xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision 9f23cbd6cae82fd77edfad7173432fa8dccd0a95)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Rick Macklem at The University of Guelph.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  */
35 
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
38 
39 /*
40  * Rpc op calls, generally called from the vnode op calls or through the
41  * buffer cache, for NFS v2, 3 and 4.
42  * These do not normally make any changes to vnode arguments or use
43  * structures that might change between the VFS variants. The returned
44  * arguments are all at the end, after the NFSPROC_T *p one.
45  */
46 
47 #include "opt_inet6.h"
48 
49 #include <fs/nfs/nfsport.h>
50 #include <fs/nfsclient/nfs.h>
51 #include <sys/extattr.h>
52 #include <sys/sysctl.h>
53 #include <sys/taskqueue.h>
54 
55 SYSCTL_DECL(_vfs_nfs);
56 
57 static int	nfsignore_eexist = 0;
58 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
59     &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
60 
61 static int	nfscl_dssameconn = 0;
62 SYSCTL_INT(_vfs_nfs, OID_AUTO, dssameconn, CTLFLAG_RW,
63     &nfscl_dssameconn, 0, "Use same TCP connection to multiple DSs");
64 
65 static uint64_t nfs_maxcopyrange = SSIZE_MAX;
66 SYSCTL_U64(_vfs_nfs, OID_AUTO, maxcopyrange, CTLFLAG_RW,
67     &nfs_maxcopyrange, 0, "Max size of a Copy so RPC times reasonable");
68 
69 /*
70  * Global variables
71  */
72 extern struct nfsstatsv1 nfsstatsv1;
73 extern int nfs_numnfscbd;
74 extern struct timeval nfsboottime;
75 extern u_int32_t newnfs_false, newnfs_true;
76 extern nfstype nfsv34_type[9];
77 extern int nfsrv_useacl;
78 extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
79 extern int nfscl_debuglevel;
80 extern int nfs_pnfsiothreads;
81 extern u_long sb_max_adj;
82 NFSCLSTATEMUTEX;
83 int nfstest_outofseq = 0;
84 int nfscl_assumeposixlocks = 1;
85 int nfscl_enablecallb = 0;
86 short nfsv4_cbport = NFSV4_CBPORT;
87 int nfstest_openallsetattr = 0;
88 
89 #define	DIRHDSIZ	offsetof(struct dirent, d_name)
90 
91 /*
92  * nfscl_getsameserver() can return one of three values:
93  * NFSDSP_USETHISSESSION - Use this session for the DS.
94  * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
95  *     session.
96  * NFSDSP_NOTFOUND - No matching server was found.
97  */
98 enum nfsclds_state {
99 	NFSDSP_USETHISSESSION = 0,
100 	NFSDSP_SEQTHISSESSION = 1,
101 	NFSDSP_NOTFOUND = 2,
102 };
103 
104 /*
105  * Do a write RPC on a DS data file, using this structure for the arguments,
106  * so that this function can be executed by a separate kernel process.
107  */
108 struct nfsclwritedsdorpc {
109 	int			done;
110 	int			inprog;
111 	struct task		tsk;
112 	struct vnode		*vp;
113 	int			iomode;
114 	int			must_commit;
115 	nfsv4stateid_t		*stateidp;
116 	struct nfsclds		*dsp;
117 	uint64_t		off;
118 	int			len;
119 #ifdef notyet
120 	int			advise;
121 #endif
122 	struct nfsfh		*fhp;
123 	struct mbuf		*m;
124 	int			vers;
125 	int			minorvers;
126 	struct ucred		*cred;
127 	NFSPROC_T		*p;
128 	int			err;
129 };
130 
131 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
132     struct ucred *, NFSPROC_T *, struct nfsvattr *, int *);
133 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
134     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *);
135 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
136     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
137     int);
138 static int nfsrpc_deallocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
139     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *);
140 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
141     nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
142     struct nfsvattr *, struct nfsfh **, int *, int *);
143 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
144     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
145     NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
146     int *, int *);
147 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
148     struct nfscllockowner *, u_int64_t, u_int64_t,
149     u_int32_t, struct ucred *, NFSPROC_T *, int);
150 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
151     struct acl *, nfsv4stateid_t *);
152 static int nfsrpc_layouterror(struct nfsmount *, uint8_t *, int, uint64_t,
153     uint64_t, nfsv4stateid_t *, struct ucred *, NFSPROC_T *, uint32_t,
154     uint32_t, char *);
155 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
156     uint32_t, uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
157     struct ucred *, NFSPROC_T *);
158 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
159     struct sockaddr_in6 *, sa_family_t, int, int, struct nfsclds **,
160     NFSPROC_T *);
161 static void nfscl_initsessionslots(struct nfsclsession *);
162 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
163     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
164     struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
165     NFSPROC_T *);
166 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
167     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
168     struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
169     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
170 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
171     struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
172     struct ucred *, NFSPROC_T *);
173 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
174     nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
175     struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
176 static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
177     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
178     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
179 static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
180     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
181     struct ucred *, NFSPROC_T *);
182 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
183     struct nfsclds *, struct nfsclds **, uint32_t *);
184 static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
185     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
186     NFSPROC_T *);
187 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
188     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
189 #ifdef notyet
190 static int nfsio_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
191     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
192     NFSPROC_T *);
193 static int nfsrpc_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
194     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
195 #endif
196 static int nfsrpc_allocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
197     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *);
198 static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
199     uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
200 static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
201     NFSPROC_T *);
202 static int nfsrv_parselayoutget(struct nfsmount *, struct nfsrv_descript *,
203     nfsv4stateid_t *, int *, struct nfsclflayouthead *);
204 static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
205     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
206     struct nfscldeleg **, struct ucred *, NFSPROC_T *);
207 static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
208     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
209     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
210     struct nfsfh **, int *, int *, int *);
211 static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
212     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
213     struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
214     struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
215 static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
216     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
217     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
218     struct nfsfh **, int *, int *, int *, nfsv4stateid_t *,
219     int, int, int, int *, struct nfsclflayouthead *, int *);
220 static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
221     uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
222     struct nfsclflayouthead *, struct ucred *, NFSPROC_T *);
223 static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
224     int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
225     struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
226 static int nfsrpc_copyrpc(vnode_t, off_t, vnode_t, off_t, size_t *,
227     nfsv4stateid_t *, nfsv4stateid_t *, struct nfsvattr *, int *,
228     struct nfsvattr *, int *, bool, int *, struct ucred *, NFSPROC_T *);
229 static int nfsrpc_seekrpc(vnode_t, off_t *, nfsv4stateid_t *, bool *,
230     int, struct nfsvattr *, int *, struct ucred *);
231 static struct mbuf *nfsm_split(struct mbuf *, uint64_t);
232 static void nfscl_statfs(struct vnode *, struct ucred *, NFSPROC_T *);
233 
234 int nfs_pnfsio(task_fn_t *, void *);
235 
236 /*
237  * nfs null call from vfs.
238  */
239 int
240 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
241 {
242 	int error;
243 	struct nfsrv_descript nfsd, *nd = &nfsd;
244 
245 	NFSCL_REQSTART(nd, NFSPROC_NULL, vp, NULL);
246 	error = nfscl_request(nd, vp, p, cred);
247 	if (nd->nd_repstat && !error)
248 		error = nd->nd_repstat;
249 	m_freem(nd->nd_mrep);
250 	return (error);
251 }
252 
253 /*
254  * nfs access rpc op.
255  * For nfs version 3 and 4, use the access rpc to check accessibility. If file
256  * modes are changed on the server, accesses might still fail later.
257  */
258 int
259 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
260     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
261 {
262 	int error;
263 	u_int32_t mode, rmode;
264 
265 	if (acmode & VREAD)
266 		mode = NFSACCESS_READ;
267 	else
268 		mode = 0;
269 	if (vp->v_type == VDIR) {
270 		if (acmode & VWRITE)
271 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
272 				 NFSACCESS_DELETE);
273 		if (acmode & VEXEC)
274 			mode |= NFSACCESS_LOOKUP;
275 	} else {
276 		if (acmode & VWRITE)
277 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
278 		if (acmode & VEXEC)
279 			mode |= NFSACCESS_EXECUTE;
280 	}
281 
282 	/*
283 	 * Now, just call nfsrpc_accessrpc() to do the actual RPC.
284 	 */
285 	error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode);
286 
287 	/*
288 	 * The NFS V3 spec does not clarify whether or not
289 	 * the returned access bits can be a superset of
290 	 * the ones requested, so...
291 	 */
292 	if (!error && (rmode & mode) != mode)
293 		error = EACCES;
294 	return (error);
295 }
296 
297 /*
298  * The actual rpc, separated out for Darwin.
299  */
300 int
301 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
302     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep)
303 {
304 	u_int32_t *tl;
305 	u_int32_t supported, rmode;
306 	int error;
307 	struct nfsrv_descript nfsd, *nd = &nfsd;
308 	nfsattrbit_t attrbits;
309 	struct nfsmount *nmp;
310 	struct nfsnode *np;
311 
312 	*attrflagp = 0;
313 	supported = mode;
314 	nmp = VFSTONFS(vp->v_mount);
315 	np = VTONFS(vp);
316 	if ((nmp->nm_privflag & NFSMNTP_FAKEROOTFH) != 0 &&
317 	    nmp->nm_fhsize == 0) {
318 		/* Attempt to get the actual root file handle. */
319 		error = nfsrpc_getdirpath(nmp, NFSMNT_DIRPATH(nmp), cred, p);
320 		if (error != 0)
321 			return (EACCES);
322 		if (np->n_fhp->nfh_len == NFSX_FHMAX + 1)
323 			nfscl_statfs(vp, cred, p);
324 	}
325 	NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp, cred);
326 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
327 	*tl = txdr_unsigned(mode);
328 	if (nd->nd_flag & ND_NFSV4) {
329 		/*
330 		 * And do a Getattr op.
331 		 */
332 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
333 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
334 		NFSGETATTR_ATTRBIT(&attrbits);
335 		(void) nfsrv_putattrbit(nd, &attrbits);
336 	}
337 	error = nfscl_request(nd, vp, p, cred);
338 	if (error)
339 		return (error);
340 	if (nd->nd_flag & ND_NFSV3) {
341 		error = nfscl_postop_attr(nd, nap, attrflagp);
342 		if (error)
343 			goto nfsmout;
344 	}
345 	if (!nd->nd_repstat) {
346 		if (nd->nd_flag & ND_NFSV4) {
347 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
348 			supported = fxdr_unsigned(u_int32_t, *tl++);
349 		} else {
350 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
351 		}
352 		rmode = fxdr_unsigned(u_int32_t, *tl);
353 		if (nd->nd_flag & ND_NFSV4)
354 			error = nfscl_postop_attr(nd, nap, attrflagp);
355 
356 		/*
357 		 * It's not obvious what should be done about
358 		 * unsupported access modes. For now, be paranoid
359 		 * and clear the unsupported ones.
360 		 */
361 		rmode &= supported;
362 		*rmodep = rmode;
363 	} else
364 		error = nd->nd_repstat;
365 nfsmout:
366 	m_freem(nd->nd_mrep);
367 	return (error);
368 }
369 
370 /*
371  * nfs open rpc
372  */
373 int
374 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
375 {
376 	struct nfsclopen *op;
377 	struct nfscldeleg *dp;
378 	struct nfsfh *nfhp;
379 	struct nfsnode *np = VTONFS(vp);
380 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
381 	u_int32_t mode, clidrev;
382 	int ret, newone, error, expireret = 0, retrycnt;
383 
384 	/*
385 	 * For NFSv4, Open Ops are only done on Regular Files.
386 	 */
387 	if (vp->v_type != VREG)
388 		return (0);
389 	mode = 0;
390 	if (amode & FREAD)
391 		mode |= NFSV4OPEN_ACCESSREAD;
392 	if (amode & FWRITE)
393 		mode |= NFSV4OPEN_ACCESSWRITE;
394 	nfhp = np->n_fhp;
395 
396 	retrycnt = 0;
397 #ifdef notdef
398 { char name[100]; int namel;
399 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
400 bcopy(NFS4NODENAME(np->n_v4), name, namel);
401 name[namel] = '\0';
402 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
403 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
404 else printf(" fhl=0\n");
405 }
406 #endif
407 	do {
408 	    dp = NULL;
409 	    error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
410 		cred, p, NULL, &op, &newone, &ret, 1, true);
411 	    if (error) {
412 		return (error);
413 	    }
414 	    if (nmp->nm_clp != NULL)
415 		clidrev = nmp->nm_clp->nfsc_clientidrev;
416 	    else
417 		clidrev = 0;
418 	    if (ret == NFSCLOPEN_DOOPEN) {
419 		if (np->n_v4 != NULL) {
420 			/*
421 			 * For the first attempt, try and get a layout, if
422 			 * pNFS is enabled for the mount.
423 			 */
424 			if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
425 			    nfs_numnfscbd == 0 ||
426 			    (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
427 				error = nfsrpc_openrpc(nmp, vp,
428 				    np->n_v4->n4_data,
429 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
430 				    np->n_fhp->nfh_len, mode, op,
431 				    NFS4NODENAME(np->n_v4),
432 				    np->n_v4->n4_namelen,
433 				    &dp, 0, 0x0, cred, p, 0, 0);
434 			else
435 				error = nfsrpc_getopenlayout(nmp, vp,
436 				    np->n_v4->n4_data,
437 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
438 				    np->n_fhp->nfh_len, mode, op,
439 				    NFS4NODENAME(np->n_v4),
440 				    np->n_v4->n4_namelen, &dp, cred, p);
441 			if (dp != NULL) {
442 				NFSLOCKNODE(np);
443 				np->n_flag &= ~NDELEGMOD;
444 				/*
445 				 * Invalidate the attribute cache, so that
446 				 * attributes that pre-date the issue of a
447 				 * delegation are not cached, since the
448 				 * cached attributes will remain valid while
449 				 * the delegation is held.
450 				 */
451 				NFSINVALATTRCACHE(np);
452 				NFSUNLOCKNODE(np);
453 				(void) nfscl_deleg(nmp->nm_mountp,
454 				    op->nfso_own->nfsow_clp,
455 				    nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
456 			}
457 		} else {
458 			error = EIO;
459 		}
460 		newnfs_copyincred(cred, &op->nfso_cred);
461 	    } else if (ret == NFSCLOPEN_SETCRED)
462 		/*
463 		 * This is a new local open on a delegation. It needs
464 		 * to have credentials so that an open can be done
465 		 * against the server during recovery.
466 		 */
467 		newnfs_copyincred(cred, &op->nfso_cred);
468 
469 	    /*
470 	     * nfso_opencnt is the count of how many VOP_OPEN()s have
471 	     * been done on this Open successfully and a VOP_CLOSE()
472 	     * is expected for each of these.
473 	     * If error is non-zero, don't increment it, since the Open
474 	     * hasn't succeeded yet.
475 	     */
476 	    if (!error) {
477 		op->nfso_opencnt++;
478 		if (NFSHASNFSV4N(nmp) && NFSHASONEOPENOWN(nmp)) {
479 		    NFSLOCKNODE(np);
480 		    np->n_openstateid = op;
481 		    NFSUNLOCKNODE(np);
482 		}
483 	    }
484 	    nfscl_openrelease(nmp, op, error, newone);
485 	    if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
486 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
487 		error == NFSERR_BADSESSION) {
488 		(void) nfs_catnap(PZERO, error, "nfs_open");
489 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
490 		&& clidrev != 0) {
491 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
492 		retrycnt++;
493 	    }
494 	} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
495 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
496 	    error == NFSERR_BADSESSION ||
497 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
498 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
499 	if (error && retrycnt >= 4)
500 		error = EIO;
501 	return (error);
502 }
503 
504 /*
505  * the actual open rpc
506  */
507 int
508 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
509     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
510     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
511     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
512     int syscred, int recursed)
513 {
514 	u_int32_t *tl;
515 	struct nfsrv_descript nfsd, *nd = &nfsd;
516 	struct nfscldeleg *dp, *ndp = NULL;
517 	struct nfsvattr nfsva;
518 	u_int32_t rflags, deleg;
519 	nfsattrbit_t attrbits;
520 	int error, ret, acesize, limitby;
521 	struct nfsclsession *tsep;
522 
523 	dp = *dpp;
524 	*dpp = NULL;
525 	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0,
526 	    cred);
527 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
528 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
529 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
530 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
531 	tsep = nfsmnt_mdssession(nmp);
532 	*tl++ = tsep->nfsess_clientid.lval[0];
533 	*tl = tsep->nfsess_clientid.lval[1];
534 	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
535 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
536 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
537 	if (reclaim) {
538 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
539 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
540 		*tl = txdr_unsigned(delegtype);
541 	} else {
542 		if (dp != NULL) {
543 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
544 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
545 			if (NFSHASNFSV4N(nmp))
546 				*tl++ = 0;
547 			else
548 				*tl++ = dp->nfsdl_stateid.seqid;
549 			*tl++ = dp->nfsdl_stateid.other[0];
550 			*tl++ = dp->nfsdl_stateid.other[1];
551 			*tl = dp->nfsdl_stateid.other[2];
552 		} else {
553 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
554 		}
555 		(void) nfsm_strtom(nd, name, namelen);
556 	}
557 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
558 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
559 	NFSZERO_ATTRBIT(&attrbits);
560 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
561 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
562 	(void) nfsrv_putattrbit(nd, &attrbits);
563 	if (syscred)
564 		nd->nd_flag |= ND_USEGSSNAME;
565 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
566 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
567 	if (error)
568 		return (error);
569 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
570 	if (!nd->nd_repstat) {
571 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
572 		    6 * NFSX_UNSIGNED);
573 		op->nfso_stateid.seqid = *tl++;
574 		op->nfso_stateid.other[0] = *tl++;
575 		op->nfso_stateid.other[1] = *tl++;
576 		op->nfso_stateid.other[2] = *tl;
577 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
578 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
579 		if (error)
580 			goto nfsmout;
581 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
582 		deleg = fxdr_unsigned(u_int32_t, *tl);
583 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
584 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
585 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
586 			      NFSCLFLAGS_FIRSTDELEG))
587 				op->nfso_own->nfsow_clp->nfsc_flags |=
588 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
589 			ndp = malloc(
590 			    sizeof (struct nfscldeleg) + newfhlen,
591 			    M_NFSCLDELEG, M_WAITOK);
592 			LIST_INIT(&ndp->nfsdl_owner);
593 			LIST_INIT(&ndp->nfsdl_lock);
594 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
595 			ndp->nfsdl_fhlen = newfhlen;
596 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
597 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
598 			nfscl_lockinit(&ndp->nfsdl_rwlock);
599 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
600 			    NFSX_UNSIGNED);
601 			ndp->nfsdl_stateid.seqid = *tl++;
602 			ndp->nfsdl_stateid.other[0] = *tl++;
603 			ndp->nfsdl_stateid.other[1] = *tl++;
604 			ndp->nfsdl_stateid.other[2] = *tl++;
605 			ret = fxdr_unsigned(int, *tl);
606 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
607 				ndp->nfsdl_flags = NFSCLDL_WRITE;
608 				/*
609 				 * Indicates how much the file can grow.
610 				 */
611 				NFSM_DISSECT(tl, u_int32_t *,
612 				    3 * NFSX_UNSIGNED);
613 				limitby = fxdr_unsigned(int, *tl++);
614 				switch (limitby) {
615 				case NFSV4OPEN_LIMITSIZE:
616 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
617 					break;
618 				case NFSV4OPEN_LIMITBLOCKS:
619 					ndp->nfsdl_sizelimit =
620 					    fxdr_unsigned(u_int64_t, *tl++);
621 					ndp->nfsdl_sizelimit *=
622 					    fxdr_unsigned(u_int64_t, *tl);
623 					break;
624 				default:
625 					error = NFSERR_BADXDR;
626 					goto nfsmout;
627 				}
628 			} else {
629 				ndp->nfsdl_flags = NFSCLDL_READ;
630 			}
631 			if (ret)
632 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
633 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, false,
634 			    &ret, &acesize, p);
635 			if (error)
636 				goto nfsmout;
637 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
638 			error = NFSERR_BADXDR;
639 			goto nfsmout;
640 		}
641 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
642 		error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
643 		    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
644 		    NULL, NULL, NULL, p, cred);
645 		if (error)
646 			goto nfsmout;
647 		if (ndp != NULL) {
648 			ndp->nfsdl_change = nfsva.na_filerev;
649 			ndp->nfsdl_modtime = nfsva.na_mtime;
650 			ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
651 		}
652 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
653 		    do {
654 			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
655 			    cred, p);
656 			if (ret == NFSERR_DELAY)
657 			    (void) nfs_catnap(PZERO, ret, "nfs_open");
658 		    } while (ret == NFSERR_DELAY);
659 		    error = ret;
660 		}
661 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
662 		    nfscl_assumeposixlocks)
663 		    op->nfso_posixlock = 1;
664 		else
665 		    op->nfso_posixlock = 0;
666 
667 		/*
668 		 * If the server is handing out delegations, but we didn't
669 		 * get one because an OpenConfirm was required, try the
670 		 * Open again, to get a delegation. This is a harmless no-op,
671 		 * from a server's point of view.
672 		 */
673 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
674 		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
675 		    && !error && dp == NULL && ndp == NULL && !recursed) {
676 		    do {
677 			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
678 			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
679 			    cred, p, syscred, 1);
680 			if (ret == NFSERR_DELAY)
681 			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
682 		    } while (ret == NFSERR_DELAY);
683 		    if (ret) {
684 			if (ndp != NULL) {
685 				free(ndp, M_NFSCLDELEG);
686 				ndp = NULL;
687 			}
688 			if (ret == NFSERR_STALECLIENTID ||
689 			    ret == NFSERR_STALEDONTRECOVER ||
690 			    ret == NFSERR_BADSESSION)
691 				error = ret;
692 		    }
693 		}
694 	}
695 	if (nd->nd_repstat != 0 && error == 0)
696 		error = nd->nd_repstat;
697 	if (error == NFSERR_STALECLIENTID)
698 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
699 nfsmout:
700 	if (!error)
701 		*dpp = ndp;
702 	else if (ndp != NULL)
703 		free(ndp, M_NFSCLDELEG);
704 	m_freem(nd->nd_mrep);
705 	return (error);
706 }
707 
708 /*
709  * open downgrade rpc
710  */
711 int
712 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
713     struct ucred *cred, NFSPROC_T *p)
714 {
715 	u_int32_t *tl;
716 	struct nfsrv_descript nfsd, *nd = &nfsd;
717 	int error;
718 
719 	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp, cred);
720 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
721 	if (NFSHASNFSV4N(VFSTONFS(vp->v_mount)))
722 		*tl++ = 0;
723 	else
724 		*tl++ = op->nfso_stateid.seqid;
725 	*tl++ = op->nfso_stateid.other[0];
726 	*tl++ = op->nfso_stateid.other[1];
727 	*tl++ = op->nfso_stateid.other[2];
728 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
729 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
730 	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
731 	error = nfscl_request(nd, vp, p, cred);
732 	if (error)
733 		return (error);
734 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
735 	if (!nd->nd_repstat) {
736 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
737 		op->nfso_stateid.seqid = *tl++;
738 		op->nfso_stateid.other[0] = *tl++;
739 		op->nfso_stateid.other[1] = *tl++;
740 		op->nfso_stateid.other[2] = *tl;
741 	}
742 	if (nd->nd_repstat && error == 0)
743 		error = nd->nd_repstat;
744 	if (error == NFSERR_STALESTATEID)
745 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
746 nfsmout:
747 	m_freem(nd->nd_mrep);
748 	return (error);
749 }
750 
751 /*
752  * V4 Close operation.
753  */
754 int
755 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
756 {
757 	struct nfsclclient *clp;
758 	int error;
759 
760 	if (vp->v_type != VREG)
761 		return (0);
762 	if (doclose)
763 		error = nfscl_doclose(vp, &clp, p);
764 	else {
765 		error = nfscl_getclose(vp, &clp);
766 		if (error == 0)
767 			nfscl_clientrelease(clp);
768 	}
769 	return (error);
770 }
771 
772 /*
773  * Close the open.
774  */
775 int
776 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p,
777     bool loop_on_delayed, bool freeop)
778 {
779 	struct nfsrv_descript nfsd, *nd = &nfsd;
780 	struct nfscllockowner *lp, *nlp;
781 	struct nfscllock *lop, *nlop;
782 	struct ucred *tcred;
783 	u_int64_t off = 0, len = 0;
784 	u_int32_t type = NFSV4LOCKT_READ;
785 	int error, do_unlock, trycnt;
786 
787 	tcred = newnfs_getcred();
788 	newnfs_copycred(&op->nfso_cred, tcred);
789 	/*
790 	 * (Theoretically this could be done in the same
791 	 *  compound as the close, but having multiple
792 	 *  sequenced Ops in the same compound might be
793 	 *  too scary for some servers.)
794 	 */
795 	if (op->nfso_posixlock) {
796 		off = 0;
797 		len = NFS64BITSSET;
798 		type = NFSV4LOCKT_READ;
799 	}
800 
801 	/*
802 	 * Since this function is only called from VOP_INACTIVE(), no
803 	 * other thread will be manipulating this Open. As such, the
804 	 * lock lists are not being changed by other threads, so it should
805 	 * be safe to do this without locking.
806 	 */
807 	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
808 		do_unlock = 1;
809 		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
810 			if (op->nfso_posixlock == 0) {
811 				off = lop->nfslo_first;
812 				len = lop->nfslo_end - lop->nfslo_first;
813 				if (lop->nfslo_type == F_WRLCK)
814 					type = NFSV4LOCKT_WRITE;
815 				else
816 					type = NFSV4LOCKT_READ;
817 			}
818 			if (do_unlock) {
819 				trycnt = 0;
820 				do {
821 					error = nfsrpc_locku(nd, nmp, lp, off,
822 					    len, type, tcred, p, 0);
823 					if ((nd->nd_repstat == NFSERR_GRACE ||
824 					    nd->nd_repstat == NFSERR_DELAY) &&
825 					    error == 0)
826 						(void) nfs_catnap(PZERO,
827 						    (int)nd->nd_repstat,
828 						    "nfs_close");
829 				} while ((nd->nd_repstat == NFSERR_GRACE ||
830 				    nd->nd_repstat == NFSERR_DELAY) &&
831 				    error == 0 && trycnt++ < 5);
832 				if (op->nfso_posixlock)
833 					do_unlock = 0;
834 			}
835 			nfscl_freelock(lop, 0);
836 		}
837 		/*
838 		 * Do a ReleaseLockOwner.
839 		 * The lock owner name nfsl_owner may be used by other opens for
840 		 * other files but the lock_owner4 name that nfsrpc_rellockown()
841 		 * puts on the wire has the file handle for this file appended
842 		 * to it, so it can be done now.
843 		 */
844 		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
845 		    lp->nfsl_open->nfso_fhlen, tcred, p);
846 	}
847 
848 	/*
849 	 * There could be other Opens for different files on the same
850 	 * OpenOwner, so locking is required.
851 	 */
852 	NFSLOCKCLSTATE();
853 	nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
854 	NFSUNLOCKCLSTATE();
855 	do {
856 		error = nfscl_tryclose(op, tcred, nmp, p, loop_on_delayed);
857 		if (error == NFSERR_GRACE)
858 			(void) nfs_catnap(PZERO, error, "nfs_close");
859 	} while (error == NFSERR_GRACE);
860 	NFSLOCKCLSTATE();
861 	nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
862 
863 	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
864 		nfscl_freelockowner(lp, 0);
865 	if (freeop && error != NFSERR_DELAY)
866 		nfscl_freeopen(op, 0, true);
867 	NFSUNLOCKCLSTATE();
868 	NFSFREECRED(tcred);
869 	return (error);
870 }
871 
872 /*
873  * The actual Close RPC.
874  */
875 int
876 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
877     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
878     int syscred)
879 {
880 	u_int32_t *tl;
881 	int error;
882 
883 	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
884 	    op->nfso_fhlen, NULL, NULL, 0, 0, cred);
885 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
886 	if (NFSHASNFSV4N(nmp)) {
887 		*tl++ = 0;
888 		*tl++ = 0;
889 	} else {
890 		*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
891 		*tl++ = op->nfso_stateid.seqid;
892 	}
893 	*tl++ = op->nfso_stateid.other[0];
894 	*tl++ = op->nfso_stateid.other[1];
895 	*tl = op->nfso_stateid.other[2];
896 	if (syscred)
897 		nd->nd_flag |= ND_USEGSSNAME;
898 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
899 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
900 	if (error)
901 		return (error);
902 	if (!NFSHASNFSV4N(nmp))
903 		NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
904 	if (nd->nd_repstat == 0)
905 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
906 	error = nd->nd_repstat;
907 	if (!NFSHASNFSV4N(nmp) && error == NFSERR_STALESTATEID)
908 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
909 nfsmout:
910 	m_freem(nd->nd_mrep);
911 	return (error);
912 }
913 
914 /*
915  * V4 Open Confirm RPC.
916  */
917 int
918 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
919     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
920 {
921 	u_int32_t *tl;
922 	struct nfsrv_descript nfsd, *nd = &nfsd;
923 	struct nfsmount *nmp;
924 	int error;
925 
926 	nmp = VFSTONFS(vp->v_mount);
927 	if (NFSHASNFSV4N(nmp))
928 		return (0);		/* No confirmation for NFSv4.1. */
929 	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
930 	    0, 0, NULL);
931 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
932 	*tl++ = op->nfso_stateid.seqid;
933 	*tl++ = op->nfso_stateid.other[0];
934 	*tl++ = op->nfso_stateid.other[1];
935 	*tl++ = op->nfso_stateid.other[2];
936 	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
937 	error = nfscl_request(nd, vp, p, cred);
938 	if (error)
939 		return (error);
940 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
941 	if (!nd->nd_repstat) {
942 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
943 		op->nfso_stateid.seqid = *tl++;
944 		op->nfso_stateid.other[0] = *tl++;
945 		op->nfso_stateid.other[1] = *tl++;
946 		op->nfso_stateid.other[2] = *tl;
947 	}
948 	error = nd->nd_repstat;
949 	if (error == NFSERR_STALESTATEID)
950 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
951 nfsmout:
952 	m_freem(nd->nd_mrep);
953 	return (error);
954 }
955 
956 /*
957  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
958  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
959  */
960 int
961 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
962     bool *retokp, struct ucred *cred, NFSPROC_T *p)
963 {
964 	u_int32_t *tl;
965 	struct nfsrv_descript nfsd;
966 	struct nfsrv_descript *nd = &nfsd;
967 	u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
968 	u_short port;
969 	int error, isinet6 = 0, callblen;
970 	nfsquad_t confirm;
971 	static u_int32_t rev = 0;
972 	struct nfsclds *dsp, *odsp;
973 	struct in6_addr a6;
974 	struct nfsclsession *tsep;
975 	struct rpc_reconupcall recon;
976 	struct nfscl_reconarg *rcp;
977 
978 	if (nfsboottime.tv_sec == 0)
979 		NFSSETBOOTTIME(nfsboottime);
980 	if (NFSHASNFSV4N(nmp)) {
981 		error = NFSERR_BADSESSION;
982 		odsp = dsp = NULL;
983 		if (retokp != NULL) {
984 			NFSLOCKMNT(nmp);
985 			odsp = TAILQ_FIRST(&nmp->nm_sess);
986 			NFSUNLOCKMNT(nmp);
987 		}
988 		if (odsp != NULL) {
989 			/*
990 			 * When a session already exists, first try a
991 			 * CreateSession with the extant ClientID.
992 			 */
993 			dsp = malloc(sizeof(struct nfsclds) +
994 			    odsp->nfsclds_servownlen + 1, M_NFSCLDS,
995 			    M_WAITOK | M_ZERO);
996 			dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
997 			dsp->nfsclds_servownlen = odsp->nfsclds_servownlen;
998 			dsp->nfsclds_sess.nfsess_clientid =
999 			    odsp->nfsclds_sess.nfsess_clientid;
1000 			dsp->nfsclds_sess.nfsess_sequenceid =
1001 			    odsp->nfsclds_sess.nfsess_sequenceid + 1;
1002 			dsp->nfsclds_flags = odsp->nfsclds_flags;
1003 			if (dsp->nfsclds_servownlen > 0)
1004 				memcpy(dsp->nfsclds_serverown,
1005 				    odsp->nfsclds_serverown,
1006 				    dsp->nfsclds_servownlen + 1);
1007 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
1008 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
1009 			    NULL, MTX_DEF);
1010 			nfscl_initsessionslots(&dsp->nfsclds_sess);
1011 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
1012 			    &nmp->nm_sockreq, NULL,
1013 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
1014 			NFSCL_DEBUG(1, "create session for extant "
1015 			    "ClientID=%d\n", error);
1016 			if (error != 0) {
1017 				nfscl_freenfsclds(dsp);
1018 				dsp = NULL;
1019 				/*
1020 				 * If *retokp is true, return any error other
1021 				 * than NFSERR_STALECLIENTID,
1022 				 * NFSERR_BADSESSION or NFSERR_STALEDONTRECOVER
1023 				 * so that nfscl_recover() will not loop.
1024 				 */
1025 				if (*retokp)
1026 					return (NFSERR_IO);
1027 			} else
1028 				*retokp = true;
1029 		} else if (retokp != NULL && *retokp)
1030 			return (NFSERR_IO);
1031 		if (error != 0) {
1032 			/*
1033 			 * Either there was no previous session or the
1034 			 * CreateSession attempt failed, so...
1035 			 * do an ExchangeID followed by the CreateSession.
1036 			 */
1037 			clp->nfsc_rev = rev++;
1038 			error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq, 0,
1039 			    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp,
1040 			    cred, p);
1041 			NFSCL_DEBUG(1, "aft exch=%d\n", error);
1042 			if (error == 0)
1043 				error = nfsrpc_createsession(nmp,
1044 				    &dsp->nfsclds_sess, &nmp->nm_sockreq, NULL,
1045 				    dsp->nfsclds_sess.nfsess_sequenceid, 1,
1046 				    cred, p);
1047 			NFSCL_DEBUG(1, "aft createsess=%d\n", error);
1048 		}
1049 		if (error == 0) {
1050 			/*
1051 			 * If the session supports a backchannel, set up
1052 			 * the BindConnectionToSession call in the krpc
1053 			 * so that it is done on a reconnection.
1054 			 */
1055 			if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0) {
1056 				rcp = mem_alloc(sizeof(*rcp));
1057 				rcp->minorvers = nmp->nm_minorvers;
1058 				memcpy(rcp->sessionid,
1059 				    dsp->nfsclds_sess.nfsess_sessionid,
1060 				    NFSX_V4SESSIONID);
1061 				recon.call = nfsrpc_bindconnsess;
1062 				recon.arg = rcp;
1063 				CLNT_CONTROL(nmp->nm_client, CLSET_RECONUPCALL,
1064 				    &recon);
1065 			}
1066 
1067 			NFSLOCKMNT(nmp);
1068 			/*
1069 			 * The old sessions cannot be safely free'd
1070 			 * here, since they may still be used by
1071 			 * in-progress RPCs.
1072 			 */
1073 			tsep = NULL;
1074 			if (TAILQ_FIRST(&nmp->nm_sess) != NULL) {
1075 				/*
1076 				 * Mark the old session defunct.  Needed
1077 				 * when called from nfscl_hasexpired().
1078 				 */
1079 				tsep = NFSMNT_MDSSESSION(nmp);
1080 				tsep->nfsess_defunct = 1;
1081 			}
1082 			TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
1083 			    nfsclds_list);
1084 			/*
1085 			 * Wake up RPCs waiting for a slot on the
1086 			 * old session. These will then fail with
1087 			 * NFSERR_BADSESSION and be retried with the
1088 			 * new session by nfsv4_setsequence().
1089 			 * Also wakeup() processes waiting for the
1090 			 * new session.
1091 			 */
1092 			if (tsep != NULL)
1093 				wakeup(&tsep->nfsess_slots);
1094 			wakeup(&nmp->nm_sess);
1095 			NFSUNLOCKMNT(nmp);
1096 		} else if (dsp != NULL)
1097 			nfscl_freenfsclds(dsp);
1098 		if (error == 0 && reclaim == 0) {
1099 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
1100 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
1101 			if (error == NFSERR_COMPLETEALREADY ||
1102 			    error == NFSERR_NOTSUPP)
1103 				/* Ignore this error. */
1104 				error = 0;
1105 		}
1106 		return (error);
1107 	} else if (retokp != NULL && *retokp)
1108 		return (NFSERR_IO);
1109 	clp->nfsc_rev = rev++;
1110 
1111 	/*
1112 	 * Allocate a single session structure for NFSv4.0, because some of
1113 	 * the fields are used by NFSv4.0 although it doesn't do a session.
1114 	 */
1115 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
1116 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
1117 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
1118 	NFSLOCKMNT(nmp);
1119 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
1120 	tsep = NFSMNT_MDSSESSION(nmp);
1121 	NFSUNLOCKMNT(nmp);
1122 
1123 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0,
1124 	    NULL);
1125 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1126 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
1127 	*tl = txdr_unsigned(clp->nfsc_rev);
1128 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1129 
1130 	/*
1131 	 * set up the callback address
1132 	 */
1133 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1134 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
1135 	callblen = strlen(nfsv4_callbackaddr);
1136 	if (callblen == 0)
1137 		cp = nfscl_getmyip(nmp, &a6, &isinet6);
1138 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1139 	    (callblen > 0 || cp != NULL)) {
1140 		port = htons(nfsv4_cbport);
1141 		cp2 = (u_int8_t *)&port;
1142 #ifdef INET6
1143 		if ((callblen > 0 &&
1144 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1145 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1146 
1147 			(void) nfsm_strtom(nd, "tcp6", 4);
1148 			if (callblen == 0) {
1149 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1150 				ip6add = ip6buf;
1151 			} else {
1152 				ip6add = nfsv4_callbackaddr;
1153 			}
1154 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1155 			    ip6add, cp2[0], cp2[1]);
1156 		} else
1157 #endif
1158 		{
1159 			(void) nfsm_strtom(nd, "tcp", 3);
1160 			if (callblen == 0)
1161 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1162 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1163 				    cp[2], cp[3], cp2[0], cp2[1]);
1164 			else
1165 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1166 				    "%s.%d.%d", nfsv4_callbackaddr,
1167 				    cp2[0], cp2[1]);
1168 		}
1169 		(void) nfsm_strtom(nd, addr, strlen(addr));
1170 	} else {
1171 		(void) nfsm_strtom(nd, "tcp", 3);
1172 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1173 	}
1174 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1175 	*tl = txdr_unsigned(clp->nfsc_cbident);
1176 	nd->nd_flag |= ND_USEGSSNAME;
1177 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1178 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1179 	if (error)
1180 		return (error);
1181 	if (nd->nd_repstat == 0) {
1182 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1183 	    tsep->nfsess_clientid.lval[0] = *tl++;
1184 	    tsep->nfsess_clientid.lval[1] = *tl++;
1185 	    confirm.lval[0] = *tl++;
1186 	    confirm.lval[1] = *tl;
1187 	    m_freem(nd->nd_mrep);
1188 	    nd->nd_mrep = NULL;
1189 
1190 	    /*
1191 	     * and confirm it.
1192 	     */
1193 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1194 		NULL, 0, 0, NULL);
1195 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1196 	    *tl++ = tsep->nfsess_clientid.lval[0];
1197 	    *tl++ = tsep->nfsess_clientid.lval[1];
1198 	    *tl++ = confirm.lval[0];
1199 	    *tl = confirm.lval[1];
1200 	    nd->nd_flag |= ND_USEGSSNAME;
1201 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1202 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1203 	    if (error)
1204 		return (error);
1205 	    m_freem(nd->nd_mrep);
1206 	    nd->nd_mrep = NULL;
1207 	}
1208 	error = nd->nd_repstat;
1209 nfsmout:
1210 	m_freem(nd->nd_mrep);
1211 	return (error);
1212 }
1213 
1214 /*
1215  * nfs getattr call.
1216  */
1217 int
1218 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1219     struct nfsvattr *nap)
1220 {
1221 	struct nfsrv_descript nfsd, *nd = &nfsd;
1222 	int error;
1223 	nfsattrbit_t attrbits;
1224 	struct nfsnode *np;
1225 	struct nfsmount *nmp;
1226 
1227 	nmp = VFSTONFS(vp->v_mount);
1228 	np = VTONFS(vp);
1229 	if ((nmp->nm_privflag & NFSMNTP_FAKEROOTFH) != 0 &&
1230 	    nmp->nm_fhsize == 0) {
1231 		/* Attempt to get the actual root file handle. */
1232 		error = nfsrpc_getdirpath(nmp, NFSMNT_DIRPATH(nmp), cred, p);
1233 		if (error != 0)
1234 			return (EACCES);
1235 		if (np->n_fhp->nfh_len == NFSX_FHMAX + 1)
1236 			nfscl_statfs(vp, cred, p);
1237 	}
1238 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp, cred);
1239 	if (nd->nd_flag & ND_NFSV4) {
1240 		NFSGETATTR_ATTRBIT(&attrbits);
1241 		(void) nfsrv_putattrbit(nd, &attrbits);
1242 	}
1243 	error = nfscl_request(nd, vp, p, cred);
1244 	if (error)
1245 		return (error);
1246 	if (!nd->nd_repstat)
1247 		error = nfsm_loadattr(nd, nap);
1248 	else
1249 		error = nd->nd_repstat;
1250 	m_freem(nd->nd_mrep);
1251 	return (error);
1252 }
1253 
1254 /*
1255  * nfs getattr call with non-vnode arguments.
1256  */
1257 int
1258 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1259     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1260     uint32_t *leasep)
1261 {
1262 	struct nfsrv_descript nfsd, *nd = &nfsd;
1263 	int error, vers = NFS_VER2;
1264 	nfsattrbit_t attrbits;
1265 
1266 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0,
1267 	    cred);
1268 	if (nd->nd_flag & ND_NFSV4) {
1269 		vers = NFS_VER4;
1270 		NFSGETATTR_ATTRBIT(&attrbits);
1271 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1272 		(void) nfsrv_putattrbit(nd, &attrbits);
1273 	} else if (nd->nd_flag & ND_NFSV3) {
1274 		vers = NFS_VER3;
1275 	}
1276 	if (syscred)
1277 		nd->nd_flag |= ND_USEGSSNAME;
1278 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1279 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1280 	if (error)
1281 		return (error);
1282 	if (nd->nd_repstat == 0) {
1283 		if ((nd->nd_flag & ND_NFSV4) != 0)
1284 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1285 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1286 			    NULL, NULL);
1287 		else
1288 			error = nfsm_loadattr(nd, nap);
1289 	} else
1290 		error = nd->nd_repstat;
1291 	m_freem(nd->nd_mrep);
1292 	return (error);
1293 }
1294 
1295 /*
1296  * Do an nfs setattr operation.
1297  */
1298 int
1299 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1300     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp)
1301 {
1302 	int error, expireret = 0, openerr, retrycnt;
1303 	u_int32_t clidrev = 0, mode;
1304 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1305 	struct nfsfh *nfhp;
1306 	nfsv4stateid_t stateid;
1307 	void *lckp;
1308 
1309 	if (nmp->nm_clp != NULL)
1310 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1311 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1312 		mode = NFSV4OPEN_ACCESSWRITE;
1313 	else
1314 		mode = NFSV4OPEN_ACCESSREAD;
1315 	retrycnt = 0;
1316 	do {
1317 		lckp = NULL;
1318 		openerr = 1;
1319 		if (NFSHASNFSV4(nmp)) {
1320 			nfhp = VTONFS(vp)->n_fhp;
1321 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1322 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1323 			if (error && vp->v_type == VREG &&
1324 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1325 			     nfstest_openallsetattr)) {
1326 				/*
1327 				 * No Open stateid, so try and open the file
1328 				 * now.
1329 				 */
1330 				if (mode == NFSV4OPEN_ACCESSWRITE)
1331 					openerr = nfsrpc_open(vp, FWRITE, cred,
1332 					    p);
1333 				else
1334 					openerr = nfsrpc_open(vp, FREAD, cred,
1335 					    p);
1336 				if (!openerr)
1337 					(void) nfscl_getstateid(vp,
1338 					    nfhp->nfh_fh, nfhp->nfh_len,
1339 					    mode, 0, cred, p, &stateid, &lckp);
1340 			}
1341 		}
1342 		if (vap != NULL)
1343 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1344 			    rnap, attrflagp);
1345 		else
1346 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid);
1347 		if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1348 			NFSLOCKMNT(nmp);
1349 			nmp->nm_state |= NFSSTA_OPENMODE;
1350 			NFSUNLOCKMNT(nmp);
1351 		}
1352 		if (error == NFSERR_STALESTATEID)
1353 			nfscl_initiate_recovery(nmp->nm_clp);
1354 		if (lckp != NULL)
1355 			nfscl_lockderef(lckp);
1356 		if (!openerr)
1357 			(void) nfsrpc_close(vp, 0, p);
1358 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1359 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1360 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1361 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1362 		} else if ((error == NFSERR_EXPIRED ||
1363 		    ((!NFSHASINT(nmp) || !NFSHASNFSV4N(nmp)) &&
1364 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
1365 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1366 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp) &&
1367 		    NFSHASNFSV4N(nmp)) {
1368 			error = EIO;
1369 		}
1370 		retrycnt++;
1371 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1372 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1373 	    error == NFSERR_BADSESSION ||
1374 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1375 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1376 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1377 	    (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1378 	     retrycnt < 4));
1379 	if (error && retrycnt >= 4)
1380 		error = EIO;
1381 	return (error);
1382 }
1383 
1384 static int
1385 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1386     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1387     struct nfsvattr *rnap, int *attrflagp)
1388 {
1389 	u_int32_t *tl;
1390 	struct nfsrv_descript nfsd, *nd = &nfsd;
1391 	int error;
1392 	nfsattrbit_t attrbits;
1393 
1394 	*attrflagp = 0;
1395 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp, cred);
1396 	if (nd->nd_flag & ND_NFSV4)
1397 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1398 	vap->va_type = vp->v_type;
1399 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1400 	if (nd->nd_flag & ND_NFSV3) {
1401 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1402 		*tl = newnfs_false;
1403 	} else if (nd->nd_flag & ND_NFSV4) {
1404 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1405 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1406 		NFSGETATTR_ATTRBIT(&attrbits);
1407 		(void) nfsrv_putattrbit(nd, &attrbits);
1408 	}
1409 	error = nfscl_request(nd, vp, p, cred);
1410 	if (error)
1411 		return (error);
1412 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1413 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, NULL);
1414 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1415 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1416 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1417 		error = nfscl_postop_attr(nd, rnap, attrflagp);
1418 	m_freem(nd->nd_mrep);
1419 	if (nd->nd_repstat && !error)
1420 		error = nd->nd_repstat;
1421 	return (error);
1422 }
1423 
1424 /*
1425  * nfs lookup rpc
1426  */
1427 int
1428 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1429     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1430     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, uint32_t openmode)
1431 {
1432 	uint32_t deleg, rflags, *tl;
1433 	struct nfsrv_descript nfsd, *nd = &nfsd;
1434 	struct nfsmount *nmp;
1435 	struct nfsnode *np;
1436 	struct nfsfh *nfhp;
1437 	nfsattrbit_t attrbits;
1438 	int error = 0, lookupp = 0, newone, ret, retop;
1439 	uint8_t own[NFSV4CL_LOCKNAMELEN];
1440 	struct nfsclopen *op;
1441 	struct nfscldeleg *ndp;
1442 	nfsv4stateid_t stateid;
1443 
1444 	*attrflagp = 0;
1445 	*dattrflagp = 0;
1446 	if (dvp->v_type != VDIR)
1447 		return (ENOTDIR);
1448 	nmp = VFSTONFS(dvp->v_mount);
1449 	if (len > NFS_MAXNAMLEN)
1450 		return (ENAMETOOLONG);
1451 	if (NFSHASNFSV4(nmp) && len == 1 &&
1452 		name[0] == '.') {
1453 		/*
1454 		 * Just return the current dir's fh.
1455 		 */
1456 		np = VTONFS(dvp);
1457 		nfhp = malloc(sizeof (struct nfsfh) +
1458 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1459 		nfhp->nfh_len = np->n_fhp->nfh_len;
1460 		NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1461 		*nfhpp = nfhp;
1462 		return (0);
1463 	}
1464 	if (NFSHASNFSV4(nmp) && len == 2 &&
1465 		name[0] == '.' && name[1] == '.') {
1466 		lookupp = 1;
1467 		openmode = 0;
1468 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp, cred);
1469 	} else if (openmode != 0) {
1470 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPOPEN, dvp, cred);
1471 		nfsm_strtom(nd, name, len);
1472 	} else {
1473 		NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp, cred);
1474 		(void) nfsm_strtom(nd, name, len);
1475 	}
1476 	if (nd->nd_flag & ND_NFSV4) {
1477 		NFSGETATTR_ATTRBIT(&attrbits);
1478 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1479 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1480 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1481 		(void) nfsrv_putattrbit(nd, &attrbits);
1482 		if (openmode != 0) {
1483 			/* Test for a VREG file. */
1484 			NFSZERO_ATTRBIT(&attrbits);
1485 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
1486 			NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
1487 			*tl = txdr_unsigned(NFSV4OP_VERIFY);
1488 			nfsrv_putattrbit(nd, &attrbits);
1489 			NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1490 			*tl++ = txdr_unsigned(NFSX_UNSIGNED);
1491 			*tl = vtonfsv34_type(VREG);
1492 
1493 			/* Attempt the Open for VREG. */
1494 			nfscl_filllockowner(NULL, own, F_POSIX);
1495 			NFSM_BUILD(tl, uint32_t *, 6 * NFSX_UNSIGNED);
1496 			*tl++ = txdr_unsigned(NFSV4OP_OPEN);
1497 			*tl++ = 0;		/* seqid, ignored. */
1498 			*tl++ = txdr_unsigned(openmode);
1499 			*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
1500 			*tl++ = 0;		/* ClientID, ignored. */
1501 			*tl = 0;
1502 			nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN);
1503 			NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1504 			*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
1505 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMFH);
1506 		}
1507 	}
1508 	error = nfscl_request(nd, dvp, p, cred);
1509 	if (error)
1510 		return (error);
1511 	ndp = NULL;
1512 	if (nd->nd_repstat) {
1513 		/*
1514 		 * When an NFSv4 Lookupp returns ENOENT, it means that
1515 		 * the lookup is at the root of an fs, so return this dir.
1516 		 */
1517 		if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1518 		    np = VTONFS(dvp);
1519 		    nfhp = malloc(sizeof (struct nfsfh) +
1520 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1521 		    nfhp->nfh_len = np->n_fhp->nfh_len;
1522 		    NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1523 		    *nfhpp = nfhp;
1524 		    m_freem(nd->nd_mrep);
1525 		    return (0);
1526 		}
1527 		if (nd->nd_flag & ND_NFSV3)
1528 		    error = nfscl_postop_attr(nd, dnap, dattrflagp);
1529 		else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1530 		    ND_NFSV4) {
1531 			/* Load the directory attributes. */
1532 			error = nfsm_loadattr(nd, dnap);
1533 			if (error != 0)
1534 				goto nfsmout;
1535 			*dattrflagp = 1;
1536 		}
1537 		/* Check Lookup operation reply status. */
1538 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1539 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1540 			if (*++tl != 0)
1541 				goto nfsmout;
1542 		}
1543 		/* Look for GetFH reply. */
1544 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1545 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1546 			if (*++tl != 0)
1547 				goto nfsmout;
1548 			error = nfsm_getfh(nd, nfhpp);
1549 			if (error)
1550 				goto nfsmout;
1551 		}
1552 		/* Look for Getattr reply. */
1553 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1554 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1555 			if (*++tl != 0)
1556 				goto nfsmout;
1557 			error = nfsm_loadattr(nd, nap);
1558 			if (error == 0) {
1559 				/*
1560 				 * We have now successfully completed the
1561 				 * lookup, so set nd_repstat to 0.
1562 				 */
1563 				nd->nd_repstat = 0;
1564 				*attrflagp = 1;
1565 			}
1566 		}
1567 		goto nfsmout;
1568 	}
1569 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1570 		/* Load the directory attributes. */
1571 		error = nfsm_loadattr(nd, dnap);
1572 		if (error != 0)
1573 			goto nfsmout;
1574 		*dattrflagp = 1;
1575 		/* Skip over the Lookup and GetFH operation status values. */
1576 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1577 	}
1578 	error = nfsm_getfh(nd, nfhpp);
1579 	if (error)
1580 		goto nfsmout;
1581 
1582 	error = nfscl_postop_attr(nd, nap, attrflagp);
1583 	if (openmode != 0 && error == 0) {
1584 		NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID +
1585 		    10 * NFSX_UNSIGNED);
1586 		tl += 4;	/* Skip over Verify+Open status. */
1587 		stateid.seqid = *tl++;
1588 		stateid.other[0] = *tl++;
1589 		stateid.other[1] = *tl++;
1590 		stateid.other[2] = *tl;
1591 		rflags = fxdr_unsigned(uint32_t, *(tl + 6));
1592 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1593 		if (error != 0)
1594 			goto nfsmout;
1595 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
1596 		deleg = fxdr_unsigned(uint32_t, *tl);
1597 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
1598 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
1599 			/*
1600 			 * Just need to fill in the fields used by
1601 			 * nfscl_trydelegreturn().
1602 			 * Mark the mount point as acquiring
1603 			 * delegations, so NFSPROC_LOOKUPOPEN will
1604 			 * no longer be done.
1605 			 */
1606 			NFSLOCKMNT(nmp);
1607 			nmp->nm_privflag |= NFSMNTP_DELEGISSUED;
1608 			NFSUNLOCKMNT(nmp);
1609 			ndp = malloc(sizeof(struct nfscldeleg) +
1610 			    (*nfhpp)->nfh_len, M_NFSCLDELEG, M_WAITOK);
1611 			ndp->nfsdl_fhlen = (*nfhpp)->nfh_len;
1612 			NFSBCOPY((*nfhpp)->nfh_fh, ndp->nfsdl_fh,
1613 			    ndp->nfsdl_fhlen);
1614 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
1615 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
1616 			ndp->nfsdl_stateid.seqid = *tl++;
1617 			ndp->nfsdl_stateid.other[0] = *tl++;
1618 			ndp->nfsdl_stateid.other[1] = *tl++;
1619 			ndp->nfsdl_stateid.other[2] = *tl++;
1620 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
1621 			error = NFSERR_BADXDR;
1622 			goto nfsmout;
1623 		}
1624 		ret = nfscl_open(dvp, (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len,
1625 		    openmode, 0, cred, p, NULL, &op, &newone, &retop, 1, true);
1626 		if (ret != 0)
1627 			goto nfsmout;
1628 		if (newone != 0) {
1629 			op->nfso_stateid.seqid = stateid.seqid;
1630 			op->nfso_stateid.other[0] = stateid.other[0];
1631 			op->nfso_stateid.other[1] = stateid.other[1];
1632 			op->nfso_stateid.other[2] = stateid.other[2];
1633 			op->nfso_mode = openmode;
1634 		} else {
1635 			op->nfso_stateid.seqid = stateid.seqid;
1636 			if (retop == NFSCLOPEN_DOOPEN)
1637 				op->nfso_mode |= openmode;
1638 		}
1639 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
1640 		    nfscl_assumeposixlocks)
1641 			op->nfso_posixlock = 1;
1642 		else
1643 			op->nfso_posixlock = 0;
1644 		nfscl_openrelease(nmp, op, 0, 0);
1645 		if (ndp != NULL) {
1646 			/*
1647 			 * Since we do not have the vnode, we
1648 			 * cannot invalidate cached attributes.
1649 			 * Just return the delegation.
1650 			 */
1651 			nfscl_trydelegreturn(ndp, cred, nmp, p);
1652 		}
1653 	}
1654 	if ((nd->nd_flag & ND_NFSV3) && !error)
1655 		error = nfscl_postop_attr(nd, dnap, dattrflagp);
1656 nfsmout:
1657 	m_freem(nd->nd_mrep);
1658 	if (!error && nd->nd_repstat)
1659 		error = nd->nd_repstat;
1660 	free(ndp, M_NFSCLDELEG);
1661 	return (error);
1662 }
1663 
1664 /*
1665  * Do a readlink rpc.
1666  */
1667 int
1668 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1669     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
1670 {
1671 	u_int32_t *tl;
1672 	struct nfsrv_descript nfsd, *nd = &nfsd;
1673 	struct nfsnode *np = VTONFS(vp);
1674 	nfsattrbit_t attrbits;
1675 	int error, len, cangetattr = 1;
1676 
1677 	*attrflagp = 0;
1678 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp, cred);
1679 	if (nd->nd_flag & ND_NFSV4) {
1680 		/*
1681 		 * And do a Getattr op.
1682 		 */
1683 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1684 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1685 		NFSGETATTR_ATTRBIT(&attrbits);
1686 		(void) nfsrv_putattrbit(nd, &attrbits);
1687 	}
1688 	error = nfscl_request(nd, vp, p, cred);
1689 	if (error)
1690 		return (error);
1691 	if (nd->nd_flag & ND_NFSV3)
1692 		error = nfscl_postop_attr(nd, nap, attrflagp);
1693 	if (!nd->nd_repstat && !error) {
1694 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1695 		/*
1696 		 * This seems weird to me, but must have been added to
1697 		 * FreeBSD for some reason. The only thing I can think of
1698 		 * is that there was/is some server that replies with
1699 		 * more link data than it should?
1700 		 */
1701 		if (len == NFS_MAXPATHLEN) {
1702 			NFSLOCKNODE(np);
1703 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1704 				len = np->n_size;
1705 				cangetattr = 0;
1706 			}
1707 			NFSUNLOCKNODE(np);
1708 		}
1709 		error = nfsm_mbufuio(nd, uiop, len);
1710 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1711 			error = nfscl_postop_attr(nd, nap, attrflagp);
1712 	}
1713 	if (nd->nd_repstat && !error)
1714 		error = nd->nd_repstat;
1715 nfsmout:
1716 	m_freem(nd->nd_mrep);
1717 	return (error);
1718 }
1719 
1720 /*
1721  * Read operation.
1722  */
1723 int
1724 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1725     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
1726 {
1727 	int error, expireret = 0, retrycnt;
1728 	u_int32_t clidrev = 0;
1729 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1730 	struct nfsnode *np = VTONFS(vp);
1731 	struct ucred *newcred;
1732 	struct nfsfh *nfhp = NULL;
1733 	nfsv4stateid_t stateid;
1734 	void *lckp;
1735 
1736 	if (nmp->nm_clp != NULL)
1737 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1738 	newcred = cred;
1739 	if (NFSHASNFSV4(nmp)) {
1740 		nfhp = np->n_fhp;
1741 		newcred = NFSNEWCRED(cred);
1742 	}
1743 	retrycnt = 0;
1744 	do {
1745 		lckp = NULL;
1746 		if (NFSHASNFSV4(nmp))
1747 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1748 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1749 			    &lckp);
1750 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1751 		    attrflagp);
1752 		if (error == NFSERR_OPENMODE) {
1753 			NFSLOCKMNT(nmp);
1754 			nmp->nm_state |= NFSSTA_OPENMODE;
1755 			NFSUNLOCKMNT(nmp);
1756 		}
1757 		if (error == NFSERR_STALESTATEID)
1758 			nfscl_initiate_recovery(nmp->nm_clp);
1759 		if (lckp != NULL)
1760 			nfscl_lockderef(lckp);
1761 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1762 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1763 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1764 			(void) nfs_catnap(PZERO, error, "nfs_read");
1765 		} else if ((error == NFSERR_EXPIRED ||
1766 		    ((!NFSHASINT(nmp) || !NFSHASNFSV4N(nmp)) &&
1767 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
1768 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1769 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp) &&
1770 		    NFSHASNFSV4N(nmp)) {
1771 			error = EIO;
1772 		}
1773 		retrycnt++;
1774 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1775 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1776 	    error == NFSERR_BADSESSION ||
1777 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1778 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1779 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1780 	    (error == NFSERR_OPENMODE && retrycnt < 4));
1781 	if (error && retrycnt >= 4)
1782 		error = EIO;
1783 	if (NFSHASNFSV4(nmp))
1784 		NFSFREECRED(newcred);
1785 	return (error);
1786 }
1787 
1788 /*
1789  * The actual read RPC.
1790  */
1791 static int
1792 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1793     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1794     int *attrflagp)
1795 {
1796 	u_int32_t *tl;
1797 	int error = 0, len, retlen, tsiz, eof = 0;
1798 	struct nfsrv_descript nfsd;
1799 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1800 	struct nfsrv_descript *nd = &nfsd;
1801 	int rsize;
1802 	off_t tmp_off;
1803 
1804 	*attrflagp = 0;
1805 	tsiz = uiop->uio_resid;
1806 	tmp_off = uiop->uio_offset + tsiz;
1807 	NFSLOCKMNT(nmp);
1808 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1809 		NFSUNLOCKMNT(nmp);
1810 		return (EFBIG);
1811 	}
1812 	rsize = nmp->nm_rsize;
1813 	NFSUNLOCKMNT(nmp);
1814 	nd->nd_mrep = NULL;
1815 	while (tsiz > 0) {
1816 		*attrflagp = 0;
1817 		len = (tsiz > rsize) ? rsize : tsiz;
1818 		NFSCL_REQSTART(nd, NFSPROC_READ, vp, cred);
1819 		if (nd->nd_flag & ND_NFSV4)
1820 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1821 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1822 		if (nd->nd_flag & ND_NFSV2) {
1823 			*tl++ = txdr_unsigned(uiop->uio_offset);
1824 			*tl++ = txdr_unsigned(len);
1825 			*tl = 0;
1826 		} else {
1827 			txdr_hyper(uiop->uio_offset, tl);
1828 			*(tl + 2) = txdr_unsigned(len);
1829 		}
1830 		/*
1831 		 * Since I can't do a Getattr for NFSv4 for Write, there
1832 		 * doesn't seem any point in doing one here, either.
1833 		 * (See the comment in nfsrpc_writerpc() for more info.)
1834 		 */
1835 		error = nfscl_request(nd, vp, p, cred);
1836 		if (error)
1837 			return (error);
1838 		if (nd->nd_flag & ND_NFSV3) {
1839 			error = nfscl_postop_attr(nd, nap, attrflagp);
1840 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1841 			error = nfsm_loadattr(nd, nap);
1842 			if (!error)
1843 				*attrflagp = 1;
1844 		}
1845 		if (nd->nd_repstat || error) {
1846 			if (!error)
1847 				error = nd->nd_repstat;
1848 			goto nfsmout;
1849 		}
1850 		if (nd->nd_flag & ND_NFSV3) {
1851 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1852 			eof = fxdr_unsigned(int, *(tl + 1));
1853 		} else if (nd->nd_flag & ND_NFSV4) {
1854 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1855 			eof = fxdr_unsigned(int, *tl);
1856 		}
1857 		NFSM_STRSIZ(retlen, len);
1858 		error = nfsm_mbufuio(nd, uiop, retlen);
1859 		if (error)
1860 			goto nfsmout;
1861 		m_freem(nd->nd_mrep);
1862 		nd->nd_mrep = NULL;
1863 		tsiz -= retlen;
1864 		if (!(nd->nd_flag & ND_NFSV2)) {
1865 			if (eof || retlen == 0)
1866 				tsiz = 0;
1867 		} else if (retlen < len)
1868 			tsiz = 0;
1869 	}
1870 	return (0);
1871 nfsmout:
1872 	if (nd->nd_mrep != NULL)
1873 		m_freem(nd->nd_mrep);
1874 	return (error);
1875 }
1876 
1877 /*
1878  * nfs write operation
1879  * When called_from_strategy != 0, it should return EIO for an error that
1880  * indicates recovery is in progress, so that the buffer will be left
1881  * dirty and be written back to the server later. If it loops around,
1882  * the recovery thread could get stuck waiting for the buffer and recovery
1883  * will then deadlock.
1884  */
1885 int
1886 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1887     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1888     int called_from_strategy, int ioflag)
1889 {
1890 	int error, expireret = 0, retrycnt, nostateid;
1891 	u_int32_t clidrev = 0;
1892 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1893 	struct nfsnode *np = VTONFS(vp);
1894 	struct ucred *newcred;
1895 	struct nfsfh *nfhp = NULL;
1896 	nfsv4stateid_t stateid;
1897 	void *lckp;
1898 
1899 	KASSERT(*must_commit >= 0 && *must_commit <= 2,
1900 	    ("nfsrpc_write: must_commit out of range=%d", *must_commit));
1901 	if (nmp->nm_clp != NULL)
1902 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1903 	newcred = cred;
1904 	if (NFSHASNFSV4(nmp)) {
1905 		newcred = NFSNEWCRED(cred);
1906 		nfhp = np->n_fhp;
1907 	}
1908 	retrycnt = 0;
1909 	do {
1910 		lckp = NULL;
1911 		nostateid = 0;
1912 		if (NFSHASNFSV4(nmp)) {
1913 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1914 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1915 			    &lckp);
1916 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1917 			    stateid.other[2] == 0) {
1918 				nostateid = 1;
1919 				NFSCL_DEBUG(1, "stateid0 in write\n");
1920 			}
1921 		}
1922 
1923 		/*
1924 		 * If there is no stateid for NFSv4, it means this is an
1925 		 * extraneous write after close. Basically a poorly
1926 		 * implemented buffer cache. Just don't do the write.
1927 		 */
1928 		if (nostateid)
1929 			error = 0;
1930 		else
1931 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1932 			    newcred, &stateid, p, nap, attrflagp, ioflag);
1933 		if (error == NFSERR_STALESTATEID)
1934 			nfscl_initiate_recovery(nmp->nm_clp);
1935 		if (lckp != NULL)
1936 			nfscl_lockderef(lckp);
1937 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1938 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1939 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1940 			(void) nfs_catnap(PZERO, error, "nfs_write");
1941 		} else if ((error == NFSERR_EXPIRED ||
1942 		    ((!NFSHASINT(nmp) || !NFSHASNFSV4N(nmp)) &&
1943 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
1944 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1945 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp) &&
1946 		    NFSHASNFSV4N(nmp)) {
1947 			error = EIO;
1948 		}
1949 		retrycnt++;
1950 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1951 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1952 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1953 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1954 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1955 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1956 	if (error != 0 && (retrycnt >= 4 ||
1957 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1958 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1959 		error = EIO;
1960 	if (NFSHASNFSV4(nmp))
1961 		NFSFREECRED(newcred);
1962 	return (error);
1963 }
1964 
1965 /*
1966  * The actual write RPC.
1967  */
1968 static int
1969 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1970     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1971     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, int ioflag)
1972 {
1973 	u_int32_t *tl;
1974 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1975 	struct nfsnode *np = VTONFS(vp);
1976 	int error = 0, len, rlen, commit, committed = NFSWRITE_FILESYNC;
1977 	int wccflag = 0;
1978 	int32_t backup;
1979 	struct nfsrv_descript *nd;
1980 	nfsattrbit_t attrbits;
1981 	uint64_t tmp_off;
1982 	ssize_t tsiz, wsize;
1983 	bool do_append;
1984 
1985 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1986 	*attrflagp = 0;
1987 	tsiz = uiop->uio_resid;
1988 	tmp_off = uiop->uio_offset + tsiz;
1989 	NFSLOCKMNT(nmp);
1990 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1991 		NFSUNLOCKMNT(nmp);
1992 		return (EFBIG);
1993 	}
1994 	wsize = nmp->nm_wsize;
1995 	do_append = false;
1996 	if ((ioflag & IO_APPEND) != 0 && NFSHASNFSV4(nmp) && !NFSHASPNFS(nmp))
1997 		do_append = true;
1998 	NFSUNLOCKMNT(nmp);
1999 	nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK);
2000 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
2001 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
2002 	while (tsiz > 0) {
2003 		*attrflagp = 0;
2004 		len = (tsiz > wsize) ? wsize : tsiz;
2005 		if (do_append)
2006 			NFSCL_REQSTART(nd, NFSPROC_APPENDWRITE, vp, cred);
2007 		else
2008 			NFSCL_REQSTART(nd, NFSPROC_WRITE, vp, cred);
2009 		if (nd->nd_flag & ND_NFSV4) {
2010 			if (do_append) {
2011 				NFSZERO_ATTRBIT(&attrbits);
2012 				NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
2013 				nfsrv_putattrbit(nd, &attrbits);
2014 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED +
2015 				    NFSX_HYPER);
2016 				*tl++ = txdr_unsigned(NFSX_HYPER);
2017 				txdr_hyper(uiop->uio_offset, tl); tl += 2;
2018 				*tl = txdr_unsigned(NFSV4OP_WRITE);
2019 			}
2020 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
2021 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
2022 			txdr_hyper(uiop->uio_offset, tl);
2023 			tl += 2;
2024 			*tl++ = txdr_unsigned(*iomode);
2025 			*tl = txdr_unsigned(len);
2026 		} else if (nd->nd_flag & ND_NFSV3) {
2027 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
2028 			txdr_hyper(uiop->uio_offset, tl);
2029 			tl += 2;
2030 			*tl++ = txdr_unsigned(len);
2031 			*tl++ = txdr_unsigned(*iomode);
2032 			*tl = txdr_unsigned(len);
2033 		} else {
2034 			u_int32_t x;
2035 
2036 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2037 			/*
2038 			 * Not sure why someone changed this, since the
2039 			 * RFC clearly states that "beginoffset" and
2040 			 * "totalcount" are ignored, but it wouldn't
2041 			 * surprise me if there's a busted server out there.
2042 			 */
2043 			/* Set both "begin" and "current" to non-garbage. */
2044 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
2045 			*tl++ = x;      /* "begin offset" */
2046 			*tl++ = x;      /* "current offset" */
2047 			x = txdr_unsigned(len);
2048 			*tl++ = x;      /* total to this offset */
2049 			*tl = x;        /* size of this write */
2050 		}
2051 		nfsm_uiombuf(nd, uiop, len);
2052 		/*
2053 		 * Although it is tempting to do a normal Getattr Op in the
2054 		 * NFSv4 compound, the result can be a nearly hung client
2055 		 * system if the Getattr asks for Owner and/or OwnerGroup.
2056 		 * It occurs when the client can't map either the Owner or
2057 		 * Owner_group name in the Getattr reply to a uid/gid. When
2058 		 * there is a cache miss, the kernel does an upcall to the
2059 		 * nfsuserd. Then, it can try and read the local /etc/passwd
2060 		 * or /etc/group file. It can then block in getnewbuf(),
2061 		 * waiting for dirty writes to be pushed to the NFS server.
2062 		 * The only reason this doesn't result in a complete
2063 		 * deadlock, is that the upcall times out and allows
2064 		 * the write to complete. However, progress is so slow
2065 		 * that it might just as well be deadlocked.
2066 		 * As such, we get the rest of the attributes, but not
2067 		 * Owner or Owner_group.
2068 		 * nb: nfscl_loadattrcache() needs to be told that these
2069 		 *     partial attributes from a write rpc are being
2070 		 *     passed in, via a argument flag.
2071 		 */
2072 		if (nd->nd_flag & ND_NFSV4) {
2073 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
2074 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2075 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2076 			(void) nfsrv_putattrbit(nd, &attrbits);
2077 		}
2078 		error = nfscl_request(nd, vp, p, cred);
2079 		if (error) {
2080 			free(nd, M_TEMP);
2081 			return (error);
2082 		}
2083 		if (nd->nd_repstat) {
2084 			/*
2085 			 * In case the rpc gets retried, roll
2086 			 * the uio fields changed by nfsm_uiombuf()
2087 			 * back.
2088 			 */
2089 			uiop->uio_offset -= len;
2090 			uiop->uio_resid += len;
2091 			uiop->uio_iov->iov_base =
2092 			    (char *)uiop->uio_iov->iov_base - len;
2093 			uiop->uio_iov->iov_len += len;
2094 		}
2095 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2096 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
2097 			    &wccflag, &tmp_off);
2098 			if (error)
2099 				goto nfsmout;
2100 		}
2101 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2102 		    (ND_NFSV4 | ND_NOMOREDATA) &&
2103 		    nd->nd_repstat == NFSERR_NOTSAME && do_append) {
2104 			/*
2105 			 * Verify of the file's size failed, so redo the
2106 			 * write using the file's size as returned in
2107 			 * the wcc attributes.
2108 			 */
2109 			if (tmp_off + tsiz <= nmp->nm_maxfilesize) {
2110 				do_append = false;
2111 				uiop->uio_offset = tmp_off;
2112 				m_freem(nd->nd_mrep);
2113 				nd->nd_mrep = NULL;
2114 				continue;
2115 			} else
2116 				nd->nd_repstat = EFBIG;
2117 		}
2118 		if (!nd->nd_repstat) {
2119 			if (do_append) {
2120 				/* Strip off the Write reply status. */
2121 				do_append = false;
2122 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
2123 			}
2124 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2125 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
2126 					+ NFSX_VERF);
2127 				rlen = fxdr_unsigned(int, *tl++);
2128 				if (rlen == 0) {
2129 					error = NFSERR_IO;
2130 					goto nfsmout;
2131 				} else if (rlen < len) {
2132 					backup = len - rlen;
2133 					uiop->uio_iov->iov_base =
2134 					    (char *)uiop->uio_iov->iov_base -
2135 					    backup;
2136 					uiop->uio_iov->iov_len += backup;
2137 					uiop->uio_offset -= backup;
2138 					uiop->uio_resid += backup;
2139 					len = rlen;
2140 				}
2141 				commit = fxdr_unsigned(int, *tl++);
2142 
2143 				/*
2144 				 * Return the lowest commitment level
2145 				 * obtained by any of the RPCs.
2146 				 */
2147 				if (committed == NFSWRITE_FILESYNC)
2148 					committed = commit;
2149 				else if (committed == NFSWRITE_DATASYNC &&
2150 					commit == NFSWRITE_UNSTABLE)
2151 					committed = commit;
2152 				NFSLOCKMNT(nmp);
2153 				if (!NFSHASWRITEVERF(nmp)) {
2154 					NFSBCOPY((caddr_t)tl,
2155 					    (caddr_t)&nmp->nm_verf[0],
2156 					    NFSX_VERF);
2157 					NFSSETWRITEVERF(nmp);
2158 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
2159 				    NFSX_VERF) && *must_commit != 2) {
2160 					*must_commit = 1;
2161 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
2162 				}
2163 				NFSUNLOCKMNT(nmp);
2164 			}
2165 			if (nd->nd_flag & ND_NFSV4)
2166 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2167 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
2168 				error = nfsm_loadattr(nd, nap);
2169 				if (!error)
2170 					*attrflagp = NFS_LATTR_NOSHRINK;
2171 			}
2172 		} else {
2173 			error = nd->nd_repstat;
2174 		}
2175 		if (error)
2176 			goto nfsmout;
2177 		NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
2178 		m_freem(nd->nd_mrep);
2179 		nd->nd_mrep = NULL;
2180 		tsiz -= len;
2181 	}
2182 nfsmout:
2183 	if (nd->nd_mrep != NULL)
2184 		m_freem(nd->nd_mrep);
2185 	*iomode = committed;
2186 	if (nd->nd_repstat && !error)
2187 		error = nd->nd_repstat;
2188 	free(nd, M_TEMP);
2189 	return (error);
2190 }
2191 
2192 /*
2193  * Do an nfs deallocate operation.
2194  */
2195 int
2196 nfsrpc_deallocate(vnode_t vp, off_t offs, off_t len, struct nfsvattr *nap,
2197     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
2198 {
2199 	int error, expireret = 0, openerr, retrycnt;
2200 	uint32_t clidrev = 0;
2201 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
2202 	struct nfsfh *nfhp;
2203 	nfsv4stateid_t stateid;
2204 	void *lckp;
2205 
2206 	if (nmp->nm_clp != NULL)
2207 		clidrev = nmp->nm_clp->nfsc_clientidrev;
2208 	retrycnt = 0;
2209 	do {
2210 		lckp = NULL;
2211 		openerr = 1;
2212 		nfhp = VTONFS(vp)->n_fhp;
2213 		error = nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
2214 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
2215 		if (error != 0) {
2216 			/*
2217 			 * No Open stateid, so try and open the file
2218 			 * now.
2219 			 */
2220 			openerr = nfsrpc_open(vp, FWRITE, cred, p);
2221 			if (openerr == 0)
2222 				nfscl_getstateid(vp, nfhp->nfh_fh,
2223 				    nfhp->nfh_len, NFSV4OPEN_ACCESSWRITE, 0,
2224 				    cred, p, &stateid, &lckp);
2225 		}
2226 		error = nfsrpc_deallocaterpc(vp, offs, len, &stateid, nap,
2227 		    attrflagp, cred, p);
2228 		if (error == NFSERR_STALESTATEID)
2229 			nfscl_initiate_recovery(nmp->nm_clp);
2230 		if (lckp != NULL)
2231 			nfscl_lockderef(lckp);
2232 		if (openerr == 0)
2233 			nfsrpc_close(vp, 0, p);
2234 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2235 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2236 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
2237 			(void) nfs_catnap(PZERO, error, "nfs_deallocate");
2238 		} else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) &&
2239 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
2240 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2241 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) {
2242 			error = EIO;
2243 		}
2244 		retrycnt++;
2245 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2246 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2247 	    error == NFSERR_BADSESSION ||
2248 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
2249 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2250 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
2251 	if (error && retrycnt >= 4)
2252 		error = EIO;
2253 	return (error);
2254 }
2255 
2256 /*
2257  * The actual deallocate RPC.
2258  */
2259 static int
2260 nfsrpc_deallocaterpc(vnode_t vp, off_t offs, off_t len,
2261     nfsv4stateid_t *stateidp, struct nfsvattr *nap, int *attrflagp,
2262     struct ucred *cred, NFSPROC_T *p)
2263 {
2264 	uint32_t *tl;
2265 	struct nfsnode *np = VTONFS(vp);
2266 	int error, wccflag;
2267 	struct nfsrv_descript nfsd;
2268 	struct nfsrv_descript *nd = &nfsd;
2269 	nfsattrbit_t attrbits;
2270 
2271 	*attrflagp = 0;
2272 	NFSCL_REQSTART(nd, NFSPROC_DEALLOCATE, vp, cred);
2273 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
2274 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
2275 	txdr_hyper(offs, tl);
2276 	tl += 2;
2277 	txdr_hyper(len, tl);
2278 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
2279 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
2280 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2281 	nfsrv_putattrbit(nd, &attrbits);
2282 	error = nfscl_request(nd, vp, p, cred);
2283 	if (error != 0)
2284 		return (error);
2285 	wccflag = 0;
2286 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, &wccflag, NULL);
2287 	if (error != 0)
2288 		goto nfsmout;
2289 	if (nd->nd_repstat == 0) {
2290 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
2291 		error = nfsm_loadattr(nd, nap);
2292 		if (error != 0)
2293 			goto nfsmout;
2294 		*attrflagp = NFS_LATTR_NOSHRINK;
2295 	}
2296 	NFSWRITERPC_SETTIME(wccflag, np, nap, 1);
2297 nfsmout:
2298 	m_freem(nd->nd_mrep);
2299 	if (nd->nd_repstat != 0 && error == 0)
2300 		error = nd->nd_repstat;
2301 	return (error);
2302 }
2303 
2304 /*
2305  * nfs mknod rpc
2306  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
2307  * mode set to specify the file type and the size field for rdev.
2308  */
2309 int
2310 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2311     u_int32_t rdev, __enum_uint8(vtype) vtyp, struct ucred *cred, NFSPROC_T *p,
2312     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2313     int *attrflagp, int *dattrflagp)
2314 {
2315 	u_int32_t *tl;
2316 	int error = 0;
2317 	struct nfsrv_descript nfsd, *nd = &nfsd;
2318 	nfsattrbit_t attrbits;
2319 
2320 	*nfhpp = NULL;
2321 	*attrflagp = 0;
2322 	*dattrflagp = 0;
2323 	if (namelen > NFS_MAXNAMLEN)
2324 		return (ENAMETOOLONG);
2325 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp, cred);
2326 	if (nd->nd_flag & ND_NFSV4) {
2327 		if (vtyp == VBLK || vtyp == VCHR) {
2328 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
2329 			*tl++ = vtonfsv34_type(vtyp);
2330 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2331 			*tl = txdr_unsigned(NFSMINOR(rdev));
2332 		} else {
2333 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2334 			*tl = vtonfsv34_type(vtyp);
2335 		}
2336 	}
2337 	(void) nfsm_strtom(nd, name, namelen);
2338 	if (nd->nd_flag & ND_NFSV3) {
2339 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2340 		*tl = vtonfsv34_type(vtyp);
2341 	}
2342 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2343 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2344 	if ((nd->nd_flag & ND_NFSV3) &&
2345 	    (vtyp == VCHR || vtyp == VBLK)) {
2346 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2347 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2348 		*tl = txdr_unsigned(NFSMINOR(rdev));
2349 	}
2350 	if (nd->nd_flag & ND_NFSV4) {
2351 		NFSGETATTR_ATTRBIT(&attrbits);
2352 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2353 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2354 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2355 		(void) nfsrv_putattrbit(nd, &attrbits);
2356 	}
2357 	if (nd->nd_flag & ND_NFSV2)
2358 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
2359 	error = nfscl_request(nd, dvp, p, cred);
2360 	if (error)
2361 		return (error);
2362 	if (nd->nd_flag & ND_NFSV4)
2363 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2364 	if (!nd->nd_repstat) {
2365 		if (nd->nd_flag & ND_NFSV4) {
2366 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2367 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2368 			if (error)
2369 				goto nfsmout;
2370 		}
2371 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2372 		if (error)
2373 			goto nfsmout;
2374 	}
2375 	if (nd->nd_flag & ND_NFSV3)
2376 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2377 	if (!error && nd->nd_repstat)
2378 		error = nd->nd_repstat;
2379 nfsmout:
2380 	m_freem(nd->nd_mrep);
2381 	return (error);
2382 }
2383 
2384 /*
2385  * nfs file create call
2386  * Mostly just call the approriate routine. (I separated out v4, so that
2387  * error recovery wouldn't be as difficult.)
2388  */
2389 int
2390 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2391     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2392     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2393     int *attrflagp, int *dattrflagp)
2394 {
2395 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
2396 	struct nfsclowner *owp;
2397 	struct nfscldeleg *dp;
2398 	struct nfsmount *nmp = VFSTONFS(dvp->v_mount);
2399 	u_int32_t clidrev;
2400 
2401 	if (NFSHASNFSV4(nmp)) {
2402 	    retrycnt = 0;
2403 	    do {
2404 		dp = NULL;
2405 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
2406 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
2407 		    NULL, 1, true);
2408 		if (error)
2409 			return (error);
2410 		if (nmp->nm_clp != NULL)
2411 			clidrev = nmp->nm_clp->nfsc_clientidrev;
2412 		else
2413 			clidrev = 0;
2414 		if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
2415 		    nfs_numnfscbd == 0 || retrycnt > 0)
2416 			error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
2417 			  fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2418 			  attrflagp, dattrflagp, &unlocked);
2419 		else
2420 			error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
2421 			  cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2422 			  attrflagp, dattrflagp, &unlocked);
2423 		/*
2424 		 * There is no need to invalidate cached attributes here,
2425 		 * since new post-delegation issue attributes are always
2426 		 * returned by nfsrpc_createv4() and these will update the
2427 		 * attribute cache.
2428 		 */
2429 		if (dp != NULL)
2430 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2431 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2432 		nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2433 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2434 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2435 		    error == NFSERR_BADSESSION) {
2436 			(void) nfs_catnap(PZERO, error, "nfs_open");
2437 		} else if ((error == NFSERR_EXPIRED ||
2438 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2439 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2440 			retrycnt++;
2441 		}
2442 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2443 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2444 		error == NFSERR_BADSESSION ||
2445 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2446 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
2447 	    if (error && retrycnt >= 4)
2448 		    error = EIO;
2449 	} else {
2450 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2451 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp);
2452 	}
2453 	return (error);
2454 }
2455 
2456 /*
2457  * The create rpc for v2 and 3.
2458  */
2459 static int
2460 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2461     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2462     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2463     int *attrflagp, int *dattrflagp)
2464 {
2465 	u_int32_t *tl;
2466 	int error = 0;
2467 	struct nfsrv_descript nfsd, *nd = &nfsd;
2468 
2469 	*nfhpp = NULL;
2470 	*attrflagp = 0;
2471 	*dattrflagp = 0;
2472 	if (namelen > NFS_MAXNAMLEN)
2473 		return (ENAMETOOLONG);
2474 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp, cred);
2475 	(void) nfsm_strtom(nd, name, namelen);
2476 	if (nd->nd_flag & ND_NFSV3) {
2477 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2478 		if (fmode & O_EXCL) {
2479 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2480 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2481 			*tl++ = cverf.lval[0];
2482 			*tl = cverf.lval[1];
2483 		} else {
2484 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2485 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
2486 		}
2487 	} else {
2488 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2489 	}
2490 	error = nfscl_request(nd, dvp, p, cred);
2491 	if (error)
2492 		return (error);
2493 	if (nd->nd_repstat == 0) {
2494 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2495 		if (error)
2496 			goto nfsmout;
2497 	}
2498 	if (nd->nd_flag & ND_NFSV3)
2499 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2500 	if (nd->nd_repstat != 0 && error == 0)
2501 		error = nd->nd_repstat;
2502 nfsmout:
2503 	m_freem(nd->nd_mrep);
2504 	return (error);
2505 }
2506 
2507 static int
2508 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2509     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2510     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2511     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2512     int *dattrflagp, int *unlockedp)
2513 {
2514 	u_int32_t *tl;
2515 	int error = 0, deleg, newone, ret, acesize, limitby;
2516 	struct nfsrv_descript nfsd, *nd = &nfsd;
2517 	struct nfsclopen *op;
2518 	struct nfscldeleg *dp = NULL;
2519 	struct nfsnode *np;
2520 	struct nfsfh *nfhp;
2521 	nfsattrbit_t attrbits;
2522 	nfsv4stateid_t stateid;
2523 	u_int32_t rflags;
2524 	struct nfsmount *nmp;
2525 	struct nfsclsession *tsep;
2526 
2527 	nmp = VFSTONFS(dvp->v_mount);
2528 	np = VTONFS(dvp);
2529 	*unlockedp = 0;
2530 	*nfhpp = NULL;
2531 	*dpp = NULL;
2532 	*attrflagp = 0;
2533 	*dattrflagp = 0;
2534 	if (namelen > NFS_MAXNAMLEN)
2535 		return (ENAMETOOLONG);
2536 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp, cred);
2537 	/*
2538 	 * For V4, this is actually an Open op.
2539 	 */
2540 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2541 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
2542 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2543 	    NFSV4OPEN_ACCESSREAD);
2544 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2545 	tsep = nfsmnt_mdssession(nmp);
2546 	*tl++ = tsep->nfsess_clientid.lval[0];
2547 	*tl = tsep->nfsess_clientid.lval[1];
2548 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2549 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2550 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2551 	if (fmode & O_EXCL) {
2552 		if (NFSHASNFSV4N(nmp)) {
2553 			if (NFSHASSESSPERSIST(nmp)) {
2554 				/* Use GUARDED for persistent sessions. */
2555 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
2556 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2557 			} else {
2558 				/* Otherwise, use EXCLUSIVE4_1. */
2559 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2560 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2561 				*tl++ = cverf.lval[0];
2562 				*tl = cverf.lval[1];
2563 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2564 			}
2565 		} else {
2566 			/* NFSv4.0 */
2567 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2568 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2569 			*tl++ = cverf.lval[0];
2570 			*tl = cverf.lval[1];
2571 		}
2572 	} else {
2573 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2574 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2575 	}
2576 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2577 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2578 	(void) nfsm_strtom(nd, name, namelen);
2579 	/* Get the new file's handle and attributes. */
2580 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2581 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2582 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2583 	NFSGETATTR_ATTRBIT(&attrbits);
2584 	(void) nfsrv_putattrbit(nd, &attrbits);
2585 	/* Get the directory's post-op attributes. */
2586 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2587 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2588 	(void)nfsm_fhtom(nmp, nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2589 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2590 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2591 	(void) nfsrv_putattrbit(nd, &attrbits);
2592 	error = nfscl_request(nd, dvp, p, cred);
2593 	if (error)
2594 		return (error);
2595 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2596 	if (nd->nd_repstat == 0) {
2597 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2598 		    6 * NFSX_UNSIGNED);
2599 		stateid.seqid = *tl++;
2600 		stateid.other[0] = *tl++;
2601 		stateid.other[1] = *tl++;
2602 		stateid.other[2] = *tl;
2603 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2604 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2605 		if (error)
2606 			goto nfsmout;
2607 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2608 		deleg = fxdr_unsigned(int, *tl);
2609 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2610 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2611 			if (!(owp->nfsow_clp->nfsc_flags &
2612 			      NFSCLFLAGS_FIRSTDELEG))
2613 				owp->nfsow_clp->nfsc_flags |=
2614 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2615 			dp = malloc(
2616 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2617 			    M_NFSCLDELEG, M_WAITOK);
2618 			LIST_INIT(&dp->nfsdl_owner);
2619 			LIST_INIT(&dp->nfsdl_lock);
2620 			dp->nfsdl_clp = owp->nfsow_clp;
2621 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2622 			nfscl_lockinit(&dp->nfsdl_rwlock);
2623 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2624 			    NFSX_UNSIGNED);
2625 			dp->nfsdl_stateid.seqid = *tl++;
2626 			dp->nfsdl_stateid.other[0] = *tl++;
2627 			dp->nfsdl_stateid.other[1] = *tl++;
2628 			dp->nfsdl_stateid.other[2] = *tl++;
2629 			ret = fxdr_unsigned(int, *tl);
2630 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2631 				dp->nfsdl_flags = NFSCLDL_WRITE;
2632 				/*
2633 				 * Indicates how much the file can grow.
2634 				 */
2635 				NFSM_DISSECT(tl, u_int32_t *,
2636 				    3 * NFSX_UNSIGNED);
2637 				limitby = fxdr_unsigned(int, *tl++);
2638 				switch (limitby) {
2639 				case NFSV4OPEN_LIMITSIZE:
2640 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2641 					break;
2642 				case NFSV4OPEN_LIMITBLOCKS:
2643 					dp->nfsdl_sizelimit =
2644 					    fxdr_unsigned(u_int64_t, *tl++);
2645 					dp->nfsdl_sizelimit *=
2646 					    fxdr_unsigned(u_int64_t, *tl);
2647 					break;
2648 				default:
2649 					error = NFSERR_BADXDR;
2650 					goto nfsmout;
2651 				}
2652 			} else {
2653 				dp->nfsdl_flags = NFSCLDL_READ;
2654 			}
2655 			if (ret)
2656 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2657 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, false,
2658 			    &ret, &acesize, p);
2659 			if (error)
2660 				goto nfsmout;
2661 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2662 			error = NFSERR_BADXDR;
2663 			goto nfsmout;
2664 		}
2665 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2666 		if (error)
2667 			goto nfsmout;
2668 		/* Get rid of the PutFH and Getattr status values. */
2669 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2670 		/* Load the directory attributes. */
2671 		error = nfsm_loadattr(nd, dnap);
2672 		if (error)
2673 			goto nfsmout;
2674 		*dattrflagp = 1;
2675 		if (dp != NULL && *attrflagp) {
2676 			dp->nfsdl_change = nnap->na_filerev;
2677 			dp->nfsdl_modtime = nnap->na_mtime;
2678 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2679 		}
2680 		/*
2681 		 * We can now complete the Open state.
2682 		 */
2683 		nfhp = *nfhpp;
2684 		if (dp != NULL) {
2685 			dp->nfsdl_fhlen = nfhp->nfh_len;
2686 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2687 		}
2688 		/*
2689 		 * Get an Open structure that will be
2690 		 * attached to the OpenOwner, acquired already.
2691 		 */
2692 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2693 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2694 		    cred, p, NULL, &op, &newone, NULL, 0, false);
2695 		if (error)
2696 			goto nfsmout;
2697 		op->nfso_stateid = stateid;
2698 		newnfs_copyincred(cred, &op->nfso_cred);
2699 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2700 		    do {
2701 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2702 			    nfhp->nfh_len, op, cred, p);
2703 			if (ret == NFSERR_DELAY)
2704 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2705 		    } while (ret == NFSERR_DELAY);
2706 		    error = ret;
2707 		}
2708 
2709 		/*
2710 		 * If the server is handing out delegations, but we didn't
2711 		 * get one because an OpenConfirm was required, try the
2712 		 * Open again, to get a delegation. This is a harmless no-op,
2713 		 * from a server's point of view.
2714 		 */
2715 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2716 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2717 		    !error && dp == NULL) {
2718 		    do {
2719 			ret = nfsrpc_openrpc(VFSTONFS(dvp->v_mount), dvp,
2720 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2721 			    nfhp->nfh_fh, nfhp->nfh_len,
2722 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2723 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2724 			if (ret == NFSERR_DELAY)
2725 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2726 		    } while (ret == NFSERR_DELAY);
2727 		    if (ret) {
2728 			if (dp != NULL) {
2729 				free(dp, M_NFSCLDELEG);
2730 				dp = NULL;
2731 			}
2732 			if (ret == NFSERR_STALECLIENTID ||
2733 			    ret == NFSERR_STALEDONTRECOVER ||
2734 			    ret == NFSERR_BADSESSION)
2735 				error = ret;
2736 		    }
2737 		}
2738 		nfscl_openrelease(nmp, op, error, newone);
2739 		*unlockedp = 1;
2740 	}
2741 	if (nd->nd_repstat != 0 && error == 0)
2742 		error = nd->nd_repstat;
2743 	if (error == NFSERR_STALECLIENTID)
2744 		nfscl_initiate_recovery(owp->nfsow_clp);
2745 nfsmout:
2746 	if (!error)
2747 		*dpp = dp;
2748 	else if (dp != NULL)
2749 		free(dp, M_NFSCLDELEG);
2750 	m_freem(nd->nd_mrep);
2751 	return (error);
2752 }
2753 
2754 /*
2755  * Nfs remove rpc
2756  */
2757 int
2758 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2759     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp)
2760 {
2761 	u_int32_t *tl;
2762 	struct nfsrv_descript nfsd, *nd = &nfsd;
2763 	struct nfsnode *np;
2764 	struct nfsmount *nmp;
2765 	nfsv4stateid_t dstateid;
2766 	int error, ret = 0, i;
2767 
2768 	*dattrflagp = 0;
2769 	if (namelen > NFS_MAXNAMLEN)
2770 		return (ENAMETOOLONG);
2771 	nmp = VFSTONFS(dvp->v_mount);
2772 tryagain:
2773 	if (NFSHASNFSV4(nmp) && ret == 0) {
2774 		ret = nfscl_removedeleg(vp, p, &dstateid);
2775 		if (ret == 1) {
2776 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp, cred);
2777 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2778 			    NFSX_UNSIGNED);
2779 			if (NFSHASNFSV4N(nmp))
2780 				*tl++ = 0;
2781 			else
2782 				*tl++ = dstateid.seqid;
2783 			*tl++ = dstateid.other[0];
2784 			*tl++ = dstateid.other[1];
2785 			*tl++ = dstateid.other[2];
2786 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2787 			np = VTONFS(dvp);
2788 			(void)nfsm_fhtom(nmp, nd, np->n_fhp->nfh_fh,
2789 			    np->n_fhp->nfh_len, 0);
2790 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2791 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2792 		}
2793 	} else {
2794 		ret = 0;
2795 	}
2796 	if (ret == 0)
2797 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp, cred);
2798 	(void) nfsm_strtom(nd, name, namelen);
2799 	error = nfscl_request(nd, dvp, p, cred);
2800 	if (error)
2801 		return (error);
2802 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2803 		/* For NFSv4, parse out any Delereturn replies. */
2804 		if (ret > 0 && nd->nd_repstat != 0 &&
2805 		    (nd->nd_flag & ND_NOMOREDATA)) {
2806 			/*
2807 			 * If the Delegreturn failed, try again without
2808 			 * it. The server will Recall, as required.
2809 			 */
2810 			m_freem(nd->nd_mrep);
2811 			goto tryagain;
2812 		}
2813 		for (i = 0; i < (ret * 2); i++) {
2814 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2815 			    ND_NFSV4) {
2816 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2817 			    if (*(tl + 1))
2818 				nd->nd_flag |= ND_NOMOREDATA;
2819 			}
2820 		}
2821 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2822 	}
2823 	if (nd->nd_repstat && !error)
2824 		error = nd->nd_repstat;
2825 nfsmout:
2826 	m_freem(nd->nd_mrep);
2827 	return (error);
2828 }
2829 
2830 /*
2831  * Do an nfs rename rpc.
2832  */
2833 int
2834 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2835     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2836     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2837     int *fattrflagp, int *tattrflagp)
2838 {
2839 	u_int32_t *tl;
2840 	struct nfsrv_descript nfsd, *nd = &nfsd;
2841 	struct nfsmount *nmp;
2842 	struct nfsnode *np;
2843 	nfsattrbit_t attrbits;
2844 	nfsv4stateid_t fdstateid, tdstateid;
2845 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2846 
2847 	*fattrflagp = 0;
2848 	*tattrflagp = 0;
2849 	nmp = VFSTONFS(fdvp->v_mount);
2850 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2851 		return (ENAMETOOLONG);
2852 tryagain:
2853 	if (NFSHASNFSV4(nmp) && ret == 0) {
2854 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2855 		    &tdstateid, &gottd, p);
2856 		if (gotfd && gottd) {
2857 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp, cred);
2858 		} else if (gotfd) {
2859 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp, cred);
2860 		} else if (gottd) {
2861 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp, cred);
2862 		}
2863 		if (gotfd) {
2864 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2865 			if (NFSHASNFSV4N(nmp))
2866 				*tl++ = 0;
2867 			else
2868 				*tl++ = fdstateid.seqid;
2869 			*tl++ = fdstateid.other[0];
2870 			*tl++ = fdstateid.other[1];
2871 			*tl = fdstateid.other[2];
2872 			if (gottd) {
2873 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2874 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2875 				np = VTONFS(tvp);
2876 				(void)nfsm_fhtom(nmp, nd, np->n_fhp->nfh_fh,
2877 				    np->n_fhp->nfh_len, 0);
2878 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2879 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2880 			}
2881 		}
2882 		if (gottd) {
2883 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2884 			if (NFSHASNFSV4N(nmp))
2885 				*tl++ = 0;
2886 			else
2887 				*tl++ = tdstateid.seqid;
2888 			*tl++ = tdstateid.other[0];
2889 			*tl++ = tdstateid.other[1];
2890 			*tl = tdstateid.other[2];
2891 		}
2892 		if (ret > 0) {
2893 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2894 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2895 			np = VTONFS(fdvp);
2896 			(void)nfsm_fhtom(nmp, nd, np->n_fhp->nfh_fh,
2897 			    np->n_fhp->nfh_len, 0);
2898 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2899 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2900 		}
2901 	} else {
2902 		ret = 0;
2903 	}
2904 	if (ret == 0)
2905 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp, cred);
2906 	if (nd->nd_flag & ND_NFSV4) {
2907 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2908 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2909 		NFSWCCATTR_ATTRBIT(&attrbits);
2910 		(void) nfsrv_putattrbit(nd, &attrbits);
2911 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2912 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2913 		(void)nfsm_fhtom(nmp, nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2914 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2915 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2916 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2917 		(void) nfsrv_putattrbit(nd, &attrbits);
2918 		nd->nd_flag |= ND_V4WCCATTR;
2919 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2920 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2921 	}
2922 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2923 	if (!(nd->nd_flag & ND_NFSV4))
2924 		(void)nfsm_fhtom(nmp, nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2925 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2926 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2927 	error = nfscl_request(nd, fdvp, p, cred);
2928 	if (error)
2929 		return (error);
2930 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2931 		/* For NFSv4, parse out any Delereturn replies. */
2932 		if (ret > 0 && nd->nd_repstat != 0 &&
2933 		    (nd->nd_flag & ND_NOMOREDATA)) {
2934 			/*
2935 			 * If the Delegreturn failed, try again without
2936 			 * it. The server will Recall, as required.
2937 			 */
2938 			m_freem(nd->nd_mrep);
2939 			goto tryagain;
2940 		}
2941 		for (i = 0; i < (ret * 2); i++) {
2942 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2943 			    ND_NFSV4) {
2944 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2945 			    if (*(tl + 1)) {
2946 				if (i == 0 && ret > 1) {
2947 				    /*
2948 				     * If the Delegreturn failed, try again
2949 				     * without it. The server will Recall, as
2950 				     * required.
2951 				     * If ret > 1, the first iteration of this
2952 				     * loop is the second DelegReturn result.
2953 				     */
2954 				    m_freem(nd->nd_mrep);
2955 				    goto tryagain;
2956 				} else {
2957 				    nd->nd_flag |= ND_NOMOREDATA;
2958 				}
2959 			    }
2960 			}
2961 		}
2962 		/* Now, the first wcc attribute reply. */
2963 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2964 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2965 			if (*(tl + 1))
2966 				nd->nd_flag |= ND_NOMOREDATA;
2967 		}
2968 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL, NULL);
2969 		/* and the second wcc attribute reply. */
2970 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2971 		    !error) {
2972 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2973 			if (*(tl + 1))
2974 				nd->nd_flag |= ND_NOMOREDATA;
2975 		}
2976 		if (!error)
2977 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2978 			    NULL, NULL);
2979 	}
2980 	if (nd->nd_repstat && !error)
2981 		error = nd->nd_repstat;
2982 nfsmout:
2983 	m_freem(nd->nd_mrep);
2984 	return (error);
2985 }
2986 
2987 /*
2988  * nfs hard link create rpc
2989  */
2990 int
2991 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2992     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2993     struct nfsvattr *nap, int *attrflagp, int *dattrflagp)
2994 {
2995 	u_int32_t *tl;
2996 	struct nfsrv_descript nfsd, *nd = &nfsd;
2997 	nfsattrbit_t attrbits;
2998 	int error = 0;
2999 
3000 	*attrflagp = 0;
3001 	*dattrflagp = 0;
3002 	if (namelen > NFS_MAXNAMLEN)
3003 		return (ENAMETOOLONG);
3004 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp, cred);
3005 	if (nd->nd_flag & ND_NFSV4) {
3006 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3007 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
3008 	}
3009 	(void)nfsm_fhtom(VFSTONFS(dvp->v_mount), nd, VTONFS(dvp)->n_fhp->nfh_fh,
3010 		VTONFS(dvp)->n_fhp->nfh_len, 0);
3011 	if (nd->nd_flag & ND_NFSV4) {
3012 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3013 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3014 		NFSWCCATTR_ATTRBIT(&attrbits);
3015 		(void) nfsrv_putattrbit(nd, &attrbits);
3016 		nd->nd_flag |= ND_V4WCCATTR;
3017 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3018 		*tl = txdr_unsigned(NFSV4OP_LINK);
3019 	}
3020 	(void) nfsm_strtom(nd, name, namelen);
3021 	error = nfscl_request(nd, vp, p, cred);
3022 	if (error)
3023 		return (error);
3024 	if (nd->nd_flag & ND_NFSV3) {
3025 		error = nfscl_postop_attr(nd, nap, attrflagp);
3026 		if (!error)
3027 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
3028 			    NULL, NULL);
3029 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
3030 		/*
3031 		 * First, parse out the PutFH and Getattr result.
3032 		 */
3033 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3034 		if (!(*(tl + 1)))
3035 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3036 		if (*(tl + 1))
3037 			nd->nd_flag |= ND_NOMOREDATA;
3038 		/*
3039 		 * Get the pre-op attributes.
3040 		 */
3041 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3042 	}
3043 	if (nd->nd_repstat && !error)
3044 		error = nd->nd_repstat;
3045 nfsmout:
3046 	m_freem(nd->nd_mrep);
3047 	return (error);
3048 }
3049 
3050 /*
3051  * nfs symbolic link create rpc
3052  */
3053 int
3054 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
3055     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
3056     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
3057     int *dattrflagp)
3058 {
3059 	u_int32_t *tl;
3060 	struct nfsrv_descript nfsd, *nd = &nfsd;
3061 	struct nfsmount *nmp;
3062 	int slen, error = 0;
3063 
3064 	*nfhpp = NULL;
3065 	*attrflagp = 0;
3066 	*dattrflagp = 0;
3067 	nmp = VFSTONFS(dvp->v_mount);
3068 	slen = strlen(target);
3069 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
3070 		return (ENAMETOOLONG);
3071 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp, cred);
3072 	if (nd->nd_flag & ND_NFSV4) {
3073 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3074 		*tl = txdr_unsigned(NFLNK);
3075 		(void) nfsm_strtom(nd, target, slen);
3076 	}
3077 	(void) nfsm_strtom(nd, name, namelen);
3078 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
3079 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
3080 	if (!(nd->nd_flag & ND_NFSV4))
3081 		(void) nfsm_strtom(nd, target, slen);
3082 	if (nd->nd_flag & ND_NFSV2)
3083 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
3084 	error = nfscl_request(nd, dvp, p, cred);
3085 	if (error)
3086 		return (error);
3087 	if (nd->nd_flag & ND_NFSV4)
3088 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3089 	if ((nd->nd_flag & ND_NFSV3) && !error) {
3090 		if (!nd->nd_repstat)
3091 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
3092 		if (!error)
3093 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
3094 			    NULL, NULL);
3095 	}
3096 	if (nd->nd_repstat && !error)
3097 		error = nd->nd_repstat;
3098 	m_freem(nd->nd_mrep);
3099 	/*
3100 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
3101 	 * Only do this if vfs.nfs.ignore_eexist is set.
3102 	 * Never do this for NFSv4.1 or later minor versions, since sessions
3103 	 * should guarantee "exactly once" RPC semantics.
3104 	 */
3105 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
3106 	    nmp->nm_minorvers == 0))
3107 		error = 0;
3108 	return (error);
3109 }
3110 
3111 /*
3112  * nfs make dir rpc
3113  */
3114 int
3115 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
3116     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
3117     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
3118     int *dattrflagp)
3119 {
3120 	u_int32_t *tl;
3121 	struct nfsrv_descript nfsd, *nd = &nfsd;
3122 	nfsattrbit_t attrbits;
3123 	int error = 0;
3124 	struct nfsfh *fhp;
3125 	struct nfsmount *nmp;
3126 
3127 	*nfhpp = NULL;
3128 	*attrflagp = 0;
3129 	*dattrflagp = 0;
3130 	nmp = VFSTONFS(dvp->v_mount);
3131 	fhp = VTONFS(dvp)->n_fhp;
3132 	if (namelen > NFS_MAXNAMLEN)
3133 		return (ENAMETOOLONG);
3134 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp, cred);
3135 	if (nd->nd_flag & ND_NFSV4) {
3136 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3137 		*tl = txdr_unsigned(NFDIR);
3138 	}
3139 	(void) nfsm_strtom(nd, name, namelen);
3140 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
3141 	if (nd->nd_flag & ND_NFSV4) {
3142 		NFSGETATTR_ATTRBIT(&attrbits);
3143 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3144 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3145 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3146 		(void) nfsrv_putattrbit(nd, &attrbits);
3147 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3148 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
3149 		(void)nfsm_fhtom(nmp, nd, fhp->nfh_fh, fhp->nfh_len, 0);
3150 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3151 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3152 		(void) nfsrv_putattrbit(nd, &attrbits);
3153 	}
3154 	error = nfscl_request(nd, dvp, p, cred);
3155 	if (error)
3156 		return (error);
3157 	if (nd->nd_flag & ND_NFSV4)
3158 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3159 	if (!nd->nd_repstat && !error) {
3160 		if (nd->nd_flag & ND_NFSV4) {
3161 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3162 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
3163 		}
3164 		if (!error)
3165 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
3166 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
3167 			/* Get rid of the PutFH and Getattr status values. */
3168 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
3169 			/* Load the directory attributes. */
3170 			error = nfsm_loadattr(nd, dnap);
3171 			if (error == 0)
3172 				*dattrflagp = 1;
3173 		}
3174 	}
3175 	if ((nd->nd_flag & ND_NFSV3) && !error)
3176 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3177 	if (nd->nd_repstat && !error)
3178 		error = nd->nd_repstat;
3179 nfsmout:
3180 	m_freem(nd->nd_mrep);
3181 	/*
3182 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
3183 	 * Only do this if vfs.nfs.ignore_eexist is set.
3184 	 * Never do this for NFSv4.1 or later minor versions, since sessions
3185 	 * should guarantee "exactly once" RPC semantics.
3186 	 */
3187 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
3188 	    nmp->nm_minorvers == 0))
3189 		error = 0;
3190 	return (error);
3191 }
3192 
3193 /*
3194  * nfs remove directory call
3195  */
3196 int
3197 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
3198     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp)
3199 {
3200 	struct nfsrv_descript nfsd, *nd = &nfsd;
3201 	int error = 0;
3202 
3203 	*dattrflagp = 0;
3204 	if (namelen > NFS_MAXNAMLEN)
3205 		return (ENAMETOOLONG);
3206 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp, cred);
3207 	(void) nfsm_strtom(nd, name, namelen);
3208 	error = nfscl_request(nd, dvp, p, cred);
3209 	if (error)
3210 		return (error);
3211 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
3212 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3213 	if (nd->nd_repstat && !error)
3214 		error = nd->nd_repstat;
3215 	m_freem(nd->nd_mrep);
3216 	/*
3217 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
3218 	 */
3219 	if (error == ENOENT)
3220 		error = 0;
3221 	return (error);
3222 }
3223 
3224 /*
3225  * Readdir rpc.
3226  * Always returns with either uio_resid unchanged, if you are at the
3227  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
3228  * filled in.
3229  * I felt this would allow caching of directory blocks more easily
3230  * than returning a pertially filled block.
3231  * Directory offset cookies:
3232  * Oh my, what to do with them...
3233  * I can think of three ways to deal with them:
3234  * 1 - have the layer above these RPCs maintain a map between logical
3235  *     directory byte offsets and the NFS directory offset cookies
3236  * 2 - pass the opaque directory offset cookies up into userland
3237  *     and let the libc functions deal with them, via the system call
3238  * 3 - return them to userland in the "struct dirent", so future versions
3239  *     of libc can use them and do whatever is necessary to make things work
3240  *     above these rpc calls, in the meantime
3241  * For now, I do #3 by "hiding" the directory offset cookies after the
3242  * d_name field in struct dirent. This is space inside d_reclen that
3243  * will be ignored by anything that doesn't know about them.
3244  * The directory offset cookies are filled in as the last 8 bytes of
3245  * each directory entry, after d_name. Someday, the userland libc
3246  * functions may be able to use these. In the meantime, it satisfies
3247  * OpenBSD's requirements for cookies being returned.
3248  * If expects the directory offset cookie for the read to be in uio_offset
3249  * and returns the one for the next entry after this directory block in
3250  * there, as well.
3251  */
3252 int
3253 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3254     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3255     int *eofp)
3256 {
3257 	int len, left;
3258 	struct dirent *dp = NULL;
3259 	u_int32_t *tl;
3260 	nfsquad_t cookie, ncookie;
3261 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3262 	struct nfsnode *dnp = VTONFS(vp);
3263 	struct nfsvattr nfsva;
3264 	struct nfsrv_descript nfsd, *nd = &nfsd;
3265 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3266 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
3267 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3268 	char *cp;
3269 	nfsattrbit_t attrbits, dattrbits;
3270 	u_int32_t rderr, *tl2 = NULL;
3271 	size_t tresid;
3272 
3273 	KASSERT(uiop->uio_iovcnt == 1 &&
3274 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3275 	    ("nfs readdirrpc bad uio"));
3276 	ncookie.lval[0] = ncookie.lval[1] = 0;
3277 	/*
3278 	 * There is no point in reading a lot more than uio_resid, however
3279 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
3280 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
3281 	 * will never make readsize > nm_readdirsize.
3282 	 */
3283 	readsize = nmp->nm_readdirsize;
3284 	if (readsize > uiop->uio_resid)
3285 		readsize = uiop->uio_resid + DIRBLKSIZ;
3286 
3287 	*attrflagp = 0;
3288 	if (eofp)
3289 		*eofp = 0;
3290 	tresid = uiop->uio_resid;
3291 	cookie.lval[0] = cookiep->nfsuquad[0];
3292 	cookie.lval[1] = cookiep->nfsuquad[1];
3293 	nd->nd_mrep = NULL;
3294 
3295 	/*
3296 	 * For NFSv4, first create the "." and ".." entries.
3297 	 */
3298 	if (NFSHASNFSV4(nmp)) {
3299 		reqsize = 6 * NFSX_UNSIGNED;
3300 		NFSGETATTR_ATTRBIT(&dattrbits);
3301 		NFSZERO_ATTRBIT(&attrbits);
3302 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3303 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
3304 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3305 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3306 			NFSSETBIT_ATTRBIT(&attrbits,
3307 			    NFSATTRBIT_MOUNTEDONFILEID);
3308 			gotmnton = 1;
3309 		} else {
3310 			/*
3311 			 * Must fake it. Use the fileno, except when the
3312 			 * fsid is != to that of the directory. For that
3313 			 * case, generate a fake fileno that is not the same.
3314 			 */
3315 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3316 			gotmnton = 0;
3317 		}
3318 
3319 		/*
3320 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3321 		 */
3322 		if (uiop->uio_offset == 0) {
3323 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp, cred);
3324 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3325 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3326 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3327 			(void) nfsrv_putattrbit(nd, &attrbits);
3328 			error = nfscl_request(nd, vp, p, cred);
3329 			if (error)
3330 			    return (error);
3331 			dotfileid = 0;	/* Fake out the compiler. */
3332 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3333 			    error = nfsm_loadattr(nd, &nfsva);
3334 			    if (error != 0)
3335 				goto nfsmout;
3336 			    dotfileid = nfsva.na_fileid;
3337 			}
3338 			if (nd->nd_repstat == 0) {
3339 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3340 			    len = fxdr_unsigned(int, *(tl + 4));
3341 			    if (len > 0 && len <= NFSX_V4FHMAX)
3342 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3343 			    else
3344 				error = EPERM;
3345 			    if (!error) {
3346 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3347 				nfsva.na_mntonfileno = UINT64_MAX;
3348 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3349 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3350 				    NULL, NULL, NULL, p, cred);
3351 				if (error) {
3352 				    dotdotfileid = dotfileid;
3353 				} else if (gotmnton) {
3354 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3355 					dotdotfileid = nfsva.na_mntonfileno;
3356 				    else
3357 					dotdotfileid = nfsva.na_fileid;
3358 				} else if (nfsva.na_filesid[0] ==
3359 				    dnp->n_vattr.na_filesid[0] &&
3360 				    nfsva.na_filesid[1] ==
3361 				    dnp->n_vattr.na_filesid[1]) {
3362 				    dotdotfileid = nfsva.na_fileid;
3363 				} else {
3364 				    do {
3365 					fakefileno--;
3366 				    } while (fakefileno ==
3367 					nfsva.na_fileid);
3368 				    dotdotfileid = fakefileno;
3369 				}
3370 			    }
3371 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3372 			    /*
3373 			     * Lookupp returns NFSERR_NOENT when we are
3374 			     * at the root, so just use the current dir.
3375 			     */
3376 			    nd->nd_repstat = 0;
3377 			    dotdotfileid = dotfileid;
3378 			} else {
3379 			    error = nd->nd_repstat;
3380 			}
3381 			m_freem(nd->nd_mrep);
3382 			if (error)
3383 			    return (error);
3384 			nd->nd_mrep = NULL;
3385 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3386 			dp->d_pad0 = dp->d_pad1 = 0;
3387 			dp->d_off = 0;
3388 			dp->d_type = DT_DIR;
3389 			dp->d_fileno = dotfileid;
3390 			dp->d_namlen = 1;
3391 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3392 			dp->d_name[0] = '.';
3393 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3394 			/*
3395 			 * Just make these offset cookie 0.
3396 			 */
3397 			tl = (u_int32_t *)&dp->d_name[8];
3398 			*tl++ = 0;
3399 			*tl = 0;
3400 			blksiz += dp->d_reclen;
3401 			uiop->uio_resid -= dp->d_reclen;
3402 			uiop->uio_offset += dp->d_reclen;
3403 			uiop->uio_iov->iov_base =
3404 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3405 			uiop->uio_iov->iov_len -= dp->d_reclen;
3406 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3407 			dp->d_pad0 = dp->d_pad1 = 0;
3408 			dp->d_off = 0;
3409 			dp->d_type = DT_DIR;
3410 			dp->d_fileno = dotdotfileid;
3411 			dp->d_namlen = 2;
3412 			*((uint64_t *)dp->d_name) = 0;
3413 			dp->d_name[0] = '.';
3414 			dp->d_name[1] = '.';
3415 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3416 			/*
3417 			 * Just make these offset cookie 0.
3418 			 */
3419 			tl = (u_int32_t *)&dp->d_name[8];
3420 			*tl++ = 0;
3421 			*tl = 0;
3422 			blksiz += dp->d_reclen;
3423 			uiop->uio_resid -= dp->d_reclen;
3424 			uiop->uio_offset += dp->d_reclen;
3425 			uiop->uio_iov->iov_base =
3426 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3427 			uiop->uio_iov->iov_len -= dp->d_reclen;
3428 		}
3429 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3430 	} else {
3431 		reqsize = 5 * NFSX_UNSIGNED;
3432 	}
3433 
3434 	/*
3435 	 * Loop around doing readdir rpc's of size readsize.
3436 	 * The stopping criteria is EOF or buffer full.
3437 	 */
3438 	while (more_dirs && bigenough) {
3439 		*attrflagp = 0;
3440 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp, cred);
3441 		if (nd->nd_flag & ND_NFSV2) {
3442 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3443 			*tl++ = cookie.lval[1];
3444 			*tl = txdr_unsigned(readsize);
3445 		} else {
3446 			NFSM_BUILD(tl, u_int32_t *, reqsize);
3447 			*tl++ = cookie.lval[0];
3448 			*tl++ = cookie.lval[1];
3449 			if (cookie.qval == 0) {
3450 				*tl++ = 0;
3451 				*tl++ = 0;
3452 			} else {
3453 				NFSLOCKNODE(dnp);
3454 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
3455 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
3456 				NFSUNLOCKNODE(dnp);
3457 			}
3458 			if (nd->nd_flag & ND_NFSV4) {
3459 				*tl++ = txdr_unsigned(readsize);
3460 				*tl = txdr_unsigned(readsize);
3461 				(void) nfsrv_putattrbit(nd, &attrbits);
3462 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3463 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
3464 				(void) nfsrv_putattrbit(nd, &dattrbits);
3465 			} else {
3466 				*tl = txdr_unsigned(readsize);
3467 			}
3468 		}
3469 		error = nfscl_request(nd, vp, p, cred);
3470 		if (error)
3471 			return (error);
3472 		if (!(nd->nd_flag & ND_NFSV2)) {
3473 			if (nd->nd_flag & ND_NFSV3)
3474 				error = nfscl_postop_attr(nd, nap, attrflagp);
3475 			if (!nd->nd_repstat && !error) {
3476 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3477 				NFSLOCKNODE(dnp);
3478 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
3479 				dnp->n_cookieverf.nfsuquad[1] = *tl;
3480 				NFSUNLOCKNODE(dnp);
3481 			}
3482 		}
3483 		if (nd->nd_repstat || error) {
3484 			if (!error)
3485 				error = nd->nd_repstat;
3486 			goto nfsmout;
3487 		}
3488 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3489 		more_dirs = fxdr_unsigned(int, *tl);
3490 		if (!more_dirs)
3491 			tryformoredirs = 0;
3492 
3493 		/* loop through the dir entries, doctoring them to 4bsd form */
3494 		while (more_dirs && bigenough) {
3495 			if (nd->nd_flag & ND_NFSV4) {
3496 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3497 				ncookie.lval[0] = *tl++;
3498 				ncookie.lval[1] = *tl++;
3499 				len = fxdr_unsigned(int, *tl);
3500 			} else if (nd->nd_flag & ND_NFSV3) {
3501 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3502 				nfsva.na_fileid = fxdr_hyper(tl);
3503 				tl += 2;
3504 				len = fxdr_unsigned(int, *tl);
3505 			} else {
3506 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3507 				nfsva.na_fileid = fxdr_unsigned(uint64_t,
3508 				    *tl++);
3509 				len = fxdr_unsigned(int, *tl);
3510 			}
3511 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3512 				error = EBADRPC;
3513 				goto nfsmout;
3514 			}
3515 			tlen = roundup2(len, 8);
3516 			if (tlen == len)
3517 				tlen += 8;  /* To ensure null termination. */
3518 			left = DIRBLKSIZ - blksiz;
3519 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3520 				NFSBZERO(uiop->uio_iov->iov_base, left);
3521 				dp->d_reclen += left;
3522 				uiop->uio_iov->iov_base =
3523 				    (char *)uiop->uio_iov->iov_base + left;
3524 				uiop->uio_iov->iov_len -= left;
3525 				uiop->uio_resid -= left;
3526 				uiop->uio_offset += left;
3527 				blksiz = 0;
3528 			}
3529 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3530 			    uiop->uio_resid)
3531 				bigenough = 0;
3532 			if (bigenough) {
3533 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3534 				dp->d_pad0 = dp->d_pad1 = 0;
3535 				dp->d_off = 0;
3536 				dp->d_namlen = len;
3537 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3538 				    NFSX_HYPER;
3539 				dp->d_type = DT_UNKNOWN;
3540 				blksiz += dp->d_reclen;
3541 				if (blksiz == DIRBLKSIZ)
3542 					blksiz = 0;
3543 				uiop->uio_resid -= DIRHDSIZ;
3544 				uiop->uio_offset += DIRHDSIZ;
3545 				uiop->uio_iov->iov_base =
3546 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3547 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3548 				error = nfsm_mbufuio(nd, uiop, len);
3549 				if (error)
3550 					goto nfsmout;
3551 				cp = uiop->uio_iov->iov_base;
3552 				tlen -= len;
3553 				NFSBZERO(cp, tlen);
3554 				cp += tlen;	/* points to cookie storage */
3555 				tl2 = (u_int32_t *)cp;
3556 				uiop->uio_iov->iov_base =
3557 				    (char *)uiop->uio_iov->iov_base + tlen +
3558 				    NFSX_HYPER;
3559 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3560 				uiop->uio_resid -= tlen + NFSX_HYPER;
3561 				uiop->uio_offset += (tlen + NFSX_HYPER);
3562 			} else {
3563 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3564 				if (error)
3565 					goto nfsmout;
3566 			}
3567 			if (nd->nd_flag & ND_NFSV4) {
3568 				rderr = 0;
3569 				nfsva.na_mntonfileno = UINT64_MAX;
3570 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3571 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3572 				    NULL, NULL, &rderr, p, cred);
3573 				if (error)
3574 					goto nfsmout;
3575 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3576 			} else if (nd->nd_flag & ND_NFSV3) {
3577 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3578 				ncookie.lval[0] = *tl++;
3579 				ncookie.lval[1] = *tl++;
3580 			} else {
3581 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3582 				ncookie.lval[0] = 0;
3583 				ncookie.lval[1] = *tl++;
3584 			}
3585 			if (bigenough) {
3586 			    if (nd->nd_flag & ND_NFSV4) {
3587 				if (rderr) {
3588 				    dp->d_fileno = 0;
3589 				} else {
3590 				    if (gotmnton) {
3591 					if (nfsva.na_mntonfileno != UINT64_MAX)
3592 					    dp->d_fileno = nfsva.na_mntonfileno;
3593 					else
3594 					    dp->d_fileno = nfsva.na_fileid;
3595 				    } else if (nfsva.na_filesid[0] ==
3596 					dnp->n_vattr.na_filesid[0] &&
3597 					nfsva.na_filesid[1] ==
3598 					dnp->n_vattr.na_filesid[1]) {
3599 					dp->d_fileno = nfsva.na_fileid;
3600 				    } else {
3601 					do {
3602 					    fakefileno--;
3603 					} while (fakefileno ==
3604 					    nfsva.na_fileid);
3605 					dp->d_fileno = fakefileno;
3606 				    }
3607 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3608 				}
3609 			    } else {
3610 				dp->d_fileno = nfsva.na_fileid;
3611 			    }
3612 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3613 				ncookie.lval[0];
3614 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3615 				ncookie.lval[1];
3616 			}
3617 			more_dirs = fxdr_unsigned(int, *tl);
3618 		}
3619 		/*
3620 		 * If at end of rpc data, get the eof boolean
3621 		 */
3622 		if (!more_dirs) {
3623 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3624 			eof = fxdr_unsigned(int, *tl);
3625 			if (tryformoredirs)
3626 				more_dirs = !eof;
3627 			if (nd->nd_flag & ND_NFSV4) {
3628 				error = nfscl_postop_attr(nd, nap, attrflagp);
3629 				if (error)
3630 					goto nfsmout;
3631 			}
3632 		}
3633 		m_freem(nd->nd_mrep);
3634 		nd->nd_mrep = NULL;
3635 	}
3636 	/*
3637 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3638 	 * by increasing d_reclen for the last record.
3639 	 */
3640 	if (blksiz > 0) {
3641 		left = DIRBLKSIZ - blksiz;
3642 		NFSBZERO(uiop->uio_iov->iov_base, left);
3643 		dp->d_reclen += left;
3644 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3645 		    left;
3646 		uiop->uio_iov->iov_len -= left;
3647 		uiop->uio_resid -= left;
3648 		uiop->uio_offset += left;
3649 	}
3650 
3651 	/*
3652 	 * If returning no data, assume end of file.
3653 	 * If not bigenough, return not end of file, since you aren't
3654 	 *    returning all the data
3655 	 * Otherwise, return the eof flag from the server.
3656 	 */
3657 	if (eofp) {
3658 		if (tresid == ((size_t)(uiop->uio_resid)))
3659 			*eofp = 1;
3660 		else if (!bigenough)
3661 			*eofp = 0;
3662 		else
3663 			*eofp = eof;
3664 	}
3665 
3666 	/*
3667 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3668 	 */
3669 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3670 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3671 		NFSBZERO(dp, DIRBLKSIZ);
3672 		dp->d_type = DT_UNKNOWN;
3673 		tl = (u_int32_t *)&dp->d_name[4];
3674 		*tl++ = cookie.lval[0];
3675 		*tl = cookie.lval[1];
3676 		dp->d_reclen = DIRBLKSIZ;
3677 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3678 		    DIRBLKSIZ;
3679 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3680 		uiop->uio_resid -= DIRBLKSIZ;
3681 		uiop->uio_offset += DIRBLKSIZ;
3682 	}
3683 
3684 nfsmout:
3685 	if (nd->nd_mrep != NULL)
3686 		m_freem(nd->nd_mrep);
3687 	return (error);
3688 }
3689 
3690 /*
3691  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3692  * (Also used for NFS V4 when mount flag set.)
3693  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3694  */
3695 int
3696 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3697     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3698     int *eofp)
3699 {
3700 	int len, left;
3701 	struct dirent *dp = NULL;
3702 	u_int32_t *tl;
3703 	vnode_t newvp = NULLVP;
3704 	struct nfsrv_descript nfsd, *nd = &nfsd;
3705 	struct nameidata nami, *ndp = &nami;
3706 	struct componentname *cnp = &ndp->ni_cnd;
3707 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3708 	struct nfsnode *dnp = VTONFS(vp), *np;
3709 	struct nfsvattr nfsva;
3710 	struct nfsfh *nfhp;
3711 	nfsquad_t cookie, ncookie;
3712 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3713 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3714 	int isdotdot = 0, unlocknewvp = 0;
3715 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3716 	u_int64_t fileno = 0;
3717 	char *cp;
3718 	nfsattrbit_t attrbits, dattrbits;
3719 	size_t tresid;
3720 	u_int32_t *tl2 = NULL, rderr;
3721 	struct timespec dctime, ts;
3722 	bool attr_ok;
3723 
3724 	KASSERT(uiop->uio_iovcnt == 1 &&
3725 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3726 	    ("nfs readdirplusrpc bad uio"));
3727 	ncookie.lval[0] = ncookie.lval[1] = 0;
3728 	timespecclear(&dctime);
3729 	*attrflagp = 0;
3730 	if (eofp != NULL)
3731 		*eofp = 0;
3732 	ndp->ni_dvp = vp;
3733 	nd->nd_mrep = NULL;
3734 	cookie.lval[0] = cookiep->nfsuquad[0];
3735 	cookie.lval[1] = cookiep->nfsuquad[1];
3736 	tresid = uiop->uio_resid;
3737 
3738 	/*
3739 	 * For NFSv4, first create the "." and ".." entries.
3740 	 */
3741 	if (NFSHASNFSV4(nmp)) {
3742 		NFSGETATTR_ATTRBIT(&dattrbits);
3743 		NFSZERO_ATTRBIT(&attrbits);
3744 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3745 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3746 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3747 			NFSSETBIT_ATTRBIT(&attrbits,
3748 			    NFSATTRBIT_MOUNTEDONFILEID);
3749 			gotmnton = 1;
3750 		} else {
3751 			/*
3752 			 * Must fake it. Use the fileno, except when the
3753 			 * fsid is != to that of the directory. For that
3754 			 * case, generate a fake fileno that is not the same.
3755 			 */
3756 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3757 			gotmnton = 0;
3758 		}
3759 
3760 		/*
3761 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3762 		 */
3763 		if (uiop->uio_offset == 0) {
3764 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp, cred);
3765 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3766 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3767 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3768 			(void) nfsrv_putattrbit(nd, &attrbits);
3769 			error = nfscl_request(nd, vp, p, cred);
3770 			if (error)
3771 			    return (error);
3772 			dotfileid = 0;	/* Fake out the compiler. */
3773 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3774 			    error = nfsm_loadattr(nd, &nfsva);
3775 			    if (error != 0)
3776 				goto nfsmout;
3777 			    dctime = nfsva.na_ctime;
3778 			    dotfileid = nfsva.na_fileid;
3779 			}
3780 			if (nd->nd_repstat == 0) {
3781 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3782 			    len = fxdr_unsigned(int, *(tl + 4));
3783 			    if (len > 0 && len <= NFSX_V4FHMAX)
3784 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3785 			    else
3786 				error = EPERM;
3787 			    if (!error) {
3788 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3789 				nfsva.na_mntonfileno = UINT64_MAX;
3790 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3791 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3792 				    NULL, NULL, NULL, p, cred);
3793 				if (error) {
3794 				    dotdotfileid = dotfileid;
3795 				} else if (gotmnton) {
3796 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3797 					dotdotfileid = nfsva.na_mntonfileno;
3798 				    else
3799 					dotdotfileid = nfsva.na_fileid;
3800 				} else if (nfsva.na_filesid[0] ==
3801 				    dnp->n_vattr.na_filesid[0] &&
3802 				    nfsva.na_filesid[1] ==
3803 				    dnp->n_vattr.na_filesid[1]) {
3804 				    dotdotfileid = nfsva.na_fileid;
3805 				} else {
3806 				    do {
3807 					fakefileno--;
3808 				    } while (fakefileno ==
3809 					nfsva.na_fileid);
3810 				    dotdotfileid = fakefileno;
3811 				}
3812 			    }
3813 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3814 			    /*
3815 			     * Lookupp returns NFSERR_NOENT when we are
3816 			     * at the root, so just use the current dir.
3817 			     */
3818 			    nd->nd_repstat = 0;
3819 			    dotdotfileid = dotfileid;
3820 			} else {
3821 			    error = nd->nd_repstat;
3822 			}
3823 			m_freem(nd->nd_mrep);
3824 			if (error)
3825 			    return (error);
3826 			nd->nd_mrep = NULL;
3827 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3828 			dp->d_pad0 = dp->d_pad1 = 0;
3829 			dp->d_off = 0;
3830 			dp->d_type = DT_DIR;
3831 			dp->d_fileno = dotfileid;
3832 			dp->d_namlen = 1;
3833 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3834 			dp->d_name[0] = '.';
3835 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3836 			/*
3837 			 * Just make these offset cookie 0.
3838 			 */
3839 			tl = (u_int32_t *)&dp->d_name[8];
3840 			*tl++ = 0;
3841 			*tl = 0;
3842 			blksiz += dp->d_reclen;
3843 			uiop->uio_resid -= dp->d_reclen;
3844 			uiop->uio_offset += dp->d_reclen;
3845 			uiop->uio_iov->iov_base =
3846 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3847 			uiop->uio_iov->iov_len -= dp->d_reclen;
3848 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3849 			dp->d_pad0 = dp->d_pad1 = 0;
3850 			dp->d_off = 0;
3851 			dp->d_type = DT_DIR;
3852 			dp->d_fileno = dotdotfileid;
3853 			dp->d_namlen = 2;
3854 			*((uint64_t *)dp->d_name) = 0;
3855 			dp->d_name[0] = '.';
3856 			dp->d_name[1] = '.';
3857 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3858 			/*
3859 			 * Just make these offset cookie 0.
3860 			 */
3861 			tl = (u_int32_t *)&dp->d_name[8];
3862 			*tl++ = 0;
3863 			*tl = 0;
3864 			blksiz += dp->d_reclen;
3865 			uiop->uio_resid -= dp->d_reclen;
3866 			uiop->uio_offset += dp->d_reclen;
3867 			uiop->uio_iov->iov_base =
3868 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3869 			uiop->uio_iov->iov_len -= dp->d_reclen;
3870 		}
3871 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3872 		if (gotmnton)
3873 			NFSSETBIT_ATTRBIT(&attrbits,
3874 			    NFSATTRBIT_MOUNTEDONFILEID);
3875 		if (!NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3876 		    NFSATTRBIT_TIMECREATE))
3877 			NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMECREATE);
3878 	}
3879 
3880 	/*
3881 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3882 	 * The stopping criteria is EOF or buffer full.
3883 	 */
3884 	while (more_dirs && bigenough) {
3885 		*attrflagp = 0;
3886 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp, cred);
3887  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3888 		*tl++ = cookie.lval[0];
3889 		*tl++ = cookie.lval[1];
3890 		if (cookie.qval == 0) {
3891 			*tl++ = 0;
3892 			*tl++ = 0;
3893 		} else {
3894 			NFSLOCKNODE(dnp);
3895 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3896 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3897 			NFSUNLOCKNODE(dnp);
3898 		}
3899 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3900 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3901 		if (nd->nd_flag & ND_NFSV4) {
3902 			(void) nfsrv_putattrbit(nd, &attrbits);
3903 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3904 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3905 			(void) nfsrv_putattrbit(nd, &dattrbits);
3906 		}
3907 		nanouptime(&ts);
3908 		error = nfscl_request(nd, vp, p, cred);
3909 		if (error)
3910 			return (error);
3911 		if (nd->nd_flag & ND_NFSV3)
3912 			error = nfscl_postop_attr(nd, nap, attrflagp);
3913 		if (nd->nd_repstat || error) {
3914 			if (!error)
3915 				error = nd->nd_repstat;
3916 			goto nfsmout;
3917 		}
3918 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3919 			dctime = nap->na_ctime;
3920 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3921 		NFSLOCKNODE(dnp);
3922 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3923 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3924 		NFSUNLOCKNODE(dnp);
3925 		more_dirs = fxdr_unsigned(int, *tl);
3926 		if (!more_dirs)
3927 			tryformoredirs = 0;
3928 
3929 		/* loop through the dir entries, doctoring them to 4bsd form */
3930 		while (more_dirs && bigenough) {
3931 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3932 			if (nd->nd_flag & ND_NFSV4) {
3933 				ncookie.lval[0] = *tl++;
3934 				ncookie.lval[1] = *tl++;
3935 			} else {
3936 				fileno = fxdr_hyper(tl);
3937 				tl += 2;
3938 			}
3939 			len = fxdr_unsigned(int, *tl);
3940 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3941 				error = EBADRPC;
3942 				goto nfsmout;
3943 			}
3944 			tlen = roundup2(len, 8);
3945 			if (tlen == len)
3946 				tlen += 8;  /* To ensure null termination. */
3947 			left = DIRBLKSIZ - blksiz;
3948 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3949 				NFSBZERO(uiop->uio_iov->iov_base, left);
3950 				dp->d_reclen += left;
3951 				uiop->uio_iov->iov_base =
3952 				    (char *)uiop->uio_iov->iov_base + left;
3953 				uiop->uio_iov->iov_len -= left;
3954 				uiop->uio_resid -= left;
3955 				uiop->uio_offset += left;
3956 				blksiz = 0;
3957 			}
3958 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3959 			    uiop->uio_resid)
3960 				bigenough = 0;
3961 			if (bigenough) {
3962 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3963 				dp->d_pad0 = dp->d_pad1 = 0;
3964 				dp->d_off = 0;
3965 				dp->d_namlen = len;
3966 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3967 				    NFSX_HYPER;
3968 				dp->d_type = DT_UNKNOWN;
3969 				blksiz += dp->d_reclen;
3970 				if (blksiz == DIRBLKSIZ)
3971 					blksiz = 0;
3972 				uiop->uio_resid -= DIRHDSIZ;
3973 				uiop->uio_offset += DIRHDSIZ;
3974 				uiop->uio_iov->iov_base =
3975 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3976 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3977 				cnp->cn_nameptr = uiop->uio_iov->iov_base;
3978 				cnp->cn_namelen = len;
3979 				NFSCNHASHZERO(cnp);
3980 				error = nfsm_mbufuio(nd, uiop, len);
3981 				if (error)
3982 					goto nfsmout;
3983 				cp = uiop->uio_iov->iov_base;
3984 				tlen -= len;
3985 				NFSBZERO(cp, tlen);
3986 				cp += tlen;	/* points to cookie storage */
3987 				tl2 = (u_int32_t *)cp;
3988 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3989 				    cnp->cn_nameptr[1] == '.')
3990 					isdotdot = 1;
3991 				else
3992 					isdotdot = 0;
3993 				uiop->uio_iov->iov_base =
3994 				    (char *)uiop->uio_iov->iov_base + tlen +
3995 				    NFSX_HYPER;
3996 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3997 				uiop->uio_resid -= tlen + NFSX_HYPER;
3998 				uiop->uio_offset += (tlen + NFSX_HYPER);
3999 			} else {
4000 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
4001 				if (error)
4002 					goto nfsmout;
4003 			}
4004 			nfhp = NULL;
4005 			if (nd->nd_flag & ND_NFSV3) {
4006 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
4007 				ncookie.lval[0] = *tl++;
4008 				ncookie.lval[1] = *tl++;
4009 				attrflag = fxdr_unsigned(int, *tl);
4010 				if (attrflag) {
4011 				  error = nfsm_loadattr(nd, &nfsva);
4012 				  if (error)
4013 					goto nfsmout;
4014 				}
4015 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
4016 				if (*tl) {
4017 					error = nfsm_getfh(nd, &nfhp);
4018 					if (error)
4019 					    goto nfsmout;
4020 				}
4021 				if (!attrflag && nfhp != NULL) {
4022 					free(nfhp, M_NFSFH);
4023 					nfhp = NULL;
4024 				}
4025 			} else {
4026 				rderr = 0;
4027 				nfsva.na_mntonfileno = 0xffffffff;
4028 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
4029 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
4030 				    NULL, NULL, &rderr, p, cred);
4031 				if (error)
4032 					goto nfsmout;
4033 			}
4034 
4035 			if (bigenough) {
4036 			    if (nd->nd_flag & ND_NFSV4) {
4037 				if (rderr) {
4038 				    dp->d_fileno = 0;
4039 				} else if (gotmnton) {
4040 				    if (nfsva.na_mntonfileno != 0xffffffff)
4041 					dp->d_fileno = nfsva.na_mntonfileno;
4042 				    else
4043 					dp->d_fileno = nfsva.na_fileid;
4044 				} else if (nfsva.na_filesid[0] ==
4045 				    dnp->n_vattr.na_filesid[0] &&
4046 				    nfsva.na_filesid[1] ==
4047 				    dnp->n_vattr.na_filesid[1]) {
4048 				    dp->d_fileno = nfsva.na_fileid;
4049 				} else {
4050 				    do {
4051 					fakefileno--;
4052 				    } while (fakefileno ==
4053 					nfsva.na_fileid);
4054 				    dp->d_fileno = fakefileno;
4055 				}
4056 			    } else {
4057 				dp->d_fileno = fileno;
4058 			    }
4059 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
4060 				ncookie.lval[0];
4061 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
4062 				ncookie.lval[1];
4063 
4064 			    if (nfhp != NULL) {
4065 				attr_ok = true;
4066 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
4067 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
4068 				    VREF(vp);
4069 				    newvp = vp;
4070 				    unlocknewvp = 0;
4071 				    free(nfhp, M_NFSFH);
4072 				    np = dnp;
4073 				} else if (isdotdot != 0) {
4074 				    /*
4075 				     * Skip doing a nfscl_nget() call for "..".
4076 				     * There's a race between acquiring the nfs
4077 				     * node here and lookups that look for the
4078 				     * directory being read (in the parent).
4079 				     * It would try to get a lock on ".." here,
4080 				     * owning the lock on the directory being
4081 				     * read. Lookup will hold the lock on ".."
4082 				     * and try to acquire the lock on the
4083 				     * directory being read.
4084 				     * If the directory is unlocked/relocked,
4085 				     * then there is a LOR with the buflock
4086 				     * vp is relocked.
4087 				     */
4088 				    free(nfhp, M_NFSFH);
4089 				} else {
4090 				    error = nfscl_nget(vp->v_mount, vp,
4091 				      nfhp, cnp, p, &np, LK_EXCLUSIVE);
4092 				    if (!error) {
4093 					newvp = NFSTOV(np);
4094 					unlocknewvp = 1;
4095 					/*
4096 					 * If n_localmodtime >= time before RPC,
4097 					 * then a file modification operation,
4098 					 * such as VOP_SETATTR() of size, has
4099 					 * occurred while the Lookup RPC and
4100 					 * acquisition of the vnode happened. As
4101 					 * such, the attributes might be stale,
4102 					 * with possibly an incorrect size.
4103 					 */
4104 					NFSLOCKNODE(np);
4105 					if (timespecisset(
4106 					    &np->n_localmodtime) &&
4107 					    timespeccmp(&np->n_localmodtime,
4108 					    &ts, >=)) {
4109 					    NFSCL_DEBUG(4, "nfsrpc_readdirplus:"
4110 						" localmod stale attributes\n");
4111 					    attr_ok = false;
4112 					}
4113 					NFSUNLOCKNODE(np);
4114 				    }
4115 				}
4116 				nfhp = NULL;
4117 				if (newvp != NULLVP) {
4118 				    if (attr_ok)
4119 					error = nfscl_loadattrcache(&newvp,
4120 					    &nfsva, NULL, 0, 0);
4121 				    if (error) {
4122 					if (unlocknewvp)
4123 					    vput(newvp);
4124 					else
4125 					    vrele(newvp);
4126 					goto nfsmout;
4127 				    }
4128 				    dp->d_type =
4129 					vtonfs_dtype(np->n_vattr.na_type);
4130 				    ndp->ni_vp = newvp;
4131 				    NFSCNHASH(cnp, HASHINIT);
4132 				    if (cnp->cn_namelen <= NCHNAMLEN &&
4133 					ndp->ni_dvp != ndp->ni_vp &&
4134 					(newvp->v_type != VDIR ||
4135 					 dctime.tv_sec != 0)) {
4136 					cache_enter_time_flags(ndp->ni_dvp,
4137 					    ndp->ni_vp, cnp,
4138 					    &nfsva.na_ctime,
4139 					    newvp->v_type != VDIR ? NULL :
4140 					    &dctime, VFS_CACHE_DROPOLD);
4141 				    }
4142 				    if (unlocknewvp)
4143 					vput(newvp);
4144 				    else
4145 					vrele(newvp);
4146 				    newvp = NULLVP;
4147 				}
4148 			    }
4149 			} else if (nfhp != NULL) {
4150 			    free(nfhp, M_NFSFH);
4151 			}
4152 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
4153 			more_dirs = fxdr_unsigned(int, *tl);
4154 		}
4155 		/*
4156 		 * If at end of rpc data, get the eof boolean
4157 		 */
4158 		if (!more_dirs) {
4159 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
4160 			eof = fxdr_unsigned(int, *tl);
4161 			if (tryformoredirs)
4162 				more_dirs = !eof;
4163 			if (nd->nd_flag & ND_NFSV4) {
4164 				error = nfscl_postop_attr(nd, nap, attrflagp);
4165 				if (error)
4166 					goto nfsmout;
4167 			}
4168 		}
4169 		m_freem(nd->nd_mrep);
4170 		nd->nd_mrep = NULL;
4171 	}
4172 	/*
4173 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
4174 	 * by increasing d_reclen for the last record.
4175 	 */
4176 	if (blksiz > 0) {
4177 		left = DIRBLKSIZ - blksiz;
4178 		NFSBZERO(uiop->uio_iov->iov_base, left);
4179 		dp->d_reclen += left;
4180 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
4181 		    left;
4182 		uiop->uio_iov->iov_len -= left;
4183 		uiop->uio_resid -= left;
4184 		uiop->uio_offset += left;
4185 	}
4186 
4187 	/*
4188 	 * If returning no data, assume end of file.
4189 	 * If not bigenough, return not end of file, since you aren't
4190 	 *    returning all the data
4191 	 * Otherwise, return the eof flag from the server.
4192 	 */
4193 	if (eofp != NULL) {
4194 		if (tresid == uiop->uio_resid)
4195 			*eofp = 1;
4196 		else if (!bigenough)
4197 			*eofp = 0;
4198 		else
4199 			*eofp = eof;
4200 	}
4201 
4202 	/*
4203 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
4204 	 */
4205 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
4206 		dp = (struct dirent *)uiop->uio_iov->iov_base;
4207 		NFSBZERO(dp, DIRBLKSIZ);
4208 		dp->d_type = DT_UNKNOWN;
4209 		tl = (u_int32_t *)&dp->d_name[4];
4210 		*tl++ = cookie.lval[0];
4211 		*tl = cookie.lval[1];
4212 		dp->d_reclen = DIRBLKSIZ;
4213 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
4214 		    DIRBLKSIZ;
4215 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
4216 		uiop->uio_resid -= DIRBLKSIZ;
4217 		uiop->uio_offset += DIRBLKSIZ;
4218 	}
4219 
4220 nfsmout:
4221 	if (nd->nd_mrep != NULL)
4222 		m_freem(nd->nd_mrep);
4223 	return (error);
4224 }
4225 
4226 /*
4227  * Nfs commit rpc
4228  */
4229 int
4230 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
4231     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4232 {
4233 	u_int32_t *tl;
4234 	struct nfsrv_descript nfsd, *nd = &nfsd;
4235 	nfsattrbit_t attrbits;
4236 	int error;
4237 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4238 
4239 	*attrflagp = 0;
4240 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp, cred);
4241 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
4242 	txdr_hyper(offset, tl);
4243 	tl += 2;
4244 	*tl = txdr_unsigned(cnt);
4245 	if (nd->nd_flag & ND_NFSV4) {
4246 		/*
4247 		 * And do a Getattr op.
4248 		 */
4249 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4250 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
4251 		NFSGETATTR_ATTRBIT(&attrbits);
4252 		(void) nfsrv_putattrbit(nd, &attrbits);
4253 	}
4254 	error = nfscl_request(nd, vp, p, cred);
4255 	if (error)
4256 		return (error);
4257 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, NULL);
4258 	if (!error && !nd->nd_repstat) {
4259 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
4260 		NFSLOCKMNT(nmp);
4261 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
4262 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
4263 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
4264 		}
4265 		NFSUNLOCKMNT(nmp);
4266 		if (nd->nd_flag & ND_NFSV4)
4267 			error = nfscl_postop_attr(nd, nap, attrflagp);
4268 	}
4269 nfsmout:
4270 	if (!error && nd->nd_repstat)
4271 		error = nd->nd_repstat;
4272 	m_freem(nd->nd_mrep);
4273 	return (error);
4274 }
4275 
4276 /*
4277  * NFS byte range lock rpc.
4278  * (Mostly just calls one of the three lower level RPC routines.)
4279  */
4280 int
4281 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
4282     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4283 {
4284 	struct nfscllockowner *lp;
4285 	struct nfsclclient *clp;
4286 	struct nfsfh *nfhp;
4287 	struct nfsrv_descript nfsd, *nd = &nfsd;
4288 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4289 	u_int64_t off, len;
4290 	off_t start, end;
4291 	u_int32_t clidrev = 0;
4292 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
4293 	int callcnt, dorpc;
4294 
4295 	/*
4296 	 * Convert the flock structure into a start and end and do POSIX
4297 	 * bounds checking.
4298 	 */
4299 	switch (fl->l_whence) {
4300 	case SEEK_SET:
4301 	case SEEK_CUR:
4302 		/*
4303 		 * Caller is responsible for adding any necessary offset
4304 		 * when SEEK_CUR is used.
4305 		 */
4306 		start = fl->l_start;
4307 		off = fl->l_start;
4308 		break;
4309 	case SEEK_END:
4310 		start = size + fl->l_start;
4311 		off = size + fl->l_start;
4312 		break;
4313 	default:
4314 		return (EINVAL);
4315 	}
4316 	if (start < 0)
4317 		return (EINVAL);
4318 	if (fl->l_len != 0) {
4319 		end = start + fl->l_len - 1;
4320 		if (end < start)
4321 			return (EINVAL);
4322 	}
4323 
4324 	len = fl->l_len;
4325 	if (len == 0)
4326 		len = NFS64BITSSET;
4327 	retrycnt = 0;
4328 	do {
4329 	    nd->nd_repstat = 0;
4330 	    if (op == F_GETLK) {
4331 		error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
4332 		if (error)
4333 			return (error);
4334 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
4335 		if (!error) {
4336 			clidrev = clp->nfsc_clientidrev;
4337 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
4338 			    p, id, flags);
4339 		} else if (error == -1) {
4340 			error = 0;
4341 		}
4342 		nfscl_clientrelease(clp);
4343 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
4344 		/*
4345 		 * We must loop around for all lockowner cases.
4346 		 */
4347 		callcnt = 0;
4348 		error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
4349 		if (error)
4350 			return (error);
4351 		do {
4352 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
4353 			clp, id, flags, &lp, &dorpc);
4354 		    /*
4355 		     * If it returns a NULL lp, we're done.
4356 		     */
4357 		    if (lp == NULL) {
4358 			if (callcnt == 0)
4359 			    nfscl_clientrelease(clp);
4360 			else
4361 			    nfscl_releasealllocks(clp, vp, p, id, flags);
4362 			return (error);
4363 		    }
4364 		    if (nmp->nm_clp != NULL)
4365 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4366 		    else
4367 			clidrev = 0;
4368 		    /*
4369 		     * If the server doesn't support Posix lock semantics,
4370 		     * only allow locks on the entire file, since it won't
4371 		     * handle overlapping byte ranges.
4372 		     * There might still be a problem when a lock
4373 		     * upgrade/downgrade (read<->write) occurs, since the
4374 		     * server "might" expect an unlock first?
4375 		     */
4376 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
4377 			(off == 0 && len == NFS64BITSSET))) {
4378 			/*
4379 			 * Since the lock records will go away, we must
4380 			 * wait for grace and delay here.
4381 			 */
4382 			do {
4383 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
4384 				NFSV4LOCKT_READ, cred, p, 0);
4385 			    if ((nd->nd_repstat == NFSERR_GRACE ||
4386 				 nd->nd_repstat == NFSERR_DELAY) &&
4387 				error == 0)
4388 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
4389 				    "nfs_advlock");
4390 			} while ((nd->nd_repstat == NFSERR_GRACE ||
4391 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
4392 		    }
4393 		    callcnt++;
4394 		} while (error == 0 && nd->nd_repstat == 0);
4395 		nfscl_releasealllocks(clp, vp, p, id, flags);
4396 	    } else if (op == F_SETLK) {
4397 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
4398 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
4399 		if (error || donelocally) {
4400 			return (error);
4401 		}
4402 		if (nmp->nm_clp != NULL)
4403 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4404 		else
4405 			clidrev = 0;
4406 		nfhp = VTONFS(vp)->n_fhp;
4407 		if (!lp->nfsl_open->nfso_posixlock &&
4408 		    (off != 0 || len != NFS64BITSSET)) {
4409 			error = EINVAL;
4410 		} else {
4411 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
4412 			    nfhp->nfh_len, lp, newone, reclaim, off,
4413 			    len, fl->l_type, cred, p, 0);
4414 		}
4415 		if (!error)
4416 			error = nd->nd_repstat;
4417 		nfscl_lockrelease(lp, error, newone);
4418 	    } else {
4419 		error = EINVAL;
4420 	    }
4421 	    if (!error)
4422 	        error = nd->nd_repstat;
4423 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
4424 		error == NFSERR_STALEDONTRECOVER ||
4425 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4426 		error == NFSERR_BADSESSION) {
4427 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
4428 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
4429 		&& clidrev != 0) {
4430 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
4431 		retrycnt++;
4432 	    }
4433 	} while (error == NFSERR_GRACE ||
4434 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4435 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
4436 	    error == NFSERR_BADSESSION ||
4437 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
4438 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
4439 	if (error && retrycnt >= 4)
4440 		error = EIO;
4441 	return (error);
4442 }
4443 
4444 /*
4445  * The lower level routine for the LockT case.
4446  */
4447 int
4448 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
4449     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
4450     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4451 {
4452 	u_int32_t *tl;
4453 	int error, type, size;
4454 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4455 	struct nfsnode *np;
4456 	struct nfsmount *nmp;
4457 	struct nfsclsession *tsep;
4458 
4459 	nmp = VFSTONFS(vp->v_mount);
4460 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp, cred);
4461 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4462 	if (fl->l_type == F_RDLCK)
4463 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4464 	else
4465 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4466 	txdr_hyper(off, tl);
4467 	tl += 2;
4468 	txdr_hyper(len, tl);
4469 	tl += 2;
4470 	tsep = nfsmnt_mdssession(nmp);
4471 	*tl++ = tsep->nfsess_clientid.lval[0];
4472 	*tl = tsep->nfsess_clientid.lval[1];
4473 	nfscl_filllockowner(id, own, flags);
4474 	np = VTONFS(vp);
4475 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4476 	    np->n_fhp->nfh_len);
4477 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4478 	error = nfscl_request(nd, vp, p, cred);
4479 	if (error)
4480 		return (error);
4481 	if (nd->nd_repstat == 0) {
4482 		fl->l_type = F_UNLCK;
4483 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4484 		nd->nd_repstat = 0;
4485 		fl->l_whence = SEEK_SET;
4486 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4487 		fl->l_start = fxdr_hyper(tl);
4488 		tl += 2;
4489 		len = fxdr_hyper(tl);
4490 		tl += 2;
4491 		if (len == NFS64BITSSET)
4492 			fl->l_len = 0;
4493 		else
4494 			fl->l_len = len;
4495 		type = fxdr_unsigned(int, *tl++);
4496 		if (type == NFSV4LOCKT_WRITE)
4497 			fl->l_type = F_WRLCK;
4498 		else
4499 			fl->l_type = F_RDLCK;
4500 		/*
4501 		 * XXX For now, I have no idea what to do with the
4502 		 * conflicting lock_owner, so I'll just set the pid == 0
4503 		 * and skip over the lock_owner.
4504 		 */
4505 		fl->l_pid = (pid_t)0;
4506 		tl += 2;
4507 		size = fxdr_unsigned(int, *tl);
4508 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4509 			error = EBADRPC;
4510 		if (!error)
4511 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4512 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4513 		nfscl_initiate_recovery(clp);
4514 nfsmout:
4515 	m_freem(nd->nd_mrep);
4516 	return (error);
4517 }
4518 
4519 /*
4520  * Lower level function that performs the LockU RPC.
4521  */
4522 static int
4523 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4524     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4525     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4526 {
4527 	u_int32_t *tl;
4528 	int error;
4529 
4530 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4531 	    lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0, cred);
4532 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4533 	*tl++ = txdr_unsigned(type);
4534 	*tl = txdr_unsigned(lp->nfsl_seqid);
4535 	if (nfstest_outofseq &&
4536 	    (arc4random() % nfstest_outofseq) == 0)
4537 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
4538 	tl++;
4539 	if (NFSHASNFSV4N(nmp))
4540 		*tl++ = 0;
4541 	else
4542 		*tl++ = lp->nfsl_stateid.seqid;
4543 	*tl++ = lp->nfsl_stateid.other[0];
4544 	*tl++ = lp->nfsl_stateid.other[1];
4545 	*tl++ = lp->nfsl_stateid.other[2];
4546 	txdr_hyper(off, tl);
4547 	tl += 2;
4548 	txdr_hyper(len, tl);
4549 	if (syscred)
4550 		nd->nd_flag |= ND_USEGSSNAME;
4551 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4552 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4553 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4554 	if (error)
4555 		return (error);
4556 	if (nd->nd_repstat == 0) {
4557 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4558 		lp->nfsl_stateid.seqid = *tl++;
4559 		lp->nfsl_stateid.other[0] = *tl++;
4560 		lp->nfsl_stateid.other[1] = *tl++;
4561 		lp->nfsl_stateid.other[2] = *tl;
4562 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4563 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4564 nfsmout:
4565 	m_freem(nd->nd_mrep);
4566 	return (error);
4567 }
4568 
4569 /*
4570  * The actual Lock RPC.
4571  */
4572 int
4573 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4574     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4575     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4576     NFSPROC_T *p, int syscred)
4577 {
4578 	u_int32_t *tl;
4579 	int error, size;
4580 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4581 	struct nfsclsession *tsep;
4582 
4583 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0,
4584 	    cred);
4585 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4586 	if (type == F_RDLCK)
4587 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4588 	else
4589 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4590 	*tl++ = txdr_unsigned(reclaim);
4591 	txdr_hyper(off, tl);
4592 	tl += 2;
4593 	txdr_hyper(len, tl);
4594 	tl += 2;
4595 	if (newone) {
4596 	    *tl = newnfs_true;
4597 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4598 		2 * NFSX_UNSIGNED + NFSX_HYPER);
4599 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4600 	    if (NFSHASNFSV4N(nmp))
4601 		*tl++ = 0;
4602 	    else
4603 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
4604 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4605 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4606 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4607 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
4608 	    tsep = nfsmnt_mdssession(nmp);
4609 	    *tl++ = tsep->nfsess_clientid.lval[0];
4610 	    *tl = tsep->nfsess_clientid.lval[1];
4611 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4612 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4613 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4614 	} else {
4615 	    *tl = newnfs_false;
4616 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4617 	    if (NFSHASNFSV4N(nmp))
4618 		*tl++ = 0;
4619 	    else
4620 		*tl++ = lp->nfsl_stateid.seqid;
4621 	    *tl++ = lp->nfsl_stateid.other[0];
4622 	    *tl++ = lp->nfsl_stateid.other[1];
4623 	    *tl++ = lp->nfsl_stateid.other[2];
4624 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4625 	    if (nfstest_outofseq &&
4626 		(arc4random() % nfstest_outofseq) == 0)
4627 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4628 	}
4629 	if (syscred)
4630 		nd->nd_flag |= ND_USEGSSNAME;
4631 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4632 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4633 	if (error)
4634 		return (error);
4635 	if (newone)
4636 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4637 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4638 	if (nd->nd_repstat == 0) {
4639 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4640 		lp->nfsl_stateid.seqid = *tl++;
4641 		lp->nfsl_stateid.other[0] = *tl++;
4642 		lp->nfsl_stateid.other[1] = *tl++;
4643 		lp->nfsl_stateid.other[2] = *tl;
4644 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4645 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4646 		size = fxdr_unsigned(int, *(tl + 7));
4647 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4648 			error = EBADRPC;
4649 		if (!error)
4650 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4651 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4652 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4653 nfsmout:
4654 	m_freem(nd->nd_mrep);
4655 	return (error);
4656 }
4657 
4658 /*
4659  * nfs statfs rpc
4660  * (always called with the vp for the mount point)
4661  */
4662 int
4663 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4664     uint32_t *leasep, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap,
4665     int *attrflagp)
4666 {
4667 	u_int32_t *tl = NULL;
4668 	struct nfsrv_descript nfsd, *nd = &nfsd;
4669 	struct nfsmount *nmp;
4670 	nfsattrbit_t attrbits;
4671 	int error;
4672 
4673 	*attrflagp = 0;
4674 	nmp = VFSTONFS(vp->v_mount);
4675 	if (NFSHASNFSV4(nmp)) {
4676 		/*
4677 		 * For V4, you actually do a getattr.
4678 		 */
4679 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp, cred);
4680 		if (leasep != NULL)
4681 			NFSROOTFS_GETATTRBIT(&attrbits);
4682 		else
4683 			NFSSTATFS_GETATTRBIT(&attrbits);
4684 		(void) nfsrv_putattrbit(nd, &attrbits);
4685 		nd->nd_flag |= ND_USEGSSNAME;
4686 		error = nfscl_request(nd, vp, p, cred);
4687 		if (error)
4688 			return (error);
4689 		if (nd->nd_repstat == 0) {
4690 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4691 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, leasep, NULL,
4692 			    p, cred);
4693 			if (!error) {
4694 				nmp->nm_fsid[0] = nap->na_filesid[0];
4695 				nmp->nm_fsid[1] = nap->na_filesid[1];
4696 				NFSSETHASSETFSID(nmp);
4697 				*attrflagp = 1;
4698 			}
4699 		} else {
4700 			error = nd->nd_repstat;
4701 		}
4702 		if (error)
4703 			goto nfsmout;
4704 	} else {
4705 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp, NULL);
4706 		error = nfscl_request(nd, vp, p, cred);
4707 		if (error)
4708 			return (error);
4709 		if (nd->nd_flag & ND_NFSV3) {
4710 			error = nfscl_postop_attr(nd, nap, attrflagp);
4711 			if (error)
4712 				goto nfsmout;
4713 		}
4714 		if (nd->nd_repstat) {
4715 			error = nd->nd_repstat;
4716 			goto nfsmout;
4717 		}
4718 		NFSM_DISSECT(tl, u_int32_t *,
4719 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4720 	}
4721 	if (NFSHASNFSV3(nmp)) {
4722 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4723 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4724 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4725 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4726 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4727 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4728 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4729 	} else if (NFSHASNFSV4(nmp) == 0) {
4730 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4731 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4732 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4733 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4734 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4735 	}
4736 nfsmout:
4737 	m_freem(nd->nd_mrep);
4738 	return (error);
4739 }
4740 
4741 /*
4742  * nfs pathconf rpc
4743  */
4744 int
4745 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4746     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4747 {
4748 	struct nfsrv_descript nfsd, *nd = &nfsd;
4749 	struct nfsmount *nmp;
4750 	u_int32_t *tl;
4751 	nfsattrbit_t attrbits;
4752 	int error;
4753 	struct nfsnode *np;
4754 
4755 	*attrflagp = 0;
4756 	nmp = VFSTONFS(vp->v_mount);
4757 	if (NFSHASNFSV4(nmp)) {
4758 		np = VTONFS(vp);
4759 		if ((nmp->nm_privflag & NFSMNTP_FAKEROOTFH) != 0 &&
4760 		    nmp->nm_fhsize == 0) {
4761 			/* Attempt to get the actual root file handle. */
4762 			error = nfsrpc_getdirpath(nmp, NFSMNT_DIRPATH(nmp),
4763 			    cred, p);
4764 			if (error != 0)
4765 				return (EACCES);
4766 			if (np->n_fhp->nfh_len == NFSX_FHMAX + 1)
4767 				nfscl_statfs(vp, cred, p);
4768 		}
4769 		/*
4770 		 * For V4, you actually do a getattr.
4771 		 */
4772 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp, cred);
4773 		NFSPATHCONF_GETATTRBIT(&attrbits);
4774 		(void) nfsrv_putattrbit(nd, &attrbits);
4775 		nd->nd_flag |= ND_USEGSSNAME;
4776 		error = nfscl_request(nd, vp, p, cred);
4777 		if (error)
4778 			return (error);
4779 		if (nd->nd_repstat == 0) {
4780 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4781 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4782 			    cred);
4783 			if (!error)
4784 				*attrflagp = 1;
4785 		} else {
4786 			error = nd->nd_repstat;
4787 		}
4788 	} else {
4789 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp, NULL);
4790 		error = nfscl_request(nd, vp, p, cred);
4791 		if (error)
4792 			return (error);
4793 		error = nfscl_postop_attr(nd, nap, attrflagp);
4794 		if (nd->nd_repstat && !error)
4795 			error = nd->nd_repstat;
4796 		if (!error) {
4797 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4798 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4799 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4800 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4801 			pc->pc_chownrestricted =
4802 			    fxdr_unsigned(u_int32_t, *tl++);
4803 			pc->pc_caseinsensitive =
4804 			    fxdr_unsigned(u_int32_t, *tl++);
4805 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4806 		}
4807 	}
4808 nfsmout:
4809 	m_freem(nd->nd_mrep);
4810 	return (error);
4811 }
4812 
4813 /*
4814  * nfs version 3 fsinfo rpc call
4815  */
4816 int
4817 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4818     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4819 {
4820 	u_int32_t *tl;
4821 	struct nfsrv_descript nfsd, *nd = &nfsd;
4822 	int error;
4823 
4824 	*attrflagp = 0;
4825 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp, NULL);
4826 	error = nfscl_request(nd, vp, p, cred);
4827 	if (error)
4828 		return (error);
4829 	error = nfscl_postop_attr(nd, nap, attrflagp);
4830 	if (nd->nd_repstat && !error)
4831 		error = nd->nd_repstat;
4832 	if (!error) {
4833 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4834 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4835 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4836 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4837 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4838 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4839 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4840 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4841 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4842 		tl += 2;
4843 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4844 		tl += 2;
4845 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4846 	}
4847 nfsmout:
4848 	m_freem(nd->nd_mrep);
4849 	return (error);
4850 }
4851 
4852 /*
4853  * This function performs the Renew RPC.
4854  */
4855 int
4856 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4857     NFSPROC_T *p)
4858 {
4859 	u_int32_t *tl;
4860 	struct nfsrv_descript nfsd;
4861 	struct nfsrv_descript *nd = &nfsd;
4862 	struct nfsmount *nmp;
4863 	int error;
4864 	struct nfssockreq *nrp;
4865 	struct nfsclsession *tsep;
4866 
4867 	nmp = clp->nfsc_nmp;
4868 	if (nmp == NULL)
4869 		return (0);
4870 	if (dsp == NULL)
4871 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4872 		    0, cred);
4873 	else
4874 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4875 		    &dsp->nfsclds_sess, 0, 0, NULL);
4876 	if (!NFSHASNFSV4N(nmp)) {
4877 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4878 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4879 		tsep = nfsmnt_mdssession(nmp);
4880 		*tl++ = tsep->nfsess_clientid.lval[0];
4881 		*tl = tsep->nfsess_clientid.lval[1];
4882 	}
4883 	nrp = NULL;
4884 	if (dsp != NULL)
4885 		nrp = dsp->nfsclds_sockp;
4886 	if (nrp == NULL)
4887 		/* If NULL, use the MDS socket. */
4888 		nrp = &nmp->nm_sockreq;
4889 	nd->nd_flag |= ND_USEGSSNAME;
4890 	if (dsp == NULL)
4891 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4892 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4893 	else {
4894 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4895 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4896 		if (error == ENXIO)
4897 			nfscl_cancelreqs(dsp);
4898 	}
4899 	if (error)
4900 		return (error);
4901 	error = nd->nd_repstat;
4902 	m_freem(nd->nd_mrep);
4903 	return (error);
4904 }
4905 
4906 /*
4907  * This function performs the Releaselockowner RPC.
4908  */
4909 int
4910 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4911     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4912 {
4913 	struct nfsrv_descript nfsd, *nd = &nfsd;
4914 	u_int32_t *tl;
4915 	int error;
4916 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4917 	struct nfsclsession *tsep;
4918 
4919 	if (NFSHASNFSV4N(nmp)) {
4920 		/* For NFSv4.1, do a FreeStateID. */
4921 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4922 		    NULL, 0, 0, cred);
4923 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4924 	} else {
4925 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4926 		    NULL, 0, 0, NULL);
4927 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4928 		tsep = nfsmnt_mdssession(nmp);
4929 		*tl++ = tsep->nfsess_clientid.lval[0];
4930 		*tl = tsep->nfsess_clientid.lval[1];
4931 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4932 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4933 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4934 	}
4935 	nd->nd_flag |= ND_USEGSSNAME;
4936 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4937 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4938 	if (error)
4939 		return (error);
4940 	error = nd->nd_repstat;
4941 	m_freem(nd->nd_mrep);
4942 	return (error);
4943 }
4944 
4945 /*
4946  * This function performs the Compound to get the mount pt FH.
4947  */
4948 int
4949 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4950     NFSPROC_T *p)
4951 {
4952 	u_int32_t *tl;
4953 	struct nfsrv_descript nfsd;
4954 	struct nfsrv_descript *nd = &nfsd;
4955 	u_char *cp, *cp2, *fhp;
4956 	int error, cnt, len, setnil;
4957 	u_int32_t *opcntp;
4958 
4959 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4960 	    0, NULL);
4961 	cp = dirpath;
4962 	cnt = 0;
4963 	do {
4964 		setnil = 0;
4965 		while (*cp == '/')
4966 			cp++;
4967 		cp2 = cp;
4968 		while (*cp2 != '\0' && *cp2 != '/')
4969 			cp2++;
4970 		if (*cp2 == '/') {
4971 			setnil = 1;
4972 			*cp2 = '\0';
4973 		}
4974 		if (cp2 != cp) {
4975 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4976 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4977 			nfsm_strtom(nd, cp, strlen(cp));
4978 			cnt++;
4979 		}
4980 		if (setnil)
4981 			*cp2++ = '/';
4982 		cp = cp2;
4983 	} while (*cp != '\0');
4984 	if (NFSHASNFSV4N(nmp))
4985 		/* Has a Sequence Op done by nfscl_reqstart(). */
4986 		*opcntp = txdr_unsigned(3 + cnt);
4987 	else
4988 		*opcntp = txdr_unsigned(2 + cnt);
4989 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4990 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4991 	nd->nd_flag |= ND_USEGSSNAME;
4992 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4993 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4994 	if (error)
4995 		return (error);
4996 	if (nd->nd_repstat == 0) {
4997 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4998 		tl += (2 + 2 * cnt);
4999 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
5000 			len > NFSX_FHMAX) {
5001 			nd->nd_repstat = NFSERR_BADXDR;
5002 		} else {
5003 			fhp = malloc(len + 1, M_TEMP, M_WAITOK);
5004 			nd->nd_repstat = nfsrv_mtostr(nd, fhp, len);
5005 			if (nd->nd_repstat == 0) {
5006 				NFSLOCKMNT(nmp);
5007 				if (nmp->nm_fhsize == 0) {
5008 					NFSBCOPY(fhp, nmp->nm_fh, len);
5009 					nmp->nm_fhsize = len;
5010 				}
5011 				NFSUNLOCKMNT(nmp);
5012 			}
5013 			free(fhp, M_TEMP);
5014 		}
5015 	}
5016 	error = nd->nd_repstat;
5017 nfsmout:
5018 	m_freem(nd->nd_mrep);
5019 	return (error);
5020 }
5021 
5022 /*
5023  * This function performs the Delegreturn RPC.
5024  */
5025 int
5026 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
5027     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
5028 {
5029 	u_int32_t *tl;
5030 	struct nfsrv_descript nfsd;
5031 	struct nfsrv_descript *nd = &nfsd;
5032 	int error;
5033 
5034 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
5035 	    dp->nfsdl_fhlen, NULL, NULL, 0, 0, cred);
5036 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
5037 	if (NFSHASNFSV4N(nmp))
5038 		*tl++ = 0;
5039 	else
5040 		*tl++ = dp->nfsdl_stateid.seqid;
5041 	*tl++ = dp->nfsdl_stateid.other[0];
5042 	*tl++ = dp->nfsdl_stateid.other[1];
5043 	*tl = dp->nfsdl_stateid.other[2];
5044 	if (syscred)
5045 		nd->nd_flag |= ND_USEGSSNAME;
5046 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5047 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5048 	if (error)
5049 		return (error);
5050 	error = nd->nd_repstat;
5051 	m_freem(nd->nd_mrep);
5052 	return (error);
5053 }
5054 
5055 /*
5056  * nfs getacl call.
5057  */
5058 int
5059 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p, struct acl *aclp)
5060 {
5061 	struct nfsrv_descript nfsd, *nd = &nfsd;
5062 	int error;
5063 	nfsattrbit_t attrbits;
5064 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5065 
5066 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
5067 		return (EOPNOTSUPP);
5068 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp, cred);
5069 	NFSZERO_ATTRBIT(&attrbits);
5070 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
5071 	(void) nfsrv_putattrbit(nd, &attrbits);
5072 	error = nfscl_request(nd, vp, p, cred);
5073 	if (error)
5074 		return (error);
5075 	if (!nd->nd_repstat)
5076 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
5077 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
5078 	else
5079 		error = nd->nd_repstat;
5080 	m_freem(nd->nd_mrep);
5081 	return (error);
5082 }
5083 
5084 /*
5085  * nfs setacl call.
5086  */
5087 int
5088 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p, struct acl *aclp)
5089 {
5090 	int error;
5091 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5092 
5093 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
5094 		return (EOPNOTSUPP);
5095 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL);
5096 	return (error);
5097 }
5098 
5099 /*
5100  * nfs setacl call.
5101  */
5102 static int
5103 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
5104     struct acl *aclp, nfsv4stateid_t *stateidp)
5105 {
5106 	struct nfsrv_descript nfsd, *nd = &nfsd;
5107 	int error;
5108 	nfsattrbit_t attrbits;
5109 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5110 
5111 	if (!NFSHASNFSV4(nmp))
5112 		return (EOPNOTSUPP);
5113 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp, cred);
5114 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
5115 	NFSZERO_ATTRBIT(&attrbits);
5116 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
5117 	(void) nfsv4_fillattr(nd, vp->v_mount, vp, aclp, NULL, NULL, 0,
5118 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
5119 	error = nfscl_request(nd, vp, p, cred);
5120 	if (error)
5121 		return (error);
5122 	/* Don't care about the pre/postop attributes */
5123 	m_freem(nd->nd_mrep);
5124 	return (nd->nd_repstat);
5125 }
5126 
5127 /*
5128  * Do the NFSv4.1 Exchange ID.
5129  */
5130 int
5131 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
5132     struct nfssockreq *nrp, int minorvers, uint32_t exchflags,
5133     struct nfsclds **dspp, struct ucred *cred, NFSPROC_T *p)
5134 {
5135 	uint32_t *tl, v41flags;
5136 	struct nfsrv_descript nfsd;
5137 	struct nfsrv_descript *nd = &nfsd;
5138 	struct nfsclds *dsp;
5139 	struct timespec verstime;
5140 	int error, len;
5141 
5142 	*dspp = NULL;
5143 	if (minorvers == 0)
5144 		minorvers = nmp->nm_minorvers;
5145 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL,
5146 	    NFS_VER4, minorvers, NULL);
5147 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5148 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
5149 	*tl = txdr_unsigned(clp->nfsc_rev);
5150 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
5151 
5152 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
5153 	*tl++ = txdr_unsigned(exchflags);
5154 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
5155 
5156 	/* Set the implementation id4 */
5157 	*tl = txdr_unsigned(1);
5158 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
5159 	(void) nfsm_strtom(nd, version, strlen(version));
5160 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
5161 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
5162 	verstime.tv_nsec = 0;
5163 	txdr_nfsv4time(&verstime, tl);
5164 	nd->nd_flag |= ND_USEGSSNAME;
5165 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
5166 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5167 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
5168 	    (int)nd->nd_repstat);
5169 	if (error != 0)
5170 		return (error);
5171 	if (nd->nd_repstat == 0) {
5172 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
5173 		len = fxdr_unsigned(int, *(tl + 7));
5174 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
5175 			error = NFSERR_BADXDR;
5176 			goto nfsmout;
5177 		}
5178 		dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
5179 		    M_WAITOK | M_ZERO);
5180 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
5181 		dsp->nfsclds_servownlen = len;
5182 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
5183 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
5184 		dsp->nfsclds_sess.nfsess_sequenceid =
5185 		    fxdr_unsigned(uint32_t, *tl++);
5186 		v41flags = fxdr_unsigned(uint32_t, *tl);
5187 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
5188 		    NFSHASPNFSOPT(nmp)) {
5189 			NFSCL_DEBUG(1, "set PNFS\n");
5190 			NFSLOCKMNT(nmp);
5191 			nmp->nm_state |= NFSSTA_PNFS;
5192 			NFSUNLOCKMNT(nmp);
5193 			dsp->nfsclds_flags |= NFSCLDS_MDS;
5194 		}
5195 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
5196 			dsp->nfsclds_flags |= NFSCLDS_DS;
5197 		if (minorvers == NFSV42_MINORVERSION)
5198 			dsp->nfsclds_flags |= NFSCLDS_MINORV2;
5199 		if (len > 0)
5200 			nd->nd_repstat = nfsrv_mtostr(nd,
5201 			    dsp->nfsclds_serverown, len);
5202 		if (nd->nd_repstat == 0) {
5203 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5204 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5205 			    NULL, MTX_DEF);
5206 			nfscl_initsessionslots(&dsp->nfsclds_sess);
5207 			*dspp = dsp;
5208 		} else
5209 			free(dsp, M_NFSCLDS);
5210 	}
5211 	error = nd->nd_repstat;
5212 nfsmout:
5213 	m_freem(nd->nd_mrep);
5214 	return (error);
5215 }
5216 
5217 /*
5218  * Do the NFSv4.1 Create Session.
5219  */
5220 int
5221 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
5222     struct nfssockreq *nrp, struct nfsclds *dsp, uint32_t sequenceid, int mds,
5223     struct ucred *cred, NFSPROC_T *p)
5224 {
5225 	uint32_t crflags, maxval, *tl;
5226 	struct nfsrv_descript nfsd;
5227 	struct nfsrv_descript *nd = &nfsd;
5228 	int error, irdcnt, minorvers;
5229 
5230 	/* Make sure nm_rsize, nm_wsize is set. */
5231 	if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
5232 		nmp->nm_rsize = NFS_MAXBSIZE;
5233 	if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
5234 		nmp->nm_wsize = NFS_MAXBSIZE;
5235 	if (dsp == NULL)
5236 		minorvers = nmp->nm_minorvers;
5237 	else if ((dsp->nfsclds_flags & NFSCLDS_MINORV2) != 0)
5238 		minorvers = NFSV42_MINORVERSION;
5239 	else
5240 		minorvers = NFSV41_MINORVERSION;
5241 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL,
5242 	    NFS_VER4, minorvers, NULL);
5243 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5244 	*tl++ = sep->nfsess_clientid.lval[0];
5245 	*tl++ = sep->nfsess_clientid.lval[1];
5246 	*tl++ = txdr_unsigned(sequenceid);
5247 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
5248 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
5249 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
5250 	*tl = txdr_unsigned(crflags);
5251 
5252 	/* Fill in fore channel attributes. */
5253 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5254 	*tl++ = 0;				/* Header pad size */
5255 	if ((nd->nd_flag & ND_NFSV42) != 0 && mds != 0 && sb_max_adj >=
5256 	    nmp->nm_wsize && sb_max_adj >= nmp->nm_rsize) {
5257 		/*
5258 		 * NFSv4.2 Extended Attribute operations may want to do
5259 		 * requests/replies that are larger than nm_rsize/nm_wsize.
5260 		 */
5261 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
5262 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
5263 	} else {
5264 		*tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);
5265 		*tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);
5266 	}
5267 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
5268 	*tl++ = txdr_unsigned(20);		/* Max operations */
5269 	*tl++ = txdr_unsigned(64);		/* Max slots */
5270 	*tl = 0;				/* No rdma ird */
5271 
5272 	/* Fill in back channel attributes. */
5273 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5274 	*tl++ = 0;				/* Header pad size */
5275 	*tl++ = txdr_unsigned(10000);		/* Max request size */
5276 	*tl++ = txdr_unsigned(10000);		/* Max response size */
5277 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
5278 	*tl++ = txdr_unsigned(4);		/* Max operations */
5279 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
5280 	*tl = 0;				/* No rdma ird */
5281 
5282 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
5283 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
5284 
5285 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
5286 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
5287 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
5288 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
5289 	*tl++ = 0;				/* Null machine name */
5290 	*tl++ = 0;				/* Uid == 0 */
5291 	*tl++ = 0;				/* Gid == 0 */
5292 	*tl = 0;				/* No additional gids */
5293 	nd->nd_flag |= ND_USEGSSNAME;
5294 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
5295 	    NFS_VER4, NULL, 1, NULL, NULL);
5296 	if (error != 0)
5297 		return (error);
5298 	if (nd->nd_repstat == 0) {
5299 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
5300 		    2 * NFSX_UNSIGNED);
5301 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
5302 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
5303 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
5304 		crflags = fxdr_unsigned(uint32_t, *tl);
5305 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
5306 			NFSLOCKMNT(nmp);
5307 			nmp->nm_state |= NFSSTA_SESSPERSIST;
5308 			NFSUNLOCKMNT(nmp);
5309 		}
5310 
5311 		/* Get the fore channel slot count. */
5312 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5313 		tl++;			/* Skip the header pad size. */
5314 
5315 		/* Make sure nm_wsize is small enough. */
5316 		maxval = fxdr_unsigned(uint32_t, *tl++);
5317 		while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
5318 			if (nmp->nm_wsize > 8096)
5319 				nmp->nm_wsize /= 2;
5320 			else
5321 				break;
5322 		}
5323 		sep->nfsess_maxreq = maxval;
5324 
5325 		/* Make sure nm_rsize is small enough. */
5326 		maxval = fxdr_unsigned(uint32_t, *tl++);
5327 		while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
5328 			if (nmp->nm_rsize > 8096)
5329 				nmp->nm_rsize /= 2;
5330 			else
5331 				break;
5332 		}
5333 		sep->nfsess_maxresp = maxval;
5334 
5335 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
5336 		tl++;
5337 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
5338 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
5339 		irdcnt = fxdr_unsigned(int, *tl);
5340 		if (irdcnt < 0 || irdcnt > 1) {
5341 			error = NFSERR_BADXDR;
5342 			goto nfsmout;
5343 		}
5344 		if (irdcnt > 0)
5345 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
5346 
5347 		/* and the back channel slot count. */
5348 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5349 		tl += 5;
5350 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
5351 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
5352 	}
5353 	error = nd->nd_repstat;
5354 nfsmout:
5355 	m_freem(nd->nd_mrep);
5356 	return (error);
5357 }
5358 
5359 /*
5360  * Do the NFSv4.1 Destroy Client.
5361  */
5362 int
5363 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
5364     struct ucred *cred, NFSPROC_T *p)
5365 {
5366 	uint32_t *tl;
5367 	struct nfsrv_descript nfsd;
5368 	struct nfsrv_descript *nd = &nfsd;
5369 	int error;
5370 	struct nfsclsession *tsep;
5371 
5372 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
5373 	    0, NULL);
5374 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5375 	tsep = nfsmnt_mdssession(nmp);
5376 	*tl++ = tsep->nfsess_clientid.lval[0];
5377 	*tl = tsep->nfsess_clientid.lval[1];
5378 	nd->nd_flag |= ND_USEGSSNAME;
5379 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5380 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5381 	if (error != 0)
5382 		return (error);
5383 	error = nd->nd_repstat;
5384 	m_freem(nd->nd_mrep);
5385 	return (error);
5386 }
5387 
5388 /*
5389  * Do the NFSv4.1 LayoutGet.
5390  */
5391 static int
5392 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
5393     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
5394     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
5395     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p)
5396 {
5397 	struct nfsrv_descript nfsd, *nd = &nfsd;
5398 	int error;
5399 
5400 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
5401 	    0, cred);
5402 	nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
5403 	    layouttype, layoutlen, 0);
5404 	nd->nd_flag |= ND_USEGSSNAME;
5405 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5406 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5407 	NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
5408 	if (error != 0)
5409 		return (error);
5410 	if (nd->nd_repstat == 0)
5411 		error = nfsrv_parselayoutget(nmp, nd, stateidp, retonclosep,
5412 		    flhp);
5413 	if (error == 0 && nd->nd_repstat != 0)
5414 		error = nd->nd_repstat;
5415 	m_freem(nd->nd_mrep);
5416 	return (error);
5417 }
5418 
5419 /*
5420  * Do the NFSv4.1 Get Device Info.
5421  */
5422 int
5423 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
5424     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
5425     NFSPROC_T *p)
5426 {
5427 	uint32_t cnt, *tl, vers, minorvers;
5428 	struct nfsrv_descript nfsd;
5429 	struct nfsrv_descript *nd = &nfsd;
5430 	struct sockaddr_in sin, ssin;
5431 	struct sockaddr_in6 sin6, ssin6;
5432 	struct nfsclds *dsp = NULL, **dspp, **gotdspp;
5433 	struct nfscldevinfo *ndi;
5434 	int addrcnt = 0, bitcnt, error, gotminor, gotvers, i, isudp, j;
5435 	int stripecnt;
5436 	uint8_t stripeindex;
5437 	sa_family_t af, safilled;
5438 
5439 	ssin.sin_port = 0;		/* To shut up compiler. */
5440 	ssin.sin_addr.s_addr = 0;	/* ditto */
5441 	*ndip = NULL;
5442 	ndi = NULL;
5443 	gotdspp = NULL;
5444 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
5445 	    0, cred);
5446 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5447 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
5448 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5449 	*tl++ = txdr_unsigned(layouttype);
5450 	*tl++ = txdr_unsigned(100000);
5451 	if (notifybitsp != NULL && *notifybitsp != 0) {
5452 		*tl = txdr_unsigned(1);		/* One word of bits. */
5453 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5454 		*tl = txdr_unsigned(*notifybitsp);
5455 	} else
5456 		*tl = txdr_unsigned(0);
5457 	nd->nd_flag |= ND_USEGSSNAME;
5458 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5459 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5460 	if (error != 0)
5461 		return (error);
5462 	if (nd->nd_repstat == 0) {
5463 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5464 		if (layouttype != fxdr_unsigned(int, *tl))
5465 			printf("EEK! devinfo layout type not same!\n");
5466 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
5467 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5468 			stripecnt = fxdr_unsigned(int, *tl);
5469 			NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
5470 			if (stripecnt < 1 || stripecnt > 4096) {
5471 				printf("pNFS File layout devinfo stripecnt %d:"
5472 				    " out of range\n", stripecnt);
5473 				error = NFSERR_BADXDR;
5474 				goto nfsmout;
5475 			}
5476 			NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
5477 			    NFSX_UNSIGNED);
5478 			addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
5479 			NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
5480 			if (addrcnt < 1 || addrcnt > 128) {
5481 				printf("NFS devinfo addrcnt %d: out of range\n",
5482 				    addrcnt);
5483 				error = NFSERR_BADXDR;
5484 				goto nfsmout;
5485 			}
5486 
5487 			/*
5488 			 * Now we know how many stripe indices and addresses, so
5489 			 * we can allocate the structure the correct size.
5490 			 */
5491 			i = (stripecnt * sizeof(uint8_t)) /
5492 			    sizeof(struct nfsclds *) + 1;
5493 			NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5494 			ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5495 			    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5496 			    M_ZERO);
5497 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5498 			    NFSX_V4DEVICEID);
5499 			ndi->nfsdi_refcnt = 0;
5500 			ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5501 			ndi->nfsdi_stripecnt = stripecnt;
5502 			ndi->nfsdi_addrcnt = addrcnt;
5503 			/* Fill in the stripe indices. */
5504 			for (i = 0; i < stripecnt; i++) {
5505 				stripeindex = fxdr_unsigned(uint8_t, *tl++);
5506 				NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5507 				if (stripeindex >= addrcnt) {
5508 					printf("pNFS File Layout devinfo"
5509 					    " stripeindex %d: too big\n",
5510 					    (int)stripeindex);
5511 					error = NFSERR_BADXDR;
5512 					goto nfsmout;
5513 				}
5514 				nfsfldi_setstripeindex(ndi, i, stripeindex);
5515 			}
5516 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
5517 			/* For Flex File, we only get one address list. */
5518 			ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5519 			    M_NFSDEVINFO, M_WAITOK | M_ZERO);
5520 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5521 			    NFSX_V4DEVICEID);
5522 			ndi->nfsdi_refcnt = 0;
5523 			ndi->nfsdi_flags = NFSDI_FLEXFILE;
5524 			addrcnt = ndi->nfsdi_addrcnt = 1;
5525 		}
5526 
5527 		/* Now, dissect the server address(es). */
5528 		safilled = AF_UNSPEC;
5529 		for (i = 0; i < addrcnt; i++) {
5530 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5531 			cnt = fxdr_unsigned(uint32_t, *tl);
5532 			if (cnt == 0) {
5533 				printf("NFS devinfo 0 len addrlist\n");
5534 				error = NFSERR_BADXDR;
5535 				goto nfsmout;
5536 			}
5537 			dspp = nfsfldi_addr(ndi, i);
5538 			safilled = AF_UNSPEC;
5539 			for (j = 0; j < cnt; j++) {
5540 				error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5541 				    &isudp);
5542 				if (error != 0 && error != EPERM) {
5543 					error = NFSERR_BADXDR;
5544 					goto nfsmout;
5545 				}
5546 				if (error == 0 && isudp == 0) {
5547 					/*
5548 					 * The priority is:
5549 					 * - Same address family.
5550 					 * Save the address and dspp, so that
5551 					 * the connection can be done after
5552 					 * parsing is complete.
5553 					 */
5554 					if (safilled == AF_UNSPEC ||
5555 					    (af == nmp->nm_nam->sa_family &&
5556 					     safilled != nmp->nm_nam->sa_family)
5557 					   ) {
5558 						if (af == AF_INET)
5559 							ssin = sin;
5560 						else
5561 							ssin6 = sin6;
5562 						safilled = af;
5563 						gotdspp = dspp;
5564 					}
5565 				}
5566 			}
5567 		}
5568 
5569 		gotvers = NFS_VER4;	/* Default NFSv4.1 for File Layout. */
5570 		gotminor = NFSV41_MINORVERSION;
5571 		/* For Flex File, we will take one of the versions to use. */
5572 		if (layouttype == NFSLAYOUT_FLEXFILE) {
5573 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5574 			j = fxdr_unsigned(int, *tl);
5575 			if (j < 1 || j > NFSDEV_MAXVERS) {
5576 				printf("pNFS: too many versions\n");
5577 				error = NFSERR_BADXDR;
5578 				goto nfsmout;
5579 			}
5580 			gotvers = 0;
5581 			gotminor = 0;
5582 			for (i = 0; i < j; i++) {
5583 				NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5584 				vers = fxdr_unsigned(uint32_t, *tl++);
5585 				minorvers = fxdr_unsigned(uint32_t, *tl++);
5586 				if (vers == NFS_VER3)
5587 					minorvers = 0;
5588 				if ((vers == NFS_VER4 && ((minorvers ==
5589 				    NFSV41_MINORVERSION && gotminor == 0) ||
5590 				    minorvers == NFSV42_MINORVERSION)) ||
5591 				    (vers == NFS_VER3 && gotvers == 0)) {
5592 					gotvers = vers;
5593 					gotminor = minorvers;
5594 					/* We'll take this one. */
5595 					ndi->nfsdi_versindex = i;
5596 					ndi->nfsdi_vers = vers;
5597 					ndi->nfsdi_minorvers = minorvers;
5598 					ndi->nfsdi_rsize = fxdr_unsigned(
5599 					    uint32_t, *tl++);
5600 					ndi->nfsdi_wsize = fxdr_unsigned(
5601 					    uint32_t, *tl++);
5602 					if (*tl == newnfs_true)
5603 						ndi->nfsdi_flags |=
5604 						    NFSDI_TIGHTCOUPLED;
5605 					else
5606 						ndi->nfsdi_flags &=
5607 						    ~NFSDI_TIGHTCOUPLED;
5608 				}
5609 			}
5610 			if (gotvers == 0) {
5611 				printf("pNFS: no NFSv3, NFSv4.1 or NFSv4.2\n");
5612 				error = NFSERR_BADXDR;
5613 				goto nfsmout;
5614 			}
5615 		}
5616 
5617 		/* And the notify bits. */
5618 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5619 		bitcnt = fxdr_unsigned(int, *tl);
5620 		if (bitcnt > 0) {
5621 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5622 			if (notifybitsp != NULL)
5623 				*notifybitsp =
5624 				    fxdr_unsigned(uint32_t, *tl);
5625 		}
5626 		if (safilled != AF_UNSPEC) {
5627 			KASSERT(ndi != NULL, ("ndi is NULL"));
5628 			*ndip = ndi;
5629 		} else
5630 			error = EPERM;
5631 		if (error == 0) {
5632 			/*
5633 			 * Now we can do a TCP connection for the correct
5634 			 * NFS version and IP address.
5635 			 */
5636 			error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5637 			    gotvers, gotminor, &dsp, p);
5638 		}
5639 		if (error == 0) {
5640 			KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5641 			*gotdspp = dsp;
5642 		}
5643 	}
5644 	if (nd->nd_repstat != 0 && error == 0)
5645 		error = nd->nd_repstat;
5646 nfsmout:
5647 	if (error != 0 && ndi != NULL)
5648 		nfscl_freedevinfo(ndi);
5649 	m_freem(nd->nd_mrep);
5650 	return (error);
5651 }
5652 
5653 /*
5654  * Do the NFSv4.1 LayoutCommit.
5655  */
5656 int
5657 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5658     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5659     int layouttype, struct ucred *cred, NFSPROC_T *p)
5660 {
5661 	uint32_t *tl;
5662 	struct nfsrv_descript nfsd, *nd = &nfsd;
5663 	int error;
5664 
5665 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5666 	    0, 0, cred);
5667 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5668 	    NFSX_STATEID);
5669 	txdr_hyper(off, tl);
5670 	tl += 2;
5671 	txdr_hyper(len, tl);
5672 	tl += 2;
5673 	if (reclaim != 0)
5674 		*tl++ = newnfs_true;
5675 	else
5676 		*tl++ = newnfs_false;
5677 	*tl++ = txdr_unsigned(stateidp->seqid);
5678 	*tl++ = stateidp->other[0];
5679 	*tl++ = stateidp->other[1];
5680 	*tl++ = stateidp->other[2];
5681 	*tl++ = newnfs_true;
5682 	if (lastbyte < off)
5683 		lastbyte = off;
5684 	else if (lastbyte >= (off + len))
5685 		lastbyte = off + len - 1;
5686 	txdr_hyper(lastbyte, tl);
5687 	tl += 2;
5688 	*tl++ = newnfs_false;
5689 	*tl++ = txdr_unsigned(layouttype);
5690 	/* All supported layouts are 0 length. */
5691 	*tl = txdr_unsigned(0);
5692 	nd->nd_flag |= ND_USEGSSNAME;
5693 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5694 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5695 	if (error != 0)
5696 		return (error);
5697 	error = nd->nd_repstat;
5698 	m_freem(nd->nd_mrep);
5699 	return (error);
5700 }
5701 
5702 /*
5703  * Do the NFSv4.1 LayoutReturn.
5704  */
5705 int
5706 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5707     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5708     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5709     uint32_t stat, uint32_t op, char *devid)
5710 {
5711 	uint32_t *tl;
5712 	struct nfsrv_descript nfsd, *nd = &nfsd;
5713 	uint64_t tu64;
5714 	int error;
5715 
5716 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5717 	    0, 0, cred);
5718 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5719 	if (reclaim != 0)
5720 		*tl++ = newnfs_true;
5721 	else
5722 		*tl++ = newnfs_false;
5723 	*tl++ = txdr_unsigned(layouttype);
5724 	*tl++ = txdr_unsigned(iomode);
5725 	*tl = txdr_unsigned(layoutreturn);
5726 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5727 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5728 		    NFSX_UNSIGNED);
5729 		txdr_hyper(offset, tl);
5730 		tl += 2;
5731 		txdr_hyper(len, tl);
5732 		tl += 2;
5733 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5734 		*tl++ = txdr_unsigned(stateidp->seqid);
5735 		*tl++ = stateidp->other[0];
5736 		*tl++ = stateidp->other[1];
5737 		*tl++ = stateidp->other[2];
5738 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5739 			*tl = txdr_unsigned(0);
5740 		else if (layouttype == NFSLAYOUT_FLEXFILE) {
5741 			if (stat != 0) {
5742 				*tl = txdr_unsigned(2 * NFSX_HYPER +
5743 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5744 				    NFSX_UNSIGNED);
5745 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5746 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5747 				    NFSX_UNSIGNED);
5748 				*tl++ = txdr_unsigned(1);	/* One error. */
5749 				tu64 = 0;			/* Offset. */
5750 				txdr_hyper(tu64, tl); tl += 2;
5751 				tu64 = UINT64_MAX;		/* Length. */
5752 				txdr_hyper(tu64, tl); tl += 2;
5753 				NFSBCOPY(stateidp, tl, NFSX_STATEID);
5754 				tl += (NFSX_STATEID / NFSX_UNSIGNED);
5755 				*tl++ = txdr_unsigned(1);	/* One error. */
5756 				NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5757 				tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5758 				*tl++ = txdr_unsigned(stat);
5759 				*tl++ = txdr_unsigned(op);
5760 			} else {
5761 				*tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5762 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5763 				/* No ioerrs. */
5764 				*tl++ = 0;
5765 			}
5766 			*tl = 0;	/* No stats yet. */
5767 		}
5768 	}
5769 	nd->nd_flag |= ND_USEGSSNAME;
5770 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5771 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5772 	if (error != 0)
5773 		return (error);
5774 	if (nd->nd_repstat == 0) {
5775 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5776 		if (*tl != 0) {
5777 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5778 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5779 			stateidp->other[0] = *tl++;
5780 			stateidp->other[1] = *tl++;
5781 			stateidp->other[2] = *tl;
5782 		}
5783 	} else
5784 		error = nd->nd_repstat;
5785 nfsmout:
5786 	m_freem(nd->nd_mrep);
5787 	return (error);
5788 }
5789 
5790 /*
5791  * Do the NFSv4.2 LayoutError.
5792  */
5793 static int
5794 nfsrpc_layouterror(struct nfsmount *nmp, uint8_t *fh, int fhlen, uint64_t offset,
5795     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5796     uint32_t stat, uint32_t op, char *devid)
5797 {
5798 	uint32_t *tl;
5799 	struct nfsrv_descript nfsd, *nd = &nfsd;
5800 	int error;
5801 
5802 	nfscl_reqstart(nd, NFSPROC_LAYOUTERROR, nmp, fh, fhlen, NULL, NULL,
5803 	    0, 0, cred);
5804 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5805 	    NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5806 	txdr_hyper(offset, tl); tl += 2;
5807 	txdr_hyper(len, tl); tl += 2;
5808 	*tl++ = txdr_unsigned(stateidp->seqid);
5809 	*tl++ = stateidp->other[0];
5810 	*tl++ = stateidp->other[1];
5811 	*tl++ = stateidp->other[2];
5812 	*tl++ = txdr_unsigned(1);
5813 	NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5814 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5815 	*tl++ = txdr_unsigned(stat);
5816 	*tl = txdr_unsigned(op);
5817 	nd->nd_flag |= ND_USEGSSNAME;
5818 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5819 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5820 	if (error != 0)
5821 		return (error);
5822 	if (nd->nd_repstat != 0)
5823 		error = nd->nd_repstat;
5824 	m_freem(nd->nd_mrep);
5825 	return (error);
5826 }
5827 
5828 /*
5829  * Acquire a layout and devinfo, if possible. The caller must have acquired
5830  * a reference count on the nfsclclient structure before calling this.
5831  * Return the layout in lypp with a reference count on it, if successful.
5832  */
5833 static int
5834 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5835     int iomode, uint32_t rw, uint32_t *notifybitsp, nfsv4stateid_t *stateidp,
5836     uint64_t off, struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5837 {
5838 	struct nfscllayout *lyp;
5839 	struct nfsclflayout *flp;
5840 	struct nfsclflayouthead flh;
5841 	int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5842 	nfsv4stateid_t stateid;
5843 	struct nfsclsession *tsep;
5844 
5845 	*lypp = NULL;
5846 	if (NFSHASFLEXFILE(nmp))
5847 		layouttype = NFSLAYOUT_FLEXFILE;
5848 	else
5849 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
5850 	/*
5851 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5852 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5853 	 * flp == NULL.
5854 	 */
5855 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5856 	    off, rw, &flp, &recalled);
5857 	islocked = 0;
5858 	if (lyp == NULL || flp == NULL) {
5859 		if (recalled != 0)
5860 			return (EIO);
5861 		LIST_INIT(&flh);
5862 		tsep = nfsmnt_mdssession(nmp);
5863 		layoutlen = tsep->nfsess_maxcache -
5864 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5865 		if (lyp == NULL) {
5866 			stateid.seqid = 0;
5867 			stateid.other[0] = stateidp->other[0];
5868 			stateid.other[1] = stateidp->other[1];
5869 			stateid.other[2] = stateidp->other[2];
5870 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5871 			    nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5872 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5873 			    &retonclose, &flh, cred, p);
5874 		} else {
5875 			islocked = 1;
5876 			stateid.seqid = lyp->nfsly_stateid.seqid;
5877 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5878 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5879 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5880 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5881 			    nfhp->nfh_len, iomode, off, UINT64_MAX,
5882 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5883 			    &retonclose, &flh, cred, p);
5884 		}
5885 		error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5886 		    nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5887 		    &flh, layouttype, error, NULL, cred, p);
5888 		if (error == 0)
5889 			*lypp = lyp;
5890 		else if (islocked != 0)
5891 			nfscl_rellayout(lyp, 1);
5892 	} else
5893 		*lypp = lyp;
5894 	return (error);
5895 }
5896 
5897 /*
5898  * Do a TCP connection plus exchange id and create session.
5899  * If successful, a "struct nfsclds" is linked into the list for the
5900  * mount point and a pointer to it is returned.
5901  */
5902 static int
5903 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5904     struct sockaddr_in6 *sin6, sa_family_t af, int vers, int minorvers,
5905     struct nfsclds **dspp, NFSPROC_T *p)
5906 {
5907 	struct sockaddr_in *msad, *sad;
5908 	struct sockaddr_in6 *msad6, *sad6;
5909 	struct nfsclclient *clp;
5910 	struct nfssockreq *nrp;
5911 	struct nfsclds *dsp, *tdsp;
5912 	int error, firsttry;
5913 	enum nfsclds_state retv;
5914 	uint32_t sequenceid = 0;
5915 
5916 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5917 	    ("nfsrpc_fillsa: NULL nr_cred"));
5918 	NFSLOCKCLSTATE();
5919 	clp = nmp->nm_clp;
5920 	NFSUNLOCKCLSTATE();
5921 	if (clp == NULL)
5922 		return (EPERM);
5923 	if (af == AF_INET) {
5924 		NFSLOCKMNT(nmp);
5925 		/*
5926 		 * Check to see if we already have a session for this
5927 		 * address that is usable for a DS.
5928 		 * Note that the MDS's address is in a different place
5929 		 * than the sessions already acquired for DS's.
5930 		 */
5931 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5932 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5933 		while (tdsp != NULL) {
5934 			if (msad != NULL && msad->sin_family == AF_INET &&
5935 			    sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5936 			    sin->sin_port == msad->sin_port &&
5937 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5938 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5939 				*dspp = tdsp;
5940 				NFSUNLOCKMNT(nmp);
5941 				NFSCL_DEBUG(4, "fnd same addr\n");
5942 				return (0);
5943 			}
5944 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5945 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5946 				msad = (struct sockaddr_in *)
5947 				    tdsp->nfsclds_sockp->nr_nam;
5948 			else
5949 				msad = NULL;
5950 		}
5951 		NFSUNLOCKMNT(nmp);
5952 
5953 		/* No IP address match, so look for new/trunked one. */
5954 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5955 		sad->sin_len = sizeof(*sad);
5956 		sad->sin_family = AF_INET;
5957 		sad->sin_port = sin->sin_port;
5958 		sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5959 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5960 		nrp->nr_nam = (struct sockaddr *)sad;
5961 	} else if (af == AF_INET6) {
5962 		NFSLOCKMNT(nmp);
5963 		/*
5964 		 * Check to see if we already have a session for this
5965 		 * address that is usable for a DS.
5966 		 * Note that the MDS's address is in a different place
5967 		 * than the sessions already acquired for DS's.
5968 		 */
5969 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5970 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5971 		while (tdsp != NULL) {
5972 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5973 			    IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5974 			    &msad6->sin6_addr) &&
5975 			    sin6->sin6_port == msad6->sin6_port &&
5976 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5977 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5978 				*dspp = tdsp;
5979 				NFSUNLOCKMNT(nmp);
5980 				return (0);
5981 			}
5982 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5983 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5984 				msad6 = (struct sockaddr_in6 *)
5985 				    tdsp->nfsclds_sockp->nr_nam;
5986 			else
5987 				msad6 = NULL;
5988 		}
5989 		NFSUNLOCKMNT(nmp);
5990 
5991 		/* No IP address match, so look for new/trunked one. */
5992 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5993 		sad6->sin6_len = sizeof(*sad6);
5994 		sad6->sin6_family = AF_INET6;
5995 		sad6->sin6_port = sin6->sin6_port;
5996 		NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5997 		    sizeof(struct in6_addr));
5998 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5999 		nrp->nr_nam = (struct sockaddr *)sad6;
6000 	} else
6001 		return (EPERM);
6002 
6003 	nrp->nr_sotype = SOCK_STREAM;
6004 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
6005 	nrp->nr_prog = NFS_PROG;
6006 	nrp->nr_vers = vers;
6007 
6008 	/*
6009 	 * Use the credentials that were used for the mount, which are
6010 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
6011 	 * Ref. counting the credentials with crhold() is probably not
6012 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
6013 	 * unmount, but I did it anyhow.
6014 	 */
6015 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
6016 	error = newnfs_connect(nmp, nrp, NULL, p, 0, false, &nrp->nr_client);
6017 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
6018 
6019 	dsp = NULL;
6020 	/* Now, do the exchangeid and create session. */
6021 	if (error == 0) {
6022 		if (vers == NFS_VER4) {
6023 			firsttry = 0;
6024 			do {
6025 				error = nfsrpc_exchangeid(nmp, clp, nrp,
6026 				    minorvers, NFSV4EXCH_USEPNFSDS, &dsp,
6027 				    nrp->nr_cred, p);
6028 				NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
6029 				if (error == NFSERR_MINORVERMISMATCH)
6030 					minorvers = NFSV42_MINORVERSION;
6031 			} while (error == NFSERR_MINORVERMISMATCH &&
6032 			    firsttry++ == 0);
6033 			if (error != 0)
6034 				newnfs_disconnect(NULL, nrp);
6035 		} else {
6036 			dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
6037 			    M_WAITOK | M_ZERO);
6038 			dsp->nfsclds_flags |= NFSCLDS_DS;
6039 			dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
6040 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
6041 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
6042 			    NULL, MTX_DEF);
6043 		}
6044 	}
6045 	if (error == 0) {
6046 		dsp->nfsclds_sockp = nrp;
6047 		if (vers == NFS_VER4) {
6048 			NFSLOCKMNT(nmp);
6049 			retv = nfscl_getsameserver(nmp, dsp, &tdsp,
6050 			    &sequenceid);
6051 			NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
6052 			if (retv == NFSDSP_USETHISSESSION &&
6053 			    nfscl_dssameconn != 0) {
6054 				NFSLOCKDS(tdsp);
6055 				tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
6056 				NFSUNLOCKDS(tdsp);
6057 				NFSUNLOCKMNT(nmp);
6058 				/*
6059 				 * If there is already a session for this
6060 				 * server, use it.
6061 				 */
6062 				newnfs_disconnect(NULL, nrp);
6063 				nfscl_freenfsclds(dsp);
6064 				*dspp = tdsp;
6065 				return (0);
6066 			}
6067 			if (retv == NFSDSP_NOTFOUND)
6068 				sequenceid =
6069 				    dsp->nfsclds_sess.nfsess_sequenceid;
6070 			NFSUNLOCKMNT(nmp);
6071 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
6072 			    nrp, dsp, sequenceid, 0, nrp->nr_cred, p);
6073 			NFSCL_DEBUG(3, "DS createsess=%d\n", error);
6074 		}
6075 	} else {
6076 		NFSFREECRED(nrp->nr_cred);
6077 		NFSFREEMUTEX(&nrp->nr_mtx);
6078 		free(nrp->nr_nam, M_SONAME);
6079 		free(nrp, M_NFSSOCKREQ);
6080 	}
6081 	if (error == 0) {
6082 		NFSCL_DEBUG(3, "add DS session\n");
6083 		/*
6084 		 * Put it at the end of the list. That way the list
6085 		 * is ordered by when the entry was added. This matters
6086 		 * since the one done first is the one that should be
6087 		 * used for sequencid'ing any subsequent create sessions.
6088 		 */
6089 		NFSLOCKMNT(nmp);
6090 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
6091 		NFSUNLOCKMNT(nmp);
6092 		*dspp = dsp;
6093 	} else if (dsp != NULL) {
6094 		newnfs_disconnect(NULL, nrp);
6095 		nfscl_freenfsclds(dsp);
6096 	}
6097 	return (error);
6098 }
6099 
6100 /*
6101  * Do the NFSv4.1 Reclaim Complete.
6102  */
6103 int
6104 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
6105 {
6106 	uint32_t *tl;
6107 	struct nfsrv_descript nfsd;
6108 	struct nfsrv_descript *nd = &nfsd;
6109 	int error;
6110 
6111 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
6112 	    0, cred);
6113 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
6114 	*tl = newnfs_false;
6115 	nd->nd_flag |= ND_USEGSSNAME;
6116 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
6117 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
6118 	if (error != 0)
6119 		return (error);
6120 	error = nd->nd_repstat;
6121 	m_freem(nd->nd_mrep);
6122 	return (error);
6123 }
6124 
6125 /*
6126  * Initialize the slot tables for a session.
6127  */
6128 static void
6129 nfscl_initsessionslots(struct nfsclsession *sep)
6130 {
6131 	int i;
6132 
6133 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
6134 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
6135 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
6136 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
6137 	}
6138 	for (i = 0; i < 64; i++)
6139 		sep->nfsess_slotseq[i] = 0;
6140 	sep->nfsess_slots = 0;
6141 	sep->nfsess_badslots = 0;
6142 }
6143 
6144 /*
6145  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
6146  */
6147 int
6148 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6149     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
6150 {
6151 	struct nfsnode *np = VTONFS(vp);
6152 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6153 	struct nfscllayout *layp;
6154 	struct nfscldevinfo *dip;
6155 	struct nfsclflayout *rflp;
6156 	struct mbuf *m, *m2;
6157 	struct nfsclwritedsdorpc *drpc, *tdrpc;
6158 	nfsv4stateid_t stateid;
6159 	struct ucred *newcred;
6160 	uint64_t lastbyte, len, off, oresid, xfer;
6161 	int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
6162 	void *lckp;
6163 	uint8_t *dev;
6164 	void *iovbase = NULL;
6165 	size_t iovlen = 0;
6166 	off_t offs = 0;
6167 	ssize_t resid = 0;
6168 	uint32_t op;
6169 
6170 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
6171 	    (np->n_flag & NNOLAYOUT) != 0)
6172 		return (EIO);
6173 	/* Now, get a reference cnt on the clientid for this mount. */
6174 	if (nfscl_getref(nmp) == 0)
6175 		return (EIO);
6176 
6177 	/* Find an appropriate stateid. */
6178 	newcred = NFSNEWCRED(cred);
6179 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
6180 	    rwaccess, 1, newcred, p, &stateid, &lckp);
6181 	if (error != 0) {
6182 		NFSFREECRED(newcred);
6183 		nfscl_relref(nmp);
6184 		return (error);
6185 	}
6186 	/* Search for a layout for this file. */
6187 	off = uiop->uio_offset;
6188 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
6189 	    np->n_fhp->nfh_len, off, rwaccess, &rflp, &recalled);
6190 	if (layp == NULL || rflp == NULL) {
6191 		if (recalled != 0) {
6192 			NFSFREECRED(newcred);
6193 			if (lckp != NULL)
6194 				nfscl_lockderef(lckp);
6195 			nfscl_relref(nmp);
6196 			return (EIO);
6197 		}
6198 		if (layp != NULL) {
6199 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
6200 			layp = NULL;
6201 		}
6202 		/* Try and get a Layout, if it is supported. */
6203 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
6204 		    (np->n_flag & NWRITEOPENED) != 0)
6205 			iolaymode = NFSLAYOUTIOMODE_RW;
6206 		else
6207 			iolaymode = NFSLAYOUTIOMODE_READ;
6208 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
6209 		    rwaccess, NULL, &stateid, off, &layp, newcred, p);
6210 		if (error != 0) {
6211 			NFSLOCKNODE(np);
6212 			np->n_flag |= NNOLAYOUT;
6213 			NFSUNLOCKNODE(np);
6214 			if (lckp != NULL)
6215 				nfscl_lockderef(lckp);
6216 			NFSFREECRED(newcred);
6217 			if (layp != NULL)
6218 				nfscl_rellayout(layp, 0);
6219 			nfscl_relref(nmp);
6220 			return (error);
6221 		}
6222 	}
6223 
6224 	/*
6225 	 * Loop around finding a layout that works for the first part of
6226 	 * this I/O operation, and then call the function that actually
6227 	 * does the RPC.
6228 	 */
6229 	eof = 0;
6230 	len = (uint64_t)uiop->uio_resid;
6231 	while (len > 0 && error == 0 && eof == 0) {
6232 		off = uiop->uio_offset;
6233 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
6234 		if (error == 0) {
6235 			oresid = xfer = (uint64_t)uiop->uio_resid;
6236 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
6237 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
6238 			/*
6239 			 * For Flex File layout with mirrored DSs, select one
6240 			 * of them at random for reads. For writes and commits,
6241 			 * do all mirrors.
6242 			 */
6243 			m = NULL;
6244 			tdrpc = drpc = NULL;
6245 			firstmirror = 0;
6246 			mirrorcnt = 1;
6247 			if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
6248 			    (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
6249 				if (rwaccess == NFSV4OPEN_ACCESSREAD) {
6250 					firstmirror = arc4random() % mirrorcnt;
6251 					mirrorcnt = firstmirror + 1;
6252 				} else {
6253 					if (docommit == 0) {
6254 						/*
6255 						 * Save values, so uiop can be
6256 						 * rolled back upon a write
6257 						 * error.
6258 						 */
6259 						offs = uiop->uio_offset;
6260 						resid = uiop->uio_resid;
6261 						iovbase =
6262 						    uiop->uio_iov->iov_base;
6263 						iovlen = uiop->uio_iov->iov_len;
6264 						m = nfsm_uiombuflist(uiop, len,
6265 						    0);
6266 					}
6267 					tdrpc = drpc = malloc(sizeof(*drpc) *
6268 					    (mirrorcnt - 1), M_TEMP, M_WAITOK |
6269 					    M_ZERO);
6270 				}
6271 			}
6272 			for (i = firstmirror; i < mirrorcnt && error == 0; i++){
6273 				m2 = NULL;
6274 				if (m != NULL && i < mirrorcnt - 1)
6275 					m2 = m_copym(m, 0, M_COPYALL, M_WAITOK);
6276 				else {
6277 					m2 = m;
6278 					m = NULL;
6279 				}
6280 				if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
6281 					dev = rflp->nfsfl_ffm[i].dev;
6282 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
6283 					    rflp->nfsfl_ffm[i].devp);
6284 				} else {
6285 					dev = rflp->nfsfl_dev;
6286 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
6287 					    rflp->nfsfl_devp);
6288 				}
6289 				if (dip != NULL) {
6290 					if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
6291 					    != 0)
6292 						error = nfscl_dofflayoutio(vp,
6293 						    uiop, iomode, must_commit,
6294 						    &eof, &stateid, rwaccess,
6295 						    dip, layp, rflp, off, xfer,
6296 						    i, docommit, m2, tdrpc,
6297 						    newcred, p);
6298 					else
6299 						error = nfscl_doflayoutio(vp,
6300 						    uiop, iomode, must_commit,
6301 						    &eof, &stateid, rwaccess,
6302 						    dip, layp, rflp, off, xfer,
6303 						    docommit, newcred, p);
6304 					nfscl_reldevinfo(dip);
6305 				} else {
6306 					if (m2 != NULL)
6307 						m_freem(m2);
6308 					error = EIO;
6309 				}
6310 				tdrpc++;
6311 			}
6312 			if (m != NULL)
6313 				m_freem(m);
6314 			tdrpc = drpc;
6315 			timo = hz / 50;		/* Wait for 20msec. */
6316 			if (timo < 1)
6317 				timo = 1;
6318 			for (i = firstmirror; i < mirrorcnt - 1 &&
6319 			    tdrpc != NULL; i++, tdrpc++) {
6320 				/*
6321 				 * For the unused drpc entries, both inprog and
6322 				 * err == 0, so this loop won't break.
6323 				 */
6324 				while (tdrpc->inprog != 0 && tdrpc->done == 0)
6325 					tsleep(&tdrpc->tsk, PVFS, "clrpcio",
6326 					    timo);
6327 				if (error == 0 && tdrpc->err != 0)
6328 					error = tdrpc->err;
6329 				if (rwaccess != NFSV4OPEN_ACCESSREAD &&
6330 				    docommit == 0 && *must_commit == 0 &&
6331 				    tdrpc->must_commit == 1)
6332 					*must_commit = 1;
6333 			}
6334 			free(drpc, M_TEMP);
6335 			if (error == 0) {
6336 				if (mirrorcnt > 1 && rwaccess ==
6337 				    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
6338 					NFSLOCKCLSTATE();
6339 					layp->nfsly_flags |= NFSLY_WRITTEN;
6340 					NFSUNLOCKCLSTATE();
6341 				}
6342 				lastbyte = off + xfer - 1;
6343 				NFSLOCKCLSTATE();
6344 				if (lastbyte > layp->nfsly_lastbyte)
6345 					layp->nfsly_lastbyte = lastbyte;
6346 				NFSUNLOCKCLSTATE();
6347 			} else if (error == NFSERR_OPENMODE &&
6348 			    rwaccess == NFSV4OPEN_ACCESSREAD) {
6349 				NFSLOCKMNT(nmp);
6350 				nmp->nm_state |= NFSSTA_OPENMODE;
6351 				NFSUNLOCKMNT(nmp);
6352 			} else if ((error == NFSERR_NOSPC ||
6353 			    error == NFSERR_IO || error == NFSERR_NXIO) &&
6354 			    nmp->nm_minorvers == NFSV42_MINORVERSION) {
6355 				if (docommit != 0)
6356 					op = NFSV4OP_COMMIT;
6357 				else if (rwaccess == NFSV4OPEN_ACCESSREAD)
6358 					op = NFSV4OP_READ;
6359 				else
6360 					op = NFSV4OP_WRITE;
6361 				nfsrpc_layouterror(nmp, np->n_fhp->nfh_fh,
6362 				    np->n_fhp->nfh_len, off, xfer,
6363 				    &layp->nfsly_stateid, newcred, p, error, op,
6364 				    dip->nfsdi_deviceid);
6365 				error = EIO;
6366 			} else
6367 				error = EIO;
6368 			if (error == 0)
6369 				len -= (oresid - (uint64_t)uiop->uio_resid);
6370 			else if (mirrorcnt > 1 && rwaccess ==
6371 			    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
6372 				/*
6373 				 * In case the rpc gets retried, roll the
6374 				 * uio fields changed by nfsm_uiombuflist()
6375 				 * back.
6376 				 */
6377 				uiop->uio_offset = offs;
6378 				uiop->uio_resid = resid;
6379 				uiop->uio_iov->iov_base = iovbase;
6380 				uiop->uio_iov->iov_len = iovlen;
6381 			}
6382 		}
6383 	}
6384 	if (lckp != NULL)
6385 		nfscl_lockderef(lckp);
6386 	NFSFREECRED(newcred);
6387 	nfscl_rellayout(layp, 0);
6388 	nfscl_relref(nmp);
6389 	return (error);
6390 }
6391 
6392 /*
6393  * Find a file layout that will handle the first bytes of the requested
6394  * range and return the information from it needed to the I/O operation.
6395  */
6396 int
6397 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
6398     struct nfsclflayout **retflpp)
6399 {
6400 	struct nfsclflayout *flp, *nflp, *rflp;
6401 	uint32_t rw;
6402 
6403 	rflp = NULL;
6404 	rw = rwaccess;
6405 	/* For reading, do the Read list first and then the Write list. */
6406 	do {
6407 		if (rw == NFSV4OPEN_ACCESSREAD)
6408 			flp = LIST_FIRST(&lyp->nfsly_flayread);
6409 		else
6410 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
6411 		while (flp != NULL) {
6412 			nflp = LIST_NEXT(flp, nfsfl_list);
6413 			if (flp->nfsfl_off > off)
6414 				break;
6415 			if (flp->nfsfl_end > off &&
6416 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
6417 				rflp = flp;
6418 			flp = nflp;
6419 		}
6420 		if (rw == NFSV4OPEN_ACCESSREAD)
6421 			rw = NFSV4OPEN_ACCESSWRITE;
6422 		else
6423 			rw = 0;
6424 	} while (rw != 0);
6425 	if (rflp != NULL) {
6426 		/* This one covers the most bytes starting at off. */
6427 		*retflpp = rflp;
6428 		return (0);
6429 	}
6430 	return (EIO);
6431 }
6432 
6433 /*
6434  * Do I/O using an NFSv4.1 or NFSv4.2 file layout.
6435  */
6436 static int
6437 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6438     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6439     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6440     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
6441 {
6442 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
6443 	int commit_thru_mds, error, stripe_index, stripe_pos, minorvers;
6444 	struct nfsnode *np;
6445 	struct nfsfh *fhp;
6446 	struct nfsclds **dspp;
6447 
6448 	np = VTONFS(vp);
6449 	rel_off = off - flp->nfsfl_patoff;
6450 	stripe_unit_size = flp->nfsfl_util & NFSFLAYUTIL_STRIPE_MASK;
6451 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
6452 	    dp->nfsdi_stripecnt;
6453 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
6454 	error = 0;
6455 
6456 	/* Loop around, doing I/O for each stripe unit. */
6457 	while (len > 0 && error == 0) {
6458 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
6459 		dspp = nfsfldi_addr(dp, stripe_index);
6460 		if (((*dspp)->nfsclds_flags & NFSCLDS_MINORV2) != 0)
6461 			minorvers = NFSV42_MINORVERSION;
6462 		else
6463 			minorvers = NFSV41_MINORVERSION;
6464 		if (len > transfer && docommit == 0)
6465 			xfer = transfer;
6466 		else
6467 			xfer = len;
6468 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
6469 			/* Dense layout. */
6470 			if (stripe_pos >= flp->nfsfl_fhcnt)
6471 				return (EIO);
6472 			fhp = flp->nfsfl_fh[stripe_pos];
6473 			io_off = (rel_off / (stripe_unit_size *
6474 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
6475 			    rel_off % stripe_unit_size;
6476 		} else {
6477 			/* Sparse layout. */
6478 			if (flp->nfsfl_fhcnt > 1) {
6479 				if (stripe_index >= flp->nfsfl_fhcnt)
6480 					return (EIO);
6481 				fhp = flp->nfsfl_fh[stripe_index];
6482 			} else if (flp->nfsfl_fhcnt == 1)
6483 				fhp = flp->nfsfl_fh[0];
6484 			else
6485 				fhp = np->n_fhp;
6486 			io_off = off;
6487 		}
6488 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
6489 			commit_thru_mds = 1;
6490 			if (docommit != 0)
6491 				error = EIO;
6492 		} else {
6493 			commit_thru_mds = 0;
6494 			NFSLOCKNODE(np);
6495 			np->n_flag |= NDSCOMMIT;
6496 			NFSUNLOCKNODE(np);
6497 		}
6498 		if (docommit != 0) {
6499 			if (error == 0)
6500 				error = nfsrpc_commitds(vp, io_off, xfer,
6501 				    *dspp, fhp, NFS_VER4, minorvers, cred, p);
6502 			if (error == 0) {
6503 				/*
6504 				 * Set both eof and uio_resid = 0 to end any
6505 				 * loops.
6506 				 */
6507 				*eofp = 1;
6508 				uiop->uio_resid = 0;
6509 			} else {
6510 				NFSLOCKNODE(np);
6511 				np->n_flag &= ~NDSCOMMIT;
6512 				NFSUNLOCKNODE(np);
6513 			}
6514 		} else if (rwflag == NFSV4OPEN_ACCESSREAD)
6515 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6516 			    io_off, xfer, fhp, 0, NFS_VER4, minorvers, cred, p);
6517 		else {
6518 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
6519 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
6520 			    0, NFS_VER4, minorvers, cred, p);
6521 			if (error == 0) {
6522 				NFSLOCKCLSTATE();
6523 				lyp->nfsly_flags |= NFSLY_WRITTEN;
6524 				NFSUNLOCKCLSTATE();
6525 			}
6526 		}
6527 		if (error == 0) {
6528 			transfer = stripe_unit_size;
6529 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
6530 			len -= xfer;
6531 			off += xfer;
6532 		}
6533 	}
6534 	return (error);
6535 }
6536 
6537 /*
6538  * Do I/O using an NFSv4.1 flex file layout.
6539  */
6540 static int
6541 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6542     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6543     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6544     uint64_t len, int mirror, int docommit, struct mbuf *mp,
6545     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6546 {
6547 	uint64_t xfer;
6548 	int error;
6549 	struct nfsnode *np;
6550 	struct nfsfh *fhp;
6551 	struct nfsclds **dspp;
6552 	struct ucred *tcred;
6553 	struct mbuf *m, *m2;
6554 	uint32_t copylen;
6555 
6556 	np = VTONFS(vp);
6557 	error = 0;
6558 	NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6559 	    (uintmax_t)len);
6560 	/* Loop around, doing I/O for each stripe unit. */
6561 	while (len > 0 && error == 0) {
6562 		dspp = nfsfldi_addr(dp, 0);
6563 		fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6564 		stateidp = &flp->nfsfl_ffm[mirror].st;
6565 		NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6566 		    mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6567 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6568 			tcred = NFSNEWCRED(cred);
6569 			tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6570 			tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6571 			tcred->cr_ngroups = 1;
6572 		} else
6573 			tcred = cred;
6574 		if (rwflag == NFSV4OPEN_ACCESSREAD)
6575 			copylen = dp->nfsdi_rsize;
6576 		else {
6577 			copylen = dp->nfsdi_wsize;
6578 			if (len > copylen && mp != NULL) {
6579 				/*
6580 				 * When a mirrored configuration needs to do
6581 				 * multiple writes to each mirror, all writes
6582 				 * except the last one must be a multiple of
6583 				 * 4 bytes.  This is required so that the XDR
6584 				 * does not need padding.
6585 				 * If possible, clip the size to an exact
6586 				 * multiple of the mbuf length, so that the
6587 				 * split will be on an mbuf boundary.
6588 				 */
6589 				copylen &= 0xfffffffc;
6590 				if (copylen > mp->m_len)
6591 					copylen = copylen / mp->m_len *
6592 					    mp->m_len;
6593 			}
6594 		}
6595 		NFSLOCKNODE(np);
6596 		np->n_flag |= NDSCOMMIT;
6597 		NFSUNLOCKNODE(np);
6598 		if (len > copylen && docommit == 0)
6599 			xfer = copylen;
6600 		else
6601 			xfer = len;
6602 		if (docommit != 0) {
6603 			if (error == 0) {
6604 				/*
6605 				 * Do last mirrored DS commit with this thread.
6606 				 */
6607 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6608 					error = nfsio_commitds(vp, off, xfer,
6609 					    *dspp, fhp, dp->nfsdi_vers,
6610 					    dp->nfsdi_minorvers, drpc, tcred,
6611 					    p);
6612 				else
6613 					error = nfsrpc_commitds(vp, off, xfer,
6614 					    *dspp, fhp, dp->nfsdi_vers,
6615 					    dp->nfsdi_minorvers, tcred, p);
6616 				NFSCL_DEBUG(4, "commitds=%d\n", error);
6617 				if (error != 0 && error != EACCES && error !=
6618 				    ESTALE) {
6619 					NFSCL_DEBUG(4,
6620 					    "DS layreterr for commit\n");
6621 					nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6622 					    lyp, *dspp);
6623 				}
6624 			}
6625 			NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6626 			if (error == 0) {
6627 				/*
6628 				 * Set both eof and uio_resid = 0 to end any
6629 				 * loops.
6630 				 */
6631 				*eofp = 1;
6632 				uiop->uio_resid = 0;
6633 			} else {
6634 				NFSLOCKNODE(np);
6635 				np->n_flag &= ~NDSCOMMIT;
6636 				NFSUNLOCKNODE(np);
6637 			}
6638 		} else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6639 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6640 			    off, xfer, fhp, 1, dp->nfsdi_vers,
6641 			    dp->nfsdi_minorvers, tcred, p);
6642 			NFSCL_DEBUG(4, "readds=%d\n", error);
6643 			if (error != 0 && error != EACCES && error != ESTALE) {
6644 				NFSCL_DEBUG(4, "DS layreterr for read\n");
6645 				nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6646 				    *dspp);
6647 			}
6648 		} else {
6649 			if (flp->nfsfl_mirrorcnt == 1) {
6650 				error = nfsrpc_writeds(vp, uiop, iomode,
6651 				    must_commit, stateidp, *dspp, off, xfer,
6652 				    fhp, 0, 1, dp->nfsdi_vers,
6653 				    dp->nfsdi_minorvers, tcred, p);
6654 				if (error == 0) {
6655 					NFSLOCKCLSTATE();
6656 					lyp->nfsly_flags |= NFSLY_WRITTEN;
6657 					NFSUNLOCKCLSTATE();
6658 				}
6659 			} else {
6660 				m = mp;
6661 				if (xfer < len) {
6662 					/* The mbuf list must be split. */
6663 					m2 = nfsm_split(mp, xfer);
6664 					if (m2 != NULL)
6665 						mp = m2;
6666 					else {
6667 						m_freem(mp);
6668 						error = EIO;
6669 					}
6670 				}
6671 				NFSCL_DEBUG(4, "mcopy len=%jd xfer=%jd\n",
6672 				    (uintmax_t)len, (uintmax_t)xfer);
6673 				/*
6674 				 * Do last write to a mirrored DS with this
6675 				 * thread.
6676 				 */
6677 				if (error == 0) {
6678 					if (mirror < flp->nfsfl_mirrorcnt - 1)
6679 						error = nfsio_writedsmir(vp,
6680 						    iomode, must_commit,
6681 						    stateidp, *dspp, off,
6682 						    xfer, fhp, m,
6683 						    dp->nfsdi_vers,
6684 						    dp->nfsdi_minorvers, drpc,
6685 						    tcred, p);
6686 					else
6687 						error = nfsrpc_writedsmir(vp,
6688 						    iomode, must_commit,
6689 						    stateidp, *dspp, off,
6690 						    xfer, fhp, m,
6691 						    dp->nfsdi_vers,
6692 						    dp->nfsdi_minorvers, tcred,
6693 						    p);
6694 				}
6695 				NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6696 				if (error != 0 && error != EACCES && error !=
6697 				    ESTALE) {
6698 					NFSCL_DEBUG(4,
6699 					    "DS layreterr for write\n");
6700 					nfscl_dserr(NFSV4OP_WRITE, error, dp,
6701 					    lyp, *dspp);
6702 				}
6703 			}
6704 		}
6705 		NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6706 		if (error == 0) {
6707 			len -= xfer;
6708 			off += xfer;
6709 		}
6710 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6711 			NFSFREECRED(tcred);
6712 	}
6713 	NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6714 	return (error);
6715 }
6716 
6717 /*
6718  * The actual read RPC done to a DS.
6719  */
6720 static int
6721 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6722     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6723     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6724 {
6725 	uint32_t *tl;
6726 	int attrflag, error, retlen;
6727 	struct nfsrv_descript nfsd;
6728 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6729 	struct nfsrv_descript *nd = &nfsd;
6730 	struct nfssockreq *nrp;
6731 	struct nfsvattr na;
6732 
6733 	nd->nd_mrep = NULL;
6734 	if (vers == 0 || vers == NFS_VER4) {
6735 		nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6736 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6737 		    NULL);
6738 		vers = NFS_VER4;
6739 		NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6740 		if (flex != 0)
6741 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6742 		else
6743 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6744 	} else {
6745 		nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6746 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6747 		    NULL);
6748 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READ]);
6749 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READDS]);
6750 		NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6751 	}
6752 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6753 	txdr_hyper(io_off, tl);
6754 	*(tl + 2) = txdr_unsigned(len);
6755 	nrp = dsp->nfsclds_sockp;
6756 	NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6757 	if (nrp == NULL)
6758 		/* If NULL, use the MDS socket. */
6759 		nrp = &nmp->nm_sockreq;
6760 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6761 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6762 	NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6763 	    error);
6764 	if (error != 0)
6765 		return (error);
6766 	if (vers == NFS_VER3) {
6767 		error = nfscl_postop_attr(nd, &na, &attrflag);
6768 		NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6769 		if (error != 0)
6770 			goto nfsmout;
6771 	}
6772 	if (nd->nd_repstat != 0) {
6773 		error = nd->nd_repstat;
6774 		goto nfsmout;
6775 	}
6776 	if (vers == NFS_VER3) {
6777 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6778 		*eofp = fxdr_unsigned(int, *(tl + 1));
6779 	} else {
6780 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6781 		*eofp = fxdr_unsigned(int, *tl);
6782 	}
6783 	NFSM_STRSIZ(retlen, len);
6784 	NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6785 	error = nfsm_mbufuio(nd, uiop, retlen);
6786 nfsmout:
6787 	if (nd->nd_mrep != NULL)
6788 		m_freem(nd->nd_mrep);
6789 	return (error);
6790 }
6791 
6792 /*
6793  * The actual write RPC done to a DS.
6794  */
6795 static int
6796 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6797     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6798     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6799     struct ucred *cred, NFSPROC_T *p)
6800 {
6801 	uint32_t *tl;
6802 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6803 	int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6804 	int32_t backup;
6805 	struct nfsrv_descript nfsd;
6806 	struct nfsrv_descript *nd = &nfsd;
6807 	struct nfssockreq *nrp;
6808 	struct nfsvattr na;
6809 
6810 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6811 	nd->nd_mrep = NULL;
6812 	if (vers == 0 || vers == NFS_VER4) {
6813 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6814 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6815 		    NULL);
6816 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6817 		vers = NFS_VER4;
6818 		if (flex != 0)
6819 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6820 		else
6821 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6822 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6823 	} else {
6824 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6825 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6826 		    NULL);
6827 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6828 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6829 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6830 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6831 	}
6832 	txdr_hyper(io_off, tl);
6833 	tl += 2;
6834 	if (vers == NFS_VER3)
6835 		*tl++ = txdr_unsigned(len);
6836 	*tl++ = txdr_unsigned(*iomode);
6837 	*tl = txdr_unsigned(len);
6838 	nfsm_uiombuf(nd, uiop, len);
6839 	nrp = dsp->nfsclds_sockp;
6840 	if (nrp == NULL)
6841 		/* If NULL, use the MDS socket. */
6842 		nrp = &nmp->nm_sockreq;
6843 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6844 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6845 	NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6846 	    nd->nd_repstat);
6847 	if (error != 0)
6848 		return (error);
6849 	if (nd->nd_repstat != 0) {
6850 		/*
6851 		 * In case the rpc gets retried, roll
6852 		 * the uio fields changed by nfsm_uiombuf()
6853 		 * back.
6854 		 */
6855 		uiop->uio_offset -= len;
6856 		uiop->uio_resid += len;
6857 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base - len;
6858 		uiop->uio_iov->iov_len += len;
6859 		error = nd->nd_repstat;
6860 	} else {
6861 		if (vers == NFS_VER3) {
6862 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6863 			    NULL);
6864 			NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6865 			if (error != 0)
6866 				goto nfsmout;
6867 		}
6868 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6869 		rlen = fxdr_unsigned(int, *tl++);
6870 		NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6871 		if (rlen == 0) {
6872 			error = NFSERR_IO;
6873 			goto nfsmout;
6874 		} else if (rlen < len) {
6875 			backup = len - rlen;
6876 			uiop->uio_iov->iov_base =
6877 			    (char *)uiop->uio_iov->iov_base - backup;
6878 			uiop->uio_iov->iov_len += backup;
6879 			uiop->uio_offset -= backup;
6880 			uiop->uio_resid += backup;
6881 			len = rlen;
6882 		}
6883 		commit = fxdr_unsigned(int, *tl++);
6884 
6885 		/*
6886 		 * Return the lowest commitment level
6887 		 * obtained by any of the RPCs.
6888 		 */
6889 		if (committed == NFSWRITE_FILESYNC)
6890 			committed = commit;
6891 		else if (committed == NFSWRITE_DATASYNC &&
6892 		    commit == NFSWRITE_UNSTABLE)
6893 			committed = commit;
6894 		if (commit_thru_mds != 0) {
6895 			NFSLOCKMNT(nmp);
6896 			if (!NFSHASWRITEVERF(nmp)) {
6897 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6898 				NFSSETWRITEVERF(nmp);
6899 			} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF) &&
6900 			    *must_commit != 2) {
6901 				*must_commit = 1;
6902 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6903 			}
6904 			NFSUNLOCKMNT(nmp);
6905 		} else {
6906 			NFSLOCKDS(dsp);
6907 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6908 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6909 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6910 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF) &&
6911 			    *must_commit != 2) {
6912 				*must_commit = 1;
6913 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6914 			}
6915 			NFSUNLOCKDS(dsp);
6916 		}
6917 	}
6918 nfsmout:
6919 	if (nd->nd_mrep != NULL)
6920 		m_freem(nd->nd_mrep);
6921 	*iomode = committed;
6922 	if (nd->nd_repstat != 0 && error == 0)
6923 		error = nd->nd_repstat;
6924 	return (error);
6925 }
6926 
6927 /*
6928  * The actual write RPC done to a DS.
6929  * This variant is called from a separate kernel process for mirrors.
6930  * Any short write is considered an IO error.
6931  */
6932 static int
6933 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6934     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6935     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6936     struct ucred *cred, NFSPROC_T *p)
6937 {
6938 	uint32_t *tl;
6939 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6940 	int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6941 	struct nfsrv_descript nfsd;
6942 	struct nfsrv_descript *nd = &nfsd;
6943 	struct nfssockreq *nrp;
6944 	struct nfsvattr na;
6945 
6946 	nd->nd_mrep = NULL;
6947 	if (vers == 0 || vers == NFS_VER4) {
6948 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6949 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6950 		    NULL);
6951 		vers = NFS_VER4;
6952 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6953 		    minorvers);
6954 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6955 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6956 	} else {
6957 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6958 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6959 		    NULL);
6960 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6961 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6962 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6963 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6964 	}
6965 	txdr_hyper(io_off, tl);
6966 	tl += 2;
6967 	if (vers == NFS_VER3)
6968 		*tl++ = txdr_unsigned(len);
6969 	*tl++ = txdr_unsigned(*iomode);
6970 	*tl = txdr_unsigned(len);
6971 	if (len > 0) {
6972 		/* Put data in mbuf chain. */
6973 		nd->nd_mb->m_next = m;
6974 	}
6975 	nrp = dsp->nfsclds_sockp;
6976 	if (nrp == NULL)
6977 		/* If NULL, use the MDS socket. */
6978 		nrp = &nmp->nm_sockreq;
6979 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6980 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6981 	NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6982 	    nd->nd_repstat);
6983 	if (error != 0)
6984 		return (error);
6985 	if (nd->nd_repstat != 0)
6986 		error = nd->nd_repstat;
6987 	else {
6988 		if (vers == NFS_VER3) {
6989 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6990 			    NULL);
6991 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6992 			    error);
6993 			if (error != 0)
6994 				goto nfsmout;
6995 		}
6996 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6997 		rlen = fxdr_unsigned(int, *tl++);
6998 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6999 		    rlen);
7000 		if (rlen != len) {
7001 			error = NFSERR_IO;
7002 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
7003 			    len, rlen);
7004 			goto nfsmout;
7005 		}
7006 		commit = fxdr_unsigned(int, *tl++);
7007 
7008 		/*
7009 		 * Return the lowest commitment level
7010 		 * obtained by any of the RPCs.
7011 		 */
7012 		if (committed == NFSWRITE_FILESYNC)
7013 			committed = commit;
7014 		else if (committed == NFSWRITE_DATASYNC &&
7015 		    commit == NFSWRITE_UNSTABLE)
7016 			committed = commit;
7017 		NFSLOCKDS(dsp);
7018 		if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
7019 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
7020 			dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
7021 		} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF) &&
7022 		    *must_commit != 2) {
7023 			*must_commit = 1;
7024 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
7025 		}
7026 		NFSUNLOCKDS(dsp);
7027 	}
7028 nfsmout:
7029 	if (nd->nd_mrep != NULL)
7030 		m_freem(nd->nd_mrep);
7031 	*iomode = committed;
7032 	if (nd->nd_repstat != 0 && error == 0)
7033 		error = nd->nd_repstat;
7034 	return (error);
7035 }
7036 
7037 /*
7038  * Start up the thread that will execute nfsrpc_writedsmir().
7039  */
7040 static void
7041 start_writedsmir(void *arg, int pending)
7042 {
7043 	struct nfsclwritedsdorpc *drpc;
7044 
7045 	drpc = (struct nfsclwritedsdorpc *)arg;
7046 	drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
7047 	    &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
7048 	    drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
7049 	    drpc->p);
7050 	drpc->done = 1;
7051 	crfree(drpc->cred);
7052 	NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
7053 }
7054 
7055 /*
7056  * Set up the write DS mirror call for the pNFS I/O thread.
7057  */
7058 static int
7059 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
7060     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
7061     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
7062     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7063 {
7064 	int error, ret;
7065 
7066 	error = 0;
7067 	drpc->done = 0;
7068 	drpc->vp = vp;
7069 	drpc->iomode = *iomode;
7070 	drpc->must_commit = *must_commit;
7071 	drpc->stateidp = stateidp;
7072 	drpc->dsp = dsp;
7073 	drpc->off = off;
7074 	drpc->len = len;
7075 	drpc->fhp = fhp;
7076 	drpc->m = m;
7077 	drpc->vers = vers;
7078 	drpc->minorvers = minorvers;
7079 	drpc->cred = crhold(cred);
7080 	drpc->p = p;
7081 	drpc->inprog = 0;
7082 	ret = EIO;
7083 	if (nfs_pnfsiothreads != 0) {
7084 		ret = nfs_pnfsio(start_writedsmir, drpc);
7085 		NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
7086 	}
7087 	if (ret != 0) {
7088 		error = nfsrpc_writedsmir(vp, iomode, &drpc->must_commit,
7089 		    stateidp, dsp, off, len, fhp, m, vers, minorvers, cred, p);
7090 		crfree(drpc->cred);
7091 	}
7092 	NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
7093 	return (error);
7094 }
7095 
7096 /*
7097  * Free up the nfsclds structure.
7098  */
7099 void
7100 nfscl_freenfsclds(struct nfsclds *dsp)
7101 {
7102 	int i;
7103 
7104 	if (dsp == NULL)
7105 		return;
7106 	if (dsp->nfsclds_sockp != NULL) {
7107 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
7108 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
7109 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
7110 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
7111 	}
7112 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
7113 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
7114 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
7115 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
7116 			m_freem(
7117 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
7118 	}
7119 	free(dsp, M_NFSCLDS);
7120 }
7121 
7122 static enum nfsclds_state
7123 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
7124     struct nfsclds **retdspp, uint32_t *sequencep)
7125 {
7126 	struct nfsclds *dsp;
7127 	int fndseq;
7128 
7129 	/*
7130 	 * Search the list of nfsclds structures for one with the same
7131 	 * server.
7132 	 */
7133 	fndseq = 0;
7134 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
7135 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
7136 		    dsp->nfsclds_servownlen != 0 &&
7137 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
7138 		    dsp->nfsclds_servownlen) &&
7139 		    dsp->nfsclds_sess.nfsess_defunct == 0) {
7140 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
7141 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
7142 			    dsp->nfsclds_flags);
7143 			if (fndseq == 0) {
7144 				/* Get sequenceid# from first entry. */
7145 				*sequencep =
7146 				    dsp->nfsclds_sess.nfsess_sequenceid;
7147 				fndseq = 1;
7148 			}
7149 			/* Server major id matches. */
7150 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
7151 				*retdspp = dsp;
7152 				return (NFSDSP_USETHISSESSION);
7153 			}
7154 		}
7155 	}
7156 	if (fndseq != 0)
7157 		return (NFSDSP_SEQTHISSESSION);
7158 	return (NFSDSP_NOTFOUND);
7159 }
7160 
7161 /*
7162  * NFS commit rpc to a NFSv4.1 DS.
7163  */
7164 static int
7165 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
7166     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
7167     NFSPROC_T *p)
7168 {
7169 	uint32_t *tl;
7170 	struct nfsrv_descript nfsd, *nd = &nfsd;
7171 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7172 	struct nfssockreq *nrp;
7173 	struct nfsvattr na;
7174 	int attrflag, error;
7175 
7176 	nd->nd_mrep = NULL;
7177 	if (vers == 0 || vers == NFS_VER4) {
7178 		nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
7179 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
7180 		    NULL);
7181 		vers = NFS_VER4;
7182 	} else {
7183 		nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
7184 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
7185 		    NULL);
7186 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMIT]);
7187 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMITDS]);
7188 	}
7189 	NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
7190 	    minorvers);
7191 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
7192 	txdr_hyper(offset, tl);
7193 	tl += 2;
7194 	*tl = txdr_unsigned(cnt);
7195 	nrp = dsp->nfsclds_sockp;
7196 	if (nrp == NULL)
7197 		/* If NULL, use the MDS socket. */
7198 		nrp = &nmp->nm_sockreq;
7199 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7200 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7201 	NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
7202 	    nd->nd_repstat);
7203 	if (error != 0)
7204 		return (error);
7205 	if (nd->nd_repstat == 0) {
7206 		if (vers == NFS_VER3) {
7207 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
7208 			    NULL);
7209 			NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
7210 			if (error != 0)
7211 				goto nfsmout;
7212 		}
7213 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
7214 		NFSLOCKDS(dsp);
7215 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
7216 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
7217 			error = NFSERR_STALEWRITEVERF;
7218 		}
7219 		NFSUNLOCKDS(dsp);
7220 	}
7221 nfsmout:
7222 	if (error == 0 && nd->nd_repstat != 0)
7223 		error = nd->nd_repstat;
7224 	m_freem(nd->nd_mrep);
7225 	return (error);
7226 }
7227 
7228 /*
7229  * Start up the thread that will execute nfsrpc_commitds().
7230  */
7231 static void
7232 start_commitds(void *arg, int pending)
7233 {
7234 	struct nfsclwritedsdorpc *drpc;
7235 
7236 	drpc = (struct nfsclwritedsdorpc *)arg;
7237 	drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
7238 	    drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
7239 	    drpc->p);
7240 	drpc->done = 1;
7241 	crfree(drpc->cred);
7242 	NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
7243 }
7244 
7245 /*
7246  * Set up the commit DS mirror call for the pNFS I/O thread.
7247  */
7248 static int
7249 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
7250     struct nfsfh *fhp, int vers, int minorvers,
7251     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7252 {
7253 	int error, ret;
7254 
7255 	error = 0;
7256 	drpc->done = 0;
7257 	drpc->vp = vp;
7258 	drpc->off = offset;
7259 	drpc->len = cnt;
7260 	drpc->dsp = dsp;
7261 	drpc->fhp = fhp;
7262 	drpc->vers = vers;
7263 	drpc->minorvers = minorvers;
7264 	drpc->cred = crhold(cred);
7265 	drpc->p = p;
7266 	drpc->inprog = 0;
7267 	ret = EIO;
7268 	if (nfs_pnfsiothreads != 0) {
7269 		ret = nfs_pnfsio(start_commitds, drpc);
7270 		NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
7271 	}
7272 	if (ret != 0) {
7273 		error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
7274 		    minorvers, cred, p);
7275 		crfree(drpc->cred);
7276 	}
7277 	NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
7278 	return (error);
7279 }
7280 
7281 /*
7282  * NFS Advise rpc
7283  */
7284 int
7285 nfsrpc_advise(vnode_t vp, off_t offset, uint64_t cnt, int advise,
7286     struct ucred *cred, NFSPROC_T *p)
7287 {
7288 	u_int32_t *tl;
7289 	struct nfsrv_descript nfsd, *nd = &nfsd;
7290 	nfsattrbit_t hints;
7291 	int error;
7292 
7293 	NFSZERO_ATTRBIT(&hints);
7294 	if (advise == POSIX_FADV_WILLNEED)
7295 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
7296 	else if (advise == POSIX_FADV_DONTNEED)
7297 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
7298 	else
7299 		return (0);
7300 	NFSCL_REQSTART(nd, NFSPROC_IOADVISE, vp, cred);
7301 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
7302 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
7303 	txdr_hyper(offset, tl);
7304 	tl += 2;
7305 	txdr_hyper(cnt, tl);
7306 	nfsrv_putattrbit(nd, &hints);
7307 	error = nfscl_request(nd, vp, p, cred);
7308 	if (error != 0)
7309 		return (error);
7310 	if (nd->nd_repstat != 0)
7311 		error = nd->nd_repstat;
7312 	m_freem(nd->nd_mrep);
7313 	return (error);
7314 }
7315 
7316 #ifdef notyet
7317 /*
7318  * NFS advise rpc to a NFSv4.2 DS.
7319  */
7320 static int
7321 nfsrpc_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
7322     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
7323     struct ucred *cred, NFSPROC_T *p)
7324 {
7325 	uint32_t *tl;
7326 	struct nfsrv_descript nfsd, *nd = &nfsd;
7327 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7328 	struct nfssockreq *nrp;
7329 	nfsattrbit_t hints;
7330 	int error;
7331 
7332 	/* For NFS DSs prior to NFSv4.2, just return OK. */
7333 	if (vers == NFS_VER3 || minorversion < NFSV42_MINORVERSION)
7334 		return (0);
7335 	NFSZERO_ATTRBIT(&hints);
7336 	if (advise == POSIX_FADV_WILLNEED)
7337 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
7338 	else if (advise == POSIX_FADV_DONTNEED)
7339 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
7340 	else
7341 		return (0);
7342 	nd->nd_mrep = NULL;
7343 	nfscl_reqstart(nd, NFSPROC_IOADVISEDS, nmp, fhp->nfh_fh,
7344 	    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers, NULL);
7345 	vers = NFS_VER4;
7346 	NFSCL_DEBUG(4, "nfsrpc_adviseds: vers=%d minvers=%d\n", vers,
7347 	    minorvers);
7348 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
7349 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
7350 	txdr_hyper(offset, tl);
7351 	tl += 2;
7352 	*tl = txdr_unsigned(cnt);
7353 	nfsrv_putattrbit(nd, &hints);
7354 	nrp = dsp->nfsclds_sockp;
7355 	if (nrp == NULL)
7356 		/* If NULL, use the MDS socket. */
7357 		nrp = &nmp->nm_sockreq;
7358 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7359 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7360 	NFSCL_DEBUG(4, "nfsrpc_adviseds: err=%d stat=%d\n", error,
7361 	    nd->nd_repstat);
7362 	if (error != 0)
7363 		return (error);
7364 	if (nd->nd_repstat != 0)
7365 		error = nd->nd_repstat;
7366 	m_freem(nd->nd_mrep);
7367 	return (error);
7368 }
7369 
7370 /*
7371  * Start up the thread that will execute nfsrpc_commitds().
7372  */
7373 static void
7374 start_adviseds(void *arg, int pending)
7375 {
7376 	struct nfsclwritedsdorpc *drpc;
7377 
7378 	drpc = (struct nfsclwritedsdorpc *)arg;
7379 	drpc->err = nfsrpc_adviseds(drpc->vp, drpc->off, drpc->len,
7380 	    drpc->advise, drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers,
7381 	    drpc->cred, drpc->p);
7382 	drpc->done = 1;
7383 	crfree(drpc->cred);
7384 	NFSCL_DEBUG(4, "start_adviseds: err=%d\n", drpc->err);
7385 }
7386 
7387 /*
7388  * Set up the advise DS mirror call for the pNFS I/O thread.
7389  */
7390 static int
7391 nfsio_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
7392     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
7393     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7394 {
7395 	int error, ret;
7396 
7397 	error = 0;
7398 	drpc->done = 0;
7399 	drpc->vp = vp;
7400 	drpc->off = offset;
7401 	drpc->len = cnt;
7402 	drpc->advise = advise;
7403 	drpc->dsp = dsp;
7404 	drpc->fhp = fhp;
7405 	drpc->vers = vers;
7406 	drpc->minorvers = minorvers;
7407 	drpc->cred = crhold(cred);
7408 	drpc->p = p;
7409 	drpc->inprog = 0;
7410 	ret = EIO;
7411 	if (nfs_pnfsiothreads != 0) {
7412 		ret = nfs_pnfsio(start_adviseds, drpc);
7413 		NFSCL_DEBUG(4, "nfsio_adviseds: nfs_pnfsio=%d\n", ret);
7414 	}
7415 	if (ret != 0) {
7416 		error = nfsrpc_adviseds(vp, offset, cnt, advise, dsp, fhp, vers,
7417 		    minorvers, cred, p);
7418 		crfree(drpc->cred);
7419 	}
7420 	NFSCL_DEBUG(4, "nfsio_adviseds: error=%d\n", error);
7421 	return (error);
7422 }
7423 #endif	/* notyet */
7424 
7425 /*
7426  * Do the Allocate operation, retrying for recovery.
7427  */
7428 int
7429 nfsrpc_allocate(vnode_t vp, off_t off, off_t len, struct nfsvattr *nap,
7430     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
7431 {
7432 	int error, expireret = 0, retrycnt, nostateid;
7433 	uint32_t clidrev = 0;
7434 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7435 	struct nfsfh *nfhp = NULL;
7436 	nfsv4stateid_t stateid;
7437 	off_t tmp_off;
7438 	void *lckp;
7439 
7440 	if (len < 0)
7441 		return (EINVAL);
7442 	if (len == 0)
7443 		return (0);
7444 	tmp_off = off + len;
7445 	NFSLOCKMNT(nmp);
7446 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < off) {
7447 		NFSUNLOCKMNT(nmp);
7448 		return (EFBIG);
7449 	}
7450 	if (nmp->nm_clp != NULL)
7451 		clidrev = nmp->nm_clp->nfsc_clientidrev;
7452 	NFSUNLOCKMNT(nmp);
7453 	nfhp = VTONFS(vp)->n_fhp;
7454 	retrycnt = 0;
7455 	do {
7456 		lckp = NULL;
7457 		nostateid = 0;
7458 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
7459 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
7460 		if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
7461 		    stateid.other[2] == 0) {
7462 			nostateid = 1;
7463 			NFSCL_DEBUG(1, "stateid0 in allocate\n");
7464 		}
7465 
7466 		/*
7467 		 * Not finding a stateid should probably never happen,
7468 		 * but just return an error for this case.
7469 		 */
7470 		if (nostateid != 0)
7471 			error = EIO;
7472 		else
7473 			error = nfsrpc_allocaterpc(vp, off, len, &stateid,
7474 			    nap, attrflagp, cred, p);
7475 		if (error == NFSERR_STALESTATEID)
7476 			nfscl_initiate_recovery(nmp->nm_clp);
7477 		if (lckp != NULL)
7478 			nfscl_lockderef(lckp);
7479 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
7480 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
7481 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
7482 			(void) nfs_catnap(PZERO, error, "nfs_allocate");
7483 		} else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) &&
7484 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
7485 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
7486 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) {
7487 			error = EIO;
7488 		}
7489 		retrycnt++;
7490 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
7491 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
7492 	    error == NFSERR_STALEDONTRECOVER ||
7493 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
7494 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
7495 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
7496 	if (error != 0 && retrycnt >= 4)
7497 		error = EIO;
7498 	return (error);
7499 }
7500 
7501 /*
7502  * The allocate RPC.
7503  */
7504 static int
7505 nfsrpc_allocaterpc(vnode_t vp, off_t off, off_t len, nfsv4stateid_t *stateidp,
7506     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
7507 {
7508 	uint32_t *tl;
7509 	int error;
7510 	struct nfsrv_descript nfsd;
7511 	struct nfsrv_descript *nd = &nfsd;
7512 	nfsattrbit_t attrbits;
7513 
7514 	*attrflagp = 0;
7515 	NFSCL_REQSTART(nd, NFSPROC_ALLOCATE, vp, cred);
7516 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7517 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
7518 	txdr_hyper(off, tl); tl += 2;
7519 	txdr_hyper(len, tl); tl += 2;
7520 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7521 	NFSGETATTR_ATTRBIT(&attrbits);
7522 	nfsrv_putattrbit(nd, &attrbits);
7523 	error = nfscl_request(nd, vp, p, cred);
7524 	if (error != 0)
7525 		return (error);
7526 	if (nd->nd_repstat == 0) {
7527 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7528 		error = nfsm_loadattr(nd, nap);
7529 		if (error == 0)
7530 			*attrflagp = NFS_LATTR_NOSHRINK;
7531 	} else
7532 		error = nd->nd_repstat;
7533 nfsmout:
7534 	m_freem(nd->nd_mrep);
7535 	return (error);
7536 }
7537 
7538 /*
7539  * Set up the XDR arguments for the LayoutGet operation.
7540  */
7541 static void
7542 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
7543     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
7544     int layoutlen, int usecurstateid)
7545 {
7546 	uint32_t *tl;
7547 
7548 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
7549 	    NFSX_STATEID);
7550 	*tl++ = newnfs_false;		/* Don't signal availability. */
7551 	*tl++ = txdr_unsigned(layouttype);
7552 	*tl++ = txdr_unsigned(iomode);
7553 	txdr_hyper(offset, tl);
7554 	tl += 2;
7555 	txdr_hyper(len, tl);
7556 	tl += 2;
7557 	txdr_hyper(minlen, tl);
7558 	tl += 2;
7559 	if (usecurstateid != 0) {
7560 		/* Special stateid for Current stateid. */
7561 		*tl++ = txdr_unsigned(1);
7562 		*tl++ = 0;
7563 		*tl++ = 0;
7564 		*tl++ = 0;
7565 	} else {
7566 		*tl++ = txdr_unsigned(stateidp->seqid);
7567 		NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
7568 		*tl++ = stateidp->other[0];
7569 		*tl++ = stateidp->other[1];
7570 		*tl++ = stateidp->other[2];
7571 	}
7572 	*tl = txdr_unsigned(layoutlen);
7573 }
7574 
7575 /*
7576  * Parse the reply for a successful LayoutGet operation.
7577  */
7578 static int
7579 nfsrv_parselayoutget(struct nfsmount *nmp, struct nfsrv_descript *nd,
7580     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp)
7581 {
7582 	uint32_t *tl;
7583 	struct nfsclflayout *flp, *prevflp, *tflp;
7584 	int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
7585 	int m, mirrorcnt;
7586 	uint64_t retlen, off;
7587 	struct nfsfh *nfhp;
7588 	uint8_t *cp;
7589 	uid_t user;
7590 	gid_t grp;
7591 
7592 	NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
7593 	error = 0;
7594 	flp = NULL;
7595 	gotiomode = -1;
7596 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
7597 	if (*tl++ != 0)
7598 		*retonclosep = 1;
7599 	else
7600 		*retonclosep = 0;
7601 	stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
7602 	NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
7603 	    (int)stateidp->seqid);
7604 	stateidp->other[0] = *tl++;
7605 	stateidp->other[1] = *tl++;
7606 	stateidp->other[2] = *tl++;
7607 	cnt = fxdr_unsigned(int, *tl);
7608 	NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
7609 	if (cnt <= 0 || cnt > 10000) {
7610 		/* Don't accept more than 10000 layouts in reply. */
7611 		error = NFSERR_BADXDR;
7612 		goto nfsmout;
7613 	}
7614 	for (i = 0; i < cnt; i++) {
7615 		/* Dissect to the layout type. */
7616 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
7617 		    3 * NFSX_UNSIGNED);
7618 		off = fxdr_hyper(tl); tl += 2;
7619 		retlen = fxdr_hyper(tl); tl += 2;
7620 		iomode = fxdr_unsigned(int, *tl++);
7621 		laytype = fxdr_unsigned(int, *tl);
7622 		NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
7623 		    (uintmax_t)off, (uintmax_t)retlen, iomode);
7624 		/* Ignore length of layout body for now. */
7625 		if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
7626 			/* Parse the File layout up to fhcnt. */
7627 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
7628 			    NFSX_HYPER + NFSX_V4DEVICEID);
7629 			fhcnt = fxdr_unsigned(int, *(tl + 4 +
7630 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
7631 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7632 			if (fhcnt < 0 || fhcnt > 100) {
7633 				/* Don't accept more than 100 file handles. */
7634 				error = NFSERR_BADXDR;
7635 				goto nfsmout;
7636 			}
7637 			if (fhcnt > 0)
7638 				flp = malloc(sizeof(*flp) + fhcnt *
7639 				    sizeof(struct nfsfh *), M_NFSFLAYOUT,
7640 				    M_WAITOK);
7641 			else
7642 				flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
7643 				    M_WAITOK);
7644 			flp->nfsfl_flags = NFSFL_FILE;
7645 			flp->nfsfl_fhcnt = 0;
7646 			flp->nfsfl_devp = NULL;
7647 			flp->nfsfl_off = off;
7648 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7649 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7650 			else
7651 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7652 			flp->nfsfl_iomode = iomode;
7653 			if (gotiomode == -1)
7654 				gotiomode = flp->nfsfl_iomode;
7655 			/* Ignore layout body length for now. */
7656 			NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
7657 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
7658 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
7659 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
7660 			mtx_lock(&nmp->nm_mtx);
7661 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_util &
7662 			    NFSFLAYUTIL_IOADVISE_THRU_MDS) != 0)
7663 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7664 			mtx_unlock(&nmp->nm_mtx);
7665 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
7666 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
7667 			NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
7668 			    flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
7669 			for (j = 0; j < fhcnt; j++) {
7670 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7671 				nfhlen = fxdr_unsigned(int, *tl);
7672 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
7673 					error = NFSERR_BADXDR;
7674 					goto nfsmout;
7675 				}
7676 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
7677 				    M_NFSFH, M_WAITOK);
7678 				flp->nfsfl_fh[j] = nfhp;
7679 				flp->nfsfl_fhcnt++;
7680 				nfhp->nfh_len = nfhlen;
7681 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
7682 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
7683 			}
7684 		} else if (laytype == NFSLAYOUT_FLEXFILE) {
7685 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
7686 			    NFSX_HYPER);
7687 			mirrorcnt = fxdr_unsigned(int, *(tl + 2));
7688 			NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
7689 			if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
7690 				error = NFSERR_BADXDR;
7691 				goto nfsmout;
7692 			}
7693 			flp = malloc(sizeof(*flp) + mirrorcnt *
7694 			    sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
7695 			flp->nfsfl_flags = NFSFL_FLEXFILE;
7696 			flp->nfsfl_mirrorcnt = mirrorcnt;
7697 			for (j = 0; j < mirrorcnt; j++)
7698 				flp->nfsfl_ffm[j].devp = NULL;
7699 			flp->nfsfl_off = off;
7700 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7701 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7702 			else
7703 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7704 			flp->nfsfl_iomode = iomode;
7705 			if (gotiomode == -1)
7706 				gotiomode = flp->nfsfl_iomode;
7707 			flp->nfsfl_stripeunit = fxdr_hyper(tl);
7708 			NFSCL_DEBUG(4, "stripeunit=%ju\n",
7709 			    (uintmax_t)flp->nfsfl_stripeunit);
7710 			for (j = 0; j < mirrorcnt; j++) {
7711 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7712 				k = fxdr_unsigned(int, *tl);
7713 				if (k < 1 || k > 128) {
7714 					error = NFSERR_BADXDR;
7715 					goto nfsmout;
7716 				}
7717 				NFSCL_DEBUG(4, "servercnt=%d\n", k);
7718 				for (l = 0; l < k; l++) {
7719 					NFSM_DISSECT(tl, uint32_t *,
7720 					    NFSX_V4DEVICEID + NFSX_STATEID +
7721 					    2 * NFSX_UNSIGNED);
7722 					if (l == 0) {
7723 						/* Just use the first server. */
7724 						NFSBCOPY(tl,
7725 						    flp->nfsfl_ffm[j].dev,
7726 						    NFSX_V4DEVICEID);
7727 						tl += (NFSX_V4DEVICEID /
7728 						    NFSX_UNSIGNED);
7729 						tl++;
7730 						flp->nfsfl_ffm[j].st.seqid =
7731 						    *tl++;
7732 						flp->nfsfl_ffm[j].st.other[0] =
7733 						    *tl++;
7734 						flp->nfsfl_ffm[j].st.other[1] =
7735 						    *tl++;
7736 						flp->nfsfl_ffm[j].st.other[2] =
7737 						    *tl++;
7738 						NFSCL_DEBUG(4, "st.seqid=%u "
7739 						 "st.o0=0x%x st.o1=0x%x "
7740 						 "st.o2=0x%x\n",
7741 						 flp->nfsfl_ffm[j].st.seqid,
7742 						 flp->nfsfl_ffm[j].st.other[0],
7743 						 flp->nfsfl_ffm[j].st.other[1],
7744 						 flp->nfsfl_ffm[j].st.other[2]);
7745 					} else
7746 						tl += ((NFSX_V4DEVICEID +
7747 						    NFSX_STATEID +
7748 						    NFSX_UNSIGNED) /
7749 						    NFSX_UNSIGNED);
7750 					fhcnt = fxdr_unsigned(int, *tl);
7751 					NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7752 					if (fhcnt < 1 ||
7753 					    fhcnt > NFSDEV_MAXVERS) {
7754 						error = NFSERR_BADXDR;
7755 						goto nfsmout;
7756 					}
7757 					for (m = 0; m < fhcnt; m++) {
7758 						NFSM_DISSECT(tl, uint32_t *,
7759 						    NFSX_UNSIGNED);
7760 						nfhlen = fxdr_unsigned(int,
7761 						    *tl);
7762 						NFSCL_DEBUG(4, "nfhlen=%d\n",
7763 						    nfhlen);
7764 						if (nfhlen <= 0 || nfhlen >
7765 						    NFSX_V4FHMAX) {
7766 							error = NFSERR_BADXDR;
7767 							goto nfsmout;
7768 						}
7769 						NFSM_DISSECT(cp, uint8_t *,
7770 						    NFSM_RNDUP(nfhlen));
7771 						if (l == 0) {
7772 							flp->nfsfl_ffm[j].fhcnt
7773 							    = fhcnt;
7774 							nfhp = malloc(
7775 							    sizeof(*nfhp) +
7776 							    nfhlen - 1, M_NFSFH,
7777 							    M_WAITOK);
7778 							flp->nfsfl_ffm[j].fh[m]
7779 							    = nfhp;
7780 							nfhp->nfh_len = nfhlen;
7781 							NFSBCOPY(cp,
7782 							    nfhp->nfh_fh,
7783 							    nfhlen);
7784 							NFSCL_DEBUG(4,
7785 							    "got fh\n");
7786 						}
7787 					}
7788 					/* Now, get the ffsd_user/ffds_group. */
7789 					error = nfsrv_parseug(nd, 0, &user,
7790 					    &grp, curthread);
7791 					NFSCL_DEBUG(4, "after parseu=%d\n",
7792 					    error);
7793 					if (error == 0)
7794 						error = nfsrv_parseug(nd, 1,
7795 						    &user, &grp, curthread);
7796 					NFSCL_DEBUG(4, "aft parseg=%d\n",
7797 					    grp);
7798 					if (error != 0)
7799 						goto nfsmout;
7800 					NFSCL_DEBUG(4, "user=%d group=%d\n",
7801 					    user, grp);
7802 					if (l == 0) {
7803 						flp->nfsfl_ffm[j].user = user;
7804 						flp->nfsfl_ffm[j].group = grp;
7805 						NFSCL_DEBUG(4,
7806 						    "usr=%d grp=%d\n", user,
7807 						    grp);
7808 					}
7809 				}
7810 			}
7811 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7812 			flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
7813 #ifdef notnow
7814 			/*
7815 			 * At this time, there is no flag.
7816 			 * NFSFLEXFLAG_IOADVISE_THRU_MDS might need to be
7817 			 * added, or it may never exist?
7818 			 */
7819 			mtx_lock(&nmp->nm_mtx);
7820 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_fflags &
7821 			    NFSFLEXFLAG_IOADVISE_THRU_MDS) != 0)
7822 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7823 			mtx_unlock(&nmp->nm_mtx);
7824 #endif
7825 			flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
7826 			NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
7827 			    flp->nfsfl_fflags, flp->nfsfl_statshint);
7828 		} else {
7829 			error = NFSERR_BADXDR;
7830 			goto nfsmout;
7831 		}
7832 		if (flp->nfsfl_iomode == gotiomode) {
7833 			/* Keep the list in increasing offset order. */
7834 			tflp = LIST_FIRST(flhp);
7835 			prevflp = NULL;
7836 			while (tflp != NULL &&
7837 			    tflp->nfsfl_off < flp->nfsfl_off) {
7838 				prevflp = tflp;
7839 				tflp = LIST_NEXT(tflp, nfsfl_list);
7840 			}
7841 			if (prevflp == NULL)
7842 				LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
7843 			else
7844 				LIST_INSERT_AFTER(prevflp, flp,
7845 				    nfsfl_list);
7846 			NFSCL_DEBUG(4, "flp inserted\n");
7847 		} else {
7848 			printf("nfscl_layoutget(): got wrong iomode\n");
7849 			nfscl_freeflayout(flp);
7850 		}
7851 		flp = NULL;
7852 	}
7853 nfsmout:
7854 	NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
7855 	if (error != 0 && flp != NULL)
7856 		nfscl_freeflayout(flp);
7857 	return (error);
7858 }
7859 
7860 /*
7861  * Parse a user/group digit string.
7862  */
7863 static int
7864 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
7865     NFSPROC_T *p)
7866 {
7867 	uint32_t *tl;
7868 	char *cp, *str, str0[NFSV4_SMALLSTR + 1];
7869 	uint32_t len = 0;
7870 	int error = 0;
7871 
7872 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7873 	len = fxdr_unsigned(uint32_t, *tl);
7874 	str = NULL;
7875 	if (len > NFSV4_OPAQUELIMIT) {
7876 		error = NFSERR_BADXDR;
7877 		goto nfsmout;
7878 	}
7879 	NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
7880 	if (len == 0) {
7881 		if (dogrp != 0)
7882 			*gidp = GID_NOGROUP;
7883 		else
7884 			*uidp = UID_NOBODY;
7885 		return (0);
7886 	}
7887 	if (len > NFSV4_SMALLSTR)
7888 		str = malloc(len + 1, M_TEMP, M_WAITOK);
7889 	else
7890 		str = str0;
7891 	NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7892 	NFSBCOPY(cp, str, len);
7893 	str[len] = '\0';
7894 	NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7895 	if (dogrp != 0)
7896 		error = nfsv4_strtogid(nd, str, len, gidp);
7897 	else
7898 		error = nfsv4_strtouid(nd, str, len, uidp);
7899 nfsmout:
7900 	if (len > NFSV4_SMALLSTR)
7901 		free(str, M_TEMP);
7902 	NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7903 	return (error);
7904 }
7905 
7906 /*
7907  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7908  * so that it does both an Open and a Layoutget.
7909  */
7910 static int
7911 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7912     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7913     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7914     struct ucred *cred, NFSPROC_T *p)
7915 {
7916 	struct nfscllayout *lyp;
7917 	struct nfsclflayout *flp;
7918 	struct nfsclflayouthead flh;
7919 	int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7920 	int layouttype, laystat;
7921 	nfsv4stateid_t stateid;
7922 	struct nfsclsession *tsep;
7923 
7924 	error = 0;
7925 	if (NFSHASFLEXFILE(nmp))
7926 		layouttype = NFSLAYOUT_FLEXFILE;
7927 	else
7928 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7929 	/*
7930 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7931 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7932 	 * flp == NULL.
7933 	 */
7934 	lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, mode, &flp,
7935 	    &recalled);
7936 	NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7937 	if (lyp == NULL)
7938 		islocked = 0;
7939 	else if (flp != NULL)
7940 		islocked = 1;
7941 	else
7942 		islocked = 2;
7943 	if ((lyp == NULL || flp == NULL) && recalled == 0) {
7944 		LIST_INIT(&flh);
7945 		tsep = nfsmnt_mdssession(nmp);
7946 		layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7947 		    3 * NFSX_UNSIGNED);
7948 		if (lyp == NULL)
7949 			usecurstateid = 1;
7950 		else {
7951 			usecurstateid = 0;
7952 			stateid.seqid = lyp->nfsly_stateid.seqid;
7953 			stateid.other[0] = lyp->nfsly_stateid.other[0];
7954 			stateid.other[1] = lyp->nfsly_stateid.other[1];
7955 			stateid.other[2] = lyp->nfsly_stateid.other[2];
7956 		}
7957 		error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7958 		    newfhp, newfhlen, mode, op, name, namelen,
7959 		    dpp, &stateid, usecurstateid, layouttype, layoutlen,
7960 		    &retonclose, &flh, &laystat, cred, p);
7961 		NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7962 		    laystat, error);
7963 		laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7964 		    &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7965 		    &islocked, cred, p);
7966 	} else
7967 		error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7968 		    mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7969 	if (islocked == 2)
7970 		nfscl_rellayout(lyp, 1);
7971 	else if (islocked == 1)
7972 		nfscl_rellayout(lyp, 0);
7973 	return (error);
7974 }
7975 
7976 /*
7977  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7978  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
7979  * handled by nfsrpc_openrpc().
7980  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
7981  * can be NULL.
7982  */
7983 static int
7984 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7985     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7986     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7987     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7988     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7989     int *laystatp, struct ucred *cred, NFSPROC_T *p)
7990 {
7991 	uint32_t *tl;
7992 	struct nfsrv_descript nfsd, *nd = &nfsd;
7993 	struct nfscldeleg *ndp = NULL;
7994 	struct nfsvattr nfsva;
7995 	struct nfsclsession *tsep;
7996 	uint32_t rflags, deleg;
7997 	nfsattrbit_t attrbits;
7998 	int error, ret, acesize, limitby, iomode;
7999 
8000 	*dpp = NULL;
8001 	*laystatp = ENXIO;
8002 	nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
8003 	    0, 0, cred);
8004 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
8005 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
8006 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
8007 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
8008 	tsep = nfsmnt_mdssession(nmp);
8009 	*tl++ = tsep->nfsess_clientid.lval[0];
8010 	*tl = tsep->nfsess_clientid.lval[1];
8011 	nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
8012 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8013 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
8014 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
8015 	nfsm_strtom(nd, name, namelen);
8016 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8017 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8018 	NFSZERO_ATTRBIT(&attrbits);
8019 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
8020 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
8021 	nfsrv_putattrbit(nd, &attrbits);
8022 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8023 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
8024 	if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
8025 		iomode = NFSLAYOUTIOMODE_RW;
8026 	else
8027 		iomode = NFSLAYOUTIOMODE_READ;
8028 	nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
8029 	    layouttype, layoutlen, usecurstateid);
8030 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
8031 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
8032 	if (error != 0)
8033 		return (error);
8034 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
8035 	if (nd->nd_repstat != 0)
8036 		*laystatp = nd->nd_repstat;
8037 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8038 		/* ND_NOMOREDATA will be set if the Open operation failed. */
8039 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8040 		    6 * NFSX_UNSIGNED);
8041 		op->nfso_stateid.seqid = *tl++;
8042 		op->nfso_stateid.other[0] = *tl++;
8043 		op->nfso_stateid.other[1] = *tl++;
8044 		op->nfso_stateid.other[2] = *tl;
8045 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
8046 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
8047 		if (error != 0)
8048 			goto nfsmout;
8049 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
8050 		deleg = fxdr_unsigned(u_int32_t, *tl);
8051 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
8052 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
8053 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
8054 			      NFSCLFLAGS_FIRSTDELEG))
8055 				op->nfso_own->nfsow_clp->nfsc_flags |=
8056 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
8057 			ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
8058 			    M_NFSCLDELEG, M_WAITOK);
8059 			LIST_INIT(&ndp->nfsdl_owner);
8060 			LIST_INIT(&ndp->nfsdl_lock);
8061 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
8062 			ndp->nfsdl_fhlen = newfhlen;
8063 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
8064 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
8065 			nfscl_lockinit(&ndp->nfsdl_rwlock);
8066 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8067 			    NFSX_UNSIGNED);
8068 			ndp->nfsdl_stateid.seqid = *tl++;
8069 			ndp->nfsdl_stateid.other[0] = *tl++;
8070 			ndp->nfsdl_stateid.other[1] = *tl++;
8071 			ndp->nfsdl_stateid.other[2] = *tl++;
8072 			ret = fxdr_unsigned(int, *tl);
8073 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
8074 				ndp->nfsdl_flags = NFSCLDL_WRITE;
8075 				/*
8076 				 * Indicates how much the file can grow.
8077 				 */
8078 				NFSM_DISSECT(tl, u_int32_t *,
8079 				    3 * NFSX_UNSIGNED);
8080 				limitby = fxdr_unsigned(int, *tl++);
8081 				switch (limitby) {
8082 				case NFSV4OPEN_LIMITSIZE:
8083 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
8084 					break;
8085 				case NFSV4OPEN_LIMITBLOCKS:
8086 					ndp->nfsdl_sizelimit =
8087 					    fxdr_unsigned(u_int64_t, *tl++);
8088 					ndp->nfsdl_sizelimit *=
8089 					    fxdr_unsigned(u_int64_t, *tl);
8090 					break;
8091 				default:
8092 					error = NFSERR_BADXDR;
8093 					goto nfsmout;
8094 				};
8095 			} else
8096 				ndp->nfsdl_flags = NFSCLDL_READ;
8097 			if (ret != 0)
8098 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
8099 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, false,
8100 			    &ret, &acesize, p);
8101 			if (error != 0)
8102 				goto nfsmout;
8103 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
8104 			error = NFSERR_BADXDR;
8105 			goto nfsmout;
8106 		}
8107 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
8108 		    nfscl_assumeposixlocks)
8109 			op->nfso_posixlock = 1;
8110 		else
8111 			op->nfso_posixlock = 0;
8112 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8113 		/* If the 2nd element == NFS_OK, the Getattr succeeded. */
8114 		if (*++tl == 0) {
8115 			error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
8116 			    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
8117 			    NULL, NULL, NULL, p, cred);
8118 			if (error != 0)
8119 				goto nfsmout;
8120 			if (ndp != NULL) {
8121 				ndp->nfsdl_change = nfsva.na_filerev;
8122 				ndp->nfsdl_modtime = nfsva.na_mtime;
8123 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
8124 				*dpp = ndp;
8125 				ndp = NULL;
8126 			}
8127 			/*
8128 			 * At this point, the Open has succeeded, so set
8129 			 * nd_repstat = NFS_OK.  If the Layoutget failed,
8130 			 * this function just won't return a layout.
8131 			 */
8132 			if (nd->nd_repstat == 0) {
8133 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8134 				*laystatp = fxdr_unsigned(int, *++tl);
8135 				if (*laystatp == 0) {
8136 					error = nfsrv_parselayoutget(nmp, nd,
8137 					    stateidp, retonclosep, flhp);
8138 					if (error != 0)
8139 						*laystatp = error;
8140 				}
8141 			} else
8142 				nd->nd_repstat = 0;	/* Return 0 for Open. */
8143 		}
8144 	}
8145 	if (nd->nd_repstat != 0 && error == 0)
8146 		error = nd->nd_repstat;
8147 nfsmout:
8148 	free(ndp, M_NFSCLDELEG);
8149 	m_freem(nd->nd_mrep);
8150 	return (error);
8151 }
8152 
8153 /*
8154  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
8155  * Used only for mounts with pNFS enabled.
8156  */
8157 static int
8158 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
8159     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
8160     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
8161     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
8162     int *dattrflagp, int *unlockedp, nfsv4stateid_t *stateidp,
8163     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
8164     struct nfsclflayouthead *flhp, int *laystatp)
8165 {
8166 	uint32_t *tl;
8167 	int error = 0, deleg, newone, ret, acesize, limitby;
8168 	struct nfsrv_descript nfsd, *nd = &nfsd;
8169 	struct nfsclopen *op;
8170 	struct nfscldeleg *dp = NULL;
8171 	struct nfsnode *np;
8172 	struct nfsfh *nfhp;
8173 	struct nfsclsession *tsep;
8174 	nfsattrbit_t attrbits;
8175 	nfsv4stateid_t stateid;
8176 	struct nfsmount *nmp;
8177 
8178 	nmp = VFSTONFS(dvp->v_mount);
8179 	np = VTONFS(dvp);
8180 	*laystatp = ENXIO;
8181 	*unlockedp = 0;
8182 	*nfhpp = NULL;
8183 	*dpp = NULL;
8184 	*attrflagp = 0;
8185 	*dattrflagp = 0;
8186 	if (namelen > NFS_MAXNAMLEN)
8187 		return (ENAMETOOLONG);
8188 	NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp, cred);
8189 	/*
8190 	 * For V4, this is actually an Open op.
8191 	 */
8192 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
8193 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
8194 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
8195 	    NFSV4OPEN_ACCESSREAD);
8196 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
8197 	tsep = nfsmnt_mdssession(nmp);
8198 	*tl++ = tsep->nfsess_clientid.lval[0];
8199 	*tl = tsep->nfsess_clientid.lval[1];
8200 	nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
8201 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8202 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
8203 	if ((fmode & O_EXCL) != 0) {
8204 		if (NFSHASSESSPERSIST(nmp)) {
8205 			/* Use GUARDED for persistent sessions. */
8206 			*tl = txdr_unsigned(NFSCREATE_GUARDED);
8207 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
8208 		} else {
8209 			/* Otherwise, use EXCLUSIVE4_1. */
8210 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
8211 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
8212 			*tl++ = cverf.lval[0];
8213 			*tl = cverf.lval[1];
8214 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
8215 		}
8216 	} else {
8217 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
8218 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
8219 	}
8220 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8221 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
8222 	nfsm_strtom(nd, name, namelen);
8223 	/* Get the new file's handle and attributes, plus save the FH. */
8224 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
8225 	*tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
8226 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
8227 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8228 	NFSGETATTR_ATTRBIT(&attrbits);
8229 	nfsrv_putattrbit(nd, &attrbits);
8230 	/* Get the directory's post-op attributes. */
8231 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8232 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8233 	(void)nfsm_fhtom(nmp, nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
8234 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8235 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8236 	nfsrv_putattrbit(nd, &attrbits);
8237 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8238 	*tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
8239 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
8240 	nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
8241 	    layouttype, layoutlen, usecurstateid);
8242 	error = nfscl_request(nd, dvp, p, cred);
8243 	if (error != 0)
8244 		return (error);
8245 	NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
8246 	    error);
8247 	if (nd->nd_repstat != 0)
8248 		*laystatp = nd->nd_repstat;
8249 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
8250 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8251 		NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
8252 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8253 		    6 * NFSX_UNSIGNED);
8254 		stateid.seqid = *tl++;
8255 		stateid.other[0] = *tl++;
8256 		stateid.other[1] = *tl++;
8257 		stateid.other[2] = *tl;
8258 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
8259 		if (error != 0)
8260 			goto nfsmout;
8261 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
8262 		deleg = fxdr_unsigned(int, *tl);
8263 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
8264 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
8265 			if (!(owp->nfsow_clp->nfsc_flags &
8266 			      NFSCLFLAGS_FIRSTDELEG))
8267 				owp->nfsow_clp->nfsc_flags |=
8268 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
8269 			dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
8270 			    M_NFSCLDELEG, M_WAITOK);
8271 			LIST_INIT(&dp->nfsdl_owner);
8272 			LIST_INIT(&dp->nfsdl_lock);
8273 			dp->nfsdl_clp = owp->nfsow_clp;
8274 			newnfs_copyincred(cred, &dp->nfsdl_cred);
8275 			nfscl_lockinit(&dp->nfsdl_rwlock);
8276 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8277 			    NFSX_UNSIGNED);
8278 			dp->nfsdl_stateid.seqid = *tl++;
8279 			dp->nfsdl_stateid.other[0] = *tl++;
8280 			dp->nfsdl_stateid.other[1] = *tl++;
8281 			dp->nfsdl_stateid.other[2] = *tl++;
8282 			ret = fxdr_unsigned(int, *tl);
8283 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
8284 				dp->nfsdl_flags = NFSCLDL_WRITE;
8285 				/*
8286 				 * Indicates how much the file can grow.
8287 				 */
8288 				NFSM_DISSECT(tl, u_int32_t *,
8289 				    3 * NFSX_UNSIGNED);
8290 				limitby = fxdr_unsigned(int, *tl++);
8291 				switch (limitby) {
8292 				case NFSV4OPEN_LIMITSIZE:
8293 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
8294 					break;
8295 				case NFSV4OPEN_LIMITBLOCKS:
8296 					dp->nfsdl_sizelimit =
8297 					    fxdr_unsigned(u_int64_t, *tl++);
8298 					dp->nfsdl_sizelimit *=
8299 					    fxdr_unsigned(u_int64_t, *tl);
8300 					break;
8301 				default:
8302 					error = NFSERR_BADXDR;
8303 					goto nfsmout;
8304 				};
8305 			} else {
8306 				dp->nfsdl_flags = NFSCLDL_READ;
8307 			}
8308 			if (ret != 0)
8309 				dp->nfsdl_flags |= NFSCLDL_RECALL;
8310 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, false,
8311 			    &ret, &acesize, p);
8312 			if (error != 0)
8313 				goto nfsmout;
8314 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
8315 			error = NFSERR_BADXDR;
8316 			goto nfsmout;
8317 		}
8318 
8319 		/* Now, we should have the status for the SaveFH. */
8320 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8321 		if (*++tl == 0) {
8322 			NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
8323 			/*
8324 			 * Now, process the GetFH and Getattr for the newly
8325 			 * created file. nfscl_mtofh() will set
8326 			 * ND_NOMOREDATA if these weren't successful.
8327 			 */
8328 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
8329 			NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
8330 			if (error != 0)
8331 				goto nfsmout;
8332 		} else
8333 			nd->nd_flag |= ND_NOMOREDATA;
8334 		/* Now we have the PutFH and Getattr for the directory. */
8335 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8336 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8337 			if (*++tl != 0)
8338 				nd->nd_flag |= ND_NOMOREDATA;
8339 			else {
8340 				NFSM_DISSECT(tl, uint32_t *, 2 *
8341 				    NFSX_UNSIGNED);
8342 				if (*++tl != 0)
8343 					nd->nd_flag |= ND_NOMOREDATA;
8344 			}
8345 		}
8346 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8347 			/* Load the directory attributes. */
8348 			error = nfsm_loadattr(nd, dnap);
8349 			NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
8350 			if (error != 0)
8351 				goto nfsmout;
8352 			*dattrflagp = 1;
8353 			if (dp != NULL && *attrflagp != 0) {
8354 				dp->nfsdl_change = nnap->na_filerev;
8355 				dp->nfsdl_modtime = nnap->na_mtime;
8356 				dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
8357 			}
8358 			/*
8359 			 * We can now complete the Open state.
8360 			 */
8361 			nfhp = *nfhpp;
8362 			if (dp != NULL) {
8363 				dp->nfsdl_fhlen = nfhp->nfh_len;
8364 				NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
8365 				    nfhp->nfh_len);
8366 			}
8367 			/*
8368 			 * Get an Open structure that will be
8369 			 * attached to the OpenOwner, acquired already.
8370 			 */
8371 			error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
8372 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
8373 			    cred, p, NULL, &op, &newone, NULL, 0, false);
8374 			if (error != 0)
8375 				goto nfsmout;
8376 			op->nfso_stateid = stateid;
8377 			newnfs_copyincred(cred, &op->nfso_cred);
8378 
8379 			nfscl_openrelease(nmp, op, error, newone);
8380 			*unlockedp = 1;
8381 
8382 			/* Now, handle the RestoreFH and LayoutGet. */
8383 			if (nd->nd_repstat == 0) {
8384 				NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
8385 				*laystatp = fxdr_unsigned(int, *(tl + 3));
8386 				if (*laystatp == 0) {
8387 					error = nfsrv_parselayoutget(nmp, nd,
8388 					    stateidp, retonclosep, flhp);
8389 					if (error != 0)
8390 						*laystatp = error;
8391 				}
8392 				NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
8393 				    error);
8394 			} else
8395 				nd->nd_repstat = 0;
8396 		}
8397 	}
8398 	if (nd->nd_repstat != 0 && error == 0)
8399 		error = nd->nd_repstat;
8400 	if (error == NFSERR_STALECLIENTID)
8401 		nfscl_initiate_recovery(owp->nfsow_clp);
8402 nfsmout:
8403 	NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
8404 	if (error == 0)
8405 		*dpp = dp;
8406 	else
8407 		free(dp, M_NFSCLDELEG);
8408 	m_freem(nd->nd_mrep);
8409 	return (error);
8410 }
8411 
8412 /*
8413  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
8414  */
8415 static int
8416 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
8417     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
8418     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
8419     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
8420     int *dattrflagp, int *unlockedp)
8421 {
8422 	struct nfscllayout *lyp;
8423 	struct nfsclflayouthead flh;
8424 	struct nfsfh *nfhp;
8425 	struct nfsclsession *tsep;
8426 	struct nfsmount *nmp;
8427 	nfsv4stateid_t stateid;
8428 	int error, layoutlen, layouttype, retonclose, laystat;
8429 
8430 	error = 0;
8431 	nmp = VFSTONFS(dvp->v_mount);
8432 	if (NFSHASFLEXFILE(nmp))
8433 		layouttype = NFSLAYOUT_FLEXFILE;
8434 	else
8435 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
8436 	LIST_INIT(&flh);
8437 	tsep = nfsmnt_mdssession(nmp);
8438 	layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
8439 	error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
8440 	    owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
8441 	    unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
8442 	    &flh, &laystat);
8443 	NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
8444 	    laystat, error);
8445 	lyp = NULL;
8446 	if (laystat == 0) {
8447 		nfhp = *nfhpp;
8448 		laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
8449 		    nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
8450 		    layouttype, laystat, NULL, cred, p);
8451 	} else
8452 		laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
8453 		    retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
8454 		    cred, p);
8455 	if (laystat == 0)
8456 		nfscl_rellayout(lyp, 0);
8457 	return (error);
8458 }
8459 
8460 /*
8461  * Process the results of a layoutget() operation.
8462  */
8463 static int
8464 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
8465     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
8466     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
8467     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
8468 {
8469 	struct nfsclflayout *tflp;
8470 	struct nfscldevinfo *dip;
8471 	uint8_t *dev;
8472 	int i, mirrorcnt;
8473 
8474 	if (laystat == NFSERR_UNKNLAYOUTTYPE) {
8475 		NFSLOCKMNT(nmp);
8476 		if (!NFSHASFLEXFILE(nmp)) {
8477 			/* Switch to using Flex File Layout. */
8478 			nmp->nm_state |= NFSSTA_FLEXFILE;
8479 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
8480 			/* Disable pNFS. */
8481 			NFSCL_DEBUG(1, "disable PNFS\n");
8482 			nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
8483 		}
8484 		NFSUNLOCKMNT(nmp);
8485 	}
8486 	if (laystat == 0) {
8487 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
8488 		LIST_FOREACH(tflp, flhp, nfsfl_list) {
8489 			if (layouttype == NFSLAYOUT_FLEXFILE)
8490 				mirrorcnt = tflp->nfsfl_mirrorcnt;
8491 			else
8492 				mirrorcnt = 1;
8493 			for (i = 0; i < mirrorcnt; i++) {
8494 				laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
8495 				NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
8496 				if (laystat != 0) {
8497 					if (layouttype == NFSLAYOUT_FLEXFILE)
8498 						dev = tflp->nfsfl_ffm[i].dev;
8499 					else
8500 						dev = tflp->nfsfl_dev;
8501 					laystat = nfsrpc_getdeviceinfo(nmp, dev,
8502 					    layouttype, notifybit, &dip, cred,
8503 					    p);
8504 					NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
8505 					    laystat);
8506 					if (laystat != 0)
8507 						goto out;
8508 					laystat = nfscl_adddevinfo(nmp, dip, i,
8509 					    tflp);
8510 					if (laystat != 0)
8511 						printf("nfsrpc_layoutgetresout"
8512 						    ": cannot add\n");
8513 				}
8514 			}
8515 		}
8516 	}
8517 out:
8518 	if (laystat == 0) {
8519 		/*
8520 		 * nfscl_layout() always returns with the nfsly_lock
8521 		 * set to a refcnt (shared lock).
8522 		 * Passing in dvp is sufficient, since it is only used to
8523 		 * get the fsid for the file system.
8524 		 */
8525 		laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
8526 		    layouttype, retonclose, flhp, lypp, cred, p);
8527 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
8528 		    laystat);
8529 		if (laystat == 0 && islockedp != NULL)
8530 			*islockedp = 1;
8531 	}
8532 	return (laystat);
8533 }
8534 
8535 /*
8536  * nfs copy_file_range operation.
8537  */
8538 int
8539 nfsrpc_copy_file_range(vnode_t invp, off_t *inoffp, vnode_t outvp,
8540     off_t *outoffp, size_t *lenp, unsigned int flags, int *inattrflagp,
8541     struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap,
8542     struct ucred *cred, bool consecutive, bool *must_commitp)
8543 {
8544 	int commit, error, expireret = 0, retrycnt;
8545 	u_int32_t clidrev = 0;
8546 	struct nfsmount *nmp = VFSTONFS(invp->v_mount);
8547 	struct nfsfh *innfhp = NULL, *outnfhp = NULL;
8548 	nfsv4stateid_t instateid, outstateid;
8549 	void *inlckp, *outlckp;
8550 
8551 	if (nmp->nm_clp != NULL)
8552 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8553 	innfhp = VTONFS(invp)->n_fhp;
8554 	outnfhp = VTONFS(outvp)->n_fhp;
8555 	retrycnt = 0;
8556 	do {
8557 		/* Get both stateids. */
8558 		inlckp = NULL;
8559 		nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len,
8560 		    NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid,
8561 		    &inlckp);
8562 		outlckp = NULL;
8563 		nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len,
8564 		    NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid,
8565 		    &outlckp);
8566 
8567 		error = nfsrpc_copyrpc(invp, *inoffp, outvp, *outoffp, lenp,
8568 		    &instateid, &outstateid, innap, inattrflagp, outnap,
8569 		    outattrflagp, consecutive, &commit, cred, curthread);
8570 		if (error == 0) {
8571 			if (commit != NFSWRITE_FILESYNC)
8572 				*must_commitp = true;
8573 			*inoffp += *lenp;
8574 			*outoffp += *lenp;
8575 		} else if (error == NFSERR_STALESTATEID)
8576 			nfscl_initiate_recovery(nmp->nm_clp);
8577 		if (inlckp != NULL)
8578 			nfscl_lockderef(inlckp);
8579 		if (outlckp != NULL)
8580 			nfscl_lockderef(outlckp);
8581 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8582 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8583 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8584 			(void) nfs_catnap(PZERO, error, "nfs_cfr");
8585 		} else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) &&
8586 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
8587 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8588 			    curthread);
8589 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) {
8590 			error = EIO;
8591 		}
8592 		retrycnt++;
8593 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
8594 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8595 	      error == NFSERR_STALEDONTRECOVER ||
8596 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8597 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8598 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
8599 	if (error != 0 && (retrycnt >= 4 ||
8600 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8601 	      error == NFSERR_STALEDONTRECOVER))
8602 		error = EIO;
8603 	return (error);
8604 }
8605 
8606 /*
8607  * The copy RPC.
8608  */
8609 static int
8610 nfsrpc_copyrpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff,
8611     size_t *lenp, nfsv4stateid_t *instateidp, nfsv4stateid_t *outstateidp,
8612     struct nfsvattr *innap, int *inattrflagp, struct nfsvattr *outnap,
8613     int *outattrflagp, bool consecutive, int *commitp, struct ucred *cred,
8614     NFSPROC_T *p)
8615 {
8616 	uint32_t *tl;
8617 	int error;
8618 	struct nfsrv_descript nfsd;
8619 	struct nfsrv_descript *nd = &nfsd;
8620 	struct nfsmount *nmp;
8621 	nfsattrbit_t attrbits;
8622 	uint64_t len;
8623 
8624 	nmp = VFSTONFS(outvp->v_mount);
8625 	*inattrflagp = *outattrflagp = 0;
8626 	*commitp = NFSWRITE_UNSTABLE;
8627 	len = *lenp;
8628 	*lenp = 0;
8629 	if (len > nfs_maxcopyrange)
8630 		len = nfs_maxcopyrange;
8631 	NFSCL_REQSTART(nd, NFSPROC_COPY, invp, cred);
8632 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8633 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8634 	NFSGETATTR_ATTRBIT(&attrbits);
8635 	nfsrv_putattrbit(nd, &attrbits);
8636 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8637 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8638 	(void)nfsm_fhtom(nmp, nd, VTONFS(outvp)->n_fhp->nfh_fh,
8639 	    VTONFS(outvp)->n_fhp->nfh_len, 0);
8640 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8641 	*tl = txdr_unsigned(NFSV4OP_COPY);
8642 	nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
8643 	nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID);
8644 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + 4 * NFSX_UNSIGNED);
8645 	txdr_hyper(inoff, tl); tl += 2;
8646 	txdr_hyper(outoff, tl); tl += 2;
8647 	txdr_hyper(len, tl); tl += 2;
8648 	if (consecutive)
8649 		*tl++ = newnfs_true;
8650 	else
8651 		*tl++ = newnfs_false;
8652 	*tl++ = newnfs_true;
8653 	*tl++ = 0;
8654 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8655 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
8656 	nfsrv_putattrbit(nd, &attrbits);
8657 	error = nfscl_request(nd, invp, p, cred);
8658 	if (error != 0)
8659 		return (error);
8660 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8661 		/* Get the input file's attributes. */
8662 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8663 		if (*(tl + 1) == 0) {
8664 			error = nfsm_loadattr(nd, innap);
8665 			if (error != 0)
8666 				goto nfsmout;
8667 			*inattrflagp = 1;
8668 		} else
8669 			nd->nd_flag |= ND_NOMOREDATA;
8670 	}
8671 	/* Skip over return stat for PutFH. */
8672 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8673 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8674 		if (*++tl != 0)
8675 			nd->nd_flag |= ND_NOMOREDATA;
8676 	}
8677 	/* Skip over return stat for Copy. */
8678 	if ((nd->nd_flag & ND_NOMOREDATA) == 0)
8679 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8680 	if (nd->nd_repstat == 0) {
8681 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8682 		if (*tl != 0) {
8683 			/* There should be no callback ids. */
8684 			error = NFSERR_BADXDR;
8685 			goto nfsmout;
8686 		}
8687 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED +
8688 		    NFSX_VERF);
8689 		len = fxdr_hyper(tl); tl += 2;
8690 		*commitp = fxdr_unsigned(int, *tl++);
8691 		NFSLOCKMNT(nmp);
8692 		if (!NFSHASWRITEVERF(nmp)) {
8693 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8694 			NFSSETWRITEVERF(nmp);
8695 	    	} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
8696 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8697 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
8698 		}
8699 		NFSUNLOCKMNT(nmp);
8700 		tl += (NFSX_VERF / NFSX_UNSIGNED);
8701 		if (nd->nd_repstat == 0 && *++tl != newnfs_true)
8702 			/* Must be a synchronous copy. */
8703 			nd->nd_repstat = NFSERR_NOTSUPP;
8704 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8705 		error = nfsm_loadattr(nd, outnap);
8706 		if (error == 0)
8707 			*outattrflagp = NFS_LATTR_NOSHRINK;
8708 		if (nd->nd_repstat == 0)
8709 			*lenp = len;
8710 	} else if (nd->nd_repstat == NFSERR_OFFLOADNOREQS) {
8711 		/*
8712 		 * For the case where consecutive is not supported, but
8713 		 * synchronous is supported, we can try consecutive == false
8714 		 * by returning this error.  Otherwise, return NFSERR_NOTSUPP,
8715 		 * since Copy cannot be done.
8716 		 */
8717 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8718 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8719 			if (!consecutive || *++tl == newnfs_false)
8720 				nd->nd_repstat = NFSERR_NOTSUPP;
8721 		} else
8722 			nd->nd_repstat = NFSERR_BADXDR;
8723 	}
8724 	if (error == 0)
8725 		error = nd->nd_repstat;
8726 nfsmout:
8727 	m_freem(nd->nd_mrep);
8728 	return (error);
8729 }
8730 
8731 /*
8732  * Seek operation.
8733  */
8734 int
8735 nfsrpc_seek(vnode_t vp, off_t *offp, bool *eofp, int content,
8736     struct ucred *cred, struct nfsvattr *nap, int *attrflagp)
8737 {
8738 	int error, expireret = 0, retrycnt;
8739 	u_int32_t clidrev = 0;
8740 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
8741 	struct nfsnode *np = VTONFS(vp);
8742 	struct nfsfh *nfhp = NULL;
8743 	nfsv4stateid_t stateid;
8744 	void *lckp;
8745 
8746 	if (nmp->nm_clp != NULL)
8747 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8748 	nfhp = np->n_fhp;
8749 	retrycnt = 0;
8750 	do {
8751 		lckp = NULL;
8752 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
8753 		    NFSV4OPEN_ACCESSREAD, 0, cred, curthread, &stateid, &lckp);
8754 		error = nfsrpc_seekrpc(vp, offp, &stateid, eofp, content,
8755 		    nap, attrflagp, cred);
8756 		if (error == NFSERR_STALESTATEID)
8757 			nfscl_initiate_recovery(nmp->nm_clp);
8758 		if (lckp != NULL)
8759 			nfscl_lockderef(lckp);
8760 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8761 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8762 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8763 			(void) nfs_catnap(PZERO, error, "nfs_seek");
8764 		} else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) &&
8765 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
8766 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8767 			    curthread);
8768 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) {
8769 			error = EIO;
8770 		}
8771 		retrycnt++;
8772 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8773 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8774 	    error == NFSERR_BADSESSION ||
8775 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8776 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8777 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
8778 	    (error == NFSERR_OPENMODE && retrycnt < 4));
8779 	if (error && retrycnt >= 4)
8780 		error = EIO;
8781 	return (error);
8782 }
8783 
8784 /*
8785  * The seek RPC.
8786  */
8787 static int
8788 nfsrpc_seekrpc(vnode_t vp, off_t *offp, nfsv4stateid_t *stateidp, bool *eofp,
8789     int content, struct nfsvattr *nap, int *attrflagp, struct ucred *cred)
8790 {
8791 	uint32_t *tl;
8792 	int error;
8793 	struct nfsrv_descript nfsd;
8794 	struct nfsrv_descript *nd = &nfsd;
8795 	nfsattrbit_t attrbits;
8796 
8797 	*attrflagp = 0;
8798 	NFSCL_REQSTART(nd, NFSPROC_SEEK, vp, cred);
8799 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
8800 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8801 	txdr_hyper(*offp, tl); tl += 2;
8802 	*tl++ = txdr_unsigned(content);
8803 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8804 	NFSGETATTR_ATTRBIT(&attrbits);
8805 	nfsrv_putattrbit(nd, &attrbits);
8806 	error = nfscl_request(nd, vp, curthread, cred);
8807 	if (error != 0)
8808 		return (error);
8809 	if (nd->nd_repstat == 0) {
8810 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED + NFSX_HYPER);
8811 		if (*tl++ == newnfs_true)
8812 			*eofp = true;
8813 		else
8814 			*eofp = false;
8815 		*offp = fxdr_hyper(tl);
8816 		/* Just skip over Getattr op status. */
8817 		error = nfsm_loadattr(nd, nap);
8818 		if (error == 0)
8819 			*attrflagp = 1;
8820 	}
8821 	error = nd->nd_repstat;
8822 nfsmout:
8823 	m_freem(nd->nd_mrep);
8824 	return (error);
8825 }
8826 
8827 /*
8828  * The getextattr RPC.
8829  */
8830 int
8831 nfsrpc_getextattr(vnode_t vp, const char *name, struct uio *uiop, ssize_t *lenp,
8832     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8833 {
8834 	uint32_t *tl;
8835 	int error;
8836 	struct nfsrv_descript nfsd;
8837 	struct nfsrv_descript *nd = &nfsd;
8838 	nfsattrbit_t attrbits;
8839 	uint32_t len, len2;
8840 
8841 	*attrflagp = 0;
8842 	NFSCL_REQSTART(nd, NFSPROC_GETEXTATTR, vp, cred);
8843 	nfsm_strtom(nd, name, strlen(name));
8844 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8845 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8846 	NFSGETATTR_ATTRBIT(&attrbits);
8847 	nfsrv_putattrbit(nd, &attrbits);
8848 	error = nfscl_request(nd, vp, p, cred);
8849 	if (error != 0)
8850 		return (error);
8851 	if (nd->nd_repstat == 0) {
8852 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8853 		len = fxdr_unsigned(uint32_t, *tl);
8854 		/* Sanity check lengths. */
8855 		if (uiop != NULL && len > 0 && len <= IOSIZE_MAX &&
8856 		    uiop->uio_resid <= UINT32_MAX) {
8857 			len2 = uiop->uio_resid;
8858 			if (len2 >= len)
8859 				error = nfsm_mbufuio(nd, uiop, len);
8860 			else {
8861 				error = nfsm_mbufuio(nd, uiop, len2);
8862 				if (error == 0) {
8863 					/*
8864 					 * nfsm_mbufuio() advances to a multiple
8865 					 * of 4, so round up len2 as well.  Then
8866 					 * we need to advance over the rest of
8867 					 * the data, rounding up the remaining
8868 					 * length.
8869 					 */
8870 					len2 = NFSM_RNDUP(len2);
8871 					len2 = NFSM_RNDUP(len - len2);
8872 					if (len2 > 0)
8873 						error = nfsm_advance(nd, len2,
8874 						    -1);
8875 				}
8876 			}
8877 		} else if (uiop == NULL && len > 0) {
8878 			/* Just wants the length and not the data. */
8879 			error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8880 		} else if (len > 0)
8881 			error = ENOATTR;
8882 		if (error != 0)
8883 			goto nfsmout;
8884 		*lenp = len;
8885 		/* Just skip over Getattr op status. */
8886 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8887 		error = nfsm_loadattr(nd, nap);
8888 		if (error == 0)
8889 			*attrflagp = 1;
8890 	}
8891 	if (error == 0)
8892 		error = nd->nd_repstat;
8893 nfsmout:
8894 	m_freem(nd->nd_mrep);
8895 	return (error);
8896 }
8897 
8898 /*
8899  * The setextattr RPC.
8900  */
8901 int
8902 nfsrpc_setextattr(vnode_t vp, const char *name, struct uio *uiop,
8903     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8904 {
8905 	uint32_t *tl;
8906 	int error;
8907 	struct nfsrv_descript nfsd;
8908 	struct nfsrv_descript *nd = &nfsd;
8909 	nfsattrbit_t attrbits;
8910 
8911 	*attrflagp = 0;
8912 	NFSCL_REQSTART(nd, NFSPROC_SETEXTATTR, vp, cred);
8913 	if (uiop->uio_resid > nd->nd_maxreq) {
8914 		/* nd_maxreq is set by NFSCL_REQSTART(). */
8915 		m_freem(nd->nd_mreq);
8916 		return (EINVAL);
8917 	}
8918 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8919 	*tl = txdr_unsigned(NFSV4SXATTR_EITHER);
8920 	nfsm_strtom(nd, name, strlen(name));
8921 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8922 	*tl = txdr_unsigned(uiop->uio_resid);
8923 	nfsm_uiombuf(nd, uiop, uiop->uio_resid);
8924 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8925 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8926 	NFSGETATTR_ATTRBIT(&attrbits);
8927 	nfsrv_putattrbit(nd, &attrbits);
8928 	error = nfscl_request(nd, vp, p, cred);
8929 	if (error != 0)
8930 		return (error);
8931 	if (nd->nd_repstat == 0) {
8932 		/* Just skip over the reply and Getattr op status. */
8933 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8934 		    NFSX_UNSIGNED);
8935 		error = nfsm_loadattr(nd, nap);
8936 		if (error == 0)
8937 			*attrflagp = 1;
8938 	}
8939 	if (error == 0)
8940 		error = nd->nd_repstat;
8941 nfsmout:
8942 	m_freem(nd->nd_mrep);
8943 	return (error);
8944 }
8945 
8946 /*
8947  * The removeextattr RPC.
8948  */
8949 int
8950 nfsrpc_rmextattr(vnode_t vp, const char *name, struct nfsvattr *nap,
8951     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8952 {
8953 	uint32_t *tl;
8954 	int error;
8955 	struct nfsrv_descript nfsd;
8956 	struct nfsrv_descript *nd = &nfsd;
8957 	nfsattrbit_t attrbits;
8958 
8959 	*attrflagp = 0;
8960 	NFSCL_REQSTART(nd, NFSPROC_RMEXTATTR, vp, cred);
8961 	nfsm_strtom(nd, name, strlen(name));
8962 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8963 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8964 	NFSGETATTR_ATTRBIT(&attrbits);
8965 	nfsrv_putattrbit(nd, &attrbits);
8966 	error = nfscl_request(nd, vp, p, cred);
8967 	if (error != 0)
8968 		return (error);
8969 	if (nd->nd_repstat == 0) {
8970 		/* Just skip over the reply and Getattr op status. */
8971 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8972 		    NFSX_UNSIGNED);
8973 		error = nfsm_loadattr(nd, nap);
8974 		if (error == 0)
8975 			*attrflagp = 1;
8976 	}
8977 	if (error == 0)
8978 		error = nd->nd_repstat;
8979 nfsmout:
8980 	m_freem(nd->nd_mrep);
8981 	return (error);
8982 }
8983 
8984 /*
8985  * The listextattr RPC.
8986  */
8987 int
8988 nfsrpc_listextattr(vnode_t vp, uint64_t *cookiep, struct uio *uiop,
8989     size_t *lenp, bool *eofp, struct nfsvattr *nap, int *attrflagp,
8990     struct ucred *cred, NFSPROC_T *p)
8991 {
8992 	uint32_t *tl;
8993 	int cnt, error, i, len;
8994 	struct nfsrv_descript nfsd;
8995 	struct nfsrv_descript *nd = &nfsd;
8996 	nfsattrbit_t attrbits;
8997 	u_char c;
8998 
8999 	*attrflagp = 0;
9000 	NFSCL_REQSTART(nd, NFSPROC_LISTEXTATTR, vp, cred);
9001 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
9002 	txdr_hyper(*cookiep, tl); tl += 2;
9003 	*tl++ = txdr_unsigned(*lenp);
9004 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
9005 	NFSGETATTR_ATTRBIT(&attrbits);
9006 	nfsrv_putattrbit(nd, &attrbits);
9007 	error = nfscl_request(nd, vp, p, cred);
9008 	if (error != 0)
9009 		return (error);
9010 	*eofp = true;
9011 	*lenp = 0;
9012 	if (nd->nd_repstat == 0) {
9013 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
9014 		*cookiep = fxdr_hyper(tl); tl += 2;
9015 		cnt = fxdr_unsigned(int, *tl);
9016 		if (cnt < 0) {
9017 			error = EBADRPC;
9018 			goto nfsmout;
9019 		}
9020 		for (i = 0; i < cnt; i++) {
9021 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
9022 			len = fxdr_unsigned(int, *tl);
9023 			if (len <= 0 || len > EXTATTR_MAXNAMELEN) {
9024 				error = EBADRPC;
9025 				goto nfsmout;
9026 			}
9027 			if (uiop == NULL)
9028 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
9029 			else if (uiop->uio_resid >= len + 1) {
9030 				c = len;
9031 				error = uiomove(&c, sizeof(c), uiop);
9032 				if (error == 0)
9033 					error = nfsm_mbufuio(nd, uiop, len);
9034 			} else {
9035 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
9036 				*eofp = false;
9037 			}
9038 			if (error != 0)
9039 				goto nfsmout;
9040 			*lenp += (len + 1);
9041 		}
9042 		/* Get the eof and skip over the Getattr op status. */
9043 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
9044 		/*
9045 		 * *eofp is set false above, because it wasn't able to copy
9046 		 * all of the reply.
9047 		 */
9048 		if (*eofp && *tl == 0)
9049 			*eofp = false;
9050 		error = nfsm_loadattr(nd, nap);
9051 		if (error == 0)
9052 			*attrflagp = 1;
9053 	}
9054 	if (error == 0)
9055 		error = nd->nd_repstat;
9056 nfsmout:
9057 	m_freem(nd->nd_mrep);
9058 	return (error);
9059 }
9060 
9061 /*
9062  * Split an mbuf list.  For non-M_EXTPG mbufs, just use m_split().
9063  */
9064 static struct mbuf *
9065 nfsm_split(struct mbuf *mp, uint64_t xfer)
9066 {
9067 	struct mbuf *m, *m2;
9068 	vm_page_t pg;
9069 	int i, j, left, pgno, plen, trim;
9070 	char *cp, *cp2;
9071 
9072 	if ((mp->m_flags & M_EXTPG) == 0) {
9073 		m = m_split(mp, xfer, M_WAITOK);
9074 		return (m);
9075 	}
9076 
9077 	/* Find the correct mbuf to split at. */
9078 	for (m = mp; m != NULL && xfer > m->m_len; m = m->m_next)
9079 		xfer -= m->m_len;
9080 	if (m == NULL)
9081 		return (NULL);
9082 
9083 	/* If xfer == m->m_len, we can just split the mbuf list. */
9084 	if (xfer == m->m_len) {
9085 		m2 = m->m_next;
9086 		m->m_next = NULL;
9087 		return (m2);
9088 	}
9089 
9090 	/* Find the page to split at. */
9091 	pgno = 0;
9092 	left = xfer;
9093 	do {
9094 		if (pgno == 0)
9095 			plen = m_epg_pagelen(m, 0, m->m_epg_1st_off);
9096 		else
9097 			plen = m_epg_pagelen(m, pgno, 0);
9098 		if (left <= plen)
9099 			break;
9100 		left -= plen;
9101 		pgno++;
9102 	} while (pgno < m->m_epg_npgs);
9103 	if (pgno == m->m_epg_npgs)
9104 		panic("nfsm_split: eroneous ext_pgs mbuf");
9105 
9106 	m2 = mb_alloc_ext_pgs(M_WAITOK, mb_free_mext_pgs);
9107 	m2->m_epg_flags |= EPG_FLAG_ANON;
9108 
9109 	/*
9110 	 * If left < plen, allocate a new page for the new mbuf
9111 	 * and copy the data after left in the page to this new
9112 	 * page.
9113 	 */
9114 	if (left < plen) {
9115 		pg = vm_page_alloc_noobj(VM_ALLOC_WAITOK | VM_ALLOC_NODUMP |
9116 		    VM_ALLOC_WIRED);
9117 		m2->m_epg_pa[0] = VM_PAGE_TO_PHYS(pg);
9118 		m2->m_epg_npgs = 1;
9119 
9120 		/* Copy the data after left to the new page. */
9121 		trim = plen - left;
9122 		cp = (char *)(void *)PHYS_TO_DMAP(m->m_epg_pa[pgno]);
9123 		if (pgno == 0)
9124 			cp += m->m_epg_1st_off;
9125 		cp += left;
9126 		cp2 = (char *)(void *)PHYS_TO_DMAP(m2->m_epg_pa[0]);
9127 		if (pgno == m->m_epg_npgs - 1)
9128 			m2->m_epg_last_len = trim;
9129 		else {
9130 			cp2 += PAGE_SIZE - trim;
9131 			m2->m_epg_1st_off = PAGE_SIZE - trim;
9132 			m2->m_epg_last_len = m->m_epg_last_len;
9133 		}
9134 		memcpy(cp2, cp, trim);
9135 		m2->m_len = trim;
9136 	} else {
9137 		m2->m_len = 0;
9138 		m2->m_epg_last_len = m->m_epg_last_len;
9139 	}
9140 
9141 	/* Move the pages beyond pgno to the new mbuf. */
9142 	for (i = pgno + 1, j = m2->m_epg_npgs; i < m->m_epg_npgs; i++, j++) {
9143 		m2->m_epg_pa[j] = m->m_epg_pa[i];
9144 		/* Never moves page 0. */
9145 		m2->m_len += m_epg_pagelen(m, i, 0);
9146 	}
9147 	m2->m_epg_npgs = j;
9148 	m->m_epg_npgs = pgno + 1;
9149 	m->m_epg_last_len = left;
9150 	m->m_len = xfer;
9151 
9152 	m2->m_next = m->m_next;
9153 	m->m_next = NULL;
9154 	return (m2);
9155 }
9156 
9157 /*
9158  * Do the NFSv4.1 Bind Connection to Session.
9159  * Called from the reconnect layer of the krpc (sys/rpc/clnt_rc.c).
9160  */
9161 void
9162 nfsrpc_bindconnsess(CLIENT *cl, void *arg, struct ucred *cr)
9163 {
9164 	struct nfscl_reconarg *rcp = (struct nfscl_reconarg *)arg;
9165 	uint32_t res, *tl;
9166 	struct nfsrv_descript nfsd;
9167 	struct nfsrv_descript *nd = &nfsd;
9168 	struct rpc_callextra ext;
9169 	struct timeval utimeout;
9170 	enum clnt_stat stat;
9171 	int error;
9172 
9173 	nfscl_reqstart(nd, NFSPROC_BINDCONNTOSESS, NULL, NULL, 0, NULL, NULL,
9174 	    NFS_VER4, rcp->minorvers, NULL);
9175 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID + 2 * NFSX_UNSIGNED);
9176 	memcpy(tl, rcp->sessionid, NFSX_V4SESSIONID);
9177 	tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
9178 	*tl++ = txdr_unsigned(NFSCDFC4_FORE_OR_BOTH);
9179 	*tl = newnfs_false;
9180 
9181 	memset(&ext, 0, sizeof(ext));
9182 	utimeout.tv_sec = 30;
9183 	utimeout.tv_usec = 0;
9184 	ext.rc_auth = authunix_create(cr);
9185 	nd->nd_mrep = NULL;
9186 	stat = CLNT_CALL_MBUF(cl, &ext, NFSV4PROC_COMPOUND, nd->nd_mreq,
9187 	    &nd->nd_mrep, utimeout);
9188 	AUTH_DESTROY(ext.rc_auth);
9189 	if (stat != RPC_SUCCESS) {
9190 		printf("nfsrpc_bindconnsess: call failed stat=%d\n", stat);
9191 		return;
9192 	}
9193 	if (nd->nd_mrep == NULL) {
9194 		printf("nfsrpc_bindconnsess: no reply args\n");
9195 		return;
9196 	}
9197 	error = 0;
9198 	newnfs_realign(&nd->nd_mrep, M_WAITOK);
9199 	nd->nd_md = nd->nd_mrep;
9200 	nd->nd_dpos = mtod(nd->nd_md, char *);
9201 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
9202 	nd->nd_repstat = fxdr_unsigned(uint32_t, *tl++);
9203 	if (nd->nd_repstat == NFSERR_OK) {
9204 		res = fxdr_unsigned(uint32_t, *tl);
9205 		if (res > 0 && (error = nfsm_advance(nd, NFSM_RNDUP(res),
9206 		    -1)) != 0)
9207 			goto nfsmout;
9208 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
9209 		    4 * NFSX_UNSIGNED);
9210 		tl += 3;
9211 		if (!NFSBCMP(tl, rcp->sessionid, NFSX_V4SESSIONID)) {
9212 			tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
9213 			res = fxdr_unsigned(uint32_t, *tl);
9214 			if (res != NFSCDFS4_BOTH)
9215 				printf("nfsrpc_bindconnsess: did not "
9216 				    "return FS4_BOTH\n");
9217 		} else
9218 			printf("nfsrpc_bindconnsess: not same "
9219 			    "sessionid\n");
9220 	} else if (nd->nd_repstat != NFSERR_BADSESSION)
9221 		printf("nfsrpc_bindconnsess: returned %d\n", nd->nd_repstat);
9222 nfsmout:
9223 	if (error != 0)
9224 		printf("nfsrpc_bindconnsess: reply bad xdr\n");
9225 	m_freem(nd->nd_mrep);
9226 }
9227 
9228 /*
9229  * Do roughly what nfs_statfs() does for NFSv4, but when called with a shared
9230  * locked vnode.
9231  */
9232 static void
9233 nfscl_statfs(struct vnode *vp, struct ucred *cred, NFSPROC_T *td)
9234 {
9235 	struct nfsvattr nfsva;
9236 	struct nfsfsinfo fs;
9237 	struct nfsstatfs sb;
9238 	struct mount *mp;
9239 	struct nfsmount *nmp;
9240 	uint32_t lease;
9241 	int attrflag, error;
9242 
9243 	mp = vp->v_mount;
9244 	nmp = VFSTONFS(mp);
9245 	error = nfsrpc_statfs(vp, &sb, &fs, &lease, cred, td, &nfsva,
9246 	    &attrflag);
9247 	if (attrflag != 0)
9248 		(void) nfscl_loadattrcache(&vp, &nfsva, NULL, 0, 1);
9249 	if (error == 0) {
9250 		NFSLOCKCLSTATE();
9251 		if (nmp->nm_clp != NULL)
9252 			nmp->nm_clp->nfsc_renew = NFSCL_RENEW(lease);
9253 		NFSUNLOCKCLSTATE();
9254 		mtx_lock(&nmp->nm_mtx);
9255 		nfscl_loadfsinfo(nmp, &fs);
9256 		nfscl_loadsbinfo(nmp, &sb, &mp->mnt_stat);
9257 		mp->mnt_stat.f_iosize = newnfs_iosize(nmp);
9258 		mtx_unlock(&nmp->nm_mtx);
9259 	}
9260 }
9261