xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision 868a59a7e296bed53dd7b986e79964eed8a59f71)
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 
833 	if (nfsboottime.tv_sec == 0)
834 		NFSSETBOOTTIME(nfsboottime);
835 	clp->nfsc_rev = rev++;
836 	if (NFSHASNFSV4N(nmp)) {
837 		error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq,
838 		    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp, cred, p);
839 		NFSCL_DEBUG(1, "aft exch=%d\n", error);
840 		if (error == 0) {
841 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
842 			    &nmp->nm_sockreq,
843 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
844 			if (error == 0) {
845 				NFSLOCKMNT(nmp);
846 				TAILQ_FOREACH_SAFE(tdsp, &nmp->nm_sess,
847 				    nfsclds_list, ndsp)
848 					nfscl_freenfsclds(tdsp);
849 				TAILQ_INIT(&nmp->nm_sess);
850 				TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
851 				    nfsclds_list);
852 				NFSUNLOCKMNT(nmp);
853 			} else
854 				nfscl_freenfsclds(dsp);
855 			NFSCL_DEBUG(1, "aft createsess=%d\n", error);
856 		}
857 		if (error == 0 && reclaim == 0) {
858 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
859 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
860 			if (error == NFSERR_COMPLETEALREADY ||
861 			    error == NFSERR_NOTSUPP)
862 				/* Ignore this error. */
863 				error = 0;
864 		}
865 		return (error);
866 	}
867 
868 	/*
869 	 * Allocate a single session structure for NFSv4.0, because some of
870 	 * the fields are used by NFSv4.0 although it doesn't do a session.
871 	 */
872 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
873 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
874 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
875 	NFSLOCKMNT(nmp);
876 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
877 	NFSUNLOCKMNT(nmp);
878 
879 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL);
880 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
881 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
882 	*tl = txdr_unsigned(clp->nfsc_rev);
883 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
884 
885 	/*
886 	 * set up the callback address
887 	 */
888 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
889 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
890 	callblen = strlen(nfsv4_callbackaddr);
891 	if (callblen == 0)
892 		cp = nfscl_getmyip(nmp, &isinet6);
893 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
894 	    (callblen > 0 || cp != NULL)) {
895 		port = htons(nfsv4_cbport);
896 		cp2 = (u_int8_t *)&port;
897 #ifdef INET6
898 		if ((callblen > 0 &&
899 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
900 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
901 
902 			(void) nfsm_strtom(nd, "tcp6", 4);
903 			if (callblen == 0) {
904 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
905 				ip6add = ip6buf;
906 			} else {
907 				ip6add = nfsv4_callbackaddr;
908 			}
909 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
910 			    ip6add, cp2[0], cp2[1]);
911 		} else
912 #endif
913 		{
914 			(void) nfsm_strtom(nd, "tcp", 3);
915 			if (callblen == 0)
916 				snprintf(addr, INET6_ADDRSTRLEN + 9,
917 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
918 				    cp[2], cp[3], cp2[0], cp2[1]);
919 			else
920 				snprintf(addr, INET6_ADDRSTRLEN + 9,
921 				    "%s.%d.%d", nfsv4_callbackaddr,
922 				    cp2[0], cp2[1]);
923 		}
924 		(void) nfsm_strtom(nd, addr, strlen(addr));
925 	} else {
926 		(void) nfsm_strtom(nd, "tcp", 3);
927 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
928 	}
929 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
930 	*tl = txdr_unsigned(clp->nfsc_cbident);
931 	nd->nd_flag |= ND_USEGSSNAME;
932 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
933 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
934 	if (error)
935 		return (error);
936 	if (nd->nd_repstat == 0) {
937 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
938 	    NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0] = *tl++;
939 	    NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1] = *tl++;
940 	    confirm.lval[0] = *tl++;
941 	    confirm.lval[1] = *tl;
942 	    mbuf_freem(nd->nd_mrep);
943 	    nd->nd_mrep = NULL;
944 
945 	    /*
946 	     * and confirm it.
947 	     */
948 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
949 		NULL);
950 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
951 	    *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
952 	    *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
953 	    *tl++ = confirm.lval[0];
954 	    *tl = confirm.lval[1];
955 	    nd->nd_flag |= ND_USEGSSNAME;
956 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
957 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
958 	    if (error)
959 		return (error);
960 	    mbuf_freem(nd->nd_mrep);
961 	    nd->nd_mrep = NULL;
962 	    if (nd->nd_repstat == 0) {
963 		nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, nmp->nm_fh,
964 		    nmp->nm_fhsize, NULL, NULL);
965 		NFSZERO_ATTRBIT(&attrbits);
966 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
967 		(void) nfsrv_putattrbit(nd, &attrbits);
968 		nd->nd_flag |= ND_USEGSSNAME;
969 		error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
970 		    cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
971 		if (error)
972 		    return (error);
973 		if (nd->nd_repstat == 0) {
974 		    error = nfsv4_loadattr(nd, NULL, NULL, NULL, NULL, 0, NULL,
975 			NULL, NULL, NULL, NULL, 0, NULL, &lease, NULL, p, cred);
976 		    if (error)
977 			goto nfsmout;
978 		    clp->nfsc_renew = NFSCL_RENEW(lease);
979 		    clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew;
980 		    clp->nfsc_clientidrev++;
981 		    if (clp->nfsc_clientidrev == 0)
982 			clp->nfsc_clientidrev++;
983 		}
984 	    }
985 	}
986 	error = nd->nd_repstat;
987 nfsmout:
988 	mbuf_freem(nd->nd_mrep);
989 	return (error);
990 }
991 
992 /*
993  * nfs getattr call.
994  */
995 APPLESTATIC int
996 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
997     struct nfsvattr *nap, void *stuff)
998 {
999 	struct nfsrv_descript nfsd, *nd = &nfsd;
1000 	int error;
1001 	nfsattrbit_t attrbits;
1002 
1003 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1004 	if (nd->nd_flag & ND_NFSV4) {
1005 		NFSGETATTR_ATTRBIT(&attrbits);
1006 		(void) nfsrv_putattrbit(nd, &attrbits);
1007 	}
1008 	error = nfscl_request(nd, vp, p, cred, stuff);
1009 	if (error)
1010 		return (error);
1011 	if (!nd->nd_repstat)
1012 		error = nfsm_loadattr(nd, nap);
1013 	else
1014 		error = nd->nd_repstat;
1015 	mbuf_freem(nd->nd_mrep);
1016 	return (error);
1017 }
1018 
1019 /*
1020  * nfs getattr call with non-vnode arguemnts.
1021  */
1022 APPLESTATIC int
1023 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1024     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1025     uint32_t *leasep)
1026 {
1027 	struct nfsrv_descript nfsd, *nd = &nfsd;
1028 	int error, vers = NFS_VER2;
1029 	nfsattrbit_t attrbits;
1030 
1031 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL);
1032 	if (nd->nd_flag & ND_NFSV4) {
1033 		vers = NFS_VER4;
1034 		NFSGETATTR_ATTRBIT(&attrbits);
1035 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1036 		(void) nfsrv_putattrbit(nd, &attrbits);
1037 	} else if (nd->nd_flag & ND_NFSV3) {
1038 		vers = NFS_VER3;
1039 	}
1040 	if (syscred)
1041 		nd->nd_flag |= ND_USEGSSNAME;
1042 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1043 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1044 	if (error)
1045 		return (error);
1046 	if (nd->nd_repstat == 0) {
1047 		if ((nd->nd_flag & ND_NFSV4) != 0)
1048 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1049 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1050 			    NULL, NULL);
1051 		else
1052 			error = nfsm_loadattr(nd, nap);
1053 	} else
1054 		error = nd->nd_repstat;
1055 	mbuf_freem(nd->nd_mrep);
1056 	return (error);
1057 }
1058 
1059 /*
1060  * Do an nfs setattr operation.
1061  */
1062 APPLESTATIC int
1063 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1064     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1065     void *stuff)
1066 {
1067 	int error, expireret = 0, openerr, retrycnt;
1068 	u_int32_t clidrev = 0, mode;
1069 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1070 	struct nfsfh *nfhp;
1071 	nfsv4stateid_t stateid;
1072 	void *lckp;
1073 
1074 	if (nmp->nm_clp != NULL)
1075 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1076 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1077 		mode = NFSV4OPEN_ACCESSWRITE;
1078 	else
1079 		mode = NFSV4OPEN_ACCESSREAD;
1080 	retrycnt = 0;
1081 	do {
1082 		lckp = NULL;
1083 		openerr = 1;
1084 		if (NFSHASNFSV4(nmp)) {
1085 			nfhp = VTONFS(vp)->n_fhp;
1086 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1087 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1088 			if (error && vnode_vtype(vp) == VREG &&
1089 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1090 			     nfstest_openallsetattr)) {
1091 				/*
1092 				 * No Open stateid, so try and open the file
1093 				 * now.
1094 				 */
1095 				if (mode == NFSV4OPEN_ACCESSWRITE)
1096 					openerr = nfsrpc_open(vp, FWRITE, cred,
1097 					    p);
1098 				else
1099 					openerr = nfsrpc_open(vp, FREAD, cred,
1100 					    p);
1101 				if (!openerr)
1102 					(void) nfscl_getstateid(vp,
1103 					    nfhp->nfh_fh, nfhp->nfh_len,
1104 					    mode, 0, cred, p, &stateid, &lckp);
1105 			}
1106 		}
1107 		if (vap != NULL)
1108 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1109 			    rnap, attrflagp, stuff);
1110 		else
1111 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1112 			    stuff);
1113 		if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
1114 			nfscl_initiate_recovery(nmp->nm_clp);
1115 		if (lckp != NULL)
1116 			nfscl_lockderef(lckp);
1117 		if (!openerr)
1118 			(void) nfsrpc_close(vp, 0, p);
1119 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1120 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1121 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1122 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1123 		} else if ((error == NFSERR_EXPIRED ||
1124 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1125 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1126 		}
1127 		retrycnt++;
1128 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1129 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1130 	    error == NFSERR_BADSESSION ||
1131 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1132 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1133 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1134 	if (error && retrycnt >= 4)
1135 		error = EIO;
1136 	return (error);
1137 }
1138 
1139 static int
1140 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1141     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1142     struct nfsvattr *rnap, int *attrflagp, void *stuff)
1143 {
1144 	u_int32_t *tl;
1145 	struct nfsrv_descript nfsd, *nd = &nfsd;
1146 	int error;
1147 	nfsattrbit_t attrbits;
1148 
1149 	*attrflagp = 0;
1150 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1151 	if (nd->nd_flag & ND_NFSV4)
1152 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1153 	vap->va_type = vnode_vtype(vp);
1154 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1155 	if (nd->nd_flag & ND_NFSV3) {
1156 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1157 		*tl = newnfs_false;
1158 	} else if (nd->nd_flag & ND_NFSV4) {
1159 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1160 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1161 		NFSGETATTR_ATTRBIT(&attrbits);
1162 		(void) nfsrv_putattrbit(nd, &attrbits);
1163 	}
1164 	error = nfscl_request(nd, vp, p, cred, stuff);
1165 	if (error)
1166 		return (error);
1167 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1168 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1169 	if ((nd->nd_flag & ND_NFSV4) && !error)
1170 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1171 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1172 		error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1173 	mbuf_freem(nd->nd_mrep);
1174 	if (nd->nd_repstat && !error)
1175 		error = nd->nd_repstat;
1176 	return (error);
1177 }
1178 
1179 /*
1180  * nfs lookup rpc
1181  */
1182 APPLESTATIC int
1183 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1184     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1185     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff)
1186 {
1187 	u_int32_t *tl;
1188 	struct nfsrv_descript nfsd, *nd = &nfsd;
1189 	struct nfsmount *nmp;
1190 	struct nfsnode *np;
1191 	struct nfsfh *nfhp;
1192 	nfsattrbit_t attrbits;
1193 	int error = 0, lookupp = 0;
1194 
1195 	*attrflagp = 0;
1196 	*dattrflagp = 0;
1197 	if (vnode_vtype(dvp) != VDIR)
1198 		return (ENOTDIR);
1199 	nmp = VFSTONFS(vnode_mount(dvp));
1200 	if (len > NFS_MAXNAMLEN)
1201 		return (ENAMETOOLONG);
1202 	if (NFSHASNFSV4(nmp) && len == 1 &&
1203 		name[0] == '.') {
1204 		/*
1205 		 * Just return the current dir's fh.
1206 		 */
1207 		np = VTONFS(dvp);
1208 		MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) +
1209 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1210 		nfhp->nfh_len = np->n_fhp->nfh_len;
1211 		NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1212 		*nfhpp = nfhp;
1213 		return (0);
1214 	}
1215 	if (NFSHASNFSV4(nmp) && len == 2 &&
1216 		name[0] == '.' && name[1] == '.') {
1217 		lookupp = 1;
1218 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1219 	} else {
1220 		NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1221 		(void) nfsm_strtom(nd, name, len);
1222 	}
1223 	if (nd->nd_flag & ND_NFSV4) {
1224 		NFSGETATTR_ATTRBIT(&attrbits);
1225 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1226 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1227 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1228 		(void) nfsrv_putattrbit(nd, &attrbits);
1229 	}
1230 	error = nfscl_request(nd, dvp, p, cred, stuff);
1231 	if (error)
1232 		return (error);
1233 	if (nd->nd_repstat) {
1234 		/*
1235 		 * When an NFSv4 Lookupp returns ENOENT, it means that
1236 		 * the lookup is at the root of an fs, so return this dir.
1237 		 */
1238 		if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1239 		    np = VTONFS(dvp);
1240 		    MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) +
1241 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1242 		    nfhp->nfh_len = np->n_fhp->nfh_len;
1243 		    NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1244 		    *nfhpp = nfhp;
1245 		    mbuf_freem(nd->nd_mrep);
1246 		    return (0);
1247 		}
1248 		if (nd->nd_flag & ND_NFSV3)
1249 		    error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1250 		else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1251 		    ND_NFSV4) {
1252 			/* Load the directory attributes. */
1253 			error = nfsm_loadattr(nd, dnap);
1254 			if (error == 0)
1255 				*dattrflagp = 1;
1256 		}
1257 		goto nfsmout;
1258 	}
1259 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1260 		/* Load the directory attributes. */
1261 		error = nfsm_loadattr(nd, dnap);
1262 		if (error != 0)
1263 			goto nfsmout;
1264 		*dattrflagp = 1;
1265 		/* Skip over the Lookup and GetFH operation status values. */
1266 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1267 	}
1268 	error = nfsm_getfh(nd, nfhpp);
1269 	if (error)
1270 		goto nfsmout;
1271 
1272 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1273 	if ((nd->nd_flag & ND_NFSV3) && !error)
1274 		error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1275 nfsmout:
1276 	mbuf_freem(nd->nd_mrep);
1277 	if (!error && nd->nd_repstat)
1278 		error = nd->nd_repstat;
1279 	return (error);
1280 }
1281 
1282 /*
1283  * Do a readlink rpc.
1284  */
1285 APPLESTATIC int
1286 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1287     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1288 {
1289 	u_int32_t *tl;
1290 	struct nfsrv_descript nfsd, *nd = &nfsd;
1291 	struct nfsnode *np = VTONFS(vp);
1292 	nfsattrbit_t attrbits;
1293 	int error, len, cangetattr = 1;
1294 
1295 	*attrflagp = 0;
1296 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1297 	if (nd->nd_flag & ND_NFSV4) {
1298 		/*
1299 		 * And do a Getattr op.
1300 		 */
1301 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1302 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1303 		NFSGETATTR_ATTRBIT(&attrbits);
1304 		(void) nfsrv_putattrbit(nd, &attrbits);
1305 	}
1306 	error = nfscl_request(nd, vp, p, cred, stuff);
1307 	if (error)
1308 		return (error);
1309 	if (nd->nd_flag & ND_NFSV3)
1310 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1311 	if (!nd->nd_repstat && !error) {
1312 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1313 		/*
1314 		 * This seems weird to me, but must have been added to
1315 		 * FreeBSD for some reason. The only thing I can think of
1316 		 * is that there was/is some server that replies with
1317 		 * more link data than it should?
1318 		 */
1319 		if (len == NFS_MAXPATHLEN) {
1320 			NFSLOCKNODE(np);
1321 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1322 				len = np->n_size;
1323 				cangetattr = 0;
1324 			}
1325 			NFSUNLOCKNODE(np);
1326 		}
1327 		error = nfsm_mbufuio(nd, uiop, len);
1328 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1329 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1330 	}
1331 	if (nd->nd_repstat && !error)
1332 		error = nd->nd_repstat;
1333 nfsmout:
1334 	mbuf_freem(nd->nd_mrep);
1335 	return (error);
1336 }
1337 
1338 /*
1339  * Read operation.
1340  */
1341 APPLESTATIC int
1342 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1343     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1344 {
1345 	int error, expireret = 0, retrycnt;
1346 	u_int32_t clidrev = 0;
1347 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1348 	struct nfsnode *np = VTONFS(vp);
1349 	struct ucred *newcred;
1350 	struct nfsfh *nfhp = NULL;
1351 	nfsv4stateid_t stateid;
1352 	void *lckp;
1353 
1354 	if (nmp->nm_clp != NULL)
1355 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1356 	newcred = cred;
1357 	if (NFSHASNFSV4(nmp)) {
1358 		nfhp = np->n_fhp;
1359 		newcred = NFSNEWCRED(cred);
1360 	}
1361 	retrycnt = 0;
1362 	do {
1363 		lckp = NULL;
1364 		if (NFSHASNFSV4(nmp))
1365 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1366 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1367 			    &lckp);
1368 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1369 		    attrflagp, stuff);
1370 		if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
1371 			nfscl_initiate_recovery(nmp->nm_clp);
1372 		if (lckp != NULL)
1373 			nfscl_lockderef(lckp);
1374 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1375 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1376 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1377 			(void) nfs_catnap(PZERO, error, "nfs_read");
1378 		} else if ((error == NFSERR_EXPIRED ||
1379 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1380 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1381 		}
1382 		retrycnt++;
1383 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1384 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1385 	    error == NFSERR_BADSESSION ||
1386 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1387 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1388 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1389 	if (error && retrycnt >= 4)
1390 		error = EIO;
1391 	if (NFSHASNFSV4(nmp))
1392 		NFSFREECRED(newcred);
1393 	return (error);
1394 }
1395 
1396 /*
1397  * The actual read RPC.
1398  */
1399 static int
1400 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1401     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1402     int *attrflagp, void *stuff)
1403 {
1404 	u_int32_t *tl;
1405 	int error = 0, len, retlen, tsiz, eof = 0;
1406 	struct nfsrv_descript nfsd;
1407 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1408 	struct nfsrv_descript *nd = &nfsd;
1409 	int rsize;
1410 	off_t tmp_off;
1411 
1412 	*attrflagp = 0;
1413 	tsiz = uio_uio_resid(uiop);
1414 	tmp_off = uiop->uio_offset + tsiz;
1415 	NFSLOCKMNT(nmp);
1416 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1417 		NFSUNLOCKMNT(nmp);
1418 		return (EFBIG);
1419 	}
1420 	rsize = nmp->nm_rsize;
1421 	NFSUNLOCKMNT(nmp);
1422 	nd->nd_mrep = NULL;
1423 	while (tsiz > 0) {
1424 		*attrflagp = 0;
1425 		len = (tsiz > rsize) ? rsize : tsiz;
1426 		NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1427 		if (nd->nd_flag & ND_NFSV4)
1428 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1429 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1430 		if (nd->nd_flag & ND_NFSV2) {
1431 			*tl++ = txdr_unsigned(uiop->uio_offset);
1432 			*tl++ = txdr_unsigned(len);
1433 			*tl = 0;
1434 		} else {
1435 			txdr_hyper(uiop->uio_offset, tl);
1436 			*(tl + 2) = txdr_unsigned(len);
1437 		}
1438 		/*
1439 		 * Since I can't do a Getattr for NFSv4 for Write, there
1440 		 * doesn't seem any point in doing one here, either.
1441 		 * (See the comment in nfsrpc_writerpc() for more info.)
1442 		 */
1443 		error = nfscl_request(nd, vp, p, cred, stuff);
1444 		if (error)
1445 			return (error);
1446 		if (nd->nd_flag & ND_NFSV3) {
1447 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1448 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1449 			error = nfsm_loadattr(nd, nap);
1450 			if (!error)
1451 				*attrflagp = 1;
1452 		}
1453 		if (nd->nd_repstat || error) {
1454 			if (!error)
1455 				error = nd->nd_repstat;
1456 			goto nfsmout;
1457 		}
1458 		if (nd->nd_flag & ND_NFSV3) {
1459 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1460 			eof = fxdr_unsigned(int, *(tl + 1));
1461 		} else if (nd->nd_flag & ND_NFSV4) {
1462 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1463 			eof = fxdr_unsigned(int, *tl);
1464 		}
1465 		NFSM_STRSIZ(retlen, len);
1466 		error = nfsm_mbufuio(nd, uiop, retlen);
1467 		if (error)
1468 			goto nfsmout;
1469 		mbuf_freem(nd->nd_mrep);
1470 		nd->nd_mrep = NULL;
1471 		tsiz -= retlen;
1472 		if (!(nd->nd_flag & ND_NFSV2)) {
1473 			if (eof || retlen == 0)
1474 				tsiz = 0;
1475 		} else if (retlen < len)
1476 			tsiz = 0;
1477 	}
1478 	return (0);
1479 nfsmout:
1480 	if (nd->nd_mrep != NULL)
1481 		mbuf_freem(nd->nd_mrep);
1482 	return (error);
1483 }
1484 
1485 /*
1486  * nfs write operation
1487  * When called_from_strategy != 0, it should return EIO for an error that
1488  * indicates recovery is in progress, so that the buffer will be left
1489  * dirty and be written back to the server later. If it loops around,
1490  * the recovery thread could get stuck waiting for the buffer and recovery
1491  * will then deadlock.
1492  */
1493 APPLESTATIC int
1494 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1495     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1496     void *stuff, int called_from_strategy)
1497 {
1498 	int error, expireret = 0, retrycnt, nostateid;
1499 	u_int32_t clidrev = 0;
1500 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1501 	struct nfsnode *np = VTONFS(vp);
1502 	struct ucred *newcred;
1503 	struct nfsfh *nfhp = NULL;
1504 	nfsv4stateid_t stateid;
1505 	void *lckp;
1506 
1507 	*must_commit = 0;
1508 	if (nmp->nm_clp != NULL)
1509 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1510 	newcred = cred;
1511 	if (NFSHASNFSV4(nmp)) {
1512 		newcred = NFSNEWCRED(cred);
1513 		nfhp = np->n_fhp;
1514 	}
1515 	retrycnt = 0;
1516 	do {
1517 		lckp = NULL;
1518 		nostateid = 0;
1519 		if (NFSHASNFSV4(nmp)) {
1520 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1521 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1522 			    &lckp);
1523 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1524 			    stateid.other[2] == 0) {
1525 				nostateid = 1;
1526 				NFSCL_DEBUG(1, "stateid0 in write\n");
1527 			}
1528 		}
1529 
1530 		/*
1531 		 * If there is no stateid for NFSv4, it means this is an
1532 		 * extraneous write after close. Basically a poorly
1533 		 * implemented buffer cache. Just don't do the write.
1534 		 */
1535 		if (nostateid)
1536 			error = 0;
1537 		else
1538 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1539 			    newcred, &stateid, p, nap, attrflagp, stuff);
1540 		if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
1541 			nfscl_initiate_recovery(nmp->nm_clp);
1542 		if (lckp != NULL)
1543 			nfscl_lockderef(lckp);
1544 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1545 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1546 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1547 			(void) nfs_catnap(PZERO, error, "nfs_write");
1548 		} else if ((error == NFSERR_EXPIRED ||
1549 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1550 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1551 		}
1552 		retrycnt++;
1553 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1554 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1555 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1556 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1557 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1558 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1559 	if (error != 0 && (retrycnt >= 4 ||
1560 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1561 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1562 		error = EIO;
1563 	if (NFSHASNFSV4(nmp))
1564 		NFSFREECRED(newcred);
1565 	return (error);
1566 }
1567 
1568 /*
1569  * The actual write RPC.
1570  */
1571 static int
1572 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1573     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1574     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1575 {
1576 	u_int32_t *tl;
1577 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1578 	struct nfsnode *np = VTONFS(vp);
1579 	int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1580 	int wccflag = 0, wsize;
1581 	int32_t backup;
1582 	struct nfsrv_descript nfsd;
1583 	struct nfsrv_descript *nd = &nfsd;
1584 	nfsattrbit_t attrbits;
1585 	off_t tmp_off;
1586 
1587 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1588 	*attrflagp = 0;
1589 	tsiz = uio_uio_resid(uiop);
1590 	tmp_off = uiop->uio_offset + tsiz;
1591 	NFSLOCKMNT(nmp);
1592 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1593 		NFSUNLOCKMNT(nmp);
1594 		return (EFBIG);
1595 	}
1596 	wsize = nmp->nm_wsize;
1597 	NFSUNLOCKMNT(nmp);
1598 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
1599 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
1600 	while (tsiz > 0) {
1601 		*attrflagp = 0;
1602 		len = (tsiz > wsize) ? wsize : tsiz;
1603 		NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1604 		if (nd->nd_flag & ND_NFSV4) {
1605 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1606 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1607 			txdr_hyper(uiop->uio_offset, tl);
1608 			tl += 2;
1609 			*tl++ = txdr_unsigned(*iomode);
1610 			*tl = txdr_unsigned(len);
1611 		} else if (nd->nd_flag & ND_NFSV3) {
1612 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1613 			txdr_hyper(uiop->uio_offset, tl);
1614 			tl += 2;
1615 			*tl++ = txdr_unsigned(len);
1616 			*tl++ = txdr_unsigned(*iomode);
1617 			*tl = txdr_unsigned(len);
1618 		} else {
1619 			u_int32_t x;
1620 
1621 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1622 			/*
1623 			 * Not sure why someone changed this, since the
1624 			 * RFC clearly states that "beginoffset" and
1625 			 * "totalcount" are ignored, but it wouldn't
1626 			 * surprise me if there's a busted server out there.
1627 			 */
1628 			/* Set both "begin" and "current" to non-garbage. */
1629 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1630 			*tl++ = x;      /* "begin offset" */
1631 			*tl++ = x;      /* "current offset" */
1632 			x = txdr_unsigned(len);
1633 			*tl++ = x;      /* total to this offset */
1634 			*tl = x;        /* size of this write */
1635 
1636 		}
1637 		nfsm_uiombuf(nd, uiop, len);
1638 		/*
1639 		 * Although it is tempting to do a normal Getattr Op in the
1640 		 * NFSv4 compound, the result can be a nearly hung client
1641 		 * system if the Getattr asks for Owner and/or OwnerGroup.
1642 		 * It occurs when the client can't map either the Owner or
1643 		 * Owner_group name in the Getattr reply to a uid/gid. When
1644 		 * there is a cache miss, the kernel does an upcall to the
1645 		 * nfsuserd. Then, it can try and read the local /etc/passwd
1646 		 * or /etc/group file. It can then block in getnewbuf(),
1647 		 * waiting for dirty writes to be pushed to the NFS server.
1648 		 * The only reason this doesn't result in a complete
1649 		 * deadlock, is that the upcall times out and allows
1650 		 * the write to complete. However, progress is so slow
1651 		 * that it might just as well be deadlocked.
1652 		 * As such, we get the rest of the attributes, but not
1653 		 * Owner or Owner_group.
1654 		 * nb: nfscl_loadattrcache() needs to be told that these
1655 		 *     partial attributes from a write rpc are being
1656 		 *     passed in, via a argument flag.
1657 		 */
1658 		if (nd->nd_flag & ND_NFSV4) {
1659 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
1660 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1661 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
1662 			(void) nfsrv_putattrbit(nd, &attrbits);
1663 		}
1664 		error = nfscl_request(nd, vp, p, cred, stuff);
1665 		if (error)
1666 			return (error);
1667 		if (nd->nd_repstat) {
1668 			/*
1669 			 * In case the rpc gets retried, roll
1670 			 * the uio fileds changed by nfsm_uiombuf()
1671 			 * back.
1672 			 */
1673 			uiop->uio_offset -= len;
1674 			uio_uio_resid_add(uiop, len);
1675 			uio_iov_base_add(uiop, -len);
1676 			uio_iov_len_add(uiop, len);
1677 		}
1678 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1679 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1680 			    &wccflag, stuff);
1681 			if (error)
1682 				goto nfsmout;
1683 		}
1684 		if (!nd->nd_repstat) {
1685 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1686 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1687 					+ NFSX_VERF);
1688 				rlen = fxdr_unsigned(int, *tl++);
1689 				if (rlen == 0) {
1690 					error = NFSERR_IO;
1691 					goto nfsmout;
1692 				} else if (rlen < len) {
1693 					backup = len - rlen;
1694 					uio_iov_base_add(uiop, -(backup));
1695 					uio_iov_len_add(uiop, backup);
1696 					uiop->uio_offset -= backup;
1697 					uio_uio_resid_add(uiop, backup);
1698 					len = rlen;
1699 				}
1700 				commit = fxdr_unsigned(int, *tl++);
1701 
1702 				/*
1703 				 * Return the lowest committment level
1704 				 * obtained by any of the RPCs.
1705 				 */
1706 				if (committed == NFSWRITE_FILESYNC)
1707 					committed = commit;
1708 				else if (committed == NFSWRITE_DATASYNC &&
1709 					commit == NFSWRITE_UNSTABLE)
1710 					committed = commit;
1711 				NFSLOCKMNT(nmp);
1712 				if (!NFSHASWRITEVERF(nmp)) {
1713 					NFSBCOPY((caddr_t)tl,
1714 					    (caddr_t)&nmp->nm_verf[0],
1715 					    NFSX_VERF);
1716 					NFSSETWRITEVERF(nmp);
1717 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
1718 				    NFSX_VERF)) {
1719 					*must_commit = 1;
1720 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1721 				}
1722 				NFSUNLOCKMNT(nmp);
1723 			}
1724 			if (nd->nd_flag & ND_NFSV4)
1725 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1726 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1727 				error = nfsm_loadattr(nd, nap);
1728 				if (!error)
1729 					*attrflagp = NFS_LATTR_NOSHRINK;
1730 			}
1731 		} else {
1732 			error = nd->nd_repstat;
1733 		}
1734 		if (error)
1735 			goto nfsmout;
1736 		NFSWRITERPC_SETTIME(wccflag, np, (nd->nd_flag & ND_NFSV4));
1737 		mbuf_freem(nd->nd_mrep);
1738 		nd->nd_mrep = NULL;
1739 		tsiz -= len;
1740 	}
1741 nfsmout:
1742 	if (nd->nd_mrep != NULL)
1743 		mbuf_freem(nd->nd_mrep);
1744 	*iomode = committed;
1745 	if (nd->nd_repstat && !error)
1746 		error = nd->nd_repstat;
1747 	return (error);
1748 }
1749 
1750 /*
1751  * nfs mknod rpc
1752  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1753  * mode set to specify the file type and the size field for rdev.
1754  */
1755 APPLESTATIC int
1756 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1757     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1758     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1759     int *attrflagp, int *dattrflagp, void *dstuff)
1760 {
1761 	u_int32_t *tl;
1762 	int error = 0;
1763 	struct nfsrv_descript nfsd, *nd = &nfsd;
1764 	nfsattrbit_t attrbits;
1765 
1766 	*nfhpp = NULL;
1767 	*attrflagp = 0;
1768 	*dattrflagp = 0;
1769 	if (namelen > NFS_MAXNAMLEN)
1770 		return (ENAMETOOLONG);
1771 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1772 	if (nd->nd_flag & ND_NFSV4) {
1773 		if (vtyp == VBLK || vtyp == VCHR) {
1774 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1775 			*tl++ = vtonfsv34_type(vtyp);
1776 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
1777 			*tl = txdr_unsigned(NFSMINOR(rdev));
1778 		} else {
1779 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1780 			*tl = vtonfsv34_type(vtyp);
1781 		}
1782 	}
1783 	(void) nfsm_strtom(nd, name, namelen);
1784 	if (nd->nd_flag & ND_NFSV3) {
1785 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1786 		*tl = vtonfsv34_type(vtyp);
1787 	}
1788 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1789 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
1790 	if ((nd->nd_flag & ND_NFSV3) &&
1791 	    (vtyp == VCHR || vtyp == VBLK)) {
1792 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1793 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
1794 		*tl = txdr_unsigned(NFSMINOR(rdev));
1795 	}
1796 	if (nd->nd_flag & ND_NFSV4) {
1797 		NFSGETATTR_ATTRBIT(&attrbits);
1798 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1799 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1800 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1801 		(void) nfsrv_putattrbit(nd, &attrbits);
1802 	}
1803 	if (nd->nd_flag & ND_NFSV2)
1804 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
1805 	error = nfscl_request(nd, dvp, p, cred, dstuff);
1806 	if (error)
1807 		return (error);
1808 	if (nd->nd_flag & ND_NFSV4)
1809 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1810 	if (!nd->nd_repstat) {
1811 		if (nd->nd_flag & ND_NFSV4) {
1812 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1813 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1814 			if (error)
1815 				goto nfsmout;
1816 		}
1817 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
1818 		if (error)
1819 			goto nfsmout;
1820 	}
1821 	if (nd->nd_flag & ND_NFSV3)
1822 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1823 	if (!error && nd->nd_repstat)
1824 		error = nd->nd_repstat;
1825 nfsmout:
1826 	mbuf_freem(nd->nd_mrep);
1827 	return (error);
1828 }
1829 
1830 /*
1831  * nfs file create call
1832  * Mostly just call the approriate routine. (I separated out v4, so that
1833  * error recovery wouldn't be as difficult.)
1834  */
1835 APPLESTATIC int
1836 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1837     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
1838     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1839     int *attrflagp, int *dattrflagp, void *dstuff)
1840 {
1841 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
1842 	struct nfsclowner *owp;
1843 	struct nfscldeleg *dp;
1844 	struct nfsmount *nmp = VFSTONFS(vnode_mount(dvp));
1845 	u_int32_t clidrev;
1846 
1847 	if (NFSHASNFSV4(nmp)) {
1848 	    retrycnt = 0;
1849 	    do {
1850 		dp = NULL;
1851 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
1852 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
1853 		    NULL, 1);
1854 		if (error)
1855 			return (error);
1856 		if (nmp->nm_clp != NULL)
1857 			clidrev = nmp->nm_clp->nfsc_clientidrev;
1858 		else
1859 			clidrev = 0;
1860 		error = nfsrpc_createv4(dvp, name, namelen, vap, cverf, fmode,
1861 		  owp, &dp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
1862 		  dstuff, &unlocked);
1863 		/*
1864 		 * There is no need to invalidate cached attributes here,
1865 		 * since new post-delegation issue attributes are always
1866 		 * returned by nfsrpc_createv4() and these will update the
1867 		 * attribute cache.
1868 		 */
1869 		if (dp != NULL)
1870 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
1871 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
1872 		nfscl_ownerrelease(owp, error, newone, unlocked);
1873 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
1874 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1875 		    error == NFSERR_BADSESSION) {
1876 			(void) nfs_catnap(PZERO, error, "nfs_open");
1877 		} else if ((error == NFSERR_EXPIRED ||
1878 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1879 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1880 			retrycnt++;
1881 		}
1882 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
1883 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1884 		error == NFSERR_BADSESSION ||
1885 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1886 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
1887 	    if (error && retrycnt >= 4)
1888 		    error = EIO;
1889 	} else {
1890 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
1891 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
1892 		    dstuff);
1893 	}
1894 	return (error);
1895 }
1896 
1897 /*
1898  * The create rpc for v2 and 3.
1899  */
1900 static int
1901 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1902     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
1903     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1904     int *attrflagp, int *dattrflagp, void *dstuff)
1905 {
1906 	u_int32_t *tl;
1907 	int error = 0;
1908 	struct nfsrv_descript nfsd, *nd = &nfsd;
1909 
1910 	*nfhpp = NULL;
1911 	*attrflagp = 0;
1912 	*dattrflagp = 0;
1913 	if (namelen > NFS_MAXNAMLEN)
1914 		return (ENAMETOOLONG);
1915 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
1916 	(void) nfsm_strtom(nd, name, namelen);
1917 	if (nd->nd_flag & ND_NFSV3) {
1918 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1919 		if (fmode & O_EXCL) {
1920 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
1921 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
1922 			*tl++ = cverf.lval[0];
1923 			*tl = cverf.lval[1];
1924 		} else {
1925 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
1926 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
1927 		}
1928 	} else {
1929 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
1930 	}
1931 	error = nfscl_request(nd, dvp, p, cred, dstuff);
1932 	if (error)
1933 		return (error);
1934 	if (nd->nd_repstat == 0) {
1935 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
1936 		if (error)
1937 			goto nfsmout;
1938 	}
1939 	if (nd->nd_flag & ND_NFSV3)
1940 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1941 	if (nd->nd_repstat != 0 && error == 0)
1942 		error = nd->nd_repstat;
1943 nfsmout:
1944 	mbuf_freem(nd->nd_mrep);
1945 	return (error);
1946 }
1947 
1948 static int
1949 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1950     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
1951     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
1952     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
1953     int *dattrflagp, void *dstuff, int *unlockedp)
1954 {
1955 	u_int32_t *tl;
1956 	int error = 0, deleg, newone, ret, acesize, limitby;
1957 	struct nfsrv_descript nfsd, *nd = &nfsd;
1958 	struct nfsclopen *op;
1959 	struct nfscldeleg *dp = NULL;
1960 	struct nfsnode *np;
1961 	struct nfsfh *nfhp;
1962 	nfsattrbit_t attrbits;
1963 	nfsv4stateid_t stateid;
1964 	u_int32_t rflags;
1965 	struct nfsmount *nmp;
1966 
1967 	nmp = VFSTONFS(dvp->v_mount);
1968 	np = VTONFS(dvp);
1969 	*unlockedp = 0;
1970 	*nfhpp = NULL;
1971 	*dpp = NULL;
1972 	*attrflagp = 0;
1973 	*dattrflagp = 0;
1974 	if (namelen > NFS_MAXNAMLEN)
1975 		return (ENAMETOOLONG);
1976 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
1977 	/*
1978 	 * For V4, this is actually an Open op.
1979 	 */
1980 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1981 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
1982 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
1983 	    NFSV4OPEN_ACCESSREAD);
1984 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
1985 	*tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
1986 	*tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
1987 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
1988 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1989 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
1990 	if (fmode & O_EXCL) {
1991 		if (NFSHASNFSV4N(nmp)) {
1992 			if (NFSHASSESSPERSIST(nmp)) {
1993 				/* Use GUARDED for persistent sessions. */
1994 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
1995 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
1996 			} else {
1997 				/* Otherwise, use EXCLUSIVE4_1. */
1998 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
1999 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2000 				*tl++ = cverf.lval[0];
2001 				*tl = cverf.lval[1];
2002 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2003 			}
2004 		} else {
2005 			/* NFSv4.0 */
2006 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2007 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2008 			*tl++ = cverf.lval[0];
2009 			*tl = cverf.lval[1];
2010 		}
2011 	} else {
2012 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2013 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2014 	}
2015 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2016 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2017 	(void) nfsm_strtom(nd, name, namelen);
2018 	/* Get the new file's handle and attributes. */
2019 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2020 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2021 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2022 	NFSGETATTR_ATTRBIT(&attrbits);
2023 	(void) nfsrv_putattrbit(nd, &attrbits);
2024 	/* Get the directory's post-op attributes. */
2025 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2026 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2027 	(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2028 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2029 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2030 	(void) nfsrv_putattrbit(nd, &attrbits);
2031 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2032 	if (error)
2033 		return (error);
2034 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2035 	if (nd->nd_repstat == 0) {
2036 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2037 		    6 * NFSX_UNSIGNED);
2038 		stateid.seqid = *tl++;
2039 		stateid.other[0] = *tl++;
2040 		stateid.other[1] = *tl++;
2041 		stateid.other[2] = *tl;
2042 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2043 		(void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2044 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2045 		deleg = fxdr_unsigned(int, *tl);
2046 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2047 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2048 			if (!(owp->nfsow_clp->nfsc_flags &
2049 			      NFSCLFLAGS_FIRSTDELEG))
2050 				owp->nfsow_clp->nfsc_flags |=
2051 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2052 			MALLOC(dp, struct nfscldeleg *,
2053 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2054 			    M_NFSCLDELEG, M_WAITOK);
2055 			LIST_INIT(&dp->nfsdl_owner);
2056 			LIST_INIT(&dp->nfsdl_lock);
2057 			dp->nfsdl_clp = owp->nfsow_clp;
2058 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2059 			nfscl_lockinit(&dp->nfsdl_rwlock);
2060 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2061 			    NFSX_UNSIGNED);
2062 			dp->nfsdl_stateid.seqid = *tl++;
2063 			dp->nfsdl_stateid.other[0] = *tl++;
2064 			dp->nfsdl_stateid.other[1] = *tl++;
2065 			dp->nfsdl_stateid.other[2] = *tl++;
2066 			ret = fxdr_unsigned(int, *tl);
2067 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2068 				dp->nfsdl_flags = NFSCLDL_WRITE;
2069 				/*
2070 				 * Indicates how much the file can grow.
2071 				 */
2072 				NFSM_DISSECT(tl, u_int32_t *,
2073 				    3 * NFSX_UNSIGNED);
2074 				limitby = fxdr_unsigned(int, *tl++);
2075 				switch (limitby) {
2076 				case NFSV4OPEN_LIMITSIZE:
2077 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2078 					break;
2079 				case NFSV4OPEN_LIMITBLOCKS:
2080 					dp->nfsdl_sizelimit =
2081 					    fxdr_unsigned(u_int64_t, *tl++);
2082 					dp->nfsdl_sizelimit *=
2083 					    fxdr_unsigned(u_int64_t, *tl);
2084 					break;
2085 				default:
2086 					error = NFSERR_BADXDR;
2087 					goto nfsmout;
2088 				};
2089 			} else {
2090 				dp->nfsdl_flags = NFSCLDL_READ;
2091 			}
2092 			if (ret)
2093 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2094 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2095 			    &acesize, p);
2096 			if (error)
2097 				goto nfsmout;
2098 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2099 			error = NFSERR_BADXDR;
2100 			goto nfsmout;
2101 		}
2102 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2103 		if (error)
2104 			goto nfsmout;
2105 		/* Get rid of the PutFH and Getattr status values. */
2106 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2107 		/* Load the directory attributes. */
2108 		error = nfsm_loadattr(nd, dnap);
2109 		if (error)
2110 			goto nfsmout;
2111 		*dattrflagp = 1;
2112 		if (dp != NULL && *attrflagp) {
2113 			dp->nfsdl_change = nnap->na_filerev;
2114 			dp->nfsdl_modtime = nnap->na_mtime;
2115 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2116 		}
2117 		/*
2118 		 * We can now complete the Open state.
2119 		 */
2120 		nfhp = *nfhpp;
2121 		if (dp != NULL) {
2122 			dp->nfsdl_fhlen = nfhp->nfh_len;
2123 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2124 		}
2125 		/*
2126 		 * Get an Open structure that will be
2127 		 * attached to the OpenOwner, acquired already.
2128 		 */
2129 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2130 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2131 		    cred, p, NULL, &op, &newone, NULL, 0);
2132 		if (error)
2133 			goto nfsmout;
2134 		op->nfso_stateid = stateid;
2135 		newnfs_copyincred(cred, &op->nfso_cred);
2136 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2137 		    do {
2138 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2139 			    nfhp->nfh_len, op, cred, p);
2140 			if (ret == NFSERR_DELAY)
2141 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2142 		    } while (ret == NFSERR_DELAY);
2143 		    error = ret;
2144 		}
2145 
2146 		/*
2147 		 * If the server is handing out delegations, but we didn't
2148 		 * get one because an OpenConfirm was required, try the
2149 		 * Open again, to get a delegation. This is a harmless no-op,
2150 		 * from a server's point of view.
2151 		 */
2152 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2153 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2154 		    !error && dp == NULL) {
2155 		    do {
2156 			ret = nfsrpc_openrpc(VFSTONFS(vnode_mount(dvp)), dvp,
2157 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2158 			    nfhp->nfh_fh, nfhp->nfh_len,
2159 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2160 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2161 			if (ret == NFSERR_DELAY)
2162 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2163 		    } while (ret == NFSERR_DELAY);
2164 		    if (ret) {
2165 			if (dp != NULL) {
2166 				FREE((caddr_t)dp, M_NFSCLDELEG);
2167 				dp = NULL;
2168 			}
2169 			if (ret == NFSERR_STALECLIENTID ||
2170 			    ret == NFSERR_STALEDONTRECOVER ||
2171 			    ret == NFSERR_BADSESSION)
2172 				error = ret;
2173 		    }
2174 		}
2175 		nfscl_openrelease(op, error, newone);
2176 		*unlockedp = 1;
2177 	}
2178 	if (nd->nd_repstat != 0 && error == 0)
2179 		error = nd->nd_repstat;
2180 	if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
2181 		nfscl_initiate_recovery(owp->nfsow_clp);
2182 nfsmout:
2183 	if (!error)
2184 		*dpp = dp;
2185 	else if (dp != NULL)
2186 		FREE((caddr_t)dp, M_NFSCLDELEG);
2187 	mbuf_freem(nd->nd_mrep);
2188 	return (error);
2189 }
2190 
2191 /*
2192  * Nfs remove rpc
2193  */
2194 APPLESTATIC int
2195 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2196     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2197     void *dstuff)
2198 {
2199 	u_int32_t *tl;
2200 	struct nfsrv_descript nfsd, *nd = &nfsd;
2201 	struct nfsnode *np;
2202 	struct nfsmount *nmp;
2203 	nfsv4stateid_t dstateid;
2204 	int error, ret = 0, i;
2205 
2206 	*dattrflagp = 0;
2207 	if (namelen > NFS_MAXNAMLEN)
2208 		return (ENAMETOOLONG);
2209 	nmp = VFSTONFS(vnode_mount(dvp));
2210 tryagain:
2211 	if (NFSHASNFSV4(nmp) && ret == 0) {
2212 		ret = nfscl_removedeleg(vp, p, &dstateid);
2213 		if (ret == 1) {
2214 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2215 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2216 			    NFSX_UNSIGNED);
2217 			if (NFSHASNFSV4N(nmp))
2218 				*tl++ = 0;
2219 			else
2220 				*tl++ = dstateid.seqid;
2221 			*tl++ = dstateid.other[0];
2222 			*tl++ = dstateid.other[1];
2223 			*tl++ = dstateid.other[2];
2224 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2225 			np = VTONFS(dvp);
2226 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2227 			    np->n_fhp->nfh_len, 0);
2228 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2229 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2230 		}
2231 	} else {
2232 		ret = 0;
2233 	}
2234 	if (ret == 0)
2235 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2236 	(void) nfsm_strtom(nd, name, namelen);
2237 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2238 	if (error)
2239 		return (error);
2240 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2241 		/* For NFSv4, parse out any Delereturn replies. */
2242 		if (ret > 0 && nd->nd_repstat != 0 &&
2243 		    (nd->nd_flag & ND_NOMOREDATA)) {
2244 			/*
2245 			 * If the Delegreturn failed, try again without
2246 			 * it. The server will Recall, as required.
2247 			 */
2248 			mbuf_freem(nd->nd_mrep);
2249 			goto tryagain;
2250 		}
2251 		for (i = 0; i < (ret * 2); i++) {
2252 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2253 			    ND_NFSV4) {
2254 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2255 			    if (*(tl + 1))
2256 				nd->nd_flag |= ND_NOMOREDATA;
2257 			}
2258 		}
2259 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2260 	}
2261 	if (nd->nd_repstat && !error)
2262 		error = nd->nd_repstat;
2263 nfsmout:
2264 	mbuf_freem(nd->nd_mrep);
2265 	return (error);
2266 }
2267 
2268 /*
2269  * Do an nfs rename rpc.
2270  */
2271 APPLESTATIC int
2272 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2273     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2274     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2275     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2276 {
2277 	u_int32_t *tl;
2278 	struct nfsrv_descript nfsd, *nd = &nfsd;
2279 	struct nfsmount *nmp;
2280 	struct nfsnode *np;
2281 	nfsattrbit_t attrbits;
2282 	nfsv4stateid_t fdstateid, tdstateid;
2283 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2284 
2285 	*fattrflagp = 0;
2286 	*tattrflagp = 0;
2287 	nmp = VFSTONFS(vnode_mount(fdvp));
2288 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2289 		return (ENAMETOOLONG);
2290 tryagain:
2291 	if (NFSHASNFSV4(nmp) && ret == 0) {
2292 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2293 		    &tdstateid, &gottd, p);
2294 		if (gotfd && gottd) {
2295 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2296 		} else if (gotfd) {
2297 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2298 		} else if (gottd) {
2299 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2300 		}
2301 		if (gotfd) {
2302 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2303 			if (NFSHASNFSV4N(nmp))
2304 				*tl++ = 0;
2305 			else
2306 				*tl++ = fdstateid.seqid;
2307 			*tl++ = fdstateid.other[0];
2308 			*tl++ = fdstateid.other[1];
2309 			*tl = fdstateid.other[2];
2310 			if (gottd) {
2311 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2312 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2313 				np = VTONFS(tvp);
2314 				(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2315 				    np->n_fhp->nfh_len, 0);
2316 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2317 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2318 			}
2319 		}
2320 		if (gottd) {
2321 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2322 			if (NFSHASNFSV4N(nmp))
2323 				*tl++ = 0;
2324 			else
2325 				*tl++ = tdstateid.seqid;
2326 			*tl++ = tdstateid.other[0];
2327 			*tl++ = tdstateid.other[1];
2328 			*tl = tdstateid.other[2];
2329 		}
2330 		if (ret > 0) {
2331 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2332 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2333 			np = VTONFS(fdvp);
2334 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2335 			    np->n_fhp->nfh_len, 0);
2336 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2337 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2338 		}
2339 	} else {
2340 		ret = 0;
2341 	}
2342 	if (ret == 0)
2343 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2344 	if (nd->nd_flag & ND_NFSV4) {
2345 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2346 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2347 		NFSWCCATTR_ATTRBIT(&attrbits);
2348 		(void) nfsrv_putattrbit(nd, &attrbits);
2349 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2350 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2351 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2352 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2353 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2354 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2355 		(void) nfsrv_putattrbit(nd, &attrbits);
2356 		nd->nd_flag |= ND_V4WCCATTR;
2357 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2358 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2359 	}
2360 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2361 	if (!(nd->nd_flag & ND_NFSV4))
2362 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2363 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2364 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2365 	error = nfscl_request(nd, fdvp, p, cred, fstuff);
2366 	if (error)
2367 		return (error);
2368 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2369 		/* For NFSv4, parse out any Delereturn replies. */
2370 		if (ret > 0 && nd->nd_repstat != 0 &&
2371 		    (nd->nd_flag & ND_NOMOREDATA)) {
2372 			/*
2373 			 * If the Delegreturn failed, try again without
2374 			 * it. The server will Recall, as required.
2375 			 */
2376 			mbuf_freem(nd->nd_mrep);
2377 			goto tryagain;
2378 		}
2379 		for (i = 0; i < (ret * 2); i++) {
2380 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2381 			    ND_NFSV4) {
2382 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2383 			    if (*(tl + 1)) {
2384 				if (i == 0 && ret > 1) {
2385 				    /*
2386 				     * If the Delegreturn failed, try again
2387 				     * without it. The server will Recall, as
2388 				     * required.
2389 				     * If ret > 1, the first iteration of this
2390 				     * loop is the second DelegReturn result.
2391 				     */
2392 				    mbuf_freem(nd->nd_mrep);
2393 				    goto tryagain;
2394 				} else {
2395 				    nd->nd_flag |= ND_NOMOREDATA;
2396 				}
2397 			    }
2398 			}
2399 		}
2400 		/* Now, the first wcc attribute reply. */
2401 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2402 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2403 			if (*(tl + 1))
2404 				nd->nd_flag |= ND_NOMOREDATA;
2405 		}
2406 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2407 		    fstuff);
2408 		/* and the second wcc attribute reply. */
2409 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2410 		    !error) {
2411 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2412 			if (*(tl + 1))
2413 				nd->nd_flag |= ND_NOMOREDATA;
2414 		}
2415 		if (!error)
2416 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2417 			    NULL, tstuff);
2418 	}
2419 	if (nd->nd_repstat && !error)
2420 		error = nd->nd_repstat;
2421 nfsmout:
2422 	mbuf_freem(nd->nd_mrep);
2423 	return (error);
2424 }
2425 
2426 /*
2427  * nfs hard link create rpc
2428  */
2429 APPLESTATIC int
2430 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2431     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2432     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2433 {
2434 	u_int32_t *tl;
2435 	struct nfsrv_descript nfsd, *nd = &nfsd;
2436 	nfsattrbit_t attrbits;
2437 	int error = 0;
2438 
2439 	*attrflagp = 0;
2440 	*dattrflagp = 0;
2441 	if (namelen > NFS_MAXNAMLEN)
2442 		return (ENAMETOOLONG);
2443 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2444 	if (nd->nd_flag & ND_NFSV4) {
2445 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2446 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2447 	}
2448 	(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2449 		VTONFS(dvp)->n_fhp->nfh_len, 0);
2450 	if (nd->nd_flag & ND_NFSV4) {
2451 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2452 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2453 		NFSWCCATTR_ATTRBIT(&attrbits);
2454 		(void) nfsrv_putattrbit(nd, &attrbits);
2455 		nd->nd_flag |= ND_V4WCCATTR;
2456 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2457 		*tl = txdr_unsigned(NFSV4OP_LINK);
2458 	}
2459 	(void) nfsm_strtom(nd, name, namelen);
2460 	error = nfscl_request(nd, vp, p, cred, dstuff);
2461 	if (error)
2462 		return (error);
2463 	if (nd->nd_flag & ND_NFSV3) {
2464 		error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2465 		if (!error)
2466 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2467 			    NULL, dstuff);
2468 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2469 		/*
2470 		 * First, parse out the PutFH and Getattr result.
2471 		 */
2472 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2473 		if (!(*(tl + 1)))
2474 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2475 		if (*(tl + 1))
2476 			nd->nd_flag |= ND_NOMOREDATA;
2477 		/*
2478 		 * Get the pre-op attributes.
2479 		 */
2480 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2481 	}
2482 	if (nd->nd_repstat && !error)
2483 		error = nd->nd_repstat;
2484 nfsmout:
2485 	mbuf_freem(nd->nd_mrep);
2486 	return (error);
2487 }
2488 
2489 /*
2490  * nfs symbolic link create rpc
2491  */
2492 APPLESTATIC int
2493 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, char *target,
2494     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2495     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2496     int *dattrflagp, void *dstuff)
2497 {
2498 	u_int32_t *tl;
2499 	struct nfsrv_descript nfsd, *nd = &nfsd;
2500 	struct nfsmount *nmp;
2501 	int slen, error = 0;
2502 
2503 	*nfhpp = NULL;
2504 	*attrflagp = 0;
2505 	*dattrflagp = 0;
2506 	nmp = VFSTONFS(vnode_mount(dvp));
2507 	slen = strlen(target);
2508 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2509 		return (ENAMETOOLONG);
2510 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2511 	if (nd->nd_flag & ND_NFSV4) {
2512 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2513 		*tl = txdr_unsigned(NFLNK);
2514 		(void) nfsm_strtom(nd, target, slen);
2515 	}
2516 	(void) nfsm_strtom(nd, name, namelen);
2517 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2518 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2519 	if (!(nd->nd_flag & ND_NFSV4))
2520 		(void) nfsm_strtom(nd, target, slen);
2521 	if (nd->nd_flag & ND_NFSV2)
2522 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2523 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2524 	if (error)
2525 		return (error);
2526 	if (nd->nd_flag & ND_NFSV4)
2527 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2528 	if ((nd->nd_flag & ND_NFSV3) && !error) {
2529 		if (!nd->nd_repstat)
2530 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2531 		if (!error)
2532 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2533 			    NULL, dstuff);
2534 	}
2535 	if (nd->nd_repstat && !error)
2536 		error = nd->nd_repstat;
2537 	mbuf_freem(nd->nd_mrep);
2538 	/*
2539 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2540 	 * Only do this if vfs.nfs.ignore_eexist is set.
2541 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2542 	 * should guarantee "exactly once" RPC semantics.
2543 	 */
2544 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2545 	    nmp->nm_minorvers == 0))
2546 		error = 0;
2547 	return (error);
2548 }
2549 
2550 /*
2551  * nfs make dir rpc
2552  */
2553 APPLESTATIC int
2554 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2555     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2556     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2557     int *dattrflagp, void *dstuff)
2558 {
2559 	u_int32_t *tl;
2560 	struct nfsrv_descript nfsd, *nd = &nfsd;
2561 	nfsattrbit_t attrbits;
2562 	int error = 0;
2563 	struct nfsfh *fhp;
2564 	struct nfsmount *nmp;
2565 
2566 	*nfhpp = NULL;
2567 	*attrflagp = 0;
2568 	*dattrflagp = 0;
2569 	nmp = VFSTONFS(vnode_mount(dvp));
2570 	fhp = VTONFS(dvp)->n_fhp;
2571 	if (namelen > NFS_MAXNAMLEN)
2572 		return (ENAMETOOLONG);
2573 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2574 	if (nd->nd_flag & ND_NFSV4) {
2575 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2576 		*tl = txdr_unsigned(NFDIR);
2577 	}
2578 	(void) nfsm_strtom(nd, name, namelen);
2579 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2580 	if (nd->nd_flag & ND_NFSV4) {
2581 		NFSGETATTR_ATTRBIT(&attrbits);
2582 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2583 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2584 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2585 		(void) nfsrv_putattrbit(nd, &attrbits);
2586 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2587 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2588 		(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2589 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2590 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2591 		(void) nfsrv_putattrbit(nd, &attrbits);
2592 	}
2593 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2594 	if (error)
2595 		return (error);
2596 	if (nd->nd_flag & ND_NFSV4)
2597 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2598 	if (!nd->nd_repstat && !error) {
2599 		if (nd->nd_flag & ND_NFSV4) {
2600 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2601 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2602 		}
2603 		if (!error)
2604 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2605 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2606 			/* Get rid of the PutFH and Getattr status values. */
2607 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2608 			/* Load the directory attributes. */
2609 			error = nfsm_loadattr(nd, dnap);
2610 			if (error == 0)
2611 				*dattrflagp = 1;
2612 		}
2613 	}
2614 	if ((nd->nd_flag & ND_NFSV3) && !error)
2615 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2616 	if (nd->nd_repstat && !error)
2617 		error = nd->nd_repstat;
2618 nfsmout:
2619 	mbuf_freem(nd->nd_mrep);
2620 	/*
2621 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2622 	 * Only do this if vfs.nfs.ignore_eexist is set.
2623 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2624 	 * should guarantee "exactly once" RPC semantics.
2625 	 */
2626 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2627 	    nmp->nm_minorvers == 0))
2628 		error = 0;
2629 	return (error);
2630 }
2631 
2632 /*
2633  * nfs remove directory call
2634  */
2635 APPLESTATIC int
2636 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2637     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2638 {
2639 	struct nfsrv_descript nfsd, *nd = &nfsd;
2640 	int error = 0;
2641 
2642 	*dattrflagp = 0;
2643 	if (namelen > NFS_MAXNAMLEN)
2644 		return (ENAMETOOLONG);
2645 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2646 	(void) nfsm_strtom(nd, name, namelen);
2647 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2648 	if (error)
2649 		return (error);
2650 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2651 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2652 	if (nd->nd_repstat && !error)
2653 		error = nd->nd_repstat;
2654 	mbuf_freem(nd->nd_mrep);
2655 	/*
2656 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2657 	 */
2658 	if (error == ENOENT)
2659 		error = 0;
2660 	return (error);
2661 }
2662 
2663 /*
2664  * Readdir rpc.
2665  * Always returns with either uio_resid unchanged, if you are at the
2666  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2667  * filled in.
2668  * I felt this would allow caching of directory blocks more easily
2669  * than returning a pertially filled block.
2670  * Directory offset cookies:
2671  * Oh my, what to do with them...
2672  * I can think of three ways to deal with them:
2673  * 1 - have the layer above these RPCs maintain a map between logical
2674  *     directory byte offsets and the NFS directory offset cookies
2675  * 2 - pass the opaque directory offset cookies up into userland
2676  *     and let the libc functions deal with them, via the system call
2677  * 3 - return them to userland in the "struct dirent", so future versions
2678  *     of libc can use them and do whatever is necessary to amke things work
2679  *     above these rpc calls, in the meantime
2680  * For now, I do #3 by "hiding" the directory offset cookies after the
2681  * d_name field in struct dirent. This is space inside d_reclen that
2682  * will be ignored by anything that doesn't know about them.
2683  * The directory offset cookies are filled in as the last 8 bytes of
2684  * each directory entry, after d_name. Someday, the userland libc
2685  * functions may be able to use these. In the meantime, it satisfies
2686  * OpenBSD's requirements for cookies being returned.
2687  * If expects the directory offset cookie for the read to be in uio_offset
2688  * and returns the one for the next entry after this directory block in
2689  * there, as well.
2690  */
2691 APPLESTATIC int
2692 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2693     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2694     int *eofp, void *stuff)
2695 {
2696 	int len, left;
2697 	struct dirent *dp = NULL;
2698 	u_int32_t *tl;
2699 	nfsquad_t cookie, ncookie;
2700 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
2701 	struct nfsnode *dnp = VTONFS(vp);
2702 	struct nfsvattr nfsva;
2703 	struct nfsrv_descript nfsd, *nd = &nfsd;
2704 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2705 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2706 	long dotfileid, dotdotfileid = 0;
2707 	u_int32_t fakefileno = 0xffffffff, rderr;
2708 	char *cp;
2709 	nfsattrbit_t attrbits, dattrbits;
2710 	u_int32_t *tl2 = NULL;
2711 	size_t tresid;
2712 
2713 	KASSERT(uiop->uio_iovcnt == 1 &&
2714 	    (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
2715 	    ("nfs readdirrpc bad uio"));
2716 
2717 	/*
2718 	 * There is no point in reading a lot more than uio_resid, however
2719 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2720 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2721 	 * will never make readsize > nm_readdirsize.
2722 	 */
2723 	readsize = nmp->nm_readdirsize;
2724 	if (readsize > uio_uio_resid(uiop))
2725 		readsize = uio_uio_resid(uiop) + DIRBLKSIZ;
2726 
2727 	*attrflagp = 0;
2728 	if (eofp)
2729 		*eofp = 0;
2730 	tresid = uio_uio_resid(uiop);
2731 	cookie.lval[0] = cookiep->nfsuquad[0];
2732 	cookie.lval[1] = cookiep->nfsuquad[1];
2733 	nd->nd_mrep = NULL;
2734 
2735 	/*
2736 	 * For NFSv4, first create the "." and ".." entries.
2737 	 */
2738 	if (NFSHASNFSV4(nmp)) {
2739 		reqsize = 6 * NFSX_UNSIGNED;
2740 		NFSGETATTR_ATTRBIT(&dattrbits);
2741 		NFSZERO_ATTRBIT(&attrbits);
2742 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2743 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2744 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2745 		    NFSATTRBIT_MOUNTEDONFILEID)) {
2746 			NFSSETBIT_ATTRBIT(&attrbits,
2747 			    NFSATTRBIT_MOUNTEDONFILEID);
2748 			gotmnton = 1;
2749 		} else {
2750 			/*
2751 			 * Must fake it. Use the fileno, except when the
2752 			 * fsid is != to that of the directory. For that
2753 			 * case, generate a fake fileno that is not the same.
2754 			 */
2755 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2756 			gotmnton = 0;
2757 		}
2758 
2759 		/*
2760 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2761 		 */
2762 		if (uiop->uio_offset == 0) {
2763 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2764 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2765 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2766 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2767 			(void) nfsrv_putattrbit(nd, &attrbits);
2768 			error = nfscl_request(nd, vp, p, cred, stuff);
2769 			if (error)
2770 			    return (error);
2771 			dotfileid = 0;	/* Fake out the compiler. */
2772 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
2773 			    error = nfsm_loadattr(nd, &nfsva);
2774 			    if (error != 0)
2775 				goto nfsmout;
2776 			    dotfileid = nfsva.na_fileid;
2777 			}
2778 			if (nd->nd_repstat == 0) {
2779 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2780 			    len = fxdr_unsigned(int, *(tl + 4));
2781 			    if (len > 0 && len <= NFSX_V4FHMAX)
2782 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2783 			    else
2784 				error = EPERM;
2785 			    if (!error) {
2786 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2787 				nfsva.na_mntonfileno = 0xffffffff;
2788 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
2789 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
2790 				    NULL, NULL, NULL, p, cred);
2791 				if (error) {
2792 				    dotdotfileid = dotfileid;
2793 				} else if (gotmnton) {
2794 				    if (nfsva.na_mntonfileno != 0xffffffff)
2795 					dotdotfileid = nfsva.na_mntonfileno;
2796 				    else
2797 					dotdotfileid = nfsva.na_fileid;
2798 				} else if (nfsva.na_filesid[0] ==
2799 				    dnp->n_vattr.na_filesid[0] &&
2800 				    nfsva.na_filesid[1] ==
2801 				    dnp->n_vattr.na_filesid[1]) {
2802 				    dotdotfileid = nfsva.na_fileid;
2803 				} else {
2804 				    do {
2805 					fakefileno--;
2806 				    } while (fakefileno ==
2807 					nfsva.na_fileid);
2808 				    dotdotfileid = fakefileno;
2809 				}
2810 			    }
2811 			} else if (nd->nd_repstat == NFSERR_NOENT) {
2812 			    /*
2813 			     * Lookupp returns NFSERR_NOENT when we are
2814 			     * at the root, so just use the current dir.
2815 			     */
2816 			    nd->nd_repstat = 0;
2817 			    dotdotfileid = dotfileid;
2818 			} else {
2819 			    error = nd->nd_repstat;
2820 			}
2821 			mbuf_freem(nd->nd_mrep);
2822 			if (error)
2823 			    return (error);
2824 			nd->nd_mrep = NULL;
2825 			dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
2826 			dp->d_type = DT_DIR;
2827 			dp->d_fileno = dotfileid;
2828 			dp->d_namlen = 1;
2829 			dp->d_name[0] = '.';
2830 			dp->d_name[1] = '\0';
2831 			dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
2832 			/*
2833 			 * Just make these offset cookie 0.
2834 			 */
2835 			tl = (u_int32_t *)&dp->d_name[4];
2836 			*tl++ = 0;
2837 			*tl = 0;
2838 			blksiz += dp->d_reclen;
2839 			uio_uio_resid_add(uiop, -(dp->d_reclen));
2840 			uiop->uio_offset += dp->d_reclen;
2841 			uio_iov_base_add(uiop, dp->d_reclen);
2842 			uio_iov_len_add(uiop, -(dp->d_reclen));
2843 			dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
2844 			dp->d_type = DT_DIR;
2845 			dp->d_fileno = dotdotfileid;
2846 			dp->d_namlen = 2;
2847 			dp->d_name[0] = '.';
2848 			dp->d_name[1] = '.';
2849 			dp->d_name[2] = '\0';
2850 			dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
2851 			/*
2852 			 * Just make these offset cookie 0.
2853 			 */
2854 			tl = (u_int32_t *)&dp->d_name[4];
2855 			*tl++ = 0;
2856 			*tl = 0;
2857 			blksiz += dp->d_reclen;
2858 			uio_uio_resid_add(uiop, -(dp->d_reclen));
2859 			uiop->uio_offset += dp->d_reclen;
2860 			uio_iov_base_add(uiop, dp->d_reclen);
2861 			uio_iov_len_add(uiop, -(dp->d_reclen));
2862 		}
2863 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
2864 	} else {
2865 		reqsize = 5 * NFSX_UNSIGNED;
2866 	}
2867 
2868 
2869 	/*
2870 	 * Loop around doing readdir rpc's of size readsize.
2871 	 * The stopping criteria is EOF or buffer full.
2872 	 */
2873 	while (more_dirs && bigenough) {
2874 		*attrflagp = 0;
2875 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
2876 		if (nd->nd_flag & ND_NFSV2) {
2877 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2878 			*tl++ = cookie.lval[1];
2879 			*tl = txdr_unsigned(readsize);
2880 		} else {
2881 			NFSM_BUILD(tl, u_int32_t *, reqsize);
2882 			*tl++ = cookie.lval[0];
2883 			*tl++ = cookie.lval[1];
2884 			if (cookie.qval == 0) {
2885 				*tl++ = 0;
2886 				*tl++ = 0;
2887 			} else {
2888 				NFSLOCKNODE(dnp);
2889 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
2890 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
2891 				NFSUNLOCKNODE(dnp);
2892 			}
2893 			if (nd->nd_flag & ND_NFSV4) {
2894 				*tl++ = txdr_unsigned(readsize);
2895 				*tl = txdr_unsigned(readsize);
2896 				(void) nfsrv_putattrbit(nd, &attrbits);
2897 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2898 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
2899 				(void) nfsrv_putattrbit(nd, &dattrbits);
2900 			} else {
2901 				*tl = txdr_unsigned(readsize);
2902 			}
2903 		}
2904 		error = nfscl_request(nd, vp, p, cred, stuff);
2905 		if (error)
2906 			return (error);
2907 		if (!(nd->nd_flag & ND_NFSV2)) {
2908 			if (nd->nd_flag & ND_NFSV3)
2909 				error = nfscl_postop_attr(nd, nap, attrflagp,
2910 				    stuff);
2911 			if (!nd->nd_repstat && !error) {
2912 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
2913 				NFSLOCKNODE(dnp);
2914 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
2915 				dnp->n_cookieverf.nfsuquad[1] = *tl;
2916 				NFSUNLOCKNODE(dnp);
2917 			}
2918 		}
2919 		if (nd->nd_repstat || error) {
2920 			if (!error)
2921 				error = nd->nd_repstat;
2922 			goto nfsmout;
2923 		}
2924 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2925 		more_dirs = fxdr_unsigned(int, *tl);
2926 		if (!more_dirs)
2927 			tryformoredirs = 0;
2928 
2929 		/* loop thru the dir entries, doctoring them to 4bsd form */
2930 		while (more_dirs && bigenough) {
2931 			if (nd->nd_flag & ND_NFSV4) {
2932 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
2933 				ncookie.lval[0] = *tl++;
2934 				ncookie.lval[1] = *tl++;
2935 				len = fxdr_unsigned(int, *tl);
2936 			} else if (nd->nd_flag & ND_NFSV3) {
2937 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
2938 				nfsva.na_fileid = fxdr_hyper(tl);
2939 				tl += 2;
2940 				len = fxdr_unsigned(int, *tl);
2941 			} else {
2942 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2943 				nfsva.na_fileid =
2944 				    fxdr_unsigned(long, *tl++);
2945 				len = fxdr_unsigned(int, *tl);
2946 			}
2947 			if (len <= 0 || len > NFS_MAXNAMLEN) {
2948 				error = EBADRPC;
2949 				goto nfsmout;
2950 			}
2951 			tlen = NFSM_RNDUP(len);
2952 			if (tlen == len)
2953 				tlen += 4;  /* To ensure null termination */
2954 			left = DIRBLKSIZ - blksiz;
2955 			if ((int)(tlen + DIRHDSIZ + NFSX_HYPER) > left) {
2956 				dp->d_reclen += left;
2957 				uio_iov_base_add(uiop, left);
2958 				uio_iov_len_add(uiop, -(left));
2959 				uio_uio_resid_add(uiop, -(left));
2960 				uiop->uio_offset += left;
2961 				blksiz = 0;
2962 			}
2963 			if ((int)(tlen + DIRHDSIZ + NFSX_HYPER) > uio_uio_resid(uiop))
2964 				bigenough = 0;
2965 			if (bigenough) {
2966 				dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
2967 				dp->d_namlen = len;
2968 				dp->d_reclen = tlen + DIRHDSIZ + NFSX_HYPER;
2969 				dp->d_type = DT_UNKNOWN;
2970 				blksiz += dp->d_reclen;
2971 				if (blksiz == DIRBLKSIZ)
2972 					blksiz = 0;
2973 				uio_uio_resid_add(uiop, -(DIRHDSIZ));
2974 				uiop->uio_offset += DIRHDSIZ;
2975 				uio_iov_base_add(uiop, DIRHDSIZ);
2976 				uio_iov_len_add(uiop, -(DIRHDSIZ));
2977 				error = nfsm_mbufuio(nd, uiop, len);
2978 				if (error)
2979 					goto nfsmout;
2980 				cp = CAST_DOWN(caddr_t, uio_iov_base(uiop));
2981 				tlen -= len;
2982 				*cp = '\0';	/* null terminate */
2983 				cp += tlen;	/* points to cookie storage */
2984 				tl2 = (u_int32_t *)cp;
2985 				uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
2986 				uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
2987 				uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
2988 				uiop->uio_offset += (tlen + NFSX_HYPER);
2989 			} else {
2990 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2991 				if (error)
2992 					goto nfsmout;
2993 			}
2994 			if (nd->nd_flag & ND_NFSV4) {
2995 				rderr = 0;
2996 				nfsva.na_mntonfileno = 0xffffffff;
2997 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
2998 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
2999 				    NULL, NULL, &rderr, p, cred);
3000 				if (error)
3001 					goto nfsmout;
3002 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3003 			} else if (nd->nd_flag & ND_NFSV3) {
3004 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3005 				ncookie.lval[0] = *tl++;
3006 				ncookie.lval[1] = *tl++;
3007 			} else {
3008 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3009 				ncookie.lval[0] = 0;
3010 				ncookie.lval[1] = *tl++;
3011 			}
3012 			if (bigenough) {
3013 			    if (nd->nd_flag & ND_NFSV4) {
3014 				if (rderr) {
3015 				    dp->d_fileno = 0;
3016 				} else {
3017 				    if (gotmnton) {
3018 					if (nfsva.na_mntonfileno != 0xffffffff)
3019 					    dp->d_fileno = nfsva.na_mntonfileno;
3020 					else
3021 					    dp->d_fileno = nfsva.na_fileid;
3022 				    } else if (nfsva.na_filesid[0] ==
3023 					dnp->n_vattr.na_filesid[0] &&
3024 					nfsva.na_filesid[1] ==
3025 					dnp->n_vattr.na_filesid[1]) {
3026 					dp->d_fileno = nfsva.na_fileid;
3027 				    } else {
3028 					do {
3029 					    fakefileno--;
3030 					} while (fakefileno ==
3031 					    nfsva.na_fileid);
3032 					dp->d_fileno = fakefileno;
3033 				    }
3034 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3035 				}
3036 			    } else {
3037 				dp->d_fileno = nfsva.na_fileid;
3038 			    }
3039 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3040 				ncookie.lval[0];
3041 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3042 				ncookie.lval[1];
3043 			}
3044 			more_dirs = fxdr_unsigned(int, *tl);
3045 		}
3046 		/*
3047 		 * If at end of rpc data, get the eof boolean
3048 		 */
3049 		if (!more_dirs) {
3050 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3051 			eof = fxdr_unsigned(int, *tl);
3052 			if (tryformoredirs)
3053 				more_dirs = !eof;
3054 			if (nd->nd_flag & ND_NFSV4) {
3055 				error = nfscl_postop_attr(nd, nap, attrflagp,
3056 				    stuff);
3057 				if (error)
3058 					goto nfsmout;
3059 			}
3060 		}
3061 		mbuf_freem(nd->nd_mrep);
3062 		nd->nd_mrep = NULL;
3063 	}
3064 	/*
3065 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3066 	 * by increasing d_reclen for the last record.
3067 	 */
3068 	if (blksiz > 0) {
3069 		left = DIRBLKSIZ - blksiz;
3070 		dp->d_reclen += left;
3071 		uio_iov_base_add(uiop, left);
3072 		uio_iov_len_add(uiop, -(left));
3073 		uio_uio_resid_add(uiop, -(left));
3074 		uiop->uio_offset += left;
3075 	}
3076 
3077 	/*
3078 	 * If returning no data, assume end of file.
3079 	 * If not bigenough, return not end of file, since you aren't
3080 	 *    returning all the data
3081 	 * Otherwise, return the eof flag from the server.
3082 	 */
3083 	if (eofp) {
3084 		if (tresid == ((size_t)(uio_uio_resid(uiop))))
3085 			*eofp = 1;
3086 		else if (!bigenough)
3087 			*eofp = 0;
3088 		else
3089 			*eofp = eof;
3090 	}
3091 
3092 	/*
3093 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3094 	 */
3095 	while (uio_uio_resid(uiop) > 0 && ((size_t)(uio_uio_resid(uiop))) != tresid) {
3096 		dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
3097 		dp->d_type = DT_UNKNOWN;
3098 		dp->d_fileno = 0;
3099 		dp->d_namlen = 0;
3100 		dp->d_name[0] = '\0';
3101 		tl = (u_int32_t *)&dp->d_name[4];
3102 		*tl++ = cookie.lval[0];
3103 		*tl = cookie.lval[1];
3104 		dp->d_reclen = DIRBLKSIZ;
3105 		uio_iov_base_add(uiop, DIRBLKSIZ);
3106 		uio_iov_len_add(uiop, -(DIRBLKSIZ));
3107 		uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3108 		uiop->uio_offset += DIRBLKSIZ;
3109 	}
3110 
3111 nfsmout:
3112 	if (nd->nd_mrep != NULL)
3113 		mbuf_freem(nd->nd_mrep);
3114 	return (error);
3115 }
3116 
3117 #ifndef APPLE
3118 /*
3119  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3120  * (Also used for NFS V4 when mount flag set.)
3121  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3122  */
3123 APPLESTATIC int
3124 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3125     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3126     int *eofp, void *stuff)
3127 {
3128 	int len, left;
3129 	struct dirent *dp = NULL;
3130 	u_int32_t *tl;
3131 	vnode_t newvp = NULLVP;
3132 	struct nfsrv_descript nfsd, *nd = &nfsd;
3133 	struct nameidata nami, *ndp = &nami;
3134 	struct componentname *cnp = &ndp->ni_cnd;
3135 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3136 	struct nfsnode *dnp = VTONFS(vp), *np;
3137 	struct nfsvattr nfsva;
3138 	struct nfsfh *nfhp;
3139 	nfsquad_t cookie, ncookie;
3140 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3141 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3142 	int isdotdot = 0, unlocknewvp = 0;
3143 	long dotfileid, dotdotfileid = 0, fileno = 0;
3144 	char *cp;
3145 	nfsattrbit_t attrbits, dattrbits;
3146 	size_t tresid;
3147 	u_int32_t *tl2 = NULL, fakefileno = 0xffffffff, rderr;
3148 	struct timespec dctime;
3149 
3150 	KASSERT(uiop->uio_iovcnt == 1 &&
3151 	    (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
3152 	    ("nfs readdirplusrpc bad uio"));
3153 	timespecclear(&dctime);
3154 	*attrflagp = 0;
3155 	if (eofp != NULL)
3156 		*eofp = 0;
3157 	ndp->ni_dvp = vp;
3158 	nd->nd_mrep = NULL;
3159 	cookie.lval[0] = cookiep->nfsuquad[0];
3160 	cookie.lval[1] = cookiep->nfsuquad[1];
3161 	tresid = uio_uio_resid(uiop);
3162 
3163 	/*
3164 	 * For NFSv4, first create the "." and ".." entries.
3165 	 */
3166 	if (NFSHASNFSV4(nmp)) {
3167 		NFSGETATTR_ATTRBIT(&dattrbits);
3168 		NFSZERO_ATTRBIT(&attrbits);
3169 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3170 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3171 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3172 			NFSSETBIT_ATTRBIT(&attrbits,
3173 			    NFSATTRBIT_MOUNTEDONFILEID);
3174 			gotmnton = 1;
3175 		} else {
3176 			/*
3177 			 * Must fake it. Use the fileno, except when the
3178 			 * fsid is != to that of the directory. For that
3179 			 * case, generate a fake fileno that is not the same.
3180 			 */
3181 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3182 			gotmnton = 0;
3183 		}
3184 
3185 		/*
3186 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3187 		 */
3188 		if (uiop->uio_offset == 0) {
3189 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3190 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3191 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3192 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3193 			(void) nfsrv_putattrbit(nd, &attrbits);
3194 			error = nfscl_request(nd, vp, p, cred, stuff);
3195 			if (error)
3196 			    return (error);
3197 			dotfileid = 0;	/* Fake out the compiler. */
3198 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3199 			    error = nfsm_loadattr(nd, &nfsva);
3200 			    if (error != 0)
3201 				goto nfsmout;
3202 			    dctime = nfsva.na_ctime;
3203 			    dotfileid = nfsva.na_fileid;
3204 			}
3205 			if (nd->nd_repstat == 0) {
3206 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3207 			    len = fxdr_unsigned(int, *(tl + 4));
3208 			    if (len > 0 && len <= NFSX_V4FHMAX)
3209 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3210 			    else
3211 				error = EPERM;
3212 			    if (!error) {
3213 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3214 				nfsva.na_mntonfileno = 0xffffffff;
3215 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3216 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3217 				    NULL, NULL, NULL, p, cred);
3218 				if (error) {
3219 				    dotdotfileid = dotfileid;
3220 				} else if (gotmnton) {
3221 				    if (nfsva.na_mntonfileno != 0xffffffff)
3222 					dotdotfileid = nfsva.na_mntonfileno;
3223 				    else
3224 					dotdotfileid = nfsva.na_fileid;
3225 				} else if (nfsva.na_filesid[0] ==
3226 				    dnp->n_vattr.na_filesid[0] &&
3227 				    nfsva.na_filesid[1] ==
3228 				    dnp->n_vattr.na_filesid[1]) {
3229 				    dotdotfileid = nfsva.na_fileid;
3230 				} else {
3231 				    do {
3232 					fakefileno--;
3233 				    } while (fakefileno ==
3234 					nfsva.na_fileid);
3235 				    dotdotfileid = fakefileno;
3236 				}
3237 			    }
3238 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3239 			    /*
3240 			     * Lookupp returns NFSERR_NOENT when we are
3241 			     * at the root, so just use the current dir.
3242 			     */
3243 			    nd->nd_repstat = 0;
3244 			    dotdotfileid = dotfileid;
3245 			} else {
3246 			    error = nd->nd_repstat;
3247 			}
3248 			mbuf_freem(nd->nd_mrep);
3249 			if (error)
3250 			    return (error);
3251 			nd->nd_mrep = NULL;
3252 			dp = (struct dirent *)uio_iov_base(uiop);
3253 			dp->d_type = DT_DIR;
3254 			dp->d_fileno = dotfileid;
3255 			dp->d_namlen = 1;
3256 			dp->d_name[0] = '.';
3257 			dp->d_name[1] = '\0';
3258 			dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
3259 			/*
3260 			 * Just make these offset cookie 0.
3261 			 */
3262 			tl = (u_int32_t *)&dp->d_name[4];
3263 			*tl++ = 0;
3264 			*tl = 0;
3265 			blksiz += dp->d_reclen;
3266 			uio_uio_resid_add(uiop, -(dp->d_reclen));
3267 			uiop->uio_offset += dp->d_reclen;
3268 			uio_iov_base_add(uiop, dp->d_reclen);
3269 			uio_iov_len_add(uiop, -(dp->d_reclen));
3270 			dp = (struct dirent *)uio_iov_base(uiop);
3271 			dp->d_type = DT_DIR;
3272 			dp->d_fileno = dotdotfileid;
3273 			dp->d_namlen = 2;
3274 			dp->d_name[0] = '.';
3275 			dp->d_name[1] = '.';
3276 			dp->d_name[2] = '\0';
3277 			dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
3278 			/*
3279 			 * Just make these offset cookie 0.
3280 			 */
3281 			tl = (u_int32_t *)&dp->d_name[4];
3282 			*tl++ = 0;
3283 			*tl = 0;
3284 			blksiz += dp->d_reclen;
3285 			uio_uio_resid_add(uiop, -(dp->d_reclen));
3286 			uiop->uio_offset += dp->d_reclen;
3287 			uio_iov_base_add(uiop, dp->d_reclen);
3288 			uio_iov_len_add(uiop, -(dp->d_reclen));
3289 		}
3290 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3291 		if (gotmnton)
3292 			NFSSETBIT_ATTRBIT(&attrbits,
3293 			    NFSATTRBIT_MOUNTEDONFILEID);
3294 	}
3295 
3296 	/*
3297 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3298 	 * The stopping criteria is EOF or buffer full.
3299 	 */
3300 	while (more_dirs && bigenough) {
3301 		*attrflagp = 0;
3302 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3303  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3304 		*tl++ = cookie.lval[0];
3305 		*tl++ = cookie.lval[1];
3306 		if (cookie.qval == 0) {
3307 			*tl++ = 0;
3308 			*tl++ = 0;
3309 		} else {
3310 			NFSLOCKNODE(dnp);
3311 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3312 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3313 			NFSUNLOCKNODE(dnp);
3314 		}
3315 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3316 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3317 		if (nd->nd_flag & ND_NFSV4) {
3318 			(void) nfsrv_putattrbit(nd, &attrbits);
3319 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3320 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3321 			(void) nfsrv_putattrbit(nd, &dattrbits);
3322 		}
3323 		error = nfscl_request(nd, vp, p, cred, stuff);
3324 		if (error)
3325 			return (error);
3326 		if (nd->nd_flag & ND_NFSV3)
3327 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3328 		if (nd->nd_repstat || error) {
3329 			if (!error)
3330 				error = nd->nd_repstat;
3331 			goto nfsmout;
3332 		}
3333 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3334 			dctime = nap->na_ctime;
3335 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3336 		NFSLOCKNODE(dnp);
3337 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3338 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3339 		NFSUNLOCKNODE(dnp);
3340 		more_dirs = fxdr_unsigned(int, *tl);
3341 		if (!more_dirs)
3342 			tryformoredirs = 0;
3343 
3344 		/* loop thru the dir entries, doctoring them to 4bsd form */
3345 		while (more_dirs && bigenough) {
3346 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3347 			if (nd->nd_flag & ND_NFSV4) {
3348 				ncookie.lval[0] = *tl++;
3349 				ncookie.lval[1] = *tl++;
3350 			} else {
3351 				fileno = fxdr_unsigned(long, *++tl);
3352 				tl++;
3353 			}
3354 			len = fxdr_unsigned(int, *tl);
3355 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3356 				error = EBADRPC;
3357 				goto nfsmout;
3358 			}
3359 			tlen = NFSM_RNDUP(len);
3360 			if (tlen == len)
3361 				tlen += 4;  /* To ensure null termination */
3362 			left = DIRBLKSIZ - blksiz;
3363 			if ((tlen + DIRHDSIZ + NFSX_HYPER) > left) {
3364 				dp->d_reclen += left;
3365 				uio_iov_base_add(uiop, left);
3366 				uio_iov_len_add(uiop, -(left));
3367 				uio_uio_resid_add(uiop, -(left));
3368 				uiop->uio_offset += left;
3369 				blksiz = 0;
3370 			}
3371 			if ((tlen + DIRHDSIZ + NFSX_HYPER) > uio_uio_resid(uiop))
3372 				bigenough = 0;
3373 			if (bigenough) {
3374 				dp = (struct dirent *)uio_iov_base(uiop);
3375 				dp->d_namlen = len;
3376 				dp->d_reclen = tlen + DIRHDSIZ + NFSX_HYPER;
3377 				dp->d_type = DT_UNKNOWN;
3378 				blksiz += dp->d_reclen;
3379 				if (blksiz == DIRBLKSIZ)
3380 					blksiz = 0;
3381 				uio_uio_resid_add(uiop, -(DIRHDSIZ));
3382 				uiop->uio_offset += DIRHDSIZ;
3383 				uio_iov_base_add(uiop, DIRHDSIZ);
3384 				uio_iov_len_add(uiop, -(DIRHDSIZ));
3385 				cnp->cn_nameptr = uio_iov_base(uiop);
3386 				cnp->cn_namelen = len;
3387 				NFSCNHASHZERO(cnp);
3388 				error = nfsm_mbufuio(nd, uiop, len);
3389 				if (error)
3390 					goto nfsmout;
3391 				cp = uio_iov_base(uiop);
3392 				tlen -= len;
3393 				*cp = '\0';
3394 				cp += tlen;	/* points to cookie storage */
3395 				tl2 = (u_int32_t *)cp;
3396 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3397 				    cnp->cn_nameptr[1] == '.')
3398 					isdotdot = 1;
3399 				else
3400 					isdotdot = 0;
3401 				uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3402 				uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3403 				uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3404 				uiop->uio_offset += (tlen + NFSX_HYPER);
3405 			} else {
3406 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3407 				if (error)
3408 					goto nfsmout;
3409 			}
3410 			nfhp = NULL;
3411 			if (nd->nd_flag & ND_NFSV3) {
3412 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3413 				ncookie.lval[0] = *tl++;
3414 				ncookie.lval[1] = *tl++;
3415 				attrflag = fxdr_unsigned(int, *tl);
3416 				if (attrflag) {
3417 				  error = nfsm_loadattr(nd, &nfsva);
3418 				  if (error)
3419 					goto nfsmout;
3420 				}
3421 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3422 				if (*tl) {
3423 					error = nfsm_getfh(nd, &nfhp);
3424 					if (error)
3425 					    goto nfsmout;
3426 				}
3427 				if (!attrflag && nfhp != NULL) {
3428 					FREE((caddr_t)nfhp, M_NFSFH);
3429 					nfhp = NULL;
3430 				}
3431 			} else {
3432 				rderr = 0;
3433 				nfsva.na_mntonfileno = 0xffffffff;
3434 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3435 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3436 				    NULL, NULL, &rderr, p, cred);
3437 				if (error)
3438 					goto nfsmout;
3439 			}
3440 
3441 			if (bigenough) {
3442 			    if (nd->nd_flag & ND_NFSV4) {
3443 				if (rderr) {
3444 				    dp->d_fileno = 0;
3445 				} else if (gotmnton) {
3446 				    if (nfsva.na_mntonfileno != 0xffffffff)
3447 					dp->d_fileno = nfsva.na_mntonfileno;
3448 				    else
3449 					dp->d_fileno = nfsva.na_fileid;
3450 				} else if (nfsva.na_filesid[0] ==
3451 				    dnp->n_vattr.na_filesid[0] &&
3452 				    nfsva.na_filesid[1] ==
3453 				    dnp->n_vattr.na_filesid[1]) {
3454 				    dp->d_fileno = nfsva.na_fileid;
3455 				} else {
3456 				    do {
3457 					fakefileno--;
3458 				    } while (fakefileno ==
3459 					nfsva.na_fileid);
3460 				    dp->d_fileno = fakefileno;
3461 				}
3462 			    } else {
3463 				dp->d_fileno = fileno;
3464 			    }
3465 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3466 				ncookie.lval[0];
3467 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3468 				ncookie.lval[1];
3469 
3470 			    if (nfhp != NULL) {
3471 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3472 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3473 				    VREF(vp);
3474 				    newvp = vp;
3475 				    unlocknewvp = 0;
3476 				    FREE((caddr_t)nfhp, M_NFSFH);
3477 				    np = dnp;
3478 				} else if (isdotdot != 0) {
3479 				    /*
3480 				     * Skip doing a nfscl_nget() call for "..".
3481 				     * There's a race between acquiring the nfs
3482 				     * node here and lookups that look for the
3483 				     * directory being read (in the parent).
3484 				     * It would try to get a lock on ".." here,
3485 				     * owning the lock on the directory being
3486 				     * read. Lookup will hold the lock on ".."
3487 				     * and try to acquire the lock on the
3488 				     * directory being read.
3489 				     * If the directory is unlocked/relocked,
3490 				     * then there is a LOR with the buflock
3491 				     * vp is relocked.
3492 				     */
3493 				    free(nfhp, M_NFSFH);
3494 				} else {
3495 				    error = nfscl_nget(vnode_mount(vp), vp,
3496 				      nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3497 				    if (!error) {
3498 					newvp = NFSTOV(np);
3499 					unlocknewvp = 1;
3500 				    }
3501 				}
3502 				nfhp = NULL;
3503 				if (newvp != NULLVP) {
3504 				    error = nfscl_loadattrcache(&newvp,
3505 					&nfsva, NULL, NULL, 0, 0);
3506 				    if (error) {
3507 					if (unlocknewvp)
3508 					    vput(newvp);
3509 					else
3510 					    vrele(newvp);
3511 					goto nfsmout;
3512 				    }
3513 				    dp->d_type =
3514 					vtonfs_dtype(np->n_vattr.na_type);
3515 				    ndp->ni_vp = newvp;
3516 				    NFSCNHASH(cnp, HASHINIT);
3517 				    if (cnp->cn_namelen <= NCHNAMLEN &&
3518 					(newvp->v_type != VDIR ||
3519 					 dctime.tv_sec != 0)) {
3520 					cache_enter_time(ndp->ni_dvp,
3521 					    ndp->ni_vp, cnp,
3522 					    &nfsva.na_ctime,
3523 					    newvp->v_type != VDIR ? NULL :
3524 					    &dctime);
3525 				    }
3526 				    if (unlocknewvp)
3527 					vput(newvp);
3528 				    else
3529 					vrele(newvp);
3530 				    newvp = NULLVP;
3531 				}
3532 			    }
3533 			} else if (nfhp != NULL) {
3534 			    FREE((caddr_t)nfhp, M_NFSFH);
3535 			}
3536 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3537 			more_dirs = fxdr_unsigned(int, *tl);
3538 		}
3539 		/*
3540 		 * If at end of rpc data, get the eof boolean
3541 		 */
3542 		if (!more_dirs) {
3543 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3544 			eof = fxdr_unsigned(int, *tl);
3545 			if (tryformoredirs)
3546 				more_dirs = !eof;
3547 			if (nd->nd_flag & ND_NFSV4) {
3548 				error = nfscl_postop_attr(nd, nap, attrflagp,
3549 				    stuff);
3550 				if (error)
3551 					goto nfsmout;
3552 			}
3553 		}
3554 		mbuf_freem(nd->nd_mrep);
3555 		nd->nd_mrep = NULL;
3556 	}
3557 	/*
3558 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3559 	 * by increasing d_reclen for the last record.
3560 	 */
3561 	if (blksiz > 0) {
3562 		left = DIRBLKSIZ - blksiz;
3563 		dp->d_reclen += left;
3564 		uio_iov_base_add(uiop, left);
3565 		uio_iov_len_add(uiop, -(left));
3566 		uio_uio_resid_add(uiop, -(left));
3567 		uiop->uio_offset += left;
3568 	}
3569 
3570 	/*
3571 	 * If returning no data, assume end of file.
3572 	 * If not bigenough, return not end of file, since you aren't
3573 	 *    returning all the data
3574 	 * Otherwise, return the eof flag from the server.
3575 	 */
3576 	if (eofp != NULL) {
3577 		if (tresid == uio_uio_resid(uiop))
3578 			*eofp = 1;
3579 		else if (!bigenough)
3580 			*eofp = 0;
3581 		else
3582 			*eofp = eof;
3583 	}
3584 
3585 	/*
3586 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3587 	 */
3588 	while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3589 		dp = (struct dirent *)uio_iov_base(uiop);
3590 		dp->d_type = DT_UNKNOWN;
3591 		dp->d_fileno = 0;
3592 		dp->d_namlen = 0;
3593 		dp->d_name[0] = '\0';
3594 		tl = (u_int32_t *)&dp->d_name[4];
3595 		*tl++ = cookie.lval[0];
3596 		*tl = cookie.lval[1];
3597 		dp->d_reclen = DIRBLKSIZ;
3598 		uio_iov_base_add(uiop, DIRBLKSIZ);
3599 		uio_iov_len_add(uiop, -(DIRBLKSIZ));
3600 		uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3601 		uiop->uio_offset += DIRBLKSIZ;
3602 	}
3603 
3604 nfsmout:
3605 	if (nd->nd_mrep != NULL)
3606 		mbuf_freem(nd->nd_mrep);
3607 	return (error);
3608 }
3609 #endif	/* !APPLE */
3610 
3611 /*
3612  * Nfs commit rpc
3613  */
3614 APPLESTATIC int
3615 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
3616     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
3617 {
3618 	u_int32_t *tl;
3619 	struct nfsrv_descript nfsd, *nd = &nfsd;
3620 	nfsattrbit_t attrbits;
3621 	int error;
3622 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3623 
3624 	*attrflagp = 0;
3625 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
3626 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3627 	txdr_hyper(offset, tl);
3628 	tl += 2;
3629 	*tl = txdr_unsigned(cnt);
3630 	if (nd->nd_flag & ND_NFSV4) {
3631 		/*
3632 		 * And do a Getattr op.
3633 		 */
3634 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3635 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3636 		NFSGETATTR_ATTRBIT(&attrbits);
3637 		(void) nfsrv_putattrbit(nd, &attrbits);
3638 	}
3639 	error = nfscl_request(nd, vp, p, cred, stuff);
3640 	if (error)
3641 		return (error);
3642 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
3643 	if (!error && !nd->nd_repstat) {
3644 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
3645 		NFSLOCKMNT(nmp);
3646 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
3647 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
3648 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
3649 		}
3650 		NFSUNLOCKMNT(nmp);
3651 		if (nd->nd_flag & ND_NFSV4)
3652 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3653 	}
3654 nfsmout:
3655 	if (!error && nd->nd_repstat)
3656 		error = nd->nd_repstat;
3657 	mbuf_freem(nd->nd_mrep);
3658 	return (error);
3659 }
3660 
3661 /*
3662  * NFS byte range lock rpc.
3663  * (Mostly just calls one of the three lower level RPC routines.)
3664  */
3665 APPLESTATIC int
3666 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
3667     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3668 {
3669 	struct nfscllockowner *lp;
3670 	struct nfsclclient *clp;
3671 	struct nfsfh *nfhp;
3672 	struct nfsrv_descript nfsd, *nd = &nfsd;
3673 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3674 	u_int64_t off, len;
3675 	off_t start, end;
3676 	u_int32_t clidrev = 0;
3677 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
3678 	int callcnt, dorpc;
3679 
3680 	/*
3681 	 * Convert the flock structure into a start and end and do POSIX
3682 	 * bounds checking.
3683 	 */
3684 	switch (fl->l_whence) {
3685 	case SEEK_SET:
3686 	case SEEK_CUR:
3687 		/*
3688 		 * Caller is responsible for adding any necessary offset
3689 		 * when SEEK_CUR is used.
3690 		 */
3691 		start = fl->l_start;
3692 		off = fl->l_start;
3693 		break;
3694 	case SEEK_END:
3695 		start = size + fl->l_start;
3696 		off = size + fl->l_start;
3697 		break;
3698 	default:
3699 		return (EINVAL);
3700 	};
3701 	if (start < 0)
3702 		return (EINVAL);
3703 	if (fl->l_len != 0) {
3704 		end = start + fl->l_len - 1;
3705 		if (end < start)
3706 			return (EINVAL);
3707 	}
3708 
3709 	len = fl->l_len;
3710 	if (len == 0)
3711 		len = NFS64BITSSET;
3712 	retrycnt = 0;
3713 	do {
3714 	    nd->nd_repstat = 0;
3715 	    if (op == F_GETLK) {
3716 		error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3717 		if (error)
3718 			return (error);
3719 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
3720 		if (!error) {
3721 			clidrev = clp->nfsc_clientidrev;
3722 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
3723 			    p, id, flags);
3724 		} else if (error == -1) {
3725 			error = 0;
3726 		}
3727 		nfscl_clientrelease(clp);
3728 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
3729 		/*
3730 		 * We must loop around for all lockowner cases.
3731 		 */
3732 		callcnt = 0;
3733 		error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3734 		if (error)
3735 			return (error);
3736 		do {
3737 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
3738 			clp, id, flags, &lp, &dorpc);
3739 		    /*
3740 		     * If it returns a NULL lp, we're done.
3741 		     */
3742 		    if (lp == NULL) {
3743 			if (callcnt == 0)
3744 			    nfscl_clientrelease(clp);
3745 			else
3746 			    nfscl_releasealllocks(clp, vp, p, id, flags);
3747 			return (error);
3748 		    }
3749 		    if (nmp->nm_clp != NULL)
3750 			clidrev = nmp->nm_clp->nfsc_clientidrev;
3751 		    else
3752 			clidrev = 0;
3753 		    /*
3754 		     * If the server doesn't support Posix lock semantics,
3755 		     * only allow locks on the entire file, since it won't
3756 		     * handle overlapping byte ranges.
3757 		     * There might still be a problem when a lock
3758 		     * upgrade/downgrade (read<->write) occurs, since the
3759 		     * server "might" expect an unlock first?
3760 		     */
3761 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
3762 			(off == 0 && len == NFS64BITSSET))) {
3763 			/*
3764 			 * Since the lock records will go away, we must
3765 			 * wait for grace and delay here.
3766 			 */
3767 			do {
3768 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
3769 				NFSV4LOCKT_READ, cred, p, 0);
3770 			    if ((nd->nd_repstat == NFSERR_GRACE ||
3771 				 nd->nd_repstat == NFSERR_DELAY) &&
3772 				error == 0)
3773 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
3774 				    "nfs_advlock");
3775 			} while ((nd->nd_repstat == NFSERR_GRACE ||
3776 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
3777 		    }
3778 		    callcnt++;
3779 		} while (error == 0 && nd->nd_repstat == 0);
3780 		nfscl_releasealllocks(clp, vp, p, id, flags);
3781 	    } else if (op == F_SETLK) {
3782 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
3783 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
3784 		if (error || donelocally) {
3785 			return (error);
3786 		}
3787 		if (nmp->nm_clp != NULL)
3788 			clidrev = nmp->nm_clp->nfsc_clientidrev;
3789 		else
3790 			clidrev = 0;
3791 		nfhp = VTONFS(vp)->n_fhp;
3792 		if (!lp->nfsl_open->nfso_posixlock &&
3793 		    (off != 0 || len != NFS64BITSSET)) {
3794 			error = EINVAL;
3795 		} else {
3796 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
3797 			    nfhp->nfh_len, lp, newone, reclaim, off,
3798 			    len, fl->l_type, cred, p, 0);
3799 		}
3800 		if (!error)
3801 			error = nd->nd_repstat;
3802 		nfscl_lockrelease(lp, error, newone);
3803 	    } else {
3804 		error = EINVAL;
3805 	    }
3806 	    if (!error)
3807 	        error = nd->nd_repstat;
3808 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
3809 		error == NFSERR_STALEDONTRECOVER ||
3810 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3811 		error == NFSERR_BADSESSION) {
3812 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
3813 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
3814 		&& clidrev != 0) {
3815 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
3816 		retrycnt++;
3817 	    }
3818 	} while (error == NFSERR_GRACE ||
3819 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3820 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
3821 	    error == NFSERR_BADSESSION ||
3822 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
3823 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
3824 	if (error && retrycnt >= 4)
3825 		error = EIO;
3826 	return (error);
3827 }
3828 
3829 /*
3830  * The lower level routine for the LockT case.
3831  */
3832 APPLESTATIC int
3833 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
3834     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
3835     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3836 {
3837 	u_int32_t *tl;
3838 	int error, type, size;
3839 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
3840 	struct nfsnode *np;
3841 	struct nfsmount *nmp;
3842 
3843 	nmp = VFSTONFS(vp->v_mount);
3844 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
3845 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
3846 	if (fl->l_type == F_RDLCK)
3847 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
3848 	else
3849 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
3850 	txdr_hyper(off, tl);
3851 	tl += 2;
3852 	txdr_hyper(len, tl);
3853 	tl += 2;
3854 	*tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
3855 	*tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
3856 	nfscl_filllockowner(id, own, flags);
3857 	np = VTONFS(vp);
3858 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
3859 	    np->n_fhp->nfh_len);
3860 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
3861 	error = nfscl_request(nd, vp, p, cred, NULL);
3862 	if (error)
3863 		return (error);
3864 	if (nd->nd_repstat == 0) {
3865 		fl->l_type = F_UNLCK;
3866 	} else if (nd->nd_repstat == NFSERR_DENIED) {
3867 		nd->nd_repstat = 0;
3868 		fl->l_whence = SEEK_SET;
3869 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
3870 		fl->l_start = fxdr_hyper(tl);
3871 		tl += 2;
3872 		len = fxdr_hyper(tl);
3873 		tl += 2;
3874 		if (len == NFS64BITSSET)
3875 			fl->l_len = 0;
3876 		else
3877 			fl->l_len = len;
3878 		type = fxdr_unsigned(int, *tl++);
3879 		if (type == NFSV4LOCKT_WRITE)
3880 			fl->l_type = F_WRLCK;
3881 		else
3882 			fl->l_type = F_RDLCK;
3883 		/*
3884 		 * XXX For now, I have no idea what to do with the
3885 		 * conflicting lock_owner, so I'll just set the pid == 0
3886 		 * and skip over the lock_owner.
3887 		 */
3888 		fl->l_pid = (pid_t)0;
3889 		tl += 2;
3890 		size = fxdr_unsigned(int, *tl);
3891 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
3892 			error = EBADRPC;
3893 		if (!error)
3894 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
3895 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID ||
3896 	    nd->nd_repstat == NFSERR_BADSESSION)
3897 		nfscl_initiate_recovery(clp);
3898 nfsmout:
3899 	mbuf_freem(nd->nd_mrep);
3900 	return (error);
3901 }
3902 
3903 /*
3904  * Lower level function that performs the LockU RPC.
3905  */
3906 static int
3907 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
3908     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
3909     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
3910 {
3911 	u_int32_t *tl;
3912 	int error;
3913 
3914 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
3915 	    lp->nfsl_open->nfso_fhlen, NULL, NULL);
3916 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
3917 	*tl++ = txdr_unsigned(type);
3918 	*tl = txdr_unsigned(lp->nfsl_seqid);
3919 	if (nfstest_outofseq &&
3920 	    (arc4random() % nfstest_outofseq) == 0)
3921 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
3922 	tl++;
3923 	if (NFSHASNFSV4N(nmp))
3924 		*tl++ = 0;
3925 	else
3926 		*tl++ = lp->nfsl_stateid.seqid;
3927 	*tl++ = lp->nfsl_stateid.other[0];
3928 	*tl++ = lp->nfsl_stateid.other[1];
3929 	*tl++ = lp->nfsl_stateid.other[2];
3930 	txdr_hyper(off, tl);
3931 	tl += 2;
3932 	txdr_hyper(len, tl);
3933 	if (syscred)
3934 		nd->nd_flag |= ND_USEGSSNAME;
3935 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
3936 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
3937 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
3938 	if (error)
3939 		return (error);
3940 	if (nd->nd_repstat == 0) {
3941 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
3942 		lp->nfsl_stateid.seqid = *tl++;
3943 		lp->nfsl_stateid.other[0] = *tl++;
3944 		lp->nfsl_stateid.other[1] = *tl++;
3945 		lp->nfsl_stateid.other[2] = *tl;
3946 	} else if (nd->nd_repstat == NFSERR_STALESTATEID ||
3947 	    nd->nd_repstat == NFSERR_BADSESSION)
3948 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
3949 nfsmout:
3950 	mbuf_freem(nd->nd_mrep);
3951 	return (error);
3952 }
3953 
3954 /*
3955  * The actual Lock RPC.
3956  */
3957 APPLESTATIC int
3958 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
3959     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
3960     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
3961     NFSPROC_T *p, int syscred)
3962 {
3963 	u_int32_t *tl;
3964 	int error, size;
3965 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
3966 
3967 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL);
3968 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
3969 	if (type == F_RDLCK)
3970 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
3971 	else
3972 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
3973 	*tl++ = txdr_unsigned(reclaim);
3974 	txdr_hyper(off, tl);
3975 	tl += 2;
3976 	txdr_hyper(len, tl);
3977 	tl += 2;
3978 	if (newone) {
3979 	    *tl = newnfs_true;
3980 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
3981 		2 * NFSX_UNSIGNED + NFSX_HYPER);
3982 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
3983 	    if (NFSHASNFSV4N(nmp))
3984 		*tl++ = 0;
3985 	    else
3986 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
3987 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
3988 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
3989 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
3990 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
3991 	    *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
3992 	    *tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
3993 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
3994 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
3995 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
3996 	} else {
3997 	    *tl = newnfs_false;
3998 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
3999 	    if (NFSHASNFSV4N(nmp))
4000 		*tl++ = 0;
4001 	    else
4002 		*tl++ = lp->nfsl_stateid.seqid;
4003 	    *tl++ = lp->nfsl_stateid.other[0];
4004 	    *tl++ = lp->nfsl_stateid.other[1];
4005 	    *tl++ = lp->nfsl_stateid.other[2];
4006 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4007 	    if (nfstest_outofseq &&
4008 		(arc4random() % nfstest_outofseq) == 0)
4009 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4010 	}
4011 	if (syscred)
4012 		nd->nd_flag |= ND_USEGSSNAME;
4013 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4014 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4015 	if (error)
4016 		return (error);
4017 	if (newone)
4018 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4019 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4020 	if (nd->nd_repstat == 0) {
4021 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4022 		lp->nfsl_stateid.seqid = *tl++;
4023 		lp->nfsl_stateid.other[0] = *tl++;
4024 		lp->nfsl_stateid.other[1] = *tl++;
4025 		lp->nfsl_stateid.other[2] = *tl;
4026 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4027 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4028 		size = fxdr_unsigned(int, *(tl + 7));
4029 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4030 			error = EBADRPC;
4031 		if (!error)
4032 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4033 	} else if (nd->nd_repstat == NFSERR_STALESTATEID ||
4034 	    nd->nd_repstat == NFSERR_BADSESSION)
4035 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4036 nfsmout:
4037 	mbuf_freem(nd->nd_mrep);
4038 	return (error);
4039 }
4040 
4041 /*
4042  * nfs statfs rpc
4043  * (always called with the vp for the mount point)
4044  */
4045 APPLESTATIC int
4046 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4047     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4048     void *stuff)
4049 {
4050 	u_int32_t *tl = NULL;
4051 	struct nfsrv_descript nfsd, *nd = &nfsd;
4052 	struct nfsmount *nmp;
4053 	nfsattrbit_t attrbits;
4054 	int error;
4055 
4056 	*attrflagp = 0;
4057 	nmp = VFSTONFS(vnode_mount(vp));
4058 	if (NFSHASNFSV4(nmp)) {
4059 		/*
4060 		 * For V4, you actually do a getattr.
4061 		 */
4062 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4063 		NFSSTATFS_GETATTRBIT(&attrbits);
4064 		(void) nfsrv_putattrbit(nd, &attrbits);
4065 		nd->nd_flag |= ND_USEGSSNAME;
4066 		error = nfscl_request(nd, vp, p, cred, stuff);
4067 		if (error)
4068 			return (error);
4069 		if (nd->nd_repstat == 0) {
4070 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4071 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4072 			    cred);
4073 			if (!error) {
4074 				nmp->nm_fsid[0] = nap->na_filesid[0];
4075 				nmp->nm_fsid[1] = nap->na_filesid[1];
4076 				NFSSETHASSETFSID(nmp);
4077 				*attrflagp = 1;
4078 			}
4079 		} else {
4080 			error = nd->nd_repstat;
4081 		}
4082 		if (error)
4083 			goto nfsmout;
4084 	} else {
4085 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4086 		error = nfscl_request(nd, vp, p, cred, stuff);
4087 		if (error)
4088 			return (error);
4089 		if (nd->nd_flag & ND_NFSV3) {
4090 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4091 			if (error)
4092 				goto nfsmout;
4093 		}
4094 		if (nd->nd_repstat) {
4095 			error = nd->nd_repstat;
4096 			goto nfsmout;
4097 		}
4098 		NFSM_DISSECT(tl, u_int32_t *,
4099 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4100 	}
4101 	if (NFSHASNFSV3(nmp)) {
4102 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4103 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4104 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4105 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4106 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4107 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4108 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4109 	} else if (NFSHASNFSV4(nmp) == 0) {
4110 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4111 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4112 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4113 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4114 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4115 	}
4116 nfsmout:
4117 	mbuf_freem(nd->nd_mrep);
4118 	return (error);
4119 }
4120 
4121 /*
4122  * nfs pathconf rpc
4123  */
4124 APPLESTATIC int
4125 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4126     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4127     void *stuff)
4128 {
4129 	struct nfsrv_descript nfsd, *nd = &nfsd;
4130 	struct nfsmount *nmp;
4131 	u_int32_t *tl;
4132 	nfsattrbit_t attrbits;
4133 	int error;
4134 
4135 	*attrflagp = 0;
4136 	nmp = VFSTONFS(vnode_mount(vp));
4137 	if (NFSHASNFSV4(nmp)) {
4138 		/*
4139 		 * For V4, you actually do a getattr.
4140 		 */
4141 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4142 		NFSPATHCONF_GETATTRBIT(&attrbits);
4143 		(void) nfsrv_putattrbit(nd, &attrbits);
4144 		nd->nd_flag |= ND_USEGSSNAME;
4145 		error = nfscl_request(nd, vp, p, cred, stuff);
4146 		if (error)
4147 			return (error);
4148 		if (nd->nd_repstat == 0) {
4149 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4150 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4151 			    cred);
4152 			if (!error)
4153 				*attrflagp = 1;
4154 		} else {
4155 			error = nd->nd_repstat;
4156 		}
4157 	} else {
4158 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4159 		error = nfscl_request(nd, vp, p, cred, stuff);
4160 		if (error)
4161 			return (error);
4162 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4163 		if (nd->nd_repstat && !error)
4164 			error = nd->nd_repstat;
4165 		if (!error) {
4166 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4167 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4168 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4169 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4170 			pc->pc_chownrestricted =
4171 			    fxdr_unsigned(u_int32_t, *tl++);
4172 			pc->pc_caseinsensitive =
4173 			    fxdr_unsigned(u_int32_t, *tl++);
4174 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4175 		}
4176 	}
4177 nfsmout:
4178 	mbuf_freem(nd->nd_mrep);
4179 	return (error);
4180 }
4181 
4182 /*
4183  * nfs version 3 fsinfo rpc call
4184  */
4185 APPLESTATIC int
4186 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4187     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4188 {
4189 	u_int32_t *tl;
4190 	struct nfsrv_descript nfsd, *nd = &nfsd;
4191 	int error;
4192 
4193 	*attrflagp = 0;
4194 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4195 	error = nfscl_request(nd, vp, p, cred, stuff);
4196 	if (error)
4197 		return (error);
4198 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4199 	if (nd->nd_repstat && !error)
4200 		error = nd->nd_repstat;
4201 	if (!error) {
4202 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4203 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4204 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4205 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4206 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4207 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4208 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4209 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4210 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4211 		tl += 2;
4212 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4213 		tl += 2;
4214 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4215 	}
4216 nfsmout:
4217 	mbuf_freem(nd->nd_mrep);
4218 	return (error);
4219 }
4220 
4221 /*
4222  * This function performs the Renew RPC.
4223  */
4224 APPLESTATIC int
4225 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4226     NFSPROC_T *p)
4227 {
4228 	u_int32_t *tl;
4229 	struct nfsrv_descript nfsd;
4230 	struct nfsrv_descript *nd = &nfsd;
4231 	struct nfsmount *nmp;
4232 	int error;
4233 	struct nfssockreq *nrp;
4234 
4235 	nmp = clp->nfsc_nmp;
4236 	if (nmp == NULL)
4237 		return (0);
4238 	nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4239 	    &dsp->nfsclds_sess);
4240 	if (!NFSHASNFSV4N(nmp)) {
4241 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4242 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4243 		*tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
4244 		*tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
4245 	}
4246 	nrp = dsp->nfsclds_sockp;
4247 	if (nrp == NULL)
4248 		/* If NULL, use the MDS socket. */
4249 		nrp = &nmp->nm_sockreq;
4250 	nd->nd_flag |= ND_USEGSSNAME;
4251 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4252 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4253 	if (error)
4254 		return (error);
4255 	error = nd->nd_repstat;
4256 	mbuf_freem(nd->nd_mrep);
4257 	return (error);
4258 }
4259 
4260 /*
4261  * This function performs the Releaselockowner RPC.
4262  */
4263 APPLESTATIC int
4264 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4265     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4266 {
4267 	struct nfsrv_descript nfsd, *nd = &nfsd;
4268 	u_int32_t *tl;
4269 	int error;
4270 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4271 
4272 	if (NFSHASNFSV4N(nmp)) {
4273 		/* For NFSv4.1, do a FreeStateID. */
4274 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4275 		    NULL);
4276 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4277 	} else {
4278 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4279 		    NULL);
4280 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4281 		*tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
4282 		*tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
4283 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4284 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4285 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4286 	}
4287 	nd->nd_flag |= ND_USEGSSNAME;
4288 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4289 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4290 	if (error)
4291 		return (error);
4292 	error = nd->nd_repstat;
4293 	mbuf_freem(nd->nd_mrep);
4294 	return (error);
4295 }
4296 
4297 /*
4298  * This function performs the Compound to get the mount pt FH.
4299  */
4300 APPLESTATIC int
4301 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4302     NFSPROC_T *p)
4303 {
4304 	u_int32_t *tl;
4305 	struct nfsrv_descript nfsd;
4306 	struct nfsrv_descript *nd = &nfsd;
4307 	u_char *cp, *cp2;
4308 	int error, cnt, len, setnil;
4309 	u_int32_t *opcntp;
4310 
4311 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL);
4312 	cp = dirpath;
4313 	cnt = 0;
4314 	do {
4315 		setnil = 0;
4316 		while (*cp == '/')
4317 			cp++;
4318 		cp2 = cp;
4319 		while (*cp2 != '\0' && *cp2 != '/')
4320 			cp2++;
4321 		if (*cp2 == '/') {
4322 			setnil = 1;
4323 			*cp2 = '\0';
4324 		}
4325 		if (cp2 != cp) {
4326 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4327 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4328 			nfsm_strtom(nd, cp, strlen(cp));
4329 			cnt++;
4330 		}
4331 		if (setnil)
4332 			*cp2++ = '/';
4333 		cp = cp2;
4334 	} while (*cp != '\0');
4335 	if (NFSHASNFSV4N(nmp))
4336 		/* Has a Sequence Op done by nfscl_reqstart(). */
4337 		*opcntp = txdr_unsigned(3 + cnt);
4338 	else
4339 		*opcntp = txdr_unsigned(2 + cnt);
4340 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4341 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4342 	nd->nd_flag |= ND_USEGSSNAME;
4343 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4344 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4345 	if (error)
4346 		return (error);
4347 	if (nd->nd_repstat == 0) {
4348 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4349 		tl += (2 + 2 * cnt);
4350 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4351 			len > NFSX_FHMAX) {
4352 			nd->nd_repstat = NFSERR_BADXDR;
4353 		} else {
4354 			nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4355 			if (nd->nd_repstat == 0)
4356 				nmp->nm_fhsize = len;
4357 		}
4358 	}
4359 	error = nd->nd_repstat;
4360 nfsmout:
4361 	mbuf_freem(nd->nd_mrep);
4362 	return (error);
4363 }
4364 
4365 /*
4366  * This function performs the Delegreturn RPC.
4367  */
4368 APPLESTATIC int
4369 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4370     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4371 {
4372 	u_int32_t *tl;
4373 	struct nfsrv_descript nfsd;
4374 	struct nfsrv_descript *nd = &nfsd;
4375 	int error;
4376 
4377 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4378 	    dp->nfsdl_fhlen, NULL, NULL);
4379 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4380 	if (NFSHASNFSV4N(nmp))
4381 		*tl++ = 0;
4382 	else
4383 		*tl++ = dp->nfsdl_stateid.seqid;
4384 	*tl++ = dp->nfsdl_stateid.other[0];
4385 	*tl++ = dp->nfsdl_stateid.other[1];
4386 	*tl = dp->nfsdl_stateid.other[2];
4387 	if (syscred)
4388 		nd->nd_flag |= ND_USEGSSNAME;
4389 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4390 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4391 	if (error)
4392 		return (error);
4393 	error = nd->nd_repstat;
4394 	mbuf_freem(nd->nd_mrep);
4395 	return (error);
4396 }
4397 
4398 /*
4399  * nfs getacl call.
4400  */
4401 APPLESTATIC int
4402 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4403     struct acl *aclp, void *stuff)
4404 {
4405 	struct nfsrv_descript nfsd, *nd = &nfsd;
4406 	int error;
4407 	nfsattrbit_t attrbits;
4408 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4409 
4410 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4411 		return (EOPNOTSUPP);
4412 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4413 	NFSZERO_ATTRBIT(&attrbits);
4414 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4415 	(void) nfsrv_putattrbit(nd, &attrbits);
4416 	error = nfscl_request(nd, vp, p, cred, stuff);
4417 	if (error)
4418 		return (error);
4419 	if (!nd->nd_repstat)
4420 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4421 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4422 	else
4423 		error = nd->nd_repstat;
4424 	mbuf_freem(nd->nd_mrep);
4425 	return (error);
4426 }
4427 
4428 /*
4429  * nfs setacl call.
4430  */
4431 APPLESTATIC int
4432 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4433     struct acl *aclp, void *stuff)
4434 {
4435 	int error;
4436 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4437 
4438 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4439 		return (EOPNOTSUPP);
4440 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4441 	return (error);
4442 }
4443 
4444 /*
4445  * nfs setacl call.
4446  */
4447 static int
4448 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4449     struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4450 {
4451 	struct nfsrv_descript nfsd, *nd = &nfsd;
4452 	int error;
4453 	nfsattrbit_t attrbits;
4454 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4455 
4456 	if (!NFSHASNFSV4(nmp))
4457 		return (EOPNOTSUPP);
4458 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4459 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4460 	NFSZERO_ATTRBIT(&attrbits);
4461 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4462 	(void) nfsv4_fillattr(nd, vnode_mount(vp), vp, aclp, NULL, NULL, 0,
4463 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0);
4464 	error = nfscl_request(nd, vp, p, cred, stuff);
4465 	if (error)
4466 		return (error);
4467 	/* Don't care about the pre/postop attributes */
4468 	mbuf_freem(nd->nd_mrep);
4469 	return (nd->nd_repstat);
4470 }
4471 
4472 /*
4473  * Do the NFSv4.1 Exchange ID.
4474  */
4475 int
4476 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
4477     struct nfssockreq *nrp, uint32_t exchflags, struct nfsclds **dspp,
4478     struct ucred *cred, NFSPROC_T *p)
4479 {
4480 	uint32_t *tl, v41flags;
4481 	struct nfsrv_descript nfsd;
4482 	struct nfsrv_descript *nd = &nfsd;
4483 	struct nfsclds *dsp;
4484 	struct timespec verstime;
4485 	int error, len;
4486 
4487 	*dspp = NULL;
4488 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL);
4489 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4490 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
4491 	*tl = txdr_unsigned(clp->nfsc_rev);
4492 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
4493 
4494 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4495 	*tl++ = txdr_unsigned(exchflags);
4496 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
4497 
4498 	/* Set the implementation id4 */
4499 	*tl = txdr_unsigned(1);
4500 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
4501 	(void) nfsm_strtom(nd, version, strlen(version));
4502 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
4503 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
4504 	verstime.tv_nsec = 0;
4505 	txdr_nfsv4time(&verstime, tl);
4506 	nd->nd_flag |= ND_USEGSSNAME;
4507 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4508 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4509 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
4510 	    (int)nd->nd_repstat);
4511 	if (error != 0)
4512 		return (error);
4513 	if (nd->nd_repstat == 0) {
4514 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
4515 		len = fxdr_unsigned(int, *(tl + 7));
4516 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
4517 			error = NFSERR_BADXDR;
4518 			goto nfsmout;
4519 		}
4520 		dsp = malloc(sizeof(struct nfsclds) + len, M_NFSCLDS,
4521 		    M_WAITOK | M_ZERO);
4522 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
4523 		dsp->nfsclds_servownlen = len;
4524 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
4525 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
4526 		dsp->nfsclds_sess.nfsess_sequenceid =
4527 		    fxdr_unsigned(uint32_t, *tl++);
4528 		v41flags = fxdr_unsigned(uint32_t, *tl);
4529 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
4530 		    NFSHASPNFSOPT(nmp)) {
4531 			NFSCL_DEBUG(1, "set PNFS\n");
4532 			NFSLOCKMNT(nmp);
4533 			nmp->nm_state |= NFSSTA_PNFS;
4534 			NFSUNLOCKMNT(nmp);
4535 			dsp->nfsclds_flags |= NFSCLDS_MDS;
4536 		}
4537 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
4538 			dsp->nfsclds_flags |= NFSCLDS_DS;
4539 		if (len > 0)
4540 			nd->nd_repstat = nfsrv_mtostr(nd,
4541 			    dsp->nfsclds_serverown, len);
4542 		if (nd->nd_repstat == 0) {
4543 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
4544 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
4545 			    NULL, MTX_DEF);
4546 			nfscl_initsessionslots(&dsp->nfsclds_sess);
4547 			*dspp = dsp;
4548 		} else
4549 			free(dsp, M_NFSCLDS);
4550 	}
4551 	error = nd->nd_repstat;
4552 nfsmout:
4553 	mbuf_freem(nd->nd_mrep);
4554 	return (error);
4555 }
4556 
4557 /*
4558  * Do the NFSv4.1 Create Session.
4559  */
4560 int
4561 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
4562     struct nfssockreq *nrp, uint32_t sequenceid, int mds, struct ucred *cred,
4563     NFSPROC_T *p)
4564 {
4565 	uint32_t crflags, *tl;
4566 	struct nfsrv_descript nfsd;
4567 	struct nfsrv_descript *nd = &nfsd;
4568 	int error, irdcnt;
4569 
4570 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL);
4571 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
4572 	*tl++ = sep->nfsess_clientid.lval[0];
4573 	*tl++ = sep->nfsess_clientid.lval[1];
4574 	*tl++ = txdr_unsigned(sequenceid);
4575 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
4576 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0)
4577 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
4578 	*tl = txdr_unsigned(crflags);
4579 
4580 	/* Fill in fore channel attributes. */
4581 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4582 	*tl++ = 0;				/* Header pad size */
4583 	*tl++ = txdr_unsigned(100000);		/* Max request size */
4584 	*tl++ = txdr_unsigned(100000);		/* Max response size */
4585 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4586 	*tl++ = txdr_unsigned(20);		/* Max operations */
4587 	*tl++ = txdr_unsigned(64);		/* Max slots */
4588 	*tl = 0;				/* No rdma ird */
4589 
4590 	/* Fill in back channel attributes. */
4591 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4592 	*tl++ = 0;				/* Header pad size */
4593 	*tl++ = txdr_unsigned(10000);		/* Max request size */
4594 	*tl++ = txdr_unsigned(10000);		/* Max response size */
4595 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4596 	*tl++ = txdr_unsigned(4);		/* Max operations */
4597 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
4598 	*tl = 0;				/* No rdma ird */
4599 
4600 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
4601 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
4602 
4603 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
4604 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
4605 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
4606 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
4607 	*tl++ = 0;				/* Null machine name */
4608 	*tl++ = 0;				/* Uid == 0 */
4609 	*tl++ = 0;				/* Gid == 0 */
4610 	*tl = 0;				/* No additional gids */
4611 	nd->nd_flag |= ND_USEGSSNAME;
4612 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
4613 	    NFS_VER4, NULL, 1, NULL, NULL);
4614 	if (error != 0)
4615 		return (error);
4616 	if (nd->nd_repstat == 0) {
4617 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4618 		    2 * NFSX_UNSIGNED);
4619 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
4620 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
4621 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
4622 		crflags = fxdr_unsigned(uint32_t, *tl);
4623 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
4624 			NFSLOCKMNT(nmp);
4625 			nmp->nm_state |= NFSSTA_SESSPERSIST;
4626 			NFSUNLOCKMNT(nmp);
4627 		}
4628 
4629 		/* Get the fore channel slot count. */
4630 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4631 		tl += 3;		/* Skip the other counts. */
4632 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
4633 		tl++;
4634 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
4635 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
4636 		irdcnt = fxdr_unsigned(int, *tl);
4637 		if (irdcnt > 0)
4638 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
4639 
4640 		/* and the back channel slot count. */
4641 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4642 		tl += 5;
4643 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
4644 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
4645 	}
4646 	error = nd->nd_repstat;
4647 nfsmout:
4648 	mbuf_freem(nd->nd_mrep);
4649 	return (error);
4650 }
4651 
4652 /*
4653  * Do the NFSv4.1 Destroy Session.
4654  */
4655 int
4656 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
4657     struct ucred *cred, NFSPROC_T *p)
4658 {
4659 	uint32_t *tl;
4660 	struct nfsrv_descript nfsd;
4661 	struct nfsrv_descript *nd = &nfsd;
4662 	int error;
4663 
4664 	nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL);
4665 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
4666 	bcopy(NFSMNT_MDSSESSION(nmp)->nfsess_sessionid, tl, NFSX_V4SESSIONID);
4667 	nd->nd_flag |= ND_USEGSSNAME;
4668 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4669 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4670 	if (error != 0)
4671 		return (error);
4672 	error = nd->nd_repstat;
4673 	mbuf_freem(nd->nd_mrep);
4674 	return (error);
4675 }
4676 
4677 /*
4678  * Do the NFSv4.1 Destroy Client.
4679  */
4680 int
4681 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
4682     struct ucred *cred, NFSPROC_T *p)
4683 {
4684 	uint32_t *tl;
4685 	struct nfsrv_descript nfsd;
4686 	struct nfsrv_descript *nd = &nfsd;
4687 	int error;
4688 
4689 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL);
4690 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4691 	*tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
4692 	*tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
4693 	nd->nd_flag |= ND_USEGSSNAME;
4694 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4695 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4696 	if (error != 0)
4697 		return (error);
4698 	error = nd->nd_repstat;
4699 	mbuf_freem(nd->nd_mrep);
4700 	return (error);
4701 }
4702 
4703 /*
4704  * Do the NFSv4.1 LayoutGet.
4705  */
4706 int
4707 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
4708     uint64_t offset, uint64_t len, uint64_t minlen, int layoutlen,
4709     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp,
4710     struct ucred *cred, NFSPROC_T *p, void *stuff)
4711 {
4712 	uint32_t *tl;
4713 	struct nfsrv_descript nfsd, *nd = &nfsd;
4714 	struct nfsfh *nfhp;
4715 	struct nfsclflayout *flp, *prevflp, *tflp;
4716 	int cnt, error, gotiomode, fhcnt, nfhlen, i, j;
4717 	uint8_t *cp;
4718 	uint64_t retlen;
4719 
4720 	flp = NULL;
4721 	gotiomode = -1;
4722 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL);
4723 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
4724 	    NFSX_STATEID);
4725 	*tl++ = newnfs_false;		/* Don't signal availability. */
4726 	*tl++ = txdr_unsigned(NFSLAYOUT_NFSV4_1_FILES);
4727 	*tl++ = txdr_unsigned(iomode);
4728 	txdr_hyper(offset, tl);
4729 	tl += 2;
4730 	txdr_hyper(len, tl);
4731 	tl += 2;
4732 	txdr_hyper(minlen, tl);
4733 	tl += 2;
4734 	*tl++ = txdr_unsigned(stateidp->seqid);
4735 	NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
4736 	*tl++ = stateidp->other[0];
4737 	*tl++ = stateidp->other[1];
4738 	*tl++ = stateidp->other[2];
4739 	*tl = txdr_unsigned(layoutlen);
4740 	nd->nd_flag |= ND_USEGSSNAME;
4741 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4742 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4743 	if (error != 0)
4744 		return (error);
4745 	if (nd->nd_repstat == 0) {
4746 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
4747 		if (*tl++ != 0)
4748 			*retonclosep = 1;
4749 		else
4750 			*retonclosep = 0;
4751 		stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
4752 		NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
4753 		    (int)stateidp->seqid);
4754 		stateidp->other[0] = *tl++;
4755 		stateidp->other[1] = *tl++;
4756 		stateidp->other[2] = *tl++;
4757 		cnt = fxdr_unsigned(int, *tl);
4758 		NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
4759 		if (cnt <= 0 || cnt > 10000) {
4760 			/* Don't accept more than 10000 layouts in reply. */
4761 			error = NFSERR_BADXDR;
4762 			goto nfsmout;
4763 		}
4764 		for (i = 0; i < cnt; i++) {
4765 			/* Dissect all the way to the file handle cnt. */
4766 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_HYPER +
4767 			    6 * NFSX_UNSIGNED + NFSX_V4DEVICEID);
4768 			fhcnt = fxdr_unsigned(int, *(tl + 11 +
4769 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
4770 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
4771 			if (fhcnt < 0 || fhcnt > 100) {
4772 				/* Don't accept more than 100 file handles. */
4773 				error = NFSERR_BADXDR;
4774 				goto nfsmout;
4775 			}
4776 			if (fhcnt > 1)
4777 				flp = malloc(sizeof(*flp) + (fhcnt - 1) *
4778 				    sizeof(struct nfsfh *),
4779 				    M_NFSFLAYOUT, M_WAITOK);
4780 			else
4781 				flp = malloc(sizeof(*flp),
4782 				    M_NFSFLAYOUT, M_WAITOK);
4783 			flp->nfsfl_flags = 0;
4784 			flp->nfsfl_fhcnt = 0;
4785 			flp->nfsfl_devp = NULL;
4786 			flp->nfsfl_off = fxdr_hyper(tl); tl += 2;
4787 			retlen = fxdr_hyper(tl); tl += 2;
4788 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
4789 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
4790 			else
4791 				flp->nfsfl_end = flp->nfsfl_off + retlen;
4792 			flp->nfsfl_iomode = fxdr_unsigned(int, *tl++);
4793 			if (gotiomode == -1)
4794 				gotiomode = flp->nfsfl_iomode;
4795 			NFSCL_DEBUG(4, "layg reqiom=%d retiom=%d\n", iomode,
4796 			    (int)flp->nfsfl_iomode);
4797 			if (fxdr_unsigned(int, *tl++) !=
4798 			    NFSLAYOUT_NFSV4_1_FILES) {
4799 				printf("NFSv4.1: got non-files layout\n");
4800 				error = NFSERR_BADXDR;
4801 				goto nfsmout;
4802 			}
4803 			NFSBCOPY(++tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
4804 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
4805 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
4806 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
4807 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
4808 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
4809 			if (fxdr_unsigned(int, *tl) != fhcnt) {
4810 				printf("EEK! bad fhcnt\n");
4811 				error = NFSERR_BADXDR;
4812 				goto nfsmout;
4813 			}
4814 			for (j = 0; j < fhcnt; j++) {
4815 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4816 				nfhlen = fxdr_unsigned(int, *tl);
4817 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
4818 					error = NFSERR_BADXDR;
4819 					goto nfsmout;
4820 				}
4821 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
4822 				    M_NFSFH, M_WAITOK);
4823 				flp->nfsfl_fh[j] = nfhp;
4824 				flp->nfsfl_fhcnt++;
4825 				nfhp->nfh_len = nfhlen;
4826 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
4827 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
4828 			}
4829 			if (flp->nfsfl_iomode == gotiomode) {
4830 				/* Keep the list in increasing offset order. */
4831 				tflp = LIST_FIRST(flhp);
4832 				prevflp = NULL;
4833 				while (tflp != NULL &&
4834 				    tflp->nfsfl_off < flp->nfsfl_off) {
4835 					prevflp = tflp;
4836 					tflp = LIST_NEXT(tflp, nfsfl_list);
4837 				}
4838 				if (prevflp == NULL)
4839 					LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
4840 				else
4841 					LIST_INSERT_AFTER(prevflp, flp,
4842 					    nfsfl_list);
4843 			} else {
4844 				printf("nfscl_layoutget(): got wrong iomode\n");
4845 				nfscl_freeflayout(flp);
4846 			}
4847 			flp = NULL;
4848 		}
4849 	}
4850 	if (nd->nd_repstat != 0 && error == 0)
4851 		error = nd->nd_repstat;
4852 nfsmout:
4853 	if (error != 0 && flp != NULL)
4854 		nfscl_freeflayout(flp);
4855 	mbuf_freem(nd->nd_mrep);
4856 	return (error);
4857 }
4858 
4859 /*
4860  * Do the NFSv4.1 Get Device Info.
4861  */
4862 int
4863 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
4864     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
4865     NFSPROC_T *p)
4866 {
4867 	uint32_t cnt, *tl;
4868 	struct nfsrv_descript nfsd;
4869 	struct nfsrv_descript *nd = &nfsd;
4870 	struct sockaddr_storage ss;
4871 	struct nfsclds *dsp = NULL, **dspp;
4872 	struct nfscldevinfo *ndi;
4873 	int addrcnt, bitcnt, error, i, isudp, j, pos, safilled, stripecnt;
4874 	uint8_t stripeindex;
4875 
4876 	*ndip = NULL;
4877 	ndi = NULL;
4878 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL);
4879 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
4880 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
4881 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
4882 	*tl++ = txdr_unsigned(layouttype);
4883 	*tl++ = txdr_unsigned(100000);
4884 	if (notifybitsp != NULL && *notifybitsp != 0) {
4885 		*tl = txdr_unsigned(1);		/* One word of bits. */
4886 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
4887 		*tl = txdr_unsigned(*notifybitsp);
4888 	} else
4889 		*tl = txdr_unsigned(0);
4890 	nd->nd_flag |= ND_USEGSSNAME;
4891 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4892 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4893 	if (error != 0)
4894 		return (error);
4895 	if (nd->nd_repstat == 0) {
4896 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4897 		if (layouttype != fxdr_unsigned(int, *tl++))
4898 			printf("EEK! devinfo layout type not same!\n");
4899 		stripecnt = fxdr_unsigned(int, *++tl);
4900 		NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
4901 		if (stripecnt < 1 || stripecnt > 4096) {
4902 			printf("NFS devinfo stripecnt %d: out of range\n",
4903 			    stripecnt);
4904 			error = NFSERR_BADXDR;
4905 			goto nfsmout;
4906 		}
4907 		NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) * NFSX_UNSIGNED);
4908 		addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
4909 		NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
4910 		if (addrcnt < 1 || addrcnt > 128) {
4911 			printf("NFS devinfo addrcnt %d: out of range\n",
4912 			    addrcnt);
4913 			error = NFSERR_BADXDR;
4914 			goto nfsmout;
4915 		}
4916 
4917 		/*
4918 		 * Now we know how many stripe indices and addresses, so
4919 		 * we can allocate the structure the correct size.
4920 		 */
4921 		i = (stripecnt * sizeof(uint8_t)) / sizeof(struct nfsclds *)
4922 		    + 1;
4923 		NFSCL_DEBUG(4, "stripeindices=%d\n", i);
4924 		ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
4925 		    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK | M_ZERO);
4926 		NFSBCOPY(deviceid, ndi->nfsdi_deviceid, NFSX_V4DEVICEID);
4927 		ndi->nfsdi_refcnt = 0;
4928 		ndi->nfsdi_stripecnt = stripecnt;
4929 		ndi->nfsdi_addrcnt = addrcnt;
4930 		/* Fill in the stripe indices. */
4931 		for (i = 0; i < stripecnt; i++) {
4932 			stripeindex = fxdr_unsigned(uint8_t, *tl++);
4933 			NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
4934 			if (stripeindex >= addrcnt) {
4935 				printf("NFS devinfo stripeindex %d: too big\n",
4936 				    (int)stripeindex);
4937 				error = NFSERR_BADXDR;
4938 				goto nfsmout;
4939 			}
4940 			nfsfldi_setstripeindex(ndi, i, stripeindex);
4941 		}
4942 
4943 		/* Now, dissect the server address(es). */
4944 		safilled = 0;
4945 		for (i = 0; i < addrcnt; i++) {
4946 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4947 			cnt = fxdr_unsigned(uint32_t, *tl);
4948 			if (cnt == 0) {
4949 				printf("NFS devinfo 0 len addrlist\n");
4950 				error = NFSERR_BADXDR;
4951 				goto nfsmout;
4952 			}
4953 			dspp = nfsfldi_addr(ndi, i);
4954 			pos = arc4random() % cnt;	/* Choose one. */
4955 			safilled = 0;
4956 			for (j = 0; j < cnt; j++) {
4957 				error = nfsv4_getipaddr(nd, &ss, &isudp);
4958 				if (error != 0 && error != EPERM) {
4959 					error = NFSERR_BADXDR;
4960 					goto nfsmout;
4961 				}
4962 				if (error == 0 && isudp == 0) {
4963 					/*
4964 					 * The algorithm is:
4965 					 * - use "pos" entry if it is of the
4966 					 *   same af_family or none of them
4967 					 *   is of the same af_family
4968 					 * else
4969 					 * - use the first one of the same
4970 					 *   af_family.
4971 					 */
4972 					if ((safilled == 0 && ss.ss_family ==
4973 					     nmp->nm_nam->sa_family) ||
4974 					    (j == pos &&
4975 					     (safilled == 0 || ss.ss_family ==
4976 					      nmp->nm_nam->sa_family)) ||
4977 					    (safilled == 1 && ss.ss_family ==
4978 					     nmp->nm_nam->sa_family)) {
4979 						error = nfsrpc_fillsa(nmp, &ss,
4980 						    &dsp, p);
4981 						if (error == 0) {
4982 							*dspp = dsp;
4983 							if (ss.ss_family ==
4984 							 nmp->nm_nam->sa_family)
4985 								safilled = 2;
4986 							else
4987 								safilled = 1;
4988 						}
4989 					}
4990 				}
4991 			}
4992 			if (safilled == 0)
4993 				break;
4994 		}
4995 
4996 		/* And the notify bits. */
4997 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4998 		if (safilled != 0) {
4999 			bitcnt = fxdr_unsigned(int, *tl);
5000 			if (bitcnt > 0) {
5001 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5002 				if (notifybitsp != NULL)
5003 					*notifybitsp =
5004 					    fxdr_unsigned(uint32_t, *tl);
5005 			}
5006 			*ndip = ndi;
5007 		} else
5008 			error = EPERM;
5009 	}
5010 	if (nd->nd_repstat != 0)
5011 		error = nd->nd_repstat;
5012 nfsmout:
5013 	if (error != 0 && ndi != NULL)
5014 		nfscl_freedevinfo(ndi);
5015 	mbuf_freem(nd->nd_mrep);
5016 	return (error);
5017 }
5018 
5019 /*
5020  * Do the NFSv4.1 LayoutCommit.
5021  */
5022 int
5023 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5024     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5025     int layouttype, int layoutupdatecnt, uint8_t *layp, struct ucred *cred,
5026     NFSPROC_T *p, void *stuff)
5027 {
5028 	uint32_t *tl;
5029 	struct nfsrv_descript nfsd, *nd = &nfsd;
5030 	int error, outcnt, i;
5031 	uint8_t *cp;
5032 
5033 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL);
5034 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5035 	    NFSX_STATEID);
5036 	txdr_hyper(off, tl);
5037 	tl += 2;
5038 	txdr_hyper(len, tl);
5039 	tl += 2;
5040 	if (reclaim != 0)
5041 		*tl++ = newnfs_true;
5042 	else
5043 		*tl++ = newnfs_false;
5044 	*tl++ = txdr_unsigned(stateidp->seqid);
5045 	*tl++ = stateidp->other[0];
5046 	*tl++ = stateidp->other[1];
5047 	*tl++ = stateidp->other[2];
5048 	*tl++ = newnfs_true;
5049 	if (lastbyte < off)
5050 		lastbyte = off;
5051 	else if (lastbyte >= (off + len))
5052 		lastbyte = off + len - 1;
5053 	txdr_hyper(lastbyte, tl);
5054 	tl += 2;
5055 	*tl++ = newnfs_false;
5056 	*tl++ = txdr_unsigned(layouttype);
5057 	*tl = txdr_unsigned(layoutupdatecnt);
5058 	if (layoutupdatecnt > 0) {
5059 		KASSERT(layouttype != NFSLAYOUT_NFSV4_1_FILES,
5060 		    ("Must be nil for Files Layout"));
5061 		outcnt = NFSM_RNDUP(layoutupdatecnt);
5062 		NFSM_BUILD(cp, uint8_t *, outcnt);
5063 		NFSBCOPY(layp, cp, layoutupdatecnt);
5064 		cp += layoutupdatecnt;
5065 		for (i = 0; i < (outcnt - layoutupdatecnt); i++)
5066 			*cp++ = 0x0;
5067 	}
5068 	nd->nd_flag |= ND_USEGSSNAME;
5069 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5070 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5071 	if (error != 0)
5072 		return (error);
5073 	error = nd->nd_repstat;
5074 	mbuf_freem(nd->nd_mrep);
5075 	return (error);
5076 }
5077 
5078 /*
5079  * Do the NFSv4.1 LayoutReturn.
5080  */
5081 int
5082 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5083     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5084     uint64_t len, nfsv4stateid_t *stateidp, int layoutcnt, uint32_t *layp,
5085     struct ucred *cred, NFSPROC_T *p, void *stuff)
5086 {
5087 	uint32_t *tl;
5088 	struct nfsrv_descript nfsd, *nd = &nfsd;
5089 	int error, outcnt, i;
5090 	uint8_t *cp;
5091 
5092 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL);
5093 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5094 	if (reclaim != 0)
5095 		*tl++ = newnfs_true;
5096 	else
5097 		*tl++ = newnfs_false;
5098 	*tl++ = txdr_unsigned(layouttype);
5099 	*tl++ = txdr_unsigned(iomode);
5100 	*tl = txdr_unsigned(layoutreturn);
5101 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5102 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5103 		    NFSX_UNSIGNED);
5104 		txdr_hyper(offset, tl);
5105 		tl += 2;
5106 		txdr_hyper(len, tl);
5107 		tl += 2;
5108 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5109 		*tl++ = txdr_unsigned(stateidp->seqid);
5110 		*tl++ = stateidp->other[0];
5111 		*tl++ = stateidp->other[1];
5112 		*tl++ = stateidp->other[2];
5113 		*tl = txdr_unsigned(layoutcnt);
5114 		if (layoutcnt > 0) {
5115 			outcnt = NFSM_RNDUP(layoutcnt);
5116 			NFSM_BUILD(cp, uint8_t *, outcnt);
5117 			NFSBCOPY(layp, cp, layoutcnt);
5118 			cp += layoutcnt;
5119 			for (i = 0; i < (outcnt - layoutcnt); i++)
5120 				*cp++ = 0x0;
5121 		}
5122 	}
5123 	nd->nd_flag |= ND_USEGSSNAME;
5124 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5125 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5126 	if (error != 0)
5127 		return (error);
5128 	if (nd->nd_repstat == 0) {
5129 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5130 		if (*tl != 0) {
5131 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5132 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5133 			stateidp->other[0] = *tl++;
5134 			stateidp->other[1] = *tl++;
5135 			stateidp->other[2] = *tl;
5136 		}
5137 	} else
5138 		error = nd->nd_repstat;
5139 nfsmout:
5140 	mbuf_freem(nd->nd_mrep);
5141 	return (error);
5142 }
5143 
5144 /*
5145  * Acquire a layout and devinfo, if possible. The caller must have acquired
5146  * a reference count on the nfsclclient structure before calling this.
5147  * Return the layout in lypp with a reference count on it, if successful.
5148  */
5149 static int
5150 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5151     int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
5152     struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5153 {
5154 	struct nfscllayout *lyp;
5155 	struct nfsclflayout *flp, *tflp;
5156 	struct nfscldevinfo *dip;
5157 	struct nfsclflayouthead flh;
5158 	int error = 0, islocked, layoutlen, recalled, retonclose;
5159 	nfsv4stateid_t stateid;
5160 
5161 	*lypp = NULL;
5162 	/*
5163 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5164 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5165 	 * flp == NULL.
5166 	 */
5167 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5168 	    off, &flp, &recalled);
5169 	islocked = 0;
5170 	if (lyp == NULL || flp == NULL) {
5171 		if (recalled != 0)
5172 			return (EIO);
5173 		LIST_INIT(&flh);
5174 		layoutlen = NFSMNT_MDSSESSION(nmp)->nfsess_maxcache -
5175 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5176 		if (lyp == NULL) {
5177 			stateid.seqid = 0;
5178 			stateid.other[0] = stateidp->other[0];
5179 			stateid.other[1] = stateidp->other[1];
5180 			stateid.other[2] = stateidp->other[2];
5181 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5182 			    nfhp->nfh_len, iomode, (uint64_t)0, INT64_MAX,
5183 			    (uint64_t)0, layoutlen, &stateid, &retonclose,
5184 			    &flh, cred, p, NULL);
5185 		} else {
5186 			islocked = 1;
5187 			stateid.seqid = lyp->nfsly_stateid.seqid;
5188 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5189 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5190 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5191 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5192 			    nfhp->nfh_len, iomode, off, INT64_MAX,
5193 			    (uint64_t)0, layoutlen, &stateid, &retonclose,
5194 			    &flh, cred, p, NULL);
5195 		}
5196 		if (error == 0)
5197 			LIST_FOREACH(tflp, &flh, nfsfl_list) {
5198 				error = nfscl_adddevinfo(nmp, NULL, tflp);
5199 				if (error != 0) {
5200 					error = nfsrpc_getdeviceinfo(nmp,
5201 					    tflp->nfsfl_dev,
5202 					    NFSLAYOUT_NFSV4_1_FILES,
5203 					    notifybitsp, &dip, cred, p);
5204 					if (error != 0)
5205 						break;
5206 					error = nfscl_adddevinfo(nmp, dip,
5207 					    tflp);
5208 					if (error != 0)
5209 						printf(
5210 						    "getlayout: cannot add\n");
5211 				}
5212 			}
5213 		if (error == 0) {
5214 			/*
5215 			 * nfscl_layout() always returns with the nfsly_lock
5216 			 * set to a refcnt (shared lock).
5217 			 */
5218 			error = nfscl_layout(nmp, vp, nfhp->nfh_fh,
5219 			    nfhp->nfh_len, &stateid, retonclose, &flh, &lyp,
5220 			    cred, p);
5221 			if (error == 0)
5222 				*lypp = lyp;
5223 		} else if (islocked != 0)
5224 			nfsv4_unlock(&lyp->nfsly_lock, 0);
5225 	} else
5226 		*lypp = lyp;
5227 	return (error);
5228 }
5229 
5230 /*
5231  * Do a TCP connection plus exchange id and create session.
5232  * If successful, a "struct nfsclds" is linked into the list for the
5233  * mount point and a pointer to it is returned.
5234  */
5235 static int
5236 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_storage *ssp,
5237     struct nfsclds **dspp, NFSPROC_T *p)
5238 {
5239 	struct sockaddr_in *msad, *sad, *ssd;
5240 	struct sockaddr_in6 *msad6, *sad6, *ssd6;
5241 	struct nfsclclient *clp;
5242 	struct nfssockreq *nrp;
5243 	struct nfsclds *dsp, *tdsp;
5244 	int error;
5245 	enum nfsclds_state retv;
5246 	uint32_t sequenceid;
5247 
5248 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5249 	    ("nfsrpc_fillsa: NULL nr_cred"));
5250 	NFSLOCKCLSTATE();
5251 	clp = nmp->nm_clp;
5252 	NFSUNLOCKCLSTATE();
5253 	if (clp == NULL)
5254 		return (EPERM);
5255 	if (ssp->ss_family == AF_INET) {
5256 		ssd = (struct sockaddr_in *)ssp;
5257 		NFSLOCKMNT(nmp);
5258 
5259 		/*
5260 		 * Check to see if we already have a session for this
5261 		 * address that is usable for a DS.
5262 		 * Note that the MDS's address is in a different place
5263 		 * than the sessions already acquired for DS's.
5264 		 */
5265 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5266 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5267 		while (tdsp != NULL) {
5268 			if (msad != NULL && msad->sin_family == AF_INET &&
5269 			    ssd->sin_addr.s_addr == msad->sin_addr.s_addr &&
5270 			    ssd->sin_port == msad->sin_port &&
5271 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0) {
5272 				*dspp = tdsp;
5273 				NFSUNLOCKMNT(nmp);
5274 				NFSCL_DEBUG(4, "fnd same addr\n");
5275 				return (0);
5276 			}
5277 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5278 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5279 				msad = (struct sockaddr_in *)
5280 				    tdsp->nfsclds_sockp->nr_nam;
5281 			else
5282 				msad = NULL;
5283 		}
5284 		NFSUNLOCKMNT(nmp);
5285 
5286 		/* No IP address match, so look for new/trunked one. */
5287 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5288 		sad->sin_len = sizeof(*sad);
5289 		sad->sin_family = AF_INET;
5290 		sad->sin_port = ssd->sin_port;
5291 		sad->sin_addr.s_addr = ssd->sin_addr.s_addr;
5292 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5293 		nrp->nr_nam = (struct sockaddr *)sad;
5294 	} else if (ssp->ss_family == AF_INET6) {
5295 		ssd6 = (struct sockaddr_in6 *)ssp;
5296 		NFSLOCKMNT(nmp);
5297 
5298 		/*
5299 		 * Check to see if we already have a session for this
5300 		 * address that is usable for a DS.
5301 		 * Note that the MDS's address is in a different place
5302 		 * than the sessions already acquired for DS's.
5303 		 */
5304 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5305 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5306 		while (tdsp != NULL) {
5307 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5308 			    IN6_ARE_ADDR_EQUAL(&ssd6->sin6_addr,
5309 			    &msad6->sin6_addr) &&
5310 			    ssd6->sin6_port == msad6->sin6_port &&
5311 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0) {
5312 				*dspp = tdsp;
5313 				NFSUNLOCKMNT(nmp);
5314 				return (0);
5315 			}
5316 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5317 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5318 				msad6 = (struct sockaddr_in6 *)
5319 				    tdsp->nfsclds_sockp->nr_nam;
5320 			else
5321 				msad6 = NULL;
5322 		}
5323 		NFSUNLOCKMNT(nmp);
5324 
5325 		/* No IP address match, so look for new/trunked one. */
5326 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5327 		sad6->sin6_len = sizeof(*sad6);
5328 		sad6->sin6_family = AF_INET6;
5329 		sad6->sin6_port = ssd6->sin6_port;
5330 		NFSBCOPY(&ssd6->sin6_addr, &sad6->sin6_addr,
5331 		    sizeof(struct in6_addr));
5332 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5333 		nrp->nr_nam = (struct sockaddr *)sad6;
5334 	} else
5335 		return (EPERM);
5336 
5337 	nrp->nr_sotype = SOCK_STREAM;
5338 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5339 	nrp->nr_prog = NFS_PROG;
5340 	nrp->nr_vers = NFS_VER4;
5341 
5342 	/*
5343 	 * Use the credentials that were used for the mount, which are
5344 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5345 	 * Ref. counting the credentials with crhold() is probably not
5346 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5347 	 * unmount, but I did it anyhow.
5348 	 */
5349 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5350 	error = newnfs_connect(nmp, nrp, NULL, p, 0);
5351 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
5352 
5353 	/* Now, do the exchangeid and create session. */
5354 	if (error == 0)
5355 		error = nfsrpc_exchangeid(nmp, clp, nrp, NFSV4EXCH_USEPNFSDS,
5356 		    &dsp, nrp->nr_cred, p);
5357 	NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5358 	if (error == 0) {
5359 		dsp->nfsclds_sockp = nrp;
5360 		NFSLOCKMNT(nmp);
5361 		retv = nfscl_getsameserver(nmp, dsp, &tdsp);
5362 		NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5363 		if (retv == NFSDSP_USETHISSESSION) {
5364 			NFSUNLOCKMNT(nmp);
5365 			/*
5366 			 * If there is already a session for this server,
5367 			 * use it.
5368 			 */
5369 			(void)newnfs_disconnect(nrp);
5370 			nfscl_freenfsclds(dsp);
5371 			*dspp = tdsp;
5372 			return (0);
5373 		}
5374 		if (retv == NFSDSP_SEQTHISSESSION)
5375 			sequenceid = tdsp->nfsclds_sess.nfsess_sequenceid;
5376 		else
5377 			sequenceid = dsp->nfsclds_sess.nfsess_sequenceid;
5378 		NFSUNLOCKMNT(nmp);
5379 		error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5380 		    nrp, sequenceid, 0, nrp->nr_cred, p);
5381 		NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5382 	} else {
5383 		NFSFREECRED(nrp->nr_cred);
5384 		NFSFREEMUTEX(&nrp->nr_mtx);
5385 		free(nrp->nr_nam, M_SONAME);
5386 		free(nrp, M_NFSSOCKREQ);
5387 	}
5388 	if (error == 0) {
5389 		NFSCL_DEBUG(3, "add DS session\n");
5390 		/*
5391 		 * Put it at the end of the list. That way the list
5392 		 * is ordered by when the entry was added. This matters
5393 		 * since the one done first is the one that should be
5394 		 * used for sequencid'ing any subsequent create sessions.
5395 		 */
5396 		NFSLOCKMNT(nmp);
5397 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5398 		NFSUNLOCKMNT(nmp);
5399 		*dspp = dsp;
5400 	} else if (dsp != NULL)
5401 		nfscl_freenfsclds(dsp);
5402 	return (error);
5403 }
5404 
5405 /*
5406  * Do the NFSv4.1 Reclaim Complete.
5407  */
5408 int
5409 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5410 {
5411 	uint32_t *tl;
5412 	struct nfsrv_descript nfsd;
5413 	struct nfsrv_descript *nd = &nfsd;
5414 	int error;
5415 
5416 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL);
5417 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5418 	*tl = newnfs_false;
5419 	nd->nd_flag |= ND_USEGSSNAME;
5420 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5421 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5422 	if (error != 0)
5423 		return (error);
5424 	error = nd->nd_repstat;
5425 	mbuf_freem(nd->nd_mrep);
5426 	return (error);
5427 }
5428 
5429 /*
5430  * Initialize the slot tables for a session.
5431  */
5432 static void
5433 nfscl_initsessionslots(struct nfsclsession *sep)
5434 {
5435 	int i;
5436 
5437 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
5438 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5439 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5440 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5441 	}
5442 	for (i = 0; i < 64; i++)
5443 		sep->nfsess_slotseq[i] = 0;
5444 	sep->nfsess_slots = 0;
5445 }
5446 
5447 /*
5448  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
5449  */
5450 int
5451 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5452     uint32_t rwaccess, struct ucred *cred, NFSPROC_T *p)
5453 {
5454 	struct nfsnode *np = VTONFS(vp);
5455 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5456 	struct nfscllayout *layp;
5457 	struct nfscldevinfo *dip;
5458 	struct nfsclflayout *rflp;
5459 	nfsv4stateid_t stateid;
5460 	struct ucred *newcred;
5461 	uint64_t lastbyte, len, off, oresid, xfer;
5462 	int eof, error, iolaymode, recalled;
5463 	void *lckp;
5464 
5465 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
5466 	    (np->n_flag & NNOLAYOUT) != 0)
5467 		return (EIO);
5468 	/* Now, get a reference cnt on the clientid for this mount. */
5469 	if (nfscl_getref(nmp) == 0)
5470 		return (EIO);
5471 
5472 	/* Find an appropriate stateid. */
5473 	newcred = NFSNEWCRED(cred);
5474 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
5475 	    rwaccess, 1, newcred, p, &stateid, &lckp);
5476 	if (error != 0) {
5477 		NFSFREECRED(newcred);
5478 		nfscl_relref(nmp);
5479 		return (error);
5480 	}
5481 	/* Search for a layout for this file. */
5482 	off = uiop->uio_offset;
5483 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
5484 	    np->n_fhp->nfh_len, off, &rflp, &recalled);
5485 	if (layp == NULL || rflp == NULL) {
5486 		if (recalled != 0) {
5487 			NFSFREECRED(newcred);
5488 			nfscl_relref(nmp);
5489 			return (EIO);
5490 		}
5491 		if (layp != NULL) {
5492 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
5493 			layp = NULL;
5494 		}
5495 		/* Try and get a Layout, if it is supported. */
5496 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
5497 		    (np->n_flag & NWRITEOPENED) != 0)
5498 			iolaymode = NFSLAYOUTIOMODE_RW;
5499 		else
5500 			iolaymode = NFSLAYOUTIOMODE_READ;
5501 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
5502 		    NULL, &stateid, off, &layp, newcred, p);
5503 		if (error != 0) {
5504 			NFSLOCKNODE(np);
5505 			np->n_flag |= NNOLAYOUT;
5506 			NFSUNLOCKNODE(np);
5507 			if (lckp != NULL)
5508 				nfscl_lockderef(lckp);
5509 			NFSFREECRED(newcred);
5510 			if (layp != NULL)
5511 				nfscl_rellayout(layp, 0);
5512 			nfscl_relref(nmp);
5513 			return (error);
5514 		}
5515 	}
5516 
5517 	/*
5518 	 * Loop around finding a layout that works for the first part of
5519 	 * this I/O operation, and then call the function that actually
5520 	 * does the RPC.
5521 	 */
5522 	eof = 0;
5523 	len = (uint64_t)uiop->uio_resid;
5524 	while (len > 0 && error == 0 && eof == 0) {
5525 		off = uiop->uio_offset;
5526 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
5527 		if (error == 0) {
5528 			oresid = xfer = (uint64_t)uiop->uio_resid;
5529 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
5530 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
5531 			dip = nfscl_getdevinfo(nmp->nm_clp, rflp->nfsfl_dev,
5532 			    rflp->nfsfl_devp);
5533 			if (dip != NULL) {
5534 				error = nfscl_doflayoutio(vp, uiop, iomode,
5535 				    must_commit, &eof, &stateid, rwaccess, dip,
5536 				    layp, rflp, off, xfer, newcred, p);
5537 				nfscl_reldevinfo(dip);
5538 				lastbyte = off + xfer - 1;
5539 				if (error == 0) {
5540 					NFSLOCKCLSTATE();
5541 					if (lastbyte > layp->nfsly_lastbyte)
5542 						layp->nfsly_lastbyte = lastbyte;
5543 					NFSUNLOCKCLSTATE();
5544 				}
5545 			} else
5546 				error = EIO;
5547 			if (error == 0)
5548 				len -= (oresid - (uint64_t)uiop->uio_resid);
5549 		}
5550 	}
5551 	if (lckp != NULL)
5552 		nfscl_lockderef(lckp);
5553 	NFSFREECRED(newcred);
5554 	nfscl_rellayout(layp, 0);
5555 	nfscl_relref(nmp);
5556 	return (error);
5557 }
5558 
5559 /*
5560  * Find a file layout that will handle the first bytes of the requested
5561  * range and return the information from it needed to to the I/O operation.
5562  */
5563 int
5564 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
5565     struct nfsclflayout **retflpp)
5566 {
5567 	struct nfsclflayout *flp, *nflp, *rflp;
5568 	uint32_t rw;
5569 
5570 	rflp = NULL;
5571 	rw = rwaccess;
5572 	/* For reading, do the Read list first and then the Write list. */
5573 	do {
5574 		if (rw == NFSV4OPEN_ACCESSREAD)
5575 			flp = LIST_FIRST(&lyp->nfsly_flayread);
5576 		else
5577 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
5578 		while (flp != NULL) {
5579 			nflp = LIST_NEXT(flp, nfsfl_list);
5580 			if (flp->nfsfl_off > off)
5581 				break;
5582 			if (flp->nfsfl_end > off &&
5583 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
5584 				rflp = flp;
5585 			flp = nflp;
5586 		}
5587 		if (rw == NFSV4OPEN_ACCESSREAD)
5588 			rw = NFSV4OPEN_ACCESSWRITE;
5589 		else
5590 			rw = 0;
5591 	} while (rw != 0);
5592 	if (rflp != NULL) {
5593 		/* This one covers the most bytes starting at off. */
5594 		*retflpp = rflp;
5595 		return (0);
5596 	}
5597 	return (EIO);
5598 }
5599 
5600 /*
5601  * Do I/O using an NFSv4.1 file layout.
5602  */
5603 static int
5604 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5605     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
5606     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
5607     uint64_t len, struct ucred *cred, NFSPROC_T *p)
5608 {
5609 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
5610 	int commit_thru_mds, error = 0, stripe_index, stripe_pos;
5611 	struct nfsnode *np;
5612 	struct nfsfh *fhp;
5613 	struct nfsclds **dspp;
5614 
5615 	np = VTONFS(vp);
5616 	rel_off = off - flp->nfsfl_patoff;
5617 	stripe_unit_size = (flp->nfsfl_util >> 6) & 0x3ffffff;
5618 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
5619 	    dp->nfsdi_stripecnt;
5620 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
5621 
5622 	/* Loop around, doing I/O for each stripe unit. */
5623 	while (len > 0 && error == 0) {
5624 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
5625 		dspp = nfsfldi_addr(dp, stripe_index);
5626 		if (len > transfer)
5627 			xfer = transfer;
5628 		else
5629 			xfer = len;
5630 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
5631 			/* Dense layout. */
5632 			if (stripe_pos >= flp->nfsfl_fhcnt)
5633 				return (EIO);
5634 			fhp = flp->nfsfl_fh[stripe_pos];
5635 			io_off = (rel_off / (stripe_unit_size *
5636 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
5637 			    rel_off % stripe_unit_size;
5638 		} else {
5639 			/* Sparse layout. */
5640 			if (flp->nfsfl_fhcnt > 1) {
5641 				if (stripe_index >= flp->nfsfl_fhcnt)
5642 					return (EIO);
5643 				fhp = flp->nfsfl_fh[stripe_index];
5644 			} else if (flp->nfsfl_fhcnt == 1)
5645 				fhp = flp->nfsfl_fh[0];
5646 			else
5647 				fhp = np->n_fhp;
5648 			io_off = off;
5649 		}
5650 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0)
5651 			commit_thru_mds = 1;
5652 		else
5653 			commit_thru_mds = 0;
5654 		if (rwflag == FREAD)
5655 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
5656 			    io_off, xfer, fhp, cred, p);
5657 		else {
5658 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
5659 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
5660 			    cred, p);
5661 			if (error == 0) {
5662 				NFSLOCKCLSTATE();
5663 				lyp->nfsly_flags |= NFSLY_WRITTEN;
5664 				NFSUNLOCKCLSTATE();
5665 			}
5666 		}
5667 		if (error == 0) {
5668 			transfer = stripe_unit_size;
5669 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
5670 			len -= xfer;
5671 			off += xfer;
5672 		}
5673 	}
5674 	return (error);
5675 }
5676 
5677 /*
5678  * The actual read RPC done to a DS.
5679  */
5680 static int
5681 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
5682     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp,
5683     struct ucred *cred, NFSPROC_T *p)
5684 {
5685 	uint32_t *tl;
5686 	int error, retlen;
5687 	struct nfsrv_descript nfsd;
5688 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5689 	struct nfsrv_descript *nd = &nfsd;
5690 	struct nfssockreq *nrp;
5691 
5692 	nd->nd_mrep = NULL;
5693 	nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh, fhp->nfh_len,
5694 	    NULL, &dsp->nfsclds_sess);
5695 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
5696 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
5697 	txdr_hyper(io_off, tl);
5698 	*(tl + 2) = txdr_unsigned(len);
5699 	nrp = dsp->nfsclds_sockp;
5700 	if (nrp == NULL)
5701 		/* If NULL, use the MDS socket. */
5702 		nrp = &nmp->nm_sockreq;
5703 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
5704 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5705 	if (error != 0)
5706 		return (error);
5707 	if (nd->nd_repstat != 0) {
5708 		error = nd->nd_repstat;
5709 		goto nfsmout;
5710 	}
5711 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5712 	*eofp = fxdr_unsigned(int, *tl);
5713 	NFSM_STRSIZ(retlen, len);
5714 	error = nfsm_mbufuio(nd, uiop, retlen);
5715 nfsmout:
5716 	if (nd->nd_mrep != NULL)
5717 		mbuf_freem(nd->nd_mrep);
5718 	return (error);
5719 }
5720 
5721 /*
5722  * The actual write RPC done to a DS.
5723  */
5724 static int
5725 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5726     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
5727     struct nfsfh *fhp, int commit_thru_mds, struct ucred *cred, NFSPROC_T *p)
5728 {
5729 	uint32_t *tl;
5730 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5731 	int error, rlen, commit, committed = NFSWRITE_FILESYNC;
5732 	int32_t backup;
5733 	struct nfsrv_descript nfsd;
5734 	struct nfsrv_descript *nd = &nfsd;
5735 	struct nfssockreq *nrp;
5736 
5737 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
5738 	nd->nd_mrep = NULL;
5739 	nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh, fhp->nfh_len,
5740 	    NULL, &dsp->nfsclds_sess);
5741 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
5742 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
5743 	txdr_hyper(io_off, tl);
5744 	tl += 2;
5745 	*tl++ = txdr_unsigned(*iomode);
5746 	*tl = txdr_unsigned(len);
5747 	nfsm_uiombuf(nd, uiop, len);
5748 	nrp = dsp->nfsclds_sockp;
5749 	if (nrp == NULL)
5750 		/* If NULL, use the MDS socket. */
5751 		nrp = &nmp->nm_sockreq;
5752 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
5753 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5754 	if (error != 0)
5755 		return (error);
5756 	if (nd->nd_repstat != 0) {
5757 		/*
5758 		 * In case the rpc gets retried, roll
5759 		 * the uio fileds changed by nfsm_uiombuf()
5760 		 * back.
5761 		 */
5762 		uiop->uio_offset -= len;
5763 		uio_uio_resid_add(uiop, len);
5764 		uio_iov_base_add(uiop, -len);
5765 		uio_iov_len_add(uiop, len);
5766 		error = nd->nd_repstat;
5767 	} else {
5768 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
5769 		rlen = fxdr_unsigned(int, *tl++);
5770 		if (rlen == 0) {
5771 			error = NFSERR_IO;
5772 			goto nfsmout;
5773 		} else if (rlen < len) {
5774 			backup = len - rlen;
5775 			uio_iov_base_add(uiop, -(backup));
5776 			uio_iov_len_add(uiop, backup);
5777 			uiop->uio_offset -= backup;
5778 			uio_uio_resid_add(uiop, backup);
5779 			len = rlen;
5780 		}
5781 		commit = fxdr_unsigned(int, *tl++);
5782 
5783 		/*
5784 		 * Return the lowest committment level
5785 		 * obtained by any of the RPCs.
5786 		 */
5787 		if (committed == NFSWRITE_FILESYNC)
5788 			committed = commit;
5789 		else if (committed == NFSWRITE_DATASYNC &&
5790 		    commit == NFSWRITE_UNSTABLE)
5791 			committed = commit;
5792 		if (commit_thru_mds != 0) {
5793 			NFSLOCKMNT(nmp);
5794 			if (!NFSHASWRITEVERF(nmp)) {
5795 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
5796 				NFSSETWRITEVERF(nmp);
5797 	    		} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
5798 				*must_commit = 1;
5799 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
5800 			}
5801 			NFSUNLOCKMNT(nmp);
5802 		} else {
5803 			NFSLOCKDS(dsp);
5804 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
5805 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
5806 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
5807 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
5808 				*must_commit = 1;
5809 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
5810 			}
5811 			NFSUNLOCKDS(dsp);
5812 		}
5813 	}
5814 nfsmout:
5815 	if (nd->nd_mrep != NULL)
5816 		mbuf_freem(nd->nd_mrep);
5817 	*iomode = committed;
5818 	if (nd->nd_repstat != 0 && error == 0)
5819 		error = nd->nd_repstat;
5820 	return (error);
5821 }
5822 
5823 /*
5824  * Free up the nfsclds structure.
5825  */
5826 void
5827 nfscl_freenfsclds(struct nfsclds *dsp)
5828 {
5829 	int i;
5830 
5831 	if (dsp == NULL)
5832 		return;
5833 	if (dsp->nfsclds_sockp != NULL) {
5834 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
5835 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
5836 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
5837 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
5838 	}
5839 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
5840 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
5841 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
5842 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
5843 			m_freem(
5844 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
5845 	}
5846 	free(dsp, M_NFSCLDS);
5847 }
5848 
5849 static enum nfsclds_state
5850 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
5851     struct nfsclds **retdspp)
5852 {
5853 	struct nfsclds *dsp, *cur_dsp;
5854 
5855 	/*
5856 	 * Search the list of nfsclds structures for one with the same
5857 	 * server.
5858 	 */
5859 	cur_dsp = NULL;
5860 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
5861 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
5862 		    dsp->nfsclds_servownlen != 0 &&
5863 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
5864 		    dsp->nfsclds_servownlen)) {
5865 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
5866 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
5867 			    dsp->nfsclds_flags);
5868 			/* Server major id matches. */
5869 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
5870 				*retdspp = dsp;
5871 				return (NFSDSP_USETHISSESSION);
5872 			}
5873 
5874 			/*
5875 			 * Note the first match, so it can be used for
5876 			 * sequence'ing new sessions.
5877 			 */
5878 			if (cur_dsp == NULL)
5879 				cur_dsp = dsp;
5880 		}
5881 	}
5882 	if (cur_dsp != NULL) {
5883 		*retdspp = cur_dsp;
5884 		return (NFSDSP_SEQTHISSESSION);
5885 	}
5886 	return (NFSDSP_NOTFOUND);
5887 }
5888 
5889 #ifdef notyet
5890 /*
5891  * NFS commit rpc to a DS.
5892  */
5893 static int
5894 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
5895     struct nfsfh *fhp, struct ucred *cred, NFSPROC_T *p, void *stuff)
5896 {
5897 	uint32_t *tl;
5898 	struct nfsrv_descript nfsd, *nd = &nfsd;
5899 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5900 	struct nfssockreq *nrp;
5901 	int error;
5902 
5903 	nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh, fhp->nfh_len,
5904 	    NULL, &dsp->nfsclds_sess);
5905 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
5906 	txdr_hyper(offset, tl);
5907 	tl += 2;
5908 	*tl = txdr_unsigned(cnt);
5909 	nrp = dsp->nfsclds_sockp;
5910 	if (nrp == NULL)
5911 		/* If NULL, use the MDS socket. */
5912 		nrp = &nmp->nm_sockreq;
5913 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
5914 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5915 	if (error)
5916 		return (error);
5917 	if (nd->nd_repstat == 0) {
5918 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
5919 		NFSLOCKDS(dsp);
5920 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
5921 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
5922 			error = NFSERR_STALEWRITEVERF;
5923 		}
5924 		NFSUNLOCKDS(dsp);
5925 	}
5926 nfsmout:
5927 	if (error == 0 && nd->nd_repstat != 0)
5928 		error = nd->nd_repstat;
5929 	mbuf_freem(nd->nd_mrep);
5930 	return (error);
5931 }
5932 #endif
5933 
5934