)
8000 			fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ;
8001 	}
8002 
8003 	/*
8004 	 * If the current mode has deny write and the new mode
8005 	 * does not, decrement the number of deny write mode bits
8006 	 * and if it goes to zero turn off the deny write bit
8007 	 * on the file.
8008 	 */
8009 	if ((sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) &&
8010 	    (deny & OPEN4_SHARE_DENY_WRITE) == 0) {
8011 		fp->rf_deny_write--;
8012 		if (fp->rf_deny_write == 0)
8013 			fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE;
8014 	}
8015 
8016 	/*
8017 	 * If the current mode has access read and the new mode
8018 	 * does not, decrement the number of access read mode bits
8019 	 * and if it goes to zero turn off the access read bit
8020 	 * on the file.  set fflags to FREAD for the call to
8021 	 * vn_open_downgrade().
8022 	 */
8023 	if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) &&
8024 	    (access & OPEN4_SHARE_ACCESS_READ) == 0) {
8025 		fp->rf_access_read--;
8026 		if (fp->rf_access_read == 0)
8027 			fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ;
8028 		fflags |= FREAD;
8029 	}
8030 
8031 	/*
8032 	 * If the current mode has access write and the new mode
8033 	 * does not, decrement the number of access write mode bits
8034 	 * and if it goes to zero turn off the access write bit
8035 	 * on the file.  set fflags to FWRITE for the call to
8036 	 * vn_open_downgrade().
8037 	 */
8038 	if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) &&
8039 	    (access & OPEN4_SHARE_ACCESS_WRITE) == 0) {
8040 		fp->rf_access_write--;
8041 		if (fp->rf_access_write == 0)
8042 			fp->rf_share_deny &= ~OPEN4_SHARE_ACCESS_WRITE;
8043 		fflags |= FWRITE;
8044 	}
8045 
8046 	/* Check that the file is still accessible */
8047 	ASSERT(fp->rf_share_access);
8048 
8049 	rfs4_dbe_unlock(fp->rf_dbe);
8050 
8051 	/* now set the new open access and deny modes */
8052 	sp->rs_open_access = access;
8053 	sp->rs_open_deny = deny;
8054 
8055 	/*
8056 	 * we successfully downgraded the share lock, now we need to downgrade
8057 	 * the open. it is possible that the downgrade was only for a deny
8058 	 * mode and we have nothing else to do.
8059 	 */
8060 	if ((fflags & (FREAD|FWRITE)) != 0)
8061 		vn_open_downgrade(cs->vp, fflags);
8062 
8063 	/* Update the stateid */
8064 	next_stateid(&sp->rs_stateid);
8065 	resp->open_stateid = sp->rs_stateid.stateid;
8066 
8067 	rfs4_dbe_unlock(sp->rs_dbe);
8068 
8069 	*cs->statusp = resp->status = NFS4_OK;
8070 	/* Update the lease */
8071 	rfs4_update_lease(sp->rs_owner->ro_client);
8072 	/* And the sequence */
8073 	rfs4_update_open_sequence(sp->rs_owner);
8074 	rfs4_update_open_resp(sp->rs_owner, resop, NULL);
8075 
8076 end:
8077 	rfs4_sw_exit(&sp->rs_owner->ro_sw);
8078 	rfs4_state_rele(sp);
8079 out:
8080 	DTRACE_NFSV4_2(op__open__downgrade__done, struct compound_state *, cs,
8081 	    OPEN_DOWNGRADE4res *, resp);
8082 }
8083 
8084 static void *
8085 memstr(const void *s1, const char *s2, size_t n)
8086 {
8087 	size_t l = strlen(s2);
8088 	char *p = (char *)s1;
8089 
8090 	while (n >= l) {
8091 		if (bcmp(p, s2, l) == 0)
8092 			return (p);
8093 		p++;
8094 		n--;
8095 	}
8096 
8097 	return (NULL);
8098 }
8099 
8100 /*
8101  * The logic behind this function is detailed in the NFSv4 RFC in the
8102  * SETCLIENTID operation description under IMPLEMENTATION.  Refer to
8103  * that section for explicit guidance to server behavior for
8104  * SETCLIENTID.
8105  */
8106 void
8107 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop,
8108     struct svc_req *req, struct compound_state *cs)
8109 {
8110 	SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid;
8111 	SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid;
8112 	rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed;
8113 	rfs4_clntip_t *ci;
8114 	bool_t create;
8115 	char *addr, *netid;
8116 	int len;
8117 
8118 	DTRACE_NFSV4_2(op__setclientid__start, struct compound_state *, cs,
8119 	    SETCLIENTID4args *, args);
8120 retry:
8121 	newcp = cp_confirmed = cp_unconfirmed = NULL;
8122 
8123 	/*
8124 	 * Save the caller's IP address
8125 	 */
8126 	args->client.cl_addr =
8127 	    (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
8128 
8129 	/*
8130 	 * Record if it is a Solaris client that cannot handle referrals.
8131 	 */
8132 	if (memstr(args->client.id_val, "Solaris", args->client.id_len) &&
8133 	    !memstr(args->client.id_val, "+referrals", args->client.id_len)) {
8134 		/* Add a "yes, it's downrev" record */
8135 		create = TRUE;
8136 		ci = rfs4_find_clntip(args->client.cl_addr, &create);
8137 		ASSERT(ci != NULL);
8138 		rfs4_dbe_rele(ci->ri_dbe);
8139 	} else {
8140 		/* Remove any previous record */
8141 		rfs4_invalidate_clntip(args->client.cl_addr);
8142 	}
8143 
8144 	/*
8145 	 * In search of an EXISTING client matching the incoming
8146 	 * request to establish a new client identifier at the server
8147 	 */
8148 	create = TRUE;
8149 	cp = rfs4_findclient(&args->client, &create, NULL);
8150 
8151 	/* Should never happen */
8152 	ASSERT(cp != NULL);
8153 
8154 	if (cp == NULL) {
8155 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
8156 		goto out;
8157 	}
8158 
8159 	/*
8160 	 * Easiest case. Client identifier is newly created and is
8161 	 * unconfirmed.  Also note that for this case, no other
8162 	 * entries exist for the client identifier.  Nothing else to
8163 	 * check.  Just setup the response and respond.
8164 	 */
8165 	if (create) {
8166 		*cs->statusp = res->status = NFS4_OK;
8167 		res->SETCLIENTID4res_u.resok4.clientid = cp->rc_clientid;
8168 		res->SETCLIENTID4res_u.resok4.setclientid_confirm =
8169 		    cp->rc_confirm_verf;
8170 		/* Setup callback information; CB_NULL confirmation later */
8171 		rfs4_client_setcb(cp, &args->callback, args->callback_ident);
8172 
8173 		rfs4_client_rele(cp);
8174 		goto out;
8175 	}
8176 
8177 	/*
8178 	 * An existing, confirmed client may exist but it may not have
8179 	 * been active for at least one lease period.  If so, then
8180 	 * "close" the client and create a new client identifier
8181 	 */
8182 	if (rfs4_lease_expired(cp)) {
8183 		rfs4_client_close(cp);
8184 		goto retry;
8185 	}
8186 
8187 	if (cp->rc_need_confirm == TRUE)
8188 		cp_unconfirmed = cp;
8189 	else
8190 		cp_confirmed = cp;
8191 
8192 	cp = NULL;
8193 
8194 	/*
8195 	 * We have a confirmed client, now check for an
8196 	 * unconfimred entry
8197 	 */
8198 	if (cp_confirmed) {
8199 		/* If creds don't match then client identifier is inuse */
8200 		if (!creds_ok(cp_confirmed->rc_cr_set, req, cs)) {
8201 			rfs4_cbinfo_t *cbp;
8202 			/*
8203 			 * Some one else has established this client
8204 			 * id. Try and say * who they are. We will use
8205 			 * the call back address supplied by * the
8206 			 * first client.
8207 			 */
8208 			*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
8209 
8210 			addr = netid = NULL;
8211 
8212 			cbp = &cp_confirmed->rc_cbinfo;
8213 			if (cbp->cb_callback.cb_location.r_addr &&
8214 			    cbp->cb_callback.cb_location.r_netid) {
8215 				cb_client4 *cbcp = &cbp->cb_callback;
8216 
8217 				len = strlen(cbcp->cb_location.r_addr)+1;
8218 				addr = kmem_alloc(len, KM_SLEEP);
8219 				bcopy(cbcp->cb_location.r_addr, addr, len);
8220 				len = strlen(cbcp->cb_location.r_netid)+1;
8221 				netid = kmem_alloc(len, KM_SLEEP);
8222 				bcopy(cbcp->cb_location.r_netid, netid, len);
8223 			}
8224 
8225 			res->SETCLIENTID4res_u.client_using.r_addr = addr;
8226 			res->SETCLIENTID4res_u.client_using.r_netid = netid;
8227 
8228 			rfs4_client_rele(cp_confirmed);
8229 		}
8230 
8231 		/*
8232 		 * Confirmed, creds match, and verifier matches; must
8233 		 * be an update of the callback info
8234 		 */
8235 		if (cp_confirmed->rc_nfs_client.verifier ==
8236 		    args->client.verifier) {
8237 			/* Setup callback information */
8238 			rfs4_client_setcb(cp_confirmed, &args->callback,
8239 			    args->callback_ident);
8240 
8241 			/* everything okay -- move ahead */
8242 			*cs->statusp = res->status = NFS4_OK;
8243 			res->SETCLIENTID4res_u.resok4.clientid =
8244 			    cp_confirmed->rc_clientid;
8245 
8246 			/* update the confirm_verifier and return it */
8247 			rfs4_client_scv_next(cp_confirmed);
8248 			res->SETCLIENTID4res_u.resok4.setclientid_confirm =
8249 			    cp_confirmed->rc_confirm_verf;
8250 
8251 			rfs4_client_rele(cp_confirmed);
8252 			goto out;
8253 		}
8254 
8255 		/*
8256 		 * Creds match but the verifier doesn't.  Must search
8257 		 * for an unconfirmed client that would be replaced by
8258 		 * this request.
8259 		 */
8260 		create = FALSE;
8261 		cp_unconfirmed = rfs4_findclient(&args->client, &create,
8262 		    cp_confirmed);
8263 	}
8264 
8265 	/*
8266 	 * At this point, we have taken care of the brand new client
8267 	 * struct, INUSE case, update of an existing, and confirmed
8268 	 * client struct.
8269 	 */
8270 
8271 	/*
8272 	 * check to see if things have changed while we originally
8273 	 * picked up the client struct.  If they have, then return and
8274 	 * retry the processing of this SETCLIENTID request.
8275 	 */
8276 	if (cp_unconfirmed) {
8277 		rfs4_dbe_lock(cp_unconfirmed->rc_dbe);
8278 		if (!cp_unconfirmed->rc_need_confirm) {
8279 			rfs4_dbe_unlock(cp_unconfirmed->rc_dbe);
8280 			rfs4_client_rele(cp_unconfirmed);
8281 			if (cp_confirmed)
8282 				rfs4_client_rele(cp_confirmed);
8283 			goto retry;
8284 		}
8285 		/* do away with the old unconfirmed one */
8286 		rfs4_dbe_invalidate(cp_unconfirmed->rc_dbe);
8287 		rfs4_dbe_unlock(cp_unconfirmed->rc_dbe);
8288 		rfs4_client_rele(cp_unconfirmed);
8289 		cp_unconfirmed = NULL;
8290 	}
8291 
8292 	/*
8293 	 * This search will temporarily hide the confirmed client
8294 	 * struct while a new client struct is created as the
8295 	 * unconfirmed one.
8296 	 */
8297 	create = TRUE;
8298 	newcp = rfs4_findclient(&args->client, &create, cp_confirmed);
8299 
8300 	ASSERT(newcp != NULL);
8301 
8302 	if (newcp == NULL) {
8303 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
8304 		rfs4_client_rele(cp_confirmed);
8305 		goto out;
8306 	}
8307 
8308 	/*
8309 	 * If one was not created, then a similar request must be in
8310 	 * process so release and start over with this one
8311 	 */
8312 	if (create != TRUE) {
8313 		rfs4_client_rele(newcp);
8314 		if (cp_confirmed)
8315 			rfs4_client_rele(cp_confirmed);
8316 		goto retry;
8317 	}
8318 
8319 	*cs->statusp = res->status = NFS4_OK;
8320 	res->SETCLIENTID4res_u.resok4.clientid = newcp->rc_clientid;
8321 	res->SETCLIENTID4res_u.resok4.setclientid_confirm =
8322 	    newcp->rc_confirm_verf;
8323 	/* Setup callback information; CB_NULL confirmation later */
8324 	rfs4_client_setcb(newcp, &args->callback, args->callback_ident);
8325 
8326 	newcp->rc_cp_confirmed = cp_confirmed;
8327 
8328 	rfs4_client_rele(newcp);
8329 
8330 out:
8331 	DTRACE_NFSV4_2(op__setclientid__done, struct compound_state *, cs,
8332 	    SETCLIENTID4res *, res);
8333 }
8334 
8335 /*ARGSUSED*/
8336 void
8337 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
8338     struct svc_req *req, struct compound_state *cs)
8339 {
8340 	SETCLIENTID_CONFIRM4args *args =
8341 	    &argop->nfs_argop4_u.opsetclientid_confirm;
8342 	SETCLIENTID_CONFIRM4res *res =
8343 	    &resop->nfs_resop4_u.opsetclientid_confirm;
8344 	rfs4_client_t *cp, *cptoclose = NULL;
8345 
8346 	DTRACE_NFSV4_2(op__setclientid__confirm__start,
8347 	    struct compound_state *, cs,
8348 	    SETCLIENTID_CONFIRM4args *, args);
8349 
8350 	*cs->statusp = res->status = NFS4_OK;
8351 
8352 	cp = rfs4_findclient_by_id(args->clientid, TRUE);
8353 
8354 	if (cp == NULL) {
8355 		*cs->statusp = res->status =
8356 		    rfs4_check_clientid(&args->clientid, 1);
8357 		goto out;
8358 	}
8359 
8360 	if (!creds_ok(cp, req, cs)) {
8361 		*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
8362 		rfs4_client_rele(cp);
8363 		goto out;
8364 	}
8365 
8366 	/* If the verifier doesn't match, the record doesn't match */
8367 	if (cp->rc_confirm_verf != args->setclientid_confirm) {
8368 		*cs->statusp = res->status = NFS4ERR_STALE_CLIENTID;
8369 		rfs4_client_rele(cp);
8370 		goto out;
8371 	}
8372 
8373 	rfs4_dbe_lock(cp->rc_dbe);
8374 	cp->rc_need_confirm = FALSE;
8375 	if (cp->rc_cp_confirmed) {
8376 		cptoclose = cp->rc_cp_confirmed;
8377 		cptoclose->rc_ss_remove = 1;
8378 		cp->rc_cp_confirmed = NULL;
8379 	}
8380 
8381 	/*
8382 	 * Update the client's associated server instance, if it's changed
8383 	 * since the client was created.
8384 	 */
8385 	if (rfs4_servinst(cp) != rfs4_cur_servinst)
8386 		rfs4_servinst_assign(cp, rfs4_cur_servinst);
8387 
8388 	/*
8389 	 * Record clientid in stable storage.
8390 	 * Must be done after server instance has been assigned.
8391 	 */
8392 	rfs4_ss_clid(cp);
8393 
8394 	rfs4_dbe_unlock(cp->rc_dbe);
8395 
8396 	if (cptoclose)
8397 		/* don't need to rele, client_close does it */
8398 		rfs4_client_close(cptoclose);
8399 
8400 	/* If needed, initiate CB_NULL call for callback path */
8401 	rfs4_deleg_cb_check(cp);
8402 	rfs4_update_lease(cp);
8403 
8404 	/*
8405 	 * Check to see if client can perform reclaims
8406 	 */
8407 	rfs4_ss_chkclid(cp);
8408 
8409 	rfs4_client_rele(cp);
8410 
8411 out:
8412 	DTRACE_NFSV4_2(op__setclientid__confirm__done,
8413 	    struct compound_state *, cs,
8414 	    SETCLIENTID_CONFIRM4 *, res);
8415 }
8416 
8417 
8418 /*ARGSUSED*/
8419 void
8420 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop,
8421     struct svc_req *req, struct compound_state *cs)
8422 {
8423 	CLOSE4args *args = &argop->nfs_argop4_u.opclose;
8424 	CLOSE4res *resp = &resop->nfs_resop4_u.opclose;
8425 	rfs4_state_t *sp;
8426 	nfsstat4 status;
8427 
8428 	DTRACE_NFSV4_2(op__close__start, struct compound_state *, cs,
8429 	    CLOSE4args *, args);
8430 
8431 	if (cs->vp == NULL) {
8432 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8433 		goto out;
8434 	}
8435 
8436 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID);
8437 	if (status != NFS4_OK) {
8438 		*cs->statusp = resp->status = status;
8439 		goto out;
8440 	}
8441 
8442 	/* Ensure specified filehandle matches */
8443 	if (cs->vp != sp->rs_finfo->rf_vp) {
8444 		rfs4_state_rele(sp);
8445 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8446 		goto out;
8447 	}
8448 
8449 	/* hold off other access to open_owner while we tinker */
8450 	rfs4_sw_enter(&sp->rs_owner->ro_sw);
8451 
8452 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
8453 	case NFS4_CHECK_STATEID_OKAY:
8454 		if (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
8455 		    resop) != NFS4_CHKSEQ_OKAY) {
8456 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8457 			goto end;
8458 		}
8459 		break;
8460 	case NFS4_CHECK_STATEID_OLD:
8461 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8462 		goto end;
8463 	case NFS4_CHECK_STATEID_BAD:
8464 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8465 		goto end;
8466 	case NFS4_CHECK_STATEID_EXPIRED:
8467 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8468 		goto end;
8469 	case NFS4_CHECK_STATEID_CLOSED:
8470 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8471 		goto end;
8472 	case NFS4_CHECK_STATEID_UNCONFIRMED:
8473 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8474 		goto end;
8475 	case NFS4_CHECK_STATEID_REPLAY:
8476 		/* Check the sequence id for the open owner */
8477 		switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
8478 		    resop)) {
8479 		case NFS4_CHKSEQ_OKAY:
8480 			/*
8481 			 * This is replayed stateid; if seqid matches
8482 			 * next expected, then client is using wrong seqid.
8483 			 */
8484 			/* FALL THROUGH */
8485 		case NFS4_CHKSEQ_BAD:
8486 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8487 			goto end;
8488 		case NFS4_CHKSEQ_REPLAY:
8489 			/*
8490 			 * Note this case is the duplicate case so
8491 			 * resp->status is already set.
8492 			 */
8493 			*cs->statusp = resp->status;
8494 			rfs4_update_lease(sp->rs_owner->ro_client);
8495 			goto end;
8496 		}
8497 		break;
8498 	default:
8499 		ASSERT(FALSE);
8500 		break;
8501 	}
8502 
8503 	rfs4_dbe_lock(sp->rs_dbe);
8504 
8505 	/* Update the stateid. */
8506 	next_stateid(&sp->rs_stateid);
8507 	resp->open_stateid = sp->rs_stateid.stateid;
8508 
8509 	rfs4_dbe_unlock(sp->rs_dbe);
8510 
8511 	rfs4_update_lease(sp->rs_owner->ro_client);
8512 	rfs4_update_open_sequence(sp->rs_owner);
8513 	rfs4_update_open_resp(sp->rs_owner, resop, NULL);
8514 
8515 	rfs4_state_close(sp, FALSE, FALSE, cs->cr);
8516 
8517 	*cs->statusp = resp->status = status;
8518 
8519 end:
8520 	rfs4_sw_exit(&sp->rs_owner->ro_sw);
8521 	rfs4_state_rele(sp);
8522 out:
8523 	DTRACE_NFSV4_2(op__close__done, struct compound_state *, cs,
8524 	    CLOSE4res *, resp);
8525 }
8526 
8527 /*
8528  * Manage the counts on the file struct and close all file locks
8529  */
8530 /*ARGSUSED*/
8531 void
8532 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr,
8533     bool_t close_of_client)
8534 {
8535 	rfs4_file_t *fp = sp->rs_finfo;
8536 	rfs4_lo_state_t *lsp;
8537 	int fflags = 0;
8538 
8539 	/*
8540 	 * If this call is part of the larger closing down of client
8541 	 * state then it is just easier to release all locks
8542 	 * associated with this client instead of going through each
8543 	 * individual file and cleaning locks there.
8544 	 */
8545 	if (close_of_client) {
8546 		if (sp->rs_owner->ro_client->rc_unlksys_completed == FALSE &&
8547 		    !list_is_empty(&sp->rs_lostatelist) &&
8548 		    sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) {
8549 			/* Is the PxFS kernel module loaded? */
8550 			if (lm_remove_file_locks != NULL) {
8551 				int new_sysid;
8552 
8553 				/* Encode the cluster nodeid in new sysid */
8554 				new_sysid = sp->rs_owner->ro_client->rc_sysidt;
8555 				lm_set_nlmid_flk(&new_sysid);
8556 
8557 				/*
8558 				 * This PxFS routine removes file locks for a
8559 				 * client over all nodes of a cluster.
8560 				 */
8561 				NFS4_DEBUG(rfs4_debug, (CE_NOTE,
8562 				    "lm_remove_file_locks(sysid=0x%x)\n",
8563 				    new_sysid));
8564 				(*lm_remove_file_locks)(new_sysid);
8565 			} else {
8566 				struct flock64 flk;
8567 
8568 				/* Release all locks for this client */
8569 				flk.l_type = F_UNLKSYS;
8570 				flk.l_whence = 0;
8571 				flk.l_start = 0;
8572 				flk.l_len = 0;
8573 				flk.l_sysid =
8574 				    sp->rs_owner->ro_client->rc_sysidt;
8575 				flk.l_pid = 0;
8576 				(void) VOP_FRLOCK(sp->rs_finfo->rf_vp, F_SETLK,
8577 				    &flk, F_REMOTELOCK | FREAD | FWRITE,
8578 				    (u_offset_t)0, NULL, CRED(), NULL);
8579 			}
8580 
8581 			sp->rs_owner->ro_client->rc_unlksys_completed = TRUE;
8582 		}
8583 	}
8584 
8585 	/*
8586 	 * Release all locks on this file by this lock owner or at
8587 	 * least mark the locks as having been released
8588 	 */
8589 	for (lsp = list_head(&sp->rs_lostatelist); lsp != NULL;
8590 	    lsp = list_next(&sp->rs_lostatelist, lsp)) {
8591 		lsp->rls_locks_cleaned = TRUE;
8592 
8593 		/* Was this already taken care of above? */
8594 		if (!close_of_client &&
8595 		    sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID)
8596 			(void) cleanlocks(sp->rs_finfo->rf_vp,
8597 			    lsp->rls_locker->rl_pid,
8598 			    lsp->rls_locker->rl_client->rc_sysidt);
8599 	}
8600 
8601 	/*
8602 	 * Release any shrlocks associated with this open state ID.
8603 	 * This must be done before the rfs4_state gets marked closed.
8604 	 */
8605 	if (sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID)
8606 		(void) rfs4_unshare(sp);
8607 
8608 	if (sp->rs_open_access) {
8609 		rfs4_dbe_lock(fp->rf_dbe);
8610 
8611 		/*
8612 		 * Decrement the count for each access and deny bit that this
8613 		 * state has contributed to the file.
8614 		 * If the file counts go to zero
8615 		 * clear the appropriate bit in the appropriate mask.
8616 		 */
8617 		if (sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) {
8618 			fp->rf_access_read--;
8619 			fflags |= FREAD;
8620 			if (fp->rf_access_read == 0)
8621 				fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ;
8622 		}
8623 		if (sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) {
8624 			fp->rf_access_write--;
8625 			fflags |= FWRITE;
8626 			if (fp->rf_access_write == 0)
8627 				fp->rf_share_access &=
8628 				    ~OPEN4_SHARE_ACCESS_WRITE;
8629 		}
8630 		if (sp->rs_open_deny & OPEN4_SHARE_DENY_READ) {
8631 			fp->rf_deny_read--;
8632 			if (fp->rf_deny_read == 0)
8633 				fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ;
8634 		}
8635 		if (sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) {
8636 			fp->rf_deny_write--;
8637 			if (fp->rf_deny_write == 0)
8638 				fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE;
8639 		}
8640 
8641 		(void) VOP_CLOSE(fp->rf_vp, fflags, 1, (offset_t)0, cr, NULL);
8642 
8643 		rfs4_dbe_unlock(fp->rf_dbe);
8644 
8645 		sp->rs_open_access = 0;
8646 		sp->rs_open_deny = 0;
8647 	}
8648 }
8649 
8650 /*
8651  * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure.
8652  */
8653 static nfsstat4
8654 lock_denied(LOCK4denied *dp, struct flock64 *flk)
8655 {
8656 	rfs4_lockowner_t *lo;
8657 	rfs4_client_t *cp;
8658 	uint32_t len;
8659 
8660 	lo = rfs4_findlockowner_by_pid(flk->l_pid);
8661 	if (lo != NULL) {
8662 		cp = lo->rl_client;
8663 		if (rfs4_lease_expired(cp)) {
8664 			rfs4_lockowner_rele(lo);
8665 			rfs4_dbe_hold(cp->rc_dbe);
8666 			rfs4_client_close(cp);
8667 			return (NFS4ERR_EXPIRED);
8668 		}
8669 		dp->owner.clientid = lo->rl_owner.clientid;
8670 		len = lo->rl_owner.owner_len;
8671 		dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
8672 		bcopy(lo->rl_owner.owner_val, dp->owner.owner_val, len);
8673 		dp->owner.owner_len = len;
8674 		rfs4_lockowner_rele(lo);
8675 		goto finish;
8676 	}
8677 
8678 	/*
8679 	 * Its not a NFS4 lock. We take advantage that the upper 32 bits
8680 	 * of the client id contain the boot time for a NFS4 lock. So we
8681 	 * fabricate and identity by setting clientid to the sysid, and
8682 	 * the lock owner to the pid.
8683 	 */
8684 	dp->owner.clientid = flk->l_sysid;
8685 	len = sizeof (pid_t);
8686 	dp->owner.owner_len = len;
8687 	dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
8688 	bcopy(&flk->l_pid, dp->owner.owner_val, len);
8689 finish:
8690 	dp->offset = flk->l_start;
8691 	dp->length = flk->l_len;
8692 
8693 	if (flk->l_type == F_RDLCK)
8694 		dp->locktype = READ_LT;
8695 	else if (flk->l_type == F_WRLCK)
8696 		dp->locktype = WRITE_LT;
8697 	else
8698 		return (NFS4ERR_INVAL);	/* no mapping from POSIX ltype to v4 */
8699 
8700 	return (NFS4_OK);
8701 }
8702 
8703 /*
8704  * The NFSv4.0 LOCK operation does not support the blocking lock (at the
8705  * NFSv4.0 protocol level) so the client needs to resend the LOCK request in a
8706  * case the lock is denied by the NFSv4.0 server.  NFSv4.0 clients are prepared
8707  * for that (obviously); they are sending the LOCK requests with some delays
8708  * between the attempts.  See nfs4frlock() and nfs4_block_and_wait() for the
8709  * locking and delay implementation at the client side.
8710  *
8711  * To make the life of the clients easier, the NFSv4.0 server tries to do some
8712  * fast retries on its own (the for loop below) in a hope the lock will be
8713  * available soon.  And if not, the client won't need to resend the LOCK
8714  * requests so fast to check the lock availability.  This basically saves some
8715  * network traffic and tries to make sure the client gets the lock ASAP.
8716  */
8717 static int
8718 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred)
8719 {
8720 	int error;
8721 	struct flock64 flk;
8722 	int i;
8723 	clock_t delaytime;
8724 	int cmd;
8725 	int spin_cnt = 0;
8726 
8727 	cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK;
8728 retry:
8729 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
8730 
8731 	for (i = 0; i < rfs4_maxlock_tries; i++) {
8732 		LOCK_PRINT(rfs4_debug, "setlock", cmd, flock);
8733 		error = VOP_FRLOCK(vp, cmd,
8734 		    flock, flag, (u_offset_t)0, NULL, cred, NULL);
8735 
8736 		if (error != EAGAIN && error != EACCES)
8737 			break;
8738 
8739 		if (i < rfs4_maxlock_tries - 1) {
8740 			delay(delaytime);
8741 			delaytime *= 2;
8742 		}
8743 	}
8744 
8745 	if (error == EAGAIN || error == EACCES) {
8746 		/* Get the owner of the lock */
8747 		flk = *flock;
8748 		LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk);
8749 		if (VOP_FRLOCK(vp, F_GETLK, &flk, flag, 0, NULL, cred,
8750 		    NULL) == 0) {
8751 			/*
8752 			 * There's a race inherent in the current VOP_FRLOCK
8753 			 * design where:
8754 			 * a: "other guy" takes a lock that conflicts with a
8755 			 * lock we want
8756 			 * b: we attempt to take our lock (non-blocking) and
8757 			 * the attempt fails.
8758 			 * c: "other guy" releases the conflicting lock
8759 			 * d: we ask what lock conflicts with the lock we want,
8760 			 * getting F_UNLCK (no lock blocks us)
8761 			 *
8762 			 * If we retry the non-blocking lock attempt in this
8763 			 * case (restart at step 'b') there's some possibility
8764 			 * that many such attempts might fail.  However a test
8765 			 * designed to actually provoke this race shows that
8766 			 * the vast majority of cases require no retry, and
8767 			 * only a few took as many as three retries.  Here's
8768 			 * the test outcome:
8769 			 *
8770 			 *	   number of retries    how many times we needed
8771 			 *				that many retries
8772 			 *	   0			79461
8773 			 *	   1			  862
8774 			 *	   2			   49
8775 			 *	   3			    5
8776 			 *
8777 			 * Given those empirical results, we arbitrarily limit
8778 			 * the retry count to ten.
8779 			 *
8780 			 * If we actually make to ten retries and give up,
8781 			 * nothing catastrophic happens, but we're unable to
8782 			 * return the information about the conflicting lock to
8783 			 * the NFS client.  That's an acceptable trade off vs.
8784 			 * letting this retry loop run forever.
8785 			 */
8786 			if (flk.l_type == F_UNLCK) {
8787 				if (spin_cnt++ < 10) {
8788 					/* No longer locked, retry */
8789 					goto retry;
8790 				}
8791 			} else {
8792 				*flock = flk;
8793 				LOCK_PRINT(rfs4_debug, "setlock(blocking lock)",
8794 				    F_GETLK, &flk);
8795 			}
8796 		}
8797 	}
8798 
8799 	return (error);
8800 }
8801 
8802 /*ARGSUSED*/
8803 static nfsstat4
8804 rfs4_do_lock(rfs4_lo_state_t *lsp, nfs_lock_type4 locktype,
8805     offset4 offset, length4 length, cred_t *cred, nfs_resop4 *resop)
8806 {
8807 	nfsstat4 status;
8808 	rfs4_lockowner_t *lo = lsp->rls_locker;
8809 	rfs4_state_t *sp = lsp->rls_state;
8810 	struct flock64 flock;
8811 	int16_t ltype;
8812 	int flag;
8813 	int error;
8814 	sysid_t sysid;
8815 	LOCK4res *lres;
8816 	vnode_t *vp;
8817 
8818 	if (rfs4_lease_expired(lo->rl_client)) {
8819 		return (NFS4ERR_EXPIRED);
8820 	}
8821 
8822 	if ((status = rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK)
8823 		return (status);
8824 
8825 	/* Check for zero length. To lock to end of file use all ones for V4 */
8826 	if (length == 0)
8827 		return (NFS4ERR_INVAL);
8828 	else if (length == (length4)(~0))
8829 		length = 0;		/* Posix to end of file  */
8830 
8831 retry:
8832 	rfs4_dbe_lock(sp->rs_dbe);
8833 	if (sp->rs_closed == TRUE) {
8834 		rfs4_dbe_unlock(sp->rs_dbe);
8835 		return (NFS4ERR_OLD_STATEID);
8836 	}
8837 
8838 	if (resop->resop != OP_LOCKU) {
8839 		switch (locktype) {
8840 		case READ_LT:
8841 		case READW_LT:
8842 			if ((sp->rs_share_access
8843 			    & OPEN4_SHARE_ACCESS_READ) == 0) {
8844 				rfs4_dbe_unlock(sp->rs_dbe);
8845 
8846 				return (NFS4ERR_OPENMODE);
8847 			}
8848 			ltype = F_RDLCK;
8849 			break;
8850 		case WRITE_LT:
8851 		case WRITEW_LT:
8852 			if ((sp->rs_share_access
8853 			    & OPEN4_SHARE_ACCESS_WRITE) == 0) {
8854 				rfs4_dbe_unlock(sp->rs_dbe);
8855 
8856 				return (NFS4ERR_OPENMODE);
8857 			}
8858 			ltype = F_WRLCK;
8859 			break;
8860 		}
8861 	} else
8862 		ltype = F_UNLCK;
8863 
8864 	flock.l_type = ltype;
8865 	flock.l_whence = 0;		/* SEEK_SET */
8866 	flock.l_start = offset;
8867 	flock.l_len = length;
8868 	flock.l_sysid = sysid;
8869 	flock.l_pid = lsp->rls_locker->rl_pid;
8870 
8871 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
8872 	if (flock.l_len < 0 || flock.l_start < 0) {
8873 		rfs4_dbe_unlock(sp->rs_dbe);
8874 		return (NFS4ERR_INVAL);
8875 	}
8876 
8877 	/*
8878 	 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and
8879 	 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE.
8880 	 */
8881 	flag = (int)sp->rs_share_access | F_REMOTELOCK;
8882 
8883 	vp = sp->rs_finfo->rf_vp;
8884 	VN_HOLD(vp);
8885 
8886 	/*
8887 	 * We need to unlock sp before we call the underlying filesystem to
8888 	 * acquire the file lock.
8889 	 */
8890 	rfs4_dbe_unlock(sp->rs_dbe);
8891 
8892 	error = setlock(vp, &flock, flag, cred);
8893 
8894 	/*
8895 	 * Make sure the file is still open.  In a case the file was closed in
8896 	 * the meantime, clean the lock we acquired using the setlock() call
8897 	 * above, and return the appropriate error.
8898 	 */
8899 	rfs4_dbe_lock(sp->rs_dbe);
8900 	if (sp->rs_closed == TRUE) {
8901 		cleanlocks(vp, lsp->rls_locker->rl_pid, sysid);
8902 		rfs4_dbe_unlock(sp->rs_dbe);
8903 
8904 		VN_RELE(vp);
8905 
8906 		return (NFS4ERR_OLD_STATEID);
8907 	}
8908 	rfs4_dbe_unlock(sp->rs_dbe);
8909 
8910 	VN_RELE(vp);
8911 
8912 	if (error == 0) {
8913 		rfs4_dbe_lock(lsp->rls_dbe);
8914 		next_stateid(&lsp->rls_lockid);
8915 		rfs4_dbe_unlock(lsp->rls_dbe);
8916 	}
8917 
8918 	/*
8919 	 * N.B. We map error values to nfsv4 errors. This is differrent
8920 	 * than puterrno4 routine.
8921 	 */
8922 	switch (error) {
8923 	case 0:
8924 		status = NFS4_OK;
8925 		break;
8926 	case EAGAIN:
8927 	case EACCES:		/* Old value */
8928 		/* Can only get here if op is OP_LOCK */
8929 		ASSERT(resop->resop == OP_LOCK);
8930 		lres = &resop->nfs_resop4_u.oplock;
8931 		status = NFS4ERR_DENIED;
8932 		if (lock_denied(&lres->LOCK4res_u.denied, &flock)
8933 		    == NFS4ERR_EXPIRED)
8934 			goto retry;
8935 		break;
8936 	case ENOLCK:
8937 		status = NFS4ERR_DELAY;
8938 		break;
8939 	case EOVERFLOW:
8940 		status = NFS4ERR_INVAL;
8941 		break;
8942 	case EINVAL:
8943 		status = NFS4ERR_NOTSUPP;
8944 		break;
8945 	default:
8946 		status = NFS4ERR_SERVERFAULT;
8947 		break;
8948 	}
8949 
8950 	return (status);
8951 }
8952 
8953 /*ARGSUSED*/
8954 void
8955 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop,
8956     struct svc_req *req, struct compound_state *cs)
8957 {
8958 	LOCK4args *args = &argop->nfs_argop4_u.oplock;
8959 	LOCK4res *resp = &resop->nfs_resop4_u.oplock;
8960 	nfsstat4 status;
8961 	stateid4 *stateid;
8962 	rfs4_lockowner_t *lo;
8963 	rfs4_client_t *cp;
8964 	rfs4_state_t *sp = NULL;
8965 	rfs4_lo_state_t *lsp = NULL;
8966 	bool_t ls_sw_held = FALSE;
8967 	bool_t create = TRUE;
8968 	bool_t lcreate = TRUE;
8969 	bool_t dup_lock = FALSE;
8970 	int rc;
8971 
8972 	DTRACE_NFSV4_2(op__lock__start, struct compound_state *, cs,
8973 	    LOCK4args *, args);
8974 
8975 	if (cs->vp == NULL) {
8976 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8977 		DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8978 		    cs, LOCK4res *, resp);
8979 		return;
8980 	}
8981 
8982 	if (args->locker.new_lock_owner) {
8983 		/* Create a new lockowner for this instance */
8984 		open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner;
8985 
8986 		NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner"));
8987 
8988 		stateid = &olo->open_stateid;
8989 		status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID);
8990 		if (status != NFS4_OK) {
8991 			NFS4_DEBUG(rfs4_debug,
8992 			    (CE_NOTE, "Get state failed in lock %d", status));
8993 			*cs->statusp = resp->status = status;
8994 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8995 			    cs, LOCK4res *, resp);
8996 			return;
8997 		}
8998 
8999 		/* Ensure specified filehandle matches */
9000 		if (cs->vp != sp->rs_finfo->rf_vp) {
9001 			rfs4_state_rele(sp);
9002 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9003 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
9004 			    cs, LOCK4res *, resp);
9005 			return;
9006 		}
9007 
9008 		/* hold off other access to open_owner while we tinker */
9009 		rfs4_sw_enter(&sp->rs_owner->ro_sw);
9010 
9011 		switch (rc = rfs4_check_stateid_seqid(sp, stateid)) {
9012 		case NFS4_CHECK_STATEID_OLD:
9013 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9014 			goto end;
9015 		case NFS4_CHECK_STATEID_BAD:
9016 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9017 			goto end;
9018 		case NFS4_CHECK_STATEID_EXPIRED:
9019 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
9020 			goto end;
9021 		case NFS4_CHECK_STATEID_UNCONFIRMED:
9022 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9023 			goto end;
9024 		case NFS4_CHECK_STATEID_CLOSED:
9025 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9026 			goto end;
9027 		case NFS4_CHECK_STATEID_OKAY:
9028 		case NFS4_CHECK_STATEID_REPLAY:
9029 			switch (rfs4_check_olo_seqid(olo->open_seqid,
9030 			    sp->rs_owner, resop)) {
9031 			case NFS4_CHKSEQ_OKAY:
9032 				if (rc == NFS4_CHECK_STATEID_OKAY)
9033 					break;
9034 				/*
9035 				 * This is replayed stateid; if seqid
9036 				 * matches next expected, then client
9037 				 * is using wrong seqid.
9038 				 */
9039 				/* FALLTHROUGH */
9040 			case NFS4_CHKSEQ_BAD:
9041 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9042 				goto end;
9043 			case NFS4_CHKSEQ_REPLAY:
9044 				/* This is a duplicate LOCK request */
9045 				dup_lock = TRUE;
9046 
9047 				/*
9048 				 * For a duplicate we do not want to
9049 				 * create a new lockowner as it should
9050 				 * already exist.
9051 				 * Turn off the lockowner create flag.
9052 				 */
9053 				lcreate = FALSE;
9054 			}
9055 			break;
9056 		}
9057 
9058 		lo = rfs4_findlockowner(&olo->lock_owner, &lcreate);
9059 		if (lo == NULL) {
9060 			NFS4_DEBUG(rfs4_debug,
9061 			    (CE_NOTE, "rfs4_op_lock: no lock owner"));
9062 			*cs->statusp = resp->status = NFS4ERR_RESOURCE;
9063 			goto end;
9064 		}
9065 
9066 		lsp = rfs4_findlo_state_by_owner(lo, sp, &create);
9067 		if (lsp == NULL) {
9068 			rfs4_update_lease(sp->rs_owner->ro_client);
9069 			/*
9070 			 * Only update theh open_seqid if this is not
9071 			 * a duplicate request
9072 			 */
9073 			if (dup_lock == FALSE) {
9074 				rfs4_update_open_sequence(sp->rs_owner);
9075 			}
9076 
9077 			NFS4_DEBUG(rfs4_debug,
9078 			    (CE_NOTE, "rfs4_op_lock: no state"));
9079 			*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
9080 			rfs4_update_open_resp(sp->rs_owner, resop, NULL);
9081 			rfs4_lockowner_rele(lo);
9082 			goto end;
9083 		}
9084 
9085 		/*
9086 		 * This is the new_lock_owner branch and the client is
9087 		 * supposed to be associating a new lock_owner with
9088 		 * the open file at this point.  If we find that a
9089 		 * lock_owner/state association already exists and a
9090 		 * successful LOCK request was returned to the client,
9091 		 * an error is returned to the client since this is
9092 		 * not appropriate.  The client should be using the
9093 		 * existing lock_owner branch.
9094 		 */
9095 		if (dup_lock == FALSE && create == FALSE) {
9096 			if (lsp->rls_lock_completed == TRUE) {
9097 				*cs->statusp =
9098 				    resp->status = NFS4ERR_BAD_SEQID;
9099 				rfs4_lockowner_rele(lo);
9100 				goto end;
9101 			}
9102 		}
9103 
9104 		rfs4_update_lease(sp->rs_owner->ro_client);
9105 
9106 		/*
9107 		 * Only update theh open_seqid if this is not
9108 		 * a duplicate request
9109 		 */
9110 		if (dup_lock == FALSE) {
9111 			rfs4_update_open_sequence(sp->rs_owner);
9112 		}
9113 
9114 		/*
9115 		 * If this is a duplicate lock request, just copy the
9116 		 * previously saved reply and return.
9117 		 */
9118 		if (dup_lock == TRUE) {
9119 			/* verify that lock_seqid's match */
9120 			if (lsp->rls_seqid != olo->lock_seqid) {
9121 				NFS4_DEBUG(rfs4_debug,
9122 				    (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad"
9123 				    "lsp->seqid=%d old->seqid=%d",
9124 				    lsp->rls_seqid, olo->lock_seqid));
9125 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9126 			} else {
9127 				rfs4_copy_reply(resop, &lsp->rls_reply);
9128 				/*
9129 				 * Make sure to copy the just
9130 				 * retrieved reply status into the
9131 				 * overall compound status
9132 				 */
9133 				*cs->statusp = resp->status;
9134 			}
9135 			rfs4_lockowner_rele(lo);
9136 			goto end;
9137 		}
9138 
9139 		rfs4_dbe_lock(lsp->rls_dbe);
9140 
9141 		/* Make sure to update the lock sequence id */
9142 		lsp->rls_seqid = olo->lock_seqid;
9143 
9144 		NFS4_DEBUG(rfs4_debug,
9145 		    (CE_NOTE, "Lock seqid established as %d", lsp->rls_seqid));
9146 
9147 		/*
9148 		 * This is used to signify the newly created lockowner
9149 		 * stateid and its sequence number.  The checks for
9150 		 * sequence number and increment don't occur on the
9151 		 * very first lock request for a lockowner.
9152 		 */
9153 		lsp->rls_skip_seqid_check = TRUE;
9154 
9155 		/* hold off other access to lsp while we tinker */
9156 		rfs4_sw_enter(&lsp->rls_sw);
9157 		ls_sw_held = TRUE;
9158 
9159 		rfs4_dbe_unlock(lsp->rls_dbe);
9160 
9161 		rfs4_lockowner_rele(lo);
9162 	} else {
9163 		stateid = &args->locker.locker4_u.lock_owner.lock_stateid;
9164 		/* get lsp and hold the lock on the underlying file struct */
9165 		if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE))
9166 		    != NFS4_OK) {
9167 			*cs->statusp = resp->status = status;
9168 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
9169 			    cs, LOCK4res *, resp);
9170 			return;
9171 		}
9172 		create = FALSE;	/* We didn't create lsp */
9173 
9174 		/* Ensure specified filehandle matches */
9175 		if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) {
9176 			rfs4_lo_state_rele(lsp, TRUE);
9177 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9178 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
9179 			    cs, LOCK4res *, resp);
9180 			return;
9181 		}
9182 
9183 		/* hold off other access to lsp while we tinker */
9184 		rfs4_sw_enter(&lsp->rls_sw);
9185 		ls_sw_held = TRUE;
9186 
9187 		switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
9188 		/*
9189 		 * The stateid looks like it was okay (expected to be
9190 		 * the next one)
9191 		 */
9192 		case NFS4_CHECK_STATEID_OKAY:
9193 			/*
9194 			 * The sequence id is now checked.  Determine
9195 			 * if this is a replay or if it is in the
9196 			 * expected (next) sequence.  In the case of a
9197 			 * replay, there are two replay conditions
9198 			 * that may occur.  The first is the normal
9199 			 * condition where a LOCK is done with a
9200 			 * NFS4_OK response and the stateid is
9201 			 * updated.  That case is handled below when
9202 			 * the stateid is identified as a REPLAY.  The
9203 			 * second is the case where an error is
9204 			 * returned, like NFS4ERR_DENIED, and the
9205 			 * sequence number is updated but the stateid
9206 			 * is not updated.  This second case is dealt
9207 			 * with here.  So it may seem odd that the
9208 			 * stateid is okay but the sequence id is a
9209 			 * replay but it is okay.
9210 			 */
9211 			switch (rfs4_check_lock_seqid(
9212 			    args->locker.locker4_u.lock_owner.lock_seqid,
9213 			    lsp, resop)) {
9214 			case NFS4_CHKSEQ_REPLAY:
9215 				if (resp->status != NFS4_OK) {
9216 					/*
9217 					 * Here is our replay and need
9218 					 * to verify that the last
9219 					 * response was an error.
9220 					 */
9221 					*cs->statusp = resp->status;
9222 					goto end;
9223 				}
9224 				/*
9225 				 * This is done since the sequence id
9226 				 * looked like a replay but it didn't
9227 				 * pass our check so a BAD_SEQID is
9228 				 * returned as a result.
9229 				 */
9230 				/*FALLTHROUGH*/
9231 			case NFS4_CHKSEQ_BAD:
9232 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9233 				goto end;
9234 			case NFS4_CHKSEQ_OKAY:
9235 				/* Everything looks okay move ahead */
9236 				break;
9237 			}
9238 			break;
9239 		case NFS4_CHECK_STATEID_OLD:
9240 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9241 			goto end;
9242 		case NFS4_CHECK_STATEID_BAD:
9243 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9244 			goto end;
9245 		case NFS4_CHECK_STATEID_EXPIRED:
9246 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
9247 			goto end;
9248 		case NFS4_CHECK_STATEID_CLOSED:
9249 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9250 			goto end;
9251 		case NFS4_CHECK_STATEID_REPLAY:
9252 			switch (rfs4_check_lock_seqid(
9253 			    args->locker.locker4_u.lock_owner.lock_seqid,
9254 			    lsp, resop)) {
9255 			case NFS4_CHKSEQ_OKAY:
9256 				/*
9257 				 * This is a replayed stateid; if
9258 				 * seqid matches the next expected,
9259 				 * then client is using wrong seqid.
9260 				 */
9261 			case NFS4_CHKSEQ_BAD:
9262 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9263 				goto end;
9264 			case NFS4_CHKSEQ_REPLAY:
9265 				rfs4_update_lease(lsp->rls_locker->rl_client);
9266 				*cs->statusp = status = resp->status;
9267 				goto end;
9268 			}
9269 			break;
9270 		default:
9271 			ASSERT(FALSE);
9272 			break;
9273 		}
9274 
9275 		rfs4_update_lock_sequence(lsp);
9276 		rfs4_update_lease(lsp->rls_locker->rl_client);
9277 	}
9278 
9279 	/*
9280 	 * NFS4 only allows locking on regular files, so
9281 	 * verify type of object.
9282 	 */
9283 	if (cs->vp->v_type != VREG) {
9284 		if (cs->vp->v_type == VDIR)
9285 			status = NFS4ERR_ISDIR;
9286 		else
9287 			status = NFS4ERR_INVAL;
9288 		goto out;
9289 	}
9290 
9291 	cp = lsp->rls_state->rs_owner->ro_client;
9292 
9293 	if (rfs4_clnt_in_grace(cp) && !args->reclaim) {
9294 		status = NFS4ERR_GRACE;
9295 		goto out;
9296 	}
9297 
9298 	if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->rc_can_reclaim) {
9299 		status = NFS4ERR_NO_GRACE;
9300 		goto out;
9301 	}
9302 
9303 	if (!rfs4_clnt_in_grace(cp) && args->reclaim) {
9304 		status = NFS4ERR_NO_GRACE;
9305 		goto out;
9306 	}
9307 
9308 	if (lsp->rls_state->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE)
9309 		cs->deleg = TRUE;
9310 
9311 	status = rfs4_do_lock(lsp, args->locktype,
9312 	    args->offset, args->length, cs->cr, resop);
9313 
9314 out:
9315 	lsp->rls_skip_seqid_check = FALSE;
9316 
9317 	*cs->statusp = resp->status = status;
9318 
9319 	if (status == NFS4_OK) {
9320 		resp->LOCK4res_u.lock_stateid = lsp->rls_lockid.stateid;
9321 		lsp->rls_lock_completed = TRUE;
9322 	}
9323 	/*
9324 	 * Only update the "OPEN" response here if this was a new
9325 	 * lock_owner
9326 	 */
9327 	if (sp)
9328 		rfs4_update_open_resp(sp->rs_owner, resop, NULL);
9329 
9330 	rfs4_update_lock_resp(lsp, resop);
9331 
9332 end:
9333 	if (lsp) {
9334 		if (ls_sw_held)
9335 			rfs4_sw_exit(&lsp->rls_sw);
9336 		/*
9337 		 * If an sp obtained, then the lsp does not represent
9338 		 * a lock on the file struct.
9339 		 */
9340 		if (sp != NULL)
9341 			rfs4_lo_state_rele(lsp, FALSE);
9342 		else
9343 			rfs4_lo_state_rele(lsp, TRUE);
9344 	}
9345 	if (sp) {
9346 		rfs4_sw_exit(&sp->rs_owner->ro_sw);
9347 		rfs4_state_rele(sp);
9348 	}
9349 
9350 	DTRACE_NFSV4_2(op__lock__done, struct compound_state *, cs,
9351 	    LOCK4res *, resp);
9352 }
9353 
9354 /* free function for LOCK/LOCKT */
9355 static void
9356 lock_denied_free(nfs_resop4 *resop)
9357 {
9358 	LOCK4denied *dp = NULL;
9359 
9360 	switch (resop->resop) {
9361 	case OP_LOCK:
9362 		if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED)
9363 			dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied;
9364 		break;
9365 	case OP_LOCKT:
9366 		if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED)
9367 			dp = &resop->nfs_resop4_u.oplockt.denied;
9368 		break;
9369 	default:
9370 		break;
9371 	}
9372 
9373 	if (dp)
9374 		kmem_free(dp->owner.owner_val, dp->owner.owner_len);
9375 }
9376 
9377 /*ARGSUSED*/
9378 void
9379 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop,
9380     struct svc_req *req, struct compound_state *cs)
9381 {
9382 	LOCKU4args *args = &argop->nfs_argop4_u.oplocku;
9383 	LOCKU4res *resp = &resop->nfs_resop4_u.oplocku;
9384 	nfsstat4 status;
9385 	stateid4 *stateid = &args->lock_stateid;
9386 	rfs4_lo_state_t *lsp;
9387 
9388 	DTRACE_NFSV4_2(op__locku__start, struct compound_state *, cs,
9389 	    LOCKU4args *, args);
9390 
9391 	if (cs->vp == NULL) {
9392 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
9393 		DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
9394 		    LOCKU4res *, resp);
9395 		return;
9396 	}
9397 
9398 	if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) {
9399 		*cs->statusp = resp->status = status;
9400 		DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
9401 		    LOCKU4res *, resp);
9402 		return;
9403 	}
9404 
9405 	/* Ensure specified filehandle matches */
9406 	if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) {
9407 		rfs4_lo_state_rele(lsp, TRUE);
9408 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9409 		DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
9410 		    LOCKU4res *, resp);
9411 		return;
9412 	}
9413 
9414 	/* hold off other access to lsp while we tinker */
9415 	rfs4_sw_enter(&lsp->rls_sw);
9416 
9417 	switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
9418 	case NFS4_CHECK_STATEID_OKAY:
9419 		if (rfs4_check_lock_seqid(args->seqid, lsp, resop)
9420 		    != NFS4_CHKSEQ_OKAY) {
9421 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9422 			goto end;
9423 		}
9424 		break;
9425 	case NFS4_CHECK_STATEID_OLD:
9426 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9427 		goto end;
9428 	case NFS4_CHECK_STATEID_BAD:
9429 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9430 		goto end;
9431 	case NFS4_CHECK_STATEID_EXPIRED:
9432 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
9433 		goto end;
9434 	case NFS4_CHECK_STATEID_CLOSED:
9435 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9436 		goto end;
9437 	case NFS4_CHECK_STATEID_REPLAY:
9438 		switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) {
9439 		case NFS4_CHKSEQ_OKAY:
9440 				/*
9441 				 * This is a replayed stateid; if
9442 				 * seqid matches the next expected,
9443 				 * then client is using wrong seqid.
9444 				 */
9445 		case NFS4_CHKSEQ_BAD:
9446 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9447 			goto end;
9448 		case NFS4_CHKSEQ_REPLAY:
9449 			rfs4_update_lease(lsp->rls_locker->rl_client);
9450 			*cs->statusp = status = resp->status;
9451 			goto end;
9452 		}
9453 		break;
9454 	default:
9455 		ASSERT(FALSE);
9456 		break;
9457 	}
9458 
9459 	rfs4_update_lock_sequence(lsp);
9460 	rfs4_update_lease(lsp->rls_locker->rl_client);
9461 
9462 	/*
9463 	 * NFS4 only allows locking on regular files, so
9464 	 * verify type of object.
9465 	 */
9466 	if (cs->vp->v_type != VREG) {
9467 		if (cs->vp->v_type == VDIR)
9468 			status = NFS4ERR_ISDIR;
9469 		else
9470 			status = NFS4ERR_INVAL;
9471 		goto out;
9472 	}
9473 
9474 	if (rfs4_clnt_in_grace(lsp->rls_state->rs_owner->ro_client)) {
9475 		status = NFS4ERR_GRACE;
9476 		goto out;
9477 	}
9478 
9479 	status = rfs4_do_lock(lsp, args->locktype,
9480 	    args->offset, args->length, cs->cr, resop);
9481 
9482 out:
9483 	*cs->statusp = resp->status = status;
9484 
9485 	if (status == NFS4_OK)
9486 		resp->lock_stateid = lsp->rls_lockid.stateid;
9487 
9488 	rfs4_update_lock_resp(lsp, resop);
9489 
9490 end:
9491 	rfs4_sw_exit(&lsp->rls_sw);
9492 	rfs4_lo_state_rele(lsp, TRUE);
9493 
9494 	DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
9495 	    LOCKU4res *, resp);
9496 }
9497 
9498 /*
9499  * LOCKT is a best effort routine, the client can not be guaranteed that
9500  * the status return is still in effect by the time the reply is received.
9501  * They are numerous race conditions in this routine, but we are not required
9502  * and can not be accurate.
9503  */
9504 /*ARGSUSED*/
9505 void
9506 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop,
9507     struct svc_req *req, struct compound_state *cs)
9508 {
9509 	LOCKT4args *args = &argop->nfs_argop4_u.oplockt;
9510 	LOCKT4res *resp = &resop->nfs_resop4_u.oplockt;
9511 	rfs4_lockowner_t *lo;
9512 	rfs4_client_t *cp;
9513 	bool_t create = FALSE;
9514 	struct flock64 flk;
9515 	int error;
9516 	int flag = FREAD | FWRITE;
9517 	int ltype;
9518 	length4 posix_length;
9519 	sysid_t sysid;
9520 	pid_t pid;
9521 
9522 	DTRACE_NFSV4_2(op__lockt__start, struct compound_state *, cs,
9523 	    LOCKT4args *, args);
9524 
9525 	if (cs->vp == NULL) {
9526 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
9527 		goto out;
9528 	}
9529 
9530 	/*
9531 	 * NFS4 only allows locking on regular files, so
9532 	 * verify type of object.
9533 	 */
9534 	if (cs->vp->v_type != VREG) {
9535 		if (cs->vp->v_type == VDIR)
9536 			*cs->statusp = resp->status = NFS4ERR_ISDIR;
9537 		else
9538 			*cs->statusp = resp->status =  NFS4ERR_INVAL;
9539 		goto out;
9540 	}
9541 
9542 	/*
9543 	 * Check out the clientid to ensure the server knows about it
9544 	 * so that we correctly inform the client of a server reboot.
9545 	 */
9546 	if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE))
9547 	    == NULL) {
9548 		*cs->statusp = resp->status =
9549 		    rfs4_check_clientid(&args->owner.clientid, 0);
9550 		goto out;
9551 	}
9552 	if (rfs4_lease_expired(cp)) {
9553 		rfs4_client_close(cp);
9554 		/*
9555 		 * Protocol doesn't allow returning NFS4ERR_STALE as
9556 		 * other operations do on this check so STALE_CLIENTID
9557 		 * is returned instead
9558 		 */
9559 		*cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
9560 		goto out;
9561 	}
9562 
9563 	if (rfs4_clnt_in_grace(cp) && !(cp->rc_can_reclaim)) {
9564 		*cs->statusp = resp->status = NFS4ERR_GRACE;
9565 		rfs4_client_rele(cp);
9566 		goto out;
9567 	}
9568 	rfs4_client_rele(cp);
9569 
9570 	resp->status = NFS4_OK;
9571 
9572 	switch (args->locktype) {
9573 	case READ_LT:
9574 	case READW_LT:
9575 		ltype = F_RDLCK;
9576 		break;
9577 	case WRITE_LT:
9578 	case WRITEW_LT:
9579 		ltype = F_WRLCK;
9580 		break;
9581 	}
9582 
9583 	posix_length = args->length;
9584 	/* Check for zero length. To lock to end of file use all ones for V4 */
9585 	if (posix_length == 0) {
9586 		*cs->statusp = resp->status = NFS4ERR_INVAL;
9587 		goto out;
9588 	} else if (posix_length == (length4)(~0)) {
9589 		posix_length = 0;	/* Posix to end of file  */
9590 	}
9591 
9592 	/* Find or create a lockowner */
9593 	lo = rfs4_findlockowner(&args->owner, &create);
9594 
9595 	if (lo) {
9596 		pid = lo->rl_pid;
9597 		if ((resp->status =
9598 		    rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK)
9599 			goto err;
9600 	} else {
9601 		pid = 0;
9602 		sysid = lockt_sysid;
9603 	}
9604 retry:
9605 	flk.l_type = ltype;
9606 	flk.l_whence = 0;		/* SEEK_SET */
9607 	flk.l_start = args->offset;
9608 	flk.l_len = posix_length;
9609 	flk.l_sysid = sysid;
9610 	flk.l_pid = pid;
9611 	flag |= F_REMOTELOCK;
9612 
9613 	LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk);
9614 
9615 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
9616 	if (flk.l_len < 0 || flk.l_start < 0) {
9617 		resp->status = NFS4ERR_INVAL;
9618 		goto err;
9619 	}
9620 	error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0,
9621 	    NULL, cs->cr, NULL);
9622 
9623 	/*
9624 	 * N.B. We map error values to nfsv4 errors. This is differrent
9625 	 * than puterrno4 routine.
9626 	 */
9627 	switch (error) {
9628 	case 0:
9629 		if (flk.l_type == F_UNLCK)
9630 			resp->status = NFS4_OK;
9631 		else {
9632 			if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED)
9633 				goto retry;
9634 			resp->status = NFS4ERR_DENIED;
9635 		}
9636 		break;
9637 	case EOVERFLOW:
9638 		resp->status = NFS4ERR_INVAL;
9639 		break;
9640 	case EINVAL:
9641 		resp->status = NFS4ERR_NOTSUPP;
9642 		break;
9643 	default:
9644 		cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)",
9645 		    error);
9646 		resp->status = NFS4ERR_SERVERFAULT;
9647 		break;
9648 	}
9649 
9650 err:
9651 	if (lo)
9652 		rfs4_lockowner_rele(lo);
9653 	*cs->statusp = resp->status;
9654 out:
9655 	DTRACE_NFSV4_2(op__lockt__done, struct compound_state *, cs,
9656 	    LOCKT4res *, resp);
9657 }
9658 
9659 int
9660 rfs4_share(rfs4_state_t *sp, uint32_t access, uint32_t deny)
9661 {
9662 	int err;
9663 	int cmd;
9664 	vnode_t *vp;
9665 	struct shrlock shr;
9666 	struct shr_locowner shr_loco;
9667 	int fflags = 0;
9668 
9669 	ASSERT(rfs4_dbe_islocked(sp->rs_dbe));
9670 	ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID);
9671 
9672 	if (sp->rs_closed)
9673 		return (NFS4ERR_OLD_STATEID);
9674 
9675 	vp = sp->rs_finfo->rf_vp;
9676 	ASSERT(vp);
9677 
9678 	shr.s_access = shr.s_deny = 0;
9679 
9680 	if (access & OPEN4_SHARE_ACCESS_READ) {
9681 		fflags |= FREAD;
9682 		shr.s_access |= F_RDACC;
9683 	}
9684 	if (access & OPEN4_SHARE_ACCESS_WRITE) {
9685 		fflags |= FWRITE;
9686 		shr.s_access |= F_WRACC;
9687 	}
9688 	ASSERT(shr.s_access);
9689 
9690 	if (deny & OPEN4_SHARE_DENY_READ)
9691 		shr.s_deny |= F_RDDNY;
9692 	if (deny & OPEN4_SHARE_DENY_WRITE)
9693 		shr.s_deny |= F_WRDNY;
9694 
9695 	shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe);
9696 	shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt;
9697 	shr_loco.sl_pid = shr.s_pid;
9698 	shr_loco.sl_id = shr.s_sysid;
9699 	shr.s_owner = (caddr_t)&shr_loco;
9700 	shr.s_own_len = sizeof (shr_loco);
9701 
9702 	cmd = nbl_need_check(vp) ? F_SHARE_NBMAND : F_SHARE;
9703 
9704 	err = VOP_SHRLOCK(vp, cmd, &shr, fflags, CRED(), NULL);
9705 	if (err != 0) {
9706 		if (err == EAGAIN)
9707 			err = NFS4ERR_SHARE_DENIED;
9708 		else
9709 			err = puterrno4(err);
9710 		return (err);
9711 	}
9712 
9713 	sp->rs_share_access |= access;
9714 	sp->rs_share_deny |= deny;
9715 
9716 	return (0);
9717 }
9718 
9719 int
9720 rfs4_unshare(rfs4_state_t *sp)
9721 {
9722 	int err;
9723 	struct shrlock shr;
9724 	struct shr_locowner shr_loco;
9725 
9726 	ASSERT(rfs4_dbe_islocked(sp->rs_dbe));
9727 
9728 	if (sp->rs_closed || sp->rs_share_access == 0)
9729 		return (0);
9730 
9731 	ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID);
9732 	ASSERT(sp->rs_finfo->rf_vp);
9733 
9734 	shr.s_access = shr.s_deny = 0;
9735 	shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe);
9736 	shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt;
9737 	shr_loco.sl_pid = shr.s_pid;
9738 	shr_loco.sl_id = shr.s_sysid;
9739 	shr.s_owner = (caddr_t)&shr_loco;
9740 	shr.s_own_len = sizeof (shr_loco);
9741 
9742 	err = VOP_SHRLOCK(sp->rs_finfo->rf_vp, F_UNSHARE, &shr, 0, CRED(),
9743 	    NULL);
9744 	if (err != 0) {
9745 		err = puterrno4(err);
9746 		return (err);
9747 	}
9748 
9749 	sp->rs_share_access = 0;
9750 	sp->rs_share_deny = 0;
9751 
9752 	return (0);
9753 
9754 }
9755 
9756 static int
9757 rdma_setup_read_data4(READ4args *args, READ4res *rok)
9758 {
9759 	struct clist	*wcl;
9760 	count4		count = rok->data_len;
9761 	int		wlist_len;
9762 
9763 	wcl = args->wlist;
9764 	if (rdma_setup_read_chunks(wcl, count, &wlist_len) == FALSE) {
9765 		return (FALSE);
9766 	}
9767 	wcl = args->wlist;
9768 	rok->wlist_len = wlist_len;
9769 	rok->wlist = wcl;
9770 	return (TRUE);
9771 }
9772 
9773 /* tunable to disable server referrals */
9774 int rfs4_no_referrals = 0;
9775 
9776 /*
9777  * Find an NFS record in reparse point data.
9778  * Returns 0 for success and <0 or an errno value on failure.
9779  */
9780 int
9781 vn_find_nfs_record(vnode_t *vp, nvlist_t **nvlp, char **svcp, char **datap)
9782 {
9783 	int err;
9784 	char *stype, *val;
9785 	nvlist_t *nvl;
9786 	nvpair_t *curr;
9787 
9788 	if ((nvl = reparse_init()) == NULL)
9789 		return (-1);
9790 
9791 	if ((err = reparse_vnode_parse(vp, nvl)) != 0) {
9792 		reparse_free(nvl);
9793 		return (err);
9794 	}
9795 
9796 	curr = NULL;
9797 	while ((curr = nvlist_next_nvpair(nvl, curr)) != NULL) {
9798 		if ((stype = nvpair_name(curr)) == NULL) {
9799 			reparse_free(nvl);
9800 			return (-2);
9801 		}
9802 		if (strncasecmp(stype, "NFS", 3) == 0)
9803 			break;
9804 	}
9805 
9806 	if ((curr == NULL) ||
9807 	    (nvpair_value_string(curr, &val))) {
9808 		reparse_free(nvl);
9809 		return (-3);
9810 	}
9811 	*nvlp = nvl;
9812 	*svcp = stype;
9813 	*datap = val;
9814 	return (0);
9815 }
9816 
9817 int
9818 vn_is_nfs_reparse(vnode_t *vp, cred_t *cr)
9819 {
9820 	nvlist_t *nvl;
9821 	char *s, *d;
9822 
9823 	if (rfs4_no_referrals != 0)
9824 		return (B_FALSE);
9825 
9826 	if (vn_is_reparse(vp, cr, NULL) == B_FALSE)
9827 		return (B_FALSE);
9828 
9829 	if (vn_find_nfs_record(vp, &nvl, &s, &d) != 0)
9830 		return (B_FALSE);
9831 
9832 	reparse_free(nvl);
9833 
9834 	return (B_TRUE);
9835 }
9836 
9837 /*
9838  * There is a user-level copy of this routine in ref_subr.c.
9839  * Changes should be kept in sync.
9840  */
9841 static int
9842 nfs4_create_components(char *path, component4 *comp4)
9843 {
9844 	int slen, plen, ncomp;
9845 	char *ori_path, *nxtc, buf[MAXNAMELEN];
9846 
9847 	if (path == NULL)
9848 		return (0);
9849 
9850 	plen = strlen(path) + 1;	/* include the terminator */
9851 	ori_path = path;
9852 	ncomp = 0;
9853 
9854 	/* count number of components in the path */
9855 	for (nxtc = path; nxtc < ori_path + plen; nxtc++) {
9856 		if (*nxtc == '/' || *nxtc == '\0' || *nxtc == '\n') {
9857 			if ((slen = nxtc - path) == 0) {
9858 				path = nxtc + 1;
9859 				continue;
9860 			}
9861 
9862 			if (comp4 != NULL) {
9863 				bcopy(path, buf, slen);
9864 				buf[slen] = '\0';
9865 				(void) str_to_utf8(buf, &comp4[ncomp]);
9866 			}
9867 
9868 			ncomp++;	/* 1 valid component */
9869 			path = nxtc + 1;
9870 		}
9871 		if (*nxtc == '\0' || *nxtc == '\n')
9872 			break;
9873 	}
9874 
9875 	return (ncomp);
9876 }
9877 
9878 /*
9879  * There is a user-level copy of this routine in ref_subr.c.
9880  * Changes should be kept in sync.
9881  */
9882 static int
9883 make_pathname4(char *path, pathname4 *pathname)
9884 {
9885 	int ncomp;
9886 	component4 *comp4;
9887 
9888 	if (pathname == NULL)
9889 		return (0);
9890 
9891 	if (path == NULL) {
9892 		pathname->pathname4_val = NULL;
9893 		pathname->pathname4_len = 0;
9894 		return (0);
9895 	}
9896 
9897 	/* count number of components to alloc buffer */
9898 	if ((ncomp = nfs4_create_components(path, NULL)) == 0) {
9899 		pathname->pathname4_val = NULL;
9900 		pathname->pathname4_len = 0;
9901 		return (0);
9902 	}
9903 	comp4 = kmem_zalloc(ncomp * sizeof (component4), KM_SLEEP);
9904 
9905 	/* copy components into allocated buffer */
9906 	ncomp = nfs4_create_components(path, comp4);
9907 
9908 	pathname->pathname4_val = comp4;
9909 	pathname->pathname4_len = ncomp;
9910 
9911 	return (ncomp);
9912 }
9913 
9914 #define	xdr_fs_locations4 xdr_fattr4_fs_locations
9915 
9916 fs_locations4 *
9917 fetch_referral(vnode_t *vp, cred_t *cr)
9918 {
9919 	nvlist_t *nvl;
9920 	char *stype, *sdata;
9921 	fs_locations4 *result;
9922 	char buf[1024];
9923 	size_t bufsize;
9924 	XDR xdr;
9925 	int err;
9926 
9927 	/*
9928 	 * Check attrs to ensure it's a reparse point
9929 	 */
9930 	if (vn_is_reparse(vp, cr, NULL) == B_FALSE)
9931 		return (NULL);
9932 
9933 	/*
9934 	 * Look for an NFS record and get the type and data
9935 	 */
9936 	if (vn_find_nfs_record(vp, &nvl, &stype, &sdata) != 0)
9937 		return (NULL);
9938 
9939 	/*
9940 	 * With the type and data, upcall to get the referral
9941 	 */
9942 	bufsize = sizeof (buf);
9943 	bzero(buf, sizeof (buf));
9944 	err = reparse_kderef((const char *)stype, (const char *)sdata,
9945 	    buf, &bufsize);
9946 	reparse_free(nvl);
9947 
9948 	DTRACE_PROBE4(nfs4serv__func__referral__upcall,
9949 	    char *, stype, char *, sdata, char *, buf, int, err);
9950 	if (err) {
9951 		cmn_err(CE_NOTE,
9952 		    "reparsed daemon not running: unable to get referral (%d)",
9953 		    err);
9954 		return (NULL);
9955 	}
9956 
9957 	/*
9958 	 * We get an XDR'ed record back from the kderef call
9959 	 */
9960 	xdrmem_create(&xdr, buf, bufsize, XDR_DECODE);
9961 	result = kmem_alloc(sizeof (fs_locations4), KM_SLEEP);
9962 	err = xdr_fs_locations4(&xdr, result);
9963 	XDR_DESTROY(&xdr);
9964 	if (err != TRUE) {
9965 		DTRACE_PROBE1(nfs4serv__func__referral__upcall__xdrfail,
9966 		    int, err);
9967 		return (NULL);
9968 	}
9969 
9970 	/*
9971 	 * Look at path to recover fs_root, ignoring the leading '/'
9972 	 */
9973 	(void) make_pathname4(vp->v_path, &result->fs_root);
9974 
9975 	return (result);
9976 }
9977 
9978 char *
9979 build_symlink(vnode_t *vp, cred_t *cr, size_t *strsz)
9980 {
9981 	fs_locations4 *fsl;
9982 	fs_location4 *fs;
9983 	char *server, *path, *symbuf;
9984 	static char *prefix = "/net/";
9985 	int i, size, npaths;
9986 	uint_t len;
9987 
9988 	/* Get the referral */
9989 	if ((fsl = fetch_referral(vp, cr)) == NULL)
9990 		return (NULL);
9991 
9992 	/* Deal with only the first location and first server */
9993 	fs = &fsl->locations_val[0];
9994 	server = utf8_to_str(&fs->server_val[0], &len, NULL);
9995 	if (server == NULL) {
9996 		rfs4_free_fs_locations4(fsl);
9997 		kmem_free(fsl, sizeof (fs_locations4));
9998 		return (NULL);
9999 	}
10000 
10001 	/* Figure out size for "/net/" + host + /path/path/path + NULL */
10002 	size = strlen(prefix) + len;
10003 	for (i = 0; i < fs->rootpath.pathname4_len; i++)
10004 		size += fs->rootpath.pathname4_val[i].utf8string_len + 1;
10005 
10006 	/* Allocate the symlink buffer and fill it */
10007 	symbuf = kmem_zalloc(size, KM_SLEEP);
10008 	(void) strcat(symbuf, prefix);
10009 	(void) strcat(symbuf, server);
10010 	kmem_free(server, len);
10011 
10012 	npaths = 0;
10013 	for (i = 0; i < fs->rootpath.pathname4_len; i++) {
10014 		path = utf8_to_str(&fs->rootpath.pathname4_val[i], &len, NULL);
10015 		if (path == NULL)
10016 			continue;
10017 		(void) strcat(symbuf, "/");
10018 		(void) strcat(symbuf, path);
10019 		npaths++;
10020 		kmem_free(path, len);
10021 	}
10022 
10023 	rfs4_free_fs_locations4(fsl);
10024 	kmem_free(fsl, sizeof (fs_locations4));
10025 
10026 	if (strsz != NULL)
10027 		*strsz = size;
10028 	return (symbuf);
10029 }
10030 
10031 /*
10032  * Check to see if we have a downrev Solaris client, so that we
10033  * can send it a symlink instead of a referral.
10034  */
10035 int
10036 client_is_downrev(struct svc_req *req)
10037 {
10038 	struct sockaddr *ca;
10039 	rfs4_clntip_t *ci;
10040 	bool_t create = FALSE;
10041 	int is_downrev;
10042 
10043 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
10044 	ASSERT(ca);
10045 	ci = rfs4_find_clntip(ca, &create);
10046 	if (ci == NULL)
10047 		return (0);
10048 	is_downrev = ci->ri_no_referrals;
10049 	rfs4_dbe_rele(ci->ri_dbe);
10050 	return (is_downrev);
10051 }
10052