xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision 47ef2a131091508e049ab10cad7f91a3c1342cd9)
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 /*
38  * Rpc op calls, generally called from the vnode op calls or through the
39  * buffer cache, for NFS v2, 3 and 4.
40  * These do not normally make any changes to vnode arguments or use
41  * structures that might change between the VFS variants. The returned
42  * arguments are all at the end, after the NFSPROC_T *p one.
43  */
44 
45 #include "opt_inet6.h"
46 
47 #include <fs/nfs/nfsport.h>
48 #include <fs/nfsclient/nfs.h>
49 #include <sys/extattr.h>
50 #include <sys/sysctl.h>
51 #include <sys/taskqueue.h>
52 
53 SYSCTL_DECL(_vfs_nfs);
54 
55 static int	nfsignore_eexist = 0;
56 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
57     &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
58 
59 static int	nfscl_dssameconn = 0;
60 SYSCTL_INT(_vfs_nfs, OID_AUTO, dssameconn, CTLFLAG_RW,
61     &nfscl_dssameconn, 0, "Use same TCP connection to multiple DSs");
62 
63 static uint64_t nfs_maxcopyrange = SSIZE_MAX;
64 SYSCTL_U64(_vfs_nfs, OID_AUTO, maxcopyrange, CTLFLAG_RW,
65     &nfs_maxcopyrange, 0, "Max size of a Copy so RPC times reasonable");
66 
67 /*
68  * Global variables
69  */
70 extern struct nfsstatsv1 nfsstatsv1;
71 extern int nfs_numnfscbd;
72 extern struct timeval nfsboottime;
73 extern u_int32_t newnfs_false, newnfs_true;
74 extern nfstype nfsv34_type[9];
75 extern int nfsrv_useacl;
76 extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
77 extern int nfscl_debuglevel;
78 extern int nfs_pnfsiothreads;
79 extern u_long sb_max_adj;
80 NFSCLSTATEMUTEX;
81 int nfstest_outofseq = 0;
82 int nfscl_assumeposixlocks = 1;
83 int nfscl_enablecallb = 0;
84 short nfsv4_cbport = NFSV4_CBPORT;
85 int nfstest_openallsetattr = 0;
86 
87 #define	DIRHDSIZ	offsetof(struct dirent, d_name)
88 
89 /*
90  * nfscl_getsameserver() can return one of three values:
91  * NFSDSP_USETHISSESSION - Use this session for the DS.
92  * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
93  *     session.
94  * NFSDSP_NOTFOUND - No matching server was found.
95  */
96 enum nfsclds_state {
97 	NFSDSP_USETHISSESSION = 0,
98 	NFSDSP_SEQTHISSESSION = 1,
99 	NFSDSP_NOTFOUND = 2,
100 };
101 
102 /*
103  * Do a write RPC on a DS data file, using this structure for the arguments,
104  * so that this function can be executed by a separate kernel process.
105  */
106 struct nfsclwritedsdorpc {
107 	int			done;
108 	int			inprog;
109 	struct task		tsk;
110 	struct vnode		*vp;
111 	int			iomode;
112 	int			must_commit;
113 	nfsv4stateid_t		*stateidp;
114 	struct nfsclds		*dsp;
115 	uint64_t		off;
116 	int			len;
117 #ifdef notyet
118 	int			advise;
119 #endif
120 	struct nfsfh		*fhp;
121 	struct mbuf		*m;
122 	int			vers;
123 	int			minorvers;
124 	struct ucred		*cred;
125 	NFSPROC_T		*p;
126 	int			err;
127 };
128 
129 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
130     struct ucred *, NFSPROC_T *, struct nfsvattr *, int *);
131 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
132     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *);
133 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
134     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
135     int);
136 static int nfsrpc_deallocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
137     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *);
138 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
139     nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
140     struct nfsvattr *, struct nfsfh **, int *, int *);
141 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
142     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
143     NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
144     int *, int *);
145 static bool nfscl_invalidfname(bool, char *, int);
146 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
147     struct nfscllockowner *, u_int64_t, u_int64_t,
148     u_int32_t, struct ucred *, NFSPROC_T *, int);
149 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
150     struct acl *, nfsv4stateid_t *);
151 static int nfsrpc_layouterror(struct nfsmount *, uint8_t *, int, uint64_t,
152     uint64_t, nfsv4stateid_t *, struct ucred *, NFSPROC_T *, uint32_t,
153     uint32_t, char *);
154 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
155     uint32_t, uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
156     struct ucred *, NFSPROC_T *);
157 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
158     struct sockaddr_in6 *, sa_family_t, int, int, struct nfsclds **,
159     NFSPROC_T *);
160 static void nfscl_initsessionslots(struct nfsclsession *);
161 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
162     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
163     struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
164     NFSPROC_T *);
165 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
166     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
167     struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
168     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
169 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
170     struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
171     struct ucred *, NFSPROC_T *);
172 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
173     nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
174     struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
175 static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
176     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
177     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
178 static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
179     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
180     struct ucred *, NFSPROC_T *);
181 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
182     struct nfsclds *, struct nfsclds **, uint32_t *);
183 static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
184     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
185     NFSPROC_T *);
186 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
187     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
188 #ifdef notyet
189 static int nfsio_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
190     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
191     NFSPROC_T *);
192 static int nfsrpc_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
193     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
194 #endif
195 static int nfsrpc_allocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
196     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *);
197 static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
198     uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
199 static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
200     NFSPROC_T *);
201 static int nfsrv_parselayoutget(struct nfsmount *, struct nfsrv_descript *,
202     nfsv4stateid_t *, int *, struct nfsclflayouthead *);
203 static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
204     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
205     struct nfscldeleg **, struct ucred *, NFSPROC_T *);
206 static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
207     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
208     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
209     struct nfsfh **, int *, int *, int *);
210 static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
211     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
212     struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
213     struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
214 static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
215     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
216     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
217     struct nfsfh **, int *, int *, int *, nfsv4stateid_t *,
218     int, int, int, int *, struct nfsclflayouthead *, int *);
219 static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
220     uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
221     struct nfsclflayouthead *, struct ucred *, NFSPROC_T *);
222 static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
223     int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
224     struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
225 static int nfsrpc_copyrpc(vnode_t, off_t, vnode_t, off_t, size_t *,
226     nfsv4stateid_t *, nfsv4stateid_t *, struct nfsvattr *, int *,
227     struct nfsvattr *, int *, bool, int *, struct ucred *, NFSPROC_T *);
228 static int nfsrpc_seekrpc(vnode_t, off_t *, nfsv4stateid_t *, bool *,
229     int, struct nfsvattr *, int *, struct ucred *);
230 static struct mbuf *nfsm_split(struct mbuf *, uint64_t);
231 static void nfscl_statfs(struct vnode *, struct ucred *, NFSPROC_T *);
232 
233 int nfs_pnfsio(task_fn_t *, void *);
234 
235 /*
236  * nfs null call from vfs.
237  */
238 int
239 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
240 {
241 	int error;
242 	struct nfsrv_descript nfsd, *nd = &nfsd;
243 
244 	NFSCL_REQSTART(nd, NFSPROC_NULL, vp, NULL);
245 	error = nfscl_request(nd, vp, p, cred);
246 	if (nd->nd_repstat && !error)
247 		error = nd->nd_repstat;
248 	m_freem(nd->nd_mrep);
249 	return (error);
250 }
251 
252 /*
253  * nfs access rpc op.
254  * For nfs version 3 and 4, use the access rpc to check accessibility. If file
255  * modes are changed on the server, accesses might still fail later.
256  */
257 int
258 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
259     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
260 {
261 	int error;
262 	u_int32_t mode, rmode;
263 
264 	if (acmode & VREAD)
265 		mode = NFSACCESS_READ;
266 	else
267 		mode = 0;
268 	if (vp->v_type == VDIR) {
269 		if (acmode & VWRITE)
270 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
271 				 NFSACCESS_DELETE);
272 		if (acmode & VEXEC)
273 			mode |= NFSACCESS_LOOKUP;
274 	} else {
275 		if (acmode & VWRITE)
276 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
277 		if (acmode & VEXEC)
278 			mode |= NFSACCESS_EXECUTE;
279 	}
280 
281 	/*
282 	 * Now, just call nfsrpc_accessrpc() to do the actual RPC.
283 	 */
284 	error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode);
285 
286 	/*
287 	 * The NFS V3 spec does not clarify whether or not
288 	 * the returned access bits can be a superset of
289 	 * the ones requested, so...
290 	 */
291 	if (!error && (rmode & mode) != mode)
292 		error = EACCES;
293 	return (error);
294 }
295 
296 /*
297  * The actual rpc, separated out for Darwin.
298  */
299 int
300 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
301     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep)
302 {
303 	u_int32_t *tl;
304 	u_int32_t supported, rmode;
305 	int error;
306 	struct nfsrv_descript nfsd, *nd = &nfsd;
307 	nfsattrbit_t attrbits;
308 	struct nfsmount *nmp;
309 	struct nfsnode *np;
310 
311 	*attrflagp = 0;
312 	supported = mode;
313 	nmp = VFSTONFS(vp->v_mount);
314 	np = VTONFS(vp);
315 	if ((nmp->nm_privflag & NFSMNTP_FAKEROOTFH) != 0 &&
316 	    nmp->nm_fhsize == 0) {
317 		/* Attempt to get the actual root file handle. */
318 		error = nfsrpc_getdirpath(nmp, NFSMNT_DIRPATH(nmp), cred, p);
319 		if (error != 0)
320 			return (EACCES);
321 		if (np->n_fhp->nfh_len == NFSX_FHMAX + 1)
322 			nfscl_statfs(vp, cred, p);
323 	}
324 	NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp, cred);
325 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
326 	*tl = txdr_unsigned(mode);
327 	if (nd->nd_flag & ND_NFSV4) {
328 		/*
329 		 * And do a Getattr op.
330 		 */
331 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
332 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
333 		NFSGETATTR_ATTRBIT(&attrbits);
334 		(void) nfsrv_putattrbit(nd, &attrbits);
335 	}
336 	error = nfscl_request(nd, vp, p, cred);
337 	if (error)
338 		return (error);
339 	if (nd->nd_flag & ND_NFSV3) {
340 		error = nfscl_postop_attr(nd, nap, attrflagp);
341 		if (error)
342 			goto nfsmout;
343 	}
344 	if (!nd->nd_repstat) {
345 		if (nd->nd_flag & ND_NFSV4) {
346 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
347 			supported = fxdr_unsigned(u_int32_t, *tl++);
348 		} else {
349 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
350 		}
351 		rmode = fxdr_unsigned(u_int32_t, *tl);
352 		if (nd->nd_flag & ND_NFSV4)
353 			error = nfscl_postop_attr(nd, nap, attrflagp);
354 
355 		/*
356 		 * It's not obvious what should be done about
357 		 * unsupported access modes. For now, be paranoid
358 		 * and clear the unsupported ones.
359 		 */
360 		rmode &= supported;
361 		*rmodep = rmode;
362 	} else
363 		error = nd->nd_repstat;
364 nfsmout:
365 	m_freem(nd->nd_mrep);
366 	return (error);
367 }
368 
369 /*
370  * nfs open rpc
371  */
372 int
373 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
374 {
375 	struct nfsclopen *op;
376 	struct nfscldeleg *dp;
377 	struct nfsfh *nfhp;
378 	struct nfsnode *np = VTONFS(vp);
379 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
380 	u_int32_t mode, clidrev;
381 	int ret, newone, error, expireret = 0, retrycnt;
382 
383 	/*
384 	 * For NFSv4, Open Ops are only done on Regular Files.
385 	 */
386 	if (vp->v_type != VREG)
387 		return (0);
388 	mode = 0;
389 	if (amode & FREAD)
390 		mode |= NFSV4OPEN_ACCESSREAD;
391 	if (amode & FWRITE)
392 		mode |= NFSV4OPEN_ACCESSWRITE;
393 	if (NFSHASNFSV4N(nmp)) {
394 		if (!NFSHASPNFS(nmp) && nfscl_enablecallb != 0 &&
395 		    nfs_numnfscbd > 0) {
396 			if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
397 				mode |= NFSV4OPEN_WANTWRITEDELEG;
398 			else
399 				mode |= NFSV4OPEN_WANTANYDELEG;
400 		} else
401 			mode |= NFSV4OPEN_WANTNODELEG;
402 	}
403 	nfhp = np->n_fhp;
404 
405 	retrycnt = 0;
406 	do {
407 	    dp = NULL;
408 	    error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len,
409 		(mode & NFSV4OPEN_ACCESSBOTH), 1, cred, p, NULL,
410 		&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 if (NFSHASNFSV4N(nmp)) {
458 			/*
459 			 * For the first attempt, try and get a layout, if
460 			 * pNFS is enabled for the mount.
461 			 */
462 			if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
463 			    nfs_numnfscbd == 0 ||
464 			    (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
465 				error = nfsrpc_openrpc(nmp, vp, nfhp->nfh_fh,
466 				    nfhp->nfh_len, nfhp->nfh_fh, nfhp->nfh_len,
467 				    mode, op, NULL, 0, &dp, 0, 0x0, cred, p, 0,
468 				    0);
469 			else
470 				error = nfsrpc_getopenlayout(nmp, vp,
471 				    nfhp->nfh_fh, nfhp->nfh_len, nfhp->nfh_fh,
472 				    nfhp->nfh_len, mode, op, NULL, 0, &dp,
473 				    cred, p);
474 			if (dp != NULL) {
475 				NFSLOCKNODE(np);
476 				np->n_flag &= ~NDELEGMOD;
477 				/*
478 				 * Invalidate the attribute cache, so that
479 				 * attributes that pre-date the issue of a
480 				 * delegation are not cached, since the
481 				 * cached attributes will remain valid while
482 				 * the delegation is held.
483 				 */
484 				NFSINVALATTRCACHE(np);
485 				NFSUNLOCKNODE(np);
486 				(void) nfscl_deleg(nmp->nm_mountp,
487 				    op->nfso_own->nfsow_clp,
488 				    nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
489 			}
490 		} else {
491 			error = EIO;
492 		}
493 		newnfs_copyincred(cred, &op->nfso_cred);
494 	    } else if (ret == NFSCLOPEN_SETCRED)
495 		/*
496 		 * This is a new local open on a delegation. It needs
497 		 * to have credentials so that an open can be done
498 		 * against the server during recovery.
499 		 */
500 		newnfs_copyincred(cred, &op->nfso_cred);
501 
502 	    /*
503 	     * nfso_opencnt is the count of how many VOP_OPEN()s have
504 	     * been done on this Open successfully and a VOP_CLOSE()
505 	     * is expected for each of these.
506 	     * If error is non-zero, don't increment it, since the Open
507 	     * hasn't succeeded yet.
508 	     */
509 	    if (!error) {
510 		op->nfso_opencnt++;
511 		if (NFSHASNFSV4N(nmp) && NFSHASONEOPENOWN(nmp)) {
512 		    NFSLOCKNODE(np);
513 		    np->n_openstateid = op;
514 		    NFSUNLOCKNODE(np);
515 		}
516 	    }
517 	    nfscl_openrelease(nmp, op, error, newone);
518 	    if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
519 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
520 		error == NFSERR_BADSESSION) {
521 		(void) nfs_catnap(PZERO, error, "nfs_open");
522 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
523 		&& clidrev != 0) {
524 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
525 		retrycnt++;
526 	    }
527 	} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
528 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
529 	    error == NFSERR_BADSESSION ||
530 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
531 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
532 	if (error && retrycnt >= 4)
533 		error = EIO;
534 	return (error);
535 }
536 
537 /*
538  * the actual open rpc
539  */
540 int
541 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
542     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
543     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
544     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
545     int syscred, int recursed)
546 {
547 	u_int32_t *tl;
548 	struct nfsrv_descript nfsd, *nd = &nfsd;
549 	struct nfscldeleg *dp, *ndp = NULL;
550 	struct nfsvattr nfsva;
551 	u_int32_t rflags, deleg;
552 	nfsattrbit_t attrbits;
553 	int error, ret, acesize, limitby;
554 	struct nfsclsession *tsep;
555 
556 	dp = *dpp;
557 	*dpp = NULL;
558 	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0,
559 	    cred);
560 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
561 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
562 	*tl++ = txdr_unsigned(mode & (NFSV4OPEN_ACCESSBOTH |
563 	    NFSV4OPEN_WANTDELEGMASK));
564 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
565 	tsep = nfsmnt_mdssession(nmp);
566 	*tl++ = tsep->nfsess_clientid.lval[0];
567 	*tl = tsep->nfsess_clientid.lval[1];
568 	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
569 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
570 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
571 	if (reclaim) {
572 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
573 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
574 		*tl = txdr_unsigned(delegtype);
575 	} else {
576 		if (dp != NULL) {
577 			if (NFSHASNFSV4N(nmp))
578 				*tl = txdr_unsigned(
579 				    NFSV4OPEN_CLAIMDELEGATECURFH);
580 			else
581 				*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
582 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
583 			if (NFSHASNFSV4N(nmp))
584 				*tl++ = 0;
585 			else
586 				*tl++ = dp->nfsdl_stateid.seqid;
587 			*tl++ = dp->nfsdl_stateid.other[0];
588 			*tl++ = dp->nfsdl_stateid.other[1];
589 			*tl = dp->nfsdl_stateid.other[2];
590 			if (!NFSHASNFSV4N(nmp))
591 				(void)nfsm_strtom(nd, name, namelen);
592 		} else if (NFSHASNFSV4N(nmp)) {
593 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMFH);
594 		} else {
595 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
596 			(void)nfsm_strtom(nd, name, namelen);
597 		}
598 	}
599 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
600 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
601 	NFSZERO_ATTRBIT(&attrbits);
602 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
603 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
604 	(void) nfsrv_putattrbit(nd, &attrbits);
605 	if (syscred)
606 		nd->nd_flag |= ND_USEGSSNAME;
607 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
608 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
609 	if (error)
610 		return (error);
611 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
612 	if (nd->nd_repstat == 0 || (nd->nd_repstat == NFSERR_DELAY &&
613 	    reclaim != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0)) {
614 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
615 		    6 * NFSX_UNSIGNED);
616 		op->nfso_stateid.seqid = *tl++;
617 		op->nfso_stateid.other[0] = *tl++;
618 		op->nfso_stateid.other[1] = *tl++;
619 		op->nfso_stateid.other[2] = *tl;
620 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
621 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
622 		if (error)
623 			goto nfsmout;
624 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
625 		deleg = fxdr_unsigned(u_int32_t, *tl);
626 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
627 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
628 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
629 			      NFSCLFLAGS_FIRSTDELEG))
630 				op->nfso_own->nfsow_clp->nfsc_flags |=
631 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
632 			ndp = malloc(
633 			    sizeof (struct nfscldeleg) + newfhlen,
634 			    M_NFSCLDELEG, M_WAITOK);
635 			LIST_INIT(&ndp->nfsdl_owner);
636 			LIST_INIT(&ndp->nfsdl_lock);
637 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
638 			ndp->nfsdl_fhlen = newfhlen;
639 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
640 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
641 			nfscl_lockinit(&ndp->nfsdl_rwlock);
642 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
643 			    NFSX_UNSIGNED);
644 			ndp->nfsdl_stateid.seqid = *tl++;
645 			ndp->nfsdl_stateid.other[0] = *tl++;
646 			ndp->nfsdl_stateid.other[1] = *tl++;
647 			ndp->nfsdl_stateid.other[2] = *tl++;
648 			ret = fxdr_unsigned(int, *tl);
649 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
650 				ndp->nfsdl_flags = NFSCLDL_WRITE;
651 				/*
652 				 * Indicates how much the file can grow.
653 				 */
654 				NFSM_DISSECT(tl, u_int32_t *,
655 				    3 * NFSX_UNSIGNED);
656 				limitby = fxdr_unsigned(int, *tl++);
657 				switch (limitby) {
658 				case NFSV4OPEN_LIMITSIZE:
659 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
660 					break;
661 				case NFSV4OPEN_LIMITBLOCKS:
662 					ndp->nfsdl_sizelimit =
663 					    fxdr_unsigned(u_int64_t, *tl++);
664 					ndp->nfsdl_sizelimit *=
665 					    fxdr_unsigned(u_int64_t, *tl);
666 					break;
667 				default:
668 					error = NFSERR_BADXDR;
669 					goto nfsmout;
670 				}
671 			} else {
672 				ndp->nfsdl_flags = NFSCLDL_READ;
673 			}
674 			if (ret)
675 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
676 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, false,
677 			    &ret, &acesize, p);
678 			if (error)
679 				goto nfsmout;
680 		} else if (deleg == NFSV4OPEN_DELEGATENONEEXT &&
681 		    NFSHASNFSV4N(nmp)) {
682 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
683 			deleg = fxdr_unsigned(uint32_t, *tl);
684 			if (deleg == NFSV4OPEN_CONTENTION ||
685 			    deleg == NFSV4OPEN_RESOURCE)
686 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
687 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
688 			error = NFSERR_BADXDR;
689 			goto nfsmout;
690 		}
691 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
692 		/* If the 2nd element == NFS_OK, the Getattr succeeded. */
693 		if (*++tl == 0) {
694 			KASSERT(nd->nd_repstat == 0,
695 			    ("nfsrpc_openrpc: Getattr repstat"));
696 			error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
697 			    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
698 			    NULL, NULL, NULL, p, cred);
699 			if (error)
700 				goto nfsmout;
701 		}
702 		if (ndp != NULL) {
703 			if (reclaim != 0 && dp != NULL) {
704 				ndp->nfsdl_change = dp->nfsdl_change;
705 				ndp->nfsdl_modtime = dp->nfsdl_modtime;
706 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
707 			} else if (nd->nd_repstat == 0) {
708 				ndp->nfsdl_change = nfsva.na_filerev;
709 				ndp->nfsdl_modtime = nfsva.na_mtime;
710 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
711 			} else
712 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
713 		}
714 		nd->nd_repstat = 0;
715 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
716 		    do {
717 			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
718 			    cred, p);
719 			if (ret == NFSERR_DELAY)
720 			    (void) nfs_catnap(PZERO, ret, "nfs_open");
721 		    } while (ret == NFSERR_DELAY);
722 		    error = ret;
723 		}
724 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
725 		    nfscl_assumeposixlocks)
726 		    op->nfso_posixlock = 1;
727 		else
728 		    op->nfso_posixlock = 0;
729 
730 		/*
731 		 * If the server is handing out delegations, but we didn't
732 		 * get one because an OpenConfirm was required, try the
733 		 * Open again, to get a delegation. This is a harmless no-op,
734 		 * from a server's point of view.
735 		 */
736 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
737 		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
738 		    && !error && dp == NULL && ndp == NULL && !recursed) {
739 		    do {
740 			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
741 			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
742 			    cred, p, syscred, 1);
743 			if (ret == NFSERR_DELAY)
744 			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
745 		    } while (ret == NFSERR_DELAY);
746 		    if (ret) {
747 			if (ndp != NULL) {
748 				free(ndp, M_NFSCLDELEG);
749 				ndp = NULL;
750 			}
751 			if (ret == NFSERR_STALECLIENTID ||
752 			    ret == NFSERR_STALEDONTRECOVER ||
753 			    ret == NFSERR_BADSESSION)
754 				error = ret;
755 		    }
756 		}
757 	}
758 	if (nd->nd_repstat != 0 && error == 0)
759 		error = nd->nd_repstat;
760 	if (error == NFSERR_STALECLIENTID)
761 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
762 nfsmout:
763 	if (!error)
764 		*dpp = ndp;
765 	else if (ndp != NULL)
766 		free(ndp, M_NFSCLDELEG);
767 	m_freem(nd->nd_mrep);
768 	return (error);
769 }
770 
771 /*
772  * open downgrade rpc
773  */
774 int
775 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
776     struct ucred *cred, NFSPROC_T *p)
777 {
778 	u_int32_t *tl;
779 	struct nfsrv_descript nfsd, *nd = &nfsd;
780 	int error;
781 
782 	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp, cred);
783 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
784 	if (NFSHASNFSV4N(VFSTONFS(vp->v_mount)))
785 		*tl++ = 0;
786 	else
787 		*tl++ = op->nfso_stateid.seqid;
788 	*tl++ = op->nfso_stateid.other[0];
789 	*tl++ = op->nfso_stateid.other[1];
790 	*tl++ = op->nfso_stateid.other[2];
791 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
792 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
793 	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
794 	error = nfscl_request(nd, vp, p, cred);
795 	if (error)
796 		return (error);
797 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
798 	if (!nd->nd_repstat) {
799 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
800 		op->nfso_stateid.seqid = *tl++;
801 		op->nfso_stateid.other[0] = *tl++;
802 		op->nfso_stateid.other[1] = *tl++;
803 		op->nfso_stateid.other[2] = *tl;
804 	}
805 	if (nd->nd_repstat && error == 0)
806 		error = nd->nd_repstat;
807 	if (error == NFSERR_STALESTATEID)
808 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
809 nfsmout:
810 	m_freem(nd->nd_mrep);
811 	return (error);
812 }
813 
814 /*
815  * V4 Close operation.
816  */
817 int
818 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
819 {
820 	struct nfsclclient *clp;
821 	int error;
822 
823 	if (vp->v_type != VREG)
824 		return (0);
825 	if (doclose)
826 		error = nfscl_doclose(vp, &clp, p);
827 	else {
828 		error = nfscl_getclose(vp, &clp);
829 		if (error == 0)
830 			nfscl_clientrelease(clp);
831 	}
832 	return (error);
833 }
834 
835 /*
836  * Close the open.
837  */
838 int
839 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p,
840     bool loop_on_delayed, bool freeop)
841 {
842 	struct nfsrv_descript nfsd, *nd = &nfsd;
843 	struct nfscllockowner *lp, *nlp;
844 	struct nfscllock *lop, *nlop;
845 	struct ucred *tcred;
846 	u_int64_t off = 0, len = 0;
847 	u_int32_t type = NFSV4LOCKT_READ;
848 	int error, do_unlock, trycnt;
849 	bool own_not_null;
850 
851 	tcred = newnfs_getcred();
852 	newnfs_copycred(&op->nfso_cred, tcred);
853 	/*
854 	 * (Theoretically this could be done in the same
855 	 *  compound as the close, but having multiple
856 	 *  sequenced Ops in the same compound might be
857 	 *  too scary for some servers.)
858 	 */
859 	if (op->nfso_posixlock) {
860 		off = 0;
861 		len = NFS64BITSSET;
862 		type = NFSV4LOCKT_READ;
863 	}
864 
865 	/*
866 	 * Since this function is only called from VOP_INACTIVE(), no
867 	 * other thread will be manipulating this Open. As such, the
868 	 * lock lists are not being changed by other threads, so it should
869 	 * be safe to do this without locking.
870 	 */
871 	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
872 		do_unlock = 1;
873 		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
874 			if (op->nfso_posixlock == 0) {
875 				off = lop->nfslo_first;
876 				len = lop->nfslo_end - lop->nfslo_first;
877 				if (lop->nfslo_type == F_WRLCK)
878 					type = NFSV4LOCKT_WRITE;
879 				else
880 					type = NFSV4LOCKT_READ;
881 			}
882 			if (do_unlock) {
883 				trycnt = 0;
884 				do {
885 					error = nfsrpc_locku(nd, nmp, lp, off,
886 					    len, type, tcred, p, 0);
887 					if ((nd->nd_repstat == NFSERR_GRACE ||
888 					    nd->nd_repstat == NFSERR_DELAY) &&
889 					    error == 0)
890 						(void) nfs_catnap(PZERO,
891 						    (int)nd->nd_repstat,
892 						    "nfs_close");
893 				} while ((nd->nd_repstat == NFSERR_GRACE ||
894 				    nd->nd_repstat == NFSERR_DELAY) &&
895 				    error == 0 && trycnt++ < 5);
896 				if (op->nfso_posixlock)
897 					do_unlock = 0;
898 			}
899 			nfscl_freelock(lop, 0);
900 		}
901 		/*
902 		 * Do a ReleaseLockOwner.
903 		 * The lock owner name nfsl_owner may be used by other opens for
904 		 * other files but the lock_owner4 name that nfsrpc_rellockown()
905 		 * puts on the wire has the file handle for this file appended
906 		 * to it, so it can be done now.
907 		 */
908 		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
909 		    lp->nfsl_open->nfso_fhlen, tcred, p);
910 	}
911 
912 	/*
913 	 * There could be other Opens for different files on the same
914 	 * OpenOwner, so locking is required.
915 	 */
916 	own_not_null = false;
917 	if (op->nfso_own != NULL) {
918 		own_not_null = true;
919 		NFSLOCKCLSTATE();
920 		nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
921 		NFSUNLOCKCLSTATE();
922 	}
923 	do {
924 		error = nfscl_tryclose(op, tcred, nmp, p, loop_on_delayed);
925 		if (error == NFSERR_GRACE)
926 			(void) nfs_catnap(PZERO, error, "nfs_close");
927 	} while (error == NFSERR_GRACE);
928 	if (own_not_null) {
929 		NFSLOCKCLSTATE();
930 		nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
931 	}
932 
933 	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
934 		nfscl_freelockowner(lp, 0);
935 	if (freeop && error != NFSERR_DELAY)
936 		nfscl_freeopen(op, 0, true);
937 	if (own_not_null)
938 		NFSUNLOCKCLSTATE();
939 	NFSFREECRED(tcred);
940 	return (error);
941 }
942 
943 /*
944  * The actual Close RPC.
945  */
946 int
947 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
948     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
949     int syscred)
950 {
951 	u_int32_t *tl;
952 	int error;
953 
954 	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
955 	    op->nfso_fhlen, NULL, NULL, 0, 0, cred);
956 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
957 	if (NFSHASNFSV4N(nmp)) {
958 		*tl++ = 0;
959 		*tl++ = 0;
960 	} else {
961 		*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
962 		*tl++ = op->nfso_stateid.seqid;
963 	}
964 	*tl++ = op->nfso_stateid.other[0];
965 	*tl++ = op->nfso_stateid.other[1];
966 	*tl = op->nfso_stateid.other[2];
967 	if (syscred)
968 		nd->nd_flag |= ND_USEGSSNAME;
969 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
970 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
971 	if (error)
972 		return (error);
973 	if (!NFSHASNFSV4N(nmp))
974 		NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
975 	if (nd->nd_repstat == 0)
976 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
977 	error = nd->nd_repstat;
978 	if (!NFSHASNFSV4N(nmp) && error == NFSERR_STALESTATEID)
979 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
980 nfsmout:
981 	m_freem(nd->nd_mrep);
982 	return (error);
983 }
984 
985 /*
986  * V4 Open Confirm RPC.
987  */
988 int
989 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
990     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
991 {
992 	u_int32_t *tl;
993 	struct nfsrv_descript nfsd, *nd = &nfsd;
994 	struct nfsmount *nmp;
995 	int error;
996 
997 	nmp = VFSTONFS(vp->v_mount);
998 	if (NFSHASNFSV4N(nmp))
999 		return (0);		/* No confirmation for NFSv4.1. */
1000 	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
1001 	    0, 0, NULL);
1002 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
1003 	*tl++ = op->nfso_stateid.seqid;
1004 	*tl++ = op->nfso_stateid.other[0];
1005 	*tl++ = op->nfso_stateid.other[1];
1006 	*tl++ = op->nfso_stateid.other[2];
1007 	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
1008 	error = nfscl_request(nd, vp, p, cred);
1009 	if (error)
1010 		return (error);
1011 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
1012 	if (!nd->nd_repstat) {
1013 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
1014 		op->nfso_stateid.seqid = *tl++;
1015 		op->nfso_stateid.other[0] = *tl++;
1016 		op->nfso_stateid.other[1] = *tl++;
1017 		op->nfso_stateid.other[2] = *tl;
1018 	}
1019 	error = nd->nd_repstat;
1020 	if (error == NFSERR_STALESTATEID)
1021 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
1022 nfsmout:
1023 	m_freem(nd->nd_mrep);
1024 	return (error);
1025 }
1026 
1027 /*
1028  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
1029  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
1030  */
1031 int
1032 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
1033     bool *retokp, struct ucred *cred, NFSPROC_T *p)
1034 {
1035 	u_int32_t *tl;
1036 	struct nfsrv_descript nfsd;
1037 	struct nfsrv_descript *nd = &nfsd;
1038 	u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
1039 	u_short port;
1040 	int error, isinet6 = 0, callblen;
1041 	nfsquad_t confirm;
1042 	static u_int32_t rev = 0;
1043 	struct nfsclds *dsp, *odsp;
1044 	struct in6_addr a6;
1045 	struct nfsclsession *tsep;
1046 	struct rpc_reconupcall recon;
1047 	struct nfscl_reconarg *rcp;
1048 
1049 	if (nfsboottime.tv_sec == 0)
1050 		NFSSETBOOTTIME(nfsboottime);
1051 	if (NFSHASNFSV4N(nmp)) {
1052 		error = NFSERR_BADSESSION;
1053 		odsp = dsp = NULL;
1054 		if (retokp != NULL) {
1055 			NFSLOCKMNT(nmp);
1056 			odsp = TAILQ_FIRST(&nmp->nm_sess);
1057 			NFSUNLOCKMNT(nmp);
1058 		}
1059 		if (odsp != NULL) {
1060 			/*
1061 			 * When a session already exists, first try a
1062 			 * CreateSession with the extant ClientID.
1063 			 */
1064 			dsp = malloc(sizeof(struct nfsclds) +
1065 			    odsp->nfsclds_servownlen + 1, M_NFSCLDS,
1066 			    M_WAITOK | M_ZERO);
1067 			dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
1068 			dsp->nfsclds_servownlen = odsp->nfsclds_servownlen;
1069 			dsp->nfsclds_sess.nfsess_clientid =
1070 			    odsp->nfsclds_sess.nfsess_clientid;
1071 			dsp->nfsclds_sess.nfsess_sequenceid =
1072 			    odsp->nfsclds_sess.nfsess_sequenceid + 1;
1073 			dsp->nfsclds_flags = odsp->nfsclds_flags;
1074 			if (dsp->nfsclds_servownlen > 0)
1075 				memcpy(dsp->nfsclds_serverown,
1076 				    odsp->nfsclds_serverown,
1077 				    dsp->nfsclds_servownlen + 1);
1078 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
1079 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
1080 			    NULL, MTX_DEF);
1081 			nfscl_initsessionslots(&dsp->nfsclds_sess);
1082 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
1083 			    &nmp->nm_sockreq, NULL,
1084 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
1085 			NFSCL_DEBUG(1, "create session for extant "
1086 			    "ClientID=%d\n", error);
1087 			if (error != 0) {
1088 				nfscl_freenfsclds(dsp);
1089 				dsp = NULL;
1090 				/*
1091 				 * If *retokp is true, return any error other
1092 				 * than NFSERR_STALECLIENTID,
1093 				 * NFSERR_BADSESSION or NFSERR_STALEDONTRECOVER
1094 				 * so that nfscl_recover() will not loop.
1095 				 */
1096 				if (*retokp)
1097 					return (NFSERR_IO);
1098 			} else
1099 				*retokp = true;
1100 		} else if (retokp != NULL && *retokp)
1101 			return (NFSERR_IO);
1102 		if (error != 0) {
1103 			/*
1104 			 * Either there was no previous session or the
1105 			 * CreateSession attempt failed, so...
1106 			 * do an ExchangeID followed by the CreateSession.
1107 			 */
1108 			clp->nfsc_rev = rev++;
1109 			error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq, 0,
1110 			    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp,
1111 			    cred, p);
1112 			NFSCL_DEBUG(1, "aft exch=%d\n", error);
1113 			if (error == 0)
1114 				error = nfsrpc_createsession(nmp,
1115 				    &dsp->nfsclds_sess, &nmp->nm_sockreq, NULL,
1116 				    dsp->nfsclds_sess.nfsess_sequenceid, 1,
1117 				    cred, p);
1118 			NFSCL_DEBUG(1, "aft createsess=%d\n", error);
1119 		}
1120 		if (error == 0) {
1121 			/*
1122 			 * If the session supports a backchannel, set up
1123 			 * the BindConnectionToSession call in the krpc
1124 			 * so that it is done on a reconnection.
1125 			 */
1126 			if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0) {
1127 				rcp = mem_alloc(sizeof(*rcp));
1128 				rcp->minorvers = nmp->nm_minorvers;
1129 				memcpy(rcp->sessionid,
1130 				    dsp->nfsclds_sess.nfsess_sessionid,
1131 				    NFSX_V4SESSIONID);
1132 				recon.call = nfsrpc_bindconnsess;
1133 				recon.arg = rcp;
1134 				CLNT_CONTROL(nmp->nm_client, CLSET_RECONUPCALL,
1135 				    &recon);
1136 			}
1137 
1138 			NFSLOCKMNT(nmp);
1139 			/*
1140 			 * The old sessions cannot be safely free'd
1141 			 * here, since they may still be used by
1142 			 * in-progress RPCs.
1143 			 */
1144 			tsep = NULL;
1145 			if (TAILQ_FIRST(&nmp->nm_sess) != NULL) {
1146 				/*
1147 				 * Mark the old session defunct.  Needed
1148 				 * when called from nfscl_hasexpired().
1149 				 */
1150 				tsep = NFSMNT_MDSSESSION(nmp);
1151 				tsep->nfsess_defunct = 1;
1152 			}
1153 			TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
1154 			    nfsclds_list);
1155 			/*
1156 			 * Wake up RPCs waiting for a slot on the
1157 			 * old session. These will then fail with
1158 			 * NFSERR_BADSESSION and be retried with the
1159 			 * new session by nfsv4_setsequence().
1160 			 * Also wakeup() processes waiting for the
1161 			 * new session.
1162 			 */
1163 			if (tsep != NULL)
1164 				wakeup(&tsep->nfsess_slots);
1165 			wakeup(&nmp->nm_sess);
1166 			NFSUNLOCKMNT(nmp);
1167 		} else if (dsp != NULL)
1168 			nfscl_freenfsclds(dsp);
1169 		if (error == 0 && reclaim == 0) {
1170 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
1171 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
1172 			if (error == NFSERR_COMPLETEALREADY ||
1173 			    error == NFSERR_NOTSUPP)
1174 				/* Ignore this error. */
1175 				error = 0;
1176 		}
1177 		return (error);
1178 	} else if (retokp != NULL && *retokp)
1179 		return (NFSERR_IO);
1180 	clp->nfsc_rev = rev++;
1181 
1182 	/*
1183 	 * Allocate a single session structure for NFSv4.0, because some of
1184 	 * the fields are used by NFSv4.0 although it doesn't do a session.
1185 	 */
1186 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
1187 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
1188 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
1189 	NFSLOCKMNT(nmp);
1190 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
1191 	tsep = NFSMNT_MDSSESSION(nmp);
1192 	NFSUNLOCKMNT(nmp);
1193 
1194 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0,
1195 	    NULL);
1196 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1197 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
1198 	*tl = txdr_unsigned(clp->nfsc_rev);
1199 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1200 
1201 	/*
1202 	 * set up the callback address
1203 	 */
1204 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1205 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
1206 	callblen = strlen(nfsv4_callbackaddr);
1207 	if (callblen == 0)
1208 		cp = nfscl_getmyip(nmp, &a6, &isinet6);
1209 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1210 	    (callblen > 0 || cp != NULL)) {
1211 		port = htons(nfsv4_cbport);
1212 		cp2 = (u_int8_t *)&port;
1213 #ifdef INET6
1214 		if ((callblen > 0 &&
1215 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1216 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1217 
1218 			(void) nfsm_strtom(nd, "tcp6", 4);
1219 			if (callblen == 0) {
1220 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1221 				ip6add = ip6buf;
1222 			} else {
1223 				ip6add = nfsv4_callbackaddr;
1224 			}
1225 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1226 			    ip6add, cp2[0], cp2[1]);
1227 		} else
1228 #endif
1229 		{
1230 			(void) nfsm_strtom(nd, "tcp", 3);
1231 			if (callblen == 0)
1232 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1233 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1234 				    cp[2], cp[3], cp2[0], cp2[1]);
1235 			else
1236 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1237 				    "%s.%d.%d", nfsv4_callbackaddr,
1238 				    cp2[0], cp2[1]);
1239 		}
1240 		(void) nfsm_strtom(nd, addr, strlen(addr));
1241 	} else {
1242 		(void) nfsm_strtom(nd, "tcp", 3);
1243 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1244 	}
1245 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1246 	*tl = txdr_unsigned(clp->nfsc_cbident);
1247 	nd->nd_flag |= ND_USEGSSNAME;
1248 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1249 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1250 	if (error)
1251 		return (error);
1252 	if (nd->nd_repstat == 0) {
1253 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1254 	    tsep->nfsess_clientid.lval[0] = *tl++;
1255 	    tsep->nfsess_clientid.lval[1] = *tl++;
1256 	    confirm.lval[0] = *tl++;
1257 	    confirm.lval[1] = *tl;
1258 	    m_freem(nd->nd_mrep);
1259 	    nd->nd_mrep = NULL;
1260 
1261 	    /*
1262 	     * and confirm it.
1263 	     */
1264 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1265 		NULL, 0, 0, NULL);
1266 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1267 	    *tl++ = tsep->nfsess_clientid.lval[0];
1268 	    *tl++ = tsep->nfsess_clientid.lval[1];
1269 	    *tl++ = confirm.lval[0];
1270 	    *tl = confirm.lval[1];
1271 	    nd->nd_flag |= ND_USEGSSNAME;
1272 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1273 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1274 	    if (error)
1275 		return (error);
1276 	    m_freem(nd->nd_mrep);
1277 	    nd->nd_mrep = NULL;
1278 	}
1279 	error = nd->nd_repstat;
1280 nfsmout:
1281 	m_freem(nd->nd_mrep);
1282 	return (error);
1283 }
1284 
1285 /*
1286  * nfs getattr call.
1287  */
1288 int
1289 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1290     struct nfsvattr *nap)
1291 {
1292 	struct nfsrv_descript nfsd, *nd = &nfsd;
1293 	int error;
1294 	nfsattrbit_t attrbits;
1295 	struct nfsnode *np;
1296 	struct nfsmount *nmp;
1297 
1298 	nmp = VFSTONFS(vp->v_mount);
1299 	np = VTONFS(vp);
1300 	if ((nmp->nm_privflag & NFSMNTP_FAKEROOTFH) != 0 &&
1301 	    nmp->nm_fhsize == 0) {
1302 		/* Attempt to get the actual root file handle. */
1303 		error = nfsrpc_getdirpath(nmp, NFSMNT_DIRPATH(nmp), cred, p);
1304 		if (error != 0)
1305 			return (EACCES);
1306 		if (np->n_fhp->nfh_len == NFSX_FHMAX + 1)
1307 			nfscl_statfs(vp, cred, p);
1308 	}
1309 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp, cred);
1310 	if (nd->nd_flag & ND_NFSV4) {
1311 		NFSGETATTR_ATTRBIT(&attrbits);
1312 		(void) nfsrv_putattrbit(nd, &attrbits);
1313 	}
1314 	error = nfscl_request(nd, vp, p, cred);
1315 	if (error)
1316 		return (error);
1317 	if (!nd->nd_repstat)
1318 		error = nfsm_loadattr(nd, nap);
1319 	else
1320 		error = nd->nd_repstat;
1321 	m_freem(nd->nd_mrep);
1322 	return (error);
1323 }
1324 
1325 /*
1326  * nfs getattr call with non-vnode arguments.
1327  */
1328 int
1329 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1330     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1331     uint32_t *leasep)
1332 {
1333 	struct nfsrv_descript nfsd, *nd = &nfsd;
1334 	int error, vers = NFS_VER2;
1335 	nfsattrbit_t attrbits;
1336 
1337 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0,
1338 	    cred);
1339 	if (nd->nd_flag & ND_NFSV4) {
1340 		vers = NFS_VER4;
1341 		NFSGETATTR_ATTRBIT(&attrbits);
1342 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1343 		(void) nfsrv_putattrbit(nd, &attrbits);
1344 	} else if (nd->nd_flag & ND_NFSV3) {
1345 		vers = NFS_VER3;
1346 	}
1347 	if (syscred)
1348 		nd->nd_flag |= ND_USEGSSNAME;
1349 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1350 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1351 	if (error)
1352 		return (error);
1353 	if (nd->nd_repstat == 0) {
1354 		if ((nd->nd_flag & ND_NFSV4) != 0)
1355 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1356 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1357 			    NULL, NULL);
1358 		else
1359 			error = nfsm_loadattr(nd, nap);
1360 	} else
1361 		error = nd->nd_repstat;
1362 	m_freem(nd->nd_mrep);
1363 	return (error);
1364 }
1365 
1366 /*
1367  * Do an nfs setattr operation.
1368  */
1369 int
1370 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1371     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp)
1372 {
1373 	int error, expireret = 0, openerr, retrycnt;
1374 	u_int32_t clidrev = 0, mode;
1375 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1376 	struct nfsfh *nfhp;
1377 	nfsv4stateid_t stateid;
1378 	void *lckp;
1379 
1380 	if (nmp->nm_clp != NULL)
1381 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1382 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1383 		mode = NFSV4OPEN_ACCESSWRITE;
1384 	else
1385 		mode = NFSV4OPEN_ACCESSREAD;
1386 	retrycnt = 0;
1387 	do {
1388 		lckp = NULL;
1389 		openerr = 1;
1390 		if (NFSHASNFSV4(nmp)) {
1391 			nfhp = VTONFS(vp)->n_fhp;
1392 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1393 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1394 			if (error && vp->v_type == VREG &&
1395 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1396 			     nfstest_openallsetattr)) {
1397 				/*
1398 				 * No Open stateid, so try and open the file
1399 				 * now.
1400 				 */
1401 				if (mode == NFSV4OPEN_ACCESSWRITE)
1402 					openerr = nfsrpc_open(vp, FWRITE, cred,
1403 					    p);
1404 				else
1405 					openerr = nfsrpc_open(vp, FREAD, cred,
1406 					    p);
1407 				if (!openerr)
1408 					(void) nfscl_getstateid(vp,
1409 					    nfhp->nfh_fh, nfhp->nfh_len,
1410 					    mode, 0, cred, p, &stateid, &lckp);
1411 			}
1412 		}
1413 		if (vap != NULL)
1414 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1415 			    rnap, attrflagp);
1416 		else
1417 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid);
1418 		if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1419 			NFSLOCKMNT(nmp);
1420 			nmp->nm_state |= NFSSTA_OPENMODE;
1421 			NFSUNLOCKMNT(nmp);
1422 		}
1423 		if (error == NFSERR_STALESTATEID)
1424 			nfscl_initiate_recovery(nmp->nm_clp);
1425 		if (lckp != NULL)
1426 			nfscl_lockderef(lckp);
1427 		if (!openerr)
1428 			(void) nfsrpc_close(vp, 0, p);
1429 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1430 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1431 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1432 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1433 		} else if ((error == NFSERR_EXPIRED ||
1434 		    ((!NFSHASINT(nmp) || !NFSHASNFSV4N(nmp)) &&
1435 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
1436 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1437 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp) &&
1438 		    NFSHASNFSV4N(nmp)) {
1439 			error = EIO;
1440 		}
1441 		retrycnt++;
1442 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1443 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1444 	    error == NFSERR_BADSESSION ||
1445 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1446 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1447 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1448 	    (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1449 	     retrycnt < 4));
1450 	if (error && retrycnt >= 4)
1451 		error = EIO;
1452 	return (error);
1453 }
1454 
1455 static int
1456 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1457     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1458     struct nfsvattr *rnap, int *attrflagp)
1459 {
1460 	u_int32_t *tl;
1461 	struct nfsrv_descript nfsd, *nd = &nfsd;
1462 	int error;
1463 	nfsattrbit_t attrbits;
1464 
1465 	*attrflagp = 0;
1466 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp, cred);
1467 	if (nd->nd_flag & ND_NFSV4)
1468 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1469 	vap->va_type = vp->v_type;
1470 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1471 	if (nd->nd_flag & ND_NFSV3) {
1472 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1473 		*tl = newnfs_false;
1474 	} else if (nd->nd_flag & ND_NFSV4) {
1475 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1476 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1477 		NFSGETATTR_ATTRBIT(&attrbits);
1478 		(void) nfsrv_putattrbit(nd, &attrbits);
1479 	}
1480 	error = nfscl_request(nd, vp, p, cred);
1481 	if (error)
1482 		return (error);
1483 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1484 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, NULL);
1485 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1486 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1487 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1488 		error = nfscl_postop_attr(nd, rnap, attrflagp);
1489 	m_freem(nd->nd_mrep);
1490 	if (nd->nd_repstat && !error)
1491 		error = nd->nd_repstat;
1492 	return (error);
1493 }
1494 
1495 /*
1496  * nfs lookup rpc
1497  */
1498 int
1499 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1500     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1501     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, uint32_t openmode)
1502 {
1503 	uint32_t deleg, rflags, *tl;
1504 	struct nfsrv_descript nfsd, *nd = &nfsd;
1505 	struct nfsmount *nmp;
1506 	struct nfsnode *np;
1507 	struct nfsfh *nfhp;
1508 	nfsattrbit_t attrbits;
1509 	int error = 0, lookupp = 0, newone, ret, retop;
1510 	uint8_t own[NFSV4CL_LOCKNAMELEN];
1511 	struct nfsclopen *op;
1512 	struct nfscldeleg *ndp;
1513 	nfsv4stateid_t stateid;
1514 
1515 	*attrflagp = 0;
1516 	*dattrflagp = 0;
1517 	if (dvp->v_type != VDIR)
1518 		return (ENOTDIR);
1519 	nmp = VFSTONFS(dvp->v_mount);
1520 	if (len > NFS_MAXNAMLEN)
1521 		return (ENAMETOOLONG);
1522 	if (NFSHASNFSV4(nmp) && len == 1 &&
1523 		name[0] == '.') {
1524 		/*
1525 		 * Just return the current dir's fh.
1526 		 */
1527 		np = VTONFS(dvp);
1528 		nfhp = malloc(sizeof (struct nfsfh) +
1529 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1530 		nfhp->nfh_len = np->n_fhp->nfh_len;
1531 		NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1532 		*nfhpp = nfhp;
1533 		return (0);
1534 	}
1535 	if (NFSHASNFSV4(nmp) && len == 2 &&
1536 		name[0] == '.' && name[1] == '.') {
1537 		lookupp = 1;
1538 		openmode = 0;
1539 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp, cred);
1540 	} else if (openmode != 0) {
1541 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPOPEN, dvp, cred);
1542 		nfsm_strtom(nd, name, len);
1543 	} else {
1544 		NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp, cred);
1545 		(void) nfsm_strtom(nd, name, len);
1546 	}
1547 	if (nd->nd_flag & ND_NFSV4) {
1548 		NFSGETATTR_ATTRBIT(&attrbits);
1549 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1550 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1551 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1552 		(void) nfsrv_putattrbit(nd, &attrbits);
1553 		if (openmode != 0) {
1554 			/* Test for a VREG file. */
1555 			NFSZERO_ATTRBIT(&attrbits);
1556 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
1557 			NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
1558 			*tl = txdr_unsigned(NFSV4OP_VERIFY);
1559 			nfsrv_putattrbit(nd, &attrbits);
1560 			NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1561 			*tl++ = txdr_unsigned(NFSX_UNSIGNED);
1562 			*tl = vtonfsv34_type(VREG);
1563 
1564 			/* Attempt the Open for VREG. */
1565 			nfscl_filllockowner(NULL, own, F_POSIX);
1566 			NFSM_BUILD(tl, uint32_t *, 6 * NFSX_UNSIGNED);
1567 			*tl++ = txdr_unsigned(NFSV4OP_OPEN);
1568 			*tl++ = 0;		/* seqid, ignored. */
1569 			*tl++ = txdr_unsigned(openmode | NFSV4OPEN_WANTNODELEG);
1570 			*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
1571 			*tl++ = 0;		/* ClientID, ignored. */
1572 			*tl = 0;
1573 			nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN);
1574 			NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1575 			*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
1576 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMFH);
1577 		}
1578 	}
1579 	error = nfscl_request(nd, dvp, p, cred);
1580 	if (error)
1581 		return (error);
1582 	ndp = NULL;
1583 	if (nd->nd_repstat) {
1584 		/*
1585 		 * When an NFSv4 Lookupp returns ENOENT, it means that
1586 		 * the lookup is at the root of an fs, so return this dir.
1587 		 */
1588 		if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1589 		    np = VTONFS(dvp);
1590 		    nfhp = malloc(sizeof (struct nfsfh) +
1591 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1592 		    nfhp->nfh_len = np->n_fhp->nfh_len;
1593 		    NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1594 		    *nfhpp = nfhp;
1595 		    m_freem(nd->nd_mrep);
1596 		    return (0);
1597 		}
1598 		if (nd->nd_flag & ND_NFSV3)
1599 		    error = nfscl_postop_attr(nd, dnap, dattrflagp);
1600 		else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1601 		    ND_NFSV4) {
1602 			/* Load the directory attributes. */
1603 			error = nfsm_loadattr(nd, dnap);
1604 			if (error != 0)
1605 				goto nfsmout;
1606 			*dattrflagp = 1;
1607 		}
1608 		/* Check Lookup operation reply status. */
1609 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1610 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1611 			if (*++tl != 0)
1612 				goto nfsmout;
1613 		}
1614 		/* Look for GetFH reply. */
1615 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1616 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1617 			if (*++tl != 0)
1618 				goto nfsmout;
1619 			error = nfsm_getfh(nd, nfhpp);
1620 			if (error)
1621 				goto nfsmout;
1622 		}
1623 		/* Look for Getattr reply. */
1624 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1625 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1626 			if (*++tl != 0)
1627 				goto nfsmout;
1628 			error = nfsm_loadattr(nd, nap);
1629 			if (error == 0) {
1630 				/*
1631 				 * We have now successfully completed the
1632 				 * lookup, so set nd_repstat to 0.
1633 				 */
1634 				nd->nd_repstat = 0;
1635 				*attrflagp = 1;
1636 			}
1637 		}
1638 		goto nfsmout;
1639 	}
1640 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1641 		/* Load the directory attributes. */
1642 		error = nfsm_loadattr(nd, dnap);
1643 		if (error != 0)
1644 			goto nfsmout;
1645 		*dattrflagp = 1;
1646 		/* Skip over the Lookup and GetFH operation status values. */
1647 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1648 	}
1649 	error = nfsm_getfh(nd, nfhpp);
1650 	if (error)
1651 		goto nfsmout;
1652 
1653 	error = nfscl_postop_attr(nd, nap, attrflagp);
1654 	if (openmode != 0 && error == 0) {
1655 		NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID +
1656 		    10 * NFSX_UNSIGNED);
1657 		tl += 4;	/* Skip over Verify+Open status. */
1658 		stateid.seqid = *tl++;
1659 		stateid.other[0] = *tl++;
1660 		stateid.other[1] = *tl++;
1661 		stateid.other[2] = *tl;
1662 		rflags = fxdr_unsigned(uint32_t, *(tl + 6));
1663 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1664 		if (error != 0)
1665 			goto nfsmout;
1666 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
1667 		deleg = fxdr_unsigned(uint32_t, *tl);
1668 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
1669 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
1670 			/*
1671 			 * Just need to fill in the fields used by
1672 			 * nfscl_trydelegreturn().
1673 			 * Mark the mount point as acquiring
1674 			 * delegations, so NFSPROC_LOOKUPOPEN will
1675 			 * no longer be done.
1676 			 */
1677 			NFSLOCKMNT(nmp);
1678 			nmp->nm_privflag |= NFSMNTP_DELEGISSUED;
1679 			NFSUNLOCKMNT(nmp);
1680 			ndp = malloc(sizeof(struct nfscldeleg) +
1681 			    (*nfhpp)->nfh_len, M_NFSCLDELEG, M_WAITOK);
1682 			ndp->nfsdl_fhlen = (*nfhpp)->nfh_len;
1683 			NFSBCOPY((*nfhpp)->nfh_fh, ndp->nfsdl_fh,
1684 			    ndp->nfsdl_fhlen);
1685 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
1686 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
1687 			ndp->nfsdl_stateid.seqid = *tl++;
1688 			ndp->nfsdl_stateid.other[0] = *tl++;
1689 			ndp->nfsdl_stateid.other[1] = *tl++;
1690 			ndp->nfsdl_stateid.other[2] = *tl++;
1691 		} else if (deleg == NFSV4OPEN_DELEGATENONEEXT &&
1692 		    NFSHASNFSV4N(nmp)) {
1693 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
1694 			deleg = fxdr_unsigned(uint32_t, *tl);
1695 			if (deleg == NFSV4OPEN_CONTENTION ||
1696 			    deleg == NFSV4OPEN_RESOURCE)
1697 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
1698 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
1699 			error = NFSERR_BADXDR;
1700 			goto nfsmout;
1701 		}
1702 		ret = nfscl_open(dvp, (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len,
1703 		    openmode, 0, cred, p, NULL, &op, &newone, &retop, 1, true);
1704 		if (ret != 0)
1705 			goto nfsmout;
1706 		if (newone != 0) {
1707 			op->nfso_stateid.seqid = stateid.seqid;
1708 			op->nfso_stateid.other[0] = stateid.other[0];
1709 			op->nfso_stateid.other[1] = stateid.other[1];
1710 			op->nfso_stateid.other[2] = stateid.other[2];
1711 			op->nfso_mode = openmode;
1712 		} else {
1713 			op->nfso_stateid.seqid = stateid.seqid;
1714 			if (retop == NFSCLOPEN_DOOPEN)
1715 				op->nfso_mode |= openmode;
1716 		}
1717 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
1718 		    nfscl_assumeposixlocks)
1719 			op->nfso_posixlock = 1;
1720 		else
1721 			op->nfso_posixlock = 0;
1722 		nfscl_openrelease(nmp, op, 0, 0);
1723 		if (ndp != NULL) {
1724 			/*
1725 			 * Since we do not have the vnode, we
1726 			 * cannot invalidate cached attributes.
1727 			 * Just return the delegation.
1728 			 */
1729 			nfscl_trydelegreturn(ndp, cred, nmp, p);
1730 		}
1731 	}
1732 	if ((nd->nd_flag & ND_NFSV3) && !error)
1733 		error = nfscl_postop_attr(nd, dnap, dattrflagp);
1734 nfsmout:
1735 	m_freem(nd->nd_mrep);
1736 	if (!error && nd->nd_repstat)
1737 		error = nd->nd_repstat;
1738 	free(ndp, M_NFSCLDELEG);
1739 	return (error);
1740 }
1741 
1742 /*
1743  * Do a readlink rpc.
1744  */
1745 int
1746 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1747     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
1748 {
1749 	u_int32_t *tl;
1750 	struct nfsrv_descript nfsd, *nd = &nfsd;
1751 	struct nfsnode *np = VTONFS(vp);
1752 	nfsattrbit_t attrbits;
1753 	int error, len, cangetattr = 1;
1754 
1755 	*attrflagp = 0;
1756 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp, cred);
1757 	if (nd->nd_flag & ND_NFSV4) {
1758 		/*
1759 		 * And do a Getattr op.
1760 		 */
1761 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1762 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1763 		NFSGETATTR_ATTRBIT(&attrbits);
1764 		(void) nfsrv_putattrbit(nd, &attrbits);
1765 	}
1766 	error = nfscl_request(nd, vp, p, cred);
1767 	if (error)
1768 		return (error);
1769 	if (nd->nd_flag & ND_NFSV3)
1770 		error = nfscl_postop_attr(nd, nap, attrflagp);
1771 	if (!nd->nd_repstat && !error) {
1772 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1773 		/*
1774 		 * This seems weird to me, but must have been added to
1775 		 * FreeBSD for some reason. The only thing I can think of
1776 		 * is that there was/is some server that replies with
1777 		 * more link data than it should?
1778 		 */
1779 		if (len == NFS_MAXPATHLEN) {
1780 			NFSLOCKNODE(np);
1781 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1782 				len = np->n_size;
1783 				cangetattr = 0;
1784 			}
1785 			NFSUNLOCKNODE(np);
1786 		}
1787 		error = nfsm_mbufuio(nd, uiop, len);
1788 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1789 			error = nfscl_postop_attr(nd, nap, attrflagp);
1790 	}
1791 	if (nd->nd_repstat && !error)
1792 		error = nd->nd_repstat;
1793 nfsmout:
1794 	m_freem(nd->nd_mrep);
1795 	return (error);
1796 }
1797 
1798 /*
1799  * Read operation.
1800  */
1801 int
1802 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1803     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
1804 {
1805 	int error, expireret = 0, retrycnt;
1806 	u_int32_t clidrev = 0;
1807 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1808 	struct nfsnode *np = VTONFS(vp);
1809 	struct ucred *newcred;
1810 	struct nfsfh *nfhp = NULL;
1811 	nfsv4stateid_t stateid;
1812 	void *lckp;
1813 
1814 	if (nmp->nm_clp != NULL)
1815 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1816 	newcred = cred;
1817 	if (NFSHASNFSV4(nmp)) {
1818 		nfhp = np->n_fhp;
1819 		newcred = NFSNEWCRED(cred);
1820 	}
1821 	retrycnt = 0;
1822 	do {
1823 		lckp = NULL;
1824 		if (NFSHASNFSV4(nmp))
1825 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1826 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1827 			    &lckp);
1828 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1829 		    attrflagp);
1830 		if (error == NFSERR_OPENMODE) {
1831 			NFSLOCKMNT(nmp);
1832 			nmp->nm_state |= NFSSTA_OPENMODE;
1833 			NFSUNLOCKMNT(nmp);
1834 		}
1835 		if (error == NFSERR_STALESTATEID)
1836 			nfscl_initiate_recovery(nmp->nm_clp);
1837 		if (lckp != NULL)
1838 			nfscl_lockderef(lckp);
1839 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1840 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1841 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1842 			(void) nfs_catnap(PZERO, error, "nfs_read");
1843 		} else if ((error == NFSERR_EXPIRED ||
1844 		    ((!NFSHASINT(nmp) || !NFSHASNFSV4N(nmp)) &&
1845 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
1846 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1847 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp) &&
1848 		    NFSHASNFSV4N(nmp)) {
1849 			error = EIO;
1850 		}
1851 		retrycnt++;
1852 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1853 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1854 	    error == NFSERR_BADSESSION ||
1855 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1856 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1857 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1858 	    (error == NFSERR_OPENMODE && retrycnt < 4));
1859 	if (error && retrycnt >= 4)
1860 		error = EIO;
1861 	if (NFSHASNFSV4(nmp))
1862 		NFSFREECRED(newcred);
1863 	return (error);
1864 }
1865 
1866 /*
1867  * The actual read RPC.
1868  */
1869 static int
1870 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1871     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1872     int *attrflagp)
1873 {
1874 	u_int32_t *tl;
1875 	int error = 0, len, retlen, tsiz, eof = 0;
1876 	struct nfsrv_descript nfsd;
1877 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1878 	struct nfsrv_descript *nd = &nfsd;
1879 	int rsize;
1880 	off_t tmp_off;
1881 
1882 	*attrflagp = 0;
1883 	tsiz = uiop->uio_resid;
1884 	tmp_off = uiop->uio_offset + tsiz;
1885 	NFSLOCKMNT(nmp);
1886 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1887 		NFSUNLOCKMNT(nmp);
1888 		return (EFBIG);
1889 	}
1890 	rsize = nmp->nm_rsize;
1891 	NFSUNLOCKMNT(nmp);
1892 	nd->nd_mrep = NULL;
1893 	while (tsiz > 0) {
1894 		*attrflagp = 0;
1895 		len = (tsiz > rsize) ? rsize : tsiz;
1896 		NFSCL_REQSTART(nd, NFSPROC_READ, vp, cred);
1897 		if (nd->nd_flag & ND_NFSV4)
1898 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1899 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1900 		if (nd->nd_flag & ND_NFSV2) {
1901 			*tl++ = txdr_unsigned(uiop->uio_offset);
1902 			*tl++ = txdr_unsigned(len);
1903 			*tl = 0;
1904 		} else {
1905 			txdr_hyper(uiop->uio_offset, tl);
1906 			*(tl + 2) = txdr_unsigned(len);
1907 		}
1908 		/*
1909 		 * Since I can't do a Getattr for NFSv4 for Write, there
1910 		 * doesn't seem any point in doing one here, either.
1911 		 * (See the comment in nfsrpc_writerpc() for more info.)
1912 		 */
1913 		error = nfscl_request(nd, vp, p, cred);
1914 		if (error)
1915 			return (error);
1916 		if (nd->nd_flag & ND_NFSV3) {
1917 			error = nfscl_postop_attr(nd, nap, attrflagp);
1918 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1919 			error = nfsm_loadattr(nd, nap);
1920 			if (!error)
1921 				*attrflagp = 1;
1922 		}
1923 		if (nd->nd_repstat || error) {
1924 			if (!error)
1925 				error = nd->nd_repstat;
1926 			goto nfsmout;
1927 		}
1928 		if (nd->nd_flag & ND_NFSV3) {
1929 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1930 			eof = fxdr_unsigned(int, *(tl + 1));
1931 		} else if (nd->nd_flag & ND_NFSV4) {
1932 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1933 			eof = fxdr_unsigned(int, *tl);
1934 		}
1935 		NFSM_STRSIZ(retlen, len);
1936 		error = nfsm_mbufuio(nd, uiop, retlen);
1937 		if (error)
1938 			goto nfsmout;
1939 		m_freem(nd->nd_mrep);
1940 		nd->nd_mrep = NULL;
1941 		tsiz -= retlen;
1942 		if (!(nd->nd_flag & ND_NFSV2)) {
1943 			if (eof || retlen == 0)
1944 				tsiz = 0;
1945 		} else if (retlen < len)
1946 			tsiz = 0;
1947 	}
1948 	return (0);
1949 nfsmout:
1950 	if (nd->nd_mrep != NULL)
1951 		m_freem(nd->nd_mrep);
1952 	return (error);
1953 }
1954 
1955 /*
1956  * nfs write operation
1957  * When called_from_strategy != 0, it should return EIO for an error that
1958  * indicates recovery is in progress, so that the buffer will be left
1959  * dirty and be written back to the server later. If it loops around,
1960  * the recovery thread could get stuck waiting for the buffer and recovery
1961  * will then deadlock.
1962  */
1963 int
1964 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1965     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1966     int called_from_strategy, int ioflag)
1967 {
1968 	int error, expireret = 0, retrycnt, nostateid;
1969 	u_int32_t clidrev = 0;
1970 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1971 	struct nfsnode *np = VTONFS(vp);
1972 	struct ucred *newcred;
1973 	struct nfsfh *nfhp = NULL;
1974 	nfsv4stateid_t stateid;
1975 	void *lckp;
1976 
1977 	KASSERT(*must_commit >= 0 && *must_commit <= 2,
1978 	    ("nfsrpc_write: must_commit out of range=%d", *must_commit));
1979 	if (nmp->nm_clp != NULL)
1980 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1981 	newcred = cred;
1982 	if (NFSHASNFSV4(nmp)) {
1983 		newcred = NFSNEWCRED(cred);
1984 		nfhp = np->n_fhp;
1985 	}
1986 	retrycnt = 0;
1987 	do {
1988 		lckp = NULL;
1989 		nostateid = 0;
1990 		if (NFSHASNFSV4(nmp)) {
1991 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1992 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1993 			    &lckp);
1994 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1995 			    stateid.other[2] == 0) {
1996 				nostateid = 1;
1997 				NFSCL_DEBUG(1, "stateid0 in write\n");
1998 			}
1999 		}
2000 
2001 		/*
2002 		 * If there is no stateid for NFSv4, it means this is an
2003 		 * extraneous write after close. Basically a poorly
2004 		 * implemented buffer cache. Just don't do the write.
2005 		 */
2006 		if (nostateid)
2007 			error = 0;
2008 		else
2009 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
2010 			    newcred, &stateid, p, nap, attrflagp, ioflag);
2011 		if (error == NFSERR_STALESTATEID)
2012 			nfscl_initiate_recovery(nmp->nm_clp);
2013 		if (lckp != NULL)
2014 			nfscl_lockderef(lckp);
2015 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2016 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2017 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
2018 			(void) nfs_catnap(PZERO, error, "nfs_write");
2019 		} else if ((error == NFSERR_EXPIRED ||
2020 		    ((!NFSHASINT(nmp) || !NFSHASNFSV4N(nmp)) &&
2021 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
2022 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2023 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp) &&
2024 		    NFSHASNFSV4N(nmp)) {
2025 			error = EIO;
2026 		}
2027 		retrycnt++;
2028 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
2029 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
2030 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
2031 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
2032 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2033 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
2034 	if (error != 0 && (retrycnt >= 4 ||
2035 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
2036 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
2037 		error = EIO;
2038 	if (NFSHASNFSV4(nmp))
2039 		NFSFREECRED(newcred);
2040 	return (error);
2041 }
2042 
2043 /*
2044  * The actual write RPC.
2045  */
2046 static int
2047 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
2048     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
2049     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, int ioflag)
2050 {
2051 	u_int32_t *tl;
2052 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
2053 	struct nfsnode *np = VTONFS(vp);
2054 	int error = 0, len, rlen, commit, committed = NFSWRITE_FILESYNC;
2055 	int wccflag = 0;
2056 	int32_t backup;
2057 	struct nfsrv_descript *nd;
2058 	nfsattrbit_t attrbits;
2059 	uint64_t tmp_off;
2060 	ssize_t tsiz, wsize;
2061 	bool do_append;
2062 
2063 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
2064 	*attrflagp = 0;
2065 	tsiz = uiop->uio_resid;
2066 	tmp_off = uiop->uio_offset + tsiz;
2067 	NFSLOCKMNT(nmp);
2068 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
2069 		NFSUNLOCKMNT(nmp);
2070 		return (EFBIG);
2071 	}
2072 	wsize = nmp->nm_wsize;
2073 	do_append = false;
2074 	if ((ioflag & IO_APPEND) != 0 && NFSHASNFSV4(nmp) && !NFSHASPNFS(nmp))
2075 		do_append = true;
2076 	NFSUNLOCKMNT(nmp);
2077 	nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK);
2078 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
2079 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
2080 	while (tsiz > 0) {
2081 		*attrflagp = 0;
2082 		len = (tsiz > wsize) ? wsize : tsiz;
2083 		if (do_append)
2084 			NFSCL_REQSTART(nd, NFSPROC_APPENDWRITE, vp, cred);
2085 		else
2086 			NFSCL_REQSTART(nd, NFSPROC_WRITE, vp, cred);
2087 		if (nd->nd_flag & ND_NFSV4) {
2088 			if (do_append) {
2089 				NFSZERO_ATTRBIT(&attrbits);
2090 				NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
2091 				nfsrv_putattrbit(nd, &attrbits);
2092 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED +
2093 				    NFSX_HYPER);
2094 				*tl++ = txdr_unsigned(NFSX_HYPER);
2095 				txdr_hyper(uiop->uio_offset, tl); tl += 2;
2096 				*tl = txdr_unsigned(NFSV4OP_WRITE);
2097 			}
2098 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
2099 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
2100 			txdr_hyper(uiop->uio_offset, tl);
2101 			tl += 2;
2102 			*tl++ = txdr_unsigned(*iomode);
2103 			*tl = txdr_unsigned(len);
2104 		} else if (nd->nd_flag & ND_NFSV3) {
2105 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
2106 			txdr_hyper(uiop->uio_offset, tl);
2107 			tl += 2;
2108 			*tl++ = txdr_unsigned(len);
2109 			*tl++ = txdr_unsigned(*iomode);
2110 			*tl = txdr_unsigned(len);
2111 		} else {
2112 			u_int32_t x;
2113 
2114 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2115 			/*
2116 			 * Not sure why someone changed this, since the
2117 			 * RFC clearly states that "beginoffset" and
2118 			 * "totalcount" are ignored, but it wouldn't
2119 			 * surprise me if there's a busted server out there.
2120 			 */
2121 			/* Set both "begin" and "current" to non-garbage. */
2122 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
2123 			*tl++ = x;      /* "begin offset" */
2124 			*tl++ = x;      /* "current offset" */
2125 			x = txdr_unsigned(len);
2126 			*tl++ = x;      /* total to this offset */
2127 			*tl = x;        /* size of this write */
2128 		}
2129 		error = nfsm_uiombuf(nd, uiop, len);
2130 		if (error != 0) {
2131 			m_freem(nd->nd_mreq);
2132 			free(nd, M_TEMP);
2133 			return (error);
2134 		}
2135 		/*
2136 		 * Although it is tempting to do a normal Getattr Op in the
2137 		 * NFSv4 compound, the result can be a nearly hung client
2138 		 * system if the Getattr asks for Owner and/or OwnerGroup.
2139 		 * It occurs when the client can't map either the Owner or
2140 		 * Owner_group name in the Getattr reply to a uid/gid. When
2141 		 * there is a cache miss, the kernel does an upcall to the
2142 		 * nfsuserd. Then, it can try and read the local /etc/passwd
2143 		 * or /etc/group file. It can then block in getnewbuf(),
2144 		 * waiting for dirty writes to be pushed to the NFS server.
2145 		 * The only reason this doesn't result in a complete
2146 		 * deadlock, is that the upcall times out and allows
2147 		 * the write to complete. However, progress is so slow
2148 		 * that it might just as well be deadlocked.
2149 		 * As such, we get the rest of the attributes, but not
2150 		 * Owner or Owner_group.
2151 		 * nb: nfscl_loadattrcache() needs to be told that these
2152 		 *     partial attributes from a write rpc are being
2153 		 *     passed in, via a argument flag.
2154 		 */
2155 		if (nd->nd_flag & ND_NFSV4) {
2156 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
2157 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2158 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2159 			(void) nfsrv_putattrbit(nd, &attrbits);
2160 		}
2161 		error = nfscl_request(nd, vp, p, cred);
2162 		if (error) {
2163 			free(nd, M_TEMP);
2164 			return (error);
2165 		}
2166 		if (nd->nd_repstat) {
2167 			/*
2168 			 * In case the rpc gets retried, roll
2169 			 * the uio fields changed by nfsm_uiombuf()
2170 			 * back.
2171 			 */
2172 			uiop->uio_offset -= len;
2173 			uiop->uio_resid += len;
2174 			uiop->uio_iov->iov_base =
2175 			    (char *)uiop->uio_iov->iov_base - len;
2176 			uiop->uio_iov->iov_len += len;
2177 		}
2178 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2179 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
2180 			    &wccflag, &tmp_off);
2181 			if (error)
2182 				goto nfsmout;
2183 		}
2184 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2185 		    (ND_NFSV4 | ND_NOMOREDATA) &&
2186 		    nd->nd_repstat == NFSERR_NOTSAME && do_append) {
2187 			/*
2188 			 * Verify of the file's size failed, so redo the
2189 			 * write using the file's size as returned in
2190 			 * the wcc attributes.
2191 			 */
2192 			if (tmp_off + tsiz <= nmp->nm_maxfilesize) {
2193 				do_append = false;
2194 				uiop->uio_offset = tmp_off;
2195 				m_freem(nd->nd_mrep);
2196 				nd->nd_mrep = NULL;
2197 				continue;
2198 			} else
2199 				nd->nd_repstat = EFBIG;
2200 		}
2201 		if (!nd->nd_repstat) {
2202 			if (do_append) {
2203 				/* Strip off the Write reply status. */
2204 				do_append = false;
2205 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
2206 			}
2207 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2208 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
2209 					+ NFSX_VERF);
2210 				rlen = fxdr_unsigned(int, *tl++);
2211 				if (rlen == 0) {
2212 					error = NFSERR_IO;
2213 					goto nfsmout;
2214 				} else if (rlen < len) {
2215 					backup = len - rlen;
2216 					uiop->uio_iov->iov_base =
2217 					    (char *)uiop->uio_iov->iov_base -
2218 					    backup;
2219 					uiop->uio_iov->iov_len += backup;
2220 					uiop->uio_offset -= backup;
2221 					uiop->uio_resid += backup;
2222 					len = rlen;
2223 				}
2224 				commit = fxdr_unsigned(int, *tl++);
2225 
2226 				/*
2227 				 * Return the lowest commitment level
2228 				 * obtained by any of the RPCs.
2229 				 */
2230 				if (committed == NFSWRITE_FILESYNC)
2231 					committed = commit;
2232 				else if (committed == NFSWRITE_DATASYNC &&
2233 					commit == NFSWRITE_UNSTABLE)
2234 					committed = commit;
2235 				NFSLOCKMNT(nmp);
2236 				if (!NFSHASWRITEVERF(nmp)) {
2237 					NFSBCOPY((caddr_t)tl,
2238 					    (caddr_t)&nmp->nm_verf[0],
2239 					    NFSX_VERF);
2240 					NFSSETWRITEVERF(nmp);
2241 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
2242 				    NFSX_VERF) && *must_commit != 2) {
2243 					*must_commit = 1;
2244 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
2245 				}
2246 				NFSUNLOCKMNT(nmp);
2247 			}
2248 			if (nd->nd_flag & ND_NFSV4)
2249 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2250 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
2251 				error = nfsm_loadattr(nd, nap);
2252 				if (!error)
2253 					*attrflagp = NFS_LATTR_NOSHRINK;
2254 			}
2255 		} else {
2256 			error = nd->nd_repstat;
2257 		}
2258 		if (error)
2259 			goto nfsmout;
2260 		NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
2261 		m_freem(nd->nd_mrep);
2262 		nd->nd_mrep = NULL;
2263 		tsiz -= len;
2264 	}
2265 nfsmout:
2266 	if (nd->nd_mrep != NULL)
2267 		m_freem(nd->nd_mrep);
2268 	*iomode = committed;
2269 	if (nd->nd_repstat && !error)
2270 		error = nd->nd_repstat;
2271 	free(nd, M_TEMP);
2272 	return (error);
2273 }
2274 
2275 /*
2276  * Do an nfs deallocate operation.
2277  */
2278 int
2279 nfsrpc_deallocate(vnode_t vp, off_t offs, off_t len, struct nfsvattr *nap,
2280     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
2281 {
2282 	int error, expireret = 0, openerr, retrycnt;
2283 	uint32_t clidrev = 0;
2284 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
2285 	struct nfsfh *nfhp;
2286 	nfsv4stateid_t stateid;
2287 	void *lckp;
2288 
2289 	if (nmp->nm_clp != NULL)
2290 		clidrev = nmp->nm_clp->nfsc_clientidrev;
2291 	retrycnt = 0;
2292 	do {
2293 		lckp = NULL;
2294 		openerr = 1;
2295 		nfhp = VTONFS(vp)->n_fhp;
2296 		error = nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
2297 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
2298 		if (error != 0) {
2299 			/*
2300 			 * No Open stateid, so try and open the file
2301 			 * now.
2302 			 */
2303 			openerr = nfsrpc_open(vp, FWRITE, cred, p);
2304 			if (openerr == 0)
2305 				nfscl_getstateid(vp, nfhp->nfh_fh,
2306 				    nfhp->nfh_len, NFSV4OPEN_ACCESSWRITE, 0,
2307 				    cred, p, &stateid, &lckp);
2308 		}
2309 		error = nfsrpc_deallocaterpc(vp, offs, len, &stateid, nap,
2310 		    attrflagp, cred, p);
2311 		if (error == NFSERR_STALESTATEID)
2312 			nfscl_initiate_recovery(nmp->nm_clp);
2313 		if (lckp != NULL)
2314 			nfscl_lockderef(lckp);
2315 		if (openerr == 0)
2316 			nfsrpc_close(vp, 0, p);
2317 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2318 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2319 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
2320 			(void) nfs_catnap(PZERO, error, "nfs_deallocate");
2321 		} else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) &&
2322 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
2323 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2324 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) {
2325 			error = EIO;
2326 		}
2327 		retrycnt++;
2328 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2329 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2330 	    error == NFSERR_BADSESSION ||
2331 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
2332 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2333 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
2334 	if (error && retrycnt >= 4)
2335 		error = EIO;
2336 	return (error);
2337 }
2338 
2339 /*
2340  * The actual deallocate RPC.
2341  */
2342 static int
2343 nfsrpc_deallocaterpc(vnode_t vp, off_t offs, off_t len,
2344     nfsv4stateid_t *stateidp, struct nfsvattr *nap, int *attrflagp,
2345     struct ucred *cred, NFSPROC_T *p)
2346 {
2347 	uint32_t *tl;
2348 	struct nfsnode *np = VTONFS(vp);
2349 	int error, wccflag;
2350 	struct nfsrv_descript nfsd;
2351 	struct nfsrv_descript *nd = &nfsd;
2352 	nfsattrbit_t attrbits;
2353 
2354 	*attrflagp = 0;
2355 	NFSCL_REQSTART(nd, NFSPROC_DEALLOCATE, vp, cred);
2356 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
2357 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
2358 	txdr_hyper(offs, tl);
2359 	tl += 2;
2360 	txdr_hyper(len, tl);
2361 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
2362 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
2363 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2364 	nfsrv_putattrbit(nd, &attrbits);
2365 	error = nfscl_request(nd, vp, p, cred);
2366 	if (error != 0)
2367 		return (error);
2368 	wccflag = 0;
2369 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, &wccflag, NULL);
2370 	if (error != 0)
2371 		goto nfsmout;
2372 	if (nd->nd_repstat == 0) {
2373 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
2374 		error = nfsm_loadattr(nd, nap);
2375 		if (error != 0)
2376 			goto nfsmout;
2377 		*attrflagp = NFS_LATTR_NOSHRINK;
2378 	}
2379 	NFSWRITERPC_SETTIME(wccflag, np, nap, 1);
2380 nfsmout:
2381 	m_freem(nd->nd_mrep);
2382 	if (nd->nd_repstat != 0 && error == 0)
2383 		error = nd->nd_repstat;
2384 	return (error);
2385 }
2386 
2387 /*
2388  * nfs mknod rpc
2389  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
2390  * mode set to specify the file type and the size field for rdev.
2391  */
2392 int
2393 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2394     u_int32_t rdev, __enum_uint8(vtype) vtyp, struct ucred *cred, NFSPROC_T *p,
2395     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2396     int *attrflagp, int *dattrflagp)
2397 {
2398 	u_int32_t *tl;
2399 	int error = 0;
2400 	struct nfsrv_descript nfsd, *nd = &nfsd;
2401 	nfsattrbit_t attrbits;
2402 
2403 	*nfhpp = NULL;
2404 	*attrflagp = 0;
2405 	*dattrflagp = 0;
2406 	if (namelen > NFS_MAXNAMLEN)
2407 		return (ENAMETOOLONG);
2408 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp, cred);
2409 	if (nd->nd_flag & ND_NFSV4) {
2410 		if (vtyp == VBLK || vtyp == VCHR) {
2411 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
2412 			*tl++ = vtonfsv34_type(vtyp);
2413 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2414 			*tl = txdr_unsigned(NFSMINOR(rdev));
2415 		} else {
2416 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2417 			*tl = vtonfsv34_type(vtyp);
2418 		}
2419 	}
2420 	(void) nfsm_strtom(nd, name, namelen);
2421 	if (nd->nd_flag & ND_NFSV3) {
2422 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2423 		*tl = vtonfsv34_type(vtyp);
2424 	}
2425 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2426 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_NEWFILE, 0);
2427 	if ((nd->nd_flag & ND_NFSV3) &&
2428 	    (vtyp == VCHR || vtyp == VBLK)) {
2429 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2430 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2431 		*tl = txdr_unsigned(NFSMINOR(rdev));
2432 	}
2433 	if (nd->nd_flag & ND_NFSV4) {
2434 		NFSGETATTR_ATTRBIT(&attrbits);
2435 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2436 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2437 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2438 		(void) nfsrv_putattrbit(nd, &attrbits);
2439 	}
2440 	if (nd->nd_flag & ND_NFSV2)
2441 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
2442 	error = nfscl_request(nd, dvp, p, cred);
2443 	if (error)
2444 		return (error);
2445 	if (nd->nd_flag & ND_NFSV4)
2446 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2447 	if (!nd->nd_repstat) {
2448 		if (nd->nd_flag & ND_NFSV4) {
2449 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2450 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2451 			if (error)
2452 				goto nfsmout;
2453 		}
2454 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2455 		if (error)
2456 			goto nfsmout;
2457 	}
2458 	if (nd->nd_flag & ND_NFSV3)
2459 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2460 	if (!error && nd->nd_repstat)
2461 		error = nd->nd_repstat;
2462 nfsmout:
2463 	m_freem(nd->nd_mrep);
2464 	return (error);
2465 }
2466 
2467 /*
2468  * nfs file create call
2469  * Mostly just call the approriate routine. (I separated out v4, so that
2470  * error recovery wouldn't be as difficult.)
2471  */
2472 int
2473 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2474     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2475     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2476     int *attrflagp, int *dattrflagp)
2477 {
2478 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
2479 	struct nfsclowner *owp;
2480 	struct nfscldeleg *dp;
2481 	struct nfsmount *nmp = VFSTONFS(dvp->v_mount);
2482 	u_int32_t clidrev;
2483 
2484 	if (NFSHASNFSV4(nmp)) {
2485 	    retrycnt = 0;
2486 	    do {
2487 		dp = NULL;
2488 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
2489 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
2490 		    NULL, 1, true);
2491 		if (error)
2492 			return (error);
2493 		if (nmp->nm_clp != NULL)
2494 			clidrev = nmp->nm_clp->nfsc_clientidrev;
2495 		else
2496 			clidrev = 0;
2497 		if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
2498 		    nfs_numnfscbd == 0 || retrycnt > 0)
2499 			error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
2500 			  fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2501 			  attrflagp, dattrflagp, &unlocked);
2502 		else
2503 			error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
2504 			  cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2505 			  attrflagp, dattrflagp, &unlocked);
2506 		/*
2507 		 * There is no need to invalidate cached attributes here,
2508 		 * since new post-delegation issue attributes are always
2509 		 * returned by nfsrpc_createv4() and these will update the
2510 		 * attribute cache.
2511 		 */
2512 		if (dp != NULL)
2513 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2514 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2515 		nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2516 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2517 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2518 		    error == NFSERR_BADSESSION) {
2519 			(void) nfs_catnap(PZERO, error, "nfs_open");
2520 		} else if ((error == NFSERR_EXPIRED ||
2521 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2522 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2523 			retrycnt++;
2524 		}
2525 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2526 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2527 		error == NFSERR_BADSESSION ||
2528 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2529 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
2530 	    if (error && retrycnt >= 4)
2531 		    error = EIO;
2532 	} else {
2533 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2534 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp);
2535 	}
2536 	return (error);
2537 }
2538 
2539 /*
2540  * The create rpc for v2 and 3.
2541  */
2542 static int
2543 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2544     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2545     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2546     int *attrflagp, int *dattrflagp)
2547 {
2548 	u_int32_t *tl;
2549 	int error = 0;
2550 	struct nfsrv_descript nfsd, *nd = &nfsd;
2551 
2552 	*nfhpp = NULL;
2553 	*attrflagp = 0;
2554 	*dattrflagp = 0;
2555 	if (namelen > NFS_MAXNAMLEN)
2556 		return (ENAMETOOLONG);
2557 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp, cred);
2558 	(void) nfsm_strtom(nd, name, namelen);
2559 	if (nd->nd_flag & ND_NFSV3) {
2560 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2561 		if (fmode & O_EXCL) {
2562 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2563 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2564 			*tl++ = cverf.lval[0];
2565 			*tl = cverf.lval[1];
2566 		} else {
2567 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2568 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
2569 		}
2570 	} else {
2571 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2572 	}
2573 	error = nfscl_request(nd, dvp, p, cred);
2574 	if (error)
2575 		return (error);
2576 	if (nd->nd_repstat == 0) {
2577 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2578 		if (error)
2579 			goto nfsmout;
2580 	}
2581 	if (nd->nd_flag & ND_NFSV3)
2582 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2583 	if (nd->nd_repstat != 0 && error == 0)
2584 		error = nd->nd_repstat;
2585 nfsmout:
2586 	m_freem(nd->nd_mrep);
2587 	return (error);
2588 }
2589 
2590 static int
2591 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2592     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2593     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2594     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2595     int *dattrflagp, int *unlockedp)
2596 {
2597 	u_int32_t *tl;
2598 	int error = 0, deleg, newone, ret, acesize, limitby;
2599 	struct nfsrv_descript nfsd, *nd = &nfsd;
2600 	struct nfsclopen *op;
2601 	struct nfscldeleg *dp = NULL;
2602 	struct nfsnode *np;
2603 	struct nfsfh *nfhp;
2604 	nfsattrbit_t attrbits;
2605 	nfsv4stateid_t stateid;
2606 	u_int32_t rflags;
2607 	struct nfsmount *nmp;
2608 	struct nfsclsession *tsep;
2609 
2610 	nmp = VFSTONFS(dvp->v_mount);
2611 	np = VTONFS(dvp);
2612 	*unlockedp = 0;
2613 	*nfhpp = NULL;
2614 	*dpp = NULL;
2615 	*attrflagp = 0;
2616 	*dattrflagp = 0;
2617 	if (namelen > NFS_MAXNAMLEN)
2618 		return (ENAMETOOLONG);
2619 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp, cred);
2620 	/*
2621 	 * For V4, this is actually an Open op.
2622 	 */
2623 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2624 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
2625 	if (NFSHASNFSV4N(nmp)) {
2626 		if (!NFSHASPNFS(nmp) && nfscl_enablecallb != 0 &&
2627 		    nfs_numnfscbd > 0)
2628 			*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2629 			    NFSV4OPEN_ACCESSREAD | NFSV4OPEN_WANTWRITEDELEG);
2630 		else
2631 			*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2632 			    NFSV4OPEN_ACCESSREAD | NFSV4OPEN_WANTNODELEG);
2633 	} else
2634 		*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2635 		    NFSV4OPEN_ACCESSREAD);
2636 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2637 	tsep = nfsmnt_mdssession(nmp);
2638 	*tl++ = tsep->nfsess_clientid.lval[0];
2639 	*tl = tsep->nfsess_clientid.lval[1];
2640 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2641 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2642 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2643 	if (fmode & O_EXCL) {
2644 		if (NFSHASNFSV4N(nmp)) {
2645 			if (NFSHASSESSPERSIST(nmp)) {
2646 				/* Use GUARDED for persistent sessions. */
2647 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
2648 				nfscl_fillsattr(nd, vap, dvp, NFSSATTR_NEWFILE,
2649 				    0);
2650 			} else {
2651 				/* Otherwise, use EXCLUSIVE4_1. */
2652 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2653 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2654 				*tl++ = cverf.lval[0];
2655 				*tl = cverf.lval[1];
2656 				nfscl_fillsattr(nd, vap, dvp, NFSSATTR_NEWFILE,
2657 				    0);
2658 			}
2659 		} else {
2660 			/* NFSv4.0 */
2661 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2662 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2663 			*tl++ = cverf.lval[0];
2664 			*tl = cverf.lval[1];
2665 		}
2666 	} else {
2667 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2668 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_NEWFILE, 0);
2669 	}
2670 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2671 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2672 	(void) nfsm_strtom(nd, name, namelen);
2673 	/* Get the new file's handle and attributes. */
2674 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2675 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2676 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2677 	NFSGETATTR_ATTRBIT(&attrbits);
2678 	(void) nfsrv_putattrbit(nd, &attrbits);
2679 	/* Get the directory's post-op attributes. */
2680 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2681 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2682 	(void)nfsm_fhtom(nmp, nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2683 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2684 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2685 	(void) nfsrv_putattrbit(nd, &attrbits);
2686 	error = nfscl_request(nd, dvp, p, cred);
2687 	if (error)
2688 		return (error);
2689 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2690 	if (nd->nd_repstat == 0) {
2691 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2692 		    6 * NFSX_UNSIGNED);
2693 		stateid.seqid = *tl++;
2694 		stateid.other[0] = *tl++;
2695 		stateid.other[1] = *tl++;
2696 		stateid.other[2] = *tl;
2697 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2698 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2699 		if (error)
2700 			goto nfsmout;
2701 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2702 		deleg = fxdr_unsigned(int, *tl);
2703 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2704 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2705 			if (!(owp->nfsow_clp->nfsc_flags &
2706 			      NFSCLFLAGS_FIRSTDELEG))
2707 				owp->nfsow_clp->nfsc_flags |=
2708 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2709 			dp = malloc(
2710 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2711 			    M_NFSCLDELEG, M_WAITOK);
2712 			LIST_INIT(&dp->nfsdl_owner);
2713 			LIST_INIT(&dp->nfsdl_lock);
2714 			dp->nfsdl_clp = owp->nfsow_clp;
2715 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2716 			nfscl_lockinit(&dp->nfsdl_rwlock);
2717 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2718 			    NFSX_UNSIGNED);
2719 			dp->nfsdl_stateid.seqid = *tl++;
2720 			dp->nfsdl_stateid.other[0] = *tl++;
2721 			dp->nfsdl_stateid.other[1] = *tl++;
2722 			dp->nfsdl_stateid.other[2] = *tl++;
2723 			ret = fxdr_unsigned(int, *tl);
2724 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2725 				dp->nfsdl_flags = NFSCLDL_WRITE;
2726 				/*
2727 				 * Indicates how much the file can grow.
2728 				 */
2729 				NFSM_DISSECT(tl, u_int32_t *,
2730 				    3 * NFSX_UNSIGNED);
2731 				limitby = fxdr_unsigned(int, *tl++);
2732 				switch (limitby) {
2733 				case NFSV4OPEN_LIMITSIZE:
2734 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2735 					break;
2736 				case NFSV4OPEN_LIMITBLOCKS:
2737 					dp->nfsdl_sizelimit =
2738 					    fxdr_unsigned(u_int64_t, *tl++);
2739 					dp->nfsdl_sizelimit *=
2740 					    fxdr_unsigned(u_int64_t, *tl);
2741 					break;
2742 				default:
2743 					error = NFSERR_BADXDR;
2744 					goto nfsmout;
2745 				}
2746 			} else {
2747 				dp->nfsdl_flags = NFSCLDL_READ;
2748 			}
2749 			if (ret)
2750 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2751 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, false,
2752 			    &ret, &acesize, p);
2753 			if (error)
2754 				goto nfsmout;
2755 		} else if (deleg == NFSV4OPEN_DELEGATENONEEXT &&
2756 		    NFSHASNFSV4N(nmp)) {
2757 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
2758 			deleg = fxdr_unsigned(uint32_t, *tl);
2759 			if (deleg == NFSV4OPEN_CONTENTION ||
2760 			    deleg == NFSV4OPEN_RESOURCE)
2761 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
2762 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2763 			error = NFSERR_BADXDR;
2764 			goto nfsmout;
2765 		}
2766 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2767 		if (error)
2768 			goto nfsmout;
2769 		/* Get rid of the PutFH and Getattr status values. */
2770 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2771 		/* Load the directory attributes. */
2772 		error = nfsm_loadattr(nd, dnap);
2773 		if (error)
2774 			goto nfsmout;
2775 		*dattrflagp = 1;
2776 		if (dp != NULL && *attrflagp) {
2777 			dp->nfsdl_change = nnap->na_filerev;
2778 			dp->nfsdl_modtime = nnap->na_mtime;
2779 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2780 		}
2781 		/*
2782 		 * We can now complete the Open state.
2783 		 */
2784 		nfhp = *nfhpp;
2785 		if (dp != NULL) {
2786 			dp->nfsdl_fhlen = nfhp->nfh_len;
2787 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2788 		}
2789 		/*
2790 		 * Get an Open structure that will be
2791 		 * attached to the OpenOwner, acquired already.
2792 		 */
2793 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2794 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2795 		    cred, p, NULL, &op, &newone, NULL, 0, false);
2796 		if (error)
2797 			goto nfsmout;
2798 		op->nfso_stateid = stateid;
2799 		newnfs_copyincred(cred, &op->nfso_cred);
2800 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2801 		    do {
2802 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2803 			    nfhp->nfh_len, op, cred, p);
2804 			if (ret == NFSERR_DELAY)
2805 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2806 		    } while (ret == NFSERR_DELAY);
2807 		    error = ret;
2808 		}
2809 
2810 		/*
2811 		 * If the server is handing out delegations, but we didn't
2812 		 * get one because an OpenConfirm was required, try the
2813 		 * Open again, to get a delegation. This is a harmless no-op,
2814 		 * from a server's point of view.
2815 		 */
2816 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2817 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2818 		    !error && dp == NULL) {
2819 		    KASSERT(!NFSHASNFSV4N(nmp),
2820 			("nfsrpc_createv4: result confirm"));
2821 		    do {
2822 			ret = nfsrpc_openrpc(VFSTONFS(dvp->v_mount), dvp,
2823 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2824 			    nfhp->nfh_fh, nfhp->nfh_len,
2825 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2826 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2827 			if (ret == NFSERR_DELAY)
2828 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2829 		    } while (ret == NFSERR_DELAY);
2830 		    if (ret) {
2831 			if (dp != NULL) {
2832 				free(dp, M_NFSCLDELEG);
2833 				dp = NULL;
2834 			}
2835 			if (ret == NFSERR_STALECLIENTID ||
2836 			    ret == NFSERR_STALEDONTRECOVER ||
2837 			    ret == NFSERR_BADSESSION)
2838 				error = ret;
2839 		    }
2840 		}
2841 		nfscl_openrelease(nmp, op, error, newone);
2842 		*unlockedp = 1;
2843 	}
2844 	if (nd->nd_repstat != 0 && error == 0)
2845 		error = nd->nd_repstat;
2846 	if (error == NFSERR_STALECLIENTID)
2847 		nfscl_initiate_recovery(owp->nfsow_clp);
2848 nfsmout:
2849 	if (!error)
2850 		*dpp = dp;
2851 	else if (dp != NULL)
2852 		free(dp, M_NFSCLDELEG);
2853 	m_freem(nd->nd_mrep);
2854 	return (error);
2855 }
2856 
2857 /*
2858  * Nfs remove rpc
2859  */
2860 int
2861 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2862     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp)
2863 {
2864 	u_int32_t *tl;
2865 	struct nfsrv_descript nfsd, *nd = &nfsd;
2866 	struct nfsnode *np;
2867 	struct nfsmount *nmp;
2868 	nfsv4stateid_t dstateid;
2869 	int error, ret = 0, i;
2870 
2871 	*dattrflagp = 0;
2872 	if (namelen > NFS_MAXNAMLEN)
2873 		return (ENAMETOOLONG);
2874 	nmp = VFSTONFS(dvp->v_mount);
2875 tryagain:
2876 	if (NFSHASNFSV4(nmp) && ret == 0) {
2877 		ret = nfscl_removedeleg(vp, p, &dstateid);
2878 		if (ret == 1) {
2879 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp, cred);
2880 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2881 			    NFSX_UNSIGNED);
2882 			if (NFSHASNFSV4N(nmp))
2883 				*tl++ = 0;
2884 			else
2885 				*tl++ = dstateid.seqid;
2886 			*tl++ = dstateid.other[0];
2887 			*tl++ = dstateid.other[1];
2888 			*tl++ = dstateid.other[2];
2889 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2890 			np = VTONFS(dvp);
2891 			(void)nfsm_fhtom(nmp, nd, np->n_fhp->nfh_fh,
2892 			    np->n_fhp->nfh_len, 0);
2893 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2894 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2895 		}
2896 	} else {
2897 		ret = 0;
2898 	}
2899 	if (ret == 0)
2900 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp, cred);
2901 	(void) nfsm_strtom(nd, name, namelen);
2902 	error = nfscl_request(nd, dvp, p, cred);
2903 	if (error)
2904 		return (error);
2905 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2906 		/* For NFSv4, parse out any Delereturn replies. */
2907 		if (ret > 0 && nd->nd_repstat != 0 &&
2908 		    (nd->nd_flag & ND_NOMOREDATA)) {
2909 			/*
2910 			 * If the Delegreturn failed, try again without
2911 			 * it. The server will Recall, as required.
2912 			 */
2913 			m_freem(nd->nd_mrep);
2914 			goto tryagain;
2915 		}
2916 		for (i = 0; i < (ret * 2); i++) {
2917 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2918 			    ND_NFSV4) {
2919 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2920 			    if (*(tl + 1))
2921 				nd->nd_flag |= ND_NOMOREDATA;
2922 			}
2923 		}
2924 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2925 	}
2926 	if (nd->nd_repstat && !error)
2927 		error = nd->nd_repstat;
2928 nfsmout:
2929 	m_freem(nd->nd_mrep);
2930 	return (error);
2931 }
2932 
2933 /*
2934  * Do an nfs rename rpc.
2935  */
2936 int
2937 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2938     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2939     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2940     int *fattrflagp, int *tattrflagp)
2941 {
2942 	u_int32_t *tl;
2943 	struct nfsrv_descript nfsd, *nd = &nfsd;
2944 	struct nfsmount *nmp;
2945 	struct nfsnode *np;
2946 	nfsattrbit_t attrbits;
2947 	nfsv4stateid_t fdstateid, tdstateid;
2948 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2949 
2950 	*fattrflagp = 0;
2951 	*tattrflagp = 0;
2952 	nmp = VFSTONFS(fdvp->v_mount);
2953 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2954 		return (ENAMETOOLONG);
2955 tryagain:
2956 	if (NFSHASNFSV4(nmp) && ret == 0) {
2957 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2958 		    &tdstateid, &gottd, p);
2959 		if (gotfd && gottd) {
2960 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp, cred);
2961 		} else if (gotfd) {
2962 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp, cred);
2963 		} else if (gottd) {
2964 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp, cred);
2965 		}
2966 		if (gotfd) {
2967 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2968 			if (NFSHASNFSV4N(nmp))
2969 				*tl++ = 0;
2970 			else
2971 				*tl++ = fdstateid.seqid;
2972 			*tl++ = fdstateid.other[0];
2973 			*tl++ = fdstateid.other[1];
2974 			*tl = fdstateid.other[2];
2975 			if (gottd) {
2976 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2977 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2978 				np = VTONFS(tvp);
2979 				(void)nfsm_fhtom(nmp, nd, np->n_fhp->nfh_fh,
2980 				    np->n_fhp->nfh_len, 0);
2981 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2982 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2983 			}
2984 		}
2985 		if (gottd) {
2986 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2987 			if (NFSHASNFSV4N(nmp))
2988 				*tl++ = 0;
2989 			else
2990 				*tl++ = tdstateid.seqid;
2991 			*tl++ = tdstateid.other[0];
2992 			*tl++ = tdstateid.other[1];
2993 			*tl = tdstateid.other[2];
2994 		}
2995 		if (ret > 0) {
2996 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2997 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2998 			np = VTONFS(fdvp);
2999 			(void)nfsm_fhtom(nmp, nd, np->n_fhp->nfh_fh,
3000 			    np->n_fhp->nfh_len, 0);
3001 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3002 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
3003 		}
3004 	} else {
3005 		ret = 0;
3006 	}
3007 	if (ret == 0)
3008 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp, cred);
3009 	if (nd->nd_flag & ND_NFSV4) {
3010 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3011 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3012 		NFSWCCATTR_ATTRBIT(&attrbits);
3013 		(void) nfsrv_putattrbit(nd, &attrbits);
3014 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3015 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
3016 		(void)nfsm_fhtom(nmp, nd, VTONFS(tdvp)->n_fhp->nfh_fh,
3017 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
3018 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3019 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3020 		(void) nfsrv_putattrbit(nd, &attrbits);
3021 		nd->nd_flag |= ND_V4WCCATTR;
3022 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3023 		*tl = txdr_unsigned(NFSV4OP_RENAME);
3024 	}
3025 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
3026 	if (!(nd->nd_flag & ND_NFSV4))
3027 		(void)nfsm_fhtom(nmp, nd, VTONFS(tdvp)->n_fhp->nfh_fh,
3028 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
3029 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
3030 	error = nfscl_request(nd, fdvp, p, cred);
3031 	if (error)
3032 		return (error);
3033 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
3034 		/* For NFSv4, parse out any Delereturn replies. */
3035 		if (ret > 0 && nd->nd_repstat != 0 &&
3036 		    (nd->nd_flag & ND_NOMOREDATA)) {
3037 			/*
3038 			 * If the Delegreturn failed, try again without
3039 			 * it. The server will Recall, as required.
3040 			 */
3041 			m_freem(nd->nd_mrep);
3042 			goto tryagain;
3043 		}
3044 		for (i = 0; i < (ret * 2); i++) {
3045 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
3046 			    ND_NFSV4) {
3047 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3048 			    if (*(tl + 1)) {
3049 				if (i == 1 && ret > 1) {
3050 				    /*
3051 				     * If the Delegreturn failed, try again
3052 				     * without it. The server will Recall, as
3053 				     * required.
3054 				     * If ret > 1, the second iteration of this
3055 				     * loop is the second DelegReturn result.
3056 				     */
3057 				    m_freem(nd->nd_mrep);
3058 				    goto tryagain;
3059 				} else {
3060 				    nd->nd_flag |= ND_NOMOREDATA;
3061 				}
3062 			    }
3063 			}
3064 		}
3065 		/* Now, the first wcc attribute reply. */
3066 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
3067 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3068 			if (*(tl + 1))
3069 				nd->nd_flag |= ND_NOMOREDATA;
3070 		}
3071 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL, NULL);
3072 		/* and the second wcc attribute reply. */
3073 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
3074 		    !error) {
3075 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3076 			if (*(tl + 1))
3077 				nd->nd_flag |= ND_NOMOREDATA;
3078 		}
3079 		if (!error)
3080 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
3081 			    NULL, NULL);
3082 	}
3083 	if (nd->nd_repstat && !error)
3084 		error = nd->nd_repstat;
3085 nfsmout:
3086 	m_freem(nd->nd_mrep);
3087 	return (error);
3088 }
3089 
3090 /*
3091  * nfs hard link create rpc
3092  */
3093 int
3094 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
3095     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
3096     struct nfsvattr *nap, int *attrflagp, int *dattrflagp)
3097 {
3098 	u_int32_t *tl;
3099 	struct nfsrv_descript nfsd, *nd = &nfsd;
3100 	nfsattrbit_t attrbits;
3101 	int error = 0;
3102 
3103 	*attrflagp = 0;
3104 	*dattrflagp = 0;
3105 	if (namelen > NFS_MAXNAMLEN)
3106 		return (ENAMETOOLONG);
3107 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp, cred);
3108 	if (nd->nd_flag & ND_NFSV4) {
3109 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3110 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
3111 	}
3112 	(void)nfsm_fhtom(VFSTONFS(dvp->v_mount), nd, VTONFS(dvp)->n_fhp->nfh_fh,
3113 		VTONFS(dvp)->n_fhp->nfh_len, 0);
3114 	if (nd->nd_flag & ND_NFSV4) {
3115 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3116 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3117 		NFSWCCATTR_ATTRBIT(&attrbits);
3118 		(void) nfsrv_putattrbit(nd, &attrbits);
3119 		nd->nd_flag |= ND_V4WCCATTR;
3120 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3121 		*tl = txdr_unsigned(NFSV4OP_LINK);
3122 	}
3123 	(void) nfsm_strtom(nd, name, namelen);
3124 	error = nfscl_request(nd, vp, p, cred);
3125 	if (error)
3126 		return (error);
3127 	if (nd->nd_flag & ND_NFSV3) {
3128 		error = nfscl_postop_attr(nd, nap, attrflagp);
3129 		if (!error)
3130 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
3131 			    NULL, NULL);
3132 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
3133 		/*
3134 		 * First, parse out the PutFH and Getattr result.
3135 		 */
3136 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3137 		if (!(*(tl + 1)))
3138 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3139 		if (*(tl + 1))
3140 			nd->nd_flag |= ND_NOMOREDATA;
3141 		/*
3142 		 * Get the pre-op attributes.
3143 		 */
3144 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3145 	}
3146 	if (nd->nd_repstat && !error)
3147 		error = nd->nd_repstat;
3148 nfsmout:
3149 	m_freem(nd->nd_mrep);
3150 	return (error);
3151 }
3152 
3153 /*
3154  * nfs symbolic link create rpc
3155  */
3156 int
3157 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
3158     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
3159     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
3160     int *dattrflagp)
3161 {
3162 	u_int32_t *tl;
3163 	struct nfsrv_descript nfsd, *nd = &nfsd;
3164 	struct nfsmount *nmp;
3165 	int slen, error = 0;
3166 
3167 	*nfhpp = NULL;
3168 	*attrflagp = 0;
3169 	*dattrflagp = 0;
3170 	nmp = VFSTONFS(dvp->v_mount);
3171 	slen = strlen(target);
3172 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
3173 		return (ENAMETOOLONG);
3174 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp, cred);
3175 	if (nd->nd_flag & ND_NFSV4) {
3176 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3177 		*tl = txdr_unsigned(NFLNK);
3178 		(void) nfsm_strtom(nd, target, slen);
3179 	}
3180 	(void) nfsm_strtom(nd, name, namelen);
3181 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
3182 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
3183 	if (!(nd->nd_flag & ND_NFSV4))
3184 		(void) nfsm_strtom(nd, target, slen);
3185 	if (nd->nd_flag & ND_NFSV2)
3186 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
3187 	error = nfscl_request(nd, dvp, p, cred);
3188 	if (error)
3189 		return (error);
3190 	if (nd->nd_flag & ND_NFSV4)
3191 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3192 	if ((nd->nd_flag & ND_NFSV3) && !error) {
3193 		if (!nd->nd_repstat)
3194 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
3195 		if (!error)
3196 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
3197 			    NULL, NULL);
3198 	}
3199 	if (nd->nd_repstat && !error)
3200 		error = nd->nd_repstat;
3201 	m_freem(nd->nd_mrep);
3202 	/*
3203 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
3204 	 * Only do this if vfs.nfs.ignore_eexist is set.
3205 	 * Never do this for NFSv4.1 or later minor versions, since sessions
3206 	 * should guarantee "exactly once" RPC semantics.
3207 	 */
3208 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
3209 	    nmp->nm_minorvers == 0))
3210 		error = 0;
3211 	return (error);
3212 }
3213 
3214 /*
3215  * nfs make dir rpc
3216  */
3217 int
3218 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
3219     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
3220     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
3221     int *dattrflagp)
3222 {
3223 	u_int32_t *tl;
3224 	struct nfsrv_descript nfsd, *nd = &nfsd;
3225 	nfsattrbit_t attrbits;
3226 	int error = 0;
3227 	struct nfsfh *fhp;
3228 	struct nfsmount *nmp;
3229 
3230 	*nfhpp = NULL;
3231 	*attrflagp = 0;
3232 	*dattrflagp = 0;
3233 	nmp = VFSTONFS(dvp->v_mount);
3234 	fhp = VTONFS(dvp)->n_fhp;
3235 	if (namelen > NFS_MAXNAMLEN)
3236 		return (ENAMETOOLONG);
3237 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp, cred);
3238 	if (nd->nd_flag & ND_NFSV4) {
3239 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3240 		*tl = txdr_unsigned(NFDIR);
3241 	}
3242 	(void) nfsm_strtom(nd, name, namelen);
3243 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1 | NFSSATTR_NEWFILE, 0);
3244 	if (nd->nd_flag & ND_NFSV4) {
3245 		NFSGETATTR_ATTRBIT(&attrbits);
3246 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3247 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3248 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3249 		(void) nfsrv_putattrbit(nd, &attrbits);
3250 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3251 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
3252 		(void)nfsm_fhtom(nmp, nd, fhp->nfh_fh, fhp->nfh_len, 0);
3253 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3254 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3255 		(void) nfsrv_putattrbit(nd, &attrbits);
3256 	}
3257 	error = nfscl_request(nd, dvp, p, cred);
3258 	if (error)
3259 		return (error);
3260 	if (nd->nd_flag & ND_NFSV4)
3261 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3262 	if (!nd->nd_repstat && !error) {
3263 		if (nd->nd_flag & ND_NFSV4) {
3264 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3265 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
3266 		}
3267 		if (!error)
3268 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
3269 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
3270 			/* Get rid of the PutFH and Getattr status values. */
3271 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
3272 			/* Load the directory attributes. */
3273 			error = nfsm_loadattr(nd, dnap);
3274 			if (error == 0)
3275 				*dattrflagp = 1;
3276 		}
3277 	}
3278 	if ((nd->nd_flag & ND_NFSV3) && !error)
3279 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3280 	if (nd->nd_repstat && !error)
3281 		error = nd->nd_repstat;
3282 nfsmout:
3283 	m_freem(nd->nd_mrep);
3284 	/*
3285 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
3286 	 * Only do this if vfs.nfs.ignore_eexist is set.
3287 	 * Never do this for NFSv4.1 or later minor versions, since sessions
3288 	 * should guarantee "exactly once" RPC semantics.
3289 	 */
3290 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
3291 	    nmp->nm_minorvers == 0))
3292 		error = 0;
3293 	return (error);
3294 }
3295 
3296 /*
3297  * nfs remove directory call
3298  */
3299 int
3300 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
3301     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp)
3302 {
3303 	struct nfsrv_descript nfsd, *nd = &nfsd;
3304 	int error = 0;
3305 
3306 	*dattrflagp = 0;
3307 	if (namelen > NFS_MAXNAMLEN)
3308 		return (ENAMETOOLONG);
3309 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp, cred);
3310 	(void) nfsm_strtom(nd, name, namelen);
3311 	error = nfscl_request(nd, dvp, p, cred);
3312 	if (error)
3313 		return (error);
3314 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
3315 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3316 	if (nd->nd_repstat && !error)
3317 		error = nd->nd_repstat;
3318 	m_freem(nd->nd_mrep);
3319 	/*
3320 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
3321 	 */
3322 	if (error == ENOENT)
3323 		error = 0;
3324 	return (error);
3325 }
3326 
3327 /*
3328  * Check to make sure the file name in a Readdir reply is valid.
3329  */
3330 static bool
3331 nfscl_invalidfname(bool is_v4, char *name, int len)
3332 {
3333 	int i;
3334 	char *cp;
3335 
3336 	if (is_v4 && ((len == 1 && name[0] == '.') ||
3337 	    (len == 2 && name[0] == '.' && name[1] == '.'))) {
3338 		printf("Readdir NFSv4 reply has dot or dotdot in it\n");
3339 		return (true);
3340 	}
3341 	cp = name;
3342 	for (i = 0; i < len; i++, cp++) {
3343 		if (*cp == '/' || *cp == '\0') {
3344 			printf("Readdir reply file name had imbedded / or nul"
3345 			    " byte\n");
3346 			return (true);
3347 		}
3348 	}
3349 	return (false);
3350 }
3351 
3352 /*
3353  * Readdir rpc.
3354  * Always returns with either uio_resid unchanged, if you are at the
3355  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
3356  * filled in.
3357  * I felt this would allow caching of directory blocks more easily
3358  * than returning a pertially filled block.
3359  * Directory offset cookies:
3360  * Oh my, what to do with them...
3361  * I can think of three ways to deal with them:
3362  * 1 - have the layer above these RPCs maintain a map between logical
3363  *     directory byte offsets and the NFS directory offset cookies
3364  * 2 - pass the opaque directory offset cookies up into userland
3365  *     and let the libc functions deal with them, via the system call
3366  * 3 - return them to userland in the "struct dirent", so future versions
3367  *     of libc can use them and do whatever is necessary to make things work
3368  *     above these rpc calls, in the meantime
3369  * For now, I do #3 by "hiding" the directory offset cookies after the
3370  * d_name field in struct dirent. This is space inside d_reclen that
3371  * will be ignored by anything that doesn't know about them.
3372  * The directory offset cookies are filled in as the last 8 bytes of
3373  * each directory entry, after d_name. Someday, the userland libc
3374  * functions may be able to use these. In the meantime, it satisfies
3375  * OpenBSD's requirements for cookies being returned.
3376  * If expects the directory offset cookie for the read to be in uio_offset
3377  * and returns the one for the next entry after this directory block in
3378  * there, as well.
3379  */
3380 int
3381 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3382     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3383     int *eofp)
3384 {
3385 	int len, left;
3386 	struct dirent *dp = NULL;
3387 	u_int32_t *tl;
3388 	nfsquad_t cookie, ncookie;
3389 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3390 	struct nfsnode *dnp = VTONFS(vp);
3391 	struct nfsvattr nfsva;
3392 	struct nfsrv_descript nfsd, *nd = &nfsd;
3393 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3394 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
3395 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3396 	char *cp;
3397 	nfsattrbit_t attrbits, dattrbits;
3398 	u_int32_t rderr, *tl2 = NULL;
3399 	size_t tresid;
3400 
3401 	KASSERT(uiop->uio_iovcnt == 1 &&
3402 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3403 	    ("nfs readdirrpc bad uio"));
3404 	KASSERT(uiop->uio_segflg == UIO_SYSSPACE,
3405 	    ("nfsrpc_readdir: uio userspace"));
3406 	ncookie.lval[0] = ncookie.lval[1] = 0;
3407 	/*
3408 	 * There is no point in reading a lot more than uio_resid, however
3409 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
3410 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
3411 	 * will never make readsize > nm_readdirsize.
3412 	 */
3413 	readsize = nmp->nm_readdirsize;
3414 	if (readsize > uiop->uio_resid)
3415 		readsize = uiop->uio_resid + DIRBLKSIZ;
3416 
3417 	*attrflagp = 0;
3418 	if (eofp)
3419 		*eofp = 0;
3420 	tresid = uiop->uio_resid;
3421 	cookie.lval[0] = cookiep->nfsuquad[0];
3422 	cookie.lval[1] = cookiep->nfsuquad[1];
3423 	nd->nd_mrep = NULL;
3424 
3425 	/*
3426 	 * For NFSv4, first create the "." and ".." entries.
3427 	 */
3428 	if (NFSHASNFSV4(nmp)) {
3429 		reqsize = 6 * NFSX_UNSIGNED;
3430 		NFSGETATTR_ATTRBIT(&dattrbits);
3431 		NFSZERO_ATTRBIT(&attrbits);
3432 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3433 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
3434 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3435 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3436 			NFSSETBIT_ATTRBIT(&attrbits,
3437 			    NFSATTRBIT_MOUNTEDONFILEID);
3438 			gotmnton = 1;
3439 		} else {
3440 			/*
3441 			 * Must fake it. Use the fileno, except when the
3442 			 * fsid is != to that of the directory. For that
3443 			 * case, generate a fake fileno that is not the same.
3444 			 */
3445 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3446 			gotmnton = 0;
3447 		}
3448 
3449 		/*
3450 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3451 		 */
3452 		if (uiop->uio_offset == 0) {
3453 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp, cred);
3454 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3455 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3456 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3457 			(void) nfsrv_putattrbit(nd, &attrbits);
3458 			error = nfscl_request(nd, vp, p, cred);
3459 			if (error)
3460 			    return (error);
3461 			dotfileid = 0;	/* Fake out the compiler. */
3462 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3463 			    error = nfsm_loadattr(nd, &nfsva);
3464 			    if (error != 0)
3465 				goto nfsmout;
3466 			    dotfileid = nfsva.na_fileid;
3467 			}
3468 			if (nd->nd_repstat == 0) {
3469 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3470 			    len = fxdr_unsigned(int, *(tl + 4));
3471 			    if (len > 0 && len <= NFSX_V4FHMAX)
3472 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3473 			    else
3474 				error = EPERM;
3475 			    if (!error) {
3476 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3477 				nfsva.na_mntonfileno = UINT64_MAX;
3478 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3479 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3480 				    NULL, NULL, NULL, p, cred);
3481 				if (error) {
3482 				    dotdotfileid = dotfileid;
3483 				} else if (gotmnton) {
3484 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3485 					dotdotfileid = nfsva.na_mntonfileno;
3486 				    else
3487 					dotdotfileid = nfsva.na_fileid;
3488 				} else if (nfsva.na_filesid[0] ==
3489 				    dnp->n_vattr.na_filesid[0] &&
3490 				    nfsva.na_filesid[1] ==
3491 				    dnp->n_vattr.na_filesid[1]) {
3492 				    dotdotfileid = nfsva.na_fileid;
3493 				} else {
3494 				    do {
3495 					fakefileno--;
3496 				    } while (fakefileno ==
3497 					nfsva.na_fileid);
3498 				    dotdotfileid = fakefileno;
3499 				}
3500 			    }
3501 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3502 			    /*
3503 			     * Lookupp returns NFSERR_NOENT when we are
3504 			     * at the root, so just use the current dir.
3505 			     */
3506 			    nd->nd_repstat = 0;
3507 			    dotdotfileid = dotfileid;
3508 			} else {
3509 			    error = nd->nd_repstat;
3510 			}
3511 			m_freem(nd->nd_mrep);
3512 			if (error)
3513 			    return (error);
3514 			nd->nd_mrep = NULL;
3515 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3516 			dp->d_pad0 = dp->d_pad1 = 0;
3517 			dp->d_off = 0;
3518 			dp->d_type = DT_DIR;
3519 			dp->d_fileno = dotfileid;
3520 			dp->d_namlen = 1;
3521 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3522 			dp->d_name[0] = '.';
3523 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3524 			/*
3525 			 * Just make these offset cookie 0.
3526 			 */
3527 			tl = (u_int32_t *)&dp->d_name[8];
3528 			*tl++ = 0;
3529 			*tl = 0;
3530 			blksiz += dp->d_reclen;
3531 			uiop->uio_resid -= dp->d_reclen;
3532 			uiop->uio_offset += dp->d_reclen;
3533 			uiop->uio_iov->iov_base =
3534 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3535 			uiop->uio_iov->iov_len -= dp->d_reclen;
3536 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3537 			dp->d_pad0 = dp->d_pad1 = 0;
3538 			dp->d_off = 0;
3539 			dp->d_type = DT_DIR;
3540 			dp->d_fileno = dotdotfileid;
3541 			dp->d_namlen = 2;
3542 			*((uint64_t *)dp->d_name) = 0;
3543 			dp->d_name[0] = '.';
3544 			dp->d_name[1] = '.';
3545 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3546 			/*
3547 			 * Just make these offset cookie 0.
3548 			 */
3549 			tl = (u_int32_t *)&dp->d_name[8];
3550 			*tl++ = 0;
3551 			*tl = 0;
3552 			blksiz += dp->d_reclen;
3553 			uiop->uio_resid -= dp->d_reclen;
3554 			uiop->uio_offset += dp->d_reclen;
3555 			uiop->uio_iov->iov_base =
3556 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3557 			uiop->uio_iov->iov_len -= dp->d_reclen;
3558 		}
3559 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3560 	} else {
3561 		reqsize = 5 * NFSX_UNSIGNED;
3562 	}
3563 
3564 	/*
3565 	 * Loop around doing readdir rpc's of size readsize.
3566 	 * The stopping criteria is EOF or buffer full.
3567 	 */
3568 	while (more_dirs && bigenough) {
3569 		*attrflagp = 0;
3570 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp, cred);
3571 		if (nd->nd_flag & ND_NFSV2) {
3572 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3573 			*tl++ = cookie.lval[1];
3574 			*tl = txdr_unsigned(readsize);
3575 		} else {
3576 			NFSM_BUILD(tl, u_int32_t *, reqsize);
3577 			*tl++ = cookie.lval[0];
3578 			*tl++ = cookie.lval[1];
3579 			if (cookie.qval == 0) {
3580 				*tl++ = 0;
3581 				*tl++ = 0;
3582 			} else {
3583 				NFSLOCKNODE(dnp);
3584 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
3585 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
3586 				NFSUNLOCKNODE(dnp);
3587 			}
3588 			if (nd->nd_flag & ND_NFSV4) {
3589 				*tl++ = txdr_unsigned(readsize);
3590 				*tl = txdr_unsigned(readsize);
3591 				(void) nfsrv_putattrbit(nd, &attrbits);
3592 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3593 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
3594 				(void) nfsrv_putattrbit(nd, &dattrbits);
3595 			} else {
3596 				*tl = txdr_unsigned(readsize);
3597 			}
3598 		}
3599 		error = nfscl_request(nd, vp, p, cred);
3600 		if (error)
3601 			return (error);
3602 		if (!(nd->nd_flag & ND_NFSV2)) {
3603 			if (nd->nd_flag & ND_NFSV3)
3604 				error = nfscl_postop_attr(nd, nap, attrflagp);
3605 			if (!nd->nd_repstat && !error) {
3606 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3607 				NFSLOCKNODE(dnp);
3608 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
3609 				dnp->n_cookieverf.nfsuquad[1] = *tl;
3610 				NFSUNLOCKNODE(dnp);
3611 			}
3612 		}
3613 		if (nd->nd_repstat || error) {
3614 			if (!error)
3615 				error = nd->nd_repstat;
3616 			goto nfsmout;
3617 		}
3618 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3619 		more_dirs = fxdr_unsigned(int, *tl);
3620 		if (!more_dirs)
3621 			tryformoredirs = 0;
3622 
3623 		/* loop through the dir entries, doctoring them to 4bsd form */
3624 		while (more_dirs && bigenough) {
3625 			if (nd->nd_flag & ND_NFSV4) {
3626 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3627 				ncookie.lval[0] = *tl++;
3628 				ncookie.lval[1] = *tl++;
3629 				len = fxdr_unsigned(int, *tl);
3630 			} else if (nd->nd_flag & ND_NFSV3) {
3631 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3632 				nfsva.na_fileid = fxdr_hyper(tl);
3633 				tl += 2;
3634 				len = fxdr_unsigned(int, *tl);
3635 			} else {
3636 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3637 				nfsva.na_fileid = fxdr_unsigned(uint64_t,
3638 				    *tl++);
3639 				len = fxdr_unsigned(int, *tl);
3640 			}
3641 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3642 				error = EBADRPC;
3643 				goto nfsmout;
3644 			}
3645 			tlen = roundup2(len, 8);
3646 			if (tlen == len)
3647 				tlen += 8;  /* To ensure null termination. */
3648 			left = DIRBLKSIZ - blksiz;
3649 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3650 				NFSBZERO(uiop->uio_iov->iov_base, left);
3651 				dp->d_reclen += left;
3652 				uiop->uio_iov->iov_base =
3653 				    (char *)uiop->uio_iov->iov_base + left;
3654 				uiop->uio_iov->iov_len -= left;
3655 				uiop->uio_resid -= left;
3656 				uiop->uio_offset += left;
3657 				blksiz = 0;
3658 			}
3659 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3660 			    uiop->uio_resid)
3661 				bigenough = 0;
3662 			if (bigenough) {
3663 				struct iovec saviov;
3664 				off_t savoff;
3665 				ssize_t savresid;
3666 				int savblksiz;
3667 
3668 				saviov.iov_base = uiop->uio_iov->iov_base;
3669 				saviov.iov_len = uiop->uio_iov->iov_len;
3670 				savoff = uiop->uio_offset;
3671 				savresid = uiop->uio_resid;
3672 				savblksiz = blksiz;
3673 
3674 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3675 				dp->d_pad0 = dp->d_pad1 = 0;
3676 				dp->d_off = 0;
3677 				dp->d_namlen = len;
3678 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3679 				    NFSX_HYPER;
3680 				dp->d_type = DT_UNKNOWN;
3681 				blksiz += dp->d_reclen;
3682 				if (blksiz == DIRBLKSIZ)
3683 					blksiz = 0;
3684 				uiop->uio_resid -= DIRHDSIZ;
3685 				uiop->uio_offset += DIRHDSIZ;
3686 				uiop->uio_iov->iov_base =
3687 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3688 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3689 				cp = uiop->uio_iov->iov_base;
3690 				error = nfsm_mbufuio(nd, uiop, len);
3691 				if (error)
3692 					goto nfsmout;
3693 				/* Check for an invalid file name. */
3694 				if (nfscl_invalidfname(
3695 				    (nd->nd_flag & ND_NFSV4) != 0, cp, len)) {
3696 					/* Skip over this entry. */
3697 					uiop->uio_iov->iov_base =
3698 					    saviov.iov_base;
3699 					uiop->uio_iov->iov_len =
3700 					    saviov.iov_len;
3701 					uiop->uio_offset = savoff;
3702 					uiop->uio_resid = savresid;
3703 					blksiz = savblksiz;
3704 				} else {
3705 					cp = uiop->uio_iov->iov_base;
3706 					tlen -= len;
3707 					NFSBZERO(cp, tlen);
3708 					cp += tlen; /* points to cookie store */
3709 					tl2 = (u_int32_t *)cp;
3710 					uiop->uio_iov->iov_base =
3711 					    (char *)uiop->uio_iov->iov_base +
3712 					    tlen + NFSX_HYPER;
3713 					uiop->uio_iov->iov_len -= tlen +
3714 					    NFSX_HYPER;
3715 					uiop->uio_resid -= tlen + NFSX_HYPER;
3716 					uiop->uio_offset += (tlen + NFSX_HYPER);
3717 				}
3718 			} else {
3719 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3720 				if (error)
3721 					goto nfsmout;
3722 			}
3723 			if (nd->nd_flag & ND_NFSV4) {
3724 				rderr = 0;
3725 				nfsva.na_mntonfileno = UINT64_MAX;
3726 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3727 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3728 				    NULL, NULL, &rderr, p, cred);
3729 				if (error)
3730 					goto nfsmout;
3731 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3732 			} else if (nd->nd_flag & ND_NFSV3) {
3733 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3734 				ncookie.lval[0] = *tl++;
3735 				ncookie.lval[1] = *tl++;
3736 			} else {
3737 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3738 				ncookie.lval[0] = 0;
3739 				ncookie.lval[1] = *tl++;
3740 			}
3741 			if (bigenough) {
3742 			    if (nd->nd_flag & ND_NFSV4) {
3743 				if (rderr) {
3744 				    dp->d_fileno = 0;
3745 				} else {
3746 				    if (gotmnton) {
3747 					if (nfsva.na_mntonfileno != UINT64_MAX)
3748 					    dp->d_fileno = nfsva.na_mntonfileno;
3749 					else
3750 					    dp->d_fileno = nfsva.na_fileid;
3751 				    } else if (nfsva.na_filesid[0] ==
3752 					dnp->n_vattr.na_filesid[0] &&
3753 					nfsva.na_filesid[1] ==
3754 					dnp->n_vattr.na_filesid[1]) {
3755 					dp->d_fileno = nfsva.na_fileid;
3756 				    } else {
3757 					do {
3758 					    fakefileno--;
3759 					} while (fakefileno ==
3760 					    nfsva.na_fileid);
3761 					dp->d_fileno = fakefileno;
3762 				    }
3763 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3764 				}
3765 			    } else {
3766 				dp->d_fileno = nfsva.na_fileid;
3767 			    }
3768 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3769 				ncookie.lval[0];
3770 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3771 				ncookie.lval[1];
3772 			}
3773 			more_dirs = fxdr_unsigned(int, *tl);
3774 		}
3775 		/*
3776 		 * If at end of rpc data, get the eof boolean
3777 		 */
3778 		if (!more_dirs) {
3779 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3780 			eof = fxdr_unsigned(int, *tl);
3781 			if (tryformoredirs)
3782 				more_dirs = !eof;
3783 			if (nd->nd_flag & ND_NFSV4) {
3784 				error = nfscl_postop_attr(nd, nap, attrflagp);
3785 				if (error)
3786 					goto nfsmout;
3787 			}
3788 		}
3789 		m_freem(nd->nd_mrep);
3790 		nd->nd_mrep = NULL;
3791 	}
3792 	/*
3793 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3794 	 * by increasing d_reclen for the last record.
3795 	 */
3796 	if (blksiz > 0) {
3797 		left = DIRBLKSIZ - blksiz;
3798 		NFSBZERO(uiop->uio_iov->iov_base, left);
3799 		dp->d_reclen += left;
3800 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3801 		    left;
3802 		uiop->uio_iov->iov_len -= left;
3803 		uiop->uio_resid -= left;
3804 		uiop->uio_offset += left;
3805 	}
3806 
3807 	/*
3808 	 * If returning no data, assume end of file.
3809 	 * If not bigenough, return not end of file, since you aren't
3810 	 *    returning all the data
3811 	 * Otherwise, return the eof flag from the server.
3812 	 */
3813 	if (eofp) {
3814 		if (tresid == ((size_t)(uiop->uio_resid)))
3815 			*eofp = 1;
3816 		else if (!bigenough)
3817 			*eofp = 0;
3818 		else
3819 			*eofp = eof;
3820 	}
3821 
3822 	/*
3823 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3824 	 */
3825 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3826 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3827 		NFSBZERO(dp, DIRBLKSIZ);
3828 		dp->d_type = DT_UNKNOWN;
3829 		tl = (u_int32_t *)&dp->d_name[4];
3830 		*tl++ = cookie.lval[0];
3831 		*tl = cookie.lval[1];
3832 		dp->d_reclen = DIRBLKSIZ;
3833 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3834 		    DIRBLKSIZ;
3835 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3836 		uiop->uio_resid -= DIRBLKSIZ;
3837 		uiop->uio_offset += DIRBLKSIZ;
3838 	}
3839 
3840 nfsmout:
3841 	if (nd->nd_mrep != NULL)
3842 		m_freem(nd->nd_mrep);
3843 	return (error);
3844 }
3845 
3846 /*
3847  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3848  * (Also used for NFS V4 when mount flag set.)
3849  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3850  */
3851 int
3852 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3853     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3854     int *eofp)
3855 {
3856 	int len, left;
3857 	struct dirent *dp = NULL;
3858 	u_int32_t *tl;
3859 	vnode_t newvp = NULLVP;
3860 	struct nfsrv_descript nfsd, *nd = &nfsd;
3861 	struct nameidata nami, *ndp = &nami;
3862 	struct componentname *cnp = &ndp->ni_cnd;
3863 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3864 	struct nfsnode *dnp = VTONFS(vp), *np;
3865 	struct nfsvattr nfsva;
3866 	struct nfsfh *nfhp;
3867 	nfsquad_t cookie, ncookie;
3868 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3869 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3870 	int isdotdot = 0, unlocknewvp = 0;
3871 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3872 	u_int64_t fileno = 0;
3873 	char *cp;
3874 	nfsattrbit_t attrbits, dattrbits;
3875 	size_t tresid;
3876 	u_int32_t *tl2 = NULL, rderr;
3877 	struct timespec dctime, ts;
3878 	bool attr_ok;
3879 
3880 	KASSERT(uiop->uio_iovcnt == 1 &&
3881 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3882 	    ("nfs readdirplusrpc bad uio"));
3883 	KASSERT(uiop->uio_segflg == UIO_SYSSPACE,
3884 	    ("nfsrpc_readdirplus: uio userspace"));
3885 	ncookie.lval[0] = ncookie.lval[1] = 0;
3886 	timespecclear(&dctime);
3887 	*attrflagp = 0;
3888 	if (eofp != NULL)
3889 		*eofp = 0;
3890 	ndp->ni_dvp = vp;
3891 	nd->nd_mrep = NULL;
3892 	cookie.lval[0] = cookiep->nfsuquad[0];
3893 	cookie.lval[1] = cookiep->nfsuquad[1];
3894 	tresid = uiop->uio_resid;
3895 
3896 	/*
3897 	 * For NFSv4, first create the "." and ".." entries.
3898 	 */
3899 	if (NFSHASNFSV4(nmp)) {
3900 		NFSGETATTR_ATTRBIT(&dattrbits);
3901 		NFSZERO_ATTRBIT(&attrbits);
3902 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3903 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3904 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3905 			NFSSETBIT_ATTRBIT(&attrbits,
3906 			    NFSATTRBIT_MOUNTEDONFILEID);
3907 			gotmnton = 1;
3908 		} else {
3909 			/*
3910 			 * Must fake it. Use the fileno, except when the
3911 			 * fsid is != to that of the directory. For that
3912 			 * case, generate a fake fileno that is not the same.
3913 			 */
3914 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3915 			gotmnton = 0;
3916 		}
3917 
3918 		/*
3919 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3920 		 */
3921 		if (uiop->uio_offset == 0) {
3922 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp, cred);
3923 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3924 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3925 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3926 			(void) nfsrv_putattrbit(nd, &attrbits);
3927 			error = nfscl_request(nd, vp, p, cred);
3928 			if (error)
3929 			    return (error);
3930 			dotfileid = 0;	/* Fake out the compiler. */
3931 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3932 			    error = nfsm_loadattr(nd, &nfsva);
3933 			    if (error != 0)
3934 				goto nfsmout;
3935 			    dctime = nfsva.na_ctime;
3936 			    dotfileid = nfsva.na_fileid;
3937 			}
3938 			if (nd->nd_repstat == 0) {
3939 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3940 			    len = fxdr_unsigned(int, *(tl + 4));
3941 			    if (len > 0 && len <= NFSX_V4FHMAX)
3942 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3943 			    else
3944 				error = EPERM;
3945 			    if (!error) {
3946 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3947 				nfsva.na_mntonfileno = UINT64_MAX;
3948 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3949 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3950 				    NULL, NULL, NULL, p, cred);
3951 				if (error) {
3952 				    dotdotfileid = dotfileid;
3953 				} else if (gotmnton) {
3954 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3955 					dotdotfileid = nfsva.na_mntonfileno;
3956 				    else
3957 					dotdotfileid = nfsva.na_fileid;
3958 				} else if (nfsva.na_filesid[0] ==
3959 				    dnp->n_vattr.na_filesid[0] &&
3960 				    nfsva.na_filesid[1] ==
3961 				    dnp->n_vattr.na_filesid[1]) {
3962 				    dotdotfileid = nfsva.na_fileid;
3963 				} else {
3964 				    do {
3965 					fakefileno--;
3966 				    } while (fakefileno ==
3967 					nfsva.na_fileid);
3968 				    dotdotfileid = fakefileno;
3969 				}
3970 			    }
3971 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3972 			    /*
3973 			     * Lookupp returns NFSERR_NOENT when we are
3974 			     * at the root, so just use the current dir.
3975 			     */
3976 			    nd->nd_repstat = 0;
3977 			    dotdotfileid = dotfileid;
3978 			} else {
3979 			    error = nd->nd_repstat;
3980 			}
3981 			m_freem(nd->nd_mrep);
3982 			if (error)
3983 			    return (error);
3984 			nd->nd_mrep = NULL;
3985 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3986 			dp->d_pad0 = dp->d_pad1 = 0;
3987 			dp->d_off = 0;
3988 			dp->d_type = DT_DIR;
3989 			dp->d_fileno = dotfileid;
3990 			dp->d_namlen = 1;
3991 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3992 			dp->d_name[0] = '.';
3993 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3994 			/*
3995 			 * Just make these offset cookie 0.
3996 			 */
3997 			tl = (u_int32_t *)&dp->d_name[8];
3998 			*tl++ = 0;
3999 			*tl = 0;
4000 			blksiz += dp->d_reclen;
4001 			uiop->uio_resid -= dp->d_reclen;
4002 			uiop->uio_offset += dp->d_reclen;
4003 			uiop->uio_iov->iov_base =
4004 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
4005 			uiop->uio_iov->iov_len -= dp->d_reclen;
4006 			dp = (struct dirent *)uiop->uio_iov->iov_base;
4007 			dp->d_pad0 = dp->d_pad1 = 0;
4008 			dp->d_off = 0;
4009 			dp->d_type = DT_DIR;
4010 			dp->d_fileno = dotdotfileid;
4011 			dp->d_namlen = 2;
4012 			*((uint64_t *)dp->d_name) = 0;
4013 			dp->d_name[0] = '.';
4014 			dp->d_name[1] = '.';
4015 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
4016 			/*
4017 			 * Just make these offset cookie 0.
4018 			 */
4019 			tl = (u_int32_t *)&dp->d_name[8];
4020 			*tl++ = 0;
4021 			*tl = 0;
4022 			blksiz += dp->d_reclen;
4023 			uiop->uio_resid -= dp->d_reclen;
4024 			uiop->uio_offset += dp->d_reclen;
4025 			uiop->uio_iov->iov_base =
4026 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
4027 			uiop->uio_iov->iov_len -= dp->d_reclen;
4028 		}
4029 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
4030 		if (gotmnton)
4031 			NFSSETBIT_ATTRBIT(&attrbits,
4032 			    NFSATTRBIT_MOUNTEDONFILEID);
4033 		if (!NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
4034 		    NFSATTRBIT_TIMECREATE))
4035 			NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMECREATE);
4036 	}
4037 
4038 	/*
4039 	 * Loop around doing readdir rpc's of size nm_readdirsize.
4040 	 * The stopping criteria is EOF or buffer full.
4041 	 */
4042 	while (more_dirs && bigenough) {
4043 		*attrflagp = 0;
4044 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp, cred);
4045  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
4046 		*tl++ = cookie.lval[0];
4047 		*tl++ = cookie.lval[1];
4048 		if (cookie.qval == 0) {
4049 			*tl++ = 0;
4050 			*tl++ = 0;
4051 		} else {
4052 			NFSLOCKNODE(dnp);
4053 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
4054 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
4055 			NFSUNLOCKNODE(dnp);
4056 		}
4057 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
4058 		*tl = txdr_unsigned(nmp->nm_readdirsize);
4059 		if (nd->nd_flag & ND_NFSV4) {
4060 			(void) nfsrv_putattrbit(nd, &attrbits);
4061 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4062 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
4063 			(void) nfsrv_putattrbit(nd, &dattrbits);
4064 		}
4065 		nanouptime(&ts);
4066 		error = nfscl_request(nd, vp, p, cred);
4067 		if (error)
4068 			return (error);
4069 		if (nd->nd_flag & ND_NFSV3)
4070 			error = nfscl_postop_attr(nd, nap, attrflagp);
4071 		if (nd->nd_repstat || error) {
4072 			if (!error)
4073 				error = nd->nd_repstat;
4074 			goto nfsmout;
4075 		}
4076 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
4077 			dctime = nap->na_ctime;
4078 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
4079 		NFSLOCKNODE(dnp);
4080 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
4081 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
4082 		NFSUNLOCKNODE(dnp);
4083 		more_dirs = fxdr_unsigned(int, *tl);
4084 		if (!more_dirs)
4085 			tryformoredirs = 0;
4086 
4087 		/* loop through the dir entries, doctoring them to 4bsd form */
4088 		while (more_dirs && bigenough) {
4089 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
4090 			if (nd->nd_flag & ND_NFSV4) {
4091 				ncookie.lval[0] = *tl++;
4092 				ncookie.lval[1] = *tl++;
4093 			} else {
4094 				fileno = fxdr_hyper(tl);
4095 				tl += 2;
4096 			}
4097 			len = fxdr_unsigned(int, *tl);
4098 			if (len <= 0 || len > NFS_MAXNAMLEN) {
4099 				error = EBADRPC;
4100 				goto nfsmout;
4101 			}
4102 			tlen = roundup2(len, 8);
4103 			if (tlen == len)
4104 				tlen += 8;  /* To ensure null termination. */
4105 			left = DIRBLKSIZ - blksiz;
4106 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
4107 				NFSBZERO(uiop->uio_iov->iov_base, left);
4108 				dp->d_reclen += left;
4109 				uiop->uio_iov->iov_base =
4110 				    (char *)uiop->uio_iov->iov_base + left;
4111 				uiop->uio_iov->iov_len -= left;
4112 				uiop->uio_resid -= left;
4113 				uiop->uio_offset += left;
4114 				blksiz = 0;
4115 			}
4116 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
4117 			    uiop->uio_resid)
4118 				bigenough = 0;
4119 			if (bigenough) {
4120 				struct iovec saviov;
4121 				off_t savoff;
4122 				ssize_t savresid;
4123 				int savblksiz;
4124 
4125 				saviov.iov_base = uiop->uio_iov->iov_base;
4126 				saviov.iov_len = uiop->uio_iov->iov_len;
4127 				savoff = uiop->uio_offset;
4128 				savresid = uiop->uio_resid;
4129 				savblksiz = blksiz;
4130 
4131 				dp = (struct dirent *)uiop->uio_iov->iov_base;
4132 				dp->d_pad0 = dp->d_pad1 = 0;
4133 				dp->d_off = 0;
4134 				dp->d_namlen = len;
4135 				dp->d_reclen = _GENERIC_DIRLEN(len) +
4136 				    NFSX_HYPER;
4137 				dp->d_type = DT_UNKNOWN;
4138 				blksiz += dp->d_reclen;
4139 				if (blksiz == DIRBLKSIZ)
4140 					blksiz = 0;
4141 				uiop->uio_resid -= DIRHDSIZ;
4142 				uiop->uio_offset += DIRHDSIZ;
4143 				uiop->uio_iov->iov_base =
4144 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
4145 				uiop->uio_iov->iov_len -= DIRHDSIZ;
4146 				cnp->cn_nameptr = uiop->uio_iov->iov_base;
4147 				cnp->cn_namelen = len;
4148 				NFSCNHASHZERO(cnp);
4149 				cp = uiop->uio_iov->iov_base;
4150 				error = nfsm_mbufuio(nd, uiop, len);
4151 				if (error)
4152 					goto nfsmout;
4153 				/* Check for an invalid file name. */
4154 				if (nfscl_invalidfname(
4155 				    (nd->nd_flag & ND_NFSV4) != 0, cp, len)) {
4156 					/* Skip over this entry. */
4157 					uiop->uio_iov->iov_base =
4158 					    saviov.iov_base;
4159 					uiop->uio_iov->iov_len =
4160 					    saviov.iov_len;
4161 					uiop->uio_offset = savoff;
4162 					uiop->uio_resid = savresid;
4163 					blksiz = savblksiz;
4164 				} else {
4165 					cp = uiop->uio_iov->iov_base;
4166 					tlen -= len;
4167 					NFSBZERO(cp, tlen);
4168 					cp += tlen; /* points to cookie store */
4169 					tl2 = (u_int32_t *)cp;
4170 					if (len == 2 &&
4171 					    cnp->cn_nameptr[0] == '.' &&
4172 					    cnp->cn_nameptr[1] == '.')
4173 						isdotdot = 1;
4174 					else
4175 						isdotdot = 0;
4176 					uiop->uio_iov->iov_base =
4177 					    (char *)uiop->uio_iov->iov_base +
4178 					    tlen + NFSX_HYPER;
4179 					uiop->uio_iov->iov_len -= tlen +
4180 					    NFSX_HYPER;
4181 					uiop->uio_resid -= tlen + NFSX_HYPER;
4182 					uiop->uio_offset += (tlen + NFSX_HYPER);
4183 				}
4184 			} else {
4185 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
4186 				if (error)
4187 					goto nfsmout;
4188 			}
4189 			nfhp = NULL;
4190 			if (nd->nd_flag & ND_NFSV3) {
4191 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
4192 				ncookie.lval[0] = *tl++;
4193 				ncookie.lval[1] = *tl++;
4194 				attrflag = fxdr_unsigned(int, *tl);
4195 				if (attrflag) {
4196 				  error = nfsm_loadattr(nd, &nfsva);
4197 				  if (error)
4198 					goto nfsmout;
4199 				}
4200 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
4201 				if (*tl) {
4202 					error = nfsm_getfh(nd, &nfhp);
4203 					if (error)
4204 					    goto nfsmout;
4205 				}
4206 				if (!attrflag && nfhp != NULL) {
4207 					free(nfhp, M_NFSFH);
4208 					nfhp = NULL;
4209 				}
4210 			} else {
4211 				rderr = 0;
4212 				nfsva.na_mntonfileno = 0xffffffff;
4213 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
4214 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
4215 				    NULL, NULL, &rderr, p, cred);
4216 				if (error)
4217 					goto nfsmout;
4218 			}
4219 
4220 			if (bigenough) {
4221 			    if (nd->nd_flag & ND_NFSV4) {
4222 				if (rderr) {
4223 				    dp->d_fileno = 0;
4224 				} else if (gotmnton) {
4225 				    if (nfsva.na_mntonfileno != 0xffffffff)
4226 					dp->d_fileno = nfsva.na_mntonfileno;
4227 				    else
4228 					dp->d_fileno = nfsva.na_fileid;
4229 				} else if (nfsva.na_filesid[0] ==
4230 				    dnp->n_vattr.na_filesid[0] &&
4231 				    nfsva.na_filesid[1] ==
4232 				    dnp->n_vattr.na_filesid[1]) {
4233 				    dp->d_fileno = nfsva.na_fileid;
4234 				} else {
4235 				    do {
4236 					fakefileno--;
4237 				    } while (fakefileno ==
4238 					nfsva.na_fileid);
4239 				    dp->d_fileno = fakefileno;
4240 				}
4241 			    } else {
4242 				dp->d_fileno = fileno;
4243 			    }
4244 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
4245 				ncookie.lval[0];
4246 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
4247 				ncookie.lval[1];
4248 
4249 			    if (nfhp != NULL) {
4250 				attr_ok = true;
4251 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
4252 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
4253 				    VREF(vp);
4254 				    newvp = vp;
4255 				    unlocknewvp = 0;
4256 				    free(nfhp, M_NFSFH);
4257 				    np = dnp;
4258 				} else if (isdotdot != 0) {
4259 				    /*
4260 				     * Skip doing a nfscl_nget() call for "..".
4261 				     * There's a race between acquiring the nfs
4262 				     * node here and lookups that look for the
4263 				     * directory being read (in the parent).
4264 				     * It would try to get a lock on ".." here,
4265 				     * owning the lock on the directory being
4266 				     * read. Lookup will hold the lock on ".."
4267 				     * and try to acquire the lock on the
4268 				     * directory being read.
4269 				     * If the directory is unlocked/relocked,
4270 				     * then there is a LOR with the buflock
4271 				     * vp is relocked.
4272 				     */
4273 				    free(nfhp, M_NFSFH);
4274 				} else {
4275 				    error = nfscl_nget(vp->v_mount, vp,
4276 				      nfhp, cnp, p, &np, LK_EXCLUSIVE);
4277 				    if (!error) {
4278 					newvp = NFSTOV(np);
4279 					unlocknewvp = 1;
4280 					/*
4281 					 * If n_localmodtime >= time before RPC,
4282 					 * then a file modification operation,
4283 					 * such as VOP_SETATTR() of size, has
4284 					 * occurred while the Lookup RPC and
4285 					 * acquisition of the vnode happened. As
4286 					 * such, the attributes might be stale,
4287 					 * with possibly an incorrect size.
4288 					 */
4289 					NFSLOCKNODE(np);
4290 					if (timespecisset(
4291 					    &np->n_localmodtime) &&
4292 					    timespeccmp(&np->n_localmodtime,
4293 					    &ts, >=)) {
4294 					    NFSCL_DEBUG(4, "nfsrpc_readdirplus:"
4295 						" localmod stale attributes\n");
4296 					    attr_ok = false;
4297 					}
4298 					NFSUNLOCKNODE(np);
4299 				    }
4300 				}
4301 				nfhp = NULL;
4302 				if (newvp != NULLVP) {
4303 				    if (attr_ok)
4304 					error = nfscl_loadattrcache(&newvp,
4305 					    &nfsva, NULL, 0, 0);
4306 				    if (error) {
4307 					if (unlocknewvp)
4308 					    vput(newvp);
4309 					else
4310 					    vrele(newvp);
4311 					goto nfsmout;
4312 				    }
4313 				    dp->d_type =
4314 					vtonfs_dtype(np->n_vattr.na_type);
4315 				    ndp->ni_vp = newvp;
4316 				    NFSCNHASH(cnp, HASHINIT);
4317 				    if (cnp->cn_namelen <= NCHNAMLEN &&
4318 					ndp->ni_dvp != ndp->ni_vp &&
4319 					(newvp->v_type != VDIR ||
4320 					 dctime.tv_sec != 0)) {
4321 					cache_enter_time_flags(ndp->ni_dvp,
4322 					    ndp->ni_vp, cnp,
4323 					    &nfsva.na_ctime,
4324 					    newvp->v_type != VDIR ? NULL :
4325 					    &dctime, VFS_CACHE_DROPOLD);
4326 				    }
4327 				    if (unlocknewvp)
4328 					vput(newvp);
4329 				    else
4330 					vrele(newvp);
4331 				    newvp = NULLVP;
4332 				}
4333 			    }
4334 			} else if (nfhp != NULL) {
4335 			    free(nfhp, M_NFSFH);
4336 			}
4337 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
4338 			more_dirs = fxdr_unsigned(int, *tl);
4339 		}
4340 		/*
4341 		 * If at end of rpc data, get the eof boolean
4342 		 */
4343 		if (!more_dirs) {
4344 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
4345 			eof = fxdr_unsigned(int, *tl);
4346 			if (tryformoredirs)
4347 				more_dirs = !eof;
4348 			if (nd->nd_flag & ND_NFSV4) {
4349 				error = nfscl_postop_attr(nd, nap, attrflagp);
4350 				if (error)
4351 					goto nfsmout;
4352 			}
4353 		}
4354 		m_freem(nd->nd_mrep);
4355 		nd->nd_mrep = NULL;
4356 	}
4357 	/*
4358 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
4359 	 * by increasing d_reclen for the last record.
4360 	 */
4361 	if (blksiz > 0) {
4362 		left = DIRBLKSIZ - blksiz;
4363 		NFSBZERO(uiop->uio_iov->iov_base, left);
4364 		dp->d_reclen += left;
4365 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
4366 		    left;
4367 		uiop->uio_iov->iov_len -= left;
4368 		uiop->uio_resid -= left;
4369 		uiop->uio_offset += left;
4370 	}
4371 
4372 	/*
4373 	 * If returning no data, assume end of file.
4374 	 * If not bigenough, return not end of file, since you aren't
4375 	 *    returning all the data
4376 	 * Otherwise, return the eof flag from the server.
4377 	 */
4378 	if (eofp != NULL) {
4379 		if (tresid == uiop->uio_resid)
4380 			*eofp = 1;
4381 		else if (!bigenough)
4382 			*eofp = 0;
4383 		else
4384 			*eofp = eof;
4385 	}
4386 
4387 	/*
4388 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
4389 	 */
4390 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
4391 		dp = (struct dirent *)uiop->uio_iov->iov_base;
4392 		NFSBZERO(dp, DIRBLKSIZ);
4393 		dp->d_type = DT_UNKNOWN;
4394 		tl = (u_int32_t *)&dp->d_name[4];
4395 		*tl++ = cookie.lval[0];
4396 		*tl = cookie.lval[1];
4397 		dp->d_reclen = DIRBLKSIZ;
4398 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
4399 		    DIRBLKSIZ;
4400 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
4401 		uiop->uio_resid -= DIRBLKSIZ;
4402 		uiop->uio_offset += DIRBLKSIZ;
4403 	}
4404 
4405 nfsmout:
4406 	if (nd->nd_mrep != NULL)
4407 		m_freem(nd->nd_mrep);
4408 	return (error);
4409 }
4410 
4411 /*
4412  * Nfs commit rpc
4413  */
4414 int
4415 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
4416     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4417 {
4418 	u_int32_t *tl;
4419 	struct nfsrv_descript nfsd, *nd = &nfsd;
4420 	nfsattrbit_t attrbits;
4421 	int error;
4422 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4423 
4424 	*attrflagp = 0;
4425 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp, cred);
4426 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
4427 	txdr_hyper(offset, tl);
4428 	tl += 2;
4429 	*tl = txdr_unsigned(cnt);
4430 	if (nd->nd_flag & ND_NFSV4) {
4431 		/*
4432 		 * And do a Getattr op.
4433 		 */
4434 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4435 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
4436 		NFSGETATTR_ATTRBIT(&attrbits);
4437 		(void) nfsrv_putattrbit(nd, &attrbits);
4438 	}
4439 	error = nfscl_request(nd, vp, p, cred);
4440 	if (error)
4441 		return (error);
4442 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, NULL);
4443 	if (!error && !nd->nd_repstat) {
4444 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
4445 		NFSLOCKMNT(nmp);
4446 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
4447 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
4448 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
4449 		}
4450 		NFSUNLOCKMNT(nmp);
4451 		if (nd->nd_flag & ND_NFSV4)
4452 			error = nfscl_postop_attr(nd, nap, attrflagp);
4453 	}
4454 nfsmout:
4455 	if (!error && nd->nd_repstat)
4456 		error = nd->nd_repstat;
4457 	m_freem(nd->nd_mrep);
4458 	return (error);
4459 }
4460 
4461 /*
4462  * NFS byte range lock rpc.
4463  * (Mostly just calls one of the three lower level RPC routines.)
4464  */
4465 int
4466 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
4467     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4468 {
4469 	struct nfscllockowner *lp;
4470 	struct nfsclclient *clp;
4471 	struct nfsfh *nfhp;
4472 	struct nfsrv_descript nfsd, *nd = &nfsd;
4473 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4474 	u_int64_t off, len;
4475 	off_t start, end;
4476 	u_int32_t clidrev = 0;
4477 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
4478 	int callcnt, dorpc;
4479 
4480 	/*
4481 	 * Convert the flock structure into a start and end and do POSIX
4482 	 * bounds checking.
4483 	 */
4484 	switch (fl->l_whence) {
4485 	case SEEK_SET:
4486 	case SEEK_CUR:
4487 		/*
4488 		 * Caller is responsible for adding any necessary offset
4489 		 * when SEEK_CUR is used.
4490 		 */
4491 		start = fl->l_start;
4492 		off = fl->l_start;
4493 		break;
4494 	case SEEK_END:
4495 		start = size + fl->l_start;
4496 		off = size + fl->l_start;
4497 		break;
4498 	default:
4499 		return (EINVAL);
4500 	}
4501 	if (start < 0)
4502 		return (EINVAL);
4503 	if (fl->l_len != 0) {
4504 		end = start + fl->l_len - 1;
4505 		if (end < start)
4506 			return (EINVAL);
4507 	}
4508 
4509 	len = fl->l_len;
4510 	if (len == 0)
4511 		len = NFS64BITSSET;
4512 	retrycnt = 0;
4513 	do {
4514 	    nd->nd_repstat = 0;
4515 	    if (op == F_GETLK) {
4516 		error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
4517 		if (error)
4518 			return (error);
4519 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
4520 		if (!error) {
4521 			clidrev = clp->nfsc_clientidrev;
4522 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
4523 			    p, id, flags);
4524 		} else if (error == -1) {
4525 			error = 0;
4526 		}
4527 		nfscl_clientrelease(clp);
4528 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
4529 		/*
4530 		 * We must loop around for all lockowner cases.
4531 		 */
4532 		callcnt = 0;
4533 		error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
4534 		if (error)
4535 			return (error);
4536 		do {
4537 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
4538 			clp, id, flags, &lp, &dorpc);
4539 		    /*
4540 		     * If it returns a NULL lp, we're done.
4541 		     */
4542 		    if (lp == NULL) {
4543 			if (callcnt == 0)
4544 			    nfscl_clientrelease(clp);
4545 			else
4546 			    nfscl_releasealllocks(clp, vp, p, id, flags);
4547 			return (error);
4548 		    }
4549 		    if (nmp->nm_clp != NULL)
4550 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4551 		    else
4552 			clidrev = 0;
4553 		    /*
4554 		     * If the server doesn't support Posix lock semantics,
4555 		     * only allow locks on the entire file, since it won't
4556 		     * handle overlapping byte ranges.
4557 		     * There might still be a problem when a lock
4558 		     * upgrade/downgrade (read<->write) occurs, since the
4559 		     * server "might" expect an unlock first?
4560 		     */
4561 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
4562 			(off == 0 && len == NFS64BITSSET))) {
4563 			/*
4564 			 * Since the lock records will go away, we must
4565 			 * wait for grace and delay here.
4566 			 */
4567 			do {
4568 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
4569 				NFSV4LOCKT_READ, cred, p, 0);
4570 			    if ((nd->nd_repstat == NFSERR_GRACE ||
4571 				 nd->nd_repstat == NFSERR_DELAY) &&
4572 				error == 0)
4573 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
4574 				    "nfs_advlock");
4575 			} while ((nd->nd_repstat == NFSERR_GRACE ||
4576 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
4577 		    }
4578 		    callcnt++;
4579 		} while (error == 0 && nd->nd_repstat == 0);
4580 		nfscl_releasealllocks(clp, vp, p, id, flags);
4581 	    } else if (op == F_SETLK) {
4582 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
4583 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
4584 		if (error || donelocally) {
4585 			return (error);
4586 		}
4587 		if (nmp->nm_clp != NULL)
4588 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4589 		else
4590 			clidrev = 0;
4591 		nfhp = VTONFS(vp)->n_fhp;
4592 		if (!lp->nfsl_open->nfso_posixlock &&
4593 		    (off != 0 || len != NFS64BITSSET)) {
4594 			error = EINVAL;
4595 		} else {
4596 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
4597 			    nfhp->nfh_len, lp, newone, reclaim, off,
4598 			    len, fl->l_type, cred, p, 0);
4599 		}
4600 		if (!error)
4601 			error = nd->nd_repstat;
4602 		nfscl_lockrelease(lp, error, newone);
4603 	    } else {
4604 		error = EINVAL;
4605 	    }
4606 	    if (!error)
4607 	        error = nd->nd_repstat;
4608 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
4609 		error == NFSERR_STALEDONTRECOVER ||
4610 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4611 		error == NFSERR_BADSESSION) {
4612 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
4613 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
4614 		&& clidrev != 0) {
4615 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
4616 		retrycnt++;
4617 	    }
4618 	} while (error == NFSERR_GRACE ||
4619 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4620 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
4621 	    error == NFSERR_BADSESSION ||
4622 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
4623 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
4624 	if (error && retrycnt >= 4)
4625 		error = EIO;
4626 	return (error);
4627 }
4628 
4629 /*
4630  * The lower level routine for the LockT case.
4631  */
4632 int
4633 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
4634     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
4635     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4636 {
4637 	u_int32_t *tl;
4638 	int error, type, size;
4639 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4640 	struct nfsnode *np;
4641 	struct nfsmount *nmp;
4642 	struct nfsclsession *tsep;
4643 
4644 	nmp = VFSTONFS(vp->v_mount);
4645 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp, cred);
4646 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4647 	if (fl->l_type == F_RDLCK)
4648 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4649 	else
4650 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4651 	txdr_hyper(off, tl);
4652 	tl += 2;
4653 	txdr_hyper(len, tl);
4654 	tl += 2;
4655 	tsep = nfsmnt_mdssession(nmp);
4656 	*tl++ = tsep->nfsess_clientid.lval[0];
4657 	*tl = tsep->nfsess_clientid.lval[1];
4658 	nfscl_filllockowner(id, own, flags);
4659 	np = VTONFS(vp);
4660 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4661 	    np->n_fhp->nfh_len);
4662 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4663 	error = nfscl_request(nd, vp, p, cred);
4664 	if (error)
4665 		return (error);
4666 	if (nd->nd_repstat == 0) {
4667 		fl->l_type = F_UNLCK;
4668 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4669 		nd->nd_repstat = 0;
4670 		fl->l_whence = SEEK_SET;
4671 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4672 		fl->l_start = fxdr_hyper(tl);
4673 		tl += 2;
4674 		len = fxdr_hyper(tl);
4675 		tl += 2;
4676 		if (len == NFS64BITSSET)
4677 			fl->l_len = 0;
4678 		else
4679 			fl->l_len = len;
4680 		type = fxdr_unsigned(int, *tl++);
4681 		if (type == NFSV4LOCKT_WRITE)
4682 			fl->l_type = F_WRLCK;
4683 		else
4684 			fl->l_type = F_RDLCK;
4685 		/*
4686 		 * XXX For now, I have no idea what to do with the
4687 		 * conflicting lock_owner, so I'll just set the pid == 0
4688 		 * and skip over the lock_owner.
4689 		 */
4690 		fl->l_pid = (pid_t)0;
4691 		tl += 2;
4692 		size = fxdr_unsigned(int, *tl);
4693 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4694 			error = EBADRPC;
4695 		if (!error)
4696 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4697 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4698 		nfscl_initiate_recovery(clp);
4699 nfsmout:
4700 	m_freem(nd->nd_mrep);
4701 	return (error);
4702 }
4703 
4704 /*
4705  * Lower level function that performs the LockU RPC.
4706  */
4707 static int
4708 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4709     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4710     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4711 {
4712 	u_int32_t *tl;
4713 	int error;
4714 
4715 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4716 	    lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0, cred);
4717 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4718 	*tl++ = txdr_unsigned(type);
4719 	*tl = txdr_unsigned(lp->nfsl_seqid);
4720 	if (nfstest_outofseq &&
4721 	    (arc4random() % nfstest_outofseq) == 0)
4722 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
4723 	tl++;
4724 	if (NFSHASNFSV4N(nmp))
4725 		*tl++ = 0;
4726 	else
4727 		*tl++ = lp->nfsl_stateid.seqid;
4728 	*tl++ = lp->nfsl_stateid.other[0];
4729 	*tl++ = lp->nfsl_stateid.other[1];
4730 	*tl++ = lp->nfsl_stateid.other[2];
4731 	txdr_hyper(off, tl);
4732 	tl += 2;
4733 	txdr_hyper(len, tl);
4734 	if (syscred)
4735 		nd->nd_flag |= ND_USEGSSNAME;
4736 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4737 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4738 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4739 	if (error)
4740 		return (error);
4741 	if (nd->nd_repstat == 0) {
4742 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4743 		lp->nfsl_stateid.seqid = *tl++;
4744 		lp->nfsl_stateid.other[0] = *tl++;
4745 		lp->nfsl_stateid.other[1] = *tl++;
4746 		lp->nfsl_stateid.other[2] = *tl;
4747 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4748 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4749 nfsmout:
4750 	m_freem(nd->nd_mrep);
4751 	return (error);
4752 }
4753 
4754 /*
4755  * The actual Lock RPC.
4756  */
4757 int
4758 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4759     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4760     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4761     NFSPROC_T *p, int syscred)
4762 {
4763 	u_int32_t *tl;
4764 	int error, size;
4765 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4766 	struct nfsclsession *tsep;
4767 
4768 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0,
4769 	    cred);
4770 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4771 	if (type == F_RDLCK)
4772 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4773 	else
4774 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4775 	*tl++ = txdr_unsigned(reclaim);
4776 	txdr_hyper(off, tl);
4777 	tl += 2;
4778 	txdr_hyper(len, tl);
4779 	tl += 2;
4780 	if (newone) {
4781 	    *tl = newnfs_true;
4782 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4783 		2 * NFSX_UNSIGNED + NFSX_HYPER);
4784 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4785 	    if (NFSHASNFSV4N(nmp))
4786 		*tl++ = 0;
4787 	    else
4788 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
4789 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4790 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4791 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4792 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
4793 	    tsep = nfsmnt_mdssession(nmp);
4794 	    *tl++ = tsep->nfsess_clientid.lval[0];
4795 	    *tl = tsep->nfsess_clientid.lval[1];
4796 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4797 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4798 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4799 	} else {
4800 	    *tl = newnfs_false;
4801 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4802 	    if (NFSHASNFSV4N(nmp))
4803 		*tl++ = 0;
4804 	    else
4805 		*tl++ = lp->nfsl_stateid.seqid;
4806 	    *tl++ = lp->nfsl_stateid.other[0];
4807 	    *tl++ = lp->nfsl_stateid.other[1];
4808 	    *tl++ = lp->nfsl_stateid.other[2];
4809 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4810 	    if (nfstest_outofseq &&
4811 		(arc4random() % nfstest_outofseq) == 0)
4812 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4813 	}
4814 	if (syscred)
4815 		nd->nd_flag |= ND_USEGSSNAME;
4816 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4817 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4818 	if (error)
4819 		return (error);
4820 	if (newone)
4821 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4822 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4823 	if (nd->nd_repstat == 0) {
4824 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4825 		lp->nfsl_stateid.seqid = *tl++;
4826 		lp->nfsl_stateid.other[0] = *tl++;
4827 		lp->nfsl_stateid.other[1] = *tl++;
4828 		lp->nfsl_stateid.other[2] = *tl;
4829 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4830 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4831 		size = fxdr_unsigned(int, *(tl + 7));
4832 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4833 			error = EBADRPC;
4834 		if (!error)
4835 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4836 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4837 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4838 nfsmout:
4839 	m_freem(nd->nd_mrep);
4840 	return (error);
4841 }
4842 
4843 /*
4844  * nfs statfs rpc
4845  * (always called with the vp for the mount point)
4846  */
4847 int
4848 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4849     uint32_t *leasep, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap,
4850     int *attrflagp)
4851 {
4852 	u_int32_t *tl = NULL;
4853 	struct nfsrv_descript nfsd, *nd = &nfsd;
4854 	struct nfsmount *nmp;
4855 	nfsattrbit_t attrbits;
4856 	int error;
4857 
4858 	*attrflagp = 0;
4859 	nmp = VFSTONFS(vp->v_mount);
4860 	if (NFSHASNFSV4(nmp)) {
4861 		/*
4862 		 * For V4, you actually do a getattr.
4863 		 */
4864 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp, cred);
4865 		if (leasep != NULL)
4866 			NFSROOTFS_GETATTRBIT(&attrbits);
4867 		else
4868 			NFSSTATFS_GETATTRBIT(&attrbits);
4869 		(void) nfsrv_putattrbit(nd, &attrbits);
4870 		nd->nd_flag |= ND_USEGSSNAME;
4871 		error = nfscl_request(nd, vp, p, cred);
4872 		if (error)
4873 			return (error);
4874 		if (nd->nd_repstat == 0) {
4875 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4876 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, leasep, NULL,
4877 			    p, cred);
4878 			if (!error) {
4879 				nmp->nm_fsid[0] = nap->na_filesid[0];
4880 				nmp->nm_fsid[1] = nap->na_filesid[1];
4881 				NFSSETHASSETFSID(nmp);
4882 				*attrflagp = 1;
4883 			}
4884 		} else {
4885 			error = nd->nd_repstat;
4886 		}
4887 		if (error)
4888 			goto nfsmout;
4889 	} else {
4890 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp, NULL);
4891 		error = nfscl_request(nd, vp, p, cred);
4892 		if (error)
4893 			return (error);
4894 		if (nd->nd_flag & ND_NFSV3) {
4895 			error = nfscl_postop_attr(nd, nap, attrflagp);
4896 			if (error)
4897 				goto nfsmout;
4898 		}
4899 		if (nd->nd_repstat) {
4900 			error = nd->nd_repstat;
4901 			goto nfsmout;
4902 		}
4903 		NFSM_DISSECT(tl, u_int32_t *,
4904 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4905 	}
4906 	if (NFSHASNFSV3(nmp)) {
4907 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4908 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4909 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4910 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4911 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4912 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4913 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4914 	} else if (NFSHASNFSV4(nmp) == 0) {
4915 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4916 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4917 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4918 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4919 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4920 	}
4921 nfsmout:
4922 	m_freem(nd->nd_mrep);
4923 	return (error);
4924 }
4925 
4926 /*
4927  * nfs pathconf rpc
4928  */
4929 int
4930 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4931     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4932 {
4933 	struct nfsrv_descript nfsd, *nd = &nfsd;
4934 	struct nfsmount *nmp;
4935 	u_int32_t *tl;
4936 	nfsattrbit_t attrbits;
4937 	int error;
4938 	struct nfsnode *np;
4939 
4940 	*attrflagp = 0;
4941 	nmp = VFSTONFS(vp->v_mount);
4942 	if (NFSHASNFSV4(nmp)) {
4943 		np = VTONFS(vp);
4944 		if ((nmp->nm_privflag & NFSMNTP_FAKEROOTFH) != 0 &&
4945 		    nmp->nm_fhsize == 0) {
4946 			/* Attempt to get the actual root file handle. */
4947 			error = nfsrpc_getdirpath(nmp, NFSMNT_DIRPATH(nmp),
4948 			    cred, p);
4949 			if (error != 0)
4950 				return (EACCES);
4951 			if (np->n_fhp->nfh_len == NFSX_FHMAX + 1)
4952 				nfscl_statfs(vp, cred, p);
4953 		}
4954 		/*
4955 		 * For V4, you actually do a getattr.
4956 		 */
4957 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp, cred);
4958 		NFSPATHCONF_GETATTRBIT(&attrbits);
4959 		(void) nfsrv_putattrbit(nd, &attrbits);
4960 		nd->nd_flag |= ND_USEGSSNAME;
4961 		error = nfscl_request(nd, vp, p, cred);
4962 		if (error)
4963 			return (error);
4964 		if (nd->nd_repstat == 0) {
4965 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4966 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4967 			    cred);
4968 			if (!error)
4969 				*attrflagp = 1;
4970 		} else {
4971 			error = nd->nd_repstat;
4972 		}
4973 	} else {
4974 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp, NULL);
4975 		error = nfscl_request(nd, vp, p, cred);
4976 		if (error)
4977 			return (error);
4978 		error = nfscl_postop_attr(nd, nap, attrflagp);
4979 		if (nd->nd_repstat && !error)
4980 			error = nd->nd_repstat;
4981 		if (!error) {
4982 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4983 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4984 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4985 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4986 			pc->pc_chownrestricted =
4987 			    fxdr_unsigned(u_int32_t, *tl++);
4988 			pc->pc_caseinsensitive =
4989 			    fxdr_unsigned(u_int32_t, *tl++);
4990 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4991 		}
4992 	}
4993 nfsmout:
4994 	m_freem(nd->nd_mrep);
4995 	return (error);
4996 }
4997 
4998 /*
4999  * nfs version 3 fsinfo rpc call
5000  */
5001 int
5002 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
5003     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
5004 {
5005 	u_int32_t *tl;
5006 	struct nfsrv_descript nfsd, *nd = &nfsd;
5007 	int error;
5008 
5009 	*attrflagp = 0;
5010 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp, NULL);
5011 	error = nfscl_request(nd, vp, p, cred);
5012 	if (error)
5013 		return (error);
5014 	error = nfscl_postop_attr(nd, nap, attrflagp);
5015 	if (nd->nd_repstat && !error)
5016 		error = nd->nd_repstat;
5017 	if (!error) {
5018 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
5019 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
5020 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
5021 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
5022 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
5023 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
5024 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
5025 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
5026 		fsp->fs_maxfilesize = fxdr_hyper(tl);
5027 		tl += 2;
5028 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
5029 		tl += 2;
5030 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
5031 	}
5032 nfsmout:
5033 	m_freem(nd->nd_mrep);
5034 	return (error);
5035 }
5036 
5037 /*
5038  * This function performs the Renew RPC.
5039  */
5040 int
5041 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
5042     NFSPROC_T *p)
5043 {
5044 	u_int32_t *tl;
5045 	struct nfsrv_descript nfsd;
5046 	struct nfsrv_descript *nd = &nfsd;
5047 	struct nfsmount *nmp;
5048 	int error;
5049 	struct nfssockreq *nrp;
5050 	struct nfsclsession *tsep;
5051 
5052 	nmp = clp->nfsc_nmp;
5053 	if (nmp == NULL)
5054 		return (0);
5055 	if (dsp == NULL)
5056 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
5057 		    0, cred);
5058 	else
5059 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
5060 		    &dsp->nfsclds_sess, 0, 0, NULL);
5061 	if (!NFSHASNFSV4N(nmp)) {
5062 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
5063 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
5064 		tsep = nfsmnt_mdssession(nmp);
5065 		*tl++ = tsep->nfsess_clientid.lval[0];
5066 		*tl = tsep->nfsess_clientid.lval[1];
5067 	}
5068 	nrp = NULL;
5069 	if (dsp != NULL)
5070 		nrp = dsp->nfsclds_sockp;
5071 	if (nrp == NULL)
5072 		/* If NULL, use the MDS socket. */
5073 		nrp = &nmp->nm_sockreq;
5074 	nd->nd_flag |= ND_USEGSSNAME;
5075 	if (dsp == NULL)
5076 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
5077 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5078 	else {
5079 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
5080 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5081 		if (error == ENXIO)
5082 			nfscl_cancelreqs(dsp);
5083 	}
5084 	if (error)
5085 		return (error);
5086 	error = nd->nd_repstat;
5087 	m_freem(nd->nd_mrep);
5088 	return (error);
5089 }
5090 
5091 /*
5092  * This function performs the Releaselockowner RPC.
5093  */
5094 int
5095 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
5096     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
5097 {
5098 	struct nfsrv_descript nfsd, *nd = &nfsd;
5099 	u_int32_t *tl;
5100 	int error;
5101 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
5102 	struct nfsclsession *tsep;
5103 
5104 	if (NFSHASNFSV4N(nmp)) {
5105 		/* For NFSv4.1, do a FreeStateID. */
5106 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
5107 		    NULL, 0, 0, cred);
5108 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
5109 	} else {
5110 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
5111 		    NULL, 0, 0, NULL);
5112 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
5113 		tsep = nfsmnt_mdssession(nmp);
5114 		*tl++ = tsep->nfsess_clientid.lval[0];
5115 		*tl = tsep->nfsess_clientid.lval[1];
5116 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
5117 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
5118 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
5119 	}
5120 	nd->nd_flag |= ND_USEGSSNAME;
5121 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5122 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5123 	if (error)
5124 		return (error);
5125 	error = nd->nd_repstat;
5126 	m_freem(nd->nd_mrep);
5127 	return (error);
5128 }
5129 
5130 /*
5131  * This function performs the Compound to get the mount pt FH.
5132  */
5133 int
5134 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
5135     NFSPROC_T *p)
5136 {
5137 	u_int32_t *tl;
5138 	struct nfsrv_descript nfsd;
5139 	struct nfsrv_descript *nd = &nfsd;
5140 	u_char *cp, *cp2, *fhp;
5141 	int error, cnt, len, setnil;
5142 	u_int32_t *opcntp;
5143 
5144 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
5145 	    0, NULL);
5146 	cp = dirpath;
5147 	cnt = 0;
5148 	do {
5149 		setnil = 0;
5150 		while (*cp == '/')
5151 			cp++;
5152 		cp2 = cp;
5153 		while (*cp2 != '\0' && *cp2 != '/')
5154 			cp2++;
5155 		if (*cp2 == '/') {
5156 			setnil = 1;
5157 			*cp2 = '\0';
5158 		}
5159 		if (cp2 != cp) {
5160 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
5161 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
5162 			nfsm_strtom(nd, cp, strlen(cp));
5163 			cnt++;
5164 		}
5165 		if (setnil)
5166 			*cp2++ = '/';
5167 		cp = cp2;
5168 	} while (*cp != '\0');
5169 	if (NFSHASNFSV4N(nmp))
5170 		/* Has a Sequence Op done by nfscl_reqstart(). */
5171 		*opcntp = txdr_unsigned(3 + cnt);
5172 	else
5173 		*opcntp = txdr_unsigned(2 + cnt);
5174 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
5175 	*tl = txdr_unsigned(NFSV4OP_GETFH);
5176 	nd->nd_flag |= ND_USEGSSNAME;
5177 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5178 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5179 	if (error)
5180 		return (error);
5181 	if (nd->nd_repstat == 0) {
5182 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
5183 		tl += (2 + 2 * cnt);
5184 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
5185 			len > NFSX_FHMAX) {
5186 			nd->nd_repstat = NFSERR_BADXDR;
5187 		} else {
5188 			fhp = malloc(len + 1, M_TEMP, M_WAITOK);
5189 			nd->nd_repstat = nfsrv_mtostr(nd, fhp, len);
5190 			if (nd->nd_repstat == 0) {
5191 				NFSLOCKMNT(nmp);
5192 				if (nmp->nm_fhsize == 0) {
5193 					NFSBCOPY(fhp, nmp->nm_fh, len);
5194 					nmp->nm_fhsize = len;
5195 				}
5196 				NFSUNLOCKMNT(nmp);
5197 			}
5198 			free(fhp, M_TEMP);
5199 		}
5200 	}
5201 	error = nd->nd_repstat;
5202 nfsmout:
5203 	m_freem(nd->nd_mrep);
5204 	return (error);
5205 }
5206 
5207 /*
5208  * This function performs the Delegreturn RPC.
5209  */
5210 int
5211 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
5212     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
5213 {
5214 	u_int32_t *tl;
5215 	struct nfsrv_descript nfsd;
5216 	struct nfsrv_descript *nd = &nfsd;
5217 	int error;
5218 
5219 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
5220 	    dp->nfsdl_fhlen, NULL, NULL, 0, 0, cred);
5221 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
5222 	if (NFSHASNFSV4N(nmp))
5223 		*tl++ = 0;
5224 	else
5225 		*tl++ = dp->nfsdl_stateid.seqid;
5226 	*tl++ = dp->nfsdl_stateid.other[0];
5227 	*tl++ = dp->nfsdl_stateid.other[1];
5228 	*tl = dp->nfsdl_stateid.other[2];
5229 	if (syscred)
5230 		nd->nd_flag |= ND_USEGSSNAME;
5231 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5232 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5233 	if (error)
5234 		return (error);
5235 	error = nd->nd_repstat;
5236 	m_freem(nd->nd_mrep);
5237 	return (error);
5238 }
5239 
5240 /*
5241  * nfs getacl call.
5242  */
5243 int
5244 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p, struct acl *aclp)
5245 {
5246 	struct nfsrv_descript nfsd, *nd = &nfsd;
5247 	int error;
5248 	nfsattrbit_t attrbits;
5249 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5250 
5251 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
5252 		return (EOPNOTSUPP);
5253 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp, cred);
5254 	NFSZERO_ATTRBIT(&attrbits);
5255 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
5256 	(void) nfsrv_putattrbit(nd, &attrbits);
5257 	error = nfscl_request(nd, vp, p, cred);
5258 	if (error)
5259 		return (error);
5260 	if (!nd->nd_repstat)
5261 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
5262 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
5263 	else
5264 		error = nd->nd_repstat;
5265 	m_freem(nd->nd_mrep);
5266 	return (error);
5267 }
5268 
5269 /*
5270  * nfs setacl call.
5271  */
5272 int
5273 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p, struct acl *aclp)
5274 {
5275 	int error;
5276 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5277 
5278 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
5279 		return (EOPNOTSUPP);
5280 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL);
5281 	return (error);
5282 }
5283 
5284 /*
5285  * nfs setacl call.
5286  */
5287 static int
5288 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
5289     struct acl *aclp, nfsv4stateid_t *stateidp)
5290 {
5291 	struct nfsrv_descript nfsd, *nd = &nfsd;
5292 	int error;
5293 	nfsattrbit_t attrbits;
5294 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5295 
5296 	if (!NFSHASNFSV4(nmp))
5297 		return (EOPNOTSUPP);
5298 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp, cred);
5299 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
5300 	NFSZERO_ATTRBIT(&attrbits);
5301 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
5302 	(void) nfsv4_fillattr(nd, vp->v_mount, vp, aclp, NULL, NULL, 0,
5303 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
5304 	error = nfscl_request(nd, vp, p, cred);
5305 	if (error)
5306 		return (error);
5307 	/* Don't care about the pre/postop attributes */
5308 	m_freem(nd->nd_mrep);
5309 	return (nd->nd_repstat);
5310 }
5311 
5312 /*
5313  * Do the NFSv4.1 Exchange ID.
5314  */
5315 int
5316 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
5317     struct nfssockreq *nrp, int minorvers, uint32_t exchflags,
5318     struct nfsclds **dspp, struct ucred *cred, NFSPROC_T *p)
5319 {
5320 	uint32_t *tl, v41flags;
5321 	struct nfsrv_descript nfsd;
5322 	struct nfsrv_descript *nd = &nfsd;
5323 	struct nfsclds *dsp;
5324 	struct timespec verstime;
5325 	int error, len;
5326 
5327 	*dspp = NULL;
5328 	if (minorvers == 0)
5329 		minorvers = nmp->nm_minorvers;
5330 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL,
5331 	    NFS_VER4, minorvers, NULL);
5332 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5333 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
5334 	*tl = txdr_unsigned(clp->nfsc_rev);
5335 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
5336 
5337 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
5338 	*tl++ = txdr_unsigned(exchflags);
5339 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
5340 
5341 	/* Set the implementation id4 */
5342 	*tl = txdr_unsigned(1);
5343 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
5344 	(void) nfsm_strtom(nd, version, strlen(version));
5345 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
5346 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
5347 	verstime.tv_nsec = 0;
5348 	txdr_nfsv4time(&verstime, tl);
5349 	nd->nd_flag |= ND_USEGSSNAME;
5350 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
5351 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5352 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
5353 	    (int)nd->nd_repstat);
5354 	if (error != 0)
5355 		return (error);
5356 	if (nd->nd_repstat == 0) {
5357 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
5358 		len = fxdr_unsigned(int, *(tl + 7));
5359 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
5360 			error = NFSERR_BADXDR;
5361 			goto nfsmout;
5362 		}
5363 		dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
5364 		    M_WAITOK | M_ZERO);
5365 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
5366 		dsp->nfsclds_servownlen = len;
5367 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
5368 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
5369 		dsp->nfsclds_sess.nfsess_sequenceid =
5370 		    fxdr_unsigned(uint32_t, *tl++);
5371 		v41flags = fxdr_unsigned(uint32_t, *tl);
5372 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
5373 		    NFSHASPNFSOPT(nmp)) {
5374 			NFSCL_DEBUG(1, "set PNFS\n");
5375 			NFSLOCKMNT(nmp);
5376 			nmp->nm_state |= NFSSTA_PNFS;
5377 			NFSUNLOCKMNT(nmp);
5378 			dsp->nfsclds_flags |= NFSCLDS_MDS;
5379 		}
5380 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
5381 			dsp->nfsclds_flags |= NFSCLDS_DS;
5382 		if (minorvers == NFSV42_MINORVERSION)
5383 			dsp->nfsclds_flags |= NFSCLDS_MINORV2;
5384 		if (len > 0)
5385 			nd->nd_repstat = nfsrv_mtostr(nd,
5386 			    dsp->nfsclds_serverown, len);
5387 		if (nd->nd_repstat == 0) {
5388 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5389 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5390 			    NULL, MTX_DEF);
5391 			nfscl_initsessionslots(&dsp->nfsclds_sess);
5392 			*dspp = dsp;
5393 		} else
5394 			free(dsp, M_NFSCLDS);
5395 	}
5396 	error = nd->nd_repstat;
5397 nfsmout:
5398 	m_freem(nd->nd_mrep);
5399 	return (error);
5400 }
5401 
5402 /*
5403  * Do the NFSv4.1 Create Session.
5404  */
5405 int
5406 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
5407     struct nfssockreq *nrp, struct nfsclds *dsp, uint32_t sequenceid, int mds,
5408     struct ucred *cred, NFSPROC_T *p)
5409 {
5410 	uint32_t crflags, maxval, *tl;
5411 	struct nfsrv_descript nfsd;
5412 	struct nfsrv_descript *nd = &nfsd;
5413 	int error, irdcnt, minorvers;
5414 
5415 	/* Make sure nm_rsize, nm_wsize is set. */
5416 	if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
5417 		nmp->nm_rsize = NFS_MAXBSIZE;
5418 	if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
5419 		nmp->nm_wsize = NFS_MAXBSIZE;
5420 	if (dsp == NULL)
5421 		minorvers = nmp->nm_minorvers;
5422 	else if ((dsp->nfsclds_flags & NFSCLDS_MINORV2) != 0)
5423 		minorvers = NFSV42_MINORVERSION;
5424 	else
5425 		minorvers = NFSV41_MINORVERSION;
5426 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL,
5427 	    NFS_VER4, minorvers, NULL);
5428 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5429 	*tl++ = sep->nfsess_clientid.lval[0];
5430 	*tl++ = sep->nfsess_clientid.lval[1];
5431 	*tl++ = txdr_unsigned(sequenceid);
5432 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
5433 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
5434 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
5435 	*tl = txdr_unsigned(crflags);
5436 
5437 	/* Fill in fore channel attributes. */
5438 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5439 	*tl++ = 0;				/* Header pad size */
5440 	if ((nd->nd_flag & ND_NFSV42) != 0 && mds != 0 && sb_max_adj >=
5441 	    nmp->nm_wsize && sb_max_adj >= nmp->nm_rsize) {
5442 		/*
5443 		 * NFSv4.2 Extended Attribute operations may want to do
5444 		 * requests/replies that are larger than nm_rsize/nm_wsize.
5445 		 */
5446 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
5447 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
5448 	} else {
5449 		*tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);
5450 		*tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);
5451 	}
5452 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
5453 	*tl++ = txdr_unsigned(20);		/* Max operations */
5454 	*tl++ = txdr_unsigned(64);		/* Max slots */
5455 	*tl = 0;				/* No rdma ird */
5456 
5457 	/* Fill in back channel attributes. */
5458 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5459 	*tl++ = 0;				/* Header pad size */
5460 	*tl++ = txdr_unsigned(10000);		/* Max request size */
5461 	*tl++ = txdr_unsigned(10000);		/* Max response size */
5462 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
5463 	*tl++ = txdr_unsigned(4);		/* Max operations */
5464 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
5465 	*tl = 0;				/* No rdma ird */
5466 
5467 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
5468 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
5469 
5470 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
5471 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
5472 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
5473 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
5474 	*tl++ = 0;				/* Null machine name */
5475 	*tl++ = 0;				/* Uid == 0 */
5476 	*tl++ = 0;				/* Gid == 0 */
5477 	*tl = 0;				/* No additional gids */
5478 	nd->nd_flag |= ND_USEGSSNAME;
5479 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
5480 	    NFS_VER4, NULL, 1, NULL, NULL);
5481 	if (error != 0)
5482 		return (error);
5483 	if (nd->nd_repstat == 0) {
5484 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
5485 		    2 * NFSX_UNSIGNED);
5486 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
5487 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
5488 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
5489 		crflags = fxdr_unsigned(uint32_t, *tl);
5490 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
5491 			NFSLOCKMNT(nmp);
5492 			nmp->nm_state |= NFSSTA_SESSPERSIST;
5493 			NFSUNLOCKMNT(nmp);
5494 		}
5495 
5496 		/* Get the fore channel slot count. */
5497 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5498 		tl++;			/* Skip the header pad size. */
5499 
5500 		/* Make sure nm_wsize is small enough. */
5501 		maxval = fxdr_unsigned(uint32_t, *tl++);
5502 		while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
5503 			if (nmp->nm_wsize > 8096)
5504 				nmp->nm_wsize /= 2;
5505 			else
5506 				break;
5507 		}
5508 		sep->nfsess_maxreq = maxval;
5509 
5510 		/* Make sure nm_rsize is small enough. */
5511 		maxval = fxdr_unsigned(uint32_t, *tl++);
5512 		while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
5513 			if (nmp->nm_rsize > 8096)
5514 				nmp->nm_rsize /= 2;
5515 			else
5516 				break;
5517 		}
5518 		sep->nfsess_maxresp = maxval;
5519 
5520 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
5521 		tl++;
5522 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
5523 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
5524 		irdcnt = fxdr_unsigned(int, *tl);
5525 		if (irdcnt < 0 || irdcnt > 1) {
5526 			error = NFSERR_BADXDR;
5527 			goto nfsmout;
5528 		}
5529 		if (irdcnt > 0)
5530 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
5531 
5532 		/* and the back channel slot count. */
5533 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5534 		tl += 5;
5535 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
5536 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
5537 	}
5538 	error = nd->nd_repstat;
5539 nfsmout:
5540 	m_freem(nd->nd_mrep);
5541 	return (error);
5542 }
5543 
5544 /*
5545  * Do the NFSv4.1 Destroy Client.
5546  */
5547 int
5548 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
5549     struct ucred *cred, NFSPROC_T *p)
5550 {
5551 	uint32_t *tl;
5552 	struct nfsrv_descript nfsd;
5553 	struct nfsrv_descript *nd = &nfsd;
5554 	int error;
5555 	struct nfsclsession *tsep;
5556 
5557 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
5558 	    0, NULL);
5559 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5560 	tsep = nfsmnt_mdssession(nmp);
5561 	*tl++ = tsep->nfsess_clientid.lval[0];
5562 	*tl = tsep->nfsess_clientid.lval[1];
5563 	nd->nd_flag |= ND_USEGSSNAME;
5564 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5565 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5566 	if (error != 0)
5567 		return (error);
5568 	error = nd->nd_repstat;
5569 	m_freem(nd->nd_mrep);
5570 	return (error);
5571 }
5572 
5573 /*
5574  * Do the NFSv4.1 LayoutGet.
5575  */
5576 static int
5577 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
5578     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
5579     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
5580     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p)
5581 {
5582 	struct nfsrv_descript nfsd, *nd = &nfsd;
5583 	int error;
5584 
5585 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
5586 	    0, cred);
5587 	nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
5588 	    layouttype, layoutlen, 0);
5589 	nd->nd_flag |= ND_USEGSSNAME;
5590 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5591 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5592 	NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
5593 	if (error != 0)
5594 		return (error);
5595 	if (nd->nd_repstat == 0)
5596 		error = nfsrv_parselayoutget(nmp, nd, stateidp, retonclosep,
5597 		    flhp);
5598 	if (error == 0 && nd->nd_repstat != 0)
5599 		error = nd->nd_repstat;
5600 	m_freem(nd->nd_mrep);
5601 	return (error);
5602 }
5603 
5604 /*
5605  * Do the NFSv4.1 Get Device Info.
5606  */
5607 int
5608 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
5609     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
5610     NFSPROC_T *p)
5611 {
5612 	uint32_t cnt, *tl, vers, minorvers;
5613 	struct nfsrv_descript nfsd;
5614 	struct nfsrv_descript *nd = &nfsd;
5615 	struct sockaddr_in sin, ssin;
5616 	struct sockaddr_in6 sin6, ssin6;
5617 	struct nfsclds *dsp = NULL, **dspp, **gotdspp;
5618 	struct nfscldevinfo *ndi;
5619 	int addrcnt = 0, bitcnt, error, gotminor, gotvers, i, isudp, j;
5620 	int stripecnt;
5621 	uint8_t stripeindex;
5622 	sa_family_t af, safilled;
5623 
5624 	ssin.sin_port = 0;		/* To shut up compiler. */
5625 	ssin.sin_addr.s_addr = 0;	/* ditto */
5626 	*ndip = NULL;
5627 	ndi = NULL;
5628 	gotdspp = NULL;
5629 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
5630 	    0, cred);
5631 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5632 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
5633 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5634 	*tl++ = txdr_unsigned(layouttype);
5635 	*tl++ = txdr_unsigned(100000);
5636 	if (notifybitsp != NULL && *notifybitsp != 0) {
5637 		*tl = txdr_unsigned(1);		/* One word of bits. */
5638 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5639 		*tl = txdr_unsigned(*notifybitsp);
5640 	} else
5641 		*tl = txdr_unsigned(0);
5642 	nd->nd_flag |= ND_USEGSSNAME;
5643 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5644 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5645 	if (error != 0)
5646 		return (error);
5647 	if (nd->nd_repstat == 0) {
5648 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5649 		if (layouttype != fxdr_unsigned(int, *tl))
5650 			printf("EEK! devinfo layout type not same!\n");
5651 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
5652 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5653 			stripecnt = fxdr_unsigned(int, *tl);
5654 			NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
5655 			if (stripecnt < 1 || stripecnt > 4096) {
5656 				printf("pNFS File layout devinfo stripecnt %d:"
5657 				    " out of range\n", stripecnt);
5658 				error = NFSERR_BADXDR;
5659 				goto nfsmout;
5660 			}
5661 			NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
5662 			    NFSX_UNSIGNED);
5663 			addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
5664 			NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
5665 			if (addrcnt < 1 || addrcnt > 128) {
5666 				printf("NFS devinfo addrcnt %d: out of range\n",
5667 				    addrcnt);
5668 				error = NFSERR_BADXDR;
5669 				goto nfsmout;
5670 			}
5671 
5672 			/*
5673 			 * Now we know how many stripe indices and addresses, so
5674 			 * we can allocate the structure the correct size.
5675 			 */
5676 			i = (stripecnt * sizeof(uint8_t)) /
5677 			    sizeof(struct nfsclds *) + 1;
5678 			NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5679 			ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5680 			    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5681 			    M_ZERO);
5682 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5683 			    NFSX_V4DEVICEID);
5684 			ndi->nfsdi_refcnt = 0;
5685 			ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5686 			ndi->nfsdi_stripecnt = stripecnt;
5687 			ndi->nfsdi_addrcnt = addrcnt;
5688 			/* Fill in the stripe indices. */
5689 			for (i = 0; i < stripecnt; i++) {
5690 				stripeindex = fxdr_unsigned(uint8_t, *tl++);
5691 				NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5692 				if (stripeindex >= addrcnt) {
5693 					printf("pNFS File Layout devinfo"
5694 					    " stripeindex %d: too big\n",
5695 					    (int)stripeindex);
5696 					error = NFSERR_BADXDR;
5697 					goto nfsmout;
5698 				}
5699 				nfsfldi_setstripeindex(ndi, i, stripeindex);
5700 			}
5701 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
5702 			/* For Flex File, we only get one address list. */
5703 			ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5704 			    M_NFSDEVINFO, M_WAITOK | M_ZERO);
5705 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5706 			    NFSX_V4DEVICEID);
5707 			ndi->nfsdi_refcnt = 0;
5708 			ndi->nfsdi_flags = NFSDI_FLEXFILE;
5709 			addrcnt = ndi->nfsdi_addrcnt = 1;
5710 		}
5711 
5712 		/* Now, dissect the server address(es). */
5713 		safilled = AF_UNSPEC;
5714 		for (i = 0; i < addrcnt; i++) {
5715 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5716 			cnt = fxdr_unsigned(uint32_t, *tl);
5717 			if (cnt == 0) {
5718 				printf("NFS devinfo 0 len addrlist\n");
5719 				error = NFSERR_BADXDR;
5720 				goto nfsmout;
5721 			}
5722 			dspp = nfsfldi_addr(ndi, i);
5723 			safilled = AF_UNSPEC;
5724 			for (j = 0; j < cnt; j++) {
5725 				error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5726 				    &isudp);
5727 				if (error != 0 && error != EPERM) {
5728 					error = NFSERR_BADXDR;
5729 					goto nfsmout;
5730 				}
5731 				if (error == 0 && isudp == 0) {
5732 					/*
5733 					 * The priority is:
5734 					 * - Same address family.
5735 					 * Save the address and dspp, so that
5736 					 * the connection can be done after
5737 					 * parsing is complete.
5738 					 */
5739 					if (safilled == AF_UNSPEC ||
5740 					    (af == nmp->nm_nam->sa_family &&
5741 					     safilled != nmp->nm_nam->sa_family)
5742 					   ) {
5743 						if (af == AF_INET)
5744 							ssin = sin;
5745 						else
5746 							ssin6 = sin6;
5747 						safilled = af;
5748 						gotdspp = dspp;
5749 					}
5750 				}
5751 			}
5752 		}
5753 
5754 		gotvers = NFS_VER4;	/* Default NFSv4.1 for File Layout. */
5755 		gotminor = NFSV41_MINORVERSION;
5756 		/* For Flex File, we will take one of the versions to use. */
5757 		if (layouttype == NFSLAYOUT_FLEXFILE) {
5758 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5759 			j = fxdr_unsigned(int, *tl);
5760 			if (j < 1 || j > NFSDEV_MAXVERS) {
5761 				printf("pNFS: too many versions\n");
5762 				error = NFSERR_BADXDR;
5763 				goto nfsmout;
5764 			}
5765 			gotvers = 0;
5766 			gotminor = 0;
5767 			for (i = 0; i < j; i++) {
5768 				NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5769 				vers = fxdr_unsigned(uint32_t, *tl++);
5770 				minorvers = fxdr_unsigned(uint32_t, *tl++);
5771 				if (vers == NFS_VER3)
5772 					minorvers = 0;
5773 				if ((vers == NFS_VER4 && ((minorvers ==
5774 				    NFSV41_MINORVERSION && gotminor == 0) ||
5775 				    minorvers == NFSV42_MINORVERSION)) ||
5776 				    (vers == NFS_VER3 && gotvers == 0)) {
5777 					gotvers = vers;
5778 					gotminor = minorvers;
5779 					/* We'll take this one. */
5780 					ndi->nfsdi_versindex = i;
5781 					ndi->nfsdi_vers = vers;
5782 					ndi->nfsdi_minorvers = minorvers;
5783 					ndi->nfsdi_rsize = fxdr_unsigned(
5784 					    uint32_t, *tl++);
5785 					ndi->nfsdi_wsize = fxdr_unsigned(
5786 					    uint32_t, *tl++);
5787 					if (*tl == newnfs_true)
5788 						ndi->nfsdi_flags |=
5789 						    NFSDI_TIGHTCOUPLED;
5790 					else
5791 						ndi->nfsdi_flags &=
5792 						    ~NFSDI_TIGHTCOUPLED;
5793 				}
5794 			}
5795 			if (gotvers == 0) {
5796 				printf("pNFS: no NFSv3, NFSv4.1 or NFSv4.2\n");
5797 				error = NFSERR_BADXDR;
5798 				goto nfsmout;
5799 			}
5800 		}
5801 
5802 		/* And the notify bits. */
5803 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5804 		bitcnt = fxdr_unsigned(int, *tl);
5805 		if (bitcnt > 0) {
5806 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5807 			if (notifybitsp != NULL)
5808 				*notifybitsp =
5809 				    fxdr_unsigned(uint32_t, *tl);
5810 		}
5811 		if (safilled != AF_UNSPEC) {
5812 			KASSERT(ndi != NULL, ("ndi is NULL"));
5813 			*ndip = ndi;
5814 		} else
5815 			error = EPERM;
5816 		if (error == 0) {
5817 			/*
5818 			 * Now we can do a TCP connection for the correct
5819 			 * NFS version and IP address.
5820 			 */
5821 			error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5822 			    gotvers, gotminor, &dsp, p);
5823 		}
5824 		if (error == 0) {
5825 			KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5826 			*gotdspp = dsp;
5827 		}
5828 	}
5829 	if (nd->nd_repstat != 0 && error == 0)
5830 		error = nd->nd_repstat;
5831 nfsmout:
5832 	if (error != 0 && ndi != NULL)
5833 		nfscl_freedevinfo(ndi);
5834 	m_freem(nd->nd_mrep);
5835 	return (error);
5836 }
5837 
5838 /*
5839  * Do the NFSv4.1 LayoutCommit.
5840  */
5841 int
5842 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5843     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5844     int layouttype, struct ucred *cred, NFSPROC_T *p)
5845 {
5846 	uint32_t *tl;
5847 	struct nfsrv_descript nfsd, *nd = &nfsd;
5848 	int error;
5849 
5850 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5851 	    0, 0, cred);
5852 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5853 	    NFSX_STATEID);
5854 	txdr_hyper(off, tl);
5855 	tl += 2;
5856 	txdr_hyper(len, tl);
5857 	tl += 2;
5858 	if (reclaim != 0)
5859 		*tl++ = newnfs_true;
5860 	else
5861 		*tl++ = newnfs_false;
5862 	*tl++ = txdr_unsigned(stateidp->seqid);
5863 	*tl++ = stateidp->other[0];
5864 	*tl++ = stateidp->other[1];
5865 	*tl++ = stateidp->other[2];
5866 	*tl++ = newnfs_true;
5867 	if (lastbyte < off)
5868 		lastbyte = off;
5869 	else if (lastbyte >= (off + len))
5870 		lastbyte = off + len - 1;
5871 	txdr_hyper(lastbyte, tl);
5872 	tl += 2;
5873 	*tl++ = newnfs_false;
5874 	*tl++ = txdr_unsigned(layouttype);
5875 	/* All supported layouts are 0 length. */
5876 	*tl = txdr_unsigned(0);
5877 	nd->nd_flag |= ND_USEGSSNAME;
5878 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5879 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5880 	if (error != 0)
5881 		return (error);
5882 	error = nd->nd_repstat;
5883 	m_freem(nd->nd_mrep);
5884 	return (error);
5885 }
5886 
5887 /*
5888  * Do the NFSv4.1 LayoutReturn.
5889  */
5890 int
5891 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5892     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5893     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5894     uint32_t stat, uint32_t op, char *devid)
5895 {
5896 	uint32_t *tl;
5897 	struct nfsrv_descript nfsd, *nd = &nfsd;
5898 	uint64_t tu64;
5899 	int error;
5900 
5901 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5902 	    0, 0, cred);
5903 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5904 	if (reclaim != 0)
5905 		*tl++ = newnfs_true;
5906 	else
5907 		*tl++ = newnfs_false;
5908 	*tl++ = txdr_unsigned(layouttype);
5909 	*tl++ = txdr_unsigned(iomode);
5910 	*tl = txdr_unsigned(layoutreturn);
5911 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5912 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5913 		    NFSX_UNSIGNED);
5914 		txdr_hyper(offset, tl);
5915 		tl += 2;
5916 		txdr_hyper(len, tl);
5917 		tl += 2;
5918 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5919 		*tl++ = txdr_unsigned(stateidp->seqid);
5920 		*tl++ = stateidp->other[0];
5921 		*tl++ = stateidp->other[1];
5922 		*tl++ = stateidp->other[2];
5923 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5924 			*tl = txdr_unsigned(0);
5925 		else if (layouttype == NFSLAYOUT_FLEXFILE) {
5926 			if (stat != 0) {
5927 				*tl = txdr_unsigned(2 * NFSX_HYPER +
5928 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5929 				    NFSX_UNSIGNED);
5930 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5931 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5932 				    NFSX_UNSIGNED);
5933 				*tl++ = txdr_unsigned(1);	/* One error. */
5934 				tu64 = 0;			/* Offset. */
5935 				txdr_hyper(tu64, tl); tl += 2;
5936 				tu64 = UINT64_MAX;		/* Length. */
5937 				txdr_hyper(tu64, tl); tl += 2;
5938 				NFSBCOPY(stateidp, tl, NFSX_STATEID);
5939 				tl += (NFSX_STATEID / NFSX_UNSIGNED);
5940 				*tl++ = txdr_unsigned(1);	/* One error. */
5941 				NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5942 				tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5943 				*tl++ = txdr_unsigned(stat);
5944 				*tl++ = txdr_unsigned(op);
5945 			} else {
5946 				*tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5947 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5948 				/* No ioerrs. */
5949 				*tl++ = 0;
5950 			}
5951 			*tl = 0;	/* No stats yet. */
5952 		}
5953 	}
5954 	nd->nd_flag |= ND_USEGSSNAME;
5955 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5956 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5957 	if (error != 0)
5958 		return (error);
5959 	if (nd->nd_repstat == 0) {
5960 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5961 		if (*tl != 0) {
5962 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5963 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5964 			stateidp->other[0] = *tl++;
5965 			stateidp->other[1] = *tl++;
5966 			stateidp->other[2] = *tl;
5967 		}
5968 	} else
5969 		error = nd->nd_repstat;
5970 nfsmout:
5971 	m_freem(nd->nd_mrep);
5972 	return (error);
5973 }
5974 
5975 /*
5976  * Do the NFSv4.2 LayoutError.
5977  */
5978 static int
5979 nfsrpc_layouterror(struct nfsmount *nmp, uint8_t *fh, int fhlen, uint64_t offset,
5980     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5981     uint32_t stat, uint32_t op, char *devid)
5982 {
5983 	uint32_t *tl;
5984 	struct nfsrv_descript nfsd, *nd = &nfsd;
5985 	int error;
5986 
5987 	nfscl_reqstart(nd, NFSPROC_LAYOUTERROR, nmp, fh, fhlen, NULL, NULL,
5988 	    0, 0, cred);
5989 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5990 	    NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5991 	txdr_hyper(offset, tl); tl += 2;
5992 	txdr_hyper(len, tl); tl += 2;
5993 	*tl++ = txdr_unsigned(stateidp->seqid);
5994 	*tl++ = stateidp->other[0];
5995 	*tl++ = stateidp->other[1];
5996 	*tl++ = stateidp->other[2];
5997 	*tl++ = txdr_unsigned(1);
5998 	NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5999 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
6000 	*tl++ = txdr_unsigned(stat);
6001 	*tl = txdr_unsigned(op);
6002 	nd->nd_flag |= ND_USEGSSNAME;
6003 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
6004 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
6005 	if (error != 0)
6006 		return (error);
6007 	if (nd->nd_repstat != 0)
6008 		error = nd->nd_repstat;
6009 	m_freem(nd->nd_mrep);
6010 	return (error);
6011 }
6012 
6013 /*
6014  * Acquire a layout and devinfo, if possible. The caller must have acquired
6015  * a reference count on the nfsclclient structure before calling this.
6016  * Return the layout in lypp with a reference count on it, if successful.
6017  */
6018 static int
6019 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
6020     int iomode, uint32_t rw, uint32_t *notifybitsp, nfsv4stateid_t *stateidp,
6021     uint64_t off, struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
6022 {
6023 	struct nfscllayout *lyp;
6024 	struct nfsclflayout *flp;
6025 	struct nfsclflayouthead flh;
6026 	int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
6027 	nfsv4stateid_t stateid;
6028 	struct nfsclsession *tsep;
6029 
6030 	*lypp = NULL;
6031 	if (NFSHASFLEXFILE(nmp))
6032 		layouttype = NFSLAYOUT_FLEXFILE;
6033 	else
6034 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
6035 	/*
6036 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
6037 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
6038 	 * flp == NULL.
6039 	 */
6040 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
6041 	    off, rw, &flp, &recalled);
6042 	islocked = 0;
6043 	if (lyp == NULL || flp == NULL) {
6044 		if (recalled != 0)
6045 			return (EIO);
6046 		LIST_INIT(&flh);
6047 		tsep = nfsmnt_mdssession(nmp);
6048 		layoutlen = tsep->nfsess_maxcache -
6049 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
6050 		if (lyp == NULL) {
6051 			stateid.seqid = 0;
6052 			stateid.other[0] = stateidp->other[0];
6053 			stateid.other[1] = stateidp->other[1];
6054 			stateid.other[2] = stateidp->other[2];
6055 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
6056 			    nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
6057 			    (uint64_t)0, layouttype, layoutlen, &stateid,
6058 			    &retonclose, &flh, cred, p);
6059 		} else {
6060 			islocked = 1;
6061 			stateid.seqid = lyp->nfsly_stateid.seqid;
6062 			stateid.other[0] = lyp->nfsly_stateid.other[0];
6063 			stateid.other[1] = lyp->nfsly_stateid.other[1];
6064 			stateid.other[2] = lyp->nfsly_stateid.other[2];
6065 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
6066 			    nfhp->nfh_len, iomode, off, UINT64_MAX,
6067 			    (uint64_t)0, layouttype, layoutlen, &stateid,
6068 			    &retonclose, &flh, cred, p);
6069 		}
6070 		error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
6071 		    nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
6072 		    &flh, layouttype, error, NULL, cred, p);
6073 		if (error == 0)
6074 			*lypp = lyp;
6075 		else if (islocked != 0)
6076 			nfscl_rellayout(lyp, 1);
6077 	} else
6078 		*lypp = lyp;
6079 	return (error);
6080 }
6081 
6082 /*
6083  * Do a TCP connection plus exchange id and create session.
6084  * If successful, a "struct nfsclds" is linked into the list for the
6085  * mount point and a pointer to it is returned.
6086  */
6087 static int
6088 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
6089     struct sockaddr_in6 *sin6, sa_family_t af, int vers, int minorvers,
6090     struct nfsclds **dspp, NFSPROC_T *p)
6091 {
6092 	struct sockaddr_in *msad, *sad;
6093 	struct sockaddr_in6 *msad6, *sad6;
6094 	struct nfsclclient *clp;
6095 	struct nfssockreq *nrp;
6096 	struct nfsclds *dsp, *tdsp;
6097 	int error, firsttry;
6098 	enum nfsclds_state retv;
6099 	uint32_t sequenceid = 0;
6100 
6101 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
6102 	    ("nfsrpc_fillsa: NULL nr_cred"));
6103 	NFSLOCKCLSTATE();
6104 	clp = nmp->nm_clp;
6105 	NFSUNLOCKCLSTATE();
6106 	if (clp == NULL)
6107 		return (EPERM);
6108 	if (af == AF_INET) {
6109 		NFSLOCKMNT(nmp);
6110 		/*
6111 		 * Check to see if we already have a session for this
6112 		 * address that is usable for a DS.
6113 		 * Note that the MDS's address is in a different place
6114 		 * than the sessions already acquired for DS's.
6115 		 */
6116 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
6117 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
6118 		while (tdsp != NULL) {
6119 			if (msad != NULL && msad->sin_family == AF_INET &&
6120 			    sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
6121 			    sin->sin_port == msad->sin_port &&
6122 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
6123 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
6124 				*dspp = tdsp;
6125 				NFSUNLOCKMNT(nmp);
6126 				NFSCL_DEBUG(4, "fnd same addr\n");
6127 				return (0);
6128 			}
6129 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
6130 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
6131 				msad = (struct sockaddr_in *)
6132 				    tdsp->nfsclds_sockp->nr_nam;
6133 			else
6134 				msad = NULL;
6135 		}
6136 		NFSUNLOCKMNT(nmp);
6137 
6138 		/* No IP address match, so look for new/trunked one. */
6139 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
6140 		sad->sin_len = sizeof(*sad);
6141 		sad->sin_family = AF_INET;
6142 		sad->sin_port = sin->sin_port;
6143 		sad->sin_addr.s_addr = sin->sin_addr.s_addr;
6144 		if (NFSHASPNFS(nmp) && NFSHASKERB(nmp)) {
6145 			/* For pNFS, a separate server principal is needed. */
6146 			nrp = malloc(sizeof(*nrp) + NI_MAXSERV + NI_MAXHOST,
6147 			    M_NFSSOCKREQ, M_WAITOK | M_ZERO);
6148 			/*
6149 			 * Use the latter part of nr_srvprinc as a temporary
6150 			 * buffer for the IP address.
6151 			 */
6152 			inet_ntoa_r(sad->sin_addr,
6153 			    &nrp->nr_srvprinc[NI_MAXSERV]);
6154 			NFSCL_DEBUG(1, "nfsrpc_fillsa: DS IP=%s\n",
6155 			    &nrp->nr_srvprinc[NI_MAXSERV]);
6156 			if (!rpc_gss_ip_to_srv_principal_call(
6157 			    &nrp->nr_srvprinc[NI_MAXSERV], "nfs",
6158 			    nrp->nr_srvprinc))
6159 				nrp->nr_srvprinc[0] = '\0';
6160 			NFSCL_DEBUG(1, "nfsrpc_fillsa: srv principal=%s\n",
6161 			    nrp->nr_srvprinc);
6162 		} else
6163 			nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ,
6164 			    M_WAITOK | M_ZERO);
6165 		nrp->nr_nam = (struct sockaddr *)sad;
6166 	} else if (af == AF_INET6) {
6167 		NFSLOCKMNT(nmp);
6168 		/*
6169 		 * Check to see if we already have a session for this
6170 		 * address that is usable for a DS.
6171 		 * Note that the MDS's address is in a different place
6172 		 * than the sessions already acquired for DS's.
6173 		 */
6174 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
6175 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
6176 		while (tdsp != NULL) {
6177 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
6178 			    IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
6179 			    &msad6->sin6_addr) &&
6180 			    sin6->sin6_port == msad6->sin6_port &&
6181 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
6182 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
6183 				*dspp = tdsp;
6184 				NFSUNLOCKMNT(nmp);
6185 				return (0);
6186 			}
6187 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
6188 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
6189 				msad6 = (struct sockaddr_in6 *)
6190 				    tdsp->nfsclds_sockp->nr_nam;
6191 			else
6192 				msad6 = NULL;
6193 		}
6194 		NFSUNLOCKMNT(nmp);
6195 
6196 		/* No IP address match, so look for new/trunked one. */
6197 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
6198 		sad6->sin6_len = sizeof(*sad6);
6199 		sad6->sin6_family = AF_INET6;
6200 		sad6->sin6_port = sin6->sin6_port;
6201 		NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
6202 		    sizeof(struct in6_addr));
6203 		if (NFSHASPNFS(nmp) && NFSHASKERB(nmp)) {
6204 			/* For pNFS, a separate server principal is needed. */
6205 			nrp = malloc(sizeof(*nrp) + NI_MAXSERV + NI_MAXHOST,
6206 			    M_NFSSOCKREQ, M_WAITOK | M_ZERO);
6207 			/*
6208 			 * Use the latter part of nr_srvprinc as a temporary
6209 			 * buffer for the IP address.
6210 			 */
6211 			inet_ntop(AF_INET6, &sad6->sin6_addr,
6212 			    &nrp->nr_srvprinc[NI_MAXSERV], NI_MAXHOST);
6213 			NFSCL_DEBUG(1, "nfsrpc_fillsa: DS IP=%s\n",
6214 			    &nrp->nr_srvprinc[NI_MAXSERV]);
6215 			if (!rpc_gss_ip_to_srv_principal_call(
6216 			    &nrp->nr_srvprinc[NI_MAXSERV], "nfs",
6217 			    nrp->nr_srvprinc))
6218 				nrp->nr_srvprinc[0] = '\0';
6219 			NFSCL_DEBUG(1, "nfsrpc_fillsa: srv principal=%s\n",
6220 			    nrp->nr_srvprinc);
6221 		} else
6222 			nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ,
6223 			    M_WAITOK | M_ZERO);
6224 		nrp->nr_nam = (struct sockaddr *)sad6;
6225 	} else
6226 		return (EPERM);
6227 
6228 	nrp->nr_sotype = SOCK_STREAM;
6229 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
6230 	nrp->nr_prog = NFS_PROG;
6231 	nrp->nr_vers = vers;
6232 
6233 	/*
6234 	 * Use the credentials that were used for the mount, which are
6235 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
6236 	 * Ref. counting the credentials with crhold() is probably not
6237 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
6238 	 * unmount, but I did it anyhow.
6239 	 */
6240 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
6241 	error = newnfs_connect(nmp, nrp, NULL, p, 0, false, &nrp->nr_client);
6242 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
6243 
6244 	dsp = NULL;
6245 	/* Now, do the exchangeid and create session. */
6246 	if (error == 0) {
6247 		if (vers == NFS_VER4) {
6248 			firsttry = 0;
6249 			do {
6250 				error = nfsrpc_exchangeid(nmp, clp, nrp,
6251 				    minorvers, NFSV4EXCH_USEPNFSDS, &dsp,
6252 				    nrp->nr_cred, p);
6253 				NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
6254 				if (error == NFSERR_MINORVERMISMATCH)
6255 					minorvers = NFSV42_MINORVERSION;
6256 			} while (error == NFSERR_MINORVERMISMATCH &&
6257 			    firsttry++ == 0);
6258 			if (error != 0)
6259 				newnfs_disconnect(NULL, nrp);
6260 		} else {
6261 			dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
6262 			    M_WAITOK | M_ZERO);
6263 			dsp->nfsclds_flags |= NFSCLDS_DS;
6264 			dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
6265 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
6266 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
6267 			    NULL, MTX_DEF);
6268 		}
6269 	}
6270 	if (error == 0) {
6271 		dsp->nfsclds_sockp = nrp;
6272 		if (vers == NFS_VER4) {
6273 			NFSLOCKMNT(nmp);
6274 			retv = nfscl_getsameserver(nmp, dsp, &tdsp,
6275 			    &sequenceid);
6276 			NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
6277 			if (retv == NFSDSP_USETHISSESSION &&
6278 			    nfscl_dssameconn != 0) {
6279 				NFSLOCKDS(tdsp);
6280 				tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
6281 				NFSUNLOCKDS(tdsp);
6282 				NFSUNLOCKMNT(nmp);
6283 				/*
6284 				 * If there is already a session for this
6285 				 * server, use it.
6286 				 */
6287 				newnfs_disconnect(NULL, nrp);
6288 				nfscl_freenfsclds(dsp);
6289 				*dspp = tdsp;
6290 				return (0);
6291 			}
6292 			if (retv == NFSDSP_NOTFOUND)
6293 				sequenceid =
6294 				    dsp->nfsclds_sess.nfsess_sequenceid;
6295 			NFSUNLOCKMNT(nmp);
6296 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
6297 			    nrp, dsp, sequenceid, 0, nrp->nr_cred, p);
6298 			NFSCL_DEBUG(3, "DS createsess=%d\n", error);
6299 		}
6300 	} else {
6301 		NFSFREECRED(nrp->nr_cred);
6302 		NFSFREEMUTEX(&nrp->nr_mtx);
6303 		free(nrp->nr_nam, M_SONAME);
6304 		free(nrp, M_NFSSOCKREQ);
6305 	}
6306 	if (error == 0) {
6307 		NFSCL_DEBUG(3, "add DS session\n");
6308 		/*
6309 		 * Put it at the end of the list. That way the list
6310 		 * is ordered by when the entry was added. This matters
6311 		 * since the one done first is the one that should be
6312 		 * used for sequencid'ing any subsequent create sessions.
6313 		 */
6314 		NFSLOCKMNT(nmp);
6315 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
6316 		NFSUNLOCKMNT(nmp);
6317 		*dspp = dsp;
6318 	} else if (dsp != NULL) {
6319 		newnfs_disconnect(NULL, nrp);
6320 		nfscl_freenfsclds(dsp);
6321 	}
6322 	return (error);
6323 }
6324 
6325 /*
6326  * Do the NFSv4.1 Reclaim Complete.
6327  */
6328 int
6329 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
6330 {
6331 	uint32_t *tl;
6332 	struct nfsrv_descript nfsd;
6333 	struct nfsrv_descript *nd = &nfsd;
6334 	int error;
6335 
6336 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
6337 	    0, cred);
6338 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
6339 	*tl = newnfs_false;
6340 	nd->nd_flag |= ND_USEGSSNAME;
6341 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
6342 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
6343 	if (error != 0)
6344 		return (error);
6345 	error = nd->nd_repstat;
6346 	m_freem(nd->nd_mrep);
6347 	return (error);
6348 }
6349 
6350 /*
6351  * Initialize the slot tables for a session.
6352  */
6353 static void
6354 nfscl_initsessionslots(struct nfsclsession *sep)
6355 {
6356 	int i;
6357 
6358 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
6359 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
6360 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
6361 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
6362 	}
6363 	for (i = 0; i < 64; i++)
6364 		sep->nfsess_slotseq[i] = 0;
6365 	sep->nfsess_slots = 0;
6366 	sep->nfsess_badslots = 0;
6367 }
6368 
6369 /*
6370  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
6371  */
6372 int
6373 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6374     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
6375 {
6376 	struct nfsnode *np = VTONFS(vp);
6377 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6378 	struct nfscllayout *layp;
6379 	struct nfscldevinfo *dip;
6380 	struct nfsclflayout *rflp;
6381 	struct mbuf *m, *m2;
6382 	struct nfsclwritedsdorpc *drpc, *tdrpc;
6383 	nfsv4stateid_t stateid;
6384 	struct ucred *newcred;
6385 	uint64_t lastbyte, len, off, oresid, xfer;
6386 	int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
6387 	void *lckp;
6388 	uint8_t *dev;
6389 	void *iovbase = NULL;
6390 	size_t iovlen = 0;
6391 	off_t offs = 0;
6392 	ssize_t resid = 0;
6393 	uint32_t op;
6394 
6395 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
6396 	    (np->n_flag & NNOLAYOUT) != 0)
6397 		return (EIO);
6398 	/* Now, get a reference cnt on the clientid for this mount. */
6399 	if (nfscl_getref(nmp) == 0)
6400 		return (EIO);
6401 
6402 	/* Find an appropriate stateid. */
6403 	newcred = NFSNEWCRED(cred);
6404 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
6405 	    rwaccess, 1, newcred, p, &stateid, &lckp);
6406 	if (error != 0) {
6407 		NFSFREECRED(newcred);
6408 		nfscl_relref(nmp);
6409 		return (error);
6410 	}
6411 	/* Search for a layout for this file. */
6412 	off = uiop->uio_offset;
6413 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
6414 	    np->n_fhp->nfh_len, off, rwaccess, &rflp, &recalled);
6415 	if (layp == NULL || rflp == NULL) {
6416 		if (recalled != 0) {
6417 			NFSFREECRED(newcred);
6418 			if (lckp != NULL)
6419 				nfscl_lockderef(lckp);
6420 			nfscl_relref(nmp);
6421 			return (EIO);
6422 		}
6423 		if (layp != NULL) {
6424 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
6425 			layp = NULL;
6426 		}
6427 		/* Try and get a Layout, if it is supported. */
6428 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
6429 		    (np->n_flag & NWRITEOPENED) != 0)
6430 			iolaymode = NFSLAYOUTIOMODE_RW;
6431 		else
6432 			iolaymode = NFSLAYOUTIOMODE_READ;
6433 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
6434 		    rwaccess, NULL, &stateid, off, &layp, newcred, p);
6435 		if (error != 0) {
6436 			NFSLOCKNODE(np);
6437 			np->n_flag |= NNOLAYOUT;
6438 			NFSUNLOCKNODE(np);
6439 			if (lckp != NULL)
6440 				nfscl_lockderef(lckp);
6441 			NFSFREECRED(newcred);
6442 			if (layp != NULL)
6443 				nfscl_rellayout(layp, 0);
6444 			nfscl_relref(nmp);
6445 			return (error);
6446 		}
6447 	}
6448 
6449 	/*
6450 	 * Loop around finding a layout that works for the first part of
6451 	 * this I/O operation, and then call the function that actually
6452 	 * does the RPC.
6453 	 */
6454 	eof = 0;
6455 	len = (uint64_t)uiop->uio_resid;
6456 	while (len > 0 && error == 0 && eof == 0) {
6457 		off = uiop->uio_offset;
6458 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
6459 		if (error == 0) {
6460 			oresid = xfer = (uint64_t)uiop->uio_resid;
6461 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
6462 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
6463 			/*
6464 			 * For Flex File layout with mirrored DSs, select one
6465 			 * of them at random for reads. For writes and commits,
6466 			 * do all mirrors.
6467 			 */
6468 			m = NULL;
6469 			tdrpc = drpc = NULL;
6470 			firstmirror = 0;
6471 			mirrorcnt = 1;
6472 			if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
6473 			    (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
6474 				if (rwaccess == NFSV4OPEN_ACCESSREAD) {
6475 					firstmirror = arc4random() % mirrorcnt;
6476 					mirrorcnt = firstmirror + 1;
6477 				} else {
6478 					if (docommit == 0) {
6479 						/*
6480 						 * Save values, so uiop can be
6481 						 * rolled back upon a write
6482 						 * error.
6483 						 */
6484 						offs = uiop->uio_offset;
6485 						resid = uiop->uio_resid;
6486 						iovbase =
6487 						    uiop->uio_iov->iov_base;
6488 						iovlen = uiop->uio_iov->iov_len;
6489 						m = nfsm_uiombuflist(uiop, len,
6490 						    0);
6491 						if (m == NULL) {
6492 							error = EFAULT;
6493 							break;
6494 						}
6495 					}
6496 					tdrpc = drpc = malloc(sizeof(*drpc) *
6497 					    (mirrorcnt - 1), M_TEMP, M_WAITOK |
6498 					    M_ZERO);
6499 				}
6500 			}
6501 			for (i = firstmirror; i < mirrorcnt && error == 0; i++){
6502 				m2 = NULL;
6503 				if (m != NULL && i < mirrorcnt - 1)
6504 					m2 = m_copym(m, 0, M_COPYALL, M_WAITOK);
6505 				else {
6506 					m2 = m;
6507 					m = NULL;
6508 				}
6509 				if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
6510 					dev = rflp->nfsfl_ffm[i].dev;
6511 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
6512 					    rflp->nfsfl_ffm[i].devp);
6513 				} else {
6514 					dev = rflp->nfsfl_dev;
6515 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
6516 					    rflp->nfsfl_devp);
6517 				}
6518 				if (dip != NULL) {
6519 					if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
6520 					    != 0)
6521 						error = nfscl_dofflayoutio(vp,
6522 						    uiop, iomode, must_commit,
6523 						    &eof, &stateid, rwaccess,
6524 						    dip, layp, rflp, off, xfer,
6525 						    i, docommit, m2, tdrpc,
6526 						    newcred, p);
6527 					else
6528 						error = nfscl_doflayoutio(vp,
6529 						    uiop, iomode, must_commit,
6530 						    &eof, &stateid, rwaccess,
6531 						    dip, layp, rflp, off, xfer,
6532 						    docommit, newcred, p);
6533 					nfscl_reldevinfo(dip);
6534 				} else {
6535 					if (m2 != NULL)
6536 						m_freem(m2);
6537 					error = EIO;
6538 				}
6539 				tdrpc++;
6540 			}
6541 			if (m != NULL)
6542 				m_freem(m);
6543 			tdrpc = drpc;
6544 			timo = hz / 50;		/* Wait for 20msec. */
6545 			if (timo < 1)
6546 				timo = 1;
6547 			for (i = firstmirror; i < mirrorcnt - 1 &&
6548 			    tdrpc != NULL; i++, tdrpc++) {
6549 				/*
6550 				 * For the unused drpc entries, both inprog and
6551 				 * err == 0, so this loop won't break.
6552 				 */
6553 				while (tdrpc->inprog != 0 && tdrpc->done == 0)
6554 					tsleep(&tdrpc->tsk, PVFS, "clrpcio",
6555 					    timo);
6556 				if (error == 0 && tdrpc->err != 0)
6557 					error = tdrpc->err;
6558 				if (rwaccess != NFSV4OPEN_ACCESSREAD &&
6559 				    docommit == 0 && *must_commit == 0 &&
6560 				    tdrpc->must_commit == 1)
6561 					*must_commit = 1;
6562 			}
6563 			free(drpc, M_TEMP);
6564 			if (error == 0) {
6565 				if (mirrorcnt > 1 && rwaccess ==
6566 				    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
6567 					NFSLOCKCLSTATE();
6568 					layp->nfsly_flags |= NFSLY_WRITTEN;
6569 					NFSUNLOCKCLSTATE();
6570 				}
6571 				lastbyte = off + xfer - 1;
6572 				NFSLOCKCLSTATE();
6573 				if (lastbyte > layp->nfsly_lastbyte)
6574 					layp->nfsly_lastbyte = lastbyte;
6575 				NFSUNLOCKCLSTATE();
6576 			} else if (error == NFSERR_OPENMODE &&
6577 			    rwaccess == NFSV4OPEN_ACCESSREAD) {
6578 				NFSLOCKMNT(nmp);
6579 				nmp->nm_state |= NFSSTA_OPENMODE;
6580 				NFSUNLOCKMNT(nmp);
6581 			} else if ((error == NFSERR_NOSPC ||
6582 			    error == NFSERR_IO || error == NFSERR_NXIO) &&
6583 			    nmp->nm_minorvers == NFSV42_MINORVERSION) {
6584 				if (docommit != 0)
6585 					op = NFSV4OP_COMMIT;
6586 				else if (rwaccess == NFSV4OPEN_ACCESSREAD)
6587 					op = NFSV4OP_READ;
6588 				else
6589 					op = NFSV4OP_WRITE;
6590 				nfsrpc_layouterror(nmp, np->n_fhp->nfh_fh,
6591 				    np->n_fhp->nfh_len, off, xfer,
6592 				    &layp->nfsly_stateid, newcred, p, error, op,
6593 				    dip->nfsdi_deviceid);
6594 				error = EIO;
6595 			} else
6596 				error = EIO;
6597 			if (error == 0)
6598 				len -= (oresid - (uint64_t)uiop->uio_resid);
6599 			else if (mirrorcnt > 1 && rwaccess ==
6600 			    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
6601 				/*
6602 				 * In case the rpc gets retried, roll the
6603 				 * uio fields changed by nfsm_uiombuflist()
6604 				 * back.
6605 				 */
6606 				uiop->uio_offset = offs;
6607 				uiop->uio_resid = resid;
6608 				uiop->uio_iov->iov_base = iovbase;
6609 				uiop->uio_iov->iov_len = iovlen;
6610 			}
6611 		}
6612 	}
6613 	if (lckp != NULL)
6614 		nfscl_lockderef(lckp);
6615 	NFSFREECRED(newcred);
6616 	nfscl_rellayout(layp, 0);
6617 	nfscl_relref(nmp);
6618 	return (error);
6619 }
6620 
6621 /*
6622  * Find a file layout that will handle the first bytes of the requested
6623  * range and return the information from it needed to the I/O operation.
6624  */
6625 int
6626 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
6627     struct nfsclflayout **retflpp)
6628 {
6629 	struct nfsclflayout *flp, *nflp, *rflp;
6630 	uint32_t rw;
6631 
6632 	rflp = NULL;
6633 	rw = rwaccess;
6634 	/* For reading, do the Read list first and then the Write list. */
6635 	do {
6636 		if (rw == NFSV4OPEN_ACCESSREAD)
6637 			flp = LIST_FIRST(&lyp->nfsly_flayread);
6638 		else
6639 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
6640 		while (flp != NULL) {
6641 			nflp = LIST_NEXT(flp, nfsfl_list);
6642 			if (flp->nfsfl_off > off)
6643 				break;
6644 			if (flp->nfsfl_end > off &&
6645 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
6646 				rflp = flp;
6647 			flp = nflp;
6648 		}
6649 		if (rw == NFSV4OPEN_ACCESSREAD)
6650 			rw = NFSV4OPEN_ACCESSWRITE;
6651 		else
6652 			rw = 0;
6653 	} while (rw != 0);
6654 	if (rflp != NULL) {
6655 		/* This one covers the most bytes starting at off. */
6656 		*retflpp = rflp;
6657 		return (0);
6658 	}
6659 	return (EIO);
6660 }
6661 
6662 /*
6663  * Do I/O using an NFSv4.1 or NFSv4.2 file layout.
6664  */
6665 static int
6666 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6667     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6668     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6669     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
6670 {
6671 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
6672 	int commit_thru_mds, error, stripe_index, stripe_pos, minorvers;
6673 	struct nfsnode *np;
6674 	struct nfsfh *fhp;
6675 	struct nfsclds **dspp;
6676 
6677 	np = VTONFS(vp);
6678 	rel_off = off - flp->nfsfl_patoff;
6679 	stripe_unit_size = flp->nfsfl_util & NFSFLAYUTIL_STRIPE_MASK;
6680 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
6681 	    dp->nfsdi_stripecnt;
6682 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
6683 	error = 0;
6684 
6685 	/* Loop around, doing I/O for each stripe unit. */
6686 	while (len > 0 && error == 0) {
6687 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
6688 		dspp = nfsfldi_addr(dp, stripe_index);
6689 		if (((*dspp)->nfsclds_flags & NFSCLDS_MINORV2) != 0)
6690 			minorvers = NFSV42_MINORVERSION;
6691 		else
6692 			minorvers = NFSV41_MINORVERSION;
6693 		if (len > transfer && docommit == 0)
6694 			xfer = transfer;
6695 		else
6696 			xfer = len;
6697 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
6698 			/* Dense layout. */
6699 			if (stripe_pos >= flp->nfsfl_fhcnt)
6700 				return (EIO);
6701 			fhp = flp->nfsfl_fh[stripe_pos];
6702 			io_off = (rel_off / (stripe_unit_size *
6703 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
6704 			    rel_off % stripe_unit_size;
6705 		} else {
6706 			/* Sparse layout. */
6707 			if (flp->nfsfl_fhcnt > 1) {
6708 				if (stripe_index >= flp->nfsfl_fhcnt)
6709 					return (EIO);
6710 				fhp = flp->nfsfl_fh[stripe_index];
6711 			} else if (flp->nfsfl_fhcnt == 1)
6712 				fhp = flp->nfsfl_fh[0];
6713 			else
6714 				fhp = np->n_fhp;
6715 			io_off = off;
6716 		}
6717 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
6718 			commit_thru_mds = 1;
6719 			if (docommit != 0)
6720 				error = EIO;
6721 		} else {
6722 			commit_thru_mds = 0;
6723 			NFSLOCKNODE(np);
6724 			np->n_flag |= NDSCOMMIT;
6725 			NFSUNLOCKNODE(np);
6726 		}
6727 		if (docommit != 0) {
6728 			if (error == 0)
6729 				error = nfsrpc_commitds(vp, io_off, xfer,
6730 				    *dspp, fhp, NFS_VER4, minorvers, cred, p);
6731 			if (error == 0) {
6732 				/*
6733 				 * Set both eof and uio_resid = 0 to end any
6734 				 * loops.
6735 				 */
6736 				*eofp = 1;
6737 				uiop->uio_resid = 0;
6738 			} else {
6739 				NFSLOCKNODE(np);
6740 				np->n_flag &= ~NDSCOMMIT;
6741 				NFSUNLOCKNODE(np);
6742 			}
6743 		} else if (rwflag == NFSV4OPEN_ACCESSREAD)
6744 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6745 			    io_off, xfer, fhp, 0, NFS_VER4, minorvers, cred, p);
6746 		else {
6747 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
6748 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
6749 			    0, NFS_VER4, minorvers, cred, p);
6750 			if (error == 0) {
6751 				NFSLOCKCLSTATE();
6752 				lyp->nfsly_flags |= NFSLY_WRITTEN;
6753 				NFSUNLOCKCLSTATE();
6754 			}
6755 		}
6756 		if (error == 0) {
6757 			transfer = stripe_unit_size;
6758 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
6759 			len -= xfer;
6760 			off += xfer;
6761 		}
6762 	}
6763 	return (error);
6764 }
6765 
6766 /*
6767  * Do I/O using an NFSv4.1 flex file layout.
6768  */
6769 static int
6770 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6771     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6772     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6773     uint64_t len, int mirror, int docommit, struct mbuf *mp,
6774     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6775 {
6776 	uint64_t xfer;
6777 	int error;
6778 	struct nfsnode *np;
6779 	struct nfsfh *fhp;
6780 	struct nfsclds **dspp;
6781 	struct ucred *tcred;
6782 	struct mbuf *m, *m2;
6783 	uint32_t copylen;
6784 
6785 	np = VTONFS(vp);
6786 	error = 0;
6787 	NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6788 	    (uintmax_t)len);
6789 	/* Loop around, doing I/O for each stripe unit. */
6790 	while (len > 0 && error == 0) {
6791 		dspp = nfsfldi_addr(dp, 0);
6792 		fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6793 		stateidp = &flp->nfsfl_ffm[mirror].st;
6794 		NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6795 		    mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6796 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6797 			tcred = NFSNEWCRED(cred);
6798 			tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6799 			tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6800 			tcred->cr_ngroups = 1;
6801 		} else
6802 			tcred = cred;
6803 		if (rwflag == NFSV4OPEN_ACCESSREAD)
6804 			copylen = dp->nfsdi_rsize;
6805 		else {
6806 			copylen = dp->nfsdi_wsize;
6807 			if (len > copylen && mp != NULL) {
6808 				/*
6809 				 * When a mirrored configuration needs to do
6810 				 * multiple writes to each mirror, all writes
6811 				 * except the last one must be a multiple of
6812 				 * 4 bytes.  This is required so that the XDR
6813 				 * does not need padding.
6814 				 * If possible, clip the size to an exact
6815 				 * multiple of the mbuf length, so that the
6816 				 * split will be on an mbuf boundary.
6817 				 */
6818 				copylen &= 0xfffffffc;
6819 				if (copylen > mp->m_len)
6820 					copylen = copylen / mp->m_len *
6821 					    mp->m_len;
6822 			}
6823 		}
6824 		NFSLOCKNODE(np);
6825 		np->n_flag |= NDSCOMMIT;
6826 		NFSUNLOCKNODE(np);
6827 		if (len > copylen && docommit == 0)
6828 			xfer = copylen;
6829 		else
6830 			xfer = len;
6831 		if (docommit != 0) {
6832 			if (error == 0) {
6833 				/*
6834 				 * Do last mirrored DS commit with this thread.
6835 				 */
6836 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6837 					error = nfsio_commitds(vp, off, xfer,
6838 					    *dspp, fhp, dp->nfsdi_vers,
6839 					    dp->nfsdi_minorvers, drpc, tcred,
6840 					    p);
6841 				else
6842 					error = nfsrpc_commitds(vp, off, xfer,
6843 					    *dspp, fhp, dp->nfsdi_vers,
6844 					    dp->nfsdi_minorvers, tcred, p);
6845 				NFSCL_DEBUG(4, "commitds=%d\n", error);
6846 				if (error != 0 && error != EACCES && error !=
6847 				    ESTALE) {
6848 					NFSCL_DEBUG(4,
6849 					    "DS layreterr for commit\n");
6850 					nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6851 					    lyp, *dspp);
6852 				}
6853 			}
6854 			NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6855 			if (error == 0) {
6856 				/*
6857 				 * Set both eof and uio_resid = 0 to end any
6858 				 * loops.
6859 				 */
6860 				*eofp = 1;
6861 				uiop->uio_resid = 0;
6862 			} else {
6863 				NFSLOCKNODE(np);
6864 				np->n_flag &= ~NDSCOMMIT;
6865 				NFSUNLOCKNODE(np);
6866 			}
6867 		} else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6868 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6869 			    off, xfer, fhp, 1, dp->nfsdi_vers,
6870 			    dp->nfsdi_minorvers, tcred, p);
6871 			NFSCL_DEBUG(4, "readds=%d\n", error);
6872 			if (error != 0 && error != EACCES && error != ESTALE) {
6873 				NFSCL_DEBUG(4, "DS layreterr for read\n");
6874 				nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6875 				    *dspp);
6876 			}
6877 		} else {
6878 			if (flp->nfsfl_mirrorcnt == 1) {
6879 				error = nfsrpc_writeds(vp, uiop, iomode,
6880 				    must_commit, stateidp, *dspp, off, xfer,
6881 				    fhp, 0, 1, dp->nfsdi_vers,
6882 				    dp->nfsdi_minorvers, tcred, p);
6883 				if (error == 0) {
6884 					NFSLOCKCLSTATE();
6885 					lyp->nfsly_flags |= NFSLY_WRITTEN;
6886 					NFSUNLOCKCLSTATE();
6887 				}
6888 			} else {
6889 				m = mp;
6890 				if (xfer < len) {
6891 					/* The mbuf list must be split. */
6892 					m2 = nfsm_split(mp, xfer);
6893 					if (m2 != NULL)
6894 						mp = m2;
6895 					else {
6896 						m_freem(mp);
6897 						error = EIO;
6898 					}
6899 				}
6900 				NFSCL_DEBUG(4, "mcopy len=%jd xfer=%jd\n",
6901 				    (uintmax_t)len, (uintmax_t)xfer);
6902 				/*
6903 				 * Do last write to a mirrored DS with this
6904 				 * thread.
6905 				 */
6906 				if (error == 0) {
6907 					if (mirror < flp->nfsfl_mirrorcnt - 1)
6908 						error = nfsio_writedsmir(vp,
6909 						    iomode, must_commit,
6910 						    stateidp, *dspp, off,
6911 						    xfer, fhp, m,
6912 						    dp->nfsdi_vers,
6913 						    dp->nfsdi_minorvers, drpc,
6914 						    tcred, p);
6915 					else
6916 						error = nfsrpc_writedsmir(vp,
6917 						    iomode, must_commit,
6918 						    stateidp, *dspp, off,
6919 						    xfer, fhp, m,
6920 						    dp->nfsdi_vers,
6921 						    dp->nfsdi_minorvers, tcred,
6922 						    p);
6923 				}
6924 				NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6925 				if (error != 0 && error != EACCES && error !=
6926 				    ESTALE) {
6927 					NFSCL_DEBUG(4,
6928 					    "DS layreterr for write\n");
6929 					nfscl_dserr(NFSV4OP_WRITE, error, dp,
6930 					    lyp, *dspp);
6931 				}
6932 			}
6933 		}
6934 		NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6935 		if (error == 0) {
6936 			len -= xfer;
6937 			off += xfer;
6938 		}
6939 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6940 			NFSFREECRED(tcred);
6941 	}
6942 	NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6943 	return (error);
6944 }
6945 
6946 /*
6947  * The actual read RPC done to a DS.
6948  */
6949 static int
6950 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6951     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6952     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6953 {
6954 	uint32_t *tl;
6955 	int attrflag, error, retlen;
6956 	struct nfsrv_descript nfsd;
6957 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6958 	struct nfsrv_descript *nd = &nfsd;
6959 	struct nfssockreq *nrp;
6960 	struct nfsvattr na;
6961 
6962 	nd->nd_mrep = NULL;
6963 	if (vers == 0 || vers == NFS_VER4) {
6964 		nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6965 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6966 		    NULL);
6967 		vers = NFS_VER4;
6968 		NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6969 		if (flex != 0)
6970 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6971 		else
6972 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6973 	} else {
6974 		nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6975 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6976 		    NULL);
6977 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READ]);
6978 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READDS]);
6979 		NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6980 	}
6981 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6982 	txdr_hyper(io_off, tl);
6983 	*(tl + 2) = txdr_unsigned(len);
6984 	nrp = dsp->nfsclds_sockp;
6985 	NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6986 	if (nrp == NULL)
6987 		/* If NULL, use the MDS socket. */
6988 		nrp = &nmp->nm_sockreq;
6989 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6990 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6991 	NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6992 	    error);
6993 	if (error != 0)
6994 		return (error);
6995 	if (vers == NFS_VER3) {
6996 		error = nfscl_postop_attr(nd, &na, &attrflag);
6997 		NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6998 		if (error != 0)
6999 			goto nfsmout;
7000 	}
7001 	if (nd->nd_repstat != 0) {
7002 		error = nd->nd_repstat;
7003 		goto nfsmout;
7004 	}
7005 	if (vers == NFS_VER3) {
7006 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7007 		*eofp = fxdr_unsigned(int, *(tl + 1));
7008 	} else {
7009 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7010 		*eofp = fxdr_unsigned(int, *tl);
7011 	}
7012 	NFSM_STRSIZ(retlen, len);
7013 	NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
7014 	error = nfsm_mbufuio(nd, uiop, retlen);
7015 nfsmout:
7016 	if (nd->nd_mrep != NULL)
7017 		m_freem(nd->nd_mrep);
7018 	return (error);
7019 }
7020 
7021 /*
7022  * The actual write RPC done to a DS.
7023  */
7024 static int
7025 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
7026     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
7027     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
7028     struct ucred *cred, NFSPROC_T *p)
7029 {
7030 	uint32_t *tl;
7031 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7032 	int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
7033 	int32_t backup;
7034 	struct nfsrv_descript nfsd;
7035 	struct nfsrv_descript *nd = &nfsd;
7036 	struct nfssockreq *nrp;
7037 	struct nfsvattr na;
7038 
7039 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
7040 	nd->nd_mrep = NULL;
7041 	if (vers == 0 || vers == NFS_VER4) {
7042 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
7043 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
7044 		    NULL);
7045 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
7046 		vers = NFS_VER4;
7047 		if (flex != 0)
7048 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7049 		else
7050 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
7051 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
7052 	} else {
7053 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
7054 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
7055 		    NULL);
7056 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
7057 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
7058 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
7059 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
7060 	}
7061 	txdr_hyper(io_off, tl);
7062 	tl += 2;
7063 	if (vers == NFS_VER3)
7064 		*tl++ = txdr_unsigned(len);
7065 	*tl++ = txdr_unsigned(*iomode);
7066 	*tl = txdr_unsigned(len);
7067 	error = nfsm_uiombuf(nd, uiop, len);
7068 	if (error != 0) {
7069 		m_freem(nd->nd_mreq);
7070 		return (error);
7071 	}
7072 	nrp = dsp->nfsclds_sockp;
7073 	if (nrp == NULL)
7074 		/* If NULL, use the MDS socket. */
7075 		nrp = &nmp->nm_sockreq;
7076 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7077 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7078 	NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
7079 	    nd->nd_repstat);
7080 	if (error != 0)
7081 		return (error);
7082 	if (nd->nd_repstat != 0) {
7083 		/*
7084 		 * In case the rpc gets retried, roll
7085 		 * the uio fields changed by nfsm_uiombuf()
7086 		 * back.
7087 		 */
7088 		uiop->uio_offset -= len;
7089 		uiop->uio_resid += len;
7090 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base - len;
7091 		uiop->uio_iov->iov_len += len;
7092 		error = nd->nd_repstat;
7093 	} else {
7094 		if (vers == NFS_VER3) {
7095 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
7096 			    NULL);
7097 			NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
7098 			if (error != 0)
7099 				goto nfsmout;
7100 		}
7101 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
7102 		rlen = fxdr_unsigned(int, *tl++);
7103 		NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
7104 		if (rlen == 0) {
7105 			error = NFSERR_IO;
7106 			goto nfsmout;
7107 		} else if (rlen < len) {
7108 			backup = len - rlen;
7109 			uiop->uio_iov->iov_base =
7110 			    (char *)uiop->uio_iov->iov_base - backup;
7111 			uiop->uio_iov->iov_len += backup;
7112 			uiop->uio_offset -= backup;
7113 			uiop->uio_resid += backup;
7114 			len = rlen;
7115 		}
7116 		commit = fxdr_unsigned(int, *tl++);
7117 
7118 		/*
7119 		 * Return the lowest commitment level
7120 		 * obtained by any of the RPCs.
7121 		 */
7122 		if (committed == NFSWRITE_FILESYNC)
7123 			committed = commit;
7124 		else if (committed == NFSWRITE_DATASYNC &&
7125 		    commit == NFSWRITE_UNSTABLE)
7126 			committed = commit;
7127 		if (commit_thru_mds != 0) {
7128 			NFSLOCKMNT(nmp);
7129 			if (!NFSHASWRITEVERF(nmp)) {
7130 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
7131 				NFSSETWRITEVERF(nmp);
7132 			} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF) &&
7133 			    *must_commit != 2) {
7134 				*must_commit = 1;
7135 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
7136 			}
7137 			NFSUNLOCKMNT(nmp);
7138 		} else {
7139 			NFSLOCKDS(dsp);
7140 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
7141 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
7142 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
7143 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF) &&
7144 			    *must_commit != 2) {
7145 				*must_commit = 1;
7146 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
7147 			}
7148 			NFSUNLOCKDS(dsp);
7149 		}
7150 	}
7151 nfsmout:
7152 	if (nd->nd_mrep != NULL)
7153 		m_freem(nd->nd_mrep);
7154 	*iomode = committed;
7155 	if (nd->nd_repstat != 0 && error == 0)
7156 		error = nd->nd_repstat;
7157 	return (error);
7158 }
7159 
7160 /*
7161  * The actual write RPC done to a DS.
7162  * This variant is called from a separate kernel process for mirrors.
7163  * Any short write is considered an IO error.
7164  */
7165 static int
7166 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
7167     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
7168     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
7169     struct ucred *cred, NFSPROC_T *p)
7170 {
7171 	uint32_t *tl;
7172 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7173 	int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
7174 	struct nfsrv_descript nfsd;
7175 	struct nfsrv_descript *nd = &nfsd;
7176 	struct nfssockreq *nrp;
7177 	struct nfsvattr na;
7178 
7179 	nd->nd_mrep = NULL;
7180 	if (vers == 0 || vers == NFS_VER4) {
7181 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
7182 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
7183 		    NULL);
7184 		vers = NFS_VER4;
7185 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
7186 		    minorvers);
7187 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7188 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
7189 	} else {
7190 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
7191 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
7192 		    NULL);
7193 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
7194 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
7195 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
7196 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
7197 	}
7198 	txdr_hyper(io_off, tl);
7199 	tl += 2;
7200 	if (vers == NFS_VER3)
7201 		*tl++ = txdr_unsigned(len);
7202 	*tl++ = txdr_unsigned(*iomode);
7203 	*tl = txdr_unsigned(len);
7204 	if (len > 0) {
7205 		/* Put data in mbuf chain. */
7206 		nd->nd_mb->m_next = m;
7207 	}
7208 	nrp = dsp->nfsclds_sockp;
7209 	if (nrp == NULL)
7210 		/* If NULL, use the MDS socket. */
7211 		nrp = &nmp->nm_sockreq;
7212 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7213 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7214 	NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
7215 	    nd->nd_repstat);
7216 	if (error != 0)
7217 		return (error);
7218 	if (nd->nd_repstat != 0)
7219 		error = nd->nd_repstat;
7220 	else {
7221 		if (vers == NFS_VER3) {
7222 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
7223 			    NULL);
7224 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
7225 			    error);
7226 			if (error != 0)
7227 				goto nfsmout;
7228 		}
7229 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
7230 		rlen = fxdr_unsigned(int, *tl++);
7231 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
7232 		    rlen);
7233 		if (rlen != len) {
7234 			error = NFSERR_IO;
7235 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
7236 			    len, rlen);
7237 			goto nfsmout;
7238 		}
7239 		commit = fxdr_unsigned(int, *tl++);
7240 
7241 		/*
7242 		 * Return the lowest commitment level
7243 		 * obtained by any of the RPCs.
7244 		 */
7245 		if (committed == NFSWRITE_FILESYNC)
7246 			committed = commit;
7247 		else if (committed == NFSWRITE_DATASYNC &&
7248 		    commit == NFSWRITE_UNSTABLE)
7249 			committed = commit;
7250 		NFSLOCKDS(dsp);
7251 		if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
7252 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
7253 			dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
7254 		} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF) &&
7255 		    *must_commit != 2) {
7256 			*must_commit = 1;
7257 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
7258 		}
7259 		NFSUNLOCKDS(dsp);
7260 	}
7261 nfsmout:
7262 	if (nd->nd_mrep != NULL)
7263 		m_freem(nd->nd_mrep);
7264 	*iomode = committed;
7265 	if (nd->nd_repstat != 0 && error == 0)
7266 		error = nd->nd_repstat;
7267 	return (error);
7268 }
7269 
7270 /*
7271  * Start up the thread that will execute nfsrpc_writedsmir().
7272  */
7273 static void
7274 start_writedsmir(void *arg, int pending)
7275 {
7276 	struct nfsclwritedsdorpc *drpc;
7277 
7278 	drpc = (struct nfsclwritedsdorpc *)arg;
7279 	drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
7280 	    &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
7281 	    drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
7282 	    drpc->p);
7283 	drpc->done = 1;
7284 	crfree(drpc->cred);
7285 	NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
7286 }
7287 
7288 /*
7289  * Set up the write DS mirror call for the pNFS I/O thread.
7290  */
7291 static int
7292 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
7293     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
7294     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
7295     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7296 {
7297 	int error, ret;
7298 
7299 	error = 0;
7300 	drpc->done = 0;
7301 	drpc->vp = vp;
7302 	drpc->iomode = *iomode;
7303 	drpc->must_commit = *must_commit;
7304 	drpc->stateidp = stateidp;
7305 	drpc->dsp = dsp;
7306 	drpc->off = off;
7307 	drpc->len = len;
7308 	drpc->fhp = fhp;
7309 	drpc->m = m;
7310 	drpc->vers = vers;
7311 	drpc->minorvers = minorvers;
7312 	drpc->cred = crhold(cred);
7313 	drpc->p = p;
7314 	drpc->inprog = 0;
7315 	ret = EIO;
7316 	if (nfs_pnfsiothreads != 0) {
7317 		ret = nfs_pnfsio(start_writedsmir, drpc);
7318 		NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
7319 	}
7320 	if (ret != 0) {
7321 		error = nfsrpc_writedsmir(vp, iomode, &drpc->must_commit,
7322 		    stateidp, dsp, off, len, fhp, m, vers, minorvers, cred, p);
7323 		crfree(drpc->cred);
7324 	}
7325 	NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
7326 	return (error);
7327 }
7328 
7329 /*
7330  * Free up the nfsclds structure.
7331  */
7332 void
7333 nfscl_freenfsclds(struct nfsclds *dsp)
7334 {
7335 	int i;
7336 
7337 	if (dsp == NULL)
7338 		return;
7339 	if (dsp->nfsclds_sockp != NULL) {
7340 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
7341 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
7342 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
7343 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
7344 	}
7345 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
7346 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
7347 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
7348 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
7349 			m_freem(
7350 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
7351 	}
7352 	free(dsp, M_NFSCLDS);
7353 }
7354 
7355 static enum nfsclds_state
7356 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
7357     struct nfsclds **retdspp, uint32_t *sequencep)
7358 {
7359 	struct nfsclds *dsp;
7360 	int fndseq;
7361 
7362 	/*
7363 	 * Search the list of nfsclds structures for one with the same
7364 	 * server.
7365 	 */
7366 	fndseq = 0;
7367 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
7368 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
7369 		    dsp->nfsclds_servownlen != 0 &&
7370 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
7371 		    dsp->nfsclds_servownlen) &&
7372 		    dsp->nfsclds_sess.nfsess_defunct == 0) {
7373 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
7374 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
7375 			    dsp->nfsclds_flags);
7376 			if (fndseq == 0) {
7377 				/* Get sequenceid# from first entry. */
7378 				*sequencep =
7379 				    dsp->nfsclds_sess.nfsess_sequenceid;
7380 				fndseq = 1;
7381 			}
7382 			/* Server major id matches. */
7383 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
7384 				*retdspp = dsp;
7385 				return (NFSDSP_USETHISSESSION);
7386 			}
7387 		}
7388 	}
7389 	if (fndseq != 0)
7390 		return (NFSDSP_SEQTHISSESSION);
7391 	return (NFSDSP_NOTFOUND);
7392 }
7393 
7394 /*
7395  * NFS commit rpc to a NFSv4.1 DS.
7396  */
7397 static int
7398 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
7399     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
7400     NFSPROC_T *p)
7401 {
7402 	uint32_t *tl;
7403 	struct nfsrv_descript nfsd, *nd = &nfsd;
7404 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7405 	struct nfssockreq *nrp;
7406 	struct nfsvattr na;
7407 	int attrflag, error;
7408 
7409 	nd->nd_mrep = NULL;
7410 	if (vers == 0 || vers == NFS_VER4) {
7411 		nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
7412 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
7413 		    NULL);
7414 		vers = NFS_VER4;
7415 	} else {
7416 		nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
7417 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
7418 		    NULL);
7419 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMIT]);
7420 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMITDS]);
7421 	}
7422 	NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
7423 	    minorvers);
7424 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
7425 	txdr_hyper(offset, tl);
7426 	tl += 2;
7427 	*tl = txdr_unsigned(cnt);
7428 	nrp = dsp->nfsclds_sockp;
7429 	if (nrp == NULL)
7430 		/* If NULL, use the MDS socket. */
7431 		nrp = &nmp->nm_sockreq;
7432 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7433 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7434 	NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
7435 	    nd->nd_repstat);
7436 	if (error != 0)
7437 		return (error);
7438 	if (nd->nd_repstat == 0) {
7439 		if (vers == NFS_VER3) {
7440 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
7441 			    NULL);
7442 			NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
7443 			if (error != 0)
7444 				goto nfsmout;
7445 		}
7446 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
7447 		NFSLOCKDS(dsp);
7448 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
7449 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
7450 			error = NFSERR_STALEWRITEVERF;
7451 		}
7452 		NFSUNLOCKDS(dsp);
7453 	}
7454 nfsmout:
7455 	if (error == 0 && nd->nd_repstat != 0)
7456 		error = nd->nd_repstat;
7457 	m_freem(nd->nd_mrep);
7458 	return (error);
7459 }
7460 
7461 /*
7462  * Start up the thread that will execute nfsrpc_commitds().
7463  */
7464 static void
7465 start_commitds(void *arg, int pending)
7466 {
7467 	struct nfsclwritedsdorpc *drpc;
7468 
7469 	drpc = (struct nfsclwritedsdorpc *)arg;
7470 	drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
7471 	    drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
7472 	    drpc->p);
7473 	drpc->done = 1;
7474 	crfree(drpc->cred);
7475 	NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
7476 }
7477 
7478 /*
7479  * Set up the commit DS mirror call for the pNFS I/O thread.
7480  */
7481 static int
7482 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
7483     struct nfsfh *fhp, int vers, int minorvers,
7484     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7485 {
7486 	int error, ret;
7487 
7488 	error = 0;
7489 	drpc->done = 0;
7490 	drpc->vp = vp;
7491 	drpc->off = offset;
7492 	drpc->len = cnt;
7493 	drpc->dsp = dsp;
7494 	drpc->fhp = fhp;
7495 	drpc->vers = vers;
7496 	drpc->minorvers = minorvers;
7497 	drpc->cred = crhold(cred);
7498 	drpc->p = p;
7499 	drpc->inprog = 0;
7500 	ret = EIO;
7501 	if (nfs_pnfsiothreads != 0) {
7502 		ret = nfs_pnfsio(start_commitds, drpc);
7503 		NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
7504 	}
7505 	if (ret != 0) {
7506 		error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
7507 		    minorvers, cred, p);
7508 		crfree(drpc->cred);
7509 	}
7510 	NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
7511 	return (error);
7512 }
7513 
7514 /*
7515  * NFS Advise rpc
7516  */
7517 int
7518 nfsrpc_advise(vnode_t vp, off_t offset, uint64_t cnt, int advise,
7519     struct ucred *cred, NFSPROC_T *p)
7520 {
7521 	u_int32_t *tl;
7522 	struct nfsrv_descript nfsd, *nd = &nfsd;
7523 	nfsattrbit_t hints;
7524 	int error;
7525 
7526 	NFSZERO_ATTRBIT(&hints);
7527 	if (advise == POSIX_FADV_WILLNEED)
7528 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
7529 	else if (advise == POSIX_FADV_DONTNEED)
7530 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
7531 	else
7532 		return (0);
7533 	NFSCL_REQSTART(nd, NFSPROC_IOADVISE, vp, cred);
7534 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
7535 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
7536 	txdr_hyper(offset, tl);
7537 	tl += 2;
7538 	txdr_hyper(cnt, tl);
7539 	nfsrv_putattrbit(nd, &hints);
7540 	error = nfscl_request(nd, vp, p, cred);
7541 	if (error != 0)
7542 		return (error);
7543 	if (nd->nd_repstat != 0)
7544 		error = nd->nd_repstat;
7545 	m_freem(nd->nd_mrep);
7546 	return (error);
7547 }
7548 
7549 #ifdef notyet
7550 /*
7551  * NFS advise rpc to a NFSv4.2 DS.
7552  */
7553 static int
7554 nfsrpc_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
7555     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
7556     struct ucred *cred, NFSPROC_T *p)
7557 {
7558 	uint32_t *tl;
7559 	struct nfsrv_descript nfsd, *nd = &nfsd;
7560 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7561 	struct nfssockreq *nrp;
7562 	nfsattrbit_t hints;
7563 	int error;
7564 
7565 	/* For NFS DSs prior to NFSv4.2, just return OK. */
7566 	if (vers == NFS_VER3 || minorversion < NFSV42_MINORVERSION)
7567 		return (0);
7568 	NFSZERO_ATTRBIT(&hints);
7569 	if (advise == POSIX_FADV_WILLNEED)
7570 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
7571 	else if (advise == POSIX_FADV_DONTNEED)
7572 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
7573 	else
7574 		return (0);
7575 	nd->nd_mrep = NULL;
7576 	nfscl_reqstart(nd, NFSPROC_IOADVISEDS, nmp, fhp->nfh_fh,
7577 	    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers, NULL);
7578 	vers = NFS_VER4;
7579 	NFSCL_DEBUG(4, "nfsrpc_adviseds: vers=%d minvers=%d\n", vers,
7580 	    minorvers);
7581 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
7582 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
7583 	txdr_hyper(offset, tl);
7584 	tl += 2;
7585 	*tl = txdr_unsigned(cnt);
7586 	nfsrv_putattrbit(nd, &hints);
7587 	nrp = dsp->nfsclds_sockp;
7588 	if (nrp == NULL)
7589 		/* If NULL, use the MDS socket. */
7590 		nrp = &nmp->nm_sockreq;
7591 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7592 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7593 	NFSCL_DEBUG(4, "nfsrpc_adviseds: err=%d stat=%d\n", error,
7594 	    nd->nd_repstat);
7595 	if (error != 0)
7596 		return (error);
7597 	if (nd->nd_repstat != 0)
7598 		error = nd->nd_repstat;
7599 	m_freem(nd->nd_mrep);
7600 	return (error);
7601 }
7602 
7603 /*
7604  * Start up the thread that will execute nfsrpc_commitds().
7605  */
7606 static void
7607 start_adviseds(void *arg, int pending)
7608 {
7609 	struct nfsclwritedsdorpc *drpc;
7610 
7611 	drpc = (struct nfsclwritedsdorpc *)arg;
7612 	drpc->err = nfsrpc_adviseds(drpc->vp, drpc->off, drpc->len,
7613 	    drpc->advise, drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers,
7614 	    drpc->cred, drpc->p);
7615 	drpc->done = 1;
7616 	crfree(drpc->cred);
7617 	NFSCL_DEBUG(4, "start_adviseds: err=%d\n", drpc->err);
7618 }
7619 
7620 /*
7621  * Set up the advise DS mirror call for the pNFS I/O thread.
7622  */
7623 static int
7624 nfsio_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
7625     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
7626     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7627 {
7628 	int error, ret;
7629 
7630 	error = 0;
7631 	drpc->done = 0;
7632 	drpc->vp = vp;
7633 	drpc->off = offset;
7634 	drpc->len = cnt;
7635 	drpc->advise = advise;
7636 	drpc->dsp = dsp;
7637 	drpc->fhp = fhp;
7638 	drpc->vers = vers;
7639 	drpc->minorvers = minorvers;
7640 	drpc->cred = crhold(cred);
7641 	drpc->p = p;
7642 	drpc->inprog = 0;
7643 	ret = EIO;
7644 	if (nfs_pnfsiothreads != 0) {
7645 		ret = nfs_pnfsio(start_adviseds, drpc);
7646 		NFSCL_DEBUG(4, "nfsio_adviseds: nfs_pnfsio=%d\n", ret);
7647 	}
7648 	if (ret != 0) {
7649 		error = nfsrpc_adviseds(vp, offset, cnt, advise, dsp, fhp, vers,
7650 		    minorvers, cred, p);
7651 		crfree(drpc->cred);
7652 	}
7653 	NFSCL_DEBUG(4, "nfsio_adviseds: error=%d\n", error);
7654 	return (error);
7655 }
7656 #endif	/* notyet */
7657 
7658 /*
7659  * Do the Allocate operation, retrying for recovery.
7660  */
7661 int
7662 nfsrpc_allocate(vnode_t vp, off_t off, off_t len, struct nfsvattr *nap,
7663     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
7664 {
7665 	int error, expireret = 0, retrycnt, nostateid;
7666 	uint32_t clidrev = 0;
7667 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7668 	struct nfsfh *nfhp = NULL;
7669 	nfsv4stateid_t stateid;
7670 	off_t tmp_off;
7671 	void *lckp;
7672 
7673 	if (len < 0)
7674 		return (EINVAL);
7675 	if (len == 0)
7676 		return (0);
7677 	tmp_off = off + len;
7678 	NFSLOCKMNT(nmp);
7679 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < off) {
7680 		NFSUNLOCKMNT(nmp);
7681 		return (EFBIG);
7682 	}
7683 	if (nmp->nm_clp != NULL)
7684 		clidrev = nmp->nm_clp->nfsc_clientidrev;
7685 	NFSUNLOCKMNT(nmp);
7686 	nfhp = VTONFS(vp)->n_fhp;
7687 	retrycnt = 0;
7688 	do {
7689 		lckp = NULL;
7690 		nostateid = 0;
7691 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
7692 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
7693 		if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
7694 		    stateid.other[2] == 0) {
7695 			nostateid = 1;
7696 			NFSCL_DEBUG(1, "stateid0 in allocate\n");
7697 		}
7698 
7699 		/*
7700 		 * Not finding a stateid should probably never happen,
7701 		 * but just return an error for this case.
7702 		 */
7703 		if (nostateid != 0)
7704 			error = EIO;
7705 		else
7706 			error = nfsrpc_allocaterpc(vp, off, len, &stateid,
7707 			    nap, attrflagp, cred, p);
7708 		if (error == NFSERR_STALESTATEID)
7709 			nfscl_initiate_recovery(nmp->nm_clp);
7710 		if (lckp != NULL)
7711 			nfscl_lockderef(lckp);
7712 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
7713 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
7714 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
7715 			(void) nfs_catnap(PZERO, error, "nfs_allocate");
7716 		} else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) &&
7717 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
7718 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
7719 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) {
7720 			error = EIO;
7721 		}
7722 		retrycnt++;
7723 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
7724 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
7725 	    error == NFSERR_STALEDONTRECOVER ||
7726 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
7727 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
7728 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
7729 	if (error != 0 && retrycnt >= 4)
7730 		error = EIO;
7731 	return (error);
7732 }
7733 
7734 /*
7735  * The allocate RPC.
7736  */
7737 static int
7738 nfsrpc_allocaterpc(vnode_t vp, off_t off, off_t len, nfsv4stateid_t *stateidp,
7739     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
7740 {
7741 	uint32_t *tl;
7742 	int error;
7743 	struct nfsrv_descript nfsd;
7744 	struct nfsrv_descript *nd = &nfsd;
7745 	nfsattrbit_t attrbits;
7746 
7747 	*attrflagp = 0;
7748 	NFSCL_REQSTART(nd, NFSPROC_ALLOCATE, vp, cred);
7749 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7750 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
7751 	txdr_hyper(off, tl); tl += 2;
7752 	txdr_hyper(len, tl); tl += 2;
7753 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7754 	NFSGETATTR_ATTRBIT(&attrbits);
7755 	nfsrv_putattrbit(nd, &attrbits);
7756 	error = nfscl_request(nd, vp, p, cred);
7757 	if (error != 0)
7758 		return (error);
7759 	if (nd->nd_repstat == 0) {
7760 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7761 		error = nfsm_loadattr(nd, nap);
7762 		if (error == 0)
7763 			*attrflagp = NFS_LATTR_NOSHRINK;
7764 	} else
7765 		error = nd->nd_repstat;
7766 nfsmout:
7767 	m_freem(nd->nd_mrep);
7768 	return (error);
7769 }
7770 
7771 /*
7772  * Set up the XDR arguments for the LayoutGet operation.
7773  */
7774 static void
7775 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
7776     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
7777     int layoutlen, int usecurstateid)
7778 {
7779 	uint32_t *tl;
7780 
7781 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
7782 	    NFSX_STATEID);
7783 	*tl++ = newnfs_false;		/* Don't signal availability. */
7784 	*tl++ = txdr_unsigned(layouttype);
7785 	*tl++ = txdr_unsigned(iomode);
7786 	txdr_hyper(offset, tl);
7787 	tl += 2;
7788 	txdr_hyper(len, tl);
7789 	tl += 2;
7790 	txdr_hyper(minlen, tl);
7791 	tl += 2;
7792 	if (usecurstateid != 0) {
7793 		/* Special stateid for Current stateid. */
7794 		*tl++ = txdr_unsigned(1);
7795 		*tl++ = 0;
7796 		*tl++ = 0;
7797 		*tl++ = 0;
7798 	} else {
7799 		*tl++ = txdr_unsigned(stateidp->seqid);
7800 		NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
7801 		*tl++ = stateidp->other[0];
7802 		*tl++ = stateidp->other[1];
7803 		*tl++ = stateidp->other[2];
7804 	}
7805 	*tl = txdr_unsigned(layoutlen);
7806 }
7807 
7808 /*
7809  * Parse the reply for a successful LayoutGet operation.
7810  */
7811 static int
7812 nfsrv_parselayoutget(struct nfsmount *nmp, struct nfsrv_descript *nd,
7813     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp)
7814 {
7815 	uint32_t *tl;
7816 	struct nfsclflayout *flp, *prevflp, *tflp;
7817 	int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
7818 	int m, mirrorcnt;
7819 	uint64_t retlen, off;
7820 	struct nfsfh *nfhp;
7821 	uint8_t *cp;
7822 	uid_t user;
7823 	gid_t grp;
7824 
7825 	NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
7826 	error = 0;
7827 	flp = NULL;
7828 	gotiomode = -1;
7829 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
7830 	if (*tl++ != 0)
7831 		*retonclosep = 1;
7832 	else
7833 		*retonclosep = 0;
7834 	stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
7835 	NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
7836 	    (int)stateidp->seqid);
7837 	stateidp->other[0] = *tl++;
7838 	stateidp->other[1] = *tl++;
7839 	stateidp->other[2] = *tl++;
7840 	cnt = fxdr_unsigned(int, *tl);
7841 	NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
7842 	if (cnt <= 0 || cnt > 10000) {
7843 		/* Don't accept more than 10000 layouts in reply. */
7844 		error = NFSERR_BADXDR;
7845 		goto nfsmout;
7846 	}
7847 	for (i = 0; i < cnt; i++) {
7848 		/* Dissect to the layout type. */
7849 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
7850 		    3 * NFSX_UNSIGNED);
7851 		off = fxdr_hyper(tl); tl += 2;
7852 		retlen = fxdr_hyper(tl); tl += 2;
7853 		iomode = fxdr_unsigned(int, *tl++);
7854 		laytype = fxdr_unsigned(int, *tl);
7855 		NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
7856 		    (uintmax_t)off, (uintmax_t)retlen, iomode);
7857 		/* Ignore length of layout body for now. */
7858 		if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
7859 			/* Parse the File layout up to fhcnt. */
7860 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
7861 			    NFSX_HYPER + NFSX_V4DEVICEID);
7862 			fhcnt = fxdr_unsigned(int, *(tl + 4 +
7863 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
7864 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7865 			if (fhcnt < 0 || fhcnt > 100) {
7866 				/* Don't accept more than 100 file handles. */
7867 				error = NFSERR_BADXDR;
7868 				goto nfsmout;
7869 			}
7870 			if (fhcnt > 0)
7871 				flp = malloc(sizeof(*flp) + fhcnt *
7872 				    sizeof(struct nfsfh *), M_NFSFLAYOUT,
7873 				    M_WAITOK);
7874 			else
7875 				flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
7876 				    M_WAITOK);
7877 			flp->nfsfl_flags = NFSFL_FILE;
7878 			flp->nfsfl_fhcnt = 0;
7879 			flp->nfsfl_devp = NULL;
7880 			flp->nfsfl_off = off;
7881 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7882 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7883 			else
7884 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7885 			flp->nfsfl_iomode = iomode;
7886 			if (gotiomode == -1)
7887 				gotiomode = flp->nfsfl_iomode;
7888 			/* Ignore layout body length for now. */
7889 			NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
7890 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
7891 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
7892 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
7893 			mtx_lock(&nmp->nm_mtx);
7894 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_util &
7895 			    NFSFLAYUTIL_IOADVISE_THRU_MDS) != 0)
7896 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7897 			mtx_unlock(&nmp->nm_mtx);
7898 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
7899 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
7900 			NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
7901 			    flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
7902 			for (j = 0; j < fhcnt; j++) {
7903 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7904 				nfhlen = fxdr_unsigned(int, *tl);
7905 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
7906 					error = NFSERR_BADXDR;
7907 					goto nfsmout;
7908 				}
7909 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
7910 				    M_NFSFH, M_WAITOK);
7911 				flp->nfsfl_fh[j] = nfhp;
7912 				flp->nfsfl_fhcnt++;
7913 				nfhp->nfh_len = nfhlen;
7914 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
7915 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
7916 			}
7917 		} else if (laytype == NFSLAYOUT_FLEXFILE) {
7918 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
7919 			    NFSX_HYPER);
7920 			mirrorcnt = fxdr_unsigned(int, *(tl + 2));
7921 			NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
7922 			if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
7923 				error = NFSERR_BADXDR;
7924 				goto nfsmout;
7925 			}
7926 			flp = malloc(sizeof(*flp) + mirrorcnt *
7927 			    sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
7928 			flp->nfsfl_flags = NFSFL_FLEXFILE;
7929 			flp->nfsfl_mirrorcnt = mirrorcnt;
7930 			for (j = 0; j < mirrorcnt; j++)
7931 				flp->nfsfl_ffm[j].devp = NULL;
7932 			flp->nfsfl_off = off;
7933 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7934 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7935 			else
7936 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7937 			flp->nfsfl_iomode = iomode;
7938 			if (gotiomode == -1)
7939 				gotiomode = flp->nfsfl_iomode;
7940 			flp->nfsfl_stripeunit = fxdr_hyper(tl);
7941 			NFSCL_DEBUG(4, "stripeunit=%ju\n",
7942 			    (uintmax_t)flp->nfsfl_stripeunit);
7943 			for (j = 0; j < mirrorcnt; j++) {
7944 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7945 				k = fxdr_unsigned(int, *tl);
7946 				if (k < 1 || k > 128) {
7947 					error = NFSERR_BADXDR;
7948 					goto nfsmout;
7949 				}
7950 				NFSCL_DEBUG(4, "servercnt=%d\n", k);
7951 				for (l = 0; l < k; l++) {
7952 					NFSM_DISSECT(tl, uint32_t *,
7953 					    NFSX_V4DEVICEID + NFSX_STATEID +
7954 					    2 * NFSX_UNSIGNED);
7955 					if (l == 0) {
7956 						/* Just use the first server. */
7957 						NFSBCOPY(tl,
7958 						    flp->nfsfl_ffm[j].dev,
7959 						    NFSX_V4DEVICEID);
7960 						tl += (NFSX_V4DEVICEID /
7961 						    NFSX_UNSIGNED);
7962 						tl++;
7963 						flp->nfsfl_ffm[j].st.seqid =
7964 						    *tl++;
7965 						flp->nfsfl_ffm[j].st.other[0] =
7966 						    *tl++;
7967 						flp->nfsfl_ffm[j].st.other[1] =
7968 						    *tl++;
7969 						flp->nfsfl_ffm[j].st.other[2] =
7970 						    *tl++;
7971 						NFSCL_DEBUG(4, "st.seqid=%u "
7972 						 "st.o0=0x%x st.o1=0x%x "
7973 						 "st.o2=0x%x\n",
7974 						 flp->nfsfl_ffm[j].st.seqid,
7975 						 flp->nfsfl_ffm[j].st.other[0],
7976 						 flp->nfsfl_ffm[j].st.other[1],
7977 						 flp->nfsfl_ffm[j].st.other[2]);
7978 					} else
7979 						tl += ((NFSX_V4DEVICEID +
7980 						    NFSX_STATEID +
7981 						    NFSX_UNSIGNED) /
7982 						    NFSX_UNSIGNED);
7983 					fhcnt = fxdr_unsigned(int, *tl);
7984 					NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7985 					if (fhcnt < 1 ||
7986 					    fhcnt > NFSDEV_MAXVERS) {
7987 						error = NFSERR_BADXDR;
7988 						goto nfsmout;
7989 					}
7990 					for (m = 0; m < fhcnt; m++) {
7991 						NFSM_DISSECT(tl, uint32_t *,
7992 						    NFSX_UNSIGNED);
7993 						nfhlen = fxdr_unsigned(int,
7994 						    *tl);
7995 						NFSCL_DEBUG(4, "nfhlen=%d\n",
7996 						    nfhlen);
7997 						if (nfhlen <= 0 || nfhlen >
7998 						    NFSX_V4FHMAX) {
7999 							error = NFSERR_BADXDR;
8000 							goto nfsmout;
8001 						}
8002 						NFSM_DISSECT(cp, uint8_t *,
8003 						    NFSM_RNDUP(nfhlen));
8004 						if (l == 0) {
8005 							flp->nfsfl_ffm[j].fhcnt
8006 							    = fhcnt;
8007 							nfhp = malloc(
8008 							    sizeof(*nfhp) +
8009 							    nfhlen - 1, M_NFSFH,
8010 							    M_WAITOK);
8011 							flp->nfsfl_ffm[j].fh[m]
8012 							    = nfhp;
8013 							nfhp->nfh_len = nfhlen;
8014 							NFSBCOPY(cp,
8015 							    nfhp->nfh_fh,
8016 							    nfhlen);
8017 							NFSCL_DEBUG(4,
8018 							    "got fh\n");
8019 						}
8020 					}
8021 					/* Now, get the ffsd_user/ffds_group. */
8022 					error = nfsrv_parseug(nd, 0, &user,
8023 					    &grp, curthread);
8024 					NFSCL_DEBUG(4, "after parseu=%d\n",
8025 					    error);
8026 					if (error == 0)
8027 						error = nfsrv_parseug(nd, 1,
8028 						    &user, &grp, curthread);
8029 					NFSCL_DEBUG(4, "aft parseg=%d\n",
8030 					    grp);
8031 					if (error != 0)
8032 						goto nfsmout;
8033 					NFSCL_DEBUG(4, "user=%d group=%d\n",
8034 					    user, grp);
8035 					if (l == 0) {
8036 						flp->nfsfl_ffm[j].user = user;
8037 						flp->nfsfl_ffm[j].group = grp;
8038 						NFSCL_DEBUG(4,
8039 						    "usr=%d grp=%d\n", user,
8040 						    grp);
8041 					}
8042 				}
8043 			}
8044 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8045 			flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
8046 #ifdef notnow
8047 			/*
8048 			 * At this time, there is no flag.
8049 			 * NFSFLEXFLAG_IOADVISE_THRU_MDS might need to be
8050 			 * added, or it may never exist?
8051 			 */
8052 			mtx_lock(&nmp->nm_mtx);
8053 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_fflags &
8054 			    NFSFLEXFLAG_IOADVISE_THRU_MDS) != 0)
8055 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
8056 			mtx_unlock(&nmp->nm_mtx);
8057 #endif
8058 			flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
8059 			NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
8060 			    flp->nfsfl_fflags, flp->nfsfl_statshint);
8061 		} else {
8062 			error = NFSERR_BADXDR;
8063 			goto nfsmout;
8064 		}
8065 		if (flp->nfsfl_iomode == gotiomode) {
8066 			/* Keep the list in increasing offset order. */
8067 			tflp = LIST_FIRST(flhp);
8068 			prevflp = NULL;
8069 			while (tflp != NULL &&
8070 			    tflp->nfsfl_off < flp->nfsfl_off) {
8071 				prevflp = tflp;
8072 				tflp = LIST_NEXT(tflp, nfsfl_list);
8073 			}
8074 			if (prevflp == NULL)
8075 				LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
8076 			else
8077 				LIST_INSERT_AFTER(prevflp, flp,
8078 				    nfsfl_list);
8079 			NFSCL_DEBUG(4, "flp inserted\n");
8080 		} else {
8081 			printf("nfscl_layoutget(): got wrong iomode\n");
8082 			nfscl_freeflayout(flp);
8083 		}
8084 		flp = NULL;
8085 	}
8086 nfsmout:
8087 	NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
8088 	if (error != 0 && flp != NULL)
8089 		nfscl_freeflayout(flp);
8090 	return (error);
8091 }
8092 
8093 /*
8094  * Parse a user/group digit string.
8095  */
8096 static int
8097 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
8098     NFSPROC_T *p)
8099 {
8100 	uint32_t *tl;
8101 	char *cp, *str, str0[NFSV4_SMALLSTR + 1];
8102 	uint32_t len = 0;
8103 	int error = 0;
8104 
8105 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8106 	len = fxdr_unsigned(uint32_t, *tl);
8107 	str = NULL;
8108 	if (len > NFSV4_OPAQUELIMIT) {
8109 		error = NFSERR_BADXDR;
8110 		goto nfsmout;
8111 	}
8112 	NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
8113 	if (len == 0) {
8114 		if (dogrp != 0)
8115 			*gidp = GID_NOGROUP;
8116 		else
8117 			*uidp = UID_NOBODY;
8118 		return (0);
8119 	}
8120 	if (len > NFSV4_SMALLSTR)
8121 		str = malloc(len + 1, M_TEMP, M_WAITOK);
8122 	else
8123 		str = str0;
8124 	NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
8125 	NFSBCOPY(cp, str, len);
8126 	str[len] = '\0';
8127 	NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
8128 	if (dogrp != 0)
8129 		error = nfsv4_strtogid(nd, str, len, gidp);
8130 	else
8131 		error = nfsv4_strtouid(nd, str, len, uidp);
8132 nfsmout:
8133 	if (len > NFSV4_SMALLSTR)
8134 		free(str, M_TEMP);
8135 	NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
8136 	return (error);
8137 }
8138 
8139 /*
8140  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
8141  * so that it does both an Open and a Layoutget.
8142  */
8143 static int
8144 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
8145     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
8146     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
8147     struct ucred *cred, NFSPROC_T *p)
8148 {
8149 	struct nfscllayout *lyp;
8150 	struct nfsclflayout *flp;
8151 	struct nfsclflayouthead flh;
8152 	int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
8153 	int layouttype, laystat;
8154 	nfsv4stateid_t stateid;
8155 	struct nfsclsession *tsep;
8156 
8157 	error = 0;
8158 	if (NFSHASFLEXFILE(nmp))
8159 		layouttype = NFSLAYOUT_FLEXFILE;
8160 	else
8161 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
8162 	/*
8163 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
8164 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
8165 	 * flp == NULL.
8166 	 */
8167 	lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, mode, &flp,
8168 	    &recalled);
8169 	NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
8170 	if (lyp == NULL)
8171 		islocked = 0;
8172 	else if (flp != NULL)
8173 		islocked = 1;
8174 	else
8175 		islocked = 2;
8176 	if ((lyp == NULL || flp == NULL) && recalled == 0) {
8177 		LIST_INIT(&flh);
8178 		tsep = nfsmnt_mdssession(nmp);
8179 		layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
8180 		    3 * NFSX_UNSIGNED);
8181 		if (lyp == NULL)
8182 			usecurstateid = 1;
8183 		else {
8184 			usecurstateid = 0;
8185 			stateid.seqid = lyp->nfsly_stateid.seqid;
8186 			stateid.other[0] = lyp->nfsly_stateid.other[0];
8187 			stateid.other[1] = lyp->nfsly_stateid.other[1];
8188 			stateid.other[2] = lyp->nfsly_stateid.other[2];
8189 		}
8190 		error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
8191 		    newfhp, newfhlen, mode, op, name, namelen,
8192 		    dpp, &stateid, usecurstateid, layouttype, layoutlen,
8193 		    &retonclose, &flh, &laystat, cred, p);
8194 		NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
8195 		    laystat, error);
8196 		laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
8197 		    &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
8198 		    &islocked, cred, p);
8199 	} else
8200 		error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
8201 		    mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
8202 	if (islocked == 2)
8203 		nfscl_rellayout(lyp, 1);
8204 	else if (islocked == 1)
8205 		nfscl_rellayout(lyp, 0);
8206 	return (error);
8207 }
8208 
8209 /*
8210  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
8211  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
8212  * handled by nfsrpc_openrpc().
8213  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
8214  * can be NULL.
8215  */
8216 static int
8217 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
8218     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
8219     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
8220     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
8221     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
8222     int *laystatp, struct ucred *cred, NFSPROC_T *p)
8223 {
8224 	uint32_t *tl;
8225 	struct nfsrv_descript nfsd, *nd = &nfsd;
8226 	struct nfscldeleg *ndp = NULL;
8227 	struct nfsvattr nfsva;
8228 	struct nfsclsession *tsep;
8229 	uint32_t rflags, deleg;
8230 	nfsattrbit_t attrbits;
8231 	int error, ret, acesize, limitby, iomode;
8232 
8233 	*dpp = NULL;
8234 	*laystatp = ENXIO;
8235 	nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
8236 	    0, 0, cred);
8237 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
8238 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
8239 	*tl++ = txdr_unsigned(mode & (NFSV4OPEN_ACCESSBOTH |
8240 	    NFSV4OPEN_WANTDELEGMASK));
8241 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
8242 	tsep = nfsmnt_mdssession(nmp);
8243 	*tl++ = tsep->nfsess_clientid.lval[0];
8244 	*tl = tsep->nfsess_clientid.lval[1];
8245 	nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
8246 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8247 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
8248 	if (NFSHASNFSV4N(nmp)) {
8249 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMFH);
8250 	} else {
8251 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
8252 		nfsm_strtom(nd, name, namelen);
8253 	}
8254 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8255 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8256 	NFSZERO_ATTRBIT(&attrbits);
8257 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
8258 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
8259 	nfsrv_putattrbit(nd, &attrbits);
8260 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8261 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
8262 	if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
8263 		iomode = NFSLAYOUTIOMODE_RW;
8264 	else
8265 		iomode = NFSLAYOUTIOMODE_READ;
8266 	nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
8267 	    layouttype, layoutlen, usecurstateid);
8268 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
8269 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
8270 	if (error != 0)
8271 		return (error);
8272 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
8273 	if (nd->nd_repstat != 0)
8274 		*laystatp = nd->nd_repstat;
8275 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8276 		/* ND_NOMOREDATA will be set if the Open operation failed. */
8277 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8278 		    6 * NFSX_UNSIGNED);
8279 		op->nfso_stateid.seqid = *tl++;
8280 		op->nfso_stateid.other[0] = *tl++;
8281 		op->nfso_stateid.other[1] = *tl++;
8282 		op->nfso_stateid.other[2] = *tl;
8283 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
8284 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
8285 		if (error != 0)
8286 			goto nfsmout;
8287 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
8288 		deleg = fxdr_unsigned(u_int32_t, *tl);
8289 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
8290 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
8291 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
8292 			      NFSCLFLAGS_FIRSTDELEG))
8293 				op->nfso_own->nfsow_clp->nfsc_flags |=
8294 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
8295 			ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
8296 			    M_NFSCLDELEG, M_WAITOK);
8297 			LIST_INIT(&ndp->nfsdl_owner);
8298 			LIST_INIT(&ndp->nfsdl_lock);
8299 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
8300 			ndp->nfsdl_fhlen = newfhlen;
8301 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
8302 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
8303 			nfscl_lockinit(&ndp->nfsdl_rwlock);
8304 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8305 			    NFSX_UNSIGNED);
8306 			ndp->nfsdl_stateid.seqid = *tl++;
8307 			ndp->nfsdl_stateid.other[0] = *tl++;
8308 			ndp->nfsdl_stateid.other[1] = *tl++;
8309 			ndp->nfsdl_stateid.other[2] = *tl++;
8310 			ret = fxdr_unsigned(int, *tl);
8311 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
8312 				ndp->nfsdl_flags = NFSCLDL_WRITE;
8313 				/*
8314 				 * Indicates how much the file can grow.
8315 				 */
8316 				NFSM_DISSECT(tl, u_int32_t *,
8317 				    3 * NFSX_UNSIGNED);
8318 				limitby = fxdr_unsigned(int, *tl++);
8319 				switch (limitby) {
8320 				case NFSV4OPEN_LIMITSIZE:
8321 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
8322 					break;
8323 				case NFSV4OPEN_LIMITBLOCKS:
8324 					ndp->nfsdl_sizelimit =
8325 					    fxdr_unsigned(u_int64_t, *tl++);
8326 					ndp->nfsdl_sizelimit *=
8327 					    fxdr_unsigned(u_int64_t, *tl);
8328 					break;
8329 				default:
8330 					error = NFSERR_BADXDR;
8331 					goto nfsmout;
8332 				};
8333 			} else
8334 				ndp->nfsdl_flags = NFSCLDL_READ;
8335 			if (ret != 0)
8336 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
8337 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, false,
8338 			    &ret, &acesize, p);
8339 			if (error != 0)
8340 				goto nfsmout;
8341 		} else if (deleg == NFSV4OPEN_DELEGATENONEEXT &&
8342 		    NFSHASNFSV4N(nmp)) {
8343 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8344 			deleg = fxdr_unsigned(uint32_t, *tl);
8345 			if (deleg == NFSV4OPEN_CONTENTION ||
8346 			    deleg == NFSV4OPEN_RESOURCE)
8347 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8348 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
8349 			error = NFSERR_BADXDR;
8350 			goto nfsmout;
8351 		}
8352 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
8353 		    nfscl_assumeposixlocks)
8354 			op->nfso_posixlock = 1;
8355 		else
8356 			op->nfso_posixlock = 0;
8357 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8358 		/* If the 2nd element == NFS_OK, the Getattr succeeded. */
8359 		if (*++tl == 0) {
8360 			error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
8361 			    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
8362 			    NULL, NULL, NULL, p, cred);
8363 			if (error != 0)
8364 				goto nfsmout;
8365 			if (ndp != NULL) {
8366 				ndp->nfsdl_change = nfsva.na_filerev;
8367 				ndp->nfsdl_modtime = nfsva.na_mtime;
8368 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
8369 				*dpp = ndp;
8370 				ndp = NULL;
8371 			}
8372 			/*
8373 			 * At this point, the Open has succeeded, so set
8374 			 * nd_repstat = NFS_OK.  If the Layoutget failed,
8375 			 * this function just won't return a layout.
8376 			 */
8377 			if (nd->nd_repstat == 0) {
8378 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8379 				*laystatp = fxdr_unsigned(int, *++tl);
8380 				if (*laystatp == 0) {
8381 					error = nfsrv_parselayoutget(nmp, nd,
8382 					    stateidp, retonclosep, flhp);
8383 					if (error != 0)
8384 						*laystatp = error;
8385 				}
8386 			} else
8387 				nd->nd_repstat = 0;	/* Return 0 for Open. */
8388 		}
8389 	}
8390 	if (nd->nd_repstat != 0 && error == 0)
8391 		error = nd->nd_repstat;
8392 nfsmout:
8393 	free(ndp, M_NFSCLDELEG);
8394 	m_freem(nd->nd_mrep);
8395 	return (error);
8396 }
8397 
8398 /*
8399  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
8400  * Used only for mounts with pNFS enabled.
8401  */
8402 static int
8403 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
8404     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
8405     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
8406     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
8407     int *dattrflagp, int *unlockedp, nfsv4stateid_t *stateidp,
8408     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
8409     struct nfsclflayouthead *flhp, int *laystatp)
8410 {
8411 	uint32_t *tl;
8412 	int error = 0, deleg, newone, ret, acesize, limitby;
8413 	struct nfsrv_descript nfsd, *nd = &nfsd;
8414 	struct nfsclopen *op;
8415 	struct nfscldeleg *dp = NULL;
8416 	struct nfsnode *np;
8417 	struct nfsfh *nfhp;
8418 	struct nfsclsession *tsep;
8419 	nfsattrbit_t attrbits;
8420 	nfsv4stateid_t stateid;
8421 	struct nfsmount *nmp;
8422 
8423 	nmp = VFSTONFS(dvp->v_mount);
8424 	np = VTONFS(dvp);
8425 	*laystatp = ENXIO;
8426 	*unlockedp = 0;
8427 	*nfhpp = NULL;
8428 	*dpp = NULL;
8429 	*attrflagp = 0;
8430 	*dattrflagp = 0;
8431 	if (namelen > NFS_MAXNAMLEN)
8432 		return (ENAMETOOLONG);
8433 	NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp, cred);
8434 	/*
8435 	 * For V4, this is actually an Open op.
8436 	 */
8437 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
8438 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
8439 	if (NFSHASNFSV4N(nmp)) {
8440 		if (!NFSHASPNFS(nmp) && nfscl_enablecallb != 0 &&
8441 		    nfs_numnfscbd > 0)
8442 			*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
8443 			    NFSV4OPEN_ACCESSREAD | NFSV4OPEN_WANTWRITEDELEG);
8444 		else
8445 			*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
8446 			    NFSV4OPEN_ACCESSREAD | NFSV4OPEN_WANTNODELEG);
8447 	} else
8448 		*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
8449 		    NFSV4OPEN_ACCESSREAD);
8450 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
8451 	tsep = nfsmnt_mdssession(nmp);
8452 	*tl++ = tsep->nfsess_clientid.lval[0];
8453 	*tl = tsep->nfsess_clientid.lval[1];
8454 	nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
8455 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8456 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
8457 	if ((fmode & O_EXCL) != 0) {
8458 		if (NFSHASSESSPERSIST(nmp)) {
8459 			/* Use GUARDED for persistent sessions. */
8460 			*tl = txdr_unsigned(NFSCREATE_GUARDED);
8461 			nfscl_fillsattr(nd, vap, dvp, NFSSATTR_NEWFILE, 0);
8462 		} else {
8463 			/* Otherwise, use EXCLUSIVE4_1. */
8464 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
8465 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
8466 			*tl++ = cverf.lval[0];
8467 			*tl = cverf.lval[1];
8468 			nfscl_fillsattr(nd, vap, dvp, NFSSATTR_NEWFILE, 0);
8469 		}
8470 	} else {
8471 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
8472 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_NEWFILE, 0);
8473 	}
8474 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8475 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
8476 	nfsm_strtom(nd, name, namelen);
8477 	/* Get the new file's handle and attributes, plus save the FH. */
8478 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
8479 	*tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
8480 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
8481 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8482 	NFSGETATTR_ATTRBIT(&attrbits);
8483 	nfsrv_putattrbit(nd, &attrbits);
8484 	/* Get the directory's post-op attributes. */
8485 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8486 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8487 	(void)nfsm_fhtom(nmp, nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
8488 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8489 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8490 	nfsrv_putattrbit(nd, &attrbits);
8491 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8492 	*tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
8493 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
8494 	nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
8495 	    layouttype, layoutlen, usecurstateid);
8496 	error = nfscl_request(nd, dvp, p, cred);
8497 	if (error != 0)
8498 		return (error);
8499 	NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
8500 	    error);
8501 	if (nd->nd_repstat != 0)
8502 		*laystatp = nd->nd_repstat;
8503 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
8504 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8505 		NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
8506 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8507 		    6 * NFSX_UNSIGNED);
8508 		stateid.seqid = *tl++;
8509 		stateid.other[0] = *tl++;
8510 		stateid.other[1] = *tl++;
8511 		stateid.other[2] = *tl;
8512 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
8513 		if (error != 0)
8514 			goto nfsmout;
8515 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
8516 		deleg = fxdr_unsigned(int, *tl);
8517 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
8518 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
8519 			if (!(owp->nfsow_clp->nfsc_flags &
8520 			      NFSCLFLAGS_FIRSTDELEG))
8521 				owp->nfsow_clp->nfsc_flags |=
8522 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
8523 			dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
8524 			    M_NFSCLDELEG, M_WAITOK);
8525 			LIST_INIT(&dp->nfsdl_owner);
8526 			LIST_INIT(&dp->nfsdl_lock);
8527 			dp->nfsdl_clp = owp->nfsow_clp;
8528 			newnfs_copyincred(cred, &dp->nfsdl_cred);
8529 			nfscl_lockinit(&dp->nfsdl_rwlock);
8530 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8531 			    NFSX_UNSIGNED);
8532 			dp->nfsdl_stateid.seqid = *tl++;
8533 			dp->nfsdl_stateid.other[0] = *tl++;
8534 			dp->nfsdl_stateid.other[1] = *tl++;
8535 			dp->nfsdl_stateid.other[2] = *tl++;
8536 			ret = fxdr_unsigned(int, *tl);
8537 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
8538 				dp->nfsdl_flags = NFSCLDL_WRITE;
8539 				/*
8540 				 * Indicates how much the file can grow.
8541 				 */
8542 				NFSM_DISSECT(tl, u_int32_t *,
8543 				    3 * NFSX_UNSIGNED);
8544 				limitby = fxdr_unsigned(int, *tl++);
8545 				switch (limitby) {
8546 				case NFSV4OPEN_LIMITSIZE:
8547 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
8548 					break;
8549 				case NFSV4OPEN_LIMITBLOCKS:
8550 					dp->nfsdl_sizelimit =
8551 					    fxdr_unsigned(u_int64_t, *tl++);
8552 					dp->nfsdl_sizelimit *=
8553 					    fxdr_unsigned(u_int64_t, *tl);
8554 					break;
8555 				default:
8556 					error = NFSERR_BADXDR;
8557 					goto nfsmout;
8558 				};
8559 			} else {
8560 				dp->nfsdl_flags = NFSCLDL_READ;
8561 			}
8562 			if (ret != 0)
8563 				dp->nfsdl_flags |= NFSCLDL_RECALL;
8564 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, false,
8565 			    &ret, &acesize, p);
8566 			if (error != 0)
8567 				goto nfsmout;
8568 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
8569 			error = NFSERR_BADXDR;
8570 			goto nfsmout;
8571 		}
8572 
8573 		/* Now, we should have the status for the SaveFH. */
8574 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8575 		if (*++tl == 0) {
8576 			NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
8577 			/*
8578 			 * Now, process the GetFH and Getattr for the newly
8579 			 * created file. nfscl_mtofh() will set
8580 			 * ND_NOMOREDATA if these weren't successful.
8581 			 */
8582 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
8583 			NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
8584 			if (error != 0)
8585 				goto nfsmout;
8586 		} else
8587 			nd->nd_flag |= ND_NOMOREDATA;
8588 		/* Now we have the PutFH and Getattr for the directory. */
8589 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8590 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8591 			if (*++tl != 0)
8592 				nd->nd_flag |= ND_NOMOREDATA;
8593 			else {
8594 				NFSM_DISSECT(tl, uint32_t *, 2 *
8595 				    NFSX_UNSIGNED);
8596 				if (*++tl != 0)
8597 					nd->nd_flag |= ND_NOMOREDATA;
8598 			}
8599 		}
8600 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8601 			/* Load the directory attributes. */
8602 			error = nfsm_loadattr(nd, dnap);
8603 			NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
8604 			if (error != 0)
8605 				goto nfsmout;
8606 			*dattrflagp = 1;
8607 			if (dp != NULL && *attrflagp != 0) {
8608 				dp->nfsdl_change = nnap->na_filerev;
8609 				dp->nfsdl_modtime = nnap->na_mtime;
8610 				dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
8611 			}
8612 			/*
8613 			 * We can now complete the Open state.
8614 			 */
8615 			nfhp = *nfhpp;
8616 			if (dp != NULL) {
8617 				dp->nfsdl_fhlen = nfhp->nfh_len;
8618 				NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
8619 				    nfhp->nfh_len);
8620 			}
8621 			/*
8622 			 * Get an Open structure that will be
8623 			 * attached to the OpenOwner, acquired already.
8624 			 */
8625 			error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
8626 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
8627 			    cred, p, NULL, &op, &newone, NULL, 0, false);
8628 			if (error != 0)
8629 				goto nfsmout;
8630 			op->nfso_stateid = stateid;
8631 			newnfs_copyincred(cred, &op->nfso_cred);
8632 
8633 			nfscl_openrelease(nmp, op, error, newone);
8634 			*unlockedp = 1;
8635 
8636 			/* Now, handle the RestoreFH and LayoutGet. */
8637 			if (nd->nd_repstat == 0) {
8638 				NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
8639 				*laystatp = fxdr_unsigned(int, *(tl + 3));
8640 				if (*laystatp == 0) {
8641 					error = nfsrv_parselayoutget(nmp, nd,
8642 					    stateidp, retonclosep, flhp);
8643 					if (error != 0)
8644 						*laystatp = error;
8645 				}
8646 				NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
8647 				    error);
8648 			} else
8649 				nd->nd_repstat = 0;
8650 		}
8651 	}
8652 	if (nd->nd_repstat != 0 && error == 0)
8653 		error = nd->nd_repstat;
8654 	if (error == NFSERR_STALECLIENTID)
8655 		nfscl_initiate_recovery(owp->nfsow_clp);
8656 nfsmout:
8657 	NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
8658 	if (error == 0)
8659 		*dpp = dp;
8660 	else
8661 		free(dp, M_NFSCLDELEG);
8662 	m_freem(nd->nd_mrep);
8663 	return (error);
8664 }
8665 
8666 /*
8667  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
8668  */
8669 static int
8670 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
8671     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
8672     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
8673     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
8674     int *dattrflagp, int *unlockedp)
8675 {
8676 	struct nfscllayout *lyp;
8677 	struct nfsclflayouthead flh;
8678 	struct nfsfh *nfhp;
8679 	struct nfsclsession *tsep;
8680 	struct nfsmount *nmp;
8681 	nfsv4stateid_t stateid;
8682 	int error, layoutlen, layouttype, retonclose, laystat;
8683 
8684 	error = 0;
8685 	nmp = VFSTONFS(dvp->v_mount);
8686 	if (NFSHASFLEXFILE(nmp))
8687 		layouttype = NFSLAYOUT_FLEXFILE;
8688 	else
8689 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
8690 	LIST_INIT(&flh);
8691 	tsep = nfsmnt_mdssession(nmp);
8692 	layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
8693 	error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
8694 	    owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
8695 	    unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
8696 	    &flh, &laystat);
8697 	NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
8698 	    laystat, error);
8699 	lyp = NULL;
8700 	if (laystat == 0) {
8701 		nfhp = *nfhpp;
8702 		laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
8703 		    nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
8704 		    layouttype, laystat, NULL, cred, p);
8705 	} else
8706 		laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
8707 		    retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
8708 		    cred, p);
8709 	if (laystat == 0)
8710 		nfscl_rellayout(lyp, 0);
8711 	return (error);
8712 }
8713 
8714 /*
8715  * Process the results of a layoutget() operation.
8716  */
8717 static int
8718 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
8719     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
8720     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
8721     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
8722 {
8723 	struct nfsclflayout *tflp;
8724 	struct nfscldevinfo *dip;
8725 	uint8_t *dev;
8726 	int i, mirrorcnt;
8727 
8728 	if (laystat == NFSERR_UNKNLAYOUTTYPE) {
8729 		NFSLOCKMNT(nmp);
8730 		if (!NFSHASFLEXFILE(nmp)) {
8731 			/* Switch to using Flex File Layout. */
8732 			nmp->nm_state |= NFSSTA_FLEXFILE;
8733 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
8734 			/* Disable pNFS. */
8735 			NFSCL_DEBUG(1, "disable PNFS\n");
8736 			nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
8737 		}
8738 		NFSUNLOCKMNT(nmp);
8739 	}
8740 	if (laystat == 0) {
8741 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
8742 		LIST_FOREACH(tflp, flhp, nfsfl_list) {
8743 			if (layouttype == NFSLAYOUT_FLEXFILE)
8744 				mirrorcnt = tflp->nfsfl_mirrorcnt;
8745 			else
8746 				mirrorcnt = 1;
8747 			for (i = 0; i < mirrorcnt; i++) {
8748 				laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
8749 				NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
8750 				if (laystat != 0) {
8751 					if (layouttype == NFSLAYOUT_FLEXFILE)
8752 						dev = tflp->nfsfl_ffm[i].dev;
8753 					else
8754 						dev = tflp->nfsfl_dev;
8755 					laystat = nfsrpc_getdeviceinfo(nmp, dev,
8756 					    layouttype, notifybit, &dip, cred,
8757 					    p);
8758 					NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
8759 					    laystat);
8760 					if (laystat != 0)
8761 						goto out;
8762 					laystat = nfscl_adddevinfo(nmp, dip, i,
8763 					    tflp);
8764 					if (laystat != 0)
8765 						printf("nfsrpc_layoutgetresout"
8766 						    ": cannot add\n");
8767 				}
8768 			}
8769 		}
8770 	}
8771 out:
8772 	if (laystat == 0) {
8773 		/*
8774 		 * nfscl_layout() always returns with the nfsly_lock
8775 		 * set to a refcnt (shared lock).
8776 		 * Passing in dvp is sufficient, since it is only used to
8777 		 * get the fsid for the file system.
8778 		 */
8779 		laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
8780 		    layouttype, retonclose, flhp, lypp, cred, p);
8781 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
8782 		    laystat);
8783 		if (laystat == 0 && islockedp != NULL)
8784 			*islockedp = 1;
8785 	}
8786 	return (laystat);
8787 }
8788 
8789 /*
8790  * nfs copy_file_range operation.
8791  */
8792 int
8793 nfsrpc_copy_file_range(vnode_t invp, off_t *inoffp, vnode_t outvp,
8794     off_t *outoffp, size_t *lenp, unsigned int flags, int *inattrflagp,
8795     struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap,
8796     struct ucred *cred, bool consecutive, bool *must_commitp)
8797 {
8798 	int commit, error, expireret = 0, retrycnt;
8799 	u_int32_t clidrev = 0;
8800 	struct nfsmount *nmp = VFSTONFS(invp->v_mount);
8801 	struct nfsfh *innfhp = NULL, *outnfhp = NULL;
8802 	nfsv4stateid_t instateid, outstateid;
8803 	void *inlckp, *outlckp;
8804 
8805 	if (nmp->nm_clp != NULL)
8806 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8807 	innfhp = VTONFS(invp)->n_fhp;
8808 	outnfhp = VTONFS(outvp)->n_fhp;
8809 	retrycnt = 0;
8810 	do {
8811 		/* Get both stateids. */
8812 		inlckp = NULL;
8813 		nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len,
8814 		    NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid,
8815 		    &inlckp);
8816 		outlckp = NULL;
8817 		nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len,
8818 		    NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid,
8819 		    &outlckp);
8820 
8821 		error = nfsrpc_copyrpc(invp, *inoffp, outvp, *outoffp, lenp,
8822 		    &instateid, &outstateid, innap, inattrflagp, outnap,
8823 		    outattrflagp, consecutive, &commit, cred, curthread);
8824 		if (error == 0) {
8825 			if (commit != NFSWRITE_FILESYNC)
8826 				*must_commitp = true;
8827 			*inoffp += *lenp;
8828 			*outoffp += *lenp;
8829 		} else if (error == NFSERR_STALESTATEID)
8830 			nfscl_initiate_recovery(nmp->nm_clp);
8831 		if (inlckp != NULL)
8832 			nfscl_lockderef(inlckp);
8833 		if (outlckp != NULL)
8834 			nfscl_lockderef(outlckp);
8835 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8836 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8837 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8838 			(void) nfs_catnap(PZERO, error, "nfs_cfr");
8839 		} else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) &&
8840 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
8841 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8842 			    curthread);
8843 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) {
8844 			error = EIO;
8845 		}
8846 		retrycnt++;
8847 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
8848 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8849 	      error == NFSERR_STALEDONTRECOVER ||
8850 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8851 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8852 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
8853 	if (error != 0 && (retrycnt >= 4 ||
8854 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8855 	      error == NFSERR_STALEDONTRECOVER))
8856 		error = EIO;
8857 	return (error);
8858 }
8859 
8860 /*
8861  * The copy RPC.
8862  */
8863 static int
8864 nfsrpc_copyrpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff,
8865     size_t *lenp, nfsv4stateid_t *instateidp, nfsv4stateid_t *outstateidp,
8866     struct nfsvattr *innap, int *inattrflagp, struct nfsvattr *outnap,
8867     int *outattrflagp, bool consecutive, int *commitp, struct ucred *cred,
8868     NFSPROC_T *p)
8869 {
8870 	uint32_t *tl, *opcntp;
8871 	int error;
8872 	struct nfsrv_descript nfsd;
8873 	struct nfsrv_descript *nd = &nfsd;
8874 	struct nfsmount *nmp;
8875 	nfsattrbit_t attrbits;
8876 	struct vattr va;
8877 	uint64_t len;
8878 
8879 	nmp = VFSTONFS(invp->v_mount);
8880 	*inattrflagp = *outattrflagp = 0;
8881 	*commitp = NFSWRITE_UNSTABLE;
8882 	len = *lenp;
8883 	*lenp = 0;
8884 	if (len > nfs_maxcopyrange)
8885 		len = nfs_maxcopyrange;
8886 	nfscl_reqstart(nd, NFSPROC_COPY, nmp, VTONFS(invp)->n_fhp->nfh_fh,
8887 	    VTONFS(invp)->n_fhp->nfh_len, &opcntp, NULL, 0, 0, cred);
8888 	/*
8889 	 * First do a Setattr of atime to the server's clock
8890 	 * time.  The FreeBSD "collective" was of the opinion
8891 	 * that setting atime was necessary for this syscall.
8892 	 * Do the Setattr before the Copy, so that it can be
8893 	 * handled well if the server replies NFSERR_DELAY to
8894 	 * the Setattr operation.
8895 	 */
8896 	if ((nmp->nm_mountp->mnt_flag & MNT_NOATIME) == 0) {
8897 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8898 		*tl = txdr_unsigned(NFSV4OP_SETATTR);
8899 		nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
8900 		VATTR_NULL(&va);
8901 		va.va_atime.tv_sec = va.va_atime.tv_nsec = 0;
8902 		va.va_vaflags = VA_UTIMES_NULL;
8903 		nfscl_fillsattr(nd, &va, invp, 0, 0);
8904 		/* Bump opcnt from 7 to 8. */
8905 		*opcntp = txdr_unsigned(8);
8906 	}
8907 
8908 	/* Now Getattr the invp attributes. */
8909 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8910 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8911 	NFSGETATTR_ATTRBIT(&attrbits);
8912 	nfsrv_putattrbit(nd, &attrbits);
8913 
8914 	/* Set outvp. */
8915 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8916 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8917 	(void)nfsm_fhtom(nmp, nd, VTONFS(outvp)->n_fhp->nfh_fh,
8918 	    VTONFS(outvp)->n_fhp->nfh_len, 0);
8919 
8920 	/* Do the Copy. */
8921 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8922 	*tl = txdr_unsigned(NFSV4OP_COPY);
8923 	nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
8924 	nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID);
8925 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + 4 * NFSX_UNSIGNED);
8926 	txdr_hyper(inoff, tl); tl += 2;
8927 	txdr_hyper(outoff, tl); tl += 2;
8928 	txdr_hyper(len, tl); tl += 2;
8929 	if (consecutive)
8930 		*tl++ = newnfs_true;
8931 	else
8932 		*tl++ = newnfs_false;
8933 	*tl++ = newnfs_true;
8934 	*tl++ = 0;
8935 
8936 	/* Get the outvp attributes. */
8937 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8938 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
8939 	nfsrv_putattrbit(nd, &attrbits);
8940 
8941 	error = nfscl_request(nd, invp, p, cred);
8942 	if (error != 0)
8943 		return (error);
8944 	/* Skip over the Setattr reply. */
8945 	if ((nd->nd_flag & ND_NOMOREDATA) == 0 &&
8946 	    (nmp->nm_mountp->mnt_flag & MNT_NOATIME) == 0) {
8947 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8948 		if (*(tl + 1) == 0) {
8949 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
8950 			if (error != 0)
8951 				goto nfsmout;
8952 		} else
8953 			nd->nd_flag |= ND_NOMOREDATA;
8954 	}
8955 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8956 		/* Get the input file's attributes. */
8957 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8958 		if (*(tl + 1) == 0) {
8959 			error = nfsm_loadattr(nd, innap);
8960 			if (error != 0)
8961 				goto nfsmout;
8962 			*inattrflagp = 1;
8963 		} else
8964 			nd->nd_flag |= ND_NOMOREDATA;
8965 	}
8966 	/* Skip over return stat for PutFH. */
8967 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8968 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8969 		if (*++tl != 0)
8970 			nd->nd_flag |= ND_NOMOREDATA;
8971 	}
8972 	/* Skip over return stat for Copy. */
8973 	if ((nd->nd_flag & ND_NOMOREDATA) == 0)
8974 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8975 	if (nd->nd_repstat == 0) {
8976 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8977 		if (*tl != 0) {
8978 			/* There should be no callback ids. */
8979 			error = NFSERR_BADXDR;
8980 			goto nfsmout;
8981 		}
8982 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED +
8983 		    NFSX_VERF);
8984 		len = fxdr_hyper(tl); tl += 2;
8985 		*commitp = fxdr_unsigned(int, *tl++);
8986 		NFSLOCKMNT(nmp);
8987 		if (!NFSHASWRITEVERF(nmp)) {
8988 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8989 			NFSSETWRITEVERF(nmp);
8990 	    	} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
8991 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8992 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
8993 		}
8994 		NFSUNLOCKMNT(nmp);
8995 		tl += (NFSX_VERF / NFSX_UNSIGNED);
8996 		if (nd->nd_repstat == 0 && *++tl != newnfs_true)
8997 			/* Must be a synchronous copy. */
8998 			nd->nd_repstat = NFSERR_NOTSUPP;
8999 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
9000 		error = nfsm_loadattr(nd, outnap);
9001 		if (error == 0)
9002 			*outattrflagp = NFS_LATTR_NOSHRINK;
9003 		if (nd->nd_repstat == 0)
9004 			*lenp = len;
9005 	} else if (nd->nd_repstat == NFSERR_OFFLOADNOREQS) {
9006 		/*
9007 		 * For the case where consecutive is not supported, but
9008 		 * synchronous is supported, we can try consecutive == false
9009 		 * by returning this error.  Otherwise, return NFSERR_NOTSUPP,
9010 		 * since Copy cannot be done.
9011 		 */
9012 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
9013 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
9014 			if (!consecutive || *++tl == newnfs_false)
9015 				nd->nd_repstat = NFSERR_NOTSUPP;
9016 		} else
9017 			nd->nd_repstat = NFSERR_BADXDR;
9018 	}
9019 	if (error == 0)
9020 		error = nd->nd_repstat;
9021 nfsmout:
9022 	m_freem(nd->nd_mrep);
9023 	return (error);
9024 }
9025 
9026 /*
9027  * Seek operation.
9028  */
9029 int
9030 nfsrpc_seek(vnode_t vp, off_t *offp, bool *eofp, int content,
9031     struct ucred *cred, struct nfsvattr *nap, int *attrflagp)
9032 {
9033 	int error, expireret = 0, retrycnt;
9034 	u_int32_t clidrev = 0;
9035 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
9036 	struct nfsnode *np = VTONFS(vp);
9037 	struct nfsfh *nfhp = NULL;
9038 	nfsv4stateid_t stateid;
9039 	void *lckp;
9040 
9041 	if (nmp->nm_clp != NULL)
9042 		clidrev = nmp->nm_clp->nfsc_clientidrev;
9043 	nfhp = np->n_fhp;
9044 	retrycnt = 0;
9045 	do {
9046 		lckp = NULL;
9047 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
9048 		    NFSV4OPEN_ACCESSREAD, 0, cred, curthread, &stateid, &lckp);
9049 		error = nfsrpc_seekrpc(vp, offp, &stateid, eofp, content,
9050 		    nap, attrflagp, cred);
9051 		if (error == NFSERR_STALESTATEID)
9052 			nfscl_initiate_recovery(nmp->nm_clp);
9053 		if (lckp != NULL)
9054 			nfscl_lockderef(lckp);
9055 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
9056 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
9057 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
9058 			(void) nfs_catnap(PZERO, error, "nfs_seek");
9059 		} else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) &&
9060 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
9061 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
9062 			    curthread);
9063 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) {
9064 			error = EIO;
9065 		}
9066 		retrycnt++;
9067 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
9068 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
9069 	    error == NFSERR_BADSESSION ||
9070 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
9071 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
9072 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
9073 	    (error == NFSERR_OPENMODE && retrycnt < 4));
9074 	if (error && retrycnt >= 4)
9075 		error = EIO;
9076 	return (error);
9077 }
9078 
9079 /*
9080  * The seek RPC.
9081  */
9082 static int
9083 nfsrpc_seekrpc(vnode_t vp, off_t *offp, nfsv4stateid_t *stateidp, bool *eofp,
9084     int content, struct nfsvattr *nap, int *attrflagp, struct ucred *cred)
9085 {
9086 	uint32_t *tl;
9087 	int error;
9088 	struct nfsrv_descript nfsd;
9089 	struct nfsrv_descript *nd = &nfsd;
9090 	nfsattrbit_t attrbits;
9091 
9092 	*attrflagp = 0;
9093 	NFSCL_REQSTART(nd, NFSPROC_SEEK, vp, cred);
9094 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
9095 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
9096 	txdr_hyper(*offp, tl); tl += 2;
9097 	*tl++ = txdr_unsigned(content);
9098 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
9099 	NFSGETATTR_ATTRBIT(&attrbits);
9100 	nfsrv_putattrbit(nd, &attrbits);
9101 	error = nfscl_request(nd, vp, curthread, cred);
9102 	if (error != 0)
9103 		return (error);
9104 	if (nd->nd_repstat == 0) {
9105 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED + NFSX_HYPER);
9106 		if (*tl++ == newnfs_true)
9107 			*eofp = true;
9108 		else
9109 			*eofp = false;
9110 		*offp = fxdr_hyper(tl);
9111 		/* Just skip over Getattr op status. */
9112 		error = nfsm_loadattr(nd, nap);
9113 		if (error == 0)
9114 			*attrflagp = 1;
9115 	}
9116 	error = nd->nd_repstat;
9117 nfsmout:
9118 	m_freem(nd->nd_mrep);
9119 	return (error);
9120 }
9121 
9122 /*
9123  * The getextattr RPC.
9124  */
9125 int
9126 nfsrpc_getextattr(vnode_t vp, const char *name, struct uio *uiop, ssize_t *lenp,
9127     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
9128 {
9129 	uint32_t *tl;
9130 	int error;
9131 	struct nfsrv_descript nfsd;
9132 	struct nfsrv_descript *nd = &nfsd;
9133 	nfsattrbit_t attrbits;
9134 	uint32_t len, len2;
9135 
9136 	*attrflagp = 0;
9137 	NFSCL_REQSTART(nd, NFSPROC_GETEXTATTR, vp, cred);
9138 	nfsm_strtom(nd, name, strlen(name));
9139 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
9140 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
9141 	NFSGETATTR_ATTRBIT(&attrbits);
9142 	nfsrv_putattrbit(nd, &attrbits);
9143 	error = nfscl_request(nd, vp, p, cred);
9144 	if (error != 0)
9145 		return (error);
9146 	if (nd->nd_repstat == 0) {
9147 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
9148 		len = fxdr_unsigned(uint32_t, *tl);
9149 		/* Sanity check lengths. */
9150 		if (uiop != NULL && len > 0 && len <= IOSIZE_MAX &&
9151 		    uiop->uio_resid <= UINT32_MAX) {
9152 			len2 = uiop->uio_resid;
9153 			if (len2 >= len)
9154 				error = nfsm_mbufuio(nd, uiop, len);
9155 			else {
9156 				error = nfsm_mbufuio(nd, uiop, len2);
9157 				if (error == 0) {
9158 					/*
9159 					 * nfsm_mbufuio() advances to a multiple
9160 					 * of 4, so round up len2 as well.  Then
9161 					 * we need to advance over the rest of
9162 					 * the data, rounding up the remaining
9163 					 * length.
9164 					 */
9165 					len2 = NFSM_RNDUP(len2);
9166 					len2 = NFSM_RNDUP(len - len2);
9167 					if (len2 > 0)
9168 						error = nfsm_advance(nd, len2,
9169 						    -1);
9170 				}
9171 			}
9172 		} else if (uiop == NULL && len > 0) {
9173 			/* Just wants the length and not the data. */
9174 			error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
9175 		} else if (len > 0)
9176 			error = ENOATTR;
9177 		if (error != 0)
9178 			goto nfsmout;
9179 		*lenp = len;
9180 		/* Just skip over Getattr op status. */
9181 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
9182 		error = nfsm_loadattr(nd, nap);
9183 		if (error == 0)
9184 			*attrflagp = 1;
9185 	}
9186 	if (error == 0)
9187 		error = nd->nd_repstat;
9188 nfsmout:
9189 	m_freem(nd->nd_mrep);
9190 	return (error);
9191 }
9192 
9193 /*
9194  * The setextattr RPC.
9195  */
9196 int
9197 nfsrpc_setextattr(vnode_t vp, const char *name, struct uio *uiop,
9198     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
9199 {
9200 	uint32_t *tl;
9201 	int error;
9202 	struct nfsrv_descript nfsd;
9203 	struct nfsrv_descript *nd = &nfsd;
9204 	nfsattrbit_t attrbits;
9205 
9206 	*attrflagp = 0;
9207 	NFSCL_REQSTART(nd, NFSPROC_SETEXTATTR, vp, cred);
9208 	if (uiop->uio_resid > nd->nd_maxreq) {
9209 		/* nd_maxreq is set by NFSCL_REQSTART(). */
9210 		m_freem(nd->nd_mreq);
9211 		return (EINVAL);
9212 	}
9213 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
9214 	*tl = txdr_unsigned(NFSV4SXATTR_EITHER);
9215 	nfsm_strtom(nd, name, strlen(name));
9216 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
9217 	*tl = txdr_unsigned(uiop->uio_resid);
9218 	error = nfsm_uiombuf(nd, uiop, uiop->uio_resid);
9219 	if (error != 0) {
9220 		m_freem(nd->nd_mreq);
9221 		return (error);
9222 	}
9223 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
9224 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
9225 	NFSGETATTR_ATTRBIT(&attrbits);
9226 	nfsrv_putattrbit(nd, &attrbits);
9227 	error = nfscl_request(nd, vp, p, cred);
9228 	if (error != 0)
9229 		return (error);
9230 	if (nd->nd_repstat == 0) {
9231 		/* Just skip over the reply and Getattr op status. */
9232 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
9233 		    NFSX_UNSIGNED);
9234 		error = nfsm_loadattr(nd, nap);
9235 		if (error == 0)
9236 			*attrflagp = 1;
9237 	}
9238 	if (error == 0)
9239 		error = nd->nd_repstat;
9240 nfsmout:
9241 	m_freem(nd->nd_mrep);
9242 	return (error);
9243 }
9244 
9245 /*
9246  * The removeextattr RPC.
9247  */
9248 int
9249 nfsrpc_rmextattr(vnode_t vp, const char *name, struct nfsvattr *nap,
9250     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
9251 {
9252 	uint32_t *tl;
9253 	int error;
9254 	struct nfsrv_descript nfsd;
9255 	struct nfsrv_descript *nd = &nfsd;
9256 	nfsattrbit_t attrbits;
9257 
9258 	*attrflagp = 0;
9259 	NFSCL_REQSTART(nd, NFSPROC_RMEXTATTR, vp, cred);
9260 	nfsm_strtom(nd, name, strlen(name));
9261 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
9262 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
9263 	NFSGETATTR_ATTRBIT(&attrbits);
9264 	nfsrv_putattrbit(nd, &attrbits);
9265 	error = nfscl_request(nd, vp, p, cred);
9266 	if (error != 0)
9267 		return (error);
9268 	if (nd->nd_repstat == 0) {
9269 		/* Just skip over the reply and Getattr op status. */
9270 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
9271 		    NFSX_UNSIGNED);
9272 		error = nfsm_loadattr(nd, nap);
9273 		if (error == 0)
9274 			*attrflagp = 1;
9275 	}
9276 	if (error == 0)
9277 		error = nd->nd_repstat;
9278 nfsmout:
9279 	m_freem(nd->nd_mrep);
9280 	return (error);
9281 }
9282 
9283 /*
9284  * The listextattr RPC.
9285  */
9286 int
9287 nfsrpc_listextattr(vnode_t vp, uint64_t *cookiep, struct uio *uiop,
9288     size_t *lenp, bool *eofp, struct nfsvattr *nap, int *attrflagp,
9289     struct ucred *cred, NFSPROC_T *p)
9290 {
9291 	uint32_t *tl;
9292 	int cnt, error, i, len;
9293 	struct nfsrv_descript nfsd;
9294 	struct nfsrv_descript *nd = &nfsd;
9295 	nfsattrbit_t attrbits;
9296 	u_char c;
9297 
9298 	*attrflagp = 0;
9299 	NFSCL_REQSTART(nd, NFSPROC_LISTEXTATTR, vp, cred);
9300 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
9301 	txdr_hyper(*cookiep, tl); tl += 2;
9302 	*tl++ = txdr_unsigned(*lenp);
9303 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
9304 	NFSGETATTR_ATTRBIT(&attrbits);
9305 	nfsrv_putattrbit(nd, &attrbits);
9306 	error = nfscl_request(nd, vp, p, cred);
9307 	if (error != 0)
9308 		return (error);
9309 	*eofp = true;
9310 	*lenp = 0;
9311 	if (nd->nd_repstat == 0) {
9312 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
9313 		*cookiep = fxdr_hyper(tl); tl += 2;
9314 		cnt = fxdr_unsigned(int, *tl);
9315 		if (cnt < 0) {
9316 			error = EBADRPC;
9317 			goto nfsmout;
9318 		}
9319 		for (i = 0; i < cnt; i++) {
9320 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
9321 			len = fxdr_unsigned(int, *tl);
9322 			if (len <= 0 || len > EXTATTR_MAXNAMELEN) {
9323 				error = EBADRPC;
9324 				goto nfsmout;
9325 			}
9326 			if (uiop == NULL)
9327 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
9328 			else if (uiop->uio_resid >= len + 1) {
9329 				c = len;
9330 				error = uiomove(&c, sizeof(c), uiop);
9331 				if (error == 0)
9332 					error = nfsm_mbufuio(nd, uiop, len);
9333 			} else {
9334 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
9335 				*eofp = false;
9336 			}
9337 			if (error != 0)
9338 				goto nfsmout;
9339 			*lenp += (len + 1);
9340 		}
9341 		/* Get the eof and skip over the Getattr op status. */
9342 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
9343 		/*
9344 		 * *eofp is set false above, because it wasn't able to copy
9345 		 * all of the reply.
9346 		 */
9347 		if (*eofp && *tl == 0)
9348 			*eofp = false;
9349 		error = nfsm_loadattr(nd, nap);
9350 		if (error == 0)
9351 			*attrflagp = 1;
9352 	}
9353 	if (error == 0)
9354 		error = nd->nd_repstat;
9355 nfsmout:
9356 	m_freem(nd->nd_mrep);
9357 	return (error);
9358 }
9359 
9360 /*
9361  * Split an mbuf list.  For non-M_EXTPG mbufs, just use m_split().
9362  */
9363 static struct mbuf *
9364 nfsm_split(struct mbuf *mp, uint64_t xfer)
9365 {
9366 	struct mbuf *m, *m2;
9367 	vm_page_t pg;
9368 	int i, j, left, pgno, plen, trim;
9369 	char *cp, *cp2;
9370 
9371 	if ((mp->m_flags & M_EXTPG) == 0) {
9372 		m = m_split(mp, xfer, M_WAITOK);
9373 		return (m);
9374 	}
9375 
9376 	/* Find the correct mbuf to split at. */
9377 	for (m = mp; m != NULL && xfer > m->m_len; m = m->m_next)
9378 		xfer -= m->m_len;
9379 	if (m == NULL)
9380 		return (NULL);
9381 
9382 	/* If xfer == m->m_len, we can just split the mbuf list. */
9383 	if (xfer == m->m_len) {
9384 		m2 = m->m_next;
9385 		m->m_next = NULL;
9386 		return (m2);
9387 	}
9388 
9389 	/* Find the page to split at. */
9390 	pgno = 0;
9391 	left = xfer;
9392 	do {
9393 		if (pgno == 0)
9394 			plen = m_epg_pagelen(m, 0, m->m_epg_1st_off);
9395 		else
9396 			plen = m_epg_pagelen(m, pgno, 0);
9397 		if (left <= plen)
9398 			break;
9399 		left -= plen;
9400 		pgno++;
9401 	} while (pgno < m->m_epg_npgs);
9402 	if (pgno == m->m_epg_npgs)
9403 		panic("nfsm_split: eroneous ext_pgs mbuf");
9404 
9405 	m2 = mb_alloc_ext_pgs(M_WAITOK, mb_free_mext_pgs, 0);
9406 	m2->m_epg_flags |= EPG_FLAG_ANON;
9407 
9408 	/*
9409 	 * If left < plen, allocate a new page for the new mbuf
9410 	 * and copy the data after left in the page to this new
9411 	 * page.
9412 	 */
9413 	if (left < plen) {
9414 		pg = vm_page_alloc_noobj(VM_ALLOC_WAITOK | VM_ALLOC_NODUMP |
9415 		    VM_ALLOC_WIRED);
9416 		m2->m_epg_pa[0] = VM_PAGE_TO_PHYS(pg);
9417 		m2->m_epg_npgs = 1;
9418 
9419 		/* Copy the data after left to the new page. */
9420 		trim = plen - left;
9421 		cp = (char *)(void *)PHYS_TO_DMAP(m->m_epg_pa[pgno]);
9422 		if (pgno == 0)
9423 			cp += m->m_epg_1st_off;
9424 		cp += left;
9425 		cp2 = (char *)(void *)PHYS_TO_DMAP(m2->m_epg_pa[0]);
9426 		if (pgno == m->m_epg_npgs - 1)
9427 			m2->m_epg_last_len = trim;
9428 		else {
9429 			cp2 += PAGE_SIZE - trim;
9430 			m2->m_epg_1st_off = PAGE_SIZE - trim;
9431 			m2->m_epg_last_len = m->m_epg_last_len;
9432 		}
9433 		memcpy(cp2, cp, trim);
9434 		m2->m_len = trim;
9435 	} else {
9436 		m2->m_len = 0;
9437 		m2->m_epg_last_len = m->m_epg_last_len;
9438 	}
9439 
9440 	/* Move the pages beyond pgno to the new mbuf. */
9441 	for (i = pgno + 1, j = m2->m_epg_npgs; i < m->m_epg_npgs; i++, j++) {
9442 		m2->m_epg_pa[j] = m->m_epg_pa[i];
9443 		/* Never moves page 0. */
9444 		m2->m_len += m_epg_pagelen(m, i, 0);
9445 	}
9446 	m2->m_epg_npgs = j;
9447 	m->m_epg_npgs = pgno + 1;
9448 	m->m_epg_last_len = left;
9449 	m->m_len = xfer;
9450 
9451 	m2->m_next = m->m_next;
9452 	m->m_next = NULL;
9453 	return (m2);
9454 }
9455 
9456 /*
9457  * Do the NFSv4.1 Bind Connection to Session.
9458  * Called from the reconnect layer of the krpc (sys/rpc/clnt_rc.c).
9459  */
9460 void
9461 nfsrpc_bindconnsess(CLIENT *cl, void *arg, struct ucred *cr)
9462 {
9463 	struct nfscl_reconarg *rcp = (struct nfscl_reconarg *)arg;
9464 	uint32_t res, *tl;
9465 	struct nfsrv_descript nfsd;
9466 	struct nfsrv_descript *nd = &nfsd;
9467 	struct rpc_callextra ext;
9468 	struct timeval utimeout;
9469 	enum clnt_stat stat;
9470 	int error;
9471 
9472 	nfscl_reqstart(nd, NFSPROC_BINDCONNTOSESS, NULL, NULL, 0, NULL, NULL,
9473 	    NFS_VER4, rcp->minorvers, NULL);
9474 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID + 2 * NFSX_UNSIGNED);
9475 	memcpy(tl, rcp->sessionid, NFSX_V4SESSIONID);
9476 	tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
9477 	*tl++ = txdr_unsigned(NFSCDFC4_FORE_OR_BOTH);
9478 	*tl = newnfs_false;
9479 
9480 	memset(&ext, 0, sizeof(ext));
9481 	utimeout.tv_sec = 30;
9482 	utimeout.tv_usec = 0;
9483 	ext.rc_auth = authunix_create(cr);
9484 	nd->nd_mrep = NULL;
9485 	stat = CLNT_CALL_MBUF(cl, &ext, NFSV4PROC_COMPOUND, nd->nd_mreq,
9486 	    &nd->nd_mrep, utimeout);
9487 	AUTH_DESTROY(ext.rc_auth);
9488 	if (stat != RPC_SUCCESS) {
9489 		printf("nfsrpc_bindconnsess: call failed stat=%d\n", stat);
9490 		return;
9491 	}
9492 	if (nd->nd_mrep == NULL) {
9493 		printf("nfsrpc_bindconnsess: no reply args\n");
9494 		return;
9495 	}
9496 	error = 0;
9497 	newnfs_realign(&nd->nd_mrep, M_WAITOK);
9498 	nd->nd_md = nd->nd_mrep;
9499 	nd->nd_dpos = mtod(nd->nd_md, char *);
9500 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
9501 	nd->nd_repstat = fxdr_unsigned(uint32_t, *tl++);
9502 	if (nd->nd_repstat == NFSERR_OK) {
9503 		res = fxdr_unsigned(uint32_t, *tl);
9504 		if (res > 0 && (error = nfsm_advance(nd, NFSM_RNDUP(res),
9505 		    -1)) != 0)
9506 			goto nfsmout;
9507 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
9508 		    4 * NFSX_UNSIGNED);
9509 		tl += 3;
9510 		if (!NFSBCMP(tl, rcp->sessionid, NFSX_V4SESSIONID)) {
9511 			tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
9512 			res = fxdr_unsigned(uint32_t, *tl);
9513 			if (res != NFSCDFS4_BOTH)
9514 				printf("nfsrpc_bindconnsess: did not "
9515 				    "return FS4_BOTH\n");
9516 		} else
9517 			printf("nfsrpc_bindconnsess: not same "
9518 			    "sessionid\n");
9519 	} else if (nd->nd_repstat != NFSERR_BADSESSION)
9520 		printf("nfsrpc_bindconnsess: returned %d\n", nd->nd_repstat);
9521 nfsmout:
9522 	if (error != 0)
9523 		printf("nfsrpc_bindconnsess: reply bad xdr\n");
9524 	m_freem(nd->nd_mrep);
9525 }
9526 
9527 /*
9528  * Do roughly what nfs_statfs() does for NFSv4, but when called with a shared
9529  * locked vnode.
9530  */
9531 static void
9532 nfscl_statfs(struct vnode *vp, struct ucred *cred, NFSPROC_T *td)
9533 {
9534 	struct nfsvattr nfsva;
9535 	struct nfsfsinfo fs;
9536 	struct nfsstatfs sb;
9537 	struct mount *mp;
9538 	struct nfsmount *nmp;
9539 	uint32_t lease;
9540 	int attrflag, error;
9541 
9542 	mp = vp->v_mount;
9543 	nmp = VFSTONFS(mp);
9544 	error = nfsrpc_statfs(vp, &sb, &fs, &lease, cred, td, &nfsva,
9545 	    &attrflag);
9546 	if (attrflag != 0)
9547 		(void) nfscl_loadattrcache(&vp, &nfsva, NULL, 0, 1);
9548 	if (error == 0) {
9549 		NFSLOCKCLSTATE();
9550 		if (nmp->nm_clp != NULL)
9551 			nmp->nm_clp->nfsc_renew = NFSCL_RENEW(lease);
9552 		NFSUNLOCKCLSTATE();
9553 		mtx_lock(&nmp->nm_mtx);
9554 		nfscl_loadfsinfo(nmp, &fs);
9555 		nfscl_loadsbinfo(nmp, &sb, &mp->mnt_stat);
9556 		mp->mnt_stat.f_iosize = newnfs_iosize(nmp);
9557 		mtx_unlock(&nmp->nm_mtx);
9558 	}
9559 }
9560