xref: /titanic_50/usr/src/uts/common/fs/nfs/nfs4_srv.c (revision 1665390b481087ca2aaff06a502c62b3ee91d530)
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 	} else {
4288 		if ((error = VOP_REMOVE(dvp, name, cs->cr, NULL, 0)) == 0 &&
4289 		    fp != NULL) {
4290 			struct vattr va;
4291 			vnode_t *tvp;
4292 
4293 			rfs4_dbe_lock(fp->rf_dbe);
4294 			tvp = fp->rf_vp;
4295 			if (tvp)
4296 				VN_HOLD(tvp);
4297 			rfs4_dbe_unlock(fp->rf_dbe);
4298 
4299 			if (tvp) {
4300 				/*
4301 				 * This is va_seq safe because we are not
4302 				 * manipulating dvp.
4303 				 */
4304 				va.va_mask = AT_NLINK;
4305 				if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) &&
4306 				    va.va_nlink == 0) {
4307 					/* Remove state on file remove */
4308 					if (in_crit) {
4309 						nbl_end_crit(vp);
4310 						in_crit = 0;
4311 					}
4312 					rfs4_close_all_state(fp);
4313 				}
4314 				VN_RELE(tvp);
4315 			}
4316 		}
4317 	}
4318 
4319 	if (in_crit)
4320 		nbl_end_crit(vp);
4321 	VN_RELE(vp);
4322 
4323 	if (fp) {
4324 		rfs4_clear_dont_grant(fp);
4325 		rfs4_file_rele(fp);
4326 	}
4327 	if (nm != name)
4328 		kmem_free(name, MAXPATHLEN + 1);
4329 	kmem_free(nm, len);
4330 
4331 	if (error) {
4332 		*cs->statusp = resp->status = puterrno4(error);
4333 		goto out;
4334 	}
4335 
4336 	/*
4337 	 * Get the initial "after" sequence number, if it fails, set to zero
4338 	 */
4339 	idva.va_mask = AT_SEQ;
4340 	if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL))
4341 		idva.va_seq = 0;
4342 
4343 	/*
4344 	 * Force modified data and metadata out to stable storage.
4345 	 */
4346 	(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
4347 
4348 	/*
4349 	 * Get "after" change value, if it fails, simply return the
4350 	 * before value.
4351 	 */
4352 	adva.va_mask = AT_CTIME|AT_SEQ;
4353 	if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) {
4354 		adva.va_ctime = bdva.va_ctime;
4355 		adva.va_seq = 0;
4356 	}
4357 
4358 	NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime)
4359 
4360 	/*
4361 	 * The cinfo.atomic = TRUE only if we have
4362 	 * non-zero va_seq's, and it has incremented by exactly one
4363 	 * during the VOP_REMOVE/RMDIR and it didn't change during
4364 	 * the VOP_FSYNC.
4365 	 */
4366 	if (bdva.va_seq && idva.va_seq && adva.va_seq &&
4367 	    idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq)
4368 		resp->cinfo.atomic = TRUE;
4369 	else
4370 		resp->cinfo.atomic = FALSE;
4371 
4372 	*cs->statusp = resp->status = NFS4_OK;
4373 
4374 out:
4375 	DTRACE_NFSV4_2(op__remove__done, struct compound_state *, cs,
4376 	    REMOVE4res *, resp);
4377 }
4378 
4379 /*
4380  * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory,
4381  *		oldname and newname.
4382  *	res: status. If success - CURRENT_FH unchanged, return change_info
4383  *		for both from and target directories.
4384  */
4385 /* ARGSUSED */
4386 static void
4387 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4388     struct compound_state *cs)
4389 {
4390 	RENAME4args *args = &argop->nfs_argop4_u.oprename;
4391 	RENAME4res *resp = &resop->nfs_resop4_u.oprename;
4392 	int error;
4393 	vnode_t *odvp;
4394 	vnode_t *ndvp;
4395 	vnode_t *srcvp, *targvp;
4396 	struct vattr obdva, oidva, oadva;
4397 	struct vattr nbdva, nidva, nadva;
4398 	char *onm, *nnm;
4399 	uint_t olen, nlen;
4400 	rfs4_file_t *fp, *sfp;
4401 	int in_crit_src, in_crit_targ;
4402 	int fp_rele_grant_hold, sfp_rele_grant_hold;
4403 	bslabel_t *clabel;
4404 	struct sockaddr *ca;
4405 	char *converted_onm = NULL;
4406 	char *converted_nnm = NULL;
4407 	nfsstat4 status;
4408 
4409 	DTRACE_NFSV4_2(op__rename__start, struct compound_state *, cs,
4410 	    RENAME4args *, args);
4411 
4412 	fp = sfp = NULL;
4413 	srcvp = targvp = NULL;
4414 	in_crit_src = in_crit_targ = 0;
4415 	fp_rele_grant_hold = sfp_rele_grant_hold = 0;
4416 
4417 	/* CURRENT_FH: target directory */
4418 	ndvp = cs->vp;
4419 	if (ndvp == NULL) {
4420 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4421 		goto out;
4422 	}
4423 
4424 	/* SAVED_FH: from directory */
4425 	odvp = cs->saved_vp;
4426 	if (odvp == NULL) {
4427 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4428 		goto out;
4429 	}
4430 
4431 	if (cs->access == CS_ACCESS_DENIED) {
4432 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4433 		goto out;
4434 	}
4435 
4436 	/*
4437 	 * If there is an unshared filesystem mounted on this vnode,
4438 	 * do not allow to rename objects in this directory.
4439 	 */
4440 	if (vn_ismntpt(odvp)) {
4441 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4442 		goto out;
4443 	}
4444 
4445 	/*
4446 	 * If there is an unshared filesystem mounted on this vnode,
4447 	 * do not allow to rename to this directory.
4448 	 */
4449 	if (vn_ismntpt(ndvp)) {
4450 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
4451 		goto out;
4452 	}
4453 
4454 	if (odvp->v_type != VDIR || ndvp->v_type != VDIR) {
4455 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
4456 		goto out;
4457 	}
4458 
4459 	if (cs->saved_exi != cs->exi) {
4460 		*cs->statusp = resp->status = NFS4ERR_XDEV;
4461 		goto out;
4462 	}
4463 
4464 	status = utf8_dir_verify(&args->oldname);
4465 	if (status != NFS4_OK) {
4466 		*cs->statusp = resp->status = status;
4467 		goto out;
4468 	}
4469 
4470 	status = utf8_dir_verify(&args->newname);
4471 	if (status != NFS4_OK) {
4472 		*cs->statusp = resp->status = status;
4473 		goto out;
4474 	}
4475 
4476 	onm = utf8_to_fn(&args->oldname, &olen, NULL);
4477 	if (onm == NULL) {
4478 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4479 		goto out;
4480 	}
4481 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
4482 	nlen = MAXPATHLEN + 1;
4483 	converted_onm = nfscmd_convname(ca, cs->exi, onm, NFSCMD_CONV_INBOUND,
4484 	    nlen);
4485 
4486 	if (converted_onm == NULL) {
4487 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4488 		kmem_free(onm, olen);
4489 		goto out;
4490 	}
4491 
4492 	nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4493 	if (nnm == NULL) {
4494 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4495 		if (onm != converted_onm)
4496 			kmem_free(converted_onm, MAXPATHLEN + 1);
4497 		kmem_free(onm, olen);
4498 		goto out;
4499 	}
4500 	converted_nnm = nfscmd_convname(ca, cs->exi, nnm, NFSCMD_CONV_INBOUND,
4501 	    MAXPATHLEN  + 1);
4502 
4503 	if (converted_nnm == NULL) {
4504 		*cs->statusp = resp->status = NFS4ERR_INVAL;
4505 		kmem_free(nnm, nlen);
4506 		nnm = NULL;
4507 		if (onm != converted_onm)
4508 			kmem_free(converted_onm, MAXPATHLEN + 1);
4509 		kmem_free(onm, olen);
4510 		goto out;
4511 	}
4512 
4513 
4514 	if (olen > MAXNAMELEN || nlen > MAXNAMELEN) {
4515 		*cs->statusp = resp->status = NFS4ERR_NAMETOOLONG;
4516 		kmem_free(onm, olen);
4517 		kmem_free(nnm, nlen);
4518 		goto out;
4519 	}
4520 
4521 
4522 	if (rdonly4(req, cs)) {
4523 		*cs->statusp = resp->status = NFS4ERR_ROFS;
4524 		if (onm != converted_onm)
4525 			kmem_free(converted_onm, MAXPATHLEN + 1);
4526 		kmem_free(onm, olen);
4527 		if (nnm != converted_nnm)
4528 			kmem_free(converted_nnm, MAXPATHLEN + 1);
4529 		kmem_free(nnm, nlen);
4530 		goto out;
4531 	}
4532 
4533 	/* check label of the target dir */
4534 	if (is_system_labeled()) {
4535 		ASSERT(req->rq_label != NULL);
4536 		clabel = req->rq_label;
4537 		DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *,
4538 		    "got client label from request(1)",
4539 		    struct svc_req *, req);
4540 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
4541 			if (!do_rfs_label_check(clabel, ndvp,
4542 			    EQUALITY_CHECK, cs->exi)) {
4543 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
4544 				goto err_out;
4545 			}
4546 		}
4547 	}
4548 
4549 	/*
4550 	 * Is the source a file and have a delegation?
4551 	 * We don't need to acquire va_seq before these lookups, if
4552 	 * it causes an update, cinfo.before will not match, which will
4553 	 * trigger a cache flush even if atomic is TRUE.
4554 	 */
4555 	if (sfp = rfs4_lookup_and_findfile(odvp, converted_onm, &srcvp,
4556 	    &error, cs->cr)) {
4557 		if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE,
4558 		    NULL)) {
4559 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4560 			goto err_out;
4561 		}
4562 	}
4563 
4564 	if (srcvp == NULL) {
4565 		*cs->statusp = resp->status = puterrno4(error);
4566 		if (onm != converted_onm)
4567 			kmem_free(converted_onm, MAXPATHLEN + 1);
4568 		kmem_free(onm, olen);
4569 		if (nnm != converted_nnm)
4570 			kmem_free(converted_nnm, MAXPATHLEN + 1);
4571 		kmem_free(nnm, nlen);
4572 		goto out;
4573 	}
4574 
4575 	sfp_rele_grant_hold = 1;
4576 
4577 	/* Does the destination exist and a file and have a delegation? */
4578 	if (fp = rfs4_lookup_and_findfile(ndvp, converted_nnm, &targvp,
4579 	    NULL, cs->cr)) {
4580 		if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE,
4581 		    NULL)) {
4582 			*cs->statusp = resp->status = NFS4ERR_DELAY;
4583 			goto err_out;
4584 		}
4585 	}
4586 	fp_rele_grant_hold = 1;
4587 
4588 
4589 	/* Check for NBMAND lock on both source and target */
4590 	if (nbl_need_check(srcvp)) {
4591 		nbl_start_crit(srcvp, RW_READER);
4592 		in_crit_src = 1;
4593 		if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) {
4594 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4595 			goto err_out;
4596 		}
4597 	}
4598 
4599 	if (targvp && nbl_need_check(targvp)) {
4600 		nbl_start_crit(targvp, RW_READER);
4601 		in_crit_targ = 1;
4602 		if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) {
4603 			*cs->statusp = resp->status = NFS4ERR_FILE_OPEN;
4604 			goto err_out;
4605 		}
4606 	}
4607 
4608 	/* Get source "before" change value */
4609 	obdva.va_mask = AT_CTIME|AT_SEQ;
4610 	error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL);
4611 	if (!error) {
4612 		nbdva.va_mask = AT_CTIME|AT_SEQ;
4613 		error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL);
4614 	}
4615 	if (error) {
4616 		*cs->statusp = resp->status = puterrno4(error);
4617 		goto err_out;
4618 	}
4619 
4620 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime)
4621 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime)
4622 
4623 	if ((error = VOP_RENAME(odvp, converted_onm, ndvp, converted_nnm,
4624 	    cs->cr, NULL, 0)) == 0 && fp != NULL) {
4625 		struct vattr va;
4626 		vnode_t *tvp;
4627 
4628 		rfs4_dbe_lock(fp->rf_dbe);
4629 		tvp = fp->rf_vp;
4630 		if (tvp)
4631 			VN_HOLD(tvp);
4632 		rfs4_dbe_unlock(fp->rf_dbe);
4633 
4634 		if (tvp) {
4635 			va.va_mask = AT_NLINK;
4636 			if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) &&
4637 			    va.va_nlink == 0) {
4638 				/* The file is gone and so should the state */
4639 				if (in_crit_targ) {
4640 					nbl_end_crit(targvp);
4641 					in_crit_targ = 0;
4642 				}
4643 				rfs4_close_all_state(fp);
4644 			}
4645 			VN_RELE(tvp);
4646 		}
4647 	}
4648 	if (error == 0)
4649 		vn_renamepath(ndvp, srcvp, nnm, nlen - 1);
4650 
4651 	if (in_crit_src)
4652 		nbl_end_crit(srcvp);
4653 	if (srcvp)
4654 		VN_RELE(srcvp);
4655 	if (in_crit_targ)
4656 		nbl_end_crit(targvp);
4657 	if (targvp)
4658 		VN_RELE(targvp);
4659 
4660 	if (sfp) {
4661 		rfs4_clear_dont_grant(sfp);
4662 		rfs4_file_rele(sfp);
4663 	}
4664 	if (fp) {
4665 		rfs4_clear_dont_grant(fp);
4666 		rfs4_file_rele(fp);
4667 	}
4668 
4669 	if (converted_onm != onm)
4670 		kmem_free(converted_onm, MAXPATHLEN + 1);
4671 	kmem_free(onm, olen);
4672 	if (converted_nnm != nnm)
4673 		kmem_free(converted_nnm, MAXPATHLEN + 1);
4674 	kmem_free(nnm, nlen);
4675 
4676 	/*
4677 	 * Get the initial "after" sequence number, if it fails, set to zero
4678 	 */
4679 	oidva.va_mask = AT_SEQ;
4680 	if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL))
4681 		oidva.va_seq = 0;
4682 
4683 	nidva.va_mask = AT_SEQ;
4684 	if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL))
4685 		nidva.va_seq = 0;
4686 
4687 	/*
4688 	 * Force modified data and metadata out to stable storage.
4689 	 */
4690 	(void) VOP_FSYNC(odvp, 0, cs->cr, NULL);
4691 	(void) VOP_FSYNC(ndvp, 0, cs->cr, NULL);
4692 
4693 	if (error) {
4694 		*cs->statusp = resp->status = puterrno4(error);
4695 		goto out;
4696 	}
4697 
4698 	/*
4699 	 * Get "after" change values, if it fails, simply return the
4700 	 * before value.
4701 	 */
4702 	oadva.va_mask = AT_CTIME|AT_SEQ;
4703 	if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) {
4704 		oadva.va_ctime = obdva.va_ctime;
4705 		oadva.va_seq = 0;
4706 	}
4707 
4708 	nadva.va_mask = AT_CTIME|AT_SEQ;
4709 	if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) {
4710 		nadva.va_ctime = nbdva.va_ctime;
4711 		nadva.va_seq = 0;
4712 	}
4713 
4714 	NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime)
4715 	NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime)
4716 
4717 	/*
4718 	 * The cinfo.atomic = TRUE only if we have
4719 	 * non-zero va_seq's, and it has incremented by exactly one
4720 	 * during the VOP_RENAME and it didn't change during the VOP_FSYNC.
4721 	 */
4722 	if (obdva.va_seq && oidva.va_seq && oadva.va_seq &&
4723 	    oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq)
4724 		resp->source_cinfo.atomic = TRUE;
4725 	else
4726 		resp->source_cinfo.atomic = FALSE;
4727 
4728 	if (nbdva.va_seq && nidva.va_seq && nadva.va_seq &&
4729 	    nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq)
4730 		resp->target_cinfo.atomic = TRUE;
4731 	else
4732 		resp->target_cinfo.atomic = FALSE;
4733 
4734 #ifdef	VOLATILE_FH_TEST
4735 	{
4736 	extern void add_volrnm_fh(struct exportinfo *, vnode_t *);
4737 
4738 	/*
4739 	 * Add the renamed file handle to the volatile rename list
4740 	 */
4741 	if (cs->exi->exi_export.ex_flags & EX_VOLRNM) {
4742 		/* file handles may expire on rename */
4743 		vnode_t *vp;
4744 
4745 		nnm = utf8_to_fn(&args->newname, &nlen, NULL);
4746 		/*
4747 		 * Already know that nnm will be a valid string
4748 		 */
4749 		error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr,
4750 		    NULL, NULL, NULL);
4751 		kmem_free(nnm, nlen);
4752 		if (!error) {
4753 			add_volrnm_fh(cs->exi, vp);
4754 			VN_RELE(vp);
4755 		}
4756 	}
4757 	}
4758 #endif	/* VOLATILE_FH_TEST */
4759 
4760 	*cs->statusp = resp->status = NFS4_OK;
4761 out:
4762 	DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs,
4763 	    RENAME4res *, resp);
4764 	return;
4765 
4766 err_out:
4767 	if (onm != converted_onm)
4768 		kmem_free(converted_onm, MAXPATHLEN + 1);
4769 	if (onm != NULL)
4770 		kmem_free(onm, olen);
4771 	if (nnm != converted_nnm)
4772 		kmem_free(converted_nnm, MAXPATHLEN + 1);
4773 	if (nnm != NULL)
4774 		kmem_free(nnm, nlen);
4775 
4776 	if (in_crit_src) nbl_end_crit(srcvp);
4777 	if (in_crit_targ) nbl_end_crit(targvp);
4778 	if (targvp) VN_RELE(targvp);
4779 	if (srcvp) VN_RELE(srcvp);
4780 	if (sfp) {
4781 		if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp);
4782 		rfs4_file_rele(sfp);
4783 	}
4784 	if (fp) {
4785 		if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp);
4786 		rfs4_file_rele(fp);
4787 	}
4788 
4789 	DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs,
4790 	    RENAME4res *, resp);
4791 }
4792 
4793 /* ARGSUSED */
4794 static void
4795 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4796     struct compound_state *cs)
4797 {
4798 	RENEW4args *args = &argop->nfs_argop4_u.oprenew;
4799 	RENEW4res *resp = &resop->nfs_resop4_u.oprenew;
4800 	rfs4_client_t *cp;
4801 
4802 	DTRACE_NFSV4_2(op__renew__start, struct compound_state *, cs,
4803 	    RENEW4args *, args);
4804 
4805 	if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) {
4806 		*cs->statusp = resp->status =
4807 		    rfs4_check_clientid(&args->clientid, 0);
4808 		goto out;
4809 	}
4810 
4811 	if (rfs4_lease_expired(cp)) {
4812 		rfs4_client_rele(cp);
4813 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
4814 		goto out;
4815 	}
4816 
4817 	rfs4_update_lease(cp);
4818 
4819 	mutex_enter(cp->rc_cbinfo.cb_lock);
4820 	if (cp->rc_cbinfo.cb_notified_of_cb_path_down == FALSE) {
4821 		cp->rc_cbinfo.cb_notified_of_cb_path_down = TRUE;
4822 		*cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN;
4823 	} else {
4824 		*cs->statusp = resp->status = NFS4_OK;
4825 	}
4826 	mutex_exit(cp->rc_cbinfo.cb_lock);
4827 
4828 	rfs4_client_rele(cp);
4829 
4830 out:
4831 	DTRACE_NFSV4_2(op__renew__done, struct compound_state *, cs,
4832 	    RENEW4res *, resp);
4833 }
4834 
4835 /* ARGSUSED */
4836 static void
4837 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req,
4838     struct compound_state *cs)
4839 {
4840 	RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh;
4841 
4842 	DTRACE_NFSV4_1(op__restorefh__start, struct compound_state *, cs);
4843 
4844 	/* No need to check cs->access - we are not accessing any object */
4845 	if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) {
4846 		*cs->statusp = resp->status = NFS4ERR_RESTOREFH;
4847 		goto out;
4848 	}
4849 	if (cs->vp != NULL) {
4850 		VN_RELE(cs->vp);
4851 	}
4852 	cs->vp = cs->saved_vp;
4853 	cs->saved_vp = NULL;
4854 	cs->exi = cs->saved_exi;
4855 	nfs_fh4_copy(&cs->saved_fh, &cs->fh);
4856 	*cs->statusp = resp->status = NFS4_OK;
4857 	cs->deleg = FALSE;
4858 
4859 out:
4860 	DTRACE_NFSV4_2(op__restorefh__done, struct compound_state *, cs,
4861 	    RESTOREFH4res *, resp);
4862 }
4863 
4864 /* ARGSUSED */
4865 static void
4866 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
4867     struct compound_state *cs)
4868 {
4869 	SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh;
4870 
4871 	DTRACE_NFSV4_1(op__savefh__start, struct compound_state *, cs);
4872 
4873 	/* No need to check cs->access - we are not accessing any object */
4874 	if (cs->vp == NULL) {
4875 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
4876 		goto out;
4877 	}
4878 	if (cs->saved_vp != NULL) {
4879 		VN_RELE(cs->saved_vp);
4880 	}
4881 	cs->saved_vp = cs->vp;
4882 	VN_HOLD(cs->saved_vp);
4883 	cs->saved_exi = cs->exi;
4884 	/*
4885 	 * since SAVEFH is fairly rare, don't alloc space for its fh
4886 	 * unless necessary.
4887 	 */
4888 	if (cs->saved_fh.nfs_fh4_val == NULL) {
4889 		cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP);
4890 	}
4891 	nfs_fh4_copy(&cs->fh, &cs->saved_fh);
4892 	*cs->statusp = resp->status = NFS4_OK;
4893 
4894 out:
4895 	DTRACE_NFSV4_2(op__savefh__done, struct compound_state *, cs,
4896 	    SAVEFH4res *, resp);
4897 }
4898 
4899 /*
4900  * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to
4901  * return the bitmap of attrs that were set successfully. It is also
4902  * called by Verify/Nverify to test the vattr/vfsstat attrs. It should
4903  * always be called only after rfs4_do_set_attrs().
4904  *
4905  * Verify that the attributes are same as the expected ones. sargp->vap
4906  * and sargp->sbp contain the input attributes as translated from fattr4.
4907  *
4908  * This function verifies only the attrs that correspond to a vattr or
4909  * vfsstat struct. That is because of the extra step needed to get the
4910  * corresponding system structs. Other attributes have already been set or
4911  * verified by do_rfs4_set_attrs.
4912  *
4913  * Return 0 if all attrs match, -1 if some don't, error if error processing.
4914  */
4915 static int
4916 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp,
4917     bitmap4 *resp, struct nfs4_ntov_table *ntovp)
4918 {
4919 	int error, ret_error = 0;
4920 	int i, k;
4921 	uint_t sva_mask = sargp->vap->va_mask;
4922 	uint_t vbit;
4923 	union nfs4_attr_u *na;
4924 	uint8_t *amap;
4925 	bool_t getsb = ntovp->vfsstat;
4926 
4927 	if (sva_mask != 0) {
4928 		/*
4929 		 * Okay to overwrite sargp->vap because we verify based
4930 		 * on the incoming values.
4931 		 */
4932 		ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0,
4933 		    sargp->cs->cr, NULL);
4934 		if (ret_error) {
4935 			if (resp == NULL)
4936 				return (ret_error);
4937 			/*
4938 			 * Must return bitmap of successful attrs
4939 			 */
4940 			sva_mask = 0;	/* to prevent checking vap later */
4941 		} else {
4942 			/*
4943 			 * Some file systems clobber va_mask. it is probably
4944 			 * wrong of them to do so, nonethless we practice
4945 			 * defensive coding.
4946 			 * See bug id 4276830.
4947 			 */
4948 			sargp->vap->va_mask = sva_mask;
4949 		}
4950 	}
4951 
4952 	if (getsb) {
4953 		/*
4954 		 * Now get the superblock and loop on the bitmap, as there is
4955 		 * no simple way of translating from superblock to bitmap4.
4956 		 */
4957 		ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp);
4958 		if (ret_error) {
4959 			if (resp == NULL)
4960 				goto errout;
4961 			getsb = FALSE;
4962 		}
4963 	}
4964 
4965 	/*
4966 	 * Now loop and verify each attribute which getattr returned
4967 	 * whether it's the same as the input.
4968 	 */
4969 	if (resp == NULL && !getsb && (sva_mask == 0))
4970 		goto errout;
4971 
4972 	na = ntovp->na;
4973 	amap = ntovp->amap;
4974 	k = 0;
4975 	for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) {
4976 		k = *amap;
4977 		ASSERT(nfs4_ntov_map[k].nval == k);
4978 		vbit = nfs4_ntov_map[k].vbit;
4979 
4980 		/*
4981 		 * If vattr attribute but VOP_GETATTR failed, or it's
4982 		 * superblock attribute but VFS_STATVFS failed, skip
4983 		 */
4984 		if (vbit) {
4985 			if ((vbit & sva_mask) == 0)
4986 				continue;
4987 		} else if (!(getsb && nfs4_ntov_map[k].vfsstat)) {
4988 			continue;
4989 		}
4990 		error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na);
4991 		if (resp != NULL) {
4992 			if (error)
4993 				ret_error = -1;	/* not all match */
4994 			else	/* update response bitmap */
4995 				*resp |= nfs4_ntov_map[k].fbit;
4996 			continue;
4997 		}
4998 		if (error) {
4999 			ret_error = -1;	/* not all match */
5000 			break;
5001 		}
5002 	}
5003 errout:
5004 	return (ret_error);
5005 }
5006 
5007 /*
5008  * Decode the attribute to be set/verified. If the attr requires a sys op
5009  * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't
5010  * call the sv_getit function for it, because the sys op hasn't yet been done.
5011  * Return 0 for success, error code if failed.
5012  *
5013  * Note: the decoded arg is not freed here but in nfs4_ntov_table_free.
5014  */
5015 static int
5016 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp,
5017     int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap)
5018 {
5019 	int error = 0;
5020 	bool_t set_later;
5021 
5022 	sargp->vap->va_mask |= nfs4_ntov_map[k].vbit;
5023 
5024 	if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) {
5025 		set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat;
5026 		/*
5027 		 * don't verify yet if a vattr or sb dependent attr,
5028 		 * because we don't have their sys values yet.
5029 		 * Will be done later.
5030 		 */
5031 		if (! (set_later && (cmd == NFS4ATTR_VERIT))) {
5032 			/*
5033 			 * ACLs are a special case, since setting the MODE
5034 			 * conflicts with setting the ACL.  We delay setting
5035 			 * the ACL until all other attributes have been set.
5036 			 * The ACL gets set in do_rfs4_op_setattr().
5037 			 */
5038 			if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) {
5039 				error = (*nfs4_ntov_map[k].sv_getit)(cmd,
5040 				    sargp, nap);
5041 				if (error) {
5042 					xdr_free(nfs4_ntov_map[k].xfunc,
5043 					    (caddr_t)nap);
5044 				}
5045 			}
5046 		}
5047 	} else {
5048 #ifdef  DEBUG
5049 		cmn_err(CE_NOTE, "decode_fattr4_attr: error "
5050 		    "decoding attribute %d\n", k);
5051 #endif
5052 		error = EINVAL;
5053 	}
5054 	if (!error && resp_bval && !set_later) {
5055 		*resp_bval |= nfs4_ntov_map[k].fbit;
5056 	}
5057 
5058 	return (error);
5059 }
5060 
5061 /*
5062  * Set vattr based on incoming fattr4 attrs - used by setattr.
5063  * Set response mask. Ignore any values that are not writable vattr attrs.
5064  */
5065 static nfsstat4
5066 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
5067     struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp,
5068     nfs4_attr_cmd_t cmd)
5069 {
5070 	int error = 0;
5071 	int i;
5072 	char *attrs = fattrp->attrlist4;
5073 	uint32_t attrslen = fattrp->attrlist4_len;
5074 	XDR xdr;
5075 	nfsstat4 status = NFS4_OK;
5076 	vnode_t *vp = cs->vp;
5077 	union nfs4_attr_u *na;
5078 	uint8_t *amap;
5079 
5080 #ifndef lint
5081 	/*
5082 	 * Make sure that maximum attribute number can be expressed as an
5083 	 * 8 bit quantity.
5084 	 */
5085 	ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1));
5086 #endif
5087 
5088 	if (vp == NULL) {
5089 		if (resp)
5090 			*resp = 0;
5091 		return (NFS4ERR_NOFILEHANDLE);
5092 	}
5093 	if (cs->access == CS_ACCESS_DENIED) {
5094 		if (resp)
5095 			*resp = 0;
5096 		return (NFS4ERR_ACCESS);
5097 	}
5098 
5099 	sargp->op = cmd;
5100 	sargp->cs = cs;
5101 	sargp->flag = 0;	/* may be set later */
5102 	sargp->vap->va_mask = 0;
5103 	sargp->rdattr_error = NFS4_OK;
5104 	sargp->rdattr_error_req = FALSE;
5105 	/* sargp->sbp is set by the caller */
5106 
5107 	xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE);
5108 
5109 	na = ntovp->na;
5110 	amap = ntovp->amap;
5111 
5112 	/*
5113 	 * The following loop iterates on the nfs4_ntov_map checking
5114 	 * if the fbit is set in the requested bitmap.
5115 	 * If set then we process the arguments using the
5116 	 * rfs4_fattr4 conversion functions to populate the setattr
5117 	 * vattr and va_mask. Any settable attrs that are not using vattr
5118 	 * will be set in this loop.
5119 	 */
5120 	for (i = 0; i < nfs4_ntov_map_size; i++) {
5121 		if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) {
5122 			continue;
5123 		}
5124 		/*
5125 		 * If setattr, must be a writable attr.
5126 		 * If verify/nverify, must be a readable attr.
5127 		 */
5128 		if ((error = (*nfs4_ntov_map[i].sv_getit)(
5129 		    NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) {
5130 			/*
5131 			 * Client tries to set/verify an
5132 			 * unsupported attribute, tries to set
5133 			 * a read only attr or verify a write
5134 			 * only one - error!
5135 			 */
5136 			break;
5137 		}
5138 		/*
5139 		 * Decode the attribute to set/verify
5140 		 */
5141 		error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval,
5142 		    &xdr, resp ? resp : NULL, na);
5143 		if (error)
5144 			break;
5145 		*amap++ = (uint8_t)nfs4_ntov_map[i].nval;
5146 		na++;
5147 		(ntovp->attrcnt)++;
5148 		if (nfs4_ntov_map[i].vfsstat)
5149 			ntovp->vfsstat = TRUE;
5150 	}
5151 
5152 	if (error != 0)
5153 		status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP :
5154 		    puterrno4(error));
5155 	/* xdrmem_destroy(&xdrs); */	/* NO-OP */
5156 	return (status);
5157 }
5158 
5159 static nfsstat4
5160 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs,
5161     stateid4 *stateid)
5162 {
5163 	int error = 0;
5164 	struct nfs4_svgetit_arg sarg;
5165 	bool_t trunc;
5166 
5167 	nfsstat4 status = NFS4_OK;
5168 	cred_t *cr = cs->cr;
5169 	vnode_t *vp = cs->vp;
5170 	struct nfs4_ntov_table ntov;
5171 	struct statvfs64 sb;
5172 	struct vattr bva;
5173 	struct flock64 bf;
5174 	int in_crit = 0;
5175 	uint_t saved_mask = 0;
5176 	caller_context_t ct;
5177 
5178 	*resp = 0;
5179 	sarg.sbp = &sb;
5180 	sarg.is_referral = B_FALSE;
5181 	nfs4_ntov_table_init(&ntov);
5182 	status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov,
5183 	    NFS4ATTR_SETIT);
5184 	if (status != NFS4_OK) {
5185 		/*
5186 		 * failed set attrs
5187 		 */
5188 		goto done;
5189 	}
5190 	if ((sarg.vap->va_mask == 0) &&
5191 	    (! (fattrp->attrmask & FATTR4_ACL_MASK))) {
5192 		/*
5193 		 * no further work to be done
5194 		 */
5195 		goto done;
5196 	}
5197 
5198 	/*
5199 	 * If we got a request to set the ACL and the MODE, only
5200 	 * allow changing VSUID, VSGID, and VSVTX.  Attempting
5201 	 * to change any other bits, along with setting an ACL,
5202 	 * gives NFS4ERR_INVAL.
5203 	 */
5204 	if ((fattrp->attrmask & FATTR4_ACL_MASK) &&
5205 	    (fattrp->attrmask & FATTR4_MODE_MASK)) {
5206 		vattr_t va;
5207 
5208 		va.va_mask = AT_MODE;
5209 		error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL);
5210 		if (error) {
5211 			status = puterrno4(error);
5212 			goto done;
5213 		}
5214 		if ((sarg.vap->va_mode ^ va.va_mode) &
5215 		    ~(VSUID | VSGID | VSVTX)) {
5216 			status = NFS4ERR_INVAL;
5217 			goto done;
5218 		}
5219 	}
5220 
5221 	/* Check stateid only if size has been set */
5222 	if (sarg.vap->va_mask & AT_SIZE) {
5223 		trunc = (sarg.vap->va_size == 0);
5224 		status = rfs4_check_stateid(FWRITE, cs->vp, stateid,
5225 		    trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct);
5226 		if (status != NFS4_OK)
5227 			goto done;
5228 	} else {
5229 		ct.cc_sysid = 0;
5230 		ct.cc_pid = 0;
5231 		ct.cc_caller_id = nfs4_srv_caller_id;
5232 		ct.cc_flags = CC_DONTBLOCK;
5233 	}
5234 
5235 	/* XXX start of possible race with delegations */
5236 
5237 	/*
5238 	 * We need to specially handle size changes because it is
5239 	 * possible for the client to create a file with read-only
5240 	 * modes, but with the file opened for writing. If the client
5241 	 * then tries to set the file size, e.g. ftruncate(3C),
5242 	 * fcntl(F_FREESP), the normal access checking done in
5243 	 * VOP_SETATTR would prevent the client from doing it even though
5244 	 * it should be allowed to do so.  To get around this, we do the
5245 	 * access checking for ourselves and use VOP_SPACE which doesn't
5246 	 * do the access checking.
5247 	 * Also the client should not be allowed to change the file
5248 	 * size if there is a conflicting non-blocking mandatory lock in
5249 	 * the region of the change.
5250 	 */
5251 	if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) {
5252 		u_offset_t offset;
5253 		ssize_t length;
5254 
5255 		/*
5256 		 * ufs_setattr clears AT_SIZE from vap->va_mask, but
5257 		 * before returning, sarg.vap->va_mask is used to
5258 		 * generate the setattr reply bitmap.  We also clear
5259 		 * AT_SIZE below before calling VOP_SPACE.  For both
5260 		 * of these cases, the va_mask needs to be saved here
5261 		 * and restored after calling VOP_SETATTR.
5262 		 */
5263 		saved_mask = sarg.vap->va_mask;
5264 
5265 		/*
5266 		 * Check any possible conflict due to NBMAND locks.
5267 		 * Get into critical region before VOP_GETATTR, so the
5268 		 * size attribute is valid when checking conflicts.
5269 		 */
5270 		if (nbl_need_check(vp)) {
5271 			nbl_start_crit(vp, RW_READER);
5272 			in_crit = 1;
5273 		}
5274 
5275 		bva.va_mask = AT_UID|AT_SIZE;
5276 		if (error = VOP_GETATTR(vp, &bva, 0, cr, &ct)) {
5277 			status = puterrno4(error);
5278 			goto done;
5279 		}
5280 
5281 		if (in_crit) {
5282 			if (sarg.vap->va_size < bva.va_size) {
5283 				offset = sarg.vap->va_size;
5284 				length = bva.va_size - sarg.vap->va_size;
5285 			} else {
5286 				offset = bva.va_size;
5287 				length = sarg.vap->va_size - bva.va_size;
5288 			}
5289 			if (nbl_conflict(vp, NBL_WRITE, offset, length, 0,
5290 			    &ct)) {
5291 				status = NFS4ERR_LOCKED;
5292 				goto done;
5293 			}
5294 		}
5295 
5296 		if (crgetuid(cr) == bva.va_uid) {
5297 			sarg.vap->va_mask &= ~AT_SIZE;
5298 			bf.l_type = F_WRLCK;
5299 			bf.l_whence = 0;
5300 			bf.l_start = (off64_t)sarg.vap->va_size;
5301 			bf.l_len = 0;
5302 			bf.l_sysid = 0;
5303 			bf.l_pid = 0;
5304 			error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE,
5305 			    (offset_t)sarg.vap->va_size, cr, &ct);
5306 		}
5307 	}
5308 
5309 	if (!error && sarg.vap->va_mask != 0)
5310 		error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct);
5311 
5312 	/* restore va_mask -- ufs_setattr clears AT_SIZE */
5313 	if (saved_mask & AT_SIZE)
5314 		sarg.vap->va_mask |= AT_SIZE;
5315 
5316 	/*
5317 	 * If an ACL was being set, it has been delayed until now,
5318 	 * in order to set the mode (via the VOP_SETATTR() above) first.
5319 	 */
5320 	if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) {
5321 		int i;
5322 
5323 		for (i = 0; i < NFS4_MAXNUM_ATTRS; i++)
5324 			if (ntov.amap[i] == FATTR4_ACL)
5325 				break;
5326 		if (i < NFS4_MAXNUM_ATTRS) {
5327 			error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)(
5328 			    NFS4ATTR_SETIT, &sarg, &ntov.na[i]);
5329 			if (error == 0) {
5330 				*resp |= FATTR4_ACL_MASK;
5331 			} else if (error == ENOTSUP) {
5332 				(void) rfs4_verify_attr(&sarg, resp, &ntov);
5333 				status = NFS4ERR_ATTRNOTSUPP;
5334 				goto done;
5335 			}
5336 		} else {
5337 			NFS4_DEBUG(rfs4_debug,
5338 			    (CE_NOTE, "do_rfs4_op_setattr: "
5339 			    "unable to find ACL in fattr4"));
5340 			error = EINVAL;
5341 		}
5342 	}
5343 
5344 	if (error) {
5345 		/* check if a monitor detected a delegation conflict */
5346 		if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK))
5347 			status = NFS4ERR_DELAY;
5348 		else
5349 			status = puterrno4(error);
5350 
5351 		/*
5352 		 * Set the response bitmap when setattr failed.
5353 		 * If VOP_SETATTR partially succeeded, test by doing a
5354 		 * VOP_GETATTR on the object and comparing the data
5355 		 * to the setattr arguments.
5356 		 */
5357 		(void) rfs4_verify_attr(&sarg, resp, &ntov);
5358 	} else {
5359 		/*
5360 		 * Force modified metadata out to stable storage.
5361 		 */
5362 		(void) VOP_FSYNC(vp, FNODSYNC, cr, &ct);
5363 		/*
5364 		 * Set response bitmap
5365 		 */
5366 		nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp);
5367 	}
5368 
5369 /* Return early and already have a NFSv4 error */
5370 done:
5371 	/*
5372 	 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr
5373 	 * conversion sets both readable and writeable NFS4 attrs
5374 	 * for AT_MTIME and AT_ATIME.  The line below masks out
5375 	 * unrequested attrs from the setattr result bitmap.  This
5376 	 * is placed after the done: label to catch the ATTRNOTSUP
5377 	 * case.
5378 	 */
5379 	*resp &= fattrp->attrmask;
5380 
5381 	if (in_crit)
5382 		nbl_end_crit(vp);
5383 
5384 	nfs4_ntov_table_free(&ntov, &sarg);
5385 
5386 	return (status);
5387 }
5388 
5389 /* ARGSUSED */
5390 static void
5391 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5392     struct compound_state *cs)
5393 {
5394 	SETATTR4args *args = &argop->nfs_argop4_u.opsetattr;
5395 	SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr;
5396 	bslabel_t *clabel;
5397 
5398 	DTRACE_NFSV4_2(op__setattr__start, struct compound_state *, cs,
5399 	    SETATTR4args *, args);
5400 
5401 	if (cs->vp == NULL) {
5402 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5403 		goto out;
5404 	}
5405 
5406 	/*
5407 	 * If there is an unshared filesystem mounted on this vnode,
5408 	 * do not allow to setattr on this vnode.
5409 	 */
5410 	if (vn_ismntpt(cs->vp)) {
5411 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5412 		goto out;
5413 	}
5414 
5415 	resp->attrsset = 0;
5416 
5417 	if (rdonly4(req, cs)) {
5418 		*cs->statusp = resp->status = NFS4ERR_ROFS;
5419 		goto out;
5420 	}
5421 
5422 	/* check label before setting attributes */
5423 	if (is_system_labeled()) {
5424 		ASSERT(req->rq_label != NULL);
5425 		clabel = req->rq_label;
5426 		DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *,
5427 		    "got client label from request(1)",
5428 		    struct svc_req *, req);
5429 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
5430 			if (!do_rfs_label_check(clabel, cs->vp,
5431 			    EQUALITY_CHECK, cs->exi)) {
5432 				*cs->statusp = resp->status = NFS4ERR_ACCESS;
5433 				goto out;
5434 			}
5435 		}
5436 	}
5437 
5438 	*cs->statusp = resp->status =
5439 	    do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs,
5440 	    &args->stateid);
5441 
5442 out:
5443 	DTRACE_NFSV4_2(op__setattr__done, struct compound_state *, cs,
5444 	    SETATTR4res *, resp);
5445 }
5446 
5447 /* ARGSUSED */
5448 static void
5449 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5450     struct compound_state *cs)
5451 {
5452 	/*
5453 	 * verify and nverify are exactly the same, except that nverify
5454 	 * succeeds when some argument changed, and verify succeeds when
5455 	 * when none changed.
5456 	 */
5457 
5458 	VERIFY4args  *args = &argop->nfs_argop4_u.opverify;
5459 	VERIFY4res *resp = &resop->nfs_resop4_u.opverify;
5460 
5461 	int error;
5462 	struct nfs4_svgetit_arg sarg;
5463 	struct statvfs64 sb;
5464 	struct nfs4_ntov_table ntov;
5465 
5466 	DTRACE_NFSV4_2(op__verify__start, struct compound_state *, cs,
5467 	    VERIFY4args *, args);
5468 
5469 	if (cs->vp == NULL) {
5470 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5471 		goto out;
5472 	}
5473 
5474 	sarg.sbp = &sb;
5475 	sarg.is_referral = B_FALSE;
5476 	nfs4_ntov_table_init(&ntov);
5477 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
5478 	    &sarg, &ntov, NFS4ATTR_VERIT);
5479 	if (resp->status != NFS4_OK) {
5480 		/*
5481 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
5482 		 * so could return -1 for "no match".
5483 		 */
5484 		if (resp->status == -1)
5485 			resp->status = NFS4ERR_NOT_SAME;
5486 		goto done;
5487 	}
5488 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
5489 	switch (error) {
5490 	case 0:
5491 		resp->status = NFS4_OK;
5492 		break;
5493 	case -1:
5494 		resp->status = NFS4ERR_NOT_SAME;
5495 		break;
5496 	default:
5497 		resp->status = puterrno4(error);
5498 		break;
5499 	}
5500 done:
5501 	*cs->statusp = resp->status;
5502 	nfs4_ntov_table_free(&ntov, &sarg);
5503 out:
5504 	DTRACE_NFSV4_2(op__verify__done, struct compound_state *, cs,
5505 	    VERIFY4res *, resp);
5506 }
5507 
5508 /* ARGSUSED */
5509 static void
5510 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5511     struct compound_state *cs)
5512 {
5513 	/*
5514 	 * verify and nverify are exactly the same, except that nverify
5515 	 * succeeds when some argument changed, and verify succeeds when
5516 	 * when none changed.
5517 	 */
5518 
5519 	NVERIFY4args  *args = &argop->nfs_argop4_u.opnverify;
5520 	NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify;
5521 
5522 	int error;
5523 	struct nfs4_svgetit_arg sarg;
5524 	struct statvfs64 sb;
5525 	struct nfs4_ntov_table ntov;
5526 
5527 	DTRACE_NFSV4_2(op__nverify__start, struct compound_state *, cs,
5528 	    NVERIFY4args *, args);
5529 
5530 	if (cs->vp == NULL) {
5531 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5532 		DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs,
5533 		    NVERIFY4res *, resp);
5534 		return;
5535 	}
5536 	sarg.sbp = &sb;
5537 	sarg.is_referral = B_FALSE;
5538 	nfs4_ntov_table_init(&ntov);
5539 	resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs,
5540 	    &sarg, &ntov, NFS4ATTR_VERIT);
5541 	if (resp->status != NFS4_OK) {
5542 		/*
5543 		 * do_rfs4_set_attrs will try to verify systemwide attrs,
5544 		 * so could return -1 for "no match".
5545 		 */
5546 		if (resp->status == -1)
5547 			resp->status = NFS4_OK;
5548 		goto done;
5549 	}
5550 	error = rfs4_verify_attr(&sarg, NULL, &ntov);
5551 	switch (error) {
5552 	case 0:
5553 		resp->status = NFS4ERR_SAME;
5554 		break;
5555 	case -1:
5556 		resp->status = NFS4_OK;
5557 		break;
5558 	default:
5559 		resp->status = puterrno4(error);
5560 		break;
5561 	}
5562 done:
5563 	*cs->statusp = resp->status;
5564 	nfs4_ntov_table_free(&ntov, &sarg);
5565 
5566 	DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs,
5567 	    NVERIFY4res *, resp);
5568 }
5569 
5570 /*
5571  * XXX - This should live in an NFS header file.
5572  */
5573 #define	MAX_IOVECS	12
5574 
5575 /* ARGSUSED */
5576 static void
5577 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req,
5578     struct compound_state *cs)
5579 {
5580 	WRITE4args *args = &argop->nfs_argop4_u.opwrite;
5581 	WRITE4res *resp = &resop->nfs_resop4_u.opwrite;
5582 	int error;
5583 	vnode_t *vp;
5584 	struct vattr bva;
5585 	u_offset_t rlimit;
5586 	struct uio uio;
5587 	struct iovec iov[MAX_IOVECS];
5588 	struct iovec *iovp;
5589 	int iovcnt;
5590 	int ioflag;
5591 	cred_t *savecred, *cr;
5592 	bool_t *deleg = &cs->deleg;
5593 	nfsstat4 stat;
5594 	int in_crit = 0;
5595 	caller_context_t ct;
5596 
5597 	DTRACE_NFSV4_2(op__write__start, struct compound_state *, cs,
5598 	    WRITE4args *, args);
5599 
5600 	vp = cs->vp;
5601 	if (vp == NULL) {
5602 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
5603 		goto out;
5604 	}
5605 	if (cs->access == CS_ACCESS_DENIED) {
5606 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5607 		goto out;
5608 	}
5609 
5610 	cr = cs->cr;
5611 
5612 	if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE,
5613 	    deleg, TRUE, &ct)) != NFS4_OK) {
5614 		*cs->statusp = resp->status = stat;
5615 		goto out;
5616 	}
5617 
5618 	/*
5619 	 * We have to enter the critical region before calling VOP_RWLOCK
5620 	 * to avoid a deadlock with ufs.
5621 	 */
5622 	if (nbl_need_check(vp)) {
5623 		nbl_start_crit(vp, RW_READER);
5624 		in_crit = 1;
5625 		if (nbl_conflict(vp, NBL_WRITE,
5626 		    args->offset, args->data_len, 0, &ct)) {
5627 			*cs->statusp = resp->status = NFS4ERR_LOCKED;
5628 			goto out;
5629 		}
5630 	}
5631 
5632 	bva.va_mask = AT_MODE | AT_UID;
5633 	error = VOP_GETATTR(vp, &bva, 0, cr, &ct);
5634 
5635 	/*
5636 	 * If we can't get the attributes, then we can't do the
5637 	 * right access checking.  So, we'll fail the request.
5638 	 */
5639 	if (error) {
5640 		*cs->statusp = resp->status = puterrno4(error);
5641 		goto out;
5642 	}
5643 
5644 	if (rdonly4(req, cs)) {
5645 		*cs->statusp = resp->status = NFS4ERR_ROFS;
5646 		goto out;
5647 	}
5648 
5649 	if (vp->v_type != VREG) {
5650 		*cs->statusp = resp->status =
5651 		    ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL);
5652 		goto out;
5653 	}
5654 
5655 	if (crgetuid(cr) != bva.va_uid &&
5656 	    (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) {
5657 		*cs->statusp = resp->status = puterrno4(error);
5658 		goto out;
5659 	}
5660 
5661 	if (MANDLOCK(vp, bva.va_mode)) {
5662 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
5663 		goto out;
5664 	}
5665 
5666 	if (args->data_len == 0) {
5667 		*cs->statusp = resp->status = NFS4_OK;
5668 		resp->count = 0;
5669 		resp->committed = args->stable;
5670 		resp->writeverf = Write4verf;
5671 		goto out;
5672 	}
5673 
5674 	if (args->mblk != NULL) {
5675 		mblk_t *m;
5676 		uint_t bytes, round_len;
5677 
5678 		iovcnt = 0;
5679 		bytes = 0;
5680 		round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT);
5681 		for (m = args->mblk;
5682 		    m != NULL && bytes < round_len;
5683 		    m = m->b_cont) {
5684 			iovcnt++;
5685 			bytes += MBLKL(m);
5686 		}
5687 #ifdef DEBUG
5688 		/* should have ended on an mblk boundary */
5689 		if (bytes != round_len) {
5690 			printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n",
5691 			    bytes, round_len, args->data_len);
5692 			printf("args=%p, args->mblk=%p, m=%p", (void *)args,
5693 			    (void *)args->mblk, (void *)m);
5694 			ASSERT(bytes == round_len);
5695 		}
5696 #endif
5697 		if (iovcnt <= MAX_IOVECS) {
5698 			iovp = iov;
5699 		} else {
5700 			iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP);
5701 		}
5702 		mblk_to_iov(args->mblk, iovcnt, iovp);
5703 	} else if (args->rlist != NULL) {
5704 		iovcnt = 1;
5705 		iovp = iov;
5706 		iovp->iov_base = (char *)((args->rlist)->u.c_daddr3);
5707 		iovp->iov_len = args->data_len;
5708 	} else {
5709 		iovcnt = 1;
5710 		iovp = iov;
5711 		iovp->iov_base = args->data_val;
5712 		iovp->iov_len = args->data_len;
5713 	}
5714 
5715 	uio.uio_iov = iovp;
5716 	uio.uio_iovcnt = iovcnt;
5717 
5718 	uio.uio_segflg = UIO_SYSSPACE;
5719 	uio.uio_extflg = UIO_COPY_DEFAULT;
5720 	uio.uio_loffset = args->offset;
5721 	uio.uio_resid = args->data_len;
5722 	uio.uio_llimit = curproc->p_fsz_ctl;
5723 	rlimit = uio.uio_llimit - args->offset;
5724 	if (rlimit < (u_offset_t)uio.uio_resid)
5725 		uio.uio_resid = (int)rlimit;
5726 
5727 	if (args->stable == UNSTABLE4)
5728 		ioflag = 0;
5729 	else if (args->stable == FILE_SYNC4)
5730 		ioflag = FSYNC;
5731 	else if (args->stable == DATA_SYNC4)
5732 		ioflag = FDSYNC;
5733 	else {
5734 		if (iovp != iov)
5735 			kmem_free(iovp, sizeof (*iovp) * iovcnt);
5736 		*cs->statusp = resp->status = NFS4ERR_INVAL;
5737 		goto out;
5738 	}
5739 
5740 	/*
5741 	 * We're changing creds because VM may fault and we need
5742 	 * the cred of the current thread to be used if quota
5743 	 * checking is enabled.
5744 	 */
5745 	savecred = curthread->t_cred;
5746 	curthread->t_cred = cr;
5747 	error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct);
5748 	curthread->t_cred = savecred;
5749 
5750 	if (iovp != iov)
5751 		kmem_free(iovp, sizeof (*iovp) * iovcnt);
5752 
5753 	if (error) {
5754 		*cs->statusp = resp->status = puterrno4(error);
5755 		goto out;
5756 	}
5757 
5758 	*cs->statusp = resp->status = NFS4_OK;
5759 	resp->count = args->data_len - uio.uio_resid;
5760 
5761 	if (ioflag == 0)
5762 		resp->committed = UNSTABLE4;
5763 	else
5764 		resp->committed = FILE_SYNC4;
5765 
5766 	resp->writeverf = Write4verf;
5767 
5768 out:
5769 	if (in_crit)
5770 		nbl_end_crit(vp);
5771 
5772 	DTRACE_NFSV4_2(op__write__done, struct compound_state *, cs,
5773 	    WRITE4res *, resp);
5774 }
5775 
5776 
5777 /* XXX put in a header file */
5778 extern int	sec_svc_getcred(struct svc_req *, cred_t *,  caddr_t *, int *);
5779 
5780 void
5781 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi,
5782     struct svc_req *req, cred_t *cr, int *rv)
5783 {
5784 	uint_t i;
5785 	struct compound_state cs;
5786 
5787 	if (rv != NULL)
5788 		*rv = 0;
5789 	rfs4_init_compound_state(&cs);
5790 	/*
5791 	 * Form a reply tag by copying over the reqeuest tag.
5792 	 */
5793 	resp->tag.utf8string_val =
5794 	    kmem_alloc(args->tag.utf8string_len, KM_SLEEP);
5795 	resp->tag.utf8string_len = args->tag.utf8string_len;
5796 	bcopy(args->tag.utf8string_val, resp->tag.utf8string_val,
5797 	    resp->tag.utf8string_len);
5798 
5799 	cs.statusp = &resp->status;
5800 	cs.req = req;
5801 	resp->array = NULL;
5802 	resp->array_len = 0;
5803 
5804 	/*
5805 	 * XXX for now, minorversion should be zero
5806 	 */
5807 	if (args->minorversion != NFS4_MINORVERSION) {
5808 		DTRACE_NFSV4_2(compound__start, struct compound_state *,
5809 		    &cs, COMPOUND4args *, args);
5810 		resp->status = NFS4ERR_MINOR_VERS_MISMATCH;
5811 		DTRACE_NFSV4_2(compound__done, struct compound_state *,
5812 		    &cs, COMPOUND4res *, resp);
5813 		return;
5814 	}
5815 
5816 	if (args->array_len == 0) {
5817 		resp->status = NFS4_OK;
5818 		return;
5819 	}
5820 
5821 	ASSERT(exi == NULL);
5822 	ASSERT(cr == NULL);
5823 
5824 	cr = crget();
5825 	ASSERT(cr != NULL);
5826 
5827 	if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) {
5828 		DTRACE_NFSV4_2(compound__start, struct compound_state *,
5829 		    &cs, COMPOUND4args *, args);
5830 		crfree(cr);
5831 		DTRACE_NFSV4_2(compound__done, struct compound_state *,
5832 		    &cs, COMPOUND4res *, resp);
5833 		svcerr_badcred(req->rq_xprt);
5834 		if (rv != NULL)
5835 			*rv = 1;
5836 		return;
5837 	}
5838 	resp->array_len = args->array_len;
5839 	resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4),
5840 	    KM_SLEEP);
5841 
5842 	cs.basecr = cr;
5843 
5844 	DTRACE_NFSV4_2(compound__start, struct compound_state *, &cs,
5845 	    COMPOUND4args *, args);
5846 
5847 	/*
5848 	 * For now, NFS4 compound processing must be protected by
5849 	 * exported_lock because it can access more than one exportinfo
5850 	 * per compound and share/unshare can now change multiple
5851 	 * exinfo structs.  The NFS2/3 code only refs 1 exportinfo
5852 	 * per proc (excluding public exinfo), and exi_count design
5853 	 * is sufficient to protect concurrent execution of NFS2/3
5854 	 * ops along with unexport.  This lock will be removed as
5855 	 * part of the NFSv4 phase 2 namespace redesign work.
5856 	 */
5857 	rw_enter(&exported_lock, RW_READER);
5858 
5859 	/*
5860 	 * If this is the first compound we've seen, we need to start all
5861 	 * new instances' grace periods.
5862 	 */
5863 	if (rfs4_seen_first_compound == 0) {
5864 		rfs4_grace_start_new();
5865 		/*
5866 		 * This must be set after rfs4_grace_start_new(), otherwise
5867 		 * another thread could proceed past here before the former
5868 		 * is finished.
5869 		 */
5870 		rfs4_seen_first_compound = 1;
5871 	}
5872 
5873 	for (i = 0; i < args->array_len && cs.cont; i++) {
5874 		nfs_argop4 *argop;
5875 		nfs_resop4 *resop;
5876 		uint_t op;
5877 
5878 		argop = &args->array[i];
5879 		resop = &resp->array[i];
5880 		resop->resop = argop->argop;
5881 		op = (uint_t)resop->resop;
5882 
5883 		if (op < rfsv4disp_cnt) {
5884 			kstat_t *ksp = rfsprocio_v4_ptr[op];
5885 			kstat_t *exi_ksp = NULL;
5886 
5887 			/*
5888 			 * Count the individual ops here; NULL and COMPOUND
5889 			 * are counted in common_dispatch()
5890 			 */
5891 			rfsproccnt_v4_ptr[op].value.ui64++;
5892 
5893 			if (ksp != NULL) {
5894 				mutex_enter(ksp->ks_lock);
5895 				kstat_runq_enter(KSTAT_IO_PTR(ksp));
5896 				mutex_exit(ksp->ks_lock);
5897 			}
5898 
5899 			switch (rfsv4disptab[op].op_type) {
5900 			case NFS4_OP_CFH:
5901 				resop->exi = cs.exi;
5902 				break;
5903 			case NFS4_OP_SFH:
5904 				resop->exi = cs.saved_exi;
5905 				break;
5906 			default:
5907 				ASSERT(resop->exi == NULL);
5908 				break;
5909 			}
5910 
5911 			if (resop->exi != NULL) {
5912 				exi_ksp = resop->exi->exi_kstats->
5913 				    rfsprocio_v4_ptr[op];
5914 				if (exi_ksp != NULL) {
5915 					mutex_enter(exi_ksp->ks_lock);
5916 					kstat_runq_enter(KSTAT_IO_PTR(exi_ksp));
5917 					mutex_exit(exi_ksp->ks_lock);
5918 				}
5919 			}
5920 
5921 			NFS4_DEBUG(rfs4_debug > 1,
5922 			    (CE_NOTE, "Executing %s", rfs4_op_string[op]));
5923 			(*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs);
5924 			NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d",
5925 			    rfs4_op_string[op], *cs.statusp));
5926 			if (*cs.statusp != NFS4_OK)
5927 				cs.cont = FALSE;
5928 
5929 			if (rfsv4disptab[op].op_type == NFS4_OP_POSTCFH &&
5930 			    *cs.statusp == NFS4_OK &&
5931 			    (resop->exi = cs.exi) != NULL) {
5932 				exi_ksp = resop->exi->exi_kstats->
5933 				    rfsprocio_v4_ptr[op];
5934 			}
5935 
5936 			if (exi_ksp != NULL) {
5937 				mutex_enter(exi_ksp->ks_lock);
5938 				KSTAT_IO_PTR(exi_ksp)->nwritten +=
5939 				    argop->opsize;
5940 				KSTAT_IO_PTR(exi_ksp)->writes++;
5941 				if (rfsv4disptab[op].op_type != NFS4_OP_POSTCFH)
5942 					kstat_runq_exit(KSTAT_IO_PTR(exi_ksp));
5943 				mutex_exit(exi_ksp->ks_lock);
5944 
5945 				exi_hold(resop->exi);
5946 			} else {
5947 				resop->exi = NULL;
5948 			}
5949 
5950 			if (ksp != NULL) {
5951 				mutex_enter(ksp->ks_lock);
5952 				kstat_runq_exit(KSTAT_IO_PTR(ksp));
5953 				mutex_exit(ksp->ks_lock);
5954 			}
5955 		} else {
5956 			/*
5957 			 * This is effectively dead code since XDR code
5958 			 * will have already returned BADXDR if op doesn't
5959 			 * decode to legal value.  This only done for a
5960 			 * day when XDR code doesn't verify v4 opcodes.
5961 			 */
5962 			op = OP_ILLEGAL;
5963 			rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++;
5964 
5965 			rfs4_op_illegal(argop, resop, req, &cs);
5966 			cs.cont = FALSE;
5967 		}
5968 
5969 		/*
5970 		 * If not at last op, and if we are to stop, then
5971 		 * compact the results array.
5972 		 */
5973 		if ((i + 1) < args->array_len && !cs.cont) {
5974 			nfs_resop4 *new_res = kmem_alloc(
5975 			    (i + 1) * sizeof (nfs_resop4), KM_SLEEP);
5976 			bcopy(resp->array,
5977 			    new_res, (i + 1) * sizeof (nfs_resop4));
5978 			kmem_free(resp->array,
5979 			    args->array_len * sizeof (nfs_resop4));
5980 
5981 			resp->array_len = i + 1;
5982 			resp->array = new_res;
5983 		}
5984 	}
5985 
5986 	rw_exit(&exported_lock);
5987 
5988 	DTRACE_NFSV4_2(compound__done, struct compound_state *, &cs,
5989 	    COMPOUND4res *, resp);
5990 
5991 	if (cs.vp)
5992 		VN_RELE(cs.vp);
5993 	if (cs.saved_vp)
5994 		VN_RELE(cs.saved_vp);
5995 	if (cs.saved_fh.nfs_fh4_val)
5996 		kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE);
5997 
5998 	if (cs.basecr)
5999 		crfree(cs.basecr);
6000 	if (cs.cr)
6001 		crfree(cs.cr);
6002 	/*
6003 	 * done with this compound request, free the label
6004 	 */
6005 
6006 	if (req->rq_label != NULL) {
6007 		kmem_free(req->rq_label, sizeof (bslabel_t));
6008 		req->rq_label = NULL;
6009 	}
6010 }
6011 
6012 /*
6013  * XXX because of what appears to be duplicate calls to rfs4_compound_free
6014  * XXX zero out the tag and array values. Need to investigate why the
6015  * XXX calls occur, but at least prevent the panic for now.
6016  */
6017 void
6018 rfs4_compound_free(COMPOUND4res *resp)
6019 {
6020 	uint_t i;
6021 
6022 	if (resp->tag.utf8string_val) {
6023 		UTF8STRING_FREE(resp->tag)
6024 	}
6025 
6026 	for (i = 0; i < resp->array_len; i++) {
6027 		nfs_resop4 *resop;
6028 		uint_t op;
6029 
6030 		resop = &resp->array[i];
6031 		op = (uint_t)resop->resop;
6032 		if (op < rfsv4disp_cnt) {
6033 			(*rfsv4disptab[op].dis_resfree)(resop);
6034 		}
6035 	}
6036 	if (resp->array != NULL) {
6037 		kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4));
6038 	}
6039 }
6040 
6041 /*
6042  * Process the value of the compound request rpc flags, as a bit-AND
6043  * of the individual per-op flags (idempotent, allowork, publicfh_ok)
6044  */
6045 void
6046 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp)
6047 {
6048 	int i;
6049 	int flag = RPC_ALL;
6050 
6051 	for (i = 0; flag && i < args->array_len; i++) {
6052 		uint_t op;
6053 
6054 		op = (uint_t)args->array[i].argop;
6055 
6056 		if (op < rfsv4disp_cnt)
6057 			flag &= rfsv4disptab[op].dis_flags;
6058 		else
6059 			flag = 0;
6060 	}
6061 	*flagp = flag;
6062 }
6063 
6064 void
6065 rfs4_compound_kstat_args(COMPOUND4args *args)
6066 {
6067 	int i;
6068 
6069 	for (i = 0; i < args->array_len; i++) {
6070 		uint_t op = (uint_t)args->array[i].argop;
6071 
6072 		if (op < rfsv4disp_cnt) {
6073 			kstat_t *ksp = rfsprocio_v4_ptr[op];
6074 
6075 			if (ksp != NULL) {
6076 				mutex_enter(ksp->ks_lock);
6077 				KSTAT_IO_PTR(ksp)->nwritten +=
6078 				    args->array[i].opsize;
6079 				KSTAT_IO_PTR(ksp)->writes++;
6080 				mutex_exit(ksp->ks_lock);
6081 			}
6082 		}
6083 	}
6084 }
6085 
6086 void
6087 rfs4_compound_kstat_res(COMPOUND4res *res)
6088 {
6089 	int i;
6090 
6091 	for (i = 0; i < res->array_len; i++) {
6092 		uint_t op = (uint_t)res->array[i].resop;
6093 
6094 		if (op < rfsv4disp_cnt) {
6095 			kstat_t *ksp = rfsprocio_v4_ptr[op];
6096 			struct exportinfo *exi = res->array[i].exi;
6097 
6098 			if (ksp != NULL) {
6099 				mutex_enter(ksp->ks_lock);
6100 				KSTAT_IO_PTR(ksp)->nread +=
6101 				    res->array[i].opsize;
6102 				KSTAT_IO_PTR(ksp)->reads++;
6103 				mutex_exit(ksp->ks_lock);
6104 			}
6105 
6106 			if (exi != NULL) {
6107 				kstat_t *exi_ksp;
6108 
6109 				rw_enter(&exported_lock, RW_READER);
6110 
6111 				exi_ksp = exi->exi_kstats->rfsprocio_v4_ptr[op];
6112 				if (exi_ksp != NULL) {
6113 					mutex_enter(exi_ksp->ks_lock);
6114 					KSTAT_IO_PTR(exi_ksp)->nread +=
6115 					    res->array[i].opsize;
6116 					KSTAT_IO_PTR(exi_ksp)->reads++;
6117 					mutex_exit(exi_ksp->ks_lock);
6118 				}
6119 
6120 				rw_exit(&exported_lock);
6121 
6122 				exi_rele(exi);
6123 			}
6124 		}
6125 	}
6126 }
6127 
6128 nfsstat4
6129 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp)
6130 {
6131 	nfsstat4 e;
6132 
6133 	rfs4_dbe_lock(cp->rc_dbe);
6134 
6135 	if (cp->rc_sysidt != LM_NOSYSID) {
6136 		*sp = cp->rc_sysidt;
6137 		e = NFS4_OK;
6138 
6139 	} else if ((cp->rc_sysidt = lm_alloc_sysidt()) != LM_NOSYSID) {
6140 		*sp = cp->rc_sysidt;
6141 		e = NFS4_OK;
6142 
6143 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
6144 		    "rfs4_client_sysid: allocated 0x%x\n", *sp));
6145 	} else
6146 		e = NFS4ERR_DELAY;
6147 
6148 	rfs4_dbe_unlock(cp->rc_dbe);
6149 	return (e);
6150 }
6151 
6152 #if defined(DEBUG) && ! defined(lint)
6153 static void lock_print(char *str, int operation, struct flock64 *flk)
6154 {
6155 	char *op, *type;
6156 
6157 	switch (operation) {
6158 	case F_GETLK: op = "F_GETLK";
6159 		break;
6160 	case F_SETLK: op = "F_SETLK";
6161 		break;
6162 	case F_SETLK_NBMAND: op = "F_SETLK_NBMAND";
6163 		break;
6164 	default: op = "F_UNKNOWN";
6165 		break;
6166 	}
6167 	switch (flk->l_type) {
6168 	case F_UNLCK: type = "F_UNLCK";
6169 		break;
6170 	case F_RDLCK: type = "F_RDLCK";
6171 		break;
6172 	case F_WRLCK: type = "F_WRLCK";
6173 		break;
6174 	default: type = "F_UNKNOWN";
6175 		break;
6176 	}
6177 
6178 	ASSERT(flk->l_whence == 0);
6179 	cmn_err(CE_NOTE, "%s:  %s, type = %s, off = %llx len = %llx pid = %d",
6180 	    str, op, type, (longlong_t)flk->l_start,
6181 	    flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid);
6182 }
6183 
6184 #define	LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f)
6185 #else
6186 #define	LOCK_PRINT(d, s, t, f)
6187 #endif
6188 
6189 /*ARGSUSED*/
6190 static bool_t
6191 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs)
6192 {
6193 	return (TRUE);
6194 }
6195 
6196 /*
6197  * Look up the pathname using the vp in cs as the directory vnode.
6198  * cs->vp will be the vnode for the file on success
6199  */
6200 
6201 static nfsstat4
6202 rfs4_lookup(component4 *component, struct svc_req *req,
6203     struct compound_state *cs)
6204 {
6205 	char *nm;
6206 	uint32_t len;
6207 	nfsstat4 status;
6208 	struct sockaddr *ca;
6209 	char *name;
6210 
6211 	if (cs->vp == NULL) {
6212 		return (NFS4ERR_NOFILEHANDLE);
6213 	}
6214 	if (cs->vp->v_type != VDIR) {
6215 		return (NFS4ERR_NOTDIR);
6216 	}
6217 
6218 	status = utf8_dir_verify(component);
6219 	if (status != NFS4_OK)
6220 		return (status);
6221 
6222 	nm = utf8_to_fn(component, &len, NULL);
6223 	if (nm == NULL) {
6224 		return (NFS4ERR_INVAL);
6225 	}
6226 
6227 	if (len > MAXNAMELEN) {
6228 		kmem_free(nm, len);
6229 		return (NFS4ERR_NAMETOOLONG);
6230 	}
6231 
6232 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
6233 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
6234 	    MAXPATHLEN + 1);
6235 
6236 	if (name == NULL) {
6237 		kmem_free(nm, len);
6238 		return (NFS4ERR_INVAL);
6239 	}
6240 
6241 	status = do_rfs4_op_lookup(name, req, cs);
6242 
6243 	if (name != nm)
6244 		kmem_free(name, MAXPATHLEN + 1);
6245 
6246 	kmem_free(nm, len);
6247 
6248 	return (status);
6249 }
6250 
6251 static nfsstat4
6252 rfs4_lookupfile(component4 *component, struct svc_req *req,
6253     struct compound_state *cs, uint32_t access, change_info4 *cinfo)
6254 {
6255 	nfsstat4 status;
6256 	vnode_t *dvp = cs->vp;
6257 	vattr_t bva, ava, fva;
6258 	int error;
6259 
6260 	/* Get "before" change value */
6261 	bva.va_mask = AT_CTIME|AT_SEQ;
6262 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL);
6263 	if (error)
6264 		return (puterrno4(error));
6265 
6266 	/* rfs4_lookup may VN_RELE directory */
6267 	VN_HOLD(dvp);
6268 
6269 	status = rfs4_lookup(component, req, cs);
6270 	if (status != NFS4_OK) {
6271 		VN_RELE(dvp);
6272 		return (status);
6273 	}
6274 
6275 	/*
6276 	 * Get "after" change value, if it fails, simply return the
6277 	 * before value.
6278 	 */
6279 	ava.va_mask = AT_CTIME|AT_SEQ;
6280 	if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) {
6281 		ava.va_ctime = bva.va_ctime;
6282 		ava.va_seq = 0;
6283 	}
6284 	VN_RELE(dvp);
6285 
6286 	/*
6287 	 * Validate the file is a file
6288 	 */
6289 	fva.va_mask = AT_TYPE|AT_MODE;
6290 	error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL);
6291 	if (error)
6292 		return (puterrno4(error));
6293 
6294 	if (fva.va_type != VREG) {
6295 		if (fva.va_type == VDIR)
6296 			return (NFS4ERR_ISDIR);
6297 		if (fva.va_type == VLNK)
6298 			return (NFS4ERR_SYMLINK);
6299 		return (NFS4ERR_INVAL);
6300 	}
6301 
6302 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime);
6303 	NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
6304 
6305 	/*
6306 	 * It is undefined if VOP_LOOKUP will change va_seq, so
6307 	 * cinfo.atomic = TRUE only if we have
6308 	 * non-zero va_seq's, and they have not changed.
6309 	 */
6310 	if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq)
6311 		cinfo->atomic = TRUE;
6312 	else
6313 		cinfo->atomic = FALSE;
6314 
6315 	/* Check for mandatory locking */
6316 	cs->mandlock = MANDLOCK(cs->vp, fva.va_mode);
6317 	return (check_open_access(access, cs, req));
6318 }
6319 
6320 static nfsstat4
6321 create_vnode(vnode_t *dvp, char *nm,  vattr_t *vap, createmode4 mode,
6322     timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created)
6323 {
6324 	int error;
6325 	nfsstat4 status = NFS4_OK;
6326 	vattr_t va;
6327 
6328 tryagain:
6329 
6330 	/*
6331 	 * The file open mode used is VWRITE.  If the client needs
6332 	 * some other semantic, then it should do the access checking
6333 	 * itself.  It would have been nice to have the file open mode
6334 	 * passed as part of the arguments.
6335 	 */
6336 
6337 	*created = TRUE;
6338 	error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL);
6339 
6340 	if (error) {
6341 		*created = FALSE;
6342 
6343 		/*
6344 		 * If we got something other than file already exists
6345 		 * then just return this error.  Otherwise, we got
6346 		 * EEXIST.  If we were doing a GUARDED create, then
6347 		 * just return this error.  Otherwise, we need to
6348 		 * make sure that this wasn't a duplicate of an
6349 		 * exclusive create request.
6350 		 *
6351 		 * The assumption is made that a non-exclusive create
6352 		 * request will never return EEXIST.
6353 		 */
6354 
6355 		if (error != EEXIST || mode == GUARDED4) {
6356 			status = puterrno4(error);
6357 			return (status);
6358 		}
6359 		error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr,
6360 		    NULL, NULL, NULL);
6361 
6362 		if (error) {
6363 			/*
6364 			 * We couldn't find the file that we thought that
6365 			 * we just created.  So, we'll just try creating
6366 			 * it again.
6367 			 */
6368 			if (error == ENOENT)
6369 				goto tryagain;
6370 
6371 			status = puterrno4(error);
6372 			return (status);
6373 		}
6374 
6375 		if (mode == UNCHECKED4) {
6376 			/* existing object must be regular file */
6377 			if ((*vpp)->v_type != VREG) {
6378 				if ((*vpp)->v_type == VDIR)
6379 					status = NFS4ERR_ISDIR;
6380 				else if ((*vpp)->v_type == VLNK)
6381 					status = NFS4ERR_SYMLINK;
6382 				else
6383 					status = NFS4ERR_INVAL;
6384 				VN_RELE(*vpp);
6385 				return (status);
6386 			}
6387 
6388 			return (NFS4_OK);
6389 		}
6390 
6391 		/* Check for duplicate request */
6392 		ASSERT(mtime != 0);
6393 		va.va_mask = AT_MTIME;
6394 		error = VOP_GETATTR(*vpp, &va, 0, cr, NULL);
6395 		if (!error) {
6396 			/* We found the file */
6397 			if (va.va_mtime.tv_sec != mtime->tv_sec ||
6398 			    va.va_mtime.tv_nsec != mtime->tv_nsec) {
6399 				/* but its not our creation */
6400 				VN_RELE(*vpp);
6401 				return (NFS4ERR_EXIST);
6402 			}
6403 			*created = TRUE; /* retrans of create == created */
6404 			return (NFS4_OK);
6405 		}
6406 		VN_RELE(*vpp);
6407 		return (NFS4ERR_EXIST);
6408 	}
6409 
6410 	return (NFS4_OK);
6411 }
6412 
6413 static nfsstat4
6414 check_open_access(uint32_t access, struct compound_state *cs,
6415     struct svc_req *req)
6416 {
6417 	int error;
6418 	vnode_t *vp;
6419 	bool_t readonly;
6420 	cred_t *cr = cs->cr;
6421 
6422 	/* For now we don't allow mandatory locking as per V2/V3 */
6423 	if (cs->access == CS_ACCESS_DENIED || cs->mandlock) {
6424 		return (NFS4ERR_ACCESS);
6425 	}
6426 
6427 	vp = cs->vp;
6428 	ASSERT(cr != NULL && vp->v_type == VREG);
6429 
6430 	/*
6431 	 * If the file system is exported read only and we are trying
6432 	 * to open for write, then return NFS4ERR_ROFS
6433 	 */
6434 
6435 	readonly = rdonly4(req, cs);
6436 
6437 	if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly)
6438 		return (NFS4ERR_ROFS);
6439 
6440 	if (access & OPEN4_SHARE_ACCESS_READ) {
6441 		if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) &&
6442 		    (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) {
6443 			return (NFS4ERR_ACCESS);
6444 		}
6445 	}
6446 
6447 	if (access & OPEN4_SHARE_ACCESS_WRITE) {
6448 		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
6449 		if (error)
6450 			return (NFS4ERR_ACCESS);
6451 	}
6452 
6453 	return (NFS4_OK);
6454 }
6455 
6456 static nfsstat4
6457 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs,
6458     change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid)
6459 {
6460 	struct nfs4_svgetit_arg sarg;
6461 	struct nfs4_ntov_table ntov;
6462 
6463 	bool_t ntov_table_init = FALSE;
6464 	struct statvfs64 sb;
6465 	nfsstat4 status;
6466 	vnode_t *vp;
6467 	vattr_t bva, ava, iva, cva, *vap;
6468 	vnode_t *dvp;
6469 	timespec32_t *mtime;
6470 	char *nm = NULL;
6471 	uint_t buflen;
6472 	bool_t created;
6473 	bool_t setsize = FALSE;
6474 	len_t reqsize;
6475 	int error;
6476 	bool_t trunc;
6477 	caller_context_t ct;
6478 	component4 *component;
6479 	bslabel_t *clabel;
6480 	struct sockaddr *ca;
6481 	char *name = NULL;
6482 
6483 	sarg.sbp = &sb;
6484 	sarg.is_referral = B_FALSE;
6485 
6486 	dvp = cs->vp;
6487 
6488 	/* Check if the file system is read only */
6489 	if (rdonly4(req, cs))
6490 		return (NFS4ERR_ROFS);
6491 
6492 	/* check the label of including directory */
6493 	if (is_system_labeled()) {
6494 		ASSERT(req->rq_label != NULL);
6495 		clabel = req->rq_label;
6496 		DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *,
6497 		    "got client label from request(1)",
6498 		    struct svc_req *, req);
6499 		if (!blequal(&l_admin_low->tsl_label, clabel)) {
6500 			if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK,
6501 			    cs->exi)) {
6502 				return (NFS4ERR_ACCESS);
6503 			}
6504 		}
6505 	}
6506 
6507 	/*
6508 	 * Get the last component of path name in nm. cs will reference
6509 	 * the including directory on success.
6510 	 */
6511 	component = &args->open_claim4_u.file;
6512 	status = utf8_dir_verify(component);
6513 	if (status != NFS4_OK)
6514 		return (status);
6515 
6516 	nm = utf8_to_fn(component, &buflen, NULL);
6517 
6518 	if (nm == NULL)
6519 		return (NFS4ERR_RESOURCE);
6520 
6521 	if (buflen > MAXNAMELEN) {
6522 		kmem_free(nm, buflen);
6523 		return (NFS4ERR_NAMETOOLONG);
6524 	}
6525 
6526 	bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ;
6527 	error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL);
6528 	if (error) {
6529 		kmem_free(nm, buflen);
6530 		return (puterrno4(error));
6531 	}
6532 
6533 	if (bva.va_type != VDIR) {
6534 		kmem_free(nm, buflen);
6535 		return (NFS4ERR_NOTDIR);
6536 	}
6537 
6538 	NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime)
6539 
6540 	switch (args->mode) {
6541 	case GUARDED4:
6542 		/*FALLTHROUGH*/
6543 	case UNCHECKED4:
6544 		nfs4_ntov_table_init(&ntov);
6545 		ntov_table_init = TRUE;
6546 
6547 		*attrset = 0;
6548 		status = do_rfs4_set_attrs(attrset,
6549 		    &args->createhow4_u.createattrs,
6550 		    cs, &sarg, &ntov, NFS4ATTR_SETIT);
6551 
6552 		if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) &&
6553 		    sarg.vap->va_type != VREG) {
6554 			if (sarg.vap->va_type == VDIR)
6555 				status = NFS4ERR_ISDIR;
6556 			else if (sarg.vap->va_type == VLNK)
6557 				status = NFS4ERR_SYMLINK;
6558 			else
6559 				status = NFS4ERR_INVAL;
6560 		}
6561 
6562 		if (status != NFS4_OK) {
6563 			kmem_free(nm, buflen);
6564 			nfs4_ntov_table_free(&ntov, &sarg);
6565 			*attrset = 0;
6566 			return (status);
6567 		}
6568 
6569 		vap = sarg.vap;
6570 		vap->va_type = VREG;
6571 		vap->va_mask |= AT_TYPE;
6572 
6573 		if ((vap->va_mask & AT_MODE) == 0) {
6574 			vap->va_mask |= AT_MODE;
6575 			vap->va_mode = (mode_t)0600;
6576 		}
6577 
6578 		if (vap->va_mask & AT_SIZE) {
6579 
6580 			/* Disallow create with a non-zero size */
6581 
6582 			if ((reqsize = sarg.vap->va_size) != 0) {
6583 				kmem_free(nm, buflen);
6584 				nfs4_ntov_table_free(&ntov, &sarg);
6585 				*attrset = 0;
6586 				return (NFS4ERR_INVAL);
6587 			}
6588 			setsize = TRUE;
6589 		}
6590 		break;
6591 
6592 	case EXCLUSIVE4:
6593 		/* prohibit EXCL create of named attributes */
6594 		if (dvp->v_flag & V_XATTRDIR) {
6595 			kmem_free(nm, buflen);
6596 			*attrset = 0;
6597 			return (NFS4ERR_INVAL);
6598 		}
6599 
6600 		cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE;
6601 		cva.va_type = VREG;
6602 		/*
6603 		 * Ensure no time overflows. Assumes underlying
6604 		 * filesystem supports at least 32 bits.
6605 		 * Truncate nsec to usec resolution to allow valid
6606 		 * compares even if the underlying filesystem truncates.
6607 		 */
6608 		mtime = (timespec32_t *)&args->createhow4_u.createverf;
6609 		cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX;
6610 		cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000;
6611 		cva.va_mode = (mode_t)0;
6612 		vap = &cva;
6613 
6614 		/*
6615 		 * For EXCL create, attrset is set to the server attr
6616 		 * used to cache the client's verifier.
6617 		 */
6618 		*attrset = FATTR4_TIME_MODIFY_MASK;
6619 		break;
6620 	}
6621 
6622 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
6623 	name = nfscmd_convname(ca, cs->exi, nm, NFSCMD_CONV_INBOUND,
6624 	    MAXPATHLEN  + 1);
6625 
6626 	if (name == NULL) {
6627 		kmem_free(nm, buflen);
6628 		return (NFS4ERR_SERVERFAULT);
6629 	}
6630 
6631 	status = create_vnode(dvp, name, vap, args->mode, mtime,
6632 	    cs->cr, &vp, &created);
6633 	if (nm != name)
6634 		kmem_free(name, MAXPATHLEN + 1);
6635 	kmem_free(nm, buflen);
6636 
6637 	if (status != NFS4_OK) {
6638 		if (ntov_table_init)
6639 			nfs4_ntov_table_free(&ntov, &sarg);
6640 		*attrset = 0;
6641 		return (status);
6642 	}
6643 
6644 	trunc = (setsize && !created);
6645 
6646 	if (args->mode != EXCLUSIVE4) {
6647 		bitmap4 createmask = args->createhow4_u.createattrs.attrmask;
6648 
6649 		/*
6650 		 * True verification that object was created with correct
6651 		 * attrs is impossible.  The attrs could have been changed
6652 		 * immediately after object creation.  If attributes did
6653 		 * not verify, the only recourse for the server is to
6654 		 * destroy the object.  Maybe if some attrs (like gid)
6655 		 * are set incorrectly, the object should be destroyed;
6656 		 * however, seems bad as a default policy.  Do we really
6657 		 * want to destroy an object over one of the times not
6658 		 * verifying correctly?  For these reasons, the server
6659 		 * currently sets bits in attrset for createattrs
6660 		 * that were set; however, no verification is done.
6661 		 *
6662 		 * vmask_to_nmask accounts for vattr bits set on create
6663 		 *	[do_rfs4_set_attrs() only sets resp bits for
6664 		 *	 non-vattr/vfs bits.]
6665 		 * Mask off any bits we set by default so as not to return
6666 		 * more attrset bits than were requested in createattrs
6667 		 */
6668 		if (created) {
6669 			nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset);
6670 			*attrset &= createmask;
6671 		} else {
6672 			/*
6673 			 * We did not create the vnode (we tried but it
6674 			 * already existed).  In this case, the only createattr
6675 			 * that the spec allows the server to set is size,
6676 			 * and even then, it can only be set if it is 0.
6677 			 */
6678 			*attrset = 0;
6679 			if (trunc)
6680 				*attrset = FATTR4_SIZE_MASK;
6681 		}
6682 	}
6683 	if (ntov_table_init)
6684 		nfs4_ntov_table_free(&ntov, &sarg);
6685 
6686 	/*
6687 	 * Get the initial "after" sequence number, if it fails,
6688 	 * set to zero, time to before.
6689 	 */
6690 	iva.va_mask = AT_CTIME|AT_SEQ;
6691 	if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) {
6692 		iva.va_seq = 0;
6693 		iva.va_ctime = bva.va_ctime;
6694 	}
6695 
6696 	/*
6697 	 * create_vnode attempts to create the file exclusive,
6698 	 * if it already exists the VOP_CREATE will fail and
6699 	 * may not increase va_seq. It is atomic if
6700 	 * we haven't changed the directory, but if it has changed
6701 	 * we don't know what changed it.
6702 	 */
6703 	if (!created) {
6704 		if (bva.va_seq && iva.va_seq &&
6705 		    bva.va_seq == iva.va_seq)
6706 			cinfo->atomic = TRUE;
6707 		else
6708 			cinfo->atomic = FALSE;
6709 		NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime);
6710 	} else {
6711 		/*
6712 		 * The entry was created, we need to sync the
6713 		 * directory metadata.
6714 		 */
6715 		(void) VOP_FSYNC(dvp, 0, cs->cr, NULL);
6716 
6717 		/*
6718 		 * Get "after" change value, if it fails, simply return the
6719 		 * before value.
6720 		 */
6721 		ava.va_mask = AT_CTIME|AT_SEQ;
6722 		if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) {
6723 			ava.va_ctime = bva.va_ctime;
6724 			ava.va_seq = 0;
6725 		}
6726 
6727 		NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime);
6728 
6729 		/*
6730 		 * The cinfo->atomic = TRUE only if we have
6731 		 * non-zero va_seq's, and it has incremented by exactly one
6732 		 * during the create_vnode and it didn't
6733 		 * change during the VOP_FSYNC.
6734 		 */
6735 		if (bva.va_seq && iva.va_seq && ava.va_seq &&
6736 		    iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq)
6737 			cinfo->atomic = TRUE;
6738 		else
6739 			cinfo->atomic = FALSE;
6740 	}
6741 
6742 	/* Check for mandatory locking and that the size gets set. */
6743 	cva.va_mask = AT_MODE;
6744 	if (setsize)
6745 		cva.va_mask |= AT_SIZE;
6746 
6747 	/* Assume the worst */
6748 	cs->mandlock = TRUE;
6749 
6750 	if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) {
6751 		cs->mandlock = MANDLOCK(cs->vp, cva.va_mode);
6752 
6753 		/*
6754 		 * Truncate the file if necessary; this would be
6755 		 * the case for create over an existing file.
6756 		 */
6757 
6758 		if (trunc) {
6759 			int in_crit = 0;
6760 			rfs4_file_t *fp;
6761 			bool_t create = FALSE;
6762 
6763 			/*
6764 			 * We are writing over an existing file.
6765 			 * Check to see if we need to recall a delegation.
6766 			 */
6767 			rfs4_hold_deleg_policy();
6768 			if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) {
6769 				if (rfs4_check_delegated_byfp(FWRITE, fp,
6770 				    (reqsize == 0), FALSE, FALSE, &clientid)) {
6771 					rfs4_file_rele(fp);
6772 					rfs4_rele_deleg_policy();
6773 					VN_RELE(vp);
6774 					*attrset = 0;
6775 					return (NFS4ERR_DELAY);
6776 				}
6777 				rfs4_file_rele(fp);
6778 			}
6779 			rfs4_rele_deleg_policy();
6780 
6781 			if (nbl_need_check(vp)) {
6782 				in_crit = 1;
6783 
6784 				ASSERT(reqsize == 0);
6785 
6786 				nbl_start_crit(vp, RW_READER);
6787 				if (nbl_conflict(vp, NBL_WRITE, 0,
6788 				    cva.va_size, 0, NULL)) {
6789 					in_crit = 0;
6790 					nbl_end_crit(vp);
6791 					VN_RELE(vp);
6792 					*attrset = 0;
6793 					return (NFS4ERR_ACCESS);
6794 				}
6795 			}
6796 			ct.cc_sysid = 0;
6797 			ct.cc_pid = 0;
6798 			ct.cc_caller_id = nfs4_srv_caller_id;
6799 			ct.cc_flags = CC_DONTBLOCK;
6800 
6801 			cva.va_mask = AT_SIZE;
6802 			cva.va_size = reqsize;
6803 			(void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct);
6804 			if (in_crit)
6805 				nbl_end_crit(vp);
6806 		}
6807 	}
6808 
6809 	error = makefh4(&cs->fh, vp, cs->exi);
6810 
6811 	/*
6812 	 * Force modified data and metadata out to stable storage.
6813 	 */
6814 	(void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL);
6815 
6816 	if (error) {
6817 		VN_RELE(vp);
6818 		*attrset = 0;
6819 		return (puterrno4(error));
6820 	}
6821 
6822 	/* if parent dir is attrdir, set namedattr fh flag */
6823 	if (dvp->v_flag & V_XATTRDIR)
6824 		set_fh4_flag(&cs->fh, FH4_NAMEDATTR);
6825 
6826 	if (cs->vp)
6827 		VN_RELE(cs->vp);
6828 
6829 	cs->vp = vp;
6830 
6831 	/*
6832 	 * if we did not create the file, we will need to check
6833 	 * the access bits on the file
6834 	 */
6835 
6836 	if (!created) {
6837 		if (setsize)
6838 			args->share_access |= OPEN4_SHARE_ACCESS_WRITE;
6839 		status = check_open_access(args->share_access, cs, req);
6840 		if (status != NFS4_OK)
6841 			*attrset = 0;
6842 	}
6843 	return (status);
6844 }
6845 
6846 /*ARGSUSED*/
6847 static void
6848 rfs4_do_open(struct compound_state *cs, struct svc_req *req,
6849     rfs4_openowner_t *oo, delegreq_t deleg,
6850     uint32_t access, uint32_t deny,
6851     OPEN4res *resp, int deleg_cur)
6852 {
6853 	/* XXX Currently not using req  */
6854 	rfs4_state_t *sp;
6855 	rfs4_file_t *fp;
6856 	bool_t screate = TRUE;
6857 	bool_t fcreate = TRUE;
6858 	uint32_t open_a, share_a;
6859 	uint32_t open_d, share_d;
6860 	rfs4_deleg_state_t *dsp;
6861 	sysid_t sysid;
6862 	nfsstat4 status;
6863 	caller_context_t ct;
6864 	int fflags = 0;
6865 	int recall = 0;
6866 	int err;
6867 	int first_open;
6868 
6869 	/* get the file struct and hold a lock on it during initial open */
6870 	fp = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate);
6871 	if (fp == NULL) {
6872 		resp->status = NFS4ERR_RESOURCE;
6873 		DTRACE_PROBE1(nfss__e__do__open1, nfsstat4, resp->status);
6874 		return;
6875 	}
6876 
6877 	sp = rfs4_findstate_by_owner_file(oo, fp, &screate);
6878 	if (sp == NULL) {
6879 		resp->status = NFS4ERR_RESOURCE;
6880 		DTRACE_PROBE1(nfss__e__do__open2, nfsstat4, resp->status);
6881 		/* No need to keep any reference */
6882 		rw_exit(&fp->rf_file_rwlock);
6883 		rfs4_file_rele(fp);
6884 		return;
6885 	}
6886 
6887 	/* try to get the sysid before continuing */
6888 	if ((status = rfs4_client_sysid(oo->ro_client, &sysid)) != NFS4_OK) {
6889 		resp->status = status;
6890 		rfs4_file_rele(fp);
6891 		/* Not a fully formed open; "close" it */
6892 		if (screate == TRUE)
6893 			rfs4_state_close(sp, FALSE, FALSE, cs->cr);
6894 		rfs4_state_rele(sp);
6895 		return;
6896 	}
6897 
6898 	/* Calculate the fflags for this OPEN. */
6899 	if (access & OPEN4_SHARE_ACCESS_READ)
6900 		fflags |= FREAD;
6901 	if (access & OPEN4_SHARE_ACCESS_WRITE)
6902 		fflags |= FWRITE;
6903 
6904 	rfs4_dbe_lock(sp->rs_dbe);
6905 
6906 	/*
6907 	 * Calculate the new deny and access mode that this open is adding to
6908 	 * the file for this open owner;
6909 	 */
6910 	open_d = (deny & ~sp->rs_open_deny);
6911 	open_a = (access & ~sp->rs_open_access);
6912 
6913 	/*
6914 	 * Calculate the new share access and share deny modes that this open
6915 	 * is adding to the file for this open owner;
6916 	 */
6917 	share_a = (access & ~sp->rs_share_access);
6918 	share_d = (deny & ~sp->rs_share_deny);
6919 
6920 	first_open = (sp->rs_open_access & OPEN4_SHARE_ACCESS_BOTH) == 0;
6921 
6922 	/*
6923 	 * Check to see the client has already sent an open for this
6924 	 * open owner on this file with the same share/deny modes.
6925 	 * If so, we don't need to check for a conflict and we don't
6926 	 * need to add another shrlock.  If not, then we need to
6927 	 * check for conflicts in deny and access before checking for
6928 	 * conflicts in delegation.  We don't want to recall a
6929 	 * delegation based on an open that will eventually fail based
6930 	 * on shares modes.
6931 	 */
6932 
6933 	if (share_a || share_d) {
6934 		if ((err = rfs4_share(sp, access, deny)) != 0) {
6935 			rfs4_dbe_unlock(sp->rs_dbe);
6936 			resp->status = err;
6937 
6938 			rfs4_file_rele(fp);
6939 			/* Not a fully formed open; "close" it */
6940 			if (screate == TRUE)
6941 				rfs4_state_close(sp, FALSE, FALSE, cs->cr);
6942 			rfs4_state_rele(sp);
6943 			return;
6944 		}
6945 	}
6946 
6947 	rfs4_dbe_lock(fp->rf_dbe);
6948 
6949 	/*
6950 	 * Check to see if this file is delegated and if so, if a
6951 	 * recall needs to be done.
6952 	 */
6953 	if (rfs4_check_recall(sp, access)) {
6954 		rfs4_dbe_unlock(fp->rf_dbe);
6955 		rfs4_dbe_unlock(sp->rs_dbe);
6956 		rfs4_recall_deleg(fp, FALSE, sp->rs_owner->ro_client);
6957 		delay(NFS4_DELEGATION_CONFLICT_DELAY);
6958 		rfs4_dbe_lock(sp->rs_dbe);
6959 
6960 		/* if state closed while lock was dropped */
6961 		if (sp->rs_closed) {
6962 			if (share_a || share_d)
6963 				(void) rfs4_unshare(sp);
6964 			rfs4_dbe_unlock(sp->rs_dbe);
6965 			rfs4_file_rele(fp);
6966 			/* Not a fully formed open; "close" it */
6967 			if (screate == TRUE)
6968 				rfs4_state_close(sp, FALSE, FALSE, cs->cr);
6969 			rfs4_state_rele(sp);
6970 			resp->status = NFS4ERR_OLD_STATEID;
6971 			return;
6972 		}
6973 
6974 		rfs4_dbe_lock(fp->rf_dbe);
6975 		/* Let's see if the delegation was returned */
6976 		if (rfs4_check_recall(sp, access)) {
6977 			rfs4_dbe_unlock(fp->rf_dbe);
6978 			if (share_a || share_d)
6979 				(void) rfs4_unshare(sp);
6980 			rfs4_dbe_unlock(sp->rs_dbe);
6981 			rfs4_file_rele(fp);
6982 			rfs4_update_lease(sp->rs_owner->ro_client);
6983 
6984 			/* Not a fully formed open; "close" it */
6985 			if (screate == TRUE)
6986 				rfs4_state_close(sp, FALSE, FALSE, cs->cr);
6987 			rfs4_state_rele(sp);
6988 			resp->status = NFS4ERR_DELAY;
6989 			return;
6990 		}
6991 	}
6992 	/*
6993 	 * the share check passed and any delegation conflict has been
6994 	 * taken care of, now call vop_open.
6995 	 * if this is the first open then call vop_open with fflags.
6996 	 * if not, call vn_open_upgrade with just the upgrade flags.
6997 	 *
6998 	 * if the file has been opened already, it will have the current
6999 	 * access mode in the state struct.  if it has no share access, then
7000 	 * this is a new open.
7001 	 *
7002 	 * However, if this is open with CLAIM_DLEGATE_CUR, then don't
7003 	 * call VOP_OPEN(), just do the open upgrade.
7004 	 */
7005 	if (first_open && !deleg_cur) {
7006 		ct.cc_sysid = sysid;
7007 		ct.cc_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe);
7008 		ct.cc_caller_id = nfs4_srv_caller_id;
7009 		ct.cc_flags = CC_DONTBLOCK;
7010 		err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct);
7011 		if (err) {
7012 			rfs4_dbe_unlock(fp->rf_dbe);
7013 			if (share_a || share_d)
7014 				(void) rfs4_unshare(sp);
7015 			rfs4_dbe_unlock(sp->rs_dbe);
7016 			rfs4_file_rele(fp);
7017 
7018 			/* Not a fully formed open; "close" it */
7019 			if (screate == TRUE)
7020 				rfs4_state_close(sp, FALSE, FALSE, cs->cr);
7021 			rfs4_state_rele(sp);
7022 			/* check if a monitor detected a delegation conflict */
7023 			if (err == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK))
7024 				resp->status = NFS4ERR_DELAY;
7025 			else
7026 				resp->status = NFS4ERR_SERVERFAULT;
7027 			return;
7028 		}
7029 	} else { /* open upgrade */
7030 		/*
7031 		 * calculate the fflags for the new mode that is being added
7032 		 * by this upgrade.
7033 		 */
7034 		fflags = 0;
7035 		if (open_a & OPEN4_SHARE_ACCESS_READ)
7036 			fflags |= FREAD;
7037 		if (open_a & OPEN4_SHARE_ACCESS_WRITE)
7038 			fflags |= FWRITE;
7039 		vn_open_upgrade(cs->vp, fflags);
7040 	}
7041 	sp->rs_open_access |= access;
7042 	sp->rs_open_deny |= deny;
7043 
7044 	if (open_d & OPEN4_SHARE_DENY_READ)
7045 		fp->rf_deny_read++;
7046 	if (open_d & OPEN4_SHARE_DENY_WRITE)
7047 		fp->rf_deny_write++;
7048 	fp->rf_share_deny |= deny;
7049 
7050 	if (open_a & OPEN4_SHARE_ACCESS_READ)
7051 		fp->rf_access_read++;
7052 	if (open_a & OPEN4_SHARE_ACCESS_WRITE)
7053 		fp->rf_access_write++;
7054 	fp->rf_share_access |= access;
7055 
7056 	/*
7057 	 * Check for delegation here. if the deleg argument is not
7058 	 * DELEG_ANY, then this is a reclaim from a client and
7059 	 * we must honor the delegation requested. If necessary we can
7060 	 * set the recall flag.
7061 	 */
7062 
7063 	dsp = rfs4_grant_delegation(deleg, sp, &recall);
7064 
7065 	cs->deleg = (fp->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE);
7066 
7067 	next_stateid(&sp->rs_stateid);
7068 
7069 	resp->stateid = sp->rs_stateid.stateid;
7070 
7071 	rfs4_dbe_unlock(fp->rf_dbe);
7072 	rfs4_dbe_unlock(sp->rs_dbe);
7073 
7074 	if (dsp) {
7075 		rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall);
7076 		rfs4_deleg_state_rele(dsp);
7077 	}
7078 
7079 	rfs4_file_rele(fp);
7080 	rfs4_state_rele(sp);
7081 
7082 	resp->status = NFS4_OK;
7083 }
7084 
7085 /*ARGSUSED*/
7086 static void
7087 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req,
7088     OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
7089 {
7090 	change_info4 *cinfo = &resp->cinfo;
7091 	bitmap4 *attrset = &resp->attrset;
7092 
7093 	if (args->opentype == OPEN4_NOCREATE)
7094 		resp->status = rfs4_lookupfile(&args->open_claim4_u.file,
7095 		    req, cs, args->share_access, cinfo);
7096 	else {
7097 		/* inhibit delegation grants during exclusive create */
7098 
7099 		if (args->mode == EXCLUSIVE4)
7100 			rfs4_disable_delegation();
7101 
7102 		resp->status = rfs4_createfile(args, req, cs, cinfo, attrset,
7103 		    oo->ro_client->rc_clientid);
7104 	}
7105 
7106 	if (resp->status == NFS4_OK) {
7107 
7108 		/* cs->vp cs->fh now reference the desired file */
7109 
7110 		rfs4_do_open(cs, req, oo,
7111 		    oo->ro_need_confirm ? DELEG_NONE : DELEG_ANY,
7112 		    args->share_access, args->share_deny, resp, 0);
7113 
7114 		/*
7115 		 * If rfs4_createfile set attrset, we must
7116 		 * clear this attrset before the response is copied.
7117 		 */
7118 		if (resp->status != NFS4_OK && resp->attrset) {
7119 			resp->attrset = 0;
7120 		}
7121 	}
7122 	else
7123 		*cs->statusp = resp->status;
7124 
7125 	if (args->mode == EXCLUSIVE4)
7126 		rfs4_enable_delegation();
7127 }
7128 
7129 /*ARGSUSED*/
7130 static void
7131 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req,
7132     OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
7133 {
7134 	change_info4 *cinfo = &resp->cinfo;
7135 	vattr_t va;
7136 	vtype_t v_type = cs->vp->v_type;
7137 	int error = 0;
7138 
7139 	/* Verify that we have a regular file */
7140 	if (v_type != VREG) {
7141 		if (v_type == VDIR)
7142 			resp->status = NFS4ERR_ISDIR;
7143 		else if (v_type == VLNK)
7144 			resp->status = NFS4ERR_SYMLINK;
7145 		else
7146 			resp->status = NFS4ERR_INVAL;
7147 		return;
7148 	}
7149 
7150 	va.va_mask = AT_MODE|AT_UID;
7151 	error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL);
7152 	if (error) {
7153 		resp->status = puterrno4(error);
7154 		return;
7155 	}
7156 
7157 	cs->mandlock = MANDLOCK(cs->vp, va.va_mode);
7158 
7159 	/*
7160 	 * Check if we have access to the file, Note the the file
7161 	 * could have originally been open UNCHECKED or GUARDED
7162 	 * with mode bits that will now fail, but there is nothing
7163 	 * we can really do about that except in the case that the
7164 	 * owner of the file is the one requesting the open.
7165 	 */
7166 	if (crgetuid(cs->cr) != va.va_uid) {
7167 		resp->status = check_open_access(args->share_access, cs, req);
7168 		if (resp->status != NFS4_OK) {
7169 			return;
7170 		}
7171 	}
7172 
7173 	/*
7174 	 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero
7175 	 */
7176 	cinfo->before = 0;
7177 	cinfo->after = 0;
7178 	cinfo->atomic = FALSE;
7179 
7180 	rfs4_do_open(cs, req, oo,
7181 	    NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type),
7182 	    args->share_access, args->share_deny, resp, 0);
7183 }
7184 
7185 static void
7186 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req,
7187     OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
7188 {
7189 	int error;
7190 	nfsstat4 status;
7191 	stateid4 stateid =
7192 	    args->open_claim4_u.delegate_cur_info.delegate_stateid;
7193 	rfs4_deleg_state_t *dsp;
7194 
7195 	/*
7196 	 * Find the state info from the stateid and confirm that the
7197 	 * file is delegated.  If the state openowner is the same as
7198 	 * the supplied openowner we're done. If not, get the file
7199 	 * info from the found state info. Use that file info to
7200 	 * create the state for this lock owner. Note solaris doen't
7201 	 * really need the pathname to find the file. We may want to
7202 	 * lookup the pathname and make sure that the vp exist and
7203 	 * matches the vp in the file structure. However it is
7204 	 * possible that the pathname nolonger exists (local process
7205 	 * unlinks the file), so this may not be that useful.
7206 	 */
7207 
7208 	status = rfs4_get_deleg_state(&stateid, &dsp);
7209 	if (status != NFS4_OK) {
7210 		resp->status = status;
7211 		return;
7212 	}
7213 
7214 	ASSERT(dsp->rds_finfo->rf_dinfo.rd_dtype != OPEN_DELEGATE_NONE);
7215 
7216 	/*
7217 	 * New lock owner, create state. Since this was probably called
7218 	 * in response to a CB_RECALL we set deleg to DELEG_NONE
7219 	 */
7220 
7221 	ASSERT(cs->vp != NULL);
7222 	VN_RELE(cs->vp);
7223 	VN_HOLD(dsp->rds_finfo->rf_vp);
7224 	cs->vp = dsp->rds_finfo->rf_vp;
7225 
7226 	if (error = makefh4(&cs->fh, cs->vp, cs->exi)) {
7227 		rfs4_deleg_state_rele(dsp);
7228 		*cs->statusp = resp->status = puterrno4(error);
7229 		return;
7230 	}
7231 
7232 	/* Mark progress for delegation returns */
7233 	dsp->rds_finfo->rf_dinfo.rd_time_lastwrite = gethrestime_sec();
7234 	rfs4_deleg_state_rele(dsp);
7235 	rfs4_do_open(cs, req, oo, DELEG_NONE,
7236 	    args->share_access, args->share_deny, resp, 1);
7237 }
7238 
7239 /*ARGSUSED*/
7240 static void
7241 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req,
7242     OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp)
7243 {
7244 	/*
7245 	 * Lookup the pathname, it must already exist since this file
7246 	 * was delegated.
7247 	 *
7248 	 * Find the file and state info for this vp and open owner pair.
7249 	 *	check that they are in fact delegated.
7250 	 *	check that the state access and deny modes are the same.
7251 	 *
7252 	 * Return the delgation possibly seting the recall flag.
7253 	 */
7254 	rfs4_file_t *fp;
7255 	rfs4_state_t *sp;
7256 	bool_t create = FALSE;
7257 	bool_t dcreate = FALSE;
7258 	rfs4_deleg_state_t *dsp;
7259 	nfsace4 *ace;
7260 
7261 	/* Note we ignore oflags */
7262 	resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev,
7263 	    req, cs, args->share_access, &resp->cinfo);
7264 
7265 	if (resp->status != NFS4_OK) {
7266 		return;
7267 	}
7268 
7269 	/* get the file struct and hold a lock on it during initial open */
7270 	fp = rfs4_findfile_withlock(cs->vp, NULL, &create);
7271 	if (fp == NULL) {
7272 		resp->status = NFS4ERR_RESOURCE;
7273 		DTRACE_PROBE1(nfss__e__do_opendelprev1, nfsstat4, resp->status);
7274 		return;
7275 	}
7276 
7277 	sp = rfs4_findstate_by_owner_file(oo, fp, &create);
7278 	if (sp == NULL) {
7279 		resp->status = NFS4ERR_SERVERFAULT;
7280 		DTRACE_PROBE1(nfss__e__do_opendelprev2, nfsstat4, resp->status);
7281 		rw_exit(&fp->rf_file_rwlock);
7282 		rfs4_file_rele(fp);
7283 		return;
7284 	}
7285 
7286 	rfs4_dbe_lock(sp->rs_dbe);
7287 	rfs4_dbe_lock(fp->rf_dbe);
7288 	if (args->share_access != sp->rs_share_access ||
7289 	    args->share_deny != sp->rs_share_deny ||
7290 	    sp->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_NONE) {
7291 		NFS4_DEBUG(rfs4_debug,
7292 		    (CE_NOTE, "rfs4_do_opendelprev: state mixup"));
7293 		rfs4_dbe_unlock(fp->rf_dbe);
7294 		rfs4_dbe_unlock(sp->rs_dbe);
7295 		rfs4_file_rele(fp);
7296 		rfs4_state_rele(sp);
7297 		resp->status = NFS4ERR_SERVERFAULT;
7298 		return;
7299 	}
7300 	rfs4_dbe_unlock(fp->rf_dbe);
7301 	rfs4_dbe_unlock(sp->rs_dbe);
7302 
7303 	dsp = rfs4_finddeleg(sp, &dcreate);
7304 	if (dsp == NULL) {
7305 		rfs4_state_rele(sp);
7306 		rfs4_file_rele(fp);
7307 		resp->status = NFS4ERR_SERVERFAULT;
7308 		return;
7309 	}
7310 
7311 	next_stateid(&sp->rs_stateid);
7312 
7313 	resp->stateid = sp->rs_stateid.stateid;
7314 
7315 	resp->delegation.delegation_type = dsp->rds_dtype;
7316 
7317 	if (dsp->rds_dtype == OPEN_DELEGATE_READ) {
7318 		open_read_delegation4 *rv =
7319 		    &resp->delegation.open_delegation4_u.read;
7320 
7321 		rv->stateid = dsp->rds_delegid.stateid;
7322 		rv->recall = FALSE; /* no policy in place to set to TRUE */
7323 		ace = &rv->permissions;
7324 	} else {
7325 		open_write_delegation4 *rv =
7326 		    &resp->delegation.open_delegation4_u.write;
7327 
7328 		rv->stateid = dsp->rds_delegid.stateid;
7329 		rv->recall = FALSE;  /* no policy in place to set to TRUE */
7330 		ace = &rv->permissions;
7331 		rv->space_limit.limitby = NFS_LIMIT_SIZE;
7332 		rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX;
7333 	}
7334 
7335 	/* XXX For now */
7336 	ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE;
7337 	ace->flag = 0;
7338 	ace->access_mask = 0;
7339 	ace->who.utf8string_len = 0;
7340 	ace->who.utf8string_val = 0;
7341 
7342 	rfs4_deleg_state_rele(dsp);
7343 	rfs4_state_rele(sp);
7344 	rfs4_file_rele(fp);
7345 }
7346 
7347 typedef enum {
7348 	NFS4_CHKSEQ_OKAY = 0,
7349 	NFS4_CHKSEQ_REPLAY = 1,
7350 	NFS4_CHKSEQ_BAD = 2
7351 } rfs4_chkseq_t;
7352 
7353 /*
7354  * Generic function for sequence number checks.
7355  */
7356 static rfs4_chkseq_t
7357 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop,
7358     seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres)
7359 {
7360 	/* Same sequence ids and matching operations? */
7361 	if (seqid == rqst_seq && resop->resop == lastop->resop) {
7362 		if (copyres == TRUE) {
7363 			rfs4_free_reply(resop);
7364 			rfs4_copy_reply(resop, lastop);
7365 		}
7366 		NFS4_DEBUG(rfs4_debug, (CE_NOTE,
7367 		    "Replayed SEQID %d\n", seqid));
7368 		return (NFS4_CHKSEQ_REPLAY);
7369 	}
7370 
7371 	/* If the incoming sequence is not the next expected then it is bad */
7372 	if (rqst_seq != seqid + 1) {
7373 		if (rqst_seq == seqid) {
7374 			NFS4_DEBUG(rfs4_debug,
7375 			    (CE_NOTE, "BAD SEQID: Replayed sequence id "
7376 			    "but last op was %d current op is %d\n",
7377 			    lastop->resop, resop->resop));
7378 			return (NFS4_CHKSEQ_BAD);
7379 		}
7380 		NFS4_DEBUG(rfs4_debug,
7381 		    (CE_NOTE, "BAD SEQID: got %u expecting %u\n",
7382 		    rqst_seq, seqid));
7383 		return (NFS4_CHKSEQ_BAD);
7384 	}
7385 
7386 	/* Everything okay -- next expected */
7387 	return (NFS4_CHKSEQ_OKAY);
7388 }
7389 
7390 
7391 static rfs4_chkseq_t
7392 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop)
7393 {
7394 	rfs4_chkseq_t rc;
7395 
7396 	rfs4_dbe_lock(op->ro_dbe);
7397 	rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply, seqid, resop,
7398 	    TRUE);
7399 	rfs4_dbe_unlock(op->ro_dbe);
7400 
7401 	if (rc == NFS4_CHKSEQ_OKAY)
7402 		rfs4_update_lease(op->ro_client);
7403 
7404 	return (rc);
7405 }
7406 
7407 static rfs4_chkseq_t
7408 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op, nfs_resop4 *resop)
7409 {
7410 	rfs4_chkseq_t rc;
7411 
7412 	rfs4_dbe_lock(op->ro_dbe);
7413 	rc = rfs4_check_seqid(op->ro_open_seqid, &op->ro_reply,
7414 	    olo_seqid, resop, FALSE);
7415 	rfs4_dbe_unlock(op->ro_dbe);
7416 
7417 	return (rc);
7418 }
7419 
7420 static rfs4_chkseq_t
7421 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lsp, nfs_resop4 *resop)
7422 {
7423 	rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY;
7424 
7425 	rfs4_dbe_lock(lsp->rls_dbe);
7426 	if (!lsp->rls_skip_seqid_check)
7427 		rc = rfs4_check_seqid(lsp->rls_seqid, &lsp->rls_reply, seqid,
7428 		    resop, TRUE);
7429 	rfs4_dbe_unlock(lsp->rls_dbe);
7430 
7431 	return (rc);
7432 }
7433 
7434 static void
7435 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop,
7436     struct svc_req *req, struct compound_state *cs)
7437 {
7438 	OPEN4args *args = &argop->nfs_argop4_u.opopen;
7439 	OPEN4res *resp = &resop->nfs_resop4_u.opopen;
7440 	open_owner4 *owner = &args->owner;
7441 	open_claim_type4 claim = args->claim;
7442 	rfs4_client_t *cp;
7443 	rfs4_openowner_t *oo;
7444 	bool_t create;
7445 	bool_t replay = FALSE;
7446 	int can_reclaim;
7447 
7448 	DTRACE_NFSV4_2(op__open__start, struct compound_state *, cs,
7449 	    OPEN4args *, args);
7450 
7451 	if (cs->vp == NULL) {
7452 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7453 		goto end;
7454 	}
7455 
7456 	/*
7457 	 * Need to check clientid and lease expiration first based on
7458 	 * error ordering and incrementing sequence id.
7459 	 */
7460 	cp = rfs4_findclient_by_id(owner->clientid, FALSE);
7461 	if (cp == NULL) {
7462 		*cs->statusp = resp->status =
7463 		    rfs4_check_clientid(&owner->clientid, 0);
7464 		goto end;
7465 	}
7466 
7467 	if (rfs4_lease_expired(cp)) {
7468 		rfs4_client_close(cp);
7469 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7470 		goto end;
7471 	}
7472 	can_reclaim = cp->rc_can_reclaim;
7473 
7474 	/*
7475 	 * Find the open_owner for use from this point forward.  Take
7476 	 * care in updating the sequence id based on the type of error
7477 	 * being returned.
7478 	 */
7479 retry:
7480 	create = TRUE;
7481 	oo = rfs4_findopenowner(owner, &create, args->seqid);
7482 	if (oo == NULL) {
7483 		*cs->statusp = resp->status = NFS4ERR_RESOURCE;
7484 		rfs4_client_rele(cp);
7485 		goto end;
7486 	}
7487 
7488 	/* Hold off access to the sequence space while the open is done */
7489 	rfs4_sw_enter(&oo->ro_sw);
7490 
7491 	/*
7492 	 * If the open_owner existed before at the server, then check
7493 	 * the sequence id.
7494 	 */
7495 	if (!create && !oo->ro_postpone_confirm) {
7496 		switch (rfs4_check_open_seqid(args->seqid, oo, resop)) {
7497 		case NFS4_CHKSEQ_BAD:
7498 			if ((args->seqid > oo->ro_open_seqid) &&
7499 			    oo->ro_need_confirm) {
7500 				rfs4_free_opens(oo, TRUE, FALSE);
7501 				rfs4_sw_exit(&oo->ro_sw);
7502 				rfs4_openowner_rele(oo);
7503 				goto retry;
7504 			}
7505 			resp->status = NFS4ERR_BAD_SEQID;
7506 			goto out;
7507 		case NFS4_CHKSEQ_REPLAY: /* replay of previous request */
7508 			replay = TRUE;
7509 			goto out;
7510 		default:
7511 			break;
7512 		}
7513 
7514 		/*
7515 		 * Sequence was ok and open owner exists
7516 		 * check to see if we have yet to see an
7517 		 * open_confirm.
7518 		 */
7519 		if (oo->ro_need_confirm) {
7520 			rfs4_free_opens(oo, TRUE, FALSE);
7521 			rfs4_sw_exit(&oo->ro_sw);
7522 			rfs4_openowner_rele(oo);
7523 			goto retry;
7524 		}
7525 	}
7526 	/* Grace only applies to regular-type OPENs */
7527 	if (rfs4_clnt_in_grace(cp) &&
7528 	    (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) {
7529 		*cs->statusp = resp->status = NFS4ERR_GRACE;
7530 		goto out;
7531 	}
7532 
7533 	/*
7534 	 * If previous state at the server existed then can_reclaim
7535 	 * will be set. If not reply NFS4ERR_NO_GRACE to the
7536 	 * client.
7537 	 */
7538 	if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) {
7539 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
7540 		goto out;
7541 	}
7542 
7543 
7544 	/*
7545 	 * Reject the open if the client has missed the grace period
7546 	 */
7547 	if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) {
7548 		*cs->statusp = resp->status = NFS4ERR_NO_GRACE;
7549 		goto out;
7550 	}
7551 
7552 	/* Couple of up-front bookkeeping items */
7553 	if (oo->ro_need_confirm) {
7554 		/*
7555 		 * If this is a reclaim OPEN then we should not ask
7556 		 * for a confirmation of the open_owner per the
7557 		 * protocol specification.
7558 		 */
7559 		if (claim == CLAIM_PREVIOUS)
7560 			oo->ro_need_confirm = FALSE;
7561 		else
7562 			resp->rflags |= OPEN4_RESULT_CONFIRM;
7563 	}
7564 	resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX;
7565 
7566 	/*
7567 	 * If there is an unshared filesystem mounted on this vnode,
7568 	 * do not allow to open/create in this directory.
7569 	 */
7570 	if (vn_ismntpt(cs->vp)) {
7571 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
7572 		goto out;
7573 	}
7574 
7575 	/*
7576 	 * access must READ, WRITE, or BOTH.  No access is invalid.
7577 	 * deny can be READ, WRITE, BOTH, or NONE.
7578 	 * bits not defined for access/deny are invalid.
7579 	 */
7580 	if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) ||
7581 	    (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) ||
7582 	    (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) {
7583 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7584 		goto out;
7585 	}
7586 
7587 
7588 	/*
7589 	 * make sure attrset is zero before response is built.
7590 	 */
7591 	resp->attrset = 0;
7592 
7593 	switch (claim) {
7594 	case CLAIM_NULL:
7595 		rfs4_do_opennull(cs, req, args, oo, resp);
7596 		break;
7597 	case CLAIM_PREVIOUS:
7598 		rfs4_do_openprev(cs, req, args, oo, resp);
7599 		break;
7600 	case CLAIM_DELEGATE_CUR:
7601 		rfs4_do_opendelcur(cs, req, args, oo, resp);
7602 		break;
7603 	case CLAIM_DELEGATE_PREV:
7604 		rfs4_do_opendelprev(cs, req, args, oo, resp);
7605 		break;
7606 	default:
7607 		resp->status = NFS4ERR_INVAL;
7608 		break;
7609 	}
7610 
7611 out:
7612 	rfs4_client_rele(cp);
7613 
7614 	/* Catch sequence id handling here to make it a little easier */
7615 	switch (resp->status) {
7616 	case NFS4ERR_BADXDR:
7617 	case NFS4ERR_BAD_SEQID:
7618 	case NFS4ERR_BAD_STATEID:
7619 	case NFS4ERR_NOFILEHANDLE:
7620 	case NFS4ERR_RESOURCE:
7621 	case NFS4ERR_STALE_CLIENTID:
7622 	case NFS4ERR_STALE_STATEID:
7623 		/*
7624 		 * The protocol states that if any of these errors are
7625 		 * being returned, the sequence id should not be
7626 		 * incremented.  Any other return requires an
7627 		 * increment.
7628 		 */
7629 		break;
7630 	default:
7631 		/* Always update the lease in this case */
7632 		rfs4_update_lease(oo->ro_client);
7633 
7634 		/* Regular response - copy the result */
7635 		if (!replay)
7636 			rfs4_update_open_resp(oo, resop, &cs->fh);
7637 
7638 		/*
7639 		 * REPLAY case: Only if the previous response was OK
7640 		 * do we copy the filehandle.  If not OK, no
7641 		 * filehandle to copy.
7642 		 */
7643 		if (replay == TRUE &&
7644 		    resp->status == NFS4_OK &&
7645 		    oo->ro_reply_fh.nfs_fh4_val) {
7646 			/*
7647 			 * If this is a replay, we must restore the
7648 			 * current filehandle/vp to that of what was
7649 			 * returned originally.  Try our best to do
7650 			 * it.
7651 			 */
7652 			nfs_fh4_fmt_t *fh_fmtp =
7653 			    (nfs_fh4_fmt_t *)oo->ro_reply_fh.nfs_fh4_val;
7654 
7655 			cs->exi = checkexport4(&fh_fmtp->fh4_fsid,
7656 			    (fid_t *)&fh_fmtp->fh4_xlen, NULL);
7657 
7658 			if (cs->exi == NULL) {
7659 				resp->status = NFS4ERR_STALE;
7660 				goto finish;
7661 			}
7662 
7663 			VN_RELE(cs->vp);
7664 
7665 			cs->vp = nfs4_fhtovp(&oo->ro_reply_fh, cs->exi,
7666 			    &resp->status);
7667 
7668 			if (cs->vp == NULL)
7669 				goto finish;
7670 
7671 			nfs_fh4_copy(&oo->ro_reply_fh, &cs->fh);
7672 		}
7673 
7674 		/*
7675 		 * If this was a replay, no need to update the
7676 		 * sequence id. If the open_owner was not created on
7677 		 * this pass, then update.  The first use of an
7678 		 * open_owner will not bump the sequence id.
7679 		 */
7680 		if (replay == FALSE && !create)
7681 			rfs4_update_open_sequence(oo);
7682 		/*
7683 		 * If the client is receiving an error and the
7684 		 * open_owner needs to be confirmed, there is no way
7685 		 * to notify the client of this fact ignoring the fact
7686 		 * that the server has no method of returning a
7687 		 * stateid to confirm.  Therefore, the server needs to
7688 		 * mark this open_owner in a way as to avoid the
7689 		 * sequence id checking the next time the client uses
7690 		 * this open_owner.
7691 		 */
7692 		if (resp->status != NFS4_OK && oo->ro_need_confirm)
7693 			oo->ro_postpone_confirm = TRUE;
7694 		/*
7695 		 * If OK response then clear the postpone flag and
7696 		 * reset the sequence id to keep in sync with the
7697 		 * client.
7698 		 */
7699 		if (resp->status == NFS4_OK && oo->ro_postpone_confirm) {
7700 			oo->ro_postpone_confirm = FALSE;
7701 			oo->ro_open_seqid = args->seqid;
7702 		}
7703 		break;
7704 	}
7705 
7706 finish:
7707 	*cs->statusp = resp->status;
7708 
7709 	rfs4_sw_exit(&oo->ro_sw);
7710 	rfs4_openowner_rele(oo);
7711 
7712 end:
7713 	DTRACE_NFSV4_2(op__open__done, struct compound_state *, cs,
7714 	    OPEN4res *, resp);
7715 }
7716 
7717 /*ARGSUSED*/
7718 void
7719 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
7720     struct svc_req *req, struct compound_state *cs)
7721 {
7722 	OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm;
7723 	OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm;
7724 	rfs4_state_t *sp;
7725 	nfsstat4 status;
7726 
7727 	DTRACE_NFSV4_2(op__open__confirm__start, struct compound_state *, cs,
7728 	    OPEN_CONFIRM4args *, args);
7729 
7730 	if (cs->vp == NULL) {
7731 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7732 		goto out;
7733 	}
7734 
7735 	if (cs->vp->v_type != VREG) {
7736 		*cs->statusp = resp->status =
7737 		    cs->vp->v_type == VDIR ? NFS4ERR_ISDIR : NFS4ERR_INVAL;
7738 		return;
7739 	}
7740 
7741 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
7742 	if (status != NFS4_OK) {
7743 		*cs->statusp = resp->status = status;
7744 		goto out;
7745 	}
7746 
7747 	/* Ensure specified filehandle matches */
7748 	if (cs->vp != sp->rs_finfo->rf_vp) {
7749 		rfs4_state_rele(sp);
7750 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7751 		goto out;
7752 	}
7753 
7754 	/* hold off other access to open_owner while we tinker */
7755 	rfs4_sw_enter(&sp->rs_owner->ro_sw);
7756 
7757 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7758 	case NFS4_CHECK_STATEID_OKAY:
7759 		if (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
7760 		    resop) != 0) {
7761 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7762 			break;
7763 		}
7764 		/*
7765 		 * If it is the appropriate stateid and determined to
7766 		 * be "OKAY" then this means that the stateid does not
7767 		 * need to be confirmed and the client is in error for
7768 		 * sending an OPEN_CONFIRM.
7769 		 */
7770 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7771 		break;
7772 	case NFS4_CHECK_STATEID_OLD:
7773 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7774 		break;
7775 	case NFS4_CHECK_STATEID_BAD:
7776 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7777 		break;
7778 	case NFS4_CHECK_STATEID_EXPIRED:
7779 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7780 		break;
7781 	case NFS4_CHECK_STATEID_CLOSED:
7782 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7783 		break;
7784 	case NFS4_CHECK_STATEID_REPLAY:
7785 		switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
7786 		    resop)) {
7787 		case NFS4_CHKSEQ_OKAY:
7788 			/*
7789 			 * This is replayed stateid; if seqid matches
7790 			 * next expected, then client is using wrong seqid.
7791 			 */
7792 			/* fall through */
7793 		case NFS4_CHKSEQ_BAD:
7794 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7795 			break;
7796 		case NFS4_CHKSEQ_REPLAY:
7797 			/*
7798 			 * Note this case is the duplicate case so
7799 			 * resp->status is already set.
7800 			 */
7801 			*cs->statusp = resp->status;
7802 			rfs4_update_lease(sp->rs_owner->ro_client);
7803 			break;
7804 		}
7805 		break;
7806 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7807 		if (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
7808 		    resop) != NFS4_CHKSEQ_OKAY) {
7809 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7810 			break;
7811 		}
7812 		*cs->statusp = resp->status = NFS4_OK;
7813 
7814 		next_stateid(&sp->rs_stateid);
7815 		resp->open_stateid = sp->rs_stateid.stateid;
7816 		sp->rs_owner->ro_need_confirm = FALSE;
7817 		rfs4_update_lease(sp->rs_owner->ro_client);
7818 		rfs4_update_open_sequence(sp->rs_owner);
7819 		rfs4_update_open_resp(sp->rs_owner, resop, NULL);
7820 		break;
7821 	default:
7822 		ASSERT(FALSE);
7823 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7824 		break;
7825 	}
7826 	rfs4_sw_exit(&sp->rs_owner->ro_sw);
7827 	rfs4_state_rele(sp);
7828 
7829 out:
7830 	DTRACE_NFSV4_2(op__open__confirm__done, struct compound_state *, cs,
7831 	    OPEN_CONFIRM4res *, resp);
7832 }
7833 
7834 /*ARGSUSED*/
7835 void
7836 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop,
7837     struct svc_req *req, struct compound_state *cs)
7838 {
7839 	OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade;
7840 	OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade;
7841 	uint32_t access = args->share_access;
7842 	uint32_t deny = args->share_deny;
7843 	nfsstat4 status;
7844 	rfs4_state_t *sp;
7845 	rfs4_file_t *fp;
7846 	int fflags = 0;
7847 
7848 	DTRACE_NFSV4_2(op__open__downgrade__start, struct compound_state *, cs,
7849 	    OPEN_DOWNGRADE4args *, args);
7850 
7851 	if (cs->vp == NULL) {
7852 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
7853 		goto out;
7854 	}
7855 
7856 	if (cs->vp->v_type != VREG) {
7857 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7858 		return;
7859 	}
7860 
7861 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID);
7862 	if (status != NFS4_OK) {
7863 		*cs->statusp = resp->status = status;
7864 		goto out;
7865 	}
7866 
7867 	/* Ensure specified filehandle matches */
7868 	if (cs->vp != sp->rs_finfo->rf_vp) {
7869 		rfs4_state_rele(sp);
7870 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7871 		goto out;
7872 	}
7873 
7874 	/* hold off other access to open_owner while we tinker */
7875 	rfs4_sw_enter(&sp->rs_owner->ro_sw);
7876 
7877 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
7878 	case NFS4_CHECK_STATEID_OKAY:
7879 		if (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
7880 		    resop) != NFS4_CHKSEQ_OKAY) {
7881 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7882 			goto end;
7883 		}
7884 		break;
7885 	case NFS4_CHECK_STATEID_OLD:
7886 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7887 		goto end;
7888 	case NFS4_CHECK_STATEID_BAD:
7889 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7890 		goto end;
7891 	case NFS4_CHECK_STATEID_EXPIRED:
7892 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
7893 		goto end;
7894 	case NFS4_CHECK_STATEID_CLOSED:
7895 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
7896 		goto end;
7897 	case NFS4_CHECK_STATEID_UNCONFIRMED:
7898 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
7899 		goto end;
7900 	case NFS4_CHECK_STATEID_REPLAY:
7901 		/* Check the sequence id for the open owner */
7902 		switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
7903 		    resop)) {
7904 		case NFS4_CHKSEQ_OKAY:
7905 			/*
7906 			 * This is replayed stateid; if seqid matches
7907 			 * next expected, then client is using wrong seqid.
7908 			 */
7909 			/* fall through */
7910 		case NFS4_CHKSEQ_BAD:
7911 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
7912 			goto end;
7913 		case NFS4_CHKSEQ_REPLAY:
7914 			/*
7915 			 * Note this case is the duplicate case so
7916 			 * resp->status is already set.
7917 			 */
7918 			*cs->statusp = resp->status;
7919 			rfs4_update_lease(sp->rs_owner->ro_client);
7920 			goto end;
7921 		}
7922 		break;
7923 	default:
7924 		ASSERT(FALSE);
7925 		break;
7926 	}
7927 
7928 	rfs4_dbe_lock(sp->rs_dbe);
7929 	/*
7930 	 * Check that the new access modes and deny modes are valid.
7931 	 * Check that no invalid bits are set.
7932 	 */
7933 	if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) ||
7934 	    (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) {
7935 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7936 		rfs4_update_open_sequence(sp->rs_owner);
7937 		rfs4_dbe_unlock(sp->rs_dbe);
7938 		goto end;
7939 	}
7940 
7941 	/*
7942 	 * The new modes must be a subset of the current modes and
7943 	 * the access must specify at least one mode. To test that
7944 	 * the new mode is a subset of the current modes we bitwise
7945 	 * AND them together and check that the result equals the new
7946 	 * mode. For example:
7947 	 * New mode, access == R and current mode, sp->rs_open_access  == RW
7948 	 * access & sp->rs_open_access == R == access, so the new access mode
7949 	 * is valid. Consider access == RW, sp->rs_open_access = R
7950 	 * access & sp->rs_open_access == R != access, so the new access mode
7951 	 * is invalid.
7952 	 */
7953 	if ((access & sp->rs_open_access) != access ||
7954 	    (deny & sp->rs_open_deny) != deny ||
7955 	    (access &
7956 	    (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) {
7957 		*cs->statusp = resp->status = NFS4ERR_INVAL;
7958 		rfs4_update_open_sequence(sp->rs_owner);
7959 		rfs4_dbe_unlock(sp->rs_dbe);
7960 		goto end;
7961 	}
7962 
7963 	/*
7964 	 * Release any share locks associated with this stateID.
7965 	 * Strictly speaking, this violates the spec because the
7966 	 * spec effectively requires that open downgrade be atomic.
7967 	 * At present, fs_shrlock does not have this capability.
7968 	 */
7969 	(void) rfs4_unshare(sp);
7970 
7971 	status = rfs4_share(sp, access, deny);
7972 	if (status != NFS4_OK) {
7973 		*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
7974 		rfs4_update_open_sequence(sp->rs_owner);
7975 		rfs4_dbe_unlock(sp->rs_dbe);
7976 		goto end;
7977 	}
7978 
7979 	fp = sp->rs_finfo;
7980 	rfs4_dbe_lock(fp->rf_dbe);
7981 
7982 	/*
7983 	 * If the current mode has deny read and the new mode
7984 	 * does not, decrement the number of deny read mode bits
7985 	 * and if it goes to zero turn off the deny read bit
7986 	 * on the file.
7987 	 */
7988 	if ((sp->rs_open_deny & OPEN4_SHARE_DENY_READ) &&
7989 	    (deny & OPEN4_SHARE_DENY_READ) == 0) {
7990 		fp->rf_deny_read--;
7991 		if (fp->rf_deny_read == 0)
7992 			fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ;
7993 	}
7994 
7995 	/*
7996 	 * If the current mode has deny write and the new mode
7997 	 * does not, decrement the number of deny write mode bits
7998 	 * and if it goes to zero turn off the deny write bit
7999 	 * on the file.
8000 	 */
8001 	if ((sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) &&
8002 	    (deny & OPEN4_SHARE_DENY_WRITE) == 0) {
8003 		fp->rf_deny_write--;
8004 		if (fp->rf_deny_write == 0)
8005 			fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE;
8006 	}
8007 
8008 	/*
8009 	 * If the current mode has access read and the new mode
8010 	 * does not, decrement the number of access read mode bits
8011 	 * and if it goes to zero turn off the access read bit
8012 	 * on the file.  set fflags to FREAD for the call to
8013 	 * vn_open_downgrade().
8014 	 */
8015 	if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) &&
8016 	    (access & OPEN4_SHARE_ACCESS_READ) == 0) {
8017 		fp->rf_access_read--;
8018 		if (fp->rf_access_read == 0)
8019 			fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ;
8020 		fflags |= FREAD;
8021 	}
8022 
8023 	/*
8024 	 * If the current mode has access write and the new mode
8025 	 * does not, decrement the number of access write mode bits
8026 	 * and if it goes to zero turn off the access write bit
8027 	 * on the file.  set fflags to FWRITE for the call to
8028 	 * vn_open_downgrade().
8029 	 */
8030 	if ((sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) &&
8031 	    (access & OPEN4_SHARE_ACCESS_WRITE) == 0) {
8032 		fp->rf_access_write--;
8033 		if (fp->rf_access_write == 0)
8034 			fp->rf_share_deny &= ~OPEN4_SHARE_ACCESS_WRITE;
8035 		fflags |= FWRITE;
8036 	}
8037 
8038 	/* Check that the file is still accessible */
8039 	ASSERT(fp->rf_share_access);
8040 
8041 	rfs4_dbe_unlock(fp->rf_dbe);
8042 
8043 	/* now set the new open access and deny modes */
8044 	sp->rs_open_access = access;
8045 	sp->rs_open_deny = deny;
8046 
8047 	/*
8048 	 * we successfully downgraded the share lock, now we need to downgrade
8049 	 * the open. it is possible that the downgrade was only for a deny
8050 	 * mode and we have nothing else to do.
8051 	 */
8052 	if ((fflags & (FREAD|FWRITE)) != 0)
8053 		vn_open_downgrade(cs->vp, fflags);
8054 
8055 	/* Update the stateid */
8056 	next_stateid(&sp->rs_stateid);
8057 	resp->open_stateid = sp->rs_stateid.stateid;
8058 
8059 	rfs4_dbe_unlock(sp->rs_dbe);
8060 
8061 	*cs->statusp = resp->status = NFS4_OK;
8062 	/* Update the lease */
8063 	rfs4_update_lease(sp->rs_owner->ro_client);
8064 	/* And the sequence */
8065 	rfs4_update_open_sequence(sp->rs_owner);
8066 	rfs4_update_open_resp(sp->rs_owner, resop, NULL);
8067 
8068 end:
8069 	rfs4_sw_exit(&sp->rs_owner->ro_sw);
8070 	rfs4_state_rele(sp);
8071 out:
8072 	DTRACE_NFSV4_2(op__open__downgrade__done, struct compound_state *, cs,
8073 	    OPEN_DOWNGRADE4res *, resp);
8074 }
8075 
8076 static void *
8077 memstr(const void *s1, const char *s2, size_t n)
8078 {
8079 	size_t l = strlen(s2);
8080 	char *p = (char *)s1;
8081 
8082 	while (n >= l) {
8083 		if (bcmp(p, s2, l) == 0)
8084 			return (p);
8085 		p++;
8086 		n--;
8087 	}
8088 
8089 	return (NULL);
8090 }
8091 
8092 /*
8093  * The logic behind this function is detailed in the NFSv4 RFC in the
8094  * SETCLIENTID operation description under IMPLEMENTATION.  Refer to
8095  * that section for explicit guidance to server behavior for
8096  * SETCLIENTID.
8097  */
8098 void
8099 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop,
8100     struct svc_req *req, struct compound_state *cs)
8101 {
8102 	SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid;
8103 	SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid;
8104 	rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed;
8105 	rfs4_clntip_t *ci;
8106 	bool_t create;
8107 	char *addr, *netid;
8108 	int len;
8109 
8110 	DTRACE_NFSV4_2(op__setclientid__start, struct compound_state *, cs,
8111 	    SETCLIENTID4args *, args);
8112 retry:
8113 	newcp = cp_confirmed = cp_unconfirmed = NULL;
8114 
8115 	/*
8116 	 * Save the caller's IP address
8117 	 */
8118 	args->client.cl_addr =
8119 	    (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
8120 
8121 	/*
8122 	 * Record if it is a Solaris client that cannot handle referrals.
8123 	 */
8124 	if (memstr(args->client.id_val, "Solaris", args->client.id_len) &&
8125 	    !memstr(args->client.id_val, "+referrals", args->client.id_len)) {
8126 		/* Add a "yes, it's downrev" record */
8127 		create = TRUE;
8128 		ci = rfs4_find_clntip(args->client.cl_addr, &create);
8129 		ASSERT(ci != NULL);
8130 		rfs4_dbe_rele(ci->ri_dbe);
8131 	} else {
8132 		/* Remove any previous record */
8133 		rfs4_invalidate_clntip(args->client.cl_addr);
8134 	}
8135 
8136 	/*
8137 	 * In search of an EXISTING client matching the incoming
8138 	 * request to establish a new client identifier at the server
8139 	 */
8140 	create = TRUE;
8141 	cp = rfs4_findclient(&args->client, &create, NULL);
8142 
8143 	/* Should never happen */
8144 	ASSERT(cp != NULL);
8145 
8146 	if (cp == NULL) {
8147 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
8148 		goto out;
8149 	}
8150 
8151 	/*
8152 	 * Easiest case. Client identifier is newly created and is
8153 	 * unconfirmed.  Also note that for this case, no other
8154 	 * entries exist for the client identifier.  Nothing else to
8155 	 * check.  Just setup the response and respond.
8156 	 */
8157 	if (create) {
8158 		*cs->statusp = res->status = NFS4_OK;
8159 		res->SETCLIENTID4res_u.resok4.clientid = cp->rc_clientid;
8160 		res->SETCLIENTID4res_u.resok4.setclientid_confirm =
8161 		    cp->rc_confirm_verf;
8162 		/* Setup callback information; CB_NULL confirmation later */
8163 		rfs4_client_setcb(cp, &args->callback, args->callback_ident);
8164 
8165 		rfs4_client_rele(cp);
8166 		goto out;
8167 	}
8168 
8169 	/*
8170 	 * An existing, confirmed client may exist but it may not have
8171 	 * been active for at least one lease period.  If so, then
8172 	 * "close" the client and create a new client identifier
8173 	 */
8174 	if (rfs4_lease_expired(cp)) {
8175 		rfs4_client_close(cp);
8176 		goto retry;
8177 	}
8178 
8179 	if (cp->rc_need_confirm == TRUE)
8180 		cp_unconfirmed = cp;
8181 	else
8182 		cp_confirmed = cp;
8183 
8184 	cp = NULL;
8185 
8186 	/*
8187 	 * We have a confirmed client, now check for an
8188 	 * unconfimred entry
8189 	 */
8190 	if (cp_confirmed) {
8191 		/* If creds don't match then client identifier is inuse */
8192 		if (!creds_ok(cp_confirmed->rc_cr_set, req, cs)) {
8193 			rfs4_cbinfo_t *cbp;
8194 			/*
8195 			 * Some one else has established this client
8196 			 * id. Try and say * who they are. We will use
8197 			 * the call back address supplied by * the
8198 			 * first client.
8199 			 */
8200 			*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
8201 
8202 			addr = netid = NULL;
8203 
8204 			cbp = &cp_confirmed->rc_cbinfo;
8205 			if (cbp->cb_callback.cb_location.r_addr &&
8206 			    cbp->cb_callback.cb_location.r_netid) {
8207 				cb_client4 *cbcp = &cbp->cb_callback;
8208 
8209 				len = strlen(cbcp->cb_location.r_addr)+1;
8210 				addr = kmem_alloc(len, KM_SLEEP);
8211 				bcopy(cbcp->cb_location.r_addr, addr, len);
8212 				len = strlen(cbcp->cb_location.r_netid)+1;
8213 				netid = kmem_alloc(len, KM_SLEEP);
8214 				bcopy(cbcp->cb_location.r_netid, netid, len);
8215 			}
8216 
8217 			res->SETCLIENTID4res_u.client_using.r_addr = addr;
8218 			res->SETCLIENTID4res_u.client_using.r_netid = netid;
8219 
8220 			rfs4_client_rele(cp_confirmed);
8221 		}
8222 
8223 		/*
8224 		 * Confirmed, creds match, and verifier matches; must
8225 		 * be an update of the callback info
8226 		 */
8227 		if (cp_confirmed->rc_nfs_client.verifier ==
8228 		    args->client.verifier) {
8229 			/* Setup callback information */
8230 			rfs4_client_setcb(cp_confirmed, &args->callback,
8231 			    args->callback_ident);
8232 
8233 			/* everything okay -- move ahead */
8234 			*cs->statusp = res->status = NFS4_OK;
8235 			res->SETCLIENTID4res_u.resok4.clientid =
8236 			    cp_confirmed->rc_clientid;
8237 
8238 			/* update the confirm_verifier and return it */
8239 			rfs4_client_scv_next(cp_confirmed);
8240 			res->SETCLIENTID4res_u.resok4.setclientid_confirm =
8241 			    cp_confirmed->rc_confirm_verf;
8242 
8243 			rfs4_client_rele(cp_confirmed);
8244 			goto out;
8245 		}
8246 
8247 		/*
8248 		 * Creds match but the verifier doesn't.  Must search
8249 		 * for an unconfirmed client that would be replaced by
8250 		 * this request.
8251 		 */
8252 		create = FALSE;
8253 		cp_unconfirmed = rfs4_findclient(&args->client, &create,
8254 		    cp_confirmed);
8255 	}
8256 
8257 	/*
8258 	 * At this point, we have taken care of the brand new client
8259 	 * struct, INUSE case, update of an existing, and confirmed
8260 	 * client struct.
8261 	 */
8262 
8263 	/*
8264 	 * check to see if things have changed while we originally
8265 	 * picked up the client struct.  If they have, then return and
8266 	 * retry the processing of this SETCLIENTID request.
8267 	 */
8268 	if (cp_unconfirmed) {
8269 		rfs4_dbe_lock(cp_unconfirmed->rc_dbe);
8270 		if (!cp_unconfirmed->rc_need_confirm) {
8271 			rfs4_dbe_unlock(cp_unconfirmed->rc_dbe);
8272 			rfs4_client_rele(cp_unconfirmed);
8273 			if (cp_confirmed)
8274 				rfs4_client_rele(cp_confirmed);
8275 			goto retry;
8276 		}
8277 		/* do away with the old unconfirmed one */
8278 		rfs4_dbe_invalidate(cp_unconfirmed->rc_dbe);
8279 		rfs4_dbe_unlock(cp_unconfirmed->rc_dbe);
8280 		rfs4_client_rele(cp_unconfirmed);
8281 		cp_unconfirmed = NULL;
8282 	}
8283 
8284 	/*
8285 	 * This search will temporarily hide the confirmed client
8286 	 * struct while a new client struct is created as the
8287 	 * unconfirmed one.
8288 	 */
8289 	create = TRUE;
8290 	newcp = rfs4_findclient(&args->client, &create, cp_confirmed);
8291 
8292 	ASSERT(newcp != NULL);
8293 
8294 	if (newcp == NULL) {
8295 		*cs->statusp = res->status = NFS4ERR_SERVERFAULT;
8296 		rfs4_client_rele(cp_confirmed);
8297 		goto out;
8298 	}
8299 
8300 	/*
8301 	 * If one was not created, then a similar request must be in
8302 	 * process so release and start over with this one
8303 	 */
8304 	if (create != TRUE) {
8305 		rfs4_client_rele(newcp);
8306 		if (cp_confirmed)
8307 			rfs4_client_rele(cp_confirmed);
8308 		goto retry;
8309 	}
8310 
8311 	*cs->statusp = res->status = NFS4_OK;
8312 	res->SETCLIENTID4res_u.resok4.clientid = newcp->rc_clientid;
8313 	res->SETCLIENTID4res_u.resok4.setclientid_confirm =
8314 	    newcp->rc_confirm_verf;
8315 	/* Setup callback information; CB_NULL confirmation later */
8316 	rfs4_client_setcb(newcp, &args->callback, args->callback_ident);
8317 
8318 	newcp->rc_cp_confirmed = cp_confirmed;
8319 
8320 	rfs4_client_rele(newcp);
8321 
8322 out:
8323 	DTRACE_NFSV4_2(op__setclientid__done, struct compound_state *, cs,
8324 	    SETCLIENTID4res *, res);
8325 }
8326 
8327 /*ARGSUSED*/
8328 void
8329 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop,
8330     struct svc_req *req, struct compound_state *cs)
8331 {
8332 	SETCLIENTID_CONFIRM4args *args =
8333 	    &argop->nfs_argop4_u.opsetclientid_confirm;
8334 	SETCLIENTID_CONFIRM4res *res =
8335 	    &resop->nfs_resop4_u.opsetclientid_confirm;
8336 	rfs4_client_t *cp, *cptoclose = NULL;
8337 
8338 	DTRACE_NFSV4_2(op__setclientid__confirm__start,
8339 	    struct compound_state *, cs,
8340 	    SETCLIENTID_CONFIRM4args *, args);
8341 
8342 	*cs->statusp = res->status = NFS4_OK;
8343 
8344 	cp = rfs4_findclient_by_id(args->clientid, TRUE);
8345 
8346 	if (cp == NULL) {
8347 		*cs->statusp = res->status =
8348 		    rfs4_check_clientid(&args->clientid, 1);
8349 		goto out;
8350 	}
8351 
8352 	if (!creds_ok(cp, req, cs)) {
8353 		*cs->statusp = res->status = NFS4ERR_CLID_INUSE;
8354 		rfs4_client_rele(cp);
8355 		goto out;
8356 	}
8357 
8358 	/* If the verifier doesn't match, the record doesn't match */
8359 	if (cp->rc_confirm_verf != args->setclientid_confirm) {
8360 		*cs->statusp = res->status = NFS4ERR_STALE_CLIENTID;
8361 		rfs4_client_rele(cp);
8362 		goto out;
8363 	}
8364 
8365 	rfs4_dbe_lock(cp->rc_dbe);
8366 	cp->rc_need_confirm = FALSE;
8367 	if (cp->rc_cp_confirmed) {
8368 		cptoclose = cp->rc_cp_confirmed;
8369 		cptoclose->rc_ss_remove = 1;
8370 		cp->rc_cp_confirmed = NULL;
8371 	}
8372 
8373 	/*
8374 	 * Update the client's associated server instance, if it's changed
8375 	 * since the client was created.
8376 	 */
8377 	if (rfs4_servinst(cp) != rfs4_cur_servinst)
8378 		rfs4_servinst_assign(cp, rfs4_cur_servinst);
8379 
8380 	/*
8381 	 * Record clientid in stable storage.
8382 	 * Must be done after server instance has been assigned.
8383 	 */
8384 	rfs4_ss_clid(cp);
8385 
8386 	rfs4_dbe_unlock(cp->rc_dbe);
8387 
8388 	if (cptoclose)
8389 		/* don't need to rele, client_close does it */
8390 		rfs4_client_close(cptoclose);
8391 
8392 	/* If needed, initiate CB_NULL call for callback path */
8393 	rfs4_deleg_cb_check(cp);
8394 	rfs4_update_lease(cp);
8395 
8396 	/*
8397 	 * Check to see if client can perform reclaims
8398 	 */
8399 	rfs4_ss_chkclid(cp);
8400 
8401 	rfs4_client_rele(cp);
8402 
8403 out:
8404 	DTRACE_NFSV4_2(op__setclientid__confirm__done,
8405 	    struct compound_state *, cs,
8406 	    SETCLIENTID_CONFIRM4 *, res);
8407 }
8408 
8409 
8410 /*ARGSUSED*/
8411 void
8412 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop,
8413     struct svc_req *req, struct compound_state *cs)
8414 {
8415 	CLOSE4args *args = &argop->nfs_argop4_u.opclose;
8416 	CLOSE4res *resp = &resop->nfs_resop4_u.opclose;
8417 	rfs4_state_t *sp;
8418 	nfsstat4 status;
8419 
8420 	DTRACE_NFSV4_2(op__close__start, struct compound_state *, cs,
8421 	    CLOSE4args *, args);
8422 
8423 	if (cs->vp == NULL) {
8424 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8425 		goto out;
8426 	}
8427 
8428 	status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID);
8429 	if (status != NFS4_OK) {
8430 		*cs->statusp = resp->status = status;
8431 		goto out;
8432 	}
8433 
8434 	/* Ensure specified filehandle matches */
8435 	if (cs->vp != sp->rs_finfo->rf_vp) {
8436 		rfs4_state_rele(sp);
8437 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8438 		goto out;
8439 	}
8440 
8441 	/* hold off other access to open_owner while we tinker */
8442 	rfs4_sw_enter(&sp->rs_owner->ro_sw);
8443 
8444 	switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) {
8445 	case NFS4_CHECK_STATEID_OKAY:
8446 		if (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
8447 		    resop) != NFS4_CHKSEQ_OKAY) {
8448 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8449 			goto end;
8450 		}
8451 		break;
8452 	case NFS4_CHECK_STATEID_OLD:
8453 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8454 		goto end;
8455 	case NFS4_CHECK_STATEID_BAD:
8456 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8457 		goto end;
8458 	case NFS4_CHECK_STATEID_EXPIRED:
8459 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
8460 		goto end;
8461 	case NFS4_CHECK_STATEID_CLOSED:
8462 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
8463 		goto end;
8464 	case NFS4_CHECK_STATEID_UNCONFIRMED:
8465 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8466 		goto end;
8467 	case NFS4_CHECK_STATEID_REPLAY:
8468 		/* Check the sequence id for the open owner */
8469 		switch (rfs4_check_open_seqid(args->seqid, sp->rs_owner,
8470 		    resop)) {
8471 		case NFS4_CHKSEQ_OKAY:
8472 			/*
8473 			 * This is replayed stateid; if seqid matches
8474 			 * next expected, then client is using wrong seqid.
8475 			 */
8476 			/* FALL THROUGH */
8477 		case NFS4_CHKSEQ_BAD:
8478 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
8479 			goto end;
8480 		case NFS4_CHKSEQ_REPLAY:
8481 			/*
8482 			 * Note this case is the duplicate case so
8483 			 * resp->status is already set.
8484 			 */
8485 			*cs->statusp = resp->status;
8486 			rfs4_update_lease(sp->rs_owner->ro_client);
8487 			goto end;
8488 		}
8489 		break;
8490 	default:
8491 		ASSERT(FALSE);
8492 		break;
8493 	}
8494 
8495 	rfs4_dbe_lock(sp->rs_dbe);
8496 
8497 	/* Update the stateid. */
8498 	next_stateid(&sp->rs_stateid);
8499 	resp->open_stateid = sp->rs_stateid.stateid;
8500 
8501 	rfs4_dbe_unlock(sp->rs_dbe);
8502 
8503 	rfs4_update_lease(sp->rs_owner->ro_client);
8504 	rfs4_update_open_sequence(sp->rs_owner);
8505 	rfs4_update_open_resp(sp->rs_owner, resop, NULL);
8506 
8507 	rfs4_state_close(sp, FALSE, FALSE, cs->cr);
8508 
8509 	*cs->statusp = resp->status = status;
8510 
8511 end:
8512 	rfs4_sw_exit(&sp->rs_owner->ro_sw);
8513 	rfs4_state_rele(sp);
8514 out:
8515 	DTRACE_NFSV4_2(op__close__done, struct compound_state *, cs,
8516 	    CLOSE4res *, resp);
8517 }
8518 
8519 /*
8520  * Manage the counts on the file struct and close all file locks
8521  */
8522 /*ARGSUSED*/
8523 void
8524 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr,
8525     bool_t close_of_client)
8526 {
8527 	rfs4_file_t *fp = sp->rs_finfo;
8528 	rfs4_lo_state_t *lsp;
8529 	int fflags = 0;
8530 
8531 	/*
8532 	 * If this call is part of the larger closing down of client
8533 	 * state then it is just easier to release all locks
8534 	 * associated with this client instead of going through each
8535 	 * individual file and cleaning locks there.
8536 	 */
8537 	if (close_of_client) {
8538 		if (sp->rs_owner->ro_client->rc_unlksys_completed == FALSE &&
8539 		    !list_is_empty(&sp->rs_lostatelist) &&
8540 		    sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID) {
8541 			/* Is the PxFS kernel module loaded? */
8542 			if (lm_remove_file_locks != NULL) {
8543 				int new_sysid;
8544 
8545 				/* Encode the cluster nodeid in new sysid */
8546 				new_sysid = sp->rs_owner->ro_client->rc_sysidt;
8547 				lm_set_nlmid_flk(&new_sysid);
8548 
8549 				/*
8550 				 * This PxFS routine removes file locks for a
8551 				 * client over all nodes of a cluster.
8552 				 */
8553 				NFS4_DEBUG(rfs4_debug, (CE_NOTE,
8554 				    "lm_remove_file_locks(sysid=0x%x)\n",
8555 				    new_sysid));
8556 				(*lm_remove_file_locks)(new_sysid);
8557 			} else {
8558 				struct flock64 flk;
8559 
8560 				/* Release all locks for this client */
8561 				flk.l_type = F_UNLKSYS;
8562 				flk.l_whence = 0;
8563 				flk.l_start = 0;
8564 				flk.l_len = 0;
8565 				flk.l_sysid =
8566 				    sp->rs_owner->ro_client->rc_sysidt;
8567 				flk.l_pid = 0;
8568 				(void) VOP_FRLOCK(sp->rs_finfo->rf_vp, F_SETLK,
8569 				    &flk, F_REMOTELOCK | FREAD | FWRITE,
8570 				    (u_offset_t)0, NULL, CRED(), NULL);
8571 			}
8572 
8573 			sp->rs_owner->ro_client->rc_unlksys_completed = TRUE;
8574 		}
8575 	}
8576 
8577 	/*
8578 	 * Release all locks on this file by this lock owner or at
8579 	 * least mark the locks as having been released
8580 	 */
8581 	for (lsp = list_head(&sp->rs_lostatelist); lsp != NULL;
8582 	    lsp = list_next(&sp->rs_lostatelist, lsp)) {
8583 		lsp->rls_locks_cleaned = TRUE;
8584 
8585 		/* Was this already taken care of above? */
8586 		if (!close_of_client &&
8587 		    sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID)
8588 			(void) cleanlocks(sp->rs_finfo->rf_vp,
8589 			    lsp->rls_locker->rl_pid,
8590 			    lsp->rls_locker->rl_client->rc_sysidt);
8591 	}
8592 
8593 	/*
8594 	 * Release any shrlocks associated with this open state ID.
8595 	 * This must be done before the rfs4_state gets marked closed.
8596 	 */
8597 	if (sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID)
8598 		(void) rfs4_unshare(sp);
8599 
8600 	if (sp->rs_open_access) {
8601 		rfs4_dbe_lock(fp->rf_dbe);
8602 
8603 		/*
8604 		 * Decrement the count for each access and deny bit that this
8605 		 * state has contributed to the file.
8606 		 * If the file counts go to zero
8607 		 * clear the appropriate bit in the appropriate mask.
8608 		 */
8609 		if (sp->rs_open_access & OPEN4_SHARE_ACCESS_READ) {
8610 			fp->rf_access_read--;
8611 			fflags |= FREAD;
8612 			if (fp->rf_access_read == 0)
8613 				fp->rf_share_access &= ~OPEN4_SHARE_ACCESS_READ;
8614 		}
8615 		if (sp->rs_open_access & OPEN4_SHARE_ACCESS_WRITE) {
8616 			fp->rf_access_write--;
8617 			fflags |= FWRITE;
8618 			if (fp->rf_access_write == 0)
8619 				fp->rf_share_access &=
8620 				    ~OPEN4_SHARE_ACCESS_WRITE;
8621 		}
8622 		if (sp->rs_open_deny & OPEN4_SHARE_DENY_READ) {
8623 			fp->rf_deny_read--;
8624 			if (fp->rf_deny_read == 0)
8625 				fp->rf_share_deny &= ~OPEN4_SHARE_DENY_READ;
8626 		}
8627 		if (sp->rs_open_deny & OPEN4_SHARE_DENY_WRITE) {
8628 			fp->rf_deny_write--;
8629 			if (fp->rf_deny_write == 0)
8630 				fp->rf_share_deny &= ~OPEN4_SHARE_DENY_WRITE;
8631 		}
8632 
8633 		(void) VOP_CLOSE(fp->rf_vp, fflags, 1, (offset_t)0, cr, NULL);
8634 
8635 		rfs4_dbe_unlock(fp->rf_dbe);
8636 
8637 		sp->rs_open_access = 0;
8638 		sp->rs_open_deny = 0;
8639 	}
8640 }
8641 
8642 /*
8643  * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure.
8644  */
8645 static nfsstat4
8646 lock_denied(LOCK4denied *dp, struct flock64 *flk)
8647 {
8648 	rfs4_lockowner_t *lo;
8649 	rfs4_client_t *cp;
8650 	uint32_t len;
8651 
8652 	lo = rfs4_findlockowner_by_pid(flk->l_pid);
8653 	if (lo != NULL) {
8654 		cp = lo->rl_client;
8655 		if (rfs4_lease_expired(cp)) {
8656 			rfs4_lockowner_rele(lo);
8657 			rfs4_dbe_hold(cp->rc_dbe);
8658 			rfs4_client_close(cp);
8659 			return (NFS4ERR_EXPIRED);
8660 		}
8661 		dp->owner.clientid = lo->rl_owner.clientid;
8662 		len = lo->rl_owner.owner_len;
8663 		dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
8664 		bcopy(lo->rl_owner.owner_val, dp->owner.owner_val, len);
8665 		dp->owner.owner_len = len;
8666 		rfs4_lockowner_rele(lo);
8667 		goto finish;
8668 	}
8669 
8670 	/*
8671 	 * Its not a NFS4 lock. We take advantage that the upper 32 bits
8672 	 * of the client id contain the boot time for a NFS4 lock. So we
8673 	 * fabricate and identity by setting clientid to the sysid, and
8674 	 * the lock owner to the pid.
8675 	 */
8676 	dp->owner.clientid = flk->l_sysid;
8677 	len = sizeof (pid_t);
8678 	dp->owner.owner_len = len;
8679 	dp->owner.owner_val = kmem_alloc(len, KM_SLEEP);
8680 	bcopy(&flk->l_pid, dp->owner.owner_val, len);
8681 finish:
8682 	dp->offset = flk->l_start;
8683 	dp->length = flk->l_len;
8684 
8685 	if (flk->l_type == F_RDLCK)
8686 		dp->locktype = READ_LT;
8687 	else if (flk->l_type == F_WRLCK)
8688 		dp->locktype = WRITE_LT;
8689 	else
8690 		return (NFS4ERR_INVAL);	/* no mapping from POSIX ltype to v4 */
8691 
8692 	return (NFS4_OK);
8693 }
8694 
8695 /*
8696  * The NFSv4.0 LOCK operation does not support the blocking lock (at the
8697  * NFSv4.0 protocol level) so the client needs to resend the LOCK request in a
8698  * case the lock is denied by the NFSv4.0 server.  NFSv4.0 clients are prepared
8699  * for that (obviously); they are sending the LOCK requests with some delays
8700  * between the attempts.  See nfs4frlock() and nfs4_block_and_wait() for the
8701  * locking and delay implementation at the client side.
8702  *
8703  * To make the life of the clients easier, the NFSv4.0 server tries to do some
8704  * fast retries on its own (the for loop below) in a hope the lock will be
8705  * available soon.  And if not, the client won't need to resend the LOCK
8706  * requests so fast to check the lock availability.  This basically saves some
8707  * network traffic and tries to make sure the client gets the lock ASAP.
8708  */
8709 static int
8710 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred)
8711 {
8712 	int error;
8713 	struct flock64 flk;
8714 	int i;
8715 	clock_t delaytime;
8716 	int cmd;
8717 	int spin_cnt = 0;
8718 
8719 	cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK;
8720 retry:
8721 	delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay);
8722 
8723 	for (i = 0; i < rfs4_maxlock_tries; i++) {
8724 		LOCK_PRINT(rfs4_debug, "setlock", cmd, flock);
8725 		error = VOP_FRLOCK(vp, cmd,
8726 		    flock, flag, (u_offset_t)0, NULL, cred, NULL);
8727 
8728 		if (error != EAGAIN && error != EACCES)
8729 			break;
8730 
8731 		if (i < rfs4_maxlock_tries - 1) {
8732 			delay(delaytime);
8733 			delaytime *= 2;
8734 		}
8735 	}
8736 
8737 	if (error == EAGAIN || error == EACCES) {
8738 		/* Get the owner of the lock */
8739 		flk = *flock;
8740 		LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk);
8741 		if (VOP_FRLOCK(vp, F_GETLK, &flk, flag, 0, NULL, cred,
8742 		    NULL) == 0) {
8743 			/*
8744 			 * There's a race inherent in the current VOP_FRLOCK
8745 			 * design where:
8746 			 * a: "other guy" takes a lock that conflicts with a
8747 			 * lock we want
8748 			 * b: we attempt to take our lock (non-blocking) and
8749 			 * the attempt fails.
8750 			 * c: "other guy" releases the conflicting lock
8751 			 * d: we ask what lock conflicts with the lock we want,
8752 			 * getting F_UNLCK (no lock blocks us)
8753 			 *
8754 			 * If we retry the non-blocking lock attempt in this
8755 			 * case (restart at step 'b') there's some possibility
8756 			 * that many such attempts might fail.  However a test
8757 			 * designed to actually provoke this race shows that
8758 			 * the vast majority of cases require no retry, and
8759 			 * only a few took as many as three retries.  Here's
8760 			 * the test outcome:
8761 			 *
8762 			 *	   number of retries    how many times we needed
8763 			 *				that many retries
8764 			 *	   0			79461
8765 			 *	   1			  862
8766 			 *	   2			   49
8767 			 *	   3			    5
8768 			 *
8769 			 * Given those empirical results, we arbitrarily limit
8770 			 * the retry count to ten.
8771 			 *
8772 			 * If we actually make to ten retries and give up,
8773 			 * nothing catastrophic happens, but we're unable to
8774 			 * return the information about the conflicting lock to
8775 			 * the NFS client.  That's an acceptable trade off vs.
8776 			 * letting this retry loop run forever.
8777 			 */
8778 			if (flk.l_type == F_UNLCK) {
8779 				if (spin_cnt++ < 10) {
8780 					/* No longer locked, retry */
8781 					goto retry;
8782 				}
8783 			} else {
8784 				*flock = flk;
8785 				LOCK_PRINT(rfs4_debug, "setlock(blocking lock)",
8786 				    F_GETLK, &flk);
8787 			}
8788 		}
8789 	}
8790 
8791 	return (error);
8792 }
8793 
8794 /*ARGSUSED*/
8795 static nfsstat4
8796 rfs4_do_lock(rfs4_lo_state_t *lsp, nfs_lock_type4 locktype,
8797     offset4 offset, length4 length, cred_t *cred, nfs_resop4 *resop)
8798 {
8799 	nfsstat4 status;
8800 	rfs4_lockowner_t *lo = lsp->rls_locker;
8801 	rfs4_state_t *sp = lsp->rls_state;
8802 	struct flock64 flock;
8803 	int16_t ltype;
8804 	int flag;
8805 	int error;
8806 	sysid_t sysid;
8807 	LOCK4res *lres;
8808 	vnode_t *vp;
8809 
8810 	if (rfs4_lease_expired(lo->rl_client)) {
8811 		return (NFS4ERR_EXPIRED);
8812 	}
8813 
8814 	if ((status = rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK)
8815 		return (status);
8816 
8817 	/* Check for zero length. To lock to end of file use all ones for V4 */
8818 	if (length == 0)
8819 		return (NFS4ERR_INVAL);
8820 	else if (length == (length4)(~0))
8821 		length = 0;		/* Posix to end of file  */
8822 
8823 retry:
8824 	rfs4_dbe_lock(sp->rs_dbe);
8825 	if (sp->rs_closed == TRUE) {
8826 		rfs4_dbe_unlock(sp->rs_dbe);
8827 		return (NFS4ERR_OLD_STATEID);
8828 	}
8829 
8830 	if (resop->resop != OP_LOCKU) {
8831 		switch (locktype) {
8832 		case READ_LT:
8833 		case READW_LT:
8834 			if ((sp->rs_share_access
8835 			    & OPEN4_SHARE_ACCESS_READ) == 0) {
8836 				rfs4_dbe_unlock(sp->rs_dbe);
8837 
8838 				return (NFS4ERR_OPENMODE);
8839 			}
8840 			ltype = F_RDLCK;
8841 			break;
8842 		case WRITE_LT:
8843 		case WRITEW_LT:
8844 			if ((sp->rs_share_access
8845 			    & OPEN4_SHARE_ACCESS_WRITE) == 0) {
8846 				rfs4_dbe_unlock(sp->rs_dbe);
8847 
8848 				return (NFS4ERR_OPENMODE);
8849 			}
8850 			ltype = F_WRLCK;
8851 			break;
8852 		}
8853 	} else
8854 		ltype = F_UNLCK;
8855 
8856 	flock.l_type = ltype;
8857 	flock.l_whence = 0;		/* SEEK_SET */
8858 	flock.l_start = offset;
8859 	flock.l_len = length;
8860 	flock.l_sysid = sysid;
8861 	flock.l_pid = lsp->rls_locker->rl_pid;
8862 
8863 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
8864 	if (flock.l_len < 0 || flock.l_start < 0) {
8865 		rfs4_dbe_unlock(sp->rs_dbe);
8866 		return (NFS4ERR_INVAL);
8867 	}
8868 
8869 	/*
8870 	 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and
8871 	 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE.
8872 	 */
8873 	flag = (int)sp->rs_share_access | F_REMOTELOCK;
8874 
8875 	vp = sp->rs_finfo->rf_vp;
8876 	VN_HOLD(vp);
8877 
8878 	/*
8879 	 * We need to unlock sp before we call the underlying filesystem to
8880 	 * acquire the file lock.
8881 	 */
8882 	rfs4_dbe_unlock(sp->rs_dbe);
8883 
8884 	error = setlock(vp, &flock, flag, cred);
8885 
8886 	/*
8887 	 * Make sure the file is still open.  In a case the file was closed in
8888 	 * the meantime, clean the lock we acquired using the setlock() call
8889 	 * above, and return the appropriate error.
8890 	 */
8891 	rfs4_dbe_lock(sp->rs_dbe);
8892 	if (sp->rs_closed == TRUE) {
8893 		cleanlocks(vp, lsp->rls_locker->rl_pid, sysid);
8894 		rfs4_dbe_unlock(sp->rs_dbe);
8895 
8896 		VN_RELE(vp);
8897 
8898 		return (NFS4ERR_OLD_STATEID);
8899 	}
8900 	rfs4_dbe_unlock(sp->rs_dbe);
8901 
8902 	VN_RELE(vp);
8903 
8904 	if (error == 0) {
8905 		rfs4_dbe_lock(lsp->rls_dbe);
8906 		next_stateid(&lsp->rls_lockid);
8907 		rfs4_dbe_unlock(lsp->rls_dbe);
8908 	}
8909 
8910 	/*
8911 	 * N.B. We map error values to nfsv4 errors. This is differrent
8912 	 * than puterrno4 routine.
8913 	 */
8914 	switch (error) {
8915 	case 0:
8916 		status = NFS4_OK;
8917 		break;
8918 	case EAGAIN:
8919 	case EACCES:		/* Old value */
8920 		/* Can only get here if op is OP_LOCK */
8921 		ASSERT(resop->resop == OP_LOCK);
8922 		lres = &resop->nfs_resop4_u.oplock;
8923 		status = NFS4ERR_DENIED;
8924 		if (lock_denied(&lres->LOCK4res_u.denied, &flock)
8925 		    == NFS4ERR_EXPIRED)
8926 			goto retry;
8927 		break;
8928 	case ENOLCK:
8929 		status = NFS4ERR_DELAY;
8930 		break;
8931 	case EOVERFLOW:
8932 		status = NFS4ERR_INVAL;
8933 		break;
8934 	case EINVAL:
8935 		status = NFS4ERR_NOTSUPP;
8936 		break;
8937 	default:
8938 		status = NFS4ERR_SERVERFAULT;
8939 		break;
8940 	}
8941 
8942 	return (status);
8943 }
8944 
8945 /*ARGSUSED*/
8946 void
8947 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop,
8948     struct svc_req *req, struct compound_state *cs)
8949 {
8950 	LOCK4args *args = &argop->nfs_argop4_u.oplock;
8951 	LOCK4res *resp = &resop->nfs_resop4_u.oplock;
8952 	nfsstat4 status;
8953 	stateid4 *stateid;
8954 	rfs4_lockowner_t *lo;
8955 	rfs4_client_t *cp;
8956 	rfs4_state_t *sp = NULL;
8957 	rfs4_lo_state_t *lsp = NULL;
8958 	bool_t ls_sw_held = FALSE;
8959 	bool_t create = TRUE;
8960 	bool_t lcreate = TRUE;
8961 	bool_t dup_lock = FALSE;
8962 	int rc;
8963 
8964 	DTRACE_NFSV4_2(op__lock__start, struct compound_state *, cs,
8965 	    LOCK4args *, args);
8966 
8967 	if (cs->vp == NULL) {
8968 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
8969 		DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8970 		    cs, LOCK4res *, resp);
8971 		return;
8972 	}
8973 
8974 	if (args->locker.new_lock_owner) {
8975 		/* Create a new lockowner for this instance */
8976 		open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner;
8977 
8978 		NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner"));
8979 
8980 		stateid = &olo->open_stateid;
8981 		status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID);
8982 		if (status != NFS4_OK) {
8983 			NFS4_DEBUG(rfs4_debug,
8984 			    (CE_NOTE, "Get state failed in lock %d", status));
8985 			*cs->statusp = resp->status = status;
8986 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8987 			    cs, LOCK4res *, resp);
8988 			return;
8989 		}
8990 
8991 		/* Ensure specified filehandle matches */
8992 		if (cs->vp != sp->rs_finfo->rf_vp) {
8993 			rfs4_state_rele(sp);
8994 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
8995 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
8996 			    cs, LOCK4res *, resp);
8997 			return;
8998 		}
8999 
9000 		/* hold off other access to open_owner while we tinker */
9001 		rfs4_sw_enter(&sp->rs_owner->ro_sw);
9002 
9003 		switch (rc = rfs4_check_stateid_seqid(sp, stateid)) {
9004 		case NFS4_CHECK_STATEID_OLD:
9005 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9006 			goto end;
9007 		case NFS4_CHECK_STATEID_BAD:
9008 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9009 			goto end;
9010 		case NFS4_CHECK_STATEID_EXPIRED:
9011 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
9012 			goto end;
9013 		case NFS4_CHECK_STATEID_UNCONFIRMED:
9014 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9015 			goto end;
9016 		case NFS4_CHECK_STATEID_CLOSED:
9017 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9018 			goto end;
9019 		case NFS4_CHECK_STATEID_OKAY:
9020 		case NFS4_CHECK_STATEID_REPLAY:
9021 			switch (rfs4_check_olo_seqid(olo->open_seqid,
9022 			    sp->rs_owner, resop)) {
9023 			case NFS4_CHKSEQ_OKAY:
9024 				if (rc == NFS4_CHECK_STATEID_OKAY)
9025 					break;
9026 				/*
9027 				 * This is replayed stateid; if seqid
9028 				 * matches next expected, then client
9029 				 * is using wrong seqid.
9030 				 */
9031 				/* FALLTHROUGH */
9032 			case NFS4_CHKSEQ_BAD:
9033 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9034 				goto end;
9035 			case NFS4_CHKSEQ_REPLAY:
9036 				/* This is a duplicate LOCK request */
9037 				dup_lock = TRUE;
9038 
9039 				/*
9040 				 * For a duplicate we do not want to
9041 				 * create a new lockowner as it should
9042 				 * already exist.
9043 				 * Turn off the lockowner create flag.
9044 				 */
9045 				lcreate = FALSE;
9046 			}
9047 			break;
9048 		}
9049 
9050 		lo = rfs4_findlockowner(&olo->lock_owner, &lcreate);
9051 		if (lo == NULL) {
9052 			NFS4_DEBUG(rfs4_debug,
9053 			    (CE_NOTE, "rfs4_op_lock: no lock owner"));
9054 			*cs->statusp = resp->status = NFS4ERR_RESOURCE;
9055 			goto end;
9056 		}
9057 
9058 		lsp = rfs4_findlo_state_by_owner(lo, sp, &create);
9059 		if (lsp == NULL) {
9060 			rfs4_update_lease(sp->rs_owner->ro_client);
9061 			/*
9062 			 * Only update theh open_seqid if this is not
9063 			 * a duplicate request
9064 			 */
9065 			if (dup_lock == FALSE) {
9066 				rfs4_update_open_sequence(sp->rs_owner);
9067 			}
9068 
9069 			NFS4_DEBUG(rfs4_debug,
9070 			    (CE_NOTE, "rfs4_op_lock: no state"));
9071 			*cs->statusp = resp->status = NFS4ERR_SERVERFAULT;
9072 			rfs4_update_open_resp(sp->rs_owner, resop, NULL);
9073 			rfs4_lockowner_rele(lo);
9074 			goto end;
9075 		}
9076 
9077 		/*
9078 		 * This is the new_lock_owner branch and the client is
9079 		 * supposed to be associating a new lock_owner with
9080 		 * the open file at this point.  If we find that a
9081 		 * lock_owner/state association already exists and a
9082 		 * successful LOCK request was returned to the client,
9083 		 * an error is returned to the client since this is
9084 		 * not appropriate.  The client should be using the
9085 		 * existing lock_owner branch.
9086 		 */
9087 		if (dup_lock == FALSE && create == FALSE) {
9088 			if (lsp->rls_lock_completed == TRUE) {
9089 				*cs->statusp =
9090 				    resp->status = NFS4ERR_BAD_SEQID;
9091 				rfs4_lockowner_rele(lo);
9092 				goto end;
9093 			}
9094 		}
9095 
9096 		rfs4_update_lease(sp->rs_owner->ro_client);
9097 
9098 		/*
9099 		 * Only update theh open_seqid if this is not
9100 		 * a duplicate request
9101 		 */
9102 		if (dup_lock == FALSE) {
9103 			rfs4_update_open_sequence(sp->rs_owner);
9104 		}
9105 
9106 		/*
9107 		 * If this is a duplicate lock request, just copy the
9108 		 * previously saved reply and return.
9109 		 */
9110 		if (dup_lock == TRUE) {
9111 			/* verify that lock_seqid's match */
9112 			if (lsp->rls_seqid != olo->lock_seqid) {
9113 				NFS4_DEBUG(rfs4_debug,
9114 				    (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad"
9115 				    "lsp->seqid=%d old->seqid=%d",
9116 				    lsp->rls_seqid, olo->lock_seqid));
9117 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9118 			} else {
9119 				rfs4_copy_reply(resop, &lsp->rls_reply);
9120 				/*
9121 				 * Make sure to copy the just
9122 				 * retrieved reply status into the
9123 				 * overall compound status
9124 				 */
9125 				*cs->statusp = resp->status;
9126 			}
9127 			rfs4_lockowner_rele(lo);
9128 			goto end;
9129 		}
9130 
9131 		rfs4_dbe_lock(lsp->rls_dbe);
9132 
9133 		/* Make sure to update the lock sequence id */
9134 		lsp->rls_seqid = olo->lock_seqid;
9135 
9136 		NFS4_DEBUG(rfs4_debug,
9137 		    (CE_NOTE, "Lock seqid established as %d", lsp->rls_seqid));
9138 
9139 		/*
9140 		 * This is used to signify the newly created lockowner
9141 		 * stateid and its sequence number.  The checks for
9142 		 * sequence number and increment don't occur on the
9143 		 * very first lock request for a lockowner.
9144 		 */
9145 		lsp->rls_skip_seqid_check = TRUE;
9146 
9147 		/* hold off other access to lsp while we tinker */
9148 		rfs4_sw_enter(&lsp->rls_sw);
9149 		ls_sw_held = TRUE;
9150 
9151 		rfs4_dbe_unlock(lsp->rls_dbe);
9152 
9153 		rfs4_lockowner_rele(lo);
9154 	} else {
9155 		stateid = &args->locker.locker4_u.lock_owner.lock_stateid;
9156 		/* get lsp and hold the lock on the underlying file struct */
9157 		if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE))
9158 		    != NFS4_OK) {
9159 			*cs->statusp = resp->status = status;
9160 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
9161 			    cs, LOCK4res *, resp);
9162 			return;
9163 		}
9164 		create = FALSE;	/* We didn't create lsp */
9165 
9166 		/* Ensure specified filehandle matches */
9167 		if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) {
9168 			rfs4_lo_state_rele(lsp, TRUE);
9169 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9170 			DTRACE_NFSV4_2(op__lock__done, struct compound_state *,
9171 			    cs, LOCK4res *, resp);
9172 			return;
9173 		}
9174 
9175 		/* hold off other access to lsp while we tinker */
9176 		rfs4_sw_enter(&lsp->rls_sw);
9177 		ls_sw_held = TRUE;
9178 
9179 		switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
9180 		/*
9181 		 * The stateid looks like it was okay (expected to be
9182 		 * the next one)
9183 		 */
9184 		case NFS4_CHECK_STATEID_OKAY:
9185 			/*
9186 			 * The sequence id is now checked.  Determine
9187 			 * if this is a replay or if it is in the
9188 			 * expected (next) sequence.  In the case of a
9189 			 * replay, there are two replay conditions
9190 			 * that may occur.  The first is the normal
9191 			 * condition where a LOCK is done with a
9192 			 * NFS4_OK response and the stateid is
9193 			 * updated.  That case is handled below when
9194 			 * the stateid is identified as a REPLAY.  The
9195 			 * second is the case where an error is
9196 			 * returned, like NFS4ERR_DENIED, and the
9197 			 * sequence number is updated but the stateid
9198 			 * is not updated.  This second case is dealt
9199 			 * with here.  So it may seem odd that the
9200 			 * stateid is okay but the sequence id is a
9201 			 * replay but it is okay.
9202 			 */
9203 			switch (rfs4_check_lock_seqid(
9204 			    args->locker.locker4_u.lock_owner.lock_seqid,
9205 			    lsp, resop)) {
9206 			case NFS4_CHKSEQ_REPLAY:
9207 				if (resp->status != NFS4_OK) {
9208 					/*
9209 					 * Here is our replay and need
9210 					 * to verify that the last
9211 					 * response was an error.
9212 					 */
9213 					*cs->statusp = resp->status;
9214 					goto end;
9215 				}
9216 				/*
9217 				 * This is done since the sequence id
9218 				 * looked like a replay but it didn't
9219 				 * pass our check so a BAD_SEQID is
9220 				 * returned as a result.
9221 				 */
9222 				/*FALLTHROUGH*/
9223 			case NFS4_CHKSEQ_BAD:
9224 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9225 				goto end;
9226 			case NFS4_CHKSEQ_OKAY:
9227 				/* Everything looks okay move ahead */
9228 				break;
9229 			}
9230 			break;
9231 		case NFS4_CHECK_STATEID_OLD:
9232 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9233 			goto end;
9234 		case NFS4_CHECK_STATEID_BAD:
9235 			*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9236 			goto end;
9237 		case NFS4_CHECK_STATEID_EXPIRED:
9238 			*cs->statusp = resp->status = NFS4ERR_EXPIRED;
9239 			goto end;
9240 		case NFS4_CHECK_STATEID_CLOSED:
9241 			*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9242 			goto end;
9243 		case NFS4_CHECK_STATEID_REPLAY:
9244 			switch (rfs4_check_lock_seqid(
9245 			    args->locker.locker4_u.lock_owner.lock_seqid,
9246 			    lsp, resop)) {
9247 			case NFS4_CHKSEQ_OKAY:
9248 				/*
9249 				 * This is a replayed stateid; if
9250 				 * seqid matches the next expected,
9251 				 * then client is using wrong seqid.
9252 				 */
9253 			case NFS4_CHKSEQ_BAD:
9254 				*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9255 				goto end;
9256 			case NFS4_CHKSEQ_REPLAY:
9257 				rfs4_update_lease(lsp->rls_locker->rl_client);
9258 				*cs->statusp = status = resp->status;
9259 				goto end;
9260 			}
9261 			break;
9262 		default:
9263 			ASSERT(FALSE);
9264 			break;
9265 		}
9266 
9267 		rfs4_update_lock_sequence(lsp);
9268 		rfs4_update_lease(lsp->rls_locker->rl_client);
9269 	}
9270 
9271 	/*
9272 	 * NFS4 only allows locking on regular files, so
9273 	 * verify type of object.
9274 	 */
9275 	if (cs->vp->v_type != VREG) {
9276 		if (cs->vp->v_type == VDIR)
9277 			status = NFS4ERR_ISDIR;
9278 		else
9279 			status = NFS4ERR_INVAL;
9280 		goto out;
9281 	}
9282 
9283 	cp = lsp->rls_state->rs_owner->ro_client;
9284 
9285 	if (rfs4_clnt_in_grace(cp) && !args->reclaim) {
9286 		status = NFS4ERR_GRACE;
9287 		goto out;
9288 	}
9289 
9290 	if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->rc_can_reclaim) {
9291 		status = NFS4ERR_NO_GRACE;
9292 		goto out;
9293 	}
9294 
9295 	if (!rfs4_clnt_in_grace(cp) && args->reclaim) {
9296 		status = NFS4ERR_NO_GRACE;
9297 		goto out;
9298 	}
9299 
9300 	if (lsp->rls_state->rs_finfo->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE)
9301 		cs->deleg = TRUE;
9302 
9303 	status = rfs4_do_lock(lsp, args->locktype,
9304 	    args->offset, args->length, cs->cr, resop);
9305 
9306 out:
9307 	lsp->rls_skip_seqid_check = FALSE;
9308 
9309 	*cs->statusp = resp->status = status;
9310 
9311 	if (status == NFS4_OK) {
9312 		resp->LOCK4res_u.lock_stateid = lsp->rls_lockid.stateid;
9313 		lsp->rls_lock_completed = TRUE;
9314 	}
9315 	/*
9316 	 * Only update the "OPEN" response here if this was a new
9317 	 * lock_owner
9318 	 */
9319 	if (sp)
9320 		rfs4_update_open_resp(sp->rs_owner, resop, NULL);
9321 
9322 	rfs4_update_lock_resp(lsp, resop);
9323 
9324 end:
9325 	if (lsp) {
9326 		if (ls_sw_held)
9327 			rfs4_sw_exit(&lsp->rls_sw);
9328 		/*
9329 		 * If an sp obtained, then the lsp does not represent
9330 		 * a lock on the file struct.
9331 		 */
9332 		if (sp != NULL)
9333 			rfs4_lo_state_rele(lsp, FALSE);
9334 		else
9335 			rfs4_lo_state_rele(lsp, TRUE);
9336 	}
9337 	if (sp) {
9338 		rfs4_sw_exit(&sp->rs_owner->ro_sw);
9339 		rfs4_state_rele(sp);
9340 	}
9341 
9342 	DTRACE_NFSV4_2(op__lock__done, struct compound_state *, cs,
9343 	    LOCK4res *, resp);
9344 }
9345 
9346 /* free function for LOCK/LOCKT */
9347 static void
9348 lock_denied_free(nfs_resop4 *resop)
9349 {
9350 	LOCK4denied *dp = NULL;
9351 
9352 	switch (resop->resop) {
9353 	case OP_LOCK:
9354 		if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED)
9355 			dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied;
9356 		break;
9357 	case OP_LOCKT:
9358 		if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED)
9359 			dp = &resop->nfs_resop4_u.oplockt.denied;
9360 		break;
9361 	default:
9362 		break;
9363 	}
9364 
9365 	if (dp)
9366 		kmem_free(dp->owner.owner_val, dp->owner.owner_len);
9367 }
9368 
9369 /*ARGSUSED*/
9370 void
9371 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop,
9372     struct svc_req *req, struct compound_state *cs)
9373 {
9374 	LOCKU4args *args = &argop->nfs_argop4_u.oplocku;
9375 	LOCKU4res *resp = &resop->nfs_resop4_u.oplocku;
9376 	nfsstat4 status;
9377 	stateid4 *stateid = &args->lock_stateid;
9378 	rfs4_lo_state_t *lsp;
9379 
9380 	DTRACE_NFSV4_2(op__locku__start, struct compound_state *, cs,
9381 	    LOCKU4args *, args);
9382 
9383 	if (cs->vp == NULL) {
9384 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
9385 		DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
9386 		    LOCKU4res *, resp);
9387 		return;
9388 	}
9389 
9390 	if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) {
9391 		*cs->statusp = resp->status = status;
9392 		DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
9393 		    LOCKU4res *, resp);
9394 		return;
9395 	}
9396 
9397 	/* Ensure specified filehandle matches */
9398 	if (cs->vp != lsp->rls_state->rs_finfo->rf_vp) {
9399 		rfs4_lo_state_rele(lsp, TRUE);
9400 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9401 		DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
9402 		    LOCKU4res *, resp);
9403 		return;
9404 	}
9405 
9406 	/* hold off other access to lsp while we tinker */
9407 	rfs4_sw_enter(&lsp->rls_sw);
9408 
9409 	switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) {
9410 	case NFS4_CHECK_STATEID_OKAY:
9411 		if (rfs4_check_lock_seqid(args->seqid, lsp, resop)
9412 		    != NFS4_CHKSEQ_OKAY) {
9413 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9414 			goto end;
9415 		}
9416 		break;
9417 	case NFS4_CHECK_STATEID_OLD:
9418 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9419 		goto end;
9420 	case NFS4_CHECK_STATEID_BAD:
9421 		*cs->statusp = resp->status = NFS4ERR_BAD_STATEID;
9422 		goto end;
9423 	case NFS4_CHECK_STATEID_EXPIRED:
9424 		*cs->statusp = resp->status = NFS4ERR_EXPIRED;
9425 		goto end;
9426 	case NFS4_CHECK_STATEID_CLOSED:
9427 		*cs->statusp = resp->status = NFS4ERR_OLD_STATEID;
9428 		goto end;
9429 	case NFS4_CHECK_STATEID_REPLAY:
9430 		switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) {
9431 		case NFS4_CHKSEQ_OKAY:
9432 				/*
9433 				 * This is a replayed stateid; if
9434 				 * seqid matches the next expected,
9435 				 * then client is using wrong seqid.
9436 				 */
9437 		case NFS4_CHKSEQ_BAD:
9438 			*cs->statusp = resp->status = NFS4ERR_BAD_SEQID;
9439 			goto end;
9440 		case NFS4_CHKSEQ_REPLAY:
9441 			rfs4_update_lease(lsp->rls_locker->rl_client);
9442 			*cs->statusp = status = resp->status;
9443 			goto end;
9444 		}
9445 		break;
9446 	default:
9447 		ASSERT(FALSE);
9448 		break;
9449 	}
9450 
9451 	rfs4_update_lock_sequence(lsp);
9452 	rfs4_update_lease(lsp->rls_locker->rl_client);
9453 
9454 	/*
9455 	 * NFS4 only allows locking on regular files, so
9456 	 * verify type of object.
9457 	 */
9458 	if (cs->vp->v_type != VREG) {
9459 		if (cs->vp->v_type == VDIR)
9460 			status = NFS4ERR_ISDIR;
9461 		else
9462 			status = NFS4ERR_INVAL;
9463 		goto out;
9464 	}
9465 
9466 	if (rfs4_clnt_in_grace(lsp->rls_state->rs_owner->ro_client)) {
9467 		status = NFS4ERR_GRACE;
9468 		goto out;
9469 	}
9470 
9471 	status = rfs4_do_lock(lsp, args->locktype,
9472 	    args->offset, args->length, cs->cr, resop);
9473 
9474 out:
9475 	*cs->statusp = resp->status = status;
9476 
9477 	if (status == NFS4_OK)
9478 		resp->lock_stateid = lsp->rls_lockid.stateid;
9479 
9480 	rfs4_update_lock_resp(lsp, resop);
9481 
9482 end:
9483 	rfs4_sw_exit(&lsp->rls_sw);
9484 	rfs4_lo_state_rele(lsp, TRUE);
9485 
9486 	DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs,
9487 	    LOCKU4res *, resp);
9488 }
9489 
9490 /*
9491  * LOCKT is a best effort routine, the client can not be guaranteed that
9492  * the status return is still in effect by the time the reply is received.
9493  * They are numerous race conditions in this routine, but we are not required
9494  * and can not be accurate.
9495  */
9496 /*ARGSUSED*/
9497 void
9498 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop,
9499     struct svc_req *req, struct compound_state *cs)
9500 {
9501 	LOCKT4args *args = &argop->nfs_argop4_u.oplockt;
9502 	LOCKT4res *resp = &resop->nfs_resop4_u.oplockt;
9503 	rfs4_lockowner_t *lo;
9504 	rfs4_client_t *cp;
9505 	bool_t create = FALSE;
9506 	struct flock64 flk;
9507 	int error;
9508 	int flag = FREAD | FWRITE;
9509 	int ltype;
9510 	length4 posix_length;
9511 	sysid_t sysid;
9512 	pid_t pid;
9513 
9514 	DTRACE_NFSV4_2(op__lockt__start, struct compound_state *, cs,
9515 	    LOCKT4args *, args);
9516 
9517 	if (cs->vp == NULL) {
9518 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
9519 		goto out;
9520 	}
9521 
9522 	/*
9523 	 * NFS4 only allows locking on regular files, so
9524 	 * verify type of object.
9525 	 */
9526 	if (cs->vp->v_type != VREG) {
9527 		if (cs->vp->v_type == VDIR)
9528 			*cs->statusp = resp->status = NFS4ERR_ISDIR;
9529 		else
9530 			*cs->statusp = resp->status =  NFS4ERR_INVAL;
9531 		goto out;
9532 	}
9533 
9534 	/*
9535 	 * Check out the clientid to ensure the server knows about it
9536 	 * so that we correctly inform the client of a server reboot.
9537 	 */
9538 	if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE))
9539 	    == NULL) {
9540 		*cs->statusp = resp->status =
9541 		    rfs4_check_clientid(&args->owner.clientid, 0);
9542 		goto out;
9543 	}
9544 	if (rfs4_lease_expired(cp)) {
9545 		rfs4_client_close(cp);
9546 		/*
9547 		 * Protocol doesn't allow returning NFS4ERR_STALE as
9548 		 * other operations do on this check so STALE_CLIENTID
9549 		 * is returned instead
9550 		 */
9551 		*cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
9552 		goto out;
9553 	}
9554 
9555 	if (rfs4_clnt_in_grace(cp) && !(cp->rc_can_reclaim)) {
9556 		*cs->statusp = resp->status = NFS4ERR_GRACE;
9557 		rfs4_client_rele(cp);
9558 		goto out;
9559 	}
9560 	rfs4_client_rele(cp);
9561 
9562 	resp->status = NFS4_OK;
9563 
9564 	switch (args->locktype) {
9565 	case READ_LT:
9566 	case READW_LT:
9567 		ltype = F_RDLCK;
9568 		break;
9569 	case WRITE_LT:
9570 	case WRITEW_LT:
9571 		ltype = F_WRLCK;
9572 		break;
9573 	}
9574 
9575 	posix_length = args->length;
9576 	/* Check for zero length. To lock to end of file use all ones for V4 */
9577 	if (posix_length == 0) {
9578 		*cs->statusp = resp->status = NFS4ERR_INVAL;
9579 		goto out;
9580 	} else if (posix_length == (length4)(~0)) {
9581 		posix_length = 0;	/* Posix to end of file  */
9582 	}
9583 
9584 	/* Find or create a lockowner */
9585 	lo = rfs4_findlockowner(&args->owner, &create);
9586 
9587 	if (lo) {
9588 		pid = lo->rl_pid;
9589 		if ((resp->status =
9590 		    rfs4_client_sysid(lo->rl_client, &sysid)) != NFS4_OK)
9591 			goto err;
9592 	} else {
9593 		pid = 0;
9594 		sysid = lockt_sysid;
9595 	}
9596 retry:
9597 	flk.l_type = ltype;
9598 	flk.l_whence = 0;		/* SEEK_SET */
9599 	flk.l_start = args->offset;
9600 	flk.l_len = posix_length;
9601 	flk.l_sysid = sysid;
9602 	flk.l_pid = pid;
9603 	flag |= F_REMOTELOCK;
9604 
9605 	LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk);
9606 
9607 	/* Note that length4 is uint64_t but l_len and l_start are off64_t */
9608 	if (flk.l_len < 0 || flk.l_start < 0) {
9609 		resp->status = NFS4ERR_INVAL;
9610 		goto err;
9611 	}
9612 	error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0,
9613 	    NULL, cs->cr, NULL);
9614 
9615 	/*
9616 	 * N.B. We map error values to nfsv4 errors. This is differrent
9617 	 * than puterrno4 routine.
9618 	 */
9619 	switch (error) {
9620 	case 0:
9621 		if (flk.l_type == F_UNLCK)
9622 			resp->status = NFS4_OK;
9623 		else {
9624 			if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED)
9625 				goto retry;
9626 			resp->status = NFS4ERR_DENIED;
9627 		}
9628 		break;
9629 	case EOVERFLOW:
9630 		resp->status = NFS4ERR_INVAL;
9631 		break;
9632 	case EINVAL:
9633 		resp->status = NFS4ERR_NOTSUPP;
9634 		break;
9635 	default:
9636 		cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)",
9637 		    error);
9638 		resp->status = NFS4ERR_SERVERFAULT;
9639 		break;
9640 	}
9641 
9642 err:
9643 	if (lo)
9644 		rfs4_lockowner_rele(lo);
9645 	*cs->statusp = resp->status;
9646 out:
9647 	DTRACE_NFSV4_2(op__lockt__done, struct compound_state *, cs,
9648 	    LOCKT4res *, resp);
9649 }
9650 
9651 int
9652 rfs4_share(rfs4_state_t *sp, uint32_t access, uint32_t deny)
9653 {
9654 	int err;
9655 	int cmd;
9656 	vnode_t *vp;
9657 	struct shrlock shr;
9658 	struct shr_locowner shr_loco;
9659 	int fflags = 0;
9660 
9661 	ASSERT(rfs4_dbe_islocked(sp->rs_dbe));
9662 	ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID);
9663 
9664 	if (sp->rs_closed)
9665 		return (NFS4ERR_OLD_STATEID);
9666 
9667 	vp = sp->rs_finfo->rf_vp;
9668 	ASSERT(vp);
9669 
9670 	shr.s_access = shr.s_deny = 0;
9671 
9672 	if (access & OPEN4_SHARE_ACCESS_READ) {
9673 		fflags |= FREAD;
9674 		shr.s_access |= F_RDACC;
9675 	}
9676 	if (access & OPEN4_SHARE_ACCESS_WRITE) {
9677 		fflags |= FWRITE;
9678 		shr.s_access |= F_WRACC;
9679 	}
9680 	ASSERT(shr.s_access);
9681 
9682 	if (deny & OPEN4_SHARE_DENY_READ)
9683 		shr.s_deny |= F_RDDNY;
9684 	if (deny & OPEN4_SHARE_DENY_WRITE)
9685 		shr.s_deny |= F_WRDNY;
9686 
9687 	shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe);
9688 	shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt;
9689 	shr_loco.sl_pid = shr.s_pid;
9690 	shr_loco.sl_id = shr.s_sysid;
9691 	shr.s_owner = (caddr_t)&shr_loco;
9692 	shr.s_own_len = sizeof (shr_loco);
9693 
9694 	cmd = nbl_need_check(vp) ? F_SHARE_NBMAND : F_SHARE;
9695 
9696 	err = VOP_SHRLOCK(vp, cmd, &shr, fflags, CRED(), NULL);
9697 	if (err != 0) {
9698 		if (err == EAGAIN)
9699 			err = NFS4ERR_SHARE_DENIED;
9700 		else
9701 			err = puterrno4(err);
9702 		return (err);
9703 	}
9704 
9705 	sp->rs_share_access |= access;
9706 	sp->rs_share_deny |= deny;
9707 
9708 	return (0);
9709 }
9710 
9711 int
9712 rfs4_unshare(rfs4_state_t *sp)
9713 {
9714 	int err;
9715 	struct shrlock shr;
9716 	struct shr_locowner shr_loco;
9717 
9718 	ASSERT(rfs4_dbe_islocked(sp->rs_dbe));
9719 
9720 	if (sp->rs_closed || sp->rs_share_access == 0)
9721 		return (0);
9722 
9723 	ASSERT(sp->rs_owner->ro_client->rc_sysidt != LM_NOSYSID);
9724 	ASSERT(sp->rs_finfo->rf_vp);
9725 
9726 	shr.s_access = shr.s_deny = 0;
9727 	shr.s_pid = rfs4_dbe_getid(sp->rs_owner->ro_dbe);
9728 	shr.s_sysid = sp->rs_owner->ro_client->rc_sysidt;
9729 	shr_loco.sl_pid = shr.s_pid;
9730 	shr_loco.sl_id = shr.s_sysid;
9731 	shr.s_owner = (caddr_t)&shr_loco;
9732 	shr.s_own_len = sizeof (shr_loco);
9733 
9734 	err = VOP_SHRLOCK(sp->rs_finfo->rf_vp, F_UNSHARE, &shr, 0, CRED(),
9735 	    NULL);
9736 	if (err != 0) {
9737 		err = puterrno4(err);
9738 		return (err);
9739 	}
9740 
9741 	sp->rs_share_access = 0;
9742 	sp->rs_share_deny = 0;
9743 
9744 	return (0);
9745 
9746 }
9747 
9748 static int
9749 rdma_setup_read_data4(READ4args *args, READ4res *rok)
9750 {
9751 	struct clist	*wcl;
9752 	count4		count = rok->data_len;
9753 	int		wlist_len;
9754 
9755 	wcl = args->wlist;
9756 	if (rdma_setup_read_chunks(wcl, count, &wlist_len) == FALSE) {
9757 		return (FALSE);
9758 	}
9759 	wcl = args->wlist;
9760 	rok->wlist_len = wlist_len;
9761 	rok->wlist = wcl;
9762 	return (TRUE);
9763 }
9764 
9765 /* tunable to disable server referrals */
9766 int rfs4_no_referrals = 0;
9767 
9768 /*
9769  * Find an NFS record in reparse point data.
9770  * Returns 0 for success and <0 or an errno value on failure.
9771  */
9772 int
9773 vn_find_nfs_record(vnode_t *vp, nvlist_t **nvlp, char **svcp, char **datap)
9774 {
9775 	int err;
9776 	char *stype, *val;
9777 	nvlist_t *nvl;
9778 	nvpair_t *curr;
9779 
9780 	if ((nvl = reparse_init()) == NULL)
9781 		return (-1);
9782 
9783 	if ((err = reparse_vnode_parse(vp, nvl)) != 0) {
9784 		reparse_free(nvl);
9785 		return (err);
9786 	}
9787 
9788 	curr = NULL;
9789 	while ((curr = nvlist_next_nvpair(nvl, curr)) != NULL) {
9790 		if ((stype = nvpair_name(curr)) == NULL) {
9791 			reparse_free(nvl);
9792 			return (-2);
9793 		}
9794 		if (strncasecmp(stype, "NFS", 3) == 0)
9795 			break;
9796 	}
9797 
9798 	if ((curr == NULL) ||
9799 	    (nvpair_value_string(curr, &val))) {
9800 		reparse_free(nvl);
9801 		return (-3);
9802 	}
9803 	*nvlp = nvl;
9804 	*svcp = stype;
9805 	*datap = val;
9806 	return (0);
9807 }
9808 
9809 int
9810 vn_is_nfs_reparse(vnode_t *vp, cred_t *cr)
9811 {
9812 	nvlist_t *nvl;
9813 	char *s, *d;
9814 
9815 	if (rfs4_no_referrals != 0)
9816 		return (B_FALSE);
9817 
9818 	if (vn_is_reparse(vp, cr, NULL) == B_FALSE)
9819 		return (B_FALSE);
9820 
9821 	if (vn_find_nfs_record(vp, &nvl, &s, &d) != 0)
9822 		return (B_FALSE);
9823 
9824 	reparse_free(nvl);
9825 
9826 	return (B_TRUE);
9827 }
9828 
9829 /*
9830  * There is a user-level copy of this routine in ref_subr.c.
9831  * Changes should be kept in sync.
9832  */
9833 static int
9834 nfs4_create_components(char *path, component4 *comp4)
9835 {
9836 	int slen, plen, ncomp;
9837 	char *ori_path, *nxtc, buf[MAXNAMELEN];
9838 
9839 	if (path == NULL)
9840 		return (0);
9841 
9842 	plen = strlen(path) + 1;	/* include the terminator */
9843 	ori_path = path;
9844 	ncomp = 0;
9845 
9846 	/* count number of components in the path */
9847 	for (nxtc = path; nxtc < ori_path + plen; nxtc++) {
9848 		if (*nxtc == '/' || *nxtc == '\0' || *nxtc == '\n') {
9849 			if ((slen = nxtc - path) == 0) {
9850 				path = nxtc + 1;
9851 				continue;
9852 			}
9853 
9854 			if (comp4 != NULL) {
9855 				bcopy(path, buf, slen);
9856 				buf[slen] = '\0';
9857 				(void) str_to_utf8(buf, &comp4[ncomp]);
9858 			}
9859 
9860 			ncomp++;	/* 1 valid component */
9861 			path = nxtc + 1;
9862 		}
9863 		if (*nxtc == '\0' || *nxtc == '\n')
9864 			break;
9865 	}
9866 
9867 	return (ncomp);
9868 }
9869 
9870 /*
9871  * There is a user-level copy of this routine in ref_subr.c.
9872  * Changes should be kept in sync.
9873  */
9874 static int
9875 make_pathname4(char *path, pathname4 *pathname)
9876 {
9877 	int ncomp;
9878 	component4 *comp4;
9879 
9880 	if (pathname == NULL)
9881 		return (0);
9882 
9883 	if (path == NULL) {
9884 		pathname->pathname4_val = NULL;
9885 		pathname->pathname4_len = 0;
9886 		return (0);
9887 	}
9888 
9889 	/* count number of components to alloc buffer */
9890 	if ((ncomp = nfs4_create_components(path, NULL)) == 0) {
9891 		pathname->pathname4_val = NULL;
9892 		pathname->pathname4_len = 0;
9893 		return (0);
9894 	}
9895 	comp4 = kmem_zalloc(ncomp * sizeof (component4), KM_SLEEP);
9896 
9897 	/* copy components into allocated buffer */
9898 	ncomp = nfs4_create_components(path, comp4);
9899 
9900 	pathname->pathname4_val = comp4;
9901 	pathname->pathname4_len = ncomp;
9902 
9903 	return (ncomp);
9904 }
9905 
9906 #define	xdr_fs_locations4 xdr_fattr4_fs_locations
9907 
9908 fs_locations4 *
9909 fetch_referral(vnode_t *vp, cred_t *cr)
9910 {
9911 	nvlist_t *nvl;
9912 	char *stype, *sdata;
9913 	fs_locations4 *result;
9914 	char buf[1024];
9915 	size_t bufsize;
9916 	XDR xdr;
9917 	int err;
9918 
9919 	/*
9920 	 * Check attrs to ensure it's a reparse point
9921 	 */
9922 	if (vn_is_reparse(vp, cr, NULL) == B_FALSE)
9923 		return (NULL);
9924 
9925 	/*
9926 	 * Look for an NFS record and get the type and data
9927 	 */
9928 	if (vn_find_nfs_record(vp, &nvl, &stype, &sdata) != 0)
9929 		return (NULL);
9930 
9931 	/*
9932 	 * With the type and data, upcall to get the referral
9933 	 */
9934 	bufsize = sizeof (buf);
9935 	bzero(buf, sizeof (buf));
9936 	err = reparse_kderef((const char *)stype, (const char *)sdata,
9937 	    buf, &bufsize);
9938 	reparse_free(nvl);
9939 
9940 	DTRACE_PROBE4(nfs4serv__func__referral__upcall,
9941 	    char *, stype, char *, sdata, char *, buf, int, err);
9942 	if (err) {
9943 		cmn_err(CE_NOTE,
9944 		    "reparsed daemon not running: unable to get referral (%d)",
9945 		    err);
9946 		return (NULL);
9947 	}
9948 
9949 	/*
9950 	 * We get an XDR'ed record back from the kderef call
9951 	 */
9952 	xdrmem_create(&xdr, buf, bufsize, XDR_DECODE);
9953 	result = kmem_alloc(sizeof (fs_locations4), KM_SLEEP);
9954 	err = xdr_fs_locations4(&xdr, result);
9955 	XDR_DESTROY(&xdr);
9956 	if (err != TRUE) {
9957 		DTRACE_PROBE1(nfs4serv__func__referral__upcall__xdrfail,
9958 		    int, err);
9959 		return (NULL);
9960 	}
9961 
9962 	/*
9963 	 * Look at path to recover fs_root, ignoring the leading '/'
9964 	 */
9965 	(void) make_pathname4(vp->v_path, &result->fs_root);
9966 
9967 	return (result);
9968 }
9969 
9970 char *
9971 build_symlink(vnode_t *vp, cred_t *cr, size_t *strsz)
9972 {
9973 	fs_locations4 *fsl;
9974 	fs_location4 *fs;
9975 	char *server, *path, *symbuf;
9976 	static char *prefix = "/net/";
9977 	int i, size, npaths;
9978 	uint_t len;
9979 
9980 	/* Get the referral */
9981 	if ((fsl = fetch_referral(vp, cr)) == NULL)
9982 		return (NULL);
9983 
9984 	/* Deal with only the first location and first server */
9985 	fs = &fsl->locations_val[0];
9986 	server = utf8_to_str(&fs->server_val[0], &len, NULL);
9987 	if (server == NULL) {
9988 		rfs4_free_fs_locations4(fsl);
9989 		kmem_free(fsl, sizeof (fs_locations4));
9990 		return (NULL);
9991 	}
9992 
9993 	/* Figure out size for "/net/" + host + /path/path/path + NULL */
9994 	size = strlen(prefix) + len;
9995 	for (i = 0; i < fs->rootpath.pathname4_len; i++)
9996 		size += fs->rootpath.pathname4_val[i].utf8string_len + 1;
9997 
9998 	/* Allocate the symlink buffer and fill it */
9999 	symbuf = kmem_zalloc(size, KM_SLEEP);
10000 	(void) strcat(symbuf, prefix);
10001 	(void) strcat(symbuf, server);
10002 	kmem_free(server, len);
10003 
10004 	npaths = 0;
10005 	for (i = 0; i < fs->rootpath.pathname4_len; i++) {
10006 		path = utf8_to_str(&fs->rootpath.pathname4_val[i], &len, NULL);
10007 		if (path == NULL)
10008 			continue;
10009 		(void) strcat(symbuf, "/");
10010 		(void) strcat(symbuf, path);
10011 		npaths++;
10012 		kmem_free(path, len);
10013 	}
10014 
10015 	rfs4_free_fs_locations4(fsl);
10016 	kmem_free(fsl, sizeof (fs_locations4));
10017 
10018 	if (strsz != NULL)
10019 		*strsz = size;
10020 	return (symbuf);
10021 }
10022 
10023 /*
10024  * Check to see if we have a downrev Solaris client, so that we
10025  * can send it a symlink instead of a referral.
10026  */
10027 int
10028 client_is_downrev(struct svc_req *req)
10029 {
10030 	struct sockaddr *ca;
10031 	rfs4_clntip_t *ci;
10032 	bool_t create = FALSE;
10033 	int is_downrev;
10034 
10035 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
10036 	ASSERT(ca);
10037 	ci = rfs4_find_clntip(ca, &create);
10038 	if (ci == NULL)
10039 		return (0);
10040 	is_downrev = ci->ri_no_referrals;
10041 	rfs4_dbe_rele(ci->ri_dbe);
10042 	return (is_downrev);
10043 }
10044