xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision 0a7fa9d11b2007331857c9575bd9b77d86c88fe4)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Rick Macklem at The University of Guelph.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  */
35 
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
38 
39 /*
40  * Rpc op calls, generally called from the vnode op calls or through the
41  * buffer cache, for NFS v2, 3 and 4.
42  * These do not normally make any changes to vnode arguments or use
43  * structures that might change between the VFS variants. The returned
44  * arguments are all at the end, after the NFSPROC_T *p one.
45  */
46 
47 #include "opt_inet6.h"
48 
49 #include <fs/nfs/nfsport.h>
50 #include <fs/nfsclient/nfs.h>
51 #include <sys/extattr.h>
52 #include <sys/sysctl.h>
53 #include <sys/taskqueue.h>
54 
55 SYSCTL_DECL(_vfs_nfs);
56 
57 static int	nfsignore_eexist = 0;
58 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
59     &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
60 
61 static int	nfscl_dssameconn = 0;
62 SYSCTL_INT(_vfs_nfs, OID_AUTO, dssameconn, CTLFLAG_RW,
63     &nfscl_dssameconn, 0, "Use same TCP connection to multiple DSs");
64 
65 static uint64_t nfs_maxcopyrange = SSIZE_MAX;
66 SYSCTL_U64(_vfs_nfs, OID_AUTO, maxcopyrange, CTLFLAG_RW,
67     &nfs_maxcopyrange, 0, "Max size of a Copy so RPC times reasonable");
68 
69 /*
70  * Global variables
71  */
72 extern struct nfsstatsv1 nfsstatsv1;
73 extern int nfs_numnfscbd;
74 extern struct timeval nfsboottime;
75 extern u_int32_t newnfs_false, newnfs_true;
76 extern nfstype nfsv34_type[9];
77 extern int nfsrv_useacl;
78 extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
79 extern int nfscl_debuglevel;
80 extern int nfs_pnfsiothreads;
81 extern u_long sb_max_adj;
82 NFSCLSTATEMUTEX;
83 int nfstest_outofseq = 0;
84 int nfscl_assumeposixlocks = 1;
85 int nfscl_enablecallb = 0;
86 short nfsv4_cbport = NFSV4_CBPORT;
87 int nfstest_openallsetattr = 0;
88 
89 #define	DIRHDSIZ	offsetof(struct dirent, d_name)
90 
91 /*
92  * nfscl_getsameserver() can return one of three values:
93  * NFSDSP_USETHISSESSION - Use this session for the DS.
94  * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
95  *     session.
96  * NFSDSP_NOTFOUND - No matching server was found.
97  */
98 enum nfsclds_state {
99 	NFSDSP_USETHISSESSION = 0,
100 	NFSDSP_SEQTHISSESSION = 1,
101 	NFSDSP_NOTFOUND = 2,
102 };
103 
104 /*
105  * Do a write RPC on a DS data file, using this structure for the arguments,
106  * so that this function can be executed by a separate kernel process.
107  */
108 struct nfsclwritedsdorpc {
109 	int			done;
110 	int			inprog;
111 	struct task		tsk;
112 	struct vnode		*vp;
113 	int			iomode;
114 	int			must_commit;
115 	nfsv4stateid_t		*stateidp;
116 	struct nfsclds		*dsp;
117 	uint64_t		off;
118 	int			len;
119 #ifdef notyet
120 	int			advise;
121 #endif
122 	struct nfsfh		*fhp;
123 	struct mbuf		*m;
124 	int			vers;
125 	int			minorvers;
126 	struct ucred		*cred;
127 	NFSPROC_T		*p;
128 	int			err;
129 };
130 
131 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
132     struct ucred *, NFSPROC_T *, struct nfsvattr *, int *);
133 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
134     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *);
135 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
136     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
137     int);
138 static int nfsrpc_deallocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
139     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *);
140 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
141     nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
142     struct nfsvattr *, struct nfsfh **, int *, int *);
143 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
144     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
145     NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
146     int *, int *);
147 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
148     struct nfscllockowner *, u_int64_t, u_int64_t,
149     u_int32_t, struct ucred *, NFSPROC_T *, int);
150 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
151     struct acl *, nfsv4stateid_t *);
152 static int nfsrpc_layouterror(struct nfsmount *, uint8_t *, int, uint64_t,
153     uint64_t, nfsv4stateid_t *, struct ucred *, NFSPROC_T *, uint32_t,
154     uint32_t, char *);
155 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
156     uint32_t, uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
157     struct ucred *, NFSPROC_T *);
158 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
159     struct sockaddr_in6 *, sa_family_t, int, int, struct nfsclds **,
160     NFSPROC_T *);
161 static void nfscl_initsessionslots(struct nfsclsession *);
162 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
163     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
164     struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
165     NFSPROC_T *);
166 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
167     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
168     struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
169     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
170 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
171     struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
172     struct ucred *, NFSPROC_T *);
173 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
174     nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
175     struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
176 static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
177     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
178     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
179 static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
180     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
181     struct ucred *, NFSPROC_T *);
182 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
183     struct nfsclds *, struct nfsclds **, uint32_t *);
184 static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
185     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
186     NFSPROC_T *);
187 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
188     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
189 #ifdef notyet
190 static int nfsio_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
191     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
192     NFSPROC_T *);
193 static int nfsrpc_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
194     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
195 #endif
196 static int nfsrpc_allocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
197     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *);
198 static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
199     uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
200 static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
201     NFSPROC_T *);
202 static int nfsrv_parselayoutget(struct nfsmount *, struct nfsrv_descript *,
203     nfsv4stateid_t *, int *, struct nfsclflayouthead *);
204 static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
205     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
206     struct nfscldeleg **, struct ucred *, NFSPROC_T *);
207 static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
208     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
209     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
210     struct nfsfh **, int *, int *, int *);
211 static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
212     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
213     struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
214     struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
215 static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
216     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
217     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
218     struct nfsfh **, int *, int *, int *, nfsv4stateid_t *,
219     int, int, int, int *, struct nfsclflayouthead *, int *);
220 static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
221     uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
222     struct nfsclflayouthead *, struct ucred *, NFSPROC_T *);
223 static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
224     int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
225     struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
226 static int nfsrpc_copyrpc(vnode_t, off_t, vnode_t, off_t, size_t *,
227     nfsv4stateid_t *, nfsv4stateid_t *, struct nfsvattr *, int *,
228     struct nfsvattr *, int *, bool, int *, struct ucred *, NFSPROC_T *);
229 static int nfsrpc_seekrpc(vnode_t, off_t *, nfsv4stateid_t *, bool *,
230     int, struct nfsvattr *, int *, struct ucred *);
231 static struct mbuf *nfsm_split(struct mbuf *, uint64_t);
232 
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);
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 
309 	*attrflagp = 0;
310 	supported = mode;
311 	NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
312 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
313 	*tl = txdr_unsigned(mode);
314 	if (nd->nd_flag & ND_NFSV4) {
315 		/*
316 		 * And do a Getattr op.
317 		 */
318 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
319 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
320 		NFSGETATTR_ATTRBIT(&attrbits);
321 		(void) nfsrv_putattrbit(nd, &attrbits);
322 	}
323 	error = nfscl_request(nd, vp, p, cred);
324 	if (error)
325 		return (error);
326 	if (nd->nd_flag & ND_NFSV3) {
327 		error = nfscl_postop_attr(nd, nap, attrflagp);
328 		if (error)
329 			goto nfsmout;
330 	}
331 	if (!nd->nd_repstat) {
332 		if (nd->nd_flag & ND_NFSV4) {
333 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
334 			supported = fxdr_unsigned(u_int32_t, *tl++);
335 		} else {
336 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
337 		}
338 		rmode = fxdr_unsigned(u_int32_t, *tl);
339 		if (nd->nd_flag & ND_NFSV4)
340 			error = nfscl_postop_attr(nd, nap, attrflagp);
341 
342 		/*
343 		 * It's not obvious what should be done about
344 		 * unsupported access modes. For now, be paranoid
345 		 * and clear the unsupported ones.
346 		 */
347 		rmode &= supported;
348 		*rmodep = rmode;
349 	} else
350 		error = nd->nd_repstat;
351 nfsmout:
352 	m_freem(nd->nd_mrep);
353 	return (error);
354 }
355 
356 /*
357  * nfs open rpc
358  */
359 int
360 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
361 {
362 	struct nfsclopen *op;
363 	struct nfscldeleg *dp;
364 	struct nfsfh *nfhp;
365 	struct nfsnode *np = VTONFS(vp);
366 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
367 	u_int32_t mode, clidrev;
368 	int ret, newone, error, expireret = 0, retrycnt;
369 
370 	/*
371 	 * For NFSv4, Open Ops are only done on Regular Files.
372 	 */
373 	if (vp->v_type != VREG)
374 		return (0);
375 	mode = 0;
376 	if (amode & FREAD)
377 		mode |= NFSV4OPEN_ACCESSREAD;
378 	if (amode & FWRITE)
379 		mode |= NFSV4OPEN_ACCESSWRITE;
380 	nfhp = np->n_fhp;
381 
382 	retrycnt = 0;
383 #ifdef notdef
384 { char name[100]; int namel;
385 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
386 bcopy(NFS4NODENAME(np->n_v4), name, namel);
387 name[namel] = '\0';
388 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
389 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
390 else printf(" fhl=0\n");
391 }
392 #endif
393 	do {
394 	    dp = NULL;
395 	    error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
396 		cred, p, NULL, &op, &newone, &ret, 1, true);
397 	    if (error) {
398 		return (error);
399 	    }
400 	    if (nmp->nm_clp != NULL)
401 		clidrev = nmp->nm_clp->nfsc_clientidrev;
402 	    else
403 		clidrev = 0;
404 	    if (ret == NFSCLOPEN_DOOPEN) {
405 		if (np->n_v4 != NULL) {
406 			/*
407 			 * For the first attempt, try and get a layout, if
408 			 * pNFS is enabled for the mount.
409 			 */
410 			if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
411 			    nfs_numnfscbd == 0 ||
412 			    (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
413 				error = nfsrpc_openrpc(nmp, vp,
414 				    np->n_v4->n4_data,
415 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
416 				    np->n_fhp->nfh_len, mode, op,
417 				    NFS4NODENAME(np->n_v4),
418 				    np->n_v4->n4_namelen,
419 				    &dp, 0, 0x0, cred, p, 0, 0);
420 			else
421 				error = nfsrpc_getopenlayout(nmp, vp,
422 				    np->n_v4->n4_data,
423 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
424 				    np->n_fhp->nfh_len, mode, op,
425 				    NFS4NODENAME(np->n_v4),
426 				    np->n_v4->n4_namelen, &dp, cred, p);
427 			if (dp != NULL) {
428 				NFSLOCKNODE(np);
429 				np->n_flag &= ~NDELEGMOD;
430 				/*
431 				 * Invalidate the attribute cache, so that
432 				 * attributes that pre-date the issue of a
433 				 * delegation are not cached, since the
434 				 * cached attributes will remain valid while
435 				 * the delegation is held.
436 				 */
437 				NFSINVALATTRCACHE(np);
438 				NFSUNLOCKNODE(np);
439 				(void) nfscl_deleg(nmp->nm_mountp,
440 				    op->nfso_own->nfsow_clp,
441 				    nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
442 			}
443 		} else {
444 			error = EIO;
445 		}
446 		newnfs_copyincred(cred, &op->nfso_cred);
447 	    } else if (ret == NFSCLOPEN_SETCRED)
448 		/*
449 		 * This is a new local open on a delegation. It needs
450 		 * to have credentials so that an open can be done
451 		 * against the server during recovery.
452 		 */
453 		newnfs_copyincred(cred, &op->nfso_cred);
454 
455 	    /*
456 	     * nfso_opencnt is the count of how many VOP_OPEN()s have
457 	     * been done on this Open successfully and a VOP_CLOSE()
458 	     * is expected for each of these.
459 	     * If error is non-zero, don't increment it, since the Open
460 	     * hasn't succeeded yet.
461 	     */
462 	    if (!error) {
463 		op->nfso_opencnt++;
464 		if (NFSHASNFSV4N(nmp) && NFSHASONEOPENOWN(nmp)) {
465 		    NFSLOCKNODE(np);
466 		    np->n_openstateid = op;
467 		    NFSUNLOCKNODE(np);
468 		}
469 	    }
470 	    nfscl_openrelease(nmp, op, error, newone);
471 	    if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
472 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
473 		error == NFSERR_BADSESSION) {
474 		(void) nfs_catnap(PZERO, error, "nfs_open");
475 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
476 		&& clidrev != 0) {
477 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
478 		retrycnt++;
479 	    }
480 	} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
481 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
482 	    error == NFSERR_BADSESSION ||
483 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
484 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
485 	if (error && retrycnt >= 4)
486 		error = EIO;
487 	return (error);
488 }
489 
490 /*
491  * the actual open rpc
492  */
493 int
494 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
495     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
496     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
497     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
498     int syscred, int recursed)
499 {
500 	u_int32_t *tl;
501 	struct nfsrv_descript nfsd, *nd = &nfsd;
502 	struct nfscldeleg *dp, *ndp = NULL;
503 	struct nfsvattr nfsva;
504 	u_int32_t rflags, deleg;
505 	nfsattrbit_t attrbits;
506 	int error, ret, acesize, limitby;
507 	struct nfsclsession *tsep;
508 
509 	dp = *dpp;
510 	*dpp = NULL;
511 	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
512 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
513 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
514 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
515 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
516 	tsep = nfsmnt_mdssession(nmp);
517 	*tl++ = tsep->nfsess_clientid.lval[0];
518 	*tl = tsep->nfsess_clientid.lval[1];
519 	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
520 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
521 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
522 	if (reclaim) {
523 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
524 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
525 		*tl = txdr_unsigned(delegtype);
526 	} else {
527 		if (dp != NULL) {
528 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
529 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
530 			if (NFSHASNFSV4N(nmp))
531 				*tl++ = 0;
532 			else
533 				*tl++ = dp->nfsdl_stateid.seqid;
534 			*tl++ = dp->nfsdl_stateid.other[0];
535 			*tl++ = dp->nfsdl_stateid.other[1];
536 			*tl = dp->nfsdl_stateid.other[2];
537 		} else {
538 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
539 		}
540 		(void) nfsm_strtom(nd, name, namelen);
541 	}
542 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
543 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
544 	NFSZERO_ATTRBIT(&attrbits);
545 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
546 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
547 	(void) nfsrv_putattrbit(nd, &attrbits);
548 	if (syscred)
549 		nd->nd_flag |= ND_USEGSSNAME;
550 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
551 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
552 	if (error)
553 		return (error);
554 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
555 	if (!nd->nd_repstat) {
556 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
557 		    6 * NFSX_UNSIGNED);
558 		op->nfso_stateid.seqid = *tl++;
559 		op->nfso_stateid.other[0] = *tl++;
560 		op->nfso_stateid.other[1] = *tl++;
561 		op->nfso_stateid.other[2] = *tl;
562 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
563 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
564 		if (error)
565 			goto nfsmout;
566 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
567 		deleg = fxdr_unsigned(u_int32_t, *tl);
568 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
569 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
570 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
571 			      NFSCLFLAGS_FIRSTDELEG))
572 				op->nfso_own->nfsow_clp->nfsc_flags |=
573 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
574 			ndp = malloc(
575 			    sizeof (struct nfscldeleg) + newfhlen,
576 			    M_NFSCLDELEG, M_WAITOK);
577 			LIST_INIT(&ndp->nfsdl_owner);
578 			LIST_INIT(&ndp->nfsdl_lock);
579 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
580 			ndp->nfsdl_fhlen = newfhlen;
581 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
582 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
583 			nfscl_lockinit(&ndp->nfsdl_rwlock);
584 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
585 			    NFSX_UNSIGNED);
586 			ndp->nfsdl_stateid.seqid = *tl++;
587 			ndp->nfsdl_stateid.other[0] = *tl++;
588 			ndp->nfsdl_stateid.other[1] = *tl++;
589 			ndp->nfsdl_stateid.other[2] = *tl++;
590 			ret = fxdr_unsigned(int, *tl);
591 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
592 				ndp->nfsdl_flags = NFSCLDL_WRITE;
593 				/*
594 				 * Indicates how much the file can grow.
595 				 */
596 				NFSM_DISSECT(tl, u_int32_t *,
597 				    3 * NFSX_UNSIGNED);
598 				limitby = fxdr_unsigned(int, *tl++);
599 				switch (limitby) {
600 				case NFSV4OPEN_LIMITSIZE:
601 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
602 					break;
603 				case NFSV4OPEN_LIMITBLOCKS:
604 					ndp->nfsdl_sizelimit =
605 					    fxdr_unsigned(u_int64_t, *tl++);
606 					ndp->nfsdl_sizelimit *=
607 					    fxdr_unsigned(u_int64_t, *tl);
608 					break;
609 				default:
610 					error = NFSERR_BADXDR;
611 					goto nfsmout;
612 				}
613 			} else {
614 				ndp->nfsdl_flags = NFSCLDL_READ;
615 			}
616 			if (ret)
617 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
618 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, false,
619 			    &ret, &acesize, p);
620 			if (error)
621 				goto nfsmout;
622 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
623 			error = NFSERR_BADXDR;
624 			goto nfsmout;
625 		}
626 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
627 		error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
628 		    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
629 		    NULL, NULL, NULL, p, cred);
630 		if (error)
631 			goto nfsmout;
632 		if (ndp != NULL) {
633 			ndp->nfsdl_change = nfsva.na_filerev;
634 			ndp->nfsdl_modtime = nfsva.na_mtime;
635 			ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
636 		}
637 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
638 		    do {
639 			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
640 			    cred, p);
641 			if (ret == NFSERR_DELAY)
642 			    (void) nfs_catnap(PZERO, ret, "nfs_open");
643 		    } while (ret == NFSERR_DELAY);
644 		    error = ret;
645 		}
646 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
647 		    nfscl_assumeposixlocks)
648 		    op->nfso_posixlock = 1;
649 		else
650 		    op->nfso_posixlock = 0;
651 
652 		/*
653 		 * If the server is handing out delegations, but we didn't
654 		 * get one because an OpenConfirm was required, try the
655 		 * Open again, to get a delegation. This is a harmless no-op,
656 		 * from a server's point of view.
657 		 */
658 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
659 		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
660 		    && !error && dp == NULL && ndp == NULL && !recursed) {
661 		    do {
662 			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
663 			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
664 			    cred, p, syscred, 1);
665 			if (ret == NFSERR_DELAY)
666 			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
667 		    } while (ret == NFSERR_DELAY);
668 		    if (ret) {
669 			if (ndp != NULL) {
670 				free(ndp, M_NFSCLDELEG);
671 				ndp = NULL;
672 			}
673 			if (ret == NFSERR_STALECLIENTID ||
674 			    ret == NFSERR_STALEDONTRECOVER ||
675 			    ret == NFSERR_BADSESSION)
676 				error = ret;
677 		    }
678 		}
679 	}
680 	if (nd->nd_repstat != 0 && error == 0)
681 		error = nd->nd_repstat;
682 	if (error == NFSERR_STALECLIENTID)
683 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
684 nfsmout:
685 	if (!error)
686 		*dpp = ndp;
687 	else if (ndp != NULL)
688 		free(ndp, M_NFSCLDELEG);
689 	m_freem(nd->nd_mrep);
690 	return (error);
691 }
692 
693 /*
694  * open downgrade rpc
695  */
696 int
697 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
698     struct ucred *cred, NFSPROC_T *p)
699 {
700 	u_int32_t *tl;
701 	struct nfsrv_descript nfsd, *nd = &nfsd;
702 	int error;
703 
704 	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
705 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
706 	if (NFSHASNFSV4N(VFSTONFS(vp->v_mount)))
707 		*tl++ = 0;
708 	else
709 		*tl++ = op->nfso_stateid.seqid;
710 	*tl++ = op->nfso_stateid.other[0];
711 	*tl++ = op->nfso_stateid.other[1];
712 	*tl++ = op->nfso_stateid.other[2];
713 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
714 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
715 	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
716 	error = nfscl_request(nd, vp, p, cred);
717 	if (error)
718 		return (error);
719 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
720 	if (!nd->nd_repstat) {
721 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
722 		op->nfso_stateid.seqid = *tl++;
723 		op->nfso_stateid.other[0] = *tl++;
724 		op->nfso_stateid.other[1] = *tl++;
725 		op->nfso_stateid.other[2] = *tl;
726 	}
727 	if (nd->nd_repstat && error == 0)
728 		error = nd->nd_repstat;
729 	if (error == NFSERR_STALESTATEID)
730 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
731 nfsmout:
732 	m_freem(nd->nd_mrep);
733 	return (error);
734 }
735 
736 /*
737  * V4 Close operation.
738  */
739 int
740 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
741 {
742 	struct nfsclclient *clp;
743 	int error;
744 
745 	if (vp->v_type != VREG)
746 		return (0);
747 	if (doclose)
748 		error = nfscl_doclose(vp, &clp, p);
749 	else {
750 		error = nfscl_getclose(vp, &clp);
751 		if (error == 0)
752 			nfscl_clientrelease(clp);
753 	}
754 	return (error);
755 }
756 
757 /*
758  * Close the open.
759  */
760 int
761 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p,
762     bool loop_on_delayed, bool freeop)
763 {
764 	struct nfsrv_descript nfsd, *nd = &nfsd;
765 	struct nfscllockowner *lp, *nlp;
766 	struct nfscllock *lop, *nlop;
767 	struct ucred *tcred;
768 	u_int64_t off = 0, len = 0;
769 	u_int32_t type = NFSV4LOCKT_READ;
770 	int error, do_unlock, trycnt;
771 
772 	tcred = newnfs_getcred();
773 	newnfs_copycred(&op->nfso_cred, tcred);
774 	/*
775 	 * (Theoretically this could be done in the same
776 	 *  compound as the close, but having multiple
777 	 *  sequenced Ops in the same compound might be
778 	 *  too scary for some servers.)
779 	 */
780 	if (op->nfso_posixlock) {
781 		off = 0;
782 		len = NFS64BITSSET;
783 		type = NFSV4LOCKT_READ;
784 	}
785 
786 	/*
787 	 * Since this function is only called from VOP_INACTIVE(), no
788 	 * other thread will be manipulating this Open. As such, the
789 	 * lock lists are not being changed by other threads, so it should
790 	 * be safe to do this without locking.
791 	 */
792 	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
793 		do_unlock = 1;
794 		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
795 			if (op->nfso_posixlock == 0) {
796 				off = lop->nfslo_first;
797 				len = lop->nfslo_end - lop->nfslo_first;
798 				if (lop->nfslo_type == F_WRLCK)
799 					type = NFSV4LOCKT_WRITE;
800 				else
801 					type = NFSV4LOCKT_READ;
802 			}
803 			if (do_unlock) {
804 				trycnt = 0;
805 				do {
806 					error = nfsrpc_locku(nd, nmp, lp, off,
807 					    len, type, tcred, p, 0);
808 					if ((nd->nd_repstat == NFSERR_GRACE ||
809 					    nd->nd_repstat == NFSERR_DELAY) &&
810 					    error == 0)
811 						(void) nfs_catnap(PZERO,
812 						    (int)nd->nd_repstat,
813 						    "nfs_close");
814 				} while ((nd->nd_repstat == NFSERR_GRACE ||
815 				    nd->nd_repstat == NFSERR_DELAY) &&
816 				    error == 0 && trycnt++ < 5);
817 				if (op->nfso_posixlock)
818 					do_unlock = 0;
819 			}
820 			nfscl_freelock(lop, 0);
821 		}
822 		/*
823 		 * Do a ReleaseLockOwner.
824 		 * The lock owner name nfsl_owner may be used by other opens for
825 		 * other files but the lock_owner4 name that nfsrpc_rellockown()
826 		 * puts on the wire has the file handle for this file appended
827 		 * to it, so it can be done now.
828 		 */
829 		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
830 		    lp->nfsl_open->nfso_fhlen, tcred, p);
831 	}
832 
833 	/*
834 	 * There could be other Opens for different files on the same
835 	 * OpenOwner, so locking is required.
836 	 */
837 	NFSLOCKCLSTATE();
838 	nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
839 	NFSUNLOCKCLSTATE();
840 	do {
841 		error = nfscl_tryclose(op, tcred, nmp, p, loop_on_delayed);
842 		if (error == NFSERR_GRACE)
843 			(void) nfs_catnap(PZERO, error, "nfs_close");
844 	} while (error == NFSERR_GRACE);
845 	NFSLOCKCLSTATE();
846 	nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
847 
848 	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
849 		nfscl_freelockowner(lp, 0);
850 	if (freeop && error != NFSERR_DELAY)
851 		nfscl_freeopen(op, 0, true);
852 	NFSUNLOCKCLSTATE();
853 	NFSFREECRED(tcred);
854 	return (error);
855 }
856 
857 /*
858  * The actual Close RPC.
859  */
860 int
861 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
862     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
863     int syscred)
864 {
865 	u_int32_t *tl;
866 	int error;
867 
868 	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
869 	    op->nfso_fhlen, NULL, NULL, 0, 0);
870 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
871 	if (NFSHASNFSV4N(nmp)) {
872 		*tl++ = 0;
873 		*tl++ = 0;
874 	} else {
875 		*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
876 		*tl++ = op->nfso_stateid.seqid;
877 	}
878 	*tl++ = op->nfso_stateid.other[0];
879 	*tl++ = op->nfso_stateid.other[1];
880 	*tl = op->nfso_stateid.other[2];
881 	if (syscred)
882 		nd->nd_flag |= ND_USEGSSNAME;
883 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
884 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
885 	if (error)
886 		return (error);
887 	if (!NFSHASNFSV4N(nmp))
888 		NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
889 	if (nd->nd_repstat == 0)
890 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
891 	error = nd->nd_repstat;
892 	if (!NFSHASNFSV4N(nmp) && error == NFSERR_STALESTATEID)
893 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
894 nfsmout:
895 	m_freem(nd->nd_mrep);
896 	return (error);
897 }
898 
899 /*
900  * V4 Open Confirm RPC.
901  */
902 int
903 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
904     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
905 {
906 	u_int32_t *tl;
907 	struct nfsrv_descript nfsd, *nd = &nfsd;
908 	struct nfsmount *nmp;
909 	int error;
910 
911 	nmp = VFSTONFS(vp->v_mount);
912 	if (NFSHASNFSV4N(nmp))
913 		return (0);		/* No confirmation for NFSv4.1. */
914 	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
915 	    0, 0);
916 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
917 	*tl++ = op->nfso_stateid.seqid;
918 	*tl++ = op->nfso_stateid.other[0];
919 	*tl++ = op->nfso_stateid.other[1];
920 	*tl++ = op->nfso_stateid.other[2];
921 	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
922 	error = nfscl_request(nd, vp, p, cred);
923 	if (error)
924 		return (error);
925 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
926 	if (!nd->nd_repstat) {
927 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
928 		op->nfso_stateid.seqid = *tl++;
929 		op->nfso_stateid.other[0] = *tl++;
930 		op->nfso_stateid.other[1] = *tl++;
931 		op->nfso_stateid.other[2] = *tl;
932 	}
933 	error = nd->nd_repstat;
934 	if (error == NFSERR_STALESTATEID)
935 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
936 nfsmout:
937 	m_freem(nd->nd_mrep);
938 	return (error);
939 }
940 
941 /*
942  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
943  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
944  */
945 int
946 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
947     bool *retokp, struct ucred *cred, NFSPROC_T *p)
948 {
949 	u_int32_t *tl;
950 	struct nfsrv_descript nfsd;
951 	struct nfsrv_descript *nd = &nfsd;
952 	u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
953 	u_short port;
954 	int error, isinet6 = 0, callblen;
955 	nfsquad_t confirm;
956 	static u_int32_t rev = 0;
957 	struct nfsclds *dsp, *odsp;
958 	struct in6_addr a6;
959 	struct nfsclsession *tsep;
960 	struct rpc_reconupcall recon;
961 	struct nfscl_reconarg *rcp;
962 
963 	if (nfsboottime.tv_sec == 0)
964 		NFSSETBOOTTIME(nfsboottime);
965 	if (NFSHASNFSV4N(nmp)) {
966 		error = NFSERR_BADSESSION;
967 		odsp = dsp = NULL;
968 		if (retokp != NULL) {
969 			NFSLOCKMNT(nmp);
970 			odsp = TAILQ_FIRST(&nmp->nm_sess);
971 			NFSUNLOCKMNT(nmp);
972 		}
973 		if (odsp != NULL) {
974 			/*
975 			 * When a session already exists, first try a
976 			 * CreateSession with the extant ClientID.
977 			 */
978 			dsp = malloc(sizeof(struct nfsclds) +
979 			    odsp->nfsclds_servownlen + 1, M_NFSCLDS,
980 			    M_WAITOK | M_ZERO);
981 			dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
982 			dsp->nfsclds_servownlen = odsp->nfsclds_servownlen;
983 			dsp->nfsclds_sess.nfsess_clientid =
984 			    odsp->nfsclds_sess.nfsess_clientid;
985 			dsp->nfsclds_sess.nfsess_sequenceid =
986 			    odsp->nfsclds_sess.nfsess_sequenceid;
987 			dsp->nfsclds_flags = odsp->nfsclds_flags;
988 			if (dsp->nfsclds_servownlen > 0)
989 				memcpy(dsp->nfsclds_serverown,
990 				    odsp->nfsclds_serverown,
991 				    dsp->nfsclds_servownlen + 1);
992 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
993 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
994 			    NULL, MTX_DEF);
995 			nfscl_initsessionslots(&dsp->nfsclds_sess);
996 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
997 			    &nmp->nm_sockreq, NULL,
998 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
999 			NFSCL_DEBUG(1, "create session for extant "
1000 			    "ClientID=%d\n", error);
1001 			if (error != 0) {
1002 				nfscl_freenfsclds(dsp);
1003 				dsp = NULL;
1004 				/*
1005 				 * If *retokp is true, return any error other
1006 				 * than NFSERR_STALECLIENTID,
1007 				 * NFSERR_BADSESSION or NFSERR_STALEDONTRECOVER
1008 				 * so that nfscl_recover() will not loop.
1009 				 */
1010 				if (*retokp)
1011 					return (NFSERR_IO);
1012 			} else
1013 				*retokp = true;
1014 		} else if (retokp != NULL && *retokp)
1015 			return (NFSERR_IO);
1016 		if (error != 0) {
1017 			/*
1018 			 * Either there was no previous session or the
1019 			 * CreateSession attempt failed, so...
1020 			 * do an ExchangeID followed by the CreateSession.
1021 			 */
1022 			clp->nfsc_rev = rev++;
1023 			error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq, 0,
1024 			    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp,
1025 			    cred, p);
1026 			NFSCL_DEBUG(1, "aft exch=%d\n", error);
1027 			if (error == 0)
1028 				error = nfsrpc_createsession(nmp,
1029 				    &dsp->nfsclds_sess, &nmp->nm_sockreq, NULL,
1030 				    dsp->nfsclds_sess.nfsess_sequenceid, 1,
1031 				    cred, p);
1032 			NFSCL_DEBUG(1, "aft createsess=%d\n", error);
1033 		}
1034 		if (error == 0) {
1035 			/*
1036 			 * If the session supports a backchannel, set up
1037 			 * the BindConnectionToSession call in the krpc
1038 			 * so that it is done on a reconnection.
1039 			 */
1040 			if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0) {
1041 				rcp = mem_alloc(sizeof(*rcp));
1042 				rcp->minorvers = nmp->nm_minorvers;
1043 				memcpy(rcp->sessionid,
1044 				    dsp->nfsclds_sess.nfsess_sessionid,
1045 				    NFSX_V4SESSIONID);
1046 				recon.call = nfsrpc_bindconnsess;
1047 				recon.arg = rcp;
1048 				CLNT_CONTROL(nmp->nm_client, CLSET_RECONUPCALL,
1049 				    &recon);
1050 			}
1051 
1052 			NFSLOCKMNT(nmp);
1053 			/*
1054 			 * The old sessions cannot be safely free'd
1055 			 * here, since they may still be used by
1056 			 * in-progress RPCs.
1057 			 */
1058 			tsep = NULL;
1059 			if (TAILQ_FIRST(&nmp->nm_sess) != NULL) {
1060 				/*
1061 				 * Mark the old session defunct.  Needed
1062 				 * when called from nfscl_hasexpired().
1063 				 */
1064 				tsep = NFSMNT_MDSSESSION(nmp);
1065 				tsep->nfsess_defunct = 1;
1066 			}
1067 			TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
1068 			    nfsclds_list);
1069 			/*
1070 			 * Wake up RPCs waiting for a slot on the
1071 			 * old session. These will then fail with
1072 			 * NFSERR_BADSESSION and be retried with the
1073 			 * new session by nfsv4_setsequence().
1074 			 * Also wakeup() processes waiting for the
1075 			 * new session.
1076 			 */
1077 			if (tsep != NULL)
1078 				wakeup(&tsep->nfsess_slots);
1079 			wakeup(&nmp->nm_sess);
1080 			NFSUNLOCKMNT(nmp);
1081 		} else if (dsp != NULL)
1082 			nfscl_freenfsclds(dsp);
1083 		if (error == 0 && reclaim == 0) {
1084 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
1085 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
1086 			if (error == NFSERR_COMPLETEALREADY ||
1087 			    error == NFSERR_NOTSUPP)
1088 				/* Ignore this error. */
1089 				error = 0;
1090 		}
1091 		return (error);
1092 	} else if (retokp != NULL && *retokp)
1093 		return (NFSERR_IO);
1094 	clp->nfsc_rev = rev++;
1095 
1096 	/*
1097 	 * Allocate a single session structure for NFSv4.0, because some of
1098 	 * the fields are used by NFSv4.0 although it doesn't do a session.
1099 	 */
1100 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
1101 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
1102 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
1103 	NFSLOCKMNT(nmp);
1104 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
1105 	tsep = NFSMNT_MDSSESSION(nmp);
1106 	NFSUNLOCKMNT(nmp);
1107 
1108 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0);
1109 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1110 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
1111 	*tl = txdr_unsigned(clp->nfsc_rev);
1112 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1113 
1114 	/*
1115 	 * set up the callback address
1116 	 */
1117 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1118 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
1119 	callblen = strlen(nfsv4_callbackaddr);
1120 	if (callblen == 0)
1121 		cp = nfscl_getmyip(nmp, &a6, &isinet6);
1122 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1123 	    (callblen > 0 || cp != NULL)) {
1124 		port = htons(nfsv4_cbport);
1125 		cp2 = (u_int8_t *)&port;
1126 #ifdef INET6
1127 		if ((callblen > 0 &&
1128 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1129 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1130 
1131 			(void) nfsm_strtom(nd, "tcp6", 4);
1132 			if (callblen == 0) {
1133 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1134 				ip6add = ip6buf;
1135 			} else {
1136 				ip6add = nfsv4_callbackaddr;
1137 			}
1138 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1139 			    ip6add, cp2[0], cp2[1]);
1140 		} else
1141 #endif
1142 		{
1143 			(void) nfsm_strtom(nd, "tcp", 3);
1144 			if (callblen == 0)
1145 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1146 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1147 				    cp[2], cp[3], cp2[0], cp2[1]);
1148 			else
1149 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1150 				    "%s.%d.%d", nfsv4_callbackaddr,
1151 				    cp2[0], cp2[1]);
1152 		}
1153 		(void) nfsm_strtom(nd, addr, strlen(addr));
1154 	} else {
1155 		(void) nfsm_strtom(nd, "tcp", 3);
1156 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1157 	}
1158 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1159 	*tl = txdr_unsigned(clp->nfsc_cbident);
1160 	nd->nd_flag |= ND_USEGSSNAME;
1161 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1162 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1163 	if (error)
1164 		return (error);
1165 	if (nd->nd_repstat == 0) {
1166 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1167 	    tsep->nfsess_clientid.lval[0] = *tl++;
1168 	    tsep->nfsess_clientid.lval[1] = *tl++;
1169 	    confirm.lval[0] = *tl++;
1170 	    confirm.lval[1] = *tl;
1171 	    m_freem(nd->nd_mrep);
1172 	    nd->nd_mrep = NULL;
1173 
1174 	    /*
1175 	     * and confirm it.
1176 	     */
1177 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1178 		NULL, 0, 0);
1179 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1180 	    *tl++ = tsep->nfsess_clientid.lval[0];
1181 	    *tl++ = tsep->nfsess_clientid.lval[1];
1182 	    *tl++ = confirm.lval[0];
1183 	    *tl = confirm.lval[1];
1184 	    nd->nd_flag |= ND_USEGSSNAME;
1185 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1186 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1187 	    if (error)
1188 		return (error);
1189 	    m_freem(nd->nd_mrep);
1190 	    nd->nd_mrep = NULL;
1191 	}
1192 	error = nd->nd_repstat;
1193 nfsmout:
1194 	m_freem(nd->nd_mrep);
1195 	return (error);
1196 }
1197 
1198 /*
1199  * nfs getattr call.
1200  */
1201 int
1202 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1203     struct nfsvattr *nap)
1204 {
1205 	struct nfsrv_descript nfsd, *nd = &nfsd;
1206 	int error;
1207 	nfsattrbit_t attrbits;
1208 
1209 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1210 	if (nd->nd_flag & ND_NFSV4) {
1211 		NFSGETATTR_ATTRBIT(&attrbits);
1212 		(void) nfsrv_putattrbit(nd, &attrbits);
1213 	}
1214 	error = nfscl_request(nd, vp, p, cred);
1215 	if (error)
1216 		return (error);
1217 	if (!nd->nd_repstat)
1218 		error = nfsm_loadattr(nd, nap);
1219 	else
1220 		error = nd->nd_repstat;
1221 	m_freem(nd->nd_mrep);
1222 	return (error);
1223 }
1224 
1225 /*
1226  * nfs getattr call with non-vnode arguments.
1227  */
1228 int
1229 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1230     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1231     uint32_t *leasep)
1232 {
1233 	struct nfsrv_descript nfsd, *nd = &nfsd;
1234 	int error, vers = NFS_VER2;
1235 	nfsattrbit_t attrbits;
1236 
1237 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0);
1238 	if (nd->nd_flag & ND_NFSV4) {
1239 		vers = NFS_VER4;
1240 		NFSGETATTR_ATTRBIT(&attrbits);
1241 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1242 		(void) nfsrv_putattrbit(nd, &attrbits);
1243 	} else if (nd->nd_flag & ND_NFSV3) {
1244 		vers = NFS_VER3;
1245 	}
1246 	if (syscred)
1247 		nd->nd_flag |= ND_USEGSSNAME;
1248 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1249 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1250 	if (error)
1251 		return (error);
1252 	if (nd->nd_repstat == 0) {
1253 		if ((nd->nd_flag & ND_NFSV4) != 0)
1254 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1255 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1256 			    NULL, NULL);
1257 		else
1258 			error = nfsm_loadattr(nd, nap);
1259 	} else
1260 		error = nd->nd_repstat;
1261 	m_freem(nd->nd_mrep);
1262 	return (error);
1263 }
1264 
1265 /*
1266  * Do an nfs setattr operation.
1267  */
1268 int
1269 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1270     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp)
1271 {
1272 	int error, expireret = 0, openerr, retrycnt;
1273 	u_int32_t clidrev = 0, mode;
1274 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1275 	struct nfsfh *nfhp;
1276 	nfsv4stateid_t stateid;
1277 	void *lckp;
1278 
1279 	if (nmp->nm_clp != NULL)
1280 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1281 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1282 		mode = NFSV4OPEN_ACCESSWRITE;
1283 	else
1284 		mode = NFSV4OPEN_ACCESSREAD;
1285 	retrycnt = 0;
1286 	do {
1287 		lckp = NULL;
1288 		openerr = 1;
1289 		if (NFSHASNFSV4(nmp)) {
1290 			nfhp = VTONFS(vp)->n_fhp;
1291 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1292 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1293 			if (error && vp->v_type == VREG &&
1294 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1295 			     nfstest_openallsetattr)) {
1296 				/*
1297 				 * No Open stateid, so try and open the file
1298 				 * now.
1299 				 */
1300 				if (mode == NFSV4OPEN_ACCESSWRITE)
1301 					openerr = nfsrpc_open(vp, FWRITE, cred,
1302 					    p);
1303 				else
1304 					openerr = nfsrpc_open(vp, FREAD, cred,
1305 					    p);
1306 				if (!openerr)
1307 					(void) nfscl_getstateid(vp,
1308 					    nfhp->nfh_fh, nfhp->nfh_len,
1309 					    mode, 0, cred, p, &stateid, &lckp);
1310 			}
1311 		}
1312 		if (vap != NULL)
1313 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1314 			    rnap, attrflagp);
1315 		else
1316 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid);
1317 		if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1318 			NFSLOCKMNT(nmp);
1319 			nmp->nm_state |= NFSSTA_OPENMODE;
1320 			NFSUNLOCKMNT(nmp);
1321 		}
1322 		if (error == NFSERR_STALESTATEID)
1323 			nfscl_initiate_recovery(nmp->nm_clp);
1324 		if (lckp != NULL)
1325 			nfscl_lockderef(lckp);
1326 		if (!openerr)
1327 			(void) nfsrpc_close(vp, 0, p);
1328 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1329 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1330 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1331 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1332 		} else if ((error == NFSERR_EXPIRED ||
1333 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1334 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1335 		}
1336 		retrycnt++;
1337 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1338 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1339 	    error == NFSERR_BADSESSION ||
1340 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1341 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1342 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1343 	    (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1344 	     retrycnt < 4));
1345 	if (error && retrycnt >= 4)
1346 		error = EIO;
1347 	return (error);
1348 }
1349 
1350 static int
1351 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1352     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1353     struct nfsvattr *rnap, int *attrflagp)
1354 {
1355 	u_int32_t *tl;
1356 	struct nfsrv_descript nfsd, *nd = &nfsd;
1357 	int error;
1358 	nfsattrbit_t attrbits;
1359 
1360 	*attrflagp = 0;
1361 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1362 	if (nd->nd_flag & ND_NFSV4)
1363 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1364 	vap->va_type = vp->v_type;
1365 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1366 	if (nd->nd_flag & ND_NFSV3) {
1367 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1368 		*tl = newnfs_false;
1369 	} else if (nd->nd_flag & ND_NFSV4) {
1370 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1371 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1372 		NFSGETATTR_ATTRBIT(&attrbits);
1373 		(void) nfsrv_putattrbit(nd, &attrbits);
1374 	}
1375 	error = nfscl_request(nd, vp, p, cred);
1376 	if (error)
1377 		return (error);
1378 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1379 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, NULL);
1380 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1381 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1382 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1383 		error = nfscl_postop_attr(nd, rnap, attrflagp);
1384 	m_freem(nd->nd_mrep);
1385 	if (nd->nd_repstat && !error)
1386 		error = nd->nd_repstat;
1387 	return (error);
1388 }
1389 
1390 /*
1391  * nfs lookup rpc
1392  */
1393 int
1394 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1395     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1396     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, uint32_t openmode)
1397 {
1398 	uint32_t deleg, rflags, *tl;
1399 	struct nfsrv_descript nfsd, *nd = &nfsd;
1400 	struct nfsmount *nmp;
1401 	struct nfsnode *np;
1402 	struct nfsfh *nfhp;
1403 	nfsattrbit_t attrbits;
1404 	int error = 0, lookupp = 0, newone, ret, retop;
1405 	uint8_t own[NFSV4CL_LOCKNAMELEN];
1406 	struct nfsclopen *op;
1407 	struct nfscldeleg *ndp;
1408 	nfsv4stateid_t stateid;
1409 
1410 	*attrflagp = 0;
1411 	*dattrflagp = 0;
1412 	if (dvp->v_type != VDIR)
1413 		return (ENOTDIR);
1414 	nmp = VFSTONFS(dvp->v_mount);
1415 	if (len > NFS_MAXNAMLEN)
1416 		return (ENAMETOOLONG);
1417 	if (NFSHASNFSV4(nmp) && len == 1 &&
1418 		name[0] == '.') {
1419 		/*
1420 		 * Just return the current dir's fh.
1421 		 */
1422 		np = VTONFS(dvp);
1423 		nfhp = malloc(sizeof (struct nfsfh) +
1424 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1425 		nfhp->nfh_len = np->n_fhp->nfh_len;
1426 		NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1427 		*nfhpp = nfhp;
1428 		return (0);
1429 	}
1430 	if (NFSHASNFSV4(nmp) && len == 2 &&
1431 		name[0] == '.' && name[1] == '.') {
1432 		lookupp = 1;
1433 		openmode = 0;
1434 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1435 	} else if (openmode != 0) {
1436 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPOPEN, dvp);
1437 		nfsm_strtom(nd, name, len);
1438 	} else {
1439 		NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1440 		(void) nfsm_strtom(nd, name, len);
1441 	}
1442 	if (nd->nd_flag & ND_NFSV4) {
1443 		NFSGETATTR_ATTRBIT(&attrbits);
1444 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1445 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1446 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1447 		(void) nfsrv_putattrbit(nd, &attrbits);
1448 		if (openmode != 0) {
1449 			/* Test for a VREG file. */
1450 			NFSZERO_ATTRBIT(&attrbits);
1451 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
1452 			NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
1453 			*tl = txdr_unsigned(NFSV4OP_VERIFY);
1454 			nfsrv_putattrbit(nd, &attrbits);
1455 			NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1456 			*tl++ = txdr_unsigned(NFSX_UNSIGNED);
1457 			*tl = vtonfsv34_type(VREG);
1458 
1459 			/* Attempt the Open for VREG. */
1460 			nfscl_filllockowner(NULL, own, F_POSIX);
1461 			NFSM_BUILD(tl, uint32_t *, 6 * NFSX_UNSIGNED);
1462 			*tl++ = txdr_unsigned(NFSV4OP_OPEN);
1463 			*tl++ = 0;		/* seqid, ignored. */
1464 			*tl++ = txdr_unsigned(openmode);
1465 			*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
1466 			*tl++ = 0;		/* ClientID, ignored. */
1467 			*tl = 0;
1468 			nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN);
1469 			NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1470 			*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
1471 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMFH);
1472 		}
1473 	}
1474 	error = nfscl_request(nd, dvp, p, cred);
1475 	if (error)
1476 		return (error);
1477 	ndp = NULL;
1478 	if (nd->nd_repstat) {
1479 		/*
1480 		 * When an NFSv4 Lookupp returns ENOENT, it means that
1481 		 * the lookup is at the root of an fs, so return this dir.
1482 		 */
1483 		if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1484 		    np = VTONFS(dvp);
1485 		    nfhp = malloc(sizeof (struct nfsfh) +
1486 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1487 		    nfhp->nfh_len = np->n_fhp->nfh_len;
1488 		    NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1489 		    *nfhpp = nfhp;
1490 		    m_freem(nd->nd_mrep);
1491 		    return (0);
1492 		}
1493 		if (nd->nd_flag & ND_NFSV3)
1494 		    error = nfscl_postop_attr(nd, dnap, dattrflagp);
1495 		else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1496 		    ND_NFSV4) {
1497 			/* Load the directory attributes. */
1498 			error = nfsm_loadattr(nd, dnap);
1499 			if (error != 0)
1500 				goto nfsmout;
1501 			*dattrflagp = 1;
1502 		}
1503 		/* Check Lookup operation reply status. */
1504 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1505 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1506 			if (*++tl != 0)
1507 				goto nfsmout;
1508 		}
1509 		/* Look for GetFH reply. */
1510 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1511 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1512 			if (*++tl != 0)
1513 				goto nfsmout;
1514 			error = nfsm_getfh(nd, nfhpp);
1515 			if (error)
1516 				goto nfsmout;
1517 		}
1518 		/* Look for Getattr reply. */
1519 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1520 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1521 			if (*++tl != 0)
1522 				goto nfsmout;
1523 			error = nfsm_loadattr(nd, nap);
1524 			if (error == 0) {
1525 				/*
1526 				 * We have now successfully completed the
1527 				 * lookup, so set nd_repstat to 0.
1528 				 */
1529 				nd->nd_repstat = 0;
1530 				*attrflagp = 1;
1531 			}
1532 		}
1533 		goto nfsmout;
1534 	}
1535 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1536 		/* Load the directory attributes. */
1537 		error = nfsm_loadattr(nd, dnap);
1538 		if (error != 0)
1539 			goto nfsmout;
1540 		*dattrflagp = 1;
1541 		/* Skip over the Lookup and GetFH operation status values. */
1542 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1543 	}
1544 	error = nfsm_getfh(nd, nfhpp);
1545 	if (error)
1546 		goto nfsmout;
1547 
1548 	error = nfscl_postop_attr(nd, nap, attrflagp);
1549 	if (openmode != 0 && error == 0) {
1550 		NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID +
1551 		    10 * NFSX_UNSIGNED);
1552 		tl += 4;	/* Skip over Verify+Open status. */
1553 		stateid.seqid = *tl++;
1554 		stateid.other[0] = *tl++;
1555 		stateid.other[1] = *tl++;
1556 		stateid.other[2] = *tl;
1557 		rflags = fxdr_unsigned(uint32_t, *(tl + 6));
1558 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1559 		if (error != 0)
1560 			goto nfsmout;
1561 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
1562 		deleg = fxdr_unsigned(uint32_t, *tl);
1563 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
1564 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
1565 			/*
1566 			 * Just need to fill in the fields used by
1567 			 * nfscl_trydelegreturn().
1568 			 * Mark the mount point as acquiring
1569 			 * delegations, so NFSPROC_LOOKUPOPEN will
1570 			 * no longer be done.
1571 			 */
1572 			NFSLOCKMNT(nmp);
1573 			nmp->nm_privflag |= NFSMNTP_DELEGISSUED;
1574 			NFSUNLOCKMNT(nmp);
1575 			ndp = malloc(sizeof(struct nfscldeleg) +
1576 			    (*nfhpp)->nfh_len, M_NFSCLDELEG, M_WAITOK);
1577 			ndp->nfsdl_fhlen = (*nfhpp)->nfh_len;
1578 			NFSBCOPY((*nfhpp)->nfh_fh, ndp->nfsdl_fh,
1579 			    ndp->nfsdl_fhlen);
1580 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
1581 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
1582 			ndp->nfsdl_stateid.seqid = *tl++;
1583 			ndp->nfsdl_stateid.other[0] = *tl++;
1584 			ndp->nfsdl_stateid.other[1] = *tl++;
1585 			ndp->nfsdl_stateid.other[2] = *tl++;
1586 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
1587 			error = NFSERR_BADXDR;
1588 			goto nfsmout;
1589 		}
1590 		ret = nfscl_open(dvp, (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len,
1591 		    openmode, 0, cred, p, NULL, &op, &newone, &retop, 1, true);
1592 		if (ret != 0)
1593 			goto nfsmout;
1594 		if (newone != 0) {
1595 			op->nfso_stateid.seqid = stateid.seqid;
1596 			op->nfso_stateid.other[0] = stateid.other[0];
1597 			op->nfso_stateid.other[1] = stateid.other[1];
1598 			op->nfso_stateid.other[2] = stateid.other[2];
1599 			op->nfso_mode = openmode;
1600 		} else {
1601 			op->nfso_stateid.seqid = stateid.seqid;
1602 			if (retop == NFSCLOPEN_DOOPEN)
1603 				op->nfso_mode |= openmode;
1604 		}
1605 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
1606 		    nfscl_assumeposixlocks)
1607 			op->nfso_posixlock = 1;
1608 		else
1609 			op->nfso_posixlock = 0;
1610 		nfscl_openrelease(nmp, op, 0, 0);
1611 		if (ndp != NULL) {
1612 			/*
1613 			 * Since we do not have the vnode, we
1614 			 * cannot invalidate cached attributes.
1615 			 * Just return the delegation.
1616 			 */
1617 			nfscl_trydelegreturn(ndp, cred, nmp, p);
1618 		}
1619 	}
1620 	if ((nd->nd_flag & ND_NFSV3) && !error)
1621 		error = nfscl_postop_attr(nd, dnap, dattrflagp);
1622 nfsmout:
1623 	m_freem(nd->nd_mrep);
1624 	if (!error && nd->nd_repstat)
1625 		error = nd->nd_repstat;
1626 	free(ndp, M_NFSCLDELEG);
1627 	return (error);
1628 }
1629 
1630 /*
1631  * Do a readlink rpc.
1632  */
1633 int
1634 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1635     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
1636 {
1637 	u_int32_t *tl;
1638 	struct nfsrv_descript nfsd, *nd = &nfsd;
1639 	struct nfsnode *np = VTONFS(vp);
1640 	nfsattrbit_t attrbits;
1641 	int error, len, cangetattr = 1;
1642 
1643 	*attrflagp = 0;
1644 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1645 	if (nd->nd_flag & ND_NFSV4) {
1646 		/*
1647 		 * And do a Getattr op.
1648 		 */
1649 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1650 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1651 		NFSGETATTR_ATTRBIT(&attrbits);
1652 		(void) nfsrv_putattrbit(nd, &attrbits);
1653 	}
1654 	error = nfscl_request(nd, vp, p, cred);
1655 	if (error)
1656 		return (error);
1657 	if (nd->nd_flag & ND_NFSV3)
1658 		error = nfscl_postop_attr(nd, nap, attrflagp);
1659 	if (!nd->nd_repstat && !error) {
1660 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1661 		/*
1662 		 * This seems weird to me, but must have been added to
1663 		 * FreeBSD for some reason. The only thing I can think of
1664 		 * is that there was/is some server that replies with
1665 		 * more link data than it should?
1666 		 */
1667 		if (len == NFS_MAXPATHLEN) {
1668 			NFSLOCKNODE(np);
1669 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1670 				len = np->n_size;
1671 				cangetattr = 0;
1672 			}
1673 			NFSUNLOCKNODE(np);
1674 		}
1675 		error = nfsm_mbufuio(nd, uiop, len);
1676 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1677 			error = nfscl_postop_attr(nd, nap, attrflagp);
1678 	}
1679 	if (nd->nd_repstat && !error)
1680 		error = nd->nd_repstat;
1681 nfsmout:
1682 	m_freem(nd->nd_mrep);
1683 	return (error);
1684 }
1685 
1686 /*
1687  * Read operation.
1688  */
1689 int
1690 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1691     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
1692 {
1693 	int error, expireret = 0, retrycnt;
1694 	u_int32_t clidrev = 0;
1695 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1696 	struct nfsnode *np = VTONFS(vp);
1697 	struct ucred *newcred;
1698 	struct nfsfh *nfhp = NULL;
1699 	nfsv4stateid_t stateid;
1700 	void *lckp;
1701 
1702 	if (nmp->nm_clp != NULL)
1703 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1704 	newcred = cred;
1705 	if (NFSHASNFSV4(nmp)) {
1706 		nfhp = np->n_fhp;
1707 		newcred = NFSNEWCRED(cred);
1708 	}
1709 	retrycnt = 0;
1710 	do {
1711 		lckp = NULL;
1712 		if (NFSHASNFSV4(nmp))
1713 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1714 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1715 			    &lckp);
1716 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1717 		    attrflagp);
1718 		if (error == NFSERR_OPENMODE) {
1719 			NFSLOCKMNT(nmp);
1720 			nmp->nm_state |= NFSSTA_OPENMODE;
1721 			NFSUNLOCKMNT(nmp);
1722 		}
1723 		if (error == NFSERR_STALESTATEID)
1724 			nfscl_initiate_recovery(nmp->nm_clp);
1725 		if (lckp != NULL)
1726 			nfscl_lockderef(lckp);
1727 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1728 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1729 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1730 			(void) nfs_catnap(PZERO, error, "nfs_read");
1731 		} else if ((error == NFSERR_EXPIRED ||
1732 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1733 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1734 		}
1735 		retrycnt++;
1736 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1737 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1738 	    error == NFSERR_BADSESSION ||
1739 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1740 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1741 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1742 	    (error == NFSERR_OPENMODE && retrycnt < 4));
1743 	if (error && retrycnt >= 4)
1744 		error = EIO;
1745 	if (NFSHASNFSV4(nmp))
1746 		NFSFREECRED(newcred);
1747 	return (error);
1748 }
1749 
1750 /*
1751  * The actual read RPC.
1752  */
1753 static int
1754 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1755     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1756     int *attrflagp)
1757 {
1758 	u_int32_t *tl;
1759 	int error = 0, len, retlen, tsiz, eof = 0;
1760 	struct nfsrv_descript nfsd;
1761 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1762 	struct nfsrv_descript *nd = &nfsd;
1763 	int rsize;
1764 	off_t tmp_off;
1765 
1766 	*attrflagp = 0;
1767 	tsiz = uiop->uio_resid;
1768 	tmp_off = uiop->uio_offset + tsiz;
1769 	NFSLOCKMNT(nmp);
1770 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1771 		NFSUNLOCKMNT(nmp);
1772 		return (EFBIG);
1773 	}
1774 	rsize = nmp->nm_rsize;
1775 	NFSUNLOCKMNT(nmp);
1776 	nd->nd_mrep = NULL;
1777 	while (tsiz > 0) {
1778 		*attrflagp = 0;
1779 		len = (tsiz > rsize) ? rsize : tsiz;
1780 		NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1781 		if (nd->nd_flag & ND_NFSV4)
1782 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1783 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1784 		if (nd->nd_flag & ND_NFSV2) {
1785 			*tl++ = txdr_unsigned(uiop->uio_offset);
1786 			*tl++ = txdr_unsigned(len);
1787 			*tl = 0;
1788 		} else {
1789 			txdr_hyper(uiop->uio_offset, tl);
1790 			*(tl + 2) = txdr_unsigned(len);
1791 		}
1792 		/*
1793 		 * Since I can't do a Getattr for NFSv4 for Write, there
1794 		 * doesn't seem any point in doing one here, either.
1795 		 * (See the comment in nfsrpc_writerpc() for more info.)
1796 		 */
1797 		error = nfscl_request(nd, vp, p, cred);
1798 		if (error)
1799 			return (error);
1800 		if (nd->nd_flag & ND_NFSV3) {
1801 			error = nfscl_postop_attr(nd, nap, attrflagp);
1802 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1803 			error = nfsm_loadattr(nd, nap);
1804 			if (!error)
1805 				*attrflagp = 1;
1806 		}
1807 		if (nd->nd_repstat || error) {
1808 			if (!error)
1809 				error = nd->nd_repstat;
1810 			goto nfsmout;
1811 		}
1812 		if (nd->nd_flag & ND_NFSV3) {
1813 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1814 			eof = fxdr_unsigned(int, *(tl + 1));
1815 		} else if (nd->nd_flag & ND_NFSV4) {
1816 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1817 			eof = fxdr_unsigned(int, *tl);
1818 		}
1819 		NFSM_STRSIZ(retlen, len);
1820 		error = nfsm_mbufuio(nd, uiop, retlen);
1821 		if (error)
1822 			goto nfsmout;
1823 		m_freem(nd->nd_mrep);
1824 		nd->nd_mrep = NULL;
1825 		tsiz -= retlen;
1826 		if (!(nd->nd_flag & ND_NFSV2)) {
1827 			if (eof || retlen == 0)
1828 				tsiz = 0;
1829 		} else if (retlen < len)
1830 			tsiz = 0;
1831 	}
1832 	return (0);
1833 nfsmout:
1834 	if (nd->nd_mrep != NULL)
1835 		m_freem(nd->nd_mrep);
1836 	return (error);
1837 }
1838 
1839 /*
1840  * nfs write operation
1841  * When called_from_strategy != 0, it should return EIO for an error that
1842  * indicates recovery is in progress, so that the buffer will be left
1843  * dirty and be written back to the server later. If it loops around,
1844  * the recovery thread could get stuck waiting for the buffer and recovery
1845  * will then deadlock.
1846  */
1847 int
1848 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1849     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1850     int called_from_strategy, int ioflag)
1851 {
1852 	int error, expireret = 0, retrycnt, nostateid;
1853 	u_int32_t clidrev = 0;
1854 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1855 	struct nfsnode *np = VTONFS(vp);
1856 	struct ucred *newcred;
1857 	struct nfsfh *nfhp = NULL;
1858 	nfsv4stateid_t stateid;
1859 	void *lckp;
1860 
1861 	KASSERT(*must_commit >= 0 && *must_commit <= 2,
1862 	    ("nfsrpc_write: must_commit out of range=%d", *must_commit));
1863 	if (nmp->nm_clp != NULL)
1864 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1865 	newcred = cred;
1866 	if (NFSHASNFSV4(nmp)) {
1867 		newcred = NFSNEWCRED(cred);
1868 		nfhp = np->n_fhp;
1869 	}
1870 	retrycnt = 0;
1871 	do {
1872 		lckp = NULL;
1873 		nostateid = 0;
1874 		if (NFSHASNFSV4(nmp)) {
1875 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1876 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1877 			    &lckp);
1878 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1879 			    stateid.other[2] == 0) {
1880 				nostateid = 1;
1881 				NFSCL_DEBUG(1, "stateid0 in write\n");
1882 			}
1883 		}
1884 
1885 		/*
1886 		 * If there is no stateid for NFSv4, it means this is an
1887 		 * extraneous write after close. Basically a poorly
1888 		 * implemented buffer cache. Just don't do the write.
1889 		 */
1890 		if (nostateid)
1891 			error = 0;
1892 		else
1893 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1894 			    newcred, &stateid, p, nap, attrflagp, ioflag);
1895 		if (error == NFSERR_STALESTATEID)
1896 			nfscl_initiate_recovery(nmp->nm_clp);
1897 		if (lckp != NULL)
1898 			nfscl_lockderef(lckp);
1899 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1900 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1901 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1902 			(void) nfs_catnap(PZERO, error, "nfs_write");
1903 		} else if ((error == NFSERR_EXPIRED ||
1904 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1905 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1906 		}
1907 		retrycnt++;
1908 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1909 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1910 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1911 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1912 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1913 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1914 	if (error != 0 && (retrycnt >= 4 ||
1915 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1916 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1917 		error = EIO;
1918 	if (NFSHASNFSV4(nmp))
1919 		NFSFREECRED(newcred);
1920 	return (error);
1921 }
1922 
1923 /*
1924  * The actual write RPC.
1925  */
1926 static int
1927 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1928     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1929     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, int ioflag)
1930 {
1931 	u_int32_t *tl;
1932 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1933 	struct nfsnode *np = VTONFS(vp);
1934 	int error = 0, len, rlen, commit, committed = NFSWRITE_FILESYNC;
1935 	int wccflag = 0;
1936 	int32_t backup;
1937 	struct nfsrv_descript *nd;
1938 	nfsattrbit_t attrbits;
1939 	uint64_t tmp_off;
1940 	ssize_t tsiz, wsize;
1941 	bool do_append;
1942 
1943 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1944 	*attrflagp = 0;
1945 	tsiz = uiop->uio_resid;
1946 	tmp_off = uiop->uio_offset + tsiz;
1947 	NFSLOCKMNT(nmp);
1948 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1949 		NFSUNLOCKMNT(nmp);
1950 		return (EFBIG);
1951 	}
1952 	wsize = nmp->nm_wsize;
1953 	do_append = false;
1954 	if ((ioflag & IO_APPEND) != 0 && NFSHASNFSV4(nmp) && !NFSHASPNFS(nmp))
1955 		do_append = true;
1956 	NFSUNLOCKMNT(nmp);
1957 	nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK);
1958 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
1959 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
1960 	while (tsiz > 0) {
1961 		*attrflagp = 0;
1962 		len = (tsiz > wsize) ? wsize : tsiz;
1963 		if (do_append)
1964 			NFSCL_REQSTART(nd, NFSPROC_APPENDWRITE, vp);
1965 		else
1966 			NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1967 		if (nd->nd_flag & ND_NFSV4) {
1968 			if (do_append) {
1969 				NFSZERO_ATTRBIT(&attrbits);
1970 				NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
1971 				nfsrv_putattrbit(nd, &attrbits);
1972 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED +
1973 				    NFSX_HYPER);
1974 				*tl++ = txdr_unsigned(NFSX_HYPER);
1975 				txdr_hyper(uiop->uio_offset, tl); tl += 2;
1976 				*tl = txdr_unsigned(NFSV4OP_WRITE);
1977 			}
1978 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1979 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1980 			txdr_hyper(uiop->uio_offset, tl);
1981 			tl += 2;
1982 			*tl++ = txdr_unsigned(*iomode);
1983 			*tl = txdr_unsigned(len);
1984 		} else if (nd->nd_flag & ND_NFSV3) {
1985 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1986 			txdr_hyper(uiop->uio_offset, tl);
1987 			tl += 2;
1988 			*tl++ = txdr_unsigned(len);
1989 			*tl++ = txdr_unsigned(*iomode);
1990 			*tl = txdr_unsigned(len);
1991 		} else {
1992 			u_int32_t x;
1993 
1994 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1995 			/*
1996 			 * Not sure why someone changed this, since the
1997 			 * RFC clearly states that "beginoffset" and
1998 			 * "totalcount" are ignored, but it wouldn't
1999 			 * surprise me if there's a busted server out there.
2000 			 */
2001 			/* Set both "begin" and "current" to non-garbage. */
2002 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
2003 			*tl++ = x;      /* "begin offset" */
2004 			*tl++ = x;      /* "current offset" */
2005 			x = txdr_unsigned(len);
2006 			*tl++ = x;      /* total to this offset */
2007 			*tl = x;        /* size of this write */
2008 		}
2009 		nfsm_uiombuf(nd, uiop, len);
2010 		/*
2011 		 * Although it is tempting to do a normal Getattr Op in the
2012 		 * NFSv4 compound, the result can be a nearly hung client
2013 		 * system if the Getattr asks for Owner and/or OwnerGroup.
2014 		 * It occurs when the client can't map either the Owner or
2015 		 * Owner_group name in the Getattr reply to a uid/gid. When
2016 		 * there is a cache miss, the kernel does an upcall to the
2017 		 * nfsuserd. Then, it can try and read the local /etc/passwd
2018 		 * or /etc/group file. It can then block in getnewbuf(),
2019 		 * waiting for dirty writes to be pushed to the NFS server.
2020 		 * The only reason this doesn't result in a complete
2021 		 * deadlock, is that the upcall times out and allows
2022 		 * the write to complete. However, progress is so slow
2023 		 * that it might just as well be deadlocked.
2024 		 * As such, we get the rest of the attributes, but not
2025 		 * Owner or Owner_group.
2026 		 * nb: nfscl_loadattrcache() needs to be told that these
2027 		 *     partial attributes from a write rpc are being
2028 		 *     passed in, via a argument flag.
2029 		 */
2030 		if (nd->nd_flag & ND_NFSV4) {
2031 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
2032 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2033 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2034 			(void) nfsrv_putattrbit(nd, &attrbits);
2035 		}
2036 		error = nfscl_request(nd, vp, p, cred);
2037 		if (error) {
2038 			free(nd, M_TEMP);
2039 			return (error);
2040 		}
2041 		if (nd->nd_repstat) {
2042 			/*
2043 			 * In case the rpc gets retried, roll
2044 			 * the uio fields changed by nfsm_uiombuf()
2045 			 * back.
2046 			 */
2047 			uiop->uio_offset -= len;
2048 			uiop->uio_resid += len;
2049 			uiop->uio_iov->iov_base =
2050 			    (char *)uiop->uio_iov->iov_base - len;
2051 			uiop->uio_iov->iov_len += len;
2052 		}
2053 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2054 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
2055 			    &wccflag, &tmp_off);
2056 			if (error)
2057 				goto nfsmout;
2058 		}
2059 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2060 		    (ND_NFSV4 | ND_NOMOREDATA) &&
2061 		    nd->nd_repstat == NFSERR_NOTSAME && do_append) {
2062 			/*
2063 			 * Verify of the file's size failed, so redo the
2064 			 * write using the file's size as returned in
2065 			 * the wcc attributes.
2066 			 */
2067 			if (tmp_off + tsiz <= nmp->nm_maxfilesize) {
2068 				do_append = false;
2069 				uiop->uio_offset = tmp_off;
2070 				m_freem(nd->nd_mrep);
2071 				nd->nd_mrep = NULL;
2072 				continue;
2073 			} else
2074 				nd->nd_repstat = EFBIG;
2075 		}
2076 		if (!nd->nd_repstat) {
2077 			if (do_append) {
2078 				/* Strip off the Write reply status. */
2079 				do_append = false;
2080 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
2081 			}
2082 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2083 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
2084 					+ NFSX_VERF);
2085 				rlen = fxdr_unsigned(int, *tl++);
2086 				if (rlen == 0) {
2087 					error = NFSERR_IO;
2088 					goto nfsmout;
2089 				} else if (rlen < len) {
2090 					backup = len - rlen;
2091 					uiop->uio_iov->iov_base =
2092 					    (char *)uiop->uio_iov->iov_base -
2093 					    backup;
2094 					uiop->uio_iov->iov_len += backup;
2095 					uiop->uio_offset -= backup;
2096 					uiop->uio_resid += backup;
2097 					len = rlen;
2098 				}
2099 				commit = fxdr_unsigned(int, *tl++);
2100 
2101 				/*
2102 				 * Return the lowest commitment level
2103 				 * obtained by any of the RPCs.
2104 				 */
2105 				if (committed == NFSWRITE_FILESYNC)
2106 					committed = commit;
2107 				else if (committed == NFSWRITE_DATASYNC &&
2108 					commit == NFSWRITE_UNSTABLE)
2109 					committed = commit;
2110 				NFSLOCKMNT(nmp);
2111 				if (!NFSHASWRITEVERF(nmp)) {
2112 					NFSBCOPY((caddr_t)tl,
2113 					    (caddr_t)&nmp->nm_verf[0],
2114 					    NFSX_VERF);
2115 					NFSSETWRITEVERF(nmp);
2116 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
2117 				    NFSX_VERF) && *must_commit != 2) {
2118 					*must_commit = 1;
2119 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
2120 				}
2121 				NFSUNLOCKMNT(nmp);
2122 			}
2123 			if (nd->nd_flag & ND_NFSV4)
2124 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2125 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
2126 				error = nfsm_loadattr(nd, nap);
2127 				if (!error)
2128 					*attrflagp = NFS_LATTR_NOSHRINK;
2129 			}
2130 		} else {
2131 			error = nd->nd_repstat;
2132 		}
2133 		if (error)
2134 			goto nfsmout;
2135 		NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
2136 		m_freem(nd->nd_mrep);
2137 		nd->nd_mrep = NULL;
2138 		tsiz -= len;
2139 	}
2140 nfsmout:
2141 	if (nd->nd_mrep != NULL)
2142 		m_freem(nd->nd_mrep);
2143 	*iomode = committed;
2144 	if (nd->nd_repstat && !error)
2145 		error = nd->nd_repstat;
2146 	free(nd, M_TEMP);
2147 	return (error);
2148 }
2149 
2150 /*
2151  * Do an nfs deallocate operation.
2152  */
2153 int
2154 nfsrpc_deallocate(vnode_t vp, off_t offs, off_t len, struct nfsvattr *nap,
2155     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
2156 {
2157 	int error, expireret = 0, openerr, retrycnt;
2158 	uint32_t clidrev = 0;
2159 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
2160 	struct nfsfh *nfhp;
2161 	nfsv4stateid_t stateid;
2162 	void *lckp;
2163 
2164 	if (nmp->nm_clp != NULL)
2165 		clidrev = nmp->nm_clp->nfsc_clientidrev;
2166 	retrycnt = 0;
2167 	do {
2168 		lckp = NULL;
2169 		openerr = 1;
2170 		nfhp = VTONFS(vp)->n_fhp;
2171 		error = nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
2172 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
2173 		if (error != 0) {
2174 			/*
2175 			 * No Open stateid, so try and open the file
2176 			 * now.
2177 			 */
2178 			openerr = nfsrpc_open(vp, FWRITE, cred, p);
2179 			if (openerr == 0)
2180 				nfscl_getstateid(vp, nfhp->nfh_fh,
2181 				    nfhp->nfh_len, NFSV4OPEN_ACCESSWRITE, 0,
2182 				    cred, p, &stateid, &lckp);
2183 		}
2184 		error = nfsrpc_deallocaterpc(vp, offs, len, &stateid, nap,
2185 		    attrflagp, cred, p);
2186 		if (error == NFSERR_STALESTATEID)
2187 			nfscl_initiate_recovery(nmp->nm_clp);
2188 		if (lckp != NULL)
2189 			nfscl_lockderef(lckp);
2190 		if (openerr == 0)
2191 			nfsrpc_close(vp, 0, p);
2192 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2193 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2194 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
2195 			(void) nfs_catnap(PZERO, error, "nfs_deallocate");
2196 		} else if ((error == NFSERR_EXPIRED ||
2197 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2198 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2199 		}
2200 		retrycnt++;
2201 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2202 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2203 	    error == NFSERR_BADSESSION ||
2204 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
2205 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2206 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
2207 	if (error && retrycnt >= 4)
2208 		error = EIO;
2209 	return (error);
2210 }
2211 
2212 /*
2213  * The actual deallocate RPC.
2214  */
2215 static int
2216 nfsrpc_deallocaterpc(vnode_t vp, off_t offs, off_t len,
2217     nfsv4stateid_t *stateidp, struct nfsvattr *nap, int *attrflagp,
2218     struct ucred *cred, NFSPROC_T *p)
2219 {
2220 	uint32_t *tl;
2221 	struct nfsnode *np = VTONFS(vp);
2222 	int error, wccflag;
2223 	struct nfsrv_descript nfsd;
2224 	struct nfsrv_descript *nd = &nfsd;
2225 	nfsattrbit_t attrbits;
2226 
2227 	*attrflagp = 0;
2228 	NFSCL_REQSTART(nd, NFSPROC_DEALLOCATE, vp);
2229 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
2230 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
2231 	txdr_hyper(offs, tl);
2232 	tl += 2;
2233 	txdr_hyper(len, tl);
2234 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
2235 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
2236 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2237 	nfsrv_putattrbit(nd, &attrbits);
2238 	error = nfscl_request(nd, vp, p, cred);
2239 	if (error != 0)
2240 		return (error);
2241 	wccflag = 0;
2242 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, &wccflag, NULL);
2243 	if (error != 0)
2244 		goto nfsmout;
2245 	if (nd->nd_repstat == 0) {
2246 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
2247 		error = nfsm_loadattr(nd, nap);
2248 		if (error != 0)
2249 			goto nfsmout;
2250 		*attrflagp = NFS_LATTR_NOSHRINK;
2251 	}
2252 	NFSWRITERPC_SETTIME(wccflag, np, nap, 1);
2253 nfsmout:
2254 	m_freem(nd->nd_mrep);
2255 	if (nd->nd_repstat != 0 && error == 0)
2256 		error = nd->nd_repstat;
2257 	return (error);
2258 }
2259 
2260 /*
2261  * nfs mknod rpc
2262  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
2263  * mode set to specify the file type and the size field for rdev.
2264  */
2265 int
2266 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2267     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
2268     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2269     int *attrflagp, int *dattrflagp)
2270 {
2271 	u_int32_t *tl;
2272 	int error = 0;
2273 	struct nfsrv_descript nfsd, *nd = &nfsd;
2274 	nfsattrbit_t attrbits;
2275 
2276 	*nfhpp = NULL;
2277 	*attrflagp = 0;
2278 	*dattrflagp = 0;
2279 	if (namelen > NFS_MAXNAMLEN)
2280 		return (ENAMETOOLONG);
2281 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
2282 	if (nd->nd_flag & ND_NFSV4) {
2283 		if (vtyp == VBLK || vtyp == VCHR) {
2284 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
2285 			*tl++ = vtonfsv34_type(vtyp);
2286 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2287 			*tl = txdr_unsigned(NFSMINOR(rdev));
2288 		} else {
2289 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2290 			*tl = vtonfsv34_type(vtyp);
2291 		}
2292 	}
2293 	(void) nfsm_strtom(nd, name, namelen);
2294 	if (nd->nd_flag & ND_NFSV3) {
2295 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2296 		*tl = vtonfsv34_type(vtyp);
2297 	}
2298 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2299 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2300 	if ((nd->nd_flag & ND_NFSV3) &&
2301 	    (vtyp == VCHR || vtyp == VBLK)) {
2302 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2303 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2304 		*tl = txdr_unsigned(NFSMINOR(rdev));
2305 	}
2306 	if (nd->nd_flag & ND_NFSV4) {
2307 		NFSGETATTR_ATTRBIT(&attrbits);
2308 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2309 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2310 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2311 		(void) nfsrv_putattrbit(nd, &attrbits);
2312 	}
2313 	if (nd->nd_flag & ND_NFSV2)
2314 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
2315 	error = nfscl_request(nd, dvp, p, cred);
2316 	if (error)
2317 		return (error);
2318 	if (nd->nd_flag & ND_NFSV4)
2319 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2320 	if (!nd->nd_repstat) {
2321 		if (nd->nd_flag & ND_NFSV4) {
2322 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2323 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2324 			if (error)
2325 				goto nfsmout;
2326 		}
2327 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2328 		if (error)
2329 			goto nfsmout;
2330 	}
2331 	if (nd->nd_flag & ND_NFSV3)
2332 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2333 	if (!error && nd->nd_repstat)
2334 		error = nd->nd_repstat;
2335 nfsmout:
2336 	m_freem(nd->nd_mrep);
2337 	return (error);
2338 }
2339 
2340 /*
2341  * nfs file create call
2342  * Mostly just call the approriate routine. (I separated out v4, so that
2343  * error recovery wouldn't be as difficult.)
2344  */
2345 int
2346 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2347     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2348     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2349     int *attrflagp, int *dattrflagp)
2350 {
2351 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
2352 	struct nfsclowner *owp;
2353 	struct nfscldeleg *dp;
2354 	struct nfsmount *nmp = VFSTONFS(dvp->v_mount);
2355 	u_int32_t clidrev;
2356 
2357 	if (NFSHASNFSV4(nmp)) {
2358 	    retrycnt = 0;
2359 	    do {
2360 		dp = NULL;
2361 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
2362 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
2363 		    NULL, 1, true);
2364 		if (error)
2365 			return (error);
2366 		if (nmp->nm_clp != NULL)
2367 			clidrev = nmp->nm_clp->nfsc_clientidrev;
2368 		else
2369 			clidrev = 0;
2370 		if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
2371 		    nfs_numnfscbd == 0 || retrycnt > 0)
2372 			error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
2373 			  fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2374 			  attrflagp, dattrflagp, &unlocked);
2375 		else
2376 			error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
2377 			  cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2378 			  attrflagp, dattrflagp, &unlocked);
2379 		/*
2380 		 * There is no need to invalidate cached attributes here,
2381 		 * since new post-delegation issue attributes are always
2382 		 * returned by nfsrpc_createv4() and these will update the
2383 		 * attribute cache.
2384 		 */
2385 		if (dp != NULL)
2386 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2387 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2388 		nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2389 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2390 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2391 		    error == NFSERR_BADSESSION) {
2392 			(void) nfs_catnap(PZERO, error, "nfs_open");
2393 		} else if ((error == NFSERR_EXPIRED ||
2394 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2395 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2396 			retrycnt++;
2397 		}
2398 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2399 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2400 		error == NFSERR_BADSESSION ||
2401 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2402 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
2403 	    if (error && retrycnt >= 4)
2404 		    error = EIO;
2405 	} else {
2406 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2407 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp);
2408 	}
2409 	return (error);
2410 }
2411 
2412 /*
2413  * The create rpc for v2 and 3.
2414  */
2415 static int
2416 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2417     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2418     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2419     int *attrflagp, int *dattrflagp)
2420 {
2421 	u_int32_t *tl;
2422 	int error = 0;
2423 	struct nfsrv_descript nfsd, *nd = &nfsd;
2424 
2425 	*nfhpp = NULL;
2426 	*attrflagp = 0;
2427 	*dattrflagp = 0;
2428 	if (namelen > NFS_MAXNAMLEN)
2429 		return (ENAMETOOLONG);
2430 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2431 	(void) nfsm_strtom(nd, name, namelen);
2432 	if (nd->nd_flag & ND_NFSV3) {
2433 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2434 		if (fmode & O_EXCL) {
2435 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2436 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2437 			*tl++ = cverf.lval[0];
2438 			*tl = cverf.lval[1];
2439 		} else {
2440 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2441 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
2442 		}
2443 	} else {
2444 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2445 	}
2446 	error = nfscl_request(nd, dvp, p, cred);
2447 	if (error)
2448 		return (error);
2449 	if (nd->nd_repstat == 0) {
2450 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2451 		if (error)
2452 			goto nfsmout;
2453 	}
2454 	if (nd->nd_flag & ND_NFSV3)
2455 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2456 	if (nd->nd_repstat != 0 && error == 0)
2457 		error = nd->nd_repstat;
2458 nfsmout:
2459 	m_freem(nd->nd_mrep);
2460 	return (error);
2461 }
2462 
2463 static int
2464 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2465     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2466     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2467     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2468     int *dattrflagp, int *unlockedp)
2469 {
2470 	u_int32_t *tl;
2471 	int error = 0, deleg, newone, ret, acesize, limitby;
2472 	struct nfsrv_descript nfsd, *nd = &nfsd;
2473 	struct nfsclopen *op;
2474 	struct nfscldeleg *dp = NULL;
2475 	struct nfsnode *np;
2476 	struct nfsfh *nfhp;
2477 	nfsattrbit_t attrbits;
2478 	nfsv4stateid_t stateid;
2479 	u_int32_t rflags;
2480 	struct nfsmount *nmp;
2481 	struct nfsclsession *tsep;
2482 
2483 	nmp = VFSTONFS(dvp->v_mount);
2484 	np = VTONFS(dvp);
2485 	*unlockedp = 0;
2486 	*nfhpp = NULL;
2487 	*dpp = NULL;
2488 	*attrflagp = 0;
2489 	*dattrflagp = 0;
2490 	if (namelen > NFS_MAXNAMLEN)
2491 		return (ENAMETOOLONG);
2492 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2493 	/*
2494 	 * For V4, this is actually an Open op.
2495 	 */
2496 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2497 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
2498 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2499 	    NFSV4OPEN_ACCESSREAD);
2500 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2501 	tsep = nfsmnt_mdssession(nmp);
2502 	*tl++ = tsep->nfsess_clientid.lval[0];
2503 	*tl = tsep->nfsess_clientid.lval[1];
2504 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2505 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2506 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2507 	if (fmode & O_EXCL) {
2508 		if (NFSHASNFSV4N(nmp)) {
2509 			if (NFSHASSESSPERSIST(nmp)) {
2510 				/* Use GUARDED for persistent sessions. */
2511 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
2512 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2513 			} else {
2514 				/* Otherwise, use EXCLUSIVE4_1. */
2515 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2516 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2517 				*tl++ = cverf.lval[0];
2518 				*tl = cverf.lval[1];
2519 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2520 			}
2521 		} else {
2522 			/* NFSv4.0 */
2523 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2524 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2525 			*tl++ = cverf.lval[0];
2526 			*tl = cverf.lval[1];
2527 		}
2528 	} else {
2529 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2530 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2531 	}
2532 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2533 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2534 	(void) nfsm_strtom(nd, name, namelen);
2535 	/* Get the new file's handle and attributes. */
2536 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2537 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2538 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2539 	NFSGETATTR_ATTRBIT(&attrbits);
2540 	(void) nfsrv_putattrbit(nd, &attrbits);
2541 	/* Get the directory's post-op attributes. */
2542 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2543 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2544 	(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2545 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2546 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2547 	(void) nfsrv_putattrbit(nd, &attrbits);
2548 	error = nfscl_request(nd, dvp, p, cred);
2549 	if (error)
2550 		return (error);
2551 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2552 	if (nd->nd_repstat == 0) {
2553 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2554 		    6 * NFSX_UNSIGNED);
2555 		stateid.seqid = *tl++;
2556 		stateid.other[0] = *tl++;
2557 		stateid.other[1] = *tl++;
2558 		stateid.other[2] = *tl;
2559 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2560 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2561 		if (error)
2562 			goto nfsmout;
2563 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2564 		deleg = fxdr_unsigned(int, *tl);
2565 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2566 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2567 			if (!(owp->nfsow_clp->nfsc_flags &
2568 			      NFSCLFLAGS_FIRSTDELEG))
2569 				owp->nfsow_clp->nfsc_flags |=
2570 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2571 			dp = malloc(
2572 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2573 			    M_NFSCLDELEG, M_WAITOK);
2574 			LIST_INIT(&dp->nfsdl_owner);
2575 			LIST_INIT(&dp->nfsdl_lock);
2576 			dp->nfsdl_clp = owp->nfsow_clp;
2577 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2578 			nfscl_lockinit(&dp->nfsdl_rwlock);
2579 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2580 			    NFSX_UNSIGNED);
2581 			dp->nfsdl_stateid.seqid = *tl++;
2582 			dp->nfsdl_stateid.other[0] = *tl++;
2583 			dp->nfsdl_stateid.other[1] = *tl++;
2584 			dp->nfsdl_stateid.other[2] = *tl++;
2585 			ret = fxdr_unsigned(int, *tl);
2586 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2587 				dp->nfsdl_flags = NFSCLDL_WRITE;
2588 				/*
2589 				 * Indicates how much the file can grow.
2590 				 */
2591 				NFSM_DISSECT(tl, u_int32_t *,
2592 				    3 * NFSX_UNSIGNED);
2593 				limitby = fxdr_unsigned(int, *tl++);
2594 				switch (limitby) {
2595 				case NFSV4OPEN_LIMITSIZE:
2596 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2597 					break;
2598 				case NFSV4OPEN_LIMITBLOCKS:
2599 					dp->nfsdl_sizelimit =
2600 					    fxdr_unsigned(u_int64_t, *tl++);
2601 					dp->nfsdl_sizelimit *=
2602 					    fxdr_unsigned(u_int64_t, *tl);
2603 					break;
2604 				default:
2605 					error = NFSERR_BADXDR;
2606 					goto nfsmout;
2607 				}
2608 			} else {
2609 				dp->nfsdl_flags = NFSCLDL_READ;
2610 			}
2611 			if (ret)
2612 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2613 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, false,
2614 			    &ret, &acesize, p);
2615 			if (error)
2616 				goto nfsmout;
2617 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2618 			error = NFSERR_BADXDR;
2619 			goto nfsmout;
2620 		}
2621 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2622 		if (error)
2623 			goto nfsmout;
2624 		/* Get rid of the PutFH and Getattr status values. */
2625 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2626 		/* Load the directory attributes. */
2627 		error = nfsm_loadattr(nd, dnap);
2628 		if (error)
2629 			goto nfsmout;
2630 		*dattrflagp = 1;
2631 		if (dp != NULL && *attrflagp) {
2632 			dp->nfsdl_change = nnap->na_filerev;
2633 			dp->nfsdl_modtime = nnap->na_mtime;
2634 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2635 		}
2636 		/*
2637 		 * We can now complete the Open state.
2638 		 */
2639 		nfhp = *nfhpp;
2640 		if (dp != NULL) {
2641 			dp->nfsdl_fhlen = nfhp->nfh_len;
2642 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2643 		}
2644 		/*
2645 		 * Get an Open structure that will be
2646 		 * attached to the OpenOwner, acquired already.
2647 		 */
2648 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2649 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2650 		    cred, p, NULL, &op, &newone, NULL, 0, false);
2651 		if (error)
2652 			goto nfsmout;
2653 		op->nfso_stateid = stateid;
2654 		newnfs_copyincred(cred, &op->nfso_cred);
2655 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2656 		    do {
2657 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2658 			    nfhp->nfh_len, op, cred, p);
2659 			if (ret == NFSERR_DELAY)
2660 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2661 		    } while (ret == NFSERR_DELAY);
2662 		    error = ret;
2663 		}
2664 
2665 		/*
2666 		 * If the server is handing out delegations, but we didn't
2667 		 * get one because an OpenConfirm was required, try the
2668 		 * Open again, to get a delegation. This is a harmless no-op,
2669 		 * from a server's point of view.
2670 		 */
2671 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2672 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2673 		    !error && dp == NULL) {
2674 		    do {
2675 			ret = nfsrpc_openrpc(VFSTONFS(dvp->v_mount), dvp,
2676 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2677 			    nfhp->nfh_fh, nfhp->nfh_len,
2678 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2679 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2680 			if (ret == NFSERR_DELAY)
2681 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2682 		    } while (ret == NFSERR_DELAY);
2683 		    if (ret) {
2684 			if (dp != NULL) {
2685 				free(dp, M_NFSCLDELEG);
2686 				dp = NULL;
2687 			}
2688 			if (ret == NFSERR_STALECLIENTID ||
2689 			    ret == NFSERR_STALEDONTRECOVER ||
2690 			    ret == NFSERR_BADSESSION)
2691 				error = ret;
2692 		    }
2693 		}
2694 		nfscl_openrelease(nmp, op, error, newone);
2695 		*unlockedp = 1;
2696 	}
2697 	if (nd->nd_repstat != 0 && error == 0)
2698 		error = nd->nd_repstat;
2699 	if (error == NFSERR_STALECLIENTID)
2700 		nfscl_initiate_recovery(owp->nfsow_clp);
2701 nfsmout:
2702 	if (!error)
2703 		*dpp = dp;
2704 	else if (dp != NULL)
2705 		free(dp, M_NFSCLDELEG);
2706 	m_freem(nd->nd_mrep);
2707 	return (error);
2708 }
2709 
2710 /*
2711  * Nfs remove rpc
2712  */
2713 int
2714 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2715     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp)
2716 {
2717 	u_int32_t *tl;
2718 	struct nfsrv_descript nfsd, *nd = &nfsd;
2719 	struct nfsnode *np;
2720 	struct nfsmount *nmp;
2721 	nfsv4stateid_t dstateid;
2722 	int error, ret = 0, i;
2723 
2724 	*dattrflagp = 0;
2725 	if (namelen > NFS_MAXNAMLEN)
2726 		return (ENAMETOOLONG);
2727 	nmp = VFSTONFS(dvp->v_mount);
2728 tryagain:
2729 	if (NFSHASNFSV4(nmp) && ret == 0) {
2730 		ret = nfscl_removedeleg(vp, p, &dstateid);
2731 		if (ret == 1) {
2732 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2733 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2734 			    NFSX_UNSIGNED);
2735 			if (NFSHASNFSV4N(nmp))
2736 				*tl++ = 0;
2737 			else
2738 				*tl++ = dstateid.seqid;
2739 			*tl++ = dstateid.other[0];
2740 			*tl++ = dstateid.other[1];
2741 			*tl++ = dstateid.other[2];
2742 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2743 			np = VTONFS(dvp);
2744 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2745 			    np->n_fhp->nfh_len, 0);
2746 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2747 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2748 		}
2749 	} else {
2750 		ret = 0;
2751 	}
2752 	if (ret == 0)
2753 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2754 	(void) nfsm_strtom(nd, name, namelen);
2755 	error = nfscl_request(nd, dvp, p, cred);
2756 	if (error)
2757 		return (error);
2758 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2759 		/* For NFSv4, parse out any Delereturn replies. */
2760 		if (ret > 0 && nd->nd_repstat != 0 &&
2761 		    (nd->nd_flag & ND_NOMOREDATA)) {
2762 			/*
2763 			 * If the Delegreturn failed, try again without
2764 			 * it. The server will Recall, as required.
2765 			 */
2766 			m_freem(nd->nd_mrep);
2767 			goto tryagain;
2768 		}
2769 		for (i = 0; i < (ret * 2); i++) {
2770 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2771 			    ND_NFSV4) {
2772 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2773 			    if (*(tl + 1))
2774 				nd->nd_flag |= ND_NOMOREDATA;
2775 			}
2776 		}
2777 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2778 	}
2779 	if (nd->nd_repstat && !error)
2780 		error = nd->nd_repstat;
2781 nfsmout:
2782 	m_freem(nd->nd_mrep);
2783 	return (error);
2784 }
2785 
2786 /*
2787  * Do an nfs rename rpc.
2788  */
2789 int
2790 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2791     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2792     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2793     int *fattrflagp, int *tattrflagp)
2794 {
2795 	u_int32_t *tl;
2796 	struct nfsrv_descript nfsd, *nd = &nfsd;
2797 	struct nfsmount *nmp;
2798 	struct nfsnode *np;
2799 	nfsattrbit_t attrbits;
2800 	nfsv4stateid_t fdstateid, tdstateid;
2801 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2802 
2803 	*fattrflagp = 0;
2804 	*tattrflagp = 0;
2805 	nmp = VFSTONFS(fdvp->v_mount);
2806 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2807 		return (ENAMETOOLONG);
2808 tryagain:
2809 	if (NFSHASNFSV4(nmp) && ret == 0) {
2810 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2811 		    &tdstateid, &gottd, p);
2812 		if (gotfd && gottd) {
2813 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2814 		} else if (gotfd) {
2815 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2816 		} else if (gottd) {
2817 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2818 		}
2819 		if (gotfd) {
2820 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2821 			if (NFSHASNFSV4N(nmp))
2822 				*tl++ = 0;
2823 			else
2824 				*tl++ = fdstateid.seqid;
2825 			*tl++ = fdstateid.other[0];
2826 			*tl++ = fdstateid.other[1];
2827 			*tl = fdstateid.other[2];
2828 			if (gottd) {
2829 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2830 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2831 				np = VTONFS(tvp);
2832 				(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2833 				    np->n_fhp->nfh_len, 0);
2834 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2835 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2836 			}
2837 		}
2838 		if (gottd) {
2839 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2840 			if (NFSHASNFSV4N(nmp))
2841 				*tl++ = 0;
2842 			else
2843 				*tl++ = tdstateid.seqid;
2844 			*tl++ = tdstateid.other[0];
2845 			*tl++ = tdstateid.other[1];
2846 			*tl = tdstateid.other[2];
2847 		}
2848 		if (ret > 0) {
2849 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2850 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2851 			np = VTONFS(fdvp);
2852 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2853 			    np->n_fhp->nfh_len, 0);
2854 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2855 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2856 		}
2857 	} else {
2858 		ret = 0;
2859 	}
2860 	if (ret == 0)
2861 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2862 	if (nd->nd_flag & ND_NFSV4) {
2863 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2864 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2865 		NFSWCCATTR_ATTRBIT(&attrbits);
2866 		(void) nfsrv_putattrbit(nd, &attrbits);
2867 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2868 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2869 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2870 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2871 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2872 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2873 		(void) nfsrv_putattrbit(nd, &attrbits);
2874 		nd->nd_flag |= ND_V4WCCATTR;
2875 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2876 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2877 	}
2878 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2879 	if (!(nd->nd_flag & ND_NFSV4))
2880 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2881 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2882 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2883 	error = nfscl_request(nd, fdvp, p, cred);
2884 	if (error)
2885 		return (error);
2886 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2887 		/* For NFSv4, parse out any Delereturn replies. */
2888 		if (ret > 0 && nd->nd_repstat != 0 &&
2889 		    (nd->nd_flag & ND_NOMOREDATA)) {
2890 			/*
2891 			 * If the Delegreturn failed, try again without
2892 			 * it. The server will Recall, as required.
2893 			 */
2894 			m_freem(nd->nd_mrep);
2895 			goto tryagain;
2896 		}
2897 		for (i = 0; i < (ret * 2); i++) {
2898 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2899 			    ND_NFSV4) {
2900 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2901 			    if (*(tl + 1)) {
2902 				if (i == 0 && ret > 1) {
2903 				    /*
2904 				     * If the Delegreturn failed, try again
2905 				     * without it. The server will Recall, as
2906 				     * required.
2907 				     * If ret > 1, the first iteration of this
2908 				     * loop is the second DelegReturn result.
2909 				     */
2910 				    m_freem(nd->nd_mrep);
2911 				    goto tryagain;
2912 				} else {
2913 				    nd->nd_flag |= ND_NOMOREDATA;
2914 				}
2915 			    }
2916 			}
2917 		}
2918 		/* Now, the first wcc attribute reply. */
2919 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2920 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2921 			if (*(tl + 1))
2922 				nd->nd_flag |= ND_NOMOREDATA;
2923 		}
2924 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL, NULL);
2925 		/* and the second wcc attribute reply. */
2926 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2927 		    !error) {
2928 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2929 			if (*(tl + 1))
2930 				nd->nd_flag |= ND_NOMOREDATA;
2931 		}
2932 		if (!error)
2933 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2934 			    NULL, NULL);
2935 	}
2936 	if (nd->nd_repstat && !error)
2937 		error = nd->nd_repstat;
2938 nfsmout:
2939 	m_freem(nd->nd_mrep);
2940 	return (error);
2941 }
2942 
2943 /*
2944  * nfs hard link create rpc
2945  */
2946 int
2947 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2948     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2949     struct nfsvattr *nap, int *attrflagp, int *dattrflagp)
2950 {
2951 	u_int32_t *tl;
2952 	struct nfsrv_descript nfsd, *nd = &nfsd;
2953 	nfsattrbit_t attrbits;
2954 	int error = 0;
2955 
2956 	*attrflagp = 0;
2957 	*dattrflagp = 0;
2958 	if (namelen > NFS_MAXNAMLEN)
2959 		return (ENAMETOOLONG);
2960 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2961 	if (nd->nd_flag & ND_NFSV4) {
2962 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2963 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2964 	}
2965 	(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2966 		VTONFS(dvp)->n_fhp->nfh_len, 0);
2967 	if (nd->nd_flag & ND_NFSV4) {
2968 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2969 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2970 		NFSWCCATTR_ATTRBIT(&attrbits);
2971 		(void) nfsrv_putattrbit(nd, &attrbits);
2972 		nd->nd_flag |= ND_V4WCCATTR;
2973 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2974 		*tl = txdr_unsigned(NFSV4OP_LINK);
2975 	}
2976 	(void) nfsm_strtom(nd, name, namelen);
2977 	error = nfscl_request(nd, vp, p, cred);
2978 	if (error)
2979 		return (error);
2980 	if (nd->nd_flag & ND_NFSV3) {
2981 		error = nfscl_postop_attr(nd, nap, attrflagp);
2982 		if (!error)
2983 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2984 			    NULL, NULL);
2985 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2986 		/*
2987 		 * First, parse out the PutFH and Getattr result.
2988 		 */
2989 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2990 		if (!(*(tl + 1)))
2991 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2992 		if (*(tl + 1))
2993 			nd->nd_flag |= ND_NOMOREDATA;
2994 		/*
2995 		 * Get the pre-op attributes.
2996 		 */
2997 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2998 	}
2999 	if (nd->nd_repstat && !error)
3000 		error = nd->nd_repstat;
3001 nfsmout:
3002 	m_freem(nd->nd_mrep);
3003 	return (error);
3004 }
3005 
3006 /*
3007  * nfs symbolic link create rpc
3008  */
3009 int
3010 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
3011     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
3012     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
3013     int *dattrflagp)
3014 {
3015 	u_int32_t *tl;
3016 	struct nfsrv_descript nfsd, *nd = &nfsd;
3017 	struct nfsmount *nmp;
3018 	int slen, error = 0;
3019 
3020 	*nfhpp = NULL;
3021 	*attrflagp = 0;
3022 	*dattrflagp = 0;
3023 	nmp = VFSTONFS(dvp->v_mount);
3024 	slen = strlen(target);
3025 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
3026 		return (ENAMETOOLONG);
3027 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
3028 	if (nd->nd_flag & ND_NFSV4) {
3029 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3030 		*tl = txdr_unsigned(NFLNK);
3031 		(void) nfsm_strtom(nd, target, slen);
3032 	}
3033 	(void) nfsm_strtom(nd, name, namelen);
3034 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
3035 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
3036 	if (!(nd->nd_flag & ND_NFSV4))
3037 		(void) nfsm_strtom(nd, target, slen);
3038 	if (nd->nd_flag & ND_NFSV2)
3039 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
3040 	error = nfscl_request(nd, dvp, p, cred);
3041 	if (error)
3042 		return (error);
3043 	if (nd->nd_flag & ND_NFSV4)
3044 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3045 	if ((nd->nd_flag & ND_NFSV3) && !error) {
3046 		if (!nd->nd_repstat)
3047 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
3048 		if (!error)
3049 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
3050 			    NULL, NULL);
3051 	}
3052 	if (nd->nd_repstat && !error)
3053 		error = nd->nd_repstat;
3054 	m_freem(nd->nd_mrep);
3055 	/*
3056 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
3057 	 * Only do this if vfs.nfs.ignore_eexist is set.
3058 	 * Never do this for NFSv4.1 or later minor versions, since sessions
3059 	 * should guarantee "exactly once" RPC semantics.
3060 	 */
3061 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
3062 	    nmp->nm_minorvers == 0))
3063 		error = 0;
3064 	return (error);
3065 }
3066 
3067 /*
3068  * nfs make dir rpc
3069  */
3070 int
3071 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
3072     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
3073     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
3074     int *dattrflagp)
3075 {
3076 	u_int32_t *tl;
3077 	struct nfsrv_descript nfsd, *nd = &nfsd;
3078 	nfsattrbit_t attrbits;
3079 	int error = 0;
3080 	struct nfsfh *fhp;
3081 	struct nfsmount *nmp;
3082 
3083 	*nfhpp = NULL;
3084 	*attrflagp = 0;
3085 	*dattrflagp = 0;
3086 	nmp = VFSTONFS(dvp->v_mount);
3087 	fhp = VTONFS(dvp)->n_fhp;
3088 	if (namelen > NFS_MAXNAMLEN)
3089 		return (ENAMETOOLONG);
3090 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
3091 	if (nd->nd_flag & ND_NFSV4) {
3092 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3093 		*tl = txdr_unsigned(NFDIR);
3094 	}
3095 	(void) nfsm_strtom(nd, name, namelen);
3096 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
3097 	if (nd->nd_flag & ND_NFSV4) {
3098 		NFSGETATTR_ATTRBIT(&attrbits);
3099 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3100 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3101 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3102 		(void) nfsrv_putattrbit(nd, &attrbits);
3103 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3104 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
3105 		(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
3106 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3107 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3108 		(void) nfsrv_putattrbit(nd, &attrbits);
3109 	}
3110 	error = nfscl_request(nd, dvp, p, cred);
3111 	if (error)
3112 		return (error);
3113 	if (nd->nd_flag & ND_NFSV4)
3114 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3115 	if (!nd->nd_repstat && !error) {
3116 		if (nd->nd_flag & ND_NFSV4) {
3117 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3118 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
3119 		}
3120 		if (!error)
3121 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
3122 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
3123 			/* Get rid of the PutFH and Getattr status values. */
3124 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
3125 			/* Load the directory attributes. */
3126 			error = nfsm_loadattr(nd, dnap);
3127 			if (error == 0)
3128 				*dattrflagp = 1;
3129 		}
3130 	}
3131 	if ((nd->nd_flag & ND_NFSV3) && !error)
3132 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3133 	if (nd->nd_repstat && !error)
3134 		error = nd->nd_repstat;
3135 nfsmout:
3136 	m_freem(nd->nd_mrep);
3137 	/*
3138 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
3139 	 * Only do this if vfs.nfs.ignore_eexist is set.
3140 	 * Never do this for NFSv4.1 or later minor versions, since sessions
3141 	 * should guarantee "exactly once" RPC semantics.
3142 	 */
3143 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
3144 	    nmp->nm_minorvers == 0))
3145 		error = 0;
3146 	return (error);
3147 }
3148 
3149 /*
3150  * nfs remove directory call
3151  */
3152 int
3153 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
3154     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp)
3155 {
3156 	struct nfsrv_descript nfsd, *nd = &nfsd;
3157 	int error = 0;
3158 
3159 	*dattrflagp = 0;
3160 	if (namelen > NFS_MAXNAMLEN)
3161 		return (ENAMETOOLONG);
3162 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
3163 	(void) nfsm_strtom(nd, name, namelen);
3164 	error = nfscl_request(nd, dvp, p, cred);
3165 	if (error)
3166 		return (error);
3167 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
3168 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3169 	if (nd->nd_repstat && !error)
3170 		error = nd->nd_repstat;
3171 	m_freem(nd->nd_mrep);
3172 	/*
3173 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
3174 	 */
3175 	if (error == ENOENT)
3176 		error = 0;
3177 	return (error);
3178 }
3179 
3180 /*
3181  * Readdir rpc.
3182  * Always returns with either uio_resid unchanged, if you are at the
3183  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
3184  * filled in.
3185  * I felt this would allow caching of directory blocks more easily
3186  * than returning a pertially filled block.
3187  * Directory offset cookies:
3188  * Oh my, what to do with them...
3189  * I can think of three ways to deal with them:
3190  * 1 - have the layer above these RPCs maintain a map between logical
3191  *     directory byte offsets and the NFS directory offset cookies
3192  * 2 - pass the opaque directory offset cookies up into userland
3193  *     and let the libc functions deal with them, via the system call
3194  * 3 - return them to userland in the "struct dirent", so future versions
3195  *     of libc can use them and do whatever is necessary to make things work
3196  *     above these rpc calls, in the meantime
3197  * For now, I do #3 by "hiding" the directory offset cookies after the
3198  * d_name field in struct dirent. This is space inside d_reclen that
3199  * will be ignored by anything that doesn't know about them.
3200  * The directory offset cookies are filled in as the last 8 bytes of
3201  * each directory entry, after d_name. Someday, the userland libc
3202  * functions may be able to use these. In the meantime, it satisfies
3203  * OpenBSD's requirements for cookies being returned.
3204  * If expects the directory offset cookie for the read to be in uio_offset
3205  * and returns the one for the next entry after this directory block in
3206  * there, as well.
3207  */
3208 int
3209 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3210     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3211     int *eofp)
3212 {
3213 	int len, left;
3214 	struct dirent *dp = NULL;
3215 	u_int32_t *tl;
3216 	nfsquad_t cookie, ncookie;
3217 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3218 	struct nfsnode *dnp = VTONFS(vp);
3219 	struct nfsvattr nfsva;
3220 	struct nfsrv_descript nfsd, *nd = &nfsd;
3221 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3222 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
3223 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3224 	char *cp;
3225 	nfsattrbit_t attrbits, dattrbits;
3226 	u_int32_t rderr, *tl2 = NULL;
3227 	size_t tresid;
3228 
3229 	KASSERT(uiop->uio_iovcnt == 1 &&
3230 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3231 	    ("nfs readdirrpc bad uio"));
3232 	ncookie.lval[0] = ncookie.lval[1] = 0;
3233 	/*
3234 	 * There is no point in reading a lot more than uio_resid, however
3235 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
3236 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
3237 	 * will never make readsize > nm_readdirsize.
3238 	 */
3239 	readsize = nmp->nm_readdirsize;
3240 	if (readsize > uiop->uio_resid)
3241 		readsize = uiop->uio_resid + DIRBLKSIZ;
3242 
3243 	*attrflagp = 0;
3244 	if (eofp)
3245 		*eofp = 0;
3246 	tresid = uiop->uio_resid;
3247 	cookie.lval[0] = cookiep->nfsuquad[0];
3248 	cookie.lval[1] = cookiep->nfsuquad[1];
3249 	nd->nd_mrep = NULL;
3250 
3251 	/*
3252 	 * For NFSv4, first create the "." and ".." entries.
3253 	 */
3254 	if (NFSHASNFSV4(nmp)) {
3255 		reqsize = 6 * NFSX_UNSIGNED;
3256 		NFSGETATTR_ATTRBIT(&dattrbits);
3257 		NFSZERO_ATTRBIT(&attrbits);
3258 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3259 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
3260 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3261 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3262 			NFSSETBIT_ATTRBIT(&attrbits,
3263 			    NFSATTRBIT_MOUNTEDONFILEID);
3264 			gotmnton = 1;
3265 		} else {
3266 			/*
3267 			 * Must fake it. Use the fileno, except when the
3268 			 * fsid is != to that of the directory. For that
3269 			 * case, generate a fake fileno that is not the same.
3270 			 */
3271 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3272 			gotmnton = 0;
3273 		}
3274 
3275 		/*
3276 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3277 		 */
3278 		if (uiop->uio_offset == 0) {
3279 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3280 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3281 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3282 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3283 			(void) nfsrv_putattrbit(nd, &attrbits);
3284 			error = nfscl_request(nd, vp, p, cred);
3285 			if (error)
3286 			    return (error);
3287 			dotfileid = 0;	/* Fake out the compiler. */
3288 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3289 			    error = nfsm_loadattr(nd, &nfsva);
3290 			    if (error != 0)
3291 				goto nfsmout;
3292 			    dotfileid = nfsva.na_fileid;
3293 			}
3294 			if (nd->nd_repstat == 0) {
3295 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3296 			    len = fxdr_unsigned(int, *(tl + 4));
3297 			    if (len > 0 && len <= NFSX_V4FHMAX)
3298 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3299 			    else
3300 				error = EPERM;
3301 			    if (!error) {
3302 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3303 				nfsva.na_mntonfileno = UINT64_MAX;
3304 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3305 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3306 				    NULL, NULL, NULL, p, cred);
3307 				if (error) {
3308 				    dotdotfileid = dotfileid;
3309 				} else if (gotmnton) {
3310 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3311 					dotdotfileid = nfsva.na_mntonfileno;
3312 				    else
3313 					dotdotfileid = nfsva.na_fileid;
3314 				} else if (nfsva.na_filesid[0] ==
3315 				    dnp->n_vattr.na_filesid[0] &&
3316 				    nfsva.na_filesid[1] ==
3317 				    dnp->n_vattr.na_filesid[1]) {
3318 				    dotdotfileid = nfsva.na_fileid;
3319 				} else {
3320 				    do {
3321 					fakefileno--;
3322 				    } while (fakefileno ==
3323 					nfsva.na_fileid);
3324 				    dotdotfileid = fakefileno;
3325 				}
3326 			    }
3327 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3328 			    /*
3329 			     * Lookupp returns NFSERR_NOENT when we are
3330 			     * at the root, so just use the current dir.
3331 			     */
3332 			    nd->nd_repstat = 0;
3333 			    dotdotfileid = dotfileid;
3334 			} else {
3335 			    error = nd->nd_repstat;
3336 			}
3337 			m_freem(nd->nd_mrep);
3338 			if (error)
3339 			    return (error);
3340 			nd->nd_mrep = NULL;
3341 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3342 			dp->d_pad0 = dp->d_pad1 = 0;
3343 			dp->d_off = 0;
3344 			dp->d_type = DT_DIR;
3345 			dp->d_fileno = dotfileid;
3346 			dp->d_namlen = 1;
3347 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3348 			dp->d_name[0] = '.';
3349 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3350 			/*
3351 			 * Just make these offset cookie 0.
3352 			 */
3353 			tl = (u_int32_t *)&dp->d_name[8];
3354 			*tl++ = 0;
3355 			*tl = 0;
3356 			blksiz += dp->d_reclen;
3357 			uiop->uio_resid -= dp->d_reclen;
3358 			uiop->uio_offset += dp->d_reclen;
3359 			uiop->uio_iov->iov_base =
3360 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3361 			uiop->uio_iov->iov_len -= dp->d_reclen;
3362 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3363 			dp->d_pad0 = dp->d_pad1 = 0;
3364 			dp->d_off = 0;
3365 			dp->d_type = DT_DIR;
3366 			dp->d_fileno = dotdotfileid;
3367 			dp->d_namlen = 2;
3368 			*((uint64_t *)dp->d_name) = 0;
3369 			dp->d_name[0] = '.';
3370 			dp->d_name[1] = '.';
3371 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3372 			/*
3373 			 * Just make these offset cookie 0.
3374 			 */
3375 			tl = (u_int32_t *)&dp->d_name[8];
3376 			*tl++ = 0;
3377 			*tl = 0;
3378 			blksiz += dp->d_reclen;
3379 			uiop->uio_resid -= dp->d_reclen;
3380 			uiop->uio_offset += dp->d_reclen;
3381 			uiop->uio_iov->iov_base =
3382 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3383 			uiop->uio_iov->iov_len -= dp->d_reclen;
3384 		}
3385 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3386 	} else {
3387 		reqsize = 5 * NFSX_UNSIGNED;
3388 	}
3389 
3390 	/*
3391 	 * Loop around doing readdir rpc's of size readsize.
3392 	 * The stopping criteria is EOF or buffer full.
3393 	 */
3394 	while (more_dirs && bigenough) {
3395 		*attrflagp = 0;
3396 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
3397 		if (nd->nd_flag & ND_NFSV2) {
3398 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3399 			*tl++ = cookie.lval[1];
3400 			*tl = txdr_unsigned(readsize);
3401 		} else {
3402 			NFSM_BUILD(tl, u_int32_t *, reqsize);
3403 			*tl++ = cookie.lval[0];
3404 			*tl++ = cookie.lval[1];
3405 			if (cookie.qval == 0) {
3406 				*tl++ = 0;
3407 				*tl++ = 0;
3408 			} else {
3409 				NFSLOCKNODE(dnp);
3410 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
3411 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
3412 				NFSUNLOCKNODE(dnp);
3413 			}
3414 			if (nd->nd_flag & ND_NFSV4) {
3415 				*tl++ = txdr_unsigned(readsize);
3416 				*tl = txdr_unsigned(readsize);
3417 				(void) nfsrv_putattrbit(nd, &attrbits);
3418 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3419 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
3420 				(void) nfsrv_putattrbit(nd, &dattrbits);
3421 			} else {
3422 				*tl = txdr_unsigned(readsize);
3423 			}
3424 		}
3425 		error = nfscl_request(nd, vp, p, cred);
3426 		if (error)
3427 			return (error);
3428 		if (!(nd->nd_flag & ND_NFSV2)) {
3429 			if (nd->nd_flag & ND_NFSV3)
3430 				error = nfscl_postop_attr(nd, nap, attrflagp);
3431 			if (!nd->nd_repstat && !error) {
3432 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3433 				NFSLOCKNODE(dnp);
3434 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
3435 				dnp->n_cookieverf.nfsuquad[1] = *tl;
3436 				NFSUNLOCKNODE(dnp);
3437 			}
3438 		}
3439 		if (nd->nd_repstat || error) {
3440 			if (!error)
3441 				error = nd->nd_repstat;
3442 			goto nfsmout;
3443 		}
3444 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3445 		more_dirs = fxdr_unsigned(int, *tl);
3446 		if (!more_dirs)
3447 			tryformoredirs = 0;
3448 
3449 		/* loop through the dir entries, doctoring them to 4bsd form */
3450 		while (more_dirs && bigenough) {
3451 			if (nd->nd_flag & ND_NFSV4) {
3452 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3453 				ncookie.lval[0] = *tl++;
3454 				ncookie.lval[1] = *tl++;
3455 				len = fxdr_unsigned(int, *tl);
3456 			} else if (nd->nd_flag & ND_NFSV3) {
3457 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3458 				nfsva.na_fileid = fxdr_hyper(tl);
3459 				tl += 2;
3460 				len = fxdr_unsigned(int, *tl);
3461 			} else {
3462 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3463 				nfsva.na_fileid = fxdr_unsigned(uint64_t,
3464 				    *tl++);
3465 				len = fxdr_unsigned(int, *tl);
3466 			}
3467 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3468 				error = EBADRPC;
3469 				goto nfsmout;
3470 			}
3471 			tlen = roundup2(len, 8);
3472 			if (tlen == len)
3473 				tlen += 8;  /* To ensure null termination. */
3474 			left = DIRBLKSIZ - blksiz;
3475 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3476 				NFSBZERO(uiop->uio_iov->iov_base, left);
3477 				dp->d_reclen += left;
3478 				uiop->uio_iov->iov_base =
3479 				    (char *)uiop->uio_iov->iov_base + left;
3480 				uiop->uio_iov->iov_len -= left;
3481 				uiop->uio_resid -= left;
3482 				uiop->uio_offset += left;
3483 				blksiz = 0;
3484 			}
3485 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3486 			    uiop->uio_resid)
3487 				bigenough = 0;
3488 			if (bigenough) {
3489 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3490 				dp->d_pad0 = dp->d_pad1 = 0;
3491 				dp->d_off = 0;
3492 				dp->d_namlen = len;
3493 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3494 				    NFSX_HYPER;
3495 				dp->d_type = DT_UNKNOWN;
3496 				blksiz += dp->d_reclen;
3497 				if (blksiz == DIRBLKSIZ)
3498 					blksiz = 0;
3499 				uiop->uio_resid -= DIRHDSIZ;
3500 				uiop->uio_offset += DIRHDSIZ;
3501 				uiop->uio_iov->iov_base =
3502 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3503 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3504 				error = nfsm_mbufuio(nd, uiop, len);
3505 				if (error)
3506 					goto nfsmout;
3507 				cp = uiop->uio_iov->iov_base;
3508 				tlen -= len;
3509 				NFSBZERO(cp, tlen);
3510 				cp += tlen;	/* points to cookie storage */
3511 				tl2 = (u_int32_t *)cp;
3512 				uiop->uio_iov->iov_base =
3513 				    (char *)uiop->uio_iov->iov_base + tlen +
3514 				    NFSX_HYPER;
3515 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3516 				uiop->uio_resid -= tlen + NFSX_HYPER;
3517 				uiop->uio_offset += (tlen + NFSX_HYPER);
3518 			} else {
3519 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3520 				if (error)
3521 					goto nfsmout;
3522 			}
3523 			if (nd->nd_flag & ND_NFSV4) {
3524 				rderr = 0;
3525 				nfsva.na_mntonfileno = UINT64_MAX;
3526 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3527 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3528 				    NULL, NULL, &rderr, p, cred);
3529 				if (error)
3530 					goto nfsmout;
3531 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3532 			} else if (nd->nd_flag & ND_NFSV3) {
3533 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3534 				ncookie.lval[0] = *tl++;
3535 				ncookie.lval[1] = *tl++;
3536 			} else {
3537 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3538 				ncookie.lval[0] = 0;
3539 				ncookie.lval[1] = *tl++;
3540 			}
3541 			if (bigenough) {
3542 			    if (nd->nd_flag & ND_NFSV4) {
3543 				if (rderr) {
3544 				    dp->d_fileno = 0;
3545 				} else {
3546 				    if (gotmnton) {
3547 					if (nfsva.na_mntonfileno != UINT64_MAX)
3548 					    dp->d_fileno = nfsva.na_mntonfileno;
3549 					else
3550 					    dp->d_fileno = nfsva.na_fileid;
3551 				    } else if (nfsva.na_filesid[0] ==
3552 					dnp->n_vattr.na_filesid[0] &&
3553 					nfsva.na_filesid[1] ==
3554 					dnp->n_vattr.na_filesid[1]) {
3555 					dp->d_fileno = nfsva.na_fileid;
3556 				    } else {
3557 					do {
3558 					    fakefileno--;
3559 					} while (fakefileno ==
3560 					    nfsva.na_fileid);
3561 					dp->d_fileno = fakefileno;
3562 				    }
3563 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3564 				}
3565 			    } else {
3566 				dp->d_fileno = nfsva.na_fileid;
3567 			    }
3568 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3569 				ncookie.lval[0];
3570 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3571 				ncookie.lval[1];
3572 			}
3573 			more_dirs = fxdr_unsigned(int, *tl);
3574 		}
3575 		/*
3576 		 * If at end of rpc data, get the eof boolean
3577 		 */
3578 		if (!more_dirs) {
3579 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3580 			eof = fxdr_unsigned(int, *tl);
3581 			if (tryformoredirs)
3582 				more_dirs = !eof;
3583 			if (nd->nd_flag & ND_NFSV4) {
3584 				error = nfscl_postop_attr(nd, nap, attrflagp);
3585 				if (error)
3586 					goto nfsmout;
3587 			}
3588 		}
3589 		m_freem(nd->nd_mrep);
3590 		nd->nd_mrep = NULL;
3591 	}
3592 	/*
3593 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3594 	 * by increasing d_reclen for the last record.
3595 	 */
3596 	if (blksiz > 0) {
3597 		left = DIRBLKSIZ - blksiz;
3598 		NFSBZERO(uiop->uio_iov->iov_base, left);
3599 		dp->d_reclen += left;
3600 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3601 		    left;
3602 		uiop->uio_iov->iov_len -= left;
3603 		uiop->uio_resid -= left;
3604 		uiop->uio_offset += left;
3605 	}
3606 
3607 	/*
3608 	 * If returning no data, assume end of file.
3609 	 * If not bigenough, return not end of file, since you aren't
3610 	 *    returning all the data
3611 	 * Otherwise, return the eof flag from the server.
3612 	 */
3613 	if (eofp) {
3614 		if (tresid == ((size_t)(uiop->uio_resid)))
3615 			*eofp = 1;
3616 		else if (!bigenough)
3617 			*eofp = 0;
3618 		else
3619 			*eofp = eof;
3620 	}
3621 
3622 	/*
3623 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3624 	 */
3625 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3626 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3627 		NFSBZERO(dp, DIRBLKSIZ);
3628 		dp->d_type = DT_UNKNOWN;
3629 		tl = (u_int32_t *)&dp->d_name[4];
3630 		*tl++ = cookie.lval[0];
3631 		*tl = cookie.lval[1];
3632 		dp->d_reclen = DIRBLKSIZ;
3633 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3634 		    DIRBLKSIZ;
3635 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3636 		uiop->uio_resid -= DIRBLKSIZ;
3637 		uiop->uio_offset += DIRBLKSIZ;
3638 	}
3639 
3640 nfsmout:
3641 	if (nd->nd_mrep != NULL)
3642 		m_freem(nd->nd_mrep);
3643 	return (error);
3644 }
3645 
3646 /*
3647  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3648  * (Also used for NFS V4 when mount flag set.)
3649  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3650  */
3651 int
3652 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3653     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3654     int *eofp)
3655 {
3656 	int len, left;
3657 	struct dirent *dp = NULL;
3658 	u_int32_t *tl;
3659 	vnode_t newvp = NULLVP;
3660 	struct nfsrv_descript nfsd, *nd = &nfsd;
3661 	struct nameidata nami, *ndp = &nami;
3662 	struct componentname *cnp = &ndp->ni_cnd;
3663 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3664 	struct nfsnode *dnp = VTONFS(vp), *np;
3665 	struct nfsvattr nfsva;
3666 	struct nfsfh *nfhp;
3667 	nfsquad_t cookie, ncookie;
3668 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3669 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3670 	int isdotdot = 0, unlocknewvp = 0;
3671 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3672 	u_int64_t fileno = 0;
3673 	char *cp;
3674 	nfsattrbit_t attrbits, dattrbits;
3675 	size_t tresid;
3676 	u_int32_t *tl2 = NULL, rderr;
3677 	struct timespec dctime, ts;
3678 	bool attr_ok;
3679 
3680 	KASSERT(uiop->uio_iovcnt == 1 &&
3681 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3682 	    ("nfs readdirplusrpc bad uio"));
3683 	ncookie.lval[0] = ncookie.lval[1] = 0;
3684 	timespecclear(&dctime);
3685 	*attrflagp = 0;
3686 	if (eofp != NULL)
3687 		*eofp = 0;
3688 	ndp->ni_dvp = vp;
3689 	nd->nd_mrep = NULL;
3690 	cookie.lval[0] = cookiep->nfsuquad[0];
3691 	cookie.lval[1] = cookiep->nfsuquad[1];
3692 	tresid = uiop->uio_resid;
3693 
3694 	/*
3695 	 * For NFSv4, first create the "." and ".." entries.
3696 	 */
3697 	if (NFSHASNFSV4(nmp)) {
3698 		NFSGETATTR_ATTRBIT(&dattrbits);
3699 		NFSZERO_ATTRBIT(&attrbits);
3700 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3701 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3702 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3703 			NFSSETBIT_ATTRBIT(&attrbits,
3704 			    NFSATTRBIT_MOUNTEDONFILEID);
3705 			gotmnton = 1;
3706 		} else {
3707 			/*
3708 			 * Must fake it. Use the fileno, except when the
3709 			 * fsid is != to that of the directory. For that
3710 			 * case, generate a fake fileno that is not the same.
3711 			 */
3712 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3713 			gotmnton = 0;
3714 		}
3715 
3716 		/*
3717 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3718 		 */
3719 		if (uiop->uio_offset == 0) {
3720 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3721 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3722 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3723 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3724 			(void) nfsrv_putattrbit(nd, &attrbits);
3725 			error = nfscl_request(nd, vp, p, cred);
3726 			if (error)
3727 			    return (error);
3728 			dotfileid = 0;	/* Fake out the compiler. */
3729 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3730 			    error = nfsm_loadattr(nd, &nfsva);
3731 			    if (error != 0)
3732 				goto nfsmout;
3733 			    dctime = nfsva.na_ctime;
3734 			    dotfileid = nfsva.na_fileid;
3735 			}
3736 			if (nd->nd_repstat == 0) {
3737 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3738 			    len = fxdr_unsigned(int, *(tl + 4));
3739 			    if (len > 0 && len <= NFSX_V4FHMAX)
3740 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3741 			    else
3742 				error = EPERM;
3743 			    if (!error) {
3744 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3745 				nfsva.na_mntonfileno = UINT64_MAX;
3746 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3747 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3748 				    NULL, NULL, NULL, p, cred);
3749 				if (error) {
3750 				    dotdotfileid = dotfileid;
3751 				} else if (gotmnton) {
3752 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3753 					dotdotfileid = nfsva.na_mntonfileno;
3754 				    else
3755 					dotdotfileid = nfsva.na_fileid;
3756 				} else if (nfsva.na_filesid[0] ==
3757 				    dnp->n_vattr.na_filesid[0] &&
3758 				    nfsva.na_filesid[1] ==
3759 				    dnp->n_vattr.na_filesid[1]) {
3760 				    dotdotfileid = nfsva.na_fileid;
3761 				} else {
3762 				    do {
3763 					fakefileno--;
3764 				    } while (fakefileno ==
3765 					nfsva.na_fileid);
3766 				    dotdotfileid = fakefileno;
3767 				}
3768 			    }
3769 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3770 			    /*
3771 			     * Lookupp returns NFSERR_NOENT when we are
3772 			     * at the root, so just use the current dir.
3773 			     */
3774 			    nd->nd_repstat = 0;
3775 			    dotdotfileid = dotfileid;
3776 			} else {
3777 			    error = nd->nd_repstat;
3778 			}
3779 			m_freem(nd->nd_mrep);
3780 			if (error)
3781 			    return (error);
3782 			nd->nd_mrep = NULL;
3783 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3784 			dp->d_pad0 = dp->d_pad1 = 0;
3785 			dp->d_off = 0;
3786 			dp->d_type = DT_DIR;
3787 			dp->d_fileno = dotfileid;
3788 			dp->d_namlen = 1;
3789 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3790 			dp->d_name[0] = '.';
3791 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3792 			/*
3793 			 * Just make these offset cookie 0.
3794 			 */
3795 			tl = (u_int32_t *)&dp->d_name[8];
3796 			*tl++ = 0;
3797 			*tl = 0;
3798 			blksiz += dp->d_reclen;
3799 			uiop->uio_resid -= dp->d_reclen;
3800 			uiop->uio_offset += dp->d_reclen;
3801 			uiop->uio_iov->iov_base =
3802 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3803 			uiop->uio_iov->iov_len -= dp->d_reclen;
3804 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3805 			dp->d_pad0 = dp->d_pad1 = 0;
3806 			dp->d_off = 0;
3807 			dp->d_type = DT_DIR;
3808 			dp->d_fileno = dotdotfileid;
3809 			dp->d_namlen = 2;
3810 			*((uint64_t *)dp->d_name) = 0;
3811 			dp->d_name[0] = '.';
3812 			dp->d_name[1] = '.';
3813 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3814 			/*
3815 			 * Just make these offset cookie 0.
3816 			 */
3817 			tl = (u_int32_t *)&dp->d_name[8];
3818 			*tl++ = 0;
3819 			*tl = 0;
3820 			blksiz += dp->d_reclen;
3821 			uiop->uio_resid -= dp->d_reclen;
3822 			uiop->uio_offset += dp->d_reclen;
3823 			uiop->uio_iov->iov_base =
3824 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3825 			uiop->uio_iov->iov_len -= dp->d_reclen;
3826 		}
3827 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3828 		if (gotmnton)
3829 			NFSSETBIT_ATTRBIT(&attrbits,
3830 			    NFSATTRBIT_MOUNTEDONFILEID);
3831 		if (!NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3832 		    NFSATTRBIT_TIMECREATE))
3833 			NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMECREATE);
3834 	}
3835 
3836 	/*
3837 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3838 	 * The stopping criteria is EOF or buffer full.
3839 	 */
3840 	while (more_dirs && bigenough) {
3841 		*attrflagp = 0;
3842 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3843  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3844 		*tl++ = cookie.lval[0];
3845 		*tl++ = cookie.lval[1];
3846 		if (cookie.qval == 0) {
3847 			*tl++ = 0;
3848 			*tl++ = 0;
3849 		} else {
3850 			NFSLOCKNODE(dnp);
3851 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3852 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3853 			NFSUNLOCKNODE(dnp);
3854 		}
3855 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3856 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3857 		if (nd->nd_flag & ND_NFSV4) {
3858 			(void) nfsrv_putattrbit(nd, &attrbits);
3859 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3860 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3861 			(void) nfsrv_putattrbit(nd, &dattrbits);
3862 		}
3863 		nanouptime(&ts);
3864 		error = nfscl_request(nd, vp, p, cred);
3865 		if (error)
3866 			return (error);
3867 		if (nd->nd_flag & ND_NFSV3)
3868 			error = nfscl_postop_attr(nd, nap, attrflagp);
3869 		if (nd->nd_repstat || error) {
3870 			if (!error)
3871 				error = nd->nd_repstat;
3872 			goto nfsmout;
3873 		}
3874 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3875 			dctime = nap->na_ctime;
3876 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3877 		NFSLOCKNODE(dnp);
3878 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3879 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3880 		NFSUNLOCKNODE(dnp);
3881 		more_dirs = fxdr_unsigned(int, *tl);
3882 		if (!more_dirs)
3883 			tryformoredirs = 0;
3884 
3885 		/* loop through the dir entries, doctoring them to 4bsd form */
3886 		while (more_dirs && bigenough) {
3887 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3888 			if (nd->nd_flag & ND_NFSV4) {
3889 				ncookie.lval[0] = *tl++;
3890 				ncookie.lval[1] = *tl++;
3891 			} else {
3892 				fileno = fxdr_hyper(tl);
3893 				tl += 2;
3894 			}
3895 			len = fxdr_unsigned(int, *tl);
3896 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3897 				error = EBADRPC;
3898 				goto nfsmout;
3899 			}
3900 			tlen = roundup2(len, 8);
3901 			if (tlen == len)
3902 				tlen += 8;  /* To ensure null termination. */
3903 			left = DIRBLKSIZ - blksiz;
3904 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3905 				NFSBZERO(uiop->uio_iov->iov_base, left);
3906 				dp->d_reclen += left;
3907 				uiop->uio_iov->iov_base =
3908 				    (char *)uiop->uio_iov->iov_base + left;
3909 				uiop->uio_iov->iov_len -= left;
3910 				uiop->uio_resid -= left;
3911 				uiop->uio_offset += left;
3912 				blksiz = 0;
3913 			}
3914 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3915 			    uiop->uio_resid)
3916 				bigenough = 0;
3917 			if (bigenough) {
3918 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3919 				dp->d_pad0 = dp->d_pad1 = 0;
3920 				dp->d_off = 0;
3921 				dp->d_namlen = len;
3922 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3923 				    NFSX_HYPER;
3924 				dp->d_type = DT_UNKNOWN;
3925 				blksiz += dp->d_reclen;
3926 				if (blksiz == DIRBLKSIZ)
3927 					blksiz = 0;
3928 				uiop->uio_resid -= DIRHDSIZ;
3929 				uiop->uio_offset += DIRHDSIZ;
3930 				uiop->uio_iov->iov_base =
3931 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3932 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3933 				cnp->cn_nameptr = uiop->uio_iov->iov_base;
3934 				cnp->cn_namelen = len;
3935 				NFSCNHASHZERO(cnp);
3936 				error = nfsm_mbufuio(nd, uiop, len);
3937 				if (error)
3938 					goto nfsmout;
3939 				cp = uiop->uio_iov->iov_base;
3940 				tlen -= len;
3941 				NFSBZERO(cp, tlen);
3942 				cp += tlen;	/* points to cookie storage */
3943 				tl2 = (u_int32_t *)cp;
3944 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3945 				    cnp->cn_nameptr[1] == '.')
3946 					isdotdot = 1;
3947 				else
3948 					isdotdot = 0;
3949 				uiop->uio_iov->iov_base =
3950 				    (char *)uiop->uio_iov->iov_base + tlen +
3951 				    NFSX_HYPER;
3952 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3953 				uiop->uio_resid -= tlen + NFSX_HYPER;
3954 				uiop->uio_offset += (tlen + NFSX_HYPER);
3955 			} else {
3956 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3957 				if (error)
3958 					goto nfsmout;
3959 			}
3960 			nfhp = NULL;
3961 			if (nd->nd_flag & ND_NFSV3) {
3962 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3963 				ncookie.lval[0] = *tl++;
3964 				ncookie.lval[1] = *tl++;
3965 				attrflag = fxdr_unsigned(int, *tl);
3966 				if (attrflag) {
3967 				  error = nfsm_loadattr(nd, &nfsva);
3968 				  if (error)
3969 					goto nfsmout;
3970 				}
3971 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3972 				if (*tl) {
3973 					error = nfsm_getfh(nd, &nfhp);
3974 					if (error)
3975 					    goto nfsmout;
3976 				}
3977 				if (!attrflag && nfhp != NULL) {
3978 					free(nfhp, M_NFSFH);
3979 					nfhp = NULL;
3980 				}
3981 			} else {
3982 				rderr = 0;
3983 				nfsva.na_mntonfileno = 0xffffffff;
3984 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3985 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3986 				    NULL, NULL, &rderr, p, cred);
3987 				if (error)
3988 					goto nfsmout;
3989 			}
3990 
3991 			if (bigenough) {
3992 			    if (nd->nd_flag & ND_NFSV4) {
3993 				if (rderr) {
3994 				    dp->d_fileno = 0;
3995 				} else if (gotmnton) {
3996 				    if (nfsva.na_mntonfileno != 0xffffffff)
3997 					dp->d_fileno = nfsva.na_mntonfileno;
3998 				    else
3999 					dp->d_fileno = nfsva.na_fileid;
4000 				} else if (nfsva.na_filesid[0] ==
4001 				    dnp->n_vattr.na_filesid[0] &&
4002 				    nfsva.na_filesid[1] ==
4003 				    dnp->n_vattr.na_filesid[1]) {
4004 				    dp->d_fileno = nfsva.na_fileid;
4005 				} else {
4006 				    do {
4007 					fakefileno--;
4008 				    } while (fakefileno ==
4009 					nfsva.na_fileid);
4010 				    dp->d_fileno = fakefileno;
4011 				}
4012 			    } else {
4013 				dp->d_fileno = fileno;
4014 			    }
4015 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
4016 				ncookie.lval[0];
4017 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
4018 				ncookie.lval[1];
4019 
4020 			    if (nfhp != NULL) {
4021 				attr_ok = true;
4022 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
4023 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
4024 				    VREF(vp);
4025 				    newvp = vp;
4026 				    unlocknewvp = 0;
4027 				    free(nfhp, M_NFSFH);
4028 				    np = dnp;
4029 				} else if (isdotdot != 0) {
4030 				    /*
4031 				     * Skip doing a nfscl_nget() call for "..".
4032 				     * There's a race between acquiring the nfs
4033 				     * node here and lookups that look for the
4034 				     * directory being read (in the parent).
4035 				     * It would try to get a lock on ".." here,
4036 				     * owning the lock on the directory being
4037 				     * read. Lookup will hold the lock on ".."
4038 				     * and try to acquire the lock on the
4039 				     * directory being read.
4040 				     * If the directory is unlocked/relocked,
4041 				     * then there is a LOR with the buflock
4042 				     * vp is relocked.
4043 				     */
4044 				    free(nfhp, M_NFSFH);
4045 				} else {
4046 				    error = nfscl_nget(vp->v_mount, vp,
4047 				      nfhp, cnp, p, &np, LK_EXCLUSIVE);
4048 				    if (!error) {
4049 					newvp = NFSTOV(np);
4050 					unlocknewvp = 1;
4051 					/*
4052 					 * If n_localmodtime >= time before RPC,
4053 					 * then a file modification operation,
4054 					 * such as VOP_SETATTR() of size, has
4055 					 * occurred while the Lookup RPC and
4056 					 * acquisition of the vnode happened. As
4057 					 * such, the attributes might be stale,
4058 					 * with possibly an incorrect size.
4059 					 */
4060 					NFSLOCKNODE(np);
4061 					if (timespecisset(
4062 					    &np->n_localmodtime) &&
4063 					    timespeccmp(&np->n_localmodtime,
4064 					    &ts, >=)) {
4065 					    NFSCL_DEBUG(4, "nfsrpc_readdirplus:"
4066 						" localmod stale attributes\n");
4067 					    attr_ok = false;
4068 					}
4069 					NFSUNLOCKNODE(np);
4070 				    }
4071 				}
4072 				nfhp = NULL;
4073 				if (newvp != NULLVP) {
4074 				    if (attr_ok)
4075 					error = nfscl_loadattrcache(&newvp,
4076 					    &nfsva, NULL, 0, 0);
4077 				    if (error) {
4078 					if (unlocknewvp)
4079 					    vput(newvp);
4080 					else
4081 					    vrele(newvp);
4082 					goto nfsmout;
4083 				    }
4084 				    dp->d_type =
4085 					vtonfs_dtype(np->n_vattr.na_type);
4086 				    ndp->ni_vp = newvp;
4087 				    NFSCNHASH(cnp, HASHINIT);
4088 				    if (cnp->cn_namelen <= NCHNAMLEN &&
4089 					ndp->ni_dvp != ndp->ni_vp &&
4090 					(newvp->v_type != VDIR ||
4091 					 dctime.tv_sec != 0)) {
4092 					cache_enter_time_flags(ndp->ni_dvp,
4093 					    ndp->ni_vp, cnp,
4094 					    &nfsva.na_ctime,
4095 					    newvp->v_type != VDIR ? NULL :
4096 					    &dctime, VFS_CACHE_DROPOLD);
4097 				    }
4098 				    if (unlocknewvp)
4099 					vput(newvp);
4100 				    else
4101 					vrele(newvp);
4102 				    newvp = NULLVP;
4103 				}
4104 			    }
4105 			} else if (nfhp != NULL) {
4106 			    free(nfhp, M_NFSFH);
4107 			}
4108 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
4109 			more_dirs = fxdr_unsigned(int, *tl);
4110 		}
4111 		/*
4112 		 * If at end of rpc data, get the eof boolean
4113 		 */
4114 		if (!more_dirs) {
4115 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
4116 			eof = fxdr_unsigned(int, *tl);
4117 			if (tryformoredirs)
4118 				more_dirs = !eof;
4119 			if (nd->nd_flag & ND_NFSV4) {
4120 				error = nfscl_postop_attr(nd, nap, attrflagp);
4121 				if (error)
4122 					goto nfsmout;
4123 			}
4124 		}
4125 		m_freem(nd->nd_mrep);
4126 		nd->nd_mrep = NULL;
4127 	}
4128 	/*
4129 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
4130 	 * by increasing d_reclen for the last record.
4131 	 */
4132 	if (blksiz > 0) {
4133 		left = DIRBLKSIZ - blksiz;
4134 		NFSBZERO(uiop->uio_iov->iov_base, left);
4135 		dp->d_reclen += left;
4136 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
4137 		    left;
4138 		uiop->uio_iov->iov_len -= left;
4139 		uiop->uio_resid -= left;
4140 		uiop->uio_offset += left;
4141 	}
4142 
4143 	/*
4144 	 * If returning no data, assume end of file.
4145 	 * If not bigenough, return not end of file, since you aren't
4146 	 *    returning all the data
4147 	 * Otherwise, return the eof flag from the server.
4148 	 */
4149 	if (eofp != NULL) {
4150 		if (tresid == uiop->uio_resid)
4151 			*eofp = 1;
4152 		else if (!bigenough)
4153 			*eofp = 0;
4154 		else
4155 			*eofp = eof;
4156 	}
4157 
4158 	/*
4159 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
4160 	 */
4161 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
4162 		dp = (struct dirent *)uiop->uio_iov->iov_base;
4163 		NFSBZERO(dp, DIRBLKSIZ);
4164 		dp->d_type = DT_UNKNOWN;
4165 		tl = (u_int32_t *)&dp->d_name[4];
4166 		*tl++ = cookie.lval[0];
4167 		*tl = cookie.lval[1];
4168 		dp->d_reclen = DIRBLKSIZ;
4169 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
4170 		    DIRBLKSIZ;
4171 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
4172 		uiop->uio_resid -= DIRBLKSIZ;
4173 		uiop->uio_offset += DIRBLKSIZ;
4174 	}
4175 
4176 nfsmout:
4177 	if (nd->nd_mrep != NULL)
4178 		m_freem(nd->nd_mrep);
4179 	return (error);
4180 }
4181 
4182 /*
4183  * Nfs commit rpc
4184  */
4185 int
4186 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
4187     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4188 {
4189 	u_int32_t *tl;
4190 	struct nfsrv_descript nfsd, *nd = &nfsd;
4191 	nfsattrbit_t attrbits;
4192 	int error;
4193 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4194 
4195 	*attrflagp = 0;
4196 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
4197 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
4198 	txdr_hyper(offset, tl);
4199 	tl += 2;
4200 	*tl = txdr_unsigned(cnt);
4201 	if (nd->nd_flag & ND_NFSV4) {
4202 		/*
4203 		 * And do a Getattr op.
4204 		 */
4205 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4206 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
4207 		NFSGETATTR_ATTRBIT(&attrbits);
4208 		(void) nfsrv_putattrbit(nd, &attrbits);
4209 	}
4210 	error = nfscl_request(nd, vp, p, cred);
4211 	if (error)
4212 		return (error);
4213 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, NULL);
4214 	if (!error && !nd->nd_repstat) {
4215 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
4216 		NFSLOCKMNT(nmp);
4217 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
4218 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
4219 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
4220 		}
4221 		NFSUNLOCKMNT(nmp);
4222 		if (nd->nd_flag & ND_NFSV4)
4223 			error = nfscl_postop_attr(nd, nap, attrflagp);
4224 	}
4225 nfsmout:
4226 	if (!error && nd->nd_repstat)
4227 		error = nd->nd_repstat;
4228 	m_freem(nd->nd_mrep);
4229 	return (error);
4230 }
4231 
4232 /*
4233  * NFS byte range lock rpc.
4234  * (Mostly just calls one of the three lower level RPC routines.)
4235  */
4236 int
4237 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
4238     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4239 {
4240 	struct nfscllockowner *lp;
4241 	struct nfsclclient *clp;
4242 	struct nfsfh *nfhp;
4243 	struct nfsrv_descript nfsd, *nd = &nfsd;
4244 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4245 	u_int64_t off, len;
4246 	off_t start, end;
4247 	u_int32_t clidrev = 0;
4248 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
4249 	int callcnt, dorpc;
4250 
4251 	/*
4252 	 * Convert the flock structure into a start and end and do POSIX
4253 	 * bounds checking.
4254 	 */
4255 	switch (fl->l_whence) {
4256 	case SEEK_SET:
4257 	case SEEK_CUR:
4258 		/*
4259 		 * Caller is responsible for adding any necessary offset
4260 		 * when SEEK_CUR is used.
4261 		 */
4262 		start = fl->l_start;
4263 		off = fl->l_start;
4264 		break;
4265 	case SEEK_END:
4266 		start = size + fl->l_start;
4267 		off = size + fl->l_start;
4268 		break;
4269 	default:
4270 		return (EINVAL);
4271 	}
4272 	if (start < 0)
4273 		return (EINVAL);
4274 	if (fl->l_len != 0) {
4275 		end = start + fl->l_len - 1;
4276 		if (end < start)
4277 			return (EINVAL);
4278 	}
4279 
4280 	len = fl->l_len;
4281 	if (len == 0)
4282 		len = NFS64BITSSET;
4283 	retrycnt = 0;
4284 	do {
4285 	    nd->nd_repstat = 0;
4286 	    if (op == F_GETLK) {
4287 		error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
4288 		if (error)
4289 			return (error);
4290 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
4291 		if (!error) {
4292 			clidrev = clp->nfsc_clientidrev;
4293 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
4294 			    p, id, flags);
4295 		} else if (error == -1) {
4296 			error = 0;
4297 		}
4298 		nfscl_clientrelease(clp);
4299 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
4300 		/*
4301 		 * We must loop around for all lockowner cases.
4302 		 */
4303 		callcnt = 0;
4304 		error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
4305 		if (error)
4306 			return (error);
4307 		do {
4308 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
4309 			clp, id, flags, &lp, &dorpc);
4310 		    /*
4311 		     * If it returns a NULL lp, we're done.
4312 		     */
4313 		    if (lp == NULL) {
4314 			if (callcnt == 0)
4315 			    nfscl_clientrelease(clp);
4316 			else
4317 			    nfscl_releasealllocks(clp, vp, p, id, flags);
4318 			return (error);
4319 		    }
4320 		    if (nmp->nm_clp != NULL)
4321 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4322 		    else
4323 			clidrev = 0;
4324 		    /*
4325 		     * If the server doesn't support Posix lock semantics,
4326 		     * only allow locks on the entire file, since it won't
4327 		     * handle overlapping byte ranges.
4328 		     * There might still be a problem when a lock
4329 		     * upgrade/downgrade (read<->write) occurs, since the
4330 		     * server "might" expect an unlock first?
4331 		     */
4332 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
4333 			(off == 0 && len == NFS64BITSSET))) {
4334 			/*
4335 			 * Since the lock records will go away, we must
4336 			 * wait for grace and delay here.
4337 			 */
4338 			do {
4339 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
4340 				NFSV4LOCKT_READ, cred, p, 0);
4341 			    if ((nd->nd_repstat == NFSERR_GRACE ||
4342 				 nd->nd_repstat == NFSERR_DELAY) &&
4343 				error == 0)
4344 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
4345 				    "nfs_advlock");
4346 			} while ((nd->nd_repstat == NFSERR_GRACE ||
4347 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
4348 		    }
4349 		    callcnt++;
4350 		} while (error == 0 && nd->nd_repstat == 0);
4351 		nfscl_releasealllocks(clp, vp, p, id, flags);
4352 	    } else if (op == F_SETLK) {
4353 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
4354 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
4355 		if (error || donelocally) {
4356 			return (error);
4357 		}
4358 		if (nmp->nm_clp != NULL)
4359 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4360 		else
4361 			clidrev = 0;
4362 		nfhp = VTONFS(vp)->n_fhp;
4363 		if (!lp->nfsl_open->nfso_posixlock &&
4364 		    (off != 0 || len != NFS64BITSSET)) {
4365 			error = EINVAL;
4366 		} else {
4367 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
4368 			    nfhp->nfh_len, lp, newone, reclaim, off,
4369 			    len, fl->l_type, cred, p, 0);
4370 		}
4371 		if (!error)
4372 			error = nd->nd_repstat;
4373 		nfscl_lockrelease(lp, error, newone);
4374 	    } else {
4375 		error = EINVAL;
4376 	    }
4377 	    if (!error)
4378 	        error = nd->nd_repstat;
4379 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
4380 		error == NFSERR_STALEDONTRECOVER ||
4381 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4382 		error == NFSERR_BADSESSION) {
4383 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
4384 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
4385 		&& clidrev != 0) {
4386 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
4387 		retrycnt++;
4388 	    }
4389 	} while (error == NFSERR_GRACE ||
4390 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4391 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
4392 	    error == NFSERR_BADSESSION ||
4393 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
4394 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
4395 	if (error && retrycnt >= 4)
4396 		error = EIO;
4397 	return (error);
4398 }
4399 
4400 /*
4401  * The lower level routine for the LockT case.
4402  */
4403 int
4404 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
4405     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
4406     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4407 {
4408 	u_int32_t *tl;
4409 	int error, type, size;
4410 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4411 	struct nfsnode *np;
4412 	struct nfsmount *nmp;
4413 	struct nfsclsession *tsep;
4414 
4415 	nmp = VFSTONFS(vp->v_mount);
4416 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
4417 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4418 	if (fl->l_type == F_RDLCK)
4419 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4420 	else
4421 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4422 	txdr_hyper(off, tl);
4423 	tl += 2;
4424 	txdr_hyper(len, tl);
4425 	tl += 2;
4426 	tsep = nfsmnt_mdssession(nmp);
4427 	*tl++ = tsep->nfsess_clientid.lval[0];
4428 	*tl = tsep->nfsess_clientid.lval[1];
4429 	nfscl_filllockowner(id, own, flags);
4430 	np = VTONFS(vp);
4431 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4432 	    np->n_fhp->nfh_len);
4433 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4434 	error = nfscl_request(nd, vp, p, cred);
4435 	if (error)
4436 		return (error);
4437 	if (nd->nd_repstat == 0) {
4438 		fl->l_type = F_UNLCK;
4439 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4440 		nd->nd_repstat = 0;
4441 		fl->l_whence = SEEK_SET;
4442 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4443 		fl->l_start = fxdr_hyper(tl);
4444 		tl += 2;
4445 		len = fxdr_hyper(tl);
4446 		tl += 2;
4447 		if (len == NFS64BITSSET)
4448 			fl->l_len = 0;
4449 		else
4450 			fl->l_len = len;
4451 		type = fxdr_unsigned(int, *tl++);
4452 		if (type == NFSV4LOCKT_WRITE)
4453 			fl->l_type = F_WRLCK;
4454 		else
4455 			fl->l_type = F_RDLCK;
4456 		/*
4457 		 * XXX For now, I have no idea what to do with the
4458 		 * conflicting lock_owner, so I'll just set the pid == 0
4459 		 * and skip over the lock_owner.
4460 		 */
4461 		fl->l_pid = (pid_t)0;
4462 		tl += 2;
4463 		size = fxdr_unsigned(int, *tl);
4464 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4465 			error = EBADRPC;
4466 		if (!error)
4467 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4468 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4469 		nfscl_initiate_recovery(clp);
4470 nfsmout:
4471 	m_freem(nd->nd_mrep);
4472 	return (error);
4473 }
4474 
4475 /*
4476  * Lower level function that performs the LockU RPC.
4477  */
4478 static int
4479 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4480     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4481     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4482 {
4483 	u_int32_t *tl;
4484 	int error;
4485 
4486 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4487 	    lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0);
4488 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4489 	*tl++ = txdr_unsigned(type);
4490 	*tl = txdr_unsigned(lp->nfsl_seqid);
4491 	if (nfstest_outofseq &&
4492 	    (arc4random() % nfstest_outofseq) == 0)
4493 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
4494 	tl++;
4495 	if (NFSHASNFSV4N(nmp))
4496 		*tl++ = 0;
4497 	else
4498 		*tl++ = lp->nfsl_stateid.seqid;
4499 	*tl++ = lp->nfsl_stateid.other[0];
4500 	*tl++ = lp->nfsl_stateid.other[1];
4501 	*tl++ = lp->nfsl_stateid.other[2];
4502 	txdr_hyper(off, tl);
4503 	tl += 2;
4504 	txdr_hyper(len, tl);
4505 	if (syscred)
4506 		nd->nd_flag |= ND_USEGSSNAME;
4507 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4508 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4509 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4510 	if (error)
4511 		return (error);
4512 	if (nd->nd_repstat == 0) {
4513 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4514 		lp->nfsl_stateid.seqid = *tl++;
4515 		lp->nfsl_stateid.other[0] = *tl++;
4516 		lp->nfsl_stateid.other[1] = *tl++;
4517 		lp->nfsl_stateid.other[2] = *tl;
4518 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4519 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4520 nfsmout:
4521 	m_freem(nd->nd_mrep);
4522 	return (error);
4523 }
4524 
4525 /*
4526  * The actual Lock RPC.
4527  */
4528 int
4529 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4530     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4531     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4532     NFSPROC_T *p, int syscred)
4533 {
4534 	u_int32_t *tl;
4535 	int error, size;
4536 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4537 	struct nfsclsession *tsep;
4538 
4539 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
4540 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4541 	if (type == F_RDLCK)
4542 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4543 	else
4544 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4545 	*tl++ = txdr_unsigned(reclaim);
4546 	txdr_hyper(off, tl);
4547 	tl += 2;
4548 	txdr_hyper(len, tl);
4549 	tl += 2;
4550 	if (newone) {
4551 	    *tl = newnfs_true;
4552 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4553 		2 * NFSX_UNSIGNED + NFSX_HYPER);
4554 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4555 	    if (NFSHASNFSV4N(nmp))
4556 		*tl++ = 0;
4557 	    else
4558 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
4559 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4560 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4561 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4562 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
4563 	    tsep = nfsmnt_mdssession(nmp);
4564 	    *tl++ = tsep->nfsess_clientid.lval[0];
4565 	    *tl = tsep->nfsess_clientid.lval[1];
4566 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4567 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4568 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4569 	} else {
4570 	    *tl = newnfs_false;
4571 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4572 	    if (NFSHASNFSV4N(nmp))
4573 		*tl++ = 0;
4574 	    else
4575 		*tl++ = lp->nfsl_stateid.seqid;
4576 	    *tl++ = lp->nfsl_stateid.other[0];
4577 	    *tl++ = lp->nfsl_stateid.other[1];
4578 	    *tl++ = lp->nfsl_stateid.other[2];
4579 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4580 	    if (nfstest_outofseq &&
4581 		(arc4random() % nfstest_outofseq) == 0)
4582 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4583 	}
4584 	if (syscred)
4585 		nd->nd_flag |= ND_USEGSSNAME;
4586 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4587 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4588 	if (error)
4589 		return (error);
4590 	if (newone)
4591 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4592 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4593 	if (nd->nd_repstat == 0) {
4594 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4595 		lp->nfsl_stateid.seqid = *tl++;
4596 		lp->nfsl_stateid.other[0] = *tl++;
4597 		lp->nfsl_stateid.other[1] = *tl++;
4598 		lp->nfsl_stateid.other[2] = *tl;
4599 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4600 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4601 		size = fxdr_unsigned(int, *(tl + 7));
4602 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4603 			error = EBADRPC;
4604 		if (!error)
4605 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4606 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4607 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4608 nfsmout:
4609 	m_freem(nd->nd_mrep);
4610 	return (error);
4611 }
4612 
4613 /*
4614  * nfs statfs rpc
4615  * (always called with the vp for the mount point)
4616  */
4617 int
4618 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4619     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4620 {
4621 	u_int32_t *tl = NULL;
4622 	struct nfsrv_descript nfsd, *nd = &nfsd;
4623 	struct nfsmount *nmp;
4624 	nfsattrbit_t attrbits;
4625 	int error;
4626 
4627 	*attrflagp = 0;
4628 	nmp = VFSTONFS(vp->v_mount);
4629 	if (NFSHASNFSV4(nmp)) {
4630 		/*
4631 		 * For V4, you actually do a getattr.
4632 		 */
4633 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4634 		NFSSTATFS_GETATTRBIT(&attrbits);
4635 		(void) nfsrv_putattrbit(nd, &attrbits);
4636 		nd->nd_flag |= ND_USEGSSNAME;
4637 		error = nfscl_request(nd, vp, p, cred);
4638 		if (error)
4639 			return (error);
4640 		if (nd->nd_repstat == 0) {
4641 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4642 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4643 			    cred);
4644 			if (!error) {
4645 				nmp->nm_fsid[0] = nap->na_filesid[0];
4646 				nmp->nm_fsid[1] = nap->na_filesid[1];
4647 				NFSSETHASSETFSID(nmp);
4648 				*attrflagp = 1;
4649 			}
4650 		} else {
4651 			error = nd->nd_repstat;
4652 		}
4653 		if (error)
4654 			goto nfsmout;
4655 	} else {
4656 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4657 		error = nfscl_request(nd, vp, p, cred);
4658 		if (error)
4659 			return (error);
4660 		if (nd->nd_flag & ND_NFSV3) {
4661 			error = nfscl_postop_attr(nd, nap, attrflagp);
4662 			if (error)
4663 				goto nfsmout;
4664 		}
4665 		if (nd->nd_repstat) {
4666 			error = nd->nd_repstat;
4667 			goto nfsmout;
4668 		}
4669 		NFSM_DISSECT(tl, u_int32_t *,
4670 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4671 	}
4672 	if (NFSHASNFSV3(nmp)) {
4673 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4674 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4675 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4676 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4677 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4678 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4679 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4680 	} else if (NFSHASNFSV4(nmp) == 0) {
4681 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4682 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4683 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4684 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4685 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4686 	}
4687 nfsmout:
4688 	m_freem(nd->nd_mrep);
4689 	return (error);
4690 }
4691 
4692 /*
4693  * nfs pathconf rpc
4694  */
4695 int
4696 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4697     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4698 {
4699 	struct nfsrv_descript nfsd, *nd = &nfsd;
4700 	struct nfsmount *nmp;
4701 	u_int32_t *tl;
4702 	nfsattrbit_t attrbits;
4703 	int error;
4704 
4705 	*attrflagp = 0;
4706 	nmp = VFSTONFS(vp->v_mount);
4707 	if (NFSHASNFSV4(nmp)) {
4708 		/*
4709 		 * For V4, you actually do a getattr.
4710 		 */
4711 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4712 		NFSPATHCONF_GETATTRBIT(&attrbits);
4713 		(void) nfsrv_putattrbit(nd, &attrbits);
4714 		nd->nd_flag |= ND_USEGSSNAME;
4715 		error = nfscl_request(nd, vp, p, cred);
4716 		if (error)
4717 			return (error);
4718 		if (nd->nd_repstat == 0) {
4719 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4720 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4721 			    cred);
4722 			if (!error)
4723 				*attrflagp = 1;
4724 		} else {
4725 			error = nd->nd_repstat;
4726 		}
4727 	} else {
4728 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4729 		error = nfscl_request(nd, vp, p, cred);
4730 		if (error)
4731 			return (error);
4732 		error = nfscl_postop_attr(nd, nap, attrflagp);
4733 		if (nd->nd_repstat && !error)
4734 			error = nd->nd_repstat;
4735 		if (!error) {
4736 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4737 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4738 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4739 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4740 			pc->pc_chownrestricted =
4741 			    fxdr_unsigned(u_int32_t, *tl++);
4742 			pc->pc_caseinsensitive =
4743 			    fxdr_unsigned(u_int32_t, *tl++);
4744 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4745 		}
4746 	}
4747 nfsmout:
4748 	m_freem(nd->nd_mrep);
4749 	return (error);
4750 }
4751 
4752 /*
4753  * nfs version 3 fsinfo rpc call
4754  */
4755 int
4756 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4757     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4758 {
4759 	u_int32_t *tl;
4760 	struct nfsrv_descript nfsd, *nd = &nfsd;
4761 	int error;
4762 
4763 	*attrflagp = 0;
4764 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4765 	error = nfscl_request(nd, vp, p, cred);
4766 	if (error)
4767 		return (error);
4768 	error = nfscl_postop_attr(nd, nap, attrflagp);
4769 	if (nd->nd_repstat && !error)
4770 		error = nd->nd_repstat;
4771 	if (!error) {
4772 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4773 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4774 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4775 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4776 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4777 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4778 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4779 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4780 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4781 		tl += 2;
4782 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4783 		tl += 2;
4784 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4785 	}
4786 nfsmout:
4787 	m_freem(nd->nd_mrep);
4788 	return (error);
4789 }
4790 
4791 /*
4792  * This function performs the Renew RPC.
4793  */
4794 int
4795 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4796     NFSPROC_T *p)
4797 {
4798 	u_int32_t *tl;
4799 	struct nfsrv_descript nfsd;
4800 	struct nfsrv_descript *nd = &nfsd;
4801 	struct nfsmount *nmp;
4802 	int error;
4803 	struct nfssockreq *nrp;
4804 	struct nfsclsession *tsep;
4805 
4806 	nmp = clp->nfsc_nmp;
4807 	if (nmp == NULL)
4808 		return (0);
4809 	if (dsp == NULL)
4810 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4811 		    0);
4812 	else
4813 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4814 		    &dsp->nfsclds_sess, 0, 0);
4815 	if (!NFSHASNFSV4N(nmp)) {
4816 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4817 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4818 		tsep = nfsmnt_mdssession(nmp);
4819 		*tl++ = tsep->nfsess_clientid.lval[0];
4820 		*tl = tsep->nfsess_clientid.lval[1];
4821 	}
4822 	nrp = NULL;
4823 	if (dsp != NULL)
4824 		nrp = dsp->nfsclds_sockp;
4825 	if (nrp == NULL)
4826 		/* If NULL, use the MDS socket. */
4827 		nrp = &nmp->nm_sockreq;
4828 	nd->nd_flag |= ND_USEGSSNAME;
4829 	if (dsp == NULL)
4830 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4831 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4832 	else {
4833 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4834 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4835 		if (error == ENXIO)
4836 			nfscl_cancelreqs(dsp);
4837 	}
4838 	if (error)
4839 		return (error);
4840 	error = nd->nd_repstat;
4841 	m_freem(nd->nd_mrep);
4842 	return (error);
4843 }
4844 
4845 /*
4846  * This function performs the Releaselockowner RPC.
4847  */
4848 int
4849 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4850     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4851 {
4852 	struct nfsrv_descript nfsd, *nd = &nfsd;
4853 	u_int32_t *tl;
4854 	int error;
4855 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4856 	struct nfsclsession *tsep;
4857 
4858 	if (NFSHASNFSV4N(nmp)) {
4859 		/* For NFSv4.1, do a FreeStateID. */
4860 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4861 		    NULL, 0, 0);
4862 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4863 	} else {
4864 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4865 		    NULL, 0, 0);
4866 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4867 		tsep = nfsmnt_mdssession(nmp);
4868 		*tl++ = tsep->nfsess_clientid.lval[0];
4869 		*tl = tsep->nfsess_clientid.lval[1];
4870 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4871 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4872 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4873 	}
4874 	nd->nd_flag |= ND_USEGSSNAME;
4875 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4876 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4877 	if (error)
4878 		return (error);
4879 	error = nd->nd_repstat;
4880 	m_freem(nd->nd_mrep);
4881 	return (error);
4882 }
4883 
4884 /*
4885  * This function performs the Compound to get the mount pt FH.
4886  */
4887 int
4888 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4889     NFSPROC_T *p)
4890 {
4891 	u_int32_t *tl;
4892 	struct nfsrv_descript nfsd;
4893 	struct nfsrv_descript *nd = &nfsd;
4894 	u_char *cp, *cp2;
4895 	int error, cnt, len, setnil;
4896 	u_int32_t *opcntp;
4897 
4898 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4899 	    0);
4900 	cp = dirpath;
4901 	cnt = 0;
4902 	do {
4903 		setnil = 0;
4904 		while (*cp == '/')
4905 			cp++;
4906 		cp2 = cp;
4907 		while (*cp2 != '\0' && *cp2 != '/')
4908 			cp2++;
4909 		if (*cp2 == '/') {
4910 			setnil = 1;
4911 			*cp2 = '\0';
4912 		}
4913 		if (cp2 != cp) {
4914 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4915 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4916 			nfsm_strtom(nd, cp, strlen(cp));
4917 			cnt++;
4918 		}
4919 		if (setnil)
4920 			*cp2++ = '/';
4921 		cp = cp2;
4922 	} while (*cp != '\0');
4923 	if (NFSHASNFSV4N(nmp))
4924 		/* Has a Sequence Op done by nfscl_reqstart(). */
4925 		*opcntp = txdr_unsigned(3 + cnt);
4926 	else
4927 		*opcntp = txdr_unsigned(2 + cnt);
4928 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4929 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4930 	nd->nd_flag |= ND_USEGSSNAME;
4931 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4932 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4933 	if (error)
4934 		return (error);
4935 	if (nd->nd_repstat == 0) {
4936 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4937 		tl += (2 + 2 * cnt);
4938 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4939 			len > NFSX_FHMAX) {
4940 			nd->nd_repstat = NFSERR_BADXDR;
4941 		} else {
4942 			nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4943 			if (nd->nd_repstat == 0)
4944 				nmp->nm_fhsize = len;
4945 		}
4946 	}
4947 	error = nd->nd_repstat;
4948 nfsmout:
4949 	m_freem(nd->nd_mrep);
4950 	return (error);
4951 }
4952 
4953 /*
4954  * This function performs the Delegreturn RPC.
4955  */
4956 int
4957 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4958     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4959 {
4960 	u_int32_t *tl;
4961 	struct nfsrv_descript nfsd;
4962 	struct nfsrv_descript *nd = &nfsd;
4963 	int error;
4964 
4965 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4966 	    dp->nfsdl_fhlen, NULL, NULL, 0, 0);
4967 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4968 	if (NFSHASNFSV4N(nmp))
4969 		*tl++ = 0;
4970 	else
4971 		*tl++ = dp->nfsdl_stateid.seqid;
4972 	*tl++ = dp->nfsdl_stateid.other[0];
4973 	*tl++ = dp->nfsdl_stateid.other[1];
4974 	*tl = dp->nfsdl_stateid.other[2];
4975 	if (syscred)
4976 		nd->nd_flag |= ND_USEGSSNAME;
4977 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4978 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4979 	if (error)
4980 		return (error);
4981 	error = nd->nd_repstat;
4982 	m_freem(nd->nd_mrep);
4983 	return (error);
4984 }
4985 
4986 /*
4987  * nfs getacl call.
4988  */
4989 int
4990 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p, struct acl *aclp)
4991 {
4992 	struct nfsrv_descript nfsd, *nd = &nfsd;
4993 	int error;
4994 	nfsattrbit_t attrbits;
4995 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4996 
4997 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4998 		return (EOPNOTSUPP);
4999 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
5000 	NFSZERO_ATTRBIT(&attrbits);
5001 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
5002 	(void) nfsrv_putattrbit(nd, &attrbits);
5003 	error = nfscl_request(nd, vp, p, cred);
5004 	if (error)
5005 		return (error);
5006 	if (!nd->nd_repstat)
5007 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
5008 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
5009 	else
5010 		error = nd->nd_repstat;
5011 	m_freem(nd->nd_mrep);
5012 	return (error);
5013 }
5014 
5015 /*
5016  * nfs setacl call.
5017  */
5018 int
5019 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p, struct acl *aclp)
5020 {
5021 	int error;
5022 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5023 
5024 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
5025 		return (EOPNOTSUPP);
5026 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL);
5027 	return (error);
5028 }
5029 
5030 /*
5031  * nfs setacl call.
5032  */
5033 static int
5034 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
5035     struct acl *aclp, nfsv4stateid_t *stateidp)
5036 {
5037 	struct nfsrv_descript nfsd, *nd = &nfsd;
5038 	int error;
5039 	nfsattrbit_t attrbits;
5040 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5041 
5042 	if (!NFSHASNFSV4(nmp))
5043 		return (EOPNOTSUPP);
5044 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
5045 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
5046 	NFSZERO_ATTRBIT(&attrbits);
5047 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
5048 	(void) nfsv4_fillattr(nd, vp->v_mount, vp, aclp, NULL, NULL, 0,
5049 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
5050 	error = nfscl_request(nd, vp, p, cred);
5051 	if (error)
5052 		return (error);
5053 	/* Don't care about the pre/postop attributes */
5054 	m_freem(nd->nd_mrep);
5055 	return (nd->nd_repstat);
5056 }
5057 
5058 /*
5059  * Do the NFSv4.1 Exchange ID.
5060  */
5061 int
5062 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
5063     struct nfssockreq *nrp, int minorvers, uint32_t exchflags,
5064     struct nfsclds **dspp, struct ucred *cred, NFSPROC_T *p)
5065 {
5066 	uint32_t *tl, v41flags;
5067 	struct nfsrv_descript nfsd;
5068 	struct nfsrv_descript *nd = &nfsd;
5069 	struct nfsclds *dsp;
5070 	struct timespec verstime;
5071 	int error, len;
5072 
5073 	*dspp = NULL;
5074 	if (minorvers == 0)
5075 		minorvers = nmp->nm_minorvers;
5076 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL,
5077 	    NFS_VER4, minorvers);
5078 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5079 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
5080 	*tl = txdr_unsigned(clp->nfsc_rev);
5081 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
5082 
5083 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
5084 	*tl++ = txdr_unsigned(exchflags);
5085 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
5086 
5087 	/* Set the implementation id4 */
5088 	*tl = txdr_unsigned(1);
5089 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
5090 	(void) nfsm_strtom(nd, version, strlen(version));
5091 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
5092 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
5093 	verstime.tv_nsec = 0;
5094 	txdr_nfsv4time(&verstime, tl);
5095 	nd->nd_flag |= ND_USEGSSNAME;
5096 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
5097 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5098 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
5099 	    (int)nd->nd_repstat);
5100 	if (error != 0)
5101 		return (error);
5102 	if (nd->nd_repstat == 0) {
5103 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
5104 		len = fxdr_unsigned(int, *(tl + 7));
5105 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
5106 			error = NFSERR_BADXDR;
5107 			goto nfsmout;
5108 		}
5109 		dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
5110 		    M_WAITOK | M_ZERO);
5111 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
5112 		dsp->nfsclds_servownlen = len;
5113 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
5114 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
5115 		dsp->nfsclds_sess.nfsess_sequenceid =
5116 		    fxdr_unsigned(uint32_t, *tl++);
5117 		v41flags = fxdr_unsigned(uint32_t, *tl);
5118 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
5119 		    NFSHASPNFSOPT(nmp)) {
5120 			NFSCL_DEBUG(1, "set PNFS\n");
5121 			NFSLOCKMNT(nmp);
5122 			nmp->nm_state |= NFSSTA_PNFS;
5123 			NFSUNLOCKMNT(nmp);
5124 			dsp->nfsclds_flags |= NFSCLDS_MDS;
5125 		}
5126 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
5127 			dsp->nfsclds_flags |= NFSCLDS_DS;
5128 		if (minorvers == NFSV42_MINORVERSION)
5129 			dsp->nfsclds_flags |= NFSCLDS_MINORV2;
5130 		if (len > 0)
5131 			nd->nd_repstat = nfsrv_mtostr(nd,
5132 			    dsp->nfsclds_serverown, len);
5133 		if (nd->nd_repstat == 0) {
5134 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5135 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5136 			    NULL, MTX_DEF);
5137 			nfscl_initsessionslots(&dsp->nfsclds_sess);
5138 			*dspp = dsp;
5139 		} else
5140 			free(dsp, M_NFSCLDS);
5141 	}
5142 	error = nd->nd_repstat;
5143 nfsmout:
5144 	m_freem(nd->nd_mrep);
5145 	return (error);
5146 }
5147 
5148 /*
5149  * Do the NFSv4.1 Create Session.
5150  */
5151 int
5152 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
5153     struct nfssockreq *nrp, struct nfsclds *dsp, uint32_t sequenceid, int mds,
5154     struct ucred *cred, NFSPROC_T *p)
5155 {
5156 	uint32_t crflags, maxval, *tl;
5157 	struct nfsrv_descript nfsd;
5158 	struct nfsrv_descript *nd = &nfsd;
5159 	int error, irdcnt, minorvers;
5160 
5161 	/* Make sure nm_rsize, nm_wsize is set. */
5162 	if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
5163 		nmp->nm_rsize = NFS_MAXBSIZE;
5164 	if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
5165 		nmp->nm_wsize = NFS_MAXBSIZE;
5166 	if (dsp == NULL)
5167 		minorvers = nmp->nm_minorvers;
5168 	else if ((dsp->nfsclds_flags & NFSCLDS_MINORV2) != 0)
5169 		minorvers = NFSV42_MINORVERSION;
5170 	else
5171 		minorvers = NFSV41_MINORVERSION;
5172 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL,
5173 	    NFS_VER4, minorvers);
5174 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5175 	*tl++ = sep->nfsess_clientid.lval[0];
5176 	*tl++ = sep->nfsess_clientid.lval[1];
5177 	*tl++ = txdr_unsigned(sequenceid);
5178 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
5179 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
5180 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
5181 	*tl = txdr_unsigned(crflags);
5182 
5183 	/* Fill in fore channel attributes. */
5184 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5185 	*tl++ = 0;				/* Header pad size */
5186 	if ((nd->nd_flag & ND_NFSV42) != 0 && mds != 0 && sb_max_adj >=
5187 	    nmp->nm_wsize && sb_max_adj >= nmp->nm_rsize) {
5188 		/*
5189 		 * NFSv4.2 Extended Attribute operations may want to do
5190 		 * requests/replies that are larger than nm_rsize/nm_wsize.
5191 		 */
5192 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
5193 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
5194 	} else {
5195 		*tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);
5196 		*tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);
5197 	}
5198 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
5199 	*tl++ = txdr_unsigned(20);		/* Max operations */
5200 	*tl++ = txdr_unsigned(64);		/* Max slots */
5201 	*tl = 0;				/* No rdma ird */
5202 
5203 	/* Fill in back channel attributes. */
5204 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5205 	*tl++ = 0;				/* Header pad size */
5206 	*tl++ = txdr_unsigned(10000);		/* Max request size */
5207 	*tl++ = txdr_unsigned(10000);		/* Max response size */
5208 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
5209 	*tl++ = txdr_unsigned(4);		/* Max operations */
5210 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
5211 	*tl = 0;				/* No rdma ird */
5212 
5213 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
5214 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
5215 
5216 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
5217 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
5218 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
5219 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
5220 	*tl++ = 0;				/* Null machine name */
5221 	*tl++ = 0;				/* Uid == 0 */
5222 	*tl++ = 0;				/* Gid == 0 */
5223 	*tl = 0;				/* No additional gids */
5224 	nd->nd_flag |= ND_USEGSSNAME;
5225 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
5226 	    NFS_VER4, NULL, 1, NULL, NULL);
5227 	if (error != 0)
5228 		return (error);
5229 	if (nd->nd_repstat == 0) {
5230 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
5231 		    2 * NFSX_UNSIGNED);
5232 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
5233 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
5234 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
5235 		crflags = fxdr_unsigned(uint32_t, *tl);
5236 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
5237 			NFSLOCKMNT(nmp);
5238 			nmp->nm_state |= NFSSTA_SESSPERSIST;
5239 			NFSUNLOCKMNT(nmp);
5240 		}
5241 
5242 		/* Get the fore channel slot count. */
5243 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5244 		tl++;			/* Skip the header pad size. */
5245 
5246 		/* Make sure nm_wsize is small enough. */
5247 		maxval = fxdr_unsigned(uint32_t, *tl++);
5248 		while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
5249 			if (nmp->nm_wsize > 8096)
5250 				nmp->nm_wsize /= 2;
5251 			else
5252 				break;
5253 		}
5254 		sep->nfsess_maxreq = maxval;
5255 
5256 		/* Make sure nm_rsize is small enough. */
5257 		maxval = fxdr_unsigned(uint32_t, *tl++);
5258 		while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
5259 			if (nmp->nm_rsize > 8096)
5260 				nmp->nm_rsize /= 2;
5261 			else
5262 				break;
5263 		}
5264 		sep->nfsess_maxresp = maxval;
5265 
5266 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
5267 		tl++;
5268 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
5269 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
5270 		irdcnt = fxdr_unsigned(int, *tl);
5271 		if (irdcnt < 0 || irdcnt > 1) {
5272 			error = NFSERR_BADXDR;
5273 			goto nfsmout;
5274 		}
5275 		if (irdcnt > 0)
5276 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
5277 
5278 		/* and the back channel slot count. */
5279 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5280 		tl += 5;
5281 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
5282 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
5283 	}
5284 	error = nd->nd_repstat;
5285 nfsmout:
5286 	m_freem(nd->nd_mrep);
5287 	return (error);
5288 }
5289 
5290 /*
5291  * Do the NFSv4.1 Destroy Session.
5292  */
5293 int
5294 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
5295     struct ucred *cred, NFSPROC_T *p)
5296 {
5297 	uint32_t *tl;
5298 	struct nfsrv_descript nfsd;
5299 	struct nfsrv_descript *nd = &nfsd;
5300 	int error;
5301 	struct nfsclsession *tsep;
5302 
5303 	nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL, 0,
5304 	    0);
5305 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
5306 	tsep = nfsmnt_mdssession(nmp);
5307 	bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
5308 	nd->nd_flag |= ND_USEGSSNAME;
5309 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5310 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5311 	if (error != 0)
5312 		return (error);
5313 	error = nd->nd_repstat;
5314 	m_freem(nd->nd_mrep);
5315 	return (error);
5316 }
5317 
5318 /*
5319  * Do the NFSv4.1 Destroy Client.
5320  */
5321 int
5322 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
5323     struct ucred *cred, NFSPROC_T *p)
5324 {
5325 	uint32_t *tl;
5326 	struct nfsrv_descript nfsd;
5327 	struct nfsrv_descript *nd = &nfsd;
5328 	int error;
5329 	struct nfsclsession *tsep;
5330 
5331 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
5332 	    0);
5333 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5334 	tsep = nfsmnt_mdssession(nmp);
5335 	*tl++ = tsep->nfsess_clientid.lval[0];
5336 	*tl = tsep->nfsess_clientid.lval[1];
5337 	nd->nd_flag |= ND_USEGSSNAME;
5338 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5339 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5340 	if (error != 0)
5341 		return (error);
5342 	error = nd->nd_repstat;
5343 	m_freem(nd->nd_mrep);
5344 	return (error);
5345 }
5346 
5347 /*
5348  * Do the NFSv4.1 LayoutGet.
5349  */
5350 static int
5351 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
5352     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
5353     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
5354     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p)
5355 {
5356 	struct nfsrv_descript nfsd, *nd = &nfsd;
5357 	int error;
5358 
5359 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
5360 	    0);
5361 	nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
5362 	    layouttype, layoutlen, 0);
5363 	nd->nd_flag |= ND_USEGSSNAME;
5364 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5365 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5366 	NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
5367 	if (error != 0)
5368 		return (error);
5369 	if (nd->nd_repstat == 0)
5370 		error = nfsrv_parselayoutget(nmp, nd, stateidp, retonclosep,
5371 		    flhp);
5372 	if (error == 0 && nd->nd_repstat != 0)
5373 		error = nd->nd_repstat;
5374 	m_freem(nd->nd_mrep);
5375 	return (error);
5376 }
5377 
5378 /*
5379  * Do the NFSv4.1 Get Device Info.
5380  */
5381 int
5382 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
5383     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
5384     NFSPROC_T *p)
5385 {
5386 	uint32_t cnt, *tl, vers, minorvers;
5387 	struct nfsrv_descript nfsd;
5388 	struct nfsrv_descript *nd = &nfsd;
5389 	struct sockaddr_in sin, ssin;
5390 	struct sockaddr_in6 sin6, ssin6;
5391 	struct nfsclds *dsp = NULL, **dspp, **gotdspp;
5392 	struct nfscldevinfo *ndi;
5393 	int addrcnt = 0, bitcnt, error, gotminor, gotvers, i, isudp, j;
5394 	int stripecnt;
5395 	uint8_t stripeindex;
5396 	sa_family_t af, safilled;
5397 
5398 	ssin.sin_port = 0;		/* To shut up compiler. */
5399 	ssin.sin_addr.s_addr = 0;	/* ditto */
5400 	*ndip = NULL;
5401 	ndi = NULL;
5402 	gotdspp = NULL;
5403 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
5404 	    0);
5405 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5406 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
5407 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5408 	*tl++ = txdr_unsigned(layouttype);
5409 	*tl++ = txdr_unsigned(100000);
5410 	if (notifybitsp != NULL && *notifybitsp != 0) {
5411 		*tl = txdr_unsigned(1);		/* One word of bits. */
5412 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5413 		*tl = txdr_unsigned(*notifybitsp);
5414 	} else
5415 		*tl = txdr_unsigned(0);
5416 	nd->nd_flag |= ND_USEGSSNAME;
5417 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5418 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5419 	if (error != 0)
5420 		return (error);
5421 	if (nd->nd_repstat == 0) {
5422 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5423 		if (layouttype != fxdr_unsigned(int, *tl))
5424 			printf("EEK! devinfo layout type not same!\n");
5425 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
5426 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5427 			stripecnt = fxdr_unsigned(int, *tl);
5428 			NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
5429 			if (stripecnt < 1 || stripecnt > 4096) {
5430 				printf("pNFS File layout devinfo stripecnt %d:"
5431 				    " out of range\n", stripecnt);
5432 				error = NFSERR_BADXDR;
5433 				goto nfsmout;
5434 			}
5435 			NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
5436 			    NFSX_UNSIGNED);
5437 			addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
5438 			NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
5439 			if (addrcnt < 1 || addrcnt > 128) {
5440 				printf("NFS devinfo addrcnt %d: out of range\n",
5441 				    addrcnt);
5442 				error = NFSERR_BADXDR;
5443 				goto nfsmout;
5444 			}
5445 
5446 			/*
5447 			 * Now we know how many stripe indices and addresses, so
5448 			 * we can allocate the structure the correct size.
5449 			 */
5450 			i = (stripecnt * sizeof(uint8_t)) /
5451 			    sizeof(struct nfsclds *) + 1;
5452 			NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5453 			ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5454 			    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5455 			    M_ZERO);
5456 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5457 			    NFSX_V4DEVICEID);
5458 			ndi->nfsdi_refcnt = 0;
5459 			ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5460 			ndi->nfsdi_stripecnt = stripecnt;
5461 			ndi->nfsdi_addrcnt = addrcnt;
5462 			/* Fill in the stripe indices. */
5463 			for (i = 0; i < stripecnt; i++) {
5464 				stripeindex = fxdr_unsigned(uint8_t, *tl++);
5465 				NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5466 				if (stripeindex >= addrcnt) {
5467 					printf("pNFS File Layout devinfo"
5468 					    " stripeindex %d: too big\n",
5469 					    (int)stripeindex);
5470 					error = NFSERR_BADXDR;
5471 					goto nfsmout;
5472 				}
5473 				nfsfldi_setstripeindex(ndi, i, stripeindex);
5474 			}
5475 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
5476 			/* For Flex File, we only get one address list. */
5477 			ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5478 			    M_NFSDEVINFO, M_WAITOK | M_ZERO);
5479 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5480 			    NFSX_V4DEVICEID);
5481 			ndi->nfsdi_refcnt = 0;
5482 			ndi->nfsdi_flags = NFSDI_FLEXFILE;
5483 			addrcnt = ndi->nfsdi_addrcnt = 1;
5484 		}
5485 
5486 		/* Now, dissect the server address(es). */
5487 		safilled = AF_UNSPEC;
5488 		for (i = 0; i < addrcnt; i++) {
5489 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5490 			cnt = fxdr_unsigned(uint32_t, *tl);
5491 			if (cnt == 0) {
5492 				printf("NFS devinfo 0 len addrlist\n");
5493 				error = NFSERR_BADXDR;
5494 				goto nfsmout;
5495 			}
5496 			dspp = nfsfldi_addr(ndi, i);
5497 			safilled = AF_UNSPEC;
5498 			for (j = 0; j < cnt; j++) {
5499 				error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5500 				    &isudp);
5501 				if (error != 0 && error != EPERM) {
5502 					error = NFSERR_BADXDR;
5503 					goto nfsmout;
5504 				}
5505 				if (error == 0 && isudp == 0) {
5506 					/*
5507 					 * The priority is:
5508 					 * - Same address family.
5509 					 * Save the address and dspp, so that
5510 					 * the connection can be done after
5511 					 * parsing is complete.
5512 					 */
5513 					if (safilled == AF_UNSPEC ||
5514 					    (af == nmp->nm_nam->sa_family &&
5515 					     safilled != nmp->nm_nam->sa_family)
5516 					   ) {
5517 						if (af == AF_INET)
5518 							ssin = sin;
5519 						else
5520 							ssin6 = sin6;
5521 						safilled = af;
5522 						gotdspp = dspp;
5523 					}
5524 				}
5525 			}
5526 		}
5527 
5528 		gotvers = NFS_VER4;	/* Default NFSv4.1 for File Layout. */
5529 		gotminor = NFSV41_MINORVERSION;
5530 		/* For Flex File, we will take one of the versions to use. */
5531 		if (layouttype == NFSLAYOUT_FLEXFILE) {
5532 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5533 			j = fxdr_unsigned(int, *tl);
5534 			if (j < 1 || j > NFSDEV_MAXVERS) {
5535 				printf("pNFS: too many versions\n");
5536 				error = NFSERR_BADXDR;
5537 				goto nfsmout;
5538 			}
5539 			gotvers = 0;
5540 			gotminor = 0;
5541 			for (i = 0; i < j; i++) {
5542 				NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5543 				vers = fxdr_unsigned(uint32_t, *tl++);
5544 				minorvers = fxdr_unsigned(uint32_t, *tl++);
5545 				if (vers == NFS_VER3)
5546 					minorvers = 0;
5547 				if ((vers == NFS_VER4 && ((minorvers ==
5548 				    NFSV41_MINORVERSION && gotminor == 0) ||
5549 				    minorvers == NFSV42_MINORVERSION)) ||
5550 				    (vers == NFS_VER3 && gotvers == 0)) {
5551 					gotvers = vers;
5552 					gotminor = minorvers;
5553 					/* We'll take this one. */
5554 					ndi->nfsdi_versindex = i;
5555 					ndi->nfsdi_vers = vers;
5556 					ndi->nfsdi_minorvers = minorvers;
5557 					ndi->nfsdi_rsize = fxdr_unsigned(
5558 					    uint32_t, *tl++);
5559 					ndi->nfsdi_wsize = fxdr_unsigned(
5560 					    uint32_t, *tl++);
5561 					if (*tl == newnfs_true)
5562 						ndi->nfsdi_flags |=
5563 						    NFSDI_TIGHTCOUPLED;
5564 					else
5565 						ndi->nfsdi_flags &=
5566 						    ~NFSDI_TIGHTCOUPLED;
5567 				}
5568 			}
5569 			if (gotvers == 0) {
5570 				printf("pNFS: no NFSv3, NFSv4.1 or NFSv4.2\n");
5571 				error = NFSERR_BADXDR;
5572 				goto nfsmout;
5573 			}
5574 		}
5575 
5576 		/* And the notify bits. */
5577 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5578 		bitcnt = fxdr_unsigned(int, *tl);
5579 		if (bitcnt > 0) {
5580 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5581 			if (notifybitsp != NULL)
5582 				*notifybitsp =
5583 				    fxdr_unsigned(uint32_t, *tl);
5584 		}
5585 		if (safilled != AF_UNSPEC) {
5586 			KASSERT(ndi != NULL, ("ndi is NULL"));
5587 			*ndip = ndi;
5588 		} else
5589 			error = EPERM;
5590 		if (error == 0) {
5591 			/*
5592 			 * Now we can do a TCP connection for the correct
5593 			 * NFS version and IP address.
5594 			 */
5595 			error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5596 			    gotvers, gotminor, &dsp, p);
5597 		}
5598 		if (error == 0) {
5599 			KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5600 			*gotdspp = dsp;
5601 		}
5602 	}
5603 	if (nd->nd_repstat != 0 && error == 0)
5604 		error = nd->nd_repstat;
5605 nfsmout:
5606 	if (error != 0 && ndi != NULL)
5607 		nfscl_freedevinfo(ndi);
5608 	m_freem(nd->nd_mrep);
5609 	return (error);
5610 }
5611 
5612 /*
5613  * Do the NFSv4.1 LayoutCommit.
5614  */
5615 int
5616 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5617     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5618     int layouttype, struct ucred *cred, NFSPROC_T *p)
5619 {
5620 	uint32_t *tl;
5621 	struct nfsrv_descript nfsd, *nd = &nfsd;
5622 	int error;
5623 
5624 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5625 	    0, 0);
5626 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5627 	    NFSX_STATEID);
5628 	txdr_hyper(off, tl);
5629 	tl += 2;
5630 	txdr_hyper(len, tl);
5631 	tl += 2;
5632 	if (reclaim != 0)
5633 		*tl++ = newnfs_true;
5634 	else
5635 		*tl++ = newnfs_false;
5636 	*tl++ = txdr_unsigned(stateidp->seqid);
5637 	*tl++ = stateidp->other[0];
5638 	*tl++ = stateidp->other[1];
5639 	*tl++ = stateidp->other[2];
5640 	*tl++ = newnfs_true;
5641 	if (lastbyte < off)
5642 		lastbyte = off;
5643 	else if (lastbyte >= (off + len))
5644 		lastbyte = off + len - 1;
5645 	txdr_hyper(lastbyte, tl);
5646 	tl += 2;
5647 	*tl++ = newnfs_false;
5648 	*tl++ = txdr_unsigned(layouttype);
5649 	/* All supported layouts are 0 length. */
5650 	*tl = txdr_unsigned(0);
5651 	nd->nd_flag |= ND_USEGSSNAME;
5652 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5653 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5654 	if (error != 0)
5655 		return (error);
5656 	error = nd->nd_repstat;
5657 	m_freem(nd->nd_mrep);
5658 	return (error);
5659 }
5660 
5661 /*
5662  * Do the NFSv4.1 LayoutReturn.
5663  */
5664 int
5665 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5666     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5667     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5668     uint32_t stat, uint32_t op, char *devid)
5669 {
5670 	uint32_t *tl;
5671 	struct nfsrv_descript nfsd, *nd = &nfsd;
5672 	uint64_t tu64;
5673 	int error;
5674 
5675 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5676 	    0, 0);
5677 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5678 	if (reclaim != 0)
5679 		*tl++ = newnfs_true;
5680 	else
5681 		*tl++ = newnfs_false;
5682 	*tl++ = txdr_unsigned(layouttype);
5683 	*tl++ = txdr_unsigned(iomode);
5684 	*tl = txdr_unsigned(layoutreturn);
5685 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5686 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5687 		    NFSX_UNSIGNED);
5688 		txdr_hyper(offset, tl);
5689 		tl += 2;
5690 		txdr_hyper(len, tl);
5691 		tl += 2;
5692 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5693 		*tl++ = txdr_unsigned(stateidp->seqid);
5694 		*tl++ = stateidp->other[0];
5695 		*tl++ = stateidp->other[1];
5696 		*tl++ = stateidp->other[2];
5697 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5698 			*tl = txdr_unsigned(0);
5699 		else if (layouttype == NFSLAYOUT_FLEXFILE) {
5700 			if (stat != 0) {
5701 				*tl = txdr_unsigned(2 * NFSX_HYPER +
5702 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5703 				    NFSX_UNSIGNED);
5704 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5705 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5706 				    NFSX_UNSIGNED);
5707 				*tl++ = txdr_unsigned(1);	/* One error. */
5708 				tu64 = 0;			/* Offset. */
5709 				txdr_hyper(tu64, tl); tl += 2;
5710 				tu64 = UINT64_MAX;		/* Length. */
5711 				txdr_hyper(tu64, tl); tl += 2;
5712 				NFSBCOPY(stateidp, tl, NFSX_STATEID);
5713 				tl += (NFSX_STATEID / NFSX_UNSIGNED);
5714 				*tl++ = txdr_unsigned(1);	/* One error. */
5715 				NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5716 				tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5717 				*tl++ = txdr_unsigned(stat);
5718 				*tl++ = txdr_unsigned(op);
5719 			} else {
5720 				*tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5721 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5722 				/* No ioerrs. */
5723 				*tl++ = 0;
5724 			}
5725 			*tl = 0;	/* No stats yet. */
5726 		}
5727 	}
5728 	nd->nd_flag |= ND_USEGSSNAME;
5729 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5730 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5731 	if (error != 0)
5732 		return (error);
5733 	if (nd->nd_repstat == 0) {
5734 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5735 		if (*tl != 0) {
5736 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5737 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5738 			stateidp->other[0] = *tl++;
5739 			stateidp->other[1] = *tl++;
5740 			stateidp->other[2] = *tl;
5741 		}
5742 	} else
5743 		error = nd->nd_repstat;
5744 nfsmout:
5745 	m_freem(nd->nd_mrep);
5746 	return (error);
5747 }
5748 
5749 /*
5750  * Do the NFSv4.2 LayoutError.
5751  */
5752 static int
5753 nfsrpc_layouterror(struct nfsmount *nmp, uint8_t *fh, int fhlen, uint64_t offset,
5754     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5755     uint32_t stat, uint32_t op, char *devid)
5756 {
5757 	uint32_t *tl;
5758 	struct nfsrv_descript nfsd, *nd = &nfsd;
5759 	int error;
5760 
5761 	nfscl_reqstart(nd, NFSPROC_LAYOUTERROR, nmp, fh, fhlen, NULL, NULL,
5762 	    0, 0);
5763 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5764 	    NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5765 	txdr_hyper(offset, tl); tl += 2;
5766 	txdr_hyper(len, tl); tl += 2;
5767 	*tl++ = txdr_unsigned(stateidp->seqid);
5768 	*tl++ = stateidp->other[0];
5769 	*tl++ = stateidp->other[1];
5770 	*tl++ = stateidp->other[2];
5771 	*tl++ = txdr_unsigned(1);
5772 	NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5773 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5774 	*tl++ = txdr_unsigned(stat);
5775 	*tl = txdr_unsigned(op);
5776 	nd->nd_flag |= ND_USEGSSNAME;
5777 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5778 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5779 	if (error != 0)
5780 		return (error);
5781 	if (nd->nd_repstat != 0)
5782 		error = nd->nd_repstat;
5783 	m_freem(nd->nd_mrep);
5784 	return (error);
5785 }
5786 
5787 /*
5788  * Acquire a layout and devinfo, if possible. The caller must have acquired
5789  * a reference count on the nfsclclient structure before calling this.
5790  * Return the layout in lypp with a reference count on it, if successful.
5791  */
5792 static int
5793 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5794     int iomode, uint32_t rw, uint32_t *notifybitsp, nfsv4stateid_t *stateidp,
5795     uint64_t off, struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5796 {
5797 	struct nfscllayout *lyp;
5798 	struct nfsclflayout *flp;
5799 	struct nfsclflayouthead flh;
5800 	int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5801 	nfsv4stateid_t stateid;
5802 	struct nfsclsession *tsep;
5803 
5804 	*lypp = NULL;
5805 	if (NFSHASFLEXFILE(nmp))
5806 		layouttype = NFSLAYOUT_FLEXFILE;
5807 	else
5808 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
5809 	/*
5810 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5811 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5812 	 * flp == NULL.
5813 	 */
5814 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5815 	    off, rw, &flp, &recalled);
5816 	islocked = 0;
5817 	if (lyp == NULL || flp == NULL) {
5818 		if (recalled != 0)
5819 			return (EIO);
5820 		LIST_INIT(&flh);
5821 		tsep = nfsmnt_mdssession(nmp);
5822 		layoutlen = tsep->nfsess_maxcache -
5823 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5824 		if (lyp == NULL) {
5825 			stateid.seqid = 0;
5826 			stateid.other[0] = stateidp->other[0];
5827 			stateid.other[1] = stateidp->other[1];
5828 			stateid.other[2] = stateidp->other[2];
5829 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5830 			    nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5831 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5832 			    &retonclose, &flh, cred, p);
5833 		} else {
5834 			islocked = 1;
5835 			stateid.seqid = lyp->nfsly_stateid.seqid;
5836 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5837 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5838 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5839 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5840 			    nfhp->nfh_len, iomode, off, UINT64_MAX,
5841 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5842 			    &retonclose, &flh, cred, p);
5843 		}
5844 		error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5845 		    nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5846 		    &flh, layouttype, error, NULL, cred, p);
5847 		if (error == 0)
5848 			*lypp = lyp;
5849 		else if (islocked != 0)
5850 			nfscl_rellayout(lyp, 1);
5851 	} else
5852 		*lypp = lyp;
5853 	return (error);
5854 }
5855 
5856 /*
5857  * Do a TCP connection plus exchange id and create session.
5858  * If successful, a "struct nfsclds" is linked into the list for the
5859  * mount point and a pointer to it is returned.
5860  */
5861 static int
5862 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5863     struct sockaddr_in6 *sin6, sa_family_t af, int vers, int minorvers,
5864     struct nfsclds **dspp, NFSPROC_T *p)
5865 {
5866 	struct sockaddr_in *msad, *sad;
5867 	struct sockaddr_in6 *msad6, *sad6;
5868 	struct nfsclclient *clp;
5869 	struct nfssockreq *nrp;
5870 	struct nfsclds *dsp, *tdsp;
5871 	int error, firsttry;
5872 	enum nfsclds_state retv;
5873 	uint32_t sequenceid = 0;
5874 
5875 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5876 	    ("nfsrpc_fillsa: NULL nr_cred"));
5877 	NFSLOCKCLSTATE();
5878 	clp = nmp->nm_clp;
5879 	NFSUNLOCKCLSTATE();
5880 	if (clp == NULL)
5881 		return (EPERM);
5882 	if (af == AF_INET) {
5883 		NFSLOCKMNT(nmp);
5884 		/*
5885 		 * Check to see if we already have a session for this
5886 		 * address that is usable for a DS.
5887 		 * Note that the MDS's address is in a different place
5888 		 * than the sessions already acquired for DS's.
5889 		 */
5890 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5891 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5892 		while (tdsp != NULL) {
5893 			if (msad != NULL && msad->sin_family == AF_INET &&
5894 			    sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5895 			    sin->sin_port == msad->sin_port &&
5896 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5897 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5898 				*dspp = tdsp;
5899 				NFSUNLOCKMNT(nmp);
5900 				NFSCL_DEBUG(4, "fnd same addr\n");
5901 				return (0);
5902 			}
5903 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5904 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5905 				msad = (struct sockaddr_in *)
5906 				    tdsp->nfsclds_sockp->nr_nam;
5907 			else
5908 				msad = NULL;
5909 		}
5910 		NFSUNLOCKMNT(nmp);
5911 
5912 		/* No IP address match, so look for new/trunked one. */
5913 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5914 		sad->sin_len = sizeof(*sad);
5915 		sad->sin_family = AF_INET;
5916 		sad->sin_port = sin->sin_port;
5917 		sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5918 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5919 		nrp->nr_nam = (struct sockaddr *)sad;
5920 	} else if (af == AF_INET6) {
5921 		NFSLOCKMNT(nmp);
5922 		/*
5923 		 * Check to see if we already have a session for this
5924 		 * address that is usable for a DS.
5925 		 * Note that the MDS's address is in a different place
5926 		 * than the sessions already acquired for DS's.
5927 		 */
5928 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5929 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5930 		while (tdsp != NULL) {
5931 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5932 			    IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5933 			    &msad6->sin6_addr) &&
5934 			    sin6->sin6_port == msad6->sin6_port &&
5935 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5936 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5937 				*dspp = tdsp;
5938 				NFSUNLOCKMNT(nmp);
5939 				return (0);
5940 			}
5941 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5942 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5943 				msad6 = (struct sockaddr_in6 *)
5944 				    tdsp->nfsclds_sockp->nr_nam;
5945 			else
5946 				msad6 = NULL;
5947 		}
5948 		NFSUNLOCKMNT(nmp);
5949 
5950 		/* No IP address match, so look for new/trunked one. */
5951 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5952 		sad6->sin6_len = sizeof(*sad6);
5953 		sad6->sin6_family = AF_INET6;
5954 		sad6->sin6_port = sin6->sin6_port;
5955 		NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5956 		    sizeof(struct in6_addr));
5957 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5958 		nrp->nr_nam = (struct sockaddr *)sad6;
5959 	} else
5960 		return (EPERM);
5961 
5962 	nrp->nr_sotype = SOCK_STREAM;
5963 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5964 	nrp->nr_prog = NFS_PROG;
5965 	nrp->nr_vers = vers;
5966 
5967 	/*
5968 	 * Use the credentials that were used for the mount, which are
5969 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5970 	 * Ref. counting the credentials with crhold() is probably not
5971 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5972 	 * unmount, but I did it anyhow.
5973 	 */
5974 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5975 	error = newnfs_connect(nmp, nrp, NULL, p, 0, false, &nrp->nr_client);
5976 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
5977 
5978 	dsp = NULL;
5979 	/* Now, do the exchangeid and create session. */
5980 	if (error == 0) {
5981 		if (vers == NFS_VER4) {
5982 			firsttry = 0;
5983 			do {
5984 				error = nfsrpc_exchangeid(nmp, clp, nrp,
5985 				    minorvers, NFSV4EXCH_USEPNFSDS, &dsp,
5986 				    nrp->nr_cred, p);
5987 				NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5988 				if (error == NFSERR_MINORVERMISMATCH)
5989 					minorvers = NFSV42_MINORVERSION;
5990 			} while (error == NFSERR_MINORVERMISMATCH &&
5991 			    firsttry++ == 0);
5992 			if (error != 0)
5993 				newnfs_disconnect(NULL, nrp);
5994 		} else {
5995 			dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
5996 			    M_WAITOK | M_ZERO);
5997 			dsp->nfsclds_flags |= NFSCLDS_DS;
5998 			dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
5999 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
6000 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
6001 			    NULL, MTX_DEF);
6002 		}
6003 	}
6004 	if (error == 0) {
6005 		dsp->nfsclds_sockp = nrp;
6006 		if (vers == NFS_VER4) {
6007 			NFSLOCKMNT(nmp);
6008 			retv = nfscl_getsameserver(nmp, dsp, &tdsp,
6009 			    &sequenceid);
6010 			NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
6011 			if (retv == NFSDSP_USETHISSESSION &&
6012 			    nfscl_dssameconn != 0) {
6013 				NFSLOCKDS(tdsp);
6014 				tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
6015 				NFSUNLOCKDS(tdsp);
6016 				NFSUNLOCKMNT(nmp);
6017 				/*
6018 				 * If there is already a session for this
6019 				 * server, use it.
6020 				 */
6021 				newnfs_disconnect(NULL, nrp);
6022 				nfscl_freenfsclds(dsp);
6023 				*dspp = tdsp;
6024 				return (0);
6025 			}
6026 			if (retv == NFSDSP_NOTFOUND)
6027 				sequenceid =
6028 				    dsp->nfsclds_sess.nfsess_sequenceid;
6029 			NFSUNLOCKMNT(nmp);
6030 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
6031 			    nrp, dsp, sequenceid, 0, nrp->nr_cred, p);
6032 			NFSCL_DEBUG(3, "DS createsess=%d\n", error);
6033 		}
6034 	} else {
6035 		NFSFREECRED(nrp->nr_cred);
6036 		NFSFREEMUTEX(&nrp->nr_mtx);
6037 		free(nrp->nr_nam, M_SONAME);
6038 		free(nrp, M_NFSSOCKREQ);
6039 	}
6040 	if (error == 0) {
6041 		NFSCL_DEBUG(3, "add DS session\n");
6042 		/*
6043 		 * Put it at the end of the list. That way the list
6044 		 * is ordered by when the entry was added. This matters
6045 		 * since the one done first is the one that should be
6046 		 * used for sequencid'ing any subsequent create sessions.
6047 		 */
6048 		NFSLOCKMNT(nmp);
6049 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
6050 		NFSUNLOCKMNT(nmp);
6051 		*dspp = dsp;
6052 	} else if (dsp != NULL) {
6053 		newnfs_disconnect(NULL, nrp);
6054 		nfscl_freenfsclds(dsp);
6055 	}
6056 	return (error);
6057 }
6058 
6059 /*
6060  * Do the NFSv4.1 Reclaim Complete.
6061  */
6062 int
6063 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
6064 {
6065 	uint32_t *tl;
6066 	struct nfsrv_descript nfsd;
6067 	struct nfsrv_descript *nd = &nfsd;
6068 	int error;
6069 
6070 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
6071 	    0);
6072 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
6073 	*tl = newnfs_false;
6074 	nd->nd_flag |= ND_USEGSSNAME;
6075 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
6076 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
6077 	if (error != 0)
6078 		return (error);
6079 	error = nd->nd_repstat;
6080 	m_freem(nd->nd_mrep);
6081 	return (error);
6082 }
6083 
6084 /*
6085  * Initialize the slot tables for a session.
6086  */
6087 static void
6088 nfscl_initsessionslots(struct nfsclsession *sep)
6089 {
6090 	int i;
6091 
6092 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
6093 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
6094 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
6095 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
6096 	}
6097 	for (i = 0; i < 64; i++)
6098 		sep->nfsess_slotseq[i] = 0;
6099 	sep->nfsess_slots = 0;
6100 }
6101 
6102 /*
6103  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
6104  */
6105 int
6106 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6107     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
6108 {
6109 	struct nfsnode *np = VTONFS(vp);
6110 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6111 	struct nfscllayout *layp;
6112 	struct nfscldevinfo *dip;
6113 	struct nfsclflayout *rflp;
6114 	struct mbuf *m, *m2;
6115 	struct nfsclwritedsdorpc *drpc, *tdrpc;
6116 	nfsv4stateid_t stateid;
6117 	struct ucred *newcred;
6118 	uint64_t lastbyte, len, off, oresid, xfer;
6119 	int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
6120 	void *lckp;
6121 	uint8_t *dev;
6122 	void *iovbase = NULL;
6123 	size_t iovlen = 0;
6124 	off_t offs = 0;
6125 	ssize_t resid = 0;
6126 	uint32_t op;
6127 
6128 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
6129 	    (np->n_flag & NNOLAYOUT) != 0)
6130 		return (EIO);
6131 	/* Now, get a reference cnt on the clientid for this mount. */
6132 	if (nfscl_getref(nmp) == 0)
6133 		return (EIO);
6134 
6135 	/* Find an appropriate stateid. */
6136 	newcred = NFSNEWCRED(cred);
6137 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
6138 	    rwaccess, 1, newcred, p, &stateid, &lckp);
6139 	if (error != 0) {
6140 		NFSFREECRED(newcred);
6141 		nfscl_relref(nmp);
6142 		return (error);
6143 	}
6144 	/* Search for a layout for this file. */
6145 	off = uiop->uio_offset;
6146 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
6147 	    np->n_fhp->nfh_len, off, rwaccess, &rflp, &recalled);
6148 	if (layp == NULL || rflp == NULL) {
6149 		if (recalled != 0) {
6150 			NFSFREECRED(newcred);
6151 			if (lckp != NULL)
6152 				nfscl_lockderef(lckp);
6153 			nfscl_relref(nmp);
6154 			return (EIO);
6155 		}
6156 		if (layp != NULL) {
6157 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
6158 			layp = NULL;
6159 		}
6160 		/* Try and get a Layout, if it is supported. */
6161 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
6162 		    (np->n_flag & NWRITEOPENED) != 0)
6163 			iolaymode = NFSLAYOUTIOMODE_RW;
6164 		else
6165 			iolaymode = NFSLAYOUTIOMODE_READ;
6166 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
6167 		    rwaccess, NULL, &stateid, off, &layp, newcred, p);
6168 		if (error != 0) {
6169 			NFSLOCKNODE(np);
6170 			np->n_flag |= NNOLAYOUT;
6171 			NFSUNLOCKNODE(np);
6172 			if (lckp != NULL)
6173 				nfscl_lockderef(lckp);
6174 			NFSFREECRED(newcred);
6175 			if (layp != NULL)
6176 				nfscl_rellayout(layp, 0);
6177 			nfscl_relref(nmp);
6178 			return (error);
6179 		}
6180 	}
6181 
6182 	/*
6183 	 * Loop around finding a layout that works for the first part of
6184 	 * this I/O operation, and then call the function that actually
6185 	 * does the RPC.
6186 	 */
6187 	eof = 0;
6188 	len = (uint64_t)uiop->uio_resid;
6189 	while (len > 0 && error == 0 && eof == 0) {
6190 		off = uiop->uio_offset;
6191 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
6192 		if (error == 0) {
6193 			oresid = xfer = (uint64_t)uiop->uio_resid;
6194 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
6195 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
6196 			/*
6197 			 * For Flex File layout with mirrored DSs, select one
6198 			 * of them at random for reads. For writes and commits,
6199 			 * do all mirrors.
6200 			 */
6201 			m = NULL;
6202 			tdrpc = drpc = NULL;
6203 			firstmirror = 0;
6204 			mirrorcnt = 1;
6205 			if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
6206 			    (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
6207 				if (rwaccess == NFSV4OPEN_ACCESSREAD) {
6208 					firstmirror = arc4random() % mirrorcnt;
6209 					mirrorcnt = firstmirror + 1;
6210 				} else {
6211 					if (docommit == 0) {
6212 						/*
6213 						 * Save values, so uiop can be
6214 						 * rolled back upon a write
6215 						 * error.
6216 						 */
6217 						offs = uiop->uio_offset;
6218 						resid = uiop->uio_resid;
6219 						iovbase =
6220 						    uiop->uio_iov->iov_base;
6221 						iovlen = uiop->uio_iov->iov_len;
6222 						m = nfsm_uiombuflist(uiop, len,
6223 						    0);
6224 					}
6225 					tdrpc = drpc = malloc(sizeof(*drpc) *
6226 					    (mirrorcnt - 1), M_TEMP, M_WAITOK |
6227 					    M_ZERO);
6228 				}
6229 			}
6230 			for (i = firstmirror; i < mirrorcnt && error == 0; i++){
6231 				m2 = NULL;
6232 				if (m != NULL && i < mirrorcnt - 1)
6233 					m2 = m_copym(m, 0, M_COPYALL, M_WAITOK);
6234 				else {
6235 					m2 = m;
6236 					m = NULL;
6237 				}
6238 				if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
6239 					dev = rflp->nfsfl_ffm[i].dev;
6240 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
6241 					    rflp->nfsfl_ffm[i].devp);
6242 				} else {
6243 					dev = rflp->nfsfl_dev;
6244 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
6245 					    rflp->nfsfl_devp);
6246 				}
6247 				if (dip != NULL) {
6248 					if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
6249 					    != 0)
6250 						error = nfscl_dofflayoutio(vp,
6251 						    uiop, iomode, must_commit,
6252 						    &eof, &stateid, rwaccess,
6253 						    dip, layp, rflp, off, xfer,
6254 						    i, docommit, m2, tdrpc,
6255 						    newcred, p);
6256 					else
6257 						error = nfscl_doflayoutio(vp,
6258 						    uiop, iomode, must_commit,
6259 						    &eof, &stateid, rwaccess,
6260 						    dip, layp, rflp, off, xfer,
6261 						    docommit, newcred, p);
6262 					nfscl_reldevinfo(dip);
6263 				} else {
6264 					if (m2 != NULL)
6265 						m_freem(m2);
6266 					error = EIO;
6267 				}
6268 				tdrpc++;
6269 			}
6270 			if (m != NULL)
6271 				m_freem(m);
6272 			tdrpc = drpc;
6273 			timo = hz / 50;		/* Wait for 20msec. */
6274 			if (timo < 1)
6275 				timo = 1;
6276 			for (i = firstmirror; i < mirrorcnt - 1 &&
6277 			    tdrpc != NULL; i++, tdrpc++) {
6278 				/*
6279 				 * For the unused drpc entries, both inprog and
6280 				 * err == 0, so this loop won't break.
6281 				 */
6282 				while (tdrpc->inprog != 0 && tdrpc->done == 0)
6283 					tsleep(&tdrpc->tsk, PVFS, "clrpcio",
6284 					    timo);
6285 				if (error == 0 && tdrpc->err != 0)
6286 					error = tdrpc->err;
6287 				if (rwaccess != NFSV4OPEN_ACCESSREAD &&
6288 				    docommit == 0 && *must_commit == 0 &&
6289 				    tdrpc->must_commit == 1)
6290 					*must_commit = 1;
6291 			}
6292 			free(drpc, M_TEMP);
6293 			if (error == 0) {
6294 				if (mirrorcnt > 1 && rwaccess ==
6295 				    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
6296 					NFSLOCKCLSTATE();
6297 					layp->nfsly_flags |= NFSLY_WRITTEN;
6298 					NFSUNLOCKCLSTATE();
6299 				}
6300 				lastbyte = off + xfer - 1;
6301 				NFSLOCKCLSTATE();
6302 				if (lastbyte > layp->nfsly_lastbyte)
6303 					layp->nfsly_lastbyte = lastbyte;
6304 				NFSUNLOCKCLSTATE();
6305 			} else if (error == NFSERR_OPENMODE &&
6306 			    rwaccess == NFSV4OPEN_ACCESSREAD) {
6307 				NFSLOCKMNT(nmp);
6308 				nmp->nm_state |= NFSSTA_OPENMODE;
6309 				NFSUNLOCKMNT(nmp);
6310 			} else if ((error == NFSERR_NOSPC ||
6311 			    error == NFSERR_IO || error == NFSERR_NXIO) &&
6312 			    nmp->nm_minorvers == NFSV42_MINORVERSION) {
6313 				if (docommit != 0)
6314 					op = NFSV4OP_COMMIT;
6315 				else if (rwaccess == NFSV4OPEN_ACCESSREAD)
6316 					op = NFSV4OP_READ;
6317 				else
6318 					op = NFSV4OP_WRITE;
6319 				nfsrpc_layouterror(nmp, np->n_fhp->nfh_fh,
6320 				    np->n_fhp->nfh_len, off, xfer,
6321 				    &layp->nfsly_stateid, newcred, p, error, op,
6322 				    dip->nfsdi_deviceid);
6323 				error = EIO;
6324 			} else
6325 				error = EIO;
6326 			if (error == 0)
6327 				len -= (oresid - (uint64_t)uiop->uio_resid);
6328 			else if (mirrorcnt > 1 && rwaccess ==
6329 			    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
6330 				/*
6331 				 * In case the rpc gets retried, roll the
6332 				 * uio fields changed by nfsm_uiombuflist()
6333 				 * back.
6334 				 */
6335 				uiop->uio_offset = offs;
6336 				uiop->uio_resid = resid;
6337 				uiop->uio_iov->iov_base = iovbase;
6338 				uiop->uio_iov->iov_len = iovlen;
6339 			}
6340 		}
6341 	}
6342 	if (lckp != NULL)
6343 		nfscl_lockderef(lckp);
6344 	NFSFREECRED(newcred);
6345 	nfscl_rellayout(layp, 0);
6346 	nfscl_relref(nmp);
6347 	return (error);
6348 }
6349 
6350 /*
6351  * Find a file layout that will handle the first bytes of the requested
6352  * range and return the information from it needed to the I/O operation.
6353  */
6354 int
6355 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
6356     struct nfsclflayout **retflpp)
6357 {
6358 	struct nfsclflayout *flp, *nflp, *rflp;
6359 	uint32_t rw;
6360 
6361 	rflp = NULL;
6362 	rw = rwaccess;
6363 	/* For reading, do the Read list first and then the Write list. */
6364 	do {
6365 		if (rw == NFSV4OPEN_ACCESSREAD)
6366 			flp = LIST_FIRST(&lyp->nfsly_flayread);
6367 		else
6368 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
6369 		while (flp != NULL) {
6370 			nflp = LIST_NEXT(flp, nfsfl_list);
6371 			if (flp->nfsfl_off > off)
6372 				break;
6373 			if (flp->nfsfl_end > off &&
6374 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
6375 				rflp = flp;
6376 			flp = nflp;
6377 		}
6378 		if (rw == NFSV4OPEN_ACCESSREAD)
6379 			rw = NFSV4OPEN_ACCESSWRITE;
6380 		else
6381 			rw = 0;
6382 	} while (rw != 0);
6383 	if (rflp != NULL) {
6384 		/* This one covers the most bytes starting at off. */
6385 		*retflpp = rflp;
6386 		return (0);
6387 	}
6388 	return (EIO);
6389 }
6390 
6391 /*
6392  * Do I/O using an NFSv4.1 or NFSv4.2 file layout.
6393  */
6394 static int
6395 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6396     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6397     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6398     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
6399 {
6400 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
6401 	int commit_thru_mds, error, stripe_index, stripe_pos, minorvers;
6402 	struct nfsnode *np;
6403 	struct nfsfh *fhp;
6404 	struct nfsclds **dspp;
6405 
6406 	np = VTONFS(vp);
6407 	rel_off = off - flp->nfsfl_patoff;
6408 	stripe_unit_size = flp->nfsfl_util & NFSFLAYUTIL_STRIPE_MASK;
6409 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
6410 	    dp->nfsdi_stripecnt;
6411 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
6412 	error = 0;
6413 
6414 	/* Loop around, doing I/O for each stripe unit. */
6415 	while (len > 0 && error == 0) {
6416 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
6417 		dspp = nfsfldi_addr(dp, stripe_index);
6418 		if (((*dspp)->nfsclds_flags & NFSCLDS_MINORV2) != 0)
6419 			minorvers = NFSV42_MINORVERSION;
6420 		else
6421 			minorvers = NFSV41_MINORVERSION;
6422 		if (len > transfer && docommit == 0)
6423 			xfer = transfer;
6424 		else
6425 			xfer = len;
6426 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
6427 			/* Dense layout. */
6428 			if (stripe_pos >= flp->nfsfl_fhcnt)
6429 				return (EIO);
6430 			fhp = flp->nfsfl_fh[stripe_pos];
6431 			io_off = (rel_off / (stripe_unit_size *
6432 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
6433 			    rel_off % stripe_unit_size;
6434 		} else {
6435 			/* Sparse layout. */
6436 			if (flp->nfsfl_fhcnt > 1) {
6437 				if (stripe_index >= flp->nfsfl_fhcnt)
6438 					return (EIO);
6439 				fhp = flp->nfsfl_fh[stripe_index];
6440 			} else if (flp->nfsfl_fhcnt == 1)
6441 				fhp = flp->nfsfl_fh[0];
6442 			else
6443 				fhp = np->n_fhp;
6444 			io_off = off;
6445 		}
6446 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
6447 			commit_thru_mds = 1;
6448 			if (docommit != 0)
6449 				error = EIO;
6450 		} else {
6451 			commit_thru_mds = 0;
6452 			NFSLOCKNODE(np);
6453 			np->n_flag |= NDSCOMMIT;
6454 			NFSUNLOCKNODE(np);
6455 		}
6456 		if (docommit != 0) {
6457 			if (error == 0)
6458 				error = nfsrpc_commitds(vp, io_off, xfer,
6459 				    *dspp, fhp, NFS_VER4, minorvers, cred, p);
6460 			if (error == 0) {
6461 				/*
6462 				 * Set both eof and uio_resid = 0 to end any
6463 				 * loops.
6464 				 */
6465 				*eofp = 1;
6466 				uiop->uio_resid = 0;
6467 			} else {
6468 				NFSLOCKNODE(np);
6469 				np->n_flag &= ~NDSCOMMIT;
6470 				NFSUNLOCKNODE(np);
6471 			}
6472 		} else if (rwflag == NFSV4OPEN_ACCESSREAD)
6473 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6474 			    io_off, xfer, fhp, 0, NFS_VER4, minorvers, cred, p);
6475 		else {
6476 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
6477 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
6478 			    0, NFS_VER4, minorvers, cred, p);
6479 			if (error == 0) {
6480 				NFSLOCKCLSTATE();
6481 				lyp->nfsly_flags |= NFSLY_WRITTEN;
6482 				NFSUNLOCKCLSTATE();
6483 			}
6484 		}
6485 		if (error == 0) {
6486 			transfer = stripe_unit_size;
6487 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
6488 			len -= xfer;
6489 			off += xfer;
6490 		}
6491 	}
6492 	return (error);
6493 }
6494 
6495 /*
6496  * Do I/O using an NFSv4.1 flex file layout.
6497  */
6498 static int
6499 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6500     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6501     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6502     uint64_t len, int mirror, int docommit, struct mbuf *mp,
6503     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6504 {
6505 	uint64_t xfer;
6506 	int error;
6507 	struct nfsnode *np;
6508 	struct nfsfh *fhp;
6509 	struct nfsclds **dspp;
6510 	struct ucred *tcred;
6511 	struct mbuf *m, *m2;
6512 	uint32_t copylen;
6513 
6514 	np = VTONFS(vp);
6515 	error = 0;
6516 	NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6517 	    (uintmax_t)len);
6518 	/* Loop around, doing I/O for each stripe unit. */
6519 	while (len > 0 && error == 0) {
6520 		dspp = nfsfldi_addr(dp, 0);
6521 		fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6522 		stateidp = &flp->nfsfl_ffm[mirror].st;
6523 		NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6524 		    mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6525 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6526 			tcred = NFSNEWCRED(cred);
6527 			tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6528 			tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6529 			tcred->cr_ngroups = 1;
6530 		} else
6531 			tcred = cred;
6532 		if (rwflag == NFSV4OPEN_ACCESSREAD)
6533 			copylen = dp->nfsdi_rsize;
6534 		else {
6535 			copylen = dp->nfsdi_wsize;
6536 			if (len > copylen && mp != NULL) {
6537 				/*
6538 				 * When a mirrored configuration needs to do
6539 				 * multiple writes to each mirror, all writes
6540 				 * except the last one must be a multiple of
6541 				 * 4 bytes.  This is required so that the XDR
6542 				 * does not need padding.
6543 				 * If possible, clip the size to an exact
6544 				 * multiple of the mbuf length, so that the
6545 				 * split will be on an mbuf boundary.
6546 				 */
6547 				copylen &= 0xfffffffc;
6548 				if (copylen > mp->m_len)
6549 					copylen = copylen / mp->m_len *
6550 					    mp->m_len;
6551 			}
6552 		}
6553 		NFSLOCKNODE(np);
6554 		np->n_flag |= NDSCOMMIT;
6555 		NFSUNLOCKNODE(np);
6556 		if (len > copylen && docommit == 0)
6557 			xfer = copylen;
6558 		else
6559 			xfer = len;
6560 		if (docommit != 0) {
6561 			if (error == 0) {
6562 				/*
6563 				 * Do last mirrored DS commit with this thread.
6564 				 */
6565 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6566 					error = nfsio_commitds(vp, off, xfer,
6567 					    *dspp, fhp, dp->nfsdi_vers,
6568 					    dp->nfsdi_minorvers, drpc, tcred,
6569 					    p);
6570 				else
6571 					error = nfsrpc_commitds(vp, off, xfer,
6572 					    *dspp, fhp, dp->nfsdi_vers,
6573 					    dp->nfsdi_minorvers, tcred, p);
6574 				NFSCL_DEBUG(4, "commitds=%d\n", error);
6575 				if (error != 0 && error != EACCES && error !=
6576 				    ESTALE) {
6577 					NFSCL_DEBUG(4,
6578 					    "DS layreterr for commit\n");
6579 					nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6580 					    lyp, *dspp);
6581 				}
6582 			}
6583 			NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6584 			if (error == 0) {
6585 				/*
6586 				 * Set both eof and uio_resid = 0 to end any
6587 				 * loops.
6588 				 */
6589 				*eofp = 1;
6590 				uiop->uio_resid = 0;
6591 			} else {
6592 				NFSLOCKNODE(np);
6593 				np->n_flag &= ~NDSCOMMIT;
6594 				NFSUNLOCKNODE(np);
6595 			}
6596 		} else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6597 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6598 			    off, xfer, fhp, 1, dp->nfsdi_vers,
6599 			    dp->nfsdi_minorvers, tcred, p);
6600 			NFSCL_DEBUG(4, "readds=%d\n", error);
6601 			if (error != 0 && error != EACCES && error != ESTALE) {
6602 				NFSCL_DEBUG(4, "DS layreterr for read\n");
6603 				nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6604 				    *dspp);
6605 			}
6606 		} else {
6607 			if (flp->nfsfl_mirrorcnt == 1) {
6608 				error = nfsrpc_writeds(vp, uiop, iomode,
6609 				    must_commit, stateidp, *dspp, off, xfer,
6610 				    fhp, 0, 1, dp->nfsdi_vers,
6611 				    dp->nfsdi_minorvers, tcred, p);
6612 				if (error == 0) {
6613 					NFSLOCKCLSTATE();
6614 					lyp->nfsly_flags |= NFSLY_WRITTEN;
6615 					NFSUNLOCKCLSTATE();
6616 				}
6617 			} else {
6618 				m = mp;
6619 				if (xfer < len) {
6620 					/* The mbuf list must be split. */
6621 					m2 = nfsm_split(mp, xfer);
6622 					if (m2 != NULL)
6623 						mp = m2;
6624 					else {
6625 						m_freem(mp);
6626 						error = EIO;
6627 					}
6628 				}
6629 				NFSCL_DEBUG(4, "mcopy len=%jd xfer=%jd\n",
6630 				    (uintmax_t)len, (uintmax_t)xfer);
6631 				/*
6632 				 * Do last write to a mirrored DS with this
6633 				 * thread.
6634 				 */
6635 				if (error == 0) {
6636 					if (mirror < flp->nfsfl_mirrorcnt - 1)
6637 						error = nfsio_writedsmir(vp,
6638 						    iomode, must_commit,
6639 						    stateidp, *dspp, off,
6640 						    xfer, fhp, m,
6641 						    dp->nfsdi_vers,
6642 						    dp->nfsdi_minorvers, drpc,
6643 						    tcred, p);
6644 					else
6645 						error = nfsrpc_writedsmir(vp,
6646 						    iomode, must_commit,
6647 						    stateidp, *dspp, off,
6648 						    xfer, fhp, m,
6649 						    dp->nfsdi_vers,
6650 						    dp->nfsdi_minorvers, tcred,
6651 						    p);
6652 				}
6653 				NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6654 				if (error != 0 && error != EACCES && error !=
6655 				    ESTALE) {
6656 					NFSCL_DEBUG(4,
6657 					    "DS layreterr for write\n");
6658 					nfscl_dserr(NFSV4OP_WRITE, error, dp,
6659 					    lyp, *dspp);
6660 				}
6661 			}
6662 		}
6663 		NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6664 		if (error == 0) {
6665 			len -= xfer;
6666 			off += xfer;
6667 		}
6668 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6669 			NFSFREECRED(tcred);
6670 	}
6671 	NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6672 	return (error);
6673 }
6674 
6675 /*
6676  * The actual read RPC done to a DS.
6677  */
6678 static int
6679 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6680     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6681     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6682 {
6683 	uint32_t *tl;
6684 	int attrflag, error, retlen;
6685 	struct nfsrv_descript nfsd;
6686 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6687 	struct nfsrv_descript *nd = &nfsd;
6688 	struct nfssockreq *nrp;
6689 	struct nfsvattr na;
6690 
6691 	nd->nd_mrep = NULL;
6692 	if (vers == 0 || vers == NFS_VER4) {
6693 		nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6694 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6695 		vers = NFS_VER4;
6696 		NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6697 		if (flex != 0)
6698 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6699 		else
6700 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6701 	} else {
6702 		nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6703 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6704 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READ]);
6705 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READDS]);
6706 		NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6707 	}
6708 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6709 	txdr_hyper(io_off, tl);
6710 	*(tl + 2) = txdr_unsigned(len);
6711 	nrp = dsp->nfsclds_sockp;
6712 	NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6713 	if (nrp == NULL)
6714 		/* If NULL, use the MDS socket. */
6715 		nrp = &nmp->nm_sockreq;
6716 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6717 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6718 	NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6719 	    error);
6720 	if (error != 0)
6721 		return (error);
6722 	if (vers == NFS_VER3) {
6723 		error = nfscl_postop_attr(nd, &na, &attrflag);
6724 		NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6725 		if (error != 0)
6726 			goto nfsmout;
6727 	}
6728 	if (nd->nd_repstat != 0) {
6729 		error = nd->nd_repstat;
6730 		goto nfsmout;
6731 	}
6732 	if (vers == NFS_VER3) {
6733 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6734 		*eofp = fxdr_unsigned(int, *(tl + 1));
6735 	} else {
6736 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6737 		*eofp = fxdr_unsigned(int, *tl);
6738 	}
6739 	NFSM_STRSIZ(retlen, len);
6740 	NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6741 	error = nfsm_mbufuio(nd, uiop, retlen);
6742 nfsmout:
6743 	if (nd->nd_mrep != NULL)
6744 		m_freem(nd->nd_mrep);
6745 	return (error);
6746 }
6747 
6748 /*
6749  * The actual write RPC done to a DS.
6750  */
6751 static int
6752 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6753     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6754     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6755     struct ucred *cred, NFSPROC_T *p)
6756 {
6757 	uint32_t *tl;
6758 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6759 	int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6760 	int32_t backup;
6761 	struct nfsrv_descript nfsd;
6762 	struct nfsrv_descript *nd = &nfsd;
6763 	struct nfssockreq *nrp;
6764 	struct nfsvattr na;
6765 
6766 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6767 	nd->nd_mrep = NULL;
6768 	if (vers == 0 || vers == NFS_VER4) {
6769 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6770 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6771 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6772 		vers = NFS_VER4;
6773 		if (flex != 0)
6774 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6775 		else
6776 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6777 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6778 	} else {
6779 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6780 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6781 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6782 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6783 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6784 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6785 	}
6786 	txdr_hyper(io_off, tl);
6787 	tl += 2;
6788 	if (vers == NFS_VER3)
6789 		*tl++ = txdr_unsigned(len);
6790 	*tl++ = txdr_unsigned(*iomode);
6791 	*tl = txdr_unsigned(len);
6792 	nfsm_uiombuf(nd, uiop, len);
6793 	nrp = dsp->nfsclds_sockp;
6794 	if (nrp == NULL)
6795 		/* If NULL, use the MDS socket. */
6796 		nrp = &nmp->nm_sockreq;
6797 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6798 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6799 	NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6800 	    nd->nd_repstat);
6801 	if (error != 0)
6802 		return (error);
6803 	if (nd->nd_repstat != 0) {
6804 		/*
6805 		 * In case the rpc gets retried, roll
6806 		 * the uio fields changed by nfsm_uiombuf()
6807 		 * back.
6808 		 */
6809 		uiop->uio_offset -= len;
6810 		uiop->uio_resid += len;
6811 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base - len;
6812 		uiop->uio_iov->iov_len += len;
6813 		error = nd->nd_repstat;
6814 	} else {
6815 		if (vers == NFS_VER3) {
6816 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6817 			    NULL);
6818 			NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6819 			if (error != 0)
6820 				goto nfsmout;
6821 		}
6822 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6823 		rlen = fxdr_unsigned(int, *tl++);
6824 		NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6825 		if (rlen == 0) {
6826 			error = NFSERR_IO;
6827 			goto nfsmout;
6828 		} else if (rlen < len) {
6829 			backup = len - rlen;
6830 			uiop->uio_iov->iov_base =
6831 			    (char *)uiop->uio_iov->iov_base - backup;
6832 			uiop->uio_iov->iov_len += backup;
6833 			uiop->uio_offset -= backup;
6834 			uiop->uio_resid += backup;
6835 			len = rlen;
6836 		}
6837 		commit = fxdr_unsigned(int, *tl++);
6838 
6839 		/*
6840 		 * Return the lowest commitment level
6841 		 * obtained by any of the RPCs.
6842 		 */
6843 		if (committed == NFSWRITE_FILESYNC)
6844 			committed = commit;
6845 		else if (committed == NFSWRITE_DATASYNC &&
6846 		    commit == NFSWRITE_UNSTABLE)
6847 			committed = commit;
6848 		if (commit_thru_mds != 0) {
6849 			NFSLOCKMNT(nmp);
6850 			if (!NFSHASWRITEVERF(nmp)) {
6851 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6852 				NFSSETWRITEVERF(nmp);
6853 			} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF) &&
6854 			    *must_commit != 2) {
6855 				*must_commit = 1;
6856 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6857 			}
6858 			NFSUNLOCKMNT(nmp);
6859 		} else {
6860 			NFSLOCKDS(dsp);
6861 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6862 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6863 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6864 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF) &&
6865 			    *must_commit != 2) {
6866 				*must_commit = 1;
6867 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6868 			}
6869 			NFSUNLOCKDS(dsp);
6870 		}
6871 	}
6872 nfsmout:
6873 	if (nd->nd_mrep != NULL)
6874 		m_freem(nd->nd_mrep);
6875 	*iomode = committed;
6876 	if (nd->nd_repstat != 0 && error == 0)
6877 		error = nd->nd_repstat;
6878 	return (error);
6879 }
6880 
6881 /*
6882  * The actual write RPC done to a DS.
6883  * This variant is called from a separate kernel process for mirrors.
6884  * Any short write is considered an IO error.
6885  */
6886 static int
6887 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6888     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6889     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6890     struct ucred *cred, NFSPROC_T *p)
6891 {
6892 	uint32_t *tl;
6893 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6894 	int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6895 	struct nfsrv_descript nfsd;
6896 	struct nfsrv_descript *nd = &nfsd;
6897 	struct nfssockreq *nrp;
6898 	struct nfsvattr na;
6899 
6900 	nd->nd_mrep = NULL;
6901 	if (vers == 0 || vers == NFS_VER4) {
6902 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6903 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6904 		vers = NFS_VER4;
6905 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6906 		    minorvers);
6907 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6908 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6909 	} else {
6910 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6911 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6912 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6913 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6914 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6915 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6916 	}
6917 	txdr_hyper(io_off, tl);
6918 	tl += 2;
6919 	if (vers == NFS_VER3)
6920 		*tl++ = txdr_unsigned(len);
6921 	*tl++ = txdr_unsigned(*iomode);
6922 	*tl = txdr_unsigned(len);
6923 	if (len > 0) {
6924 		/* Put data in mbuf chain. */
6925 		nd->nd_mb->m_next = m;
6926 	}
6927 	nrp = dsp->nfsclds_sockp;
6928 	if (nrp == NULL)
6929 		/* If NULL, use the MDS socket. */
6930 		nrp = &nmp->nm_sockreq;
6931 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6932 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6933 	NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6934 	    nd->nd_repstat);
6935 	if (error != 0)
6936 		return (error);
6937 	if (nd->nd_repstat != 0)
6938 		error = nd->nd_repstat;
6939 	else {
6940 		if (vers == NFS_VER3) {
6941 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6942 			    NULL);
6943 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6944 			    error);
6945 			if (error != 0)
6946 				goto nfsmout;
6947 		}
6948 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6949 		rlen = fxdr_unsigned(int, *tl++);
6950 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6951 		    rlen);
6952 		if (rlen != len) {
6953 			error = NFSERR_IO;
6954 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6955 			    len, rlen);
6956 			goto nfsmout;
6957 		}
6958 		commit = fxdr_unsigned(int, *tl++);
6959 
6960 		/*
6961 		 * Return the lowest commitment level
6962 		 * obtained by any of the RPCs.
6963 		 */
6964 		if (committed == NFSWRITE_FILESYNC)
6965 			committed = commit;
6966 		else if (committed == NFSWRITE_DATASYNC &&
6967 		    commit == NFSWRITE_UNSTABLE)
6968 			committed = commit;
6969 		NFSLOCKDS(dsp);
6970 		if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6971 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6972 			dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6973 		} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF) &&
6974 		    *must_commit != 2) {
6975 			*must_commit = 1;
6976 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6977 		}
6978 		NFSUNLOCKDS(dsp);
6979 	}
6980 nfsmout:
6981 	if (nd->nd_mrep != NULL)
6982 		m_freem(nd->nd_mrep);
6983 	*iomode = committed;
6984 	if (nd->nd_repstat != 0 && error == 0)
6985 		error = nd->nd_repstat;
6986 	return (error);
6987 }
6988 
6989 /*
6990  * Start up the thread that will execute nfsrpc_writedsmir().
6991  */
6992 static void
6993 start_writedsmir(void *arg, int pending)
6994 {
6995 	struct nfsclwritedsdorpc *drpc;
6996 
6997 	drpc = (struct nfsclwritedsdorpc *)arg;
6998 	drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
6999 	    &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
7000 	    drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
7001 	    drpc->p);
7002 	drpc->done = 1;
7003 	crfree(drpc->cred);
7004 	NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
7005 }
7006 
7007 /*
7008  * Set up the write DS mirror call for the pNFS I/O thread.
7009  */
7010 static int
7011 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
7012     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
7013     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
7014     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7015 {
7016 	int error, ret;
7017 
7018 	error = 0;
7019 	drpc->done = 0;
7020 	drpc->vp = vp;
7021 	drpc->iomode = *iomode;
7022 	drpc->must_commit = *must_commit;
7023 	drpc->stateidp = stateidp;
7024 	drpc->dsp = dsp;
7025 	drpc->off = off;
7026 	drpc->len = len;
7027 	drpc->fhp = fhp;
7028 	drpc->m = m;
7029 	drpc->vers = vers;
7030 	drpc->minorvers = minorvers;
7031 	drpc->cred = crhold(cred);
7032 	drpc->p = p;
7033 	drpc->inprog = 0;
7034 	ret = EIO;
7035 	if (nfs_pnfsiothreads != 0) {
7036 		ret = nfs_pnfsio(start_writedsmir, drpc);
7037 		NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
7038 	}
7039 	if (ret != 0) {
7040 		error = nfsrpc_writedsmir(vp, iomode, &drpc->must_commit,
7041 		    stateidp, dsp, off, len, fhp, m, vers, minorvers, cred, p);
7042 		crfree(drpc->cred);
7043 	}
7044 	NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
7045 	return (error);
7046 }
7047 
7048 /*
7049  * Free up the nfsclds structure.
7050  */
7051 void
7052 nfscl_freenfsclds(struct nfsclds *dsp)
7053 {
7054 	int i;
7055 
7056 	if (dsp == NULL)
7057 		return;
7058 	if (dsp->nfsclds_sockp != NULL) {
7059 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
7060 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
7061 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
7062 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
7063 	}
7064 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
7065 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
7066 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
7067 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
7068 			m_freem(
7069 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
7070 	}
7071 	free(dsp, M_NFSCLDS);
7072 }
7073 
7074 static enum nfsclds_state
7075 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
7076     struct nfsclds **retdspp, uint32_t *sequencep)
7077 {
7078 	struct nfsclds *dsp;
7079 	int fndseq;
7080 
7081 	/*
7082 	 * Search the list of nfsclds structures for one with the same
7083 	 * server.
7084 	 */
7085 	fndseq = 0;
7086 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
7087 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
7088 		    dsp->nfsclds_servownlen != 0 &&
7089 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
7090 		    dsp->nfsclds_servownlen) &&
7091 		    dsp->nfsclds_sess.nfsess_defunct == 0) {
7092 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
7093 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
7094 			    dsp->nfsclds_flags);
7095 			if (fndseq == 0) {
7096 				/* Get sequenceid# from first entry. */
7097 				*sequencep =
7098 				    dsp->nfsclds_sess.nfsess_sequenceid;
7099 				fndseq = 1;
7100 			}
7101 			/* Server major id matches. */
7102 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
7103 				*retdspp = dsp;
7104 				return (NFSDSP_USETHISSESSION);
7105 			}
7106 		}
7107 	}
7108 	if (fndseq != 0)
7109 		return (NFSDSP_SEQTHISSESSION);
7110 	return (NFSDSP_NOTFOUND);
7111 }
7112 
7113 /*
7114  * NFS commit rpc to a NFSv4.1 DS.
7115  */
7116 static int
7117 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
7118     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
7119     NFSPROC_T *p)
7120 {
7121 	uint32_t *tl;
7122 	struct nfsrv_descript nfsd, *nd = &nfsd;
7123 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7124 	struct nfssockreq *nrp;
7125 	struct nfsvattr na;
7126 	int attrflag, error;
7127 
7128 	nd->nd_mrep = NULL;
7129 	if (vers == 0 || vers == NFS_VER4) {
7130 		nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
7131 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
7132 		vers = NFS_VER4;
7133 	} else {
7134 		nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
7135 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
7136 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMIT]);
7137 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMITDS]);
7138 	}
7139 	NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
7140 	    minorvers);
7141 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
7142 	txdr_hyper(offset, tl);
7143 	tl += 2;
7144 	*tl = txdr_unsigned(cnt);
7145 	nrp = dsp->nfsclds_sockp;
7146 	if (nrp == NULL)
7147 		/* If NULL, use the MDS socket. */
7148 		nrp = &nmp->nm_sockreq;
7149 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7150 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7151 	NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
7152 	    nd->nd_repstat);
7153 	if (error != 0)
7154 		return (error);
7155 	if (nd->nd_repstat == 0) {
7156 		if (vers == NFS_VER3) {
7157 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
7158 			    NULL);
7159 			NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
7160 			if (error != 0)
7161 				goto nfsmout;
7162 		}
7163 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
7164 		NFSLOCKDS(dsp);
7165 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
7166 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
7167 			error = NFSERR_STALEWRITEVERF;
7168 		}
7169 		NFSUNLOCKDS(dsp);
7170 	}
7171 nfsmout:
7172 	if (error == 0 && nd->nd_repstat != 0)
7173 		error = nd->nd_repstat;
7174 	m_freem(nd->nd_mrep);
7175 	return (error);
7176 }
7177 
7178 /*
7179  * Start up the thread that will execute nfsrpc_commitds().
7180  */
7181 static void
7182 start_commitds(void *arg, int pending)
7183 {
7184 	struct nfsclwritedsdorpc *drpc;
7185 
7186 	drpc = (struct nfsclwritedsdorpc *)arg;
7187 	drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
7188 	    drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
7189 	    drpc->p);
7190 	drpc->done = 1;
7191 	crfree(drpc->cred);
7192 	NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
7193 }
7194 
7195 /*
7196  * Set up the commit DS mirror call for the pNFS I/O thread.
7197  */
7198 static int
7199 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
7200     struct nfsfh *fhp, int vers, int minorvers,
7201     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7202 {
7203 	int error, ret;
7204 
7205 	error = 0;
7206 	drpc->done = 0;
7207 	drpc->vp = vp;
7208 	drpc->off = offset;
7209 	drpc->len = cnt;
7210 	drpc->dsp = dsp;
7211 	drpc->fhp = fhp;
7212 	drpc->vers = vers;
7213 	drpc->minorvers = minorvers;
7214 	drpc->cred = crhold(cred);
7215 	drpc->p = p;
7216 	drpc->inprog = 0;
7217 	ret = EIO;
7218 	if (nfs_pnfsiothreads != 0) {
7219 		ret = nfs_pnfsio(start_commitds, drpc);
7220 		NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
7221 	}
7222 	if (ret != 0) {
7223 		error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
7224 		    minorvers, cred, p);
7225 		crfree(drpc->cred);
7226 	}
7227 	NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
7228 	return (error);
7229 }
7230 
7231 /*
7232  * NFS Advise rpc
7233  */
7234 int
7235 nfsrpc_advise(vnode_t vp, off_t offset, uint64_t cnt, int advise,
7236     struct ucred *cred, NFSPROC_T *p)
7237 {
7238 	u_int32_t *tl;
7239 	struct nfsrv_descript nfsd, *nd = &nfsd;
7240 	nfsattrbit_t hints;
7241 	int error;
7242 
7243 	NFSZERO_ATTRBIT(&hints);
7244 	if (advise == POSIX_FADV_WILLNEED)
7245 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
7246 	else if (advise == POSIX_FADV_DONTNEED)
7247 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
7248 	else
7249 		return (0);
7250 	NFSCL_REQSTART(nd, NFSPROC_IOADVISE, vp);
7251 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
7252 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
7253 	txdr_hyper(offset, tl);
7254 	tl += 2;
7255 	txdr_hyper(cnt, tl);
7256 	nfsrv_putattrbit(nd, &hints);
7257 	error = nfscl_request(nd, vp, p, cred);
7258 	if (error != 0)
7259 		return (error);
7260 	if (nd->nd_repstat != 0)
7261 		error = nd->nd_repstat;
7262 	m_freem(nd->nd_mrep);
7263 	return (error);
7264 }
7265 
7266 #ifdef notyet
7267 /*
7268  * NFS advise rpc to a NFSv4.2 DS.
7269  */
7270 static int
7271 nfsrpc_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
7272     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
7273     struct ucred *cred, NFSPROC_T *p)
7274 {
7275 	uint32_t *tl;
7276 	struct nfsrv_descript nfsd, *nd = &nfsd;
7277 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7278 	struct nfssockreq *nrp;
7279 	nfsattrbit_t hints;
7280 	int error;
7281 
7282 	/* For NFS DSs prior to NFSv4.2, just return OK. */
7283 	if (vers == NFS_VER3 || minorversion < NFSV42_MINORVERSION)
7284 		return (0);
7285 	NFSZERO_ATTRBIT(&hints);
7286 	if (advise == POSIX_FADV_WILLNEED)
7287 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
7288 	else if (advise == POSIX_FADV_DONTNEED)
7289 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
7290 	else
7291 		return (0);
7292 	nd->nd_mrep = NULL;
7293 	nfscl_reqstart(nd, NFSPROC_IOADVISEDS, nmp, fhp->nfh_fh,
7294 	    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
7295 	vers = NFS_VER4;
7296 	NFSCL_DEBUG(4, "nfsrpc_adviseds: vers=%d minvers=%d\n", vers,
7297 	    minorvers);
7298 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
7299 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
7300 	txdr_hyper(offset, tl);
7301 	tl += 2;
7302 	*tl = txdr_unsigned(cnt);
7303 	nfsrv_putattrbit(nd, &hints);
7304 	nrp = dsp->nfsclds_sockp;
7305 	if (nrp == NULL)
7306 		/* If NULL, use the MDS socket. */
7307 		nrp = &nmp->nm_sockreq;
7308 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7309 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7310 	NFSCL_DEBUG(4, "nfsrpc_adviseds: err=%d stat=%d\n", error,
7311 	    nd->nd_repstat);
7312 	if (error != 0)
7313 		return (error);
7314 	if (nd->nd_repstat != 0)
7315 		error = nd->nd_repstat;
7316 	m_freem(nd->nd_mrep);
7317 	return (error);
7318 }
7319 
7320 /*
7321  * Start up the thread that will execute nfsrpc_commitds().
7322  */
7323 static void
7324 start_adviseds(void *arg, int pending)
7325 {
7326 	struct nfsclwritedsdorpc *drpc;
7327 
7328 	drpc = (struct nfsclwritedsdorpc *)arg;
7329 	drpc->err = nfsrpc_adviseds(drpc->vp, drpc->off, drpc->len,
7330 	    drpc->advise, drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers,
7331 	    drpc->cred, drpc->p);
7332 	drpc->done = 1;
7333 	crfree(drpc->cred);
7334 	NFSCL_DEBUG(4, "start_adviseds: err=%d\n", drpc->err);
7335 }
7336 
7337 /*
7338  * Set up the advise DS mirror call for the pNFS I/O thread.
7339  */
7340 static int
7341 nfsio_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
7342     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
7343     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7344 {
7345 	int error, ret;
7346 
7347 	error = 0;
7348 	drpc->done = 0;
7349 	drpc->vp = vp;
7350 	drpc->off = offset;
7351 	drpc->len = cnt;
7352 	drpc->advise = advise;
7353 	drpc->dsp = dsp;
7354 	drpc->fhp = fhp;
7355 	drpc->vers = vers;
7356 	drpc->minorvers = minorvers;
7357 	drpc->cred = crhold(cred);
7358 	drpc->p = p;
7359 	drpc->inprog = 0;
7360 	ret = EIO;
7361 	if (nfs_pnfsiothreads != 0) {
7362 		ret = nfs_pnfsio(start_adviseds, drpc);
7363 		NFSCL_DEBUG(4, "nfsio_adviseds: nfs_pnfsio=%d\n", ret);
7364 	}
7365 	if (ret != 0) {
7366 		error = nfsrpc_adviseds(vp, offset, cnt, advise, dsp, fhp, vers,
7367 		    minorvers, cred, p);
7368 		crfree(drpc->cred);
7369 	}
7370 	NFSCL_DEBUG(4, "nfsio_adviseds: error=%d\n", error);
7371 	return (error);
7372 }
7373 #endif	/* notyet */
7374 
7375 /*
7376  * Do the Allocate operation, retrying for recovery.
7377  */
7378 int
7379 nfsrpc_allocate(vnode_t vp, off_t off, off_t len, struct nfsvattr *nap,
7380     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
7381 {
7382 	int error, expireret = 0, retrycnt, nostateid;
7383 	uint32_t clidrev = 0;
7384 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7385 	struct nfsfh *nfhp = NULL;
7386 	nfsv4stateid_t stateid;
7387 	off_t tmp_off;
7388 	void *lckp;
7389 
7390 	if (len < 0)
7391 		return (EINVAL);
7392 	if (len == 0)
7393 		return (0);
7394 	tmp_off = off + len;
7395 	NFSLOCKMNT(nmp);
7396 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < off) {
7397 		NFSUNLOCKMNT(nmp);
7398 		return (EFBIG);
7399 	}
7400 	if (nmp->nm_clp != NULL)
7401 		clidrev = nmp->nm_clp->nfsc_clientidrev;
7402 	NFSUNLOCKMNT(nmp);
7403 	nfhp = VTONFS(vp)->n_fhp;
7404 	retrycnt = 0;
7405 	do {
7406 		lckp = NULL;
7407 		nostateid = 0;
7408 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
7409 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
7410 		if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
7411 		    stateid.other[2] == 0) {
7412 			nostateid = 1;
7413 			NFSCL_DEBUG(1, "stateid0 in allocate\n");
7414 		}
7415 
7416 		/*
7417 		 * Not finding a stateid should probably never happen,
7418 		 * but just return an error for this case.
7419 		 */
7420 		if (nostateid != 0)
7421 			error = EIO;
7422 		else
7423 			error = nfsrpc_allocaterpc(vp, off, len, &stateid,
7424 			    nap, attrflagp, cred, p);
7425 		if (error == NFSERR_STALESTATEID)
7426 			nfscl_initiate_recovery(nmp->nm_clp);
7427 		if (lckp != NULL)
7428 			nfscl_lockderef(lckp);
7429 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
7430 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
7431 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
7432 			(void) nfs_catnap(PZERO, error, "nfs_allocate");
7433 		} else if ((error == NFSERR_EXPIRED ||
7434 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
7435 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
7436 		}
7437 		retrycnt++;
7438 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
7439 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
7440 	    error == NFSERR_STALEDONTRECOVER ||
7441 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
7442 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
7443 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
7444 	if (error != 0 && retrycnt >= 4)
7445 		error = EIO;
7446 	return (error);
7447 }
7448 
7449 /*
7450  * The allocate RPC.
7451  */
7452 static int
7453 nfsrpc_allocaterpc(vnode_t vp, off_t off, off_t len, nfsv4stateid_t *stateidp,
7454     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
7455 {
7456 	uint32_t *tl;
7457 	int error;
7458 	struct nfsrv_descript nfsd;
7459 	struct nfsrv_descript *nd = &nfsd;
7460 	nfsattrbit_t attrbits;
7461 
7462 	*attrflagp = 0;
7463 	NFSCL_REQSTART(nd, NFSPROC_ALLOCATE, vp);
7464 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7465 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
7466 	txdr_hyper(off, tl); tl += 2;
7467 	txdr_hyper(len, tl); tl += 2;
7468 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7469 	NFSGETATTR_ATTRBIT(&attrbits);
7470 	nfsrv_putattrbit(nd, &attrbits);
7471 	error = nfscl_request(nd, vp, p, cred);
7472 	if (error != 0)
7473 		return (error);
7474 	if (nd->nd_repstat == 0) {
7475 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7476 		error = nfsm_loadattr(nd, nap);
7477 		if (error == 0)
7478 			*attrflagp = NFS_LATTR_NOSHRINK;
7479 	} else
7480 		error = nd->nd_repstat;
7481 nfsmout:
7482 	m_freem(nd->nd_mrep);
7483 	return (error);
7484 }
7485 
7486 /*
7487  * Set up the XDR arguments for the LayoutGet operation.
7488  */
7489 static void
7490 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
7491     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
7492     int layoutlen, int usecurstateid)
7493 {
7494 	uint32_t *tl;
7495 
7496 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
7497 	    NFSX_STATEID);
7498 	*tl++ = newnfs_false;		/* Don't signal availability. */
7499 	*tl++ = txdr_unsigned(layouttype);
7500 	*tl++ = txdr_unsigned(iomode);
7501 	txdr_hyper(offset, tl);
7502 	tl += 2;
7503 	txdr_hyper(len, tl);
7504 	tl += 2;
7505 	txdr_hyper(minlen, tl);
7506 	tl += 2;
7507 	if (usecurstateid != 0) {
7508 		/* Special stateid for Current stateid. */
7509 		*tl++ = txdr_unsigned(1);
7510 		*tl++ = 0;
7511 		*tl++ = 0;
7512 		*tl++ = 0;
7513 	} else {
7514 		*tl++ = txdr_unsigned(stateidp->seqid);
7515 		NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
7516 		*tl++ = stateidp->other[0];
7517 		*tl++ = stateidp->other[1];
7518 		*tl++ = stateidp->other[2];
7519 	}
7520 	*tl = txdr_unsigned(layoutlen);
7521 }
7522 
7523 /*
7524  * Parse the reply for a successful LayoutGet operation.
7525  */
7526 static int
7527 nfsrv_parselayoutget(struct nfsmount *nmp, struct nfsrv_descript *nd,
7528     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp)
7529 {
7530 	uint32_t *tl;
7531 	struct nfsclflayout *flp, *prevflp, *tflp;
7532 	int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
7533 	int m, mirrorcnt;
7534 	uint64_t retlen, off;
7535 	struct nfsfh *nfhp;
7536 	uint8_t *cp;
7537 	uid_t user;
7538 	gid_t grp;
7539 
7540 	NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
7541 	error = 0;
7542 	flp = NULL;
7543 	gotiomode = -1;
7544 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
7545 	if (*tl++ != 0)
7546 		*retonclosep = 1;
7547 	else
7548 		*retonclosep = 0;
7549 	stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
7550 	NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
7551 	    (int)stateidp->seqid);
7552 	stateidp->other[0] = *tl++;
7553 	stateidp->other[1] = *tl++;
7554 	stateidp->other[2] = *tl++;
7555 	cnt = fxdr_unsigned(int, *tl);
7556 	NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
7557 	if (cnt <= 0 || cnt > 10000) {
7558 		/* Don't accept more than 10000 layouts in reply. */
7559 		error = NFSERR_BADXDR;
7560 		goto nfsmout;
7561 	}
7562 	for (i = 0; i < cnt; i++) {
7563 		/* Dissect to the layout type. */
7564 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
7565 		    3 * NFSX_UNSIGNED);
7566 		off = fxdr_hyper(tl); tl += 2;
7567 		retlen = fxdr_hyper(tl); tl += 2;
7568 		iomode = fxdr_unsigned(int, *tl++);
7569 		laytype = fxdr_unsigned(int, *tl);
7570 		NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
7571 		    (uintmax_t)off, (uintmax_t)retlen, iomode);
7572 		/* Ignore length of layout body for now. */
7573 		if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
7574 			/* Parse the File layout up to fhcnt. */
7575 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
7576 			    NFSX_HYPER + NFSX_V4DEVICEID);
7577 			fhcnt = fxdr_unsigned(int, *(tl + 4 +
7578 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
7579 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7580 			if (fhcnt < 0 || fhcnt > 100) {
7581 				/* Don't accept more than 100 file handles. */
7582 				error = NFSERR_BADXDR;
7583 				goto nfsmout;
7584 			}
7585 			if (fhcnt > 0)
7586 				flp = malloc(sizeof(*flp) + fhcnt *
7587 				    sizeof(struct nfsfh *), M_NFSFLAYOUT,
7588 				    M_WAITOK);
7589 			else
7590 				flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
7591 				    M_WAITOK);
7592 			flp->nfsfl_flags = NFSFL_FILE;
7593 			flp->nfsfl_fhcnt = 0;
7594 			flp->nfsfl_devp = NULL;
7595 			flp->nfsfl_off = off;
7596 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7597 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7598 			else
7599 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7600 			flp->nfsfl_iomode = iomode;
7601 			if (gotiomode == -1)
7602 				gotiomode = flp->nfsfl_iomode;
7603 			/* Ignore layout body length for now. */
7604 			NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
7605 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
7606 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
7607 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
7608 			mtx_lock(&nmp->nm_mtx);
7609 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_util &
7610 			    NFSFLAYUTIL_IOADVISE_THRU_MDS) != 0)
7611 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7612 			mtx_unlock(&nmp->nm_mtx);
7613 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
7614 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
7615 			NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
7616 			    flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
7617 			for (j = 0; j < fhcnt; j++) {
7618 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7619 				nfhlen = fxdr_unsigned(int, *tl);
7620 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
7621 					error = NFSERR_BADXDR;
7622 					goto nfsmout;
7623 				}
7624 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
7625 				    M_NFSFH, M_WAITOK);
7626 				flp->nfsfl_fh[j] = nfhp;
7627 				flp->nfsfl_fhcnt++;
7628 				nfhp->nfh_len = nfhlen;
7629 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
7630 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
7631 			}
7632 		} else if (laytype == NFSLAYOUT_FLEXFILE) {
7633 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
7634 			    NFSX_HYPER);
7635 			mirrorcnt = fxdr_unsigned(int, *(tl + 2));
7636 			NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
7637 			if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
7638 				error = NFSERR_BADXDR;
7639 				goto nfsmout;
7640 			}
7641 			flp = malloc(sizeof(*flp) + mirrorcnt *
7642 			    sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
7643 			flp->nfsfl_flags = NFSFL_FLEXFILE;
7644 			flp->nfsfl_mirrorcnt = mirrorcnt;
7645 			for (j = 0; j < mirrorcnt; j++)
7646 				flp->nfsfl_ffm[j].devp = NULL;
7647 			flp->nfsfl_off = off;
7648 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7649 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7650 			else
7651 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7652 			flp->nfsfl_iomode = iomode;
7653 			if (gotiomode == -1)
7654 				gotiomode = flp->nfsfl_iomode;
7655 			flp->nfsfl_stripeunit = fxdr_hyper(tl);
7656 			NFSCL_DEBUG(4, "stripeunit=%ju\n",
7657 			    (uintmax_t)flp->nfsfl_stripeunit);
7658 			for (j = 0; j < mirrorcnt; j++) {
7659 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7660 				k = fxdr_unsigned(int, *tl);
7661 				if (k < 1 || k > 128) {
7662 					error = NFSERR_BADXDR;
7663 					goto nfsmout;
7664 				}
7665 				NFSCL_DEBUG(4, "servercnt=%d\n", k);
7666 				for (l = 0; l < k; l++) {
7667 					NFSM_DISSECT(tl, uint32_t *,
7668 					    NFSX_V4DEVICEID + NFSX_STATEID +
7669 					    2 * NFSX_UNSIGNED);
7670 					if (l == 0) {
7671 						/* Just use the first server. */
7672 						NFSBCOPY(tl,
7673 						    flp->nfsfl_ffm[j].dev,
7674 						    NFSX_V4DEVICEID);
7675 						tl += (NFSX_V4DEVICEID /
7676 						    NFSX_UNSIGNED);
7677 						tl++;
7678 						flp->nfsfl_ffm[j].st.seqid =
7679 						    *tl++;
7680 						flp->nfsfl_ffm[j].st.other[0] =
7681 						    *tl++;
7682 						flp->nfsfl_ffm[j].st.other[1] =
7683 						    *tl++;
7684 						flp->nfsfl_ffm[j].st.other[2] =
7685 						    *tl++;
7686 						NFSCL_DEBUG(4, "st.seqid=%u "
7687 						 "st.o0=0x%x st.o1=0x%x "
7688 						 "st.o2=0x%x\n",
7689 						 flp->nfsfl_ffm[j].st.seqid,
7690 						 flp->nfsfl_ffm[j].st.other[0],
7691 						 flp->nfsfl_ffm[j].st.other[1],
7692 						 flp->nfsfl_ffm[j].st.other[2]);
7693 					} else
7694 						tl += ((NFSX_V4DEVICEID +
7695 						    NFSX_STATEID +
7696 						    NFSX_UNSIGNED) /
7697 						    NFSX_UNSIGNED);
7698 					fhcnt = fxdr_unsigned(int, *tl);
7699 					NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7700 					if (fhcnt < 1 ||
7701 					    fhcnt > NFSDEV_MAXVERS) {
7702 						error = NFSERR_BADXDR;
7703 						goto nfsmout;
7704 					}
7705 					for (m = 0; m < fhcnt; m++) {
7706 						NFSM_DISSECT(tl, uint32_t *,
7707 						    NFSX_UNSIGNED);
7708 						nfhlen = fxdr_unsigned(int,
7709 						    *tl);
7710 						NFSCL_DEBUG(4, "nfhlen=%d\n",
7711 						    nfhlen);
7712 						if (nfhlen <= 0 || nfhlen >
7713 						    NFSX_V4FHMAX) {
7714 							error = NFSERR_BADXDR;
7715 							goto nfsmout;
7716 						}
7717 						NFSM_DISSECT(cp, uint8_t *,
7718 						    NFSM_RNDUP(nfhlen));
7719 						if (l == 0) {
7720 							flp->nfsfl_ffm[j].fhcnt
7721 							    = fhcnt;
7722 							nfhp = malloc(
7723 							    sizeof(*nfhp) +
7724 							    nfhlen - 1, M_NFSFH,
7725 							    M_WAITOK);
7726 							flp->nfsfl_ffm[j].fh[m]
7727 							    = nfhp;
7728 							nfhp->nfh_len = nfhlen;
7729 							NFSBCOPY(cp,
7730 							    nfhp->nfh_fh,
7731 							    nfhlen);
7732 							NFSCL_DEBUG(4,
7733 							    "got fh\n");
7734 						}
7735 					}
7736 					/* Now, get the ffsd_user/ffds_group. */
7737 					error = nfsrv_parseug(nd, 0, &user,
7738 					    &grp, curthread);
7739 					NFSCL_DEBUG(4, "after parseu=%d\n",
7740 					    error);
7741 					if (error == 0)
7742 						error = nfsrv_parseug(nd, 1,
7743 						    &user, &grp, curthread);
7744 					NFSCL_DEBUG(4, "aft parseg=%d\n",
7745 					    grp);
7746 					if (error != 0)
7747 						goto nfsmout;
7748 					NFSCL_DEBUG(4, "user=%d group=%d\n",
7749 					    user, grp);
7750 					if (l == 0) {
7751 						flp->nfsfl_ffm[j].user = user;
7752 						flp->nfsfl_ffm[j].group = grp;
7753 						NFSCL_DEBUG(4,
7754 						    "usr=%d grp=%d\n", user,
7755 						    grp);
7756 					}
7757 				}
7758 			}
7759 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7760 			flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
7761 #ifdef notnow
7762 			/*
7763 			 * At this time, there is no flag.
7764 			 * NFSFLEXFLAG_IOADVISE_THRU_MDS might need to be
7765 			 * added, or it may never exist?
7766 			 */
7767 			mtx_lock(&nmp->nm_mtx);
7768 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_fflags &
7769 			    NFSFLEXFLAG_IOADVISE_THRU_MDS) != 0)
7770 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7771 			mtx_unlock(&nmp->nm_mtx);
7772 #endif
7773 			flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
7774 			NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
7775 			    flp->nfsfl_fflags, flp->nfsfl_statshint);
7776 		} else {
7777 			error = NFSERR_BADXDR;
7778 			goto nfsmout;
7779 		}
7780 		if (flp->nfsfl_iomode == gotiomode) {
7781 			/* Keep the list in increasing offset order. */
7782 			tflp = LIST_FIRST(flhp);
7783 			prevflp = NULL;
7784 			while (tflp != NULL &&
7785 			    tflp->nfsfl_off < flp->nfsfl_off) {
7786 				prevflp = tflp;
7787 				tflp = LIST_NEXT(tflp, nfsfl_list);
7788 			}
7789 			if (prevflp == NULL)
7790 				LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
7791 			else
7792 				LIST_INSERT_AFTER(prevflp, flp,
7793 				    nfsfl_list);
7794 			NFSCL_DEBUG(4, "flp inserted\n");
7795 		} else {
7796 			printf("nfscl_layoutget(): got wrong iomode\n");
7797 			nfscl_freeflayout(flp);
7798 		}
7799 		flp = NULL;
7800 	}
7801 nfsmout:
7802 	NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
7803 	if (error != 0 && flp != NULL)
7804 		nfscl_freeflayout(flp);
7805 	return (error);
7806 }
7807 
7808 /*
7809  * Parse a user/group digit string.
7810  */
7811 static int
7812 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
7813     NFSPROC_T *p)
7814 {
7815 	uint32_t *tl;
7816 	char *cp, *str, str0[NFSV4_SMALLSTR + 1];
7817 	uint32_t len = 0;
7818 	int error = 0;
7819 
7820 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7821 	len = fxdr_unsigned(uint32_t, *tl);
7822 	str = NULL;
7823 	if (len > NFSV4_OPAQUELIMIT) {
7824 		error = NFSERR_BADXDR;
7825 		goto nfsmout;
7826 	}
7827 	NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
7828 	if (len == 0) {
7829 		if (dogrp != 0)
7830 			*gidp = GID_NOGROUP;
7831 		else
7832 			*uidp = UID_NOBODY;
7833 		return (0);
7834 	}
7835 	if (len > NFSV4_SMALLSTR)
7836 		str = malloc(len + 1, M_TEMP, M_WAITOK);
7837 	else
7838 		str = str0;
7839 	NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7840 	NFSBCOPY(cp, str, len);
7841 	str[len] = '\0';
7842 	NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7843 	if (dogrp != 0)
7844 		error = nfsv4_strtogid(nd, str, len, gidp);
7845 	else
7846 		error = nfsv4_strtouid(nd, str, len, uidp);
7847 nfsmout:
7848 	if (len > NFSV4_SMALLSTR)
7849 		free(str, M_TEMP);
7850 	NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7851 	return (error);
7852 }
7853 
7854 /*
7855  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7856  * so that it does both an Open and a Layoutget.
7857  */
7858 static int
7859 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7860     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7861     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7862     struct ucred *cred, NFSPROC_T *p)
7863 {
7864 	struct nfscllayout *lyp;
7865 	struct nfsclflayout *flp;
7866 	struct nfsclflayouthead flh;
7867 	int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7868 	int layouttype, laystat;
7869 	nfsv4stateid_t stateid;
7870 	struct nfsclsession *tsep;
7871 
7872 	error = 0;
7873 	if (NFSHASFLEXFILE(nmp))
7874 		layouttype = NFSLAYOUT_FLEXFILE;
7875 	else
7876 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7877 	/*
7878 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7879 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7880 	 * flp == NULL.
7881 	 */
7882 	lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, mode, &flp,
7883 	    &recalled);
7884 	NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7885 	if (lyp == NULL)
7886 		islocked = 0;
7887 	else if (flp != NULL)
7888 		islocked = 1;
7889 	else
7890 		islocked = 2;
7891 	if ((lyp == NULL || flp == NULL) && recalled == 0) {
7892 		LIST_INIT(&flh);
7893 		tsep = nfsmnt_mdssession(nmp);
7894 		layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7895 		    3 * NFSX_UNSIGNED);
7896 		if (lyp == NULL)
7897 			usecurstateid = 1;
7898 		else {
7899 			usecurstateid = 0;
7900 			stateid.seqid = lyp->nfsly_stateid.seqid;
7901 			stateid.other[0] = lyp->nfsly_stateid.other[0];
7902 			stateid.other[1] = lyp->nfsly_stateid.other[1];
7903 			stateid.other[2] = lyp->nfsly_stateid.other[2];
7904 		}
7905 		error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7906 		    newfhp, newfhlen, mode, op, name, namelen,
7907 		    dpp, &stateid, usecurstateid, layouttype, layoutlen,
7908 		    &retonclose, &flh, &laystat, cred, p);
7909 		NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7910 		    laystat, error);
7911 		laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7912 		    &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7913 		    &islocked, cred, p);
7914 	} else
7915 		error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7916 		    mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7917 	if (islocked == 2)
7918 		nfscl_rellayout(lyp, 1);
7919 	else if (islocked == 1)
7920 		nfscl_rellayout(lyp, 0);
7921 	return (error);
7922 }
7923 
7924 /*
7925  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7926  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
7927  * handled by nfsrpc_openrpc().
7928  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
7929  * can be NULL.
7930  */
7931 static int
7932 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7933     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7934     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7935     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7936     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7937     int *laystatp, struct ucred *cred, NFSPROC_T *p)
7938 {
7939 	uint32_t *tl;
7940 	struct nfsrv_descript nfsd, *nd = &nfsd;
7941 	struct nfscldeleg *ndp = NULL;
7942 	struct nfsvattr nfsva;
7943 	struct nfsclsession *tsep;
7944 	uint32_t rflags, deleg;
7945 	nfsattrbit_t attrbits;
7946 	int error, ret, acesize, limitby, iomode;
7947 
7948 	*dpp = NULL;
7949 	*laystatp = ENXIO;
7950 	nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7951 	    0, 0);
7952 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7953 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7954 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7955 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7956 	tsep = nfsmnt_mdssession(nmp);
7957 	*tl++ = tsep->nfsess_clientid.lval[0];
7958 	*tl = tsep->nfsess_clientid.lval[1];
7959 	nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7960 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7961 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7962 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7963 	nfsm_strtom(nd, name, namelen);
7964 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7965 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7966 	NFSZERO_ATTRBIT(&attrbits);
7967 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7968 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7969 	nfsrv_putattrbit(nd, &attrbits);
7970 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7971 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7972 	if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7973 		iomode = NFSLAYOUTIOMODE_RW;
7974 	else
7975 		iomode = NFSLAYOUTIOMODE_READ;
7976 	nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7977 	    layouttype, layoutlen, usecurstateid);
7978 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7979 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7980 	if (error != 0)
7981 		return (error);
7982 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7983 	if (nd->nd_repstat != 0)
7984 		*laystatp = nd->nd_repstat;
7985 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7986 		/* ND_NOMOREDATA will be set if the Open operation failed. */
7987 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7988 		    6 * NFSX_UNSIGNED);
7989 		op->nfso_stateid.seqid = *tl++;
7990 		op->nfso_stateid.other[0] = *tl++;
7991 		op->nfso_stateid.other[1] = *tl++;
7992 		op->nfso_stateid.other[2] = *tl;
7993 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7994 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7995 		if (error != 0)
7996 			goto nfsmout;
7997 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7998 		deleg = fxdr_unsigned(u_int32_t, *tl);
7999 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
8000 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
8001 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
8002 			      NFSCLFLAGS_FIRSTDELEG))
8003 				op->nfso_own->nfsow_clp->nfsc_flags |=
8004 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
8005 			ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
8006 			    M_NFSCLDELEG, M_WAITOK);
8007 			LIST_INIT(&ndp->nfsdl_owner);
8008 			LIST_INIT(&ndp->nfsdl_lock);
8009 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
8010 			ndp->nfsdl_fhlen = newfhlen;
8011 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
8012 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
8013 			nfscl_lockinit(&ndp->nfsdl_rwlock);
8014 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8015 			    NFSX_UNSIGNED);
8016 			ndp->nfsdl_stateid.seqid = *tl++;
8017 			ndp->nfsdl_stateid.other[0] = *tl++;
8018 			ndp->nfsdl_stateid.other[1] = *tl++;
8019 			ndp->nfsdl_stateid.other[2] = *tl++;
8020 			ret = fxdr_unsigned(int, *tl);
8021 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
8022 				ndp->nfsdl_flags = NFSCLDL_WRITE;
8023 				/*
8024 				 * Indicates how much the file can grow.
8025 				 */
8026 				NFSM_DISSECT(tl, u_int32_t *,
8027 				    3 * NFSX_UNSIGNED);
8028 				limitby = fxdr_unsigned(int, *tl++);
8029 				switch (limitby) {
8030 				case NFSV4OPEN_LIMITSIZE:
8031 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
8032 					break;
8033 				case NFSV4OPEN_LIMITBLOCKS:
8034 					ndp->nfsdl_sizelimit =
8035 					    fxdr_unsigned(u_int64_t, *tl++);
8036 					ndp->nfsdl_sizelimit *=
8037 					    fxdr_unsigned(u_int64_t, *tl);
8038 					break;
8039 				default:
8040 					error = NFSERR_BADXDR;
8041 					goto nfsmout;
8042 				};
8043 			} else
8044 				ndp->nfsdl_flags = NFSCLDL_READ;
8045 			if (ret != 0)
8046 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
8047 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, false,
8048 			    &ret, &acesize, p);
8049 			if (error != 0)
8050 				goto nfsmout;
8051 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
8052 			error = NFSERR_BADXDR;
8053 			goto nfsmout;
8054 		}
8055 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
8056 		    nfscl_assumeposixlocks)
8057 			op->nfso_posixlock = 1;
8058 		else
8059 			op->nfso_posixlock = 0;
8060 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8061 		/* If the 2nd element == NFS_OK, the Getattr succeeded. */
8062 		if (*++tl == 0) {
8063 			error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
8064 			    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
8065 			    NULL, NULL, NULL, p, cred);
8066 			if (error != 0)
8067 				goto nfsmout;
8068 			if (ndp != NULL) {
8069 				ndp->nfsdl_change = nfsva.na_filerev;
8070 				ndp->nfsdl_modtime = nfsva.na_mtime;
8071 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
8072 				*dpp = ndp;
8073 				ndp = NULL;
8074 			}
8075 			/*
8076 			 * At this point, the Open has succeeded, so set
8077 			 * nd_repstat = NFS_OK.  If the Layoutget failed,
8078 			 * this function just won't return a layout.
8079 			 */
8080 			if (nd->nd_repstat == 0) {
8081 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8082 				*laystatp = fxdr_unsigned(int, *++tl);
8083 				if (*laystatp == 0) {
8084 					error = nfsrv_parselayoutget(nmp, nd,
8085 					    stateidp, retonclosep, flhp);
8086 					if (error != 0)
8087 						*laystatp = error;
8088 				}
8089 			} else
8090 				nd->nd_repstat = 0;	/* Return 0 for Open. */
8091 		}
8092 	}
8093 	if (nd->nd_repstat != 0 && error == 0)
8094 		error = nd->nd_repstat;
8095 nfsmout:
8096 	free(ndp, M_NFSCLDELEG);
8097 	m_freem(nd->nd_mrep);
8098 	return (error);
8099 }
8100 
8101 /*
8102  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
8103  * Used only for mounts with pNFS enabled.
8104  */
8105 static int
8106 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
8107     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
8108     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
8109     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
8110     int *dattrflagp, int *unlockedp, nfsv4stateid_t *stateidp,
8111     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
8112     struct nfsclflayouthead *flhp, int *laystatp)
8113 {
8114 	uint32_t *tl;
8115 	int error = 0, deleg, newone, ret, acesize, limitby;
8116 	struct nfsrv_descript nfsd, *nd = &nfsd;
8117 	struct nfsclopen *op;
8118 	struct nfscldeleg *dp = NULL;
8119 	struct nfsnode *np;
8120 	struct nfsfh *nfhp;
8121 	struct nfsclsession *tsep;
8122 	nfsattrbit_t attrbits;
8123 	nfsv4stateid_t stateid;
8124 	struct nfsmount *nmp;
8125 
8126 	nmp = VFSTONFS(dvp->v_mount);
8127 	np = VTONFS(dvp);
8128 	*laystatp = ENXIO;
8129 	*unlockedp = 0;
8130 	*nfhpp = NULL;
8131 	*dpp = NULL;
8132 	*attrflagp = 0;
8133 	*dattrflagp = 0;
8134 	if (namelen > NFS_MAXNAMLEN)
8135 		return (ENAMETOOLONG);
8136 	NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
8137 	/*
8138 	 * For V4, this is actually an Open op.
8139 	 */
8140 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
8141 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
8142 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
8143 	    NFSV4OPEN_ACCESSREAD);
8144 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
8145 	tsep = nfsmnt_mdssession(nmp);
8146 	*tl++ = tsep->nfsess_clientid.lval[0];
8147 	*tl = tsep->nfsess_clientid.lval[1];
8148 	nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
8149 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8150 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
8151 	if ((fmode & O_EXCL) != 0) {
8152 		if (NFSHASSESSPERSIST(nmp)) {
8153 			/* Use GUARDED for persistent sessions. */
8154 			*tl = txdr_unsigned(NFSCREATE_GUARDED);
8155 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
8156 		} else {
8157 			/* Otherwise, use EXCLUSIVE4_1. */
8158 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
8159 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
8160 			*tl++ = cverf.lval[0];
8161 			*tl = cverf.lval[1];
8162 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
8163 		}
8164 	} else {
8165 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
8166 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
8167 	}
8168 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8169 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
8170 	nfsm_strtom(nd, name, namelen);
8171 	/* Get the new file's handle and attributes, plus save the FH. */
8172 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
8173 	*tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
8174 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
8175 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8176 	NFSGETATTR_ATTRBIT(&attrbits);
8177 	nfsrv_putattrbit(nd, &attrbits);
8178 	/* Get the directory's post-op attributes. */
8179 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8180 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8181 	nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
8182 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8183 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8184 	nfsrv_putattrbit(nd, &attrbits);
8185 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8186 	*tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
8187 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
8188 	nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
8189 	    layouttype, layoutlen, usecurstateid);
8190 	error = nfscl_request(nd, dvp, p, cred);
8191 	if (error != 0)
8192 		return (error);
8193 	NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
8194 	    error);
8195 	if (nd->nd_repstat != 0)
8196 		*laystatp = nd->nd_repstat;
8197 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
8198 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8199 		NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
8200 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8201 		    6 * NFSX_UNSIGNED);
8202 		stateid.seqid = *tl++;
8203 		stateid.other[0] = *tl++;
8204 		stateid.other[1] = *tl++;
8205 		stateid.other[2] = *tl;
8206 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
8207 		if (error != 0)
8208 			goto nfsmout;
8209 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
8210 		deleg = fxdr_unsigned(int, *tl);
8211 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
8212 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
8213 			if (!(owp->nfsow_clp->nfsc_flags &
8214 			      NFSCLFLAGS_FIRSTDELEG))
8215 				owp->nfsow_clp->nfsc_flags |=
8216 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
8217 			dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
8218 			    M_NFSCLDELEG, M_WAITOK);
8219 			LIST_INIT(&dp->nfsdl_owner);
8220 			LIST_INIT(&dp->nfsdl_lock);
8221 			dp->nfsdl_clp = owp->nfsow_clp;
8222 			newnfs_copyincred(cred, &dp->nfsdl_cred);
8223 			nfscl_lockinit(&dp->nfsdl_rwlock);
8224 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8225 			    NFSX_UNSIGNED);
8226 			dp->nfsdl_stateid.seqid = *tl++;
8227 			dp->nfsdl_stateid.other[0] = *tl++;
8228 			dp->nfsdl_stateid.other[1] = *tl++;
8229 			dp->nfsdl_stateid.other[2] = *tl++;
8230 			ret = fxdr_unsigned(int, *tl);
8231 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
8232 				dp->nfsdl_flags = NFSCLDL_WRITE;
8233 				/*
8234 				 * Indicates how much the file can grow.
8235 				 */
8236 				NFSM_DISSECT(tl, u_int32_t *,
8237 				    3 * NFSX_UNSIGNED);
8238 				limitby = fxdr_unsigned(int, *tl++);
8239 				switch (limitby) {
8240 				case NFSV4OPEN_LIMITSIZE:
8241 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
8242 					break;
8243 				case NFSV4OPEN_LIMITBLOCKS:
8244 					dp->nfsdl_sizelimit =
8245 					    fxdr_unsigned(u_int64_t, *tl++);
8246 					dp->nfsdl_sizelimit *=
8247 					    fxdr_unsigned(u_int64_t, *tl);
8248 					break;
8249 				default:
8250 					error = NFSERR_BADXDR;
8251 					goto nfsmout;
8252 				};
8253 			} else {
8254 				dp->nfsdl_flags = NFSCLDL_READ;
8255 			}
8256 			if (ret != 0)
8257 				dp->nfsdl_flags |= NFSCLDL_RECALL;
8258 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, false,
8259 			    &ret, &acesize, p);
8260 			if (error != 0)
8261 				goto nfsmout;
8262 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
8263 			error = NFSERR_BADXDR;
8264 			goto nfsmout;
8265 		}
8266 
8267 		/* Now, we should have the status for the SaveFH. */
8268 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8269 		if (*++tl == 0) {
8270 			NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
8271 			/*
8272 			 * Now, process the GetFH and Getattr for the newly
8273 			 * created file. nfscl_mtofh() will set
8274 			 * ND_NOMOREDATA if these weren't successful.
8275 			 */
8276 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
8277 			NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
8278 			if (error != 0)
8279 				goto nfsmout;
8280 		} else
8281 			nd->nd_flag |= ND_NOMOREDATA;
8282 		/* Now we have the PutFH and Getattr for the directory. */
8283 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8284 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8285 			if (*++tl != 0)
8286 				nd->nd_flag |= ND_NOMOREDATA;
8287 			else {
8288 				NFSM_DISSECT(tl, uint32_t *, 2 *
8289 				    NFSX_UNSIGNED);
8290 				if (*++tl != 0)
8291 					nd->nd_flag |= ND_NOMOREDATA;
8292 			}
8293 		}
8294 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8295 			/* Load the directory attributes. */
8296 			error = nfsm_loadattr(nd, dnap);
8297 			NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
8298 			if (error != 0)
8299 				goto nfsmout;
8300 			*dattrflagp = 1;
8301 			if (dp != NULL && *attrflagp != 0) {
8302 				dp->nfsdl_change = nnap->na_filerev;
8303 				dp->nfsdl_modtime = nnap->na_mtime;
8304 				dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
8305 			}
8306 			/*
8307 			 * We can now complete the Open state.
8308 			 */
8309 			nfhp = *nfhpp;
8310 			if (dp != NULL) {
8311 				dp->nfsdl_fhlen = nfhp->nfh_len;
8312 				NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
8313 				    nfhp->nfh_len);
8314 			}
8315 			/*
8316 			 * Get an Open structure that will be
8317 			 * attached to the OpenOwner, acquired already.
8318 			 */
8319 			error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
8320 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
8321 			    cred, p, NULL, &op, &newone, NULL, 0, false);
8322 			if (error != 0)
8323 				goto nfsmout;
8324 			op->nfso_stateid = stateid;
8325 			newnfs_copyincred(cred, &op->nfso_cred);
8326 
8327 			nfscl_openrelease(nmp, op, error, newone);
8328 			*unlockedp = 1;
8329 
8330 			/* Now, handle the RestoreFH and LayoutGet. */
8331 			if (nd->nd_repstat == 0) {
8332 				NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
8333 				*laystatp = fxdr_unsigned(int, *(tl + 3));
8334 				if (*laystatp == 0) {
8335 					error = nfsrv_parselayoutget(nmp, nd,
8336 					    stateidp, retonclosep, flhp);
8337 					if (error != 0)
8338 						*laystatp = error;
8339 				}
8340 				NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
8341 				    error);
8342 			} else
8343 				nd->nd_repstat = 0;
8344 		}
8345 	}
8346 	if (nd->nd_repstat != 0 && error == 0)
8347 		error = nd->nd_repstat;
8348 	if (error == NFSERR_STALECLIENTID)
8349 		nfscl_initiate_recovery(owp->nfsow_clp);
8350 nfsmout:
8351 	NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
8352 	if (error == 0)
8353 		*dpp = dp;
8354 	else
8355 		free(dp, M_NFSCLDELEG);
8356 	m_freem(nd->nd_mrep);
8357 	return (error);
8358 }
8359 
8360 /*
8361  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
8362  */
8363 static int
8364 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
8365     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
8366     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
8367     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
8368     int *dattrflagp, int *unlockedp)
8369 {
8370 	struct nfscllayout *lyp;
8371 	struct nfsclflayouthead flh;
8372 	struct nfsfh *nfhp;
8373 	struct nfsclsession *tsep;
8374 	struct nfsmount *nmp;
8375 	nfsv4stateid_t stateid;
8376 	int error, layoutlen, layouttype, retonclose, laystat;
8377 
8378 	error = 0;
8379 	nmp = VFSTONFS(dvp->v_mount);
8380 	if (NFSHASFLEXFILE(nmp))
8381 		layouttype = NFSLAYOUT_FLEXFILE;
8382 	else
8383 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
8384 	LIST_INIT(&flh);
8385 	tsep = nfsmnt_mdssession(nmp);
8386 	layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
8387 	error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
8388 	    owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
8389 	    unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
8390 	    &flh, &laystat);
8391 	NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
8392 	    laystat, error);
8393 	lyp = NULL;
8394 	if (laystat == 0) {
8395 		nfhp = *nfhpp;
8396 		laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
8397 		    nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
8398 		    layouttype, laystat, NULL, cred, p);
8399 	} else
8400 		laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
8401 		    retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
8402 		    cred, p);
8403 	if (laystat == 0)
8404 		nfscl_rellayout(lyp, 0);
8405 	return (error);
8406 }
8407 
8408 /*
8409  * Process the results of a layoutget() operation.
8410  */
8411 static int
8412 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
8413     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
8414     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
8415     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
8416 {
8417 	struct nfsclflayout *tflp;
8418 	struct nfscldevinfo *dip;
8419 	uint8_t *dev;
8420 	int i, mirrorcnt;
8421 
8422 	if (laystat == NFSERR_UNKNLAYOUTTYPE) {
8423 		NFSLOCKMNT(nmp);
8424 		if (!NFSHASFLEXFILE(nmp)) {
8425 			/* Switch to using Flex File Layout. */
8426 			nmp->nm_state |= NFSSTA_FLEXFILE;
8427 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
8428 			/* Disable pNFS. */
8429 			NFSCL_DEBUG(1, "disable PNFS\n");
8430 			nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
8431 		}
8432 		NFSUNLOCKMNT(nmp);
8433 	}
8434 	if (laystat == 0) {
8435 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
8436 		LIST_FOREACH(tflp, flhp, nfsfl_list) {
8437 			if (layouttype == NFSLAYOUT_FLEXFILE)
8438 				mirrorcnt = tflp->nfsfl_mirrorcnt;
8439 			else
8440 				mirrorcnt = 1;
8441 			for (i = 0; i < mirrorcnt; i++) {
8442 				laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
8443 				NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
8444 				if (laystat != 0) {
8445 					if (layouttype == NFSLAYOUT_FLEXFILE)
8446 						dev = tflp->nfsfl_ffm[i].dev;
8447 					else
8448 						dev = tflp->nfsfl_dev;
8449 					laystat = nfsrpc_getdeviceinfo(nmp, dev,
8450 					    layouttype, notifybit, &dip, cred,
8451 					    p);
8452 					NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
8453 					    laystat);
8454 					if (laystat != 0)
8455 						goto out;
8456 					laystat = nfscl_adddevinfo(nmp, dip, i,
8457 					    tflp);
8458 					if (laystat != 0)
8459 						printf("nfsrpc_layoutgetresout"
8460 						    ": cannot add\n");
8461 				}
8462 			}
8463 		}
8464 	}
8465 out:
8466 	if (laystat == 0) {
8467 		/*
8468 		 * nfscl_layout() always returns with the nfsly_lock
8469 		 * set to a refcnt (shared lock).
8470 		 * Passing in dvp is sufficient, since it is only used to
8471 		 * get the fsid for the file system.
8472 		 */
8473 		laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
8474 		    layouttype, retonclose, flhp, lypp, cred, p);
8475 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
8476 		    laystat);
8477 		if (laystat == 0 && islockedp != NULL)
8478 			*islockedp = 1;
8479 	}
8480 	return (laystat);
8481 }
8482 
8483 /*
8484  * nfs copy_file_range operation.
8485  */
8486 int
8487 nfsrpc_copy_file_range(vnode_t invp, off_t *inoffp, vnode_t outvp,
8488     off_t *outoffp, size_t *lenp, unsigned int flags, int *inattrflagp,
8489     struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap,
8490     struct ucred *cred, bool consecutive, bool *must_commitp)
8491 {
8492 	int commit, error, expireret = 0, retrycnt;
8493 	u_int32_t clidrev = 0;
8494 	struct nfsmount *nmp = VFSTONFS(invp->v_mount);
8495 	struct nfsfh *innfhp = NULL, *outnfhp = NULL;
8496 	nfsv4stateid_t instateid, outstateid;
8497 	void *inlckp, *outlckp;
8498 
8499 	if (nmp->nm_clp != NULL)
8500 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8501 	innfhp = VTONFS(invp)->n_fhp;
8502 	outnfhp = VTONFS(outvp)->n_fhp;
8503 	retrycnt = 0;
8504 	do {
8505 		/* Get both stateids. */
8506 		inlckp = NULL;
8507 		nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len,
8508 		    NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid,
8509 		    &inlckp);
8510 		outlckp = NULL;
8511 		nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len,
8512 		    NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid,
8513 		    &outlckp);
8514 
8515 		error = nfsrpc_copyrpc(invp, *inoffp, outvp, *outoffp, lenp,
8516 		    &instateid, &outstateid, innap, inattrflagp, outnap,
8517 		    outattrflagp, consecutive, &commit, cred, curthread);
8518 		if (error == 0) {
8519 			if (commit != NFSWRITE_FILESYNC)
8520 				*must_commitp = true;
8521 			*inoffp += *lenp;
8522 			*outoffp += *lenp;
8523 		} else if (error == NFSERR_STALESTATEID)
8524 			nfscl_initiate_recovery(nmp->nm_clp);
8525 		if (inlckp != NULL)
8526 			nfscl_lockderef(inlckp);
8527 		if (outlckp != NULL)
8528 			nfscl_lockderef(outlckp);
8529 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8530 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8531 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8532 			(void) nfs_catnap(PZERO, error, "nfs_cfr");
8533 		} else if ((error == NFSERR_EXPIRED ||
8534 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8535 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8536 			    curthread);
8537 		}
8538 		retrycnt++;
8539 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
8540 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8541 	      error == NFSERR_STALEDONTRECOVER ||
8542 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8543 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8544 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
8545 	if (error != 0 && (retrycnt >= 4 ||
8546 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8547 	      error == NFSERR_STALEDONTRECOVER))
8548 		error = EIO;
8549 	return (error);
8550 }
8551 
8552 /*
8553  * The copy RPC.
8554  */
8555 static int
8556 nfsrpc_copyrpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff,
8557     size_t *lenp, nfsv4stateid_t *instateidp, nfsv4stateid_t *outstateidp,
8558     struct nfsvattr *innap, int *inattrflagp, struct nfsvattr *outnap,
8559     int *outattrflagp, bool consecutive, int *commitp, struct ucred *cred,
8560     NFSPROC_T *p)
8561 {
8562 	uint32_t *tl;
8563 	int error;
8564 	struct nfsrv_descript nfsd;
8565 	struct nfsrv_descript *nd = &nfsd;
8566 	struct nfsmount *nmp;
8567 	nfsattrbit_t attrbits;
8568 	uint64_t len;
8569 
8570 	nmp = VFSTONFS(outvp->v_mount);
8571 	*inattrflagp = *outattrflagp = 0;
8572 	*commitp = NFSWRITE_UNSTABLE;
8573 	len = *lenp;
8574 	*lenp = 0;
8575 	if (len > nfs_maxcopyrange)
8576 		len = nfs_maxcopyrange;
8577 	NFSCL_REQSTART(nd, NFSPROC_COPY, invp);
8578 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8579 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8580 	NFSGETATTR_ATTRBIT(&attrbits);
8581 	nfsrv_putattrbit(nd, &attrbits);
8582 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8583 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8584 	nfsm_fhtom(nd, VTONFS(outvp)->n_fhp->nfh_fh,
8585 	    VTONFS(outvp)->n_fhp->nfh_len, 0);
8586 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8587 	*tl = txdr_unsigned(NFSV4OP_COPY);
8588 	nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
8589 	nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID);
8590 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + 4 * NFSX_UNSIGNED);
8591 	txdr_hyper(inoff, tl); tl += 2;
8592 	txdr_hyper(outoff, tl); tl += 2;
8593 	txdr_hyper(len, tl); tl += 2;
8594 	if (consecutive)
8595 		*tl++ = newnfs_true;
8596 	else
8597 		*tl++ = newnfs_false;
8598 	*tl++ = newnfs_true;
8599 	*tl++ = 0;
8600 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8601 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
8602 	nfsrv_putattrbit(nd, &attrbits);
8603 	error = nfscl_request(nd, invp, p, cred);
8604 	if (error != 0)
8605 		return (error);
8606 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8607 		/* Get the input file's attributes. */
8608 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8609 		if (*(tl + 1) == 0) {
8610 			error = nfsm_loadattr(nd, innap);
8611 			if (error != 0)
8612 				goto nfsmout;
8613 			*inattrflagp = 1;
8614 		} else
8615 			nd->nd_flag |= ND_NOMOREDATA;
8616 	}
8617 	/* Skip over return stat for PutFH. */
8618 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8619 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8620 		if (*++tl != 0)
8621 			nd->nd_flag |= ND_NOMOREDATA;
8622 	}
8623 	/* Skip over return stat for Copy. */
8624 	if ((nd->nd_flag & ND_NOMOREDATA) == 0)
8625 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8626 	if (nd->nd_repstat == 0) {
8627 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8628 		if (*tl != 0) {
8629 			/* There should be no callback ids. */
8630 			error = NFSERR_BADXDR;
8631 			goto nfsmout;
8632 		}
8633 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED +
8634 		    NFSX_VERF);
8635 		len = fxdr_hyper(tl); tl += 2;
8636 		*commitp = fxdr_unsigned(int, *tl++);
8637 		NFSLOCKMNT(nmp);
8638 		if (!NFSHASWRITEVERF(nmp)) {
8639 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8640 			NFSSETWRITEVERF(nmp);
8641 	    	} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
8642 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8643 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
8644 		}
8645 		NFSUNLOCKMNT(nmp);
8646 		tl += (NFSX_VERF / NFSX_UNSIGNED);
8647 		if (nd->nd_repstat == 0 && *++tl != newnfs_true)
8648 			/* Must be a synchronous copy. */
8649 			nd->nd_repstat = NFSERR_NOTSUPP;
8650 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8651 		error = nfsm_loadattr(nd, outnap);
8652 		if (error == 0)
8653 			*outattrflagp = NFS_LATTR_NOSHRINK;
8654 		if (nd->nd_repstat == 0)
8655 			*lenp = len;
8656 	} else if (nd->nd_repstat == NFSERR_OFFLOADNOREQS) {
8657 		/*
8658 		 * For the case where consecutive is not supported, but
8659 		 * synchronous is supported, we can try consecutive == false
8660 		 * by returning this error.  Otherwise, return NFSERR_NOTSUPP,
8661 		 * since Copy cannot be done.
8662 		 */
8663 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8664 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8665 			if (!consecutive || *++tl == newnfs_false)
8666 				nd->nd_repstat = NFSERR_NOTSUPP;
8667 		} else
8668 			nd->nd_repstat = NFSERR_BADXDR;
8669 	}
8670 	if (error == 0)
8671 		error = nd->nd_repstat;
8672 nfsmout:
8673 	m_freem(nd->nd_mrep);
8674 	return (error);
8675 }
8676 
8677 /*
8678  * Seek operation.
8679  */
8680 int
8681 nfsrpc_seek(vnode_t vp, off_t *offp, bool *eofp, int content,
8682     struct ucred *cred, struct nfsvattr *nap, int *attrflagp)
8683 {
8684 	int error, expireret = 0, retrycnt;
8685 	u_int32_t clidrev = 0;
8686 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
8687 	struct nfsnode *np = VTONFS(vp);
8688 	struct nfsfh *nfhp = NULL;
8689 	nfsv4stateid_t stateid;
8690 	void *lckp;
8691 
8692 	if (nmp->nm_clp != NULL)
8693 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8694 	nfhp = np->n_fhp;
8695 	retrycnt = 0;
8696 	do {
8697 		lckp = NULL;
8698 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
8699 		    NFSV4OPEN_ACCESSREAD, 0, cred, curthread, &stateid, &lckp);
8700 		error = nfsrpc_seekrpc(vp, offp, &stateid, eofp, content,
8701 		    nap, attrflagp, cred);
8702 		if (error == NFSERR_STALESTATEID)
8703 			nfscl_initiate_recovery(nmp->nm_clp);
8704 		if (lckp != NULL)
8705 			nfscl_lockderef(lckp);
8706 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8707 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8708 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8709 			(void) nfs_catnap(PZERO, error, "nfs_seek");
8710 		} else if ((error == NFSERR_EXPIRED ||
8711 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8712 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8713 			    curthread);
8714 		}
8715 		retrycnt++;
8716 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8717 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8718 	    error == NFSERR_BADSESSION ||
8719 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8720 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8721 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
8722 	    (error == NFSERR_OPENMODE && retrycnt < 4));
8723 	if (error && retrycnt >= 4)
8724 		error = EIO;
8725 	return (error);
8726 }
8727 
8728 /*
8729  * The seek RPC.
8730  */
8731 static int
8732 nfsrpc_seekrpc(vnode_t vp, off_t *offp, nfsv4stateid_t *stateidp, bool *eofp,
8733     int content, struct nfsvattr *nap, int *attrflagp, struct ucred *cred)
8734 {
8735 	uint32_t *tl;
8736 	int error;
8737 	struct nfsrv_descript nfsd;
8738 	struct nfsrv_descript *nd = &nfsd;
8739 	nfsattrbit_t attrbits;
8740 
8741 	*attrflagp = 0;
8742 	NFSCL_REQSTART(nd, NFSPROC_SEEK, vp);
8743 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
8744 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8745 	txdr_hyper(*offp, tl); tl += 2;
8746 	*tl++ = txdr_unsigned(content);
8747 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8748 	NFSGETATTR_ATTRBIT(&attrbits);
8749 	nfsrv_putattrbit(nd, &attrbits);
8750 	error = nfscl_request(nd, vp, curthread, cred);
8751 	if (error != 0)
8752 		return (error);
8753 	if (nd->nd_repstat == 0) {
8754 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED + NFSX_HYPER);
8755 		if (*tl++ == newnfs_true)
8756 			*eofp = true;
8757 		else
8758 			*eofp = false;
8759 		*offp = fxdr_hyper(tl);
8760 		/* Just skip over Getattr op status. */
8761 		error = nfsm_loadattr(nd, nap);
8762 		if (error == 0)
8763 			*attrflagp = 1;
8764 	}
8765 	error = nd->nd_repstat;
8766 nfsmout:
8767 	m_freem(nd->nd_mrep);
8768 	return (error);
8769 }
8770 
8771 /*
8772  * The getextattr RPC.
8773  */
8774 int
8775 nfsrpc_getextattr(vnode_t vp, const char *name, struct uio *uiop, ssize_t *lenp,
8776     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8777 {
8778 	uint32_t *tl;
8779 	int error;
8780 	struct nfsrv_descript nfsd;
8781 	struct nfsrv_descript *nd = &nfsd;
8782 	nfsattrbit_t attrbits;
8783 	uint32_t len, len2;
8784 
8785 	*attrflagp = 0;
8786 	NFSCL_REQSTART(nd, NFSPROC_GETEXTATTR, vp);
8787 	nfsm_strtom(nd, name, strlen(name));
8788 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8789 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8790 	NFSGETATTR_ATTRBIT(&attrbits);
8791 	nfsrv_putattrbit(nd, &attrbits);
8792 	error = nfscl_request(nd, vp, p, cred);
8793 	if (error != 0)
8794 		return (error);
8795 	if (nd->nd_repstat == 0) {
8796 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8797 		len = fxdr_unsigned(uint32_t, *tl);
8798 		/* Sanity check lengths. */
8799 		if (uiop != NULL && len > 0 && len <= IOSIZE_MAX &&
8800 		    uiop->uio_resid <= UINT32_MAX) {
8801 			len2 = uiop->uio_resid;
8802 			if (len2 >= len)
8803 				error = nfsm_mbufuio(nd, uiop, len);
8804 			else {
8805 				error = nfsm_mbufuio(nd, uiop, len2);
8806 				if (error == 0) {
8807 					/*
8808 					 * nfsm_mbufuio() advances to a multiple
8809 					 * of 4, so round up len2 as well.  Then
8810 					 * we need to advance over the rest of
8811 					 * the data, rounding up the remaining
8812 					 * length.
8813 					 */
8814 					len2 = NFSM_RNDUP(len2);
8815 					len2 = NFSM_RNDUP(len - len2);
8816 					if (len2 > 0)
8817 						error = nfsm_advance(nd, len2,
8818 						    -1);
8819 				}
8820 			}
8821 		} else if (uiop == NULL && len > 0) {
8822 			/* Just wants the length and not the data. */
8823 			error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8824 		} else if (len > 0)
8825 			error = ENOATTR;
8826 		if (error != 0)
8827 			goto nfsmout;
8828 		*lenp = len;
8829 		/* Just skip over Getattr op status. */
8830 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8831 		error = nfsm_loadattr(nd, nap);
8832 		if (error == 0)
8833 			*attrflagp = 1;
8834 	}
8835 	if (error == 0)
8836 		error = nd->nd_repstat;
8837 nfsmout:
8838 	m_freem(nd->nd_mrep);
8839 	return (error);
8840 }
8841 
8842 /*
8843  * The setextattr RPC.
8844  */
8845 int
8846 nfsrpc_setextattr(vnode_t vp, const char *name, struct uio *uiop,
8847     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8848 {
8849 	uint32_t *tl;
8850 	int error;
8851 	struct nfsrv_descript nfsd;
8852 	struct nfsrv_descript *nd = &nfsd;
8853 	nfsattrbit_t attrbits;
8854 
8855 	*attrflagp = 0;
8856 	NFSCL_REQSTART(nd, NFSPROC_SETEXTATTR, vp);
8857 	if (uiop->uio_resid > nd->nd_maxreq) {
8858 		/* nd_maxreq is set by NFSCL_REQSTART(). */
8859 		m_freem(nd->nd_mreq);
8860 		return (EINVAL);
8861 	}
8862 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8863 	*tl = txdr_unsigned(NFSV4SXATTR_EITHER);
8864 	nfsm_strtom(nd, name, strlen(name));
8865 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8866 	*tl = txdr_unsigned(uiop->uio_resid);
8867 	nfsm_uiombuf(nd, uiop, uiop->uio_resid);
8868 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8869 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8870 	NFSGETATTR_ATTRBIT(&attrbits);
8871 	nfsrv_putattrbit(nd, &attrbits);
8872 	error = nfscl_request(nd, vp, p, cred);
8873 	if (error != 0)
8874 		return (error);
8875 	if (nd->nd_repstat == 0) {
8876 		/* Just skip over the reply and Getattr op status. */
8877 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8878 		    NFSX_UNSIGNED);
8879 		error = nfsm_loadattr(nd, nap);
8880 		if (error == 0)
8881 			*attrflagp = 1;
8882 	}
8883 	if (error == 0)
8884 		error = nd->nd_repstat;
8885 nfsmout:
8886 	m_freem(nd->nd_mrep);
8887 	return (error);
8888 }
8889 
8890 /*
8891  * The removeextattr RPC.
8892  */
8893 int
8894 nfsrpc_rmextattr(vnode_t vp, const char *name, struct nfsvattr *nap,
8895     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8896 {
8897 	uint32_t *tl;
8898 	int error;
8899 	struct nfsrv_descript nfsd;
8900 	struct nfsrv_descript *nd = &nfsd;
8901 	nfsattrbit_t attrbits;
8902 
8903 	*attrflagp = 0;
8904 	NFSCL_REQSTART(nd, NFSPROC_RMEXTATTR, vp);
8905 	nfsm_strtom(nd, name, strlen(name));
8906 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8907 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8908 	NFSGETATTR_ATTRBIT(&attrbits);
8909 	nfsrv_putattrbit(nd, &attrbits);
8910 	error = nfscl_request(nd, vp, p, cred);
8911 	if (error != 0)
8912 		return (error);
8913 	if (nd->nd_repstat == 0) {
8914 		/* Just skip over the reply and Getattr op status. */
8915 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8916 		    NFSX_UNSIGNED);
8917 		error = nfsm_loadattr(nd, nap);
8918 		if (error == 0)
8919 			*attrflagp = 1;
8920 	}
8921 	if (error == 0)
8922 		error = nd->nd_repstat;
8923 nfsmout:
8924 	m_freem(nd->nd_mrep);
8925 	return (error);
8926 }
8927 
8928 /*
8929  * The listextattr RPC.
8930  */
8931 int
8932 nfsrpc_listextattr(vnode_t vp, uint64_t *cookiep, struct uio *uiop,
8933     size_t *lenp, bool *eofp, struct nfsvattr *nap, int *attrflagp,
8934     struct ucred *cred, NFSPROC_T *p)
8935 {
8936 	uint32_t *tl;
8937 	int cnt, error, i, len;
8938 	struct nfsrv_descript nfsd;
8939 	struct nfsrv_descript *nd = &nfsd;
8940 	nfsattrbit_t attrbits;
8941 	u_char c;
8942 
8943 	*attrflagp = 0;
8944 	NFSCL_REQSTART(nd, NFSPROC_LISTEXTATTR, vp);
8945 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8946 	txdr_hyper(*cookiep, tl); tl += 2;
8947 	*tl++ = txdr_unsigned(*lenp);
8948 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8949 	NFSGETATTR_ATTRBIT(&attrbits);
8950 	nfsrv_putattrbit(nd, &attrbits);
8951 	error = nfscl_request(nd, vp, p, cred);
8952 	if (error != 0)
8953 		return (error);
8954 	*eofp = true;
8955 	*lenp = 0;
8956 	if (nd->nd_repstat == 0) {
8957 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
8958 		*cookiep = fxdr_hyper(tl); tl += 2;
8959 		cnt = fxdr_unsigned(int, *tl);
8960 		if (cnt < 0) {
8961 			error = EBADRPC;
8962 			goto nfsmout;
8963 		}
8964 		for (i = 0; i < cnt; i++) {
8965 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8966 			len = fxdr_unsigned(int, *tl);
8967 			if (len <= 0 || len > EXTATTR_MAXNAMELEN) {
8968 				error = EBADRPC;
8969 				goto nfsmout;
8970 			}
8971 			if (uiop == NULL)
8972 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8973 			else if (uiop->uio_resid >= len + 1) {
8974 				c = len;
8975 				error = uiomove(&c, sizeof(c), uiop);
8976 				if (error == 0)
8977 					error = nfsm_mbufuio(nd, uiop, len);
8978 			} else {
8979 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8980 				*eofp = false;
8981 			}
8982 			if (error != 0)
8983 				goto nfsmout;
8984 			*lenp += (len + 1);
8985 		}
8986 		/* Get the eof and skip over the Getattr op status. */
8987 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
8988 		/*
8989 		 * *eofp is set false above, because it wasn't able to copy
8990 		 * all of the reply.
8991 		 */
8992 		if (*eofp && *tl == 0)
8993 			*eofp = false;
8994 		error = nfsm_loadattr(nd, nap);
8995 		if (error == 0)
8996 			*attrflagp = 1;
8997 	}
8998 	if (error == 0)
8999 		error = nd->nd_repstat;
9000 nfsmout:
9001 	m_freem(nd->nd_mrep);
9002 	return (error);
9003 }
9004 
9005 /*
9006  * Split an mbuf list.  For non-M_EXTPG mbufs, just use m_split().
9007  */
9008 static struct mbuf *
9009 nfsm_split(struct mbuf *mp, uint64_t xfer)
9010 {
9011 	struct mbuf *m, *m2;
9012 	vm_page_t pg;
9013 	int i, j, left, pgno, plen, trim;
9014 	char *cp, *cp2;
9015 
9016 	if ((mp->m_flags & M_EXTPG) == 0) {
9017 		m = m_split(mp, xfer, M_WAITOK);
9018 		return (m);
9019 	}
9020 
9021 	/* Find the correct mbuf to split at. */
9022 	for (m = mp; m != NULL && xfer > m->m_len; m = m->m_next)
9023 		xfer -= m->m_len;
9024 	if (m == NULL)
9025 		return (NULL);
9026 
9027 	/* If xfer == m->m_len, we can just split the mbuf list. */
9028 	if (xfer == m->m_len) {
9029 		m2 = m->m_next;
9030 		m->m_next = NULL;
9031 		return (m2);
9032 	}
9033 
9034 	/* Find the page to split at. */
9035 	pgno = 0;
9036 	left = xfer;
9037 	do {
9038 		if (pgno == 0)
9039 			plen = m_epg_pagelen(m, 0, m->m_epg_1st_off);
9040 		else
9041 			plen = m_epg_pagelen(m, pgno, 0);
9042 		if (left <= plen)
9043 			break;
9044 		left -= plen;
9045 		pgno++;
9046 	} while (pgno < m->m_epg_npgs);
9047 	if (pgno == m->m_epg_npgs)
9048 		panic("nfsm_split: eroneous ext_pgs mbuf");
9049 
9050 	m2 = mb_alloc_ext_pgs(M_WAITOK, mb_free_mext_pgs);
9051 	m2->m_epg_flags |= EPG_FLAG_ANON;
9052 
9053 	/*
9054 	 * If left < plen, allocate a new page for the new mbuf
9055 	 * and copy the data after left in the page to this new
9056 	 * page.
9057 	 */
9058 	if (left < plen) {
9059 		pg = vm_page_alloc_noobj(VM_ALLOC_WAITOK | VM_ALLOC_NODUMP |
9060 		    VM_ALLOC_WIRED);
9061 		m2->m_epg_pa[0] = VM_PAGE_TO_PHYS(pg);
9062 		m2->m_epg_npgs = 1;
9063 
9064 		/* Copy the data after left to the new page. */
9065 		trim = plen - left;
9066 		cp = (char *)(void *)PHYS_TO_DMAP(m->m_epg_pa[pgno]);
9067 		if (pgno == 0)
9068 			cp += m->m_epg_1st_off;
9069 		cp += left;
9070 		cp2 = (char *)(void *)PHYS_TO_DMAP(m2->m_epg_pa[0]);
9071 		if (pgno == m->m_epg_npgs - 1)
9072 			m2->m_epg_last_len = trim;
9073 		else {
9074 			cp2 += PAGE_SIZE - trim;
9075 			m2->m_epg_1st_off = PAGE_SIZE - trim;
9076 			m2->m_epg_last_len = m->m_epg_last_len;
9077 		}
9078 		memcpy(cp2, cp, trim);
9079 		m2->m_len = trim;
9080 	} else {
9081 		m2->m_len = 0;
9082 		m2->m_epg_last_len = m->m_epg_last_len;
9083 	}
9084 
9085 	/* Move the pages beyond pgno to the new mbuf. */
9086 	for (i = pgno + 1, j = m2->m_epg_npgs; i < m->m_epg_npgs; i++, j++) {
9087 		m2->m_epg_pa[j] = m->m_epg_pa[i];
9088 		/* Never moves page 0. */
9089 		m2->m_len += m_epg_pagelen(m, i, 0);
9090 	}
9091 	m2->m_epg_npgs = j;
9092 	m->m_epg_npgs = pgno + 1;
9093 	m->m_epg_last_len = left;
9094 	m->m_len = xfer;
9095 
9096 	m2->m_next = m->m_next;
9097 	m->m_next = NULL;
9098 	return (m2);
9099 }
9100 
9101 /*
9102  * Do the NFSv4.1 Bind Connection to Session.
9103  * Called from the reconnect layer of the krpc (sys/rpc/clnt_rc.c).
9104  */
9105 void
9106 nfsrpc_bindconnsess(CLIENT *cl, void *arg, struct ucred *cr)
9107 {
9108 	struct nfscl_reconarg *rcp = (struct nfscl_reconarg *)arg;
9109 	uint32_t res, *tl;
9110 	struct nfsrv_descript nfsd;
9111 	struct nfsrv_descript *nd = &nfsd;
9112 	struct rpc_callextra ext;
9113 	struct timeval utimeout;
9114 	enum clnt_stat stat;
9115 	int error;
9116 
9117 	nfscl_reqstart(nd, NFSPROC_BINDCONNTOSESS, NULL, NULL, 0, NULL, NULL,
9118 	    NFS_VER4, rcp->minorvers);
9119 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID + 2 * NFSX_UNSIGNED);
9120 	memcpy(tl, rcp->sessionid, NFSX_V4SESSIONID);
9121 	tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
9122 	*tl++ = txdr_unsigned(NFSCDFC4_FORE_OR_BOTH);
9123 	*tl = newnfs_false;
9124 
9125 	memset(&ext, 0, sizeof(ext));
9126 	utimeout.tv_sec = 30;
9127 	utimeout.tv_usec = 0;
9128 	ext.rc_auth = authunix_create(cr);
9129 	nd->nd_mrep = NULL;
9130 	stat = CLNT_CALL_MBUF(cl, &ext, NFSV4PROC_COMPOUND, nd->nd_mreq,
9131 	    &nd->nd_mrep, utimeout);
9132 	AUTH_DESTROY(ext.rc_auth);
9133 	if (stat != RPC_SUCCESS) {
9134 		printf("nfsrpc_bindconnsess: call failed stat=%d\n", stat);
9135 		return;
9136 	}
9137 	if (nd->nd_mrep == NULL) {
9138 		printf("nfsrpc_bindconnsess: no reply args\n");
9139 		return;
9140 	}
9141 	error = 0;
9142 	newnfs_realign(&nd->nd_mrep, M_WAITOK);
9143 	nd->nd_md = nd->nd_mrep;
9144 	nd->nd_dpos = mtod(nd->nd_md, char *);
9145 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
9146 	nd->nd_repstat = fxdr_unsigned(uint32_t, *tl++);
9147 	if (nd->nd_repstat == NFSERR_OK) {
9148 		res = fxdr_unsigned(uint32_t, *tl);
9149 		if (res > 0 && (error = nfsm_advance(nd, NFSM_RNDUP(res),
9150 		    -1)) != 0)
9151 			goto nfsmout;
9152 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
9153 		    4 * NFSX_UNSIGNED);
9154 		tl += 3;
9155 		if (!NFSBCMP(tl, rcp->sessionid, NFSX_V4SESSIONID)) {
9156 			tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
9157 			res = fxdr_unsigned(uint32_t, *tl);
9158 			if (res != NFSCDFS4_BOTH)
9159 				printf("nfsrpc_bindconnsess: did not "
9160 				    "return FS4_BOTH\n");
9161 		} else
9162 			printf("nfsrpc_bindconnsess: not same "
9163 			    "sessionid\n");
9164 	} else if (nd->nd_repstat != NFSERR_BADSESSION)
9165 		printf("nfsrpc_bindconnsess: returned %d\n", nd->nd_repstat);
9166 nfsmout:
9167 	if (error != 0)
9168 		printf("nfsrpc_bindconnsess: reply bad xdr\n");
9169 	m_freem(nd->nd_mrep);
9170 }
9171