xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision 790b5264886e581e995fc3b8fa45ca4ab4ffd31c)
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 *, void *);
133 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
134     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
135 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
136     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
137     void *);
138 static int nfsrpc_deallocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
139     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *, void *);
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 *, void *);
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 *, void *, 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 *, void *);
152 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
153     uint32_t, uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
154     struct ucred *, NFSPROC_T *);
155 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
156     struct sockaddr_in6 *, sa_family_t, int, int, struct nfsclds **,
157     NFSPROC_T *);
158 static void nfscl_initsessionslots(struct nfsclsession *);
159 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
160     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
161     struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
162     NFSPROC_T *);
163 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
164     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
165     struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
166     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
167 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
168     struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
169     struct ucred *, NFSPROC_T *);
170 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
171     nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
172     struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
173 static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
174     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
175     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
176 static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
177     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
178     struct ucred *, NFSPROC_T *);
179 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
180     struct nfsclds *, struct nfsclds **, uint32_t *);
181 static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
182     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
183     NFSPROC_T *);
184 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
185     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
186 #ifdef notyet
187 static int nfsio_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
188     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
189     NFSPROC_T *);
190 static int nfsrpc_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
191     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
192 #endif
193 static int nfsrpc_allocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
194     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *, void *);
195 static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
196     uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
197 static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
198     NFSPROC_T *);
199 static int nfsrv_parselayoutget(struct nfsmount *, struct nfsrv_descript *,
200     nfsv4stateid_t *, int *, struct nfsclflayouthead *);
201 static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
202     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
203     struct nfscldeleg **, struct ucred *, NFSPROC_T *);
204 static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
205     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
206     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
207     struct nfsfh **, int *, int *, void *, int *);
208 static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
209     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
210     struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
211     struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
212 static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
213     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
214     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
215     struct nfsfh **, int *, int *, void *, int *, nfsv4stateid_t *,
216     int, int, int, int *, struct nfsclflayouthead *, int *);
217 static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
218     uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
219     struct nfsclflayouthead *, struct ucred *, NFSPROC_T *, void *);
220 static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
221     int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
222     struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
223 static int nfsrpc_copyrpc(vnode_t, off_t, vnode_t, off_t, size_t *,
224     nfsv4stateid_t *, nfsv4stateid_t *, struct nfsvattr *, int *,
225     struct nfsvattr *, int *, bool, int *, struct ucred *, NFSPROC_T *);
226 static int nfsrpc_seekrpc(vnode_t, off_t *, nfsv4stateid_t *, bool *,
227     int, struct nfsvattr *, int *, struct ucred *);
228 static struct mbuf *nfsm_split(struct mbuf *, uint64_t);
229 
230 int nfs_pnfsio(task_fn_t *, void *);
231 
232 /*
233  * nfs null call from vfs.
234  */
235 int
236 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
237 {
238 	int error;
239 	struct nfsrv_descript nfsd, *nd = &nfsd;
240 
241 	NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
242 	error = nfscl_request(nd, vp, p, cred, NULL);
243 	if (nd->nd_repstat && !error)
244 		error = nd->nd_repstat;
245 	m_freem(nd->nd_mrep);
246 	return (error);
247 }
248 
249 /*
250  * nfs access rpc op.
251  * For nfs version 3 and 4, use the access rpc to check accessibility. If file
252  * modes are changed on the server, accesses might still fail later.
253  */
254 int
255 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
256     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
257 {
258 	int error;
259 	u_int32_t mode, rmode;
260 
261 	if (acmode & VREAD)
262 		mode = NFSACCESS_READ;
263 	else
264 		mode = 0;
265 	if (vnode_vtype(vp) == VDIR) {
266 		if (acmode & VWRITE)
267 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
268 				 NFSACCESS_DELETE);
269 		if (acmode & VEXEC)
270 			mode |= NFSACCESS_LOOKUP;
271 	} else {
272 		if (acmode & VWRITE)
273 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
274 		if (acmode & VEXEC)
275 			mode |= NFSACCESS_EXECUTE;
276 	}
277 
278 	/*
279 	 * Now, just call nfsrpc_accessrpc() to do the actual RPC.
280 	 */
281 	error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
282 	    NULL);
283 
284 	/*
285 	 * The NFS V3 spec does not clarify whether or not
286 	 * the returned access bits can be a superset of
287 	 * the ones requested, so...
288 	 */
289 	if (!error && (rmode & mode) != mode)
290 		error = EACCES;
291 	return (error);
292 }
293 
294 /*
295  * The actual rpc, separated out for Darwin.
296  */
297 int
298 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
299     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
300     void *stuff)
301 {
302 	u_int32_t *tl;
303 	u_int32_t supported, rmode;
304 	int error;
305 	struct nfsrv_descript nfsd, *nd = &nfsd;
306 	nfsattrbit_t attrbits;
307 
308 	*attrflagp = 0;
309 	supported = mode;
310 	NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
311 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
312 	*tl = txdr_unsigned(mode);
313 	if (nd->nd_flag & ND_NFSV4) {
314 		/*
315 		 * And do a Getattr op.
316 		 */
317 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
318 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
319 		NFSGETATTR_ATTRBIT(&attrbits);
320 		(void) nfsrv_putattrbit(nd, &attrbits);
321 	}
322 	error = nfscl_request(nd, vp, p, cred, stuff);
323 	if (error)
324 		return (error);
325 	if (nd->nd_flag & ND_NFSV3) {
326 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
327 		if (error)
328 			goto nfsmout;
329 	}
330 	if (!nd->nd_repstat) {
331 		if (nd->nd_flag & ND_NFSV4) {
332 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
333 			supported = fxdr_unsigned(u_int32_t, *tl++);
334 		} else {
335 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
336 		}
337 		rmode = fxdr_unsigned(u_int32_t, *tl);
338 		if (nd->nd_flag & ND_NFSV4)
339 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
340 
341 		/*
342 		 * It's not obvious what should be done about
343 		 * unsupported access modes. For now, be paranoid
344 		 * and clear the unsupported ones.
345 		 */
346 		rmode &= supported;
347 		*rmodep = rmode;
348 	} else
349 		error = nd->nd_repstat;
350 nfsmout:
351 	m_freem(nd->nd_mrep);
352 	return (error);
353 }
354 
355 /*
356  * nfs open rpc
357  */
358 int
359 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
360 {
361 	struct nfsclopen *op;
362 	struct nfscldeleg *dp;
363 	struct nfsfh *nfhp;
364 	struct nfsnode *np = VTONFS(vp);
365 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
366 	u_int32_t mode, clidrev;
367 	int ret, newone, error, expireret = 0, retrycnt;
368 
369 	/*
370 	 * For NFSv4, Open Ops are only done on Regular Files.
371 	 */
372 	if (vnode_vtype(vp) != VREG)
373 		return (0);
374 	mode = 0;
375 	if (amode & FREAD)
376 		mode |= NFSV4OPEN_ACCESSREAD;
377 	if (amode & FWRITE)
378 		mode |= NFSV4OPEN_ACCESSWRITE;
379 	nfhp = np->n_fhp;
380 
381 	retrycnt = 0;
382 #ifdef notdef
383 { char name[100]; int namel;
384 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
385 bcopy(NFS4NODENAME(np->n_v4), name, namel);
386 name[namel] = '\0';
387 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
388 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
389 else printf(" fhl=0\n");
390 }
391 #endif
392 	do {
393 	    dp = NULL;
394 	    error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
395 		cred, p, NULL, &op, &newone, &ret, 1, true);
396 	    if (error) {
397 		return (error);
398 	    }
399 	    if (nmp->nm_clp != NULL)
400 		clidrev = nmp->nm_clp->nfsc_clientidrev;
401 	    else
402 		clidrev = 0;
403 	    if (ret == NFSCLOPEN_DOOPEN) {
404 		if (np->n_v4 != NULL) {
405 			/*
406 			 * For the first attempt, try and get a layout, if
407 			 * pNFS is enabled for the mount.
408 			 */
409 			if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
410 			    nfs_numnfscbd == 0 ||
411 			    (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
412 				error = nfsrpc_openrpc(nmp, vp,
413 				    np->n_v4->n4_data,
414 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
415 				    np->n_fhp->nfh_len, mode, op,
416 				    NFS4NODENAME(np->n_v4),
417 				    np->n_v4->n4_namelen,
418 				    &dp, 0, 0x0, cred, p, 0, 0);
419 			else
420 				error = nfsrpc_getopenlayout(nmp, vp,
421 				    np->n_v4->n4_data,
422 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
423 				    np->n_fhp->nfh_len, mode, op,
424 				    NFS4NODENAME(np->n_v4),
425 				    np->n_v4->n4_namelen, &dp, cred, p);
426 			if (dp != NULL) {
427 #ifdef APPLE
428 				OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
429 #else
430 				NFSLOCKNODE(np);
431 				np->n_flag &= ~NDELEGMOD;
432 				/*
433 				 * Invalidate the attribute cache, so that
434 				 * attributes that pre-date the issue of a
435 				 * delegation are not cached, since the
436 				 * cached attributes will remain valid while
437 				 * the delegation is held.
438 				 */
439 				NFSINVALATTRCACHE(np);
440 				NFSUNLOCKNODE(np);
441 #endif
442 				(void) nfscl_deleg(nmp->nm_mountp,
443 				    op->nfso_own->nfsow_clp,
444 				    nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
445 			}
446 		} else {
447 			error = EIO;
448 		}
449 		newnfs_copyincred(cred, &op->nfso_cred);
450 	    } else if (ret == NFSCLOPEN_SETCRED)
451 		/*
452 		 * This is a new local open on a delegation. It needs
453 		 * to have credentials so that an open can be done
454 		 * against the server during recovery.
455 		 */
456 		newnfs_copyincred(cred, &op->nfso_cred);
457 
458 	    /*
459 	     * nfso_opencnt is the count of how many VOP_OPEN()s have
460 	     * been done on this Open successfully and a VOP_CLOSE()
461 	     * is expected for each of these.
462 	     * If error is non-zero, don't increment it, since the Open
463 	     * hasn't succeeded yet.
464 	     */
465 	    if (!error) {
466 		op->nfso_opencnt++;
467 		if (NFSHASNFSV4N(nmp) && NFSHASONEOPENOWN(nmp)) {
468 		    NFSLOCKNODE(np);
469 		    np->n_openstateid = op;
470 		    NFSUNLOCKNODE(np);
471 		}
472 	    }
473 	    nfscl_openrelease(nmp, op, error, newone);
474 	    if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
475 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
476 		error == NFSERR_BADSESSION) {
477 		(void) nfs_catnap(PZERO, error, "nfs_open");
478 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
479 		&& clidrev != 0) {
480 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
481 		retrycnt++;
482 	    }
483 	} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
484 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
485 	    error == NFSERR_BADSESSION ||
486 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
487 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
488 	if (error && retrycnt >= 4)
489 		error = EIO;
490 	return (error);
491 }
492 
493 /*
494  * the actual open rpc
495  */
496 int
497 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
498     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
499     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
500     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
501     int syscred, int recursed)
502 {
503 	u_int32_t *tl;
504 	struct nfsrv_descript nfsd, *nd = &nfsd;
505 	struct nfscldeleg *dp, *ndp = NULL;
506 	struct nfsvattr nfsva;
507 	u_int32_t rflags, deleg;
508 	nfsattrbit_t attrbits;
509 	int error, ret, acesize, limitby;
510 	struct nfsclsession *tsep;
511 
512 	dp = *dpp;
513 	*dpp = NULL;
514 	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
515 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
516 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
517 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
518 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
519 	tsep = nfsmnt_mdssession(nmp);
520 	*tl++ = tsep->nfsess_clientid.lval[0];
521 	*tl = tsep->nfsess_clientid.lval[1];
522 	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
523 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
524 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
525 	if (reclaim) {
526 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
527 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
528 		*tl = txdr_unsigned(delegtype);
529 	} else {
530 		if (dp != NULL) {
531 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
532 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
533 			if (NFSHASNFSV4N(nmp))
534 				*tl++ = 0;
535 			else
536 				*tl++ = dp->nfsdl_stateid.seqid;
537 			*tl++ = dp->nfsdl_stateid.other[0];
538 			*tl++ = dp->nfsdl_stateid.other[1];
539 			*tl = dp->nfsdl_stateid.other[2];
540 		} else {
541 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
542 		}
543 		(void) nfsm_strtom(nd, name, namelen);
544 	}
545 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
546 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
547 	NFSZERO_ATTRBIT(&attrbits);
548 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
549 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
550 	(void) nfsrv_putattrbit(nd, &attrbits);
551 	if (syscred)
552 		nd->nd_flag |= ND_USEGSSNAME;
553 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
554 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
555 	if (error)
556 		return (error);
557 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
558 	if (!nd->nd_repstat) {
559 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
560 		    6 * NFSX_UNSIGNED);
561 		op->nfso_stateid.seqid = *tl++;
562 		op->nfso_stateid.other[0] = *tl++;
563 		op->nfso_stateid.other[1] = *tl++;
564 		op->nfso_stateid.other[2] = *tl;
565 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
566 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
567 		if (error)
568 			goto nfsmout;
569 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
570 		deleg = fxdr_unsigned(u_int32_t, *tl);
571 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
572 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
573 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
574 			      NFSCLFLAGS_FIRSTDELEG))
575 				op->nfso_own->nfsow_clp->nfsc_flags |=
576 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
577 			ndp = malloc(
578 			    sizeof (struct nfscldeleg) + newfhlen,
579 			    M_NFSCLDELEG, M_WAITOK);
580 			LIST_INIT(&ndp->nfsdl_owner);
581 			LIST_INIT(&ndp->nfsdl_lock);
582 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
583 			ndp->nfsdl_fhlen = newfhlen;
584 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
585 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
586 			nfscl_lockinit(&ndp->nfsdl_rwlock);
587 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
588 			    NFSX_UNSIGNED);
589 			ndp->nfsdl_stateid.seqid = *tl++;
590 			ndp->nfsdl_stateid.other[0] = *tl++;
591 			ndp->nfsdl_stateid.other[1] = *tl++;
592 			ndp->nfsdl_stateid.other[2] = *tl++;
593 			ret = fxdr_unsigned(int, *tl);
594 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
595 				ndp->nfsdl_flags = NFSCLDL_WRITE;
596 				/*
597 				 * Indicates how much the file can grow.
598 				 */
599 				NFSM_DISSECT(tl, u_int32_t *,
600 				    3 * NFSX_UNSIGNED);
601 				limitby = fxdr_unsigned(int, *tl++);
602 				switch (limitby) {
603 				case NFSV4OPEN_LIMITSIZE:
604 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
605 					break;
606 				case NFSV4OPEN_LIMITBLOCKS:
607 					ndp->nfsdl_sizelimit =
608 					    fxdr_unsigned(u_int64_t, *tl++);
609 					ndp->nfsdl_sizelimit *=
610 					    fxdr_unsigned(u_int64_t, *tl);
611 					break;
612 				default:
613 					error = NFSERR_BADXDR;
614 					goto nfsmout;
615 				}
616 			} else {
617 				ndp->nfsdl_flags = NFSCLDL_READ;
618 			}
619 			if (ret)
620 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
621 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
622 			    &acesize, p);
623 			if (error)
624 				goto nfsmout;
625 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
626 			error = NFSERR_BADXDR;
627 			goto nfsmout;
628 		}
629 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
630 		error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
631 		    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
632 		    NULL, NULL, NULL, p, cred);
633 		if (error)
634 			goto nfsmout;
635 		if (ndp != NULL) {
636 			ndp->nfsdl_change = nfsva.na_filerev;
637 			ndp->nfsdl_modtime = nfsva.na_mtime;
638 			ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
639 		}
640 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
641 		    do {
642 			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
643 			    cred, p);
644 			if (ret == NFSERR_DELAY)
645 			    (void) nfs_catnap(PZERO, ret, "nfs_open");
646 		    } while (ret == NFSERR_DELAY);
647 		    error = ret;
648 		}
649 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
650 		    nfscl_assumeposixlocks)
651 		    op->nfso_posixlock = 1;
652 		else
653 		    op->nfso_posixlock = 0;
654 
655 		/*
656 		 * If the server is handing out delegations, but we didn't
657 		 * get one because an OpenConfirm was required, try the
658 		 * Open again, to get a delegation. This is a harmless no-op,
659 		 * from a server's point of view.
660 		 */
661 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
662 		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
663 		    && !error && dp == NULL && ndp == NULL && !recursed) {
664 		    do {
665 			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
666 			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
667 			    cred, p, syscred, 1);
668 			if (ret == NFSERR_DELAY)
669 			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
670 		    } while (ret == NFSERR_DELAY);
671 		    if (ret) {
672 			if (ndp != NULL) {
673 				free(ndp, M_NFSCLDELEG);
674 				ndp = NULL;
675 			}
676 			if (ret == NFSERR_STALECLIENTID ||
677 			    ret == NFSERR_STALEDONTRECOVER ||
678 			    ret == NFSERR_BADSESSION)
679 				error = ret;
680 		    }
681 		}
682 	}
683 	if (nd->nd_repstat != 0 && error == 0)
684 		error = nd->nd_repstat;
685 	if (error == NFSERR_STALECLIENTID)
686 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
687 nfsmout:
688 	if (!error)
689 		*dpp = ndp;
690 	else if (ndp != NULL)
691 		free(ndp, M_NFSCLDELEG);
692 	m_freem(nd->nd_mrep);
693 	return (error);
694 }
695 
696 /*
697  * open downgrade rpc
698  */
699 int
700 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
701     struct ucred *cred, NFSPROC_T *p)
702 {
703 	u_int32_t *tl;
704 	struct nfsrv_descript nfsd, *nd = &nfsd;
705 	int error;
706 
707 	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
708 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
709 	if (NFSHASNFSV4N(VFSTONFS(vp->v_mount)))
710 		*tl++ = 0;
711 	else
712 		*tl++ = op->nfso_stateid.seqid;
713 	*tl++ = op->nfso_stateid.other[0];
714 	*tl++ = op->nfso_stateid.other[1];
715 	*tl++ = op->nfso_stateid.other[2];
716 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
717 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
718 	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
719 	error = nfscl_request(nd, vp, p, cred, NULL);
720 	if (error)
721 		return (error);
722 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
723 	if (!nd->nd_repstat) {
724 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
725 		op->nfso_stateid.seqid = *tl++;
726 		op->nfso_stateid.other[0] = *tl++;
727 		op->nfso_stateid.other[1] = *tl++;
728 		op->nfso_stateid.other[2] = *tl;
729 	}
730 	if (nd->nd_repstat && error == 0)
731 		error = nd->nd_repstat;
732 	if (error == NFSERR_STALESTATEID)
733 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
734 nfsmout:
735 	m_freem(nd->nd_mrep);
736 	return (error);
737 }
738 
739 /*
740  * V4 Close operation.
741  */
742 int
743 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
744 {
745 	struct nfsclclient *clp;
746 	int error;
747 
748 	if (vnode_vtype(vp) != VREG)
749 		return (0);
750 	if (doclose)
751 		error = nfscl_doclose(vp, &clp, p);
752 	else {
753 		error = nfscl_getclose(vp, &clp);
754 		if (error == 0)
755 			nfscl_clientrelease(clp);
756 	}
757 	return (error);
758 }
759 
760 /*
761  * Close the open.
762  */
763 int
764 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p,
765     bool loop_on_delayed, bool freeop)
766 {
767 	struct nfsrv_descript nfsd, *nd = &nfsd;
768 	struct nfscllockowner *lp, *nlp;
769 	struct nfscllock *lop, *nlop;
770 	struct ucred *tcred;
771 	u_int64_t off = 0, len = 0;
772 	u_int32_t type = NFSV4LOCKT_READ;
773 	int error, do_unlock, trycnt;
774 
775 	tcred = newnfs_getcred();
776 	newnfs_copycred(&op->nfso_cred, tcred);
777 	/*
778 	 * (Theoretically this could be done in the same
779 	 *  compound as the close, but having multiple
780 	 *  sequenced Ops in the same compound might be
781 	 *  too scary for some servers.)
782 	 */
783 	if (op->nfso_posixlock) {
784 		off = 0;
785 		len = NFS64BITSSET;
786 		type = NFSV4LOCKT_READ;
787 	}
788 
789 	/*
790 	 * Since this function is only called from VOP_INACTIVE(), no
791 	 * other thread will be manipulating this Open. As such, the
792 	 * lock lists are not being changed by other threads, so it should
793 	 * be safe to do this without locking.
794 	 */
795 	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
796 		do_unlock = 1;
797 		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
798 			if (op->nfso_posixlock == 0) {
799 				off = lop->nfslo_first;
800 				len = lop->nfslo_end - lop->nfslo_first;
801 				if (lop->nfslo_type == F_WRLCK)
802 					type = NFSV4LOCKT_WRITE;
803 				else
804 					type = NFSV4LOCKT_READ;
805 			}
806 			if (do_unlock) {
807 				trycnt = 0;
808 				do {
809 					error = nfsrpc_locku(nd, nmp, lp, off,
810 					    len, type, tcred, p, 0);
811 					if ((nd->nd_repstat == NFSERR_GRACE ||
812 					    nd->nd_repstat == NFSERR_DELAY) &&
813 					    error == 0)
814 						(void) nfs_catnap(PZERO,
815 						    (int)nd->nd_repstat,
816 						    "nfs_close");
817 				} while ((nd->nd_repstat == NFSERR_GRACE ||
818 				    nd->nd_repstat == NFSERR_DELAY) &&
819 				    error == 0 && trycnt++ < 5);
820 				if (op->nfso_posixlock)
821 					do_unlock = 0;
822 			}
823 			nfscl_freelock(lop, 0);
824 		}
825 		/*
826 		 * Do a ReleaseLockOwner.
827 		 * The lock owner name nfsl_owner may be used by other opens for
828 		 * other files but the lock_owner4 name that nfsrpc_rellockown()
829 		 * puts on the wire has the file handle for this file appended
830 		 * to it, so it can be done now.
831 		 */
832 		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
833 		    lp->nfsl_open->nfso_fhlen, tcred, p);
834 	}
835 
836 	/*
837 	 * There could be other Opens for different files on the same
838 	 * OpenOwner, so locking is required.
839 	 */
840 	NFSLOCKCLSTATE();
841 	nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
842 	NFSUNLOCKCLSTATE();
843 	do {
844 		error = nfscl_tryclose(op, tcred, nmp, p, loop_on_delayed);
845 		if (error == NFSERR_GRACE)
846 			(void) nfs_catnap(PZERO, error, "nfs_close");
847 	} while (error == NFSERR_GRACE);
848 	NFSLOCKCLSTATE();
849 	nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
850 
851 	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
852 		nfscl_freelockowner(lp, 0);
853 	if (freeop && error != NFSERR_DELAY)
854 		nfscl_freeopen(op, 0, true);
855 	NFSUNLOCKCLSTATE();
856 	NFSFREECRED(tcred);
857 	return (error);
858 }
859 
860 /*
861  * The actual Close RPC.
862  */
863 int
864 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
865     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
866     int syscred)
867 {
868 	u_int32_t *tl;
869 	int error;
870 
871 	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
872 	    op->nfso_fhlen, NULL, NULL, 0, 0);
873 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
874 	if (NFSHASNFSV4N(nmp)) {
875 		*tl++ = 0;
876 		*tl++ = 0;
877 	} else {
878 		*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
879 		*tl++ = op->nfso_stateid.seqid;
880 	}
881 	*tl++ = op->nfso_stateid.other[0];
882 	*tl++ = op->nfso_stateid.other[1];
883 	*tl = op->nfso_stateid.other[2];
884 	if (syscred)
885 		nd->nd_flag |= ND_USEGSSNAME;
886 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
887 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
888 	if (error)
889 		return (error);
890 	if (!NFSHASNFSV4N(nmp))
891 		NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
892 	if (nd->nd_repstat == 0)
893 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
894 	error = nd->nd_repstat;
895 	if (!NFSHASNFSV4N(nmp) && error == NFSERR_STALESTATEID)
896 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
897 nfsmout:
898 	m_freem(nd->nd_mrep);
899 	return (error);
900 }
901 
902 /*
903  * V4 Open Confirm RPC.
904  */
905 int
906 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
907     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
908 {
909 	u_int32_t *tl;
910 	struct nfsrv_descript nfsd, *nd = &nfsd;
911 	struct nfsmount *nmp;
912 	int error;
913 
914 	nmp = VFSTONFS(vp->v_mount);
915 	if (NFSHASNFSV4N(nmp))
916 		return (0);		/* No confirmation for NFSv4.1. */
917 	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
918 	    0, 0);
919 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
920 	*tl++ = op->nfso_stateid.seqid;
921 	*tl++ = op->nfso_stateid.other[0];
922 	*tl++ = op->nfso_stateid.other[1];
923 	*tl++ = op->nfso_stateid.other[2];
924 	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
925 	error = nfscl_request(nd, vp, p, cred, NULL);
926 	if (error)
927 		return (error);
928 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
929 	if (!nd->nd_repstat) {
930 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
931 		op->nfso_stateid.seqid = *tl++;
932 		op->nfso_stateid.other[0] = *tl++;
933 		op->nfso_stateid.other[1] = *tl++;
934 		op->nfso_stateid.other[2] = *tl;
935 	}
936 	error = nd->nd_repstat;
937 	if (error == NFSERR_STALESTATEID)
938 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
939 nfsmout:
940 	m_freem(nd->nd_mrep);
941 	return (error);
942 }
943 
944 /*
945  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
946  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
947  */
948 int
949 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
950     bool *retokp, struct ucred *cred, NFSPROC_T *p)
951 {
952 	u_int32_t *tl;
953 	struct nfsrv_descript nfsd;
954 	struct nfsrv_descript *nd = &nfsd;
955 	u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
956 	u_short port;
957 	int error, isinet6 = 0, callblen;
958 	nfsquad_t confirm;
959 	static u_int32_t rev = 0;
960 	struct nfsclds *dsp, *odsp;
961 	struct in6_addr a6;
962 	struct nfsclsession *tsep;
963 	struct rpc_reconupcall recon;
964 	struct nfscl_reconarg *rcp;
965 
966 	if (nfsboottime.tv_sec == 0)
967 		NFSSETBOOTTIME(nfsboottime);
968 	if (NFSHASNFSV4N(nmp)) {
969 		error = NFSERR_BADSESSION;
970 		odsp = dsp = NULL;
971 		if (retokp != NULL) {
972 			NFSLOCKMNT(nmp);
973 			odsp = TAILQ_FIRST(&nmp->nm_sess);
974 			NFSUNLOCKMNT(nmp);
975 		}
976 		if (odsp != NULL) {
977 			/*
978 			 * When a session already exists, first try a
979 			 * CreateSession with the extant ClientID.
980 			 */
981 			dsp = malloc(sizeof(struct nfsclds) +
982 			    odsp->nfsclds_servownlen + 1, M_NFSCLDS,
983 			    M_WAITOK | M_ZERO);
984 			dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
985 			dsp->nfsclds_servownlen = odsp->nfsclds_servownlen;
986 			dsp->nfsclds_sess.nfsess_clientid =
987 			    odsp->nfsclds_sess.nfsess_clientid;
988 			dsp->nfsclds_sess.nfsess_sequenceid =
989 			    odsp->nfsclds_sess.nfsess_sequenceid;
990 			dsp->nfsclds_flags = odsp->nfsclds_flags;
991 			if (dsp->nfsclds_servownlen > 0)
992 				memcpy(dsp->nfsclds_serverown,
993 				    odsp->nfsclds_serverown,
994 				    dsp->nfsclds_servownlen + 1);
995 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
996 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
997 			    NULL, MTX_DEF);
998 			nfscl_initsessionslots(&dsp->nfsclds_sess);
999 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
1000 			    &nmp->nm_sockreq, NULL,
1001 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
1002 			NFSCL_DEBUG(1, "create session for extant "
1003 			    "ClientID=%d\n", error);
1004 			if (error != 0) {
1005 				nfscl_freenfsclds(dsp);
1006 				dsp = NULL;
1007 				/*
1008 				 * If *retokp is true, return any error other
1009 				 * than NFSERR_STALECLIENTID,
1010 				 * NFSERR_BADSESSION or NFSERR_STALEDONTRECOVER
1011 				 * so that nfscl_recover() will not loop.
1012 				 */
1013 				if (*retokp)
1014 					return (NFSERR_IO);
1015 			} else
1016 				*retokp = true;
1017 		} else if (retokp != NULL && *retokp)
1018 			return (NFSERR_IO);
1019 		if (error != 0) {
1020 			/*
1021 			 * Either there was no previous session or the
1022 			 * CreateSession attempt failed, so...
1023 			 * do an ExchangeID followed by the CreateSession.
1024 			 */
1025 			clp->nfsc_rev = rev++;
1026 			error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq, 0,
1027 			    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp,
1028 			    cred, p);
1029 			NFSCL_DEBUG(1, "aft exch=%d\n", error);
1030 			if (error == 0)
1031 				error = nfsrpc_createsession(nmp,
1032 				    &dsp->nfsclds_sess, &nmp->nm_sockreq, NULL,
1033 				    dsp->nfsclds_sess.nfsess_sequenceid, 1,
1034 				    cred, p);
1035 			NFSCL_DEBUG(1, "aft createsess=%d\n", error);
1036 		}
1037 		if (error == 0) {
1038 			/*
1039 			 * If the session supports a backchannel, set up
1040 			 * the BindConnectionToSession call in the krpc
1041 			 * so that it is done on a reconnection.
1042 			 */
1043 			if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0) {
1044 				rcp = mem_alloc(sizeof(*rcp));
1045 				rcp->minorvers = nmp->nm_minorvers;
1046 				memcpy(rcp->sessionid,
1047 				    dsp->nfsclds_sess.nfsess_sessionid,
1048 				    NFSX_V4SESSIONID);
1049 				recon.call = nfsrpc_bindconnsess;
1050 				recon.arg = rcp;
1051 				CLNT_CONTROL(nmp->nm_client, CLSET_RECONUPCALL,
1052 				    &recon);
1053 			}
1054 
1055 			NFSLOCKMNT(nmp);
1056 			/*
1057 			 * The old sessions cannot be safely free'd
1058 			 * here, since they may still be used by
1059 			 * in-progress RPCs.
1060 			 */
1061 			tsep = NULL;
1062 			if (TAILQ_FIRST(&nmp->nm_sess) != NULL)
1063 				tsep = NFSMNT_MDSSESSION(nmp);
1064 			TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
1065 			    nfsclds_list);
1066 			/*
1067 			 * Wake up RPCs waiting for a slot on the
1068 			 * old session. These will then fail with
1069 			 * NFSERR_BADSESSION and be retried with the
1070 			 * new session by nfsv4_setsequence().
1071 			 * Also wakeup() processes waiting for the
1072 			 * new session.
1073 			 */
1074 			if (tsep != NULL)
1075 				wakeup(&tsep->nfsess_slots);
1076 			wakeup(&nmp->nm_sess);
1077 			NFSUNLOCKMNT(nmp);
1078 		} else if (dsp != NULL)
1079 			nfscl_freenfsclds(dsp);
1080 		if (error == 0 && reclaim == 0) {
1081 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
1082 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
1083 			if (error == NFSERR_COMPLETEALREADY ||
1084 			    error == NFSERR_NOTSUPP)
1085 				/* Ignore this error. */
1086 				error = 0;
1087 		}
1088 		return (error);
1089 	} else if (retokp != NULL && *retokp)
1090 		return (NFSERR_IO);
1091 	clp->nfsc_rev = rev++;
1092 
1093 	/*
1094 	 * Allocate a single session structure for NFSv4.0, because some of
1095 	 * the fields are used by NFSv4.0 although it doesn't do a session.
1096 	 */
1097 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
1098 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
1099 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
1100 	NFSLOCKMNT(nmp);
1101 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
1102 	tsep = NFSMNT_MDSSESSION(nmp);
1103 	NFSUNLOCKMNT(nmp);
1104 
1105 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0);
1106 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1107 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
1108 	*tl = txdr_unsigned(clp->nfsc_rev);
1109 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1110 
1111 	/*
1112 	 * set up the callback address
1113 	 */
1114 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1115 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
1116 	callblen = strlen(nfsv4_callbackaddr);
1117 	if (callblen == 0)
1118 		cp = nfscl_getmyip(nmp, &a6, &isinet6);
1119 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1120 	    (callblen > 0 || cp != NULL)) {
1121 		port = htons(nfsv4_cbport);
1122 		cp2 = (u_int8_t *)&port;
1123 #ifdef INET6
1124 		if ((callblen > 0 &&
1125 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1126 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1127 
1128 			(void) nfsm_strtom(nd, "tcp6", 4);
1129 			if (callblen == 0) {
1130 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1131 				ip6add = ip6buf;
1132 			} else {
1133 				ip6add = nfsv4_callbackaddr;
1134 			}
1135 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1136 			    ip6add, cp2[0], cp2[1]);
1137 		} else
1138 #endif
1139 		{
1140 			(void) nfsm_strtom(nd, "tcp", 3);
1141 			if (callblen == 0)
1142 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1143 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1144 				    cp[2], cp[3], cp2[0], cp2[1]);
1145 			else
1146 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1147 				    "%s.%d.%d", nfsv4_callbackaddr,
1148 				    cp2[0], cp2[1]);
1149 		}
1150 		(void) nfsm_strtom(nd, addr, strlen(addr));
1151 	} else {
1152 		(void) nfsm_strtom(nd, "tcp", 3);
1153 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1154 	}
1155 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1156 	*tl = txdr_unsigned(clp->nfsc_cbident);
1157 	nd->nd_flag |= ND_USEGSSNAME;
1158 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1159 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1160 	if (error)
1161 		return (error);
1162 	if (nd->nd_repstat == 0) {
1163 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1164 	    tsep->nfsess_clientid.lval[0] = *tl++;
1165 	    tsep->nfsess_clientid.lval[1] = *tl++;
1166 	    confirm.lval[0] = *tl++;
1167 	    confirm.lval[1] = *tl;
1168 	    m_freem(nd->nd_mrep);
1169 	    nd->nd_mrep = NULL;
1170 
1171 	    /*
1172 	     * and confirm it.
1173 	     */
1174 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1175 		NULL, 0, 0);
1176 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1177 	    *tl++ = tsep->nfsess_clientid.lval[0];
1178 	    *tl++ = tsep->nfsess_clientid.lval[1];
1179 	    *tl++ = confirm.lval[0];
1180 	    *tl = confirm.lval[1];
1181 	    nd->nd_flag |= ND_USEGSSNAME;
1182 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1183 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1184 	    if (error)
1185 		return (error);
1186 	    m_freem(nd->nd_mrep);
1187 	    nd->nd_mrep = NULL;
1188 	}
1189 	error = nd->nd_repstat;
1190 nfsmout:
1191 	m_freem(nd->nd_mrep);
1192 	return (error);
1193 }
1194 
1195 /*
1196  * nfs getattr call.
1197  */
1198 int
1199 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1200     struct nfsvattr *nap, void *stuff)
1201 {
1202 	struct nfsrv_descript nfsd, *nd = &nfsd;
1203 	int error;
1204 	nfsattrbit_t attrbits;
1205 
1206 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1207 	if (nd->nd_flag & ND_NFSV4) {
1208 		NFSGETATTR_ATTRBIT(&attrbits);
1209 		(void) nfsrv_putattrbit(nd, &attrbits);
1210 	}
1211 	error = nfscl_request(nd, vp, p, cred, stuff);
1212 	if (error)
1213 		return (error);
1214 	if (!nd->nd_repstat)
1215 		error = nfsm_loadattr(nd, nap);
1216 	else
1217 		error = nd->nd_repstat;
1218 	m_freem(nd->nd_mrep);
1219 	return (error);
1220 }
1221 
1222 /*
1223  * nfs getattr call with non-vnode arguments.
1224  */
1225 int
1226 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1227     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1228     uint32_t *leasep)
1229 {
1230 	struct nfsrv_descript nfsd, *nd = &nfsd;
1231 	int error, vers = NFS_VER2;
1232 	nfsattrbit_t attrbits;
1233 
1234 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0);
1235 	if (nd->nd_flag & ND_NFSV4) {
1236 		vers = NFS_VER4;
1237 		NFSGETATTR_ATTRBIT(&attrbits);
1238 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1239 		(void) nfsrv_putattrbit(nd, &attrbits);
1240 	} else if (nd->nd_flag & ND_NFSV3) {
1241 		vers = NFS_VER3;
1242 	}
1243 	if (syscred)
1244 		nd->nd_flag |= ND_USEGSSNAME;
1245 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1246 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1247 	if (error)
1248 		return (error);
1249 	if (nd->nd_repstat == 0) {
1250 		if ((nd->nd_flag & ND_NFSV4) != 0)
1251 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1252 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1253 			    NULL, NULL);
1254 		else
1255 			error = nfsm_loadattr(nd, nap);
1256 	} else
1257 		error = nd->nd_repstat;
1258 	m_freem(nd->nd_mrep);
1259 	return (error);
1260 }
1261 
1262 /*
1263  * Do an nfs setattr operation.
1264  */
1265 int
1266 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1267     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1268     void *stuff)
1269 {
1270 	int error, expireret = 0, openerr, retrycnt;
1271 	u_int32_t clidrev = 0, mode;
1272 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1273 	struct nfsfh *nfhp;
1274 	nfsv4stateid_t stateid;
1275 	void *lckp;
1276 
1277 	if (nmp->nm_clp != NULL)
1278 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1279 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1280 		mode = NFSV4OPEN_ACCESSWRITE;
1281 	else
1282 		mode = NFSV4OPEN_ACCESSREAD;
1283 	retrycnt = 0;
1284 	do {
1285 		lckp = NULL;
1286 		openerr = 1;
1287 		if (NFSHASNFSV4(nmp)) {
1288 			nfhp = VTONFS(vp)->n_fhp;
1289 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1290 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1291 			if (error && vnode_vtype(vp) == VREG &&
1292 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1293 			     nfstest_openallsetattr)) {
1294 				/*
1295 				 * No Open stateid, so try and open the file
1296 				 * now.
1297 				 */
1298 				if (mode == NFSV4OPEN_ACCESSWRITE)
1299 					openerr = nfsrpc_open(vp, FWRITE, cred,
1300 					    p);
1301 				else
1302 					openerr = nfsrpc_open(vp, FREAD, cred,
1303 					    p);
1304 				if (!openerr)
1305 					(void) nfscl_getstateid(vp,
1306 					    nfhp->nfh_fh, nfhp->nfh_len,
1307 					    mode, 0, cred, p, &stateid, &lckp);
1308 			}
1309 		}
1310 		if (vap != NULL)
1311 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1312 			    rnap, attrflagp, stuff);
1313 		else
1314 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1315 			    stuff);
1316 		if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1317 			NFSLOCKMNT(nmp);
1318 			nmp->nm_state |= NFSSTA_OPENMODE;
1319 			NFSUNLOCKMNT(nmp);
1320 		}
1321 		if (error == NFSERR_STALESTATEID)
1322 			nfscl_initiate_recovery(nmp->nm_clp);
1323 		if (lckp != NULL)
1324 			nfscl_lockderef(lckp);
1325 		if (!openerr)
1326 			(void) nfsrpc_close(vp, 0, p);
1327 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1328 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1329 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1330 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1331 		} else if ((error == NFSERR_EXPIRED ||
1332 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1333 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1334 		}
1335 		retrycnt++;
1336 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1337 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1338 	    error == NFSERR_BADSESSION ||
1339 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1340 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1341 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1342 	    (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1343 	     retrycnt < 4));
1344 	if (error && retrycnt >= 4)
1345 		error = EIO;
1346 	return (error);
1347 }
1348 
1349 static int
1350 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1351     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1352     struct nfsvattr *rnap, int *attrflagp, void *stuff)
1353 {
1354 	u_int32_t *tl;
1355 	struct nfsrv_descript nfsd, *nd = &nfsd;
1356 	int error;
1357 	nfsattrbit_t attrbits;
1358 
1359 	*attrflagp = 0;
1360 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1361 	if (nd->nd_flag & ND_NFSV4)
1362 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1363 	vap->va_type = vnode_vtype(vp);
1364 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1365 	if (nd->nd_flag & ND_NFSV3) {
1366 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1367 		*tl = newnfs_false;
1368 	} else if (nd->nd_flag & ND_NFSV4) {
1369 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1370 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1371 		NFSGETATTR_ATTRBIT(&attrbits);
1372 		(void) nfsrv_putattrbit(nd, &attrbits);
1373 	}
1374 	error = nfscl_request(nd, vp, p, cred, stuff);
1375 	if (error)
1376 		return (error);
1377 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1378 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1379 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1380 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1381 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1382 		error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1383 	m_freem(nd->nd_mrep);
1384 	if (nd->nd_repstat && !error)
1385 		error = nd->nd_repstat;
1386 	return (error);
1387 }
1388 
1389 /*
1390  * nfs lookup rpc
1391  */
1392 int
1393 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1394     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1395     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff,
1396     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 (vnode_vtype(dvp) != 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, stuff);
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, stuff);
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 				*dattrflagp = 1;
1501 			else
1502 				goto nfsmout;
1503 		}
1504 		/* Check Lookup operation reply status. */
1505 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1506 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1507 			if (*++tl != 0)
1508 				goto nfsmout;
1509 		}
1510 		/* Look for GetFH reply. */
1511 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1512 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1513 			if (*++tl != 0)
1514 				goto nfsmout;
1515 			error = nfsm_getfh(nd, nfhpp);
1516 			if (error)
1517 				goto nfsmout;
1518 		}
1519 		/* Look for Getattr reply. */
1520 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1521 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1522 			if (*++tl != 0)
1523 				goto nfsmout;
1524 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1525 			if (error == 0)
1526 				/* Successfully got Lookup done. */
1527 				nd->nd_repstat = 0;
1528 		}
1529 		goto nfsmout;
1530 	}
1531 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1532 		/* Load the directory attributes. */
1533 		error = nfsm_loadattr(nd, dnap);
1534 		if (error != 0)
1535 			goto nfsmout;
1536 		*dattrflagp = 1;
1537 		/* Skip over the Lookup and GetFH operation status values. */
1538 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1539 	}
1540 	error = nfsm_getfh(nd, nfhpp);
1541 	if (error)
1542 		goto nfsmout;
1543 
1544 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1545 	if (openmode != 0 && error == 0) {
1546 		NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID +
1547 		    10 * NFSX_UNSIGNED);
1548 		tl += 4;	/* Skip over Verify+Open status. */
1549 		stateid.seqid = *tl++;
1550 		stateid.other[0] = *tl++;
1551 		stateid.other[1] = *tl++;
1552 		stateid.other[2] = *tl;
1553 		rflags = fxdr_unsigned(uint32_t, *(tl + 6));
1554 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1555 		if (error != 0)
1556 			goto nfsmout;
1557 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
1558 		deleg = fxdr_unsigned(uint32_t, *tl);
1559 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
1560 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
1561 			/*
1562 			 * Just need to fill in the fields used by
1563 			 * nfscl_trydelegreturn().
1564 			 * Mark the mount point as acquiring
1565 			 * delegations, so NFSPROC_LOOKUPOPEN will
1566 			 * no longer be done.
1567 			 */
1568 			NFSLOCKMNT(nmp);
1569 			nmp->nm_privflag |= NFSMNTP_DELEGISSUED;
1570 			NFSUNLOCKMNT(nmp);
1571 			ndp = malloc(sizeof(struct nfscldeleg) +
1572 			    (*nfhpp)->nfh_len, M_NFSCLDELEG, M_WAITOK);
1573 			ndp->nfsdl_fhlen = (*nfhpp)->nfh_len;
1574 			NFSBCOPY((*nfhpp)->nfh_fh, ndp->nfsdl_fh,
1575 			    ndp->nfsdl_fhlen);
1576 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
1577 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
1578 			ndp->nfsdl_stateid.seqid = *tl++;
1579 			ndp->nfsdl_stateid.other[0] = *tl++;
1580 			ndp->nfsdl_stateid.other[1] = *tl++;
1581 			ndp->nfsdl_stateid.other[2] = *tl++;
1582 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
1583 			error = NFSERR_BADXDR;
1584 			goto nfsmout;
1585 		}
1586 		ret = nfscl_open(dvp, (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len,
1587 		    openmode, 0, cred, p, NULL, &op, &newone, &retop, 1, true);
1588 		if (ret != 0)
1589 			goto nfsmout;
1590 		if (newone != 0) {
1591 			op->nfso_stateid.seqid = stateid.seqid;
1592 			op->nfso_stateid.other[0] = stateid.other[0];
1593 			op->nfso_stateid.other[1] = stateid.other[1];
1594 			op->nfso_stateid.other[2] = stateid.other[2];
1595 			op->nfso_mode = openmode;
1596 		} else {
1597 			op->nfso_stateid.seqid = stateid.seqid;
1598 			if (retop == NFSCLOPEN_DOOPEN)
1599 				op->nfso_mode |= openmode;
1600 		}
1601 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
1602 		    nfscl_assumeposixlocks)
1603 			op->nfso_posixlock = 1;
1604 		else
1605 			op->nfso_posixlock = 0;
1606 		nfscl_openrelease(nmp, op, 0, 0);
1607 		if (ndp != NULL) {
1608 			/*
1609 			 * Since we do not have the vnode, we
1610 			 * cannot invalidate cached attributes.
1611 			 * Just return the delegation.
1612 			 */
1613 			nfscl_trydelegreturn(ndp, cred, nmp, p);
1614 		}
1615 	}
1616 	if ((nd->nd_flag & ND_NFSV3) && !error)
1617 		error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1618 nfsmout:
1619 	m_freem(nd->nd_mrep);
1620 	if (!error && nd->nd_repstat)
1621 		error = nd->nd_repstat;
1622 	free(ndp, M_NFSCLDELEG);
1623 	return (error);
1624 }
1625 
1626 /*
1627  * Do a readlink rpc.
1628  */
1629 int
1630 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1631     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1632 {
1633 	u_int32_t *tl;
1634 	struct nfsrv_descript nfsd, *nd = &nfsd;
1635 	struct nfsnode *np = VTONFS(vp);
1636 	nfsattrbit_t attrbits;
1637 	int error, len, cangetattr = 1;
1638 
1639 	*attrflagp = 0;
1640 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1641 	if (nd->nd_flag & ND_NFSV4) {
1642 		/*
1643 		 * And do a Getattr op.
1644 		 */
1645 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1646 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1647 		NFSGETATTR_ATTRBIT(&attrbits);
1648 		(void) nfsrv_putattrbit(nd, &attrbits);
1649 	}
1650 	error = nfscl_request(nd, vp, p, cred, stuff);
1651 	if (error)
1652 		return (error);
1653 	if (nd->nd_flag & ND_NFSV3)
1654 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1655 	if (!nd->nd_repstat && !error) {
1656 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1657 		/*
1658 		 * This seems weird to me, but must have been added to
1659 		 * FreeBSD for some reason. The only thing I can think of
1660 		 * is that there was/is some server that replies with
1661 		 * more link data than it should?
1662 		 */
1663 		if (len == NFS_MAXPATHLEN) {
1664 			NFSLOCKNODE(np);
1665 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1666 				len = np->n_size;
1667 				cangetattr = 0;
1668 			}
1669 			NFSUNLOCKNODE(np);
1670 		}
1671 		error = nfsm_mbufuio(nd, uiop, len);
1672 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1673 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1674 	}
1675 	if (nd->nd_repstat && !error)
1676 		error = nd->nd_repstat;
1677 nfsmout:
1678 	m_freem(nd->nd_mrep);
1679 	return (error);
1680 }
1681 
1682 /*
1683  * Read operation.
1684  */
1685 int
1686 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1687     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1688 {
1689 	int error, expireret = 0, retrycnt;
1690 	u_int32_t clidrev = 0;
1691 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1692 	struct nfsnode *np = VTONFS(vp);
1693 	struct ucred *newcred;
1694 	struct nfsfh *nfhp = NULL;
1695 	nfsv4stateid_t stateid;
1696 	void *lckp;
1697 
1698 	if (nmp->nm_clp != NULL)
1699 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1700 	newcred = cred;
1701 	if (NFSHASNFSV4(nmp)) {
1702 		nfhp = np->n_fhp;
1703 		newcred = NFSNEWCRED(cred);
1704 	}
1705 	retrycnt = 0;
1706 	do {
1707 		lckp = NULL;
1708 		if (NFSHASNFSV4(nmp))
1709 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1710 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1711 			    &lckp);
1712 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1713 		    attrflagp, stuff);
1714 		if (error == NFSERR_OPENMODE) {
1715 			NFSLOCKMNT(nmp);
1716 			nmp->nm_state |= NFSSTA_OPENMODE;
1717 			NFSUNLOCKMNT(nmp);
1718 		}
1719 		if (error == NFSERR_STALESTATEID)
1720 			nfscl_initiate_recovery(nmp->nm_clp);
1721 		if (lckp != NULL)
1722 			nfscl_lockderef(lckp);
1723 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1724 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1725 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1726 			(void) nfs_catnap(PZERO, error, "nfs_read");
1727 		} else if ((error == NFSERR_EXPIRED ||
1728 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1729 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1730 		}
1731 		retrycnt++;
1732 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1733 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1734 	    error == NFSERR_BADSESSION ||
1735 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1736 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1737 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1738 	    (error == NFSERR_OPENMODE && retrycnt < 4));
1739 	if (error && retrycnt >= 4)
1740 		error = EIO;
1741 	if (NFSHASNFSV4(nmp))
1742 		NFSFREECRED(newcred);
1743 	return (error);
1744 }
1745 
1746 /*
1747  * The actual read RPC.
1748  */
1749 static int
1750 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1751     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1752     int *attrflagp, void *stuff)
1753 {
1754 	u_int32_t *tl;
1755 	int error = 0, len, retlen, tsiz, eof = 0;
1756 	struct nfsrv_descript nfsd;
1757 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1758 	struct nfsrv_descript *nd = &nfsd;
1759 	int rsize;
1760 	off_t tmp_off;
1761 
1762 	*attrflagp = 0;
1763 	tsiz = uiop->uio_resid;
1764 	tmp_off = uiop->uio_offset + tsiz;
1765 	NFSLOCKMNT(nmp);
1766 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1767 		NFSUNLOCKMNT(nmp);
1768 		return (EFBIG);
1769 	}
1770 	rsize = nmp->nm_rsize;
1771 	NFSUNLOCKMNT(nmp);
1772 	nd->nd_mrep = NULL;
1773 	while (tsiz > 0) {
1774 		*attrflagp = 0;
1775 		len = (tsiz > rsize) ? rsize : tsiz;
1776 		NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1777 		if (nd->nd_flag & ND_NFSV4)
1778 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1779 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1780 		if (nd->nd_flag & ND_NFSV2) {
1781 			*tl++ = txdr_unsigned(uiop->uio_offset);
1782 			*tl++ = txdr_unsigned(len);
1783 			*tl = 0;
1784 		} else {
1785 			txdr_hyper(uiop->uio_offset, tl);
1786 			*(tl + 2) = txdr_unsigned(len);
1787 		}
1788 		/*
1789 		 * Since I can't do a Getattr for NFSv4 for Write, there
1790 		 * doesn't seem any point in doing one here, either.
1791 		 * (See the comment in nfsrpc_writerpc() for more info.)
1792 		 */
1793 		error = nfscl_request(nd, vp, p, cred, stuff);
1794 		if (error)
1795 			return (error);
1796 		if (nd->nd_flag & ND_NFSV3) {
1797 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1798 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1799 			error = nfsm_loadattr(nd, nap);
1800 			if (!error)
1801 				*attrflagp = 1;
1802 		}
1803 		if (nd->nd_repstat || error) {
1804 			if (!error)
1805 				error = nd->nd_repstat;
1806 			goto nfsmout;
1807 		}
1808 		if (nd->nd_flag & ND_NFSV3) {
1809 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1810 			eof = fxdr_unsigned(int, *(tl + 1));
1811 		} else if (nd->nd_flag & ND_NFSV4) {
1812 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1813 			eof = fxdr_unsigned(int, *tl);
1814 		}
1815 		NFSM_STRSIZ(retlen, len);
1816 		error = nfsm_mbufuio(nd, uiop, retlen);
1817 		if (error)
1818 			goto nfsmout;
1819 		m_freem(nd->nd_mrep);
1820 		nd->nd_mrep = NULL;
1821 		tsiz -= retlen;
1822 		if (!(nd->nd_flag & ND_NFSV2)) {
1823 			if (eof || retlen == 0)
1824 				tsiz = 0;
1825 		} else if (retlen < len)
1826 			tsiz = 0;
1827 	}
1828 	return (0);
1829 nfsmout:
1830 	if (nd->nd_mrep != NULL)
1831 		m_freem(nd->nd_mrep);
1832 	return (error);
1833 }
1834 
1835 /*
1836  * nfs write operation
1837  * When called_from_strategy != 0, it should return EIO for an error that
1838  * indicates recovery is in progress, so that the buffer will be left
1839  * dirty and be written back to the server later. If it loops around,
1840  * the recovery thread could get stuck waiting for the buffer and recovery
1841  * will then deadlock.
1842  */
1843 int
1844 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1845     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1846     void *stuff, int called_from_strategy)
1847 {
1848 	int error, expireret = 0, retrycnt, nostateid;
1849 	u_int32_t clidrev = 0;
1850 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1851 	struct nfsnode *np = VTONFS(vp);
1852 	struct ucred *newcred;
1853 	struct nfsfh *nfhp = NULL;
1854 	nfsv4stateid_t stateid;
1855 	void *lckp;
1856 
1857 	*must_commit = 0;
1858 	if (nmp->nm_clp != NULL)
1859 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1860 	newcred = cred;
1861 	if (NFSHASNFSV4(nmp)) {
1862 		newcred = NFSNEWCRED(cred);
1863 		nfhp = np->n_fhp;
1864 	}
1865 	retrycnt = 0;
1866 	do {
1867 		lckp = NULL;
1868 		nostateid = 0;
1869 		if (NFSHASNFSV4(nmp)) {
1870 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1871 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1872 			    &lckp);
1873 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1874 			    stateid.other[2] == 0) {
1875 				nostateid = 1;
1876 				NFSCL_DEBUG(1, "stateid0 in write\n");
1877 			}
1878 		}
1879 
1880 		/*
1881 		 * If there is no stateid for NFSv4, it means this is an
1882 		 * extraneous write after close. Basically a poorly
1883 		 * implemented buffer cache. Just don't do the write.
1884 		 */
1885 		if (nostateid)
1886 			error = 0;
1887 		else
1888 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1889 			    newcred, &stateid, p, nap, attrflagp, stuff);
1890 		if (error == NFSERR_STALESTATEID)
1891 			nfscl_initiate_recovery(nmp->nm_clp);
1892 		if (lckp != NULL)
1893 			nfscl_lockderef(lckp);
1894 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1895 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1896 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1897 			(void) nfs_catnap(PZERO, error, "nfs_write");
1898 		} else if ((error == NFSERR_EXPIRED ||
1899 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1900 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1901 		}
1902 		retrycnt++;
1903 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1904 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1905 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1906 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1907 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1908 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1909 	if (error != 0 && (retrycnt >= 4 ||
1910 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1911 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1912 		error = EIO;
1913 	if (NFSHASNFSV4(nmp))
1914 		NFSFREECRED(newcred);
1915 	return (error);
1916 }
1917 
1918 /*
1919  * The actual write RPC.
1920  */
1921 static int
1922 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1923     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1924     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1925 {
1926 	u_int32_t *tl;
1927 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1928 	struct nfsnode *np = VTONFS(vp);
1929 	int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1930 	int wccflag = 0, wsize;
1931 	int32_t backup;
1932 	struct nfsrv_descript nfsd;
1933 	struct nfsrv_descript *nd = &nfsd;
1934 	nfsattrbit_t attrbits;
1935 	off_t tmp_off;
1936 
1937 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1938 	*attrflagp = 0;
1939 	tsiz = uiop->uio_resid;
1940 	tmp_off = uiop->uio_offset + tsiz;
1941 	NFSLOCKMNT(nmp);
1942 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1943 		NFSUNLOCKMNT(nmp);
1944 		return (EFBIG);
1945 	}
1946 	wsize = nmp->nm_wsize;
1947 	NFSUNLOCKMNT(nmp);
1948 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
1949 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
1950 	while (tsiz > 0) {
1951 		*attrflagp = 0;
1952 		len = (tsiz > wsize) ? wsize : tsiz;
1953 		NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1954 		if (nd->nd_flag & ND_NFSV4) {
1955 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1956 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1957 			txdr_hyper(uiop->uio_offset, tl);
1958 			tl += 2;
1959 			*tl++ = txdr_unsigned(*iomode);
1960 			*tl = txdr_unsigned(len);
1961 		} else if (nd->nd_flag & ND_NFSV3) {
1962 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1963 			txdr_hyper(uiop->uio_offset, tl);
1964 			tl += 2;
1965 			*tl++ = txdr_unsigned(len);
1966 			*tl++ = txdr_unsigned(*iomode);
1967 			*tl = txdr_unsigned(len);
1968 		} else {
1969 			u_int32_t x;
1970 
1971 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1972 			/*
1973 			 * Not sure why someone changed this, since the
1974 			 * RFC clearly states that "beginoffset" and
1975 			 * "totalcount" are ignored, but it wouldn't
1976 			 * surprise me if there's a busted server out there.
1977 			 */
1978 			/* Set both "begin" and "current" to non-garbage. */
1979 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1980 			*tl++ = x;      /* "begin offset" */
1981 			*tl++ = x;      /* "current offset" */
1982 			x = txdr_unsigned(len);
1983 			*tl++ = x;      /* total to this offset */
1984 			*tl = x;        /* size of this write */
1985 		}
1986 		nfsm_uiombuf(nd, uiop, len);
1987 		/*
1988 		 * Although it is tempting to do a normal Getattr Op in the
1989 		 * NFSv4 compound, the result can be a nearly hung client
1990 		 * system if the Getattr asks for Owner and/or OwnerGroup.
1991 		 * It occurs when the client can't map either the Owner or
1992 		 * Owner_group name in the Getattr reply to a uid/gid. When
1993 		 * there is a cache miss, the kernel does an upcall to the
1994 		 * nfsuserd. Then, it can try and read the local /etc/passwd
1995 		 * or /etc/group file. It can then block in getnewbuf(),
1996 		 * waiting for dirty writes to be pushed to the NFS server.
1997 		 * The only reason this doesn't result in a complete
1998 		 * deadlock, is that the upcall times out and allows
1999 		 * the write to complete. However, progress is so slow
2000 		 * that it might just as well be deadlocked.
2001 		 * As such, we get the rest of the attributes, but not
2002 		 * Owner or Owner_group.
2003 		 * nb: nfscl_loadattrcache() needs to be told that these
2004 		 *     partial attributes from a write rpc are being
2005 		 *     passed in, via a argument flag.
2006 		 */
2007 		if (nd->nd_flag & ND_NFSV4) {
2008 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
2009 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2010 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2011 			(void) nfsrv_putattrbit(nd, &attrbits);
2012 		}
2013 		error = nfscl_request(nd, vp, p, cred, stuff);
2014 		if (error)
2015 			return (error);
2016 		if (nd->nd_repstat) {
2017 			/*
2018 			 * In case the rpc gets retried, roll
2019 			 * the uio fields changed by nfsm_uiombuf()
2020 			 * back.
2021 			 */
2022 			uiop->uio_offset -= len;
2023 			uiop->uio_resid += len;
2024 			uiop->uio_iov->iov_base =
2025 			    (char *)uiop->uio_iov->iov_base - len;
2026 			uiop->uio_iov->iov_len += len;
2027 		}
2028 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2029 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
2030 			    &wccflag, stuff);
2031 			if (error)
2032 				goto nfsmout;
2033 		}
2034 		if (!nd->nd_repstat) {
2035 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2036 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
2037 					+ NFSX_VERF);
2038 				rlen = fxdr_unsigned(int, *tl++);
2039 				if (rlen == 0) {
2040 					error = NFSERR_IO;
2041 					goto nfsmout;
2042 				} else if (rlen < len) {
2043 					backup = len - rlen;
2044 					uiop->uio_iov->iov_base =
2045 					    (char *)uiop->uio_iov->iov_base -
2046 					    backup;
2047 					uiop->uio_iov->iov_len += backup;
2048 					uiop->uio_offset -= backup;
2049 					uiop->uio_resid += backup;
2050 					len = rlen;
2051 				}
2052 				commit = fxdr_unsigned(int, *tl++);
2053 
2054 				/*
2055 				 * Return the lowest commitment level
2056 				 * obtained by any of the RPCs.
2057 				 */
2058 				if (committed == NFSWRITE_FILESYNC)
2059 					committed = commit;
2060 				else if (committed == NFSWRITE_DATASYNC &&
2061 					commit == NFSWRITE_UNSTABLE)
2062 					committed = commit;
2063 				NFSLOCKMNT(nmp);
2064 				if (!NFSHASWRITEVERF(nmp)) {
2065 					NFSBCOPY((caddr_t)tl,
2066 					    (caddr_t)&nmp->nm_verf[0],
2067 					    NFSX_VERF);
2068 					NFSSETWRITEVERF(nmp);
2069 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
2070 				    NFSX_VERF)) {
2071 					*must_commit = 1;
2072 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
2073 				}
2074 				NFSUNLOCKMNT(nmp);
2075 			}
2076 			if (nd->nd_flag & ND_NFSV4)
2077 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2078 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
2079 				error = nfsm_loadattr(nd, nap);
2080 				if (!error)
2081 					*attrflagp = NFS_LATTR_NOSHRINK;
2082 			}
2083 		} else {
2084 			error = nd->nd_repstat;
2085 		}
2086 		if (error)
2087 			goto nfsmout;
2088 		NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
2089 		m_freem(nd->nd_mrep);
2090 		nd->nd_mrep = NULL;
2091 		tsiz -= len;
2092 	}
2093 nfsmout:
2094 	if (nd->nd_mrep != NULL)
2095 		m_freem(nd->nd_mrep);
2096 	*iomode = committed;
2097 	if (nd->nd_repstat && !error)
2098 		error = nd->nd_repstat;
2099 	return (error);
2100 }
2101 
2102 /*
2103  * Do an nfs deallocate operation.
2104  */
2105 int
2106 nfsrpc_deallocate(vnode_t vp, off_t offs, off_t len, struct nfsvattr *nap,
2107     int *attrflagp, struct ucred *cred, NFSPROC_T *p, void *stuff)
2108 {
2109 	int error, expireret = 0, openerr, retrycnt;
2110 	uint32_t clidrev = 0;
2111 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
2112 	struct nfsfh *nfhp;
2113 	nfsv4stateid_t stateid;
2114 	void *lckp;
2115 
2116 	if (nmp->nm_clp != NULL)
2117 		clidrev = nmp->nm_clp->nfsc_clientidrev;
2118 	retrycnt = 0;
2119 	do {
2120 		lckp = NULL;
2121 		openerr = 1;
2122 		nfhp = VTONFS(vp)->n_fhp;
2123 		error = nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
2124 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
2125 		if (error != 0) {
2126 			/*
2127 			 * No Open stateid, so try and open the file
2128 			 * now.
2129 			 */
2130 			openerr = nfsrpc_open(vp, FWRITE, cred, p);
2131 			if (openerr == 0)
2132 				nfscl_getstateid(vp, nfhp->nfh_fh,
2133 				    nfhp->nfh_len, NFSV4OPEN_ACCESSWRITE, 0,
2134 				    cred, p, &stateid, &lckp);
2135 		}
2136 		error = nfsrpc_deallocaterpc(vp, offs, len, &stateid, nap,
2137 		    attrflagp, cred, p, stuff);
2138 		if (error == NFSERR_STALESTATEID)
2139 			nfscl_initiate_recovery(nmp->nm_clp);
2140 		if (lckp != NULL)
2141 			nfscl_lockderef(lckp);
2142 		if (openerr == 0)
2143 			nfsrpc_close(vp, 0, p);
2144 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2145 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2146 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
2147 			(void) nfs_catnap(PZERO, error, "nfs_deallocate");
2148 		} else if ((error == NFSERR_EXPIRED ||
2149 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2150 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2151 		}
2152 		retrycnt++;
2153 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2154 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2155 	    error == NFSERR_BADSESSION ||
2156 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
2157 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2158 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
2159 	if (error && retrycnt >= 4)
2160 		error = EIO;
2161 	return (error);
2162 }
2163 
2164 /*
2165  * The actual deallocate RPC.
2166  */
2167 static int
2168 nfsrpc_deallocaterpc(vnode_t vp, off_t offs, off_t len,
2169     nfsv4stateid_t *stateidp, struct nfsvattr *nap, int *attrflagp,
2170     struct ucred *cred, NFSPROC_T *p, void *stuff)
2171 {
2172 	uint32_t *tl;
2173 	struct nfsnode *np = VTONFS(vp);
2174 	int error, wccflag;
2175 	struct nfsrv_descript nfsd;
2176 	struct nfsrv_descript *nd = &nfsd;
2177 	nfsattrbit_t attrbits;
2178 
2179 	*attrflagp = 0;
2180 	NFSCL_REQSTART(nd, NFSPROC_DEALLOCATE, vp);
2181 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
2182 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
2183 	txdr_hyper(offs, tl);
2184 	tl += 2;
2185 	txdr_hyper(len, tl);
2186 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
2187 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
2188 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2189 	nfsrv_putattrbit(nd, &attrbits);
2190 	error = nfscl_request(nd, vp, p, cred, stuff);
2191 	if (error != 0)
2192 		return (error);
2193 	wccflag = 0;
2194 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, &wccflag, stuff);
2195 	if (error != 0)
2196 		goto nfsmout;
2197 	if (nd->nd_repstat == 0) {
2198 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
2199 		error = nfsm_loadattr(nd, nap);
2200 		if (error != 0)
2201 			goto nfsmout;
2202 		*attrflagp = NFS_LATTR_NOSHRINK;
2203 	}
2204 	NFSWRITERPC_SETTIME(wccflag, np, nap, 1);
2205 nfsmout:
2206 	m_freem(nd->nd_mrep);
2207 	if (nd->nd_repstat != 0 && error == 0)
2208 		error = nd->nd_repstat;
2209 	return (error);
2210 }
2211 
2212 /*
2213  * nfs mknod rpc
2214  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
2215  * mode set to specify the file type and the size field for rdev.
2216  */
2217 int
2218 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2219     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
2220     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2221     int *attrflagp, int *dattrflagp, void *dstuff)
2222 {
2223 	u_int32_t *tl;
2224 	int error = 0;
2225 	struct nfsrv_descript nfsd, *nd = &nfsd;
2226 	nfsattrbit_t attrbits;
2227 
2228 	*nfhpp = NULL;
2229 	*attrflagp = 0;
2230 	*dattrflagp = 0;
2231 	if (namelen > NFS_MAXNAMLEN)
2232 		return (ENAMETOOLONG);
2233 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
2234 	if (nd->nd_flag & ND_NFSV4) {
2235 		if (vtyp == VBLK || vtyp == VCHR) {
2236 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
2237 			*tl++ = vtonfsv34_type(vtyp);
2238 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2239 			*tl = txdr_unsigned(NFSMINOR(rdev));
2240 		} else {
2241 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2242 			*tl = vtonfsv34_type(vtyp);
2243 		}
2244 	}
2245 	(void) nfsm_strtom(nd, name, namelen);
2246 	if (nd->nd_flag & ND_NFSV3) {
2247 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2248 		*tl = vtonfsv34_type(vtyp);
2249 	}
2250 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2251 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2252 	if ((nd->nd_flag & ND_NFSV3) &&
2253 	    (vtyp == VCHR || vtyp == VBLK)) {
2254 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2255 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2256 		*tl = txdr_unsigned(NFSMINOR(rdev));
2257 	}
2258 	if (nd->nd_flag & ND_NFSV4) {
2259 		NFSGETATTR_ATTRBIT(&attrbits);
2260 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2261 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2262 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2263 		(void) nfsrv_putattrbit(nd, &attrbits);
2264 	}
2265 	if (nd->nd_flag & ND_NFSV2)
2266 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
2267 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2268 	if (error)
2269 		return (error);
2270 	if (nd->nd_flag & ND_NFSV4)
2271 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2272 	if (!nd->nd_repstat) {
2273 		if (nd->nd_flag & ND_NFSV4) {
2274 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2275 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2276 			if (error)
2277 				goto nfsmout;
2278 		}
2279 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2280 		if (error)
2281 			goto nfsmout;
2282 	}
2283 	if (nd->nd_flag & ND_NFSV3)
2284 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2285 	if (!error && nd->nd_repstat)
2286 		error = nd->nd_repstat;
2287 nfsmout:
2288 	m_freem(nd->nd_mrep);
2289 	return (error);
2290 }
2291 
2292 /*
2293  * nfs file create call
2294  * Mostly just call the approriate routine. (I separated out v4, so that
2295  * error recovery wouldn't be as difficult.)
2296  */
2297 int
2298 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2299     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2300     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2301     int *attrflagp, int *dattrflagp, void *dstuff)
2302 {
2303 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
2304 	struct nfsclowner *owp;
2305 	struct nfscldeleg *dp;
2306 	struct nfsmount *nmp = VFSTONFS(dvp->v_mount);
2307 	u_int32_t clidrev;
2308 
2309 	if (NFSHASNFSV4(nmp)) {
2310 	    retrycnt = 0;
2311 	    do {
2312 		dp = NULL;
2313 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
2314 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
2315 		    NULL, 1, true);
2316 		if (error)
2317 			return (error);
2318 		if (nmp->nm_clp != NULL)
2319 			clidrev = nmp->nm_clp->nfsc_clientidrev;
2320 		else
2321 			clidrev = 0;
2322 		if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
2323 		    nfs_numnfscbd == 0 || retrycnt > 0)
2324 			error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
2325 			  fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2326 			  attrflagp, dattrflagp, dstuff, &unlocked);
2327 		else
2328 			error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
2329 			  cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2330 			  attrflagp, dattrflagp, dstuff, &unlocked);
2331 		/*
2332 		 * There is no need to invalidate cached attributes here,
2333 		 * since new post-delegation issue attributes are always
2334 		 * returned by nfsrpc_createv4() and these will update the
2335 		 * attribute cache.
2336 		 */
2337 		if (dp != NULL)
2338 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2339 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2340 		nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2341 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2342 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2343 		    error == NFSERR_BADSESSION) {
2344 			(void) nfs_catnap(PZERO, error, "nfs_open");
2345 		} else if ((error == NFSERR_EXPIRED ||
2346 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2347 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2348 			retrycnt++;
2349 		}
2350 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2351 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2352 		error == NFSERR_BADSESSION ||
2353 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2354 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
2355 	    if (error && retrycnt >= 4)
2356 		    error = EIO;
2357 	} else {
2358 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2359 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
2360 		    dstuff);
2361 	}
2362 	return (error);
2363 }
2364 
2365 /*
2366  * The create rpc for v2 and 3.
2367  */
2368 static int
2369 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2370     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2371     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2372     int *attrflagp, int *dattrflagp, void *dstuff)
2373 {
2374 	u_int32_t *tl;
2375 	int error = 0;
2376 	struct nfsrv_descript nfsd, *nd = &nfsd;
2377 
2378 	*nfhpp = NULL;
2379 	*attrflagp = 0;
2380 	*dattrflagp = 0;
2381 	if (namelen > NFS_MAXNAMLEN)
2382 		return (ENAMETOOLONG);
2383 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2384 	(void) nfsm_strtom(nd, name, namelen);
2385 	if (nd->nd_flag & ND_NFSV3) {
2386 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2387 		if (fmode & O_EXCL) {
2388 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2389 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2390 			*tl++ = cverf.lval[0];
2391 			*tl = cverf.lval[1];
2392 		} else {
2393 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2394 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
2395 		}
2396 	} else {
2397 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2398 	}
2399 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2400 	if (error)
2401 		return (error);
2402 	if (nd->nd_repstat == 0) {
2403 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2404 		if (error)
2405 			goto nfsmout;
2406 	}
2407 	if (nd->nd_flag & ND_NFSV3)
2408 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2409 	if (nd->nd_repstat != 0 && error == 0)
2410 		error = nd->nd_repstat;
2411 nfsmout:
2412 	m_freem(nd->nd_mrep);
2413 	return (error);
2414 }
2415 
2416 static int
2417 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2418     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2419     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2420     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2421     int *dattrflagp, void *dstuff, int *unlockedp)
2422 {
2423 	u_int32_t *tl;
2424 	int error = 0, deleg, newone, ret, acesize, limitby;
2425 	struct nfsrv_descript nfsd, *nd = &nfsd;
2426 	struct nfsclopen *op;
2427 	struct nfscldeleg *dp = NULL;
2428 	struct nfsnode *np;
2429 	struct nfsfh *nfhp;
2430 	nfsattrbit_t attrbits;
2431 	nfsv4stateid_t stateid;
2432 	u_int32_t rflags;
2433 	struct nfsmount *nmp;
2434 	struct nfsclsession *tsep;
2435 
2436 	nmp = VFSTONFS(dvp->v_mount);
2437 	np = VTONFS(dvp);
2438 	*unlockedp = 0;
2439 	*nfhpp = NULL;
2440 	*dpp = NULL;
2441 	*attrflagp = 0;
2442 	*dattrflagp = 0;
2443 	if (namelen > NFS_MAXNAMLEN)
2444 		return (ENAMETOOLONG);
2445 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2446 	/*
2447 	 * For V4, this is actually an Open op.
2448 	 */
2449 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2450 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
2451 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2452 	    NFSV4OPEN_ACCESSREAD);
2453 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2454 	tsep = nfsmnt_mdssession(nmp);
2455 	*tl++ = tsep->nfsess_clientid.lval[0];
2456 	*tl = tsep->nfsess_clientid.lval[1];
2457 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2458 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2459 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2460 	if (fmode & O_EXCL) {
2461 		if (NFSHASNFSV4N(nmp)) {
2462 			if (NFSHASSESSPERSIST(nmp)) {
2463 				/* Use GUARDED for persistent sessions. */
2464 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
2465 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2466 			} else {
2467 				/* Otherwise, use EXCLUSIVE4_1. */
2468 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2469 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2470 				*tl++ = cverf.lval[0];
2471 				*tl = cverf.lval[1];
2472 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2473 			}
2474 		} else {
2475 			/* NFSv4.0 */
2476 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2477 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2478 			*tl++ = cverf.lval[0];
2479 			*tl = cverf.lval[1];
2480 		}
2481 	} else {
2482 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2483 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2484 	}
2485 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2486 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2487 	(void) nfsm_strtom(nd, name, namelen);
2488 	/* Get the new file's handle and attributes. */
2489 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2490 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2491 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2492 	NFSGETATTR_ATTRBIT(&attrbits);
2493 	(void) nfsrv_putattrbit(nd, &attrbits);
2494 	/* Get the directory's post-op attributes. */
2495 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2496 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2497 	(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2498 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2499 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2500 	(void) nfsrv_putattrbit(nd, &attrbits);
2501 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2502 	if (error)
2503 		return (error);
2504 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2505 	if (nd->nd_repstat == 0) {
2506 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2507 		    6 * NFSX_UNSIGNED);
2508 		stateid.seqid = *tl++;
2509 		stateid.other[0] = *tl++;
2510 		stateid.other[1] = *tl++;
2511 		stateid.other[2] = *tl;
2512 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2513 		(void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2514 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2515 		deleg = fxdr_unsigned(int, *tl);
2516 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2517 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2518 			if (!(owp->nfsow_clp->nfsc_flags &
2519 			      NFSCLFLAGS_FIRSTDELEG))
2520 				owp->nfsow_clp->nfsc_flags |=
2521 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2522 			dp = malloc(
2523 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2524 			    M_NFSCLDELEG, M_WAITOK);
2525 			LIST_INIT(&dp->nfsdl_owner);
2526 			LIST_INIT(&dp->nfsdl_lock);
2527 			dp->nfsdl_clp = owp->nfsow_clp;
2528 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2529 			nfscl_lockinit(&dp->nfsdl_rwlock);
2530 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2531 			    NFSX_UNSIGNED);
2532 			dp->nfsdl_stateid.seqid = *tl++;
2533 			dp->nfsdl_stateid.other[0] = *tl++;
2534 			dp->nfsdl_stateid.other[1] = *tl++;
2535 			dp->nfsdl_stateid.other[2] = *tl++;
2536 			ret = fxdr_unsigned(int, *tl);
2537 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2538 				dp->nfsdl_flags = NFSCLDL_WRITE;
2539 				/*
2540 				 * Indicates how much the file can grow.
2541 				 */
2542 				NFSM_DISSECT(tl, u_int32_t *,
2543 				    3 * NFSX_UNSIGNED);
2544 				limitby = fxdr_unsigned(int, *tl++);
2545 				switch (limitby) {
2546 				case NFSV4OPEN_LIMITSIZE:
2547 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2548 					break;
2549 				case NFSV4OPEN_LIMITBLOCKS:
2550 					dp->nfsdl_sizelimit =
2551 					    fxdr_unsigned(u_int64_t, *tl++);
2552 					dp->nfsdl_sizelimit *=
2553 					    fxdr_unsigned(u_int64_t, *tl);
2554 					break;
2555 				default:
2556 					error = NFSERR_BADXDR;
2557 					goto nfsmout;
2558 				}
2559 			} else {
2560 				dp->nfsdl_flags = NFSCLDL_READ;
2561 			}
2562 			if (ret)
2563 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2564 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2565 			    &acesize, p);
2566 			if (error)
2567 				goto nfsmout;
2568 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2569 			error = NFSERR_BADXDR;
2570 			goto nfsmout;
2571 		}
2572 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2573 		if (error)
2574 			goto nfsmout;
2575 		/* Get rid of the PutFH and Getattr status values. */
2576 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2577 		/* Load the directory attributes. */
2578 		error = nfsm_loadattr(nd, dnap);
2579 		if (error)
2580 			goto nfsmout;
2581 		*dattrflagp = 1;
2582 		if (dp != NULL && *attrflagp) {
2583 			dp->nfsdl_change = nnap->na_filerev;
2584 			dp->nfsdl_modtime = nnap->na_mtime;
2585 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2586 		}
2587 		/*
2588 		 * We can now complete the Open state.
2589 		 */
2590 		nfhp = *nfhpp;
2591 		if (dp != NULL) {
2592 			dp->nfsdl_fhlen = nfhp->nfh_len;
2593 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2594 		}
2595 		/*
2596 		 * Get an Open structure that will be
2597 		 * attached to the OpenOwner, acquired already.
2598 		 */
2599 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2600 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2601 		    cred, p, NULL, &op, &newone, NULL, 0, false);
2602 		if (error)
2603 			goto nfsmout;
2604 		op->nfso_stateid = stateid;
2605 		newnfs_copyincred(cred, &op->nfso_cred);
2606 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2607 		    do {
2608 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2609 			    nfhp->nfh_len, op, cred, p);
2610 			if (ret == NFSERR_DELAY)
2611 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2612 		    } while (ret == NFSERR_DELAY);
2613 		    error = ret;
2614 		}
2615 
2616 		/*
2617 		 * If the server is handing out delegations, but we didn't
2618 		 * get one because an OpenConfirm was required, try the
2619 		 * Open again, to get a delegation. This is a harmless no-op,
2620 		 * from a server's point of view.
2621 		 */
2622 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2623 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2624 		    !error && dp == NULL) {
2625 		    do {
2626 			ret = nfsrpc_openrpc(VFSTONFS(dvp->v_mount), dvp,
2627 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2628 			    nfhp->nfh_fh, nfhp->nfh_len,
2629 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2630 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2631 			if (ret == NFSERR_DELAY)
2632 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2633 		    } while (ret == NFSERR_DELAY);
2634 		    if (ret) {
2635 			if (dp != NULL) {
2636 				free(dp, M_NFSCLDELEG);
2637 				dp = NULL;
2638 			}
2639 			if (ret == NFSERR_STALECLIENTID ||
2640 			    ret == NFSERR_STALEDONTRECOVER ||
2641 			    ret == NFSERR_BADSESSION)
2642 				error = ret;
2643 		    }
2644 		}
2645 		nfscl_openrelease(nmp, op, error, newone);
2646 		*unlockedp = 1;
2647 	}
2648 	if (nd->nd_repstat != 0 && error == 0)
2649 		error = nd->nd_repstat;
2650 	if (error == NFSERR_STALECLIENTID)
2651 		nfscl_initiate_recovery(owp->nfsow_clp);
2652 nfsmout:
2653 	if (!error)
2654 		*dpp = dp;
2655 	else if (dp != NULL)
2656 		free(dp, M_NFSCLDELEG);
2657 	m_freem(nd->nd_mrep);
2658 	return (error);
2659 }
2660 
2661 /*
2662  * Nfs remove rpc
2663  */
2664 int
2665 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2666     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2667     void *dstuff)
2668 {
2669 	u_int32_t *tl;
2670 	struct nfsrv_descript nfsd, *nd = &nfsd;
2671 	struct nfsnode *np;
2672 	struct nfsmount *nmp;
2673 	nfsv4stateid_t dstateid;
2674 	int error, ret = 0, i;
2675 
2676 	*dattrflagp = 0;
2677 	if (namelen > NFS_MAXNAMLEN)
2678 		return (ENAMETOOLONG);
2679 	nmp = VFSTONFS(dvp->v_mount);
2680 tryagain:
2681 	if (NFSHASNFSV4(nmp) && ret == 0) {
2682 		ret = nfscl_removedeleg(vp, p, &dstateid);
2683 		if (ret == 1) {
2684 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2685 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2686 			    NFSX_UNSIGNED);
2687 			if (NFSHASNFSV4N(nmp))
2688 				*tl++ = 0;
2689 			else
2690 				*tl++ = dstateid.seqid;
2691 			*tl++ = dstateid.other[0];
2692 			*tl++ = dstateid.other[1];
2693 			*tl++ = dstateid.other[2];
2694 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2695 			np = VTONFS(dvp);
2696 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2697 			    np->n_fhp->nfh_len, 0);
2698 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2699 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2700 		}
2701 	} else {
2702 		ret = 0;
2703 	}
2704 	if (ret == 0)
2705 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2706 	(void) nfsm_strtom(nd, name, namelen);
2707 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2708 	if (error)
2709 		return (error);
2710 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2711 		/* For NFSv4, parse out any Delereturn replies. */
2712 		if (ret > 0 && nd->nd_repstat != 0 &&
2713 		    (nd->nd_flag & ND_NOMOREDATA)) {
2714 			/*
2715 			 * If the Delegreturn failed, try again without
2716 			 * it. The server will Recall, as required.
2717 			 */
2718 			m_freem(nd->nd_mrep);
2719 			goto tryagain;
2720 		}
2721 		for (i = 0; i < (ret * 2); i++) {
2722 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2723 			    ND_NFSV4) {
2724 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2725 			    if (*(tl + 1))
2726 				nd->nd_flag |= ND_NOMOREDATA;
2727 			}
2728 		}
2729 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2730 	}
2731 	if (nd->nd_repstat && !error)
2732 		error = nd->nd_repstat;
2733 nfsmout:
2734 	m_freem(nd->nd_mrep);
2735 	return (error);
2736 }
2737 
2738 /*
2739  * Do an nfs rename rpc.
2740  */
2741 int
2742 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2743     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2744     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2745     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2746 {
2747 	u_int32_t *tl;
2748 	struct nfsrv_descript nfsd, *nd = &nfsd;
2749 	struct nfsmount *nmp;
2750 	struct nfsnode *np;
2751 	nfsattrbit_t attrbits;
2752 	nfsv4stateid_t fdstateid, tdstateid;
2753 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2754 
2755 	*fattrflagp = 0;
2756 	*tattrflagp = 0;
2757 	nmp = VFSTONFS(fdvp->v_mount);
2758 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2759 		return (ENAMETOOLONG);
2760 tryagain:
2761 	if (NFSHASNFSV4(nmp) && ret == 0) {
2762 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2763 		    &tdstateid, &gottd, p);
2764 		if (gotfd && gottd) {
2765 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2766 		} else if (gotfd) {
2767 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2768 		} else if (gottd) {
2769 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2770 		}
2771 		if (gotfd) {
2772 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2773 			if (NFSHASNFSV4N(nmp))
2774 				*tl++ = 0;
2775 			else
2776 				*tl++ = fdstateid.seqid;
2777 			*tl++ = fdstateid.other[0];
2778 			*tl++ = fdstateid.other[1];
2779 			*tl = fdstateid.other[2];
2780 			if (gottd) {
2781 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2782 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2783 				np = VTONFS(tvp);
2784 				(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2785 				    np->n_fhp->nfh_len, 0);
2786 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2787 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2788 			}
2789 		}
2790 		if (gottd) {
2791 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2792 			if (NFSHASNFSV4N(nmp))
2793 				*tl++ = 0;
2794 			else
2795 				*tl++ = tdstateid.seqid;
2796 			*tl++ = tdstateid.other[0];
2797 			*tl++ = tdstateid.other[1];
2798 			*tl = tdstateid.other[2];
2799 		}
2800 		if (ret > 0) {
2801 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2802 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2803 			np = VTONFS(fdvp);
2804 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2805 			    np->n_fhp->nfh_len, 0);
2806 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2807 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2808 		}
2809 	} else {
2810 		ret = 0;
2811 	}
2812 	if (ret == 0)
2813 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2814 	if (nd->nd_flag & ND_NFSV4) {
2815 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2816 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2817 		NFSWCCATTR_ATTRBIT(&attrbits);
2818 		(void) nfsrv_putattrbit(nd, &attrbits);
2819 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2820 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2821 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2822 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2823 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2824 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2825 		(void) nfsrv_putattrbit(nd, &attrbits);
2826 		nd->nd_flag |= ND_V4WCCATTR;
2827 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2828 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2829 	}
2830 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2831 	if (!(nd->nd_flag & ND_NFSV4))
2832 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2833 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2834 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2835 	error = nfscl_request(nd, fdvp, p, cred, fstuff);
2836 	if (error)
2837 		return (error);
2838 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2839 		/* For NFSv4, parse out any Delereturn replies. */
2840 		if (ret > 0 && nd->nd_repstat != 0 &&
2841 		    (nd->nd_flag & ND_NOMOREDATA)) {
2842 			/*
2843 			 * If the Delegreturn failed, try again without
2844 			 * it. The server will Recall, as required.
2845 			 */
2846 			m_freem(nd->nd_mrep);
2847 			goto tryagain;
2848 		}
2849 		for (i = 0; i < (ret * 2); i++) {
2850 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2851 			    ND_NFSV4) {
2852 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2853 			    if (*(tl + 1)) {
2854 				if (i == 0 && ret > 1) {
2855 				    /*
2856 				     * If the Delegreturn failed, try again
2857 				     * without it. The server will Recall, as
2858 				     * required.
2859 				     * If ret > 1, the first iteration of this
2860 				     * loop is the second DelegReturn result.
2861 				     */
2862 				    m_freem(nd->nd_mrep);
2863 				    goto tryagain;
2864 				} else {
2865 				    nd->nd_flag |= ND_NOMOREDATA;
2866 				}
2867 			    }
2868 			}
2869 		}
2870 		/* Now, the first wcc attribute reply. */
2871 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2872 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2873 			if (*(tl + 1))
2874 				nd->nd_flag |= ND_NOMOREDATA;
2875 		}
2876 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2877 		    fstuff);
2878 		/* and the second wcc attribute reply. */
2879 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2880 		    !error) {
2881 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2882 			if (*(tl + 1))
2883 				nd->nd_flag |= ND_NOMOREDATA;
2884 		}
2885 		if (!error)
2886 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2887 			    NULL, tstuff);
2888 	}
2889 	if (nd->nd_repstat && !error)
2890 		error = nd->nd_repstat;
2891 nfsmout:
2892 	m_freem(nd->nd_mrep);
2893 	return (error);
2894 }
2895 
2896 /*
2897  * nfs hard link create rpc
2898  */
2899 int
2900 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2901     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2902     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2903 {
2904 	u_int32_t *tl;
2905 	struct nfsrv_descript nfsd, *nd = &nfsd;
2906 	nfsattrbit_t attrbits;
2907 	int error = 0;
2908 
2909 	*attrflagp = 0;
2910 	*dattrflagp = 0;
2911 	if (namelen > NFS_MAXNAMLEN)
2912 		return (ENAMETOOLONG);
2913 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2914 	if (nd->nd_flag & ND_NFSV4) {
2915 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2916 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2917 	}
2918 	(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2919 		VTONFS(dvp)->n_fhp->nfh_len, 0);
2920 	if (nd->nd_flag & ND_NFSV4) {
2921 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2922 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2923 		NFSWCCATTR_ATTRBIT(&attrbits);
2924 		(void) nfsrv_putattrbit(nd, &attrbits);
2925 		nd->nd_flag |= ND_V4WCCATTR;
2926 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2927 		*tl = txdr_unsigned(NFSV4OP_LINK);
2928 	}
2929 	(void) nfsm_strtom(nd, name, namelen);
2930 	error = nfscl_request(nd, vp, p, cred, dstuff);
2931 	if (error)
2932 		return (error);
2933 	if (nd->nd_flag & ND_NFSV3) {
2934 		error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2935 		if (!error)
2936 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2937 			    NULL, dstuff);
2938 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2939 		/*
2940 		 * First, parse out the PutFH and Getattr result.
2941 		 */
2942 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2943 		if (!(*(tl + 1)))
2944 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2945 		if (*(tl + 1))
2946 			nd->nd_flag |= ND_NOMOREDATA;
2947 		/*
2948 		 * Get the pre-op attributes.
2949 		 */
2950 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2951 	}
2952 	if (nd->nd_repstat && !error)
2953 		error = nd->nd_repstat;
2954 nfsmout:
2955 	m_freem(nd->nd_mrep);
2956 	return (error);
2957 }
2958 
2959 /*
2960  * nfs symbolic link create rpc
2961  */
2962 int
2963 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
2964     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2965     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2966     int *dattrflagp, void *dstuff)
2967 {
2968 	u_int32_t *tl;
2969 	struct nfsrv_descript nfsd, *nd = &nfsd;
2970 	struct nfsmount *nmp;
2971 	int slen, error = 0;
2972 
2973 	*nfhpp = NULL;
2974 	*attrflagp = 0;
2975 	*dattrflagp = 0;
2976 	nmp = VFSTONFS(dvp->v_mount);
2977 	slen = strlen(target);
2978 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2979 		return (ENAMETOOLONG);
2980 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2981 	if (nd->nd_flag & ND_NFSV4) {
2982 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2983 		*tl = txdr_unsigned(NFLNK);
2984 		(void) nfsm_strtom(nd, target, slen);
2985 	}
2986 	(void) nfsm_strtom(nd, name, namelen);
2987 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2988 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2989 	if (!(nd->nd_flag & ND_NFSV4))
2990 		(void) nfsm_strtom(nd, target, slen);
2991 	if (nd->nd_flag & ND_NFSV2)
2992 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2993 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2994 	if (error)
2995 		return (error);
2996 	if (nd->nd_flag & ND_NFSV4)
2997 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2998 	if ((nd->nd_flag & ND_NFSV3) && !error) {
2999 		if (!nd->nd_repstat)
3000 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
3001 		if (!error)
3002 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
3003 			    NULL, dstuff);
3004 	}
3005 	if (nd->nd_repstat && !error)
3006 		error = nd->nd_repstat;
3007 	m_freem(nd->nd_mrep);
3008 	/*
3009 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
3010 	 * Only do this if vfs.nfs.ignore_eexist is set.
3011 	 * Never do this for NFSv4.1 or later minor versions, since sessions
3012 	 * should guarantee "exactly once" RPC semantics.
3013 	 */
3014 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
3015 	    nmp->nm_minorvers == 0))
3016 		error = 0;
3017 	return (error);
3018 }
3019 
3020 /*
3021  * nfs make dir rpc
3022  */
3023 int
3024 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
3025     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
3026     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
3027     int *dattrflagp, void *dstuff)
3028 {
3029 	u_int32_t *tl;
3030 	struct nfsrv_descript nfsd, *nd = &nfsd;
3031 	nfsattrbit_t attrbits;
3032 	int error = 0;
3033 	struct nfsfh *fhp;
3034 	struct nfsmount *nmp;
3035 
3036 	*nfhpp = NULL;
3037 	*attrflagp = 0;
3038 	*dattrflagp = 0;
3039 	nmp = VFSTONFS(dvp->v_mount);
3040 	fhp = VTONFS(dvp)->n_fhp;
3041 	if (namelen > NFS_MAXNAMLEN)
3042 		return (ENAMETOOLONG);
3043 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
3044 	if (nd->nd_flag & ND_NFSV4) {
3045 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3046 		*tl = txdr_unsigned(NFDIR);
3047 	}
3048 	(void) nfsm_strtom(nd, name, namelen);
3049 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
3050 	if (nd->nd_flag & ND_NFSV4) {
3051 		NFSGETATTR_ATTRBIT(&attrbits);
3052 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3053 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3054 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3055 		(void) nfsrv_putattrbit(nd, &attrbits);
3056 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3057 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
3058 		(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
3059 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3060 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3061 		(void) nfsrv_putattrbit(nd, &attrbits);
3062 	}
3063 	error = nfscl_request(nd, dvp, p, cred, dstuff);
3064 	if (error)
3065 		return (error);
3066 	if (nd->nd_flag & ND_NFSV4)
3067 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
3068 	if (!nd->nd_repstat && !error) {
3069 		if (nd->nd_flag & ND_NFSV4) {
3070 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3071 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
3072 		}
3073 		if (!error)
3074 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
3075 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
3076 			/* Get rid of the PutFH and Getattr status values. */
3077 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
3078 			/* Load the directory attributes. */
3079 			error = nfsm_loadattr(nd, dnap);
3080 			if (error == 0)
3081 				*dattrflagp = 1;
3082 		}
3083 	}
3084 	if ((nd->nd_flag & ND_NFSV3) && !error)
3085 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
3086 	if (nd->nd_repstat && !error)
3087 		error = nd->nd_repstat;
3088 nfsmout:
3089 	m_freem(nd->nd_mrep);
3090 	/*
3091 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
3092 	 * Only do this if vfs.nfs.ignore_eexist is set.
3093 	 * Never do this for NFSv4.1 or later minor versions, since sessions
3094 	 * should guarantee "exactly once" RPC semantics.
3095 	 */
3096 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
3097 	    nmp->nm_minorvers == 0))
3098 		error = 0;
3099 	return (error);
3100 }
3101 
3102 /*
3103  * nfs remove directory call
3104  */
3105 int
3106 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
3107     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
3108 {
3109 	struct nfsrv_descript nfsd, *nd = &nfsd;
3110 	int error = 0;
3111 
3112 	*dattrflagp = 0;
3113 	if (namelen > NFS_MAXNAMLEN)
3114 		return (ENAMETOOLONG);
3115 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
3116 	(void) nfsm_strtom(nd, name, namelen);
3117 	error = nfscl_request(nd, dvp, p, cred, dstuff);
3118 	if (error)
3119 		return (error);
3120 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
3121 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
3122 	if (nd->nd_repstat && !error)
3123 		error = nd->nd_repstat;
3124 	m_freem(nd->nd_mrep);
3125 	/*
3126 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
3127 	 */
3128 	if (error == ENOENT)
3129 		error = 0;
3130 	return (error);
3131 }
3132 
3133 /*
3134  * Readdir rpc.
3135  * Always returns with either uio_resid unchanged, if you are at the
3136  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
3137  * filled in.
3138  * I felt this would allow caching of directory blocks more easily
3139  * than returning a pertially filled block.
3140  * Directory offset cookies:
3141  * Oh my, what to do with them...
3142  * I can think of three ways to deal with them:
3143  * 1 - have the layer above these RPCs maintain a map between logical
3144  *     directory byte offsets and the NFS directory offset cookies
3145  * 2 - pass the opaque directory offset cookies up into userland
3146  *     and let the libc functions deal with them, via the system call
3147  * 3 - return them to userland in the "struct dirent", so future versions
3148  *     of libc can use them and do whatever is necessary to make things work
3149  *     above these rpc calls, in the meantime
3150  * For now, I do #3 by "hiding" the directory offset cookies after the
3151  * d_name field in struct dirent. This is space inside d_reclen that
3152  * will be ignored by anything that doesn't know about them.
3153  * The directory offset cookies are filled in as the last 8 bytes of
3154  * each directory entry, after d_name. Someday, the userland libc
3155  * functions may be able to use these. In the meantime, it satisfies
3156  * OpenBSD's requirements for cookies being returned.
3157  * If expects the directory offset cookie for the read to be in uio_offset
3158  * and returns the one for the next entry after this directory block in
3159  * there, as well.
3160  */
3161 int
3162 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3163     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3164     int *eofp, void *stuff)
3165 {
3166 	int len, left;
3167 	struct dirent *dp = NULL;
3168 	u_int32_t *tl;
3169 	nfsquad_t cookie, ncookie;
3170 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3171 	struct nfsnode *dnp = VTONFS(vp);
3172 	struct nfsvattr nfsva;
3173 	struct nfsrv_descript nfsd, *nd = &nfsd;
3174 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3175 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
3176 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3177 	char *cp;
3178 	nfsattrbit_t attrbits, dattrbits;
3179 	u_int32_t rderr, *tl2 = NULL;
3180 	size_t tresid;
3181 
3182 	KASSERT(uiop->uio_iovcnt == 1 &&
3183 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3184 	    ("nfs readdirrpc bad uio"));
3185 	ncookie.lval[0] = ncookie.lval[1] = 0;
3186 	/*
3187 	 * There is no point in reading a lot more than uio_resid, however
3188 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
3189 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
3190 	 * will never make readsize > nm_readdirsize.
3191 	 */
3192 	readsize = nmp->nm_readdirsize;
3193 	if (readsize > uiop->uio_resid)
3194 		readsize = uiop->uio_resid + DIRBLKSIZ;
3195 
3196 	*attrflagp = 0;
3197 	if (eofp)
3198 		*eofp = 0;
3199 	tresid = uiop->uio_resid;
3200 	cookie.lval[0] = cookiep->nfsuquad[0];
3201 	cookie.lval[1] = cookiep->nfsuquad[1];
3202 	nd->nd_mrep = NULL;
3203 
3204 	/*
3205 	 * For NFSv4, first create the "." and ".." entries.
3206 	 */
3207 	if (NFSHASNFSV4(nmp)) {
3208 		reqsize = 6 * NFSX_UNSIGNED;
3209 		NFSGETATTR_ATTRBIT(&dattrbits);
3210 		NFSZERO_ATTRBIT(&attrbits);
3211 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3212 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
3213 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3214 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3215 			NFSSETBIT_ATTRBIT(&attrbits,
3216 			    NFSATTRBIT_MOUNTEDONFILEID);
3217 			gotmnton = 1;
3218 		} else {
3219 			/*
3220 			 * Must fake it. Use the fileno, except when the
3221 			 * fsid is != to that of the directory. For that
3222 			 * case, generate a fake fileno that is not the same.
3223 			 */
3224 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3225 			gotmnton = 0;
3226 		}
3227 
3228 		/*
3229 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3230 		 */
3231 		if (uiop->uio_offset == 0) {
3232 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3233 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3234 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3235 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3236 			(void) nfsrv_putattrbit(nd, &attrbits);
3237 			error = nfscl_request(nd, vp, p, cred, stuff);
3238 			if (error)
3239 			    return (error);
3240 			dotfileid = 0;	/* Fake out the compiler. */
3241 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3242 			    error = nfsm_loadattr(nd, &nfsva);
3243 			    if (error != 0)
3244 				goto nfsmout;
3245 			    dotfileid = nfsva.na_fileid;
3246 			}
3247 			if (nd->nd_repstat == 0) {
3248 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3249 			    len = fxdr_unsigned(int, *(tl + 4));
3250 			    if (len > 0 && len <= NFSX_V4FHMAX)
3251 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3252 			    else
3253 				error = EPERM;
3254 			    if (!error) {
3255 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3256 				nfsva.na_mntonfileno = UINT64_MAX;
3257 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3258 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3259 				    NULL, NULL, NULL, p, cred);
3260 				if (error) {
3261 				    dotdotfileid = dotfileid;
3262 				} else if (gotmnton) {
3263 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3264 					dotdotfileid = nfsva.na_mntonfileno;
3265 				    else
3266 					dotdotfileid = nfsva.na_fileid;
3267 				} else if (nfsva.na_filesid[0] ==
3268 				    dnp->n_vattr.na_filesid[0] &&
3269 				    nfsva.na_filesid[1] ==
3270 				    dnp->n_vattr.na_filesid[1]) {
3271 				    dotdotfileid = nfsva.na_fileid;
3272 				} else {
3273 				    do {
3274 					fakefileno--;
3275 				    } while (fakefileno ==
3276 					nfsva.na_fileid);
3277 				    dotdotfileid = fakefileno;
3278 				}
3279 			    }
3280 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3281 			    /*
3282 			     * Lookupp returns NFSERR_NOENT when we are
3283 			     * at the root, so just use the current dir.
3284 			     */
3285 			    nd->nd_repstat = 0;
3286 			    dotdotfileid = dotfileid;
3287 			} else {
3288 			    error = nd->nd_repstat;
3289 			}
3290 			m_freem(nd->nd_mrep);
3291 			if (error)
3292 			    return (error);
3293 			nd->nd_mrep = NULL;
3294 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3295 			dp->d_pad0 = dp->d_pad1 = 0;
3296 			dp->d_off = 0;
3297 			dp->d_type = DT_DIR;
3298 			dp->d_fileno = dotfileid;
3299 			dp->d_namlen = 1;
3300 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3301 			dp->d_name[0] = '.';
3302 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3303 			/*
3304 			 * Just make these offset cookie 0.
3305 			 */
3306 			tl = (u_int32_t *)&dp->d_name[8];
3307 			*tl++ = 0;
3308 			*tl = 0;
3309 			blksiz += dp->d_reclen;
3310 			uiop->uio_resid -= dp->d_reclen;
3311 			uiop->uio_offset += dp->d_reclen;
3312 			uiop->uio_iov->iov_base =
3313 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3314 			uiop->uio_iov->iov_len -= dp->d_reclen;
3315 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3316 			dp->d_pad0 = dp->d_pad1 = 0;
3317 			dp->d_off = 0;
3318 			dp->d_type = DT_DIR;
3319 			dp->d_fileno = dotdotfileid;
3320 			dp->d_namlen = 2;
3321 			*((uint64_t *)dp->d_name) = 0;
3322 			dp->d_name[0] = '.';
3323 			dp->d_name[1] = '.';
3324 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3325 			/*
3326 			 * Just make these offset cookie 0.
3327 			 */
3328 			tl = (u_int32_t *)&dp->d_name[8];
3329 			*tl++ = 0;
3330 			*tl = 0;
3331 			blksiz += dp->d_reclen;
3332 			uiop->uio_resid -= dp->d_reclen;
3333 			uiop->uio_offset += dp->d_reclen;
3334 			uiop->uio_iov->iov_base =
3335 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3336 			uiop->uio_iov->iov_len -= dp->d_reclen;
3337 		}
3338 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3339 	} else {
3340 		reqsize = 5 * NFSX_UNSIGNED;
3341 	}
3342 
3343 	/*
3344 	 * Loop around doing readdir rpc's of size readsize.
3345 	 * The stopping criteria is EOF or buffer full.
3346 	 */
3347 	while (more_dirs && bigenough) {
3348 		*attrflagp = 0;
3349 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
3350 		if (nd->nd_flag & ND_NFSV2) {
3351 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3352 			*tl++ = cookie.lval[1];
3353 			*tl = txdr_unsigned(readsize);
3354 		} else {
3355 			NFSM_BUILD(tl, u_int32_t *, reqsize);
3356 			*tl++ = cookie.lval[0];
3357 			*tl++ = cookie.lval[1];
3358 			if (cookie.qval == 0) {
3359 				*tl++ = 0;
3360 				*tl++ = 0;
3361 			} else {
3362 				NFSLOCKNODE(dnp);
3363 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
3364 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
3365 				NFSUNLOCKNODE(dnp);
3366 			}
3367 			if (nd->nd_flag & ND_NFSV4) {
3368 				*tl++ = txdr_unsigned(readsize);
3369 				*tl = txdr_unsigned(readsize);
3370 				(void) nfsrv_putattrbit(nd, &attrbits);
3371 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3372 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
3373 				(void) nfsrv_putattrbit(nd, &dattrbits);
3374 			} else {
3375 				*tl = txdr_unsigned(readsize);
3376 			}
3377 		}
3378 		error = nfscl_request(nd, vp, p, cred, stuff);
3379 		if (error)
3380 			return (error);
3381 		if (!(nd->nd_flag & ND_NFSV2)) {
3382 			if (nd->nd_flag & ND_NFSV3)
3383 				error = nfscl_postop_attr(nd, nap, attrflagp,
3384 				    stuff);
3385 			if (!nd->nd_repstat && !error) {
3386 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3387 				NFSLOCKNODE(dnp);
3388 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
3389 				dnp->n_cookieverf.nfsuquad[1] = *tl;
3390 				NFSUNLOCKNODE(dnp);
3391 			}
3392 		}
3393 		if (nd->nd_repstat || error) {
3394 			if (!error)
3395 				error = nd->nd_repstat;
3396 			goto nfsmout;
3397 		}
3398 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3399 		more_dirs = fxdr_unsigned(int, *tl);
3400 		if (!more_dirs)
3401 			tryformoredirs = 0;
3402 
3403 		/* loop through the dir entries, doctoring them to 4bsd form */
3404 		while (more_dirs && bigenough) {
3405 			if (nd->nd_flag & ND_NFSV4) {
3406 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3407 				ncookie.lval[0] = *tl++;
3408 				ncookie.lval[1] = *tl++;
3409 				len = fxdr_unsigned(int, *tl);
3410 			} else if (nd->nd_flag & ND_NFSV3) {
3411 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3412 				nfsva.na_fileid = fxdr_hyper(tl);
3413 				tl += 2;
3414 				len = fxdr_unsigned(int, *tl);
3415 			} else {
3416 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3417 				nfsva.na_fileid = fxdr_unsigned(uint64_t,
3418 				    *tl++);
3419 				len = fxdr_unsigned(int, *tl);
3420 			}
3421 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3422 				error = EBADRPC;
3423 				goto nfsmout;
3424 			}
3425 			tlen = roundup2(len, 8);
3426 			if (tlen == len)
3427 				tlen += 8;  /* To ensure null termination. */
3428 			left = DIRBLKSIZ - blksiz;
3429 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3430 				NFSBZERO(uiop->uio_iov->iov_base, left);
3431 				dp->d_reclen += left;
3432 				uiop->uio_iov->iov_base =
3433 				    (char *)uiop->uio_iov->iov_base + left;
3434 				uiop->uio_iov->iov_len -= left;
3435 				uiop->uio_resid -= left;
3436 				uiop->uio_offset += left;
3437 				blksiz = 0;
3438 			}
3439 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3440 			    uiop->uio_resid)
3441 				bigenough = 0;
3442 			if (bigenough) {
3443 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3444 				dp->d_pad0 = dp->d_pad1 = 0;
3445 				dp->d_off = 0;
3446 				dp->d_namlen = len;
3447 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3448 				    NFSX_HYPER;
3449 				dp->d_type = DT_UNKNOWN;
3450 				blksiz += dp->d_reclen;
3451 				if (blksiz == DIRBLKSIZ)
3452 					blksiz = 0;
3453 				uiop->uio_resid -= DIRHDSIZ;
3454 				uiop->uio_offset += DIRHDSIZ;
3455 				uiop->uio_iov->iov_base =
3456 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3457 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3458 				error = nfsm_mbufuio(nd, uiop, len);
3459 				if (error)
3460 					goto nfsmout;
3461 				cp = uiop->uio_iov->iov_base;
3462 				tlen -= len;
3463 				NFSBZERO(cp, tlen);
3464 				cp += tlen;	/* points to cookie storage */
3465 				tl2 = (u_int32_t *)cp;
3466 				uiop->uio_iov->iov_base =
3467 				    (char *)uiop->uio_iov->iov_base + tlen +
3468 				    NFSX_HYPER;
3469 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3470 				uiop->uio_resid -= tlen + NFSX_HYPER;
3471 				uiop->uio_offset += (tlen + NFSX_HYPER);
3472 			} else {
3473 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3474 				if (error)
3475 					goto nfsmout;
3476 			}
3477 			if (nd->nd_flag & ND_NFSV4) {
3478 				rderr = 0;
3479 				nfsva.na_mntonfileno = UINT64_MAX;
3480 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3481 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3482 				    NULL, NULL, &rderr, p, cred);
3483 				if (error)
3484 					goto nfsmout;
3485 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3486 			} else if (nd->nd_flag & ND_NFSV3) {
3487 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3488 				ncookie.lval[0] = *tl++;
3489 				ncookie.lval[1] = *tl++;
3490 			} else {
3491 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3492 				ncookie.lval[0] = 0;
3493 				ncookie.lval[1] = *tl++;
3494 			}
3495 			if (bigenough) {
3496 			    if (nd->nd_flag & ND_NFSV4) {
3497 				if (rderr) {
3498 				    dp->d_fileno = 0;
3499 				} else {
3500 				    if (gotmnton) {
3501 					if (nfsva.na_mntonfileno != UINT64_MAX)
3502 					    dp->d_fileno = nfsva.na_mntonfileno;
3503 					else
3504 					    dp->d_fileno = nfsva.na_fileid;
3505 				    } else if (nfsva.na_filesid[0] ==
3506 					dnp->n_vattr.na_filesid[0] &&
3507 					nfsva.na_filesid[1] ==
3508 					dnp->n_vattr.na_filesid[1]) {
3509 					dp->d_fileno = nfsva.na_fileid;
3510 				    } else {
3511 					do {
3512 					    fakefileno--;
3513 					} while (fakefileno ==
3514 					    nfsva.na_fileid);
3515 					dp->d_fileno = fakefileno;
3516 				    }
3517 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3518 				}
3519 			    } else {
3520 				dp->d_fileno = nfsva.na_fileid;
3521 			    }
3522 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3523 				ncookie.lval[0];
3524 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3525 				ncookie.lval[1];
3526 			}
3527 			more_dirs = fxdr_unsigned(int, *tl);
3528 		}
3529 		/*
3530 		 * If at end of rpc data, get the eof boolean
3531 		 */
3532 		if (!more_dirs) {
3533 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3534 			eof = fxdr_unsigned(int, *tl);
3535 			if (tryformoredirs)
3536 				more_dirs = !eof;
3537 			if (nd->nd_flag & ND_NFSV4) {
3538 				error = nfscl_postop_attr(nd, nap, attrflagp,
3539 				    stuff);
3540 				if (error)
3541 					goto nfsmout;
3542 			}
3543 		}
3544 		m_freem(nd->nd_mrep);
3545 		nd->nd_mrep = NULL;
3546 	}
3547 	/*
3548 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3549 	 * by increasing d_reclen for the last record.
3550 	 */
3551 	if (blksiz > 0) {
3552 		left = DIRBLKSIZ - blksiz;
3553 		NFSBZERO(uiop->uio_iov->iov_base, left);
3554 		dp->d_reclen += left;
3555 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3556 		    left;
3557 		uiop->uio_iov->iov_len -= left;
3558 		uiop->uio_resid -= left;
3559 		uiop->uio_offset += left;
3560 	}
3561 
3562 	/*
3563 	 * If returning no data, assume end of file.
3564 	 * If not bigenough, return not end of file, since you aren't
3565 	 *    returning all the data
3566 	 * Otherwise, return the eof flag from the server.
3567 	 */
3568 	if (eofp) {
3569 		if (tresid == ((size_t)(uiop->uio_resid)))
3570 			*eofp = 1;
3571 		else if (!bigenough)
3572 			*eofp = 0;
3573 		else
3574 			*eofp = eof;
3575 	}
3576 
3577 	/*
3578 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3579 	 */
3580 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3581 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3582 		NFSBZERO(dp, DIRBLKSIZ);
3583 		dp->d_type = DT_UNKNOWN;
3584 		tl = (u_int32_t *)&dp->d_name[4];
3585 		*tl++ = cookie.lval[0];
3586 		*tl = cookie.lval[1];
3587 		dp->d_reclen = DIRBLKSIZ;
3588 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3589 		    DIRBLKSIZ;
3590 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3591 		uiop->uio_resid -= DIRBLKSIZ;
3592 		uiop->uio_offset += DIRBLKSIZ;
3593 	}
3594 
3595 nfsmout:
3596 	if (nd->nd_mrep != NULL)
3597 		m_freem(nd->nd_mrep);
3598 	return (error);
3599 }
3600 
3601 #ifndef APPLE
3602 /*
3603  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3604  * (Also used for NFS V4 when mount flag set.)
3605  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3606  */
3607 int
3608 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3609     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3610     int *eofp, void *stuff)
3611 {
3612 	int len, left;
3613 	struct dirent *dp = NULL;
3614 	u_int32_t *tl;
3615 	vnode_t newvp = NULLVP;
3616 	struct nfsrv_descript nfsd, *nd = &nfsd;
3617 	struct nameidata nami, *ndp = &nami;
3618 	struct componentname *cnp = &ndp->ni_cnd;
3619 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3620 	struct nfsnode *dnp = VTONFS(vp), *np;
3621 	struct nfsvattr nfsva;
3622 	struct nfsfh *nfhp;
3623 	nfsquad_t cookie, ncookie;
3624 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3625 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3626 	int isdotdot = 0, unlocknewvp = 0;
3627 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3628 	u_int64_t fileno = 0;
3629 	char *cp;
3630 	nfsattrbit_t attrbits, dattrbits;
3631 	size_t tresid;
3632 	u_int32_t *tl2 = NULL, rderr;
3633 	struct timespec dctime;
3634 
3635 	KASSERT(uiop->uio_iovcnt == 1 &&
3636 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3637 	    ("nfs readdirplusrpc bad uio"));
3638 	ncookie.lval[0] = ncookie.lval[1] = 0;
3639 	timespecclear(&dctime);
3640 	*attrflagp = 0;
3641 	if (eofp != NULL)
3642 		*eofp = 0;
3643 	ndp->ni_dvp = vp;
3644 	nd->nd_mrep = NULL;
3645 	cookie.lval[0] = cookiep->nfsuquad[0];
3646 	cookie.lval[1] = cookiep->nfsuquad[1];
3647 	tresid = uiop->uio_resid;
3648 
3649 	/*
3650 	 * For NFSv4, first create the "." and ".." entries.
3651 	 */
3652 	if (NFSHASNFSV4(nmp)) {
3653 		NFSGETATTR_ATTRBIT(&dattrbits);
3654 		NFSZERO_ATTRBIT(&attrbits);
3655 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3656 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3657 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3658 			NFSSETBIT_ATTRBIT(&attrbits,
3659 			    NFSATTRBIT_MOUNTEDONFILEID);
3660 			gotmnton = 1;
3661 		} else {
3662 			/*
3663 			 * Must fake it. Use the fileno, except when the
3664 			 * fsid is != to that of the directory. For that
3665 			 * case, generate a fake fileno that is not the same.
3666 			 */
3667 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3668 			gotmnton = 0;
3669 		}
3670 
3671 		/*
3672 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3673 		 */
3674 		if (uiop->uio_offset == 0) {
3675 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3676 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3677 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3678 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3679 			(void) nfsrv_putattrbit(nd, &attrbits);
3680 			error = nfscl_request(nd, vp, p, cred, stuff);
3681 			if (error)
3682 			    return (error);
3683 			dotfileid = 0;	/* Fake out the compiler. */
3684 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3685 			    error = nfsm_loadattr(nd, &nfsva);
3686 			    if (error != 0)
3687 				goto nfsmout;
3688 			    dctime = nfsva.na_ctime;
3689 			    dotfileid = nfsva.na_fileid;
3690 			}
3691 			if (nd->nd_repstat == 0) {
3692 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3693 			    len = fxdr_unsigned(int, *(tl + 4));
3694 			    if (len > 0 && len <= NFSX_V4FHMAX)
3695 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3696 			    else
3697 				error = EPERM;
3698 			    if (!error) {
3699 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3700 				nfsva.na_mntonfileno = UINT64_MAX;
3701 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3702 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3703 				    NULL, NULL, NULL, p, cred);
3704 				if (error) {
3705 				    dotdotfileid = dotfileid;
3706 				} else if (gotmnton) {
3707 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3708 					dotdotfileid = nfsva.na_mntonfileno;
3709 				    else
3710 					dotdotfileid = nfsva.na_fileid;
3711 				} else if (nfsva.na_filesid[0] ==
3712 				    dnp->n_vattr.na_filesid[0] &&
3713 				    nfsva.na_filesid[1] ==
3714 				    dnp->n_vattr.na_filesid[1]) {
3715 				    dotdotfileid = nfsva.na_fileid;
3716 				} else {
3717 				    do {
3718 					fakefileno--;
3719 				    } while (fakefileno ==
3720 					nfsva.na_fileid);
3721 				    dotdotfileid = fakefileno;
3722 				}
3723 			    }
3724 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3725 			    /*
3726 			     * Lookupp returns NFSERR_NOENT when we are
3727 			     * at the root, so just use the current dir.
3728 			     */
3729 			    nd->nd_repstat = 0;
3730 			    dotdotfileid = dotfileid;
3731 			} else {
3732 			    error = nd->nd_repstat;
3733 			}
3734 			m_freem(nd->nd_mrep);
3735 			if (error)
3736 			    return (error);
3737 			nd->nd_mrep = NULL;
3738 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3739 			dp->d_pad0 = dp->d_pad1 = 0;
3740 			dp->d_off = 0;
3741 			dp->d_type = DT_DIR;
3742 			dp->d_fileno = dotfileid;
3743 			dp->d_namlen = 1;
3744 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3745 			dp->d_name[0] = '.';
3746 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3747 			/*
3748 			 * Just make these offset cookie 0.
3749 			 */
3750 			tl = (u_int32_t *)&dp->d_name[8];
3751 			*tl++ = 0;
3752 			*tl = 0;
3753 			blksiz += dp->d_reclen;
3754 			uiop->uio_resid -= dp->d_reclen;
3755 			uiop->uio_offset += dp->d_reclen;
3756 			uiop->uio_iov->iov_base =
3757 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3758 			uiop->uio_iov->iov_len -= dp->d_reclen;
3759 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3760 			dp->d_pad0 = dp->d_pad1 = 0;
3761 			dp->d_off = 0;
3762 			dp->d_type = DT_DIR;
3763 			dp->d_fileno = dotdotfileid;
3764 			dp->d_namlen = 2;
3765 			*((uint64_t *)dp->d_name) = 0;
3766 			dp->d_name[0] = '.';
3767 			dp->d_name[1] = '.';
3768 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3769 			/*
3770 			 * Just make these offset cookie 0.
3771 			 */
3772 			tl = (u_int32_t *)&dp->d_name[8];
3773 			*tl++ = 0;
3774 			*tl = 0;
3775 			blksiz += dp->d_reclen;
3776 			uiop->uio_resid -= dp->d_reclen;
3777 			uiop->uio_offset += dp->d_reclen;
3778 			uiop->uio_iov->iov_base =
3779 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3780 			uiop->uio_iov->iov_len -= dp->d_reclen;
3781 		}
3782 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3783 		if (gotmnton)
3784 			NFSSETBIT_ATTRBIT(&attrbits,
3785 			    NFSATTRBIT_MOUNTEDONFILEID);
3786 		if (!NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3787 		    NFSATTRBIT_TIMECREATE))
3788 			NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMECREATE);
3789 	}
3790 
3791 	/*
3792 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3793 	 * The stopping criteria is EOF or buffer full.
3794 	 */
3795 	while (more_dirs && bigenough) {
3796 		*attrflagp = 0;
3797 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3798  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3799 		*tl++ = cookie.lval[0];
3800 		*tl++ = cookie.lval[1];
3801 		if (cookie.qval == 0) {
3802 			*tl++ = 0;
3803 			*tl++ = 0;
3804 		} else {
3805 			NFSLOCKNODE(dnp);
3806 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3807 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3808 			NFSUNLOCKNODE(dnp);
3809 		}
3810 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3811 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3812 		if (nd->nd_flag & ND_NFSV4) {
3813 			(void) nfsrv_putattrbit(nd, &attrbits);
3814 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3815 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3816 			(void) nfsrv_putattrbit(nd, &dattrbits);
3817 		}
3818 		error = nfscl_request(nd, vp, p, cred, stuff);
3819 		if (error)
3820 			return (error);
3821 		if (nd->nd_flag & ND_NFSV3)
3822 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3823 		if (nd->nd_repstat || error) {
3824 			if (!error)
3825 				error = nd->nd_repstat;
3826 			goto nfsmout;
3827 		}
3828 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3829 			dctime = nap->na_ctime;
3830 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3831 		NFSLOCKNODE(dnp);
3832 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3833 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3834 		NFSUNLOCKNODE(dnp);
3835 		more_dirs = fxdr_unsigned(int, *tl);
3836 		if (!more_dirs)
3837 			tryformoredirs = 0;
3838 
3839 		/* loop through the dir entries, doctoring them to 4bsd form */
3840 		while (more_dirs && bigenough) {
3841 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3842 			if (nd->nd_flag & ND_NFSV4) {
3843 				ncookie.lval[0] = *tl++;
3844 				ncookie.lval[1] = *tl++;
3845 			} else {
3846 				fileno = fxdr_hyper(tl);
3847 				tl += 2;
3848 			}
3849 			len = fxdr_unsigned(int, *tl);
3850 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3851 				error = EBADRPC;
3852 				goto nfsmout;
3853 			}
3854 			tlen = roundup2(len, 8);
3855 			if (tlen == len)
3856 				tlen += 8;  /* To ensure null termination. */
3857 			left = DIRBLKSIZ - blksiz;
3858 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3859 				NFSBZERO(uiop->uio_iov->iov_base, left);
3860 				dp->d_reclen += left;
3861 				uiop->uio_iov->iov_base =
3862 				    (char *)uiop->uio_iov->iov_base + left;
3863 				uiop->uio_iov->iov_len -= left;
3864 				uiop->uio_resid -= left;
3865 				uiop->uio_offset += left;
3866 				blksiz = 0;
3867 			}
3868 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3869 			    uiop->uio_resid)
3870 				bigenough = 0;
3871 			if (bigenough) {
3872 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3873 				dp->d_pad0 = dp->d_pad1 = 0;
3874 				dp->d_off = 0;
3875 				dp->d_namlen = len;
3876 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3877 				    NFSX_HYPER;
3878 				dp->d_type = DT_UNKNOWN;
3879 				blksiz += dp->d_reclen;
3880 				if (blksiz == DIRBLKSIZ)
3881 					blksiz = 0;
3882 				uiop->uio_resid -= DIRHDSIZ;
3883 				uiop->uio_offset += DIRHDSIZ;
3884 				uiop->uio_iov->iov_base =
3885 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3886 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3887 				cnp->cn_nameptr = uiop->uio_iov->iov_base;
3888 				cnp->cn_namelen = len;
3889 				NFSCNHASHZERO(cnp);
3890 				error = nfsm_mbufuio(nd, uiop, len);
3891 				if (error)
3892 					goto nfsmout;
3893 				cp = uiop->uio_iov->iov_base;
3894 				tlen -= len;
3895 				NFSBZERO(cp, tlen);
3896 				cp += tlen;	/* points to cookie storage */
3897 				tl2 = (u_int32_t *)cp;
3898 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3899 				    cnp->cn_nameptr[1] == '.')
3900 					isdotdot = 1;
3901 				else
3902 					isdotdot = 0;
3903 				uiop->uio_iov->iov_base =
3904 				    (char *)uiop->uio_iov->iov_base + tlen +
3905 				    NFSX_HYPER;
3906 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3907 				uiop->uio_resid -= tlen + NFSX_HYPER;
3908 				uiop->uio_offset += (tlen + NFSX_HYPER);
3909 			} else {
3910 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3911 				if (error)
3912 					goto nfsmout;
3913 			}
3914 			nfhp = NULL;
3915 			if (nd->nd_flag & ND_NFSV3) {
3916 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3917 				ncookie.lval[0] = *tl++;
3918 				ncookie.lval[1] = *tl++;
3919 				attrflag = fxdr_unsigned(int, *tl);
3920 				if (attrflag) {
3921 				  error = nfsm_loadattr(nd, &nfsva);
3922 				  if (error)
3923 					goto nfsmout;
3924 				}
3925 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3926 				if (*tl) {
3927 					error = nfsm_getfh(nd, &nfhp);
3928 					if (error)
3929 					    goto nfsmout;
3930 				}
3931 				if (!attrflag && nfhp != NULL) {
3932 					free(nfhp, M_NFSFH);
3933 					nfhp = NULL;
3934 				}
3935 			} else {
3936 				rderr = 0;
3937 				nfsva.na_mntonfileno = 0xffffffff;
3938 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3939 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3940 				    NULL, NULL, &rderr, p, cred);
3941 				if (error)
3942 					goto nfsmout;
3943 			}
3944 
3945 			if (bigenough) {
3946 			    if (nd->nd_flag & ND_NFSV4) {
3947 				if (rderr) {
3948 				    dp->d_fileno = 0;
3949 				} else if (gotmnton) {
3950 				    if (nfsva.na_mntonfileno != 0xffffffff)
3951 					dp->d_fileno = nfsva.na_mntonfileno;
3952 				    else
3953 					dp->d_fileno = nfsva.na_fileid;
3954 				} else if (nfsva.na_filesid[0] ==
3955 				    dnp->n_vattr.na_filesid[0] &&
3956 				    nfsva.na_filesid[1] ==
3957 				    dnp->n_vattr.na_filesid[1]) {
3958 				    dp->d_fileno = nfsva.na_fileid;
3959 				} else {
3960 				    do {
3961 					fakefileno--;
3962 				    } while (fakefileno ==
3963 					nfsva.na_fileid);
3964 				    dp->d_fileno = fakefileno;
3965 				}
3966 			    } else {
3967 				dp->d_fileno = fileno;
3968 			    }
3969 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3970 				ncookie.lval[0];
3971 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3972 				ncookie.lval[1];
3973 
3974 			    if (nfhp != NULL) {
3975 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3976 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3977 				    VREF(vp);
3978 				    newvp = vp;
3979 				    unlocknewvp = 0;
3980 				    free(nfhp, M_NFSFH);
3981 				    np = dnp;
3982 				} else if (isdotdot != 0) {
3983 				    /*
3984 				     * Skip doing a nfscl_nget() call for "..".
3985 				     * There's a race between acquiring the nfs
3986 				     * node here and lookups that look for the
3987 				     * directory being read (in the parent).
3988 				     * It would try to get a lock on ".." here,
3989 				     * owning the lock on the directory being
3990 				     * read. Lookup will hold the lock on ".."
3991 				     * and try to acquire the lock on the
3992 				     * directory being read.
3993 				     * If the directory is unlocked/relocked,
3994 				     * then there is a LOR with the buflock
3995 				     * vp is relocked.
3996 				     */
3997 				    free(nfhp, M_NFSFH);
3998 				} else {
3999 				    error = nfscl_nget(vp->v_mount, vp,
4000 				      nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
4001 				    if (!error) {
4002 					newvp = NFSTOV(np);
4003 					unlocknewvp = 1;
4004 				    }
4005 				}
4006 				nfhp = NULL;
4007 				if (newvp != NULLVP) {
4008 				    error = nfscl_loadattrcache(&newvp,
4009 					&nfsva, NULL, NULL, 0, 0);
4010 				    if (error) {
4011 					if (unlocknewvp)
4012 					    vput(newvp);
4013 					else
4014 					    vrele(newvp);
4015 					goto nfsmout;
4016 				    }
4017 				    dp->d_type =
4018 					vtonfs_dtype(np->n_vattr.na_type);
4019 				    ndp->ni_vp = newvp;
4020 				    NFSCNHASH(cnp, HASHINIT);
4021 				    if (cnp->cn_namelen <= NCHNAMLEN &&
4022 					ndp->ni_dvp != ndp->ni_vp &&
4023 					(newvp->v_type != VDIR ||
4024 					 dctime.tv_sec != 0)) {
4025 					cache_enter_time_flags(ndp->ni_dvp,
4026 					    ndp->ni_vp, cnp,
4027 					    &nfsva.na_ctime,
4028 					    newvp->v_type != VDIR ? NULL :
4029 					    &dctime, VFS_CACHE_DROPOLD);
4030 				    }
4031 				    if (unlocknewvp)
4032 					vput(newvp);
4033 				    else
4034 					vrele(newvp);
4035 				    newvp = NULLVP;
4036 				}
4037 			    }
4038 			} else if (nfhp != NULL) {
4039 			    free(nfhp, M_NFSFH);
4040 			}
4041 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
4042 			more_dirs = fxdr_unsigned(int, *tl);
4043 		}
4044 		/*
4045 		 * If at end of rpc data, get the eof boolean
4046 		 */
4047 		if (!more_dirs) {
4048 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
4049 			eof = fxdr_unsigned(int, *tl);
4050 			if (tryformoredirs)
4051 				more_dirs = !eof;
4052 			if (nd->nd_flag & ND_NFSV4) {
4053 				error = nfscl_postop_attr(nd, nap, attrflagp,
4054 				    stuff);
4055 				if (error)
4056 					goto nfsmout;
4057 			}
4058 		}
4059 		m_freem(nd->nd_mrep);
4060 		nd->nd_mrep = NULL;
4061 	}
4062 	/*
4063 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
4064 	 * by increasing d_reclen for the last record.
4065 	 */
4066 	if (blksiz > 0) {
4067 		left = DIRBLKSIZ - blksiz;
4068 		NFSBZERO(uiop->uio_iov->iov_base, left);
4069 		dp->d_reclen += left;
4070 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
4071 		    left;
4072 		uiop->uio_iov->iov_len -= left;
4073 		uiop->uio_resid -= left;
4074 		uiop->uio_offset += left;
4075 	}
4076 
4077 	/*
4078 	 * If returning no data, assume end of file.
4079 	 * If not bigenough, return not end of file, since you aren't
4080 	 *    returning all the data
4081 	 * Otherwise, return the eof flag from the server.
4082 	 */
4083 	if (eofp != NULL) {
4084 		if (tresid == uiop->uio_resid)
4085 			*eofp = 1;
4086 		else if (!bigenough)
4087 			*eofp = 0;
4088 		else
4089 			*eofp = eof;
4090 	}
4091 
4092 	/*
4093 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
4094 	 */
4095 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
4096 		dp = (struct dirent *)uiop->uio_iov->iov_base;
4097 		NFSBZERO(dp, DIRBLKSIZ);
4098 		dp->d_type = DT_UNKNOWN;
4099 		tl = (u_int32_t *)&dp->d_name[4];
4100 		*tl++ = cookie.lval[0];
4101 		*tl = cookie.lval[1];
4102 		dp->d_reclen = DIRBLKSIZ;
4103 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
4104 		    DIRBLKSIZ;
4105 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
4106 		uiop->uio_resid -= DIRBLKSIZ;
4107 		uiop->uio_offset += DIRBLKSIZ;
4108 	}
4109 
4110 nfsmout:
4111 	if (nd->nd_mrep != NULL)
4112 		m_freem(nd->nd_mrep);
4113 	return (error);
4114 }
4115 #endif	/* !APPLE */
4116 
4117 /*
4118  * Nfs commit rpc
4119  */
4120 int
4121 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
4122     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4123 {
4124 	u_int32_t *tl;
4125 	struct nfsrv_descript nfsd, *nd = &nfsd;
4126 	nfsattrbit_t attrbits;
4127 	int error;
4128 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4129 
4130 	*attrflagp = 0;
4131 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
4132 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
4133 	txdr_hyper(offset, tl);
4134 	tl += 2;
4135 	*tl = txdr_unsigned(cnt);
4136 	if (nd->nd_flag & ND_NFSV4) {
4137 		/*
4138 		 * And do a Getattr op.
4139 		 */
4140 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4141 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
4142 		NFSGETATTR_ATTRBIT(&attrbits);
4143 		(void) nfsrv_putattrbit(nd, &attrbits);
4144 	}
4145 	error = nfscl_request(nd, vp, p, cred, stuff);
4146 	if (error)
4147 		return (error);
4148 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
4149 	if (!error && !nd->nd_repstat) {
4150 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
4151 		NFSLOCKMNT(nmp);
4152 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
4153 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
4154 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
4155 		}
4156 		NFSUNLOCKMNT(nmp);
4157 		if (nd->nd_flag & ND_NFSV4)
4158 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4159 	}
4160 nfsmout:
4161 	if (!error && nd->nd_repstat)
4162 		error = nd->nd_repstat;
4163 	m_freem(nd->nd_mrep);
4164 	return (error);
4165 }
4166 
4167 /*
4168  * NFS byte range lock rpc.
4169  * (Mostly just calls one of the three lower level RPC routines.)
4170  */
4171 int
4172 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
4173     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4174 {
4175 	struct nfscllockowner *lp;
4176 	struct nfsclclient *clp;
4177 	struct nfsfh *nfhp;
4178 	struct nfsrv_descript nfsd, *nd = &nfsd;
4179 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4180 	u_int64_t off, len;
4181 	off_t start, end;
4182 	u_int32_t clidrev = 0;
4183 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
4184 	int callcnt, dorpc;
4185 
4186 	/*
4187 	 * Convert the flock structure into a start and end and do POSIX
4188 	 * bounds checking.
4189 	 */
4190 	switch (fl->l_whence) {
4191 	case SEEK_SET:
4192 	case SEEK_CUR:
4193 		/*
4194 		 * Caller is responsible for adding any necessary offset
4195 		 * when SEEK_CUR is used.
4196 		 */
4197 		start = fl->l_start;
4198 		off = fl->l_start;
4199 		break;
4200 	case SEEK_END:
4201 		start = size + fl->l_start;
4202 		off = size + fl->l_start;
4203 		break;
4204 	default:
4205 		return (EINVAL);
4206 	}
4207 	if (start < 0)
4208 		return (EINVAL);
4209 	if (fl->l_len != 0) {
4210 		end = start + fl->l_len - 1;
4211 		if (end < start)
4212 			return (EINVAL);
4213 	}
4214 
4215 	len = fl->l_len;
4216 	if (len == 0)
4217 		len = NFS64BITSSET;
4218 	retrycnt = 0;
4219 	do {
4220 	    nd->nd_repstat = 0;
4221 	    if (op == F_GETLK) {
4222 		error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
4223 		if (error)
4224 			return (error);
4225 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
4226 		if (!error) {
4227 			clidrev = clp->nfsc_clientidrev;
4228 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
4229 			    p, id, flags);
4230 		} else if (error == -1) {
4231 			error = 0;
4232 		}
4233 		nfscl_clientrelease(clp);
4234 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
4235 		/*
4236 		 * We must loop around for all lockowner cases.
4237 		 */
4238 		callcnt = 0;
4239 		error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
4240 		if (error)
4241 			return (error);
4242 		do {
4243 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
4244 			clp, id, flags, &lp, &dorpc);
4245 		    /*
4246 		     * If it returns a NULL lp, we're done.
4247 		     */
4248 		    if (lp == NULL) {
4249 			if (callcnt == 0)
4250 			    nfscl_clientrelease(clp);
4251 			else
4252 			    nfscl_releasealllocks(clp, vp, p, id, flags);
4253 			return (error);
4254 		    }
4255 		    if (nmp->nm_clp != NULL)
4256 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4257 		    else
4258 			clidrev = 0;
4259 		    /*
4260 		     * If the server doesn't support Posix lock semantics,
4261 		     * only allow locks on the entire file, since it won't
4262 		     * handle overlapping byte ranges.
4263 		     * There might still be a problem when a lock
4264 		     * upgrade/downgrade (read<->write) occurs, since the
4265 		     * server "might" expect an unlock first?
4266 		     */
4267 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
4268 			(off == 0 && len == NFS64BITSSET))) {
4269 			/*
4270 			 * Since the lock records will go away, we must
4271 			 * wait for grace and delay here.
4272 			 */
4273 			do {
4274 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
4275 				NFSV4LOCKT_READ, cred, p, 0);
4276 			    if ((nd->nd_repstat == NFSERR_GRACE ||
4277 				 nd->nd_repstat == NFSERR_DELAY) &&
4278 				error == 0)
4279 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
4280 				    "nfs_advlock");
4281 			} while ((nd->nd_repstat == NFSERR_GRACE ||
4282 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
4283 		    }
4284 		    callcnt++;
4285 		} while (error == 0 && nd->nd_repstat == 0);
4286 		nfscl_releasealllocks(clp, vp, p, id, flags);
4287 	    } else if (op == F_SETLK) {
4288 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
4289 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
4290 		if (error || donelocally) {
4291 			return (error);
4292 		}
4293 		if (nmp->nm_clp != NULL)
4294 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4295 		else
4296 			clidrev = 0;
4297 		nfhp = VTONFS(vp)->n_fhp;
4298 		if (!lp->nfsl_open->nfso_posixlock &&
4299 		    (off != 0 || len != NFS64BITSSET)) {
4300 			error = EINVAL;
4301 		} else {
4302 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
4303 			    nfhp->nfh_len, lp, newone, reclaim, off,
4304 			    len, fl->l_type, cred, p, 0);
4305 		}
4306 		if (!error)
4307 			error = nd->nd_repstat;
4308 		nfscl_lockrelease(lp, error, newone);
4309 	    } else {
4310 		error = EINVAL;
4311 	    }
4312 	    if (!error)
4313 	        error = nd->nd_repstat;
4314 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
4315 		error == NFSERR_STALEDONTRECOVER ||
4316 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4317 		error == NFSERR_BADSESSION) {
4318 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
4319 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
4320 		&& clidrev != 0) {
4321 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
4322 		retrycnt++;
4323 	    }
4324 	} while (error == NFSERR_GRACE ||
4325 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4326 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
4327 	    error == NFSERR_BADSESSION ||
4328 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
4329 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
4330 	if (error && retrycnt >= 4)
4331 		error = EIO;
4332 	return (error);
4333 }
4334 
4335 /*
4336  * The lower level routine for the LockT case.
4337  */
4338 int
4339 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
4340     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
4341     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4342 {
4343 	u_int32_t *tl;
4344 	int error, type, size;
4345 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4346 	struct nfsnode *np;
4347 	struct nfsmount *nmp;
4348 	struct nfsclsession *tsep;
4349 
4350 	nmp = VFSTONFS(vp->v_mount);
4351 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
4352 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4353 	if (fl->l_type == F_RDLCK)
4354 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4355 	else
4356 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4357 	txdr_hyper(off, tl);
4358 	tl += 2;
4359 	txdr_hyper(len, tl);
4360 	tl += 2;
4361 	tsep = nfsmnt_mdssession(nmp);
4362 	*tl++ = tsep->nfsess_clientid.lval[0];
4363 	*tl = tsep->nfsess_clientid.lval[1];
4364 	nfscl_filllockowner(id, own, flags);
4365 	np = VTONFS(vp);
4366 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4367 	    np->n_fhp->nfh_len);
4368 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4369 	error = nfscl_request(nd, vp, p, cred, NULL);
4370 	if (error)
4371 		return (error);
4372 	if (nd->nd_repstat == 0) {
4373 		fl->l_type = F_UNLCK;
4374 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4375 		nd->nd_repstat = 0;
4376 		fl->l_whence = SEEK_SET;
4377 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4378 		fl->l_start = fxdr_hyper(tl);
4379 		tl += 2;
4380 		len = fxdr_hyper(tl);
4381 		tl += 2;
4382 		if (len == NFS64BITSSET)
4383 			fl->l_len = 0;
4384 		else
4385 			fl->l_len = len;
4386 		type = fxdr_unsigned(int, *tl++);
4387 		if (type == NFSV4LOCKT_WRITE)
4388 			fl->l_type = F_WRLCK;
4389 		else
4390 			fl->l_type = F_RDLCK;
4391 		/*
4392 		 * XXX For now, I have no idea what to do with the
4393 		 * conflicting lock_owner, so I'll just set the pid == 0
4394 		 * and skip over the lock_owner.
4395 		 */
4396 		fl->l_pid = (pid_t)0;
4397 		tl += 2;
4398 		size = fxdr_unsigned(int, *tl);
4399 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4400 			error = EBADRPC;
4401 		if (!error)
4402 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4403 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4404 		nfscl_initiate_recovery(clp);
4405 nfsmout:
4406 	m_freem(nd->nd_mrep);
4407 	return (error);
4408 }
4409 
4410 /*
4411  * Lower level function that performs the LockU RPC.
4412  */
4413 static int
4414 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4415     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4416     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4417 {
4418 	u_int32_t *tl;
4419 	int error;
4420 
4421 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4422 	    lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0);
4423 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4424 	*tl++ = txdr_unsigned(type);
4425 	*tl = txdr_unsigned(lp->nfsl_seqid);
4426 	if (nfstest_outofseq &&
4427 	    (arc4random() % nfstest_outofseq) == 0)
4428 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
4429 	tl++;
4430 	if (NFSHASNFSV4N(nmp))
4431 		*tl++ = 0;
4432 	else
4433 		*tl++ = lp->nfsl_stateid.seqid;
4434 	*tl++ = lp->nfsl_stateid.other[0];
4435 	*tl++ = lp->nfsl_stateid.other[1];
4436 	*tl++ = lp->nfsl_stateid.other[2];
4437 	txdr_hyper(off, tl);
4438 	tl += 2;
4439 	txdr_hyper(len, tl);
4440 	if (syscred)
4441 		nd->nd_flag |= ND_USEGSSNAME;
4442 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4443 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4444 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4445 	if (error)
4446 		return (error);
4447 	if (nd->nd_repstat == 0) {
4448 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4449 		lp->nfsl_stateid.seqid = *tl++;
4450 		lp->nfsl_stateid.other[0] = *tl++;
4451 		lp->nfsl_stateid.other[1] = *tl++;
4452 		lp->nfsl_stateid.other[2] = *tl;
4453 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4454 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4455 nfsmout:
4456 	m_freem(nd->nd_mrep);
4457 	return (error);
4458 }
4459 
4460 /*
4461  * The actual Lock RPC.
4462  */
4463 int
4464 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4465     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4466     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4467     NFSPROC_T *p, int syscred)
4468 {
4469 	u_int32_t *tl;
4470 	int error, size;
4471 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4472 	struct nfsclsession *tsep;
4473 
4474 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
4475 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4476 	if (type == F_RDLCK)
4477 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4478 	else
4479 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4480 	*tl++ = txdr_unsigned(reclaim);
4481 	txdr_hyper(off, tl);
4482 	tl += 2;
4483 	txdr_hyper(len, tl);
4484 	tl += 2;
4485 	if (newone) {
4486 	    *tl = newnfs_true;
4487 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4488 		2 * NFSX_UNSIGNED + NFSX_HYPER);
4489 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4490 	    if (NFSHASNFSV4N(nmp))
4491 		*tl++ = 0;
4492 	    else
4493 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
4494 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4495 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4496 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4497 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
4498 	    tsep = nfsmnt_mdssession(nmp);
4499 	    *tl++ = tsep->nfsess_clientid.lval[0];
4500 	    *tl = tsep->nfsess_clientid.lval[1];
4501 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4502 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4503 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4504 	} else {
4505 	    *tl = newnfs_false;
4506 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4507 	    if (NFSHASNFSV4N(nmp))
4508 		*tl++ = 0;
4509 	    else
4510 		*tl++ = lp->nfsl_stateid.seqid;
4511 	    *tl++ = lp->nfsl_stateid.other[0];
4512 	    *tl++ = lp->nfsl_stateid.other[1];
4513 	    *tl++ = lp->nfsl_stateid.other[2];
4514 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4515 	    if (nfstest_outofseq &&
4516 		(arc4random() % nfstest_outofseq) == 0)
4517 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4518 	}
4519 	if (syscred)
4520 		nd->nd_flag |= ND_USEGSSNAME;
4521 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4522 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4523 	if (error)
4524 		return (error);
4525 	if (newone)
4526 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4527 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4528 	if (nd->nd_repstat == 0) {
4529 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4530 		lp->nfsl_stateid.seqid = *tl++;
4531 		lp->nfsl_stateid.other[0] = *tl++;
4532 		lp->nfsl_stateid.other[1] = *tl++;
4533 		lp->nfsl_stateid.other[2] = *tl;
4534 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4535 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4536 		size = fxdr_unsigned(int, *(tl + 7));
4537 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4538 			error = EBADRPC;
4539 		if (!error)
4540 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4541 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4542 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4543 nfsmout:
4544 	m_freem(nd->nd_mrep);
4545 	return (error);
4546 }
4547 
4548 /*
4549  * nfs statfs rpc
4550  * (always called with the vp for the mount point)
4551  */
4552 int
4553 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4554     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4555     void *stuff)
4556 {
4557 	u_int32_t *tl = NULL;
4558 	struct nfsrv_descript nfsd, *nd = &nfsd;
4559 	struct nfsmount *nmp;
4560 	nfsattrbit_t attrbits;
4561 	int error;
4562 
4563 	*attrflagp = 0;
4564 	nmp = VFSTONFS(vp->v_mount);
4565 	if (NFSHASNFSV4(nmp)) {
4566 		/*
4567 		 * For V4, you actually do a getattr.
4568 		 */
4569 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4570 		NFSSTATFS_GETATTRBIT(&attrbits);
4571 		(void) nfsrv_putattrbit(nd, &attrbits);
4572 		nd->nd_flag |= ND_USEGSSNAME;
4573 		error = nfscl_request(nd, vp, p, cred, stuff);
4574 		if (error)
4575 			return (error);
4576 		if (nd->nd_repstat == 0) {
4577 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4578 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4579 			    cred);
4580 			if (!error) {
4581 				nmp->nm_fsid[0] = nap->na_filesid[0];
4582 				nmp->nm_fsid[1] = nap->na_filesid[1];
4583 				NFSSETHASSETFSID(nmp);
4584 				*attrflagp = 1;
4585 			}
4586 		} else {
4587 			error = nd->nd_repstat;
4588 		}
4589 		if (error)
4590 			goto nfsmout;
4591 	} else {
4592 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4593 		error = nfscl_request(nd, vp, p, cred, stuff);
4594 		if (error)
4595 			return (error);
4596 		if (nd->nd_flag & ND_NFSV3) {
4597 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4598 			if (error)
4599 				goto nfsmout;
4600 		}
4601 		if (nd->nd_repstat) {
4602 			error = nd->nd_repstat;
4603 			goto nfsmout;
4604 		}
4605 		NFSM_DISSECT(tl, u_int32_t *,
4606 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4607 	}
4608 	if (NFSHASNFSV3(nmp)) {
4609 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4610 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4611 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4612 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4613 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4614 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4615 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4616 	} else if (NFSHASNFSV4(nmp) == 0) {
4617 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4618 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4619 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4620 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4621 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4622 	}
4623 nfsmout:
4624 	m_freem(nd->nd_mrep);
4625 	return (error);
4626 }
4627 
4628 /*
4629  * nfs pathconf rpc
4630  */
4631 int
4632 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4633     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4634     void *stuff)
4635 {
4636 	struct nfsrv_descript nfsd, *nd = &nfsd;
4637 	struct nfsmount *nmp;
4638 	u_int32_t *tl;
4639 	nfsattrbit_t attrbits;
4640 	int error;
4641 
4642 	*attrflagp = 0;
4643 	nmp = VFSTONFS(vp->v_mount);
4644 	if (NFSHASNFSV4(nmp)) {
4645 		/*
4646 		 * For V4, you actually do a getattr.
4647 		 */
4648 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4649 		NFSPATHCONF_GETATTRBIT(&attrbits);
4650 		(void) nfsrv_putattrbit(nd, &attrbits);
4651 		nd->nd_flag |= ND_USEGSSNAME;
4652 		error = nfscl_request(nd, vp, p, cred, stuff);
4653 		if (error)
4654 			return (error);
4655 		if (nd->nd_repstat == 0) {
4656 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4657 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4658 			    cred);
4659 			if (!error)
4660 				*attrflagp = 1;
4661 		} else {
4662 			error = nd->nd_repstat;
4663 		}
4664 	} else {
4665 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4666 		error = nfscl_request(nd, vp, p, cred, stuff);
4667 		if (error)
4668 			return (error);
4669 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4670 		if (nd->nd_repstat && !error)
4671 			error = nd->nd_repstat;
4672 		if (!error) {
4673 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4674 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4675 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4676 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4677 			pc->pc_chownrestricted =
4678 			    fxdr_unsigned(u_int32_t, *tl++);
4679 			pc->pc_caseinsensitive =
4680 			    fxdr_unsigned(u_int32_t, *tl++);
4681 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4682 		}
4683 	}
4684 nfsmout:
4685 	m_freem(nd->nd_mrep);
4686 	return (error);
4687 }
4688 
4689 /*
4690  * nfs version 3 fsinfo rpc call
4691  */
4692 int
4693 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4694     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4695 {
4696 	u_int32_t *tl;
4697 	struct nfsrv_descript nfsd, *nd = &nfsd;
4698 	int error;
4699 
4700 	*attrflagp = 0;
4701 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4702 	error = nfscl_request(nd, vp, p, cred, stuff);
4703 	if (error)
4704 		return (error);
4705 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4706 	if (nd->nd_repstat && !error)
4707 		error = nd->nd_repstat;
4708 	if (!error) {
4709 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4710 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4711 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4712 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4713 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4714 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4715 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4716 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4717 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4718 		tl += 2;
4719 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4720 		tl += 2;
4721 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4722 	}
4723 nfsmout:
4724 	m_freem(nd->nd_mrep);
4725 	return (error);
4726 }
4727 
4728 /*
4729  * This function performs the Renew RPC.
4730  */
4731 int
4732 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4733     NFSPROC_T *p)
4734 {
4735 	u_int32_t *tl;
4736 	struct nfsrv_descript nfsd;
4737 	struct nfsrv_descript *nd = &nfsd;
4738 	struct nfsmount *nmp;
4739 	int error;
4740 	struct nfssockreq *nrp;
4741 	struct nfsclsession *tsep;
4742 
4743 	nmp = clp->nfsc_nmp;
4744 	if (nmp == NULL)
4745 		return (0);
4746 	if (dsp == NULL)
4747 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4748 		    0);
4749 	else
4750 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4751 		    &dsp->nfsclds_sess, 0, 0);
4752 	if (!NFSHASNFSV4N(nmp)) {
4753 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4754 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4755 		tsep = nfsmnt_mdssession(nmp);
4756 		*tl++ = tsep->nfsess_clientid.lval[0];
4757 		*tl = tsep->nfsess_clientid.lval[1];
4758 	}
4759 	nrp = NULL;
4760 	if (dsp != NULL)
4761 		nrp = dsp->nfsclds_sockp;
4762 	if (nrp == NULL)
4763 		/* If NULL, use the MDS socket. */
4764 		nrp = &nmp->nm_sockreq;
4765 	nd->nd_flag |= ND_USEGSSNAME;
4766 	if (dsp == NULL)
4767 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4768 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4769 	else {
4770 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4771 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4772 		if (error == ENXIO)
4773 			nfscl_cancelreqs(dsp);
4774 	}
4775 	if (error)
4776 		return (error);
4777 	error = nd->nd_repstat;
4778 	m_freem(nd->nd_mrep);
4779 	return (error);
4780 }
4781 
4782 /*
4783  * This function performs the Releaselockowner RPC.
4784  */
4785 int
4786 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4787     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4788 {
4789 	struct nfsrv_descript nfsd, *nd = &nfsd;
4790 	u_int32_t *tl;
4791 	int error;
4792 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4793 	struct nfsclsession *tsep;
4794 
4795 	if (NFSHASNFSV4N(nmp)) {
4796 		/* For NFSv4.1, do a FreeStateID. */
4797 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4798 		    NULL, 0, 0);
4799 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4800 	} else {
4801 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4802 		    NULL, 0, 0);
4803 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4804 		tsep = nfsmnt_mdssession(nmp);
4805 		*tl++ = tsep->nfsess_clientid.lval[0];
4806 		*tl = tsep->nfsess_clientid.lval[1];
4807 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4808 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4809 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4810 	}
4811 	nd->nd_flag |= ND_USEGSSNAME;
4812 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4813 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4814 	if (error)
4815 		return (error);
4816 	error = nd->nd_repstat;
4817 	m_freem(nd->nd_mrep);
4818 	return (error);
4819 }
4820 
4821 /*
4822  * This function performs the Compound to get the mount pt FH.
4823  */
4824 int
4825 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4826     NFSPROC_T *p)
4827 {
4828 	u_int32_t *tl;
4829 	struct nfsrv_descript nfsd;
4830 	struct nfsrv_descript *nd = &nfsd;
4831 	u_char *cp, *cp2;
4832 	int error, cnt, len, setnil;
4833 	u_int32_t *opcntp;
4834 
4835 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4836 	    0);
4837 	cp = dirpath;
4838 	cnt = 0;
4839 	do {
4840 		setnil = 0;
4841 		while (*cp == '/')
4842 			cp++;
4843 		cp2 = cp;
4844 		while (*cp2 != '\0' && *cp2 != '/')
4845 			cp2++;
4846 		if (*cp2 == '/') {
4847 			setnil = 1;
4848 			*cp2 = '\0';
4849 		}
4850 		if (cp2 != cp) {
4851 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4852 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4853 			nfsm_strtom(nd, cp, strlen(cp));
4854 			cnt++;
4855 		}
4856 		if (setnil)
4857 			*cp2++ = '/';
4858 		cp = cp2;
4859 	} while (*cp != '\0');
4860 	if (NFSHASNFSV4N(nmp))
4861 		/* Has a Sequence Op done by nfscl_reqstart(). */
4862 		*opcntp = txdr_unsigned(3 + cnt);
4863 	else
4864 		*opcntp = txdr_unsigned(2 + cnt);
4865 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4866 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4867 	nd->nd_flag |= ND_USEGSSNAME;
4868 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4869 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4870 	if (error)
4871 		return (error);
4872 	if (nd->nd_repstat == 0) {
4873 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4874 		tl += (2 + 2 * cnt);
4875 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4876 			len > NFSX_FHMAX) {
4877 			nd->nd_repstat = NFSERR_BADXDR;
4878 		} else {
4879 			nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4880 			if (nd->nd_repstat == 0)
4881 				nmp->nm_fhsize = len;
4882 		}
4883 	}
4884 	error = nd->nd_repstat;
4885 nfsmout:
4886 	m_freem(nd->nd_mrep);
4887 	return (error);
4888 }
4889 
4890 /*
4891  * This function performs the Delegreturn RPC.
4892  */
4893 int
4894 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4895     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4896 {
4897 	u_int32_t *tl;
4898 	struct nfsrv_descript nfsd;
4899 	struct nfsrv_descript *nd = &nfsd;
4900 	int error;
4901 
4902 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4903 	    dp->nfsdl_fhlen, NULL, NULL, 0, 0);
4904 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4905 	if (NFSHASNFSV4N(nmp))
4906 		*tl++ = 0;
4907 	else
4908 		*tl++ = dp->nfsdl_stateid.seqid;
4909 	*tl++ = dp->nfsdl_stateid.other[0];
4910 	*tl++ = dp->nfsdl_stateid.other[1];
4911 	*tl = dp->nfsdl_stateid.other[2];
4912 	if (syscred)
4913 		nd->nd_flag |= ND_USEGSSNAME;
4914 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4915 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4916 	if (error)
4917 		return (error);
4918 	error = nd->nd_repstat;
4919 	m_freem(nd->nd_mrep);
4920 	return (error);
4921 }
4922 
4923 /*
4924  * nfs getacl call.
4925  */
4926 int
4927 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4928     struct acl *aclp, void *stuff)
4929 {
4930 	struct nfsrv_descript nfsd, *nd = &nfsd;
4931 	int error;
4932 	nfsattrbit_t attrbits;
4933 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4934 
4935 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4936 		return (EOPNOTSUPP);
4937 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4938 	NFSZERO_ATTRBIT(&attrbits);
4939 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4940 	(void) nfsrv_putattrbit(nd, &attrbits);
4941 	error = nfscl_request(nd, vp, p, cred, stuff);
4942 	if (error)
4943 		return (error);
4944 	if (!nd->nd_repstat)
4945 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4946 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4947 	else
4948 		error = nd->nd_repstat;
4949 	m_freem(nd->nd_mrep);
4950 	return (error);
4951 }
4952 
4953 /*
4954  * nfs setacl call.
4955  */
4956 int
4957 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4958     struct acl *aclp, void *stuff)
4959 {
4960 	int error;
4961 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4962 
4963 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4964 		return (EOPNOTSUPP);
4965 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4966 	return (error);
4967 }
4968 
4969 /*
4970  * nfs setacl call.
4971  */
4972 static int
4973 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4974     struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4975 {
4976 	struct nfsrv_descript nfsd, *nd = &nfsd;
4977 	int error;
4978 	nfsattrbit_t attrbits;
4979 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4980 
4981 	if (!NFSHASNFSV4(nmp))
4982 		return (EOPNOTSUPP);
4983 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4984 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4985 	NFSZERO_ATTRBIT(&attrbits);
4986 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4987 	(void) nfsv4_fillattr(nd, vp->v_mount, vp, aclp, NULL, NULL, 0,
4988 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
4989 	error = nfscl_request(nd, vp, p, cred, stuff);
4990 	if (error)
4991 		return (error);
4992 	/* Don't care about the pre/postop attributes */
4993 	m_freem(nd->nd_mrep);
4994 	return (nd->nd_repstat);
4995 }
4996 
4997 /*
4998  * Do the NFSv4.1 Exchange ID.
4999  */
5000 int
5001 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
5002     struct nfssockreq *nrp, int minorvers, uint32_t exchflags,
5003     struct nfsclds **dspp, struct ucred *cred, NFSPROC_T *p)
5004 {
5005 	uint32_t *tl, v41flags;
5006 	struct nfsrv_descript nfsd;
5007 	struct nfsrv_descript *nd = &nfsd;
5008 	struct nfsclds *dsp;
5009 	struct timespec verstime;
5010 	int error, len;
5011 
5012 	*dspp = NULL;
5013 	if (minorvers == 0)
5014 		minorvers = nmp->nm_minorvers;
5015 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL,
5016 	    NFS_VER4, minorvers);
5017 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5018 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
5019 	*tl = txdr_unsigned(clp->nfsc_rev);
5020 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
5021 
5022 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
5023 	*tl++ = txdr_unsigned(exchflags);
5024 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
5025 
5026 	/* Set the implementation id4 */
5027 	*tl = txdr_unsigned(1);
5028 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
5029 	(void) nfsm_strtom(nd, version, strlen(version));
5030 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
5031 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
5032 	verstime.tv_nsec = 0;
5033 	txdr_nfsv4time(&verstime, tl);
5034 	nd->nd_flag |= ND_USEGSSNAME;
5035 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
5036 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5037 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
5038 	    (int)nd->nd_repstat);
5039 	if (error != 0)
5040 		return (error);
5041 	if (nd->nd_repstat == 0) {
5042 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
5043 		len = fxdr_unsigned(int, *(tl + 7));
5044 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
5045 			error = NFSERR_BADXDR;
5046 			goto nfsmout;
5047 		}
5048 		dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
5049 		    M_WAITOK | M_ZERO);
5050 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
5051 		dsp->nfsclds_servownlen = len;
5052 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
5053 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
5054 		dsp->nfsclds_sess.nfsess_sequenceid =
5055 		    fxdr_unsigned(uint32_t, *tl++);
5056 		v41flags = fxdr_unsigned(uint32_t, *tl);
5057 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
5058 		    NFSHASPNFSOPT(nmp)) {
5059 			NFSCL_DEBUG(1, "set PNFS\n");
5060 			NFSLOCKMNT(nmp);
5061 			nmp->nm_state |= NFSSTA_PNFS;
5062 			NFSUNLOCKMNT(nmp);
5063 			dsp->nfsclds_flags |= NFSCLDS_MDS;
5064 		}
5065 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
5066 			dsp->nfsclds_flags |= NFSCLDS_DS;
5067 		if (minorvers == NFSV42_MINORVERSION)
5068 			dsp->nfsclds_flags |= NFSCLDS_MINORV2;
5069 		if (len > 0)
5070 			nd->nd_repstat = nfsrv_mtostr(nd,
5071 			    dsp->nfsclds_serverown, len);
5072 		if (nd->nd_repstat == 0) {
5073 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5074 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5075 			    NULL, MTX_DEF);
5076 			nfscl_initsessionslots(&dsp->nfsclds_sess);
5077 			*dspp = dsp;
5078 		} else
5079 			free(dsp, M_NFSCLDS);
5080 	}
5081 	error = nd->nd_repstat;
5082 nfsmout:
5083 	m_freem(nd->nd_mrep);
5084 	return (error);
5085 }
5086 
5087 /*
5088  * Do the NFSv4.1 Create Session.
5089  */
5090 int
5091 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
5092     struct nfssockreq *nrp, struct nfsclds *dsp, uint32_t sequenceid, int mds,
5093     struct ucred *cred, NFSPROC_T *p)
5094 {
5095 	uint32_t crflags, maxval, *tl;
5096 	struct nfsrv_descript nfsd;
5097 	struct nfsrv_descript *nd = &nfsd;
5098 	int error, irdcnt, minorvers;
5099 
5100 	/* Make sure nm_rsize, nm_wsize is set. */
5101 	if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
5102 		nmp->nm_rsize = NFS_MAXBSIZE;
5103 	if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
5104 		nmp->nm_wsize = NFS_MAXBSIZE;
5105 	if (dsp == NULL)
5106 		minorvers = nmp->nm_minorvers;
5107 	else if ((dsp->nfsclds_flags & NFSCLDS_MINORV2) != 0)
5108 		minorvers = NFSV42_MINORVERSION;
5109 	else
5110 		minorvers = NFSV41_MINORVERSION;
5111 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL,
5112 	    NFS_VER4, minorvers);
5113 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5114 	*tl++ = sep->nfsess_clientid.lval[0];
5115 	*tl++ = sep->nfsess_clientid.lval[1];
5116 	*tl++ = txdr_unsigned(sequenceid);
5117 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
5118 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
5119 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
5120 	*tl = txdr_unsigned(crflags);
5121 
5122 	/* Fill in fore channel attributes. */
5123 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5124 	*tl++ = 0;				/* Header pad size */
5125 	if ((nd->nd_flag & ND_NFSV42) != 0 && mds != 0 && sb_max_adj >=
5126 	    nmp->nm_wsize && sb_max_adj >= nmp->nm_rsize) {
5127 		/*
5128 		 * NFSv4.2 Extended Attribute operations may want to do
5129 		 * requests/replies that are larger than nm_rsize/nm_wsize.
5130 		 */
5131 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
5132 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
5133 	} else {
5134 		*tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);
5135 		*tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);
5136 	}
5137 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
5138 	*tl++ = txdr_unsigned(20);		/* Max operations */
5139 	*tl++ = txdr_unsigned(64);		/* Max slots */
5140 	*tl = 0;				/* No rdma ird */
5141 
5142 	/* Fill in back channel attributes. */
5143 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5144 	*tl++ = 0;				/* Header pad size */
5145 	*tl++ = txdr_unsigned(10000);		/* Max request size */
5146 	*tl++ = txdr_unsigned(10000);		/* Max response size */
5147 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
5148 	*tl++ = txdr_unsigned(4);		/* Max operations */
5149 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
5150 	*tl = 0;				/* No rdma ird */
5151 
5152 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
5153 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
5154 
5155 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
5156 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
5157 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
5158 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
5159 	*tl++ = 0;				/* Null machine name */
5160 	*tl++ = 0;				/* Uid == 0 */
5161 	*tl++ = 0;				/* Gid == 0 */
5162 	*tl = 0;				/* No additional gids */
5163 	nd->nd_flag |= ND_USEGSSNAME;
5164 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
5165 	    NFS_VER4, NULL, 1, NULL, NULL);
5166 	if (error != 0)
5167 		return (error);
5168 	if (nd->nd_repstat == 0) {
5169 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
5170 		    2 * NFSX_UNSIGNED);
5171 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
5172 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
5173 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
5174 		crflags = fxdr_unsigned(uint32_t, *tl);
5175 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
5176 			NFSLOCKMNT(nmp);
5177 			nmp->nm_state |= NFSSTA_SESSPERSIST;
5178 			NFSUNLOCKMNT(nmp);
5179 		}
5180 
5181 		/* Get the fore channel slot count. */
5182 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5183 		tl++;			/* Skip the header pad size. */
5184 
5185 		/* Make sure nm_wsize is small enough. */
5186 		maxval = fxdr_unsigned(uint32_t, *tl++);
5187 		while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
5188 			if (nmp->nm_wsize > 8096)
5189 				nmp->nm_wsize /= 2;
5190 			else
5191 				break;
5192 		}
5193 		sep->nfsess_maxreq = maxval;
5194 
5195 		/* Make sure nm_rsize is small enough. */
5196 		maxval = fxdr_unsigned(uint32_t, *tl++);
5197 		while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
5198 			if (nmp->nm_rsize > 8096)
5199 				nmp->nm_rsize /= 2;
5200 			else
5201 				break;
5202 		}
5203 		sep->nfsess_maxresp = maxval;
5204 
5205 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
5206 		tl++;
5207 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
5208 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
5209 		irdcnt = fxdr_unsigned(int, *tl);
5210 		if (irdcnt > 0)
5211 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
5212 
5213 		/* and the back channel slot count. */
5214 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5215 		tl += 5;
5216 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
5217 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
5218 	}
5219 	error = nd->nd_repstat;
5220 nfsmout:
5221 	m_freem(nd->nd_mrep);
5222 	return (error);
5223 }
5224 
5225 /*
5226  * Do the NFSv4.1 Destroy Session.
5227  */
5228 int
5229 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
5230     struct ucred *cred, NFSPROC_T *p)
5231 {
5232 	uint32_t *tl;
5233 	struct nfsrv_descript nfsd;
5234 	struct nfsrv_descript *nd = &nfsd;
5235 	int error;
5236 	struct nfsclsession *tsep;
5237 
5238 	nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL, 0,
5239 	    0);
5240 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
5241 	tsep = nfsmnt_mdssession(nmp);
5242 	bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
5243 	nd->nd_flag |= ND_USEGSSNAME;
5244 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5245 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5246 	if (error != 0)
5247 		return (error);
5248 	error = nd->nd_repstat;
5249 	m_freem(nd->nd_mrep);
5250 	return (error);
5251 }
5252 
5253 /*
5254  * Do the NFSv4.1 Destroy Client.
5255  */
5256 int
5257 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
5258     struct ucred *cred, NFSPROC_T *p)
5259 {
5260 	uint32_t *tl;
5261 	struct nfsrv_descript nfsd;
5262 	struct nfsrv_descript *nd = &nfsd;
5263 	int error;
5264 	struct nfsclsession *tsep;
5265 
5266 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
5267 	    0);
5268 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5269 	tsep = nfsmnt_mdssession(nmp);
5270 	*tl++ = tsep->nfsess_clientid.lval[0];
5271 	*tl = tsep->nfsess_clientid.lval[1];
5272 	nd->nd_flag |= ND_USEGSSNAME;
5273 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5274 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5275 	if (error != 0)
5276 		return (error);
5277 	error = nd->nd_repstat;
5278 	m_freem(nd->nd_mrep);
5279 	return (error);
5280 }
5281 
5282 /*
5283  * Do the NFSv4.1 LayoutGet.
5284  */
5285 static int
5286 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
5287     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
5288     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
5289     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p,
5290     void *stuff)
5291 {
5292 	struct nfsrv_descript nfsd, *nd = &nfsd;
5293 	int error;
5294 
5295 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
5296 	    0);
5297 	nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
5298 	    layouttype, layoutlen, 0);
5299 	nd->nd_flag |= ND_USEGSSNAME;
5300 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5301 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5302 	NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
5303 	if (error != 0)
5304 		return (error);
5305 	if (nd->nd_repstat == 0)
5306 		error = nfsrv_parselayoutget(nmp, nd, stateidp, retonclosep,
5307 		    flhp);
5308 	if (error == 0 && nd->nd_repstat != 0)
5309 		error = nd->nd_repstat;
5310 	m_freem(nd->nd_mrep);
5311 	return (error);
5312 }
5313 
5314 /*
5315  * Do the NFSv4.1 Get Device Info.
5316  */
5317 int
5318 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
5319     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
5320     NFSPROC_T *p)
5321 {
5322 	uint32_t cnt, *tl, vers, minorvers;
5323 	struct nfsrv_descript nfsd;
5324 	struct nfsrv_descript *nd = &nfsd;
5325 	struct sockaddr_in sin, ssin;
5326 	struct sockaddr_in6 sin6, ssin6;
5327 	struct nfsclds *dsp = NULL, **dspp, **gotdspp;
5328 	struct nfscldevinfo *ndi;
5329 	int addrcnt = 0, bitcnt, error, gotminor, gotvers, i, isudp, j;
5330 	int stripecnt;
5331 	uint8_t stripeindex;
5332 	sa_family_t af, safilled;
5333 
5334 	ssin.sin_port = 0;		/* To shut up compiler. */
5335 	ssin.sin_addr.s_addr = 0;	/* ditto */
5336 	*ndip = NULL;
5337 	ndi = NULL;
5338 	gotdspp = NULL;
5339 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
5340 	    0);
5341 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5342 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
5343 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5344 	*tl++ = txdr_unsigned(layouttype);
5345 	*tl++ = txdr_unsigned(100000);
5346 	if (notifybitsp != NULL && *notifybitsp != 0) {
5347 		*tl = txdr_unsigned(1);		/* One word of bits. */
5348 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5349 		*tl = txdr_unsigned(*notifybitsp);
5350 	} else
5351 		*tl = txdr_unsigned(0);
5352 	nd->nd_flag |= ND_USEGSSNAME;
5353 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5354 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5355 	if (error != 0)
5356 		return (error);
5357 	if (nd->nd_repstat == 0) {
5358 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5359 		if (layouttype != fxdr_unsigned(int, *tl))
5360 			printf("EEK! devinfo layout type not same!\n");
5361 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
5362 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5363 			stripecnt = fxdr_unsigned(int, *tl);
5364 			NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
5365 			if (stripecnt < 1 || stripecnt > 4096) {
5366 				printf("pNFS File layout devinfo stripecnt %d:"
5367 				    " out of range\n", stripecnt);
5368 				error = NFSERR_BADXDR;
5369 				goto nfsmout;
5370 			}
5371 			NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
5372 			    NFSX_UNSIGNED);
5373 			addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
5374 			NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
5375 			if (addrcnt < 1 || addrcnt > 128) {
5376 				printf("NFS devinfo addrcnt %d: out of range\n",
5377 				    addrcnt);
5378 				error = NFSERR_BADXDR;
5379 				goto nfsmout;
5380 			}
5381 
5382 			/*
5383 			 * Now we know how many stripe indices and addresses, so
5384 			 * we can allocate the structure the correct size.
5385 			 */
5386 			i = (stripecnt * sizeof(uint8_t)) /
5387 			    sizeof(struct nfsclds *) + 1;
5388 			NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5389 			ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5390 			    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5391 			    M_ZERO);
5392 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5393 			    NFSX_V4DEVICEID);
5394 			ndi->nfsdi_refcnt = 0;
5395 			ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5396 			ndi->nfsdi_stripecnt = stripecnt;
5397 			ndi->nfsdi_addrcnt = addrcnt;
5398 			/* Fill in the stripe indices. */
5399 			for (i = 0; i < stripecnt; i++) {
5400 				stripeindex = fxdr_unsigned(uint8_t, *tl++);
5401 				NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5402 				if (stripeindex >= addrcnt) {
5403 					printf("pNFS File Layout devinfo"
5404 					    " stripeindex %d: too big\n",
5405 					    (int)stripeindex);
5406 					error = NFSERR_BADXDR;
5407 					goto nfsmout;
5408 				}
5409 				nfsfldi_setstripeindex(ndi, i, stripeindex);
5410 			}
5411 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
5412 			/* For Flex File, we only get one address list. */
5413 			ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5414 			    M_NFSDEVINFO, M_WAITOK | M_ZERO);
5415 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5416 			    NFSX_V4DEVICEID);
5417 			ndi->nfsdi_refcnt = 0;
5418 			ndi->nfsdi_flags = NFSDI_FLEXFILE;
5419 			addrcnt = ndi->nfsdi_addrcnt = 1;
5420 		}
5421 
5422 		/* Now, dissect the server address(es). */
5423 		safilled = AF_UNSPEC;
5424 		for (i = 0; i < addrcnt; i++) {
5425 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5426 			cnt = fxdr_unsigned(uint32_t, *tl);
5427 			if (cnt == 0) {
5428 				printf("NFS devinfo 0 len addrlist\n");
5429 				error = NFSERR_BADXDR;
5430 				goto nfsmout;
5431 			}
5432 			dspp = nfsfldi_addr(ndi, i);
5433 			safilled = AF_UNSPEC;
5434 			for (j = 0; j < cnt; j++) {
5435 				error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5436 				    &isudp);
5437 				if (error != 0 && error != EPERM) {
5438 					error = NFSERR_BADXDR;
5439 					goto nfsmout;
5440 				}
5441 				if (error == 0 && isudp == 0) {
5442 					/*
5443 					 * The priority is:
5444 					 * - Same address family.
5445 					 * Save the address and dspp, so that
5446 					 * the connection can be done after
5447 					 * parsing is complete.
5448 					 */
5449 					if (safilled == AF_UNSPEC ||
5450 					    (af == nmp->nm_nam->sa_family &&
5451 					     safilled != nmp->nm_nam->sa_family)
5452 					   ) {
5453 						if (af == AF_INET)
5454 							ssin = sin;
5455 						else
5456 							ssin6 = sin6;
5457 						safilled = af;
5458 						gotdspp = dspp;
5459 					}
5460 				}
5461 			}
5462 		}
5463 
5464 		gotvers = NFS_VER4;	/* Default NFSv4.1 for File Layout. */
5465 		gotminor = NFSV41_MINORVERSION;
5466 		/* For Flex File, we will take one of the versions to use. */
5467 		if (layouttype == NFSLAYOUT_FLEXFILE) {
5468 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5469 			j = fxdr_unsigned(int, *tl);
5470 			if (j < 1 || j > NFSDEV_MAXVERS) {
5471 				printf("pNFS: too many versions\n");
5472 				error = NFSERR_BADXDR;
5473 				goto nfsmout;
5474 			}
5475 			gotvers = 0;
5476 			gotminor = 0;
5477 			for (i = 0; i < j; i++) {
5478 				NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5479 				vers = fxdr_unsigned(uint32_t, *tl++);
5480 				minorvers = fxdr_unsigned(uint32_t, *tl++);
5481 				if (vers == NFS_VER3)
5482 					minorvers = 0;
5483 				if ((vers == NFS_VER4 && ((minorvers ==
5484 				    NFSV41_MINORVERSION && gotminor == 0) ||
5485 				    minorvers == NFSV42_MINORVERSION)) ||
5486 				    (vers == NFS_VER3 && gotvers == 0)) {
5487 					gotvers = vers;
5488 					gotminor = minorvers;
5489 					/* We'll take this one. */
5490 					ndi->nfsdi_versindex = i;
5491 					ndi->nfsdi_vers = vers;
5492 					ndi->nfsdi_minorvers = minorvers;
5493 					ndi->nfsdi_rsize = fxdr_unsigned(
5494 					    uint32_t, *tl++);
5495 					ndi->nfsdi_wsize = fxdr_unsigned(
5496 					    uint32_t, *tl++);
5497 					if (*tl == newnfs_true)
5498 						ndi->nfsdi_flags |=
5499 						    NFSDI_TIGHTCOUPLED;
5500 					else
5501 						ndi->nfsdi_flags &=
5502 						    ~NFSDI_TIGHTCOUPLED;
5503 				}
5504 			}
5505 			if (gotvers == 0) {
5506 				printf("pNFS: no NFSv3, NFSv4.1 or NFSv4.2\n");
5507 				error = NFSERR_BADXDR;
5508 				goto nfsmout;
5509 			}
5510 		}
5511 
5512 		/* And the notify bits. */
5513 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5514 		bitcnt = fxdr_unsigned(int, *tl);
5515 		if (bitcnt > 0) {
5516 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5517 			if (notifybitsp != NULL)
5518 				*notifybitsp =
5519 				    fxdr_unsigned(uint32_t, *tl);
5520 		}
5521 		if (safilled != AF_UNSPEC) {
5522 			KASSERT(ndi != NULL, ("ndi is NULL"));
5523 			*ndip = ndi;
5524 		} else
5525 			error = EPERM;
5526 		if (error == 0) {
5527 			/*
5528 			 * Now we can do a TCP connection for the correct
5529 			 * NFS version and IP address.
5530 			 */
5531 			error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5532 			    gotvers, gotminor, &dsp, p);
5533 		}
5534 		if (error == 0) {
5535 			KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5536 			*gotdspp = dsp;
5537 		}
5538 	}
5539 	if (nd->nd_repstat != 0 && error == 0)
5540 		error = nd->nd_repstat;
5541 nfsmout:
5542 	if (error != 0 && ndi != NULL)
5543 		nfscl_freedevinfo(ndi);
5544 	m_freem(nd->nd_mrep);
5545 	return (error);
5546 }
5547 
5548 /*
5549  * Do the NFSv4.1 LayoutCommit.
5550  */
5551 int
5552 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5553     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5554     int layouttype, struct ucred *cred, NFSPROC_T *p, void *stuff)
5555 {
5556 	uint32_t *tl;
5557 	struct nfsrv_descript nfsd, *nd = &nfsd;
5558 	int error;
5559 
5560 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5561 	    0, 0);
5562 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5563 	    NFSX_STATEID);
5564 	txdr_hyper(off, tl);
5565 	tl += 2;
5566 	txdr_hyper(len, tl);
5567 	tl += 2;
5568 	if (reclaim != 0)
5569 		*tl++ = newnfs_true;
5570 	else
5571 		*tl++ = newnfs_false;
5572 	*tl++ = txdr_unsigned(stateidp->seqid);
5573 	*tl++ = stateidp->other[0];
5574 	*tl++ = stateidp->other[1];
5575 	*tl++ = stateidp->other[2];
5576 	*tl++ = newnfs_true;
5577 	if (lastbyte < off)
5578 		lastbyte = off;
5579 	else if (lastbyte >= (off + len))
5580 		lastbyte = off + len - 1;
5581 	txdr_hyper(lastbyte, tl);
5582 	tl += 2;
5583 	*tl++ = newnfs_false;
5584 	*tl++ = txdr_unsigned(layouttype);
5585 	/* All supported layouts are 0 length. */
5586 	*tl = txdr_unsigned(0);
5587 	nd->nd_flag |= ND_USEGSSNAME;
5588 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5589 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5590 	if (error != 0)
5591 		return (error);
5592 	error = nd->nd_repstat;
5593 	m_freem(nd->nd_mrep);
5594 	return (error);
5595 }
5596 
5597 /*
5598  * Do the NFSv4.1 LayoutReturn.
5599  */
5600 int
5601 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5602     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5603     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5604     uint32_t stat, uint32_t op, char *devid)
5605 {
5606 	uint32_t *tl;
5607 	struct nfsrv_descript nfsd, *nd = &nfsd;
5608 	uint64_t tu64;
5609 	int error;
5610 
5611 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5612 	    0, 0);
5613 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5614 	if (reclaim != 0)
5615 		*tl++ = newnfs_true;
5616 	else
5617 		*tl++ = newnfs_false;
5618 	*tl++ = txdr_unsigned(layouttype);
5619 	*tl++ = txdr_unsigned(iomode);
5620 	*tl = txdr_unsigned(layoutreturn);
5621 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5622 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5623 		    NFSX_UNSIGNED);
5624 		txdr_hyper(offset, tl);
5625 		tl += 2;
5626 		txdr_hyper(len, tl);
5627 		tl += 2;
5628 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5629 		*tl++ = txdr_unsigned(stateidp->seqid);
5630 		*tl++ = stateidp->other[0];
5631 		*tl++ = stateidp->other[1];
5632 		*tl++ = stateidp->other[2];
5633 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5634 			*tl = txdr_unsigned(0);
5635 		else if (layouttype == NFSLAYOUT_FLEXFILE) {
5636 			if (stat != 0) {
5637 				*tl = txdr_unsigned(2 * NFSX_HYPER +
5638 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5639 				    NFSX_UNSIGNED);
5640 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5641 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5642 				    NFSX_UNSIGNED);
5643 				*tl++ = txdr_unsigned(1);	/* One error. */
5644 				tu64 = 0;			/* Offset. */
5645 				txdr_hyper(tu64, tl); tl += 2;
5646 				tu64 = UINT64_MAX;		/* Length. */
5647 				txdr_hyper(tu64, tl); tl += 2;
5648 				NFSBCOPY(stateidp, tl, NFSX_STATEID);
5649 				tl += (NFSX_STATEID / NFSX_UNSIGNED);
5650 				*tl++ = txdr_unsigned(1);	/* One error. */
5651 				NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5652 				tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5653 				*tl++ = txdr_unsigned(stat);
5654 				*tl++ = txdr_unsigned(op);
5655 			} else {
5656 				*tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5657 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5658 				/* No ioerrs. */
5659 				*tl++ = 0;
5660 			}
5661 			*tl = 0;	/* No stats yet. */
5662 		}
5663 	}
5664 	nd->nd_flag |= ND_USEGSSNAME;
5665 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5666 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5667 	if (error != 0)
5668 		return (error);
5669 	if (nd->nd_repstat == 0) {
5670 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5671 		if (*tl != 0) {
5672 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5673 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5674 			stateidp->other[0] = *tl++;
5675 			stateidp->other[1] = *tl++;
5676 			stateidp->other[2] = *tl;
5677 		}
5678 	} else
5679 		error = nd->nd_repstat;
5680 nfsmout:
5681 	m_freem(nd->nd_mrep);
5682 	return (error);
5683 }
5684 
5685 /*
5686  * Acquire a layout and devinfo, if possible. The caller must have acquired
5687  * a reference count on the nfsclclient structure before calling this.
5688  * Return the layout in lypp with a reference count on it, if successful.
5689  */
5690 static int
5691 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5692     int iomode, uint32_t rw, uint32_t *notifybitsp, nfsv4stateid_t *stateidp,
5693     uint64_t off, struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5694 {
5695 	struct nfscllayout *lyp;
5696 	struct nfsclflayout *flp;
5697 	struct nfsclflayouthead flh;
5698 	int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5699 	nfsv4stateid_t stateid;
5700 	struct nfsclsession *tsep;
5701 
5702 	*lypp = NULL;
5703 	if (NFSHASFLEXFILE(nmp))
5704 		layouttype = NFSLAYOUT_FLEXFILE;
5705 	else
5706 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
5707 	/*
5708 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5709 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5710 	 * flp == NULL.
5711 	 */
5712 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5713 	    off, rw, &flp, &recalled);
5714 	islocked = 0;
5715 	if (lyp == NULL || flp == NULL) {
5716 		if (recalled != 0)
5717 			return (EIO);
5718 		LIST_INIT(&flh);
5719 		tsep = nfsmnt_mdssession(nmp);
5720 		layoutlen = tsep->nfsess_maxcache -
5721 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5722 		if (lyp == NULL) {
5723 			stateid.seqid = 0;
5724 			stateid.other[0] = stateidp->other[0];
5725 			stateid.other[1] = stateidp->other[1];
5726 			stateid.other[2] = stateidp->other[2];
5727 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5728 			    nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5729 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5730 			    &retonclose, &flh, cred, p, NULL);
5731 		} else {
5732 			islocked = 1;
5733 			stateid.seqid = lyp->nfsly_stateid.seqid;
5734 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5735 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5736 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5737 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5738 			    nfhp->nfh_len, iomode, off, UINT64_MAX,
5739 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5740 			    &retonclose, &flh, cred, p, NULL);
5741 		}
5742 		error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5743 		    nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5744 		    &flh, layouttype, error, NULL, cred, p);
5745 		if (error == 0)
5746 			*lypp = lyp;
5747 		else if (islocked != 0)
5748 			nfscl_rellayout(lyp, 1);
5749 	} else
5750 		*lypp = lyp;
5751 	return (error);
5752 }
5753 
5754 /*
5755  * Do a TCP connection plus exchange id and create session.
5756  * If successful, a "struct nfsclds" is linked into the list for the
5757  * mount point and a pointer to it is returned.
5758  */
5759 static int
5760 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5761     struct sockaddr_in6 *sin6, sa_family_t af, int vers, int minorvers,
5762     struct nfsclds **dspp, NFSPROC_T *p)
5763 {
5764 	struct sockaddr_in *msad, *sad;
5765 	struct sockaddr_in6 *msad6, *sad6;
5766 	struct nfsclclient *clp;
5767 	struct nfssockreq *nrp;
5768 	struct nfsclds *dsp, *tdsp;
5769 	int error, firsttry;
5770 	enum nfsclds_state retv;
5771 	uint32_t sequenceid = 0;
5772 
5773 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5774 	    ("nfsrpc_fillsa: NULL nr_cred"));
5775 	NFSLOCKCLSTATE();
5776 	clp = nmp->nm_clp;
5777 	NFSUNLOCKCLSTATE();
5778 	if (clp == NULL)
5779 		return (EPERM);
5780 	if (af == AF_INET) {
5781 		NFSLOCKMNT(nmp);
5782 		/*
5783 		 * Check to see if we already have a session for this
5784 		 * address that is usable for a DS.
5785 		 * Note that the MDS's address is in a different place
5786 		 * than the sessions already acquired for DS's.
5787 		 */
5788 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5789 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5790 		while (tdsp != NULL) {
5791 			if (msad != NULL && msad->sin_family == AF_INET &&
5792 			    sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5793 			    sin->sin_port == msad->sin_port &&
5794 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5795 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5796 				*dspp = tdsp;
5797 				NFSUNLOCKMNT(nmp);
5798 				NFSCL_DEBUG(4, "fnd same addr\n");
5799 				return (0);
5800 			}
5801 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5802 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5803 				msad = (struct sockaddr_in *)
5804 				    tdsp->nfsclds_sockp->nr_nam;
5805 			else
5806 				msad = NULL;
5807 		}
5808 		NFSUNLOCKMNT(nmp);
5809 
5810 		/* No IP address match, so look for new/trunked one. */
5811 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5812 		sad->sin_len = sizeof(*sad);
5813 		sad->sin_family = AF_INET;
5814 		sad->sin_port = sin->sin_port;
5815 		sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5816 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5817 		nrp->nr_nam = (struct sockaddr *)sad;
5818 	} else if (af == AF_INET6) {
5819 		NFSLOCKMNT(nmp);
5820 		/*
5821 		 * Check to see if we already have a session for this
5822 		 * address that is usable for a DS.
5823 		 * Note that the MDS's address is in a different place
5824 		 * than the sessions already acquired for DS's.
5825 		 */
5826 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5827 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5828 		while (tdsp != NULL) {
5829 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5830 			    IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5831 			    &msad6->sin6_addr) &&
5832 			    sin6->sin6_port == msad6->sin6_port &&
5833 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5834 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5835 				*dspp = tdsp;
5836 				NFSUNLOCKMNT(nmp);
5837 				return (0);
5838 			}
5839 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5840 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5841 				msad6 = (struct sockaddr_in6 *)
5842 				    tdsp->nfsclds_sockp->nr_nam;
5843 			else
5844 				msad6 = NULL;
5845 		}
5846 		NFSUNLOCKMNT(nmp);
5847 
5848 		/* No IP address match, so look for new/trunked one. */
5849 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5850 		sad6->sin6_len = sizeof(*sad6);
5851 		sad6->sin6_family = AF_INET6;
5852 		sad6->sin6_port = sin6->sin6_port;
5853 		NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5854 		    sizeof(struct in6_addr));
5855 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5856 		nrp->nr_nam = (struct sockaddr *)sad6;
5857 	} else
5858 		return (EPERM);
5859 
5860 	nrp->nr_sotype = SOCK_STREAM;
5861 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5862 	nrp->nr_prog = NFS_PROG;
5863 	nrp->nr_vers = vers;
5864 
5865 	/*
5866 	 * Use the credentials that were used for the mount, which are
5867 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5868 	 * Ref. counting the credentials with crhold() is probably not
5869 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5870 	 * unmount, but I did it anyhow.
5871 	 */
5872 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5873 	error = newnfs_connect(nmp, nrp, NULL, p, 0, false, &nrp->nr_client);
5874 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
5875 
5876 	dsp = NULL;
5877 	/* Now, do the exchangeid and create session. */
5878 	if (error == 0) {
5879 		if (vers == NFS_VER4) {
5880 			firsttry = 0;
5881 			do {
5882 				error = nfsrpc_exchangeid(nmp, clp, nrp,
5883 				    minorvers, NFSV4EXCH_USEPNFSDS, &dsp,
5884 				    nrp->nr_cred, p);
5885 				NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5886 				if (error == NFSERR_MINORVERMISMATCH)
5887 					minorvers = NFSV42_MINORVERSION;
5888 			} while (error == NFSERR_MINORVERMISMATCH &&
5889 			    firsttry++ == 0);
5890 			if (error != 0)
5891 				newnfs_disconnect(NULL, nrp);
5892 		} else {
5893 			dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
5894 			    M_WAITOK | M_ZERO);
5895 			dsp->nfsclds_flags |= NFSCLDS_DS;
5896 			dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
5897 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5898 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5899 			    NULL, MTX_DEF);
5900 		}
5901 	}
5902 	if (error == 0) {
5903 		dsp->nfsclds_sockp = nrp;
5904 		if (vers == NFS_VER4) {
5905 			NFSLOCKMNT(nmp);
5906 			retv = nfscl_getsameserver(nmp, dsp, &tdsp,
5907 			    &sequenceid);
5908 			NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5909 			if (retv == NFSDSP_USETHISSESSION &&
5910 			    nfscl_dssameconn != 0) {
5911 				NFSLOCKDS(tdsp);
5912 				tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
5913 				NFSUNLOCKDS(tdsp);
5914 				NFSUNLOCKMNT(nmp);
5915 				/*
5916 				 * If there is already a session for this
5917 				 * server, use it.
5918 				 */
5919 				newnfs_disconnect(NULL, nrp);
5920 				nfscl_freenfsclds(dsp);
5921 				*dspp = tdsp;
5922 				return (0);
5923 			}
5924 			if (retv == NFSDSP_NOTFOUND)
5925 				sequenceid =
5926 				    dsp->nfsclds_sess.nfsess_sequenceid;
5927 			NFSUNLOCKMNT(nmp);
5928 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5929 			    nrp, dsp, sequenceid, 0, nrp->nr_cred, p);
5930 			NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5931 		}
5932 	} else {
5933 		NFSFREECRED(nrp->nr_cred);
5934 		NFSFREEMUTEX(&nrp->nr_mtx);
5935 		free(nrp->nr_nam, M_SONAME);
5936 		free(nrp, M_NFSSOCKREQ);
5937 	}
5938 	if (error == 0) {
5939 		NFSCL_DEBUG(3, "add DS session\n");
5940 		/*
5941 		 * Put it at the end of the list. That way the list
5942 		 * is ordered by when the entry was added. This matters
5943 		 * since the one done first is the one that should be
5944 		 * used for sequencid'ing any subsequent create sessions.
5945 		 */
5946 		NFSLOCKMNT(nmp);
5947 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5948 		NFSUNLOCKMNT(nmp);
5949 		*dspp = dsp;
5950 	} else if (dsp != NULL) {
5951 		newnfs_disconnect(NULL, nrp);
5952 		nfscl_freenfsclds(dsp);
5953 	}
5954 	return (error);
5955 }
5956 
5957 /*
5958  * Do the NFSv4.1 Reclaim Complete.
5959  */
5960 int
5961 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5962 {
5963 	uint32_t *tl;
5964 	struct nfsrv_descript nfsd;
5965 	struct nfsrv_descript *nd = &nfsd;
5966 	int error;
5967 
5968 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
5969 	    0);
5970 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5971 	*tl = newnfs_false;
5972 	nd->nd_flag |= ND_USEGSSNAME;
5973 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5974 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5975 	if (error != 0)
5976 		return (error);
5977 	error = nd->nd_repstat;
5978 	m_freem(nd->nd_mrep);
5979 	return (error);
5980 }
5981 
5982 /*
5983  * Initialize the slot tables for a session.
5984  */
5985 static void
5986 nfscl_initsessionslots(struct nfsclsession *sep)
5987 {
5988 	int i;
5989 
5990 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
5991 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5992 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5993 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5994 	}
5995 	for (i = 0; i < 64; i++)
5996 		sep->nfsess_slotseq[i] = 0;
5997 	sep->nfsess_slots = 0;
5998 }
5999 
6000 /*
6001  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
6002  */
6003 int
6004 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6005     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
6006 {
6007 	struct nfsnode *np = VTONFS(vp);
6008 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6009 	struct nfscllayout *layp;
6010 	struct nfscldevinfo *dip;
6011 	struct nfsclflayout *rflp;
6012 	struct mbuf *m, *m2;
6013 	struct nfsclwritedsdorpc *drpc, *tdrpc;
6014 	nfsv4stateid_t stateid;
6015 	struct ucred *newcred;
6016 	uint64_t lastbyte, len, off, oresid, xfer;
6017 	int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
6018 	void *lckp;
6019 	uint8_t *dev;
6020 	void *iovbase = NULL;
6021 	size_t iovlen = 0;
6022 	off_t offs = 0;
6023 	ssize_t resid = 0;
6024 
6025 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
6026 	    (np->n_flag & NNOLAYOUT) != 0)
6027 		return (EIO);
6028 	/* Now, get a reference cnt on the clientid for this mount. */
6029 	if (nfscl_getref(nmp) == 0)
6030 		return (EIO);
6031 
6032 	/* Find an appropriate stateid. */
6033 	newcred = NFSNEWCRED(cred);
6034 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
6035 	    rwaccess, 1, newcred, p, &stateid, &lckp);
6036 	if (error != 0) {
6037 		NFSFREECRED(newcred);
6038 		nfscl_relref(nmp);
6039 		return (error);
6040 	}
6041 	/* Search for a layout for this file. */
6042 	off = uiop->uio_offset;
6043 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
6044 	    np->n_fhp->nfh_len, off, rwaccess, &rflp, &recalled);
6045 	if (layp == NULL || rflp == NULL) {
6046 		if (recalled != 0) {
6047 			NFSFREECRED(newcred);
6048 			nfscl_relref(nmp);
6049 			return (EIO);
6050 		}
6051 		if (layp != NULL) {
6052 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
6053 			layp = NULL;
6054 		}
6055 		/* Try and get a Layout, if it is supported. */
6056 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
6057 		    (np->n_flag & NWRITEOPENED) != 0)
6058 			iolaymode = NFSLAYOUTIOMODE_RW;
6059 		else
6060 			iolaymode = NFSLAYOUTIOMODE_READ;
6061 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
6062 		    rwaccess, NULL, &stateid, off, &layp, newcred, p);
6063 		if (error != 0) {
6064 			NFSLOCKNODE(np);
6065 			np->n_flag |= NNOLAYOUT;
6066 			NFSUNLOCKNODE(np);
6067 			if (lckp != NULL)
6068 				nfscl_lockderef(lckp);
6069 			NFSFREECRED(newcred);
6070 			if (layp != NULL)
6071 				nfscl_rellayout(layp, 0);
6072 			nfscl_relref(nmp);
6073 			return (error);
6074 		}
6075 	}
6076 
6077 	/*
6078 	 * Loop around finding a layout that works for the first part of
6079 	 * this I/O operation, and then call the function that actually
6080 	 * does the RPC.
6081 	 */
6082 	eof = 0;
6083 	len = (uint64_t)uiop->uio_resid;
6084 	while (len > 0 && error == 0 && eof == 0) {
6085 		off = uiop->uio_offset;
6086 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
6087 		if (error == 0) {
6088 			oresid = xfer = (uint64_t)uiop->uio_resid;
6089 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
6090 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
6091 			/*
6092 			 * For Flex File layout with mirrored DSs, select one
6093 			 * of them at random for reads. For writes and commits,
6094 			 * do all mirrors.
6095 			 */
6096 			m = NULL;
6097 			tdrpc = drpc = NULL;
6098 			firstmirror = 0;
6099 			mirrorcnt = 1;
6100 			if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
6101 			    (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
6102 				if (rwaccess == NFSV4OPEN_ACCESSREAD) {
6103 					firstmirror = arc4random() % mirrorcnt;
6104 					mirrorcnt = firstmirror + 1;
6105 				} else {
6106 					if (docommit == 0) {
6107 						/*
6108 						 * Save values, so uiop can be
6109 						 * rolled back upon a write
6110 						 * error.
6111 						 */
6112 						offs = uiop->uio_offset;
6113 						resid = uiop->uio_resid;
6114 						iovbase =
6115 						    uiop->uio_iov->iov_base;
6116 						iovlen = uiop->uio_iov->iov_len;
6117 						m = nfsm_uiombuflist(uiop, len,
6118 						    0);
6119 					}
6120 					tdrpc = drpc = malloc(sizeof(*drpc) *
6121 					    (mirrorcnt - 1), M_TEMP, M_WAITOK |
6122 					    M_ZERO);
6123 				}
6124 			}
6125 			for (i = firstmirror; i < mirrorcnt && error == 0; i++){
6126 				m2 = NULL;
6127 				if (m != NULL && i < mirrorcnt - 1)
6128 					m2 = m_copym(m, 0, M_COPYALL, M_WAITOK);
6129 				else {
6130 					m2 = m;
6131 					m = NULL;
6132 				}
6133 				if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
6134 					dev = rflp->nfsfl_ffm[i].dev;
6135 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
6136 					    rflp->nfsfl_ffm[i].devp);
6137 				} else {
6138 					dev = rflp->nfsfl_dev;
6139 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
6140 					    rflp->nfsfl_devp);
6141 				}
6142 				if (dip != NULL) {
6143 					if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
6144 					    != 0)
6145 						error = nfscl_dofflayoutio(vp,
6146 						    uiop, iomode, must_commit,
6147 						    &eof, &stateid, rwaccess,
6148 						    dip, layp, rflp, off, xfer,
6149 						    i, docommit, m2, tdrpc,
6150 						    newcred, p);
6151 					else
6152 						error = nfscl_doflayoutio(vp,
6153 						    uiop, iomode, must_commit,
6154 						    &eof, &stateid, rwaccess,
6155 						    dip, layp, rflp, off, xfer,
6156 						    docommit, newcred, p);
6157 					nfscl_reldevinfo(dip);
6158 				} else {
6159 					if (m2 != NULL)
6160 						m_freem(m2);
6161 					error = EIO;
6162 				}
6163 				tdrpc++;
6164 			}
6165 			if (m != NULL)
6166 				m_freem(m);
6167 			tdrpc = drpc;
6168 			timo = hz / 50;		/* Wait for 20msec. */
6169 			if (timo < 1)
6170 				timo = 1;
6171 			for (i = firstmirror; i < mirrorcnt - 1 &&
6172 			    tdrpc != NULL; i++, tdrpc++) {
6173 				/*
6174 				 * For the unused drpc entries, both inprog and
6175 				 * err == 0, so this loop won't break.
6176 				 */
6177 				while (tdrpc->inprog != 0 && tdrpc->done == 0)
6178 					tsleep(&tdrpc->tsk, PVFS, "clrpcio",
6179 					    timo);
6180 				if (error == 0 && tdrpc->err != 0)
6181 					error = tdrpc->err;
6182 			}
6183 			free(drpc, M_TEMP);
6184 			if (error == 0) {
6185 				if (mirrorcnt > 1 && rwaccess ==
6186 				    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
6187 					NFSLOCKCLSTATE();
6188 					layp->nfsly_flags |= NFSLY_WRITTEN;
6189 					NFSUNLOCKCLSTATE();
6190 				}
6191 				lastbyte = off + xfer - 1;
6192 				NFSLOCKCLSTATE();
6193 				if (lastbyte > layp->nfsly_lastbyte)
6194 					layp->nfsly_lastbyte = lastbyte;
6195 				NFSUNLOCKCLSTATE();
6196 			} else if (error == NFSERR_OPENMODE &&
6197 			    rwaccess == NFSV4OPEN_ACCESSREAD) {
6198 				NFSLOCKMNT(nmp);
6199 				nmp->nm_state |= NFSSTA_OPENMODE;
6200 				NFSUNLOCKMNT(nmp);
6201 			} else
6202 				error = EIO;
6203 			if (error == 0)
6204 				len -= (oresid - (uint64_t)uiop->uio_resid);
6205 			else if (mirrorcnt > 1 && rwaccess ==
6206 			    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
6207 				/*
6208 				 * In case the rpc gets retried, roll the
6209 				 * uio fields changed by nfsm_uiombuflist()
6210 				 * back.
6211 				 */
6212 				uiop->uio_offset = offs;
6213 				uiop->uio_resid = resid;
6214 				uiop->uio_iov->iov_base = iovbase;
6215 				uiop->uio_iov->iov_len = iovlen;
6216 			}
6217 		}
6218 	}
6219 	if (lckp != NULL)
6220 		nfscl_lockderef(lckp);
6221 	NFSFREECRED(newcred);
6222 	nfscl_rellayout(layp, 0);
6223 	nfscl_relref(nmp);
6224 	return (error);
6225 }
6226 
6227 /*
6228  * Find a file layout that will handle the first bytes of the requested
6229  * range and return the information from it needed to the I/O operation.
6230  */
6231 int
6232 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
6233     struct nfsclflayout **retflpp)
6234 {
6235 	struct nfsclflayout *flp, *nflp, *rflp;
6236 	uint32_t rw;
6237 
6238 	rflp = NULL;
6239 	rw = rwaccess;
6240 	/* For reading, do the Read list first and then the Write list. */
6241 	do {
6242 		if (rw == NFSV4OPEN_ACCESSREAD)
6243 			flp = LIST_FIRST(&lyp->nfsly_flayread);
6244 		else
6245 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
6246 		while (flp != NULL) {
6247 			nflp = LIST_NEXT(flp, nfsfl_list);
6248 			if (flp->nfsfl_off > off)
6249 				break;
6250 			if (flp->nfsfl_end > off &&
6251 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
6252 				rflp = flp;
6253 			flp = nflp;
6254 		}
6255 		if (rw == NFSV4OPEN_ACCESSREAD)
6256 			rw = NFSV4OPEN_ACCESSWRITE;
6257 		else
6258 			rw = 0;
6259 	} while (rw != 0);
6260 	if (rflp != NULL) {
6261 		/* This one covers the most bytes starting at off. */
6262 		*retflpp = rflp;
6263 		return (0);
6264 	}
6265 	return (EIO);
6266 }
6267 
6268 /*
6269  * Do I/O using an NFSv4.1 or NFSv4.2 file layout.
6270  */
6271 static int
6272 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6273     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6274     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6275     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
6276 {
6277 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
6278 	int commit_thru_mds, error, stripe_index, stripe_pos, minorvers;
6279 	struct nfsnode *np;
6280 	struct nfsfh *fhp;
6281 	struct nfsclds **dspp;
6282 
6283 	np = VTONFS(vp);
6284 	rel_off = off - flp->nfsfl_patoff;
6285 	stripe_unit_size = flp->nfsfl_util & NFSFLAYUTIL_STRIPE_MASK;
6286 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
6287 	    dp->nfsdi_stripecnt;
6288 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
6289 	error = 0;
6290 
6291 	/* Loop around, doing I/O for each stripe unit. */
6292 	while (len > 0 && error == 0) {
6293 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
6294 		dspp = nfsfldi_addr(dp, stripe_index);
6295 		if (((*dspp)->nfsclds_flags & NFSCLDS_MINORV2) != 0)
6296 			minorvers = NFSV42_MINORVERSION;
6297 		else
6298 			minorvers = NFSV41_MINORVERSION;
6299 		if (len > transfer && docommit == 0)
6300 			xfer = transfer;
6301 		else
6302 			xfer = len;
6303 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
6304 			/* Dense layout. */
6305 			if (stripe_pos >= flp->nfsfl_fhcnt)
6306 				return (EIO);
6307 			fhp = flp->nfsfl_fh[stripe_pos];
6308 			io_off = (rel_off / (stripe_unit_size *
6309 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
6310 			    rel_off % stripe_unit_size;
6311 		} else {
6312 			/* Sparse layout. */
6313 			if (flp->nfsfl_fhcnt > 1) {
6314 				if (stripe_index >= flp->nfsfl_fhcnt)
6315 					return (EIO);
6316 				fhp = flp->nfsfl_fh[stripe_index];
6317 			} else if (flp->nfsfl_fhcnt == 1)
6318 				fhp = flp->nfsfl_fh[0];
6319 			else
6320 				fhp = np->n_fhp;
6321 			io_off = off;
6322 		}
6323 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
6324 			commit_thru_mds = 1;
6325 			if (docommit != 0)
6326 				error = EIO;
6327 		} else {
6328 			commit_thru_mds = 0;
6329 			NFSLOCKNODE(np);
6330 			np->n_flag |= NDSCOMMIT;
6331 			NFSUNLOCKNODE(np);
6332 		}
6333 		if (docommit != 0) {
6334 			if (error == 0)
6335 				error = nfsrpc_commitds(vp, io_off, xfer,
6336 				    *dspp, fhp, NFS_VER4, minorvers, cred, p);
6337 			if (error == 0) {
6338 				/*
6339 				 * Set both eof and uio_resid = 0 to end any
6340 				 * loops.
6341 				 */
6342 				*eofp = 1;
6343 				uiop->uio_resid = 0;
6344 			} else {
6345 				NFSLOCKNODE(np);
6346 				np->n_flag &= ~NDSCOMMIT;
6347 				NFSUNLOCKNODE(np);
6348 			}
6349 		} else if (rwflag == NFSV4OPEN_ACCESSREAD)
6350 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6351 			    io_off, xfer, fhp, 0, NFS_VER4, minorvers, cred, p);
6352 		else {
6353 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
6354 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
6355 			    0, NFS_VER4, minorvers, cred, p);
6356 			if (error == 0) {
6357 				NFSLOCKCLSTATE();
6358 				lyp->nfsly_flags |= NFSLY_WRITTEN;
6359 				NFSUNLOCKCLSTATE();
6360 			}
6361 		}
6362 		if (error == 0) {
6363 			transfer = stripe_unit_size;
6364 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
6365 			len -= xfer;
6366 			off += xfer;
6367 		}
6368 	}
6369 	return (error);
6370 }
6371 
6372 /*
6373  * Do I/O using an NFSv4.1 flex file layout.
6374  */
6375 static int
6376 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6377     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6378     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6379     uint64_t len, int mirror, int docommit, struct mbuf *mp,
6380     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6381 {
6382 	uint64_t xfer;
6383 	int error;
6384 	struct nfsnode *np;
6385 	struct nfsfh *fhp;
6386 	struct nfsclds **dspp;
6387 	struct ucred *tcred;
6388 	struct mbuf *m, *m2;
6389 	uint32_t copylen;
6390 
6391 	np = VTONFS(vp);
6392 	error = 0;
6393 	NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6394 	    (uintmax_t)len);
6395 	/* Loop around, doing I/O for each stripe unit. */
6396 	while (len > 0 && error == 0) {
6397 		dspp = nfsfldi_addr(dp, 0);
6398 		fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6399 		stateidp = &flp->nfsfl_ffm[mirror].st;
6400 		NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6401 		    mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6402 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6403 			tcred = NFSNEWCRED(cred);
6404 			tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6405 			tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6406 			tcred->cr_ngroups = 1;
6407 		} else
6408 			tcred = cred;
6409 		if (rwflag == NFSV4OPEN_ACCESSREAD)
6410 			copylen = dp->nfsdi_rsize;
6411 		else {
6412 			copylen = dp->nfsdi_wsize;
6413 			if (len > copylen && mp != NULL) {
6414 				/*
6415 				 * When a mirrored configuration needs to do
6416 				 * multiple writes to each mirror, all writes
6417 				 * except the last one must be a multiple of
6418 				 * 4 bytes.  This is required so that the XDR
6419 				 * does not need padding.
6420 				 * If possible, clip the size to an exact
6421 				 * multiple of the mbuf length, so that the
6422 				 * split will be on an mbuf boundary.
6423 				 */
6424 				copylen &= 0xfffffffc;
6425 				if (copylen > mp->m_len)
6426 					copylen = copylen / mp->m_len *
6427 					    mp->m_len;
6428 			}
6429 		}
6430 		NFSLOCKNODE(np);
6431 		np->n_flag |= NDSCOMMIT;
6432 		NFSUNLOCKNODE(np);
6433 		if (len > copylen && docommit == 0)
6434 			xfer = copylen;
6435 		else
6436 			xfer = len;
6437 		if (docommit != 0) {
6438 			if (error == 0) {
6439 				/*
6440 				 * Do last mirrored DS commit with this thread.
6441 				 */
6442 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6443 					error = nfsio_commitds(vp, off, xfer,
6444 					    *dspp, fhp, dp->nfsdi_vers,
6445 					    dp->nfsdi_minorvers, drpc, tcred,
6446 					    p);
6447 				else
6448 					error = nfsrpc_commitds(vp, off, xfer,
6449 					    *dspp, fhp, dp->nfsdi_vers,
6450 					    dp->nfsdi_minorvers, tcred, p);
6451 				NFSCL_DEBUG(4, "commitds=%d\n", error);
6452 				if (error != 0 && error != EACCES && error !=
6453 				    ESTALE) {
6454 					NFSCL_DEBUG(4,
6455 					    "DS layreterr for commit\n");
6456 					nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6457 					    lyp, *dspp);
6458 				}
6459 			}
6460 			NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6461 			if (error == 0) {
6462 				/*
6463 				 * Set both eof and uio_resid = 0 to end any
6464 				 * loops.
6465 				 */
6466 				*eofp = 1;
6467 				uiop->uio_resid = 0;
6468 			} else {
6469 				NFSLOCKNODE(np);
6470 				np->n_flag &= ~NDSCOMMIT;
6471 				NFSUNLOCKNODE(np);
6472 			}
6473 		} else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6474 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6475 			    off, xfer, fhp, 1, dp->nfsdi_vers,
6476 			    dp->nfsdi_minorvers, tcred, p);
6477 			NFSCL_DEBUG(4, "readds=%d\n", error);
6478 			if (error != 0 && error != EACCES && error != ESTALE) {
6479 				NFSCL_DEBUG(4, "DS layreterr for read\n");
6480 				nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6481 				    *dspp);
6482 			}
6483 		} else {
6484 			if (flp->nfsfl_mirrorcnt == 1) {
6485 				error = nfsrpc_writeds(vp, uiop, iomode,
6486 				    must_commit, stateidp, *dspp, off, xfer,
6487 				    fhp, 0, 1, dp->nfsdi_vers,
6488 				    dp->nfsdi_minorvers, tcred, p);
6489 				if (error == 0) {
6490 					NFSLOCKCLSTATE();
6491 					lyp->nfsly_flags |= NFSLY_WRITTEN;
6492 					NFSUNLOCKCLSTATE();
6493 				}
6494 			} else {
6495 				m = mp;
6496 				if (xfer < len) {
6497 					/* The mbuf list must be split. */
6498 					m2 = nfsm_split(mp, xfer);
6499 					if (m2 != NULL)
6500 						mp = m2;
6501 					else {
6502 						m_freem(mp);
6503 						error = EIO;
6504 					}
6505 				}
6506 				NFSCL_DEBUG(4, "mcopy len=%jd xfer=%jd\n",
6507 				    (uintmax_t)len, (uintmax_t)xfer);
6508 				/*
6509 				 * Do last write to a mirrored DS with this
6510 				 * thread.
6511 				 */
6512 				if (error == 0) {
6513 					if (mirror < flp->nfsfl_mirrorcnt - 1)
6514 						error = nfsio_writedsmir(vp,
6515 						    iomode, must_commit,
6516 						    stateidp, *dspp, off,
6517 						    xfer, fhp, m,
6518 						    dp->nfsdi_vers,
6519 						    dp->nfsdi_minorvers, drpc,
6520 						    tcred, p);
6521 					else
6522 						error = nfsrpc_writedsmir(vp,
6523 						    iomode, must_commit,
6524 						    stateidp, *dspp, off,
6525 						    xfer, fhp, m,
6526 						    dp->nfsdi_vers,
6527 						    dp->nfsdi_minorvers, tcred,
6528 						    p);
6529 				}
6530 				NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6531 				if (error != 0 && error != EACCES && error !=
6532 				    ESTALE) {
6533 					NFSCL_DEBUG(4,
6534 					    "DS layreterr for write\n");
6535 					nfscl_dserr(NFSV4OP_WRITE, error, dp,
6536 					    lyp, *dspp);
6537 				}
6538 			}
6539 		}
6540 		NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6541 		if (error == 0) {
6542 			len -= xfer;
6543 			off += xfer;
6544 		}
6545 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6546 			NFSFREECRED(tcred);
6547 	}
6548 	NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6549 	return (error);
6550 }
6551 
6552 /*
6553  * The actual read RPC done to a DS.
6554  */
6555 static int
6556 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6557     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6558     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6559 {
6560 	uint32_t *tl;
6561 	int attrflag, error, retlen;
6562 	struct nfsrv_descript nfsd;
6563 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6564 	struct nfsrv_descript *nd = &nfsd;
6565 	struct nfssockreq *nrp;
6566 	struct nfsvattr na;
6567 
6568 	nd->nd_mrep = NULL;
6569 	if (vers == 0 || vers == NFS_VER4) {
6570 		nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6571 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6572 		vers = NFS_VER4;
6573 		NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6574 		if (flex != 0)
6575 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6576 		else
6577 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6578 	} else {
6579 		nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6580 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6581 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READ]);
6582 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READDS]);
6583 		NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6584 	}
6585 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6586 	txdr_hyper(io_off, tl);
6587 	*(tl + 2) = txdr_unsigned(len);
6588 	nrp = dsp->nfsclds_sockp;
6589 	NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6590 	if (nrp == NULL)
6591 		/* If NULL, use the MDS socket. */
6592 		nrp = &nmp->nm_sockreq;
6593 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6594 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6595 	NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6596 	    error);
6597 	if (error != 0)
6598 		return (error);
6599 	if (vers == NFS_VER3) {
6600 		error = nfscl_postop_attr(nd, &na, &attrflag, NULL);
6601 		NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6602 		if (error != 0)
6603 			goto nfsmout;
6604 	}
6605 	if (nd->nd_repstat != 0) {
6606 		error = nd->nd_repstat;
6607 		goto nfsmout;
6608 	}
6609 	if (vers == NFS_VER3) {
6610 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6611 		*eofp = fxdr_unsigned(int, *(tl + 1));
6612 	} else {
6613 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6614 		*eofp = fxdr_unsigned(int, *tl);
6615 	}
6616 	NFSM_STRSIZ(retlen, len);
6617 	NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6618 	error = nfsm_mbufuio(nd, uiop, retlen);
6619 nfsmout:
6620 	if (nd->nd_mrep != NULL)
6621 		m_freem(nd->nd_mrep);
6622 	return (error);
6623 }
6624 
6625 /*
6626  * The actual write RPC done to a DS.
6627  */
6628 static int
6629 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6630     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6631     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6632     struct ucred *cred, NFSPROC_T *p)
6633 {
6634 	uint32_t *tl;
6635 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6636 	int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6637 	int32_t backup;
6638 	struct nfsrv_descript nfsd;
6639 	struct nfsrv_descript *nd = &nfsd;
6640 	struct nfssockreq *nrp;
6641 	struct nfsvattr na;
6642 
6643 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6644 	nd->nd_mrep = NULL;
6645 	if (vers == 0 || vers == NFS_VER4) {
6646 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6647 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6648 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6649 		vers = NFS_VER4;
6650 		if (flex != 0)
6651 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6652 		else
6653 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6654 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6655 	} else {
6656 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6657 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6658 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6659 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6660 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6661 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6662 	}
6663 	txdr_hyper(io_off, tl);
6664 	tl += 2;
6665 	if (vers == NFS_VER3)
6666 		*tl++ = txdr_unsigned(len);
6667 	*tl++ = txdr_unsigned(*iomode);
6668 	*tl = txdr_unsigned(len);
6669 	nfsm_uiombuf(nd, uiop, len);
6670 	nrp = dsp->nfsclds_sockp;
6671 	if (nrp == NULL)
6672 		/* If NULL, use the MDS socket. */
6673 		nrp = &nmp->nm_sockreq;
6674 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6675 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6676 	NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6677 	    nd->nd_repstat);
6678 	if (error != 0)
6679 		return (error);
6680 	if (nd->nd_repstat != 0) {
6681 		/*
6682 		 * In case the rpc gets retried, roll
6683 		 * the uio fields changed by nfsm_uiombuf()
6684 		 * back.
6685 		 */
6686 		uiop->uio_offset -= len;
6687 		uiop->uio_resid += len;
6688 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base - len;
6689 		uiop->uio_iov->iov_len += len;
6690 		error = nd->nd_repstat;
6691 	} else {
6692 		if (vers == NFS_VER3) {
6693 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6694 			    NULL);
6695 			NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6696 			if (error != 0)
6697 				goto nfsmout;
6698 		}
6699 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6700 		rlen = fxdr_unsigned(int, *tl++);
6701 		NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6702 		if (rlen == 0) {
6703 			error = NFSERR_IO;
6704 			goto nfsmout;
6705 		} else if (rlen < len) {
6706 			backup = len - rlen;
6707 			uiop->uio_iov->iov_base =
6708 			    (char *)uiop->uio_iov->iov_base - backup;
6709 			uiop->uio_iov->iov_len += backup;
6710 			uiop->uio_offset -= backup;
6711 			uiop->uio_resid += backup;
6712 			len = rlen;
6713 		}
6714 		commit = fxdr_unsigned(int, *tl++);
6715 
6716 		/*
6717 		 * Return the lowest commitment level
6718 		 * obtained by any of the RPCs.
6719 		 */
6720 		if (committed == NFSWRITE_FILESYNC)
6721 			committed = commit;
6722 		else if (committed == NFSWRITE_DATASYNC &&
6723 		    commit == NFSWRITE_UNSTABLE)
6724 			committed = commit;
6725 		if (commit_thru_mds != 0) {
6726 			NFSLOCKMNT(nmp);
6727 			if (!NFSHASWRITEVERF(nmp)) {
6728 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6729 				NFSSETWRITEVERF(nmp);
6730 	    		} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
6731 				*must_commit = 1;
6732 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6733 			}
6734 			NFSUNLOCKMNT(nmp);
6735 		} else {
6736 			NFSLOCKDS(dsp);
6737 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6738 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6739 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6740 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6741 				*must_commit = 1;
6742 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6743 			}
6744 			NFSUNLOCKDS(dsp);
6745 		}
6746 	}
6747 nfsmout:
6748 	if (nd->nd_mrep != NULL)
6749 		m_freem(nd->nd_mrep);
6750 	*iomode = committed;
6751 	if (nd->nd_repstat != 0 && error == 0)
6752 		error = nd->nd_repstat;
6753 	return (error);
6754 }
6755 
6756 /*
6757  * The actual write RPC done to a DS.
6758  * This variant is called from a separate kernel process for mirrors.
6759  * Any short write is considered an IO error.
6760  */
6761 static int
6762 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6763     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6764     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6765     struct ucred *cred, NFSPROC_T *p)
6766 {
6767 	uint32_t *tl;
6768 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6769 	int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6770 	struct nfsrv_descript nfsd;
6771 	struct nfsrv_descript *nd = &nfsd;
6772 	struct nfssockreq *nrp;
6773 	struct nfsvattr na;
6774 
6775 	nd->nd_mrep = NULL;
6776 	if (vers == 0 || vers == NFS_VER4) {
6777 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6778 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6779 		vers = NFS_VER4;
6780 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6781 		    minorvers);
6782 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6783 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6784 	} else {
6785 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6786 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6787 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6788 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6789 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6790 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6791 	}
6792 	txdr_hyper(io_off, tl);
6793 	tl += 2;
6794 	if (vers == NFS_VER3)
6795 		*tl++ = txdr_unsigned(len);
6796 	*tl++ = txdr_unsigned(*iomode);
6797 	*tl = txdr_unsigned(len);
6798 	if (len > 0) {
6799 		/* Put data in mbuf chain. */
6800 		nd->nd_mb->m_next = m;
6801 	}
6802 	nrp = dsp->nfsclds_sockp;
6803 	if (nrp == NULL)
6804 		/* If NULL, use the MDS socket. */
6805 		nrp = &nmp->nm_sockreq;
6806 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6807 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6808 	NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6809 	    nd->nd_repstat);
6810 	if (error != 0)
6811 		return (error);
6812 	if (nd->nd_repstat != 0)
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_writedsmir: wcc_data=%d\n",
6819 			    error);
6820 			if (error != 0)
6821 				goto nfsmout;
6822 		}
6823 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6824 		rlen = fxdr_unsigned(int, *tl++);
6825 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6826 		    rlen);
6827 		if (rlen != len) {
6828 			error = NFSERR_IO;
6829 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6830 			    len, rlen);
6831 			goto nfsmout;
6832 		}
6833 		commit = fxdr_unsigned(int, *tl++);
6834 
6835 		/*
6836 		 * Return the lowest commitment level
6837 		 * obtained by any of the RPCs.
6838 		 */
6839 		if (committed == NFSWRITE_FILESYNC)
6840 			committed = commit;
6841 		else if (committed == NFSWRITE_DATASYNC &&
6842 		    commit == NFSWRITE_UNSTABLE)
6843 			committed = commit;
6844 		NFSLOCKDS(dsp);
6845 		if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6846 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6847 			dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6848 		} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6849 			*must_commit = 1;
6850 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6851 		}
6852 		NFSUNLOCKDS(dsp);
6853 	}
6854 nfsmout:
6855 	if (nd->nd_mrep != NULL)
6856 		m_freem(nd->nd_mrep);
6857 	*iomode = committed;
6858 	if (nd->nd_repstat != 0 && error == 0)
6859 		error = nd->nd_repstat;
6860 	return (error);
6861 }
6862 
6863 /*
6864  * Start up the thread that will execute nfsrpc_writedsmir().
6865  */
6866 static void
6867 start_writedsmir(void *arg, int pending)
6868 {
6869 	struct nfsclwritedsdorpc *drpc;
6870 
6871 	drpc = (struct nfsclwritedsdorpc *)arg;
6872 	drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
6873 	    &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
6874 	    drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
6875 	    drpc->p);
6876 	drpc->done = 1;
6877 	NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
6878 }
6879 
6880 /*
6881  * Set up the write DS mirror call for the pNFS I/O thread.
6882  */
6883 static int
6884 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6885     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
6886     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6887     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6888 {
6889 	int error, ret;
6890 
6891 	error = 0;
6892 	drpc->done = 0;
6893 	drpc->vp = vp;
6894 	drpc->iomode = *iomode;
6895 	drpc->must_commit = *must_commit;
6896 	drpc->stateidp = stateidp;
6897 	drpc->dsp = dsp;
6898 	drpc->off = off;
6899 	drpc->len = len;
6900 	drpc->fhp = fhp;
6901 	drpc->m = m;
6902 	drpc->vers = vers;
6903 	drpc->minorvers = minorvers;
6904 	drpc->cred = cred;
6905 	drpc->p = p;
6906 	drpc->inprog = 0;
6907 	ret = EIO;
6908 	if (nfs_pnfsiothreads != 0) {
6909 		ret = nfs_pnfsio(start_writedsmir, drpc);
6910 		NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
6911 	}
6912 	if (ret != 0)
6913 		error = nfsrpc_writedsmir(vp, iomode, must_commit, stateidp,
6914 		    dsp, off, len, fhp, m, vers, minorvers, cred, p);
6915 	NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
6916 	return (error);
6917 }
6918 
6919 /*
6920  * Free up the nfsclds structure.
6921  */
6922 void
6923 nfscl_freenfsclds(struct nfsclds *dsp)
6924 {
6925 	int i;
6926 
6927 	if (dsp == NULL)
6928 		return;
6929 	if (dsp->nfsclds_sockp != NULL) {
6930 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
6931 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
6932 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
6933 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
6934 	}
6935 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
6936 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
6937 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
6938 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
6939 			m_freem(
6940 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
6941 	}
6942 	free(dsp, M_NFSCLDS);
6943 }
6944 
6945 static enum nfsclds_state
6946 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
6947     struct nfsclds **retdspp, uint32_t *sequencep)
6948 {
6949 	struct nfsclds *dsp;
6950 	int fndseq;
6951 
6952 	/*
6953 	 * Search the list of nfsclds structures for one with the same
6954 	 * server.
6955 	 */
6956 	fndseq = 0;
6957 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
6958 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
6959 		    dsp->nfsclds_servownlen != 0 &&
6960 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
6961 		    dsp->nfsclds_servownlen) &&
6962 		    dsp->nfsclds_sess.nfsess_defunct == 0) {
6963 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
6964 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
6965 			    dsp->nfsclds_flags);
6966 			if (fndseq == 0) {
6967 				/* Get sequenceid# from first entry. */
6968 				*sequencep =
6969 				    dsp->nfsclds_sess.nfsess_sequenceid;
6970 				fndseq = 1;
6971 			}
6972 			/* Server major id matches. */
6973 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
6974 				*retdspp = dsp;
6975 				return (NFSDSP_USETHISSESSION);
6976 			}
6977 		}
6978 	}
6979 	if (fndseq != 0)
6980 		return (NFSDSP_SEQTHISSESSION);
6981 	return (NFSDSP_NOTFOUND);
6982 }
6983 
6984 /*
6985  * NFS commit rpc to a NFSv4.1 DS.
6986  */
6987 static int
6988 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6989     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
6990     NFSPROC_T *p)
6991 {
6992 	uint32_t *tl;
6993 	struct nfsrv_descript nfsd, *nd = &nfsd;
6994 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6995 	struct nfssockreq *nrp;
6996 	struct nfsvattr na;
6997 	int attrflag, error;
6998 
6999 	nd->nd_mrep = NULL;
7000 	if (vers == 0 || vers == NFS_VER4) {
7001 		nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
7002 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
7003 		vers = NFS_VER4;
7004 	} else {
7005 		nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
7006 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
7007 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMIT]);
7008 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMITDS]);
7009 	}
7010 	NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
7011 	    minorvers);
7012 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
7013 	txdr_hyper(offset, tl);
7014 	tl += 2;
7015 	*tl = txdr_unsigned(cnt);
7016 	nrp = dsp->nfsclds_sockp;
7017 	if (nrp == NULL)
7018 		/* If NULL, use the MDS socket. */
7019 		nrp = &nmp->nm_sockreq;
7020 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7021 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7022 	NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
7023 	    nd->nd_repstat);
7024 	if (error != 0)
7025 		return (error);
7026 	if (nd->nd_repstat == 0) {
7027 		if (vers == NFS_VER3) {
7028 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
7029 			    NULL);
7030 			NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
7031 			if (error != 0)
7032 				goto nfsmout;
7033 		}
7034 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
7035 		NFSLOCKDS(dsp);
7036 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
7037 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
7038 			error = NFSERR_STALEWRITEVERF;
7039 		}
7040 		NFSUNLOCKDS(dsp);
7041 	}
7042 nfsmout:
7043 	if (error == 0 && nd->nd_repstat != 0)
7044 		error = nd->nd_repstat;
7045 	m_freem(nd->nd_mrep);
7046 	return (error);
7047 }
7048 
7049 /*
7050  * Start up the thread that will execute nfsrpc_commitds().
7051  */
7052 static void
7053 start_commitds(void *arg, int pending)
7054 {
7055 	struct nfsclwritedsdorpc *drpc;
7056 
7057 	drpc = (struct nfsclwritedsdorpc *)arg;
7058 	drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
7059 	    drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
7060 	    drpc->p);
7061 	drpc->done = 1;
7062 	NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
7063 }
7064 
7065 /*
7066  * Set up the commit DS mirror call for the pNFS I/O thread.
7067  */
7068 static int
7069 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
7070     struct nfsfh *fhp, int vers, int minorvers,
7071     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7072 {
7073 	int error, ret;
7074 
7075 	error = 0;
7076 	drpc->done = 0;
7077 	drpc->vp = vp;
7078 	drpc->off = offset;
7079 	drpc->len = cnt;
7080 	drpc->dsp = dsp;
7081 	drpc->fhp = fhp;
7082 	drpc->vers = vers;
7083 	drpc->minorvers = minorvers;
7084 	drpc->cred = cred;
7085 	drpc->p = p;
7086 	drpc->inprog = 0;
7087 	ret = EIO;
7088 	if (nfs_pnfsiothreads != 0) {
7089 		ret = nfs_pnfsio(start_commitds, drpc);
7090 		NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
7091 	}
7092 	if (ret != 0)
7093 		error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
7094 		    minorvers, cred, p);
7095 	NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
7096 	return (error);
7097 }
7098 
7099 /*
7100  * NFS Advise rpc
7101  */
7102 int
7103 nfsrpc_advise(vnode_t vp, off_t offset, uint64_t cnt, int advise,
7104     struct ucred *cred, NFSPROC_T *p)
7105 {
7106 	u_int32_t *tl;
7107 	struct nfsrv_descript nfsd, *nd = &nfsd;
7108 	nfsattrbit_t hints;
7109 	int error;
7110 
7111 	NFSZERO_ATTRBIT(&hints);
7112 	if (advise == POSIX_FADV_WILLNEED)
7113 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
7114 	else if (advise == POSIX_FADV_DONTNEED)
7115 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
7116 	else
7117 		return (0);
7118 	NFSCL_REQSTART(nd, NFSPROC_IOADVISE, vp);
7119 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
7120 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
7121 	txdr_hyper(offset, tl);
7122 	tl += 2;
7123 	txdr_hyper(cnt, tl);
7124 	nfsrv_putattrbit(nd, &hints);
7125 	error = nfscl_request(nd, vp, p, cred, NULL);
7126 	if (error != 0)
7127 		return (error);
7128 	if (nd->nd_repstat != 0)
7129 		error = nd->nd_repstat;
7130 	m_freem(nd->nd_mrep);
7131 	return (error);
7132 }
7133 
7134 #ifdef notyet
7135 /*
7136  * NFS advise rpc to a NFSv4.2 DS.
7137  */
7138 static int
7139 nfsrpc_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
7140     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
7141     struct ucred *cred, NFSPROC_T *p)
7142 {
7143 	uint32_t *tl;
7144 	struct nfsrv_descript nfsd, *nd = &nfsd;
7145 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7146 	struct nfssockreq *nrp;
7147 	nfsattrbit_t hints;
7148 	int error;
7149 
7150 	/* For NFS DSs prior to NFSv4.2, just return OK. */
7151 	if (vers == NFS_VER3 || minorversion < NFSV42_MINORVERSION)
7152 		return (0);
7153 	NFSZERO_ATTRBIT(&hints);
7154 	if (advise == POSIX_FADV_WILLNEED)
7155 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
7156 	else if (advise == POSIX_FADV_DONTNEED)
7157 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
7158 	else
7159 		return (0);
7160 	nd->nd_mrep = NULL;
7161 	nfscl_reqstart(nd, NFSPROC_IOADVISEDS, nmp, fhp->nfh_fh,
7162 	    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
7163 	vers = NFS_VER4;
7164 	NFSCL_DEBUG(4, "nfsrpc_adviseds: vers=%d minvers=%d\n", vers,
7165 	    minorvers);
7166 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
7167 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
7168 	txdr_hyper(offset, tl);
7169 	tl += 2;
7170 	*tl = txdr_unsigned(cnt);
7171 	nfsrv_putattrbit(nd, &hints);
7172 	nrp = dsp->nfsclds_sockp;
7173 	if (nrp == NULL)
7174 		/* If NULL, use the MDS socket. */
7175 		nrp = &nmp->nm_sockreq;
7176 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7177 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7178 	NFSCL_DEBUG(4, "nfsrpc_adviseds: err=%d stat=%d\n", error,
7179 	    nd->nd_repstat);
7180 	if (error != 0)
7181 		return (error);
7182 	if (nd->nd_repstat != 0)
7183 		error = nd->nd_repstat;
7184 	m_freem(nd->nd_mrep);
7185 	return (error);
7186 }
7187 
7188 /*
7189  * Start up the thread that will execute nfsrpc_commitds().
7190  */
7191 static void
7192 start_adviseds(void *arg, int pending)
7193 {
7194 	struct nfsclwritedsdorpc *drpc;
7195 
7196 	drpc = (struct nfsclwritedsdorpc *)arg;
7197 	drpc->err = nfsrpc_adviseds(drpc->vp, drpc->off, drpc->len,
7198 	    drpc->advise, drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers,
7199 	    drpc->cred, drpc->p);
7200 	drpc->done = 1;
7201 	NFSCL_DEBUG(4, "start_adviseds: err=%d\n", drpc->err);
7202 }
7203 
7204 /*
7205  * Set up the commit DS mirror call for the pNFS I/O thread.
7206  */
7207 static int
7208 nfsio_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
7209     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
7210     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7211 {
7212 	int error, ret;
7213 
7214 	error = 0;
7215 	drpc->done = 0;
7216 	drpc->vp = vp;
7217 	drpc->off = offset;
7218 	drpc->len = cnt;
7219 	drpc->advise = advise;
7220 	drpc->dsp = dsp;
7221 	drpc->fhp = fhp;
7222 	drpc->vers = vers;
7223 	drpc->minorvers = minorvers;
7224 	drpc->cred = cred;
7225 	drpc->p = p;
7226 	drpc->inprog = 0;
7227 	ret = EIO;
7228 	if (nfs_pnfsiothreads != 0) {
7229 		ret = nfs_pnfsio(start_adviseds, drpc);
7230 		NFSCL_DEBUG(4, "nfsio_adviseds: nfs_pnfsio=%d\n", ret);
7231 	}
7232 	if (ret != 0)
7233 		error = nfsrpc_adviseds(vp, offset, cnt, advise, dsp, fhp, vers,
7234 		    minorvers, cred, p);
7235 	NFSCL_DEBUG(4, "nfsio_adviseds: error=%d\n", error);
7236 	return (error);
7237 }
7238 #endif	/* notyet */
7239 
7240 /*
7241  * Do the Allocate operation, retrying for recovery.
7242  */
7243 int
7244 nfsrpc_allocate(vnode_t vp, off_t off, off_t len, struct nfsvattr *nap,
7245     int *attrflagp, struct ucred *cred, NFSPROC_T *p, void *stuff)
7246 {
7247 	int error, expireret = 0, retrycnt, nostateid;
7248 	uint32_t clidrev = 0;
7249 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7250 	struct nfsfh *nfhp = NULL;
7251 	nfsv4stateid_t stateid;
7252 	off_t tmp_off;
7253 	void *lckp;
7254 
7255 	if (len < 0)
7256 		return (EINVAL);
7257 	if (len == 0)
7258 		return (0);
7259 	tmp_off = off + len;
7260 	NFSLOCKMNT(nmp);
7261 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < off) {
7262 		NFSUNLOCKMNT(nmp);
7263 		return (EFBIG);
7264 	}
7265 	if (nmp->nm_clp != NULL)
7266 		clidrev = nmp->nm_clp->nfsc_clientidrev;
7267 	NFSUNLOCKMNT(nmp);
7268 	nfhp = VTONFS(vp)->n_fhp;
7269 	retrycnt = 0;
7270 	do {
7271 		lckp = NULL;
7272 		nostateid = 0;
7273 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
7274 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
7275 		if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
7276 		    stateid.other[2] == 0) {
7277 			nostateid = 1;
7278 			NFSCL_DEBUG(1, "stateid0 in allocate\n");
7279 		}
7280 
7281 		/*
7282 		 * Not finding a stateid should probably never happen,
7283 		 * but just return an error for this case.
7284 		 */
7285 		if (nostateid != 0)
7286 			error = EIO;
7287 		else
7288 			error = nfsrpc_allocaterpc(vp, off, len, &stateid,
7289 			    nap, attrflagp, cred, p, stuff);
7290 		if (error == NFSERR_STALESTATEID)
7291 			nfscl_initiate_recovery(nmp->nm_clp);
7292 		if (lckp != NULL)
7293 			nfscl_lockderef(lckp);
7294 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
7295 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
7296 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
7297 			(void) nfs_catnap(PZERO, error, "nfs_allocate");
7298 		} else if ((error == NFSERR_EXPIRED ||
7299 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
7300 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
7301 		}
7302 		retrycnt++;
7303 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
7304 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
7305 	    error == NFSERR_STALEDONTRECOVER ||
7306 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
7307 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
7308 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
7309 	if (error != 0 && retrycnt >= 4)
7310 		error = EIO;
7311 	return (error);
7312 }
7313 
7314 /*
7315  * The allocate RPC.
7316  */
7317 static int
7318 nfsrpc_allocaterpc(vnode_t vp, off_t off, off_t len, nfsv4stateid_t *stateidp,
7319     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p,
7320     void *stuff)
7321 {
7322 	uint32_t *tl;
7323 	int error;
7324 	struct nfsrv_descript nfsd;
7325 	struct nfsrv_descript *nd = &nfsd;
7326 	nfsattrbit_t attrbits;
7327 
7328 	*attrflagp = 0;
7329 	NFSCL_REQSTART(nd, NFSPROC_ALLOCATE, vp);
7330 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7331 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
7332 	txdr_hyper(off, tl); tl += 2;
7333 	txdr_hyper(len, tl); tl += 2;
7334 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7335 	NFSGETATTR_ATTRBIT(&attrbits);
7336 	nfsrv_putattrbit(nd, &attrbits);
7337 	error = nfscl_request(nd, vp, p, cred, stuff);
7338 	if (error != 0)
7339 		return (error);
7340 	if (nd->nd_repstat == 0) {
7341 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7342 		error = nfsm_loadattr(nd, nap);
7343 		if (error == 0)
7344 			*attrflagp = NFS_LATTR_NOSHRINK;
7345 	} else
7346 		error = nd->nd_repstat;
7347 nfsmout:
7348 	m_freem(nd->nd_mrep);
7349 	return (error);
7350 }
7351 
7352 /*
7353  * Set up the XDR arguments for the LayoutGet operation.
7354  */
7355 static void
7356 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
7357     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
7358     int layoutlen, int usecurstateid)
7359 {
7360 	uint32_t *tl;
7361 
7362 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
7363 	    NFSX_STATEID);
7364 	*tl++ = newnfs_false;		/* Don't signal availability. */
7365 	*tl++ = txdr_unsigned(layouttype);
7366 	*tl++ = txdr_unsigned(iomode);
7367 	txdr_hyper(offset, tl);
7368 	tl += 2;
7369 	txdr_hyper(len, tl);
7370 	tl += 2;
7371 	txdr_hyper(minlen, tl);
7372 	tl += 2;
7373 	if (usecurstateid != 0) {
7374 		/* Special stateid for Current stateid. */
7375 		*tl++ = txdr_unsigned(1);
7376 		*tl++ = 0;
7377 		*tl++ = 0;
7378 		*tl++ = 0;
7379 	} else {
7380 		*tl++ = txdr_unsigned(stateidp->seqid);
7381 		NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
7382 		*tl++ = stateidp->other[0];
7383 		*tl++ = stateidp->other[1];
7384 		*tl++ = stateidp->other[2];
7385 	}
7386 	*tl = txdr_unsigned(layoutlen);
7387 }
7388 
7389 /*
7390  * Parse the reply for a successful LayoutGet operation.
7391  */
7392 static int
7393 nfsrv_parselayoutget(struct nfsmount *nmp, struct nfsrv_descript *nd,
7394     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp)
7395 {
7396 	uint32_t *tl;
7397 	struct nfsclflayout *flp, *prevflp, *tflp;
7398 	int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
7399 	int m, mirrorcnt;
7400 	uint64_t retlen, off;
7401 	struct nfsfh *nfhp;
7402 	uint8_t *cp;
7403 	uid_t user;
7404 	gid_t grp;
7405 
7406 	NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
7407 	error = 0;
7408 	flp = NULL;
7409 	gotiomode = -1;
7410 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
7411 	if (*tl++ != 0)
7412 		*retonclosep = 1;
7413 	else
7414 		*retonclosep = 0;
7415 	stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
7416 	NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
7417 	    (int)stateidp->seqid);
7418 	stateidp->other[0] = *tl++;
7419 	stateidp->other[1] = *tl++;
7420 	stateidp->other[2] = *tl++;
7421 	cnt = fxdr_unsigned(int, *tl);
7422 	NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
7423 	if (cnt <= 0 || cnt > 10000) {
7424 		/* Don't accept more than 10000 layouts in reply. */
7425 		error = NFSERR_BADXDR;
7426 		goto nfsmout;
7427 	}
7428 	for (i = 0; i < cnt; i++) {
7429 		/* Dissect to the layout type. */
7430 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
7431 		    3 * NFSX_UNSIGNED);
7432 		off = fxdr_hyper(tl); tl += 2;
7433 		retlen = fxdr_hyper(tl); tl += 2;
7434 		iomode = fxdr_unsigned(int, *tl++);
7435 		laytype = fxdr_unsigned(int, *tl);
7436 		NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
7437 		    (uintmax_t)off, (uintmax_t)retlen, iomode);
7438 		/* Ignore length of layout body for now. */
7439 		if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
7440 			/* Parse the File layout up to fhcnt. */
7441 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
7442 			    NFSX_HYPER + NFSX_V4DEVICEID);
7443 			fhcnt = fxdr_unsigned(int, *(tl + 4 +
7444 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
7445 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7446 			if (fhcnt < 0 || fhcnt > 100) {
7447 				/* Don't accept more than 100 file handles. */
7448 				error = NFSERR_BADXDR;
7449 				goto nfsmout;
7450 			}
7451 			if (fhcnt > 0)
7452 				flp = malloc(sizeof(*flp) + fhcnt *
7453 				    sizeof(struct nfsfh *), M_NFSFLAYOUT,
7454 				    M_WAITOK);
7455 			else
7456 				flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
7457 				    M_WAITOK);
7458 			flp->nfsfl_flags = NFSFL_FILE;
7459 			flp->nfsfl_fhcnt = 0;
7460 			flp->nfsfl_devp = NULL;
7461 			flp->nfsfl_off = off;
7462 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7463 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7464 			else
7465 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7466 			flp->nfsfl_iomode = iomode;
7467 			if (gotiomode == -1)
7468 				gotiomode = flp->nfsfl_iomode;
7469 			/* Ignore layout body length for now. */
7470 			NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
7471 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
7472 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
7473 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
7474 			mtx_lock(&nmp->nm_mtx);
7475 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_util &
7476 			    NFSFLAYUTIL_IOADVISE_THRU_MDS) != 0)
7477 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7478 			mtx_unlock(&nmp->nm_mtx);
7479 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
7480 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
7481 			NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
7482 			    flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
7483 			for (j = 0; j < fhcnt; j++) {
7484 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7485 				nfhlen = fxdr_unsigned(int, *tl);
7486 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
7487 					error = NFSERR_BADXDR;
7488 					goto nfsmout;
7489 				}
7490 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
7491 				    M_NFSFH, M_WAITOK);
7492 				flp->nfsfl_fh[j] = nfhp;
7493 				flp->nfsfl_fhcnt++;
7494 				nfhp->nfh_len = nfhlen;
7495 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
7496 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
7497 			}
7498 		} else if (laytype == NFSLAYOUT_FLEXFILE) {
7499 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
7500 			    NFSX_HYPER);
7501 			mirrorcnt = fxdr_unsigned(int, *(tl + 2));
7502 			NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
7503 			if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
7504 				error = NFSERR_BADXDR;
7505 				goto nfsmout;
7506 			}
7507 			flp = malloc(sizeof(*flp) + mirrorcnt *
7508 			    sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
7509 			flp->nfsfl_flags = NFSFL_FLEXFILE;
7510 			flp->nfsfl_mirrorcnt = mirrorcnt;
7511 			for (j = 0; j < mirrorcnt; j++)
7512 				flp->nfsfl_ffm[j].devp = NULL;
7513 			flp->nfsfl_off = off;
7514 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7515 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7516 			else
7517 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7518 			flp->nfsfl_iomode = iomode;
7519 			if (gotiomode == -1)
7520 				gotiomode = flp->nfsfl_iomode;
7521 			flp->nfsfl_stripeunit = fxdr_hyper(tl);
7522 			NFSCL_DEBUG(4, "stripeunit=%ju\n",
7523 			    (uintmax_t)flp->nfsfl_stripeunit);
7524 			for (j = 0; j < mirrorcnt; j++) {
7525 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7526 				k = fxdr_unsigned(int, *tl);
7527 				if (k < 1 || k > 128) {
7528 					error = NFSERR_BADXDR;
7529 					goto nfsmout;
7530 				}
7531 				NFSCL_DEBUG(4, "servercnt=%d\n", k);
7532 				for (l = 0; l < k; l++) {
7533 					NFSM_DISSECT(tl, uint32_t *,
7534 					    NFSX_V4DEVICEID + NFSX_STATEID +
7535 					    2 * NFSX_UNSIGNED);
7536 					if (l == 0) {
7537 						/* Just use the first server. */
7538 						NFSBCOPY(tl,
7539 						    flp->nfsfl_ffm[j].dev,
7540 						    NFSX_V4DEVICEID);
7541 						tl += (NFSX_V4DEVICEID /
7542 						    NFSX_UNSIGNED);
7543 						tl++;
7544 						flp->nfsfl_ffm[j].st.seqid =
7545 						    *tl++;
7546 						flp->nfsfl_ffm[j].st.other[0] =
7547 						    *tl++;
7548 						flp->nfsfl_ffm[j].st.other[1] =
7549 						    *tl++;
7550 						flp->nfsfl_ffm[j].st.other[2] =
7551 						    *tl++;
7552 						NFSCL_DEBUG(4, "st.seqid=%u "
7553 						 "st.o0=0x%x st.o1=0x%x "
7554 						 "st.o2=0x%x\n",
7555 						 flp->nfsfl_ffm[j].st.seqid,
7556 						 flp->nfsfl_ffm[j].st.other[0],
7557 						 flp->nfsfl_ffm[j].st.other[1],
7558 						 flp->nfsfl_ffm[j].st.other[2]);
7559 					} else
7560 						tl += ((NFSX_V4DEVICEID +
7561 						    NFSX_STATEID +
7562 						    NFSX_UNSIGNED) /
7563 						    NFSX_UNSIGNED);
7564 					fhcnt = fxdr_unsigned(int, *tl);
7565 					NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7566 					if (fhcnt < 1 ||
7567 					    fhcnt > NFSDEV_MAXVERS) {
7568 						error = NFSERR_BADXDR;
7569 						goto nfsmout;
7570 					}
7571 					for (m = 0; m < fhcnt; m++) {
7572 						NFSM_DISSECT(tl, uint32_t *,
7573 						    NFSX_UNSIGNED);
7574 						nfhlen = fxdr_unsigned(int,
7575 						    *tl);
7576 						NFSCL_DEBUG(4, "nfhlen=%d\n",
7577 						    nfhlen);
7578 						if (nfhlen <= 0 || nfhlen >
7579 						    NFSX_V4FHMAX) {
7580 							error = NFSERR_BADXDR;
7581 							goto nfsmout;
7582 						}
7583 						NFSM_DISSECT(cp, uint8_t *,
7584 						    NFSM_RNDUP(nfhlen));
7585 						if (l == 0) {
7586 							flp->nfsfl_ffm[j].fhcnt
7587 							    = fhcnt;
7588 							nfhp = malloc(
7589 							    sizeof(*nfhp) +
7590 							    nfhlen - 1, M_NFSFH,
7591 							    M_WAITOK);
7592 							flp->nfsfl_ffm[j].fh[m]
7593 							    = nfhp;
7594 							nfhp->nfh_len = nfhlen;
7595 							NFSBCOPY(cp,
7596 							    nfhp->nfh_fh,
7597 							    nfhlen);
7598 							NFSCL_DEBUG(4,
7599 							    "got fh\n");
7600 						}
7601 					}
7602 					/* Now, get the ffsd_user/ffds_group. */
7603 					error = nfsrv_parseug(nd, 0, &user,
7604 					    &grp, curthread);
7605 					NFSCL_DEBUG(4, "after parseu=%d\n",
7606 					    error);
7607 					if (error == 0)
7608 						error = nfsrv_parseug(nd, 1,
7609 						    &user, &grp, curthread);
7610 					NFSCL_DEBUG(4, "aft parseg=%d\n",
7611 					    grp);
7612 					if (error != 0)
7613 						goto nfsmout;
7614 					NFSCL_DEBUG(4, "user=%d group=%d\n",
7615 					    user, grp);
7616 					if (l == 0) {
7617 						flp->nfsfl_ffm[j].user = user;
7618 						flp->nfsfl_ffm[j].group = grp;
7619 						NFSCL_DEBUG(4,
7620 						    "usr=%d grp=%d\n", user,
7621 						    grp);
7622 					}
7623 				}
7624 			}
7625 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7626 			flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
7627 #ifdef notnow
7628 			/*
7629 			 * At this time, there is no flag.
7630 			 * NFSFLEXFLAG_IOADVISE_THRU_MDS might need to be
7631 			 * added, or it may never exist?
7632 			 */
7633 			mtx_lock(&nmp->nm_mtx);
7634 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_fflags &
7635 			    NFSFLEXFLAG_IOADVISE_THRU_MDS) != 0)
7636 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7637 			mtx_unlock(&nmp->nm_mtx);
7638 #endif
7639 			flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
7640 			NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
7641 			    flp->nfsfl_fflags, flp->nfsfl_statshint);
7642 		} else {
7643 			error = NFSERR_BADXDR;
7644 			goto nfsmout;
7645 		}
7646 		if (flp->nfsfl_iomode == gotiomode) {
7647 			/* Keep the list in increasing offset order. */
7648 			tflp = LIST_FIRST(flhp);
7649 			prevflp = NULL;
7650 			while (tflp != NULL &&
7651 			    tflp->nfsfl_off < flp->nfsfl_off) {
7652 				prevflp = tflp;
7653 				tflp = LIST_NEXT(tflp, nfsfl_list);
7654 			}
7655 			if (prevflp == NULL)
7656 				LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
7657 			else
7658 				LIST_INSERT_AFTER(prevflp, flp,
7659 				    nfsfl_list);
7660 			NFSCL_DEBUG(4, "flp inserted\n");
7661 		} else {
7662 			printf("nfscl_layoutget(): got wrong iomode\n");
7663 			nfscl_freeflayout(flp);
7664 		}
7665 		flp = NULL;
7666 	}
7667 nfsmout:
7668 	NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
7669 	if (error != 0 && flp != NULL)
7670 		nfscl_freeflayout(flp);
7671 	return (error);
7672 }
7673 
7674 /*
7675  * Parse a user/group digit string.
7676  */
7677 static int
7678 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
7679     NFSPROC_T *p)
7680 {
7681 	uint32_t *tl;
7682 	char *cp, *str, str0[NFSV4_SMALLSTR + 1];
7683 	uint32_t len = 0;
7684 	int error = 0;
7685 
7686 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7687 	len = fxdr_unsigned(uint32_t, *tl);
7688 	str = NULL;
7689 	if (len > NFSV4_OPAQUELIMIT) {
7690 		error = NFSERR_BADXDR;
7691 		goto nfsmout;
7692 	}
7693 	NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
7694 	if (len == 0) {
7695 		if (dogrp != 0)
7696 			*gidp = GID_NOGROUP;
7697 		else
7698 			*uidp = UID_NOBODY;
7699 		return (0);
7700 	}
7701 	if (len > NFSV4_SMALLSTR)
7702 		str = malloc(len + 1, M_TEMP, M_WAITOK);
7703 	else
7704 		str = str0;
7705 	NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7706 	NFSBCOPY(cp, str, len);
7707 	str[len] = '\0';
7708 	NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7709 	if (dogrp != 0)
7710 		error = nfsv4_strtogid(nd, str, len, gidp);
7711 	else
7712 		error = nfsv4_strtouid(nd, str, len, uidp);
7713 nfsmout:
7714 	if (len > NFSV4_SMALLSTR)
7715 		free(str, M_TEMP);
7716 	NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7717 	return (error);
7718 }
7719 
7720 /*
7721  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7722  * so that it does both an Open and a Layoutget.
7723  */
7724 static int
7725 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7726     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7727     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7728     struct ucred *cred, NFSPROC_T *p)
7729 {
7730 	struct nfscllayout *lyp;
7731 	struct nfsclflayout *flp;
7732 	struct nfsclflayouthead flh;
7733 	int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7734 	int layouttype, laystat;
7735 	nfsv4stateid_t stateid;
7736 	struct nfsclsession *tsep;
7737 
7738 	error = 0;
7739 	if (NFSHASFLEXFILE(nmp))
7740 		layouttype = NFSLAYOUT_FLEXFILE;
7741 	else
7742 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7743 	/*
7744 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7745 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7746 	 * flp == NULL.
7747 	 */
7748 	lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, mode, &flp,
7749 	    &recalled);
7750 	NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7751 	if (lyp == NULL)
7752 		islocked = 0;
7753 	else if (flp != NULL)
7754 		islocked = 1;
7755 	else
7756 		islocked = 2;
7757 	if ((lyp == NULL || flp == NULL) && recalled == 0) {
7758 		LIST_INIT(&flh);
7759 		tsep = nfsmnt_mdssession(nmp);
7760 		layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7761 		    3 * NFSX_UNSIGNED);
7762 		if (lyp == NULL)
7763 			usecurstateid = 1;
7764 		else {
7765 			usecurstateid = 0;
7766 			stateid.seqid = lyp->nfsly_stateid.seqid;
7767 			stateid.other[0] = lyp->nfsly_stateid.other[0];
7768 			stateid.other[1] = lyp->nfsly_stateid.other[1];
7769 			stateid.other[2] = lyp->nfsly_stateid.other[2];
7770 		}
7771 		error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7772 		    newfhp, newfhlen, mode, op, name, namelen,
7773 		    dpp, &stateid, usecurstateid, layouttype, layoutlen,
7774 		    &retonclose, &flh, &laystat, cred, p);
7775 		NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7776 		    laystat, error);
7777 		laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7778 		    &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7779 		    &islocked, cred, p);
7780 	} else
7781 		error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7782 		    mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7783 	if (islocked == 2)
7784 		nfscl_rellayout(lyp, 1);
7785 	else if (islocked == 1)
7786 		nfscl_rellayout(lyp, 0);
7787 	return (error);
7788 }
7789 
7790 /*
7791  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7792  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
7793  * handled by nfsrpc_openrpc().
7794  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
7795  * can be NULL.
7796  */
7797 static int
7798 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7799     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7800     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7801     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7802     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7803     int *laystatp, struct ucred *cred, NFSPROC_T *p)
7804 {
7805 	uint32_t *tl;
7806 	struct nfsrv_descript nfsd, *nd = &nfsd;
7807 	struct nfscldeleg *ndp = NULL;
7808 	struct nfsvattr nfsva;
7809 	struct nfsclsession *tsep;
7810 	uint32_t rflags, deleg;
7811 	nfsattrbit_t attrbits;
7812 	int error, ret, acesize, limitby, iomode;
7813 
7814 	*dpp = NULL;
7815 	*laystatp = ENXIO;
7816 	nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7817 	    0, 0);
7818 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7819 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7820 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7821 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7822 	tsep = nfsmnt_mdssession(nmp);
7823 	*tl++ = tsep->nfsess_clientid.lval[0];
7824 	*tl = tsep->nfsess_clientid.lval[1];
7825 	nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7826 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7827 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7828 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7829 	nfsm_strtom(nd, name, namelen);
7830 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7831 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7832 	NFSZERO_ATTRBIT(&attrbits);
7833 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7834 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7835 	nfsrv_putattrbit(nd, &attrbits);
7836 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7837 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7838 	if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7839 		iomode = NFSLAYOUTIOMODE_RW;
7840 	else
7841 		iomode = NFSLAYOUTIOMODE_READ;
7842 	nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7843 	    layouttype, layoutlen, usecurstateid);
7844 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7845 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7846 	if (error != 0)
7847 		return (error);
7848 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7849 	if (nd->nd_repstat != 0)
7850 		*laystatp = nd->nd_repstat;
7851 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7852 		/* ND_NOMOREDATA will be set if the Open operation failed. */
7853 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7854 		    6 * NFSX_UNSIGNED);
7855 		op->nfso_stateid.seqid = *tl++;
7856 		op->nfso_stateid.other[0] = *tl++;
7857 		op->nfso_stateid.other[1] = *tl++;
7858 		op->nfso_stateid.other[2] = *tl;
7859 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7860 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7861 		if (error != 0)
7862 			goto nfsmout;
7863 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7864 		deleg = fxdr_unsigned(u_int32_t, *tl);
7865 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
7866 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
7867 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
7868 			      NFSCLFLAGS_FIRSTDELEG))
7869 				op->nfso_own->nfsow_clp->nfsc_flags |=
7870 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7871 			ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
7872 			    M_NFSCLDELEG, M_WAITOK);
7873 			LIST_INIT(&ndp->nfsdl_owner);
7874 			LIST_INIT(&ndp->nfsdl_lock);
7875 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
7876 			ndp->nfsdl_fhlen = newfhlen;
7877 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
7878 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
7879 			nfscl_lockinit(&ndp->nfsdl_rwlock);
7880 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7881 			    NFSX_UNSIGNED);
7882 			ndp->nfsdl_stateid.seqid = *tl++;
7883 			ndp->nfsdl_stateid.other[0] = *tl++;
7884 			ndp->nfsdl_stateid.other[1] = *tl++;
7885 			ndp->nfsdl_stateid.other[2] = *tl++;
7886 			ret = fxdr_unsigned(int, *tl);
7887 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7888 				ndp->nfsdl_flags = NFSCLDL_WRITE;
7889 				/*
7890 				 * Indicates how much the file can grow.
7891 				 */
7892 				NFSM_DISSECT(tl, u_int32_t *,
7893 				    3 * NFSX_UNSIGNED);
7894 				limitby = fxdr_unsigned(int, *tl++);
7895 				switch (limitby) {
7896 				case NFSV4OPEN_LIMITSIZE:
7897 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
7898 					break;
7899 				case NFSV4OPEN_LIMITBLOCKS:
7900 					ndp->nfsdl_sizelimit =
7901 					    fxdr_unsigned(u_int64_t, *tl++);
7902 					ndp->nfsdl_sizelimit *=
7903 					    fxdr_unsigned(u_int64_t, *tl);
7904 					break;
7905 				default:
7906 					error = NFSERR_BADXDR;
7907 					goto nfsmout;
7908 				};
7909 			} else
7910 				ndp->nfsdl_flags = NFSCLDL_READ;
7911 			if (ret != 0)
7912 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
7913 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
7914 			    &acesize, p);
7915 			if (error != 0)
7916 				goto nfsmout;
7917 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
7918 			error = NFSERR_BADXDR;
7919 			goto nfsmout;
7920 		}
7921 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
7922 		    nfscl_assumeposixlocks)
7923 			op->nfso_posixlock = 1;
7924 		else
7925 			op->nfso_posixlock = 0;
7926 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7927 		/* If the 2nd element == NFS_OK, the Getattr succeeded. */
7928 		if (*++tl == 0) {
7929 			error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
7930 			    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
7931 			    NULL, NULL, NULL, p, cred);
7932 			if (error != 0)
7933 				goto nfsmout;
7934 			if (ndp != NULL) {
7935 				ndp->nfsdl_change = nfsva.na_filerev;
7936 				ndp->nfsdl_modtime = nfsva.na_mtime;
7937 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7938 				*dpp = ndp;
7939 				ndp = NULL;
7940 			}
7941 			/*
7942 			 * At this point, the Open has succeeded, so set
7943 			 * nd_repstat = NFS_OK.  If the Layoutget failed,
7944 			 * this function just won't return a layout.
7945 			 */
7946 			if (nd->nd_repstat == 0) {
7947 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7948 				*laystatp = fxdr_unsigned(int, *++tl);
7949 				if (*laystatp == 0) {
7950 					error = nfsrv_parselayoutget(nmp, nd,
7951 					    stateidp, retonclosep, flhp);
7952 					if (error != 0)
7953 						*laystatp = error;
7954 				}
7955 			} else
7956 				nd->nd_repstat = 0;	/* Return 0 for Open. */
7957 		}
7958 	}
7959 	if (nd->nd_repstat != 0 && error == 0)
7960 		error = nd->nd_repstat;
7961 nfsmout:
7962 	free(ndp, M_NFSCLDELEG);
7963 	m_freem(nd->nd_mrep);
7964 	return (error);
7965 }
7966 
7967 /*
7968  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
7969  * Used only for mounts with pNFS enabled.
7970  */
7971 static int
7972 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7973     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7974     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7975     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7976     int *dattrflagp, void *dstuff, int *unlockedp, nfsv4stateid_t *stateidp,
7977     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
7978     struct nfsclflayouthead *flhp, int *laystatp)
7979 {
7980 	uint32_t *tl;
7981 	int error = 0, deleg, newone, ret, acesize, limitby;
7982 	struct nfsrv_descript nfsd, *nd = &nfsd;
7983 	struct nfsclopen *op;
7984 	struct nfscldeleg *dp = NULL;
7985 	struct nfsnode *np;
7986 	struct nfsfh *nfhp;
7987 	struct nfsclsession *tsep;
7988 	nfsattrbit_t attrbits;
7989 	nfsv4stateid_t stateid;
7990 	struct nfsmount *nmp;
7991 
7992 	nmp = VFSTONFS(dvp->v_mount);
7993 	np = VTONFS(dvp);
7994 	*laystatp = ENXIO;
7995 	*unlockedp = 0;
7996 	*nfhpp = NULL;
7997 	*dpp = NULL;
7998 	*attrflagp = 0;
7999 	*dattrflagp = 0;
8000 	if (namelen > NFS_MAXNAMLEN)
8001 		return (ENAMETOOLONG);
8002 	NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
8003 	/*
8004 	 * For V4, this is actually an Open op.
8005 	 */
8006 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
8007 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
8008 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
8009 	    NFSV4OPEN_ACCESSREAD);
8010 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
8011 	tsep = nfsmnt_mdssession(nmp);
8012 	*tl++ = tsep->nfsess_clientid.lval[0];
8013 	*tl = tsep->nfsess_clientid.lval[1];
8014 	nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
8015 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8016 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
8017 	if ((fmode & O_EXCL) != 0) {
8018 		if (NFSHASSESSPERSIST(nmp)) {
8019 			/* Use GUARDED for persistent sessions. */
8020 			*tl = txdr_unsigned(NFSCREATE_GUARDED);
8021 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
8022 		} else {
8023 			/* Otherwise, use EXCLUSIVE4_1. */
8024 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
8025 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
8026 			*tl++ = cverf.lval[0];
8027 			*tl = cverf.lval[1];
8028 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
8029 		}
8030 	} else {
8031 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
8032 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
8033 	}
8034 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8035 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
8036 	nfsm_strtom(nd, name, namelen);
8037 	/* Get the new file's handle and attributes, plus save the FH. */
8038 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
8039 	*tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
8040 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
8041 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8042 	NFSGETATTR_ATTRBIT(&attrbits);
8043 	nfsrv_putattrbit(nd, &attrbits);
8044 	/* Get the directory's post-op attributes. */
8045 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8046 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8047 	nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
8048 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8049 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8050 	nfsrv_putattrbit(nd, &attrbits);
8051 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8052 	*tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
8053 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
8054 	nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
8055 	    layouttype, layoutlen, usecurstateid);
8056 	error = nfscl_request(nd, dvp, p, cred, dstuff);
8057 	if (error != 0)
8058 		return (error);
8059 	NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
8060 	    error);
8061 	if (nd->nd_repstat != 0)
8062 		*laystatp = nd->nd_repstat;
8063 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
8064 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8065 		NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
8066 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8067 		    6 * NFSX_UNSIGNED);
8068 		stateid.seqid = *tl++;
8069 		stateid.other[0] = *tl++;
8070 		stateid.other[1] = *tl++;
8071 		stateid.other[2] = *tl;
8072 		nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
8073 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
8074 		deleg = fxdr_unsigned(int, *tl);
8075 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
8076 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
8077 			if (!(owp->nfsow_clp->nfsc_flags &
8078 			      NFSCLFLAGS_FIRSTDELEG))
8079 				owp->nfsow_clp->nfsc_flags |=
8080 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
8081 			dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
8082 			    M_NFSCLDELEG, M_WAITOK);
8083 			LIST_INIT(&dp->nfsdl_owner);
8084 			LIST_INIT(&dp->nfsdl_lock);
8085 			dp->nfsdl_clp = owp->nfsow_clp;
8086 			newnfs_copyincred(cred, &dp->nfsdl_cred);
8087 			nfscl_lockinit(&dp->nfsdl_rwlock);
8088 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8089 			    NFSX_UNSIGNED);
8090 			dp->nfsdl_stateid.seqid = *tl++;
8091 			dp->nfsdl_stateid.other[0] = *tl++;
8092 			dp->nfsdl_stateid.other[1] = *tl++;
8093 			dp->nfsdl_stateid.other[2] = *tl++;
8094 			ret = fxdr_unsigned(int, *tl);
8095 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
8096 				dp->nfsdl_flags = NFSCLDL_WRITE;
8097 				/*
8098 				 * Indicates how much the file can grow.
8099 				 */
8100 				NFSM_DISSECT(tl, u_int32_t *,
8101 				    3 * NFSX_UNSIGNED);
8102 				limitby = fxdr_unsigned(int, *tl++);
8103 				switch (limitby) {
8104 				case NFSV4OPEN_LIMITSIZE:
8105 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
8106 					break;
8107 				case NFSV4OPEN_LIMITBLOCKS:
8108 					dp->nfsdl_sizelimit =
8109 					    fxdr_unsigned(u_int64_t, *tl++);
8110 					dp->nfsdl_sizelimit *=
8111 					    fxdr_unsigned(u_int64_t, *tl);
8112 					break;
8113 				default:
8114 					error = NFSERR_BADXDR;
8115 					goto nfsmout;
8116 				};
8117 			} else {
8118 				dp->nfsdl_flags = NFSCLDL_READ;
8119 			}
8120 			if (ret != 0)
8121 				dp->nfsdl_flags |= NFSCLDL_RECALL;
8122 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
8123 			    &acesize, p);
8124 			if (error != 0)
8125 				goto nfsmout;
8126 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
8127 			error = NFSERR_BADXDR;
8128 			goto nfsmout;
8129 		}
8130 
8131 		/* Now, we should have the status for the SaveFH. */
8132 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8133 		if (*++tl == 0) {
8134 			NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
8135 			/*
8136 			 * Now, process the GetFH and Getattr for the newly
8137 			 * created file. nfscl_mtofh() will set
8138 			 * ND_NOMOREDATA if these weren't successful.
8139 			 */
8140 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
8141 			NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
8142 			if (error != 0)
8143 				goto nfsmout;
8144 		} else
8145 			nd->nd_flag |= ND_NOMOREDATA;
8146 		/* Now we have the PutFH and Getattr for the directory. */
8147 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8148 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8149 			if (*++tl != 0)
8150 				nd->nd_flag |= ND_NOMOREDATA;
8151 			else {
8152 				NFSM_DISSECT(tl, uint32_t *, 2 *
8153 				    NFSX_UNSIGNED);
8154 				if (*++tl != 0)
8155 					nd->nd_flag |= ND_NOMOREDATA;
8156 			}
8157 		}
8158 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8159 			/* Load the directory attributes. */
8160 			error = nfsm_loadattr(nd, dnap);
8161 			NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
8162 			if (error != 0)
8163 				goto nfsmout;
8164 			*dattrflagp = 1;
8165 			if (dp != NULL && *attrflagp != 0) {
8166 				dp->nfsdl_change = nnap->na_filerev;
8167 				dp->nfsdl_modtime = nnap->na_mtime;
8168 				dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
8169 			}
8170 			/*
8171 			 * We can now complete the Open state.
8172 			 */
8173 			nfhp = *nfhpp;
8174 			if (dp != NULL) {
8175 				dp->nfsdl_fhlen = nfhp->nfh_len;
8176 				NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
8177 				    nfhp->nfh_len);
8178 			}
8179 			/*
8180 			 * Get an Open structure that will be
8181 			 * attached to the OpenOwner, acquired already.
8182 			 */
8183 			error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
8184 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
8185 			    cred, p, NULL, &op, &newone, NULL, 0, false);
8186 			if (error != 0)
8187 				goto nfsmout;
8188 			op->nfso_stateid = stateid;
8189 			newnfs_copyincred(cred, &op->nfso_cred);
8190 
8191 			nfscl_openrelease(nmp, op, error, newone);
8192 			*unlockedp = 1;
8193 
8194 			/* Now, handle the RestoreFH and LayoutGet. */
8195 			if (nd->nd_repstat == 0) {
8196 				NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
8197 				*laystatp = fxdr_unsigned(int, *(tl + 3));
8198 				if (*laystatp == 0) {
8199 					error = nfsrv_parselayoutget(nmp, nd,
8200 					    stateidp, retonclosep, flhp);
8201 					if (error != 0)
8202 						*laystatp = error;
8203 				}
8204 				NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
8205 				    error);
8206 			} else
8207 				nd->nd_repstat = 0;
8208 		}
8209 	}
8210 	if (nd->nd_repstat != 0 && error == 0)
8211 		error = nd->nd_repstat;
8212 	if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
8213 		nfscl_initiate_recovery(owp->nfsow_clp);
8214 nfsmout:
8215 	NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
8216 	if (error == 0)
8217 		*dpp = dp;
8218 	else
8219 		free(dp, M_NFSCLDELEG);
8220 	m_freem(nd->nd_mrep);
8221 	return (error);
8222 }
8223 
8224 /*
8225  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
8226  */
8227 static int
8228 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
8229     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
8230     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
8231     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
8232     int *dattrflagp, void *dstuff, int *unlockedp)
8233 {
8234 	struct nfscllayout *lyp;
8235 	struct nfsclflayouthead flh;
8236 	struct nfsfh *nfhp;
8237 	struct nfsclsession *tsep;
8238 	struct nfsmount *nmp;
8239 	nfsv4stateid_t stateid;
8240 	int error, layoutlen, layouttype, retonclose, laystat;
8241 
8242 	error = 0;
8243 	nmp = VFSTONFS(dvp->v_mount);
8244 	if (NFSHASFLEXFILE(nmp))
8245 		layouttype = NFSLAYOUT_FLEXFILE;
8246 	else
8247 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
8248 	LIST_INIT(&flh);
8249 	tsep = nfsmnt_mdssession(nmp);
8250 	layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
8251 	error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
8252 	    owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
8253 	    dstuff, unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
8254 	    &flh, &laystat);
8255 	NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
8256 	    laystat, error);
8257 	lyp = NULL;
8258 	if (laystat == 0) {
8259 		nfhp = *nfhpp;
8260 		laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
8261 		    nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
8262 		    layouttype, laystat, NULL, cred, p);
8263 	} else
8264 		laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
8265 		    retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
8266 		    cred, p);
8267 	if (laystat == 0)
8268 		nfscl_rellayout(lyp, 0);
8269 	return (error);
8270 }
8271 
8272 /*
8273  * Process the results of a layoutget() operation.
8274  */
8275 static int
8276 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
8277     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
8278     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
8279     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
8280 {
8281 	struct nfsclflayout *tflp;
8282 	struct nfscldevinfo *dip;
8283 	uint8_t *dev;
8284 	int i, mirrorcnt;
8285 
8286 	if (laystat == NFSERR_UNKNLAYOUTTYPE) {
8287 		NFSLOCKMNT(nmp);
8288 		if (!NFSHASFLEXFILE(nmp)) {
8289 			/* Switch to using Flex File Layout. */
8290 			nmp->nm_state |= NFSSTA_FLEXFILE;
8291 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
8292 			/* Disable pNFS. */
8293 			NFSCL_DEBUG(1, "disable PNFS\n");
8294 			nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
8295 		}
8296 		NFSUNLOCKMNT(nmp);
8297 	}
8298 	if (laystat == 0) {
8299 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
8300 		LIST_FOREACH(tflp, flhp, nfsfl_list) {
8301 			if (layouttype == NFSLAYOUT_FLEXFILE)
8302 				mirrorcnt = tflp->nfsfl_mirrorcnt;
8303 			else
8304 				mirrorcnt = 1;
8305 			for (i = 0; i < mirrorcnt; i++) {
8306 				laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
8307 				NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
8308 				if (laystat != 0) {
8309 					if (layouttype == NFSLAYOUT_FLEXFILE)
8310 						dev = tflp->nfsfl_ffm[i].dev;
8311 					else
8312 						dev = tflp->nfsfl_dev;
8313 					laystat = nfsrpc_getdeviceinfo(nmp, dev,
8314 					    layouttype, notifybit, &dip, cred,
8315 					    p);
8316 					NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
8317 					    laystat);
8318 					if (laystat != 0)
8319 						goto out;
8320 					laystat = nfscl_adddevinfo(nmp, dip, i,
8321 					    tflp);
8322 					if (laystat != 0)
8323 						printf("nfsrpc_layoutgetresout"
8324 						    ": cannot add\n");
8325 				}
8326 			}
8327 		}
8328 	}
8329 out:
8330 	if (laystat == 0) {
8331 		/*
8332 		 * nfscl_layout() always returns with the nfsly_lock
8333 		 * set to a refcnt (shared lock).
8334 		 * Passing in dvp is sufficient, since it is only used to
8335 		 * get the fsid for the file system.
8336 		 */
8337 		laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
8338 		    layouttype, retonclose, flhp, lypp, cred, p);
8339 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
8340 		    laystat);
8341 		if (laystat == 0 && islockedp != NULL)
8342 			*islockedp = 1;
8343 	}
8344 	return (laystat);
8345 }
8346 
8347 /*
8348  * nfs copy_file_range operation.
8349  */
8350 int
8351 nfsrpc_copy_file_range(vnode_t invp, off_t *inoffp, vnode_t outvp,
8352     off_t *outoffp, size_t *lenp, unsigned int flags, int *inattrflagp,
8353     struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap,
8354     struct ucred *cred, bool consecutive, bool *must_commitp)
8355 {
8356 	int commit, error, expireret = 0, retrycnt;
8357 	u_int32_t clidrev = 0;
8358 	struct nfsmount *nmp = VFSTONFS(invp->v_mount);
8359 	struct nfsfh *innfhp = NULL, *outnfhp = NULL;
8360 	nfsv4stateid_t instateid, outstateid;
8361 	void *inlckp, *outlckp;
8362 
8363 	if (nmp->nm_clp != NULL)
8364 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8365 	innfhp = VTONFS(invp)->n_fhp;
8366 	outnfhp = VTONFS(outvp)->n_fhp;
8367 	retrycnt = 0;
8368 	do {
8369 		/* Get both stateids. */
8370 		inlckp = NULL;
8371 		nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len,
8372 		    NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid,
8373 		    &inlckp);
8374 		outlckp = NULL;
8375 		nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len,
8376 		    NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid,
8377 		    &outlckp);
8378 
8379 		error = nfsrpc_copyrpc(invp, *inoffp, outvp, *outoffp, lenp,
8380 		    &instateid, &outstateid, innap, inattrflagp, outnap,
8381 		    outattrflagp, consecutive, &commit, cred, curthread);
8382 		if (error == 0) {
8383 			if (commit != NFSWRITE_FILESYNC)
8384 				*must_commitp = true;
8385 			*inoffp += *lenp;
8386 			*outoffp += *lenp;
8387 		} else if (error == NFSERR_STALESTATEID)
8388 			nfscl_initiate_recovery(nmp->nm_clp);
8389 		if (inlckp != NULL)
8390 			nfscl_lockderef(inlckp);
8391 		if (outlckp != NULL)
8392 			nfscl_lockderef(outlckp);
8393 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8394 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8395 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8396 			(void) nfs_catnap(PZERO, error, "nfs_cfr");
8397 		} else if ((error == NFSERR_EXPIRED ||
8398 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8399 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8400 			    curthread);
8401 		}
8402 		retrycnt++;
8403 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
8404 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8405 	      error == NFSERR_STALEDONTRECOVER ||
8406 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8407 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8408 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
8409 	if (error != 0 && (retrycnt >= 4 ||
8410 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8411 	      error == NFSERR_STALEDONTRECOVER))
8412 		error = EIO;
8413 	return (error);
8414 }
8415 
8416 /*
8417  * The copy RPC.
8418  */
8419 static int
8420 nfsrpc_copyrpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff,
8421     size_t *lenp, nfsv4stateid_t *instateidp, nfsv4stateid_t *outstateidp,
8422     struct nfsvattr *innap, int *inattrflagp, struct nfsvattr *outnap,
8423     int *outattrflagp, bool consecutive, int *commitp, struct ucred *cred,
8424     NFSPROC_T *p)
8425 {
8426 	uint32_t *tl;
8427 	int error;
8428 	struct nfsrv_descript nfsd;
8429 	struct nfsrv_descript *nd = &nfsd;
8430 	struct nfsmount *nmp;
8431 	nfsattrbit_t attrbits;
8432 	uint64_t len;
8433 
8434 	nmp = VFSTONFS(outvp->v_mount);
8435 	*inattrflagp = *outattrflagp = 0;
8436 	*commitp = NFSWRITE_UNSTABLE;
8437 	len = *lenp;
8438 	*lenp = 0;
8439 	if (len > nfs_maxcopyrange)
8440 		len = nfs_maxcopyrange;
8441 	NFSCL_REQSTART(nd, NFSPROC_COPY, invp);
8442 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8443 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8444 	NFSGETATTR_ATTRBIT(&attrbits);
8445 	nfsrv_putattrbit(nd, &attrbits);
8446 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8447 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8448 	nfsm_fhtom(nd, VTONFS(outvp)->n_fhp->nfh_fh,
8449 	    VTONFS(outvp)->n_fhp->nfh_len, 0);
8450 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8451 	*tl = txdr_unsigned(NFSV4OP_COPY);
8452 	nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
8453 	nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID);
8454 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + 4 * NFSX_UNSIGNED);
8455 	txdr_hyper(inoff, tl); tl += 2;
8456 	txdr_hyper(outoff, tl); tl += 2;
8457 	txdr_hyper(len, tl); tl += 2;
8458 	if (consecutive)
8459 		*tl++ = newnfs_true;
8460 	else
8461 		*tl++ = newnfs_false;
8462 	*tl++ = newnfs_true;
8463 	*tl++ = 0;
8464 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8465 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
8466 	nfsrv_putattrbit(nd, &attrbits);
8467 	error = nfscl_request(nd, invp, p, cred, NULL);
8468 	if (error != 0)
8469 		return (error);
8470 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8471 		/* Get the input file's attributes. */
8472 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8473 		if (*(tl + 1) == 0) {
8474 			error = nfsm_loadattr(nd, innap);
8475 			if (error != 0)
8476 				goto nfsmout;
8477 			*inattrflagp = 1;
8478 		} else
8479 			nd->nd_flag |= ND_NOMOREDATA;
8480 	}
8481 	/* Skip over return stat for PutFH. */
8482 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8483 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8484 		if (*++tl != 0)
8485 			nd->nd_flag |= ND_NOMOREDATA;
8486 	}
8487 	/* Skip over return stat for Copy. */
8488 	if ((nd->nd_flag & ND_NOMOREDATA) == 0)
8489 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8490 	if (nd->nd_repstat == 0) {
8491 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8492 		if (*tl != 0) {
8493 			/* There should be no callback ids. */
8494 			error = NFSERR_BADXDR;
8495 			goto nfsmout;
8496 		}
8497 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED +
8498 		    NFSX_VERF);
8499 		len = fxdr_hyper(tl); tl += 2;
8500 		*commitp = fxdr_unsigned(int, *tl++);
8501 		NFSLOCKMNT(nmp);
8502 		if (!NFSHASWRITEVERF(nmp)) {
8503 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8504 			NFSSETWRITEVERF(nmp);
8505 	    	} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
8506 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8507 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
8508 		}
8509 		NFSUNLOCKMNT(nmp);
8510 		tl += (NFSX_VERF / NFSX_UNSIGNED);
8511 		if (nd->nd_repstat == 0 && *++tl != newnfs_true)
8512 			/* Must be a synchronous copy. */
8513 			nd->nd_repstat = NFSERR_NOTSUPP;
8514 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8515 		error = nfsm_loadattr(nd, outnap);
8516 		if (error == 0)
8517 			*outattrflagp = NFS_LATTR_NOSHRINK;
8518 		if (nd->nd_repstat == 0)
8519 			*lenp = len;
8520 	} else if (nd->nd_repstat == NFSERR_OFFLOADNOREQS) {
8521 		/*
8522 		 * For the case where consecutive is not supported, but
8523 		 * synchronous is supported, we can try consecutive == false
8524 		 * by returning this error.  Otherwise, return NFSERR_NOTSUPP,
8525 		 * since Copy cannot be done.
8526 		 */
8527 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8528 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8529 			if (!consecutive || *++tl == newnfs_false)
8530 				nd->nd_repstat = NFSERR_NOTSUPP;
8531 		} else
8532 			nd->nd_repstat = NFSERR_BADXDR;
8533 	}
8534 	if (error == 0)
8535 		error = nd->nd_repstat;
8536 nfsmout:
8537 	m_freem(nd->nd_mrep);
8538 	return (error);
8539 }
8540 
8541 /*
8542  * Seek operation.
8543  */
8544 int
8545 nfsrpc_seek(vnode_t vp, off_t *offp, bool *eofp, int content,
8546     struct ucred *cred, struct nfsvattr *nap, int *attrflagp)
8547 {
8548 	int error, expireret = 0, retrycnt;
8549 	u_int32_t clidrev = 0;
8550 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
8551 	struct nfsnode *np = VTONFS(vp);
8552 	struct nfsfh *nfhp = NULL;
8553 	nfsv4stateid_t stateid;
8554 	void *lckp;
8555 
8556 	if (nmp->nm_clp != NULL)
8557 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8558 	nfhp = np->n_fhp;
8559 	retrycnt = 0;
8560 	do {
8561 		lckp = NULL;
8562 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
8563 		    NFSV4OPEN_ACCESSREAD, 0, cred, curthread, &stateid, &lckp);
8564 		error = nfsrpc_seekrpc(vp, offp, &stateid, eofp, content,
8565 		    nap, attrflagp, cred);
8566 		if (error == NFSERR_STALESTATEID)
8567 			nfscl_initiate_recovery(nmp->nm_clp);
8568 		if (lckp != NULL)
8569 			nfscl_lockderef(lckp);
8570 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8571 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8572 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8573 			(void) nfs_catnap(PZERO, error, "nfs_seek");
8574 		} else if ((error == NFSERR_EXPIRED ||
8575 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8576 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8577 			    curthread);
8578 		}
8579 		retrycnt++;
8580 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8581 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8582 	    error == NFSERR_BADSESSION ||
8583 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8584 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8585 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
8586 	    (error == NFSERR_OPENMODE && retrycnt < 4));
8587 	if (error && retrycnt >= 4)
8588 		error = EIO;
8589 	return (error);
8590 }
8591 
8592 /*
8593  * The seek RPC.
8594  */
8595 static int
8596 nfsrpc_seekrpc(vnode_t vp, off_t *offp, nfsv4stateid_t *stateidp, bool *eofp,
8597     int content, struct nfsvattr *nap, int *attrflagp, struct ucred *cred)
8598 {
8599 	uint32_t *tl;
8600 	int error;
8601 	struct nfsrv_descript nfsd;
8602 	struct nfsrv_descript *nd = &nfsd;
8603 	nfsattrbit_t attrbits;
8604 
8605 	*attrflagp = 0;
8606 	NFSCL_REQSTART(nd, NFSPROC_SEEK, vp);
8607 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
8608 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8609 	txdr_hyper(*offp, tl); tl += 2;
8610 	*tl++ = txdr_unsigned(content);
8611 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8612 	NFSGETATTR_ATTRBIT(&attrbits);
8613 	nfsrv_putattrbit(nd, &attrbits);
8614 	error = nfscl_request(nd, vp, curthread, cred, NULL);
8615 	if (error != 0)
8616 		return (error);
8617 	if (nd->nd_repstat == 0) {
8618 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED + NFSX_HYPER);
8619 		if (*tl++ == newnfs_true)
8620 			*eofp = true;
8621 		else
8622 			*eofp = false;
8623 		*offp = fxdr_hyper(tl);
8624 		/* Just skip over Getattr op status. */
8625 		error = nfsm_loadattr(nd, nap);
8626 		if (error == 0)
8627 			*attrflagp = 1;
8628 	}
8629 	error = nd->nd_repstat;
8630 nfsmout:
8631 	m_freem(nd->nd_mrep);
8632 	return (error);
8633 }
8634 
8635 /*
8636  * The getextattr RPC.
8637  */
8638 int
8639 nfsrpc_getextattr(vnode_t vp, const char *name, struct uio *uiop, ssize_t *lenp,
8640     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8641 {
8642 	uint32_t *tl;
8643 	int error;
8644 	struct nfsrv_descript nfsd;
8645 	struct nfsrv_descript *nd = &nfsd;
8646 	nfsattrbit_t attrbits;
8647 	uint32_t len, len2;
8648 
8649 	*attrflagp = 0;
8650 	NFSCL_REQSTART(nd, NFSPROC_GETEXTATTR, vp);
8651 	nfsm_strtom(nd, name, strlen(name));
8652 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8653 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8654 	NFSGETATTR_ATTRBIT(&attrbits);
8655 	nfsrv_putattrbit(nd, &attrbits);
8656 	error = nfscl_request(nd, vp, p, cred, NULL);
8657 	if (error != 0)
8658 		return (error);
8659 	if (nd->nd_repstat == 0) {
8660 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8661 		len = fxdr_unsigned(uint32_t, *tl);
8662 		/* Sanity check lengths. */
8663 		if (uiop != NULL && len > 0 && len <= IOSIZE_MAX &&
8664 		    uiop->uio_resid <= UINT32_MAX) {
8665 			len2 = uiop->uio_resid;
8666 			if (len2 >= len)
8667 				error = nfsm_mbufuio(nd, uiop, len);
8668 			else {
8669 				error = nfsm_mbufuio(nd, uiop, len2);
8670 				if (error == 0) {
8671 					/*
8672 					 * nfsm_mbufuio() advances to a multiple
8673 					 * of 4, so round up len2 as well.  Then
8674 					 * we need to advance over the rest of
8675 					 * the data, rounding up the remaining
8676 					 * length.
8677 					 */
8678 					len2 = NFSM_RNDUP(len2);
8679 					len2 = NFSM_RNDUP(len - len2);
8680 					if (len2 > 0)
8681 						error = nfsm_advance(nd, len2,
8682 						    -1);
8683 				}
8684 			}
8685 		} else if (uiop == NULL && len > 0) {
8686 			/* Just wants the length and not the data. */
8687 			error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8688 		} else if (len > 0)
8689 			error = ENOATTR;
8690 		if (error != 0)
8691 			goto nfsmout;
8692 		*lenp = len;
8693 		/* Just skip over Getattr op status. */
8694 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8695 		error = nfsm_loadattr(nd, nap);
8696 		if (error == 0)
8697 			*attrflagp = 1;
8698 	}
8699 	if (error == 0)
8700 		error = nd->nd_repstat;
8701 nfsmout:
8702 	m_freem(nd->nd_mrep);
8703 	return (error);
8704 }
8705 
8706 /*
8707  * The setextattr RPC.
8708  */
8709 int
8710 nfsrpc_setextattr(vnode_t vp, const char *name, struct uio *uiop,
8711     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8712 {
8713 	uint32_t *tl;
8714 	int error;
8715 	struct nfsrv_descript nfsd;
8716 	struct nfsrv_descript *nd = &nfsd;
8717 	nfsattrbit_t attrbits;
8718 
8719 	*attrflagp = 0;
8720 	NFSCL_REQSTART(nd, NFSPROC_SETEXTATTR, vp);
8721 	if (uiop->uio_resid > nd->nd_maxreq) {
8722 		/* nd_maxreq is set by NFSCL_REQSTART(). */
8723 		m_freem(nd->nd_mreq);
8724 		return (EINVAL);
8725 	}
8726 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8727 	*tl = txdr_unsigned(NFSV4SXATTR_EITHER);
8728 	nfsm_strtom(nd, name, strlen(name));
8729 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8730 	*tl = txdr_unsigned(uiop->uio_resid);
8731 	nfsm_uiombuf(nd, uiop, uiop->uio_resid);
8732 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8733 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8734 	NFSGETATTR_ATTRBIT(&attrbits);
8735 	nfsrv_putattrbit(nd, &attrbits);
8736 	error = nfscl_request(nd, vp, p, cred, NULL);
8737 	if (error != 0)
8738 		return (error);
8739 	if (nd->nd_repstat == 0) {
8740 		/* Just skip over the reply and Getattr op status. */
8741 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8742 		    NFSX_UNSIGNED);
8743 		error = nfsm_loadattr(nd, nap);
8744 		if (error == 0)
8745 			*attrflagp = 1;
8746 	}
8747 	if (error == 0)
8748 		error = nd->nd_repstat;
8749 nfsmout:
8750 	m_freem(nd->nd_mrep);
8751 	return (error);
8752 }
8753 
8754 /*
8755  * The removeextattr RPC.
8756  */
8757 int
8758 nfsrpc_rmextattr(vnode_t vp, const char *name, struct nfsvattr *nap,
8759     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8760 {
8761 	uint32_t *tl;
8762 	int error;
8763 	struct nfsrv_descript nfsd;
8764 	struct nfsrv_descript *nd = &nfsd;
8765 	nfsattrbit_t attrbits;
8766 
8767 	*attrflagp = 0;
8768 	NFSCL_REQSTART(nd, NFSPROC_RMEXTATTR, vp);
8769 	nfsm_strtom(nd, name, strlen(name));
8770 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8771 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8772 	NFSGETATTR_ATTRBIT(&attrbits);
8773 	nfsrv_putattrbit(nd, &attrbits);
8774 	error = nfscl_request(nd, vp, p, cred, NULL);
8775 	if (error != 0)
8776 		return (error);
8777 	if (nd->nd_repstat == 0) {
8778 		/* Just skip over the reply and Getattr op status. */
8779 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8780 		    NFSX_UNSIGNED);
8781 		error = nfsm_loadattr(nd, nap);
8782 		if (error == 0)
8783 			*attrflagp = 1;
8784 	}
8785 	if (error == 0)
8786 		error = nd->nd_repstat;
8787 nfsmout:
8788 	m_freem(nd->nd_mrep);
8789 	return (error);
8790 }
8791 
8792 /*
8793  * The listextattr RPC.
8794  */
8795 int
8796 nfsrpc_listextattr(vnode_t vp, uint64_t *cookiep, struct uio *uiop,
8797     size_t *lenp, bool *eofp, struct nfsvattr *nap, int *attrflagp,
8798     struct ucred *cred, NFSPROC_T *p)
8799 {
8800 	uint32_t *tl;
8801 	int cnt, error, i, len;
8802 	struct nfsrv_descript nfsd;
8803 	struct nfsrv_descript *nd = &nfsd;
8804 	nfsattrbit_t attrbits;
8805 	u_char c;
8806 
8807 	*attrflagp = 0;
8808 	NFSCL_REQSTART(nd, NFSPROC_LISTEXTATTR, vp);
8809 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8810 	txdr_hyper(*cookiep, tl); tl += 2;
8811 	*tl++ = txdr_unsigned(*lenp);
8812 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8813 	NFSGETATTR_ATTRBIT(&attrbits);
8814 	nfsrv_putattrbit(nd, &attrbits);
8815 	error = nfscl_request(nd, vp, p, cred, NULL);
8816 	if (error != 0)
8817 		return (error);
8818 	*eofp = true;
8819 	*lenp = 0;
8820 	if (nd->nd_repstat == 0) {
8821 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
8822 		*cookiep = fxdr_hyper(tl); tl += 2;
8823 		cnt = fxdr_unsigned(int, *tl);
8824 		if (cnt < 0) {
8825 			error = EBADRPC;
8826 			goto nfsmout;
8827 		}
8828 		for (i = 0; i < cnt; i++) {
8829 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8830 			len = fxdr_unsigned(int, *tl);
8831 			if (len <= 0 || len > EXTATTR_MAXNAMELEN) {
8832 				error = EBADRPC;
8833 				goto nfsmout;
8834 			}
8835 			if (uiop == NULL)
8836 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8837 			else if (uiop->uio_resid >= len + 1) {
8838 				c = len;
8839 				error = uiomove(&c, sizeof(c), uiop);
8840 				if (error == 0)
8841 					error = nfsm_mbufuio(nd, uiop, len);
8842 			} else {
8843 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8844 				*eofp = false;
8845 			}
8846 			if (error != 0)
8847 				goto nfsmout;
8848 			*lenp += (len + 1);
8849 		}
8850 		/* Get the eof and skip over the Getattr op status. */
8851 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
8852 		/*
8853 		 * *eofp is set false above, because it wasn't able to copy
8854 		 * all of the reply.
8855 		 */
8856 		if (*eofp && *tl == 0)
8857 			*eofp = false;
8858 		error = nfsm_loadattr(nd, nap);
8859 		if (error == 0)
8860 			*attrflagp = 1;
8861 	}
8862 	if (error == 0)
8863 		error = nd->nd_repstat;
8864 nfsmout:
8865 	m_freem(nd->nd_mrep);
8866 	return (error);
8867 }
8868 
8869 /*
8870  * Split an mbuf list.  For non-M_EXTPG mbufs, just use m_split().
8871  */
8872 static struct mbuf *
8873 nfsm_split(struct mbuf *mp, uint64_t xfer)
8874 {
8875 	struct mbuf *m, *m2;
8876 	vm_page_t pg;
8877 	int i, j, left, pgno, plen, trim;
8878 	char *cp, *cp2;
8879 
8880 	if ((mp->m_flags & M_EXTPG) == 0) {
8881 		m = m_split(mp, xfer, M_WAITOK);
8882 		return (m);
8883 	}
8884 
8885 	/* Find the correct mbuf to split at. */
8886 	for (m = mp; m != NULL && xfer > m->m_len; m = m->m_next)
8887 		xfer -= m->m_len;
8888 	if (m == NULL)
8889 		return (NULL);
8890 
8891 	/* If xfer == m->m_len, we can just split the mbuf list. */
8892 	if (xfer == m->m_len) {
8893 		m2 = m->m_next;
8894 		m->m_next = NULL;
8895 		return (m2);
8896 	}
8897 
8898 	/* Find the page to split at. */
8899 	pgno = 0;
8900 	left = xfer;
8901 	do {
8902 		if (pgno == 0)
8903 			plen = m_epg_pagelen(m, 0, m->m_epg_1st_off);
8904 		else
8905 			plen = m_epg_pagelen(m, pgno, 0);
8906 		if (left <= plen)
8907 			break;
8908 		left -= plen;
8909 		pgno++;
8910 	} while (pgno < m->m_epg_npgs);
8911 	if (pgno == m->m_epg_npgs)
8912 		panic("nfsm_split: eroneous ext_pgs mbuf");
8913 
8914 	m2 = mb_alloc_ext_pgs(M_WAITOK, mb_free_mext_pgs);
8915 	m2->m_epg_flags |= EPG_FLAG_ANON;
8916 
8917 	/*
8918 	 * If left < plen, allocate a new page for the new mbuf
8919 	 * and copy the data after left in the page to this new
8920 	 * page.
8921 	 */
8922 	if (left < plen) {
8923 		pg = vm_page_alloc_noobj(VM_ALLOC_WAITOK | VM_ALLOC_NODUMP |
8924 		    VM_ALLOC_WIRED);
8925 		m2->m_epg_pa[0] = VM_PAGE_TO_PHYS(pg);
8926 		m2->m_epg_npgs = 1;
8927 
8928 		/* Copy the data after left to the new page. */
8929 		trim = plen - left;
8930 		cp = (char *)(void *)PHYS_TO_DMAP(m->m_epg_pa[pgno]);
8931 		if (pgno == 0)
8932 			cp += m->m_epg_1st_off;
8933 		cp += left;
8934 		cp2 = (char *)(void *)PHYS_TO_DMAP(m2->m_epg_pa[0]);
8935 		if (pgno == m->m_epg_npgs - 1)
8936 			m2->m_epg_last_len = trim;
8937 		else {
8938 			cp2 += PAGE_SIZE - trim;
8939 			m2->m_epg_1st_off = PAGE_SIZE - trim;
8940 			m2->m_epg_last_len = m->m_epg_last_len;
8941 		}
8942 		memcpy(cp2, cp, trim);
8943 		m2->m_len = trim;
8944 	} else {
8945 		m2->m_len = 0;
8946 		m2->m_epg_last_len = m->m_epg_last_len;
8947 	}
8948 
8949 	/* Move the pages beyond pgno to the new mbuf. */
8950 	for (i = pgno + 1, j = m2->m_epg_npgs; i < m->m_epg_npgs; i++, j++) {
8951 		m2->m_epg_pa[j] = m->m_epg_pa[i];
8952 		/* Never moves page 0. */
8953 		m2->m_len += m_epg_pagelen(m, i, 0);
8954 	}
8955 	m2->m_epg_npgs = j;
8956 	m->m_epg_npgs = pgno + 1;
8957 	m->m_epg_last_len = left;
8958 	m->m_len = xfer;
8959 
8960 	m2->m_next = m->m_next;
8961 	m->m_next = NULL;
8962 	return (m2);
8963 }
8964 
8965 /*
8966  * Do the NFSv4.1 Bind Connection to Session.
8967  * Called from the reconnect layer of the krpc (sys/rpc/clnt_rc.c).
8968  */
8969 void
8970 nfsrpc_bindconnsess(CLIENT *cl, void *arg, struct ucred *cr)
8971 {
8972 	struct nfscl_reconarg *rcp = (struct nfscl_reconarg *)arg;
8973 	uint32_t res, *tl;
8974 	struct nfsrv_descript nfsd;
8975 	struct nfsrv_descript *nd = &nfsd;
8976 	struct rpc_callextra ext;
8977 	struct timeval utimeout;
8978 	enum clnt_stat stat;
8979 	int error;
8980 
8981 	nfscl_reqstart(nd, NFSPROC_BINDCONNTOSESS, NULL, NULL, 0, NULL, NULL,
8982 	    NFS_VER4, rcp->minorvers);
8983 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID + 2 * NFSX_UNSIGNED);
8984 	memcpy(tl, rcp->sessionid, NFSX_V4SESSIONID);
8985 	tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
8986 	*tl++ = txdr_unsigned(NFSCDFC4_FORE_OR_BOTH);
8987 	*tl = newnfs_false;
8988 
8989 	memset(&ext, 0, sizeof(ext));
8990 	utimeout.tv_sec = 30;
8991 	utimeout.tv_usec = 0;
8992 	ext.rc_auth = authunix_create(cr);
8993 	nd->nd_mrep = NULL;
8994 	stat = CLNT_CALL_MBUF(cl, &ext, NFSV4PROC_COMPOUND, nd->nd_mreq,
8995 	    &nd->nd_mrep, utimeout);
8996 	AUTH_DESTROY(ext.rc_auth);
8997 	if (stat != RPC_SUCCESS) {
8998 		printf("nfsrpc_bindconnsess: call failed stat=%d\n", stat);
8999 		return;
9000 	}
9001 	if (nd->nd_mrep == NULL) {
9002 		printf("nfsrpc_bindconnsess: no reply args\n");
9003 		return;
9004 	}
9005 	error = 0;
9006 	newnfs_realign(&nd->nd_mrep, M_WAITOK);
9007 	nd->nd_md = nd->nd_mrep;
9008 	nd->nd_dpos = mtod(nd->nd_md, char *);
9009 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
9010 	nd->nd_repstat = fxdr_unsigned(uint32_t, *tl++);
9011 	if (nd->nd_repstat == NFSERR_OK) {
9012 		res = fxdr_unsigned(uint32_t, *tl);
9013 		if (res > 0 && (error = nfsm_advance(nd, NFSM_RNDUP(res),
9014 		    -1)) != 0)
9015 			goto nfsmout;
9016 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
9017 		    4 * NFSX_UNSIGNED);
9018 		tl += 3;
9019 		if (!NFSBCMP(tl, rcp->sessionid, NFSX_V4SESSIONID)) {
9020 			tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
9021 			res = fxdr_unsigned(uint32_t, *tl);
9022 			if (res != NFSCDFS4_BOTH)
9023 				printf("nfsrpc_bindconnsess: did not "
9024 				    "return FS4_BOTH\n");
9025 		} else
9026 			printf("nfsrpc_bindconnsess: not same "
9027 			    "sessionid\n");
9028 	} else if (nd->nd_repstat != NFSERR_BADSESSION)
9029 		printf("nfsrpc_bindconnsess: returned %d\n", nd->nd_repstat);
9030 nfsmout:
9031 	if (error != 0)
9032 		printf("nfsrpc_bindconnsess: reply bad xdr\n");
9033 	m_freem(nd->nd_mrep);
9034 }
9035