xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision 0b3105a37d7adcadcb720112fed4dc4e8040be99)
1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Rick Macklem at The University of Guelph.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 4. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 /*
38  * Rpc op calls, generally called from the vnode op calls or through the
39  * buffer cache, for NFS v2, 3 and 4.
40  * These do not normally make any changes to vnode arguments or use
41  * structures that might change between the VFS variants. The returned
42  * arguments are all at the end, after the NFSPROC_T *p one.
43  */
44 
45 #ifndef APPLEKEXT
46 #include "opt_inet6.h"
47 
48 #include <fs/nfs/nfsport.h>
49 #include <sys/sysctl.h>
50 
51 SYSCTL_DECL(_vfs_nfs);
52 
53 static int	nfsignore_eexist = 0;
54 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
55     &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
56 
57 /*
58  * Global variables
59  */
60 extern int nfs_numnfscbd;
61 extern struct timeval nfsboottime;
62 extern u_int32_t newnfs_false, newnfs_true;
63 extern nfstype nfsv34_type[9];
64 extern int nfsrv_useacl;
65 extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
66 extern int nfscl_debuglevel;
67 NFSCLSTATEMUTEX;
68 int nfstest_outofseq = 0;
69 int nfscl_assumeposixlocks = 1;
70 int nfscl_enablecallb = 0;
71 short nfsv4_cbport = NFSV4_CBPORT;
72 int nfstest_openallsetattr = 0;
73 #endif	/* !APPLEKEXT */
74 
75 #define	DIRHDSIZ	(sizeof (struct dirent) - (MAXNAMLEN + 1))
76 
77 /*
78  * nfscl_getsameserver() can return one of three values:
79  * NFSDSP_USETHISSESSION - Use this session for the DS.
80  * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
81  *     session.
82  * NFSDSP_NOTFOUND - No matching server was found.
83  */
84 enum nfsclds_state {
85 	NFSDSP_USETHISSESSION = 0,
86 	NFSDSP_SEQTHISSESSION = 1,
87 	NFSDSP_NOTFOUND = 2,
88 };
89 
90 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
91     struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
92 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
93     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
94 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
95     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
96     void *);
97 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
98     nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
99     struct nfsvattr *, struct nfsfh **, int *, int *, void *);
100 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
101     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
102     NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
103     int *, void *, int *);
104 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
105     struct nfscllockowner *, u_int64_t, u_int64_t,
106     u_int32_t, struct ucred *, NFSPROC_T *, int);
107 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
108     struct acl *, nfsv4stateid_t *, void *);
109 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
110     uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
111     struct ucred *, NFSPROC_T *);
112 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_storage *,
113     struct nfsclds **, NFSPROC_T *);
114 static void nfscl_initsessionslots(struct nfsclsession *);
115 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
116     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
117     struct nfsclflayout *, uint64_t, uint64_t, struct ucred *, NFSPROC_T *);
118 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
119     struct nfsclds *, uint64_t, int, struct nfsfh *, struct ucred *,
120     NFSPROC_T *);
121 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
122     nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
123     struct nfsfh *, int, struct ucred *, NFSPROC_T *);
124 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
125     struct nfsclds *, struct nfsclds **);
126 #ifdef notyet
127 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
128     struct nfsfh *, struct ucred *, NFSPROC_T *, void *);
129 #endif
130 
131 /*
132  * nfs null call from vfs.
133  */
134 APPLESTATIC int
135 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
136 {
137 	int error;
138 	struct nfsrv_descript nfsd, *nd = &nfsd;
139 
140 	NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
141 	error = nfscl_request(nd, vp, p, cred, NULL);
142 	if (nd->nd_repstat && !error)
143 		error = nd->nd_repstat;
144 	mbuf_freem(nd->nd_mrep);
145 	return (error);
146 }
147 
148 /*
149  * nfs access rpc op.
150  * For nfs version 3 and 4, use the access rpc to check accessibility. If file
151  * modes are changed on the server, accesses might still fail later.
152  */
153 APPLESTATIC int
154 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
155     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
156 {
157 	int error;
158 	u_int32_t mode, rmode;
159 
160 	if (acmode & VREAD)
161 		mode = NFSACCESS_READ;
162 	else
163 		mode = 0;
164 	if (vnode_vtype(vp) == VDIR) {
165 		if (acmode & VWRITE)
166 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
167 				 NFSACCESS_DELETE);
168 		if (acmode & VEXEC)
169 			mode |= NFSACCESS_LOOKUP;
170 	} else {
171 		if (acmode & VWRITE)
172 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
173 		if (acmode & VEXEC)
174 			mode |= NFSACCESS_EXECUTE;
175 	}
176 
177 	/*
178 	 * Now, just call nfsrpc_accessrpc() to do the actual RPC.
179 	 */
180 	error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
181 	    NULL);
182 
183 	/*
184 	 * The NFS V3 spec does not clarify whether or not
185 	 * the returned access bits can be a superset of
186 	 * the ones requested, so...
187 	 */
188 	if (!error && (rmode & mode) != mode)
189 		error = EACCES;
190 	return (error);
191 }
192 
193 /*
194  * The actual rpc, separated out for Darwin.
195  */
196 APPLESTATIC int
197 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
198     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
199     void *stuff)
200 {
201 	u_int32_t *tl;
202 	u_int32_t supported, rmode;
203 	int error;
204 	struct nfsrv_descript nfsd, *nd = &nfsd;
205 	nfsattrbit_t attrbits;
206 
207 	*attrflagp = 0;
208 	supported = mode;
209 	NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
210 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
211 	*tl = txdr_unsigned(mode);
212 	if (nd->nd_flag & ND_NFSV4) {
213 		/*
214 		 * And do a Getattr op.
215 		 */
216 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
217 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
218 		NFSGETATTR_ATTRBIT(&attrbits);
219 		(void) nfsrv_putattrbit(nd, &attrbits);
220 	}
221 	error = nfscl_request(nd, vp, p, cred, stuff);
222 	if (error)
223 		return (error);
224 	if (nd->nd_flag & ND_NFSV3) {
225 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
226 		if (error)
227 			goto nfsmout;
228 	}
229 	if (!nd->nd_repstat) {
230 		if (nd->nd_flag & ND_NFSV4) {
231 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
232 			supported = fxdr_unsigned(u_int32_t, *tl++);
233 		} else {
234 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
235 		}
236 		rmode = fxdr_unsigned(u_int32_t, *tl);
237 		if (nd->nd_flag & ND_NFSV4)
238 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
239 
240 		/*
241 		 * It's not obvious what should be done about
242 		 * unsupported access modes. For now, be paranoid
243 		 * and clear the unsupported ones.
244 		 */
245 		rmode &= supported;
246 		*rmodep = rmode;
247 	} else
248 		error = nd->nd_repstat;
249 nfsmout:
250 	mbuf_freem(nd->nd_mrep);
251 	return (error);
252 }
253 
254 /*
255  * nfs open rpc
256  */
257 APPLESTATIC int
258 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
259 {
260 	struct nfsclopen *op;
261 	struct nfscldeleg *dp;
262 	struct nfsfh *nfhp;
263 	struct nfsnode *np = VTONFS(vp);
264 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
265 	u_int32_t mode, clidrev;
266 	int ret, newone, error, expireret = 0, retrycnt;
267 
268 	/*
269 	 * For NFSv4, Open Ops are only done on Regular Files.
270 	 */
271 	if (vnode_vtype(vp) != VREG)
272 		return (0);
273 	mode = 0;
274 	if (amode & FREAD)
275 		mode |= NFSV4OPEN_ACCESSREAD;
276 	if (amode & FWRITE)
277 		mode |= NFSV4OPEN_ACCESSWRITE;
278 	nfhp = np->n_fhp;
279 
280 	retrycnt = 0;
281 #ifdef notdef
282 { char name[100]; int namel;
283 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
284 bcopy(NFS4NODENAME(np->n_v4), name, namel);
285 name[namel] = '\0';
286 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
287 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
288 else printf(" fhl=0\n");
289 }
290 #endif
291 	do {
292 	    dp = NULL;
293 	    error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
294 		cred, p, NULL, &op, &newone, &ret, 1);
295 	    if (error) {
296 		return (error);
297 	    }
298 	    if (nmp->nm_clp != NULL)
299 		clidrev = nmp->nm_clp->nfsc_clientidrev;
300 	    else
301 		clidrev = 0;
302 	    if (ret == NFSCLOPEN_DOOPEN) {
303 		if (np->n_v4 != NULL) {
304 			error = nfsrpc_openrpc(nmp, vp, np->n_v4->n4_data,
305 			   np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
306 			   np->n_fhp->nfh_len, mode, op,
307 			   NFS4NODENAME(np->n_v4), np->n_v4->n4_namelen, &dp,
308 			   0, 0x0, cred, p, 0, 0);
309 			if (dp != NULL) {
310 #ifdef APPLE
311 				OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
312 #else
313 				NFSLOCKNODE(np);
314 				np->n_flag &= ~NDELEGMOD;
315 				/*
316 				 * Invalidate the attribute cache, so that
317 				 * attributes that pre-date the issue of a
318 				 * delegation are not cached, since the
319 				 * cached attributes will remain valid while
320 				 * the delegation is held.
321 				 */
322 				NFSINVALATTRCACHE(np);
323 				NFSUNLOCKNODE(np);
324 #endif
325 				(void) nfscl_deleg(nmp->nm_mountp,
326 				    op->nfso_own->nfsow_clp,
327 				    nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
328 			}
329 		} else {
330 			error = EIO;
331 		}
332 		newnfs_copyincred(cred, &op->nfso_cred);
333 	    } else if (ret == NFSCLOPEN_SETCRED)
334 		/*
335 		 * This is a new local open on a delegation. It needs
336 		 * to have credentials so that an open can be done
337 		 * against the server during recovery.
338 		 */
339 		newnfs_copyincred(cred, &op->nfso_cred);
340 
341 	    /*
342 	     * nfso_opencnt is the count of how many VOP_OPEN()s have
343 	     * been done on this Open successfully and a VOP_CLOSE()
344 	     * is expected for each of these.
345 	     * If error is non-zero, don't increment it, since the Open
346 	     * hasn't succeeded yet.
347 	     */
348 	    if (!error)
349 		op->nfso_opencnt++;
350 	    nfscl_openrelease(op, error, newone);
351 	    if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
352 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
353 		error == NFSERR_BADSESSION) {
354 		(void) nfs_catnap(PZERO, error, "nfs_open");
355 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
356 		&& clidrev != 0) {
357 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
358 		retrycnt++;
359 	    }
360 	} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
361 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
362 	    error == NFSERR_BADSESSION ||
363 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
364 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
365 	if (error && retrycnt >= 4)
366 		error = EIO;
367 	return (error);
368 }
369 
370 /*
371  * the actual open rpc
372  */
373 APPLESTATIC int
374 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
375     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
376     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
377     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
378     int syscred, int recursed)
379 {
380 	u_int32_t *tl;
381 	struct nfsrv_descript nfsd, *nd = &nfsd;
382 	struct nfscldeleg *dp, *ndp = NULL;
383 	struct nfsvattr nfsva;
384 	u_int32_t rflags, deleg;
385 	nfsattrbit_t attrbits;
386 	int error, ret, acesize, limitby;
387 
388 	dp = *dpp;
389 	*dpp = NULL;
390 	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL);
391 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
392 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
393 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
394 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
395 	*tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
396 	*tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
397 	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
398 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
399 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
400 	if (reclaim) {
401 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
402 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
403 		*tl = txdr_unsigned(delegtype);
404 	} else {
405 		if (dp != NULL) {
406 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
407 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
408 			if (NFSHASNFSV4N(nmp))
409 				*tl++ = 0;
410 			else
411 				*tl++ = dp->nfsdl_stateid.seqid;
412 			*tl++ = dp->nfsdl_stateid.other[0];
413 			*tl++ = dp->nfsdl_stateid.other[1];
414 			*tl = dp->nfsdl_stateid.other[2];
415 		} else {
416 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
417 		}
418 		(void) nfsm_strtom(nd, name, namelen);
419 	}
420 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
421 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
422 	NFSZERO_ATTRBIT(&attrbits);
423 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
424 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
425 	(void) nfsrv_putattrbit(nd, &attrbits);
426 	if (syscred)
427 		nd->nd_flag |= ND_USEGSSNAME;
428 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
429 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
430 	if (error)
431 		return (error);
432 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
433 	if (!nd->nd_repstat) {
434 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
435 		    6 * NFSX_UNSIGNED);
436 		op->nfso_stateid.seqid = *tl++;
437 		op->nfso_stateid.other[0] = *tl++;
438 		op->nfso_stateid.other[1] = *tl++;
439 		op->nfso_stateid.other[2] = *tl;
440 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
441 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
442 		if (error)
443 			goto nfsmout;
444 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
445 		deleg = fxdr_unsigned(u_int32_t, *tl);
446 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
447 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
448 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
449 			      NFSCLFLAGS_FIRSTDELEG))
450 				op->nfso_own->nfsow_clp->nfsc_flags |=
451 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
452 			MALLOC(ndp, struct nfscldeleg *,
453 			    sizeof (struct nfscldeleg) + newfhlen,
454 			    M_NFSCLDELEG, M_WAITOK);
455 			LIST_INIT(&ndp->nfsdl_owner);
456 			LIST_INIT(&ndp->nfsdl_lock);
457 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
458 			ndp->nfsdl_fhlen = newfhlen;
459 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
460 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
461 			nfscl_lockinit(&ndp->nfsdl_rwlock);
462 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
463 			    NFSX_UNSIGNED);
464 			ndp->nfsdl_stateid.seqid = *tl++;
465 			ndp->nfsdl_stateid.other[0] = *tl++;
466 			ndp->nfsdl_stateid.other[1] = *tl++;
467 			ndp->nfsdl_stateid.other[2] = *tl++;
468 			ret = fxdr_unsigned(int, *tl);
469 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
470 				ndp->nfsdl_flags = NFSCLDL_WRITE;
471 				/*
472 				 * Indicates how much the file can grow.
473 				 */
474 				NFSM_DISSECT(tl, u_int32_t *,
475 				    3 * NFSX_UNSIGNED);
476 				limitby = fxdr_unsigned(int, *tl++);
477 				switch (limitby) {
478 				case NFSV4OPEN_LIMITSIZE:
479 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
480 					break;
481 				case NFSV4OPEN_LIMITBLOCKS:
482 					ndp->nfsdl_sizelimit =
483 					    fxdr_unsigned(u_int64_t, *tl++);
484 					ndp->nfsdl_sizelimit *=
485 					    fxdr_unsigned(u_int64_t, *tl);
486 					break;
487 				default:
488 					error = NFSERR_BADXDR;
489 					goto nfsmout;
490 				};
491 			} else {
492 				ndp->nfsdl_flags = NFSCLDL_READ;
493 			}
494 			if (ret)
495 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
496 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
497 			    &acesize, p);
498 			if (error)
499 				goto nfsmout;
500 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
501 			error = NFSERR_BADXDR;
502 			goto nfsmout;
503 		}
504 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
505 		error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
506 		    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
507 		    NULL, NULL, NULL, p, cred);
508 		if (error)
509 			goto nfsmout;
510 		if (ndp != NULL) {
511 			ndp->nfsdl_change = nfsva.na_filerev;
512 			ndp->nfsdl_modtime = nfsva.na_mtime;
513 			ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
514 		}
515 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
516 		    do {
517 			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
518 			    cred, p);
519 			if (ret == NFSERR_DELAY)
520 			    (void) nfs_catnap(PZERO, ret, "nfs_open");
521 		    } while (ret == NFSERR_DELAY);
522 		    error = ret;
523 		}
524 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
525 		    nfscl_assumeposixlocks)
526 		    op->nfso_posixlock = 1;
527 		else
528 		    op->nfso_posixlock = 0;
529 
530 		/*
531 		 * If the server is handing out delegations, but we didn't
532 		 * get one because an OpenConfirm was required, try the
533 		 * Open again, to get a delegation. This is a harmless no-op,
534 		 * from a server's point of view.
535 		 */
536 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
537 		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
538 		    && !error && dp == NULL && ndp == NULL && !recursed) {
539 		    do {
540 			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
541 			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
542 			    cred, p, syscred, 1);
543 			if (ret == NFSERR_DELAY)
544 			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
545 		    } while (ret == NFSERR_DELAY);
546 		    if (ret) {
547 			if (ndp != NULL) {
548 				FREE((caddr_t)ndp, M_NFSCLDELEG);
549 				ndp = NULL;
550 			}
551 			if (ret == NFSERR_STALECLIENTID ||
552 			    ret == NFSERR_STALEDONTRECOVER ||
553 			    ret == NFSERR_BADSESSION)
554 				error = ret;
555 		    }
556 		}
557 	}
558 	if (nd->nd_repstat != 0 && error == 0)
559 		error = nd->nd_repstat;
560 	if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
561 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
562 nfsmout:
563 	if (!error)
564 		*dpp = ndp;
565 	else if (ndp != NULL)
566 		FREE((caddr_t)ndp, M_NFSCLDELEG);
567 	mbuf_freem(nd->nd_mrep);
568 	return (error);
569 }
570 
571 /*
572  * open downgrade rpc
573  */
574 APPLESTATIC int
575 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
576     struct ucred *cred, NFSPROC_T *p)
577 {
578 	u_int32_t *tl;
579 	struct nfsrv_descript nfsd, *nd = &nfsd;
580 	int error;
581 
582 	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
583 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
584 	if (NFSHASNFSV4N(VFSTONFS(vnode_mount(vp))))
585 		*tl++ = 0;
586 	else
587 		*tl++ = op->nfso_stateid.seqid;
588 	*tl++ = op->nfso_stateid.other[0];
589 	*tl++ = op->nfso_stateid.other[1];
590 	*tl++ = op->nfso_stateid.other[2];
591 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
592 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
593 	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
594 	error = nfscl_request(nd, vp, p, cred, NULL);
595 	if (error)
596 		return (error);
597 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
598 	if (!nd->nd_repstat) {
599 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
600 		op->nfso_stateid.seqid = *tl++;
601 		op->nfso_stateid.other[0] = *tl++;
602 		op->nfso_stateid.other[1] = *tl++;
603 		op->nfso_stateid.other[2] = *tl;
604 	}
605 	if (nd->nd_repstat && error == 0)
606 		error = nd->nd_repstat;
607 	if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
608 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
609 nfsmout:
610 	mbuf_freem(nd->nd_mrep);
611 	return (error);
612 }
613 
614 /*
615  * V4 Close operation.
616  */
617 APPLESTATIC int
618 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
619 {
620 	struct nfsclclient *clp;
621 	int error;
622 
623 	if (vnode_vtype(vp) != VREG)
624 		return (0);
625 	if (doclose)
626 		error = nfscl_doclose(vp, &clp, p);
627 	else
628 		error = nfscl_getclose(vp, &clp);
629 	if (error)
630 		return (error);
631 
632 	nfscl_clientrelease(clp);
633 	return (0);
634 }
635 
636 /*
637  * Close the open.
638  */
639 APPLESTATIC void
640 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p)
641 {
642 	struct nfsrv_descript nfsd, *nd = &nfsd;
643 	struct nfscllockowner *lp, *nlp;
644 	struct nfscllock *lop, *nlop;
645 	struct ucred *tcred;
646 	u_int64_t off = 0, len = 0;
647 	u_int32_t type = NFSV4LOCKT_READ;
648 	int error, do_unlock, trycnt;
649 
650 	tcred = newnfs_getcred();
651 	newnfs_copycred(&op->nfso_cred, tcred);
652 	/*
653 	 * (Theoretically this could be done in the same
654 	 *  compound as the close, but having multiple
655 	 *  sequenced Ops in the same compound might be
656 	 *  too scary for some servers.)
657 	 */
658 	if (op->nfso_posixlock) {
659 		off = 0;
660 		len = NFS64BITSSET;
661 		type = NFSV4LOCKT_READ;
662 	}
663 
664 	/*
665 	 * Since this function is only called from VOP_INACTIVE(), no
666 	 * other thread will be manipulating this Open. As such, the
667 	 * lock lists are not being changed by other threads, so it should
668 	 * be safe to do this without locking.
669 	 */
670 	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
671 		do_unlock = 1;
672 		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
673 			if (op->nfso_posixlock == 0) {
674 				off = lop->nfslo_first;
675 				len = lop->nfslo_end - lop->nfslo_first;
676 				if (lop->nfslo_type == F_WRLCK)
677 					type = NFSV4LOCKT_WRITE;
678 				else
679 					type = NFSV4LOCKT_READ;
680 			}
681 			if (do_unlock) {
682 				trycnt = 0;
683 				do {
684 					error = nfsrpc_locku(nd, nmp, lp, off,
685 					    len, type, tcred, p, 0);
686 					if ((nd->nd_repstat == NFSERR_GRACE ||
687 					    nd->nd_repstat == NFSERR_DELAY) &&
688 					    error == 0)
689 						(void) nfs_catnap(PZERO,
690 						    (int)nd->nd_repstat,
691 						    "nfs_close");
692 				} while ((nd->nd_repstat == NFSERR_GRACE ||
693 				    nd->nd_repstat == NFSERR_DELAY) &&
694 				    error == 0 && trycnt++ < 5);
695 				if (op->nfso_posixlock)
696 					do_unlock = 0;
697 			}
698 			nfscl_freelock(lop, 0);
699 		}
700 		/*
701 		 * Do a ReleaseLockOwner.
702 		 * The lock owner name nfsl_owner may be used by other opens for
703 		 * other files but the lock_owner4 name that nfsrpc_rellockown()
704 		 * puts on the wire has the file handle for this file appended
705 		 * to it, so it can be done now.
706 		 */
707 		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
708 		    lp->nfsl_open->nfso_fhlen, tcred, p);
709 	}
710 
711 	/*
712 	 * There could be other Opens for different files on the same
713 	 * OpenOwner, so locking is required.
714 	 */
715 	NFSLOCKCLSTATE();
716 	nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
717 	NFSUNLOCKCLSTATE();
718 	do {
719 		error = nfscl_tryclose(op, tcred, nmp, p);
720 		if (error == NFSERR_GRACE)
721 			(void) nfs_catnap(PZERO, error, "nfs_close");
722 	} while (error == NFSERR_GRACE);
723 	NFSLOCKCLSTATE();
724 	nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
725 
726 	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
727 		nfscl_freelockowner(lp, 0);
728 	nfscl_freeopen(op, 0);
729 	NFSUNLOCKCLSTATE();
730 	NFSFREECRED(tcred);
731 }
732 
733 /*
734  * The actual Close RPC.
735  */
736 APPLESTATIC int
737 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
738     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
739     int syscred)
740 {
741 	u_int32_t *tl;
742 	int error;
743 
744 	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
745 	    op->nfso_fhlen, NULL, NULL);
746 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
747 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
748 	if (NFSHASNFSV4N(nmp))
749 		*tl++ = 0;
750 	else
751 		*tl++ = op->nfso_stateid.seqid;
752 	*tl++ = op->nfso_stateid.other[0];
753 	*tl++ = op->nfso_stateid.other[1];
754 	*tl = op->nfso_stateid.other[2];
755 	if (syscred)
756 		nd->nd_flag |= ND_USEGSSNAME;
757 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
758 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
759 	if (error)
760 		return (error);
761 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
762 	if (nd->nd_repstat == 0)
763 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
764 	error = nd->nd_repstat;
765 	if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
766 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
767 nfsmout:
768 	mbuf_freem(nd->nd_mrep);
769 	return (error);
770 }
771 
772 /*
773  * V4 Open Confirm RPC.
774  */
775 APPLESTATIC int
776 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
777     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
778 {
779 	u_int32_t *tl;
780 	struct nfsrv_descript nfsd, *nd = &nfsd;
781 	struct nfsmount *nmp;
782 	int error;
783 
784 	nmp = VFSTONFS(vnode_mount(vp));
785 	if (NFSHASNFSV4N(nmp))
786 		return (0);		/* No confirmation for NFSv4.1. */
787 	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL);
788 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
789 	*tl++ = op->nfso_stateid.seqid;
790 	*tl++ = op->nfso_stateid.other[0];
791 	*tl++ = op->nfso_stateid.other[1];
792 	*tl++ = op->nfso_stateid.other[2];
793 	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
794 	error = nfscl_request(nd, vp, p, cred, NULL);
795 	if (error)
796 		return (error);
797 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
798 	if (!nd->nd_repstat) {
799 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
800 		op->nfso_stateid.seqid = *tl++;
801 		op->nfso_stateid.other[0] = *tl++;
802 		op->nfso_stateid.other[1] = *tl++;
803 		op->nfso_stateid.other[2] = *tl;
804 	}
805 	error = nd->nd_repstat;
806 	if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
807 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
808 nfsmout:
809 	mbuf_freem(nd->nd_mrep);
810 	return (error);
811 }
812 
813 /*
814  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
815  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
816  */
817 APPLESTATIC int
818 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
819     struct ucred *cred, NFSPROC_T *p)
820 {
821 	u_int32_t *tl;
822 	struct nfsrv_descript nfsd;
823 	struct nfsrv_descript *nd = &nfsd;
824 	nfsattrbit_t attrbits;
825 	u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
826 	u_short port;
827 	int error, isinet6 = 0, callblen;
828 	nfsquad_t confirm;
829 	u_int32_t lease;
830 	static u_int32_t rev = 0;
831 	struct nfsclds *dsp, *ndsp, *tdsp;
832 	struct in6_addr a6;
833 
834 	if (nfsboottime.tv_sec == 0)
835 		NFSSETBOOTTIME(nfsboottime);
836 	clp->nfsc_rev = rev++;
837 	if (NFSHASNFSV4N(nmp)) {
838 		error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq,
839 		    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp, cred, p);
840 		NFSCL_DEBUG(1, "aft exch=%d\n", error);
841 		if (error == 0) {
842 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
843 			    &nmp->nm_sockreq,
844 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
845 			if (error == 0) {
846 				NFSLOCKMNT(nmp);
847 				TAILQ_FOREACH_SAFE(tdsp, &nmp->nm_sess,
848 				    nfsclds_list, ndsp)
849 					nfscl_freenfsclds(tdsp);
850 				TAILQ_INIT(&nmp->nm_sess);
851 				TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
852 				    nfsclds_list);
853 				NFSUNLOCKMNT(nmp);
854 			} else
855 				nfscl_freenfsclds(dsp);
856 			NFSCL_DEBUG(1, "aft createsess=%d\n", error);
857 		}
858 		if (error == 0 && reclaim == 0) {
859 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
860 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
861 			if (error == NFSERR_COMPLETEALREADY ||
862 			    error == NFSERR_NOTSUPP)
863 				/* Ignore this error. */
864 				error = 0;
865 		}
866 		return (error);
867 	}
868 
869 	/*
870 	 * Allocate a single session structure for NFSv4.0, because some of
871 	 * the fields are used by NFSv4.0 although it doesn't do a session.
872 	 */
873 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
874 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
875 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
876 	NFSLOCKMNT(nmp);
877 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
878 	NFSUNLOCKMNT(nmp);
879 
880 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL);
881 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
882 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
883 	*tl = txdr_unsigned(clp->nfsc_rev);
884 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
885 
886 	/*
887 	 * set up the callback address
888 	 */
889 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
890 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
891 	callblen = strlen(nfsv4_callbackaddr);
892 	if (callblen == 0)
893 		cp = nfscl_getmyip(nmp, &a6, &isinet6);
894 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
895 	    (callblen > 0 || cp != NULL)) {
896 		port = htons(nfsv4_cbport);
897 		cp2 = (u_int8_t *)&port;
898 #ifdef INET6
899 		if ((callblen > 0 &&
900 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
901 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
902 
903 			(void) nfsm_strtom(nd, "tcp6", 4);
904 			if (callblen == 0) {
905 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
906 				ip6add = ip6buf;
907 			} else {
908 				ip6add = nfsv4_callbackaddr;
909 			}
910 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
911 			    ip6add, cp2[0], cp2[1]);
912 		} else
913 #endif
914 		{
915 			(void) nfsm_strtom(nd, "tcp", 3);
916 			if (callblen == 0)
917 				snprintf(addr, INET6_ADDRSTRLEN + 9,
918 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
919 				    cp[2], cp[3], cp2[0], cp2[1]);
920 			else
921 				snprintf(addr, INET6_ADDRSTRLEN + 9,
922 				    "%s.%d.%d", nfsv4_callbackaddr,
923 				    cp2[0], cp2[1]);
924 		}
925 		(void) nfsm_strtom(nd, addr, strlen(addr));
926 	} else {
927 		(void) nfsm_strtom(nd, "tcp", 3);
928 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
929 	}
930 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
931 	*tl = txdr_unsigned(clp->nfsc_cbident);
932 	nd->nd_flag |= ND_USEGSSNAME;
933 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
934 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
935 	if (error)
936 		return (error);
937 	if (nd->nd_repstat == 0) {
938 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
939 	    NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0] = *tl++;
940 	    NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1] = *tl++;
941 	    confirm.lval[0] = *tl++;
942 	    confirm.lval[1] = *tl;
943 	    mbuf_freem(nd->nd_mrep);
944 	    nd->nd_mrep = NULL;
945 
946 	    /*
947 	     * and confirm it.
948 	     */
949 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
950 		NULL);
951 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
952 	    *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
953 	    *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
954 	    *tl++ = confirm.lval[0];
955 	    *tl = confirm.lval[1];
956 	    nd->nd_flag |= ND_USEGSSNAME;
957 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
958 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
959 	    if (error)
960 		return (error);
961 	    mbuf_freem(nd->nd_mrep);
962 	    nd->nd_mrep = NULL;
963 	    if (nd->nd_repstat == 0) {
964 		nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, nmp->nm_fh,
965 		    nmp->nm_fhsize, NULL, NULL);
966 		NFSZERO_ATTRBIT(&attrbits);
967 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
968 		(void) nfsrv_putattrbit(nd, &attrbits);
969 		nd->nd_flag |= ND_USEGSSNAME;
970 		error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
971 		    cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
972 		if (error)
973 		    return (error);
974 		if (nd->nd_repstat == 0) {
975 		    error = nfsv4_loadattr(nd, NULL, NULL, NULL, NULL, 0, NULL,
976 			NULL, NULL, NULL, NULL, 0, NULL, &lease, NULL, p, cred);
977 		    if (error)
978 			goto nfsmout;
979 		    clp->nfsc_renew = NFSCL_RENEW(lease);
980 		    clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew;
981 		    clp->nfsc_clientidrev++;
982 		    if (clp->nfsc_clientidrev == 0)
983 			clp->nfsc_clientidrev++;
984 		}
985 	    }
986 	}
987 	error = nd->nd_repstat;
988 nfsmout:
989 	mbuf_freem(nd->nd_mrep);
990 	return (error);
991 }
992 
993 /*
994  * nfs getattr call.
995  */
996 APPLESTATIC int
997 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
998     struct nfsvattr *nap, void *stuff)
999 {
1000 	struct nfsrv_descript nfsd, *nd = &nfsd;
1001 	int error;
1002 	nfsattrbit_t attrbits;
1003 
1004 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1005 	if (nd->nd_flag & ND_NFSV4) {
1006 		NFSGETATTR_ATTRBIT(&attrbits);
1007 		(void) nfsrv_putattrbit(nd, &attrbits);
1008 	}
1009 	error = nfscl_request(nd, vp, p, cred, stuff);
1010 	if (error)
1011 		return (error);
1012 	if (!nd->nd_repstat)
1013 		error = nfsm_loadattr(nd, nap);
1014 	else
1015 		error = nd->nd_repstat;
1016 	mbuf_freem(nd->nd_mrep);
1017 	return (error);
1018 }
1019 
1020 /*
1021  * nfs getattr call with non-vnode arguemnts.
1022  */
1023 APPLESTATIC int
1024 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1025     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1026     uint32_t *leasep)
1027 {
1028 	struct nfsrv_descript nfsd, *nd = &nfsd;
1029 	int error, vers = NFS_VER2;
1030 	nfsattrbit_t attrbits;
1031 
1032 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL);
1033 	if (nd->nd_flag & ND_NFSV4) {
1034 		vers = NFS_VER4;
1035 		NFSGETATTR_ATTRBIT(&attrbits);
1036 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1037 		(void) nfsrv_putattrbit(nd, &attrbits);
1038 	} else if (nd->nd_flag & ND_NFSV3) {
1039 		vers = NFS_VER3;
1040 	}
1041 	if (syscred)
1042 		nd->nd_flag |= ND_USEGSSNAME;
1043 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1044 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1045 	if (error)
1046 		return (error);
1047 	if (nd->nd_repstat == 0) {
1048 		if ((nd->nd_flag & ND_NFSV4) != 0)
1049 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1050 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1051 			    NULL, NULL);
1052 		else
1053 			error = nfsm_loadattr(nd, nap);
1054 	} else
1055 		error = nd->nd_repstat;
1056 	mbuf_freem(nd->nd_mrep);
1057 	return (error);
1058 }
1059 
1060 /*
1061  * Do an nfs setattr operation.
1062  */
1063 APPLESTATIC int
1064 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1065     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1066     void *stuff)
1067 {
1068 	int error, expireret = 0, openerr, retrycnt;
1069 	u_int32_t clidrev = 0, mode;
1070 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1071 	struct nfsfh *nfhp;
1072 	nfsv4stateid_t stateid;
1073 	void *lckp;
1074 
1075 	if (nmp->nm_clp != NULL)
1076 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1077 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1078 		mode = NFSV4OPEN_ACCESSWRITE;
1079 	else
1080 		mode = NFSV4OPEN_ACCESSREAD;
1081 	retrycnt = 0;
1082 	do {
1083 		lckp = NULL;
1084 		openerr = 1;
1085 		if (NFSHASNFSV4(nmp)) {
1086 			nfhp = VTONFS(vp)->n_fhp;
1087 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1088 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1089 			if (error && vnode_vtype(vp) == VREG &&
1090 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1091 			     nfstest_openallsetattr)) {
1092 				/*
1093 				 * No Open stateid, so try and open the file
1094 				 * now.
1095 				 */
1096 				if (mode == NFSV4OPEN_ACCESSWRITE)
1097 					openerr = nfsrpc_open(vp, FWRITE, cred,
1098 					    p);
1099 				else
1100 					openerr = nfsrpc_open(vp, FREAD, cred,
1101 					    p);
1102 				if (!openerr)
1103 					(void) nfscl_getstateid(vp,
1104 					    nfhp->nfh_fh, nfhp->nfh_len,
1105 					    mode, 0, cred, p, &stateid, &lckp);
1106 			}
1107 		}
1108 		if (vap != NULL)
1109 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1110 			    rnap, attrflagp, stuff);
1111 		else
1112 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1113 			    stuff);
1114 		if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
1115 			nfscl_initiate_recovery(nmp->nm_clp);
1116 		if (lckp != NULL)
1117 			nfscl_lockderef(lckp);
1118 		if (!openerr)
1119 			(void) nfsrpc_close(vp, 0, p);
1120 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1121 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1122 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1123 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1124 		} else if ((error == NFSERR_EXPIRED ||
1125 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1126 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1127 		}
1128 		retrycnt++;
1129 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1130 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1131 	    error == NFSERR_BADSESSION ||
1132 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1133 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1134 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1135 	if (error && retrycnt >= 4)
1136 		error = EIO;
1137 	return (error);
1138 }
1139 
1140 static int
1141 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1142     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1143     struct nfsvattr *rnap, int *attrflagp, void *stuff)
1144 {
1145 	u_int32_t *tl;
1146 	struct nfsrv_descript nfsd, *nd = &nfsd;
1147 	int error;
1148 	nfsattrbit_t attrbits;
1149 
1150 	*attrflagp = 0;
1151 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1152 	if (nd->nd_flag & ND_NFSV4)
1153 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1154 	vap->va_type = vnode_vtype(vp);
1155 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1156 	if (nd->nd_flag & ND_NFSV3) {
1157 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1158 		*tl = newnfs_false;
1159 	} else if (nd->nd_flag & ND_NFSV4) {
1160 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1161 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1162 		NFSGETATTR_ATTRBIT(&attrbits);
1163 		(void) nfsrv_putattrbit(nd, &attrbits);
1164 	}
1165 	error = nfscl_request(nd, vp, p, cred, stuff);
1166 	if (error)
1167 		return (error);
1168 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1169 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1170 	if ((nd->nd_flag & ND_NFSV4) && !error)
1171 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1172 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1173 		error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1174 	mbuf_freem(nd->nd_mrep);
1175 	if (nd->nd_repstat && !error)
1176 		error = nd->nd_repstat;
1177 	return (error);
1178 }
1179 
1180 /*
1181  * nfs lookup rpc
1182  */
1183 APPLESTATIC int
1184 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1185     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1186     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff)
1187 {
1188 	u_int32_t *tl;
1189 	struct nfsrv_descript nfsd, *nd = &nfsd;
1190 	struct nfsmount *nmp;
1191 	struct nfsnode *np;
1192 	struct nfsfh *nfhp;
1193 	nfsattrbit_t attrbits;
1194 	int error = 0, lookupp = 0;
1195 
1196 	*attrflagp = 0;
1197 	*dattrflagp = 0;
1198 	if (vnode_vtype(dvp) != VDIR)
1199 		return (ENOTDIR);
1200 	nmp = VFSTONFS(vnode_mount(dvp));
1201 	if (len > NFS_MAXNAMLEN)
1202 		return (ENAMETOOLONG);
1203 	if (NFSHASNFSV4(nmp) && len == 1 &&
1204 		name[0] == '.') {
1205 		/*
1206 		 * Just return the current dir's fh.
1207 		 */
1208 		np = VTONFS(dvp);
1209 		MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) +
1210 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1211 		nfhp->nfh_len = np->n_fhp->nfh_len;
1212 		NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1213 		*nfhpp = nfhp;
1214 		return (0);
1215 	}
1216 	if (NFSHASNFSV4(nmp) && len == 2 &&
1217 		name[0] == '.' && name[1] == '.') {
1218 		lookupp = 1;
1219 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1220 	} else {
1221 		NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1222 		(void) nfsm_strtom(nd, name, len);
1223 	}
1224 	if (nd->nd_flag & ND_NFSV4) {
1225 		NFSGETATTR_ATTRBIT(&attrbits);
1226 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1227 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1228 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1229 		(void) nfsrv_putattrbit(nd, &attrbits);
1230 	}
1231 	error = nfscl_request(nd, dvp, p, cred, stuff);
1232 	if (error)
1233 		return (error);
1234 	if (nd->nd_repstat) {
1235 		/*
1236 		 * When an NFSv4 Lookupp returns ENOENT, it means that
1237 		 * the lookup is at the root of an fs, so return this dir.
1238 		 */
1239 		if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1240 		    np = VTONFS(dvp);
1241 		    MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) +
1242 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1243 		    nfhp->nfh_len = np->n_fhp->nfh_len;
1244 		    NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1245 		    *nfhpp = nfhp;
1246 		    mbuf_freem(nd->nd_mrep);
1247 		    return (0);
1248 		}
1249 		if (nd->nd_flag & ND_NFSV3)
1250 		    error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1251 		else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1252 		    ND_NFSV4) {
1253 			/* Load the directory attributes. */
1254 			error = nfsm_loadattr(nd, dnap);
1255 			if (error == 0)
1256 				*dattrflagp = 1;
1257 		}
1258 		goto nfsmout;
1259 	}
1260 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1261 		/* Load the directory attributes. */
1262 		error = nfsm_loadattr(nd, dnap);
1263 		if (error != 0)
1264 			goto nfsmout;
1265 		*dattrflagp = 1;
1266 		/* Skip over the Lookup and GetFH operation status values. */
1267 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1268 	}
1269 	error = nfsm_getfh(nd, nfhpp);
1270 	if (error)
1271 		goto nfsmout;
1272 
1273 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1274 	if ((nd->nd_flag & ND_NFSV3) && !error)
1275 		error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1276 nfsmout:
1277 	mbuf_freem(nd->nd_mrep);
1278 	if (!error && nd->nd_repstat)
1279 		error = nd->nd_repstat;
1280 	return (error);
1281 }
1282 
1283 /*
1284  * Do a readlink rpc.
1285  */
1286 APPLESTATIC int
1287 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1288     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1289 {
1290 	u_int32_t *tl;
1291 	struct nfsrv_descript nfsd, *nd = &nfsd;
1292 	struct nfsnode *np = VTONFS(vp);
1293 	nfsattrbit_t attrbits;
1294 	int error, len, cangetattr = 1;
1295 
1296 	*attrflagp = 0;
1297 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1298 	if (nd->nd_flag & ND_NFSV4) {
1299 		/*
1300 		 * And do a Getattr op.
1301 		 */
1302 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1303 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1304 		NFSGETATTR_ATTRBIT(&attrbits);
1305 		(void) nfsrv_putattrbit(nd, &attrbits);
1306 	}
1307 	error = nfscl_request(nd, vp, p, cred, stuff);
1308 	if (error)
1309 		return (error);
1310 	if (nd->nd_flag & ND_NFSV3)
1311 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1312 	if (!nd->nd_repstat && !error) {
1313 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1314 		/*
1315 		 * This seems weird to me, but must have been added to
1316 		 * FreeBSD for some reason. The only thing I can think of
1317 		 * is that there was/is some server that replies with
1318 		 * more link data than it should?
1319 		 */
1320 		if (len == NFS_MAXPATHLEN) {
1321 			NFSLOCKNODE(np);
1322 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1323 				len = np->n_size;
1324 				cangetattr = 0;
1325 			}
1326 			NFSUNLOCKNODE(np);
1327 		}
1328 		error = nfsm_mbufuio(nd, uiop, len);
1329 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1330 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1331 	}
1332 	if (nd->nd_repstat && !error)
1333 		error = nd->nd_repstat;
1334 nfsmout:
1335 	mbuf_freem(nd->nd_mrep);
1336 	return (error);
1337 }
1338 
1339 /*
1340  * Read operation.
1341  */
1342 APPLESTATIC int
1343 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1344     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1345 {
1346 	int error, expireret = 0, retrycnt;
1347 	u_int32_t clidrev = 0;
1348 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1349 	struct nfsnode *np = VTONFS(vp);
1350 	struct ucred *newcred;
1351 	struct nfsfh *nfhp = NULL;
1352 	nfsv4stateid_t stateid;
1353 	void *lckp;
1354 
1355 	if (nmp->nm_clp != NULL)
1356 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1357 	newcred = cred;
1358 	if (NFSHASNFSV4(nmp)) {
1359 		nfhp = np->n_fhp;
1360 		newcred = NFSNEWCRED(cred);
1361 	}
1362 	retrycnt = 0;
1363 	do {
1364 		lckp = NULL;
1365 		if (NFSHASNFSV4(nmp))
1366 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1367 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1368 			    &lckp);
1369 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1370 		    attrflagp, stuff);
1371 		if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
1372 			nfscl_initiate_recovery(nmp->nm_clp);
1373 		if (lckp != NULL)
1374 			nfscl_lockderef(lckp);
1375 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1376 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1377 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1378 			(void) nfs_catnap(PZERO, error, "nfs_read");
1379 		} else if ((error == NFSERR_EXPIRED ||
1380 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1381 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1382 		}
1383 		retrycnt++;
1384 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1385 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1386 	    error == NFSERR_BADSESSION ||
1387 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1388 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1389 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1390 	if (error && retrycnt >= 4)
1391 		error = EIO;
1392 	if (NFSHASNFSV4(nmp))
1393 		NFSFREECRED(newcred);
1394 	return (error);
1395 }
1396 
1397 /*
1398  * The actual read RPC.
1399  */
1400 static int
1401 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1402     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1403     int *attrflagp, void *stuff)
1404 {
1405 	u_int32_t *tl;
1406 	int error = 0, len, retlen, tsiz, eof = 0;
1407 	struct nfsrv_descript nfsd;
1408 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1409 	struct nfsrv_descript *nd = &nfsd;
1410 	int rsize;
1411 	off_t tmp_off;
1412 
1413 	*attrflagp = 0;
1414 	tsiz = uio_uio_resid(uiop);
1415 	tmp_off = uiop->uio_offset + tsiz;
1416 	NFSLOCKMNT(nmp);
1417 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1418 		NFSUNLOCKMNT(nmp);
1419 		return (EFBIG);
1420 	}
1421 	rsize = nmp->nm_rsize;
1422 	NFSUNLOCKMNT(nmp);
1423 	nd->nd_mrep = NULL;
1424 	while (tsiz > 0) {
1425 		*attrflagp = 0;
1426 		len = (tsiz > rsize) ? rsize : tsiz;
1427 		NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1428 		if (nd->nd_flag & ND_NFSV4)
1429 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1430 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1431 		if (nd->nd_flag & ND_NFSV2) {
1432 			*tl++ = txdr_unsigned(uiop->uio_offset);
1433 			*tl++ = txdr_unsigned(len);
1434 			*tl = 0;
1435 		} else {
1436 			txdr_hyper(uiop->uio_offset, tl);
1437 			*(tl + 2) = txdr_unsigned(len);
1438 		}
1439 		/*
1440 		 * Since I can't do a Getattr for NFSv4 for Write, there
1441 		 * doesn't seem any point in doing one here, either.
1442 		 * (See the comment in nfsrpc_writerpc() for more info.)
1443 		 */
1444 		error = nfscl_request(nd, vp, p, cred, stuff);
1445 		if (error)
1446 			return (error);
1447 		if (nd->nd_flag & ND_NFSV3) {
1448 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1449 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1450 			error = nfsm_loadattr(nd, nap);
1451 			if (!error)
1452 				*attrflagp = 1;
1453 		}
1454 		if (nd->nd_repstat || error) {
1455 			if (!error)
1456 				error = nd->nd_repstat;
1457 			goto nfsmout;
1458 		}
1459 		if (nd->nd_flag & ND_NFSV3) {
1460 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1461 			eof = fxdr_unsigned(int, *(tl + 1));
1462 		} else if (nd->nd_flag & ND_NFSV4) {
1463 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1464 			eof = fxdr_unsigned(int, *tl);
1465 		}
1466 		NFSM_STRSIZ(retlen, len);
1467 		error = nfsm_mbufuio(nd, uiop, retlen);
1468 		if (error)
1469 			goto nfsmout;
1470 		mbuf_freem(nd->nd_mrep);
1471 		nd->nd_mrep = NULL;
1472 		tsiz -= retlen;
1473 		if (!(nd->nd_flag & ND_NFSV2)) {
1474 			if (eof || retlen == 0)
1475 				tsiz = 0;
1476 		} else if (retlen < len)
1477 			tsiz = 0;
1478 	}
1479 	return (0);
1480 nfsmout:
1481 	if (nd->nd_mrep != NULL)
1482 		mbuf_freem(nd->nd_mrep);
1483 	return (error);
1484 }
1485 
1486 /*
1487  * nfs write operation
1488  * When called_from_strategy != 0, it should return EIO for an error that
1489  * indicates recovery is in progress, so that the buffer will be left
1490  * dirty and be written back to the server later. If it loops around,
1491  * the recovery thread could get stuck waiting for the buffer and recovery
1492  * will then deadlock.
1493  */
1494 APPLESTATIC int
1495 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1496     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1497     void *stuff, int called_from_strategy)
1498 {
1499 	int error, expireret = 0, retrycnt, nostateid;
1500 	u_int32_t clidrev = 0;
1501 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1502 	struct nfsnode *np = VTONFS(vp);
1503 	struct ucred *newcred;
1504 	struct nfsfh *nfhp = NULL;
1505 	nfsv4stateid_t stateid;
1506 	void *lckp;
1507 
1508 	*must_commit = 0;
1509 	if (nmp->nm_clp != NULL)
1510 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1511 	newcred = cred;
1512 	if (NFSHASNFSV4(nmp)) {
1513 		newcred = NFSNEWCRED(cred);
1514 		nfhp = np->n_fhp;
1515 	}
1516 	retrycnt = 0;
1517 	do {
1518 		lckp = NULL;
1519 		nostateid = 0;
1520 		if (NFSHASNFSV4(nmp)) {
1521 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1522 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1523 			    &lckp);
1524 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1525 			    stateid.other[2] == 0) {
1526 				nostateid = 1;
1527 				NFSCL_DEBUG(1, "stateid0 in write\n");
1528 			}
1529 		}
1530 
1531 		/*
1532 		 * If there is no stateid for NFSv4, it means this is an
1533 		 * extraneous write after close. Basically a poorly
1534 		 * implemented buffer cache. Just don't do the write.
1535 		 */
1536 		if (nostateid)
1537 			error = 0;
1538 		else
1539 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1540 			    newcred, &stateid, p, nap, attrflagp, stuff);
1541 		if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
1542 			nfscl_initiate_recovery(nmp->nm_clp);
1543 		if (lckp != NULL)
1544 			nfscl_lockderef(lckp);
1545 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1546 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1547 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1548 			(void) nfs_catnap(PZERO, error, "nfs_write");
1549 		} else if ((error == NFSERR_EXPIRED ||
1550 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1551 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1552 		}
1553 		retrycnt++;
1554 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1555 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1556 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1557 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1558 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1559 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1560 	if (error != 0 && (retrycnt >= 4 ||
1561 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1562 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1563 		error = EIO;
1564 	if (NFSHASNFSV4(nmp))
1565 		NFSFREECRED(newcred);
1566 	return (error);
1567 }
1568 
1569 /*
1570  * The actual write RPC.
1571  */
1572 static int
1573 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1574     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1575     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1576 {
1577 	u_int32_t *tl;
1578 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1579 	struct nfsnode *np = VTONFS(vp);
1580 	int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1581 	int wccflag = 0, wsize;
1582 	int32_t backup;
1583 	struct nfsrv_descript nfsd;
1584 	struct nfsrv_descript *nd = &nfsd;
1585 	nfsattrbit_t attrbits;
1586 	off_t tmp_off;
1587 
1588 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1589 	*attrflagp = 0;
1590 	tsiz = uio_uio_resid(uiop);
1591 	tmp_off = uiop->uio_offset + tsiz;
1592 	NFSLOCKMNT(nmp);
1593 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1594 		NFSUNLOCKMNT(nmp);
1595 		return (EFBIG);
1596 	}
1597 	wsize = nmp->nm_wsize;
1598 	NFSUNLOCKMNT(nmp);
1599 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
1600 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
1601 	while (tsiz > 0) {
1602 		*attrflagp = 0;
1603 		len = (tsiz > wsize) ? wsize : tsiz;
1604 		NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1605 		if (nd->nd_flag & ND_NFSV4) {
1606 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1607 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1608 			txdr_hyper(uiop->uio_offset, tl);
1609 			tl += 2;
1610 			*tl++ = txdr_unsigned(*iomode);
1611 			*tl = txdr_unsigned(len);
1612 		} else if (nd->nd_flag & ND_NFSV3) {
1613 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1614 			txdr_hyper(uiop->uio_offset, tl);
1615 			tl += 2;
1616 			*tl++ = txdr_unsigned(len);
1617 			*tl++ = txdr_unsigned(*iomode);
1618 			*tl = txdr_unsigned(len);
1619 		} else {
1620 			u_int32_t x;
1621 
1622 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1623 			/*
1624 			 * Not sure why someone changed this, since the
1625 			 * RFC clearly states that "beginoffset" and
1626 			 * "totalcount" are ignored, but it wouldn't
1627 			 * surprise me if there's a busted server out there.
1628 			 */
1629 			/* Set both "begin" and "current" to non-garbage. */
1630 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1631 			*tl++ = x;      /* "begin offset" */
1632 			*tl++ = x;      /* "current offset" */
1633 			x = txdr_unsigned(len);
1634 			*tl++ = x;      /* total to this offset */
1635 			*tl = x;        /* size of this write */
1636 
1637 		}
1638 		nfsm_uiombuf(nd, uiop, len);
1639 		/*
1640 		 * Although it is tempting to do a normal Getattr Op in the
1641 		 * NFSv4 compound, the result can be a nearly hung client
1642 		 * system if the Getattr asks for Owner and/or OwnerGroup.
1643 		 * It occurs when the client can't map either the Owner or
1644 		 * Owner_group name in the Getattr reply to a uid/gid. When
1645 		 * there is a cache miss, the kernel does an upcall to the
1646 		 * nfsuserd. Then, it can try and read the local /etc/passwd
1647 		 * or /etc/group file. It can then block in getnewbuf(),
1648 		 * waiting for dirty writes to be pushed to the NFS server.
1649 		 * The only reason this doesn't result in a complete
1650 		 * deadlock, is that the upcall times out and allows
1651 		 * the write to complete. However, progress is so slow
1652 		 * that it might just as well be deadlocked.
1653 		 * As such, we get the rest of the attributes, but not
1654 		 * Owner or Owner_group.
1655 		 * nb: nfscl_loadattrcache() needs to be told that these
1656 		 *     partial attributes from a write rpc are being
1657 		 *     passed in, via a argument flag.
1658 		 */
1659 		if (nd->nd_flag & ND_NFSV4) {
1660 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
1661 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1662 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
1663 			(void) nfsrv_putattrbit(nd, &attrbits);
1664 		}
1665 		error = nfscl_request(nd, vp, p, cred, stuff);
1666 		if (error)
1667 			return (error);
1668 		if (nd->nd_repstat) {
1669 			/*
1670 			 * In case the rpc gets retried, roll
1671 			 * the uio fileds changed by nfsm_uiombuf()
1672 			 * back.
1673 			 */
1674 			uiop->uio_offset -= len;
1675 			uio_uio_resid_add(uiop, len);
1676 			uio_iov_base_add(uiop, -len);
1677 			uio_iov_len_add(uiop, len);
1678 		}
1679 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1680 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1681 			    &wccflag, stuff);
1682 			if (error)
1683 				goto nfsmout;
1684 		}
1685 		if (!nd->nd_repstat) {
1686 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1687 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1688 					+ NFSX_VERF);
1689 				rlen = fxdr_unsigned(int, *tl++);
1690 				if (rlen == 0) {
1691 					error = NFSERR_IO;
1692 					goto nfsmout;
1693 				} else if (rlen < len) {
1694 					backup = len - rlen;
1695 					uio_iov_base_add(uiop, -(backup));
1696 					uio_iov_len_add(uiop, backup);
1697 					uiop->uio_offset -= backup;
1698 					uio_uio_resid_add(uiop, backup);
1699 					len = rlen;
1700 				}
1701 				commit = fxdr_unsigned(int, *tl++);
1702 
1703 				/*
1704 				 * Return the lowest committment level
1705 				 * obtained by any of the RPCs.
1706 				 */
1707 				if (committed == NFSWRITE_FILESYNC)
1708 					committed = commit;
1709 				else if (committed == NFSWRITE_DATASYNC &&
1710 					commit == NFSWRITE_UNSTABLE)
1711 					committed = commit;
1712 				NFSLOCKMNT(nmp);
1713 				if (!NFSHASWRITEVERF(nmp)) {
1714 					NFSBCOPY((caddr_t)tl,
1715 					    (caddr_t)&nmp->nm_verf[0],
1716 					    NFSX_VERF);
1717 					NFSSETWRITEVERF(nmp);
1718 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
1719 				    NFSX_VERF)) {
1720 					*must_commit = 1;
1721 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1722 				}
1723 				NFSUNLOCKMNT(nmp);
1724 			}
1725 			if (nd->nd_flag & ND_NFSV4)
1726 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1727 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1728 				error = nfsm_loadattr(nd, nap);
1729 				if (!error)
1730 					*attrflagp = NFS_LATTR_NOSHRINK;
1731 			}
1732 		} else {
1733 			error = nd->nd_repstat;
1734 		}
1735 		if (error)
1736 			goto nfsmout;
1737 		NFSWRITERPC_SETTIME(wccflag, np, (nd->nd_flag & ND_NFSV4));
1738 		mbuf_freem(nd->nd_mrep);
1739 		nd->nd_mrep = NULL;
1740 		tsiz -= len;
1741 	}
1742 nfsmout:
1743 	if (nd->nd_mrep != NULL)
1744 		mbuf_freem(nd->nd_mrep);
1745 	*iomode = committed;
1746 	if (nd->nd_repstat && !error)
1747 		error = nd->nd_repstat;
1748 	return (error);
1749 }
1750 
1751 /*
1752  * nfs mknod rpc
1753  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1754  * mode set to specify the file type and the size field for rdev.
1755  */
1756 APPLESTATIC int
1757 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1758     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1759     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1760     int *attrflagp, int *dattrflagp, void *dstuff)
1761 {
1762 	u_int32_t *tl;
1763 	int error = 0;
1764 	struct nfsrv_descript nfsd, *nd = &nfsd;
1765 	nfsattrbit_t attrbits;
1766 
1767 	*nfhpp = NULL;
1768 	*attrflagp = 0;
1769 	*dattrflagp = 0;
1770 	if (namelen > NFS_MAXNAMLEN)
1771 		return (ENAMETOOLONG);
1772 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1773 	if (nd->nd_flag & ND_NFSV4) {
1774 		if (vtyp == VBLK || vtyp == VCHR) {
1775 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1776 			*tl++ = vtonfsv34_type(vtyp);
1777 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
1778 			*tl = txdr_unsigned(NFSMINOR(rdev));
1779 		} else {
1780 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1781 			*tl = vtonfsv34_type(vtyp);
1782 		}
1783 	}
1784 	(void) nfsm_strtom(nd, name, namelen);
1785 	if (nd->nd_flag & ND_NFSV3) {
1786 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1787 		*tl = vtonfsv34_type(vtyp);
1788 	}
1789 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1790 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
1791 	if ((nd->nd_flag & ND_NFSV3) &&
1792 	    (vtyp == VCHR || vtyp == VBLK)) {
1793 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1794 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
1795 		*tl = txdr_unsigned(NFSMINOR(rdev));
1796 	}
1797 	if (nd->nd_flag & ND_NFSV4) {
1798 		NFSGETATTR_ATTRBIT(&attrbits);
1799 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1800 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1801 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1802 		(void) nfsrv_putattrbit(nd, &attrbits);
1803 	}
1804 	if (nd->nd_flag & ND_NFSV2)
1805 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
1806 	error = nfscl_request(nd, dvp, p, cred, dstuff);
1807 	if (error)
1808 		return (error);
1809 	if (nd->nd_flag & ND_NFSV4)
1810 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1811 	if (!nd->nd_repstat) {
1812 		if (nd->nd_flag & ND_NFSV4) {
1813 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1814 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1815 			if (error)
1816 				goto nfsmout;
1817 		}
1818 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
1819 		if (error)
1820 			goto nfsmout;
1821 	}
1822 	if (nd->nd_flag & ND_NFSV3)
1823 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1824 	if (!error && nd->nd_repstat)
1825 		error = nd->nd_repstat;
1826 nfsmout:
1827 	mbuf_freem(nd->nd_mrep);
1828 	return (error);
1829 }
1830 
1831 /*
1832  * nfs file create call
1833  * Mostly just call the approriate routine. (I separated out v4, so that
1834  * error recovery wouldn't be as difficult.)
1835  */
1836 APPLESTATIC int
1837 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1838     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
1839     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1840     int *attrflagp, int *dattrflagp, void *dstuff)
1841 {
1842 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
1843 	struct nfsclowner *owp;
1844 	struct nfscldeleg *dp;
1845 	struct nfsmount *nmp = VFSTONFS(vnode_mount(dvp));
1846 	u_int32_t clidrev;
1847 
1848 	if (NFSHASNFSV4(nmp)) {
1849 	    retrycnt = 0;
1850 	    do {
1851 		dp = NULL;
1852 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
1853 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
1854 		    NULL, 1);
1855 		if (error)
1856 			return (error);
1857 		if (nmp->nm_clp != NULL)
1858 			clidrev = nmp->nm_clp->nfsc_clientidrev;
1859 		else
1860 			clidrev = 0;
1861 		error = nfsrpc_createv4(dvp, name, namelen, vap, cverf, fmode,
1862 		  owp, &dp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
1863 		  dstuff, &unlocked);
1864 		/*
1865 		 * There is no need to invalidate cached attributes here,
1866 		 * since new post-delegation issue attributes are always
1867 		 * returned by nfsrpc_createv4() and these will update the
1868 		 * attribute cache.
1869 		 */
1870 		if (dp != NULL)
1871 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
1872 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
1873 		nfscl_ownerrelease(owp, error, newone, unlocked);
1874 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
1875 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1876 		    error == NFSERR_BADSESSION) {
1877 			(void) nfs_catnap(PZERO, error, "nfs_open");
1878 		} else if ((error == NFSERR_EXPIRED ||
1879 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1880 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1881 			retrycnt++;
1882 		}
1883 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
1884 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1885 		error == NFSERR_BADSESSION ||
1886 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1887 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
1888 	    if (error && retrycnt >= 4)
1889 		    error = EIO;
1890 	} else {
1891 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
1892 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
1893 		    dstuff);
1894 	}
1895 	return (error);
1896 }
1897 
1898 /*
1899  * The create rpc for v2 and 3.
1900  */
1901 static int
1902 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1903     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
1904     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1905     int *attrflagp, int *dattrflagp, void *dstuff)
1906 {
1907 	u_int32_t *tl;
1908 	int error = 0;
1909 	struct nfsrv_descript nfsd, *nd = &nfsd;
1910 
1911 	*nfhpp = NULL;
1912 	*attrflagp = 0;
1913 	*dattrflagp = 0;
1914 	if (namelen > NFS_MAXNAMLEN)
1915 		return (ENAMETOOLONG);
1916 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
1917 	(void) nfsm_strtom(nd, name, namelen);
1918 	if (nd->nd_flag & ND_NFSV3) {
1919 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1920 		if (fmode & O_EXCL) {
1921 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
1922 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
1923 			*tl++ = cverf.lval[0];
1924 			*tl = cverf.lval[1];
1925 		} else {
1926 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
1927 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
1928 		}
1929 	} else {
1930 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
1931 	}
1932 	error = nfscl_request(nd, dvp, p, cred, dstuff);
1933 	if (error)
1934 		return (error);
1935 	if (nd->nd_repstat == 0) {
1936 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
1937 		if (error)
1938 			goto nfsmout;
1939 	}
1940 	if (nd->nd_flag & ND_NFSV3)
1941 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1942 	if (nd->nd_repstat != 0 && error == 0)
1943 		error = nd->nd_repstat;
1944 nfsmout:
1945 	mbuf_freem(nd->nd_mrep);
1946 	return (error);
1947 }
1948 
1949 static int
1950 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1951     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
1952     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
1953     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
1954     int *dattrflagp, void *dstuff, int *unlockedp)
1955 {
1956 	u_int32_t *tl;
1957 	int error = 0, deleg, newone, ret, acesize, limitby;
1958 	struct nfsrv_descript nfsd, *nd = &nfsd;
1959 	struct nfsclopen *op;
1960 	struct nfscldeleg *dp = NULL;
1961 	struct nfsnode *np;
1962 	struct nfsfh *nfhp;
1963 	nfsattrbit_t attrbits;
1964 	nfsv4stateid_t stateid;
1965 	u_int32_t rflags;
1966 	struct nfsmount *nmp;
1967 
1968 	nmp = VFSTONFS(dvp->v_mount);
1969 	np = VTONFS(dvp);
1970 	*unlockedp = 0;
1971 	*nfhpp = NULL;
1972 	*dpp = NULL;
1973 	*attrflagp = 0;
1974 	*dattrflagp = 0;
1975 	if (namelen > NFS_MAXNAMLEN)
1976 		return (ENAMETOOLONG);
1977 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
1978 	/*
1979 	 * For V4, this is actually an Open op.
1980 	 */
1981 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1982 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
1983 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
1984 	    NFSV4OPEN_ACCESSREAD);
1985 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
1986 	*tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
1987 	*tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
1988 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
1989 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1990 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
1991 	if (fmode & O_EXCL) {
1992 		if (NFSHASNFSV4N(nmp)) {
1993 			if (NFSHASSESSPERSIST(nmp)) {
1994 				/* Use GUARDED for persistent sessions. */
1995 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
1996 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
1997 			} else {
1998 				/* Otherwise, use EXCLUSIVE4_1. */
1999 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2000 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2001 				*tl++ = cverf.lval[0];
2002 				*tl = cverf.lval[1];
2003 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2004 			}
2005 		} else {
2006 			/* NFSv4.0 */
2007 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2008 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2009 			*tl++ = cverf.lval[0];
2010 			*tl = cverf.lval[1];
2011 		}
2012 	} else {
2013 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2014 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2015 	}
2016 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2017 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2018 	(void) nfsm_strtom(nd, name, namelen);
2019 	/* Get the new file's handle and attributes. */
2020 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2021 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2022 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2023 	NFSGETATTR_ATTRBIT(&attrbits);
2024 	(void) nfsrv_putattrbit(nd, &attrbits);
2025 	/* Get the directory's post-op attributes. */
2026 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2027 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2028 	(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2029 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2030 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2031 	(void) nfsrv_putattrbit(nd, &attrbits);
2032 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2033 	if (error)
2034 		return (error);
2035 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2036 	if (nd->nd_repstat == 0) {
2037 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2038 		    6 * NFSX_UNSIGNED);
2039 		stateid.seqid = *tl++;
2040 		stateid.other[0] = *tl++;
2041 		stateid.other[1] = *tl++;
2042 		stateid.other[2] = *tl;
2043 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2044 		(void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2045 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2046 		deleg = fxdr_unsigned(int, *tl);
2047 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2048 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2049 			if (!(owp->nfsow_clp->nfsc_flags &
2050 			      NFSCLFLAGS_FIRSTDELEG))
2051 				owp->nfsow_clp->nfsc_flags |=
2052 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2053 			MALLOC(dp, struct nfscldeleg *,
2054 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2055 			    M_NFSCLDELEG, M_WAITOK);
2056 			LIST_INIT(&dp->nfsdl_owner);
2057 			LIST_INIT(&dp->nfsdl_lock);
2058 			dp->nfsdl_clp = owp->nfsow_clp;
2059 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2060 			nfscl_lockinit(&dp->nfsdl_rwlock);
2061 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2062 			    NFSX_UNSIGNED);
2063 			dp->nfsdl_stateid.seqid = *tl++;
2064 			dp->nfsdl_stateid.other[0] = *tl++;
2065 			dp->nfsdl_stateid.other[1] = *tl++;
2066 			dp->nfsdl_stateid.other[2] = *tl++;
2067 			ret = fxdr_unsigned(int, *tl);
2068 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2069 				dp->nfsdl_flags = NFSCLDL_WRITE;
2070 				/*
2071 				 * Indicates how much the file can grow.
2072 				 */
2073 				NFSM_DISSECT(tl, u_int32_t *,
2074 				    3 * NFSX_UNSIGNED);
2075 				limitby = fxdr_unsigned(int, *tl++);
2076 				switch (limitby) {
2077 				case NFSV4OPEN_LIMITSIZE:
2078 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2079 					break;
2080 				case NFSV4OPEN_LIMITBLOCKS:
2081 					dp->nfsdl_sizelimit =
2082 					    fxdr_unsigned(u_int64_t, *tl++);
2083 					dp->nfsdl_sizelimit *=
2084 					    fxdr_unsigned(u_int64_t, *tl);
2085 					break;
2086 				default:
2087 					error = NFSERR_BADXDR;
2088 					goto nfsmout;
2089 				};
2090 			} else {
2091 				dp->nfsdl_flags = NFSCLDL_READ;
2092 			}
2093 			if (ret)
2094 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2095 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2096 			    &acesize, p);
2097 			if (error)
2098 				goto nfsmout;
2099 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2100 			error = NFSERR_BADXDR;
2101 			goto nfsmout;
2102 		}
2103 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2104 		if (error)
2105 			goto nfsmout;
2106 		/* Get rid of the PutFH and Getattr status values. */
2107 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2108 		/* Load the directory attributes. */
2109 		error = nfsm_loadattr(nd, dnap);
2110 		if (error)
2111 			goto nfsmout;
2112 		*dattrflagp = 1;
2113 		if (dp != NULL && *attrflagp) {
2114 			dp->nfsdl_change = nnap->na_filerev;
2115 			dp->nfsdl_modtime = nnap->na_mtime;
2116 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2117 		}
2118 		/*
2119 		 * We can now complete the Open state.
2120 		 */
2121 		nfhp = *nfhpp;
2122 		if (dp != NULL) {
2123 			dp->nfsdl_fhlen = nfhp->nfh_len;
2124 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2125 		}
2126 		/*
2127 		 * Get an Open structure that will be
2128 		 * attached to the OpenOwner, acquired already.
2129 		 */
2130 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2131 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2132 		    cred, p, NULL, &op, &newone, NULL, 0);
2133 		if (error)
2134 			goto nfsmout;
2135 		op->nfso_stateid = stateid;
2136 		newnfs_copyincred(cred, &op->nfso_cred);
2137 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2138 		    do {
2139 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2140 			    nfhp->nfh_len, op, cred, p);
2141 			if (ret == NFSERR_DELAY)
2142 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2143 		    } while (ret == NFSERR_DELAY);
2144 		    error = ret;
2145 		}
2146 
2147 		/*
2148 		 * If the server is handing out delegations, but we didn't
2149 		 * get one because an OpenConfirm was required, try the
2150 		 * Open again, to get a delegation. This is a harmless no-op,
2151 		 * from a server's point of view.
2152 		 */
2153 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2154 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2155 		    !error && dp == NULL) {
2156 		    do {
2157 			ret = nfsrpc_openrpc(VFSTONFS(vnode_mount(dvp)), dvp,
2158 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2159 			    nfhp->nfh_fh, nfhp->nfh_len,
2160 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2161 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2162 			if (ret == NFSERR_DELAY)
2163 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2164 		    } while (ret == NFSERR_DELAY);
2165 		    if (ret) {
2166 			if (dp != NULL) {
2167 				FREE((caddr_t)dp, M_NFSCLDELEG);
2168 				dp = NULL;
2169 			}
2170 			if (ret == NFSERR_STALECLIENTID ||
2171 			    ret == NFSERR_STALEDONTRECOVER ||
2172 			    ret == NFSERR_BADSESSION)
2173 				error = ret;
2174 		    }
2175 		}
2176 		nfscl_openrelease(op, error, newone);
2177 		*unlockedp = 1;
2178 	}
2179 	if (nd->nd_repstat != 0 && error == 0)
2180 		error = nd->nd_repstat;
2181 	if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
2182 		nfscl_initiate_recovery(owp->nfsow_clp);
2183 nfsmout:
2184 	if (!error)
2185 		*dpp = dp;
2186 	else if (dp != NULL)
2187 		FREE((caddr_t)dp, M_NFSCLDELEG);
2188 	mbuf_freem(nd->nd_mrep);
2189 	return (error);
2190 }
2191 
2192 /*
2193  * Nfs remove rpc
2194  */
2195 APPLESTATIC int
2196 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2197     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2198     void *dstuff)
2199 {
2200 	u_int32_t *tl;
2201 	struct nfsrv_descript nfsd, *nd = &nfsd;
2202 	struct nfsnode *np;
2203 	struct nfsmount *nmp;
2204 	nfsv4stateid_t dstateid;
2205 	int error, ret = 0, i;
2206 
2207 	*dattrflagp = 0;
2208 	if (namelen > NFS_MAXNAMLEN)
2209 		return (ENAMETOOLONG);
2210 	nmp = VFSTONFS(vnode_mount(dvp));
2211 tryagain:
2212 	if (NFSHASNFSV4(nmp) && ret == 0) {
2213 		ret = nfscl_removedeleg(vp, p, &dstateid);
2214 		if (ret == 1) {
2215 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2216 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2217 			    NFSX_UNSIGNED);
2218 			if (NFSHASNFSV4N(nmp))
2219 				*tl++ = 0;
2220 			else
2221 				*tl++ = dstateid.seqid;
2222 			*tl++ = dstateid.other[0];
2223 			*tl++ = dstateid.other[1];
2224 			*tl++ = dstateid.other[2];
2225 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2226 			np = VTONFS(dvp);
2227 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2228 			    np->n_fhp->nfh_len, 0);
2229 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2230 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2231 		}
2232 	} else {
2233 		ret = 0;
2234 	}
2235 	if (ret == 0)
2236 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2237 	(void) nfsm_strtom(nd, name, namelen);
2238 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2239 	if (error)
2240 		return (error);
2241 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2242 		/* For NFSv4, parse out any Delereturn replies. */
2243 		if (ret > 0 && nd->nd_repstat != 0 &&
2244 		    (nd->nd_flag & ND_NOMOREDATA)) {
2245 			/*
2246 			 * If the Delegreturn failed, try again without
2247 			 * it. The server will Recall, as required.
2248 			 */
2249 			mbuf_freem(nd->nd_mrep);
2250 			goto tryagain;
2251 		}
2252 		for (i = 0; i < (ret * 2); i++) {
2253 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2254 			    ND_NFSV4) {
2255 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2256 			    if (*(tl + 1))
2257 				nd->nd_flag |= ND_NOMOREDATA;
2258 			}
2259 		}
2260 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2261 	}
2262 	if (nd->nd_repstat && !error)
2263 		error = nd->nd_repstat;
2264 nfsmout:
2265 	mbuf_freem(nd->nd_mrep);
2266 	return (error);
2267 }
2268 
2269 /*
2270  * Do an nfs rename rpc.
2271  */
2272 APPLESTATIC int
2273 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2274     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2275     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2276     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2277 {
2278 	u_int32_t *tl;
2279 	struct nfsrv_descript nfsd, *nd = &nfsd;
2280 	struct nfsmount *nmp;
2281 	struct nfsnode *np;
2282 	nfsattrbit_t attrbits;
2283 	nfsv4stateid_t fdstateid, tdstateid;
2284 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2285 
2286 	*fattrflagp = 0;
2287 	*tattrflagp = 0;
2288 	nmp = VFSTONFS(vnode_mount(fdvp));
2289 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2290 		return (ENAMETOOLONG);
2291 tryagain:
2292 	if (NFSHASNFSV4(nmp) && ret == 0) {
2293 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2294 		    &tdstateid, &gottd, p);
2295 		if (gotfd && gottd) {
2296 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2297 		} else if (gotfd) {
2298 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2299 		} else if (gottd) {
2300 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2301 		}
2302 		if (gotfd) {
2303 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2304 			if (NFSHASNFSV4N(nmp))
2305 				*tl++ = 0;
2306 			else
2307 				*tl++ = fdstateid.seqid;
2308 			*tl++ = fdstateid.other[0];
2309 			*tl++ = fdstateid.other[1];
2310 			*tl = fdstateid.other[2];
2311 			if (gottd) {
2312 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2313 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2314 				np = VTONFS(tvp);
2315 				(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2316 				    np->n_fhp->nfh_len, 0);
2317 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2318 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2319 			}
2320 		}
2321 		if (gottd) {
2322 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2323 			if (NFSHASNFSV4N(nmp))
2324 				*tl++ = 0;
2325 			else
2326 				*tl++ = tdstateid.seqid;
2327 			*tl++ = tdstateid.other[0];
2328 			*tl++ = tdstateid.other[1];
2329 			*tl = tdstateid.other[2];
2330 		}
2331 		if (ret > 0) {
2332 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2333 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2334 			np = VTONFS(fdvp);
2335 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2336 			    np->n_fhp->nfh_len, 0);
2337 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2338 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2339 		}
2340 	} else {
2341 		ret = 0;
2342 	}
2343 	if (ret == 0)
2344 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2345 	if (nd->nd_flag & ND_NFSV4) {
2346 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2347 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2348 		NFSWCCATTR_ATTRBIT(&attrbits);
2349 		(void) nfsrv_putattrbit(nd, &attrbits);
2350 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2351 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2352 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2353 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2354 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2355 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2356 		(void) nfsrv_putattrbit(nd, &attrbits);
2357 		nd->nd_flag |= ND_V4WCCATTR;
2358 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2359 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2360 	}
2361 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2362 	if (!(nd->nd_flag & ND_NFSV4))
2363 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2364 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2365 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2366 	error = nfscl_request(nd, fdvp, p, cred, fstuff);
2367 	if (error)
2368 		return (error);
2369 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2370 		/* For NFSv4, parse out any Delereturn replies. */
2371 		if (ret > 0 && nd->nd_repstat != 0 &&
2372 		    (nd->nd_flag & ND_NOMOREDATA)) {
2373 			/*
2374 			 * If the Delegreturn failed, try again without
2375 			 * it. The server will Recall, as required.
2376 			 */
2377 			mbuf_freem(nd->nd_mrep);
2378 			goto tryagain;
2379 		}
2380 		for (i = 0; i < (ret * 2); i++) {
2381 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2382 			    ND_NFSV4) {
2383 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2384 			    if (*(tl + 1)) {
2385 				if (i == 0 && ret > 1) {
2386 				    /*
2387 				     * If the Delegreturn failed, try again
2388 				     * without it. The server will Recall, as
2389 				     * required.
2390 				     * If ret > 1, the first iteration of this
2391 				     * loop is the second DelegReturn result.
2392 				     */
2393 				    mbuf_freem(nd->nd_mrep);
2394 				    goto tryagain;
2395 				} else {
2396 				    nd->nd_flag |= ND_NOMOREDATA;
2397 				}
2398 			    }
2399 			}
2400 		}
2401 		/* Now, the first wcc attribute reply. */
2402 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2403 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2404 			if (*(tl + 1))
2405 				nd->nd_flag |= ND_NOMOREDATA;
2406 		}
2407 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2408 		    fstuff);
2409 		/* and the second wcc attribute reply. */
2410 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2411 		    !error) {
2412 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2413 			if (*(tl + 1))
2414 				nd->nd_flag |= ND_NOMOREDATA;
2415 		}
2416 		if (!error)
2417 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2418 			    NULL, tstuff);
2419 	}
2420 	if (nd->nd_repstat && !error)
2421 		error = nd->nd_repstat;
2422 nfsmout:
2423 	mbuf_freem(nd->nd_mrep);
2424 	return (error);
2425 }
2426 
2427 /*
2428  * nfs hard link create rpc
2429  */
2430 APPLESTATIC int
2431 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2432     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2433     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2434 {
2435 	u_int32_t *tl;
2436 	struct nfsrv_descript nfsd, *nd = &nfsd;
2437 	nfsattrbit_t attrbits;
2438 	int error = 0;
2439 
2440 	*attrflagp = 0;
2441 	*dattrflagp = 0;
2442 	if (namelen > NFS_MAXNAMLEN)
2443 		return (ENAMETOOLONG);
2444 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2445 	if (nd->nd_flag & ND_NFSV4) {
2446 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2447 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2448 	}
2449 	(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2450 		VTONFS(dvp)->n_fhp->nfh_len, 0);
2451 	if (nd->nd_flag & ND_NFSV4) {
2452 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2453 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2454 		NFSWCCATTR_ATTRBIT(&attrbits);
2455 		(void) nfsrv_putattrbit(nd, &attrbits);
2456 		nd->nd_flag |= ND_V4WCCATTR;
2457 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2458 		*tl = txdr_unsigned(NFSV4OP_LINK);
2459 	}
2460 	(void) nfsm_strtom(nd, name, namelen);
2461 	error = nfscl_request(nd, vp, p, cred, dstuff);
2462 	if (error)
2463 		return (error);
2464 	if (nd->nd_flag & ND_NFSV3) {
2465 		error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2466 		if (!error)
2467 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2468 			    NULL, dstuff);
2469 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2470 		/*
2471 		 * First, parse out the PutFH and Getattr result.
2472 		 */
2473 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2474 		if (!(*(tl + 1)))
2475 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2476 		if (*(tl + 1))
2477 			nd->nd_flag |= ND_NOMOREDATA;
2478 		/*
2479 		 * Get the pre-op attributes.
2480 		 */
2481 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2482 	}
2483 	if (nd->nd_repstat && !error)
2484 		error = nd->nd_repstat;
2485 nfsmout:
2486 	mbuf_freem(nd->nd_mrep);
2487 	return (error);
2488 }
2489 
2490 /*
2491  * nfs symbolic link create rpc
2492  */
2493 APPLESTATIC int
2494 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, char *target,
2495     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2496     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2497     int *dattrflagp, void *dstuff)
2498 {
2499 	u_int32_t *tl;
2500 	struct nfsrv_descript nfsd, *nd = &nfsd;
2501 	struct nfsmount *nmp;
2502 	int slen, error = 0;
2503 
2504 	*nfhpp = NULL;
2505 	*attrflagp = 0;
2506 	*dattrflagp = 0;
2507 	nmp = VFSTONFS(vnode_mount(dvp));
2508 	slen = strlen(target);
2509 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2510 		return (ENAMETOOLONG);
2511 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2512 	if (nd->nd_flag & ND_NFSV4) {
2513 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2514 		*tl = txdr_unsigned(NFLNK);
2515 		(void) nfsm_strtom(nd, target, slen);
2516 	}
2517 	(void) nfsm_strtom(nd, name, namelen);
2518 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2519 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2520 	if (!(nd->nd_flag & ND_NFSV4))
2521 		(void) nfsm_strtom(nd, target, slen);
2522 	if (nd->nd_flag & ND_NFSV2)
2523 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2524 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2525 	if (error)
2526 		return (error);
2527 	if (nd->nd_flag & ND_NFSV4)
2528 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2529 	if ((nd->nd_flag & ND_NFSV3) && !error) {
2530 		if (!nd->nd_repstat)
2531 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2532 		if (!error)
2533 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2534 			    NULL, dstuff);
2535 	}
2536 	if (nd->nd_repstat && !error)
2537 		error = nd->nd_repstat;
2538 	mbuf_freem(nd->nd_mrep);
2539 	/*
2540 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2541 	 * Only do this if vfs.nfs.ignore_eexist is set.
2542 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2543 	 * should guarantee "exactly once" RPC semantics.
2544 	 */
2545 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2546 	    nmp->nm_minorvers == 0))
2547 		error = 0;
2548 	return (error);
2549 }
2550 
2551 /*
2552  * nfs make dir rpc
2553  */
2554 APPLESTATIC int
2555 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2556     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2557     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2558     int *dattrflagp, void *dstuff)
2559 {
2560 	u_int32_t *tl;
2561 	struct nfsrv_descript nfsd, *nd = &nfsd;
2562 	nfsattrbit_t attrbits;
2563 	int error = 0;
2564 	struct nfsfh *fhp;
2565 	struct nfsmount *nmp;
2566 
2567 	*nfhpp = NULL;
2568 	*attrflagp = 0;
2569 	*dattrflagp = 0;
2570 	nmp = VFSTONFS(vnode_mount(dvp));
2571 	fhp = VTONFS(dvp)->n_fhp;
2572 	if (namelen > NFS_MAXNAMLEN)
2573 		return (ENAMETOOLONG);
2574 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2575 	if (nd->nd_flag & ND_NFSV4) {
2576 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2577 		*tl = txdr_unsigned(NFDIR);
2578 	}
2579 	(void) nfsm_strtom(nd, name, namelen);
2580 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2581 	if (nd->nd_flag & ND_NFSV4) {
2582 		NFSGETATTR_ATTRBIT(&attrbits);
2583 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2584 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2585 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2586 		(void) nfsrv_putattrbit(nd, &attrbits);
2587 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2588 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2589 		(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2590 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2591 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2592 		(void) nfsrv_putattrbit(nd, &attrbits);
2593 	}
2594 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2595 	if (error)
2596 		return (error);
2597 	if (nd->nd_flag & ND_NFSV4)
2598 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2599 	if (!nd->nd_repstat && !error) {
2600 		if (nd->nd_flag & ND_NFSV4) {
2601 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2602 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2603 		}
2604 		if (!error)
2605 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2606 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2607 			/* Get rid of the PutFH and Getattr status values. */
2608 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2609 			/* Load the directory attributes. */
2610 			error = nfsm_loadattr(nd, dnap);
2611 			if (error == 0)
2612 				*dattrflagp = 1;
2613 		}
2614 	}
2615 	if ((nd->nd_flag & ND_NFSV3) && !error)
2616 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2617 	if (nd->nd_repstat && !error)
2618 		error = nd->nd_repstat;
2619 nfsmout:
2620 	mbuf_freem(nd->nd_mrep);
2621 	/*
2622 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2623 	 * Only do this if vfs.nfs.ignore_eexist is set.
2624 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2625 	 * should guarantee "exactly once" RPC semantics.
2626 	 */
2627 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2628 	    nmp->nm_minorvers == 0))
2629 		error = 0;
2630 	return (error);
2631 }
2632 
2633 /*
2634  * nfs remove directory call
2635  */
2636 APPLESTATIC int
2637 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2638     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2639 {
2640 	struct nfsrv_descript nfsd, *nd = &nfsd;
2641 	int error = 0;
2642 
2643 	*dattrflagp = 0;
2644 	if (namelen > NFS_MAXNAMLEN)
2645 		return (ENAMETOOLONG);
2646 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2647 	(void) nfsm_strtom(nd, name, namelen);
2648 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2649 	if (error)
2650 		return (error);
2651 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2652 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2653 	if (nd->nd_repstat && !error)
2654 		error = nd->nd_repstat;
2655 	mbuf_freem(nd->nd_mrep);
2656 	/*
2657 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2658 	 */
2659 	if (error == ENOENT)
2660 		error = 0;
2661 	return (error);
2662 }
2663 
2664 /*
2665  * Readdir rpc.
2666  * Always returns with either uio_resid unchanged, if you are at the
2667  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2668  * filled in.
2669  * I felt this would allow caching of directory blocks more easily
2670  * than returning a pertially filled block.
2671  * Directory offset cookies:
2672  * Oh my, what to do with them...
2673  * I can think of three ways to deal with them:
2674  * 1 - have the layer above these RPCs maintain a map between logical
2675  *     directory byte offsets and the NFS directory offset cookies
2676  * 2 - pass the opaque directory offset cookies up into userland
2677  *     and let the libc functions deal with them, via the system call
2678  * 3 - return them to userland in the "struct dirent", so future versions
2679  *     of libc can use them and do whatever is necessary to amke things work
2680  *     above these rpc calls, in the meantime
2681  * For now, I do #3 by "hiding" the directory offset cookies after the
2682  * d_name field in struct dirent. This is space inside d_reclen that
2683  * will be ignored by anything that doesn't know about them.
2684  * The directory offset cookies are filled in as the last 8 bytes of
2685  * each directory entry, after d_name. Someday, the userland libc
2686  * functions may be able to use these. In the meantime, it satisfies
2687  * OpenBSD's requirements for cookies being returned.
2688  * If expects the directory offset cookie for the read to be in uio_offset
2689  * and returns the one for the next entry after this directory block in
2690  * there, as well.
2691  */
2692 APPLESTATIC int
2693 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2694     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2695     int *eofp, void *stuff)
2696 {
2697 	int len, left;
2698 	struct dirent *dp = NULL;
2699 	u_int32_t *tl;
2700 	nfsquad_t cookie, ncookie;
2701 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
2702 	struct nfsnode *dnp = VTONFS(vp);
2703 	struct nfsvattr nfsva;
2704 	struct nfsrv_descript nfsd, *nd = &nfsd;
2705 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2706 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2707 	long dotfileid, dotdotfileid = 0;
2708 	u_int32_t fakefileno = 0xffffffff, rderr;
2709 	char *cp;
2710 	nfsattrbit_t attrbits, dattrbits;
2711 	u_int32_t *tl2 = NULL;
2712 	size_t tresid;
2713 
2714 	KASSERT(uiop->uio_iovcnt == 1 &&
2715 	    (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
2716 	    ("nfs readdirrpc bad uio"));
2717 
2718 	/*
2719 	 * There is no point in reading a lot more than uio_resid, however
2720 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2721 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2722 	 * will never make readsize > nm_readdirsize.
2723 	 */
2724 	readsize = nmp->nm_readdirsize;
2725 	if (readsize > uio_uio_resid(uiop))
2726 		readsize = uio_uio_resid(uiop) + DIRBLKSIZ;
2727 
2728 	*attrflagp = 0;
2729 	if (eofp)
2730 		*eofp = 0;
2731 	tresid = uio_uio_resid(uiop);
2732 	cookie.lval[0] = cookiep->nfsuquad[0];
2733 	cookie.lval[1] = cookiep->nfsuquad[1];
2734 	nd->nd_mrep = NULL;
2735 
2736 	/*
2737 	 * For NFSv4, first create the "." and ".." entries.
2738 	 */
2739 	if (NFSHASNFSV4(nmp)) {
2740 		reqsize = 6 * NFSX_UNSIGNED;
2741 		NFSGETATTR_ATTRBIT(&dattrbits);
2742 		NFSZERO_ATTRBIT(&attrbits);
2743 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2744 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2745 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2746 		    NFSATTRBIT_MOUNTEDONFILEID)) {
2747 			NFSSETBIT_ATTRBIT(&attrbits,
2748 			    NFSATTRBIT_MOUNTEDONFILEID);
2749 			gotmnton = 1;
2750 		} else {
2751 			/*
2752 			 * Must fake it. Use the fileno, except when the
2753 			 * fsid is != to that of the directory. For that
2754 			 * case, generate a fake fileno that is not the same.
2755 			 */
2756 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2757 			gotmnton = 0;
2758 		}
2759 
2760 		/*
2761 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2762 		 */
2763 		if (uiop->uio_offset == 0) {
2764 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2765 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2766 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2767 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2768 			(void) nfsrv_putattrbit(nd, &attrbits);
2769 			error = nfscl_request(nd, vp, p, cred, stuff);
2770 			if (error)
2771 			    return (error);
2772 			dotfileid = 0;	/* Fake out the compiler. */
2773 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
2774 			    error = nfsm_loadattr(nd, &nfsva);
2775 			    if (error != 0)
2776 				goto nfsmout;
2777 			    dotfileid = nfsva.na_fileid;
2778 			}
2779 			if (nd->nd_repstat == 0) {
2780 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2781 			    len = fxdr_unsigned(int, *(tl + 4));
2782 			    if (len > 0 && len <= NFSX_V4FHMAX)
2783 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2784 			    else
2785 				error = EPERM;
2786 			    if (!error) {
2787 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2788 				nfsva.na_mntonfileno = 0xffffffff;
2789 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
2790 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
2791 				    NULL, NULL, NULL, p, cred);
2792 				if (error) {
2793 				    dotdotfileid = dotfileid;
2794 				} else if (gotmnton) {
2795 				    if (nfsva.na_mntonfileno != 0xffffffff)
2796 					dotdotfileid = nfsva.na_mntonfileno;
2797 				    else
2798 					dotdotfileid = nfsva.na_fileid;
2799 				} else if (nfsva.na_filesid[0] ==
2800 				    dnp->n_vattr.na_filesid[0] &&
2801 				    nfsva.na_filesid[1] ==
2802 				    dnp->n_vattr.na_filesid[1]) {
2803 				    dotdotfileid = nfsva.na_fileid;
2804 				} else {
2805 				    do {
2806 					fakefileno--;
2807 				    } while (fakefileno ==
2808 					nfsva.na_fileid);
2809 				    dotdotfileid = fakefileno;
2810 				}
2811 			    }
2812 			} else if (nd->nd_repstat == NFSERR_NOENT) {
2813 			    /*
2814 			     * Lookupp returns NFSERR_NOENT when we are
2815 			     * at the root, so just use the current dir.
2816 			     */
2817 			    nd->nd_repstat = 0;
2818 			    dotdotfileid = dotfileid;
2819 			} else {
2820 			    error = nd->nd_repstat;
2821 			}
2822 			mbuf_freem(nd->nd_mrep);
2823 			if (error)
2824 			    return (error);
2825 			nd->nd_mrep = NULL;
2826 			dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
2827 			dp->d_type = DT_DIR;
2828 			dp->d_fileno = dotfileid;
2829 			dp->d_namlen = 1;
2830 			dp->d_name[0] = '.';
2831 			dp->d_name[1] = '\0';
2832 			dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
2833 			/*
2834 			 * Just make these offset cookie 0.
2835 			 */
2836 			tl = (u_int32_t *)&dp->d_name[4];
2837 			*tl++ = 0;
2838 			*tl = 0;
2839 			blksiz += dp->d_reclen;
2840 			uio_uio_resid_add(uiop, -(dp->d_reclen));
2841 			uiop->uio_offset += dp->d_reclen;
2842 			uio_iov_base_add(uiop, dp->d_reclen);
2843 			uio_iov_len_add(uiop, -(dp->d_reclen));
2844 			dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
2845 			dp->d_type = DT_DIR;
2846 			dp->d_fileno = dotdotfileid;
2847 			dp->d_namlen = 2;
2848 			dp->d_name[0] = '.';
2849 			dp->d_name[1] = '.';
2850 			dp->d_name[2] = '\0';
2851 			dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
2852 			/*
2853 			 * Just make these offset cookie 0.
2854 			 */
2855 			tl = (u_int32_t *)&dp->d_name[4];
2856 			*tl++ = 0;
2857 			*tl = 0;
2858 			blksiz += dp->d_reclen;
2859 			uio_uio_resid_add(uiop, -(dp->d_reclen));
2860 			uiop->uio_offset += dp->d_reclen;
2861 			uio_iov_base_add(uiop, dp->d_reclen);
2862 			uio_iov_len_add(uiop, -(dp->d_reclen));
2863 		}
2864 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
2865 	} else {
2866 		reqsize = 5 * NFSX_UNSIGNED;
2867 	}
2868 
2869 
2870 	/*
2871 	 * Loop around doing readdir rpc's of size readsize.
2872 	 * The stopping criteria is EOF or buffer full.
2873 	 */
2874 	while (more_dirs && bigenough) {
2875 		*attrflagp = 0;
2876 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
2877 		if (nd->nd_flag & ND_NFSV2) {
2878 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2879 			*tl++ = cookie.lval[1];
2880 			*tl = txdr_unsigned(readsize);
2881 		} else {
2882 			NFSM_BUILD(tl, u_int32_t *, reqsize);
2883 			*tl++ = cookie.lval[0];
2884 			*tl++ = cookie.lval[1];
2885 			if (cookie.qval == 0) {
2886 				*tl++ = 0;
2887 				*tl++ = 0;
2888 			} else {
2889 				NFSLOCKNODE(dnp);
2890 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
2891 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
2892 				NFSUNLOCKNODE(dnp);
2893 			}
2894 			if (nd->nd_flag & ND_NFSV4) {
2895 				*tl++ = txdr_unsigned(readsize);
2896 				*tl = txdr_unsigned(readsize);
2897 				(void) nfsrv_putattrbit(nd, &attrbits);
2898 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2899 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
2900 				(void) nfsrv_putattrbit(nd, &dattrbits);
2901 			} else {
2902 				*tl = txdr_unsigned(readsize);
2903 			}
2904 		}
2905 		error = nfscl_request(nd, vp, p, cred, stuff);
2906 		if (error)
2907 			return (error);
2908 		if (!(nd->nd_flag & ND_NFSV2)) {
2909 			if (nd->nd_flag & ND_NFSV3)
2910 				error = nfscl_postop_attr(nd, nap, attrflagp,
2911 				    stuff);
2912 			if (!nd->nd_repstat && !error) {
2913 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
2914 				NFSLOCKNODE(dnp);
2915 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
2916 				dnp->n_cookieverf.nfsuquad[1] = *tl;
2917 				NFSUNLOCKNODE(dnp);
2918 			}
2919 		}
2920 		if (nd->nd_repstat || error) {
2921 			if (!error)
2922 				error = nd->nd_repstat;
2923 			goto nfsmout;
2924 		}
2925 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2926 		more_dirs = fxdr_unsigned(int, *tl);
2927 		if (!more_dirs)
2928 			tryformoredirs = 0;
2929 
2930 		/* loop thru the dir entries, doctoring them to 4bsd form */
2931 		while (more_dirs && bigenough) {
2932 			if (nd->nd_flag & ND_NFSV4) {
2933 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
2934 				ncookie.lval[0] = *tl++;
2935 				ncookie.lval[1] = *tl++;
2936 				len = fxdr_unsigned(int, *tl);
2937 			} else if (nd->nd_flag & ND_NFSV3) {
2938 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
2939 				nfsva.na_fileid = fxdr_hyper(tl);
2940 				tl += 2;
2941 				len = fxdr_unsigned(int, *tl);
2942 			} else {
2943 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2944 				nfsva.na_fileid =
2945 				    fxdr_unsigned(long, *tl++);
2946 				len = fxdr_unsigned(int, *tl);
2947 			}
2948 			if (len <= 0 || len > NFS_MAXNAMLEN) {
2949 				error = EBADRPC;
2950 				goto nfsmout;
2951 			}
2952 			tlen = NFSM_RNDUP(len);
2953 			if (tlen == len)
2954 				tlen += 4;  /* To ensure null termination */
2955 			left = DIRBLKSIZ - blksiz;
2956 			if ((int)(tlen + DIRHDSIZ + NFSX_HYPER) > left) {
2957 				dp->d_reclen += left;
2958 				uio_iov_base_add(uiop, left);
2959 				uio_iov_len_add(uiop, -(left));
2960 				uio_uio_resid_add(uiop, -(left));
2961 				uiop->uio_offset += left;
2962 				blksiz = 0;
2963 			}
2964 			if ((int)(tlen + DIRHDSIZ + NFSX_HYPER) > uio_uio_resid(uiop))
2965 				bigenough = 0;
2966 			if (bigenough) {
2967 				dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
2968 				dp->d_namlen = len;
2969 				dp->d_reclen = tlen + DIRHDSIZ + NFSX_HYPER;
2970 				dp->d_type = DT_UNKNOWN;
2971 				blksiz += dp->d_reclen;
2972 				if (blksiz == DIRBLKSIZ)
2973 					blksiz = 0;
2974 				uio_uio_resid_add(uiop, -(DIRHDSIZ));
2975 				uiop->uio_offset += DIRHDSIZ;
2976 				uio_iov_base_add(uiop, DIRHDSIZ);
2977 				uio_iov_len_add(uiop, -(DIRHDSIZ));
2978 				error = nfsm_mbufuio(nd, uiop, len);
2979 				if (error)
2980 					goto nfsmout;
2981 				cp = CAST_DOWN(caddr_t, uio_iov_base(uiop));
2982 				tlen -= len;
2983 				*cp = '\0';	/* null terminate */
2984 				cp += tlen;	/* points to cookie storage */
2985 				tl2 = (u_int32_t *)cp;
2986 				uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
2987 				uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
2988 				uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
2989 				uiop->uio_offset += (tlen + NFSX_HYPER);
2990 			} else {
2991 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2992 				if (error)
2993 					goto nfsmout;
2994 			}
2995 			if (nd->nd_flag & ND_NFSV4) {
2996 				rderr = 0;
2997 				nfsva.na_mntonfileno = 0xffffffff;
2998 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
2999 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3000 				    NULL, NULL, &rderr, p, cred);
3001 				if (error)
3002 					goto nfsmout;
3003 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3004 			} else if (nd->nd_flag & ND_NFSV3) {
3005 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3006 				ncookie.lval[0] = *tl++;
3007 				ncookie.lval[1] = *tl++;
3008 			} else {
3009 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3010 				ncookie.lval[0] = 0;
3011 				ncookie.lval[1] = *tl++;
3012 			}
3013 			if (bigenough) {
3014 			    if (nd->nd_flag & ND_NFSV4) {
3015 				if (rderr) {
3016 				    dp->d_fileno = 0;
3017 				} else {
3018 				    if (gotmnton) {
3019 					if (nfsva.na_mntonfileno != 0xffffffff)
3020 					    dp->d_fileno = nfsva.na_mntonfileno;
3021 					else
3022 					    dp->d_fileno = nfsva.na_fileid;
3023 				    } else if (nfsva.na_filesid[0] ==
3024 					dnp->n_vattr.na_filesid[0] &&
3025 					nfsva.na_filesid[1] ==
3026 					dnp->n_vattr.na_filesid[1]) {
3027 					dp->d_fileno = nfsva.na_fileid;
3028 				    } else {
3029 					do {
3030 					    fakefileno--;
3031 					} while (fakefileno ==
3032 					    nfsva.na_fileid);
3033 					dp->d_fileno = fakefileno;
3034 				    }
3035 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3036 				}
3037 			    } else {
3038 				dp->d_fileno = nfsva.na_fileid;
3039 			    }
3040 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3041 				ncookie.lval[0];
3042 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3043 				ncookie.lval[1];
3044 			}
3045 			more_dirs = fxdr_unsigned(int, *tl);
3046 		}
3047 		/*
3048 		 * If at end of rpc data, get the eof boolean
3049 		 */
3050 		if (!more_dirs) {
3051 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3052 			eof = fxdr_unsigned(int, *tl);
3053 			if (tryformoredirs)
3054 				more_dirs = !eof;
3055 			if (nd->nd_flag & ND_NFSV4) {
3056 				error = nfscl_postop_attr(nd, nap, attrflagp,
3057 				    stuff);
3058 				if (error)
3059 					goto nfsmout;
3060 			}
3061 		}
3062 		mbuf_freem(nd->nd_mrep);
3063 		nd->nd_mrep = NULL;
3064 	}
3065 	/*
3066 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3067 	 * by increasing d_reclen for the last record.
3068 	 */
3069 	if (blksiz > 0) {
3070 		left = DIRBLKSIZ - blksiz;
3071 		dp->d_reclen += left;
3072 		uio_iov_base_add(uiop, left);
3073 		uio_iov_len_add(uiop, -(left));
3074 		uio_uio_resid_add(uiop, -(left));
3075 		uiop->uio_offset += left;
3076 	}
3077 
3078 	/*
3079 	 * If returning no data, assume end of file.
3080 	 * If not bigenough, return not end of file, since you aren't
3081 	 *    returning all the data
3082 	 * Otherwise, return the eof flag from the server.
3083 	 */
3084 	if (eofp) {
3085 		if (tresid == ((size_t)(uio_uio_resid(uiop))))
3086 			*eofp = 1;
3087 		else if (!bigenough)
3088 			*eofp = 0;
3089 		else
3090 			*eofp = eof;
3091 	}
3092 
3093 	/*
3094 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3095 	 */
3096 	while (uio_uio_resid(uiop) > 0 && ((size_t)(uio_uio_resid(uiop))) != tresid) {
3097 		dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
3098 		dp->d_type = DT_UNKNOWN;
3099 		dp->d_fileno = 0;
3100 		dp->d_namlen = 0;
3101 		dp->d_name[0] = '\0';
3102 		tl = (u_int32_t *)&dp->d_name[4];
3103 		*tl++ = cookie.lval[0];
3104 		*tl = cookie.lval[1];
3105 		dp->d_reclen = DIRBLKSIZ;
3106 		uio_iov_base_add(uiop, DIRBLKSIZ);
3107 		uio_iov_len_add(uiop, -(DIRBLKSIZ));
3108 		uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3109 		uiop->uio_offset += DIRBLKSIZ;
3110 	}
3111 
3112 nfsmout:
3113 	if (nd->nd_mrep != NULL)
3114 		mbuf_freem(nd->nd_mrep);
3115 	return (error);
3116 }
3117 
3118 #ifndef APPLE
3119 /*
3120  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3121  * (Also used for NFS V4 when mount flag set.)
3122  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3123  */
3124 APPLESTATIC int
3125 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3126     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3127     int *eofp, void *stuff)
3128 {
3129 	int len, left;
3130 	struct dirent *dp = NULL;
3131 	u_int32_t *tl;
3132 	vnode_t newvp = NULLVP;
3133 	struct nfsrv_descript nfsd, *nd = &nfsd;
3134 	struct nameidata nami, *ndp = &nami;
3135 	struct componentname *cnp = &ndp->ni_cnd;
3136 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3137 	struct nfsnode *dnp = VTONFS(vp), *np;
3138 	struct nfsvattr nfsva;
3139 	struct nfsfh *nfhp;
3140 	nfsquad_t cookie, ncookie;
3141 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3142 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3143 	int isdotdot = 0, unlocknewvp = 0;
3144 	long dotfileid, dotdotfileid = 0, fileno = 0;
3145 	char *cp;
3146 	nfsattrbit_t attrbits, dattrbits;
3147 	size_t tresid;
3148 	u_int32_t *tl2 = NULL, fakefileno = 0xffffffff, rderr;
3149 	struct timespec dctime;
3150 
3151 	KASSERT(uiop->uio_iovcnt == 1 &&
3152 	    (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
3153 	    ("nfs readdirplusrpc bad uio"));
3154 	timespecclear(&dctime);
3155 	*attrflagp = 0;
3156 	if (eofp != NULL)
3157 		*eofp = 0;
3158 	ndp->ni_dvp = vp;
3159 	nd->nd_mrep = NULL;
3160 	cookie.lval[0] = cookiep->nfsuquad[0];
3161 	cookie.lval[1] = cookiep->nfsuquad[1];
3162 	tresid = uio_uio_resid(uiop);
3163 
3164 	/*
3165 	 * For NFSv4, first create the "." and ".." entries.
3166 	 */
3167 	if (NFSHASNFSV4(nmp)) {
3168 		NFSGETATTR_ATTRBIT(&dattrbits);
3169 		NFSZERO_ATTRBIT(&attrbits);
3170 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3171 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3172 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3173 			NFSSETBIT_ATTRBIT(&attrbits,
3174 			    NFSATTRBIT_MOUNTEDONFILEID);
3175 			gotmnton = 1;
3176 		} else {
3177 			/*
3178 			 * Must fake it. Use the fileno, except when the
3179 			 * fsid is != to that of the directory. For that
3180 			 * case, generate a fake fileno that is not the same.
3181 			 */
3182 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3183 			gotmnton = 0;
3184 		}
3185 
3186 		/*
3187 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3188 		 */
3189 		if (uiop->uio_offset == 0) {
3190 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3191 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3192 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3193 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3194 			(void) nfsrv_putattrbit(nd, &attrbits);
3195 			error = nfscl_request(nd, vp, p, cred, stuff);
3196 			if (error)
3197 			    return (error);
3198 			dotfileid = 0;	/* Fake out the compiler. */
3199 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3200 			    error = nfsm_loadattr(nd, &nfsva);
3201 			    if (error != 0)
3202 				goto nfsmout;
3203 			    dctime = nfsva.na_ctime;
3204 			    dotfileid = nfsva.na_fileid;
3205 			}
3206 			if (nd->nd_repstat == 0) {
3207 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3208 			    len = fxdr_unsigned(int, *(tl + 4));
3209 			    if (len > 0 && len <= NFSX_V4FHMAX)
3210 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3211 			    else
3212 				error = EPERM;
3213 			    if (!error) {
3214 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3215 				nfsva.na_mntonfileno = 0xffffffff;
3216 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3217 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3218 				    NULL, NULL, NULL, p, cred);
3219 				if (error) {
3220 				    dotdotfileid = dotfileid;
3221 				} else if (gotmnton) {
3222 				    if (nfsva.na_mntonfileno != 0xffffffff)
3223 					dotdotfileid = nfsva.na_mntonfileno;
3224 				    else
3225 					dotdotfileid = nfsva.na_fileid;
3226 				} else if (nfsva.na_filesid[0] ==
3227 				    dnp->n_vattr.na_filesid[0] &&
3228 				    nfsva.na_filesid[1] ==
3229 				    dnp->n_vattr.na_filesid[1]) {
3230 				    dotdotfileid = nfsva.na_fileid;
3231 				} else {
3232 				    do {
3233 					fakefileno--;
3234 				    } while (fakefileno ==
3235 					nfsva.na_fileid);
3236 				    dotdotfileid = fakefileno;
3237 				}
3238 			    }
3239 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3240 			    /*
3241 			     * Lookupp returns NFSERR_NOENT when we are
3242 			     * at the root, so just use the current dir.
3243 			     */
3244 			    nd->nd_repstat = 0;
3245 			    dotdotfileid = dotfileid;
3246 			} else {
3247 			    error = nd->nd_repstat;
3248 			}
3249 			mbuf_freem(nd->nd_mrep);
3250 			if (error)
3251 			    return (error);
3252 			nd->nd_mrep = NULL;
3253 			dp = (struct dirent *)uio_iov_base(uiop);
3254 			dp->d_type = DT_DIR;
3255 			dp->d_fileno = dotfileid;
3256 			dp->d_namlen = 1;
3257 			dp->d_name[0] = '.';
3258 			dp->d_name[1] = '\0';
3259 			dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
3260 			/*
3261 			 * Just make these offset cookie 0.
3262 			 */
3263 			tl = (u_int32_t *)&dp->d_name[4];
3264 			*tl++ = 0;
3265 			*tl = 0;
3266 			blksiz += dp->d_reclen;
3267 			uio_uio_resid_add(uiop, -(dp->d_reclen));
3268 			uiop->uio_offset += dp->d_reclen;
3269 			uio_iov_base_add(uiop, dp->d_reclen);
3270 			uio_iov_len_add(uiop, -(dp->d_reclen));
3271 			dp = (struct dirent *)uio_iov_base(uiop);
3272 			dp->d_type = DT_DIR;
3273 			dp->d_fileno = dotdotfileid;
3274 			dp->d_namlen = 2;
3275 			dp->d_name[0] = '.';
3276 			dp->d_name[1] = '.';
3277 			dp->d_name[2] = '\0';
3278 			dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
3279 			/*
3280 			 * Just make these offset cookie 0.
3281 			 */
3282 			tl = (u_int32_t *)&dp->d_name[4];
3283 			*tl++ = 0;
3284 			*tl = 0;
3285 			blksiz += dp->d_reclen;
3286 			uio_uio_resid_add(uiop, -(dp->d_reclen));
3287 			uiop->uio_offset += dp->d_reclen;
3288 			uio_iov_base_add(uiop, dp->d_reclen);
3289 			uio_iov_len_add(uiop, -(dp->d_reclen));
3290 		}
3291 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3292 		if (gotmnton)
3293 			NFSSETBIT_ATTRBIT(&attrbits,
3294 			    NFSATTRBIT_MOUNTEDONFILEID);
3295 	}
3296 
3297 	/*
3298 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3299 	 * The stopping criteria is EOF or buffer full.
3300 	 */
3301 	while (more_dirs && bigenough) {
3302 		*attrflagp = 0;
3303 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3304  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3305 		*tl++ = cookie.lval[0];
3306 		*tl++ = cookie.lval[1];
3307 		if (cookie.qval == 0) {
3308 			*tl++ = 0;
3309 			*tl++ = 0;
3310 		} else {
3311 			NFSLOCKNODE(dnp);
3312 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3313 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3314 			NFSUNLOCKNODE(dnp);
3315 		}
3316 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3317 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3318 		if (nd->nd_flag & ND_NFSV4) {
3319 			(void) nfsrv_putattrbit(nd, &attrbits);
3320 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3321 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3322 			(void) nfsrv_putattrbit(nd, &dattrbits);
3323 		}
3324 		error = nfscl_request(nd, vp, p, cred, stuff);
3325 		if (error)
3326 			return (error);
3327 		if (nd->nd_flag & ND_NFSV3)
3328 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3329 		if (nd->nd_repstat || error) {
3330 			if (!error)
3331 				error = nd->nd_repstat;
3332 			goto nfsmout;
3333 		}
3334 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3335 			dctime = nap->na_ctime;
3336 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3337 		NFSLOCKNODE(dnp);
3338 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3339 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3340 		NFSUNLOCKNODE(dnp);
3341 		more_dirs = fxdr_unsigned(int, *tl);
3342 		if (!more_dirs)
3343 			tryformoredirs = 0;
3344 
3345 		/* loop thru the dir entries, doctoring them to 4bsd form */
3346 		while (more_dirs && bigenough) {
3347 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3348 			if (nd->nd_flag & ND_NFSV4) {
3349 				ncookie.lval[0] = *tl++;
3350 				ncookie.lval[1] = *tl++;
3351 			} else {
3352 				fileno = fxdr_unsigned(long, *++tl);
3353 				tl++;
3354 			}
3355 			len = fxdr_unsigned(int, *tl);
3356 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3357 				error = EBADRPC;
3358 				goto nfsmout;
3359 			}
3360 			tlen = NFSM_RNDUP(len);
3361 			if (tlen == len)
3362 				tlen += 4;  /* To ensure null termination */
3363 			left = DIRBLKSIZ - blksiz;
3364 			if ((tlen + DIRHDSIZ + NFSX_HYPER) > left) {
3365 				dp->d_reclen += left;
3366 				uio_iov_base_add(uiop, left);
3367 				uio_iov_len_add(uiop, -(left));
3368 				uio_uio_resid_add(uiop, -(left));
3369 				uiop->uio_offset += left;
3370 				blksiz = 0;
3371 			}
3372 			if ((tlen + DIRHDSIZ + NFSX_HYPER) > uio_uio_resid(uiop))
3373 				bigenough = 0;
3374 			if (bigenough) {
3375 				dp = (struct dirent *)uio_iov_base(uiop);
3376 				dp->d_namlen = len;
3377 				dp->d_reclen = tlen + DIRHDSIZ + NFSX_HYPER;
3378 				dp->d_type = DT_UNKNOWN;
3379 				blksiz += dp->d_reclen;
3380 				if (blksiz == DIRBLKSIZ)
3381 					blksiz = 0;
3382 				uio_uio_resid_add(uiop, -(DIRHDSIZ));
3383 				uiop->uio_offset += DIRHDSIZ;
3384 				uio_iov_base_add(uiop, DIRHDSIZ);
3385 				uio_iov_len_add(uiop, -(DIRHDSIZ));
3386 				cnp->cn_nameptr = uio_iov_base(uiop);
3387 				cnp->cn_namelen = len;
3388 				NFSCNHASHZERO(cnp);
3389 				error = nfsm_mbufuio(nd, uiop, len);
3390 				if (error)
3391 					goto nfsmout;
3392 				cp = uio_iov_base(uiop);
3393 				tlen -= len;
3394 				*cp = '\0';
3395 				cp += tlen;	/* points to cookie storage */
3396 				tl2 = (u_int32_t *)cp;
3397 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3398 				    cnp->cn_nameptr[1] == '.')
3399 					isdotdot = 1;
3400 				else
3401 					isdotdot = 0;
3402 				uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3403 				uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3404 				uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3405 				uiop->uio_offset += (tlen + NFSX_HYPER);
3406 			} else {
3407 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3408 				if (error)
3409 					goto nfsmout;
3410 			}
3411 			nfhp = NULL;
3412 			if (nd->nd_flag & ND_NFSV3) {
3413 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3414 				ncookie.lval[0] = *tl++;
3415 				ncookie.lval[1] = *tl++;
3416 				attrflag = fxdr_unsigned(int, *tl);
3417 				if (attrflag) {
3418 				  error = nfsm_loadattr(nd, &nfsva);
3419 				  if (error)
3420 					goto nfsmout;
3421 				}
3422 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3423 				if (*tl) {
3424 					error = nfsm_getfh(nd, &nfhp);
3425 					if (error)
3426 					    goto nfsmout;
3427 				}
3428 				if (!attrflag && nfhp != NULL) {
3429 					FREE((caddr_t)nfhp, M_NFSFH);
3430 					nfhp = NULL;
3431 				}
3432 			} else {
3433 				rderr = 0;
3434 				nfsva.na_mntonfileno = 0xffffffff;
3435 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3436 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3437 				    NULL, NULL, &rderr, p, cred);
3438 				if (error)
3439 					goto nfsmout;
3440 			}
3441 
3442 			if (bigenough) {
3443 			    if (nd->nd_flag & ND_NFSV4) {
3444 				if (rderr) {
3445 				    dp->d_fileno = 0;
3446 				} else if (gotmnton) {
3447 				    if (nfsva.na_mntonfileno != 0xffffffff)
3448 					dp->d_fileno = nfsva.na_mntonfileno;
3449 				    else
3450 					dp->d_fileno = nfsva.na_fileid;
3451 				} else if (nfsva.na_filesid[0] ==
3452 				    dnp->n_vattr.na_filesid[0] &&
3453 				    nfsva.na_filesid[1] ==
3454 				    dnp->n_vattr.na_filesid[1]) {
3455 				    dp->d_fileno = nfsva.na_fileid;
3456 				} else {
3457 				    do {
3458 					fakefileno--;
3459 				    } while (fakefileno ==
3460 					nfsva.na_fileid);
3461 				    dp->d_fileno = fakefileno;
3462 				}
3463 			    } else {
3464 				dp->d_fileno = fileno;
3465 			    }
3466 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3467 				ncookie.lval[0];
3468 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3469 				ncookie.lval[1];
3470 
3471 			    if (nfhp != NULL) {
3472 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3473 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3474 				    VREF(vp);
3475 				    newvp = vp;
3476 				    unlocknewvp = 0;
3477 				    FREE((caddr_t)nfhp, M_NFSFH);
3478 				    np = dnp;
3479 				} else if (isdotdot != 0) {
3480 				    /*
3481 				     * Skip doing a nfscl_nget() call for "..".
3482 				     * There's a race between acquiring the nfs
3483 				     * node here and lookups that look for the
3484 				     * directory being read (in the parent).
3485 				     * It would try to get a lock on ".." here,
3486 				     * owning the lock on the directory being
3487 				     * read. Lookup will hold the lock on ".."
3488 				     * and try to acquire the lock on the
3489 				     * directory being read.
3490 				     * If the directory is unlocked/relocked,
3491 				     * then there is a LOR with the buflock
3492 				     * vp is relocked.
3493 				     */
3494 				    free(nfhp, M_NFSFH);
3495 				} else {
3496 				    error = nfscl_nget(vnode_mount(vp), vp,
3497 				      nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3498 				    if (!error) {
3499 					newvp = NFSTOV(np);
3500 					unlocknewvp = 1;
3501 				    }
3502 				}
3503 				nfhp = NULL;
3504 				if (newvp != NULLVP) {
3505 				    error = nfscl_loadattrcache(&newvp,
3506 					&nfsva, NULL, NULL, 0, 0);
3507 				    if (error) {
3508 					if (unlocknewvp)
3509 					    vput(newvp);
3510 					else
3511 					    vrele(newvp);
3512 					goto nfsmout;
3513 				    }
3514 				    dp->d_type =
3515 					vtonfs_dtype(np->n_vattr.na_type);
3516 				    ndp->ni_vp = newvp;
3517 				    NFSCNHASH(cnp, HASHINIT);
3518 				    if (cnp->cn_namelen <= NCHNAMLEN &&
3519 					(newvp->v_type != VDIR ||
3520 					 dctime.tv_sec != 0)) {
3521 					cache_enter_time(ndp->ni_dvp,
3522 					    ndp->ni_vp, cnp,
3523 					    &nfsva.na_ctime,
3524 					    newvp->v_type != VDIR ? NULL :
3525 					    &dctime);
3526 				    }
3527 				    if (unlocknewvp)
3528 					vput(newvp);
3529 				    else
3530 					vrele(newvp);
3531 				    newvp = NULLVP;
3532 				}
3533 			    }
3534 			} else if (nfhp != NULL) {
3535 			    FREE((caddr_t)nfhp, M_NFSFH);
3536 			}
3537 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3538 			more_dirs = fxdr_unsigned(int, *tl);
3539 		}
3540 		/*
3541 		 * If at end of rpc data, get the eof boolean
3542 		 */
3543 		if (!more_dirs) {
3544 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3545 			eof = fxdr_unsigned(int, *tl);
3546 			if (tryformoredirs)
3547 				more_dirs = !eof;
3548 			if (nd->nd_flag & ND_NFSV4) {
3549 				error = nfscl_postop_attr(nd, nap, attrflagp,
3550 				    stuff);
3551 				if (error)
3552 					goto nfsmout;
3553 			}
3554 		}
3555 		mbuf_freem(nd->nd_mrep);
3556 		nd->nd_mrep = NULL;
3557 	}
3558 	/*
3559 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3560 	 * by increasing d_reclen for the last record.
3561 	 */
3562 	if (blksiz > 0) {
3563 		left = DIRBLKSIZ - blksiz;
3564 		dp->d_reclen += left;
3565 		uio_iov_base_add(uiop, left);
3566 		uio_iov_len_add(uiop, -(left));
3567 		uio_uio_resid_add(uiop, -(left));
3568 		uiop->uio_offset += left;
3569 	}
3570 
3571 	/*
3572 	 * If returning no data, assume end of file.
3573 	 * If not bigenough, return not end of file, since you aren't
3574 	 *    returning all the data
3575 	 * Otherwise, return the eof flag from the server.
3576 	 */
3577 	if (eofp != NULL) {
3578 		if (tresid == uio_uio_resid(uiop))
3579 			*eofp = 1;
3580 		else if (!bigenough)
3581 			*eofp = 0;
3582 		else
3583 			*eofp = eof;
3584 	}
3585 
3586 	/*
3587 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3588 	 */
3589 	while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3590 		dp = (struct dirent *)uio_iov_base(uiop);
3591 		dp->d_type = DT_UNKNOWN;
3592 		dp->d_fileno = 0;
3593 		dp->d_namlen = 0;
3594 		dp->d_name[0] = '\0';
3595 		tl = (u_int32_t *)&dp->d_name[4];
3596 		*tl++ = cookie.lval[0];
3597 		*tl = cookie.lval[1];
3598 		dp->d_reclen = DIRBLKSIZ;
3599 		uio_iov_base_add(uiop, DIRBLKSIZ);
3600 		uio_iov_len_add(uiop, -(DIRBLKSIZ));
3601 		uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3602 		uiop->uio_offset += DIRBLKSIZ;
3603 	}
3604 
3605 nfsmout:
3606 	if (nd->nd_mrep != NULL)
3607 		mbuf_freem(nd->nd_mrep);
3608 	return (error);
3609 }
3610 #endif	/* !APPLE */
3611 
3612 /*
3613  * Nfs commit rpc
3614  */
3615 APPLESTATIC int
3616 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
3617     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
3618 {
3619 	u_int32_t *tl;
3620 	struct nfsrv_descript nfsd, *nd = &nfsd;
3621 	nfsattrbit_t attrbits;
3622 	int error;
3623 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3624 
3625 	*attrflagp = 0;
3626 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
3627 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3628 	txdr_hyper(offset, tl);
3629 	tl += 2;
3630 	*tl = txdr_unsigned(cnt);
3631 	if (nd->nd_flag & ND_NFSV4) {
3632 		/*
3633 		 * And do a Getattr op.
3634 		 */
3635 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3636 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3637 		NFSGETATTR_ATTRBIT(&attrbits);
3638 		(void) nfsrv_putattrbit(nd, &attrbits);
3639 	}
3640 	error = nfscl_request(nd, vp, p, cred, stuff);
3641 	if (error)
3642 		return (error);
3643 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
3644 	if (!error && !nd->nd_repstat) {
3645 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
3646 		NFSLOCKMNT(nmp);
3647 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
3648 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
3649 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
3650 		}
3651 		NFSUNLOCKMNT(nmp);
3652 		if (nd->nd_flag & ND_NFSV4)
3653 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3654 	}
3655 nfsmout:
3656 	if (!error && nd->nd_repstat)
3657 		error = nd->nd_repstat;
3658 	mbuf_freem(nd->nd_mrep);
3659 	return (error);
3660 }
3661 
3662 /*
3663  * NFS byte range lock rpc.
3664  * (Mostly just calls one of the three lower level RPC routines.)
3665  */
3666 APPLESTATIC int
3667 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
3668     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3669 {
3670 	struct nfscllockowner *lp;
3671 	struct nfsclclient *clp;
3672 	struct nfsfh *nfhp;
3673 	struct nfsrv_descript nfsd, *nd = &nfsd;
3674 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3675 	u_int64_t off, len;
3676 	off_t start, end;
3677 	u_int32_t clidrev = 0;
3678 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
3679 	int callcnt, dorpc;
3680 
3681 	/*
3682 	 * Convert the flock structure into a start and end and do POSIX
3683 	 * bounds checking.
3684 	 */
3685 	switch (fl->l_whence) {
3686 	case SEEK_SET:
3687 	case SEEK_CUR:
3688 		/*
3689 		 * Caller is responsible for adding any necessary offset
3690 		 * when SEEK_CUR is used.
3691 		 */
3692 		start = fl->l_start;
3693 		off = fl->l_start;
3694 		break;
3695 	case SEEK_END:
3696 		start = size + fl->l_start;
3697 		off = size + fl->l_start;
3698 		break;
3699 	default:
3700 		return (EINVAL);
3701 	};
3702 	if (start < 0)
3703 		return (EINVAL);
3704 	if (fl->l_len != 0) {
3705 		end = start + fl->l_len - 1;
3706 		if (end < start)
3707 			return (EINVAL);
3708 	}
3709 
3710 	len = fl->l_len;
3711 	if (len == 0)
3712 		len = NFS64BITSSET;
3713 	retrycnt = 0;
3714 	do {
3715 	    nd->nd_repstat = 0;
3716 	    if (op == F_GETLK) {
3717 		error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3718 		if (error)
3719 			return (error);
3720 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
3721 		if (!error) {
3722 			clidrev = clp->nfsc_clientidrev;
3723 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
3724 			    p, id, flags);
3725 		} else if (error == -1) {
3726 			error = 0;
3727 		}
3728 		nfscl_clientrelease(clp);
3729 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
3730 		/*
3731 		 * We must loop around for all lockowner cases.
3732 		 */
3733 		callcnt = 0;
3734 		error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3735 		if (error)
3736 			return (error);
3737 		do {
3738 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
3739 			clp, id, flags, &lp, &dorpc);
3740 		    /*
3741 		     * If it returns a NULL lp, we're done.
3742 		     */
3743 		    if (lp == NULL) {
3744 			if (callcnt == 0)
3745 			    nfscl_clientrelease(clp);
3746 			else
3747 			    nfscl_releasealllocks(clp, vp, p, id, flags);
3748 			return (error);
3749 		    }
3750 		    if (nmp->nm_clp != NULL)
3751 			clidrev = nmp->nm_clp->nfsc_clientidrev;
3752 		    else
3753 			clidrev = 0;
3754 		    /*
3755 		     * If the server doesn't support Posix lock semantics,
3756 		     * only allow locks on the entire file, since it won't
3757 		     * handle overlapping byte ranges.
3758 		     * There might still be a problem when a lock
3759 		     * upgrade/downgrade (read<->write) occurs, since the
3760 		     * server "might" expect an unlock first?
3761 		     */
3762 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
3763 			(off == 0 && len == NFS64BITSSET))) {
3764 			/*
3765 			 * Since the lock records will go away, we must
3766 			 * wait for grace and delay here.
3767 			 */
3768 			do {
3769 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
3770 				NFSV4LOCKT_READ, cred, p, 0);
3771 			    if ((nd->nd_repstat == NFSERR_GRACE ||
3772 				 nd->nd_repstat == NFSERR_DELAY) &&
3773 				error == 0)
3774 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
3775 				    "nfs_advlock");
3776 			} while ((nd->nd_repstat == NFSERR_GRACE ||
3777 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
3778 		    }
3779 		    callcnt++;
3780 		} while (error == 0 && nd->nd_repstat == 0);
3781 		nfscl_releasealllocks(clp, vp, p, id, flags);
3782 	    } else if (op == F_SETLK) {
3783 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
3784 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
3785 		if (error || donelocally) {
3786 			return (error);
3787 		}
3788 		if (nmp->nm_clp != NULL)
3789 			clidrev = nmp->nm_clp->nfsc_clientidrev;
3790 		else
3791 			clidrev = 0;
3792 		nfhp = VTONFS(vp)->n_fhp;
3793 		if (!lp->nfsl_open->nfso_posixlock &&
3794 		    (off != 0 || len != NFS64BITSSET)) {
3795 			error = EINVAL;
3796 		} else {
3797 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
3798 			    nfhp->nfh_len, lp, newone, reclaim, off,
3799 			    len, fl->l_type, cred, p, 0);
3800 		}
3801 		if (!error)
3802 			error = nd->nd_repstat;
3803 		nfscl_lockrelease(lp, error, newone);
3804 	    } else {
3805 		error = EINVAL;
3806 	    }
3807 	    if (!error)
3808 	        error = nd->nd_repstat;
3809 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
3810 		error == NFSERR_STALEDONTRECOVER ||
3811 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3812 		error == NFSERR_BADSESSION) {
3813 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
3814 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
3815 		&& clidrev != 0) {
3816 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
3817 		retrycnt++;
3818 	    }
3819 	} while (error == NFSERR_GRACE ||
3820 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3821 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
3822 	    error == NFSERR_BADSESSION ||
3823 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
3824 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
3825 	if (error && retrycnt >= 4)
3826 		error = EIO;
3827 	return (error);
3828 }
3829 
3830 /*
3831  * The lower level routine for the LockT case.
3832  */
3833 APPLESTATIC int
3834 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
3835     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
3836     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3837 {
3838 	u_int32_t *tl;
3839 	int error, type, size;
3840 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
3841 	struct nfsnode *np;
3842 	struct nfsmount *nmp;
3843 
3844 	nmp = VFSTONFS(vp->v_mount);
3845 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
3846 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
3847 	if (fl->l_type == F_RDLCK)
3848 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
3849 	else
3850 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
3851 	txdr_hyper(off, tl);
3852 	tl += 2;
3853 	txdr_hyper(len, tl);
3854 	tl += 2;
3855 	*tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
3856 	*tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
3857 	nfscl_filllockowner(id, own, flags);
3858 	np = VTONFS(vp);
3859 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
3860 	    np->n_fhp->nfh_len);
3861 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
3862 	error = nfscl_request(nd, vp, p, cred, NULL);
3863 	if (error)
3864 		return (error);
3865 	if (nd->nd_repstat == 0) {
3866 		fl->l_type = F_UNLCK;
3867 	} else if (nd->nd_repstat == NFSERR_DENIED) {
3868 		nd->nd_repstat = 0;
3869 		fl->l_whence = SEEK_SET;
3870 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
3871 		fl->l_start = fxdr_hyper(tl);
3872 		tl += 2;
3873 		len = fxdr_hyper(tl);
3874 		tl += 2;
3875 		if (len == NFS64BITSSET)
3876 			fl->l_len = 0;
3877 		else
3878 			fl->l_len = len;
3879 		type = fxdr_unsigned(int, *tl++);
3880 		if (type == NFSV4LOCKT_WRITE)
3881 			fl->l_type = F_WRLCK;
3882 		else
3883 			fl->l_type = F_RDLCK;
3884 		/*
3885 		 * XXX For now, I have no idea what to do with the
3886 		 * conflicting lock_owner, so I'll just set the pid == 0
3887 		 * and skip over the lock_owner.
3888 		 */
3889 		fl->l_pid = (pid_t)0;
3890 		tl += 2;
3891 		size = fxdr_unsigned(int, *tl);
3892 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
3893 			error = EBADRPC;
3894 		if (!error)
3895 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
3896 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID ||
3897 	    nd->nd_repstat == NFSERR_BADSESSION)
3898 		nfscl_initiate_recovery(clp);
3899 nfsmout:
3900 	mbuf_freem(nd->nd_mrep);
3901 	return (error);
3902 }
3903 
3904 /*
3905  * Lower level function that performs the LockU RPC.
3906  */
3907 static int
3908 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
3909     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
3910     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
3911 {
3912 	u_int32_t *tl;
3913 	int error;
3914 
3915 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
3916 	    lp->nfsl_open->nfso_fhlen, NULL, NULL);
3917 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
3918 	*tl++ = txdr_unsigned(type);
3919 	*tl = txdr_unsigned(lp->nfsl_seqid);
3920 	if (nfstest_outofseq &&
3921 	    (arc4random() % nfstest_outofseq) == 0)
3922 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
3923 	tl++;
3924 	if (NFSHASNFSV4N(nmp))
3925 		*tl++ = 0;
3926 	else
3927 		*tl++ = lp->nfsl_stateid.seqid;
3928 	*tl++ = lp->nfsl_stateid.other[0];
3929 	*tl++ = lp->nfsl_stateid.other[1];
3930 	*tl++ = lp->nfsl_stateid.other[2];
3931 	txdr_hyper(off, tl);
3932 	tl += 2;
3933 	txdr_hyper(len, tl);
3934 	if (syscred)
3935 		nd->nd_flag |= ND_USEGSSNAME;
3936 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
3937 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
3938 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
3939 	if (error)
3940 		return (error);
3941 	if (nd->nd_repstat == 0) {
3942 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
3943 		lp->nfsl_stateid.seqid = *tl++;
3944 		lp->nfsl_stateid.other[0] = *tl++;
3945 		lp->nfsl_stateid.other[1] = *tl++;
3946 		lp->nfsl_stateid.other[2] = *tl;
3947 	} else if (nd->nd_repstat == NFSERR_STALESTATEID ||
3948 	    nd->nd_repstat == NFSERR_BADSESSION)
3949 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
3950 nfsmout:
3951 	mbuf_freem(nd->nd_mrep);
3952 	return (error);
3953 }
3954 
3955 /*
3956  * The actual Lock RPC.
3957  */
3958 APPLESTATIC int
3959 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
3960     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
3961     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
3962     NFSPROC_T *p, int syscred)
3963 {
3964 	u_int32_t *tl;
3965 	int error, size;
3966 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
3967 
3968 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL);
3969 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
3970 	if (type == F_RDLCK)
3971 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
3972 	else
3973 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
3974 	*tl++ = txdr_unsigned(reclaim);
3975 	txdr_hyper(off, tl);
3976 	tl += 2;
3977 	txdr_hyper(len, tl);
3978 	tl += 2;
3979 	if (newone) {
3980 	    *tl = newnfs_true;
3981 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
3982 		2 * NFSX_UNSIGNED + NFSX_HYPER);
3983 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
3984 	    if (NFSHASNFSV4N(nmp))
3985 		*tl++ = 0;
3986 	    else
3987 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
3988 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
3989 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
3990 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
3991 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
3992 	    *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
3993 	    *tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
3994 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
3995 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
3996 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
3997 	} else {
3998 	    *tl = newnfs_false;
3999 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4000 	    if (NFSHASNFSV4N(nmp))
4001 		*tl++ = 0;
4002 	    else
4003 		*tl++ = lp->nfsl_stateid.seqid;
4004 	    *tl++ = lp->nfsl_stateid.other[0];
4005 	    *tl++ = lp->nfsl_stateid.other[1];
4006 	    *tl++ = lp->nfsl_stateid.other[2];
4007 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4008 	    if (nfstest_outofseq &&
4009 		(arc4random() % nfstest_outofseq) == 0)
4010 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4011 	}
4012 	if (syscred)
4013 		nd->nd_flag |= ND_USEGSSNAME;
4014 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4015 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4016 	if (error)
4017 		return (error);
4018 	if (newone)
4019 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4020 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4021 	if (nd->nd_repstat == 0) {
4022 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4023 		lp->nfsl_stateid.seqid = *tl++;
4024 		lp->nfsl_stateid.other[0] = *tl++;
4025 		lp->nfsl_stateid.other[1] = *tl++;
4026 		lp->nfsl_stateid.other[2] = *tl;
4027 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4028 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4029 		size = fxdr_unsigned(int, *(tl + 7));
4030 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4031 			error = EBADRPC;
4032 		if (!error)
4033 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4034 	} else if (nd->nd_repstat == NFSERR_STALESTATEID ||
4035 	    nd->nd_repstat == NFSERR_BADSESSION)
4036 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4037 nfsmout:
4038 	mbuf_freem(nd->nd_mrep);
4039 	return (error);
4040 }
4041 
4042 /*
4043  * nfs statfs rpc
4044  * (always called with the vp for the mount point)
4045  */
4046 APPLESTATIC int
4047 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4048     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4049     void *stuff)
4050 {
4051 	u_int32_t *tl = NULL;
4052 	struct nfsrv_descript nfsd, *nd = &nfsd;
4053 	struct nfsmount *nmp;
4054 	nfsattrbit_t attrbits;
4055 	int error;
4056 
4057 	*attrflagp = 0;
4058 	nmp = VFSTONFS(vnode_mount(vp));
4059 	if (NFSHASNFSV4(nmp)) {
4060 		/*
4061 		 * For V4, you actually do a getattr.
4062 		 */
4063 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4064 		NFSSTATFS_GETATTRBIT(&attrbits);
4065 		(void) nfsrv_putattrbit(nd, &attrbits);
4066 		nd->nd_flag |= ND_USEGSSNAME;
4067 		error = nfscl_request(nd, vp, p, cred, stuff);
4068 		if (error)
4069 			return (error);
4070 		if (nd->nd_repstat == 0) {
4071 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4072 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4073 			    cred);
4074 			if (!error) {
4075 				nmp->nm_fsid[0] = nap->na_filesid[0];
4076 				nmp->nm_fsid[1] = nap->na_filesid[1];
4077 				NFSSETHASSETFSID(nmp);
4078 				*attrflagp = 1;
4079 			}
4080 		} else {
4081 			error = nd->nd_repstat;
4082 		}
4083 		if (error)
4084 			goto nfsmout;
4085 	} else {
4086 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4087 		error = nfscl_request(nd, vp, p, cred, stuff);
4088 		if (error)
4089 			return (error);
4090 		if (nd->nd_flag & ND_NFSV3) {
4091 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4092 			if (error)
4093 				goto nfsmout;
4094 		}
4095 		if (nd->nd_repstat) {
4096 			error = nd->nd_repstat;
4097 			goto nfsmout;
4098 		}
4099 		NFSM_DISSECT(tl, u_int32_t *,
4100 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4101 	}
4102 	if (NFSHASNFSV3(nmp)) {
4103 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4104 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4105 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4106 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4107 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4108 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4109 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4110 	} else if (NFSHASNFSV4(nmp) == 0) {
4111 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4112 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4113 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4114 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4115 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4116 	}
4117 nfsmout:
4118 	mbuf_freem(nd->nd_mrep);
4119 	return (error);
4120 }
4121 
4122 /*
4123  * nfs pathconf rpc
4124  */
4125 APPLESTATIC int
4126 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4127     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4128     void *stuff)
4129 {
4130 	struct nfsrv_descript nfsd, *nd = &nfsd;
4131 	struct nfsmount *nmp;
4132 	u_int32_t *tl;
4133 	nfsattrbit_t attrbits;
4134 	int error;
4135 
4136 	*attrflagp = 0;
4137 	nmp = VFSTONFS(vnode_mount(vp));
4138 	if (NFSHASNFSV4(nmp)) {
4139 		/*
4140 		 * For V4, you actually do a getattr.
4141 		 */
4142 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4143 		NFSPATHCONF_GETATTRBIT(&attrbits);
4144 		(void) nfsrv_putattrbit(nd, &attrbits);
4145 		nd->nd_flag |= ND_USEGSSNAME;
4146 		error = nfscl_request(nd, vp, p, cred, stuff);
4147 		if (error)
4148 			return (error);
4149 		if (nd->nd_repstat == 0) {
4150 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4151 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4152 			    cred);
4153 			if (!error)
4154 				*attrflagp = 1;
4155 		} else {
4156 			error = nd->nd_repstat;
4157 		}
4158 	} else {
4159 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4160 		error = nfscl_request(nd, vp, p, cred, stuff);
4161 		if (error)
4162 			return (error);
4163 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4164 		if (nd->nd_repstat && !error)
4165 			error = nd->nd_repstat;
4166 		if (!error) {
4167 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4168 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4169 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4170 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4171 			pc->pc_chownrestricted =
4172 			    fxdr_unsigned(u_int32_t, *tl++);
4173 			pc->pc_caseinsensitive =
4174 			    fxdr_unsigned(u_int32_t, *tl++);
4175 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4176 		}
4177 	}
4178 nfsmout:
4179 	mbuf_freem(nd->nd_mrep);
4180 	return (error);
4181 }
4182 
4183 /*
4184  * nfs version 3 fsinfo rpc call
4185  */
4186 APPLESTATIC int
4187 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4188     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4189 {
4190 	u_int32_t *tl;
4191 	struct nfsrv_descript nfsd, *nd = &nfsd;
4192 	int error;
4193 
4194 	*attrflagp = 0;
4195 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4196 	error = nfscl_request(nd, vp, p, cred, stuff);
4197 	if (error)
4198 		return (error);
4199 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4200 	if (nd->nd_repstat && !error)
4201 		error = nd->nd_repstat;
4202 	if (!error) {
4203 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4204 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4205 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4206 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4207 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4208 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4209 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4210 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4211 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4212 		tl += 2;
4213 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4214 		tl += 2;
4215 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4216 	}
4217 nfsmout:
4218 	mbuf_freem(nd->nd_mrep);
4219 	return (error);
4220 }
4221 
4222 /*
4223  * This function performs the Renew RPC.
4224  */
4225 APPLESTATIC int
4226 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4227     NFSPROC_T *p)
4228 {
4229 	u_int32_t *tl;
4230 	struct nfsrv_descript nfsd;
4231 	struct nfsrv_descript *nd = &nfsd;
4232 	struct nfsmount *nmp;
4233 	int error;
4234 	struct nfssockreq *nrp;
4235 
4236 	nmp = clp->nfsc_nmp;
4237 	if (nmp == NULL)
4238 		return (0);
4239 	nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4240 	    &dsp->nfsclds_sess);
4241 	if (!NFSHASNFSV4N(nmp)) {
4242 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4243 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4244 		*tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
4245 		*tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
4246 	}
4247 	nrp = dsp->nfsclds_sockp;
4248 	if (nrp == NULL)
4249 		/* If NULL, use the MDS socket. */
4250 		nrp = &nmp->nm_sockreq;
4251 	nd->nd_flag |= ND_USEGSSNAME;
4252 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4253 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4254 	if (error)
4255 		return (error);
4256 	error = nd->nd_repstat;
4257 	mbuf_freem(nd->nd_mrep);
4258 	return (error);
4259 }
4260 
4261 /*
4262  * This function performs the Releaselockowner RPC.
4263  */
4264 APPLESTATIC int
4265 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4266     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4267 {
4268 	struct nfsrv_descript nfsd, *nd = &nfsd;
4269 	u_int32_t *tl;
4270 	int error;
4271 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4272 
4273 	if (NFSHASNFSV4N(nmp)) {
4274 		/* For NFSv4.1, do a FreeStateID. */
4275 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4276 		    NULL);
4277 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4278 	} else {
4279 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4280 		    NULL);
4281 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4282 		*tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
4283 		*tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
4284 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4285 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4286 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4287 	}
4288 	nd->nd_flag |= ND_USEGSSNAME;
4289 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4290 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4291 	if (error)
4292 		return (error);
4293 	error = nd->nd_repstat;
4294 	mbuf_freem(nd->nd_mrep);
4295 	return (error);
4296 }
4297 
4298 /*
4299  * This function performs the Compound to get the mount pt FH.
4300  */
4301 APPLESTATIC int
4302 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4303     NFSPROC_T *p)
4304 {
4305 	u_int32_t *tl;
4306 	struct nfsrv_descript nfsd;
4307 	struct nfsrv_descript *nd = &nfsd;
4308 	u_char *cp, *cp2;
4309 	int error, cnt, len, setnil;
4310 	u_int32_t *opcntp;
4311 
4312 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL);
4313 	cp = dirpath;
4314 	cnt = 0;
4315 	do {
4316 		setnil = 0;
4317 		while (*cp == '/')
4318 			cp++;
4319 		cp2 = cp;
4320 		while (*cp2 != '\0' && *cp2 != '/')
4321 			cp2++;
4322 		if (*cp2 == '/') {
4323 			setnil = 1;
4324 			*cp2 = '\0';
4325 		}
4326 		if (cp2 != cp) {
4327 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4328 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4329 			nfsm_strtom(nd, cp, strlen(cp));
4330 			cnt++;
4331 		}
4332 		if (setnil)
4333 			*cp2++ = '/';
4334 		cp = cp2;
4335 	} while (*cp != '\0');
4336 	if (NFSHASNFSV4N(nmp))
4337 		/* Has a Sequence Op done by nfscl_reqstart(). */
4338 		*opcntp = txdr_unsigned(3 + cnt);
4339 	else
4340 		*opcntp = txdr_unsigned(2 + cnt);
4341 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4342 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4343 	nd->nd_flag |= ND_USEGSSNAME;
4344 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4345 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4346 	if (error)
4347 		return (error);
4348 	if (nd->nd_repstat == 0) {
4349 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4350 		tl += (2 + 2 * cnt);
4351 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4352 			len > NFSX_FHMAX) {
4353 			nd->nd_repstat = NFSERR_BADXDR;
4354 		} else {
4355 			nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4356 			if (nd->nd_repstat == 0)
4357 				nmp->nm_fhsize = len;
4358 		}
4359 	}
4360 	error = nd->nd_repstat;
4361 nfsmout:
4362 	mbuf_freem(nd->nd_mrep);
4363 	return (error);
4364 }
4365 
4366 /*
4367  * This function performs the Delegreturn RPC.
4368  */
4369 APPLESTATIC int
4370 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4371     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4372 {
4373 	u_int32_t *tl;
4374 	struct nfsrv_descript nfsd;
4375 	struct nfsrv_descript *nd = &nfsd;
4376 	int error;
4377 
4378 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4379 	    dp->nfsdl_fhlen, NULL, NULL);
4380 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4381 	if (NFSHASNFSV4N(nmp))
4382 		*tl++ = 0;
4383 	else
4384 		*tl++ = dp->nfsdl_stateid.seqid;
4385 	*tl++ = dp->nfsdl_stateid.other[0];
4386 	*tl++ = dp->nfsdl_stateid.other[1];
4387 	*tl = dp->nfsdl_stateid.other[2];
4388 	if (syscred)
4389 		nd->nd_flag |= ND_USEGSSNAME;
4390 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4391 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4392 	if (error)
4393 		return (error);
4394 	error = nd->nd_repstat;
4395 	mbuf_freem(nd->nd_mrep);
4396 	return (error);
4397 }
4398 
4399 /*
4400  * nfs getacl call.
4401  */
4402 APPLESTATIC int
4403 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4404     struct acl *aclp, void *stuff)
4405 {
4406 	struct nfsrv_descript nfsd, *nd = &nfsd;
4407 	int error;
4408 	nfsattrbit_t attrbits;
4409 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4410 
4411 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4412 		return (EOPNOTSUPP);
4413 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4414 	NFSZERO_ATTRBIT(&attrbits);
4415 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4416 	(void) nfsrv_putattrbit(nd, &attrbits);
4417 	error = nfscl_request(nd, vp, p, cred, stuff);
4418 	if (error)
4419 		return (error);
4420 	if (!nd->nd_repstat)
4421 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4422 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4423 	else
4424 		error = nd->nd_repstat;
4425 	mbuf_freem(nd->nd_mrep);
4426 	return (error);
4427 }
4428 
4429 /*
4430  * nfs setacl call.
4431  */
4432 APPLESTATIC int
4433 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4434     struct acl *aclp, void *stuff)
4435 {
4436 	int error;
4437 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4438 
4439 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4440 		return (EOPNOTSUPP);
4441 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4442 	return (error);
4443 }
4444 
4445 /*
4446  * nfs setacl call.
4447  */
4448 static int
4449 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4450     struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4451 {
4452 	struct nfsrv_descript nfsd, *nd = &nfsd;
4453 	int error;
4454 	nfsattrbit_t attrbits;
4455 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4456 
4457 	if (!NFSHASNFSV4(nmp))
4458 		return (EOPNOTSUPP);
4459 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4460 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4461 	NFSZERO_ATTRBIT(&attrbits);
4462 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4463 	(void) nfsv4_fillattr(nd, vnode_mount(vp), vp, aclp, NULL, NULL, 0,
4464 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0);
4465 	error = nfscl_request(nd, vp, p, cred, stuff);
4466 	if (error)
4467 		return (error);
4468 	/* Don't care about the pre/postop attributes */
4469 	mbuf_freem(nd->nd_mrep);
4470 	return (nd->nd_repstat);
4471 }
4472 
4473 /*
4474  * Do the NFSv4.1 Exchange ID.
4475  */
4476 int
4477 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
4478     struct nfssockreq *nrp, uint32_t exchflags, struct nfsclds **dspp,
4479     struct ucred *cred, NFSPROC_T *p)
4480 {
4481 	uint32_t *tl, v41flags;
4482 	struct nfsrv_descript nfsd;
4483 	struct nfsrv_descript *nd = &nfsd;
4484 	struct nfsclds *dsp;
4485 	struct timespec verstime;
4486 	int error, len;
4487 
4488 	*dspp = NULL;
4489 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL);
4490 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4491 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
4492 	*tl = txdr_unsigned(clp->nfsc_rev);
4493 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
4494 
4495 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4496 	*tl++ = txdr_unsigned(exchflags);
4497 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
4498 
4499 	/* Set the implementation id4 */
4500 	*tl = txdr_unsigned(1);
4501 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
4502 	(void) nfsm_strtom(nd, version, strlen(version));
4503 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
4504 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
4505 	verstime.tv_nsec = 0;
4506 	txdr_nfsv4time(&verstime, tl);
4507 	nd->nd_flag |= ND_USEGSSNAME;
4508 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4509 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4510 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
4511 	    (int)nd->nd_repstat);
4512 	if (error != 0)
4513 		return (error);
4514 	if (nd->nd_repstat == 0) {
4515 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
4516 		len = fxdr_unsigned(int, *(tl + 7));
4517 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
4518 			error = NFSERR_BADXDR;
4519 			goto nfsmout;
4520 		}
4521 		dsp = malloc(sizeof(struct nfsclds) + len, M_NFSCLDS,
4522 		    M_WAITOK | M_ZERO);
4523 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
4524 		dsp->nfsclds_servownlen = len;
4525 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
4526 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
4527 		dsp->nfsclds_sess.nfsess_sequenceid =
4528 		    fxdr_unsigned(uint32_t, *tl++);
4529 		v41flags = fxdr_unsigned(uint32_t, *tl);
4530 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
4531 		    NFSHASPNFSOPT(nmp)) {
4532 			NFSCL_DEBUG(1, "set PNFS\n");
4533 			NFSLOCKMNT(nmp);
4534 			nmp->nm_state |= NFSSTA_PNFS;
4535 			NFSUNLOCKMNT(nmp);
4536 			dsp->nfsclds_flags |= NFSCLDS_MDS;
4537 		}
4538 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
4539 			dsp->nfsclds_flags |= NFSCLDS_DS;
4540 		if (len > 0)
4541 			nd->nd_repstat = nfsrv_mtostr(nd,
4542 			    dsp->nfsclds_serverown, len);
4543 		if (nd->nd_repstat == 0) {
4544 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
4545 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
4546 			    NULL, MTX_DEF);
4547 			nfscl_initsessionslots(&dsp->nfsclds_sess);
4548 			*dspp = dsp;
4549 		} else
4550 			free(dsp, M_NFSCLDS);
4551 	}
4552 	error = nd->nd_repstat;
4553 nfsmout:
4554 	mbuf_freem(nd->nd_mrep);
4555 	return (error);
4556 }
4557 
4558 /*
4559  * Do the NFSv4.1 Create Session.
4560  */
4561 int
4562 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
4563     struct nfssockreq *nrp, uint32_t sequenceid, int mds, struct ucred *cred,
4564     NFSPROC_T *p)
4565 {
4566 	uint32_t crflags, *tl;
4567 	struct nfsrv_descript nfsd;
4568 	struct nfsrv_descript *nd = &nfsd;
4569 	int error, irdcnt;
4570 
4571 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL);
4572 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
4573 	*tl++ = sep->nfsess_clientid.lval[0];
4574 	*tl++ = sep->nfsess_clientid.lval[1];
4575 	*tl++ = txdr_unsigned(sequenceid);
4576 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
4577 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0)
4578 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
4579 	*tl = txdr_unsigned(crflags);
4580 
4581 	/* Fill in fore channel attributes. */
4582 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4583 	*tl++ = 0;				/* Header pad size */
4584 	*tl++ = txdr_unsigned(100000);		/* Max request size */
4585 	*tl++ = txdr_unsigned(100000);		/* Max response size */
4586 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4587 	*tl++ = txdr_unsigned(20);		/* Max operations */
4588 	*tl++ = txdr_unsigned(64);		/* Max slots */
4589 	*tl = 0;				/* No rdma ird */
4590 
4591 	/* Fill in back channel attributes. */
4592 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4593 	*tl++ = 0;				/* Header pad size */
4594 	*tl++ = txdr_unsigned(10000);		/* Max request size */
4595 	*tl++ = txdr_unsigned(10000);		/* Max response size */
4596 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4597 	*tl++ = txdr_unsigned(4);		/* Max operations */
4598 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
4599 	*tl = 0;				/* No rdma ird */
4600 
4601 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
4602 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
4603 
4604 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
4605 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
4606 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
4607 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
4608 	*tl++ = 0;				/* Null machine name */
4609 	*tl++ = 0;				/* Uid == 0 */
4610 	*tl++ = 0;				/* Gid == 0 */
4611 	*tl = 0;				/* No additional gids */
4612 	nd->nd_flag |= ND_USEGSSNAME;
4613 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
4614 	    NFS_VER4, NULL, 1, NULL, NULL);
4615 	if (error != 0)
4616 		return (error);
4617 	if (nd->nd_repstat == 0) {
4618 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4619 		    2 * NFSX_UNSIGNED);
4620 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
4621 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
4622 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
4623 		crflags = fxdr_unsigned(uint32_t, *tl);
4624 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
4625 			NFSLOCKMNT(nmp);
4626 			nmp->nm_state |= NFSSTA_SESSPERSIST;
4627 			NFSUNLOCKMNT(nmp);
4628 		}
4629 
4630 		/* Get the fore channel slot count. */
4631 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4632 		tl += 3;		/* Skip the other counts. */
4633 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
4634 		tl++;
4635 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
4636 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
4637 		irdcnt = fxdr_unsigned(int, *tl);
4638 		if (irdcnt > 0)
4639 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
4640 
4641 		/* and the back channel slot count. */
4642 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4643 		tl += 5;
4644 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
4645 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
4646 	}
4647 	error = nd->nd_repstat;
4648 nfsmout:
4649 	mbuf_freem(nd->nd_mrep);
4650 	return (error);
4651 }
4652 
4653 /*
4654  * Do the NFSv4.1 Destroy Session.
4655  */
4656 int
4657 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
4658     struct ucred *cred, NFSPROC_T *p)
4659 {
4660 	uint32_t *tl;
4661 	struct nfsrv_descript nfsd;
4662 	struct nfsrv_descript *nd = &nfsd;
4663 	int error;
4664 
4665 	nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL);
4666 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
4667 	bcopy(NFSMNT_MDSSESSION(nmp)->nfsess_sessionid, tl, NFSX_V4SESSIONID);
4668 	nd->nd_flag |= ND_USEGSSNAME;
4669 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4670 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4671 	if (error != 0)
4672 		return (error);
4673 	error = nd->nd_repstat;
4674 	mbuf_freem(nd->nd_mrep);
4675 	return (error);
4676 }
4677 
4678 /*
4679  * Do the NFSv4.1 Destroy Client.
4680  */
4681 int
4682 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
4683     struct ucred *cred, NFSPROC_T *p)
4684 {
4685 	uint32_t *tl;
4686 	struct nfsrv_descript nfsd;
4687 	struct nfsrv_descript *nd = &nfsd;
4688 	int error;
4689 
4690 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL);
4691 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4692 	*tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
4693 	*tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
4694 	nd->nd_flag |= ND_USEGSSNAME;
4695 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4696 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4697 	if (error != 0)
4698 		return (error);
4699 	error = nd->nd_repstat;
4700 	mbuf_freem(nd->nd_mrep);
4701 	return (error);
4702 }
4703 
4704 /*
4705  * Do the NFSv4.1 LayoutGet.
4706  */
4707 int
4708 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
4709     uint64_t offset, uint64_t len, uint64_t minlen, int layoutlen,
4710     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp,
4711     struct ucred *cred, NFSPROC_T *p, void *stuff)
4712 {
4713 	uint32_t *tl;
4714 	struct nfsrv_descript nfsd, *nd = &nfsd;
4715 	struct nfsfh *nfhp;
4716 	struct nfsclflayout *flp, *prevflp, *tflp;
4717 	int cnt, error, gotiomode, fhcnt, nfhlen, i, j;
4718 	uint8_t *cp;
4719 	uint64_t retlen;
4720 
4721 	flp = NULL;
4722 	gotiomode = -1;
4723 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL);
4724 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
4725 	    NFSX_STATEID);
4726 	*tl++ = newnfs_false;		/* Don't signal availability. */
4727 	*tl++ = txdr_unsigned(NFSLAYOUT_NFSV4_1_FILES);
4728 	*tl++ = txdr_unsigned(iomode);
4729 	txdr_hyper(offset, tl);
4730 	tl += 2;
4731 	txdr_hyper(len, tl);
4732 	tl += 2;
4733 	txdr_hyper(minlen, tl);
4734 	tl += 2;
4735 	*tl++ = txdr_unsigned(stateidp->seqid);
4736 	NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
4737 	*tl++ = stateidp->other[0];
4738 	*tl++ = stateidp->other[1];
4739 	*tl++ = stateidp->other[2];
4740 	*tl = txdr_unsigned(layoutlen);
4741 	nd->nd_flag |= ND_USEGSSNAME;
4742 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4743 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4744 	if (error != 0)
4745 		return (error);
4746 	if (nd->nd_repstat == 0) {
4747 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
4748 		if (*tl++ != 0)
4749 			*retonclosep = 1;
4750 		else
4751 			*retonclosep = 0;
4752 		stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
4753 		NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
4754 		    (int)stateidp->seqid);
4755 		stateidp->other[0] = *tl++;
4756 		stateidp->other[1] = *tl++;
4757 		stateidp->other[2] = *tl++;
4758 		cnt = fxdr_unsigned(int, *tl);
4759 		NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
4760 		if (cnt <= 0 || cnt > 10000) {
4761 			/* Don't accept more than 10000 layouts in reply. */
4762 			error = NFSERR_BADXDR;
4763 			goto nfsmout;
4764 		}
4765 		for (i = 0; i < cnt; i++) {
4766 			/* Dissect all the way to the file handle cnt. */
4767 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_HYPER +
4768 			    6 * NFSX_UNSIGNED + NFSX_V4DEVICEID);
4769 			fhcnt = fxdr_unsigned(int, *(tl + 11 +
4770 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
4771 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
4772 			if (fhcnt < 0 || fhcnt > 100) {
4773 				/* Don't accept more than 100 file handles. */
4774 				error = NFSERR_BADXDR;
4775 				goto nfsmout;
4776 			}
4777 			if (fhcnt > 1)
4778 				flp = malloc(sizeof(*flp) + (fhcnt - 1) *
4779 				    sizeof(struct nfsfh *),
4780 				    M_NFSFLAYOUT, M_WAITOK);
4781 			else
4782 				flp = malloc(sizeof(*flp),
4783 				    M_NFSFLAYOUT, M_WAITOK);
4784 			flp->nfsfl_flags = 0;
4785 			flp->nfsfl_fhcnt = 0;
4786 			flp->nfsfl_devp = NULL;
4787 			flp->nfsfl_off = fxdr_hyper(tl); tl += 2;
4788 			retlen = fxdr_hyper(tl); tl += 2;
4789 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
4790 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
4791 			else
4792 				flp->nfsfl_end = flp->nfsfl_off + retlen;
4793 			flp->nfsfl_iomode = fxdr_unsigned(int, *tl++);
4794 			if (gotiomode == -1)
4795 				gotiomode = flp->nfsfl_iomode;
4796 			NFSCL_DEBUG(4, "layg reqiom=%d retiom=%d\n", iomode,
4797 			    (int)flp->nfsfl_iomode);
4798 			if (fxdr_unsigned(int, *tl++) !=
4799 			    NFSLAYOUT_NFSV4_1_FILES) {
4800 				printf("NFSv4.1: got non-files layout\n");
4801 				error = NFSERR_BADXDR;
4802 				goto nfsmout;
4803 			}
4804 			NFSBCOPY(++tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
4805 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
4806 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
4807 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
4808 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
4809 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
4810 			if (fxdr_unsigned(int, *tl) != fhcnt) {
4811 				printf("EEK! bad fhcnt\n");
4812 				error = NFSERR_BADXDR;
4813 				goto nfsmout;
4814 			}
4815 			for (j = 0; j < fhcnt; j++) {
4816 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4817 				nfhlen = fxdr_unsigned(int, *tl);
4818 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
4819 					error = NFSERR_BADXDR;
4820 					goto nfsmout;
4821 				}
4822 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
4823 				    M_NFSFH, M_WAITOK);
4824 				flp->nfsfl_fh[j] = nfhp;
4825 				flp->nfsfl_fhcnt++;
4826 				nfhp->nfh_len = nfhlen;
4827 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
4828 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
4829 			}
4830 			if (flp->nfsfl_iomode == gotiomode) {
4831 				/* Keep the list in increasing offset order. */
4832 				tflp = LIST_FIRST(flhp);
4833 				prevflp = NULL;
4834 				while (tflp != NULL &&
4835 				    tflp->nfsfl_off < flp->nfsfl_off) {
4836 					prevflp = tflp;
4837 					tflp = LIST_NEXT(tflp, nfsfl_list);
4838 				}
4839 				if (prevflp == NULL)
4840 					LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
4841 				else
4842 					LIST_INSERT_AFTER(prevflp, flp,
4843 					    nfsfl_list);
4844 			} else {
4845 				printf("nfscl_layoutget(): got wrong iomode\n");
4846 				nfscl_freeflayout(flp);
4847 			}
4848 			flp = NULL;
4849 		}
4850 	}
4851 	if (nd->nd_repstat != 0 && error == 0)
4852 		error = nd->nd_repstat;
4853 nfsmout:
4854 	if (error != 0 && flp != NULL)
4855 		nfscl_freeflayout(flp);
4856 	mbuf_freem(nd->nd_mrep);
4857 	return (error);
4858 }
4859 
4860 /*
4861  * Do the NFSv4.1 Get Device Info.
4862  */
4863 int
4864 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
4865     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
4866     NFSPROC_T *p)
4867 {
4868 	uint32_t cnt, *tl;
4869 	struct nfsrv_descript nfsd;
4870 	struct nfsrv_descript *nd = &nfsd;
4871 	struct sockaddr_storage ss;
4872 	struct nfsclds *dsp = NULL, **dspp;
4873 	struct nfscldevinfo *ndi;
4874 	int addrcnt, bitcnt, error, i, isudp, j, pos, safilled, stripecnt;
4875 	uint8_t stripeindex;
4876 
4877 	*ndip = NULL;
4878 	ndi = NULL;
4879 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL);
4880 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
4881 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
4882 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
4883 	*tl++ = txdr_unsigned(layouttype);
4884 	*tl++ = txdr_unsigned(100000);
4885 	if (notifybitsp != NULL && *notifybitsp != 0) {
4886 		*tl = txdr_unsigned(1);		/* One word of bits. */
4887 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
4888 		*tl = txdr_unsigned(*notifybitsp);
4889 	} else
4890 		*tl = txdr_unsigned(0);
4891 	nd->nd_flag |= ND_USEGSSNAME;
4892 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4893 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4894 	if (error != 0)
4895 		return (error);
4896 	if (nd->nd_repstat == 0) {
4897 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4898 		if (layouttype != fxdr_unsigned(int, *tl++))
4899 			printf("EEK! devinfo layout type not same!\n");
4900 		stripecnt = fxdr_unsigned(int, *++tl);
4901 		NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
4902 		if (stripecnt < 1 || stripecnt > 4096) {
4903 			printf("NFS devinfo stripecnt %d: out of range\n",
4904 			    stripecnt);
4905 			error = NFSERR_BADXDR;
4906 			goto nfsmout;
4907 		}
4908 		NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) * NFSX_UNSIGNED);
4909 		addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
4910 		NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
4911 		if (addrcnt < 1 || addrcnt > 128) {
4912 			printf("NFS devinfo addrcnt %d: out of range\n",
4913 			    addrcnt);
4914 			error = NFSERR_BADXDR;
4915 			goto nfsmout;
4916 		}
4917 
4918 		/*
4919 		 * Now we know how many stripe indices and addresses, so
4920 		 * we can allocate the structure the correct size.
4921 		 */
4922 		i = (stripecnt * sizeof(uint8_t)) / sizeof(struct nfsclds *)
4923 		    + 1;
4924 		NFSCL_DEBUG(4, "stripeindices=%d\n", i);
4925 		ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
4926 		    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK | M_ZERO);
4927 		NFSBCOPY(deviceid, ndi->nfsdi_deviceid, NFSX_V4DEVICEID);
4928 		ndi->nfsdi_refcnt = 0;
4929 		ndi->nfsdi_stripecnt = stripecnt;
4930 		ndi->nfsdi_addrcnt = addrcnt;
4931 		/* Fill in the stripe indices. */
4932 		for (i = 0; i < stripecnt; i++) {
4933 			stripeindex = fxdr_unsigned(uint8_t, *tl++);
4934 			NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
4935 			if (stripeindex >= addrcnt) {
4936 				printf("NFS devinfo stripeindex %d: too big\n",
4937 				    (int)stripeindex);
4938 				error = NFSERR_BADXDR;
4939 				goto nfsmout;
4940 			}
4941 			nfsfldi_setstripeindex(ndi, i, stripeindex);
4942 		}
4943 
4944 		/* Now, dissect the server address(es). */
4945 		safilled = 0;
4946 		for (i = 0; i < addrcnt; i++) {
4947 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4948 			cnt = fxdr_unsigned(uint32_t, *tl);
4949 			if (cnt == 0) {
4950 				printf("NFS devinfo 0 len addrlist\n");
4951 				error = NFSERR_BADXDR;
4952 				goto nfsmout;
4953 			}
4954 			dspp = nfsfldi_addr(ndi, i);
4955 			pos = arc4random() % cnt;	/* Choose one. */
4956 			safilled = 0;
4957 			for (j = 0; j < cnt; j++) {
4958 				error = nfsv4_getipaddr(nd, &ss, &isudp);
4959 				if (error != 0 && error != EPERM) {
4960 					error = NFSERR_BADXDR;
4961 					goto nfsmout;
4962 				}
4963 				if (error == 0 && isudp == 0) {
4964 					/*
4965 					 * The algorithm is:
4966 					 * - use "pos" entry if it is of the
4967 					 *   same af_family or none of them
4968 					 *   is of the same af_family
4969 					 * else
4970 					 * - use the first one of the same
4971 					 *   af_family.
4972 					 */
4973 					if ((safilled == 0 && ss.ss_family ==
4974 					     nmp->nm_nam->sa_family) ||
4975 					    (j == pos &&
4976 					     (safilled == 0 || ss.ss_family ==
4977 					      nmp->nm_nam->sa_family)) ||
4978 					    (safilled == 1 && ss.ss_family ==
4979 					     nmp->nm_nam->sa_family)) {
4980 						error = nfsrpc_fillsa(nmp, &ss,
4981 						    &dsp, p);
4982 						if (error == 0) {
4983 							*dspp = dsp;
4984 							if (ss.ss_family ==
4985 							 nmp->nm_nam->sa_family)
4986 								safilled = 2;
4987 							else
4988 								safilled = 1;
4989 						}
4990 					}
4991 				}
4992 			}
4993 			if (safilled == 0)
4994 				break;
4995 		}
4996 
4997 		/* And the notify bits. */
4998 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4999 		if (safilled != 0) {
5000 			bitcnt = fxdr_unsigned(int, *tl);
5001 			if (bitcnt > 0) {
5002 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5003 				if (notifybitsp != NULL)
5004 					*notifybitsp =
5005 					    fxdr_unsigned(uint32_t, *tl);
5006 			}
5007 			*ndip = ndi;
5008 		} else
5009 			error = EPERM;
5010 	}
5011 	if (nd->nd_repstat != 0)
5012 		error = nd->nd_repstat;
5013 nfsmout:
5014 	if (error != 0 && ndi != NULL)
5015 		nfscl_freedevinfo(ndi);
5016 	mbuf_freem(nd->nd_mrep);
5017 	return (error);
5018 }
5019 
5020 /*
5021  * Do the NFSv4.1 LayoutCommit.
5022  */
5023 int
5024 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5025     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5026     int layouttype, int layoutupdatecnt, uint8_t *layp, struct ucred *cred,
5027     NFSPROC_T *p, void *stuff)
5028 {
5029 	uint32_t *tl;
5030 	struct nfsrv_descript nfsd, *nd = &nfsd;
5031 	int error, outcnt, i;
5032 	uint8_t *cp;
5033 
5034 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL);
5035 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5036 	    NFSX_STATEID);
5037 	txdr_hyper(off, tl);
5038 	tl += 2;
5039 	txdr_hyper(len, tl);
5040 	tl += 2;
5041 	if (reclaim != 0)
5042 		*tl++ = newnfs_true;
5043 	else
5044 		*tl++ = newnfs_false;
5045 	*tl++ = txdr_unsigned(stateidp->seqid);
5046 	*tl++ = stateidp->other[0];
5047 	*tl++ = stateidp->other[1];
5048 	*tl++ = stateidp->other[2];
5049 	*tl++ = newnfs_true;
5050 	if (lastbyte < off)
5051 		lastbyte = off;
5052 	else if (lastbyte >= (off + len))
5053 		lastbyte = off + len - 1;
5054 	txdr_hyper(lastbyte, tl);
5055 	tl += 2;
5056 	*tl++ = newnfs_false;
5057 	*tl++ = txdr_unsigned(layouttype);
5058 	*tl = txdr_unsigned(layoutupdatecnt);
5059 	if (layoutupdatecnt > 0) {
5060 		KASSERT(layouttype != NFSLAYOUT_NFSV4_1_FILES,
5061 		    ("Must be nil for Files Layout"));
5062 		outcnt = NFSM_RNDUP(layoutupdatecnt);
5063 		NFSM_BUILD(cp, uint8_t *, outcnt);
5064 		NFSBCOPY(layp, cp, layoutupdatecnt);
5065 		cp += layoutupdatecnt;
5066 		for (i = 0; i < (outcnt - layoutupdatecnt); i++)
5067 			*cp++ = 0x0;
5068 	}
5069 	nd->nd_flag |= ND_USEGSSNAME;
5070 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5071 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5072 	if (error != 0)
5073 		return (error);
5074 	error = nd->nd_repstat;
5075 	mbuf_freem(nd->nd_mrep);
5076 	return (error);
5077 }
5078 
5079 /*
5080  * Do the NFSv4.1 LayoutReturn.
5081  */
5082 int
5083 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5084     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5085     uint64_t len, nfsv4stateid_t *stateidp, int layoutcnt, uint32_t *layp,
5086     struct ucred *cred, NFSPROC_T *p, void *stuff)
5087 {
5088 	uint32_t *tl;
5089 	struct nfsrv_descript nfsd, *nd = &nfsd;
5090 	int error, outcnt, i;
5091 	uint8_t *cp;
5092 
5093 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL);
5094 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5095 	if (reclaim != 0)
5096 		*tl++ = newnfs_true;
5097 	else
5098 		*tl++ = newnfs_false;
5099 	*tl++ = txdr_unsigned(layouttype);
5100 	*tl++ = txdr_unsigned(iomode);
5101 	*tl = txdr_unsigned(layoutreturn);
5102 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5103 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5104 		    NFSX_UNSIGNED);
5105 		txdr_hyper(offset, tl);
5106 		tl += 2;
5107 		txdr_hyper(len, tl);
5108 		tl += 2;
5109 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5110 		*tl++ = txdr_unsigned(stateidp->seqid);
5111 		*tl++ = stateidp->other[0];
5112 		*tl++ = stateidp->other[1];
5113 		*tl++ = stateidp->other[2];
5114 		*tl = txdr_unsigned(layoutcnt);
5115 		if (layoutcnt > 0) {
5116 			outcnt = NFSM_RNDUP(layoutcnt);
5117 			NFSM_BUILD(cp, uint8_t *, outcnt);
5118 			NFSBCOPY(layp, cp, layoutcnt);
5119 			cp += layoutcnt;
5120 			for (i = 0; i < (outcnt - layoutcnt); i++)
5121 				*cp++ = 0x0;
5122 		}
5123 	}
5124 	nd->nd_flag |= ND_USEGSSNAME;
5125 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5126 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5127 	if (error != 0)
5128 		return (error);
5129 	if (nd->nd_repstat == 0) {
5130 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5131 		if (*tl != 0) {
5132 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5133 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5134 			stateidp->other[0] = *tl++;
5135 			stateidp->other[1] = *tl++;
5136 			stateidp->other[2] = *tl;
5137 		}
5138 	} else
5139 		error = nd->nd_repstat;
5140 nfsmout:
5141 	mbuf_freem(nd->nd_mrep);
5142 	return (error);
5143 }
5144 
5145 /*
5146  * Acquire a layout and devinfo, if possible. The caller must have acquired
5147  * a reference count on the nfsclclient structure before calling this.
5148  * Return the layout in lypp with a reference count on it, if successful.
5149  */
5150 static int
5151 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5152     int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
5153     struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5154 {
5155 	struct nfscllayout *lyp;
5156 	struct nfsclflayout *flp, *tflp;
5157 	struct nfscldevinfo *dip;
5158 	struct nfsclflayouthead flh;
5159 	int error = 0, islocked, layoutlen, recalled, retonclose;
5160 	nfsv4stateid_t stateid;
5161 
5162 	*lypp = NULL;
5163 	/*
5164 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5165 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5166 	 * flp == NULL.
5167 	 */
5168 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5169 	    off, &flp, &recalled);
5170 	islocked = 0;
5171 	if (lyp == NULL || flp == NULL) {
5172 		if (recalled != 0)
5173 			return (EIO);
5174 		LIST_INIT(&flh);
5175 		layoutlen = NFSMNT_MDSSESSION(nmp)->nfsess_maxcache -
5176 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5177 		if (lyp == NULL) {
5178 			stateid.seqid = 0;
5179 			stateid.other[0] = stateidp->other[0];
5180 			stateid.other[1] = stateidp->other[1];
5181 			stateid.other[2] = stateidp->other[2];
5182 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5183 			    nfhp->nfh_len, iomode, (uint64_t)0, INT64_MAX,
5184 			    (uint64_t)0, layoutlen, &stateid, &retonclose,
5185 			    &flh, cred, p, NULL);
5186 		} else {
5187 			islocked = 1;
5188 			stateid.seqid = lyp->nfsly_stateid.seqid;
5189 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5190 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5191 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5192 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5193 			    nfhp->nfh_len, iomode, off, INT64_MAX,
5194 			    (uint64_t)0, layoutlen, &stateid, &retonclose,
5195 			    &flh, cred, p, NULL);
5196 		}
5197 		if (error == 0)
5198 			LIST_FOREACH(tflp, &flh, nfsfl_list) {
5199 				error = nfscl_adddevinfo(nmp, NULL, tflp);
5200 				if (error != 0) {
5201 					error = nfsrpc_getdeviceinfo(nmp,
5202 					    tflp->nfsfl_dev,
5203 					    NFSLAYOUT_NFSV4_1_FILES,
5204 					    notifybitsp, &dip, cred, p);
5205 					if (error != 0)
5206 						break;
5207 					error = nfscl_adddevinfo(nmp, dip,
5208 					    tflp);
5209 					if (error != 0)
5210 						printf(
5211 						    "getlayout: cannot add\n");
5212 				}
5213 			}
5214 		if (error == 0) {
5215 			/*
5216 			 * nfscl_layout() always returns with the nfsly_lock
5217 			 * set to a refcnt (shared lock).
5218 			 */
5219 			error = nfscl_layout(nmp, vp, nfhp->nfh_fh,
5220 			    nfhp->nfh_len, &stateid, retonclose, &flh, &lyp,
5221 			    cred, p);
5222 			if (error == 0)
5223 				*lypp = lyp;
5224 		} else if (islocked != 0)
5225 			nfsv4_unlock(&lyp->nfsly_lock, 0);
5226 	} else
5227 		*lypp = lyp;
5228 	return (error);
5229 }
5230 
5231 /*
5232  * Do a TCP connection plus exchange id and create session.
5233  * If successful, a "struct nfsclds" is linked into the list for the
5234  * mount point and a pointer to it is returned.
5235  */
5236 static int
5237 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_storage *ssp,
5238     struct nfsclds **dspp, NFSPROC_T *p)
5239 {
5240 	struct sockaddr_in *msad, *sad, *ssd;
5241 	struct sockaddr_in6 *msad6, *sad6, *ssd6;
5242 	struct nfsclclient *clp;
5243 	struct nfssockreq *nrp;
5244 	struct nfsclds *dsp, *tdsp;
5245 	int error;
5246 	enum nfsclds_state retv;
5247 	uint32_t sequenceid;
5248 
5249 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5250 	    ("nfsrpc_fillsa: NULL nr_cred"));
5251 	NFSLOCKCLSTATE();
5252 	clp = nmp->nm_clp;
5253 	NFSUNLOCKCLSTATE();
5254 	if (clp == NULL)
5255 		return (EPERM);
5256 	if (ssp->ss_family == AF_INET) {
5257 		ssd = (struct sockaddr_in *)ssp;
5258 		NFSLOCKMNT(nmp);
5259 
5260 		/*
5261 		 * Check to see if we already have a session for this
5262 		 * address that is usable for a DS.
5263 		 * Note that the MDS's address is in a different place
5264 		 * than the sessions already acquired for DS's.
5265 		 */
5266 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5267 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5268 		while (tdsp != NULL) {
5269 			if (msad != NULL && msad->sin_family == AF_INET &&
5270 			    ssd->sin_addr.s_addr == msad->sin_addr.s_addr &&
5271 			    ssd->sin_port == msad->sin_port &&
5272 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0) {
5273 				*dspp = tdsp;
5274 				NFSUNLOCKMNT(nmp);
5275 				NFSCL_DEBUG(4, "fnd same addr\n");
5276 				return (0);
5277 			}
5278 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5279 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5280 				msad = (struct sockaddr_in *)
5281 				    tdsp->nfsclds_sockp->nr_nam;
5282 			else
5283 				msad = NULL;
5284 		}
5285 		NFSUNLOCKMNT(nmp);
5286 
5287 		/* No IP address match, so look for new/trunked one. */
5288 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5289 		sad->sin_len = sizeof(*sad);
5290 		sad->sin_family = AF_INET;
5291 		sad->sin_port = ssd->sin_port;
5292 		sad->sin_addr.s_addr = ssd->sin_addr.s_addr;
5293 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5294 		nrp->nr_nam = (struct sockaddr *)sad;
5295 	} else if (ssp->ss_family == AF_INET6) {
5296 		ssd6 = (struct sockaddr_in6 *)ssp;
5297 		NFSLOCKMNT(nmp);
5298 
5299 		/*
5300 		 * Check to see if we already have a session for this
5301 		 * address that is usable for a DS.
5302 		 * Note that the MDS's address is in a different place
5303 		 * than the sessions already acquired for DS's.
5304 		 */
5305 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5306 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5307 		while (tdsp != NULL) {
5308 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5309 			    IN6_ARE_ADDR_EQUAL(&ssd6->sin6_addr,
5310 			    &msad6->sin6_addr) &&
5311 			    ssd6->sin6_port == msad6->sin6_port &&
5312 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0) {
5313 				*dspp = tdsp;
5314 				NFSUNLOCKMNT(nmp);
5315 				return (0);
5316 			}
5317 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5318 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5319 				msad6 = (struct sockaddr_in6 *)
5320 				    tdsp->nfsclds_sockp->nr_nam;
5321 			else
5322 				msad6 = NULL;
5323 		}
5324 		NFSUNLOCKMNT(nmp);
5325 
5326 		/* No IP address match, so look for new/trunked one. */
5327 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5328 		sad6->sin6_len = sizeof(*sad6);
5329 		sad6->sin6_family = AF_INET6;
5330 		sad6->sin6_port = ssd6->sin6_port;
5331 		NFSBCOPY(&ssd6->sin6_addr, &sad6->sin6_addr,
5332 		    sizeof(struct in6_addr));
5333 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5334 		nrp->nr_nam = (struct sockaddr *)sad6;
5335 	} else
5336 		return (EPERM);
5337 
5338 	nrp->nr_sotype = SOCK_STREAM;
5339 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5340 	nrp->nr_prog = NFS_PROG;
5341 	nrp->nr_vers = NFS_VER4;
5342 
5343 	/*
5344 	 * Use the credentials that were used for the mount, which are
5345 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5346 	 * Ref. counting the credentials with crhold() is probably not
5347 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5348 	 * unmount, but I did it anyhow.
5349 	 */
5350 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5351 	error = newnfs_connect(nmp, nrp, NULL, p, 0);
5352 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
5353 
5354 	/* Now, do the exchangeid and create session. */
5355 	if (error == 0)
5356 		error = nfsrpc_exchangeid(nmp, clp, nrp, NFSV4EXCH_USEPNFSDS,
5357 		    &dsp, nrp->nr_cred, p);
5358 	NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5359 	if (error == 0) {
5360 		dsp->nfsclds_sockp = nrp;
5361 		NFSLOCKMNT(nmp);
5362 		retv = nfscl_getsameserver(nmp, dsp, &tdsp);
5363 		NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5364 		if (retv == NFSDSP_USETHISSESSION) {
5365 			NFSUNLOCKMNT(nmp);
5366 			/*
5367 			 * If there is already a session for this server,
5368 			 * use it.
5369 			 */
5370 			(void)newnfs_disconnect(nrp);
5371 			nfscl_freenfsclds(dsp);
5372 			*dspp = tdsp;
5373 			return (0);
5374 		}
5375 		if (retv == NFSDSP_SEQTHISSESSION)
5376 			sequenceid = tdsp->nfsclds_sess.nfsess_sequenceid;
5377 		else
5378 			sequenceid = dsp->nfsclds_sess.nfsess_sequenceid;
5379 		NFSUNLOCKMNT(nmp);
5380 		error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5381 		    nrp, sequenceid, 0, nrp->nr_cred, p);
5382 		NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5383 	} else {
5384 		NFSFREECRED(nrp->nr_cred);
5385 		NFSFREEMUTEX(&nrp->nr_mtx);
5386 		free(nrp->nr_nam, M_SONAME);
5387 		free(nrp, M_NFSSOCKREQ);
5388 	}
5389 	if (error == 0) {
5390 		NFSCL_DEBUG(3, "add DS session\n");
5391 		/*
5392 		 * Put it at the end of the list. That way the list
5393 		 * is ordered by when the entry was added. This matters
5394 		 * since the one done first is the one that should be
5395 		 * used for sequencid'ing any subsequent create sessions.
5396 		 */
5397 		NFSLOCKMNT(nmp);
5398 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5399 		NFSUNLOCKMNT(nmp);
5400 		*dspp = dsp;
5401 	} else if (dsp != NULL)
5402 		nfscl_freenfsclds(dsp);
5403 	return (error);
5404 }
5405 
5406 /*
5407  * Do the NFSv4.1 Reclaim Complete.
5408  */
5409 int
5410 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5411 {
5412 	uint32_t *tl;
5413 	struct nfsrv_descript nfsd;
5414 	struct nfsrv_descript *nd = &nfsd;
5415 	int error;
5416 
5417 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL);
5418 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5419 	*tl = newnfs_false;
5420 	nd->nd_flag |= ND_USEGSSNAME;
5421 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5422 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5423 	if (error != 0)
5424 		return (error);
5425 	error = nd->nd_repstat;
5426 	mbuf_freem(nd->nd_mrep);
5427 	return (error);
5428 }
5429 
5430 /*
5431  * Initialize the slot tables for a session.
5432  */
5433 static void
5434 nfscl_initsessionslots(struct nfsclsession *sep)
5435 {
5436 	int i;
5437 
5438 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
5439 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5440 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5441 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5442 	}
5443 	for (i = 0; i < 64; i++)
5444 		sep->nfsess_slotseq[i] = 0;
5445 	sep->nfsess_slots = 0;
5446 }
5447 
5448 /*
5449  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
5450  */
5451 int
5452 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5453     uint32_t rwaccess, struct ucred *cred, NFSPROC_T *p)
5454 {
5455 	struct nfsnode *np = VTONFS(vp);
5456 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5457 	struct nfscllayout *layp;
5458 	struct nfscldevinfo *dip;
5459 	struct nfsclflayout *rflp;
5460 	nfsv4stateid_t stateid;
5461 	struct ucred *newcred;
5462 	uint64_t lastbyte, len, off, oresid, xfer;
5463 	int eof, error, iolaymode, recalled;
5464 	void *lckp;
5465 
5466 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
5467 	    (np->n_flag & NNOLAYOUT) != 0)
5468 		return (EIO);
5469 	/* Now, get a reference cnt on the clientid for this mount. */
5470 	if (nfscl_getref(nmp) == 0)
5471 		return (EIO);
5472 
5473 	/* Find an appropriate stateid. */
5474 	newcred = NFSNEWCRED(cred);
5475 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
5476 	    rwaccess, 1, newcred, p, &stateid, &lckp);
5477 	if (error != 0) {
5478 		NFSFREECRED(newcred);
5479 		nfscl_relref(nmp);
5480 		return (error);
5481 	}
5482 	/* Search for a layout for this file. */
5483 	off = uiop->uio_offset;
5484 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
5485 	    np->n_fhp->nfh_len, off, &rflp, &recalled);
5486 	if (layp == NULL || rflp == NULL) {
5487 		if (recalled != 0) {
5488 			NFSFREECRED(newcred);
5489 			nfscl_relref(nmp);
5490 			return (EIO);
5491 		}
5492 		if (layp != NULL) {
5493 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
5494 			layp = NULL;
5495 		}
5496 		/* Try and get a Layout, if it is supported. */
5497 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
5498 		    (np->n_flag & NWRITEOPENED) != 0)
5499 			iolaymode = NFSLAYOUTIOMODE_RW;
5500 		else
5501 			iolaymode = NFSLAYOUTIOMODE_READ;
5502 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
5503 		    NULL, &stateid, off, &layp, newcred, p);
5504 		if (error != 0) {
5505 			NFSLOCKNODE(np);
5506 			np->n_flag |= NNOLAYOUT;
5507 			NFSUNLOCKNODE(np);
5508 			if (lckp != NULL)
5509 				nfscl_lockderef(lckp);
5510 			NFSFREECRED(newcred);
5511 			if (layp != NULL)
5512 				nfscl_rellayout(layp, 0);
5513 			nfscl_relref(nmp);
5514 			return (error);
5515 		}
5516 	}
5517 
5518 	/*
5519 	 * Loop around finding a layout that works for the first part of
5520 	 * this I/O operation, and then call the function that actually
5521 	 * does the RPC.
5522 	 */
5523 	eof = 0;
5524 	len = (uint64_t)uiop->uio_resid;
5525 	while (len > 0 && error == 0 && eof == 0) {
5526 		off = uiop->uio_offset;
5527 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
5528 		if (error == 0) {
5529 			oresid = xfer = (uint64_t)uiop->uio_resid;
5530 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
5531 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
5532 			dip = nfscl_getdevinfo(nmp->nm_clp, rflp->nfsfl_dev,
5533 			    rflp->nfsfl_devp);
5534 			if (dip != NULL) {
5535 				error = nfscl_doflayoutio(vp, uiop, iomode,
5536 				    must_commit, &eof, &stateid, rwaccess, dip,
5537 				    layp, rflp, off, xfer, newcred, p);
5538 				nfscl_reldevinfo(dip);
5539 				lastbyte = off + xfer - 1;
5540 				if (error == 0) {
5541 					NFSLOCKCLSTATE();
5542 					if (lastbyte > layp->nfsly_lastbyte)
5543 						layp->nfsly_lastbyte = lastbyte;
5544 					NFSUNLOCKCLSTATE();
5545 				}
5546 			} else
5547 				error = EIO;
5548 			if (error == 0)
5549 				len -= (oresid - (uint64_t)uiop->uio_resid);
5550 		}
5551 	}
5552 	if (lckp != NULL)
5553 		nfscl_lockderef(lckp);
5554 	NFSFREECRED(newcred);
5555 	nfscl_rellayout(layp, 0);
5556 	nfscl_relref(nmp);
5557 	return (error);
5558 }
5559 
5560 /*
5561  * Find a file layout that will handle the first bytes of the requested
5562  * range and return the information from it needed to to the I/O operation.
5563  */
5564 int
5565 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
5566     struct nfsclflayout **retflpp)
5567 {
5568 	struct nfsclflayout *flp, *nflp, *rflp;
5569 	uint32_t rw;
5570 
5571 	rflp = NULL;
5572 	rw = rwaccess;
5573 	/* For reading, do the Read list first and then the Write list. */
5574 	do {
5575 		if (rw == NFSV4OPEN_ACCESSREAD)
5576 			flp = LIST_FIRST(&lyp->nfsly_flayread);
5577 		else
5578 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
5579 		while (flp != NULL) {
5580 			nflp = LIST_NEXT(flp, nfsfl_list);
5581 			if (flp->nfsfl_off > off)
5582 				break;
5583 			if (flp->nfsfl_end > off &&
5584 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
5585 				rflp = flp;
5586 			flp = nflp;
5587 		}
5588 		if (rw == NFSV4OPEN_ACCESSREAD)
5589 			rw = NFSV4OPEN_ACCESSWRITE;
5590 		else
5591 			rw = 0;
5592 	} while (rw != 0);
5593 	if (rflp != NULL) {
5594 		/* This one covers the most bytes starting at off. */
5595 		*retflpp = rflp;
5596 		return (0);
5597 	}
5598 	return (EIO);
5599 }
5600 
5601 /*
5602  * Do I/O using an NFSv4.1 file layout.
5603  */
5604 static int
5605 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5606     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
5607     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
5608     uint64_t len, struct ucred *cred, NFSPROC_T *p)
5609 {
5610 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
5611 	int commit_thru_mds, error = 0, stripe_index, stripe_pos;
5612 	struct nfsnode *np;
5613 	struct nfsfh *fhp;
5614 	struct nfsclds **dspp;
5615 
5616 	np = VTONFS(vp);
5617 	rel_off = off - flp->nfsfl_patoff;
5618 	stripe_unit_size = (flp->nfsfl_util >> 6) & 0x3ffffff;
5619 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
5620 	    dp->nfsdi_stripecnt;
5621 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
5622 
5623 	/* Loop around, doing I/O for each stripe unit. */
5624 	while (len > 0 && error == 0) {
5625 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
5626 		dspp = nfsfldi_addr(dp, stripe_index);
5627 		if (len > transfer)
5628 			xfer = transfer;
5629 		else
5630 			xfer = len;
5631 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
5632 			/* Dense layout. */
5633 			if (stripe_pos >= flp->nfsfl_fhcnt)
5634 				return (EIO);
5635 			fhp = flp->nfsfl_fh[stripe_pos];
5636 			io_off = (rel_off / (stripe_unit_size *
5637 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
5638 			    rel_off % stripe_unit_size;
5639 		} else {
5640 			/* Sparse layout. */
5641 			if (flp->nfsfl_fhcnt > 1) {
5642 				if (stripe_index >= flp->nfsfl_fhcnt)
5643 					return (EIO);
5644 				fhp = flp->nfsfl_fh[stripe_index];
5645 			} else if (flp->nfsfl_fhcnt == 1)
5646 				fhp = flp->nfsfl_fh[0];
5647 			else
5648 				fhp = np->n_fhp;
5649 			io_off = off;
5650 		}
5651 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0)
5652 			commit_thru_mds = 1;
5653 		else
5654 			commit_thru_mds = 0;
5655 		if (rwflag == FREAD)
5656 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
5657 			    io_off, xfer, fhp, cred, p);
5658 		else {
5659 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
5660 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
5661 			    cred, p);
5662 			if (error == 0) {
5663 				NFSLOCKCLSTATE();
5664 				lyp->nfsly_flags |= NFSLY_WRITTEN;
5665 				NFSUNLOCKCLSTATE();
5666 			}
5667 		}
5668 		if (error == 0) {
5669 			transfer = stripe_unit_size;
5670 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
5671 			len -= xfer;
5672 			off += xfer;
5673 		}
5674 	}
5675 	return (error);
5676 }
5677 
5678 /*
5679  * The actual read RPC done to a DS.
5680  */
5681 static int
5682 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
5683     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp,
5684     struct ucred *cred, NFSPROC_T *p)
5685 {
5686 	uint32_t *tl;
5687 	int error, retlen;
5688 	struct nfsrv_descript nfsd;
5689 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5690 	struct nfsrv_descript *nd = &nfsd;
5691 	struct nfssockreq *nrp;
5692 
5693 	nd->nd_mrep = NULL;
5694 	nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh, fhp->nfh_len,
5695 	    NULL, &dsp->nfsclds_sess);
5696 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
5697 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
5698 	txdr_hyper(io_off, tl);
5699 	*(tl + 2) = txdr_unsigned(len);
5700 	nrp = dsp->nfsclds_sockp;
5701 	if (nrp == NULL)
5702 		/* If NULL, use the MDS socket. */
5703 		nrp = &nmp->nm_sockreq;
5704 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
5705 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5706 	if (error != 0)
5707 		return (error);
5708 	if (nd->nd_repstat != 0) {
5709 		error = nd->nd_repstat;
5710 		goto nfsmout;
5711 	}
5712 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5713 	*eofp = fxdr_unsigned(int, *tl);
5714 	NFSM_STRSIZ(retlen, len);
5715 	error = nfsm_mbufuio(nd, uiop, retlen);
5716 nfsmout:
5717 	if (nd->nd_mrep != NULL)
5718 		mbuf_freem(nd->nd_mrep);
5719 	return (error);
5720 }
5721 
5722 /*
5723  * The actual write RPC done to a DS.
5724  */
5725 static int
5726 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5727     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
5728     struct nfsfh *fhp, int commit_thru_mds, struct ucred *cred, NFSPROC_T *p)
5729 {
5730 	uint32_t *tl;
5731 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5732 	int error, rlen, commit, committed = NFSWRITE_FILESYNC;
5733 	int32_t backup;
5734 	struct nfsrv_descript nfsd;
5735 	struct nfsrv_descript *nd = &nfsd;
5736 	struct nfssockreq *nrp;
5737 
5738 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
5739 	nd->nd_mrep = NULL;
5740 	nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh, fhp->nfh_len,
5741 	    NULL, &dsp->nfsclds_sess);
5742 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
5743 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
5744 	txdr_hyper(io_off, tl);
5745 	tl += 2;
5746 	*tl++ = txdr_unsigned(*iomode);
5747 	*tl = txdr_unsigned(len);
5748 	nfsm_uiombuf(nd, uiop, len);
5749 	nrp = dsp->nfsclds_sockp;
5750 	if (nrp == NULL)
5751 		/* If NULL, use the MDS socket. */
5752 		nrp = &nmp->nm_sockreq;
5753 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
5754 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5755 	if (error != 0)
5756 		return (error);
5757 	if (nd->nd_repstat != 0) {
5758 		/*
5759 		 * In case the rpc gets retried, roll
5760 		 * the uio fileds changed by nfsm_uiombuf()
5761 		 * back.
5762 		 */
5763 		uiop->uio_offset -= len;
5764 		uio_uio_resid_add(uiop, len);
5765 		uio_iov_base_add(uiop, -len);
5766 		uio_iov_len_add(uiop, len);
5767 		error = nd->nd_repstat;
5768 	} else {
5769 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
5770 		rlen = fxdr_unsigned(int, *tl++);
5771 		if (rlen == 0) {
5772 			error = NFSERR_IO;
5773 			goto nfsmout;
5774 		} else if (rlen < len) {
5775 			backup = len - rlen;
5776 			uio_iov_base_add(uiop, -(backup));
5777 			uio_iov_len_add(uiop, backup);
5778 			uiop->uio_offset -= backup;
5779 			uio_uio_resid_add(uiop, backup);
5780 			len = rlen;
5781 		}
5782 		commit = fxdr_unsigned(int, *tl++);
5783 
5784 		/*
5785 		 * Return the lowest committment level
5786 		 * obtained by any of the RPCs.
5787 		 */
5788 		if (committed == NFSWRITE_FILESYNC)
5789 			committed = commit;
5790 		else if (committed == NFSWRITE_DATASYNC &&
5791 		    commit == NFSWRITE_UNSTABLE)
5792 			committed = commit;
5793 		if (commit_thru_mds != 0) {
5794 			NFSLOCKMNT(nmp);
5795 			if (!NFSHASWRITEVERF(nmp)) {
5796 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
5797 				NFSSETWRITEVERF(nmp);
5798 	    		} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
5799 				*must_commit = 1;
5800 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
5801 			}
5802 			NFSUNLOCKMNT(nmp);
5803 		} else {
5804 			NFSLOCKDS(dsp);
5805 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
5806 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
5807 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
5808 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
5809 				*must_commit = 1;
5810 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
5811 			}
5812 			NFSUNLOCKDS(dsp);
5813 		}
5814 	}
5815 nfsmout:
5816 	if (nd->nd_mrep != NULL)
5817 		mbuf_freem(nd->nd_mrep);
5818 	*iomode = committed;
5819 	if (nd->nd_repstat != 0 && error == 0)
5820 		error = nd->nd_repstat;
5821 	return (error);
5822 }
5823 
5824 /*
5825  * Free up the nfsclds structure.
5826  */
5827 void
5828 nfscl_freenfsclds(struct nfsclds *dsp)
5829 {
5830 	int i;
5831 
5832 	if (dsp == NULL)
5833 		return;
5834 	if (dsp->nfsclds_sockp != NULL) {
5835 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
5836 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
5837 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
5838 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
5839 	}
5840 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
5841 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
5842 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
5843 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
5844 			m_freem(
5845 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
5846 	}
5847 	free(dsp, M_NFSCLDS);
5848 }
5849 
5850 static enum nfsclds_state
5851 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
5852     struct nfsclds **retdspp)
5853 {
5854 	struct nfsclds *dsp, *cur_dsp;
5855 
5856 	/*
5857 	 * Search the list of nfsclds structures for one with the same
5858 	 * server.
5859 	 */
5860 	cur_dsp = NULL;
5861 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
5862 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
5863 		    dsp->nfsclds_servownlen != 0 &&
5864 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
5865 		    dsp->nfsclds_servownlen)) {
5866 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
5867 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
5868 			    dsp->nfsclds_flags);
5869 			/* Server major id matches. */
5870 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
5871 				*retdspp = dsp;
5872 				return (NFSDSP_USETHISSESSION);
5873 			}
5874 
5875 			/*
5876 			 * Note the first match, so it can be used for
5877 			 * sequence'ing new sessions.
5878 			 */
5879 			if (cur_dsp == NULL)
5880 				cur_dsp = dsp;
5881 		}
5882 	}
5883 	if (cur_dsp != NULL) {
5884 		*retdspp = cur_dsp;
5885 		return (NFSDSP_SEQTHISSESSION);
5886 	}
5887 	return (NFSDSP_NOTFOUND);
5888 }
5889 
5890 #ifdef notyet
5891 /*
5892  * NFS commit rpc to a DS.
5893  */
5894 static int
5895 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
5896     struct nfsfh *fhp, struct ucred *cred, NFSPROC_T *p, void *stuff)
5897 {
5898 	uint32_t *tl;
5899 	struct nfsrv_descript nfsd, *nd = &nfsd;
5900 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5901 	struct nfssockreq *nrp;
5902 	int error;
5903 
5904 	nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh, fhp->nfh_len,
5905 	    NULL, &dsp->nfsclds_sess);
5906 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
5907 	txdr_hyper(offset, tl);
5908 	tl += 2;
5909 	*tl = txdr_unsigned(cnt);
5910 	nrp = dsp->nfsclds_sockp;
5911 	if (nrp == NULL)
5912 		/* If NULL, use the MDS socket. */
5913 		nrp = &nmp->nm_sockreq;
5914 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
5915 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5916 	if (error)
5917 		return (error);
5918 	if (nd->nd_repstat == 0) {
5919 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
5920 		NFSLOCKDS(dsp);
5921 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
5922 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
5923 			error = NFSERR_STALEWRITEVERF;
5924 		}
5925 		NFSUNLOCKDS(dsp);
5926 	}
5927 nfsmout:
5928 	if (error == 0 && nd->nd_repstat != 0)
5929 		error = nd->nd_repstat;
5930 	mbuf_freem(nd->nd_mrep);
5931 	return (error);
5932 }
5933 #endif
5934 
5935